Quantum Oscillators 047046609X, 9780470466094

Citation preview

ffirs.tex

5/4/2011

11: 40

Page i

QUANTUM OSCILLATORS

ffirs.tex

5/4/2011

11: 40

Page iii

QUANTUM OSCILLATORS OLIVIER HENRI-ROUSSEAU and PAUL BLAISE

A John Wiley & Sons, Inc., Publication

ffirs.tex

5/4/2011

11: 40

Page iv

Copyright © 2011 by John Wiley & Sons, Inc. All rights reserved Published by John Wiley & Sons, Inc., Hoboken, New Jersey Published simultaneously in Canada No part of this publication may be reproduced, stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, photocopying, recording, scanning, or otherwise, except as permitted under Section 107 or 108 of the 1976 United States Copyright Act, without either the prior written permission of the Publisher, or authorization through payment of the appropriate per-copy fee to the Copyright Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, (978) 750-8400, fax (978) 750-4470, or on the web at www.copyright.com. Requests to the Publisher for permission should be addressed to the Permissions Department, John Wiley & Sons, Inc., 111 River Street, Hoboken, NJ 07030, (201) 748-6011, fax (201) 748-6008, or online at http://www.wiley.com/go/permission. Limit of Liability/Disclaimer of Warranty: While the publisher and author have used their best efforts in preparing this book, they make no representations or warranties with respect to the accuracy or completeness of the contents of this book and specifically disclaim any implied warranties of merchantability or fitness for a particular purpose. No warranty may be created or extended by sales representatives or written sales materials. The advice and strategies contained herein may not be suitable for your situation. You should consult with a professional where appropriate. Neither the publisher nor author shall be liable for any loss of profit or any other commercial damages, including but not limited to special, incidental, consequential, or other damages. For general information on our other products and services or for technical support, please contact our Customer Care Department within the United States at (800) 762–2974, outside the United States at (317) 572–3993 or fax (317) 572–4002. Wiley also publishes its books in a variety of electronic formats. Some content that appears in print may not be available in electronic formats. For more information about Wiley products, visit our web site at www.wiley.com. Library of Congress Cataloging-in-Publication Data Henri-Rousseau, Olivier. Quantum oscillators / Olivier Henri-Rousseau and Paul Blaise. p. cm. Includes index. ISBN 978-0-470-46609-4 (cloth) 1. Harmonic oscillators. 2. Spectrum analysis. 3. Wave mechanics. I. Blaise, Paul. II. Title. QC174.2.H45 2011 541 .224–dc22

4. Hydrogen bonding.

2011008577 Printed in the United States of America 10

9

8

7

6

5

4

3

2

1

ffirs.tex

5/4/2011

11: 40

Page v

This book is dedicated to Prof. Andrzej Witkowski of the Jagellonian University of Cracow, on the occasion of his 80th birthday.

Toc.tex

18/3/2011

10: 35

Page vii

CONTENTS List of Figures xiii Preface xvii Acknowledgments xxiii

PART 1

BASIS REQUIRED FOR QUANTUM OSCILLATOR STUDIES CHAPTER 1

BASIC CONCEPTS REQUIRED FOR QUANTUM MECHANICS

1.1 Basic Concepts of Complex Vectorial Spaces 1.2 Hermitian Conjugation 8 1.3 Hermiticity and Unitarity 12 1.4 Algebra Operators 18 CHAPTER 2

2.1 2.2 2.3 2.4 2.5 2.6

3

BASIS FOR QUANTUM APPROACHES OF OSCILLATORS

Oscillator Quantization at the Historical Origin of Quantum Mechanics Quantum Mechanics Postulates and Noncommutativity 25 Heisenberg Uncertainty Relations 30 Schrödinger Picture Dynamics 37 Position or Momentum Translation Operators 45 Conclusion 54 Bibliography 55

CHAPTER 3

21

QUANTUM MECHANICS REPRESENTATIONS

3.1 Matrix Representation 57 3.2 Wave Mechanics 68 3.3 Evolution Operators 76 3.4 Density operators 88 3.5 Conclusion 104 Bibliography 106 CHAPTER 4

SIMPLE MODELS USEFUL FOR QUANTUM OSCILLATOR

PHYSICS 4.1 Particle-in-a-Box Model 107 4.2 Two-Energy-Level Systems 115

vii

Toc.tex

18/3/2011

viii 4.3

10: 35

Page viii

CONTENTS

Conclusion 128 Bibliography 128

PART II

SINGLE QUANTUM HARMONIC OSCILLATORS CHAPTER 5 ENERGY REPRESENTATION FOR QUANTUM HARMONIC OSCILLATOR

5.1 Hamiltonian Eigenkets and Eigenvalues 131 5.2 Wavefunctions Corresponding to Hamiltonian Eigenkets 5.3 Dynamics 156 5.4 Boson and fermion operators 162 5.5 Conclusion 165 Bibliography 166

CHAPTER 6

150

COHERENT STATES AND TRANSLATION OPERATORS

6.1 Coherent-State Properties 168 6.2 Poisson Density Operator 174 6.3 Average and Fluctuation of Energy 175 6.4 Coherent States as Minimizing Heisenberg Uncertainty Relations 6.5 Dynamics 180 6.6 Translation Operators 183 6.7 Coherent-State Wavefunctions 186 6.8 Franck–Condon Factors 189 6.9 Driven Harmonic Oscillators 193 6.10 Conclusion 197 Bibliography 198

CHAPTER 7

BOSON OPERATOR THEOREMS

7.1 Canonical Transformations 199 7.2 Normal and Antinormal Ordering Formalism 204 7.3 Time Evolution Operator of Driven Harmonic Oscillators 7.4 Conclusion 221 Bibliography 222

CHAPTER 8

8.1 8.2 8.3 8.4

PHASE OPERATORS AND SQUEEZED STATES

Phase Operators 223 Squeezed States 229 Bogoliubov–Valatin transformation Conclusion 241 Bibliography 241

239

217

177

Toc.tex

18/3/2011

10: 35

Page ix

CONTENTS

PART III

ANHARMONICITY CHAPTER 9

9.1 9.2 9.3 9.4 9.5 9.6

ANHARMONIC OSCILLATORS

Model for Diatomic Molecule Potentials 245 Harmonic oscillator perturbed by a Q3 potential 251 Morse Oscillator 257 Quadratic Potentials Perturbed by Cosine Functions Double-well potential and tunneling effect 267 Conclusion 277 Bibliography 277

CHAPTER 10

265

OSCILLATORS INVOLVING ANHARMONIC COUPLINGS

10.1 10.2 10.3 10.4

Fermi resonances 279 Strong Anharmonic Coupling Theory 282 Strong Anharmonic Coupling within the Adiabatic Approximation 285 Fermi Resonances and Strong Anharmonic Coupling within Adiabatic Approximation 297 10.5 Davydov and Strong Anharmonic Couplings 301 10.6 Conclusion 312 Bibliography 312

PART IV

OSCILLATOR POPULATIONS IN THERMAL EQUILIBRIUM CHAPTER 11

DYNAMICS OF A LARGE SET OF COUPLED OSCILLATORS

11.1 Dynamical Equations in the Normal Ordering Formalism 317 11.2 Solving the linear set of differential equations (11.27) 323 11.3 Obtainment of the Dynamics 325 11.4 Application to a Linear Chain 329 11.5 Conclusion 331 Bibliography 331 DENSITY OPERATORS FOR EQUILIBRIUM POPULATIONS CHAPTER 12 OF OSCILLATORS 12.1 12.2

Boltzmann’s H-Theorem 333 Evolution Toward Equilibrium of a Large Population of Weakly Coupled Harmonic Oscillators 337 12.3 Microcanonical Systems 348 12.4 Equilibrium Density Operators from Entropy Maximization 349 12.5 Conclusion 358 Bibliography 359

ix

Toc.tex

18/3/2011

x

10: 35

Page x

CONTENTS

CHAPTER 13

THERMAL PROPERTIES OF HARMONIC OSCILLATORS

13.1 Boltzmann Distribution Law inside a Large Population of Equivalent Oscillators 13.2 Thermal properties of harmonic oscillators 364 13.3 Helmholtz Potential for Anharmonic Oscillators 388 13.4 Thermal Average of Boson Operator Functions 391 13.5 Conclusion 403 Bibliography 405

PART V

QUANTUM NORMAL MODES OF VIBRATION CHAPTER 14

14.1 14.2 14.3 14.4 14.5 14.6 14.7

Maxwell Equations 409 Electromagnetic Field Hamiltonian 415 Polarized Normal Modes 418 Normal Modes of a Cavity 420 Quantization of the Electromagnetic Fields 423 Some Thermal Properties of the Quantum Fields Conclusion 442 Bibliography 442

CHAPTER 15

15.1 15.2 15.3 15.4

QUANTUM ELECTROMAGNETIC MODES

437

QUANTUM MODES IN MOLECULES AND SOLIDS

Molecular Normal Modes 443 Phonons and Normal Modes in Solids 451 Einstein and Debye Models of Heat Capacity Conclusion 464 Bibliography 464

460

PART VI

DAMPED HARMONIC OSCILLATORS CHAPTER 16

16.1 16.2 16.3 16.4 16.5 16.6 16.7

DAMPED OSCILLATORS

Quantum Model for Damped Harmonic Oscillators Second-Order Solution of Eq. (16.41) 475 Fokker–Planck Equation Corresponding to (16.114) Nonperturbative Results for Density Operator 498 Langevin Equations for Ladder Operators 503 Evolution Operators of Driven Damped Oscillators Conclusion 515 Bibliography 516

468 494

509

361

Toc.tex

18/3/2011

10: 35

Page xi

CONTENTS

PART VII

VIBRATIONAL SPECTROSCOPY CHAPTER 17

APPLICATIONS TO OSCILLATOR SPECTROSCOPY

17.1 IR Selection Rules for Molecular Oscillators 519 17.2 IR Spectra within the Linear Response Theory 534 17.3 IR Spectra of Weak H-Bonded Species 539 17.4 SD of Damped Weak H-Bonded Species 548 17.5 Approximation for Quantum Damping 550 17.6 Damped Fermi Resonances 555 17.7 H-Bonded IR Line Shapes Involving Fermi Resonance 17.8 Line Shapes of H-Bonded Cyclic Dimers 566 Bibliography 584 CHAPTER 18

APPENDIX

18.1 An Important Commutator 587 18.2 An Important Basic Canonical Transformation 587 18.3 Canonical Transformation on a Function of Operators 18.4 Glauber–Weyl Theorem 590 18.5 Commutators of Functions of the P and Q operators 18.6 Distribution Functions and Fourier Transforms 593 18.7 Lagrange Multipliers Method 604 18.8 Triple Vector Product 605 18.9 Point Groups 607 18.10 Scientific Authors Appearing in the Book 622

Index

635

561

589 591

xi

Toc.tex

18/3/2011

10: 35

Page xiii

LIST OF FIGURES 2.1 2.2 4.1 4.2

4.3 5.1 5.2 6.1

Contradiction between experiment (shaded areas) and classical prediction (lines). 22 Quantum and classical relative variance A/A. 28 Particle-in-a-box model. 109 One-dimensional particle-in-a-box model. Energy levels and corresponding wavefunctions and probability densities for the four lowest quantum numbers. 112 Correlation energy levels of two interacting energy levels. 120 Five lowest energy levels and wavefunctions. Comparison between (a) quantum harmonic oscillator and (b) particle-in-a-box model. 157 Fermion energy levels and corresponding eigenkets. 162 Time evolution of the probability density (6.115) of a coherent-state 

 units, t in ω−1 small units, and wavefunction, with Q expressed in 2mω α = 1. 190 6.2 Displaced oscillator wavefunctions generating Franck–Condon factors. 191 6.3 Stabilization of the energy of the eight lowest eigenvalues Ek (n◦ )/ω◦ with respect to n◦ . 197 9.1 Total energy of the molecular ion H+ 2 as a compromise between a repulsive electronic kinetic energy and an attractive potential energy. Energies are in electron volt and distances in Ångström. 247 9.2 Progressive stabilization of the eigenvalues appearing in Eq. (9.50) with the dimension n◦ of the truncated matrix representation (η = −0.017). 254 9.3 Relative dispersion of the difference between the energy levels and the virial theorem. 256 9.4 Five lowest wavefunctions k (ξ) of the Morse Hamiltonian compared to the five symmetric or antisymmetric lowest wavefunctions n (ξ) of the √ harmonic Hamiltonian. The length unit is Q◦◦ = h/2mω. 263 9.5 The 40√lowest energy levels of the Morse oscillator. The length unit is Q◦◦ = /2mω. 264 9.6 Energy gap between the numerical and exact eigenvalues for a Morse oscillator. 264 9.7 Comparison between the energy levels calculated by Eq. (9.100) and the wavefunctions obtained by Eq. (9.101) and the energy levels and the wavefunctions of the harmonic oscillator. 267 9.8 Ammonia molecule. 268 9.9 Double-well ammonia potential. 268 9.10 Example of double-well potential V (Q) defined by Eq. (9.103) in terms of the geometric parameters V1◦ , V2◦ , QS , Q1 , and Q2 defined in the text. 269

xiii

Toc.tex

xiv 9.11 9.12 9.13 9.14 10.1 10.2 10.3 10.4 10.5 10.6

11.1

12.1

12.2

12.3

12.4

12.5 12.6 12.7

18/3/2011

10: 35

Page xiv

LIST OF FIGURES

Representation of the six lowest wavefunctions and the corresponding energy levels for symmetrical double-well potential. 273 Influence of the double-well potential asymmetry on the eigenstates of the double-well potential Hamiltonian. 274 Schematic representation of the two wavefunctions (9.120). 275 Probability density (9.124) for different times t expressed in units ω−1 . 276 Excitation of the fast mode changing the ground state of the H-bond bridge oscillator into a coherent state. 297 Fermi resonance in H-bonded species within the adiabatic approximation. 298 Davydov coupling. 302 Degenerate modes of a centrosymmetric H-bonded dimer. 302 Davydov coupling in H-bonded centrosymmetric cyclic dimers. 303 Effects of the parity operator C2 on the ground and the first excited states of the symmetrized g and u eigenfunctions of the g and u quantum harmonic oscillators involved in the centrosymmetric cyclic dimer. 312 Classical model equivalent to the quantum one described by the Hamiltonian (11.64). A long chain of pendula of the same angular frequency ω◦ coupled by springs of angular frequency ω, where k is the force constant of the springs, l and m are, respectively, the lengths and the masses of the pendula, and g is the gravity acceleration constant. 330 Time evolution of the local energy H1 (t) of oscillator 1 of systems involving N = 2, 10, 100, and 500 oscillators computed by Eqs. (12.21) and (12.22). The time is expressed in units corresponding to the time required to attain the first zero value of the local energy. 339 Pictorial representation of the coarse-grained analysis of the energy distribution of the oscillators inside energy cells of increasing energy Ei. . The boxes indicate the energy cells, whereas the black disks represent the oscillators. The number ni (Ei ) of oscillators having energy Ei is given in the bottom boxes. εγ is the width of the energy cells given by Eq. (12.24). 340 Time evolution of the entropy of a chain of N = 100 quantum harmonic oscillators. The time is in Tθ units, with Tθ given by Eq. (12.23). The initial excitation energy of the site k = 1 is α21 = N. 341 Energy distribution of a chain of N = 1000 oscillators for several values of the cell parameter γ. The analyzing time t∞ = 1000Tθ with Tθ given by Eq. (12.23). The initial excitation energy of the site k = 1 is α21 = N. ni (E, t∞ ) is the number of oscillators having their energy calculated by Eqs. (12.21) and (12.22) within the energy cell i of width εγ given by Eq. (12.24) according to Fig. 12.2. 342 Energy distribution of N = 1000 coupled oscillators for γ = 4 and for time t∞ going from t∞ = 10Tθ to t∞ = 109 Tθ . 342 Staircase representation of the cumulative distribution functions of the probabilities (12.26). 343 Time fluctuation of B(t) around its mean value B(t) for a chain of N = 100 coupled quantum harmonic oscillators. 344

Toc.tex

18/3/2011

10: 35

Page xv

LIST OF FIGURES

12.8

xv

Linear regression −B as a function of 1/α◦2 1 from the values of expression (12.33). The solid line is the regression curve corresponding to − = 80.659 × α1◦2 − 0.0179 with a regression coefficient 1

r 2 = 0.999. 345 √ 12.9 Linear regression of B/B of B with respect to 1/ N obtained according to the values of expression (12.37). 346 12.10 Relative dispersion S/S of the entropy S as a function of the number N 3 of degrees of freedom. γ = 4, k = 1, α◦2 102 . The i = N, t∞ = 10 Tθ , Ntk = √ full line corresponds to the linear regression S/S = 0.543(1/ N) + 0.3473 with a correlation coefficient r 2 = 0.988. 347 13.1 Values of W (N1 , N2 , . . . ) calculated by Eqs. (13.5) and for NTot = 21, ETot = 21ω, for eight different configurations verifying Eqs. (13.4). For each configuration, the eight lowest energy levels Ek of the quantum harmonic oscillators are reproduced, with for each of them, as many dark circles as they are (Nk ) of oscillators having the corresponding energy Ek . 363 13.2 Thermal capacity Cv in R units for a mole of oscillators of angular frequency ω = 1000 cm−1 . 370 √ 13.3 Temperature evolution of the elongation Q(T ) (in Q◦◦ = /2mω units) of an anharmonic oscillator. Anharmonic parameter β = 0.017ω; number of basis states 75. 387 14.1 Polar spheric coordinates: x = r sin θ cos φ, y = r sin θ sin φ, and z = r cos θ; and 0 ≤ r < ∞, 0 ≤ θ ≤ π, and 0 ≤ φ ≤ 2π. r is the radial coordinate, θ and φ are respectively the inclination and azimuth angles. 422 14.2 HP electric field averaged over different coherent states of increasing eigenvalue αnε and their corresponding relative dispersion pictured by the thickness of the time dependence field function. 434 14.3 Electromagnetic field spectrum. 435 14.4 Energy density U(ω) within a cavity for different temperatures. The U(ω) are normalized with respect to the maximum of the curve at 2500 K. 438 14.5 Spectrum of the cosmic microwave background (squares) superposed on a 2.735 K black-body emission (full line). The intensities are normalized to the maximum of the curve. 440 14.6 Einstein coefficients for two energy levels. 440 15.1 Symmetry elements for a C2v molecule. 450 15.2 Three normal modes of a C2V molecule. 451 15.3 Comparison between the assumed normal mode vibrational frequency distribution σ(ω) given by Eq. (15.62) and an experimental one (solid line) dealing with aluminum at 300 K, deduced from X-ray scattering dealing with aluminum at 300 K, deduced from X-ray scattering measurements. [After C. B. Walker, Phys. Rev., 103 (1956):547–557.] 461 15.4 Temperature dependence of experimental (Handbook of Physics and Chemistry, 72 ed.) heat capacities (dots) of silver as compared to the Einstein (CvE ) and the Debye (CvD ) models as a function of the absolute temperature T . TE = 181 K, TD = 225 K. 464

Toc.tex

18/3/2011

xvi 16.1 16.2 17.1

17.2 17.3 17.4 17.5

17.6 17.7

17.8 17.9 17.10 17.11

17.12

17.13

17.14 17.15

18.1 18.2 18.3

10: 35

Page xvi

LIST OF FIGURES

Integration area over t  and t  . 486 Time evolution of the average position for the driven damped quantum harmonic oscillator. 503 Absorption or emission by a quantum harmonic oscillator mode resulting from a resonant coupling with an electromagnetic mode of the same angular frequency ω◦ . 524 IR transitions in a Morse oscillator. 527 Appearance of a hot band in the IR spectrum of a Morse oscillator. 529 IR transition splitting by Fermi resonance. 532 IR doublets of Fermi resonance for three situations: one at resonance (2ωδ = ω◦ = 3000 cm−1 ) and two symmetric ones, out of resonance (2ωδ = ω◦ ±200 cm−1 = 2800 cm−1 ) for a coupling √ 2ξωδ = 120 cm−1 . 533 Tunnel effect splitting. 534 Comparison of the adiabatic (17.89) SD with the reference nonadiabatic (17.115) one: α◦ = 1.00, T = 300 K, ω◦ = 3000 cm−1 ,  = 150 cm−1 , γ ◦ = −0.20 . 545 Spectral analysis at T = 0 K in the absence of indirect damping ω◦ = 3000 cm−1 ,  = 100 cm−1 , α◦ = 1, γ ◦ = 0.025 , γ = 0. 548 Spectral analysis at T = 0 K in the presence of damping. ω◦ = 3000 cm−1 ,  = 100 cm−1 , α◦ = 1, γ ◦ = 0.025 , γ = 0.10 . 554 Damped Fermi resonance. 556 Influence of damping on line shapes involving Fermi resonance. Comparison between profiles calculated with the help of Eq. (17.179) to the corresponding Dirac delta peaks obtained from Eq. (17.180). ω◦ = 3000 cm−1 ,  = 150 cm−1 , 2ωδ = 3150 cm−1 . 560 Influence of damping on line shapes involving Fermi resonance, calculated by Fourier transform of Eq. (17.181). ω◦ = 3000 cm−1 ,  = 150 cm−1 , 2ωδ = 3150 cm−1 . 561 νX−H spectral densities of weak H-bonded species involving a Fermi ◦ −1 −1 resonance for √ different ◦values of the ωδ . ω = 3000 cm ,  = 150 cm , ◦ α = 1.5, ξ 2 = 0.8, γ = 0.15 . 564 Line shapes obtained from Eq. (17.193) when the Fermi coupling is vanishing. 565 IR spectrum for the CD3 CO2 H dimer in the gas phase at room temperature. Parameters: T = 300 K,  = 88 cm−1 , α◦ = 1.19, ω◦ = 3100 cm−1 , V ◦ = −1.55 , η = 0.25, γ = 0.24 , γ ◦ = 0.10 . 584 − → − → − → Triple vectorial product A × ( B × C ). 606 Symmetry elements for a C2v molecule. 610 The C3v symmetry operations. 611

fpref.tex

18/3/2011

10: 29

Page xvii

PREFACE Quantum oscillators play a fundamental role in many area of physics and chemical physics, especially in infrared spectroscopy. They are encountered in molecular normal modes, or in solid-state physics with phonons, or in the quantum theory of light, with photons. Besides, quantum oscillators have the merit to be more easily exposed than the other physical systems interested by quantum mechanics because of their one-dimensional fundamental nature. However, despite the relative simplicity of quantum oscillators combined with their physical importance, there is a lack of monographs specifically devoted to them. Indeed it would be thereby of interest to dispose of a treatise widely covering the quantum properties of quantum harmonic oscillators at the following levels of increasing difficulty: (i) time-independent properties, (ii) reversible dynamics, (iii) thermal statistical equilibrium, and (iv) irreversible evolution toward equilibrium. And not only harmonic oscillators but also anharmonic ones, as well as single oscillators and anharmonically coupled oscillators. As a matter of fact, such subjects are dispersed among different books of more or less difficulty and mixed with other physical systems. The aim of the present book is to remove that which would be considered as a lack. This book will start from an undergraduate level of knowledge and then will rise progressively to a graduate one. To allow that, it is divided into seven different parts of increasing conceptual difficulties. Part I with Chapters 1–4 gives all the basic concepts required to study the different aspects of quantum oscillators. Part II, Chapters 5–8, is devoted to the properties of single quantum harmonic oscillators. Moreover, Part III deals with anharmonicity, either that of single anharmonic oscillator (Chapter 9) or that of anharmonically coupled harmonic oscillators (Chapter 10). Furthermore, Part IV, Chapters 11–13, treats the thermal properties of a large population of harmonic oscillators at statistical equilibrium. Part V concerns different kinds of quantum normal modes met either in light (Chapter 14) or in molecules and solids (Chapter 15). Finally, Part VI, Chapter 16, studies the irreversible behavior of damped quantum oscillators, whereas Part VII, Chapter 17, applies many of the results of the previous chapters to some spectroscopic properties of quantum oscillators. Its now time to be more precise with the contents of these parts. Chapter 1 summarizes the minimal mathematical properties (specially those of Hilbert spaces and of noncommuting operator algebra) required to understand quantum principles. That is the aim of Chapter 2, which, after giving the postulates of quantum mechanics, treats quantum average values and dispersion, allowing one to get the Heisenberg uncertainty relations, and develops the basic consequences of the time-dependent Schrödinger equation. Then, Chapter 3 goes further by looking at the different representations of quantum mechanics, which makes tractable the

xvii

fpref.tex

xviii

18/3/2011

10: 29

Page xviii

PREFACE

quantum generalities exposed in the previous chapter, and which will be of great help in the further studies of quantum oscillators. These quantum descriptions are matrix mechanics, wave mechanics, and time-dependent representations, that is, Schrödinger, Heisenberg, and interaction pictures, and finally the density operator representation, which may be declined according to matrix mechanics or wave mechanics and also to different time-dependent pictures. Chapter 4 ends Part I, being devoted to three different but important physical models, which will enlighten the further studies of quantum oscillators. They are the particle-in-a-box model, which is a simple and didactic introduction to energy quantization that will be met for quantum oscillators, the two-energy-level model, which will be used when studying Fermi resonances appearing in vibrational spectroscopy, and the Fermi golden rule, involving concepts that will be used in the same area of vibrational spectroscopy. Following Part 1, which deals with the basis required for quantum oscillators studies, Part II enters into the heart of the subject. Chapter 5 focuses attention on the quantum energetic representation of harmonic oscillators by solving their timeindependent Schrödinger equation using ladder operators (Boson operators), thus allowing one to determine the quantized energy levels and the corresponding Hamiltonian eigenkets, and also the action of the ladder operators on these eigenkets. It continues by obtaining the oscillator excited wavefunctions, from the corresponding ground state using the action of the ladder operators on the Hamiltonian eigenkets. After this Hamiltonian eigenket representation, Chapter 6 is concerned with coherent states, which minimize the Heisenberg uncertainty relations, and translation operators, the action of which on Hamiltonian ground states yields coherent states, by studying their properties, which are deeply interconnected, and then used to calculate Franck–Condon factors and to diagonalize the Hamiltonian of driven harmonic oscillators. Chapter 7 continues Part II by giving proofs of some Boson operator theorems, which are applied at its end to find the dynamics of a driven harmonic oscillator and which will be widely used in the following. Finally, Chapter 8 closes Part II by treating some more complicated topics such as phase operators, squeezed states, and Bogoliubov–Valatin transformation, which involve products of ladder operators. The properties of single quantum harmonic oscillators found in Part II allow us to treat anharmonicity in Part III. That is first done in Chapter 9 by studying anharmonic oscillators such as those involving Morse potentials, which are more realistic than harmonic potentials for diatomic molecules or double-well potentials leading to quantum tunneling, and in Chapter 10 by studying several harmonic oscillators involving anharmonic coupling. In this last chapter of Part III, together with Fermi resonances, is studied the strong anharmonic coupling theory encountered in the quantum theory of weak H-bonded species and allowing the adiabatic separation between low- and high-frequency anharmonically coupled oscillators, which is studied in detail. Chapter 10 ends with a study of anharmonic coupling between four oscillators, which is used to model a centrosymmetric cyclic H-bonded dimer. Parts II and III ignored the thermal properties of single or coupled quantum oscillators, considering them as isolated from the medium, what they may be, harmonic or anharmonic. The aim of Part IV is to address the thermal influence of the medium. Part IV begins this study with a somewhat unusual chapter (Chapter 11) dealing with the dynamics of very large populations of linearly coupled harmonic

fpref.tex

18/3/2011

10: 29

Page xix

PREFACE

xix

oscillators starting from an initial situation where the energy is found only on one of the oscillators. Moreover, having proven the Boltzmann H-theorem according to which the entropy increases until statistical equilibrium is attained, Chapter 12 applies the results of Chapter 11 to show how, after some characteristic time has elapsed, the statistical entropy reaches its maximum, in agreement with the Boltzmann theorem, whereas a coarse-grained energy analysis of the energy distribution of the oscillators sets reveals a Boltzmann energy distribution. Then, applying the principle of entropy maximization at statistical equilibrium, this chapter obtains the microcanonical and canonical density operators. Finally, Chapter 13 closes Part IV by studying the thermal properties of quantum harmonic oscillators (thermal average energies, heat capacities, thermal energy fluctuations) and ends with the demonstration of the expression of the thermal average of general functions of Boson operators, which contains as a special case the Bloch theorem. Chapter 11 of Part IV studies the dynamics of a large population of coupled quantum harmonic oscillators that, as calculation intermediates, are considered to be normal modes, but without taking attention to them due to the dynamics preoccupations. Since normal modes of systems of many degrees of freedom are collective harmonic motions in which all the parts are moving at the same angular frequency and the same phase, it is possible, within classical physics, to extract for such systems the classical normal modes and then to quantize them to get quantum harmonic oscillators to which it is possible to apply all the results of Parts II–IV. This is the purpose of Part V, which starts (Chapter 14) with a study of the quantum normal modes of electromagnetic fields. That may be first performed with obtaining the classical normal modes of the fields by passing for the Maxwell equations in the vacuum, from the geometrical space to the reciprocal one, using Fourier transforms, and then introduce a commutation rule between the conjugate variables of the electromagnetic field, which are the potential vector and the electric field in the reciprocal space. Then, applying the thermal properties of quantum oscillators found in Chapter 13, it is possible to derive the black-body radiation Planck law and the Stefan–Boltzmann law, and also the ratio of the Einstein coefficients. Chapter 15 completes this part devoted to normal modes by determining the classical molecular normal modes and then quantizing them, and so obtaining the normal modes of a one-dimensional solid in the reciprocal space, allowing one, on application of the thermal properties of oscillators, to obtain the Einstein and the Debye results concerning the solids heat capacity of solids. Continuing the work of Part IV devoted to thermal equilibrium, which was applied in Part V to find the thermal statistical properties of normal modes, Part VI, involving only Chapter 16, studies the irreversible behavior of harmonic oscillators, which are damped due to the influence of the medium. This irreversible influence is modeled by considering the medium, acting as a thermal bath, as a very large set of harmonic oscillators of variable angular frequencies, weakly coupled to the damped oscillator, and each constrained to remain in statistical thermal equilibrium. Then, solving within this approach the Liouville equation, and after performing the Markov approximation, the master equations governing the dynamics of the density operators of driven or undriven harmonic oscillators are obtained. This procedure allows one to derive in a subsequent section the Fokker–Planck equation for damped harmonic oscillators. Next, Chapter 16 continues, by aid of an approach similar to that used for the

fpref.tex

xx

18/3/2011

10: 29

Page xx

PREFACE

master equations by deriving the Langevin equations governing the time-dependent statistical averages of the Boson operators, and ends, using these Langevin equations, by obtaining the interaction picture time evolution operator of driven damped quantum harmonic oscillators, which allows one to get the corresponding time-dependent density operator, which may be envisaged as a consequence of the corresponding master equation governing the dynamics of damped oscillators. The book ends with Part VII corresponding to the single Chapter 17, by applying many of the properties of quantum oscillators obtained in Parts II and III (Chapter 10), Part IV (Chapter 13), and Part VI (Chapter 16), to find some important results in vibrational spectroscopy, such as the IR selection rule for quantum harmonic oscillators, and to study using linear response theory, and after having proved it, the line shapes of some physical realistic situations involving anharmonically coupled damped quantum harmonic oscillators encountered in the area of H-bonded species. Clearly, the topics studied in all these parts involve progressive levels of difficulty, varying from undergraduate to graduate. It may be of interest to list the quantum theoretical tools necessary to treat the different subjects of the book. Essential tools are kets, bras, scalar products, closure relation, linear Hermitian and unitary operators, commutators and eigenvalue equations, as well as quantum mechanical fundamentals. There exist seven postulates, concerning the notions of quantum average values and of the corresponding fluctuations leading to the Heisenberg uncertainty relations. We list the time dependence of the quantum average values leading to the Ehrenfest theorem and to the virial theorem, the different representations of quantum mechanics involving wave mechanics, matrix representation, the different time-dependent representations, that is, the Schrödinger and Heisenberg ones and also the interaction picture, all using the time evolution operators and, finally, the various density operator representations. Furthermore, there are also mathematical tools that are not specific to the subject but necessary to the understanding of some developments and that will be treated in the Appendix (Chapter 18). Among them, some commutator algebra, particularly those dealing with the position and momentum operators, some theorems concerning exponential operators as the Baker–Campbell–Hausdorff relation or the Glauber–Weyl theorem, some information about Fourier transforms and distribution functions, the Lagrange multipliers method, complex results concerning vectorial analysis, and elements dealing with the point-group theory. On the other hand, as it may be inferred from the presentation of the different parts of the book, the following quantum oscillator properties will be considered: Hamiltonian eigenkets of harmonic oscillators and their corresponding wavefunctions, ladder operators, action of these operators on the Hamiltonian eigenkets, coherent states, translation operators, squeezed states and corresponding squeezing operators, time dependence of the ladder operators, canonical transformations involving ladder operators, normal and antinormal ordering, Bogolyubov transformations, Boltzmann density operators of harmonic oscillators, and thermal quantum average values of operators, specially that of the translation operator leading to the Bloch theorem. Despite the complexity of the project, our aim is to propose a progressive course where all the demonstrations, whatever their level may be, would present no particular

fpref.tex

18/3/2011

10: 29

Page xxi

PREFACE

xxi

difficulties, and thus would be readable at various levels ranging from undergraduate to postgraduate levels. In this end, we have applied our teaching experience, which used the Gestalt psychology, according to which the main operational principle of the mind is holistic, the whole being more important than the sum of its parts, that is particularly sensitive with respect to the visual recognition of figures and whole forms instead of just a collection of simple lines and curves: We have observed that this concept is very well verified to those unfamiliar with long equations involving many intricated symbols. There are different ways to read this book. The first one concerns quantum mechanics, which, since considered from the viewpoint of oscillators, allows one to avoid all the mathematical difficulties related to the techniques for solving the secondorder partial differential equations encountered in wave mechanics. The second one gives the elements required to understand the theories dealing with the line shapes met spectroscopy more specially in the area of H-bonded species. The third one may be viewed as a simple introduction to quantization of light. The fourth one may be considered as an introduction to quantum equilibrium statistical properties of oscillators, while the fifth focuses attention on the irreversible behavior of oscillators Finally, the sixth concerns chemists interested in molecular spectroscopy. The chapters may be considered as follows: Domains Chapters Quantum 1 2 3 4 5 6 7 9 10 oscillators IR line shape 2 3 4 5 6 7 9 10 spectra Theory 2 3 5 6 7 8 of light Statistical 2 3 5 6 7 12 equilibrium Irreversibility 2 3 5 6 7 11 Molecular 1 2 5 9 10 spectroscopy

13 14 15 16 13 13 14

15

17 16

13 16 17

The cost to be paid will be the inclusion of many details in the demonstrations, which sometimes appear to the advanced readers to be superfluous. In addition, to make the equations more easily readable we have sometimes used unusual notations combined with the introduction of additive brackets, which would appear to be surprising and unnecessary for those indifferent to the didactic advantages of the Gestalt psychology, which is our option.

flast.tex

17/3/2011

19: 5

Page xxiii

ACKNOWLEDGMENTS Prof. W. Coffey (Dublin) Prof. Ph. Durand (Toulouse) Prof. J-L. Déjardin Prof. Y. Kalmykov Prof. H. Kachkachi Dr. P. M. Déjardin Dr. A. Velcescu-Ceasu Dr. P. Villalongue Dr. B. Boulil

xxiii

bins.tex

16/3/2011

16: 58

Page 1

12 10

Ek(n)/( ω)

8

Exact energy E7 E6 E5 E4 E3 E2 E1 E0

6 4 2 0

2

4

6

8 10 12 Number of basis states n

Figure 6.3 Stabilization of the energy of the eight lowest eigenvalues Ek (n◦ )/ω◦ with respect to n◦ .

bins.tex

16/3/2011

16: 58

Page 2

12 E9 10 E8 E7

Ek (n)/ ω

8 E6 6

E5 E4

4

E3 E2

2 E1 E0 0 2

4

6

8 n

10

12

14

Figure 9.2 Progressive stabilization of the eigenvalues appearing in Eq. (9.50) with the dimension n◦ of the truncated matrix representation (η = −0.017).

bins.tex

16/3/2011

ΔEk(n)

16: 58

Page 3

0.2 k0 k1

〈Ek(n)〉 0.0

k2

0.2 k3 0.4

k4 k5

0.6

Figure 9.3 theorem.

0

10

20 n

30

40

Relative dispersion of the difference between the energy levels and the virial

bins.tex

16/3/2011

16: 58

Page 4

5 E4/ ω 4 E3/ ω 3 E2/ ω 2 E1/ ω 1 E0 / ω 10

5

0 Q/Q

5

10

Figure 9.4 Five lowest wavefunctions k (ξ) of the Morse Hamiltonian compared to the five symmetric or antisymmetric lowest wavefunctions n (ξ) of the harmonic Hamiltonian. √ The length unit is Q◦◦ = h/2mw.

bins.tex

16/3/2011

16: 58

Page 5

Ek

Ek 7

7

6

6

E 5

E5 5

5

E 4

E4 4

4

E 3

E3 3

3

E 2

E2 2

2

E 1

E1 1

ω

E0 54 32 1 0 1 2 3 4 5 Q

1

ω E 0

543 21 0 1 2 3 4 5 Q

Figure 9.7 Comparison between the energy levels calculated by Eq. (9.100) and the wavefunctions obtained by Eq. (9.101) and the energy levels and the wavefunctions of the harmonic oscillator.

bins.tex

16/3/2011

16: 58

Page 6

Ek

E4 E5

E3 E2

E E0 1

0

Q

Figure 9.11 Representation of the six lowest wavefunctions and the corresponding energy levels for symmetrical double-well potential.

16: 58

Page 7

Hot band

16/3/2011

EII ωI EI

Intensity

Energy

bins.tex

ωI,II

ωII

E0 ω

2ω

q Figure 17.3 Appearance of a hot band in the IR spectrum of a Morse oscillator.

30

C3

Ty H

H

σv

Tx

Tx Tx

Tx

Ty

σv

Tx

Figure 18.3 The C3v symmetry operations.

120

C23

H 30 30

Tx

Ty Ty

σv

σ v

σv

60

60

Ty

σ v

30

3030

30 30

30

Ty Ty Tx

Tx

16: 58

Tx

Ty

30

Tx

Ty

16/3/2011

120

Ty

bins.tex Page 8

Ch001.tex

PART

10/3/2011

21: 32

Page 1

I

BASIS REQUIRED FOR QUANTUM OSCILLATOR STUDIES

Ch001.tex

10/3/2011

21: 32

Page 3

1

CHAPTER

BASIC CONCEPTS REQUIRED FOR QUANTUM MECHANICS In order to summarize the quantum basis required for the study of oscillators, it is necessary to define some mathematical notions concerning the properties of state spaces, particularly the concepts of linear operators, kets, bras, Hermiticity, eigenvalues, and eigenvectors of linear operators involved in the formulation of the different postulates. The first two sections of this chapter are devoted to this. However, it is possible to pass directly to the third section leaving for later the lecture of the previous one.

1.1 1.1.1

BASIC CONCEPTS OF COMPLEX VECTORIAL SPACES Kets, bras, and scalar products

Quantum mechanics deals with state spaces, that is, vectorial spaces involving complex scalar products that are generally of infinite dimension. Any element of these spaces is named a ket and symbolized | . . . | by inserting inside it a free notation allowing one to clearly identify this ket; for instance, |k 1 or |n. Since the space of states is vectorial, and if the kets |1  and |2  belong to the same state space, then the ket | defined by the linear superposition | = λ1 |1  + λ2 |2  where λ1 and λ2 are two scalars, belongs also to the same state space. Now, to some ket | of the state space there exists a linear functional that associates with some another ket | of this space a complex scalar A , which is the scalar product of | by |. This may be written A = |

(1.1)

All the notations inside the symbol | . . . | are designed to distinguish clearly the ket of interest. For instance, some Latin n or Greek  letters lead to the writing |n or |, but the notation may be as complex as required; for instance, |nl  or |k , the subscripts allowing to distinguish between two kets |nl  and |nj  of the same kind, and in a similar way to kets |k  and |j . In the following we shall use also as specification notations of the form: |{n}, |(n), |[n] in order to reserve the notations |nl  or |k  for kets belonging to the same basis. 1

Quantum Oscillators, First Edition. Olivier Henri-Rousseau and Paul Blaise. © 2011 John Wiley & Sons, Inc. Published 2011 by John Wiley & Sons, Inc.

3

Ch001.tex

4

10/3/2011

21: 32

Page 4

BASIC CONCEPTS REQUIRED FOR QUANTUM MECHANICS

This linear functional, which is denoted |, is named the bra, corresponding to the ket |. The bras may be viewed as belonging to a state space that is the dual space of the state space to which belong the kets, that is, the bras are the Hermitian conjugates of the corresponding kets, namely | = |†

(1.2)

superscript†

where the denotes the Hermitian conjugation. The scalar products have the following properties: (λ1 1 + λ2 2 |)| = λ∗1 1 | + λ∗2 2 | |(|λ1 1 + λ2 2 ) = λ1 |1  + λ2 |2  k |l  = l |k ∗ | > 0 | = 0

(1.3)

if |  = 0

if and only if | = 0

(1.4)

In addition, if this scalar product is normalized, we have | = 1 If the scalar product of two kets | and | is zero, the two kets | and | are said to be orthogonal: | = 0

1.1.2

Linear transformations

Let us consider the action of a linear operator A on a ket |ξ belonging to the state space. This action leads to another ket | according to A|ξ = |

(1.5)

Consider now the action of an another linear operator B acting on the same ket |ξ. Generally, it will yield another ket |: B|ξ = | In most situations, the product of two operators A and B does not commute, that is, AB  = BA The commutator of two operators A and B is symbolized2 by [A, B] ≡ AB − BA 2 The standard notation for a commutator is […, …] where the comma separates the two operators involved. Since the comma risks being unnoticed, in order to avoid this risk we have chosen to reserve as far as, the notation involving [..,..] to commutators, and to use for other situations notations of the kinds (…) or {…}.

Ch001.tex

10/3/2011

21: 32

Page 5

1.1

BASIC CONCEPTS OF COMPLEX VECTORIAL SPACES

5

In some situations, a linear operator A may act on different kets |1 , |2 , …, in such a way as it multiplies them by scalars A1 , A2 , …, more generally A|l  = Al |l 

(1.6)

The kets |l  corresponding to these special situations are the eigenvectors of the operator A while the scalars Al are the corresponding eigenvalues. Equation (1.6) is called an eigenvalue equation. The scalar Al is generally complex. When different eigenvectors exist corresponding to a same eigenvalue, then a degeneracy exists, its degree being the number of eigenvectors associated with this same eigenvalue. In the following, we shall not encounter degeneracy except for very special situations so that we shall ignore the particular treatment of this case. 1.1.2.1 Hermitian conjugate of a linear transformation The Hermitian conjugate of the linear operator A is A† . Consider a linear transformation of the form (1.5) A| = |

(1.7)

Its Hermitian conjugate is the bra |: {A|}† = | Now, the Hermitian conjugate of the linear transformation (1.7) is {A|}† = |A†

(1.8)

| = |A†

(1.9)

which is equivalent to

Consider now an eigenvalue equation of the form (1.6) A| = A|

(1.10)

Then, owing to Eq. (1.8), and because the Hermitian conjugate of a scalar is its complex conjugate, the Hermitian conjugate of Eq. (1.10) is |A† = |A∗

1.1.3

(1.11)

Basis and closure relation

A set {|n } of kets |n  of the state space is said to be orthonormal if these states satisfy n |m  = δnm

(1.12)

where δnm is the Kronecker symbol given by δmm = 1

and

δmn = 0

if

m = n

(1.13)

Ch001.tex

6

10/3/2011

21: 32

Page 6

BASIC CONCEPTS REQUIRED FOR QUANTUM MECHANICS

Again, such a set {|n } forms a basis in the state space, provided all kets |k  belonging to this space may be expanded according to |k  =

∞ 

Cnk |n 

(1.14)

n=1

where the Cnk are the expansion coefficients, which may be complex. Now, premultiply both members of Eq. (1.14) by a bra m | corresponding to some ket belonging to the basis {|n }. It reads m |k  = m |

∞ 

Cnk |n 

n=1

or m |k  =

∞ 

Cnk m |n 

n=1

Therefore, in view of Eq. (1.12), it transforms to m |k  =

∞ 

Cnk δnm

n=1

or, in view of Eq. (1.13), m |k  = Cmk

(1.15)

Then, introducing Eq. (1.15) into Eq. (1.14) we have |k  =

∞ 

n |k |n 

n=1

Furthermore, after commuting the scalar product with the ket in the second member of this equation, we have |k  =

∞ 

{|n n |} |k 

(1.16)

n=1

Now, in order for Eq. (1.16) to be satisfied, whatever may be the ket |k  appearing on both sides of this equation, it is necessary that ∞ 

|n n | = 1

(1.17)

n=1

Equation (1.17) is known as the closure relation. The closure relation (1.17) together with the orthonormality condition (1.12) are the two important properties of a basis, the first being the consequence of the second one. Now, consider the following operation: {|n n |} |k 

Ch001.tex

10/3/2011

21: 32

Page 7

1.1

BASIC CONCEPTS OF COMPLEX VECTORIAL SPACES

7

Then, using the expansion (1.16), this expression reads {|n n |}|k  = {|n n |}

∞ 

Cmk |m 

m=1

or {|n n |}|k  = |n 

∞ 

Cmk n |m 

m=1

and thus, using the orthonormality properties (1.12) {|n n |}|k  = |n 

∞ 

Cmk δnm

m=1

so that {|n n }|k  = |n Cnk Thus, |n n | acts on the ket |k  as an operator, projecting it on to the state |n . Thus it is called a projector

1.1.4

Schwarz inequality

Consider a ket | that is the superposition of two different kets | and |ξ: | = | + λ|ξ

(1.18)

where λ is a complex scalar number. The Hermitian conjugate of this equation is | = | + λ∗ ξ|

(1.19)

Consider now the norm of this ket, which cannot be negative, so that it must be written |  0

(1.20)

Then, using Eqs. (1.18) and (1.19) the norm becomes | = | + λ|ξ + λ∗ ξ| + λλ∗ ξ|ξ

(1.21)

Now, suppose that the scalar λ is given by ξ| ξ|ξ Then, according to Eq. (1.3), the complex conjugate of λ is λ=−

|ξ ξ|ξ Again, introducing Eqs. (1.22) and (1.23) in (1.21), one obtains λ∗ = −

ξ| |ξ ξ| |ξ |ξ − ξ| + ξ|ξ ξ|ξ ξ|ξ ξ|ξ ξ|ξ yielding, after an initial simplification | = | −

| = | −

ξ| |ξ ξ| |ξ − ξ| + |ξ ξ|ξ ξ|ξ ξ|ξ

(1.22)

(1.23)

(1.24)

Ch001.tex

8

10/3/2011

21: 32

Page 8

BASIC CONCEPTS REQUIRED FOR QUANTUM MECHANICS

Then, after cancellation of the two last right-hand terms, this last equation becomes ξ||ξ | = | − ξ|ξ or, in view of the inequality (1.20) |ξ|ξ − ξ||ξ  0 leading to a result that is known as the Schwarz inequality: |ξ|ξ  ξ||ξ

1.2

(1.25)

HERMITIAN CONJUGATION

1.2.1 Theorem dealing with Hermitian conjugates Consider the linear transformation B| = |ξ

(1.26)

Again, owing to Eq. (1.8), its Hermitian conjugate is |B† = ξ|

(1.27)

Then, premultiplying Eq. (1.26) by | and postmultiplying Eq. (1.27) by |, one obtains, respectively, |B| = |ξ

(1.28)

|B† | = ξ|

(1.29)

Thus, owing to Eq. (1.3), it appears that, in the present situation |ξ = ξ|∗ Thus, Eqs. (1.28) and (1.29) yield |B† | = |B|∗

1.2.2

(1.30)

Hermitian conjugate of A†

Consider the Hermitian conjugate (A† )† of the Hermitian conjugate A† of the linear operator A. First, we may write that the Hermitian conjugate of the operator A is a new operator B: B = A† Then, the Hermitian conjugate of

A†

is

(1.31)

B† :

(A† )† = B†

(1.32)

Now, premultiply the two members of Eq. (1.32) by some bra | and postmultiply them by some ket |. Then, one obtains |(A† )† | = |B† |

Ch001.tex

10/3/2011

21: 32

Page 9

1.2

HERMITIAN CONJUGATION

9

Owing to Eq. (1.30), this last expression becomes |(A† )† | = |B|∗ Again, introduce Eq. (1.31) on the right-hand side of this last result. Then, one finds |(A† )† | = |A† |∗ Moreover, using again theorem (1.30), one gets |(A† )† | = |A| Finally, since the latter must be true whatever | and | are, it follows that (A† )† = A

(1.33)

1.2.3 Successive Hermitian conjugations of a linear transformation Consider the Hermitian conjugate of a linear transformation (1.8). It is {{A|}† }† = {|A† }†

(1.34)

Now, let A| = |

|A† = |

and

(1.35)

Then, due to this last equation, Eq. (1.34) reads {{A|}† }† = |† or, in view of Eq. (1.2) {{A|}† }† = | Moreover, due to the first equation of (1.35), we also have {{A|}† }† = A|

1.2.4

Hermitian conjugate of |ξζ|

Consider the following operator and its Hermitian conjugate: A = |ξζ|

and

A† = {|ξζ|}†

(1.36)

What is the relation between A and A† ? To answer this question, premultiply both the operator and its Hermitian conjugate by the bra | and postmultiply both of them by the ket | leading, respectively, to |A| = |{|ξζ|}| and |A† | = | {|ξζ|}† | Now, according to Eq. (1.30), the operator defined by Eq. (1.36) must obey |A† | = |A|∗

Ch001.tex

10

10/3/2011

21: 32

Page 10

BASIC CONCEPTS REQUIRED FOR QUANTUM MECHANICS

Thus, in the present situation, due to the expressions (1.36), the latter takes on the form | {|ξζ|}† | = {| {|ξζ|} |}∗ After simplifying the notation in the more usual form, we have | {|ξζ|}† | = {|ξζ|}∗

(1.37)

Again, the two terms of the right-hand side of this last equation are scalars obeying |ξ∗ = ξ|

and

ζ|∗ = |ζ

Thus, Eq. (1.37) transforms to | {|ξζ|}† | = ξ||ζ Now, the two right scalars appearing on the right-hand side of this last expression do commute, so that | {|ξζ|}† | = |ζξ| Finally, since this last equation must be satisfied, whatever | and | are, one obtains the final result {|ξζ|}† = |ζξ|

(1.38)

1.2.5 Hermitian conjugate of a product of operators that do not commute Now, consider two noncommuting linear operators A and B the product of which is C, that is, AB = C

and

[A, B]  = 0

Then, seek the Hermitian conjugate (AB)† of their product AB. Hence, premultiply the product AB by the bra | and postmultiply it by the ket |. Then, considering the product AB as a new operator C, and applying the theorem (1.30), that is, |C† | = |C|∗ we have |(AB)† | = |AB|∗

(1.39)

Now, observe that the action of the operator B on the ket | and that of the operator A on the bra | are linear transformations of the type B| = |χ

and

|A = μ|

(1.40)

Then, owing to these linear transformations defining the ket |χ and the bra μ|, Eq. (1.39) reads |(AB)† | = μ|χ∗

(1.41)

Ch001.tex

10/3/2011

21: 32

Page 11

1.2

HERMITIAN CONJUGATION

11

Again, due to the relation (1.3) defining the scalar product and its complex conjugate, there is μ|χ∗ = χ|μ Hence, Eq. (1.41) takes the form |(AB)† | = χ|μ

(1.42)

Moreover, the Hermitian conjugate of the linear transformations (1.40) is |B† = χ|

and

A† | = |μ

Thus, the corresponding scalar product yields χ|μ = |B† A† | As a consequence, Eq. (1.42) becomes |(AB)† | = |B† A† | Of course, this last equation must be true for all | and | so that (AB)† = B† A†

(1.43)

1.2.6 Hermitian conjugate of a general expression involving kets, bra operators, and scalars We may summarize here the present results obtained previously and that dealt with the Hermitian conjugate in special situations given, respectively, by Eqs. (1.33), (1.38), and (1.43). For operators we have (A† )† = A;

{|ξζ|} † = |ζξ|

and

(AB)† = B† A†

For linear transformations, we have If

A| = A|  then

{A|}† = |A†

with

|A† = |A∗

Finally, for scalars, we have | = |∗

and

|B† | = |B|∗

Thus, it is possible to deduce general rules allowing one to find the Hermitian conjugate of a general expression involving linear operators kets, bra, and scalars, that is, 1. Replace (a) scalars by their complex conjugates (b) kets by the corresponding bras and vice versa (c)

linear operators by their Hermitian conjugates

2. Invert the order of the different terms, recalling that the position of the scalar is irrelevant.

Ch001.tex

12

10/3/2011

21: 32

Page 12

BASIC CONCEPTS REQUIRED FOR QUANTUM MECHANICS

As a first example, consider the following expression, which is a scalar: |A| = A Since the Hermitian conjugate of | is | and vice versa and since the Hermitian conjugate of the scalar A is its complex conjugate A∗ , the Hermitian conjugate of this expression is |A† | = A∗ Now, consider the following operator: B = λ|A|χ|μ| Applying the above rules, its Hermitian conjugate is given by B† = λ∗ |μ|χ|A† | Finally, consider the operator C, which consists of an exponential of another operator A: C = eiA

with

i2 = −1

Since the complex conjugate of the scalar i is −i, the Hermitian conjugate of the operator C is C† = (eiA )† = e−iA

1.3 1.3.1

†

(1.44)

HERMITICITY AND UNITARITY Hermitian operators

If certain linear operators A are equal to their Hermitian conjugate A† , then they are said to be Hermitian: A = A†

(1.45)

1.3.1.1 Reality of eigenvalues and orthonormality of the eigenvectors In order to show that the eigenvalues of Hermitian operators are real, let us write the eigenvalue equation of a linear operator A|i  = Ai |i 

(1.46)

where Ai is one of the eigenvalues of this operator and |i  the corresponding eigenvector. Premultiply the two members of this equation by the bra i |, conjugate to the ket |i : i |A|i  = i |Ai |i  The eigenvalue Ai being a scalar, must commute with the bra so that one may write i |A|i  = Ai i |i 

(1.47)

Ch001.tex

10/3/2011

21: 32

Page 13

1.3

HERMITICITY AND UNITARITY

13

Next, assume that the eigenvector |i  is normalized, that is, i |i  = 1 Then, Eq. (1.47) simplifies to i |A|i  = Ai

(1.48)

On the other hand, the Hermitian conjugate of this equation is i |A† |i  = A∗i

(1.49)

Again, since we have assumed that the linear operator is Hermitian, it obeys Eq. (1.45), so that i |A† |i  = i |A|i  Thus, it appears from Eqs. (1.48) and (1.49) that the eigenvalue Ai of the Hermitian operator is equal to its complex conjugate A∗i , that is, it is real since it obeys Ai = A∗i

(1.50)

Thus, we have the following property: If

A = A†

Ai = A∗i

then

(1.51)

Now, in order to show that the eigenvectors of an Hermitian operator are orthogonal, let us write the eigenvalue equation of a linear operator for two distinct eigenvalues and eigenvectors: A|i  = Ai |i 

and

A|k  = Ak |k 

The Hermitian conjugate of the first expression in this eigenvalue equation, is i |A† = A∗i i | Besides, if we assume that the operator A is Hermitian, then the eigenvalue Ai is real; then, according to Eq. (1.50), the following results hold: A|k  = Ak |k 

and

i |A = Ai i |

if A = A†

Now, premultiply the two members of the first eigenvalue equation by the bra i |, and postmultiply the two members of its Hermitian conjugate by the ket |k . Then, after commuting the eigenvalues that are scalars, one obtains the two expressions i |A|k  = Ak i |k 

and

i |A|k  = Ai i |k 

Now, substract the second expression from the first one, that is, i |A|k  − i |A|k  = (Ak − Ai )i |k  yielding (Ak − Ai )i |k  = 0 Thus, since we have assumed that the two eigenvalues are different, that is, (Ak − Ai )  = 0

Ch001.tex

14

10/3/2011

21: 32

Page 14

BASIC CONCEPTS REQUIRED FOR QUANTUM MECHANICS

hence it appears that the corresponding eigenvectors of Hermitian operators are orthogonal, leading us to write i |k  = 0

A = A†

if

with

A|k  = Ak |k 

(1.52)

1.3.1.2 Trace and invariance of the trace By definition the trace operation, denoted tr, over any operator C is3  tr{C} = n |C|n  (1.53) n

where the |n  involved in the infinite sum belong to the basis {|n }. Next, suppose that the operator C is the product of two operators A and B, which do not commute, that is, C = AB with Then, the trace takes the form tr{AB} =

[A, B] = 0

 n |AB|n 

(1.54)

n

Introduce between A and B the closure relation built up from the basis {|m }. This procedure leads to a double summation not only over n but also over m:  tr{AB} = n |A|m m |B|n  n

m

Since the terms involved in the double summation are scalar, they commute, so that  m |B|n n |A|m  tr{AB} = n

m

Then, one may omit between B and A the closure relation involving the summation over n to give  tr{AB} = m |BA|m  (1.55) m

However, owing to the definition (1.53) of the trace, the right-hand side of Eq. (1.56) yields  m |BA|m  = tr{BA} (1.56) m

Hence, comparison of Eq. (1.54) and (1.56) shows that tr{AB} = tr{BA} so that the trace operation is invariant with respect to a permutation of A and B. 3

In order to make clear what is meant by the trace operation, in the following we shall denote it, by tr{ } where all the operators involved A, B… will be inside the notation {…}. For instance, tr{AB}.

Ch001.tex

10/3/2011

21: 32

Page 15

1.3

HERMITICITY AND UNITARITY

15

1.3.1.3 Hermitization of the product AB of Hermitian operators when [A, B]  = 0 Consider the product C of two linear operators A and B: C = AB

(1.57)

Again, assume that A and B are Hermitian operators that do not commute, that is, A = A†

B = B†

[A, B]  = 0

As we shall see, their product C is not Hermitian so that it is necessary to convert it to Hermitian form. To show that the product C is not Hermitian let us write with the aid of Eq. (1.43) the Hermitian of C: C† = (AB)† = B† A† Since both operators are Hermitian, it is possible to write C† = BA

if

B = B†

A = A†

and

(1.58)

Thus, since by hypothesis the two operators do not commute, the comparison of Eqs. (1.57) and (1.58) shows that the product C is not Hermitian. Hence, it is necessary to recall that C†  = C if

[A, B]  = 0

when A = A†

B = B†

and

(1.59)

In order to write the product in Hermitian form, we consider the linear combination of C and its Hermitian conjugate, namely D = 21 (C + C† ) Then, the Hermitian conjugate D† of D is Hermitian since D† = 21 (C† + C) = D As a consequence, it appears that the linear combination of the products AB and BA is Hermitian. Hence, important property of Hermitization of the product of two Hermitian operators follows, namely D = 21 {AB + BA} = D†

1.3.2 1.3.2.1

if

A = A†

B = B†

[A, B]  = 0

(1.60)

Eigenkets of two commuting Hermitian operators First theorem Consider two operators A and B that commute, that is, [A, B] = 0

(1.61)

A|i  = Ai |i 

(1.62)

The eigenvalue equation of A is

where Ai is the scalar eigenvalue of the operator A. Now, consider the action of the product BA of the two operators on any eigenket of A BA|i  = BAi |i  = Ai B|i  = Ai |Bi 

Ch001.tex

16

10/3/2011

21: 32

Page 16

BASIC CONCEPTS REQUIRED FOR QUANTUM MECHANICS

In addition, owing to the nullity of the commutator (1.61), we have BA|i  = AB|i  = A|Bi 

(1.63)

where |Bi  is the ket obtained by the linear transformation of B over |i . Thus, by identification of the two last equations, it appears that A|Bi  = Ai |Bi 

(1.64)

This result shows that when A and B commute, and that, according to Eq. (1.62) if |i  is an eigenket of A, then, due to Eq. (1.64), |Bi  is also an eigenket of A. In a like manner, if the eigenvalue equation of B is B|k  = Bk |k  then one obtains B|Ak  = Bk |Ak 

(1.65)

showing that when A and B commute, if |k  is an eigenket of B, |Ak  is also an eigenket of B. 1.3.2.2 Second theorem Consider the two following eigenvalue equations of the same linear Hermitian operator A: A|1  = A1 |1 

and

A|2  = A2 |2 

(1.66)

where A1 and A2 are two different eigenvalues of A, that is, A1 − A2  = 0

(1.67)

Now, consider another linear operator B, which commutes with A, but which is not necessarily Hermitian, that is, [A, B] = 0 Then, owing to the nullity of this commutator, we have 1 |[A, B]|2  = 0

(1.68)

Expanding the commutator gives 1 |[A, B]|2  = 1 |AB|2  − 1 |BA|2  Then, using the first equation of (1.66) or the Hermitian conjugate of the second, one reads 1 | [A, B] |2  = (A1 − A2 ) 1 |B|2 

(1.69)

Hence, owing to Eqs. (1.67)–(1.69), it appears that 1 |B|2  = 0

(1.70)

Thus, if |1  and |2  are eigenkets of any Hermitian operator A, then Eq. (1.70) holds for any operator B that commutes with A.

Ch001.tex

10/3/2011

21: 32

Page 17

1.3

1.3.3

HERMITICITY AND UNITARITY

17

Eigenvalue equation of an exponential operator

Consider an exponential operator eξA , which is a function of the scalar ξ, and another operator A obeying the eigenvalue equation A|n  = An |n 

(1.71)

We search what is the effect of this operator on an eigenstate |n . For this purpose, we may expand on the right-hand side of this last equation the exponential operator in Taylor series, to give ξA

e |n  =

 ξk k!

k

Ak |n 

(1.72)

Now, observe that Ak |n  = Ak−1 A|n  or, in view of Eq. (1.71) Ak |n  = Ak−1 An |n  Again, after commuting the scalar An with the operator Ak−1 written Ak−2 A, one obtains Ak |n  = An Ak−2 A|n  or Ak |n  = An An Ak−2 |n  Proceeding in the same way for each power of A, one gets finally Ak |n  = Akn |n  Then, using this result, Eq. (1.72) becomes ξA

e |n  =

 ξk

Akn |n 

k!

k

Again, return to the expansion appearing on the right-hand side of this last equation to the exponential, and one obtains ξA

ξAn

e |n  = e

1.3.4

|n 

(1.73)

Unitary operators

Consider the inverse U−1 of a linear operator U. This inverse is defined by UU−1 = U−1 U = 1 Next, assume that the inverse U−1 of the linear operator U is the Hermitian conjugate of U: U−1 = U†

Ch001.tex

18

10/3/2011

21: 32

Page 18

BASIC CONCEPTS REQUIRED FOR QUANTUM MECHANICS

Then the operator U, which is said to be unitary, obeys the following relation: UU† = U† U = 1

if

U−1 = U†

(1.74)

As an example of unitary operator, consider the following expression for the linear operator U, which is an exponential of the Hermitian operator B times a real scalar λ times the imaginary number i: U = eiλB

with B = B†

λ = λ∗

i2 = −1

and

Then, using Eq. (1.44), the Hermitian conjugate of U appears to be given by U† = e−iλB On the other hand, it is obvious that the inverse of U is U−1 = e−iλB As a consequence, comparison of the two above equations shows that the Hermitian conjugate of U is its inverse, showing that U is unitary: U† = U−1

1.4

ALGEBRA OPERATORS

Here, we give some important results dealing with the algebra of operators, which are proved in Appendices 1–5. They are •

The commutator involving three noncommuting operators. A, B, and C: [A, BC] = [A, B]C + B[A, C]

•

(1.75)

The transformations 1 1 eξA Be−ξA = B + [A, B]ξ + [A, [A, B]]ξ 2 + [A, [A, [A, B]]]ξ 3 + . . . 2 3! (1.76) eξA F(B)e−ξA = F(eξA Be−ξA )

(1.77)

where ξ is a scalar and A and B are two independent linear operators that do not depend on ξ and that do not commute. •

the Glauber or Glauber–Weyl relation eξA eξB = e(A+B)ξ e+[A,B]ξ /2 with [A, [A, B]] = 0 2

and

[B, [A, B]] = 0 (1.78)

where ξ is a scalar and may be also written e(A+B)ξ = eξA eξB e−[A,B]ξ /2 = e(B+A)ξ

(1.79)

e(A+B)ξ = eξBξA e−[B,A]ξ /2 = e(B+A)ξ

(1.80)

2

2

Ch001.tex

10/3/2011

21: 32

Page 19

1.4 ALGEBRA OPERATORS

19

In the latter equations, the last terms on the right-hand side have been introduced in order to focus attention on the fact that e(B+A)ξ = e(A+B)ξ Now, we may summarize the most important results as follows: Basic equations for quantum mechanics Linear transformations and their Hermitian conjugates: A| = |

|A† = |

Hermitian operators A, unitary operators U, commutators: A = A†

U−1 = U†

with UU−1 = U−1 U = 1

[A, B] = AB − BA

Eigenvalue equations and their Hermitean conjugates: A|i  = Ai |i 

i |A† = i |A∗i

Eigenvalue equations of Hermitian operators and their Hermitean conjugates: A|i  = Ai |i  with

i |A = i |Ai  i |k  = δik and |k k | = 1

An important relation: |B† | = |B|∗ Invariance of the trace:   k |AB|k  = k |BA|k  even if

[A, B]  = 0

CH002.tex

10/3/2011

CHAPTER

21: 38

Page 21

2

BASIS FOR QUANTUM APPROACHES OF OSCILLATORS INTRODUCTION Using the mathematical basis treated in this chapter, it will be possible to discuss the quantum mechanics tools necessary for the study of the behavior of oscillators. We begin with an exposition of the postulates of quantum mechanics, which will be the purpose of Section 2.1. An important place will be given to the notions of quantum average values and to quantum fluctuations, allowing one to deduce from quantum principles the Heisenberg uncertainty relations according to which it is not possible to simultaneously know with arbitrary accuracy both the position and the momentum of any particle. In a subsequent section, some dynamic aspects will be developed allowing one both to determine the time dependence of the quantum average values and show that the Heisenberg uncertainty relations introduce a limit to the perfect knowledge assumed by classical mechanics. However, the quantum principles lead to the Ehrenfest equations, which nearly behave as the Newton equations, save that they are dealing with average values and not with exact ones, as for the classical equations. Related to these dynamic aspects, we shall prove the energy conservation, in a quantum averaged form, and the virial theorem relating the quantum average value of the kinetic and potential energies to the total energy, which holds also in classical mechanics. The last section will be devoted to some developments dealing with quantum concepts related to the connection between the position and the momentum, which will be used in Chapter 3 to relate quantum mechanics to wave mechanics.

2.1 OSCILLATOR QUANTIZATION AT THE HISTORICAL ORIGIN OF QUANTUM MECHANICS 2.1.1

Ultraviolet catastrophe

Measurements of thermal capacity of solids were discovered at the beginning of the twentieth century to be in contradiction with the principles of statistical physics Quantum Oscillators, First Edition. Olivier Henri-Rousseau and Paul Blaise. © 2011 John Wiley & Sons, Inc. Published 2011 by John Wiley & Sons, Inc.

21

CH002.tex

22

10/3/2011

21: 38

Page 22

BASIS FOR QUANTUM APPROACHES OF OSCILLATORS

Disagreement

U(ω)

Disagreement

0

500

1000

1500

Matter oscillator

2000 T (K)

0

1

(a) Figure 2.1

2

3

4

ω/1014 Hz

Light oscillator (b)

Contradiction between experiment (shaded areas) and classical prediction (lines).

based on classical mechanics: The experiments show that these thermal capacities are temperature dependent, whereas the theory assuming that they result from the partial derivative with respect to the temperature of the average oscillation energy of the atoms within the solids predicted that they ought to be constant, due to the equipartition theorem of statistical mechanics applied to classical mechanics, according to which each degree of freedom of vibration of the solid contributes the same energy amount kB T (where kB is the Boltzmann constant and T the absolute temperature). See, for instance, Fig. 2.1a. In addition, the study of the frequency distribution of the intensity of the electromagnetic radiations enclosed in a heated cavity at thermal equilibrium (black-body radiations) lead to the results that this intensity narrows to zero as the frequency increases, in utter contradiction with the classical statistical mechanics predictions (applied to Maxwell electromagnetic modes of vibration) by Rayleigh and Jeans, according to which the intensity ought to tend to infinity (the ultraviolet catastrophe). See Fig. 2.1b.

2.1.2

Planck, Einstein, and Bohr’s old quantum mechanics

To reconcile the ultraviolet catastrophe with physics, Planck (1858–1947) assumed that the walls of the black body responsible for the absorption and emission of ultraviolet light are made of small oscillators of various frequencies, the energy of which cannot vary continuously as in Newtonian mechanics, but is quantized, the energy levels En obeying En (matter oscillator) = nhν where n is an integer, ν the frequency of the microscopic oscillator, and h Planck’s constant. With this assumption of the oscillator energy quantization, Planck was able in 1901 to reproduce with great accuracy the experimental results. Moreover, some time later, Einstein proposed (1905) a theoretical interpretation of the photoelectric effect, a recent and unexplained laboratory result: It had been

CH002.tex

10/3/2011

2.1

21: 38

Page 23

OSCILLATOR QUANTIZATION AT THE HISTORICAL ORIGIN OF QUANTUM MECHANICS

23

discovered that an electron can be expelled from a material by a light radiation, when its frequency is greater than a threshold characteristic of the material, the kinetic energy of the emitted electron increasing linearly with the light frequency beyond the threshold. To interpret that Einstein assumed that light, considered at this time by the physicists as of wave nature, has also to be considered as consisting of a grain of light, the photon, the energy of which is proportional to the angular frequency ω of the light, the proportionality constant being that introduced by Planck in his theory. En (light oscillator) = nω

with

=

h 2π

A few years later, in 1913, Bohr (1885–1962), a Danish physicist, attacked the problem raised by the absorption and emission of light rays by hydrogen atoms. The frequencies of these lines, which are the same for both processes, were found by Balmer (1825–1898) to obey with a perfect precision an empirical formula, the Balmer formula, involving integer numbers. Bohr was able to theoretically reproduce the empirical Balmer formula by assuming that the angular momentum of the electron generated by its circular orbit motion around the proton is quantized, being an integer multiple of Planck’s constant already introduced in the Planck and in the Einstein theoretical approaches. Moreover, Bohr assumed that when the electron moves from one orbit to another, it performs that in a sudden and unrepresentative manner, by emitting or absorbing a quantum of light (photon) of frequency given by the absolute difference between the orbit’s energies divided by Planck’s constant. In addition, to link his theoretical approach with classical mechanics, Bohr introduced a correspondence principle, claiming that when the quantized energy levels of the electronic orbits are higher and higher, the transitions between successive energy levels involve a dynamics that approaches more and more closely the classical circular motion.

2.1.3

Heisenberg and matrix mechanics

All these works of Planck, Einstein, and Bohr called into question the continuous variation of the energy level of atoms since they assumed that energy may change only by small packets, the energy quanta involving the Planck constant. These works aroused passionate debates, some scientists thinking with Bohr that the classical mechanics of Newton would have to be rethought from top to bottom in order to deeply reflect the new realities at the scale of molecules, atoms, and their elementary constituents. Among these young physicists, Heisenberg (1901–1976) played a pioneering role, building during his thesis in 1924 a new theory. He focused on quantizing the energy of microscopic oscillators proposed by Planck. His ideas were primarily based on two kinds of square noncommutative matrices, one of which was intended to represent the position coordinate and the other one the conjugated momentum. Heisenberg completed this assumption by introducing Planck’s constant in these matrices. Heisenberg justified his assumption of noncommutative matrices representing position coordinates and momentum by the positivist postulate that it is impossible

CH002.tex

24

10/3/2011

21: 38

Page 24

BASIS FOR QUANTUM APPROACHES OF OSCILLATORS

on the atomic scale to measure the position of a particle without changing instantly ipso facto speed and therefore its momentum. Heisenberg was able, from his noncommutative matrices (recognized as such by Jordan), to find the formulas postulated by Planck for the quantization of the energy of small oscillators belonging to the atomic scale. This work may be regarded as the foundation stone of the new quantum mechanics.

2.1.4

De Broglie and wave mechanics

As seen above, Einstein introduced in his interpretation of the photoelectric effect the necessity to add corpuscular properties to the wave ones assumed for light, following the interference experiments of Young and others. This dual nature of light, Louis de Broglie (1892–1987) has extended it to matter, that is, all entities involving mass, which comprise the physical realities around us: At the same time Heisenberg was working on his thesis on the matrix mechanics, de Broglie, starting from intuitions of the Irish physicist Hamilton (1805–1865). proposed a new mechanism applying to the microscopic scale in which a wave is associated with the particle dynamics. In this new mechanics, the wavelength λ (de Broglie wavelength) of free particles (particles moving in a straight line in the absence of potential) is equal to Planck’s constant divided by the momentum p of the particles (de Broglie relation): λ=

h p

Hence, since the momentum is proportional to the product of mass times velocity, the de Broglie wavelength becomes smaller the greater the mass, so that it becomes negligible when going from the atomic and molecular scales to the human one and, a fortiori, to those of the planets and stars. In this wave mechanics, the corpuscular properties of matter are linked with the position coordinates, while the wave properties are linked to the momentum through the de Broglie wavelength. That is the origin of the term wave mechanics given to this new discipline of physics. One of the famous theoreticians of the time, the Austrian physicist Schrödinger (1887–1961), who initially despised the ideas of the young French physicist de Broglie, thereafter applied them to the hydrogen atom. By solving the partial differential equation governing in wave mechanics the electron behavior of the hydrogen atom, Schrödinger retrieved the results of Bohr concerning the empirical Balmer formula. Wave mechanics was soon experimentally confirmed by Davisson (1881–1958) and Germer (1896–1971) in connection with diffraction observations on crystals, allowing to verify the validity of the de Broglie relation. In addition the wave particle duality nature became evident via new experiments where particles having crossed separately a dispersion pattern, strike a screen by exhibit an interference pattern, thus suggesting that each isolated particle interferes with itself. This phenomenon was observed for light (photons) and also for material particles such as atoms.

CH002.tex

10/3/2011

21: 38

2.2

Page 25

QUANTUM MECHANICS POSTULATES AND NONCOMMUTATIVITY

25

2.2 QUANTUM MECHANICS POSTULATES AND NONCOMMUTATIVITY 2.2.1 The principles 2.2.1.1 First postulate At a given time, the physical state of a system is described by a ket |j (t) belonging to the state space, that is, to a vector space of infinite dimension involving complex scalar products. 2.2.1.2 Second postulate With each classical physical variable A is associated a linear operator A acting in the state space, which must be Hermitian (observable), and obeying, therefore, A = A† 2.2.1.3 Third postulate The possible measurements of an observable A are given by the eigenvalues An of this operator, that is A|n  = An |n  where |n  are the corresponding eigenkets of the eigenvectors of A. Owing to the Hermiticity of the observables, their eigenvalues are real: An = A∗n This constraint of Hermiticity on linear operators, which describe the physical variables, avoids the possibility of complex expressions involving an imaginary part in measurements of many physical variables. 2.2.1.4 Fourth postulate The transition of a system from any ket |n  to another |j  cannot be predicted in a deterministic way but only in a probabilistic one defined by a probability Pnj , which may be calculated from the squared modulus of the scalar product of the initial and final kets, that is, by Pnj = |j |n |2

(2.1)

2.2.1.5 Fifth postulate This postulate concerns situations where the eigenvalues are degenerate, which we shall not encounter here. Thus, in order to simplify, we do not give it here. 2.2.1.6 Sixth postulate There are two different equivalent ways to obtain the dynamics of a quantum system. In the first one, the kets and bras are time dependent and the operators are constants. This is the Schrödinger picture. In the second one, the kets and bras are constants, and it is the operators that are time dependent. The latter is the Heisenberg picture. The sixth postulate, in the Schrödinger picture, states that the kets describing a physical system evolve with time between two quantum jumps in a deterministic

CH002.tex

26

10/3/2011

21: 38

Page 26

BASIS FOR QUANTUM APPROACHES OF OSCILLATORS

way, which is given by the following equation named the time-dependent Schrödinger equation or more shortly the Schrödinger equation: i

∂ |j (t) = H|j (t) ∂t

with

i2 = −1

(2.2)

where H is the total quantum Hamiltonian describing the system, whereas  is the Planck constant divided by 2π. Note that in Eq. (2.2) the partial derivative with respect to time is sometimes replaced by a time derivative. However, as the ket may be affected by transformations other than that of time, for instance, translations of the origin (vide infra the translation operators), we prefer the partial derivative notation. 2.2.1.7 Seventh postulate The quantum operator A describing a classical physical variable A may be obtained as follows: 1.

Express the classical variable A in terms of the space variables Qk related to the different freedom degrees k of the system, and of their corresponding conjugate momentum Pk , that is, write A(Pk , Qk ).

2. Associate the Hermitian operators Qi and Pi , respectively, to each space variable Qk and to its corresponding conjugate momentum Pk , in order to pass from the classical expression A(Pk , Qk ) to the corresponding quantum Hermitian operator A(Pk , Qk ), that is, A(Pk , Qk ) → A(Pk , Qk ) 3.

Require that the Qk and Pk operators obey the commutation rule [Qk , Pl ] = iδkl

with

i2 = −1

(2.3)

where  is the Planck constant given by h 6.62 = × 10−34 J · S 2π 2π Some further information concerning the commutation rules are given in Section 18.5. The third and fourth postulates lead to the following important remarks: The third postulate leads one to distinguish, in the measurement of an observable, two different possibilities according to whether or not before any measurement of one of its observables, the system was in an eigenket of the measured operator. This postulate gives directly a response only in the specific situation where the system was in an eigenket of this last one. Owing to Eq. (2.3), noting that the basic physical variables P and Q do not commute, different Hermitian operators A(Pk , Qk ) and B(Pl , Ql ), both functions of P and Q, have no reasons to commute: =

[A(Pk , Qk ), B(Pl , Ql )]  = 0

(2.4)

To make clear the discussion, write the eigenvalue equations of these two Hermitian operators: A(Pk , Qk )|ν  = Aν |ν 

and

B(Pk , Qk )|μ  = Bμ |μ 

CH002.tex

10/3/2011

21: 38

2.2

Page 27

QUANTUM MECHANICS POSTULATES AND NONCOMMUTATIVITY

27

where |ν  and |μ  are, respectively, the eigenkets of A(Pk , Qk ) and B(Pk , Qk ), whereas Aν and Bμ are the corresponding eigenvalues. Of course, since these operators do not commute, they do not admit the same set of eigenvectors. Moreover, since they are Hermitian, each eigenket of one of these operators may be linearly expanded in the set of eigenkets of the other operator. For instance,  aνμ |ν  with aνμ = ν |μ  (2.5) |μ  = ν

Now, suppose that at an initial time the system is in one of the eigenstates |μ  of the B(Pk , Qk ) operator. Next, if a measurement of the Hermitian operator A(Pk , Qk ) is performed on this system, then, according to the third postulate, this measurement will yield, for instance, Aη of the different eigenvalues and Aν of the Hermitian operator A(Pk , Qk ). That implies that, after such a measurement, the system is now in the ket |η  corresponding to the eigenvalue Aη . It appears, therefore, that measurement of the operator A(Pk , Qk ) of the system, which was initially in the ket |μ , has induced a jump in the ket |η . Hence, according to the fourth postulate, this jump is not deterministic but occurs with probability Pμη = |μ |η |2 or, compare, Eq. (2.5), Pμη

 2     = aνμ ν |η    ν

so that due to the orthonormality of the eigenkets of a Hermitian operator A(Pk , Qk )  2     aνμ δην  = |aμη |2 Pμη =    ν Thus, the measurement of A(Pk , Qk ) has induced the abrupt change  aνμ |ν  → |η  ν

with the probability equal to the squared absolute value of the coefficients aμη of the expansion appearing on the left-hand side of this last equation. As a matter of fact, the measurement of A(Pk , Qk ) has induced a reduction of the left-hand-side expansion, which is called the wave packet reduction, for historical reasons related to the fact that in wave mechanics the |ν  may be related to different orthogonal wavefunctions (see the discussion in Chapter 3 dealing with wave mechanics).

2.2.2

Classical mechanics as special limit of quantum mechanics

Despite its very formal character, which is far from classical mechanics, quantum mechanics is not without a link with it. As we shall see, it is possible from the postulates of quantum mechanics, to demonstrate the following equations, named the Ehrenfest equations, which govern the dynamics of a system:   dQ(t) P(t) dP(t) ∂V (2.6) = and =− dt m dt ∂Q 

CH002.tex

28

10/3/2011

21: 38

Page 28

BASIS FOR QUANTUM APPROACHES OF OSCILLATORS

Here, Q(t) , P(t) , and (∂V/∂Q) are, respectively, the average values of the position, momentum, and potential when the system is in the quantum state characterized by the ket |. Now, these equations are very similar to Newton’s equations:  (t)  P d Q(t) = dt m

 (t) dP ∂V =−  dt ∂Q

and

(2.7)

Letters with arrow mean vectorial entities in classical mechanics, at the opposite of bold letters appearing above and meaning quantum mechanical operators. However, an important difference exists because the quantum Eqs. (2.6) govern average values, whereas the classical Eqs. (2.7) govern exact ones. Hence, if they are average values A , they indicate that dispersion of the possible values around the average values exists, which may be analyzed using the variance A, namely  (2.8) A = A2  − A2 where A2  is the average of the square of A. For the position and the momentum, the time-dependent average values are governed by Eqs. (2.6), whereas the corresponding variance are governed by the Heisenberg uncertainty relation (which will be demonstrated later). Figure 2.2 shows two situations occurring for the relative variance A/A, the left-hand-side showing a quantum behavior, whereas the right-hand-side exhibits classical behavior. (P(t)) (Q(t)) 

 2

(2.9)

The passage from the quantum mechanics to the classical mechanics occurs when (P(t)) →0 P(t)

(Q(t)) →0 Q(t)

and

(2.10)

When the size of the system is very small, of the order of the size of molecules or atoms or smaller, the quantum mechanics of Eqs. (2.6) holds. However, when this ΔA

ΔA ~1 〈A〉

P(A)

P(A)

ΔA ~0 〈A〉

ΔA

0

〈A〉 Figure 2.2

A

0

〈A〉

Quantum and classical relative variance A/A.

A

CH002.tex

10/3/2011

21: 38

2.2

Page 29

QUANTUM MECHANICS POSTULATES AND NONCOMMUTATIVITY

29

size is progressively increasing, the conditions (2.10) are more and more verified so that the quantum mechanics of Eqs. (2.6) transforms to classical mechanics (2.7). The specific physical behavior as the size of the system decreases is linked to the basic uncertainty characterizing the fundamental physical variables of small particles manifested via the following probability passage from any state of position to one of momentum, which may be the initial position and the final momentum, or vice versa, through the following relation, which will be demonstrated later |{P}|{Q}|2 =

1 2π

(2.11)

where |{Q} and |{P} are, respectively, the eigenkets of the position Q and momentum P operators defined, according to the third postulate by the continuous eigenvalue equation Q|{Q} = Q|{Q}

P|{P} = P|{P}

and

where Q and P are, respectively, the eigenvalues of Q and P, and thus the respective measured values of these operators, when the system is either in |{Q} or in |{P}, Eq. (2.11) implies therefore that, after a measurement of the position yielding Q, another measurement of the momentum may lead to all the possible values P, with the same probability, is given by 1 PQ→P = PP→Q = 2π The Heisenberg uncertainty relation (2.9) and the jump probability (2.11) are consequence of the fundamental commutator [Q, P] = i

(2.12)

Thus, the noncommutativity properties of observables, which are very general, play a fundamental role in the knowledge of the possible measurement of a physical variable. In order to appreciate the role played by the Hermitian operators in quantum mechanics, it is necessary to find the expression of the commutators [Q, F(P)] and [P, F(Q)], which are deeply linked to their behavior. In Appendix 5 are demonstrated some expressions dealing with commutators that are functions of P and Q, and that result from the basic commutator (2.12). They are the following: [Q, Pn ] = n(i)Pn−1 

∂F(P) [Q, F(P)] = (i) ∂P

(2.13)  (2.14)

[P, Qn ] = −n(i)Qn−1  [P, F(Q)] = −(i)

∂F(Q) ∂Q

 (2.15)

CH002.tex

30

10/3/2011

21: 38

Page 30

BASIS FOR QUANTUM APPROACHES OF OSCILLATORS

2.3 2.3.1

HEISENBERG UNCERTAINTY RELATIONS Mean values

Clearly, according to the third postulate, if a system is in a state |n , which is an eigenvector of some Hermitian operator A, the measurement of the physical variable associated to this operator is given by the corresponding real eigenvalue An of this operator, that is, A|n  = An |n 

with

A = A†

(2.16)

However, if the system is described by a state |k  that is not an eigenvector of the operator, we have seen that, according to the fourth postulate, there are as many possibilities to get measurements of the physical variable associated to A as there are eigenvalues of A. Then the only possibility for a measurement of A is an average value Ak given by Ak = k |A|k 

(2.17)

To show that Ak is an average value, use the closure relation of the eigenkets of the Hermitian operator A:  1= |n n | (2.18) n

Then insert it on the right-hand side of Eq. (2.17) just after A:

 |n n | |k  Ak = k |A n

This last expression reads in the more usual form on commuting the sum  k |A|n n |k  Ak = n

Again, according to the eigenvalue Eq. (2.16), this equation transforms to  k |An |n n |k  Ak = n

or, on commuting the eigenvalues An that are scalars  An k |n n |k  Ak = n

Moreover, using the fact that the two right-hand-side scalar products are complex conjugates, we have  Ak = An |k |n |2 (2.19) n

Finally, due to the fourth postulate, the right-hand-side squared modulus is the transition probability to pass from the ket |k  in which the system was initially before

CH002.tex

10/3/2011

21: 38

Page 31

2.3

HEISENBERG UNCERTAINTY RELATIONS

31

the measurement of A to the eigenket |n  of this operator A associated with the eigenvalue An to which the measurement of A has lead. |k |n |2 = Pkn

(2.20)

Thus, the left-hand side of Eq. (2.19), which is defined by Eq. (2.17), appears to be given by Ak =



Pkn An = k |A|k 

(2.21)

n

Examination of this last result shows that Ak has the properties of a statistical average value since it is the sum of the possible values of the observable A weighted by their corresponding probabilities.

2.3.2 Variation theorem It is now possible to prove the variation theorem. The sixth postulate attributes to the Hamiltonian a privileged role. Dealing with the Hamiltonian, there is, in quantum mechanics, a theorem that is of great interest concerning the energy of physical systems. Let us write the eigenvalue equation of the Hamiltonian H: H|i  = Ei |i  with

i |j  = δij

(2.22)

where Ei are the eigenvalues and |i  the corresponding eigenvectors. Now, consider the average value over any ket |l  of the difference between the Hamiltonian and the lowest eigenvalue E0 : l | (H − E0 ) |l  = l | H| l  − E0 l |l 

(2.23)

Next, assume that the ket |l  is given by the following expansion over the eigenkets |i  of the Hamiltonian, that is,   |l  = ajl |j  and l | = ali i | j

i

Then, Eq. (2.23) becomes l | (H − E0 ) |l  =

 i

ajl ali {i | H | j  − E0 i |j }

j

Furthermore, due to the eigenvalue equation (2.22) by orthogonality properties we have  ajl ali (Ej − E0 )δij l | (H − E0 )|l  = i

and thus l | (H − E0 )|l  =

j

 j

|ajl |2 (Ej − E0 )

(2.24)

CH002.tex

32

10/3/2011

21: 38

Page 32

BASIS FOR QUANTUM APPROACHES OF OSCILLATORS

Now, observe that, since E0 is the lowest eigenvalue, the right-hand-side differences are positive in the same way as the squared modulus of the expansion coefficients, that is, (Ej − E0 )  0

|ajl |2  0

and

Thus, we have from the left-hand side of Eq. (2.24), the following inequality: l |(H − E0 )|l   0 Since E0 is a scalar, one then obtains the following fundamental result: l |H|l   E0

(2.25)

Hence, the average value of the Hamiltonian performed over any one ket cannot be smaller than the lowest energy E0 . That gives the possibility to approach E0 by variational methods if it is not possible to solve exactly the eigenvalue equation (2.22) of the Hamiltonian.

2.3.3 Variance Now, observe that in probability theory and statistics, the variance A of a random variable is a measurement of the statistical dispersion averaging the squared distance of its possible values from the mean value A. Start from the variance (2.8):  Ak = A2 k − A2k (2.26) In this last equation, the second term under the square root is given by Eq. (2.21). In order to get the first one, we may begin by using the definition (2.17) of the average of some operator, by taking A2 in place of A: A2 k = k |A2 |k  which may also be written A2 k = k |AA|k 

(2.27)

Again, write the eigenvalue equation of Hermitian operators and the corresponding closure relation:  A|n  = An |n  and |n n | = 1 (2.28) n

Then, introduce in Eq. (2.27) between A and |k  this last closure relation   2 A k = k |AA |n n | |k  n

Using the eigenvalue equation appearing in (2.28), one obtains  A2 k = k |AAn |n n |k  n

Next, commuting the scalar An , this equation becomes  A2 k = An k |A|n n |k  n

CH002.tex

10/3/2011

21: 38

Page 33

2.3

HEISENBERG UNCERTAINTY RELATIONS

33

Again using in turn the eigenvalue equation (2.28), one finds  A2 k = An k |An |n n |k  n

or, An being a scalar A2 k =



A2n |k |n |2

n

Finally, using the fourth postulate given in the present context by Eq. (2.20) leads to  A2 k = Pkn A2n with Pkn = |k |n |2 n

Thus, the variance (2.26) takes the form

 2      2 Pkn An − Pkn An Ak = n

2.3.4

n

Product of two variances

2.3.4.1 Variances of two different operators before and after some shift Consider two Hermitian operators A and B, the commutator of which is obeying [A, B] = iC with

i2 = −1

(2.29)

and A = A†

B = B†

C = C†

Again, consider the average values of these operators A and B, respectively, calculated on some ket | that we shall suppose normalized: A = |A|

and

B = |B|

with

| = 1

(2.30)

Next, consider the following transformed operators: ˜ = {A − A } A

B˜ = {B − B }

(2.31)

with average values on | ˜  = |A| ˜ A

and

˜  = |B| ˜ B

(2.32)

Now, write explicitly the first of the average values (2.32), using the first equation of (2.31): ˜  = |{A − A }| A That gives ˜  = A − A | = {A − A } = 0 A

(2.33)

˜  where the normalization of | has been used. In like manner one may find that B is zero. Hence, one may write ˜ =0 A

and

˜ =0 B

(2.34)

CH002.tex

34

10/3/2011

21: 38

Page 34

BASIS FOR QUANTUM APPROACHES OF OSCILLATORS

Now, consider the corresponding squares of the variances concerning the two operators (2.31), that is, ˜ 2 = {A ˜ 2  − A ˜ 2 } A

and

˜ 2 = {B˜ 2  − B ˜ 2 } B

and

˜ 2  B˜ 2 = B

or, owing to Eq. (2.34) ˜ 2  ˜ 2 = A A

(2.35)

Furthermore, the quantum averages over | of the operators (2.31) are ˜ 2  = |A ˜ 2 | A

and

B˜ 2  = |B˜ 2 |

and

B˜ 2 = |B˜ 2 |

(2.36)

Thus, Eq. (2.35) reads ˜ 2 | ˜ 2 = |A A

Next, owing to the first equation of (2.31), the first equation of (2.36) transforms, after expanding the squared expression, into ˜ 2  = |{A2 + A2 − 2AA }| A or ˜ 2  = |A2 | + |A2 | − 2|A|A A Since | is normalized and due to the first equation of (2.30), we then have ˜ 2  = A2  + A2 − 2A A A or ˜ 2  = A2  − A2 A

(2.37)

Now, observe that the right-hand side of Eq. (2.37) is the dispersion of the operator A averaged on the ket |: A2  − A2 = A2 Thus, Eq. (2.37) yields ˜ 2  = A2 A ˜ as for A, ˜ one obtains, Thus, compare Eq. (2.35), and working in the same way for B respectively, ˜ 2 = A2 A  

and

˜ 2 = B2 B  

(2.38)

which we shall use later on. 2.3.4.2 Product of variances of A and B and Heisenberg uncertainty rela˜ and B, ˜ over |. In view of tions Now, consider the product of the variances of A Eqs. (2.30) and (2.35), it is ˜ 2 B ˜ 2 = |A ˜ 2 ||B ˜ 2 | A

CH002.tex

10/3/2011

21: 38

Page 35

2.3

HEISENBERG UNCERTAINTY RELATIONS

35

or ˜ 2 = (|A)( ˜ A|)(| ˜ ˜ B|) ˜ ˜ 2 B B)( A

(2.39)

˜ and B˜ transforms, respectively, Next, observe that the linear action of the operators A the ket | into the new kets | and |ξ according to ˜ A| = |

and

˜ B| = |ξ

(2.40)

The Hermitian conjugates of these two linear transformations are ˜ = | |A

and

˜ = ξ| |B

(2.41)

Thus, Eq. (2.39) becomes ˜ 2 B ˜ 2 = |ξ|ξ A

(2.42)

Now, observe that the Schwarz inequality (1.25) stipulates that |ξ|ξ  ξ||ξ Hence, the product of uncertainties (2.42) transforms to the following inequality: ˜ 2 B ˜ 2  ξ||ξ A

(2.43)

Again, in view of the linear transformations (2.40) and (2.41), the scalar products involved on the right-hand side of this last inequality are given by ˜ B| ˜ |ξ = |A

and

˜ A| ˜ ξ| = |B

Thus, the product of uncertainties (2.43) transforms to ˜ 2 B ˜ 2  |A ˜ B|| ˜ ˜ A| ˜ A B

(2.44)

˜ B˜ nor B˜ A ˜ are Hermitian, Moreover, keeping in mind Eq. (1.60), and since neither A it is suitable to express these products in terms of symmetric and antisymmetric combinations according to ˜B ˜ = 1 (A ˜B ˜ +B ˜ A) ˜ + 1 (A ˜ B˜ − B ˜ A) ˜ A 2 2

(2.45)

˜ and B: ˜ Remark that the antisymmetric part is just the commutator of A ˜B ˜ −B ˜ A) ˜ = [A, ˜ B] ˜ (A Now, this commutator involving the transformed operators may be expressed in terms of the initial ones using Eq. (2.31), so that ˜ B] ˜ = [(A − A ), (B − B )] [A, Then, since the average values involved in this last equation are scalars, the commutator of the transformed operators appears to be that of the nontransformed ones: ˜ B] ˜ = [A, B] [A, Thus, Eq. (2.45) transforms to ˜B ˜ = 1 (A ˜B ˜ +B ˜ A) ˜ + 1 [A, B] A 2 2

CH002.tex

36

10/3/2011

21: 38

Page 36

BASIS FOR QUANTUM APPROACHES OF OSCILLATORS

Again, owing to the assumption (2.29) we have performed for the commutators of the initial operators A and B, this last equation leads to ˜ B˜ = 1 {A ˜B ˜ + B˜ A ˜ + iC} A 2

(2.46)

˜ A, ˜ which reads Now, consider B ˜A ˜ −A ˜ B) ˜ = [B, ˜ A] ˜ = −[A, B] (B Then, due to Eq. (2.29), we have ˜A ˜ = 1 {A ˜B ˜ + B˜ A ˜ − iC} B 2

(2.47)

As a consequence of Eqs. (2.46) and (2.47), Eq. (2.44) becomes ˜ 2 B ˜ 2  1 |{A ˜B ˜ +B ˜A ˜ + iC}||{A ˜B ˜ + B˜ A ˜ − iC}| A 4 Thus ˜ 2 B ˜ 2  1 (|(A ˜B ˜ + B˜ A)| ˜ ˜ B˜ + B ˜ A)| ˜ A + i|C|)(|(A − i|C|) 4 or 2 ˜ 2 B ˜ 2  1 {|(A ˜B ˜ + B˜ A)| ˜ A + |C|2 } 4

(2.48)

Now, as it appears by inspection of this last inequality, each member of the righthand-side, however small it may be, cannot be negative since it is a squared average value. Moreover, the inequality is also satisfied when one substracts from the smallest ˜B ˜ and B˜ A. ˜ Hence, right-hand-side term, its first squared term involving the products A if the inequality (2.48) is satisfied, the following one will be a fortiori satisfied: ˜ 2 B ˜ 2  1 |C|2 A 4 Hence, owing to Eq. (2.38), it appears that the same inequality for the dispersions dealing with the initial operators A and B exists, so that it reads A2 B2  41 |C|2 or, in view of Eq. (2.29) A2 B2  −i 41 |[A, B]|2

(2.49)

Now, apply the inequality (2.49) to the coordinate operator Q and its conjugate momentum P. Then take A=Q

B=P

and

[A, B] = [Q, P] = i

Besides, due to Eq. (2.29), that is, iC = [Q, P] = i Then, the inequality (2.49) takes the form 2 P Q2

1  ||2 = 4

 2  2

CH002.tex

10/3/2011

21: 38

Page 37

2.4

SCHRÖDINGER PICTURE DYNAMICS

37

so that the product of variances of the coordinate Q and its conjugate momentum P appears, whatever the ket | describing the system P Q 

 2

(2.50)

That is the Heisenberg uncertainty relation. Of course, this relation holds for the three Cartesian coordinates of some particle, so that (Px ) (Qx ) 

 2

(Py ) (Qy ) 

 2

(Pz ) (Qz ) 

 2

An important consequence of these uncertainty relations is that the trajectory, which is fundamental in classical mechanics, has no meaning in quantum mechanics. The reason is that, to define the trajectory of some particle, it is necessary to know exactly both its position and momentum at all times. This impossibility of a precise trajectory indicates that two particles of the same kind, such as, for instance, two electrons, or two protons, or two hydrogen atoms, are indistinguishable because the only possibility to distinguish them would be their individual trajectories, which is impossible because of the uncertainty relations.

2.4

SCHRÖDINGER PICTURE DYNAMICS

Now, we shall consider some dynamic behaviors appearing in quantum mechanics as a consequence of the Schrödinger equation appearing in the sixth postulate. Recall that according to this equation, the kets and the corresponding bras are time dependent, whereas the operators are constant. Such a time description in which kets and bras are time dependent whereas the operators are constant is called the Schrödinger picture (SP) in order to differentiate it from another description named Heisenberg picture (HP) in which the operators are changing with time whereas the kets and bras remain constant. We shall first show that the Schrödinger equation preserves the conservation of the norm that is required from the physical viewpoint. Then, we shall demonstrate some fundamental dynamic equations, and, finally, two theorems, one of which is the Ehrenfest theorem, which resembles the basic Newtonian equations of classical mechanics, the only difference being that the Ehrenfest theorem governs average values instead of exact ones.

CH002.tex

38

10/3/2011

21: 38

Page 38

BASIS FOR QUANTUM APPROACHES OF OSCILLATORS

2.4.1 2.4.1.1

Norm conservation and average values time dependence Norm conservation

Consider the Schrödinger equation  ∂|(t) = H|(t) i ∂t 

(2.51)

where H is the Hamiltonian operator, which is, of course, Hermitian. If it is normalized, the norm of the ket|(t ◦ ) at time t ◦ is (t ◦ )|(t ◦ ) = 1 Of course, if the norm has to be conserved, it must be given at any time t  = t ◦ by (t)|(t) = 1 We shall show that this last equation is in agreement with the Schrödinger equation. For this purpose, we write explicitly the time derivative of the norm       ∂(t)| ∂|(t) ∂(t)|(t) = |(t) + (t)| (2.52) ∂t ∂t ∂t To calculate the time derivative of the bra involved on the first right-hand-side term of this last equation, we consider the Hermitian conjugate of Eq. (2.51)   ∂(t)| = (t)|H† −i ∂t Then, since the Hamiltonian is Hermitian, that is, H† = H, this last equation becomes   ∂(t)| = (t)|H (2.53) −i ∂t As a consequence of Eqs. (2.51) and (2.53), the time derivative of the norm (2.52) becomes   1 1 ∂(t)|(t) = − (t)|H|(t) + (t)|H|(t) ∂t i i This last result simplifies to 

∂(t)|(t) ∂t

 =0

showing, as required, that the norm is conserved along the time.

2.4.2 Time evolution of operator average value 2.4.2.1 General expression We shall now consider how the average value of some operator A calculated over any ket |(t) evolves, which is time dependent because of the Schrödinger equation. In the Schrödinger time-dependent picture, any operator A does not depend on time so that the time derivative of the average value of A over |(t) is       ∂(t)|A|(t) ∂(t)| ∂|(t) = A|(t) + (t)|A (2.54) ∂t ∂t ∂t

CH002.tex

10/3/2011

21: 38

Page 39

2.4

SCHRÖDINGER PICTURE DYNAMICS

Using Eqs. (2.51) and (2.53), this equation transforms to   ∂(t)|A|(t) 1 1 = − (t)|HA|(t) + (t)|AH|(t) ∂t i i or, in term of the commutator of H and A   ∂(t)|A|(t) i = (t)|[H, A]|(t) ∂t  which may be written in the compact form   ∂A(t) i = [H, A] ∂t 

39

(2.55)

(2.56)

with A(t) ≡ (t)|A|(t)

[H, A] = (t)|[H, A]|(t)

(2.57)

We remark that the notation A(t) does not imply that A depends on time but only means that the average value A(t) of A depends on time. Besides, observe that, in the Schrödinger time-dependent picture, some physical systems that have to be studied may appear quantum mechanically for one part and classically for another one. In such systems, which are said to be hemiquantal, there is then the possibility for any operator A to present a time dependence through its classical part. Then, Eq. (2.55) has to be generalized into     ∂(t)|A(t)|(t) i ∂A(t) = (t)|[H, A(t)]|(t) + (t)| |(t) (2.58) ∂t  ∂t 2.4.2.2 Conservation of the total energy and exchange of energies In the special situation where the operator is the Hamiltonian, and what may be the ket |(t) describing the system, Eq. (2.56) reads   ∂H i (2.59) = [H, H] = 0 ∂t  Hence, the average value of the total Hamiltonian, that is, the total energy, remains constant whichever ket |(t) describes the system. However, if the total energy is conserved, it is not true for the energies of subsystems from which any physical system is built up. Suppose, for instance, that the total Hamiltonian is the sum of two Hamiltonians, which do not mutually commute: H = H1 + H2

with

[H1 , H2 ]  = 0

Then, the commutators of H1 and H2 with H are [H1 , H] = [H1 , H2 ]

and

[H2 , H] = [H2 , H1 ] = −[H1 , H2 ]

(2.60)

Thus, due to Eqs. (2.56) and (2.60), the time dependences of the averages of the Hamiltonians of the two subsystems obey   ∂H1 (t) i (2.61) = [H2 , H1 ] ∂t 

CH002.tex

40

10/3/2011

21: 38

Page 40

BASIS FOR QUANTUM APPROACHES OF OSCILLATORS



∂H2 (t) ∂t



i = − [H2 , H1 ] 

(2.62)

We emphasize that in these last equations, they are the average values of H1 and H2 , which depend on time through (t), although the operators H1 and H2 do not depend on time. Equations (2.61) and (2.62) show that the energy moves back and forth between the two subsystems according to     ∂H2 (t) ∂H1 (t) =− ∂t ∂t in such a way as their sum remains constant according to Eq. (2.59). Of course, if one considers, respectively, in place of H1 and H2 the kinetic and potential energies, T and V of the system, the Eqs. (2.61) and (2.62) become   i ∂T(t) = [V, T] ∂t    ∂V(t) i = − [V, T] ∂t  so that the kinetic and potential energies exchange themselves with time according to     ∂V(t) ∂T(t) =− ∂t ∂t 2.4.2.3 Stationary states By definition, a stationary state is an eigenstate of the Hamiltonian, that is, it obeys the eigenvalue equation H|k (t) = Ek |k (t) The time-dependent Schrödinger equation is   ∂|k (t) = H|k (t) i ∂t For a stationary state, it reads



∂|k (t) i ∂t

 = Ek |k (t)

so that, by integration |k (t) = |k (0)e−iEk t/

(2.63)

Now, consider the average value of any operator over a stationary state. At an initial time it is A(0)k = k (0)|A|k (0)

(2.64)

A(t)k = k (t)|A|k (t)

(2.65)

At time t, it is given by

Then, owing to Eq. (2.63) and to its Hermitian conjugate, one has A(t)k = eiEk t/ k (0)|A|k (0)e−iEk t/

CH002.tex

10/3/2011

21: 38

Page 41

2.4

SCHRÖDINGER PICTURE DYNAMICS

41

and, thus, after simplification A(t)k = k (0)|A|k (0) Comparison of Eqs. (2.64) and (2.66) shows that   ∂A(t)k = 0 for any stationary state ∂t

2.4.3

(2.66)

(2.67)

Ehrenfest equations

Now, we are able to demonstrate the Ehrenfest equations governing the dynamics of the operators Q and P. Applying Eq. (2.56), one obtains, respectively,   ∂Q(t)k i (2.68) = [H, Q]k ∂t  

∂P(t)k ∂t

 =

i [H, P]k 

(2.69)

When the system involves only forces that are the derivative of a potential, the Hamiltonian H(P, Q) is as above the sum of the kinetic T(P) and potential V(Q) operators, the first one depending on P and the last one on Q: H(P, Q) = T(P) + V(Q) For a single particle, the kinetic operator is simply P2 2m Of course, the commutators of the kinetic momentum operators and that of the potential and coordinate operators, are, respectively, zero, that is T(P) =

[T(P), P] = [V(Q), Q] = 0 Thus, the commutators of the Hamiltonian with the coordinate and momentum operators are, respectively, [H, Q] =

1 2 [P , Q] 2m

[H, P] = [V(Q), P]

(2.70) (2.71)

Besides, owing to Eqs. (2.14) and (2.15), the commutators appearing on the right-hand sides of these two last equations are, respectively, [P2 , Q] = −2iP  [V(Q), P] = i

∂V ∂Q



Then, using for the two commutators (2.70) and (2.71), these two last equations, and introducing them into Eqs. (2.68) and (2.69), one obtains the final results, which

CH002.tex

42

10/3/2011

21: 38

Page 42

BASIS FOR QUANTUM APPROACHES OF OSCILLATORS

are known as the Ehrenfest equations, and which hold whatever the ket |k (t) considered for the calculation:   dQ(t)k P(t)k = m (2.72) dt 

∂ P(t)k = − ∂t



∂V ∂Q

(2.73) k

Thus, the first Ehrenfest equation looks like the Newton equation defining the momentum in terms of the velocity, whereas the second one looks like that relating the time derivative of the momentum (i.e., the acceleration) to the gradient of the potential (i.e., the force). However, the ket |k (t) considered for the calculation can never be simultaneously an eigenket of P and Q because [Q, P] = i thus, the uncertainty relations must be retained so that the Ehrenfest equations [(2.72) and (2.73)] have always to be considered mindful of the Heisenberg uncertainty relation: (P)k (Q)k 

 2

2.4.4 Virial theorem 2.4.4.1 Demonstration of the virial theorem Now, we shall prove the virial theorem, which relates the average values of the kinetic and potential operators, when the averages are performed over stationary states |k , that is, eigenstates of the Hamiltonian H obeying the eigenvalue equation H|k  = Ek |k 

(2.74)

Apply Eq. (2.56) to the product QP of the coordinate and momentum operators Q and P. Hence   ∂QPk i (2.75) = [QP, H]k ∂t The Hamiltonian H may be written, as above, as the sum of the kinetic T and potential V operators, the first only a function of P and the latter of Q. H = T(P) + V(Q) Of course, as above, the following commutators are zero: [T(P), P] = [V(Q), Q] = 0

(2.76)

Thus, the commutator appearing on the right-hand side of Eq. (2.75) reads [QP, H] = [QP, T(P)] + [QP, V(Q)]

(2.77)

CH002.tex

10/3/2011

21: 38

Page 43

2.4

SCHRÖDINGER PICTURE DYNAMICS

43

For a single particle, the kinetic operator is 1 2 (2.78) P 2m Hence, the first commutator appearing on the right-hand side of Eq. (2.77) is T(P) =

1 (2.79) [QP, P2 ] 2m The commutator appearing on the right member of this last equation may be written [QP, T(P)] =

[QP, P2 ] = (QP2 − P2 Q)P or [QP, P2 ] = [Q, P2 ]P Thus, in view of Eq. (2.13), it transforms to [QP, P2 ] = (i)2P2 Hence, the commutator (2.79) becomes P2 (2.80) m Now, consider the second commutator appearing on the right-hand side of Eq. (2.77): [QP, T(P)] = i

[QP, V(Q)] = QPV(Q) − V(Q)QP which, since Q commutes with V(Q), transforms to [QP, V(Q)] = QPV(Q) − QV(Q)P so that Eq. (2.76) reads [QP, V(Q)] = Q[P, V(Q)] Then, using Eq. (2.15), we have



∂V [QP, V(Q)] = −(i)Q ∂Q

 (2.81)

Now, using Eqs. (2.80) and (2.81) the commutator (2.77) appears to be given by    1 2 ∂V [QP, H] = (i) P −Q m ∂Q Moreover, after averaging over |k  one obtains   

1 2 ∂V [QP, H]k = (i) P k − Q m ∂Q k Finally, using this result into Eq. (2.75), we have       2 ∂Q Pk ∂V P − Q =2 ∂t 2m k ∂Q k

(2.82)

CH002.tex

44

10/3/2011

21: 38

Page 44

BASIS FOR QUANTUM APPROACHES OF OSCILLATORS

When the ket, over which the average value is performed, is stationary, Eq. (2.67) holds, so that the time dependence of the average value of the correlation between Q and P is zero:   ∂Q Pk =0 ∂t Hence, for stationary situations, Eq. (2.82) simplifies to     2 ∂V P = Q 2 2m k ∂Q k

(2.83)

Observe that the gradient of the potential may be written as a force F according to   ∂V = −F ∂Q so that, Eq. (2.83) yields 

P2 2 2m

 = −Q Fk

(2.84)

k

This equation may be generalized for many degrees of freedom j. We have  2 N N   Pj 2 = − Qj Fj k 2m j=1

k

j=1

2.4.4.2 Applications of the virial theorem 2.4.4.2.1 Systems involving harmonic potential Now, apply Eq. (2.83) to a quantum harmonic oscillator where the potential obeys V(Q) = 21 kQ2

(2.85)

where k is the force constant of the potential, which is a scalar. Then, deriving Eq. (2.85) leads to   ∂V = kQ ∂Q Besides, multiplying both terms by Q gives   ∂V = kQ2 Q ∂Q Again, averaging over the ket |k  and in view of Eq. (2.85),    ∂V = 2V(Q)k Q ∂Q k At last, owing to Eqs. (2.78) and (2.83), Eq. (2.86) yields Tk = V(Q)k

(2.86)

CH002.tex

10/3/2011

21: 38

Page 45

2.5

POSITION OR MOMENTUM TRANSLATION OPERATORS

45

On the other hand, the average value of the Hamiltonian is the sum of the kinetic and potential operators, that is, Hk = Tk + V(Q)k

(2.87)

However, since the average value of the Hamiltonian is performed over one of its eigenstates obeying Eq. (2.74), this is just the corresponding eigenvalue Ek so that Eq. (2.87) gives Ek = Tk + V(Q)k

(2.88)

Hence, one may determine the average value of the kinetic and potential operators from the value of the corresponding energy levels via Tk =

Ek = V(Q)k 2

(2.89)

2.4.4.2.2 Systems involving Coulomb potential Now consider, as a second example, a Coulomb potential involving two electrical charges q and q obeying V(Q) = −K

1 Q

with

K=

qq

4πε◦

(2.90)

where ε◦ is the vacuum permittivity, which is a scalar. Then, after deriving V with respect to Q and rearranging, it reads   ∂V 1 Q =K ∂Q Q Furthermore, the quantum average over |k  leads, by aid of Eq. (2.90), to    ∂V Q = −V(Q)k ∂Q k Now, with Eq. (2.78), the virial theorem (2.83), takes the form 2Tk = −V(Q)k Of course, since Eqs. (2.87) and (2.88) continue to apply, one may obtain from the expression for the energy levels the corresponding average values of the potential and kinetic operators by aid of V(Q)k = 2Ek

and

Tk = −Ek

2.5 POSITION OR MOMENTUM TRANSLATION OPERATORS 2.5.1 Eigenvalue equations of the position and momentum operators Consider the eigenvalue equation of the coordinate operator Q: Q|{Q} = Q|{Q}

(2.91)

CH002.tex

46

10/3/2011

21: 38

Page 46

BASIS FOR QUANTUM APPROACHES OF OSCILLATORS

The meaning of this eigenvalue equation is that when a system is in an eigenket |{Q}1 of the coordinate operator Q, the measurement of its position is given by the corresponding eigenvalue Q. Of course, since the Q operator is Hermitian, the Hermitian conjugate of Eq. (2.91) is {Q}|Q = {Q}|Q Now, observe that the possible measurements of the Q coordinate are continuous. Thus, the orthormality properties of the Q operator involving two different kets |{Q} and |{Q } must be written according to this continuous property. Hence {Q}|{Q } = δ(Q − Q )

(2.92)

Furthermore, the eigenvectors of the Hermitian operator Q form a basis that must be continuous, owing to this continuity of Q. Thus, the usual closure relation (1.17) built up on the eigenvectors, must be replaced by a new one where an infinite integral takes the place of the sum over the eigenkets. That leads to +∞ |{Q}{Q}| dQ = 1 −∞

In a similar way, we may write the eigenvalue equation of the momentum operator P and its Hermitian conjugate P|{P} = P|{P}

and

{P}|P = {P}|P

(2.93)

Here, |{P}2 is an eigenket of the momentum operator P with the eigenvalue P. Besides, owing to the continuity of the eigenvalues of the operator P, that is, of its possible measured values, the orthonormality of the eigenkets of P and the closure relation are similar to those dealing with Q, that is,



{P}|{P } = δ(P − P )

and

+∞ |{P}{P}| dP = 1

(2.94)

−∞

In the following, we shall show that the scalar product of any eigenket of the position operator Q by some eigenket of its momentum conjugate P is the same irrespective of the corresponding eigenvalues Q and P:   1 iPQ {P}|{Q} = √ exp −  2π which is consistent with the Heisenberg uncertainty relation (P) (Q) 

 2

We shall use for the eigenkets of Q, the notation |{Q} in place of the usual one |Q in order to make clearer some equations (see later).

1

In a similar way we shall use for the eigenkets of P, the notation |{P} in place of the usual one |P in order to make clearer some equations (see later).

2

CH002.tex

10/3/2011

21: 38

Page 47

2.5

POSITION OR MOMENTUM TRANSLATION OPERATORS

47

and with the basic postulate commutator [Q, P] = i Now, one has to get the expression of the unitary operators, allowing one to translate the origin of the position and momentum operators.

2.5.2

Position operator translation

First, consider the following linear operator: A(P, Q◦ )

≡

A(Q◦ )



iQ◦ P = exp − 

 (2.95)

where Q◦ is a real scalar having the dimension of a length, and P the momentum operator, conjugate to the position operator Q. Its Hermitian conjugate is  ◦ † iQ P ◦ † A(P, Q ) = exp  Since P is Hermitian, that is, P = P† , this equation transforms to  ◦  iQ P A(P, Q◦ )† = exp 

(2.96)

Thus A(P, Q◦ )† = A(P, Q◦ )−1

(2.97)

Now, the operator A(P, Q◦ ) is unitary, so that A(P, Q◦ )† A(P, Q◦ ) = 1

(2.98)

Now, calculate the commutator of the operator (2.95) with Q. Then, since A(Q◦ , P) is a function of P, in view of Eq. (2.14) it takes the form   ∂A(P, Q◦ ) ◦ (2.99) [Q, A(P, Q )] = i ∂P The right-hand side of this last equation may be obtained differentiating Eq. (2.95) to give     i ∂A(P, Q◦ ) =− Q◦ A(P, Q◦ ) ∂P  As a consequence, Eq. (2.99) becomes [Q, A(P, Q◦ )] = Q◦ A(P, Q◦ )

(2.100)

Next, writing explicitly the left-hand side of Eq. (2.100) yields QA(Q◦ ) = A(P, Q◦ )Q + Q◦ A(P, Q◦ )

(2.101)

CH002.tex

48

10/3/2011

21: 38

Page 48

BASIS FOR QUANTUM APPROACHES OF OSCILLATORS

Again, premultiply each member of Eq. (2.101) by the inverse of A, that is, A(P, Q◦ )−1 QA(P, Q◦ ) = A(P, Q◦ )−1 A(P, Q◦ )Q + Q◦ A(P, Q◦ )−1 A(P, Q◦ ) Then, after simplifying, by aid of Eqs. (2.97) and (2.98), this last expression reduces to A(P, Q◦ )−1 QA(P, Q◦ ) = Q + Q◦

(2.102)

Thus, Eq. (2.102), which is called a canonical transformation on the coordinate operator Q, translates the origin of Q by the scalar amount Q◦ . Furthermore, for an infinitesimal scalar displacement dQ◦ , Eq. (2.102) transforms to A(P, dQ◦ )−1 QA(P, dQ◦ ) = Q + dQ◦

2.5.3

(2.103)

Momentum operator translation

Now, consider the linear operator B(Q, P◦ ): B(Q, P◦ ) = exp



iP◦ Q 

 (2.104)

where P◦ is a scalar having the dimension of a momentum and Q being the Hermitian coordinate operator. The inspection of its expression shows that the operator B(P◦ ) is unitary since its inverse B(Q, P◦ )−1 is equal to its Hermitian conjugate B(Q, P◦ )† : B(Q, P◦ )† = B(Q, P◦ )−1 Calculate the commutator of this operator with the momentum operator P. Since B(P◦ , Q) is a function of Q, one may use Eq. (2.15), which leads to   ∂B(Q, P◦ ) [P, B(Q, P◦ )] = −i ∂Q Differentiating Eq. (2.104), and after identification, one obtains [P, B(Q, P◦ )] = P◦ B(Q, P◦ )

(2.105)

Then, writing explicitly the commutator, Eq. (2.105) reads PB(Q, P◦ ) = B(Q, P◦ )P + P◦ B(Q, P◦ )

(2.106)

Moreover, premultiply this equation by the inverse of B to get B(Q, P◦ )−1 PB(Q, P◦ ) = B(Q, P◦ )−1 B(Q, P◦ )P + B(Q, P◦ )−1 P◦ B(Q, P◦ ) (2.107) On simplification, this expression reduces to B(Q, P◦ )−1 PB(Q, P◦ ) = P + P◦

(2.108)

Clearly, this canonical transformation allows to translate P by the scalar amount P◦ .

CH002.tex

10/3/2011

21: 38

Page 49

2.5

2.5.4

POSITION OR MOMENTUM TRANSLATION OPERATORS

49

Quantum Galilean transformation

One may define the Galilean transformation operator according to   i S(v) = exp (mvQ − Pvt) 

(2.109)

where v is the scalar velocity. Observe that this operator is Hermitian since   i † S (v) = exp − (mvQ − Pvt) = S−1 (v)  Using the Glauber theorem (1.78), the operator (2.109) and its inverse take, respectively, the forms     i i S(v) = exp mvQ exp − Pvt eζ       i i S−1 (v) = exp Pvt exp − mvQ e−ζ   with

  1 i i ζ=− mvQ, − Pv 2  

Next, perform the following transformation on the position coordinate according to         i i i i −1 S(v) QS(v) = exp Pvt exp − mvQ Q exp mvQ exp − Pvt     Hence     i i −1 S(v) QS(v) = exp Pvt Q exp − Pvt (2.110)   Next, taking vt in place of P◦ , and using Eq. (2.108), with the aid of Eq. (2.104), Eq. (2.110) reads S(v)−1 QS(v) = Q−vt

(2.111)

On the other hand, the transformation on the P coordinate involving the unitary operator (2.109) takes the form         i i i i −1 S(v) PS(v) = exp Pvt exp − mvQ P exp mvQ exp − Pvt     Again, taking mv in place of Q◦ , and then using Eqs. (2.102) and (2.95), yields     i i −1 S(v) PS(v) = exp Pvt (P + mv) exp − Pvt   or, after simplification S(v)−1 PS(v) = P + mv

(2.112)

Equations (2.111) and (2.112) are the quantum Galilean transformations dealing with the Q and P operators

CH002.tex

50

10/3/2011

21: 38

Page 50

BASIS FOR QUANTUM APPROACHES OF OSCILLATORS

2.5.5

Action of translation operators on the Q eigenkets

Start from Eq. (2.101), that is, omitting the dependence of the translation operator on P   iQ◦ P ◦ ◦ ◦ ◦ ◦ QA(Q ) = A(Q )Q + Q A(Q ) with A(Q ) = exp − (2.113)  where A(Q◦ ) is the translation operator, Q◦ a scalar having the dimension of a length, and Q and P having their usual meaning. Now, postmultiply both members of the first equation appearing in (2.113) by an eigenket |{Q} of the position operator Q: QA(Q◦ )|{Q} = A(Q◦ )Q|{Q} + Q◦ A(Q◦ )|{Q}

(2.114)

Owing to the eigenvalue equation (2.91), this equation transforms to QA(Q◦ )|{Q} = A(Q◦ )Q|{Q} + Q◦ A(Q◦ )|{Q} or, after commuting the scalar Q, with the translation operator QA(Q◦ )|{Q} = (Q + Q◦ )A(Q◦ )|{Q}

(2.115)

Now, using the notation A(Q◦ )|{Q} ≡ |{A(Q◦ )Q} Eq. (2.115) yields Q|{A(Q◦ )Q} = (Q + Q◦ )|{A(Q◦ )Q}

(2.116)

On the other hand, the eigenvalue equation Eq. (2.91) reads Q |{Q + Q◦ } = (Q + Q◦ )|{Q + Q◦ }

(2.117)

where |{Q + Q◦ } is the corresponding eigenvector of Q. Then, by comparison of Eqs. (2.115) and (2.117) and ignoring a phase factor without interest, it appears that

or, in view of Eq. (2.95)

A(Q◦ )|{Q} = |{Q + Q◦ }

(2.118)

  iQ◦ P exp − |{Q} = |{Q + Q◦ } 

(2.119)

Of course, since P is Hermitian and Q◦ a real scalar, the Hermitian conjugate of this last expression is  ◦  iQ P {Q}| exp (2.120) = {Q + Q◦ }|  Now, remark that for the infinitesimal transformation (2.103), the translation operator (2.95) may be expanded up to first order in dQ◦ to give A(dQ◦ ) = 1 −

i ◦ dQ P 

CH002.tex

10/3/2011

21: 38

Page 51

2.5

POSITION OR MOMENTUM TRANSLATION OPERATORS

51

Next, the action of this infinitesimal translation operator on an eigenket of Q takes the form   i ◦ ◦ A(dQ )|{Q} = 1 − dQ P |{Q}  or, due to Eq. (2.118), with dQ◦ in place of Q◦ A(dQ◦ )|{Q} = |{Q + dQ◦ } Thus, by identification of these two last equations, one gets |{Q

+ dQ◦ }



 i ◦ = 1 − dQ P |{Q} 

(2.121)

Next, let |{0}Q  be the eigenket of the coordinate operator Q, corresponding to the zero eigenvalue Q|{0}Q  = 0 |{0}Q 

(2.122)

Then, Eq. (2.118) reads A(Q◦ )|{0}Q  = |{0 + Q◦ } or A(Q◦ )|{0}Q  = |{Q◦ }

(2.123)

Again, writing explicitly the translation operator by the aid of Eq. (2.95), and substituting the notation Q◦ by the more general one Q, without modifying anything, one obtains   i exp − QP |{0}Q  = |{Q} (2.124)  On the other hand, recall Eq. (2.106), that is, PB(P◦ ) = B(P◦ )P + P◦ B(P◦ ) with



iP◦ Q B(P ) = exp  ◦

(2.125)



where B(P◦ ) is the translation operator and where P◦ is a scalar having the dimension of an impulsion. Now, postmultiply Eq. (2.125) by an eigenket |{P} of P PB(P◦ )|{P} = B(P◦ )P|{P} + P◦ B(P◦ )|{P} Then, by an inference very similar to that allowing one to pass from Eq. (2.114) to (2.118), one finds B(P◦ )|{P} = |{P + P◦ }

(2.126)

CH002.tex

52

10/3/2011

21: 38

Page 52

BASIS FOR QUANTUM APPROACHES OF OSCILLATORS

Now, consider the eigenvalue equation of the momentum operator corresponding to the zero eigenvalue, that is, P|{0}P  = 0|{0}P  Then, Eq. (2.126) yields B(P◦ )|{0}P  = |{P◦ } Finally, explicitly writing the translation operator B(P◦ ), using Eq. (2.104), and taking P in place of P◦ , and |{P} in place of |{P◦ }, this last equation becomes   iPQ exp (2.127) |{0}P  = |{P} 

2.5.6

Scalar products {P} |{Q}

We have now to find the expression of the scalar product between an eigenket of Q and one of P. 2.5.6.1 A first expression To this aim, premultiply Eq. (2.124) by any bra {P}|:   iQP {P}|{Q} = {P}| exp − |{0}Q   Using Eq. (2.93) and the action of the exponential operator on the left bra, which is an eigenbra of the P operator with the eigenvalue P, one obtains from (2.93)   iQP {P}|{Q} = exp − (2.128) {P}|{0}Q   Now, observe that in this last equation, the bra {P}| may be obtained via the Hermitian conjugate of Eq. (2.127), that is,   iPQ {P}| = {0}P | exp −  Then, using this expression for the bra {P}|, Eq. (2.128) transforms to        iQP iPQ  {P}|{Q} = exp − {0}P  exp − {0} Q   

(2.129)

Next, owing to Eq. (2.122), we may remark that the eigenvalue of Q corresponding to the right-hand-side ket of this last equation, is zero, so that the corresponding eigenvalue equation involving the exponential of Q reduces to   iPQ exp − |{0}Q  = |{0}Q   Thus, the scalar product (2.129) yields  iQP {0}P |{0}Q  {P}|{Q} = exp −  

(2.130)

CH002.tex

10/3/2011

21: 38

Page 53

2.5

POSITION OR MOMENTUM TRANSLATION OPERATORS

53

2.5.6.2 Scalar products involved on the right-hand side of Eq. (2.130) To further utilize Eq. (2.130), we require the following scalar product: {0}P |{0}Q  For this purpose, let us first consider the scalar product {P }|{P

} between two different eigenkets of the momentum operator, which obeys Eq. (2.94), that is, {P }|{P

} = δ(P − P

) Introduce between the ket and the bra the closure relation on the eigenkets of the coordinate operator: ⎧ +∞ ⎫ ⎨ ⎬ {P }| |{Q}{Q}| dQ |{P

} = δ(P − P

) ⎩ ⎭ −∞

or +∞ {P }|{Q}{Q}|{P

}dQ = δ(P − P

) −∞

On the other hand, using Eq. (2.130) and its complex conjugate, this last expression yields     +∞ iQP

iQP

|{0}P |{0}Q | exp − exp dQ = δ(P − P

)   2

−∞

or   +∞ iQ(P − P

) |{0}P |{0}Q | exp − dQ = δ(P − P

)  2

(2.131)

−∞

Now, observe that according to Eq. (18.60) and keeping in mind the fact that the dimension of P is that of Q/, the integral appearing in Eq. (2.131) reads   +∞ iQ(P − P

) exp − dQ = 2πδ(P − P

) 

−∞

Thus, Eq. (2.131) simplifies to 1 2π Therefore, ignoring the unknown phase factor, which is without interest,  1 {0}P |{0}Q  = 2π Eq. (2.130) becomes    1 iPQ exp − {P}|{Q} = 2π  |{0}P |{0}Q |2 =

(2.132)

CH002.tex

54

10/3/2011

21: 38

Page 54

BASIS FOR QUANTUM APPROACHES OF OSCILLATORS

At last, the probability of passage from some ket |{Q} to any ket |{P} or vice versa is, according to the fourth postulate |{P}|{Q}|2 =

1 2π

(2.133)

That shows that whatever the value observed for the position coordinate before a measurement of the momentum, the probability to find after such a measurement some value of the momentum is the same whatever this last value and vice versa. Equation (2.133) may be viewed as the expression of the basic contingency affecting the most simple and fundamental variables appearing in physics through Lagrange’s equations.

2.6

CONCLUSION

This chapter has considered the presentation of the principles of quantum mechanics. We have introduced the important concepts of bras and kets describing the quantum states, that of Hermitean operators describing the physical variables, that of the measurement of physical variables through the eigenvalues of the corresponding Hermitian operators, and the notions of quantum average values generally relating kets and Hermitian operators. In discussing the quantum principles, large parts have been devoted to the time-dependent Schrödinger equation and to quantum averages and to their corresponding fluctuations. The quantum principles were shown to lead to a limitation of the knowledge of some physical conjugated variables, which is illustrated by the Heisenberg uncertainty relations, forbidding one to know simultaneously and exactly the position and momentum, however, preserving the main features of Newton’s laws of classical mechanics, the cost to be paid to the Heisenberg uncertainty relations being the fact that these laws govern average values of the position and momentum in place of exact ones. Now, to be useful applied to particular situations such as, for instance, oscillators, quantum mechanics requires different equivalent representations such as matrix mechanics, wave mechanics, density operator approach, and also equivalent different time-dependent representations such as the Schrödinger, the Heisenberg, and the interaction pictures. The most important results of this chapter are summarized below: Basic equations for quantum mechanics Deterministic and probalistic changes:   ∂|(t) i = H|(t) Pkl = |k |l |2 ∂t Average values, dispersions, and their dynamics:  A = |A| A = A2  − A2   ∂(t)|A|(t) i = (t)|[H, A]|(t) ∂t 

CH002.tex

10/3/2011

21: 38

Page 55

BIBLIOGRAPHY

Eigenvalue equations of the Hermitian operators and their eigenvalues and eigenvectors: A|n  = An |n  since A = A† ,

An = A∗n , n |m  = δnm ,



55

|n n | = 1

Important relations resulting from the commutation rule [Q, P] = i:     ∂F(P) ∂F(Q) [Q, F(P)] = (i) [P, F(Q)] = −(i) ∂P ∂Q (eiQ

◦ P/

if

Q|{Q} = Q|{Q} and P|{P} = P|{P} {P}|{Q}  1 −iQP/ = e P Q   2π

)Q (e−iQ

◦ P/

) = Q + Q◦

(e−iP

◦ Q/

)P(eiP

◦ Q/

) = P + P◦

BIBLIOGRAPHY C. Cohen-Tannoudji, B. Diu, and F. Laloe. Quantum Mechanics. Wiley-Interscience: New York, 2006. P. A. M. Dirac. The Principles of Quantum Mechanics, 4th ed. Oxford University Press: 1982. A. Messiah. Quantum Mechanics. Dover Publications, New York, 1999. L. I. Schiff. Quantum Mechanics, 3rd ed. McGraw-Hill: New York, 1968.

CH003.tex

10/3/2011

CHAPTER

21: 46

Page 57

3

QUANTUM MECHANICS REPRESENTATIONS INTRODUCTION In the previous chapter we obtained different simple but important results following from the postulates of quantum mechanics such as the Ehrenfest and the virial theorems, the Heisenberg uncertainty relations, and the scalar products between any eigenket of Q and another one of P, the modulus of them being the same whatever the corresponding eigenvalues. But, in order to become tractable for the study of concrete situations, it is necessary to adapt the postulates. That is the aim of what is termed different representations of quantum mechanics. Among them there are the matrix mechanics, due initially to Heisenberg, Born, Jordan, and Pauli, and the wave mechanics of Louis de Broglie and Schrödinger. There are also different time-dependent representations besides those of Schrödinger, that is, the Heisenberg picture and the different interaction pictures, which deal with time evolution operators. Finally, there are the density operator representations in which the informations dealing with the kets or the wavefunctions are introduced into an operator and which are very useful when working on many-particle systems. All these representations will be studied in the present chapter.

3.1

MATRIX REPRESENTATION

Because the postulates of quantum mechanics concern the state space, which is a vector space, the matrices play a fundamental role in quantum mechanics leading to the fact that there are matrix representations for all theoretical entities involved in the postulates, that is, for kets, bras, linear transformations, eigenvalue equations, and so on. The purpose of the present section is to consider that subject more deeply.

3.1.1

Kets and bras

First, consider the eigenvalue equation of a Hermitian operator A: A|l  = Al |l  with A = A†

and thus

l |k  = δlk

Quantum Oscillators, First Edition. Olivier Henri-Rousseau and Paul Blaise. © 2011 John Wiley & Sons, Inc. Published 2011 by John Wiley & Sons, Inc.

57

CH003.tex

58

10/3/2011

21: 46

Page 58

QUANTUM MECHANICS REPRESENTATIONS

We remember that the eigenvectors |l  of A obey the closure relation  |i i | = 1 i

Then consider a ket |k  of the state space that does not belong to the set {l } of the eigenvectors |l , and multiply it by the above closure relation:    |i i | |k  |k  = 1|k  = i

Hence |k  =

 i |k |i  i

or |k  =



aik |i  with

aik = i |k 

(3.1)

i

Owing to the convention for matrix notation in which the first index corresponds to the row and the second one to the column, and in view of Eq. (3.1), a ket |k  may be represented, in a basis {|i }, by a column vector, the components of which are the coefficients aik : ⎛ ⎞ a1k ⎜ a2k ⎟ ⎜ ⎟ ⎟ (3.2) |k  ⇔ ⎜ ⎜ ... ⎟ ⎝ aik ⎠ ... Next, one may proceed in a similar way for the bra j | corresponding to the above ket. Then, the Hermitian conjugate of Eq. (3.1) is  aji i | with aji = j |i  j | = i

This last result shows that the matrix representation of the bra j | is a row vector, the components of which are the expansion coefficients of the above equation: j |

⇔

(aj1

aj2

. . . aji

. . .)

(3.3)

Note that the expansion coefficients aik and aki are complex conjugates since they are the expressions of the complex conjugate scalar products, tha is, aij = aji∗

3.1.2

because

i |j  = j |i ∗

Scalar products

Consider the following expansions of the ket |k  and of the bra ξj | in the basis {|i } obeying i |l  = δil

(3.4)

CH003.tex

10/3/2011

21: 46

Page 59

3.1

|k  =



MATRIX REPRESENTATION

aik |i 

with

aik = i |k 

bjl l |

with

bjl = ξj |l 

59

i

ξj | =

 l

Now, the scalar product ξj |k  reads      ξj |k  = bjl l | aik |i  i

l

or  ξj |k  =

 l

bjl aik l |i 

i

so that, due to the orthonormality properties (3.4)  ξj |k  = bjl alk

(3.5)

l

Owing to the matrix convention according to which the first index refers to the row and the second to the column, expression (3.5) appears to be the matrix product between the jth line and the kth column vectors constructed, respectively, from the set of bjl and alk coefficients. ⎛ ⎞ a1k ⎜ a2k ⎟ ⎜ ⎟ ⎟ ξj |k  ⇔ (bj1 bj2 · · · bjl · · ·) ⎜ ⎜ ··· ⎟ ⎝ alk ⎠ ···

3.1.3

Operators

Consider a linear operator A. Premultiply it by the bra i | and postmultiply it by the ket |k  belonging to the same basis as the ket |l , the Hermitian conjugate of which is i |. The linear operation of A on |k  gives a new ket |k  on which the action of i | corresponds to a scalar product, the result of which is the double index scalar Aik : i |A|k  = i |k  = Aik

(3.6)

The different scalars Aik (which may be obtained by allowing the indexes of the ket and of the bra to run over the different terms of the basis) appear to be the matrix elements of a square matrix the dimension of which is generally infinite. Observe that, owing to Eq. (1.30), i |A|k  = k |A† |i ∗

(3.7)

CH003.tex

10/3/2011

60

21: 46

Page 60

QUANTUM MECHANICS REPRESENTATIONS

3.1.3.1

Hermitian operators

If the linear operator is Hermitian, that is, A = A†

then the matrix elements (3.7) simplify to i |A|k  = k |A|i ∗ Hence, from Eq. (3.6), it appears that the matrix elements are complex conjugate with respect to the diagonal part, which is real, so that Aik = A∗ki

Akk = A∗kk

and thus

so that

Akk is real

(3.8)

with A∗ki = k |A|i ∗ A matrix the elements defined by condition (3.8) is a Hermitian matrix. 3.1.3.2

Unitary operators U

−1

Consider the linear unitary operator U satisfying = U†

with

U−1 U = 1

(3.9)

Now, consider a matrix element of this operator Uik = i |U|k  and the corresponding matrix elements of its inverse and of its Hermitian conjugate. They must be equal owing to the fact that the inverse of the unitary operator is equal to its Hermitian conjugate. Hence i |U−1 |k  = i |U† |k 

(3.10)

Of course, owing to the general property (1.30), the following relation for the right-hand-side matrix element of the latter equation exists: i |U† |k  = k |U|i ∗

(3.11)

Thus, because of this last equation, Eq. (3.10) becomes i |U−1 |k  = k |U|i ∗ Thus, the following general relation between the matrix elements of the unitary operator and those of its inverse exists: Uik−1 = Uki∗

(3.12)

with Uik−1 = i |U−1 |k 

Uki∗ = k |U|i ∗ (3.13) Next, consider the matrix element built up from the definition of an inverse operator: Uki = k |U|i 

and

i |U−1 U|k  = i |1|k 

(3.14)

Since the ket |k  and the ket |i  (which is the Hermitian conjugate of the bra i |) belong to the same basis, they are orthogonal, that is, i |1|k  = i |k  = δik

CH003.tex

10/3/2011

21: 46

Page 61

3.1

MATRIX REPRESENTATION

61

so that the matrix element (3.14) obeys i |U−1 U|k  = δik

(3.15)

Next, use the closure relation on the basis {|l }  |l l | = 1 l

By inserting it in Eq. (3.15) between the unitary operator and its inverse, it yields    −1 |l l | U|k  = δik i |U l

Then, using the properties (3.9) of the unitary operator, we have  i |U† |l l |U|k  = δik

(3.16)

l

or, due to Eq. (3.11)  l |U|i ∗ l |U|k  = δik l

so that owing to Eq. (3.13) 

Uli∗ Ulk = δik

l

This last expression may be split into two equations, the first of which shows that any column labeled i of some unitary matrix is normalized and that two different columns labeled i and k of such a matrix are orthogonal:   |Uli |2 = 1 and Uli∗ Ulk = 0 if i  = k (3.17) l

l

All the matrix elements Uli , with l running over the elements of the basis, form therefore a column vector so that Eq. (3.17) may be visualized as the orthonormality properties of the column vectors from which the unitary matrix is built up. Now, taking the Hermitian conjugate of Eq. (3.16) and proceeding in a similar way, one would obtain the two following equations, expressing, respectively, that any row i of a unitary matrix is normalized and that two different rows i and k of such a matrix are orthogonal:   |Uil |2 = 1 and Uil∗ Ukl = 0 if i  = k l

l

Observe that some unitary matrices are real so that, owing to Eq. (3.12), their matrix elements Oik obey −1 = Oki Oik

CH003.tex

62

10/3/2011

21: 46

Page 62

QUANTUM MECHANICS REPRESENTATIONS

For such matrices, which are said to be orthogonal, and, owing to Eq. (3.17), their columns obey the simplified orthonormality properties (which are at the origin of their name):   2 Olk =1 and Oli Olk = 0 if l  = k l

3.1.4

l

Linear transformations

3.1.4.1 Simple linear transformations Hermitian operator B:

Consider the eigenvalue equation of the

B|l  = Bl |l  Since it is Hermitian, its eigenkets |l  are orthonormal so that the following basis {|l } can be constructed:  |i i | = 1 and i |j  = δij (3.18) i

Next, consider the following linear transformation involving the linear operator A, which does not commute with B and which transforms a ket |k  into any another one |ξq : A|k  = |ξq  with

[A, B]  = 0

(3.19)

Now, introduce the closure relation appearing in (3.18) in this linear transformation according to  A |i i |k  = |ξq  i

Again, premultiply both sides of this equation by the bra r |:  r |A|i i |k  = r |ξq  i

which reads



Ari bik = ark

(3.20)

i

with, respectively, Ari = r |A|i 

bik = i |k 

and

arq = r |ξq 

Owing to the matrix convention, and within the representation defined by the basis (3.18), Eq. (3.19) appears to be the matrix linear transformation (3.20) through ⎞ ⎛ ⎞ ⎛ ⎞ ⎛ A11 A12 · · · A1i · · · b1k a1q ⎟ ⎜ b2k ⎟ ⎜ a2q ⎟ ⎜ A21 A22 ⎟ ⎜ ⎟ ⎜ ⎟ ⎜ ⎟ ⎜ ··· ⎟ ⎜ ··· ⎟ = ⎜ ··· ⎟ ⎜ ⎟ ⎜ ⎟ ⎜ ⎠ ⎝ blk ⎠ ⎝ alq ⎠ ⎝ Ar1 ··· ··· ···

CH003.tex

10/3/2011

21: 46

Page 63

3.1

MATRIX REPRESENTATION

63

3.1.4.2 Inverse transformations Now, we shall consider the inverse of the transformation (3.19). Hence, premultiply both members of this equation by the inverse A−1 of the operator A: A−1 A|k  = A−1 |ξq  Then, after simplification, we have |k  = A−1 |ξq  Now, insert the closure relation (3.18) in the following way:  |i i |ξq  |k  = A−1 i

Premultiplying by the bra r | reads  r |A−1 |i i |ξq  r |k  = i

leading to the following matrix representation of the inverse linear transformation: ⎛ ⎞ ⎛ −1 ⎞ ⎞ ⎛ · · · A−1 ··· A11 A−1 b1k a1q 12 1i ⎜ ⎟ ⎜ ⎟ ⎟ ⎜ A−1 ⎜ b2k ⎟ ⎜A−1 ⎟ ⎜ a2q ⎟ 22 ⎜ ⎟ ⎜ 21 ⎟ ⎟ ⎜ ⎜ ··· ⎟ = ⎜··· ⎟ ⎜ ··· ⎟ ⎜ ⎟ ⎜ ⎟ ⎟ ⎜ ⎜ ⎟ ⎜ ⎟ ⎟ ⎜ a ⎝ blk ⎠ ⎝A−1 ⎠ ⎠ ⎝ iq r1 ··· · · · ··· that may be also written brk =



A−1 ri aiq

i

with aiq = i |ξq 

and

−1 A−1 ri = r |A |i 

respectively. 3.1.4.3 Unitary transformations Consider a matrix element of a matrix representation of a linear operator A in some basis {|k } defined by the eigenvalue equation of a Hermitian operator C that does not commute with A: C|k  = Ck |k  with 

|k k | = 1

C = C† and

and

[C, A]  = 0

k |l  = δkl

(3.21)

k

This element is l |A|k  = Alk

(3.22)

CH003.tex

64

10/3/2011

21: 46

Page 64

QUANTUM MECHANICS REPRESENTATIONS

Now, seek the representation of this operator within a new basis {|q } defined by the eigenvalue equation of another Hermitian operator B, which commutes neither with A nor with C: B|q  = Bq |q  with B = B† 

|q q | = 1

[B, A]  = 0

and

p |q  = δpq

and

(3.23)

q

To this end, introduce twice the unity operator inside the matrix element (3.22): l |A|k  = l |1 A 1|k  Then, using for the unity operator of the closure relation appearing in Eq. (3.23), it reads ⎧ ⎫ ⎧ ⎫ ⎨ ⎬ ⎨ ⎬ Alk = l |A|k  = l | |q q | A |p p | |k  ⎩ ⎭ ⎩ ⎭ q

p

or Alk =

 q

l |q q |A|p p |k 

p

and thus Alk =

 q

alq A˜ qp apk

(3.24)

p

with A˜ qp = q |A|p 

alq = l |q 

apk = p |k 

and

(3.25)

Owing to the matrix notation conventions, Eq. (3.24) appears as the following product of matrices: ⎛

A11 ⎜ A21 ⎜ ⎜ .. ⎜ . ⎜ ⎝ Ak1 ⎛

a11 ⎜ a21 ⎜ =⎜ ⎜· · · ⎝ al1

a12 a22 ··· al2

··· ···

⎟ ⎟ ⎟ ⎟ ⎠ ···

A1k

···

··· Ak2

⎞ ⎛

··· ··· all

A12 A22

A˜ 11 ⎜ A˜ 21 ⎜ ⎜· · · ⎜ ⎝ A˜ q1

Akk

⎞ ⎟ ⎟ ⎟ ⎟ ⎟ ⎠

···

⎞⎛ A˜ 12 ··· ··· a11 ⎟ ⎜ a21 A˜ 22 ⎟⎜ ⎟⎜··· ··· ··· ⎟⎜ ⎠ ⎝ ap1 A˜ q2 ··· ···

a12 · · · · · · a22 · · · · · · ··· ap2 ···

⎞ ⎟ ⎟ ⎟ ⎟ ⎠

··· (3.26)

CH003.tex

10/3/2011

21: 46

Page 65

3.1

MATRIX REPRESENTATION

65

Now, observe that the first and third right-hand-side matrices are unitary, which may be proved by first observing that due to Eq. (3.25), it is always possible to write   alq aqk = l |q q |k  q

q

Again, using the closure relation appearing in Eq. (3.23), and the orthonormality properties (3.21), we have  alq aqk = l |k  = δlk q

On the other hand, the unitary transformation (3.26) may be denoted A = U

−1

˜ U A

˜ is where A is the matrix representation of the operator A in the basis (3.21), A that of the same operator in the basis (3.23), and U is the unitary matrix whose elements are given by (3.25). 3.1.4.4 Eigenvalue equations operator A:

Now, write the eigenvalue equations of a linear A|k  = Ak |k 

(3.27)

Now, seek the matrix representation of this equation in the basis {|i } of the eigenkets of a Hermitian operator B, which does not commute with A: B|q  = Bq |q  with 

|q q | = 1

B = B† and

and

[B, A]  = 0

q |p  = δqp

(3.28)

q

Now, introduce this closure relation on both sides of the eigenvalue equation (3.27) according to ⎧ ⎧ ⎫ ⎫ ⎨ ⎨ ⎬ ⎬ A |q q | |k  = Ak |q q | |k  ⎩ ⎩ ⎭ ⎭ q q

so that

 q

A|q q |k  =



Ak |q q |k 

q

Again, premultiply both sides of this last equation by a bra p |:   A|q q |k  = p | Ak |q q |k  p | q

q

which may be written   p |A|q q |k  = Ak p |q q |k  q

q

CH003.tex

66

10/3/2011

21: 46

Page 66

QUANTUM MECHANICS REPRESENTATIONS

which, owing to the orthonormality properties (3.28) of the basis {|i }, transforms to   p |A|q q |k  = Ak δpq q |k  q

or

q

 p |A|q q |k  = Ak p |k  q

and thus



Apq aqk = Ak apk

(3.29)

q

with Apq = p |A|q 

apk = p |k 

and

Equation (3.29) leads to the following matrix representation in the basis (3.28) of the eigenvalue equation (3.27): ⎛ ⎛ ⎞ ⎛ ⎞⎛ ⎞ ⎞ A11 A12 A1q · · · 1 a1k a1k ⎜ A21 A22 ⎜ ⎟ ⎜ a2k ⎟ ⎟ ⎜ a2k ⎟ 1 ⎜ ⎜ ⎟ ⎜ ⎟⎜ ⎟ ⎟ ⎜ ··· ··· ··· ⎟ ⎜ ··· ⎟ ⎟ ⎜ · · · ⎟ = Ak ⎜ 1 ⎜ ⎜ ⎟ ⎜ ⎟⎜ ⎟ ⎟ ⎝ Aq1 Aq2 ⎝ ⎠ ⎝ apk ⎠ ⎠ ⎝ aqk ⎠ 1 Aqq 1 ··· ··· ··· (3.30)

3.1.5

Block matrix representation and symmetry

When some symmetry in a system exists, the matrix representation of the Hamiltonian operator takes the form of a block matrix, the study of which is the aim of the present section. As shown in section 18.9, the symmetry operations all have an inverse, so that the operators S describing them must also have an inverse S−1 obeying S−1 S = SS−1 = 1

(3.31)

Furthermore, since the Hamiltonian operator H of a system cannot be modified by symmetry operations in the same way as its corresponding classical scalar form, the action of any symmetry operator on it cannot modify it so that one may write SH = H

and

S−1 H = H

Hence, the following canonical transformation yields S−1 HS = H

(3.32)

demonstrating that the symmetry operators S commute with the Hamiltonians H, that is, [H, S] = 0

(3.33)

Now, consider a basis {|l } yielding a matrix representation of the Hamiltonian. {g} {u} Then, one may form linear combinations |k  or |j  of the kets |l  belonging

CH003.tex

10/3/2011

21: 46

Page 67

3.1

MATRIX REPRESENTATION

67

to this basis, which are such that they will be symmetric or antisymmetric with respect to the symmetry operation corresponding to the S operator, that is, constructed from the following linear combinations:  {g} {g} |k  = {Clk }|l  l

 {u} {u} {Clj }|l  |j  = l

obeying {g}

{g}

S−1 |k  = |k  {u}

{u}

S−1 |j  = −|j 

{g}

{g}

and

k |S = k |

and

j |S = −j |

{u}

{u}

(3.34) (3.35)

Here the symbols {g} (gerade) and {u} (ungerade) have been used to distinguish between the symmetric and antisymmetric linear combinations. Moreover, consider a matrix element of the Hamiltonian built up from a gerade ket and an ungerade bra. Then, insert the unity operator defined by Eq. (3.31) before and after H in such a way that {g}

{u}

{g}

{u}

j |H|k  = j |SS−1 HSS−1 |k  which, because of Eq. (3.32), simplifies to {g}

{u}

{g}

{u}

j |H|k  = j |SHS−1 |k  a result that, owing to Eqs. (3.34) and (3.35), reads {g}

{u}

{g}

{u}

j |H|k  = −j |H|k  so that {g}

{u}

{g}

{u}

j |H|k  = k |H|j  = 0

(3.36)

where, in the last step, has used Eq. (1.30) and the Hermiticity of H. Equation (3.36) expresses the fact that the matrix element of a Hamiltonian between two kets of different symmetry is zero. As an illustration, if, for instance, a subspace spanned by two gerade and two ungerade kets exists, then, according to Eq. (3.36), the matrix representation of the Hamiltonian takes on the following block form: {g}

{g}

1 |

{g} 2 | {u} 1 | {u} 2 |

|1 

{g}

|2 

{u}

|1 

{u}

|2 

{H {g} } {H12{g} } 0

0

{H {g} }

11

{H {g} }

0

0

0

0

{H {u} }

{H {u} }

0

0

{H {u} } {H {u} }

21

22

11 21

(3.37) 12 22

The interest of the symmetry is to allow size reducing of Hamiltonian matrix representations to be diagonalized.

CH003.tex

68

10/3/2011

21: 46

Page 68

QUANTUM MECHANICS REPRESENTATIONS

Now, consider any ket |{ξ} describing the state of a system of the given symmetry characterized by the symmetry operation S, which may be expressed by a linear combination of g and u state, according to  {g} {g}  {u} {u} |{ξl } = {bkl }|k  + {bjl }|j  (3.38) j

k

Then, due to the first expressions of (3.34) and (3.35), it reads, respectively {g} 1 2 {1 + S}|k 

= 21 {|k  + |k } = |k 

{u} 1 2 {1 − S}|k 

= 21 {|k  + |k } = |k 

{g}

{g}

{g}

{u}

{u}

{u}

{g} 1 2 {1 − S}|k 

= 21 {|k  − |k } = 0

{g}

{g}

{u} 1 2 {1 + S}|k 

= 21 {|k  − |k } = 0

{u}

{u}

As a consequence of these results and of Eq. (3.38), one obtains, respectively  {g} {g} 1 {bkl }|k  {1 + S}|{ξl } = 2

(3.39)

 {u} {u} 1 {bjl }|j  {1 − S}|{ξl } = 2

(3.40)

k

j

3.2

WAVE MECHANICS

Following the above exposition of the matrix representation of quantum mechanics, we now pass to wave mechanics, that is, to the representation of quantum mechanics in the basis of the eigenkets of the Q operator, which is sometimes called the Q representation of quantum mechanics. The precise foundation of wave mechanics by Louis de Broglie in 1924 was completely independent from that of quantum matrix mechanics by Heisenberg, the deep link between the two approaches being later discovered.

3.2.1

Quantum mechanics in representation {|{Q}}

In order to introduce wave mechanics, we start from the eigenvalue equation of the coordinate operator Q and its Hermitian conjugate: Q|{Q} = Q|{Q}

and

{Q}|Q = Q{Q}|

(3.41)

together with the closure relation over the eigenstates of Q and the corresponding orthonormality relations +∞ |{Q}{Q}|dQ = 1 −∞

and

{Q}|{Q } = δ(Q − Q )

(3.42)

CH003.tex

10/3/2011

21: 46

Page 69

3.2 WAVE MECHANICS

69

Now, consider the following scalar product of a ket | by any eigenket |{Q} of Q and its complex conjugate, that is, {Q}| = (Q)

and

|{Q} = ∗ (Q)

(3.43)

Here, the scalar (Q), which is by definition the representation {|{Q}} of the ket |, is named the wavefunction associated with this ket at the measured position Q. It is generally complex. The squared modulus of this scalar product is |{Q}||2 = |(Q)|2

(3.44)

Owing to the fourth postulate, the left-hand side of Eq. (3.44) corresponds to the probability for the system to jump from the ket | into the ket |{Q}, which is an eigenket of the position operator Q with the corresponding eigenvalue Q. Thus, on the right-hand side of Eq. (3.44), |(Q)|2 is the probability for the system described by the scalar function (Q) to be found at the position Q. Now, consider the scalar products of two different kets | and |, and introduce inside the scalar product the closure relation (3.42) ⎫ ⎧ +∞ ⎬ ⎨ |{Q}{Q}|dQ | | = | ⎭ ⎩ −∞

Since the integration operation commutes with the kets or the bras, the scalar product simply reads +∞ | = |{Q}{Q}|dQ −∞

Thus, in view of Eq. (3.43), it takes the form +∞ ∗ (Q)(Q) dQ | = −∞

Next, if the two kets involved in the scalar product belong to a given orthonormal basis, we have +∞ k |l  = k∗ (Q)l (Q) dQ = δkl (3.45) −∞

When applied to the norm of any ket, Eq. (3.45) reduces to the normalization condition +∞ k∗ (Q)k (Q) dQ = 1 −∞

3.2.2

Many-particle systems

The fourth postulate allows one to find the ket of a system formed by many particles, each of them being characterized by their own ket. We illustrate as follows: Consider the value of the total wavefunction Tot (Q) of two particles at any value Q of the position, the individual wavefunction of each particle being,

CH003.tex

70

10/3/2011

21: 46

Page 70

QUANTUM MECHANICS REPRESENTATIONS

respectively, {1} (Q) and {2} (Q). The probability to find the two particles at position Q may be obtained by PTot (Q) = |Tot (Q)|2 Again, since the probabilities multiply, one has |Tot (Q)|2 = |{1} (Q)|2 |{2} (Q)|2 As a consequence of each wavefunction working within its own state space, the total wavefunction may be written Tot (Q)Tot (Q)∗ = ({1} (Q){1} (Q)∗ )({2} (Q){2} (Q)∗ ) so that Tot (Q) = {1} (Q){2} (Q) Hence, since the probabilities multiply, the meaning of the wavefunction implies that the total wavefunction of a system composed of two particles may be written as the product of the wavefunctions of each particle. By generalization to N particles, we have Tot (Q) =

N 

{k} (Q)

(3.46)

k=1

Furthermore, the wavefunctions Tot (Q) and {k} (Q) are given, respectively, by the following scalar products: Tot (Q) = {Q}|Tot  {k} (Q) = {Q}|{k}  Then, Eq. (3.46) leads to {Q}| Tot  =

N 

{Q}| {k} 

k=1

This equation may be also written {Q}| Tot  = {Q}|

N 

|{k} 

k=1

Of course, this expression holds what may be the bra involved in the scalar products. Thus, it is possible to write |Tot  =

N 

|{k} 

(3.47)

k=1

That shows that the total ket of a system formed by several particles is the product of the kets of the different particles.

CH003.tex

10/3/2011

21: 46

Page 71

3.2 WAVE MECHANICS

3.2.3

71

Momentum operator in representation {|{Q}}

In order to get the action of the momentum operator P on any ket | within the {Q} representation, introduce between P and | the closure relation on the eigenkets of P: ⎧ +∞ ⎫ ⎨ ⎬ {Q}|P| = {Q}|P |{P}{P}|dP | ⎩ ⎭ −∞

which due to the eigenvalue equation of P becomes +∞ {Q}|P| = {Q}| P|{P}{P}|{}dP −∞

or, using {P}|{} = (P) +∞ {Q}|P| = P{Q}|{P}(P)dP

(3.48)

−∞

Moreover, observe that we have shown that the scalar product of an eigenket of Q by another one of P is given by Eq. (2.132), that is,   iPQ 1 exp − {P}|{Q} = √  2π so that Eq. (3.48) transforms to   +∞ iPQ P exp − (P)dP {Q}|P| = √  2π 1

(3.49)

−∞

Now, using Eq. (18.49), that is,   +∞ ∂f (Q) iQP/ Pf (P)e dQ = i √ ∂Q 2π 1

with

(Q) = f (Q)

−∞

Eq. (3.49) takes the form  {Q}|P| = i

∂(Q) ∂Q

 (3.50)

This last result shows that in the quantum representation {|{Q}}, the momentum operator is acting on a wavefunction as a partial derivative with respect to the scalar Q times /i, which may be written formally as P=

 ∂ ∂ = −i i ∂Q ∂Q

(3.51)

CH003.tex

72

10/3/2011

21: 46

Page 72

QUANTUM MECHANICS REPRESENTATIONS

Of course, in the quantum representation {|{Q}}, the action of the operator Q over some ket | reads {Q}|Q| = Q{Q}| = Q(Q)

(3.52)

Observe that the following commutator reads    ∂ ∂ [Q, P] = Q − Q i ∂Q ∂Q thereby taking into account the fact that the right-hand side of this last equation is acting on any function. Thus one has to write ∂ ∂ Q=Q +1 ∂Q ∂Q and, after simplification, the above commutator becomes [Q, P] = i That is the equivalent in the quantum representation {|{Q}} of the fundamental commutator given by the last postulate of quantum mechanics: [Q, P] = i

3.2.4 Time-independent Schrödinger equation Consider some operator function F(Q, P) of P and Q that may be separately expanded in powers of P and Q according to  {Cn Pn + Bn Qn } (3.53) F(Q, P) = n

where Cn and Bn are, respectively, the expansion coefficients that are scalars. Now, consider matrix elements of this operator:  {Q}|{Cn Pn + Bn Qn }| (3.54) {Q}|F(Q, P)| = n

Again, owing to Eqs. (3.52), and (3.50), it appears that {Q}|Qn | = Qn (Q)  {Q}|Pn | =

 ∂ i ∂Q

(3.55)

n (Q)

Hence, with Eqs. (3.55) and (3.56), Eq. (3.54) transforms to      ∂ n {Q}|F(Q, P)| = Cn + Bn Qn (Q) i ∂Q n When the operator F(Q, P) is the Hamiltonian H(Q, P) H(Q, P) = T(P) + V(Q)

(3.56)

(3.57)

CH003.tex

10/3/2011

21: 46

Page 73

3.2 WAVE MECHANICS

73

with T(P) =

P2 2m

and



V(Q) =

Bn Q n

n

Eq. (3.57) takes the form ˆ {Q}|H(Q, P)| = H(Q)

(3.58)

with Hˆ = Tˆ + Vˆ (Q)

with

Tˆ = −

2 ∂ 2 2m ∂Q2

(3.59)

and Vˆ (Q) = Bn Qn so that Hˆ = −

2 ∂ 2 + Vˆ (Q) 2m ∂Q2

(3.60)

This equation is the wave mechanics representation of the Hamiltonian, the eigenvalue equation of which is ˆ k (Q) = Ek k (Q) H or −

2 ∂ 2 k (Q) + Vˆ (Q)k (Q) = Ek k (Q) 2m ∂Q2

(3.61)

This is the time-independent Schrödinger equation, that is, the wave mechanics representation of the Hamiltonian eigenvalue equation H(Q, P)|k  = Ek |k 

3.2.5

(3.62)

Wavefunction boundary conditions

The eigenvalues Ek are the same in both Eqs. (3.61) and (3.62), whereas the connection between the eigenfunction k (Q) of H and the eigenket |k  of H(Q, P) is through the following scalar product: k (Q) = {Q}|k 

(3.63)

Recall that, the fourth postulate allows one to write |{Q}|k |2 = |k (Q)|2 ≡ P(Q)

(3.64)

Observe that P(Q) may be regarded as the probability density, which is also denoted ρ(Q). Again, since the probabilities P(Q) must obey +∞ P(Q)dQ = 1 −∞

CH003.tex

74

10/3/2011

21: 46

Page 74

QUANTUM MECHANICS REPRESENTATIONS

thus Eq. (3.64) implies that the wavefunction k (Q) has the normalization property +∞ |k (Q)|2 dQ = 1

(3.65)

−∞

˜ k (Q). In Furthermore, if the wavefunction is not normalized, it may be written as  order to be square summable, its integral must be finite according to +∞ ˜ k (Q)|2 dQ = k 2 |

with k 2 finite

−∞

Then, in order to satisfy Eq. (3.65), one has 1 ˜ k (Q) k (Q) =  k where 1/k is the normalization constant. The normalization condition implies that at infinity the wavefunction must vanish, that is, k (Q → ±∞) → 0

(3.66)

This is an essential boundary condition for the time-independent Schrödinger equation (3.61). Such a condition leads to quantized eigenvalues and thus to quantized energy levels, not only for the eigenvalue equation (3.61) but also for that (3.62), which is equivalent. Since the eigenvalue equation (3.61) has the structure of a wave equation, the {Q} representation (3.61) of the eigenvalue equation (3.62) of the Hamiltonian may be viewed as a wave mechanics equation.

3.2.6 Time-dependent Schrödinger equation From the Schrödinger equation it is possible, with help from the sixth postulate, to find the linear time-dependent operator that transforms some ket at initial time |(0) into the corresponding one |(t) at time t. To get this operator, we start from the Schrödinger equation   ∂|(t) i = H|(t) ∂t In order to solve this equation, premultiply it by some eigenbra of Q leading to   ∂{Q}|(t) i = {Q}|H|(t) ∂t or, due to Eq. (3.58),   ∂(Q, t) ˆ i (3.67) = H(Q, t) ∂t By omitting the Q dependence, after integration between t = 0 and t one obtains   (t) i ˆ ln = − Ht (0) 

CH003.tex

10/3/2011

21: 46

Page 75

3.2 WAVE MECHANICS

75

Passing to the exponential, that reads (t) ˆ = e−iHt/ (0) or ˆ

(t) = e−iHt/ (0)

3.2.7

(3.68)

Current density and continuity equation

Now, let us define the current density operator according to the Hermitian product of the two Hermitian operators P|{Q}{Q}| and |{Q}{Q}|P: 1 {P|{Q}{Q}| + |{Q}{Q}|P} 2m Now, consider the diagonal matrix elements of this operator built up from some eigenkets of any Hermitian operator, that is, J≡

J = |J| This representation of the operator is therefore   1 |P|{Q}{Q}| + hc J= 2m where hc denotes the Hermitian conjugate. Then, using Eq. (3.50) and its Hermitian conjugate, one obtains     ∂ ∂ ∗ ∗ J=− i (3.69) (Q) (Q) − (Q) (Q) 2m ∂Q ∂Q Now, in order to find the continuity equation governing the wavefunction, differentiate the current density (3.69) with respect to Q       ∂J  ∂ ∂ ∂ (3.70) =− i ∗  −  ∗ ∂Q 2m ∂Q ∂Q ∂Q One obtains, respectively,     2 ∂ ∗ ∂ ∂ ∗ ∂ ∗ ∂   =  +  ∂Q ∂Q ∂Q ∂Q ∂Q2     ∂ ∂ ∂ ∂2 ∂  ∗ =  ∗ +  2 ∗ ∂Q ∂Q ∂Q ∂Q ∂Q Thus, Eq. (3.70) transforms to      ∂2 ∂J ∂2  =− i ∗ 2  −  2 ∗ ∂Q 2m ∂Q ∂Q

(3.71)

On the other hand, consider the probability density related to the wavefunction ρ = ∗ 

(3.72)

CH003.tex

76

10/3/2011

21: 46

Page 76

QUANTUM MECHANICS REPRESENTATIONS

By differentiation with respect to time, we get      ∗ ∂ ∂ρ ∂ =  + ∗ ∂t ∂t ∂t

(3.73)

Besides, the two derivatives of the wavefunction are governed by the time-dependent Schrödinger equation (3.67) and its Hermitian conjugate, that is,    ∗  ∂ ∂ ˆ ∗ i (3.74) = H and −i = H ∂t ∂t Moreover, according to Eq. (3.60), the Hamiltonian is  2  2 ∂ Hˆ = − + Vˆ 2m ∂Q2 Hence, Eqs. (3.74) lead to    2  ∂ 2 ∂ i  + Vˆ  =− ∂t 2m ∂Q2

and −i

 ∗   ∂ 2 ∂ 2 ∗ + Vˆ ∗ =− ∂t 2m ∂Q2

These two last equations allow one to transform Eq. (3.73) into      ∂ρ ∂2 ∂2  = −i  2 ∗ − ∗ 2  ∂t 2m ∂Q ∂Q Finally, it appears from comparison with Eq. (3.71) that the following onedimensional equation is verified:     ∂ρ ∂J =− ∂t ∂Q By generalization to the three-dimensional equation, one obtains the continuity equation   ∂ρ − → (3.75) + Div J = 0 ∂t where the arrow indicates a vectorial entity.

3.3

EVOLUTION OPERATORS

As we have seen, when considering the sixth postulate of quantum mechanics dealing with the dynamics involved in quantum mechanics, there are several timedependent descriptions of quantum mechanics. In the Schrödinger picture (SP), the kets depend on time, whereas the operators do not change with it. However, another time-dependent description, the Heisenberg picture (HP) exists, where the operators depend on time whereas the kets remain constant. Finally, many other time-dependent representations of quantum mechanics exist, which are intermediate between the Schrödinger and the Heisenberg pictures, in which both the kets and the operators depend on time in subtle ways. They are named the interaction pictures. We shall first consider the time evolution operator within the Schrödinger picture.

CH003.tex

10/3/2011

21: 46

Page 77

3.3

3.3.1

EVOLUTION OPERATORS

77

Schrödinger picture

Starting from the Schrödinger equation defined by the sixth postulate and governing the dynamics of some time-dependent ket |(t) ∂ |(t) = H|(t) (3.76) ∂t where H is the total Hamiltonian of the system. Then, we introduce a linear operator U(t), the time evolution operator, allowing one to transform the ket |(0) at an initial time t = 0 into one at time t according to i

|(t) = U(t)|(0)

(3.77)

with the obvious condition U(0) = 1 Then, time differentiation of Eq. (3.77) yields   ∂ ∂U(t) |(t) = |(0) ∂t ∂t

(3.78)

Again, introduce on the right-hand side of Eq. (3.78) U(t)−1 U(t) = 1 in such a way as to write ∂ |(t) = ∂t



(3.79)

 ∂U(t) U(t)−1 U(t)|(0) ∂t

leading with the help of Eq. (3.77) and after multiplying by i, to   ∂ ∂U(t) i |(t) = i U(t)−1 |(t) ∂t ∂t

(3.80)

Thus, identification of Eqs. (3.76) and (3.80) yields   ∂U(t) H|(t) = i U(t)−1 |(t) ∂t Hence, since this latter result holds irrespective of |(t), it appears that the following relation between the operators U(t) and H exists:   ∂U(t) i (3.81) = HU(t) ∂t The foregoing partial differential equation reads dU(t) i = − H dt U(t)  which, by integration yields ln

U(t) i = − Ht U(0) 

CH003.tex

78

10/3/2011

21: 46

Page 78

QUANTUM MECHANICS REPRESENTATIONS

or due to the boundary condition (3.79) U(t) = e−iHt/

(3.82)

Observe that the Hamiltonian H being Hermitian, the evolution operator U(t) is unitary since its inverse is equal to its Hermitian conjugate, that is, U(t)−1 = U(t)† = eiHt/

(3.83)

U(t)† U(t) = U(t)−1 U(t) = 1

(3.84)

so that

Moreover, due to Eq. (3.82), Eq. (3.77) becomes |(t) = (e−iHt/ )|(0)

(3.85)

We remark that Eq. (3.68) is the wave mechanics representation of the quantum relation (3.85). Sometimes the Hamiltonian may depend on time, so that one has to solve a dynamic equation that is more complicated than (3.81) and of the form   ∂U(t) = H(t)U(t) (3.86) i ∂t Here, the Hamiltonians at different times do not commute: [H(t), H(t  )]  = 0 Moreover, it is possible to write formally a solution of Eq. (3.86) in the same way as (3.81) according to ⎫ ⎧ ⎬ ⎨ i t U(t) = Pˆ exp − H(t  ) dt  (3.87) ⎭ ⎩  0

where Pˆ is the Dyson time-ordering operator.

3.3.2

Heisenberg picture

Now, it is suitable to introduce a new time-independent picture in which (in contrast to the Schrödinger picture where the kets are time dependent and the operators constant) the kets are constant and the operators time dependent. This is the Heisenberg picture. For this purpose, we start from the Schrödinger picture equation (2.65) yielding the mean value of some operator A averaged over the time-dependent states, that is, A(t)k = k (t)SP |ASP |k (t)SP  where the superior index SP indicates that the Schrödinger picture has been used. Next, due to Eq. (3.85) and to its Hermitian conjugate, this average value reads A(t)k = k (0)|(eiHt/ )ASP (e−iHt/ )|k (0)

CH003.tex

10/3/2011

21: 46

Page 79

3.3

EVOLUTION OPERATORS

79

which may be written A(t)k = kHP |A(t)HP |kHP  where |kHP  is the time-independent ket in the Heisenberg picture, whereas A(t)HP is the time-dependent operator in this same Heisenberg picture given by A(t)HP = (eiHt/ )ASP (e−iHt/ )

(3.88)

or, due to Eqs. (3.82) and (3.83), by A(t)HP = U(t)† ASP U(t)

(3.89)

Since the operator A in the Schrödinger picture is time independent, the time derivation of each members of Eq. (3.89) gives after writing ASP = A       ∂A(t)HP ∂U(t)† ∂U(t) = AU(t) + U(t)† A (3.90) ∂t ∂t ∂t Besides, since the Hamiltonian H is Hermitian, note that the derivative with respect to time of the evolution operator (3.82) and that of its Hermitian conjugate (3.83) are, respectively     i ∂U(t) =− HU(t) ∂t      ∂U(t)† i = U(t)† H ∂t  Using these two equations, Eq. (3.90) becomes       i i ∂A(t)HP = U(t)† HAU(t) − (3.91) U(t)† AHU(t) ∂t   Next, using Eq. (3.84), that is, 1 = U(t)U(t)†

(3.92)

and, inserting on the right-hand-side term of Eq. (3.91) this unity operator, first between the Hamiltonian H and the operator A, and then between the operator A and the Hamiltonian H, one obtains       ∂A(t)HP i i † † = U(t) H{U(t)U(t) }AU(t) − U(t)† A{U(t)U(t)† }HU(t) ∂t   Thus, by changing the position of the brackets, we have     i ∂A(t)HP = ({U(t)† HU(t)}{U(t)† AU(t)} − {U(t)† AU(t)}{U(t)† HU(t)}) ∂t  (3.93) Now, observe that, according to Eqs. (3.82) and (3.83), {U(t)† HU(t)} = (eiHt/ )H(e−iHt/ ) Again, since the exponential depends on the Hamiltonian, it must commute with it, so that after simplification this unitary transformation reduces to {U(t)† HU(t)} = H

CH003.tex

80

10/3/2011

21: 46

Page 80

QUANTUM MECHANICS REPRESENTATIONS

Hence, Eq. (3.93) simplifies to   ∂A(t)HP = −(H{U(t)† AU(t)} − {U(t)† AU(t)}H) i ∂t Thus, due to the definition (3.89), this equation transforms to the final result, which is the Heisenberg equation governing the dynamics of any operator in the Heisenberg picture:   ∂A(t)HP = [AHP (t), H] i (3.94) ∂t This equation contains the same information as the Schrödinger time-dependent equation in the Schrödinger picture. It may also be of interest to take the average of this equation in a state |, in postmultiplying both terms of this equation by |, and premultiplying them by |   ∂A(t)HP | = |[AHP (t), H]| (3.95) i| ∂t This time dependence of the average value in the Heisenberg picture may be compared to that (2.55) we have obtained above in the Schrödinger picture, that is,   ∂A |(t) = (t)|[A, H]|(t) (3.96) i(t)| ∂t Comparison of the Heisenberg picture (3.95) and Schödinger picture (3.96) shows clearly the exchange of the time dependence between the operator and the kets.

3.3.3

Hamilton equations

Consider the position and momentum operators in the Heisenberg picture. To simplify the notation, we shall write Q(t)HP ≡ Q(t)

and

P(t)HP ≡ P(t)

These operators are given in the Heisenberg picture by Q(t) = U(t)† QU(t)

and

P(t) = U(t)† PU(t)

(3.97)

First, verify that the commutators of the two operators remain the same in the Heisenberg picture, where they are time dependent, as in the Schrödinger picture where they are not so. To verify that, use Eq. (3.97) to write explicitly the commutator appearing on the left-hand side, yielding [Q(t), P(t)] = U(t)† QU(t)U(t)† PU(t) − U(t)† PU(t)U(t)† QU(t) After simplification using Eq. (3.92), we have [Q(t), P(t)] = U(t)† [Q, P]U(t) so that, after using Eq. (3.92), it appears that [Q(t), P(t)] = [Q, P]

CH003.tex

10/3/2011

21: 46

Page 81

3.3

EVOLUTION OPERATORS

81

or, due to the basic commutator (2.3), of Q and P [Q(t), P(t)] = i

(3.98)

Now, from the Heisenberg dynamic equation (3.94), it is possible to obtain the dynamics governing the time dependence of the position operator and its conjugate momentum. First consider that of the Q(t) coordinate. Keeping in mind that Q(t) depends only on time, Eq. (3.94) allows one to write the differential equation   dQ(t) i = [Q, H(Q, P)] (3.99) dt Next, use the theorem (2.14), which in the present situation reads as follows:   ∂H(Q, P) [Q, H(Q, P)] = i ∂P Then, in view of this result, Eq. (3.99) takes the form     dQ(t) ∂H(Q, P) = dt ∂P

(3.100)

Now, consider the Heisenberg equation governing the dynamics of the momentum   dP(t) i = [P, H(Q, P)] (3.101) dt Next, in view of Eq. (2.15), the commutator involved in this equation reads   ∂H(Q, P) [P, H(Q, P)] = −i ∂Q Thus, Eq. (3.101) becomes



dP(t) dt



 =−

∂H(Q, P) ∂Q

 (3.102)

Both Eqs. (3.100) and (3.102), which satisfy the quantum commutator (3.98), are the quantum Hamilton equations of motion, the classical limits of which are the classical Hamilton equations  −    −  → →  dP ∂H ∂H dQ − → − → =− − = and with [ Q , P ] = 0 → − → dt dt ∂Q ∂P

3.3.4

Interaction picture

Now, consider a new time-dependent picture of quantum mechanics, the interaction picture (IP), which is intermediate between the Schrödinger and Heisenberg pictures. This picture is sometimes more practical than the pure Schrödinger and Heisenberg representations.

CH003.tex

82

10/3/2011

21: 46

Page 82

QUANTUM MECHANICS REPRESENTATIONS

3.3.4.1 Operators and kets in the interaction picture Suppose that the Hamiltonian H of a system may be split into two parts H◦ and V H = H◦ + V

(3.103)

Now, we may introduce an IP time-dependent ket through an action on a timedependent ket in the Schrödinger picture by aid of the Hermitian of the time evolution operator obtained from H◦ |(t)IP  ≡ (eiH

◦ t/

)|(t)SP 

(3.104)

|(t)SP  is the ket at time t in the Schrödinger representation, whereas |(t)IP  is the corresponding ket at the same time t in the interaction picture. Now, premultiply each member of this last equation by the inverse of the time evolution operator involved in the previous equation. (e−iH

◦ t/

)|(t)IP  = (e−iH

◦ t/

)(eiH

◦ t/

)|(t)SP 

After simplification, which leads to the equation inverse of (3.104) which allows us to pass from the IP to the SP for all kets |(t)SP  = (e−iH

◦ t/

)|(t)IP 

(3.105)

Next, take the partial time derivative of Eq. (3.104), that is,      SP   iH◦ t/ ∂|(t)IP  ∂|(t) ∂e ◦ = (eiH t/ ) + |(t)SP  ∂t ∂t ∂t The last partial time derivative appearing on the right-hand side of this equation is     ◦ ∂eiH t/ i ◦ = H◦ (eiH t/ ) ∂t  whereas the first one is given by the time-dependent Schrödinger equation defined by the sixth postulate, that is,     ∂|(t)SP  1 = H|(t)SP  ∂t i Thus, the time derivative of the IP ket becomes   ∂|(t)IP  ◦ ◦ i = (eiH t/ )H|(t)SP  − H◦ (eiH t/ )|(t)SP  ∂t Next, use for the right-hand-side SP kets, Eq. (3.105), in order to obtain an equation involving only IP kets. Hence   ∂|(t)IP  ◦ ◦ i = (eiH t/ )H(e−iH t/ )|(t)IP  − H◦ |(t)IP  (3.106) ∂t where we have performed a simplification on the right-hand-side because the Hamiltonian H◦ commutes with all function of it, that is, H◦ = (eiH

◦ t/

)H◦ (e−iH

◦ t/

)

CH003.tex

10/3/2011

21: 46

Page 83

3.3

EVOLUTION OPERATORS

83

Thus, one may use in Eq. (3.106) the right-hand side of this last equation in place of H◦ , which reads   ∂|(t)IP  ◦ ◦ ◦ ◦ = (eiH t/ )H(e−iH t/ )|(t)IP  − (eiH t/ )H◦ (e−iH t/ )|(t)IP  i ∂t Then, rearranging, one obtains   ∂|(t)IP  ◦ ◦ = (eiH t/ )(H − H◦ )(e−iH t/ )|(t)IP  i ∂t Again, in view of the partition (3.103), the previous expression reduces to   ∂|(t)IP  ◦ ◦ = (eiH t/ )V(e−iH t/ )|(t)IP  i ∂t which may be written  i

∂|(t)IP  ∂t

 = V(t)IP |(t)IP 

(3.107)

where V(t)IP is the perturbation V in the interaction picture, which is given by V(t)IP = (eiH

◦ t/

)V(e−iH

◦ t/

)

(3.108)

Observe that in the IP both the perturbation operator and the ket are time dependent at the difference of the SP and HP where it is either the ket or the operator, which evolves with time. More generally, under partition (3.103), the IP time dependence of an operator is given by A(t)IP = (eiH

◦ t/

)A(e−iH

◦ t/

)

3.3.4.2 Dynamics of IP time evolution operators Now, we may introduce an interaction picture operator U(t)IP , which transforms any SP ket at initial time t0 into the corresponding IP ket at time t, according to |(t)IP  ≡ U(t − t0 )IP |(t0 )SP 

(3.109)

U(t0 )IP = 1

(3.110)

with the stipulation that

Now, observe that Eq. (3.104) may be written |IP (t) ≡ U◦ (t − t0 )−1 |SP (t)

(3.111)

where U◦ (t − t0 ) is the time evolution operator given by U◦ (t − t0 ) = e−iH

◦ (t−t

with, of course, U◦ (t0 ) = 1

0 )/

(3.112)

CH003.tex

84

10/3/2011

21: 46

Page 84

QUANTUM MECHANICS REPRESENTATIONS

Moreover, the inverse transformation of Eq. (3.109) may be obtained by premultiplying in it both members of the inverse of the IP time evolution operator. On simplification, we have |(t0 )SP  = U(t − t0 )IP−1 |(t)IP 

(3.113)

On the other hand, taking the partial derivative of both terms of Eq. (3.109) reads     ∂|(t)IP  ∂U(t)IP = |(t0 )SP  ∂t ∂t Thus, in view of Eq. (3.113) allowing to pass from any SP ket at initial time t0 to the corresponding IP ket at time t, the equation transforms to     ∂|(t)IP  ∂U(t)IP = U(t − t0 )IP−1 |(t)IP  ∂t ∂t Then, one may replace the left-hand side of this last equation by its expression given by Eq. (3.107). After rearranging, we have   ∂U(t)IP IP IP V(t − t0 ) |(t)  = i U(t − t0 )IP−1 |(t)IP  ∂t Then, since this last linear transformation is satisfied irrespective of the IP ket at any time, we have   ∂U(t)IP U(t − t0 )IP−1 = V(t − t0 )IP i ∂t Finally, postmultiply both member of this last equation by U(t − t0 )IP . Hence, after simplification using the operator property   ∂U(t)IP i (3.114) = V(t − t0 )IP U(t − t0 )IP ∂t which, when t0 = 0 simplifies to   ∂U(t)IP i = V(t)IP U(t)IP ∂t

(3.115)

and, due to Eq. (3.110) U(0)IP = 1

(3.116)

3.3.4.3 Relation between IP and SP time evolution operators Now, observe that the linear transformation, which is inverse of that given by Eq. (3.111), may be obtained by premultiplying both terms by U◦ (t) and then simplifying the result using U◦ (t − t0 )U◦ (t − t0 )−1 = 1, leading to |(t)SP  = U◦ (t − t0 )|(t)IP 

(3.117)

Then, premultiplying both members of this last equation by U◦ (t − t0 )−1 , we have on simplification U◦ (t − t0 )−1 U◦ (t − t0 ) = 1

CH003.tex

10/3/2011

21: 46

Page 85

3.3

EVOLUTION OPERATORS

85

and U◦ (t − t0 )−1 |(t)SP  = |(t)IP 

(3.118)

Next, owing to Eq. (3.109) relating the IP ket at time t with the SP one at initial time t0 , the last equation becomes |(t)SP  ≡ U◦ (t − t0 )U(t − t0 )IP |(t0 )SP 

(3.119)

Remark that, according to Eq. (3.85), the SP ket at time t is related to the corresponding one on initial time t0 via |(t)SP  ≡ U(t − t0 )|(t0 )SP 

(3.120)

U(t − t0 ) = e−iH(t−t0 )/

(3.121)

with

Thus, comparison of Eqs. (3.119) and (3.120) shows that U(t − t0 ) = U◦ (t − t0 )U(t − t0 )IP

(3.122)

Equation (3.122) shows that the full time evolution operator U(t − t0 ), which is given by Eq. (3.121), is equal to the unperturbed time evolution operator U◦ (t − t0 ) given by Eq. (3.112) times the IP time evolution operator governed by the partial differential equation (3.114). 3.3.4.4 Perturbation expansion of the time evolution operator We shall now obtain the full time evolution operator U(t) when it is only easy to find its corresponding unperturbed time evolution operator U◦ (t). The solution of the problem requires one to get the IP time evolution operator by solution of Eq. (3.115) with the boundary condition (3.116), that is,   ∂U(t)IP = V(t)IP U(t)IP with U(0)IP = 1 i ∂t On integration between t  = 0 and t  = t, and using the boundary condition, we have  IP

U(t)

=1+

1 i

 t

V(t  )IP U(t  )IP dt 

(3.123)

0

Now, in order to solve the integral equation (3.123), one may write for U(t  )IP on its right-hand side, an expression that may be obtained from Eq. (3.123), by the replacements t  → t, and t  → t  , namely U(t  )IP = 1 +



1 i

 t 

V(t  )IP U (t  )IP dt 

0

Hence, Eq. (3.123) yields 

U(t)IP

1 = 1+ i

 t 0



1 V(t  )IP dt  + i

2  t t  0

0

VIP (t  )VIP (t  )UIP (t  )dt  dt  (3.124)

CH003.tex

86

10/3/2011

21: 46

Page 86

QUANTUM MECHANICS REPRESENTATIONS

The foregoing equation may be iterated as many times as required. If the perturbation V is very small with respect to H◦ , the third term on the right-hand side of this last equation, which is quadratic in V, may be neglected with respect to the second term, which is linear in V, leading to the first-order expansion of the IP time evolution operator given by  U(t)

=1+

IP

1 i

 t

V(t  )IP dt 

(3.125)

0

To simplify, limit the iteration by truncating the IP time evolution operator U(t  )IP at time t  appearing in Eq. (3.124), to the first term unity  IP

U(t )

 =1+

1 i

 t 

V(t  )IP U(t  )IP dt  1

0

That leads to the following second-order perturbative expansion for the IP time evolution operator:  U(t)IP 1 +

1 i

 t

V(t  )IP dt  +

0



1 i

2  t t  0

VIP (t  )VIP (t  )dt  dt 

0

Next, owing to Eqs. (3.108) and (3.112), the IP time evolution operator reads V(t)IP = U◦ (t)−1 VU◦ (t)

(3.126)

Then, using (3.126) and also Eq. (3.122) allowing to pass from U(t)IP to U(t), the full time evolution operator appears to be given by U(t) U◦ (t)   t 1 ◦ + U (t) U◦ (t  )−1 VU◦ (t  )dt  i 0



1 + i

2

U◦ (t)

 t t  0

U◦ (t  )−1 VU◦ (t  )U◦ (t  )−1 VU◦ (t  )dt  dt 

0

This result must be considered, keeping in mind Eqs. (3.82) and (3.112), that is, U(t) = (e−iHt/ )

3.3.5

and

U◦ (t) = (e−iH

◦ t/

)

Formal expression to make Eq. (3.123) tractable

Observe that the time evolution operators allow one to pass from a ket at initial time t = 0 to another at time t. |(t) = U(t)|(0)

CH003.tex

10/3/2011

21: 46

Page 87

3.3

EVOLUTION OPERATORS

87

Now, we may replace the initial time t = 0 by any time t ◦ . Then, this equation may be written |(t) = U(t, t ◦ )|(t ◦ )

(3.127)

In this equation, the time evolution operator appears to be a conditional operator, which, if the ket is |(t ◦ ) at time t = t ◦ , transforms this ket into |(t) at time t. Note that, U(t, t ◦ ) is by definition given by U(t, t ◦ ) = e−i(t−t

◦ )H/

Next, consider the following time evolution operators: U(t2 , t1 ) = e−i(t2 −t1 )H/

U(t1 , t ◦ ) = e−i(t1 −t

and

◦ )H/

(3.128)

Again, consider the third evolution operator U(t2 , t ◦ ) = e−i(t2 −t

◦ )H/

Of course, this operator may be written U(t2 , t ◦ ) = e−i{(t2 −t1 )+(t1 −t

◦ )}H/

(3.129)

Then, owing to the equation appearing in Eq. (3.128), the time evolution operator (3.129) appears to be U(t2 , t ◦ ) = U(t2 , t1 ) U(t1 , t ◦ )

(3.130)

It may be observed that Eqs. (3.127) and (3.130) are true for all kinds of time evolution operators, that is, for full, unperturbed, and IP time evolution operators. Keeping that in mind, we may return to Eq. (3.123). 

◦

U (t, t ) = 1 + IP

1 i

 t

VIP (τ, t ◦ )UIP (τ, t ◦ ) dτ

(3.131)

t◦

Next, by inversion of Eq. (3.122), one obtains UIP (t, t ◦ ) = U◦ (t, t ◦ )−1 U(t, t ◦ ) This equation allows to transform Eq. (3.131) into ◦

◦ −1

U (t, t )

◦



U(t, t ) = 1 +

1 i

 t

VIP (τ, t ◦ )UIP (τ, t ◦ ) dτ

(3.132)

t◦

Next, we may use U◦ (t, t ◦ )−1 U◦ (t, t ◦ ) = 1

(3.133)

Then, premultiplying the right-hand side of Eq. (3.132) by this last equation leads to ⎛ ⎞   t 1 U◦ (t, t ◦ )−1 U(t, t ◦ ) = U◦ (t, t ◦ )−1 U◦ (t, t ◦ ) ⎝1 + VIP (τ, t ◦ )UIP (τ, t ◦ ) dτ ⎠ i t◦

(3.134)

CH003.tex

88

10/3/2011

21: 46

Page 88

QUANTUM MECHANICS REPRESENTATIONS

Next, premultiply both members of this last equation by U◦ (t, t ◦ ). Then, owing to Eq. (3.133), Eq. (3.134) reduces to ⎛ ⎞   t 1 U(t, t ◦ ) = U◦ (t, t ◦ ) ⎝1 + VIP (τ, t ◦ )UIP (τ, t ◦ ) dτ ⎠ i t◦

which may be written as ◦

◦



◦

U(t, t ) = U (t, t ) +

1 i

 t

U◦ (t, t ◦ )VIP (τ, t ◦ )UIP (τ, t ◦ ) dτ

(3.135)

t◦

Now, observe that the unperturbed time evolution operator, just after the integral allowing one to pass from t = 0 to t may be viewed as the product: U◦ (t, t ◦ ) = [U◦ (t, τ)U◦ (τ, t ◦ )] Then, Eq. (3.135) may be written ◦

◦

◦

U(t, t ) = U (t, t ) +



1 i

 t

[U◦ (t, τ)U◦ (τ, t ◦ )]{VIP (τ, t ◦ )}UIP (τ, t ◦ ) dτ

t◦

Again, for the perturbation Hamiltonian in the interaction picture use Eq. (3.126): U(t, t ◦ ) = U◦ (t, t ◦ )   t 1 + U◦ (t, τ)U◦ (τ, t ◦ )U◦ (τ, t ◦ )−1 VU◦ (τ, t ◦ )U◦ (τ, t ◦ )−1 U◦ (τ, t ◦ ) dτ i t◦

Finally, in order to simplify this last result, we may use the property of a time evolution operator and of its inverse in the following way: U◦ (τ, t ◦ )U◦ (τ, t ◦ )−1 = 1

U◦ (τ, t ◦ )−1 U◦ (τ, t ◦ ) = 1

and

That leads to the final result of importance: U(t, t ◦ ) = U◦ (t, t ◦ ) +



1 i

 t

U◦ (t, τ)VU◦ (τ, t ◦ ) dτ

(3.136)

t◦

Note in this last equation the respective places of the times t ◦ , τ, and t,

3.4

DENSITY OPERATORS

After studying the time dependence of quantum mechanics, through the Schrödinger, Heisenberg, and interaction pictures using the time evolution operator, it is now appropriate to introduce the fundamental concept of the density operator, which is a very powerful tool in quantum mechanics.

CH003.tex

10/3/2011

21: 46

Page 89

3.4

3.4.1 3.4.1.1

DENSITY OPERATORS

89

Basic properties Definition

By definition, the density operator ρ of a statistical mixture is  ρ= Wi |i i | (3.137) i

Here the Wi are the probabilities for the states to be occupied, which are therefore real and must obey  Wi = 1 and 0 ≤ Wi ≤ 1 i

whereas the kets |i  belong to an arbitrary basis in the state space, and thus obey  |i i | = 1 (3.138) and i |k  = δik i

Note that for a pure state, all the operations are zero except one, which is equal to unity. Then the density operator expression (3.137) reduces to ρ = |i i |

(3.139)

3.4.1.2 Trace of the density operator Consider now the trace of the density operator. It is in the basis used for its description:     k | Wi |i i | |k  tr{ρ} = i

k

Then, since the Wi are scalars, owing to the orthonormality properties of the basis, this last equation transforms to  Wi k |i i |k  tr{ρ} = k

i

Again, owing to the orthonormality properties (3.138) of the basis, that reduces to  Wi tr{ρ} = i

At last, since the sum of the probabilities Wi is equal to unity, the trace appears to be simply given as tr{ρ} = 1

(3.140)

3.4.1.3 Hermiticity of the density operator The Hermitian conjugate of the density operator (3.137) is  †  ρ† = Wi |i i | (3.141) i

Again, using the rules of this section governing Hermitian conjugation, the right-hand side of this last equation is †    Wi |i i | = Wi∗ |i i | i

i

CH003.tex

90

10/3/2011

21: 46

Page 90

QUANTUM MECHANICS REPRESENTATIONS

or, since the probabilities are real, †    Wi |i i | = Wi |i i | i

i

Hence, Eq. (3.141) becomes

 ρ = †



 Wi |i i |

(3.142)

i

Thus, by comparison of Eq. (3.142) with Eq. (3.137), it appears that ρ† = ρ showing that the density operator is Hermitian. 3.4.1.4 Inequality governing the density operator in the general case of mixed states Consider the square of the density operator, which, owing to Eq. (3.137), reads   ρ2 = Wi |i i | Wk |k k | i

k

or, since the probabilities Wi are scalars,  ρ2 = Wi Wk |i i |k k | i

k

so that due to the orthonormality properties (3.138)  ρ2 = Wi Wk |i δik k | i

k

and, thus, after simplification using the properties of the Kronecker symbol, it is found that  ρ2 = Wi2 |i i | (3.143) i

Moreover, since the probabilities are smaller than unity, their squares obey the inequality Wi2 < Wi2 and it appears by comparison of Eq. (3.143) with (3.137) that ρ2 < ρ

(3.144)

For a pure state verifying Eq. (3.139), the square of the density operator reduces to ρ2 = |i i |i i | or, because of the orthonormality properties (3.138), to ρ2 = |i i | so that ρ2 = ρ

(pure state)

CH003.tex

10/3/2011

21: 46

Page 91

3.4

3.4.2

DENSITY OPERATORS

91

Density operator for many particles

Now, consider the density operator of a set of N particles. Then, according to Eq. (3.47), a ket characterizing a whole system is given by the product |k(1),l(2)...f (N)  =

(N) 

|j(r) 

(3.145)

(r)

where |l(r)  is the lth ket |l  of the rth particle. Next, for a pure case, the full density operator ρTot of the set of N particles is given by an expression of the same form as that in Eq. (3.139) in which the ket given by Eq. (3.145) plays the role of |i  in Eq. (3.139), that is, ρTot = |k(1)l(2)...f (N) k(1)l(2)....f (N) | or ρTot =

(N) 

|j(r) j(r) |

(3.146)

(r)

Again, for a mixed situation, the generalization to N particles of Eq. (3.137), leads to   ... Wk(1)l(2)...f (N) |k(1)l(2)...f (N)  k(1)l(2)....f (N) | ρTot = k(1) l(2)

f (N)

where the Wk(1)l(2)...f (N) are the joint probabilities to find the first particle (1) in the kth state |k , with the probability Wk(1) , the second particle (2) in the lth state |l  with the probability Wl(2) , and so on given by Wk(1)l(2)...f (N) = Wk(1) Wl(2) . . . On the other hand, consider a physical system that may be divided into two different subsystems. Then, the full density operator of this system may be written as the product of the density operators of the two subsystems: ρTot = ρ(1) ρ(2) By definition, the reduced density operator of one of the two subsystems is the partial trace over the subspace spanned by the other subsystem over the full density operator: ρRed(2) = tr(1) {ρTot }

3.4.3

ρRed(1) = tr(2) {ρTot }

Average values

We now show that the average value of an operator A performed over the density operator of a statistical mixed state is Aρ = tr{ρA}

(3.147)

In order to prove this equation, recall that, according to Eqs. (3.137) and (3.140), the density operator obeys  ρ= Wi |i i | and tr{ρ} = 1 i

CH003.tex

92

10/3/2011

21: 46

Page 92

QUANTUM MECHANICS REPRESENTATIONS

Then, Eq. (3.147) becomes



Aρ = tr



 Wi |i i |A

i

Perform the trace over the basis {|i }, which reads  k |Wi |i i |A|k  Aρ = i

k

Or, since the probabilities Wi are scalars,  Wi k |i i |A|k  Aρ = i

k

Finally, owing to the orthonormality properties (3.138) of the basis {|i }, that simplifies to  Aρ = Wi δki i |A|k  k

and thus Aρ =

i



Wi i |A|i 

(3.148)

i

Hence, the average of operator A over density operator ρ is the sum of all the quantum average values of operator A over the kets |i  belonging to the basis {|i }, times the corresponding probabilities Wi . Of course, for a pure density state where all the probabilities are zero, except one which is unity, the average value over the density operator (3.148) reduces to the simple quantum average value (2.21), that is, Aρ = i |A|i 

3.4.4

Entropy and density operators

Introduce the statistical entropy function through S = −kB ln ρρ where kB is the Boltzmann constant. Now, keeping in mind that the average of an operator over the density operator is given by Eq. (3.147), the statistical entropy becomes S = −kB tr{ρ ln ρ}

(3.149)

Again, writing explicitly the trace involved in Eq. (3.149) by performing the trace over the basis {|i } obeying Eq. (3.138), that is, k |l  = δkl we have S = −kB

 i

i |ρ ln ρ|i 

(3.150)

(3.151)

CH003.tex

10/3/2011

21: 46

Page 93

3.4

DENSITY OPERATORS

93

Moreover, since, owing to Eq. (3.137), the density operator of a mixed state expressed in the basis {|i } is given by  ρ= Wk |k k | k

the statistical entropy (3.151) yields 

S=−

i





Wk i |k k | ln

k

 Wl |l l | |i 

l

or, due to the orthonormality properties (3.138)     S = −kB Wi i | ln Wl |l l | |i  i

(3.152)

l

Next, to calculate the operator ln A averaged over |i , where  A= Wl |l l | l

we use the following formal expansion of the logarithm of some function A given by  ln A = C k Ak (3.153) k

where Ck are the coefficients involved in the expansion of the logarithm. Then, the logarithm involved on the right-hand side of Eq. (3.152) expands as   k     Wl |l l | = Ck Wl |l l | (3.154) ln l

k

l

Next, observe that, when k = 2, it reads  2   Wl |l l | = Wl |l l |Ws |s s | l

s

l

or, since Ws is a scalar,  2   Wl |l l | = Wl Ws |l l |s s | l

s

l

so that, due to the orthonormality property (3.150),  2   Wl |l l | = Wl |l Ws δls s | l

s

l

After simplification using the orthonormality properties (3.138) of the basis, that simplifies to  2   Wl |l l | = (Wl )2 (|l l |)2 l

l

CH003.tex

94

10/3/2011

21: 46

Page 94

QUANTUM MECHANICS REPRESENTATIONS

Again, by recurrence one obtains for any value of k  k   Wl |l l | = (Wl )k |l l | l

l

so that Eq. (3.154) takes the form      ln Wl |l l | = Ck (Wl )k |l l | l

k

l

Moreover, due to the latter result, the diagonal matrix elements of Eq. (3.154) read      i | ln Wl |l l | |i  = Ck i | (Wl )k (|l l |)|i  (3.155) l

k

l

or, in view of the orthonormality properties (3.138), one has i |(Wl )k (|l l |)|i  = (Wl )k δil Hence, after simplification using the property of δil , Eq. (3.155) becomes     i | ln Wl |l l | |i  = Ck (Wi )k l

k

Hence, according to the formal expression of the expansion (3.153), in which Wl plays now the role of the function A, we have    i | ln Wl |l l | |i  = ln Wi l

Thus, the entropy given by Eq. (3.152) transforms to the simple form  S = −kB Wi ln Wi

(3.156)

i

which is the usual statistical expression of entropy in information theory. Of course, the probabilities may depend on time, so that the statistical entropy depends also on time. Thus, Eq. (3.156) may be written for any time  S = −kB Wi (t) ln Wi (t) (3.157) i

3.4.5

Density operator representations

Start from the general expression (3.137) of the density operator ρ of a mixed state, that is,  ρ= Wi |i i | (3.158) i

where Wi is the probability for the ket |i  to be occupied. This operator may be expressed in the basis {|{Q}} of the eigenstates of the position operator as it will be now seen.

CH003.tex

10/3/2011

21: 46

Page 95

3.4

3.4.5.1 Position representation of the position operator Q, that is,

DENSITY OPERATORS

95

For this purpose, write the eigenvalue equation

Q|{Q} = Q|{Q} In the basis {|{Q}}, the matrix elements of the density operator (3.158) read  {Q}|ρ|{Q } = Wi {Q}|i i |{Q } i

The scalar products involved in this last equation are the wavefunctions given by {Q}|i  = i (Q)

i |{Q } = ∗i (Q )

and

Hence, the matrix elements, which may be denoted as ρ(Q, Q ), become  {Q}|ρ|{Q } = Wi i (Q)∗i (Q ) ≡ ρ(Q, Q )

(3.159)

i

the corresponding diagonal matrix elements denoted ρ(Q, Q) reduce to  {Q}|ρ|{Q} = Wi |i (Q)|2 ≡ ρ(Q, Q)

(3.160)

i

3.4.5.2 Momentum representation Now, write the eigenvalue equation of the momentum P as P|{P} = P|{P} In the basis of the eigenstates of the position operator, the matrix elements of the density operator are, comparing Eq. (3.158),  {P}|ρ|{P } = Wi {P}|i i |{P } i

The scalar products involved here are the wavefunctions in the momentum representation, that is, {P}|i  = i (P)

i |{P } = ∗i (P )

and

Thus, the matrix elements ρ(P, P ) become ρ(P, P ) = {P}|ρ|{P } =



Wi i (P)∗i (P )

(3.161)

i

the corresponding diagonal matrix elements being  ρ(P, P) = {P}|ρ|{P} = Wi |i (P)|2

(3.162)

i

3.4.5.3 Wigner distribution function Now, consider for one dimension in the position representation the following off-diagonal matrix elements of the density operator:  η η   η  η  ρ Q + ,Q − = Q+ |ρ| Q − 2 2 2 2

CH003.tex

96

10/3/2011

21: 46

Page 96

QUANTUM MECHANICS REPRESENTATIONS

Then it is possible to write from this matrix element the following function:   +∞  η η −iPη ρ Q + ,Q − fw (P, Q) = exp dη 2 2 

(3.163)

−∞

Next, multiply both members of this last equation by 21 π and integrate over all the momentum:   +∞ +∞ +∞  η −iPη 1 1 η exp dηdP fw (P, Q) dP = ρ Q + ,Q − 2π 2π 2 2  −∞

or

−∞ −∞

(3.164)

⎧ +∞ ⎫   ⎬ +∞ +∞   ⎨  1 η η 1 −iPη fw (P, Q) dP = ρ Q + ,Q − exp dP dη ⎭ 2π 2 2 2π ⎩  −∞

−∞

−∞

(3.165) Next, owing to the distribution theory leading to Eq. (18.60), the last integral of the right-hand part of Eq. (3.165) reads ⎧ +∞ ⎫   ⎬  1 ⎨ −iPη exp dP = δ(η) ⎭ 2π ⎩  −∞

so that Eq. (3.164) becomes 1 2π

+∞ +∞  η η fw (P, Q) dP = ρ Q + ,Q − δ(η) dη 2 2

−∞

−∞

Therefore, according to the fact that δ(η) is zero, except if η = 0, for which δ(η) = 1, and keeping in mind Eq. (3.160), this last expression reduces to 1 2π

+∞ fw (P, Q) dP = ρ(Q, Q) = f (Q)

(3.166)

−∞

The function fw (P, Q) (3.163), known as the Wigner distribution function, may be viewed as corresponding from quantum mechanics to the classical distribution function in the phase space f (P, Q). However, it must be observed that the Wigner distribution function may be negative, that is, impossible for the classical distribution function f (P, Q). This aspect is the cost to be paid by the requirement to save the Heisenberg uncertainty relations, which forbid the simultaneous knowledge of the position and of the momentum.

3.4.6

Dynamics

3.4.6.1 Schrödinger picture At the difference of the other operators, which do not depend on time in the Schrödinger picture, the density operator is time dependent

CH003.tex

10/3/2011

21: 46

Page 97

3.4

DENSITY OPERATORS

97

in this representation because it is built up from the kets and the corresponding bras, which evolve with time according to the time-dependent Schrödinger equation. 3.4.6.1.1 Populations and coherences Start from the general expression (3.137) of the density operator of a mixed state. In the Schrödinger picture, the kets and bras are time dependent, so that at the difference of the other operators of quantum mechanics, the density operators must be time dependent, that is, when time is taken into account, Eq. (3.137) must read  ρ(t)SP = Wi |i (t)i (t)| (3.167) i

where the Wi are time-dependent probabilities. Now, consider the eigenvalue equation of a Hermitian operator A A|n  = An |n  Next, consider a matrix element of the density operator in the basis {|n }: ρnm (t)SP = n |ρ(t)SP |m 

(3.168)

The time-dependent off-diagonal matrix elements of the density operator are known as coherences, whereas the diagonal corresponding ones are known as populations. Using Eq. (3.167) gives  ρnm (t)SP = Wi n | i (t)i (t)|m  i

This latter result may be also written for the coherences and for the populations, respectively  ρnm (t)SP = Wi Cni (t)Cim (t) i

ρnn (t)SP =



Wi |Cni (t)|2

i

with Cni (t) = n | i (t) 3.4.6.1.2 Liouville equation In order to get the time dependence of the density operator, first start from its expression (3.167) for a mixed state. Since the Wi are time independent, the partial derivative of Eq. (3.167) is      ∂ρ(t)SP ∂|i (t)i (t)| Wi = (3.169) ∂t ∂t i

The time derivative of the right-hand side of this last equation is, of course,       ∂|i (t)i (t)| ∂|i (t) ∂i (t)| = i (t)| + |i (t) ∂t ∂t ∂t

CH003.tex

98

10/3/2011

21: 46

Page 98

QUANTUM MECHANICS REPRESENTATIONS

Again, recall that thetime-dependent Schrödinger equation and its Hermitian conjugate are     ∂i (t)| ∂|i (t) and −i = H|i (t) = i (t)|H i ∂t ∂t where H is the Hamiltonian. Thus, Eq. (3.169) becomes     ∂ρ(t)SP = Wi H|i (t)i (t)| − Wi |i (t)i (t)|H i ∂t i

i

or, owing to Eq. (3.167) and since Wi commutes with H,   ∂ρ(t)SP = Hρ(t)SP − ρ(t)SP H i ∂t so that  i

∂ρ(t)SP ∂t

 = [H, ρ(t)SP ]

(3.170)

that is, the Liouville–Von Neumann equation also called the Liouville equation or the Von Neumann equation. Note the difference in the sign of the commutator when passing from this equation, which applies to density operator, to that of (3.94) dealing with the observables. The reason is that the density operator is not an observable but is constructed from projectors and thus from kets and bras. The sign difference between Eq. (3.170) governing the time dependence of the density operator and that of (3.95) giving the time dependence of some operators other than the density operator, in the Heisenberg picture, that is,   ∂A(t)HP i| | = |[AHP (t), H]| ∂t 3.4.6.1.3 Density operators in statistical equilibrium When an isolated system is not in statistical equilibrium, its total density operator changes with time:   ∂ρ Tot (t)SP = 0 ∂t and will continue to change until the system has attained its statistical equilibrium:   ∂ρTot (t)SP =0 ∂t In this special situation, it results from Eq. (3.170) that, at equilibrium, it is necessary that [H, ρTot (t)SP ] = 0 3.4.6.1.4 Energy representation of the density operator Owing to the appearance of the Hamiltonian H on the right-hand side of the Liouville–Von Neumann

CH003.tex

10/3/2011

21: 46

Page 99

3.4

DENSITY OPERATORS

99

Eq. (3.170), it may be of interest to consider the matrix representation of this equation on the basis of the eigenvectors of the Hamiltonian. Thus, write the eigenvalue equation of the Hamiltonian: H|n  = En |n 

(3.171)

Then, on the basis {|n }, the matrix representation of the Liouville Eq. (3.170) takes the form   ∂n |ρ(t)SP |m  i = n |H ρ(t)SP |m  − n |ρ(t)SP H|m  ∂t Due to the eigenvalue equation (3.171), this equation transforms to   ∂n |ρ(t)SP |m  i = En n |ρ(t)SP |m  − Em n |ρ(t)SP |m  ∂t which may also be written using the notations (3.168) for populations and coherences   ∂ρnm (t)SP i (3.172) = (En − Em )ρnm (t)SP ∂t Then, by integration of Eq. (3.172), one obtains ρnm (t)SP = ρnm (0)SP e−i(En −Em )t/

(3.173)

Observe that it appears from Eq. (3.173) that the populations (corresponding to n = m) remain constant. 3.4.6.1.5 Canonical transformation on the density operator involving the Schrödinger evolution operator Consider the density operator at initial time t = 0. Equation (3.167) reads  ρ(0)SP = Wi |i (0)i (0)| (3.174) i

At time t, the Wi being constant, the SP density operator becomes  ρ(t)SP = |i (t)SP i (t)SP |

(3.175)

i

In the time evolution operator formalism, the time dependence of the kets and of the corresponding bras is given by Eq. (3.77): |i (t)SP  = U(t)|i (0)SP 

and

i (t)SP | = i (0)SP |U(t)†

(3.176)

where U(t) is the time evolution operator (3.82) governed by the Hamiltonian of the system, that is, U(t) = (e−iHt/ )

and

U(t)† = U(t)−1 = (eiHt/ )

The time-dependent density operator is therefore  ρ(t)SP = U(t)|i (0)i (0)|U(t)† i

Hence, in view of Eq. (3.174) ρ(t)SP = U(t)ρ(0)U(t)† = U(t)ρ(0)U(t)−1

(3.177)

CH003.tex

100

10/3/2011

21: 46

Page 100

QUANTUM MECHANICS REPRESENTATIONS

or, writing explicitly the time evolution operator ρ(t)SP = (e−iHt/ )ρ(0)SP (e+iHt/ )

(3.178)

Note that in the canonical transformation of (3.177) or (3.178), the signs have changed with respect to those appearing in the time dependence of the Heisenberg picture or observables that, according to Eqs. (3.89) and (3.88), are A(t)HP = (eiHt/ )A(e−iHt/ ) = U(t)† AU(t) 3.4.6.2 Interaction picture Liouville equation Suppose that the system that is studied involves a Hamiltonian H that may be split into an unperturbed part H◦ and a perturbation V, according to H = H◦ + V Due to the partition of the Hamiltonian, the Liouville–Von Neumann equation (3.170) takes the form   ∂ρ(t)SP i (3.179) = [H◦ , ρ(t)SP ] + [V, ρ(t)SP ] ∂t Next, keeping in mind that the SP density operator at time t is given by Eq. (3.175),  ρ(t)SP = Wi |i (t)SP i (t)SP | i

and since the Wi are constant, it is clear that the corresponding IP density operator is given by  ρ(t)IP = Wi |i (t)IP i (t)IP | (3.180) i

whereas Eq. (3.118) relating the IP and SP kets is |(t)IP  = U◦ (t)−1 |(t)SP 

(3.181)

where U◦ (t) = (e−iH

◦ t/

)

(3.182)

Hence, due to Eq. (3.181) and to its Hermitian conjugate, the IP density operator (3.180) reads ρ(t)IP ≡ U◦ (t)−1 ρ(t)SP U◦ (t)

(3.183)

Then, premultiplying this equation by U◦ (t) and postmultiplying it by its inverse, leads to U◦ (t)ρ(t)IP U◦ (t)−1 = U◦ (t)U◦ (t)−1 ρ(t)SP U◦ (t)U◦ (t)−1 so that, on simplification of the right-hand side, one obtains the relation inverse to (3.183), that is, ρ(t)SP = U◦ (t)ρ(t)IP U◦ (t)−1

(3.184)

CH003.tex

10/3/2011

21: 46

Page 101

3.4

DENSITY OPERATORS

101

Moreover, due to Eq. (3.184), Eq. (3.179) yields   ∂ρ(t)SP i = [H◦ , U◦ (t)ρ(t)IP U◦ (t)−1 ] + [V, U◦ (t)ρ(t)IP U◦ (t)−1 ] (3.185) ∂t On the other hand, the partial time derivative of Eq. (3.184), reads    ◦    ∂ρ(t)SP ∂U (t) ∂ρ(t)IP IP ◦ −1 ◦ = ρ(t) U (t) + U (t) U◦ (t)−1 ∂t ∂t ∂t  ◦ −1  ∂U (t) + U◦ (t)ρ(t)IP (3.186) ∂t Then, by identification of Eqs. (3.185) and (3.186), one obtains [H◦ , U◦ (t)ρ(t)IP U◦ (t)−1 ] + [V, U◦ (t)ρ(t)IP U◦ (t)−1 ]    ◦  ∂ρ(t)IP ∂U (t) ρ(t)IP U◦ (t)−1 + U◦ (t) U◦ (t)−1 = i ∂t ∂t  ◦ −1  ◦ IP ∂U (t) + U (t)ρ(t) ∂t

(3.187)

Moreover, observe that, according to Eq. (3.81), and since H◦ is Hermitian, the Schrödinger equation governing the dynamics of the unitary evolution operator U◦ (t) and its Hermitian conjugate is  ◦   ◦ −1  ∂U (t) ∂U (t) i = H◦ U◦ (t) and −i = U◦ (t)−1 H◦ ∂t ∂t These equations allow one to write the first and third right-hand-side terms of Eq. (3.187) according to  ◦  ∂U (t) i ρ(t)IP U◦ (t)−1 = H◦ U◦ (t)ρ(t)IP U◦−1 (t) ∂t ◦

iU (t)ρ(t)

 IP

∂U◦ (t)−1 ∂t



= −U◦ (t)ρ(t)IP U◦ (t)−1 H◦

Hence, the sum of these two terms appearing in Eq. (3.187) reads  ◦   ◦ −1  ∂U (t) ∂U (t) i ρ(t)IP U◦ (t)−1 + U◦ (t)ρ(t)IP = [H◦ , U◦ (t)ρ(t)IP U◦ (t)−1 ] ∂t ∂t (3.188) Hence, the left-hand side of Eq. (3.188) is equivalent to the first and third right-hand terms of Eq. (3.187), whereas the right-hand term of Eq. (3.188) is the same as the first commutator appearing on the left-hand side of Eq. (3.187). As a consequence, Eq. (3.187) simplifies to   ∂ρ(t)IP iU◦ (t) U◦ (t)−1 = [V, U◦ (t)ρ(t)IP U◦ (t)−1 ] (3.189) ∂t On the other hand, Eq. (3.179) may be transformed using Eq. (3.184) to   ∂ρ(t)SP i = [H◦ , ρ(t)SP ] + [V, U◦ (t)ρ(t)IP U◦ (t)−1 ] ∂t

CH003.tex

102

10/3/2011

21: 46

Page 102

QUANTUM MECHANICS REPRESENTATIONS

which, owing to Eq. (3.189), yields     ∂ρ(t)SP ∂ρ(t)IP 1 = [H◦ , ρ(t)SP ] + U◦ (t) U◦ (t)−1 ∂t i ∂t

(3.190)

On the other hand, writing explicitly the right-hand commutator of Eq. (3.189) gives   ∂ρ(t)IP U◦ (t)−1 = VU◦ (t)ρ(t)IP U◦ (t)−1 − U◦ (t)ρ(t)IP U◦ (t)−1 V iU◦ (t) ∂t Then, postmultiplying both members of this last equation by U◦ and premultiplying them by its inverse, allows us to write     ∂ρ(t)IP iU◦ (t)−1 U◦ (t) U◦ (t)−1 U◦ (t) ∂t = U◦ (t)−1 (VU◦ (t)ρ(t)IP U◦ (t)−1 )U◦ (t) − U◦ (t)−1 (U◦ (t)ρ(t)IP U◦ (t)−1 V) U◦ (t) or, on simplification   ∂ρ(t)IP = U◦ (t)−1 VU◦ (t)ρ(t)IP − ρ(t)IP U◦ (t)−1 VU◦ (t) i ∂t a result that may be written   ∂ρ(t)IP = V(t)IP ρ(t)IP − ρ(t)IP V(t)IP i ∂t

(3.191)

where VIP (t) is given, in agreement to Eq. (3.88), by V(t)IP = U◦ (t)−1 VU◦ (t)

(3.192)

Finally, Eq. (3.191) may be expressed in terms of a commutator to give   ∂ρ(t)IP i = [V(t)IP , ρ(t)IP ] ∂t

(3.193)

that is, the IP Liouville–Von Neumann equation governing the IP density operator, which involves the same sign for the Hamiltonian and density operator commutator as that appearing in the corresponding SP Liouville equation (3.170). 3.4.6.3 Integration of the IP Liouville Equation Formal integration of the IP Liouville–Von Neumann equation from t0 to t leads to the following integral equation:  ρ(t)

IP

= ρ(t0 ) + IP

1 i

 t

[V(t  − t0 )IP , ρ(t  − t0 )IP ] dt 

(3.194)

t0

with, due to Eqs. (3.192) and (3.182), V(t  − t0 )IP = eiH

◦ (t  −t

0 )/

Ve−iH

◦ (t  −t

0 )/

(3.195)

If the potential V is small with respect to H◦ , the integral equation (3.194) may be solved by successive approximations. For this purpose, observe that the

CH003.tex

10/3/2011

21: 46

Page 103

3.4

DENSITY OPERATORS

103

time-dependent IP time evolution density operator involved in the commutator appearing on the right-hand side of Eq. (3.194) may be found by using an equation similar to Eq. (3.194), that is, 

ρ(t  − t0 )IP = ρ(t0 )IP +

1 i

 t −t0 [V(t  − t0 )IP , ρ(t  − t0 )IP ] dt  t0

so that Eq. (3.194) becomes [V(t  − t0 )IP , ρ(t  − t0 )IP ]



1 = [V(t − t0 ) , ρ(t0 ) ] + i 

IP

 t

IP

[V(t  − t0 )IP , [V(t  − t0 )IP , ρ(t  − t0 )IP ]] dt 

0

Then, inserting this expression into Eq. (3.194) yields  IP

ρ(t)

= ρ(t0 ) + IP

1 i

 t

[V(t  − t0 )IP , ρ(t0 )IP ] dt 

t0

 +

1 i

2

t

t 

[V(t  − t0 )IP , [V(t  − t0 )IP , ρ(t  − t0 )IP ]] dt  dt 

t0 t0

Again, by iteration, one obtains ρ(t)

IP

1 = ρ(t0 ) + i

t

[V(t  − t0 )IP , ρ(t0 )IP ] dt 

IP

(3.196)

t0



1 + i

2

t

t 

[V(t  − t0 )IP , [V(t  − t0 )IP , ρ(t0 )IP ]] dt  dt 

t0 t0



1 + i

3  t t  t  t0 t0 t0

[V(t  − t0 )IP , [V(t  − t0 )IP , [V(t  − t0 )IP , ρ(t  − t0 )IP ]]] dt  dt  dt 

This operation may be repeated any number of times. However, if the perturbation V is weak, the treatment may be limited to the second order in the IP perturbation operator so that Eq. (3.196) becomes truncated at the second order in the perturbation according to ρ(t)

IP

1

ρ(t0 ) + i

t

IP

[V(t  − t0 )IP , ρ(t0 )IP ] dt 

t0



1 + i

2

t

t 

t0 t0

[V(t  − t0 )IP , [V(t  − t0 )IP , ρ(t0 )IP ]] dt  dt  (3.197)

CH003.tex

104

10/3/2011

21: 46

Page 104

QUANTUM MECHANICS REPRESENTATIONS

Recall that the IP perturbation operator is given by Eq. (3.192) and that, when the expression of the IP density operator has been obtained with the help of Eq. (3.197), one may retrieve the time-dependent density operator using Eq. (3.184).

3.5

CONCLUSION

This chapter, which was devoted to different representations of quantum mechanics, has lead to important useful developments. (i) Matrix representation allowing one to replace the eigenvalue equation of an Hermitian operator to be solved by a corresponding matrix eigenvalue equation susceptible to be easily numerically solved, which is of great interest for the study of quantum anharmonic oscillators. (ii) Wave mechanics, which is the representation of quantum mechanics in geometrical space, is in many situations such as atoms or molecules more tractable than matrix or quantum mechanics. Although there is less interest in quantum oscillators than matrix mechanics, it will remain of some interest in visualizing some results dealing with these oscillators. (iii) Density operator approaches are very powerful when studying many-particle systems, particularly for statistical equilibrium situations leading to thermal equilibrium, and will be widely used when studying thermal properties of quantum oscillators. (iv) Time-dependent representations other than the Schrödinger picture where the time dependence resides in the quantum states, which constitute the Heisenberg picture where the time dependence is contained in the Hermitian operators, and the interaction picture, which is a description intermediate between the Schrödinger and Heisenberg pictures and which will be very useful when studying the irreversible dynamics of quantum oscillators coupled to a thermal bath. The important results concerning the time-dependent Schrödinger, Heisenberg, and interaction pictures are collected into the two following lists: Schrödinger and Heisenberg pictures Schrödinger equation and evolution operator: i

∂ |(t)SP  = H|(t)SP  ∂t

Time-dependent ket in the Schrödinger picture: |(t)SP  = U(t)SP |(0)SP  Time-dependent evolution operator in the Schrödinger picture: U(t)SP = e−iHt/ Time-dependent operators in the Heisenberg picture: A(t)HP = U(t)SP−1 A(0)U(t)SP

CH003.tex

10/3/2011

21: 46

Page 105

3.5

CONCLUSION

105

Dynamic equation governing evolution operators in the Schrödinger picture:   ∂U(t)SP i = HU(t)SP ∂t Dynamic equation governing operators in the Heisenberg picture:   ∂A(t)HP i = [A(t)HP , H] ∂t

Interaction picture Hamiltonian partition and corresponding evolution operators: H = H◦ + V ◦ U◦ (t) = e−iH t/

and

U(t) = e−iHt/

Relation between IP and SP evolution operators: U(t)SP = U◦ (t)U(t)IP Time-dependent operators A in the interaction picture: A(t)IP = U◦ (t)−1 AU◦ (t) Dynamic equation governing the interaction picture evolution operator:   ∂U(t)IP = V(t)IP U(t)IP i ∂t Connection between SP and IP time-dependent kets: |(t)SP  = U◦ (t)|(t)IP  Those dealing with density operators are given as follows: Density operators Definition of density operators:  ρ= Wi |i i | i

Average values performed over density operators: Aρ = tr{ρA}

CH003.tex

106

10/3/2011

21: 46

Page 106

QUANTUM MECHANICS REPRESENTATIONS

Liouville equation in the Schrödinger picture:   ∂ρ(t)SP i = [H, ρ(t)SP ] ∂t Statistical entropy: S = −kB tr{ρ ln ρ}

BIBLIOGRAPHY C. Cohen-Tannoudji, B. Diu, and F. Laloe. Quantum Mechanics. Wiley-Interscience: Hoboken, NJ, 2006. A. S. Davydov. Quantum Mechanics, 2nd ed. Pergamon Press: Oxford, 1976. P. A. M. Dirac. The Principles of Quantum Mechanics, 4th ed. Oxford University Press: 1982. W. H. Louisell. Quantum Statistical Properties of Radiation. Wiley: New York, 1973. A. Messiah. Quantum Mechanics. Dover Publications: New York, 1999. L. I. Schiff. Quantum Mechanics, 3rd ed. McGraw-Hill: New York, 1968.

CH004.tex

10/3/2011

CHAPTER

21: 55

Page 107

4

SIMPLE MODELS USEFUL FOR QUANTUM OSCILLATOR PHYSICS INTRODUCTION Before studing quantum oscillators, which is the principal aim of the present book, it may be useful to apply the information of the previous chapters dealing with the basis of quantum mechanics to three simple models that will be of interest in the future. The first one is the particle-in-a-box model, which comprises a single particle enclosed in a box where the potential is zero, this same potential being infinite beyond the box walls. Applying simply the wave mechanics, we shall get quantized energy levels and their associated wavefunctions, the node number of which increases with the energy. It will also be useful to illustrate the quantization of the energy levels and the decrease of the associated wavelength when the energy rises, two concepts we shall meet when discussing the oscillators. The second model to which the present chapter is devoted deals with the interaction between two energy levels, which will be of interest when focusing attention on the local interaction between two excited states of two different oscillators, a situation that occurs in the area of Fermi resonances. Finally, the last section treats the probability for a system to pass from one of its stationary energy levels to another if a potential perturbs it. Using a formalism that will later be applied to the interaction of oscillators with the electromagnetic field, it will lead to the important Fermi golden rule.

4.1

PARTICLE-IN-A-BOX MODEL

Consider a particle of mass m enclosed in a box of volume V given by V = a x ay az

(4.1)

dimensions in which the potential is zero while it is infinite outside. Its kinetic energy is T=

1 2 (P + Py2 + Pz2 ) 2m x

Quantum Oscillators, First Edition. Olivier Henri-Rousseau and Paul Blaise. © 2011 John Wiley & Sons, Inc. Published 2011 by John Wiley & Sons, Inc.

107

CH004.tex

108

10/3/2011

21: 55

Page 108

SIMPLE MODELS USEFUL FOR QUANTUM OSCILLATOR PHYSICS

where the Pi2 are the components x, y, and z of the momentum. In the wave mechanics representation, the momentum operator obeys Eq. (3.51): P = −i

∂ ∂Q

Thus, the wave mechanics description of the particle-in-a-box kinetic operator reads  2  2 ∂ ∂2 ∂2 T =− + + 2m ∂x 2 ∂y2 ∂z2 where x, y, and z are the components of Q. Furthermore, since the potential is assumed to be zero inside the box, the wave mechanics description of the potential is simply V =0 The Hamiltonian of the system, which is the sum of the kinetic and potential operators, is in the position representation   2 2 ∂2 ∂2 ∂ H=− + + 2m ∂x 2 ∂y2 ∂z2 Now, for this three-dimensional (3D) model, the wavefunction of the particle can be written as the product of the wavefunctions along the three independent dimensions, that is, (x, y, z) = (x)( y)(z)

(4.2)

Hence, the eigenvalue equation of the Hamiltonian, that is, the time-independent Schrödinger equation, takes the form   2 ∂2 (x) ∂2 (y) ∂2 (z) − ( y)(z) + (x)(z) + (x)(y) 2m ∂x 2 ∂y2 ∂z2 = E(x)(y)(z)

(4.3)

where E is the Hamiltonian eigenvalue.

4.1.1

Solving the 3D Schrödinger equation

Now, the Hamiltonian eigenvalue E may be written as the sum of the energies along the three dimensions, that is, E = Ex + E y + E z

(4.4)

Thus, the Schrödinger equation (4.3) splits into three independent and equivalent Schrödinger equations corresponding to the three dimensions of the geometrical space, according to  2  ∂ (x) = −kx2 (x) (4.5) ∂x 2 

∂2 (y) ∂y2

 = −ky2 (y)

(4.6)

CH004.tex

10/3/2011

21: 55

Page 109

4.1



∂2 (z) ∂z2

PARTICLE-IN-A-BOX MODEL

109

 = −kz2 (z)

(4.7)

kx2 =

2m Ex 2

(4.8)

ky2 =

2m Ey 2

(4.9)

with

2m Ez (4.10) 2 Observe that since outside the box the potential is infinite, it is impossible to the particle to get out of the box (Fig. 4.1). Therefore, the probability for the particle to be outside the box is zero, and, thus, since this probability is the squared modulus of the wavefunction, the three wavefunctions satisfy the following boundary conditions, for example, leading for the x component to kz2 =

(x) = 0

if

−∞ < x  0

and

ax  x < ∞

(4.11)

(y) = 0

if −∞ < y  0

and

ay  y < ∞

(4.12)

(z) = 0

if

−∞ < z  0

and

az  z < ∞

(4.13)

It appears, therefore, that the partial differential equations (4.5)–(4.7) to be solved are subject to the boundary conditions (4.11)–(4.13). The general solution of Eq. (4.5) is of the form (x) = Ax sin (kx x) + Bx cos (kx x) z

az

0

ay

ax x Figure 4.1

Particle-in-a-box model.

y

CH004.tex

110

10/3/2011

21: 55

Page 110

SIMPLE MODELS USEFUL FOR QUANTUM OSCILLATOR PHYSICS

where Ax and Bx are two constants. Next, owing to the boundary condition (4.11) implying that, at x = 0, (0) = 0, it follows that Bx = 0, so that (x) = Ax sin (kx x)

(4.14)

Now, the same boundary condition (4.11) implying that x = a, leads to write (a) = 0 so that, since A  = 0 Eq. (4.14), it is necessary that sin (kx ax ) = 0 such a condition being verified if kx ax = nx π

(4.15)

where nx is a number that may take a priori all the integer values between 0 and ∞. That leads one to write the solution (4.14) as   nx π nx (x) = Ax sin x (4.16) ax In a similar way, one can obtain for the solutions of Eqs. (4.6) and (4.7) subject, respectively, to the boundary conditions (4.12) and (4.13)   ny π ny (y) = Ay sin y (4.17) ay  nz (z) = Az sin

nz π z az

 (4.18)

with ky ay = ny π

and

kz az = nz π

(4.19)

The normalization condition of a wavefunction, which is  a nx (x)2 dx = 1 0

reads for the wavefunction (4.16)    a nx π A2x sin2 x dx = 1 ax 0

(4.20)

Next, using the trigonometric relation sin2 (z) = Eq. (4.20) yields

 A2x

0

a

 1 2

Moreover, due to the fact that 

a 0

1 2

(1 − cos 2z)

  nx π 1 − cos 2 x dx = 1 ax   nx π cos 2 x dx = 0 ax

(4.21)

CH004.tex

10/3/2011

21: 55

Page 111

4.1

PARTICLE-IN-A-BOX MODEL

111

Eq. (4.21) reduces to 1 2 2 A x ax

so that Eq. (4.16) becomes

 nx (x) =

=1

  2 nx π sin x ax ax

In a similar way, one would obtain for the wavefunctions (4.17) and (4.18),    ny π 2 ny (y) = sin y ay ay  nz (z) =

4.1.2

  2 nz π sin z az az

(4.22)

(4.23)

(4.24)

3D Wavefunctions and energy levels

Due to Eq. (4.2), and to Eqs. (4.22)–(4.24), the total wavefunction reads       ny π nx π 23/2 nz π x sin y sin z nx ,ny ,nz (x, y, z) = √ sin ax ay az V

(4.25)

where Eq. (4.1) has been used, relating the dimensions ax , ay , and az of the box to its volume V. Of course, the solutions corresponding to nx = 0, ny = 0, or nz = 0 are without physical meaning since it would imply erroneously that the wavefunction (4.25) and thus the probability of the particle in all the box would be zero. Thus, all the quantum numbers nx , ny , and nz must be integers, starting from unity. The wavefunction (4.25) appears to be a product of stationary wavefunctions of the form nx ,ny ,nz (x, y, z) = nx (x)ny ( y)nz (z) with



  2 2π sin x ax λn x



  2 2π sin y ay λn y



  2 2π sin z az λn z

nx (x) =

ny (y) =

nz (z) =

and where the λnx , λny , and λnz are wavelengths obeying       2ay 2ax 2az λny = λnz = λnx = nx ny nz

(4.26)

CH004.tex

10/3/2011

En (h2/8ma2)

112

21: 55

Page 112

SIMPLE MODELS USEFUL FOR QUANTUM OSCILLATOR PHYSICS

25

10

5

0

|ψ5(x)|2

ψ4(x)

|ψ4(x)|2

ψ3(x)

|ψ3(x)|2

ψ2(x)

|ψ2(x)|2

ψ1(x)

|ψ1(x)|2

n⫽5

20

15

ψ5(x)

n⫽4

n⫽3

n⫽2 n⫽1 0

a/2

X

a

0

a/2

X

a

Figure 4.2 One-dimensional particle-in-a-box model. Energy levels and corresponding wavefunctions and probability densities for the four lowest quantum numbers.

In addition, Eqs. (4.8)–(4.10) combined with (4.15)–(4.19) allow us to get following results, with the energies corresponding to the x, y, and z components:    2 2  2 2 π  π 2 2 π  2 2 n nz2 E nx = n E = E = (4.27) ny nz x y 2max2 2may2 2maz2 Hence, after passing from  to h, the total energy (4.4) becomes   ny2 nz2 h2 nx2 + 2+ 2 Enx ,ny ,nz = 8m ax2 ay az

(4.28)

It must be emphasized that, since nx , nz , and nz are integers, the energy levels (4.28) are quantized, a result that will be also found later for quantum harmonic oscillators. Figure 4.2, which deals with the x component of the 3D model, gives the dimensionless energy levels and the corresponding wavefunctions for the four lowest quantum number nx . Hence, the nodes of the wavefunction are increasing with the quantum number and the corresponding energy level, a situation that will be met later for quantum harmonic oscillators and that is related to the de Broglie wavelength, we shall consider some later. Moreover, when the box is cubic, that is, when ax = ay = az = a and due to Eq. (4.1), the equation (4.28) giving the energy levels simplifies to Enx ,ny ,nz =

h2 (nx2 + ny2 + nz2 ) 8m V2/3

(4.29)

CH004.tex

10/3/2011

21: 55

Page 113

4.1

PARTICLE-IN-A-BOX MODEL

113

Then, in terms of the energy units E ◦ E◦ =

h2 8mV2/3

and for the two lowest values 1 and 2 of the quantum numbers nx , ny , and nz , the lowest energy levels (4.29) appear to be those given in the tabular expression (4.30): nx 1 1 1 2 2 2 1

ny 1 1 2 1 2 1 2

nz 1 2 1 1 1 2 2

Enx ,ny ,nz 1 6 6 6 9 9 9

(4.30)

Inspection of this data shows that different energy levels may have the same energy, that is, they are degenerate.

4.1.3

Consequences useful for quantum oscillators

As seen above, the particle-in-a-box model leads to the important result of the energy quantization. However, it leads also to some other interesting consequences, for example, the de Broglie wavelength of a quantum particle and a simple understanding of how the energy quantization disappears when the physical dimensions are progressively increasing. Observe that the energy levels (4.28) are only kinetic in nature since the potential energy is zero inside the box. Thus, they may be written as the sum of the kinetic energies along the three dimensions, that is, 1  2 Enx ,ny ,nz = Pnx + Pn2y + Pn2z 2m Then, by identification of this formal expression with Eq. (4.28), one obtains Pnx = ±

h nx 2ax

so that, due to Eq. (4.26), it appears that the wavelengths are given by h λ nx =

Pn

(4.31)

x

which is the Louis de Broglie’s relation, which has been experimentally verified for microscopic particles. Note that Fig. 4.2 reveals that the number of nodes of the wavefunctions are increasing with the quantum number nx , reflecting the fact that in agreement with Eq. (4.31), the modulus of the momentum raises when the de Broglie wavelength decreases, leading, therefore, to an enhancement of the energy since there is no potential.

CH004.tex

114

10/3/2011

21: 55

Page 114

SIMPLE MODELS USEFUL FOR QUANTUM OSCILLATOR PHYSICS

In the special case of a cubic box, when one of the quantum numbers is increased by a factor of 1, the others remaining constant, Eq. (4.29) reads  2 (nx + 1)2 + ny2 + nz2 h (4.32) Enx +1,ny ,nz = 8m V2/3 Then, the difference between the two successive energy levels (4.29) and (4.32) yields  2 (2nx + 1) h Enx ,nx +1 = Enx +1,ny ,nz − Enx ,ny ,nz = 8m V2/3 which, for large quantum numbers, may be approximated as  2 h nx Enx ,nx +1 = 4m V2/3

(4.33)

This result, which follows from quantum mechanics, holds for microscopic dimensions of atoms or molecules. But there is no reason why it should not also be true for macroscopic systems where the mass (denoted M in place of m) of the particle and the volume in which it is enclosed are those of usual experiences, so that Eq. (4.33) reads  2  nx nx ,nx +1 = h E (4.34) 4M V2/3 As an illustration, when passing from a description of an atomic electron of mass me enclosed in a volume V, which is roughly that of the atom of radius aat , to the description of a ball of mass MB of 1 kg moving in a volume V around 1 m3 , one has respectively me  10−30 kg MB  1 kg

and and

aat  10−10 m a  1m

so that Eq. (4.34) leads to nx ,nx +1 = 10−50 Enx ,nx +1 E

(4.35)

In a similar way, when passing from a description of a proton of mass mp enclosed in a nucleus of volume V, which is roughly that of the third power of the nucleus radius anu , to that of the Earth of mass ME moving around the Sun at the distance aSun  1011 m there are mp  10−27 kg

and

anu  10−15 m

ME  1025 kg

and

aSun  1011 m

so that Eq. (4.34) leads to  nx ,nx +1 = 10−104 Enx ,nx +1 E

(4.36)

Equations (4.35) and (4.36) illustrate the fact that passing from the microscopic to the macroscopic levels drastically decreases the energy gap between two successive energy levels.

CH004.tex

10/3/2011

21: 55

Page 115

4.2 TWO-ENERGY-LEVEL SYSTEMS

115

4.2 TWO-ENERGY-LEVEL SYSTEMS Now, it is time to study the model of two-energy-level systems, which will illustrate the phenomenon called quantum interference between kets, which is a simple consequence of the linear properties of quantum mechanics. Such a model will be later applied when studying the interactions dealing with anharmonically coupled oscillators. But, it is suitable to begin the present approach by starting from equations dealing with the more general model of multiple interacting energy levels.

4.2.1

Multiple interacting energy levels

Consider a system the Hamiltonian of which may be split into two parts according to H = H◦ + V

(4.37)

Now, suppose that the eigenvalue equation of H◦ is known, that is, H◦ |i  = Ei◦ |i 

(4.38)

Since H◦

is Hermitian, its eigenvectors are orthogonal so that, if they are normalized, they verify i |j  = δij

(4.39)

Owing to Eqs. (4.37)–(4.39), and in the basis {|i }, the diagonal matrix elements of the full Hamiltonian are i |H|i  = Ei◦ + i |V|i 

(4.40)

Now, due to Eq. (4.37), the off-diagonal matrix elements are i |H|j  = i |(H◦ + V)|j  In addition, owing to the eigenvalue equation (4.38), there is i |H◦ |j  = i |E ◦j |j  = E ◦j i |j  = E ◦j δij so that only part V of the Hamiltonian (4.37) couples two different eigenkets of H◦ , according to i |H|j  = i |V|j  = βij Now, writing αi = i |H|i 

and

βii = i |V|i 

(4.41)

Eq. (4.40) leads to αi = E ◦i + βii Now, the eigenvalue equation of the full Hamiltonian (4.37) is H|μ  = Eμ |μ  with

μ |ν  = δμν

(4.42)

CH004.tex

116

10/3/2011

21: 55

Page 116

SIMPLE MODELS USEFUL FOR QUANTUM OSCILLATOR PHYSICS

the matrix representation of this eigenvalue equation in the basis of the eigenkets of H◦ being ⎞⎛ ⎞ ⎛ ⎞⎛ ⎞ ⎛ β12 … β1N C1μ 1 0 … 0 C1μ α1 ⎜ ⎜ ⎟ ⎜β21 α2 … … ⎟ ⎜ C2μ ⎟ 1 … …⎟ ⎟⎜ ⎟ = Eμ ⎜0 ⎟ ⎜ C2μ ⎟ (4.43) ⎜ ⎠ ⎝ ⎠ ⎝ ⎠ ⎝ ⎝… … … … … … … … ... ... ⎠ βN1 … … αN 0 … … 1 CNμ CNμ with Cμi = μ |i  One possibility to obtain the eigenvalues Eμ and the corresponding eigenvectors is to diagonalize the left-hand matrix appearing in Eq. (4.43). However, there is yet another possibility, because this eigenvalue equation may be also written as a system of simultaneous equations: ⎧ (α1 − Eμ )C1μ + β12 C2μ + · · · + β1N CNμ =0 ⎪ ⎪ ⎪ ⎨β C =0 + (α − E )C + · · · + · · · 21 1μ 2 μ 2μ ⎪· · · ... + ··· + ··· + ··· ⎪ ⎪ ⎩ βN1 C1μ + ··· + · · · + (αN − Eμ )CNμ = 0 Then, since the coefficients Ciμ cannot be zero, this system of equations is satisfied if the corresponding determinant is zero, that is,



(α1 − E) β12 … β1N



β21 (α2 − E) … …



=0 (4.44)

… … … …



βN1 … … (αN − E)

where we have omitted the subscript μ for the unknown eigenvalues Eμ .

4.2.2

Energies of two interacting levels

In the special situation of two interacting energy levels α1 and α2 , and where β12 is real and thus equal to β21 , the matrix representation of the Hamiltonian reduces to   α1 β (4.45) H = β α2 In order to solve the eigenvalue equation H|±  = E± |±  consider the secular equation that, according to Eq. (4.44), reads



α1 − E β



= 0 with β ≡ β12

β α2 − E

Then, expanding the determinant according to the usual rule, that is, (α1 − E)(α2 − E) − β2 = 0

(4.46)

(4.47)

CH004.tex

10/3/2011

21: 55

Page 117

4.2 TWO-ENERGY-LEVEL SYSTEMS

117

the following second-order equation in E is obtained E 2 + α1 α2 − E(α1 + α2 ) − β2 = 0 the two roots of which are E± =

(α1 + α2 ) ±



(α1 + α2 )2 − 4(α1 α2 − β2 ) 2

or, after simplification (α1 + α2 ) ±



(α1 − α2 )2 + 4β2 2 Hence, the difference between the two eigenvalues is  E+ − E− = (α1 − α2 )2 + 4β2 E± =

(4.48)

(4.49)

On the other hand, the eigenvectors of H appearing in (4.42), and corresponding to the eigenvalues (4.48), are of the form |±  = C1± |1  + C2± |2 

(4.50)

whereas the orthonormality properties (4.42) of these kets leads to − |+  = 0

and

+ | +  = − |−  = 1

(4.51)

so that, due to Eq. (4.50), (C1− 1 | + C2− 2 |)(C1+ |1  + C2+ |2 ) = 0 and thus C1− C1+ 1 |1  + C2− C2+ 2 |2  + C1− C2+ 1 |2  + C1+ C2− 2 |1  = 0 Then, owing to the orthonormality conditions appearing in Eq. (4.39), this last expression reduces to C1− C1+ + C2− C2+ = 0 Likewise, the normality conditions appearing in (4.51) lead to C12− + C22− = 1

and

C12+ + C22+ = 1

(4.52)

When the two interacting levels are degenerate, that is, have the same energy, the two eigenvalues (4.48) of the Hamiltonian H reduce to E± = α ± β

when

α1 = α2 = α

(4.53)

Then, in order to get the expansion coefficients of the H eigenvectors, corresponding to these two eigenvalues, return to Eq. (4.43); however, for the special situation of two interacting levels, that is,    α − E± β C1± =0 (4.54) β α − E± C2± which leads to (α − (α ± β)) C1± = βC2±

CH004.tex

118

10/3/2011

21: 55

Page 118

SIMPLE MODELS USEFUL FOR QUANTUM OSCILLATOR PHYSICS

Rearranging, gives, respectively, for the components of the eigenvectors corresponding to the two eigenvalues C1+ =1 C2 +

and

C 1− = −1 C2 −

where, of course, the complementary equations (4.52) continue to hold.

4.2.3

Approximate solution far from degeneracy

Now, consider the special situation where |α1 − α2 | |β|

(4.55)

4.2.3.1 Eigenvalues Before applying this relation, it is convenient to write the eigenvalues (4.48) in the following form: ⎤ ⎡  1⎣ 4β2 ⎦ E± = (4.56) (α1 + α2 ) ± (α1 − α2 ) 1 + 2 (α1 − α2 )2 where the square root appears to be of the form √ 1+ε

with ε =

4β2 (α1 − α2 )2

and

ε 1

Hence, by expansion of the square root up to first order in ε, one has √ ε 1+ε1+ 2 Thus, when the condition (4.55) is verified, the eigenvalues are    1 2β2 E± = (α1 + α2 ) ± (α1 − α2 ) 1 + 2 (α1 − α2 )2 or E+ = α1 +

β2 β2 and E− = α2 − (α1 − α2 ) (α1 − α2 )

4.2.3.2 Expansion coefficients of the eigenvectors inequalities hold: α1 < 0,

α2 < 0,

(4.57)

Generally, the following

β α2

and

β E− Hence, in view of this new inequality combined with the second one appearing in (4.61), Eqs. (4.59) and (4.60) lead to the following results: |C1+ | |C1− | >> 1 and > i |V|j 

± |H|±   i |H|i  ±

i |V|j 2 i |H|i  − j |H|j 

(4.64)

with ± |H|±  = E±

i |H|i  = αi

i |V|j  = β

Equation (4.64) is the expression for the special case of a two-energy-level system, of second-order perturbation expansion of the eigenvalues of the full Hamiltonian H in terms of the matrix elements of this Hamiltonian in the basis of the Hamiltonian H◦ .

4.2.4

Dynamics

In order to get the dynamics of the system, it is convenient to write the Hamiltonian matrix (4.45) in the following form: ⎛α + α ⎞ ⎛ α −α ⎞ 1 1 2 2 0 β + ⎜ ⎟ ⎜ ⎟ 2 H =⎝ 2 + (4.65) α1 + α 2 ⎠ ⎝ α1 − α 2 ⎠ 0 β − 2 2

CH004.tex

10/3/2011

21: 55

Page 121

4.2 TWO-ENERGY-LEVEL SYSTEMS

or

 H =

α1 + α2 2



 1 +

α 1 − α2 2

121

 K

where the last right-hand-side matrix is given by ⎛ 2β ⎞ +1 ⎜ α1 − α 2 ⎟ K =⎝ ⎠ 2β −1 α1 − α 2

(4.66)

(4.67)

According to Section 1.3.2, since the two right-hand-side Hermitian matrices of Eq. (4.66) commute, they admit the same eigenvectors, so that the two following eigenvalue equations, both involving the same eigenkets |± , are satisfied: H |±  = E± |± 

K |±  = K± |± 

where E± are the Hamiltonian eigenvalues we obtained above, whereas K± are the corresponding eigenvalues of the K matrix. Hence, due to Eq. (4.66), the eigenvalues of H read     α1 + α2 α 1 − α2 E± = (4.68) + K± 2 2 Next, write the matrix (4.67) as follows:   +1 tan θ K = tan θ −1 with tan θ =

2β α1 − α 2

(4.69)

Then, since K and H , which commute, have the same set of eigenvectors |± , the matricial eigenvalue equation of K is similar to that for H given by Eq. (4.58), so that one gets       +1 tan θ C1± 1 0 C1± = K± (4.70) tan θ −1 C 2± 0 1 C2± where the Ck± are the components of the eigenvectors |±  given by Eq. (4.50). The corresponding secular determinant, which must be zero, that is,



1 − K± tan θ



tan θ −1 − K± = 0 leads by expansion to 2 K± − 1 − tan2 θ = 0

so that 2 K± = 1 + tan2 θ =

cos2 θ sin2 θ 1 + = 2 2 cos θ cos θ cos2 θ

CH004.tex

122

10/3/2011

21: 55

Page 122

SIMPLE MODELS USEFUL FOR QUANTUM OSCILLATOR PHYSICS

and thus 1 cos θ Hence, the two Hamiltonian eigenvalues (4.68) read     α1 + α 2 α 1 − α2 E± = ± 2 2 cos θ K± = ±

(4.71)

their difference being E + − E− =

α1 − α2 cos θ

so that, by inversion, cos θ appears to be cos θ =

α1 − α2 E+ − E −

(4.72)

Now, by insertion of Eq. (4.71) into Eq. (4.70), one gets for the situation corresponding to the E+ eigenvalue   1 1− C1+ + tan θ C2+ = 0 cos θ which reads ( cos θ − 1) C1+ + sin θ C2+ = 0

(4.73)

Moreover, keeping in mind the trigonometric relation 1 − cos 2θ = sin2 θ 2

(4.74)

which reads 1 − cos 2θ = 2 sin θ sin θ the term multiplying C1+ in Eq. (4.73) reads cos θ − 1 = −2 sin

    θ θ sin 2 2

(4.75)

Furthermore, the trigonometric relation sin 2θ = 2 sin θ cos θ yields sin θ = 2 sin

    θ θ cos 2 2

Hence, using Eqs. (4.75) and (4.76 ), Eq. (4.73) transforms to         θ θ θ θ −sin sin C1+ + sin cos C2+ = 0 2 2 2 2 so that C1+ cos (θ/2) = C2+ sin (θ/2)

(4.76)

CH004.tex

10/3/2011

21: 55

Page 123

123

4.2 TWO-ENERGY-LEVEL SYSTEMS

Thus, the two expansion coefficients, which are clearly normalized, read     θ θ C1+ = cos and C2+ = sin 2 2 so that Eq. (4.50) yields

    θ θ |1  + sin |2  2 2

(4.77)

    θ θ |1  + cos |2  2 2

(4.78)

|+  = cos In a similar way one would obtain

|−  = − sin

which may be verified by observing that the two normalized kets (4.77) and (4.78) are orthogonal. Then, multiplying Eq. (4.77) by sin (θ/2), and Eq. (4.78) by cos (θ/2), one obtains, after summing these results and simplification,     θ θ |2  = sin (4.79) |+  + cos |−  2 2 or, due to Eqs. (4.77) and (4.78), |2  = |+ + |2  + |− − |2  with

  θ 2 (4.80) In a similar way, after multiplying Eq. (4.77) by cos (θ/2), and Eq. (4.78) by − sin (θ/2), and adding the results, one obtains,     θ θ |1  = cos |+  − sin (4.81) |−  2 2 2 |+  = + |2  = sin

  θ 2

and

2 |−  = − |2  = cos

4.2.5 Transition probability from |1  to |2  due to the V perturbation Suppose that at an initial time the system is in the state |(0) = |1 

(4.82)

At time t, this state will transform into |(t) given, according to Eq. (3.85), by |(t) = (e−iHt/ )|(0) or, owing to the initial condition (4.82), by |(t) = (e−iHt/ )|1  and thus, according to Eq. (4.81), by     θ θ −iHt/ |(t) = cos (e (e−iHt/ )|−  )|+  − sin 2 2

CH004.tex

124

10/3/2011

21: 55

Page 124

SIMPLE MODELS USEFUL FOR QUANTUM OSCILLATOR PHYSICS

Again, owing to the Hamiltonian eigenvalue equation (4.46), this expression reads     θ θ (e−iE+ t/ )|+  − sin (e−iE− t/ )|−  |(t) = cos 2 2 Next premultiplying both terms of this last equation by the bra 2 | corresponding to the ket (4.79 ), that is,     θ θ (e−iE+ t/ )2 |+  − sin (e−iE− t/ )2 |−  2 |(t) = cos 2 2 then, owing to (4.80), it appears that     θ θ cos (e−iE+ t/ − e−iE− t/ ) 2 |(t) = sin 2 2

(4.83)

Moreover, the probability for the system to jump at time t into the state |2  being P12 (t) = |2 |(t)|2 becomes with the help of Eq. (4.83)     θ θ cos2 (2 − (e+i(E+ −E− )t/ + e−i(E+ −E− )t/ )) P12 (t) = sin2 2 2 or       θ θ (E+ − E− )t cos2 1 − cos P12 (t) = 2 sin2 2 2 

(4.84)

Furthermore, by aid of the trigonometric relations   x x 1 − cos 2x 2 and sin x = 2 sin sin x = cos 2 2 2 where x is some variable, we have     (E+ − E− )t 2 (E+ − E− )t = 2 sin 1 − cos  2 so that Eq. (4.84) may be written



P12 (t) = sin2 θ sin2

(E+ − E− ) t 2

 (4.85)

Now, since we do not know sin θ, but both tan θ and cos θ, which are, respectively, given by Eqs. (4.69) and (4.72), it is suitable to transform this last equation into   (E+ − E− )t P12 (t) = cos2 θ tan2 θ sin2 2 so that, due to Eqs. (4.69) and (4.72), the transition probability transforms to   β2 2 (E+ − E− )t P12 (t) = 4 sin (E+ − E− )2 2 Finally, owing to Eq. (4.49), we have ⎛  P12 (t) =

4β2 (α1 − α2 )2 + 4β2

sin2 ⎝

⎞ (α1 − α2 )2 + 4β2 t ⎠ 2

(4.86)

CH004.tex

10/3/2011

21: 55

Page 125

4.2 TWO-ENERGY-LEVEL SYSTEMS

125

that is, the Rabi equation. Besides, the time-dependent probability P12 (t) to jump from |1  to |2  plus that P11 (t) for the system to remain into |1  must be unity, one has P11 (t) = 1 − P12 (t) In the special situation where the two levels are degenerate, Eq. (4.86) reduces to     βt βt 2 2 P12 (t) = sin so that P11 (t) = cos when α1 = α2 (4.87)   In the following, many results of this section will be applied to Fermi resonances, a physical phenomenon that is met in situations involving anharmonic couplings between molecular oscillators.

4.2.6

Fermi golden rule

In relation with the dynamics of the double energy levels, and to end this chapter, we must now touch on the question of transition probabilities per unit time from one energy level to another one because of a coupling between them, a question that will be of importance when we later study the coupling between molecular oscillators and the electromagnetic field. Thus, consider a system described by a Hamiltonian H that may be split into two noncommuting parts H◦ and V according to H = H◦ +V

with

[H◦ , V]  = 0

the eigenvalue equation of H◦ being H◦ |k  = Ek |k 

(4.88)

k |l  = δkl

(4.89)

with

We seek the transition probability at time t for the system described by H to pass from any eigenstate of H◦ to another because of the presence of V, that is, |C(l, t|k, 0)|2 = |l (t)|k (0)|2 = k (0)|l (t)l (t)|k (0)

(4.90)

Owing to the time-dependent evolution equation, the ket |l (t) evolves with time according to |l (t) = U(t)|l (0) Now, in the interaction picture, the time evolution operator is given, in terms of the Hamiltonian H◦ by Eq. (3.122), that is, U(t) = (e−iH

◦ t/

)U(t)IP

Thus, the transition probability (4.90) becomes |C(l, t|k, 0)|2 = k (0)|(e−iH

◦ t/

)U(t)IP |l (0)l (0)|U(t)IP−1 (eiH

◦ t/

)|l (0)

Again, owing to the eigenvalue equation (4.88), the transition probability transforms to |C(l, t | k, 0)|2 = k (0)|(e−iEk t/ )U(t)IP |l (0)l (0)|U(t)IP−1 (eiEk t/ )|k (0)

CH004.tex

126

10/3/2011

21: 55

Page 126

SIMPLE MODELS USEFUL FOR QUANTUM OSCILLATOR PHYSICS

Or, after simplification |C(l, t|k, 0)|2 = |k (0)|U(t)IP |l (0)|2

(4.91)

Up to first order, the IP time evolution operator is, according to Eq. (3.125), given by   t 1 V(t )IP dt U(t)IP = 1 + i 0 where V(t)IP = (eiH

◦ t/

)V(e−iH

◦ t/

)

(4.92)

Thus, Eq. (4.91) becomes



2    t 



1 V(t )IP dt |l (0)

|C(l, t|k, 0)|2 =

k (0)| 1 + i 0

Next, using Eq. (4.92) and simplifying by the orthogonality property (4.89), the transition probability takes the form

2  2  t

1

2 iH◦ t / −iH◦ t /

(4.93) k (0)|(e )V(e )|l (0)dt

|C(l, t|k, 0)| =  0 Again, the eigenvalue equation allows one to transform this result into

2  2  t

1

2 iEk t / −iEl t /

k (0)|(e )V(e )|l (0)dt

|C(l, t|k, 0)| =

 0 or

 t

2  2



1 iωkl t |C(l, t|k, 0)|2 = |k |V|l |2

(e )dt

 0

(4.94)

with (Ek − El )  where the reference to time t = 0 has been omitted. By integration, one has   iω t  t 1 e kl − 1 iωkl t (e ) dt = i ωkl 0 ωkl =

(4.95)

(4.96)

In addition, the corresponding absolute value is

2

 t



(eiωkl t ) dt = 2 (1 − cos ωkl t)



2 ωkl 0 Moreover, by aid of the usual trigonometric relations   2 ωkl t (1 − cos ωkl t) = 2 sin 2 Eq. (4.94) becomes



|C(l, t|k, 0)| = 4|k |V|l | 2

2

sin2 (ωkl t/2) (ωkl )2

 (4.97)

CH004.tex

10/3/2011

21: 55

Page 127

4.2 TWO-ENERGY-LEVEL SYSTEMS

127

This last expression holds for any time, up to first order in V. Now, consider this expression for large times for which it is convenient to write the second right-hand-side term of Eq. (4.97) in the following way:  2  2 2 sin (xt/2) (4.98) |C(l, t|k, 0)| = 4|k |V|l | x2 with x = ωkl Then, for large time t, Eq. (4.98) reads |C(l, t|k, 0)|2 = |k |V|l |2

(4.99) 

sin2 (x/ε) x2

 (4.100)

with t 1 = with ε → 0 (4.101) 2 ε Next observe that one of the expressions of the Dirac distribution function is given by Eq. (18.57) of Section 18.6.   ε sin2 (x/ε) δ(x) = when ε → 0 π x2 which, owing to Eqs. (4.99) and (4.100), reads in the present situation   2 πt sin (x/ε) = δ(x) x2 2 so that Eq. (4.100) takes the form |C(l, t|k, 0)|2 = 4|k |V|l |2 t

π δ(x) 2

or, in view of Eqs. (4.95) and (4.99), 2π (4.102) |k |V|l |2 tδ(Ek − El )  Owing to this result, it is now possible to get the first-order transition probability per unit time, which is by definition   ∂|C(l, t|k, 0)|2 W (l, t|k, 0) = ∂t |C(l, t|k, 0)|2 =

That gives what is called the Fermi golden rule: W (l, t|k, 0) =

2π |k |V|l |2 δ(Ek − El ) 

(4.103)

an equation of the form of (4.103) will be met at the end of this book, dealing with molecular spectroscopy, when studying the interaction of molecular oscillators with electromagnetic field through a potential V involving a coupling of their dipolar moments with the electric field.

CH004.tex

10/3/2011

21: 55

Page 128

128

SIMPLE MODELS USEFUL FOR QUANTUM OSCILLATOR PHYSICS

4.3

CONCLUSION

This chapter, devoted to some quantum models, has lead to the following important results that will be useful in the subsequent studies of quantum oscillators: Particle-in-a-box energy and de Broglie relation  2  h h (nx2 + ny2 + nz2 ) Enx ny nz = λ= 8ma2 p Second-order perturbation energy: ± |H|±   i |H|i  ±

i |V|j 2 i |H|i  − j |H|j 

Rabi’s relation: P12 (t) =

4|1 |H|2 |2 (1 |H|1  − 2 |H|2 )2 + 4|1 |H|2 |2 ⎞ ⎛  (1 |H|1  − 2 |H|2 )2 + 4|1 |H|2 |2 t ⎠ × sin2 ⎝ 2

Fermi’s golden rule: W (l, t|k, 0) =

2π |k |V|l |2 δ(Ek − El ) 

Among them, the result of the particle-in-a-box model showing that waves associated to quantum states obey the de Broglie wavelength law according to which the number of nodes of the stationary waves increases with energy, a property that is to be obeyed by the energy wavefunctions of quantum oscillators. The other is the quantum interference found in the study of two-energy-state systems, which is met in the study of Fermi resonances, a physical phenomenon appearing in anharmonically coupled molecular oscillators. The latter is the time-dependent amplitude probability for a system to pass from one state to another one due to some coupling with the thermal bath, a result that will be widely used when studying coupling of molecular oscillators with the infrared (IR) electromagnetic field.

BIBLIOGRAPHY C. Cohen-Tannoudji, B. Diu, and F. Laloe. Quantum Mechanics. Wiley-Interscience: Hoboken, NJ, 2006.

CH005.tex

PART

10/3/2011

22: 1

Page 129

II

SINGLE QUANTUM HARMONIC OSCILLATORS

CH005.tex

10/3/2011

CHAPTER

22: 1

Page 131

5

ENERGY REPRESENTATION FOR QUANTUM HARMONIC OSCILLATORS INTRODUCTION The present chapter develops the basis of the quantum approach to harmonic oscillators. The dimensionless creation and annihilation operators are first introduced. Using these operators, which are Hermitian self-conjugate, it is possible to solve the eigenvalue equation of the Hamiltonian and thus to get the values of the energy levels of quantum harmonic oscillators. That also permits one to obtain the corresponding orthonormalized eigenkets, thus providing a basis for the study of quantum oscillators. Moreover, in a subsequent section, the relation governing the action of the raising and lowering operators on the eigenkets of the Hamiltonian are derived, leading to the possibility of finding how the Heisenberg uncertainty relations apply to quantum harmonic oscillators, when they are in some eigenkets of their Hamiltonian. The formalism introduced allows them to verify the validity of the virial theorem. Furthermore, a place is reserved to non-Hermitian operators (Fermion operators) playing for two-level systems a role analogous to that of creating annihilation operators (Boson operators) for quantum oscillators. Another section is devoted to the wave mechanics representation of the eigenkets of the Hamiltonian, which will permit a pictorial description of these kets in terms of wavefunctions, the corresponding number of nodes increasing with the energy. Finally, the time dependence of the creation and annihilation operators is calculated in the Heisenberg picture and applied to get the time dependence of the basic operators and of their mean values averaged over the eigenkets of the oscillator Hamiltonians.

5.1

HAMILTONIAN EIGENKETS AND EIGENVALUES

The most important result dealing with quantum harmonic oscillators is the knowledge of its quantized energy levels En , initially introduced by Planck (1901) in order to explain the spectral density of a black body via En = nω

with

n = 1, 2, . . .

Quantum Oscillators, First Edition. Olivier Henri-Rousseau and Paul Blaise. © 2011 John Wiley & Sons, Inc. Published 2011 by John Wiley & Sons, Inc.

131

CH005.tex

132

10/3/2011

22: 1

Page 132

ENERGY REPRESENTATION FOR QUANTUM HARMONIC OSCILLATORS

Later, this assumed expression of the quantized energy levels was weakly modified by Heisenberg (1925) in its matrix mechanics, which showed that they were given by   En = n + 21 ω with n = 0, 1, 2, . . . From quantum mechanics, the problem is to solve the eigenvalue equation of the Hamiltonian H. One possibility is to solve the second-order partial differential equation, which is the wave mechanics picture of this eigenvalue equation, that is, the time-independent Schrödinger equation. Such an approach supposes to have some knowledge about the theory of partial derivative equations. The other possibility is to pass from the position and momentum Hermitian operators involved in the Hamiltonian to two new dimensionless Hermitian self-conjugated operators, the ladder operators, which allow an easy resolution of the eigenvalue equation of the Hamiltonian. It is the latter approach that is chosen in the present section.

5.1.1

Hamiltonian in terms of ladder operators

5.1.1.1 Ladder operators Starting from the Hamiltonian H of a quantum harmonic oscillator of angular frequency ω and of reduced mass m coupling two masses, m1 and m2 , that is   2 1 P (5.1) + mω2 Q2 H= 2m 2 where Q is the position operator and P its conjugate momentum obeying the commutation rule [Q, P] = i where the reduced mass m of the oscillator is given by   m1 m2 m= m1 + m 2

(5.2)

In order to solve the eigenvalue equation of this Hamiltonian, it is convenient to work with the following dimensionless non-Hermitian operators, which are mutually Hermitian conjugates (the ladder operators):   mω 1 Q+i P (5.3) a= 2 2mω  a = †

 mω 1 Q−i P 2 2mω

(5.4)

Next, we calculate the commutator of these two conjugate Hermitian operators. From Eqs. (5.3) and (5.4), it reads aa† = (ηQ + iζP)(ηQ − iζP) = η2 Q2 + ζ 2 P2 + iζη[P, Q] a† a = (ηQ − iζP)(ηQ + iζP) = η2 Q2 + ζ 2 P2 − iζη[P, Q]

CH005.tex

10/3/2011

22: 1

Page 133

5.1

with



HAMILTONIAN EIGENKETS AND EIGENVALUES



mω 2

ζ=

1 2mω

aa† − a† a = 2iζη[P, Q] =

i [P, Q] 

η=

and

133

Hence, the commutator of a and a† reads

or, due to the basic commutator (2.3), aa† − a† a = [a, a† ] = 1

(5.5)

Next, by inversion of Eqs. (5.3) and (5.4), one obtains the dependence of Q and P operators with respect to a and a† , respectively,   Q= (5.6) (a† + a) 2mω  P=i

mω † (a − a) 2

(5.7)

For reasons that will be clear later, the ladder operators a† and a are, respectively, often named the raising and lowering operators or creation and annihilation operators. Then, the insertion of Eqs. (5.6) and (5.7) into Eq. (5.1) gives H=

i2 mω † 1  (a − a)2 + mω2 (a† + a)2 2m 2 2 2mω

or H=−

ω † ω † (a − a)2 + (a + a)2 4 4

Hence, ω † 2 ω † 2 ((a ) + (a)2 − a† a − aa† ) + ((a ) + (a)2 + a† a + aa† ) 4 4 and, after simplification H=−

ω † (a a + aa† ) 2 Now, the commutator (5.5) may be written H=

(5.8)

aa† = a† a + 1 so that Eq. (5.8) leads to the following fundamental expression for the Hamiltonian of the quantum harmonic oscillator:   H = ω a† a + 21 (5.9) Observe that this Hamiltonian is Hermitian, as required, since †

(a† a) = a† a

CH005.tex

134

10/3/2011

22: 1

Page 134

ENERGY REPRESENTATION FOR QUANTUM HARMONIC OSCILLATORS

Now write the Hamiltonian eigenvalue equation to be solved: H|{n} = En |{n}

(5.10)

where |{n}1 are the eigenvectors and E

n the corresponding eigenvalues, which are real because H is Hermitian. Besides, since H is Hermitian, its eigenvectors are orthogonal and, if normalized, satisfy

{n}|{m} = δnm

5.1.2

(5.11)

Resolution of the Hamiltonian eigenvalue

To solve the eigenvalue equation (5.10), define the following Hermitian operator N, which commutes with the Hamiltonian H, that is, N = a† a

with

[N, H] = 0

and

N† = N

(5.12)

5.1.2.1 Commutators [N, a], [N, a† ], and eigenvalue equation of N For this purpose, it is necessary to know the commutator [N, a] of N with the annihilation operator a: [N, a] = (a† a)a − a(a† a) Now, by changing the position of the second parenthesis, which does not modify anything, the commutator reads [N, a] = (a† a)a − (aa† )a

(5.13)

In addition, according to Eq. (5.5), that is, aa† = a† a + 1

(5.14)

Eq. (5.13) becomes [N, a] = {a† a − (a† a + 1)}a or [N, a] = −a = [a† a, a]

(5.15)

Now, calculate the commutator of N with the creation operator a† . We have [N, a† ] = (a† a)a† − a† (a† a) which, by changing the first parenthesis position, reads [N, a† ] = a† (aa† ) − a† (a† a) or, due to Eq. (5.14), [N, a† ] = a† (a† a + 1) − a† (a† a) We shall use for the writing of the eigenkets of a† a notations such as |{n}, |(n), and |[n], which are more complex than the usual ones |n, in order to allow one to distinguish easily different eigenkets belonging to different oscillators characterized by different sets of ladder operators a† a, b† b, and c† c. That will appear to be useful in the following chapters.

1

CH005.tex

10/3/2011

22: 1

Page 135

5.1

HAMILTONIAN EIGENKETS AND EIGENVALUES

135

so that [N, a† ] = a† = [a† a, a† ]

(5.16)

Next, write the eigenvalue equation of N: N|{n} = An |{n}

(5.17)

where An are the eigenvalues of N, which are real because N is Hermitian, whereas |{n} are the corresponding eigenvectors obeying Eq. (5.11), which must be, therefore, the same as those appearing in Eq. (5.10) because H and N commute. 5.1.2.2 Action of N on |ak {n} To solve the eigenvalue equation (5.17) consider the action of the commutator (5.15) on any eigenket of Eq. (5.17), that is, [N, a]|{n} = (Na − aN)|{n}

(5.18)

which, due to Eq. (5.15), reads (Na − aN)|{n} = −a|{n} and which, owing to Eq. (5.17), transforms to (Na − aAn )|{n} = −a|{n} Then, rearranging, it yields Na|{n} = aAn |{n} − a|{n} Since An is a scalar that commutes with a, we have Na|{n} = (An − 1)a|{n}

(5.19)

Now, observe that the action of a on the eigenstate |{n} yields a new state, which may be written formally as a|{n} ≡ |a{n}

(5.20)

N|a{n} = (An − 1)|a{n}

(5.21)

so that Eq. (5.19) reads

Hence, (An −1) is the eigenvalue of N corresponding to the ket resulting from the action of a on |{n}. Again, consider the action of the commutator (5.15); however, let it now act on the ket defined by Eq. (5.20), that is, [N, a]|a{n} = (Na − aN)|a{n} Then, proceeding in the same way as for passing from Eq. (5.18) to (5.19), one finds Na|a{n} = (An − 2)a|a{n} Moreover, writing a|a{n} = |a2 {n}

(5.22)

CH005.tex

136

10/3/2011

22: 1

Page 136

ENERGY REPRESENTATION FOR QUANTUM HARMONIC OSCILLATORS

Eq. (5.22) takes the form N|a2 {n} = (An − 2)|a2 {n}

(5.23)

Hence, from Eqs. (5.21) and (5.23), one obtains by recurrence N|ak {n} = (An − k)|ak {n}

with

|ak {n} ≡ ak |{n}

(5.24)

5.1.2.3 Action of N on |(a† )k {n} Now, consider the action of the commutator (5.16) on any eigenket of Eq. (5.17), that is, [N, a† ]|{n} = (Na† − a† N)|{n} Then, using Eq. (5.16) to express the left-hand-side member of this last expression, one obtains a† |{n} = (Na† − a† N)|{n} Again, using the eigenvalue equation (5.17), one gets †

a† |{n} = (Na† − a An )|{n} and thus, after commuting the scalar An with the operator a† , we have Na† |{n} = (An + 1)a† |{n} or N|a† {n} = (An + 1)|a† {n}

(5.25)

with a† |{n} ≡ |a† {n} Consider again the action of the commutator (5.16) on |a† {n}: [N, a† ]|a† {n} = (Na† − a† N)|a† {n} Then, proceeding as above, one would obtain Na† |a† {n} = (An + 2)a† |a† {n} or, changing the notation, N|(a† )2 {n} = (An + 2)|(a† )2 {n}

(5.26)

a† |a† {n} ≡ |(a† )2 {n}

(5.27)

with

Hence, from Eqs. (5.25) and (5.26), one gets by recurrence N|(a† )k {n} = (An + k)|(a† )k {n} with

|(a† )k {n} = (a† )k |{n}

(5.28)

CH005.tex

10/3/2011

22: 1

Page 137

5.1

HAMILTONIAN EIGENKETS AND EIGENVALUES

137

5.1.2.4 Discrete character of the eigenvalues An Starting from the assumed eigenvalue equation (5.17), it has been possible to prove Eqs. (5.24) and (5.28). Rewrite them for comparison: N|{n} = An |{n} N|(a)k {n} = (An − k)|(a)k {n} N|(a† )k {n} = (An + k)|(a† )k {n} with |(a)k {n} ≡ (a)k |{n}

and

|(a† )k {n} ≡ (a† )k |{n}

(5.29)

By inspection of these equations, it appears that |{n} is an eigenvector of N with the corresponding eigenvalue An . |(a)k {n} is an eigenvector of N with the corresponding eigenvalue (An − k). |(a† )k {n} is an eigenvector of N with the corresponding eigenvalue (An + k). Hence, |{n}, (a)k |{n}, and (a† )k |{n} are eigenvectors of N with the eigenvalues An (An − k) and (An + k), respectively. Thus, it may be inferred that the action of the kth power of the a operator on an eigenvector of N lowers by k the eigenvalue An of N corresponding to this eigenvector, whereas the action of the kth power of a† on the same eigenvector of N raises by k the eigenvalue An . Hence, the eigenvalues of N obey the relation An , An ± 1, An ± 2, . . . 5.1.2.5 Impossibility for An to be negative negative. Thus, consider

(5.30)

Now let us show that An cannot be

|a{n} ≡ a|{n}

(5.31)

the Hermitian conjugate of which is {n}a† | ≡ {n}|a†

(5.32)

Then, owing to the property of the norm, requiring {n}a† |a{n} ≥ 0 and according to the notations (5.31) and (5.32), we have {n}a† |a{n} = {n}|a† a|{n} Moreover, due to the definition (5.12) of N, Eq. (5.33) transforms to {n}a† |a{n} = {n}|N|{n}

(5.33)

CH005.tex

138

10/3/2011

22: 1

Page 138

ENERGY REPRESENTATION FOR QUANTUM HARMONIC OSCILLATORS

which, with the help of the eigenvalue equation (5.17), transforms to {n}a† |a{n} = {n}|An |{n} so that An being a scalar, is given by An =

{n}a† |a{n} {n}|{n}

Now, observe that because the norm cannot be negative one has, respectively, {n}a† |a{n} ≥ 0

and

{n}|{n} ≥ 0

Hence, the eigenvalues of N cannot be negative An ≥ 0

(5.34)

5.1.2.6 Nullity of the lowest eigenvalue Since the eigenvalue An cannot be negative, there exists a lowest eigenvalue A0 to which is associated an eigenvector denoted |{0}, leading to write in this special situation the eigenvalue equation (5.17) according to N|{0} = A0 |{0} Now, the action of N on the ket resulting from the action of a on the lowest state |{0} would lead, according to Eq. (5.19), to a new state, eigenvector of N with a corresponding eigenvalue (A0 − 1), which is impossible since A0 was assumed to be the lowest possible eigenvalue: Na|{0} = N|a{0} = (A0 − 1)|a{0}

Impossible

Thereby, owing to this impossibility, |{0} must be the fundamental eigenstate of N, leading to write a|{0} = |a{0} = 0

(5.35)

the Hermitian conjugate of which is {0}|a† = {0}a† | = 0 Of course, the norm between the states involved in the two above equations is {0}a† |a{0} = 0

(5.36)

Next, observe that, due to the notations (5.31) and (5.32), {0}a† |a{0} ≡ {0}|a† a|{0}

(5.37)

and, due to Eq. (5.12), that {0}|a† a|{0} = {0}|N|{0} and, at last, that, owing to Eq. (5.17), {0}|N|{0} = {0}|A0 |{0} = A0 {0}|{0}

(5.38)

CH005.tex

10/3/2011

22: 1

Page 139

5.1

HAMILTONIAN EIGENKETS AND EIGENVALUES

139

Again, if |{0} is normalized, that is, if {0}|{0} = 1 then, in view of Eqs. (5.37) and (5.38), it reads {0}a† |a{0} = A0 so that, owing to Eq. (5.36), we have A0 = 0

(5.39)

5.1.2.7 Solution of the Hamiltonian eigenvalue equation (5.17) section, we studied the eigenvalue equation (5.17), that is,

In the above

N|{n} = An |{n} for which it was shown that the eigenvalues An obey Eqs. (5.30), (5.34), and (5.39), that is, An , An ± 1, An ± 2, . . .

with

An ≥ 0

and

A0 = 0

These results show that An is of the form An = 0, 1, 2, 3, . . . Hence, writing explicitly the operator N by aid of Eq. (5.12) leads to writing the following eigenvalue equation: (a† a)|{n} = n|{n}

with n ≡ An

and

n = 0, 1, 2, 3, . . .

(5.40)

Since the Hamiltonian of the quantum harmonic oscillator is given by Eq. (5.9), that is,   H = a† a + 21 ω (5.41) and due to Eq. (5.40), we see that the following eigenvalue equation is satisfied:   H|{n} = ω n + 21 |{n} with n = 0, 1, 2, 3, . . . (5.42) The lowest eigenstate |{0} of the Hamiltonian corresponding to n = 0 is called the ground state, whereas the corresponding residual energy ω/2 is called the zeropoint energy of the oscillator. Now, according to Section 1.3.1, since the Hamiltonian operator (5.9) is Hermitian, its eigenvectors are necessarily orthogonal. Thus, if they have been normalized, they form an orthonormal basis obeying  {n}|{m} = δnm and |{n}{n}| = 1 (5.43) n

5.1.2.8 Zero-point energy as preserving the Heisenberg uncertainty relations It may be of interest to understand the role of the zero-point energy ω/2 appearing in Eq. (5.42) in the context of the Heisenberg uncertainty relations (2.9) dealing with the momentum and the position operators: P Q 

 2

(5.44)

CH005.tex

140

10/3/2011

22: 1

Page 140

ENERGY REPRESENTATION FOR QUANTUM HARMONIC OSCILLATORS

Now, suppose that the energy of the ground state |{0} is zero. Then, since the harmonic potential energy V of the oscillator cannot be negative because quadratic in the position Q, that is, V = 21 Mω2 Q2

(5.45)

and because the kinetic energy T , which is quadratic in the momentum P, is necessarily positive, that is, T=

P2 >0 2M

(5.46)

our supposition would imply that both the kinetic T and potential V energies ought separately to be zero, that is, T =V =0

(5.47)

However, as a consequence of Eqs. (5.45)–(5.47), it would then appear that P=Q=0

(5.48)

Moreover, if Eq. (5.48) was true, that would in turn imply that P and Q would be known without any uncertainty, that is, P = Q = 0 in contradiction to the Heisenberg uncertainty relations (5.44).

5.1.3

Action of ladder operators on Hamiltonian eigenkets

The solution of the eigenvalue equation of the Hermitian Hamiltonian has not only the merit that it yields energy levels of the oscillators but also the merit that it provides a basis from which it is possible to obtain matrix representations of all operators dealing with quantum oscillators. Since these operators may be written as functions of the position and momentum operators, they may be also expressed as functions of the raising and lowering operators. Therefore, it appears that the knowledge of the action of these operators on the eigenkets of the Hamiltonian will be of much interest from now on. Thus, the aim of this new section will be to treat this point. 5.1.3.1 Action of a Consider the action of a operator on |{n}. Keeping in mind Eq. (5.40) according to which n ≡ An , Eq. (5.21) reads N|a{n} = (n − 1)|a{n}

with

n = 0, 1, 2, 3, . . .

(5.49)

whereas the eigenvalue equation Eq. (5.17) allows one to write N|{n} = n|{n} N|{n − 1} = (n − 1)|{n − 1}

(5.50)

Comparison of the eigenvalue equations (5.49) and (5.50) shows that both equations involve the same operator and the same eigenvalues. Moreover, if the eigenvectors

CH005.tex

10/3/2011

22: 1

Page 141

5.1

HAMILTONIAN EIGENKETS AND EIGENVALUES

141

appearing in these eigenvalue equations are not necessarily the same, they must be proportional, so that one may write |a{n} = λn |{n − 1} where λn is a complex scalar. The Hermitian conjugate of this last expression being {n}a† | = λ∗n {n − 1}| the corresponding norm is thereby {n}a† |a{n} = |λn |2 {n − 1}|{n − 1} Next, if the right-hand-side ket is normalized, this last equation reduces to {n}a† |a{n} = |λn |2

(5.51)

On the other hand, Eqs. (5.31) and (5.32) allow one to write the left-hand side of Eq. (5.51) as {n}a† |a{n} = {n}|a† a|{n} It appears that, due to Eq. (5.40), {n}a† |a{n} = n{n}|{n} = n

(5.52)

Therefore, by identification of Eqs. (5.51) and (5.52), we have |λn |2 = n so that, ignoring the phase factor (if λn would be imaginary), which is of no interest, √ λn = n Thus, one obtains the final result of interest: a|{n} =

√

n|{n − 1}

(5.53)

As it appears, the action of operator a on any eigenstate |{n} of a† a corresponding to the eigenvalue n transforms this state into a new eigenstate |{n − 1} of a† a corresponding to the eigenvalue (n − 1). This action may be, therefore, viewed as lowering the eigenvalue of a† a by unity and thus the corresponding eigenvector. Hence, a is called a lowering operator. Observe that the Hermitian conjugate of this equation is √ (5.54) {n}|a† = n{n − 1}| Now, since |{0} is the lowest eigenket of a† a, Eqs. (5.53) and (5.54) lead to a|{0} = {0}|a† = 0

(5.55)

CH005.tex

142

10/3/2011

22: 1

Page 142

ENERGY REPRESENTATION FOR QUANTUM HARMONIC OSCILLATORS

5.1.3.2 Action of a† After finding the action of a on the Hamiltonian eigenkets, then pass to that on its Hermitian conjugate a† . In view of An = n, Eq. (5.25) reads N|a† {n} = (n + 1)|a† {n}

with

n = 0, 1, 2, 3, . . .

(5.56)

whereas the eigenvalue equation (5.17) reads, respectively, N|{n} = n|{n} N|{n + 1} = (n + 1)|{n + 1}

(5.57)

Equation (5.57) is analogous to Eq. (5.56) since the N operator and the eigenvalues (n + 1) are the same in both cases so that the kets appearing in Eqs. (5.56) and (5.57) must be at least proportional to each other, leading one to write |a† {n} = μn |{n + 1}

(5.58)

where μn is a complex scalar. The corresponding norm is, therefore, {n}a|a† {n} = |μn |2 {n + 1}|{n + 1} Again, if the eigenvectors |{n + 1} are normalized, this last expression reduces to {n}a|a† {n} = |μn |2

(5.59)

In addition, changing the writing with the aid of Eqs. (5.31) and (5.32), the lefthand-side reads {n}a|a† {n} = {n}|aa† |{n}

(5.60)

Furthermore, using the commutation rule (5.5), leading to aa† = a† a + 1 Eq. (5.60) becomes {n}a|a† {n} = {n}|(a† a + 1)|{n}

(5.61)

Moreover, using the eigenvalue equation (5.40), we have (a† a + 1)|{n} = (n + 1)|{n} Thus, Eq. (5.61) transforms to {n}a|a† {n} = (n + 1){n}|{n} or, |{n} being normalized, {n}a|a† {n} = n + 1 Finally, by identification of Eqs. (5.59) and (5.62) |μn |2 = n + 1 and ignoring an irrelevant phase factor, we have √ μn = n + 1

(5.62)

CH005.tex

10/3/2011

22: 1

Page 143

5.1

HAMILTONIAN EIGENKETS AND EIGENVALUES

143

Hence, from Eq. (5.58), and due to the notations in (5.31) and (5.32), one obtains the final result: √ a† |{n} = n + 1|{n + 1} (5.63) the Hermitian conjugate of which is {n}|a =

√

n + 1{n + 1}|

(5.64)

Observe that, according to Eq. (5.63), the action of a† on any ket |{n} is changed into the raised one |{n + 1}; hence, this operator is called a raising operator. 5.1.3.3 Action of different powers of a† and a Consider the action of different powers of a† . The action of a† on the lowest state |{0} corresponding to n = 0, Eq. (5.63) yields a† |{0} = |{1}

(5.65)

In addition, due to Eq. (5.63), the second power of a† (a† )2 |{0} = a† a† |{0} = a† |{1} or, using again Eq. (5.63), (a† )2 |{0} =

(5.66)

 1(1 + 1)|{1 + 1}

Moreover, the third power of a† yields, using Eq. (5.63),  (a† )3 |{0} = 1(1 + 1)(2 + 1)|{2 + 1} so that, by recurrence, one obtains (a† )n |{0} = the Hermitian conjugate of which is {0}|(a)n =

√ n!|{n}

(5.67)

√ n!{n}|

(5.68)

Furthermore, by inversion, Eqs. (5.67) and (5.68) read, respectively, (a† )n |{n} = √ |{0} n!

(5.69)

(a)n {n}| = {0}| √ n!

(5.70)

Next, passing to the action of different powers of a on |{n}, Eq. (5.53), allows one to write successively √ (a)|{n} = n|{n − 1} (a)2 |{n} = (a)2 |{n} =

√ n(a)|{n − 1}

 n(n − 1)|{n − 2}

CH005.tex

144

10/3/2011

22: 1

Page 144

ENERGY REPRESENTATION FOR QUANTUM HARMONIC OSCILLATORS

so that, by recurrence, one gets  (a)l |{n} = n(n − 1) · · · (n − l + 1)|{n − l} or

(5.71)

√

(a)l |{n}

=√

n! |{n − l} (n − l)!

the Hermitian conjugate of which is

(5.72)

√

{n}|(a ) = √ † l

n! {n − l}| (n − l)!

(5.73)

In a similar way, one would obtain using Eq. (5.63)  (a† )l |{n} = n(n + 1) · · · (n + l)|{n + l} or (a† )l |{n}

√ (n + l)! = √ |{n + l} (n)!

for which the Hermitian conjugate is {n}|(a)l =

5.1.4

(5.74)

√ (n + l)! {n + l}| √ (n)!

Matrix representation of ladder operators

Knowledge of the action of the ladder operators on the eigenkets and eigenbras of the Hamiltonian allows one to get the matrix representation of these operators. For this purpose, start from the eigenvalue equation N|(n) = n|(n) with

n = 0, 1, 2, . . .

(5.75)

keeping in mind that N is the Hermitian number occupation operator N = a† a

and

N = N†

since

(a† a)† = a† a

(5.76)

whereas a and a† are obeying the commutation rules [a, a† ]− ≡ aa† − a† a = 1

(5.77)

[a, a]− = [a† , a† ]− = 0

(5.78)

and that the kets form an orthogonal basis so that (n)|(m) = δmn

(5.79)

Note that the subscript − has been introduced in the expressions for commutators (5.77) and (5.78) in order to distinguish them from the anticommutators, which will appear later. At last, recall Eqs. (5.53) and (5.63), that is, √ √ a|(m) = m|(m − 1) and a† |(m) = m + 1|(m + 1) (5.80)

CH005.tex

10/3/2011

22: 1

Page 145

5.1

HAMILTONIAN EIGENKETS AND EIGENVALUES

145

Thus, in the basis defined by Eq. (5.79) and using Eq. (5.80), the matrix elements of a and a† read, respectively, √ √ (n)|a|(m) = m(n)|(m − 1) = mδn,m−1 (n)|a† |(m) =

√

m + 1(n)|(m + 1) =

√

m + 1δn,m+1

As a consequence, the matrix representations of a and a† read, after arbitrary truncation, ⎛ √ ⎛ ⎞ ⎞ 0 1√ √0 ⎜ 0 ⎜ 1 0 ⎟ ⎟ 2 √ ⎜ ⎜ ⎟ ⎟ √ † ⎜ ⎜ ⎟ ⎟ (5.81) 2 √0 a =⎜ a =⎜ and 0 3√ ⎟ ⎟ ⎝ ⎝ ⎠ 3 √0 ⎠ 0 4 0 4 0 Hence, the matrix representation of the occupation number defined by Eq. (5.76) yields ⎛ ⎞⎛ √ ⎞ 0 1 √ √0 ⎜ 1 0 ⎟⎜ 0 ⎟ 2 √ ⎜ ⎟⎜ ⎟ √ ⎜ ⎟ ⎜ 2 √0 N =⎜ 3 √ ⎟ 0 ⎟⎜ ⎟ ⎝ 3 √0 ⎠ ⎝ 0 4⎠ 4 0 0 or, after performing the matrix product, ⎛

⎞

0

⎜ 1 ⎜ 2 N =⎜ ⎜ ⎝ 3

⎟ ⎟ ⎟ ⎟ ⎠ 4

That is in agreement with the result obtained by premultiplying Eq. (5.75) by the bra (m)| to give (m)|N|(n) = n(m)|(n) = n δmn

5.1.5

Heisenberg uncertainty relations

As we have said above, Heisenberg provided the first demonstration of the quantized energy levels of harmonic oscillators and was lead to these results through his anticipation of the uncertainty relations. It is, therefore, important to answer the question of the expression of the Heisenberg uncertainty relation when computed over the eigenstates of the quantum harmonic oscillator Hamiltonian. For this purpose, first consider the required average values of Q and Q2 . Thus, start from the average value of Q over the number occupation eigenstates, which, owing to Eq. (5.6), is   {n}|Q|{n} = {n}|(a† + a)|{n} (5.82) 2mω

CH005.tex

146

10/3/2011

22: 1

Page 146

ENERGY REPRESENTATION FOR QUANTUM HARMONIC OSCILLATORS

Again, owing to Eq. (5.53), the average value of a appearing in Eq. (5.82) is zero because of the orthogonality of the eigenkets of the Hamiltonian: √ (5.83) {n}|a|{n} = n{n}|{n − 1} = 0 Of course, the Hermitian conjugate must be also zero: √ {n}|a† |{n} = n{n − 1}|{n} = 0

(5.84)

Therefore, it appears that the average value (5.82) of Q is zero, that is, {n}|Q|{n} = 0

(5.85)

Now, owing to Eq. (5.6), the average value of Q2 reads {n}|Q2 |{n} =

 {n}|(a† + a)2 |{n} 2mω

(5.86)

Moreover, the square appearing on the right-hand-side yields (a† + a)2 = a† a† + aa + a† a + aa† or, due to the commutation rule (5.77), that gives (a† + a)2 = (a† )2 + (a)2 + 2a† a + 1

(5.87)

Furthermore, owing to Eq. (5.53), the two successive actions of a on an eigenstate of the Hamiltonian, lead to √ √ √ aa|{n} = na|{n − 1} = n n − 1|{n − 2} The average value of aa is, therefore, zero, according to the orthogonality of the eigenkets of the Hamiltonian, that is,  (5.88) {n}|(aa)|{n} = n(n − 1){n}|{n − 2} = 0 Of course, the Hermitian conjugate of this last equation may be obtained by taking for the left-hand-side a† a† in place of aa and permuting, for the right-hand side, the ket and the bra of the scalar product, without changing the real scalar. That is,  (5.89) {n}|(a† a† )|{n} = n(n − 1){n − 2}|{n} = 0 which is also zero. Finally, in view of Eq. (5.40), the required average value of a† a reads {n}|(a† a)|{n} = n{n}|{n} = n

(5.90)

Thus, using Eqs. (5.86)–(5.90), one obtains {n}|Q2 |{n} =

 (2n + 1) 2mω

(5.91)

CH005.tex

10/3/2011

22: 1

Page 147

5.1

147

HAMILTONIAN EIGENKETS AND EIGENVALUES

so that, in view of Eqs. (5.85) and (5.91), the dispersion over Q appears to be     1  2 2 Q|{n} = {n}|Q |{n} − {n}|Q|{n} = n+ (5.92) mω 2 Now, consider the corresponding average value of the momentum, which, according to Eq. (9.35), is  mω {n}|P|{n} = i {n}|(a† − a)|{n} 2 Owing to Eqs. (5.19) and (5.84), it appears to be zero, that is, {n}|P|{n} = 0

(5.93)

Again, owing to Eq. (9.35), the average value of the square of the momentum is mω {n}|(a† − a)2 |{n} 2 where, according to the commutation rule, the right-hand-side operator reads {n}|P2 |{n} = −

(a† − a)2 = a† a† + aa − (2a† a + 1) Hence, by combination of Eqs. (5.88)–(5.90), one gets {n}|P2 |{n} =

mω (2n + 1) 2

Thereby, in view of Eqs. (5.93) and (5.94), the dispersion over P becomes   √  2 2 P|{n} = {n}|P |{n} − {n}|P|{n} = mω n + 21

(5.94)

(5.95)

by combining Eqs. (5.92) and (5.95), one obtains the following expression for the Heisenberg relation as applied to the eigenstates of the Hamiltonian of the harmonic oscillator:   Q|{n} P|{n} = n + 21  (5.96) which is an agreement with the Heisenberg uncertainty relation, which states that Q|{n} P|{n} ≥ 

(5.97)

5.1.6 Virial theorem We have seen that the knowledge of the average values of P2 and Q2 over the eigenstates of the number occupation operator allowed us to find the uncertainty Heisenberg relation (5.96), which holds for these states. These same average values may also allow us to verify the virial theorem studied in Section 2.4.4. This is the purpose of the present section. When applied to harmonic oscillators, the virial theorem leads to Eqs. (2.88) and (2.89) from which results the following relation between the average values of the

CH005.tex

148

10/3/2011

22: 1

Page 148

ENERGY REPRESENTATION FOR QUANTUM HARMONIC OSCILLATORS

Hamiltonian and those of the kinetic T and potential V operators, what may be the stationary state over which the averages are performed: T| = V| = 21 H|

(5.98)

To verify if our above results of the present chapter are in agreement with this theorem, first consider the average value of the harmonic potential over the states |{n}, which, being the eigenstates of the harmonic oscillator Hamiltonian, are therefore stationary: V|{n} = 21 mω2 {n}|Q2 |{n} Owing to Eq. (5.91), this takes the form

  V|{n} = 21 ω n + 21

(5.99)

On the other hand, the corresponding average value of the kinetic operator is T|{n} = or, in view of Eq. (5.94),

2 1 2m {n}|P |{n}

  T|{n} = 21 ω n + 21

(5.100)

Thus, according to Eq. (5.9), the average value of the harmonic Hamiltonian is   H|{n} = ω{n}| a† a + 21 |{n} or, due to Eq. (5.90),

  H|{n} = ω n + 21

(5.101)

Hence, as it may be observed, Eqs. (5.99)–(5.101) obey the virial theorem (5.98).

5.1.7

3D Harmonic oscillators

The previous sections dealt with 1D harmonic oscillators. The generalization of 1D results to 3D harmonic oscillators is the aim of the present section. The kinetic operator T of a 3D oscillator of reduced mass m is   Px2 + Py2 + Pz2 T= 2m where the Px , Py , and Pz are, respectively, the x, y, and z Cartesian components of the momentum operator. On the other hand, the potential operator is V = 21 m (ωx2 Q2x + ωy2 Q2y + ωz2 Q2z ) where the Qx , Qy , and Qz are, respectively, the x, y, and z Cartesian components of the position operator, obeying the commutation rules [Qk , Pl ] = i δkl

(5.102)

where k and l run for x, y, and z, whereas the ωk are the corresponding angular frequencies. Then, the Hamiltonian of the oscillator yields   Px2 + Py2 + Pz2 1 (5.103) H= + m (ωx2 Q2x + ωy2 Q2y + ωz2 Q2z ) 2m 2

CH005.tex

10/3/2011

22: 1

Page 149

5.1

HAMILTONIAN EIGENKETS AND EIGENVALUES

149

In a similar way as in Eqs. (5.6) and (5.7), one may express the position and momentum operators in terms of dimensionless non-Hermitian operators according to   Qk = (a† + ak ) (5.104) 2mω k  mω † Pk = i (ak − ak ) (5.105) 2 with the following commutation rule between ak and al† resulting from Eq. (5.102): [ak , al† ] = δkl

(5.106)

Again, proceeding as at the beginning of this chapter, the Hamiltonian (5.103) takes the form H = Hx + Hy + Hz with

  Hk = ωk ak† ak + 21

with

(5.107) k = x, y, z

Then, since each term Hk of the Hamiltonian H has the same structure as that of Eq. (5.41) of the Hamiltonian of 1D harmonic oscillators, one may write for each Hamiltonian Hk an eigenvalue equation having the same structure as that of (5.42), that is,   ωk ak† ak + 21 |{n}k  = Enk |{n}k  (5.108) with, for k = x, y, and z,

  Enk = ωk nk + 21

and

nk = 0, 1, 2, 3, . . .

In Eq. (5.108), the |{n}k  are the eigenkets of the Hk Hamiltonians, whereas the Enk are the corresponding eigenvalues. Of course, since the Hamiltonians Hk are Hermitian, their eigenkets are orthonormal: {n}k |{m}k  = δnk mk and, thereby, form a basis allowing us to write for each dimension the closure relation, that is,  |{n}k {n}k | = 1 nk

Now, as for the particle-in-a-box model, the full eigenkets of the 3D Hamiltonian (5.107) must be written as the products of the eigenkets of the 1D Hamiltonians Hk , that is |nx ny nz  = |{n}x |{n}y |{n}z 

(5.109)

whereas the corresponding eigenvalue of the 3D Hamiltonian must be the sum of the corresponding eigenvalues Enk , that is,       (5.110) Enx ny nz = ωx nx + 21 + ωy ny + 21 + ωz nz + 21

CH005.tex

150

10/3/2011

22: 1

Page 150

ENERGY REPRESENTATION FOR QUANTUM HARMONIC OSCILLATORS

Of course, the eigenkets (5.109) and the eigenvalues (5.110) are related through the eigenvalue equation H|{n}x |{n}y |{n}z         = ωx nx + 21 + ωy ny + 21 + ωz nz + 21 |{n}x |{n}y |{n}z  Moreover, when the 3D harmonic oscillator is isotropic, the eigenvalue (5.110) reduces to   Enx ny nz = ω (nx + ny + nz ) + 23 (5.111) Hence, as for the particle-in-a-box model, it appears that degeneracy occurs for all situations having the same value of (nx + ny + nz ) verifying Eq. (5.111). Owing to the equivalence between the three Hamiltonians Hx , Hy , and Hz and the Hamiltonian of the 1D harmonic oscillator, all that has been proved for 1D oscillators may be transposed to the 3D ones. Using equations similar to Eqs. (5.53) and (5.63), namely, √ ak |{n}k  = nk |{nk − 1} ak† |{n}k  =

 nk + 1|{nk + 1}

and by the aid of Eqs. (5.106) and (5.108), it is possible to reproduce for each component of the 3D oscillator the results obtained in the 1D situation, particularly those concerning the Heisenberg uncertainty relations and the virial theorem.

5.2 WAVEFUNCTIONS CORRESPONDING TO HAMILTONIAN EIGENKETS Although the kets and the corresponding wavefunctions are without direct physical meaning, it may be of interest, for the purpose of physical intuitive investigation, to visualize the forms of the wavefunctions corresponding to the eigenvectors of the Hamiltonian of quantum harmonic oscillators. One of the reasons, which will appear later, is that the number of nodes of these vibrational wavefunctions increases with the corresponding energy in a way that is deeply linked to the de Broglie wavelength rule according to which the kinetic energy increases with the number of nodes of the associated wavelength.

5.2.1

Second-order partial differential equation to be solved

In order to get the expression of the wavefunctions corresponding to the eigenkets of the harmonic Hamiltonian, consider this operator within wave mechanics that reads Hˆ = Tˆ + Vˆ where Tˆ and Vˆ are, respectively, the wave mechanical kinetic and potential operators, the first one being given by Eq. (3.51), that is, Tˆ = −

2 ∂ 2 2m ∂Q2

CH005.tex

10/3/2011

22: 1

Page 151

5.2 WAVEFUNCTIONS CORRESPONDING TO HAMILTONIAN EIGENKETS

151

and the last one simply by Vˆ = 21 mω2 Q2 According to Eq. (3.60), the Hamiltonian of the harmonic oscillator is, therefore, Hˆ = −

1 2 ∂ 2 + mω2 Q2 2 2m ∂Q 2

Then, the time-independent Schrödinger equation

reads

ˆ n (Q) = En n (Q) H

(5.112)

  1 2 ∂2 n (Q) + mω2 Q2 n (Q) = En n (Q) − 2 2m ∂Q 2

(5.113)

Of course, since the quantum mechanics and the wave mechanics are equivalent, the eigenvalues of the Hamiltonian appearing in Eq. (5.112) are given by Eq. (5.42), that is,   En = ω n + 21 (5.114) As a consequence, the eigenvalue equation (5.113) becomes      mω2 2 2 ∂2 n (Q) 1 − Q − ω n + n (Q) = 0 2m ∂Q2 2 2

(5.115)

That is the equation to be solved in order to get the expression of the wavefunction n (Q) given by the scalar product n (Q) = {Q}|{n} its boundary condition being n (Q) → 0

when

Q→∞

Next, perform the following variable change: ξ = ξ◦ Q

with

ξ◦ =

leading to ∂ξ = ξ◦ ∂Q

and

∂ ∂ = ∂Q ∂ξ





(5.116)

mω  ∂ξ ∂Q



(5.117)

= ξ◦

∂ ∂ξ

(5.118)

and thus, using in turn Eq. (5.117), to 2 mω ∂2 ∂2 ◦2 ∂ = ξ = ∂Q2 ∂ξ 2  ∂ξ 2

Thereby, using Eqs. (5.117) and (5.119), Eq. (5.115) becomes       mω2  2 1 2 mω ∂2 n (ξ) − ξ − n+ ω n (ξ) = 0 2m  ∂ξ 2 2 mω 2

(5.119)

CH005.tex

152

10/3/2011

22: 1

Page 152

ENERGY REPRESENTATION FOR QUANTUM HARMONIC OSCILLATORS

or, after simplifying by ω  2  ∂ n (ξ) − (ξ 2 − (2n + 1))n (ξ) = 0 ∂ξ 2 with, due to Eq. (5.116), the following boundary condition: n (ξ) → 0

(5.120)

ξ→∞

when

(5.121)

resulting from the fact that the wavefunctions have to be normalized in order to verify that their squared modulus must be a density probability obeying +∞ |n (ξ)|2 dξ = 1

(5.122)

−∞

5.2.2

Special solutions of Eq. (5.120)

Now, look at Eq. (5.120) for the lowest state situation corresponding to n = 0, that is,   2 ∂ 0 (ξ) − (ξ 2 − 1)0 (ξ) = 0 (5.123) ∂ξ 2 with the same boundary condition (5.121). Search for a solution of the form 0 (ξ) = (e±ξ Then, it yields   2 ∂e±ξ /2 2 = ±ξ(e±ξ /2 ) ∂ξ



2 /2

)

(5.124)

∂2 e±ξ /2 ∂ξ 2 2

 = (ξ 2 ± 1)(e±ξ

2 /2

)

so that the second partial derivative of the expression (5.124) reads  2  ∂ 0 (ξ) − (ξ 2 ± 1)0 (ξ) = 0 (5.125) ∂ξ 2 Thus, it appears that the two solutions of Eq. (5.123) are verified. But, the boundary condition (5.121) being not compatible with the + solution, the physical solution is necessarily the following one: 0 (ξ) = e−ξ

2 /2

This last equation is the unnormalized ground-state wavefunction of the Hamiltonian of the harmonic oscillator satisfying Eq. (5.115) with the ground-state energy ω/2, its normalized form being 0 (ξ) = C0 (e−ξ

2 /2

)

(5.126)

where C0 is the normalization constant of the wavefunction. The normalization constant C0 must be such that Eq. (5.122) has to be satisfied. Hence, using Eq. (5.117) in order to return from the dimensionless variable ξ to the dimensioned one Q, the normalization of the ground-state wavefunction (5.126) reads (C02 )−1

+∞  mω  = exp − Q2 dQ  −∞

CH005.tex

10/3/2011

22: 1

Page 153

5.2 WAVEFUNCTIONS CORRESPONDING TO HAMILTONIAN EIGENKETS

and, thus, after integration (C02 )−1

5.2.3

 =

π mω

 mω 1/4

C0 =

or

π

153

(5.127)

Recurrence relation between wavefunctions

Now, in order to pass from the ground-state wavefunction to the excited wavefunctions, use Eq. (5.63), that is, √ a† |{n} = n + 1|{n + 1} Then premultiplying both terms by any bra, Hermitian conjugate of some eigenket of the position operator, one obtains {Q}|{n + 1} = √

1 n+1

{Q}|a† |{n}

(5.128)

or, due to Eq. (5.4), it reads 1 1 {Q}|(ξ ◦ Q − iζ ◦ P)|{n} {Q}|{n + 1} = √ √ 2 n+1 with ◦



ξ =

mω 

 and

◦

ζ =

1 mω

(5.129)

Again, introduce the closure relation over the eigenstates of the position operator: ⎫ ⎧ +∞ ⎬ ⎨ 1 1 {Q}|(ξ ◦ Q − iζ ◦ P) {Q}|{n + 1} = √ √ |{Q }{Q }|dQ |{n} ⎭ ⎩ 2 n+1 −∞

leading to 1 1 {Q}|{n + 1} = √ √ {Q}|(ξ ◦ Q − iζ ◦ P) 2 n+1

+∞ |{Q }{Q }|{n}dQ

−∞

Hence, Eq. (5.128) becomes 1 1 n+1 (Q) = √ √ {Q}|(ξ ◦ Q − iζ ◦ P) 2 n+1

+∞ |{Q }n (Q ) dQ

−∞

with n+1 (Q) = {Q}|{n + 1}

and

n (Q ) = {Q }|{n}

Next, observe that, according to Eqs. (3.50) and (3.52) {Q}|P|{Q } = −iδ(Q − Q ) {Q}|Q|{Q } = δ(Q − Q )Q

∂ ∂Q

(5.130)

CH005.tex

154

10/3/2011

22: 1

Page 154

ENERGY REPRESENTATION FOR QUANTUM HARMONIC OSCILLATORS

As a consequence, using these expressions and the fact that, in wave mechanics, Q acts as a scalar Q, Eq. (5.130) transforms to   1 ∂ 1 n+1 (Q) = √ √ (5.131) n (Q) ξ ◦ Q − ζ ◦ ∂Q 2 n+1 Again, in view of Eqs. (5.117) and (5.118) and since [cf., Eq. (5.129)] the following relation between ζ ◦ and ξ ◦ exists ζ ◦ =

1 ξ◦

so that Eq. (5.131) takes the final recurrence form   1 1 ∂ n+1 (ξ) = √ √ n (ξ) ξ− ∂ξ 2 n+1

(5.132)

keeping in mind Eq. (5.117), that is, ξ = ξ◦ Q

5.2.4

Obtaining the lowest wavefunctions

5.2.4.1 First excited wavefunction Now, recall that the ground-state wavefunction (5.126), for reasons that will become apparent, may be formally written 0 (ξ) = C0 H0 (ξ)(e−ξ

2 /2

)

(5.133)

with H0 (ξ) = 1

(5.134)

and where C0 is the normalization constant of the wavefunction. Apply Eq. (5.132) to the ground-state wavefunction (5.133), that is, for n = 0   ∂ 1 1 2 1 (ξ) = √ √ C0 ξ − (5.135) (e−ξ /2 ) ∂ξ 2 1 The partial derivative with respect to ξ being ∂ −ξ2 /2 2 (e ) = −(ξe−ξ /2 ) ∂ξ Eq. (5.135) yields 1 (ξ) = C1 2(ξe−ξ where

2 /2

)

1 1  mω 1/4 C1 = √ C0 = √ 2 2 π

(5.136)

(5.137)

Finally, Eq. (5.136) may be written in a form similar to that of Eq. (5.133): 1 (ξ) = C1 H1 (ξ)e−ξ

2 /2

with

H1 (ξ) = 2ξ

(5.138)

CH005.tex

10/3/2011

22: 1

Page 155

5.2 WAVEFUNCTIONS CORRESPONDING TO HAMILTONIAN EIGENKETS

155

5.2.4.2 Second excited wavefunction Next, in order to get the second excitedstate wavefunction, apply Eq. (5.132) to Eq. (5.136), that is   1 1 ∂ 2 ξ− {C1 2(ξe−ξ /2 )} 2 (ξ) = √ √ ∂ξ 2 2! which reads

 2 (ξ) = 2C2 ξ(ξe

−ξ 2 /2



∂ 2 ) − (ξe−ξ /2 ) ∂ξ

(5.139)

with, in view of Eq. (5.137), 1 C2 = √ 2!



   1 1 2 1 C0 √ C1 = √ √ 2 2 2!

(5.140)

By derivation one obtains ∂ 2 2 (ξe−ξ /2 ) = (1 − ξ 2 )(e−ξ /2 ) ∂ξ

(5.141)

so that Eq. (5.139) yields 2 (ξ) = C2 (4ξ 2 − 2)(e−ξ

2 /2

)

(5.142)

or 2 (ξ) = C2 H2 (ξ)(e−ξ

2 /2

)

H2 (ξ) = 4ξ 2 − 2

with

(5.143)

5.2.4.3 Third excited wavefunction Again, to get the third excited-state wavefunction, apply Eq. (5.132) a third time to Eq. (5.142), leading to   ∂ 2 3 (ξ) = C3 ξ − {(4ξ 2 − 2)e−ξ /2 } (5.144) ∂ξ with 1 1 1 C3 = √ √ C2 = √ 2 3! 3!



1 √ 2

3 C0

which transforms to

  ∂ 2 2 3 (ξ) = C3 (4ξ 3 − 2ξ)(e−ξ /2 ) − {(4ξ 2 − 2)(e−ξ /2 )} ∂ξ

Then, obtaining by differentiation ∂ 2 2 (4ξ 2 − 2)(e−ξ /2 ) = (8ξ − (4ξ 2 − 2)ξ)(e−ξ /2 ) ∂ξ the wavefunction becomes 3 (ξ) = C3 (8ξ 3 − 12ξ)(e−ξ

2 /2

)

(5.145)

or 3 (ξ) = C3 H3 (ξ)(e−ξ

2 /2

)

with

H3 (ξ) = 8ξ 3 − 12ξ

(5.146)

CH005.tex

156

10/3/2011

22: 1

Page 156

ENERGY REPRESENTATION FOR QUANTUM HARMONIC OSCILLATORS

5.2.4.4 nth excited wavefunction Note that the functions Hk (ξ) (5.134), (5.138), (5.143), and (5.146) are the first Hermite polynomials. Besides, proceeding in a similar way for the higher excited wavefunctions, one would obtain now n (ξ) = Cn Hn (ξ)(e−ξ with

2 /2

)

(5.147)

  1 1 n C0 Cn = √ √ 2 n!

or, in view of Eq. (5.127),

  1 n  mω 1/4 1 Cn = √ √ π 2 n! and with, for n = 4, 5, and 6,

(5.148)

H4 (ξ) = 16ξ 4 − 48ξ 2 + 12 H5 (ξ) = 32ξ 5 − 160ξ 3 + 120ξ

(5.149)

H6 (ξ) = 64ξ − 480ξ + 720ξ − 120 6

4

2

5.2.4.5 Pictorial representation of the lowest wavefunctions The five lowest wavefunctions and energy levels are pictured in Fig. 5.1a, whereas the corresponding wavefunctions and energy levels of the particle-in-a-box model are shown in Fig. 5.1b. Observe that, in agreement with Eqs. (5.126), (5.138), (5.143), (5.146), and (5.147), the parity of the wavefunctions n (ξ) is alternatively changing, those characterized by even quantum numbers n, being gerade and the other ones, characterized by odd quantum numbers n, being ungerade. Observe also that this figure illustrates the node number increase of the wavefunctions when enhancing the quantum number and thus the energy, an evolution that is almost similar to that encountered in the particle-in-a-box model, as may be verified by inspection of Fig. 5.1(b).

5.3

DYNAMICS

In the previous sections we found many important results dealing with static situations of quantum harmonic oscillators. We have now to search the dynamics of these oscillators via the time dependence of the mean values of the basic operators averaged over the eigenkets of the harmonic oscillator Hamiltonian. To get these time dependent average values, it will be suitable to work within the Heisenberg picture (where the operators depend on time whereas the kets remain constant).

5.3.1

Heisenberg equations for oscillator operators

5.3.1.1 Ladder operators Heisenberg equations Look, therefore, at the Heisenberg equation governing the dynamical equation of the lowering operator a(t), which, according to Eq. (3.94), reads in the present situation   da(t) = [a(t), H] (5.150) i dt

CH005.tex

10/3/2011

22: 1

Page 157

5.3

Energy ( ω)

DYNAMICS

157

7 6

V(ξ)

5 Ψ4(ξ)

E4 ⫽ 9 2

4 nodes n ⫽ 5

E3 ⫽ 7 2

3 nodes n ⫽ 4

E2 ⫽ 5 2

2 nodes n ⫽ 3

E1 ⫽ 3 2

1 node n ⫽ 2

E0 ⫽ 1 2

0 node n ⫽ 1

4 Ψ3(ξ) 3 Ψ2(ξ) 2 Ψ1(ξ) 1 Ψ0(ξ) ⫺4

⫺2

0 (a)

2

ξ

4

0

a

x

(b)

Figure 5.1 Five lowest energy levels and wavefunctions. Comparison between (a) quantum harmonic oscillator and (b) particle-in-a-box model.

which, with the help of Eq. (5.9) defining the Hamiltonian, becomes      1 da(t) = ω a(t), a(t)† a(t) + i dt 2 or   da(t) = ω(a(t)a(t)† a(t) − a(t)† a(t)a(t)) i dt and thus



da(t) dt

 = −iω[a(t), a(t)† ]a(t)

Again, the commutation rule (5.5) holds for any time, so that [a(t), a(t)† ] = 1 and thereby



da(t) dt

 = −iωa(t)

Hence, after integration, that gives a(t) = a(0)e−iωt

(5.151)

CH005.tex

158

10/3/2011

22: 1

Page 158

ENERGY REPRESENTATION FOR QUANTUM HARMONIC OSCILLATORS

the Hermitian conjugate of which being a† (t) = a† (0)eiωt

(5.152)

Now, observe that when dealing with 3D oscillators, the latter results read ak (t) = ak (0)e−iωk t

ak† (t) = ak† (0)eiωk t

and

with k standing for the x, y, and z components. 5.3.1.2 Position and momentum Heisenberg equations In view of Eq. (5.6), and owing to Eqs. (5.151) and (5.152), the time dependence of the quantum oscillator position coordinate appears to be   Q(t) = (a† eiωt + ae−iωt ) (5.153) 2mω Passing then from the imaginary exponentials to the corresponding sine and cosine functions leads to   Q(t) = (a† cos ωt + ia† sin ωt + a cos ωt − ia sin ωt) 2mω or   ((a† + a) cos ωt + i(a† − a) sin ωt) (5.154) Q(t) = 2mω so that, due to Eqs. (5.6) and (5.7), we have Q(t) = Q(0) cos ωt + with, respectively,

 P(0) = i  Q(0) =

1 P(0) sin ωt mω

(5.155)

mω † (a − a) 2

(5.156)

mω † (a + a) 2

(5.157)

Now, consider the commutator of the position coordinate operators at different times, that is, [Q(t), Q(t  )]      = ((a† eiωt + ae−iωt )(a† eiωt + ae−iωt ) − (a† eiωt + ae−iωt ) 2mω × (a† eiωt + ae−iωt )) which, after performing the products and simplification gets [Q(t), Q(t  )] =

     (a† a(eiω(t−t ) − e−iω(t−t ) ) + aa† (e−iω(t−t ) − eiω(t −t ) )) 2mω

CH005.tex

10/3/2011

22: 1

Page 159

5.3

DYNAMICS

159

and, thus, after coming back to the sine function, we have  (a† a2i sin (ω(t − t  )) − aa† 2i sin (ω(t − t  ))) 2mω Then, with the help of the commutation rule (5.5), this last result reduces, after simplification, to [Q(t), Q(t  )] =

[Q(t), Q(t  )] = −i

 sin (ω(t − t  )) mω

(5.158)

We emphasizes that this commutator differs from zero, that is, [Q(t), Q(t  )] = 0 On the other hand, the time dependence of the momentum, which reads in view of Eq. (5.7)  mω † iωt (a e − ae−iωt ) (5.159) P(t) = i 2 transforms after passing to the trigonometric functions and using Eq. (5.7)  mω † P(t) = i ((a − a) cos ωt + i(a† + a) sin ωt) 2 so that   mω † mω † P(t) = i (a − a) cos ωt − (5.160) (a + a) sin ωt 2 2 and, therefore, due to Eqs. (5.156) and (5.157), leads to P(t) = P(0) cos ωt − mωQ(0) sin ωt

(5.161)

Hence, the commutator of P at different times is not zero.

5.3.2 Time dependence of mean values averaged on Hamiltonian eigenkets Recall that operators, unlike average values, are not directly connected with experience. Thus, it is now necessary to study the dependence of the average values of the operators defined by Eq. (5.155) or (5.154) and by Eq. (5.160) or (5.161). 5.3.2.1 Average Values of Q(t) and P(t) First, consider the average values of the momentum and position coordinates on the eigenkets of the harmonic Hamiltonian. According to Eq. (5.154), that of the position operator reads   {n}|Q(t)|{n} = ({n}|a† |{n}eiωt + {n}|a|{n}e−iωt ) 2mω Then, keeping in mind Eq. (5.53) and its Hermitian conjugate, that is, √ √ and {n}|a† = n{n − 1}| a|{n} = n|{n − 1}

CH005.tex

160

10/3/2011

22: 1

Page 160

ENERGY REPRESENTATION FOR QUANTUM HARMONIC OSCILLATORS

the average value of the position coordinate becomes  1  {n}|Q(t)|{n} = ({n − 1}|{n}eiωt + {n}|{n − 1}e−iωt ) 2 2mω Thus, owing to the orthonormality of the eigenkets of the harmonic oscillator Hamiltonian, we have {n}|Q(t)|{n} = 0

(5.162)

Now, proceeding in the same way for the momentum coordinate by the aid of Eq. (5.159), it may be easily shown that {n}|P(t)|{n} = 0

(5.163)

Note that the results (5.162) may be also found by the aid of the wave mechanics. In this quantum picture, the left-hand side of Eq. (5.162) reads +∞ {n}|Q(t)|{n} = n (Q)Q(t)n (Q) dQ −∞

where n (Q) ≡ {Q}|{n} Now, since Q commutes with n (Q) and irrespective of the time t, which may be omitted, the average value reduces to +∞ n (Q)2 Q dQ {n}|Q|{n} = −∞

Now, observe that, whatever n (Q), the parity of its square is always even, whereas that of Q is odd. Hence, the parity of the integrand, which is the product of that of n (Q)2 by that of Q, is always odd, so that the integral involving this integrand must be necessarily zero, the contribution from 0 to +∞ being canceled by that from −∞ to 0, that is, +∞ n (Q)2 Q dQ = 0 −∞

On the other hand, in the Schrödinger picture, Eq. (5.163) takes the form {n}(t)|P|{n}(t) = 0 Then, inserting a closure relation over the basis {|{Q}} before and after P, it transforms by the aid of Eq. (3.50) into +∞ ∂ n (Q)∗ n (Q) dQ = 0 −i ∂Q −∞

CH005.tex

10/3/2011

22: 1

Page 161

5.3

DYNAMICS

161

Of course, Eqs. (5.162) and (5.163) may be immediately extended to the 3D oscillator to give for the three Cartesian components k = x, y, and z: {n}|Qk (t)|{n} = {n}|Pk (t)|{n} = 0 5.3.2.2 Average values of V(t) and T(t) Now, consider the time dependence of the average potential energy. In the Heisenberg representation, this average value reads {n}|V(t)|{n} = 21 mω2 {n}|Q(t)2 |{n}

(5.164)

Next, the right-hand-side average value may be expressed in terms of the raising and lowering operators by the aid of Eq. (5.153):  {n}|(a† eiωt + ae−iωt )2 |{n} 2mω By expansion of the square involved on the right-hand side, one finds {n}|Q(t)2 |{n} =

(a† eiωt + ae−iωt )2 = (a† )2 e2iωt + (a)2 e−2iωt + a† a + aa†

(5.165)

(5.166)

or, after using the usual commutation rule (5.5) between the raising and lowering operators, Eq. (5.165) becomes  {n}|((a† )2 e2iωt + (a)2 e−2iωt + 2a† a + 1)|{n} 2mω By inserting this result into Eq. (5.164), the average value of the potential energy is {n}|Q(t)2 |{n} =

{n}|V(t)|{n} = 41 ω{n}|((a† )2 e2iωt + (a)2 e−2iωt + 2a† a + 1)|{n} so that, owing to Eqs. (5.53) and (5.63), √ √ {n}|(a)2 |{n} = n n − 1{n}|{n − 2} = 0 √ √ {n}|(a† )2 |{n} = n + 2 n + 1{n}|{n + 2} = 0 Now, due to Eq. (5.40) {n}|(a† a)|{n} = n{n}|{n} = n Hence, using these equations, the average value of the potential energy becomes   (5.167) {n}|V(t)|{n} = 21 ω n + 21 = const. Hence, the average potential energy remains constant throughout time and equal to half the energy when the oscillator is in any eigenstate |{n} of its Hamiltonian as can be directly obtained from the virial theorem. In a similar way, one would find for the mean kinetic energy averaged over an Hamiltonian eigenket     1 1 P(t)2 |{n} = ω n + = const. (5.168) {n}| 2m 2 2 with mω † iωt (a e − ae−iωt )2 P(t)2 = − 2

CH005.tex

162

10/3/2011

22: 1

Page 162

ENERGY REPRESENTATION FOR QUANTUM HARMONIC OSCILLATORS

Of course, results (5.167) and (5.168) may be generalized to the x, y, and z components of 3D harmonic oscillators, leading one to write for each component     Pk (t)2 1 1 {n}k |Vk (t)|{n}k  = {n}k | |{n}k  = ωk nk + 2m 2 2

5.4

BOSON AND FERMION OPERATORS

As observed above, the non-Hermitian annihilation and creation operators a and a† are very important in the quantum approach of harmonic oscillators. They are often called Boson operators because they are related to the Bose–Einstein statistics where the number of particles inside a nondegenerate energy level is arbitrary. On the other hand, in the study of double-energy-level systems one meets non-Hermitian operators af and af† , which play for these simple systems a role presenting analogies with those of a and a† in the quantum oscillators theory. (In all this section, the subscript f refers to Fermions and the corresponding double-energy-level systems.) These new operators are called Fermion operators because they are related to the Fermi–Dirac statistics where the number of particles inside a nondegenerate energy level can be only either zero or unity. Owing to the importance of Fermion operators description in many double-energy-level system studies, and of their deep analogy with the Boson operators, it is convenient to treat here the Fermion operators. Consider a two-energy-level system, the Hamiltonian eigenvalue equation of which is H|(k)f  = Ek |(k)f  with k = 0, 1 where Ek are the two eigenvalues, that is, E0 and E1 , with E1 > E0 , whereas |(k)f  are the corresponding eigenkets |(0)f  and |(1)f  of H, the first one being the ground state and the last one the excited state as illustrated in Fig. 5.2.

Figure 5.2

E1

|(1)f 〉

E0

|(0)f 〉

Fermion energy levels and corresponding eigenkets.

CH005.tex

10/3/2011

22: 1

Page 163

5.4

BOSON AND FERMION OPERATORS

163

When normalized and orthogonal, these states read (1)f |(1)f  = (0)f |(0)f  = 1

(5.169)

(1)f |(0)f  = 0

(5.170)

Now, by analogy with the Boson operators a† and a, introduce two kinds of non-Hermitian operators af† and af , the Fermion operators, obeying the following anticommutation rules: [af , af ]+ = [af† , af† ]+ = 0

(5.171)

[af , af† ]+ = 1

(5.172)

where the anticommutator is defined by [A, B]+ ≡ AB + BA The two Fermion operators are assumed to act on the kets |(1)f  and |(0)f  according to af |(1)f  = |(0)f 

(5.173)

af† |(0)f  = |(1)f 

(5.174)

af |(0)f  = 0

(5.175)

af† |(1)f  = 0

(5.176)

and

Now, by analogy with the Hermitian number occupation operator (5.12) of Boson operators, introduce here the Hermitian operators defined by Nf = af† af

with

Nf = Nf†

(5.177)

since (af† af )† = af† af Then, owing to Eq. (5.177), the action of Nf on the excited state reads Nf |(1)f  = af† af |(1)f  or, owing to Eq. (5.173), Nf |(1)f  = af† |(0)f  and thus, due to Eq. (5.174), Nf |(1)f  = 1|(1)f  On the other hand, the action of Nf on the ground state reads Nf |(0)f  = af† af |(0)f 

(5.178)

CH005.tex

164

10/3/2011

22: 1

Page 164

ENERGY REPRESENTATION FOR QUANTUM HARMONIC OSCILLATORS

which, according to Eq. (5.175), reads Nf |(0)f  = 0

(5.179)

Equations (5.178) and (5.179) may be viewed as corresponding to the eigenvalue equation (5.75), whereas Eqs. (5.173) and (5.174) may be put into correspondence with Eq. (5.80). Moreover, the anticommutation rules (5.172) and (5.171) dealing with the Fermion operators are in correspondence with the commutation rule of the Boson operators (5.77) and (5.78). Finally, the Hermitian operator Nf defined by Eq. (5.177) is the Fermion operator analog of the number occupation operator N (5.76) involving Boson operators. It is now of interest to give matrix representations of the Fermion operators. For this purpose, it is convenient to represent the two orthogonal kets |(0)f  and |(1)f , by two orthonormalized column vectors of dimension 2 according to     0 1 |(0)f  = and |(1)f  = (5.180) 1 0 which satisfy the orthonormality properties, since one obtains, respectively,   1 (1)f |(1)f  = (1 0) =1 0 (0)f |(0)f  = (0

  0 1) =1 1

(1)f |(0)f  = (1

0)

  0 =0 1

Then, in order to be compatible with Eqs. (5.171)–(5.179), the matrix representations of the two Fermion operators af and af† of the Hermitian operator Nf have to be chosen in such a way as     0 0 0 1 † af = and af = (5.181) 1 0 0 0 These matrix representations, which may be compared to those of the Boson operators a and a† given by Eq. (5.81), satisfy, as required, the anticommutation relation (5.172) because       0 0 0 1 0 1 0 0 + [af , af† ]+ = 1 0 0 0 0 0 1 0 leading after matrix multiplication to    0 0 1 [af , af† ]+ = + 0 1 0

0 0



 =

1 0

0 1

 = 1

In the same way, the matrix representation of the anticommutator of af with itself, as required by (5.171), reads,       0 0 0 0 0 0 0 0 [af , af ]+ = + = 0 1 0 1 0 1 0 1 0

CH005.tex

10/3/2011

22: 1

Page 165

5.5

CONCLUSION

165

One would find in a similar way that the anticommutator of af† with itself is also zero as required by (5.171). Moreover, using Eqs. (5.181), the matrix representation of the Hermitian operator Nf defined by Eq. (5.177) is  Nf =

0 0

1 0



0 1

0 0



 =

1 0

0 0

 (5.182)

Moreover, using Eqs. (5.180) and (5.182), the matrix representation of the action of the operator Nf on |(1)f  and |(0)f , reads  Nf |(1)f  =

1 0

0 0

  1 0

 and

Nf |(0)f  =

1 0

0 0

  0 1

so it appears that Eqs. (5.178) and (5.179) are satisfied, since the two expressions above yield, respectively,   1 Nf |(1)f  = = |(1)f  0 and Nf |(0)f  =

  0 =0 0

On the other hand, the matrix representation of the actions of the two Fermion operators af and af† on the two states, lead, respectively, as required by Eqs. (5.173)–(5.176), to      0 0 1 0 = = |(0)f  af |(1)f  = 1 0 0 1  af |(0)f  =  af† |(1)f  =  af† |(0)f 

5.5

=

0 0

0 1

0 0

    0 0 = =0 1 0

0 0

1 0

    1 0 = =0 0 0

1 0

    0 1 = = |(1)f  1 0

CONCLUSION

In the present chapter devoted to the study of single isolated quantum harmonic oscillators, we have obtained many important results. Among them, there are the eigenvalues of the Hamiltonian and the action of the raising and lowering operators on the orthonormalized eigenvectors of the Hamiltonian, which constitute a basis in

CH005.tex

166

10/3/2011

22: 1

Page 166

ENERGY REPRESENTATION FOR QUANTUM HARMONIC OSCILLATORS

the state space. We have also considered the time evolution of the average values performed over these kets, of different observables dealing with the oscillators. All these important results, which are convenient to know, are collected in the following list: Relations between ladder operators and position and momentum operators    mω † † Q= and P=i (a + a) (a − a) 2mω 2 [a, a† ] = 1 Eigenvalue equation of the harmonic Hamiltonian:   H = ω a† a + 21     ω a† a + 21 |{n} = ω n + 21 |{n} with n = 0, 1, 2, 3, . . . {n}|{m} = δnm Action of the ladder operators on the harmonic Hamiltonian eigenkets: √ √ a|{n} = n|{n − 1} and a† |{n} = n + 1|{n + 1} Time dependence of the Boson operators: a(t) = a(0)e−iωt

and

a† (t) = a† (0)eiωt

Finally, the analytical expressions for the vibrational wavefunctions associated with the quantized energy levels exist, which yield some knowledge concerning the corresponding somewhat “esoteric” kets.

BIBLIOGRAPHY C. Cohen-Tannoudji, B. Diu, and F. Laloe. Quantum Mechanics. Wiley-Interscience: Hoboken, NJ, 2006. H. Eyring, J. Walter, and G. E. Kimball. Quantum Chemistry. Wiley: New York, 1944. W. H. Louisell. Quantum Statistical Properties of Radiation. Wiley: New York, 1973.

CH006.tex

16/3/2011

CHAPTER

13: 23

Page 167

6

COHERENT STATES AND TRANSLATION OPERATORS INTRODUCTION The previous chapter focused essentially on the stationary orthonormal eigenstates of the harmonic oscillator Hamiltonian, which form a basis of the state space. There exist other states dealing with harmonic oscillators that also play a very fundamental role in the area of quantum harmonic oscillators. They are the coherent states |{α}, which are, by definition, the eigenkets of the lowering operator a. They are of great importance for numerous reasons. The first one is that the coherent states, whatever they are, minimize the Heisenberg uncertainty relations. Another one, which is deeply connected to the first one, is that the harmonic oscillator operators, when averaged on it, lead to behaviors that are more and more classical when the eigenvalue α corresponding to the eigenket |{α} is increasing. In addition, they are good simple examples of how the formalism of quantum mechanics operates. Moreover, since they are the eigenkets of a non-Hermitian operator, they illustrate that, unlike the number occupation operator, they do not necessarily admit real eigenvalues and furthermore are continuous and nonorthogonal. Moreover, they play a fundamental role in the area of the quantum theory of light, the average values of the electric field operators performed on them, being reached via the corresponding classical fields. For all these reasons, coherent states now merit study. Thus, the present chapter will begin by deducing its definition, the expansion of a coherent state on the eigenkets of the number occupation operator. Then, the scalar product between two coherent states will be calculated. The chapter will continue by proving that the Heisenberg uncertainty relation is always minimal for coherent states. Then, it will be shown that coherent states may be generated by the action of the translation operator. One section concerns the time dependence of coherent states. Thus, it will be possible to obtain the wave representation of coherent states and of their time dependence. In another section, it will be also possible to calculate by the aid of the translation operator, the Franck–Condon factors, that is, the scalar products between any eigenfunction of the harmonic oscillator Hamiltonian and another one that has been translated with respect to the first one. The chapter ends with the quest for the energy levels of driven harmonic oscillators, which are deeply connected to the properties of coherent states and of translation operators. This is the opportunity

Quantum Oscillators, First Edition. Olivier Henri-Rousseau and Paul Blaise. © 2011 John Wiley & Sons, Inc. Published 2011 by John Wiley & Sons, Inc.

167

CH006.tex

168

16/3/2011

13: 23

Page 168

COHERENT STATES AND TRANSLATION OPERATORS

to test numerically approximate approaches of these levels through truncated matrix representations of the driven oscillator Hamiltonian in the basis of the eigenkets of the harmonic oscillator.

6.1

COHERENT-STATE PROPERTIES

6.1.1 Definition and expansion within Hamiltonian eigenkets basis By definition, coherent states |{α} are the eigenkets of the lowering a operator, obeying therefore a|{α} = α|{α}

(6.1)

where α are the corresponding eigenvalues, the Hermitian conjugate of which is {α}|a† = α∗ {α}|

(6.2)

Observe that, since a is not Hermitian, its eigenvalues are not necessarily real and different coherent states are not necessarily orthogonal. An important information dealing with coherent states is contained in their expansion over the eigenkets of the occupation number operator, that is, of the harmonic Hamiltonian. Keeping in mind the eigenvalue equation (5.40), that is, a† a|{n} = n|{n} with, of course, since a† a is Hermitian, 1=

∞ 

|{n}{n}|

and

{m}|{n} = δmn

(6.3)

n=0

In order to get the expansion of a coherent state on this basis {|{n}}, introduce the unity operator resulting from the closure relation as follows: |{α} =

∞ 

|{n}{n}|{α}

n=0

so that |{α} =

∞ 

Cn (α)|{n}

(6.4)

n=0

where Cn (α) is the scalar product given by Cn (α) = {n}|{α} On the other hand, observe that, by action on the left of a on both sides of Eq. (6.4), one gets a|{α} =

∞  n=0

Cn (α)a|{n}

CH006.tex

16/3/2011

13: 23

Page 169

6.1

COHERENT-STATE PROPERTIES

169

Next, using Eq. (5.53) in order to find the expression of the right-hand side of this last equation, one finds a|{α} =

∞ 

√ Cn (α)) n|{n − 1}

n=0

Again, using eigenvalue equation (6.1) we have α|{α} =

∞ 

√ Cn (α) n|{n − 1}

n=0

and after using Eq. (6.4) for the left-hand side of this last equation, one gets α

∞ 

Cn (α)|{n} =

n=0

∞ 

√ Cn (α) n|{n − 1}

(6.5)

n=0

Now, observe that the first term involved on the right-hand side of Eq. (6.5) corresponding to n = 0 is zero since no eigenkets of the harmonic oscillator Hamiltonian under |{0} exist. Thus, performing the following variable change n→n+1 Eq. (6.5) reads α

∞ 

Cn (α)|{n} =

n=0

∞ 

√ Cn+1 (α) n + 1|{n}

n=0

Since this last expression must be true for each term of the sum, the following relation must be verified: √ (6.6) Cn+1 (α) n + 1 = αCn (α) which yields for n = 0

and for n = 1

√ 1C1 (α) = αC0 (α)

(6.7)

√ 2C2 (α) = αC1 (α)

Then, inserting in this last result Eq. (6.7), one obtains α2 C2 (α) = √ C0 (α) 2

(6.8)

Moreover, for n = 2, Eq. (6.6) gives √ 3C3 (α) = αC2 (α) which, using Eq. (6.8) leads to α3 C3 (α) = √ √ C0 (α) 3 2

(6.9)

CH006.tex

170

16/3/2011

13: 23

Page 170

COHERENT STATES AND TRANSLATION OPERATORS

Hence, one obtains by recurrence from Eqs. (6.7)–(6.9) αn Cn (α) = √ C0 (α) n!

(6.10)

Allowing to transform the expansion (6.4) of the coherent state into |{α} = C0 (α)

∞  αn √ |{n} n! n=0

(6.11)

Furthermore, in order to find the expression of the unknown coefficient C0 (α), use the Hermitian conjugate of Eq. (6.11), that is, ∞  (α∗ )m {α}| = C0 (α) √ {m}| m! m=0 ∗

(6.12)

allowing, with the help of Eq. (6.11), to get the norm of the coherent state   (α∗ )m (α)n {α}|{α} = |C0 (α)|2 √ √ {m}|{n} m! n! m n which, in view of the orthonormality property appearing in (6.3), reduces to {α}|{α} = |C0 (α)|2

 |α|2n n!

n

{n}|{n}

(6.13)

Again, owing to Eq. (6.3), and after imposing the coherent state to be normalized, it reads {α}|{α} = 1

{n}|{n} = 1

and

so that Eq. (6.13) reduces to |C0 (α)|2

 |α|2n n!

n

=1

which, due to the expansion properties of the exponential, yields |C0 (α)|2 = e−|α|

2

(6.14)

At last, passing from the squared absolute value |C0 to the corresponding C0 (α), and after neglecting a phase factor eiϕ without interest, one obtains (α)|2

C0 (α) = e−|α|

2 /2

(6.15)

so that the recurrence equation (6.10) becomes Cn (α) = e−|α|

2 /2

αn √ n!

which allows us to transform Eq. (6.4) into |{α} =

2 e−|α| /2

 α n  |{n} √ n! n

(6.16)

CH006.tex

16/3/2011

13: 23

Page 171

6.1

COHERENT-STATE PROPERTIES

171

This 1D result may be generalized to three dimensions, leading, for each x, y, and z components, to coherent states of the form   α nk  2 |{α}k  = e−|αk | /2 |{n}k  √k nk ! nk obeying the eigenvalue equations ak |{α}k  = αk |{α}k 

6.1.2

Scalar products and closure relations

Since coherent states are the eigenkets of a, which is non-Hermitian, different coherent states being different eigenkets of a, have no reason to be orthogonal, since this property is specific to eigenkets of Hermitian operators. Hence, because of the absence of orthogonality between two coherent states characterized by two different eigenvalues of α, it is necessary to determine their scalar product. 6.1.2.1 Scalar products For this purpose, consider a coherent state |{β} obeying an expression of the same form as Eq. (6.1), which reads a|{β} = β |{β}

(6.17)

where β is the corresponding eigenvalue. The expansion of this new coherent state is of course given by an expression similar to Eq. (6.16), so that its Hermitian conjugate reads as the bra (6.12), that is,  β∗m 2 {β}| = e−|β| /2 √ {m}| m! m Thereby, the scalar product of |{α}, defined by Eq. (6.16) and of |{β} given by Eq. (6.17), yields   αn β∗m 2 2 {β}|{α} = e−|β| /2 e−|α| /2 √ √ {m}|{n} n! m! m n Next, using the orthonormality properties {m}|{n} = δmn the above scalar product reduces to {β}|{α} = e

−|β|2 /2 −|α|2 /2

e

   αβ∗ n n! n

or, after passing to exponentials, to

  |β|2 + |α|2 αβ∗ {β}|{α} = exp − e 2

and thus {β}|{α} = e−|α−β|

2 /2

(6.18)

CH006.tex

172

16/3/2011

13: 23

Page 172

COHERENT STATES AND TRANSLATION OPERATORS

6.1.2.2 Closure relations Since coherent states are not orthogonal, they cannot be used to generate a standard discrete closure relation. However, despite this difficulty, it is possible to obtain a continuous closure relation given by 1 I= π

+∞ +∞ |{α} {α}|d Re (α) d Im (α)

(6.19)

−∞ −∞

In order to prove that Eq. (6.19) is unity, first start from the α complex eigenvalue written as α = ρeiϕ

(6.20)

where ρ and ϕ are both real, the differentiation of which yields dα = eiϕ (dρ + iρ dϕ) or, after passing to the trigonometric expression, dα = (cos ϕ + i sin ϕ) (dρ + iρ dϕ) and thus dα = d{Re (α)} + id{Im (α)} where Re (α) and Im (α) are, respectively, the real and imaginary parts of α obeying      d Re (α) cos ϕ −ρ sin ϕ dρ = (6.21) d Im (α) sin ϕ +ρ cos ϕ dϕ Next, consider the product of d Re(α) and d Im(α), namely d Re(α) d Im(α) = det (J) dρ dϕ

(6.22)

where J is the Jacobian, that is, the determinant corresponding to the matrix involved in Eq. (6.21), that is,   cos ϕ −ρ sin ϕ    J= sin ϕ +ρ cos ϕ Hence, the product (6.22) reads d Re (α) d Im (α) = ρ(cos2 ϕ + sin2 ϕ)dρ dϕ or, after simplifications d Re (α) d Im (α) = ρ dρ dϕ

(6.23)

Now, in view of Eq. (6.16) and of its Hermitian conjugate, one may write n α∗m 2  α |{α} {α}| = e−|α| √ √ |{n} {m}| n! m! m n which, using Eq. (6.20), yields |{α} {α}| = e−ρ

2

  ρn+m √ √ ei(n−m)ϕ |{n}{m}| n! m! m n

(6.24)

CH006.tex

16/3/2011

13: 23

Page 173

6.1

COHERENT-STATE PROPERTIES

173

As a consequence of Eqs. (6.23) and (6.24), the integral I appearing in Eq. (6.19) becomes +∞ 2π 1   |{n} {m}| −ρ2 n+m I= e ρ ρ dρ ei(n−m)ϕ dϕ (6.25) √ √ π m n n! m! 0

0

The last right-hand-side integral has the following solutions: 2π ei(n−m)ϕ dϕ = 2π

n=m

if

0

=

1 [ei(n−m)ϕ ]2π 0 =0 i (n − m)

if the integers n = m

so that 2π ei(n−m)ϕ dϕ = 2πδnm

(6.26)

0

Then, Eq. (6.25) reduces to I=



In

n

|{n}{n}| n!

(6.27)

with +∞ 2 e−ρ ρ2n ρ dρ

In = 2

(6.28)

−∞

Again, performing the variable change u = ρ2 the integrals (6.22) yield +∞ In = e−u un du = n! 0

Owing to this result, and according to the closure relation appearing in (5.43), the integral (6.27) reduces to  |{n}{n}| = π I=π n

Consequently, keeping in mind that, in view of Eqs. (6.20) and (6.23), ρeiϕ = α

and

ρ dρ dϕ = dRe (α) dIm (α)

it appears that the closure relation over the coherent sates (6.19) is +∞ +∞ −∞ −∞

|{α}{α}| dRe (α) dIm (α) = 1

(6.29)

CH006.tex

16/3/2011

13: 23

Page 174

174

COHERENT STATES AND TRANSLATION OPERATORS

6.2

POISSON DENSITY OPERATOR

Consider a pure density operator of oscillators described by coherent states, which according to Eq. (3.139), reads ρα = |{α}{α}|

(6.30)

with a|{α} = α|{α}

{α}|a† = {α}|α∗

and

and [a, a† ] = 1

α = α◦ eiϕ

and

(6.31)

Then, owing to Eq. (6.16), Eq. (6.30) becomes

 n     (α)m  α∗ −α◦2 ρα = e |{m}{n}| √ √ m! n! m n or, in view of Eq. (6.31), −α◦2

ρα = e

   (α◦ )m+n  |{m}{n}|ei(m−n)ϕ √ √ m! n! m n

(6.32)

Next, performing the average of this density operator over the phase ϕ according to 1 ρ¯ α = 2π

2π ρα dϕ 0

Eq. (6.32) transforms to −α◦2

ρ¯ α = e

2π    (α◦ )m+n  1 ei(m−n)ϕ dϕ |{m}{n}| √ √ 2π m! n! m n 0

which after integration using Eq. (6.26) yields  ◦ m+n  (α ) ◦2   ρ¯ α = e−α |{m}{n}|δmn √ √ m! n! m n or −α◦2

ρ¯ α = e

  (α◦ )2n  n

n!

|{n}{n}|

(6.33)

Now, observe that this density operator is diagonal in the basis {|{n}} and that its diagonal matrix elements are given by the following Poisson distribution:

 ◦ 2n −α◦2 (α ) {n}|ρ¯ α |{n} = e (6.34) n!

CH006.tex

16/3/2011

13: 23

Page 175

6.3 AVERAGE AND FLUCTUATION OF ENERGY

175

On the other hand, if the unaveraged density operator (6.32) is not diagonal, its matrix elements are given by {m}|ρα |{n} = e−α

◦2



 (α◦ )m+n ei(m−n)ϕ √ √ m! n!

(6.35)

Thus, the diagonal matrix elements (6.35) are the same as the diagonal ones (6.34) of the averaged density operator (6.33).

6.3

AVERAGE AND FLUCTUATION OF ENERGY

An important property of coherent states is that these states allow us to easily obtain the mean values of operators averaged on them, whereas another interest is that these mean values exhibit physical properties that are close to those of classical harmonic oscillators. The following sections will thus be devoted to calculate mean values of operators averaged over coherent states, the present one dealing with the average value of the Hamiltonian and of its square, allowing one to find the energy fluctuations within coherent states.

6.3.1

Average Hamiltonian

First, consider the mean value of the harmonic oscillator Hamiltonian averaged over coherent states, that is, Hα = {α}|H |{α} which becomes in view of Eq. (5.9)   Hα = ω{α}| a† a + 21 |{α} or Hα = ω{α}|a† a|{α} + 21 ω {α}|{α}

(6.36)

Moreover, in view of Eqs. (6.1) and (6.2), the right-hand-side average value appearing in Eq. (6.36) reads {α}|a† a|{α} = {α}|α∗ α|{α} = |α|2 {α}|{α}

(6.37)

so that, if the coherent states are normalized, the average value (6.36) of the energy takes the form   Hα = ω |α|2 + 21 which may be generalized to 3D oscillators:       Hα = ωx |αx |2 + 21 + ωy |αy |2 + 21 + ωz |αz |2 + 21

(6.38)

CH006.tex

176

6.3.2

16/3/2011

13: 23

Page 176

COHERENT STATES AND TRANSLATION OPERATORS

Average squared Hamiltonian

Next, consider the corresponding average value of the squared Hamiltonian, that is, H2 α = {α}|H2 |{α} which, by comparing Eq. (5.9), reads

or

2  H2 α = (ω)2 {α}| a† a + 21 |{α}

(6.39)

  H2 α = (ω)2 {α}| a† aa† a + a† a + 41 |{α}

(6.40)

Again, in view of Eqs. (6.1) and (6.2), the average value of the quadruple product of operators over coherent states takes the form {α}|a† aa† a|{α} = {α}|α∗ aa† α|{α} and thus α being a scalar, {α}|a† aa† a|{α} = |α|2 {α}|aa† |{α}

(6.41)

Furthermore, because of the basic commutator (5.5), the right-hand-side matrix element of Eq. (6.41) becomes   {α}|aa† |{α} = {α}| a† a + 1 |{α} and thus using in turn Eqs. (6.1) and (6.2),   {α}|aa† |{α} = |α|2 + 1 {α}|{α} = |α|2 + 1 so that Eq. (6.41) becomes

  {α}|a† aa† a|{α} = |α|2 |α|2 + 1

As a consequence and according to Eqs. (6.40) and (6.37), it yields  2   {α}| a† a + 21 |{α} = |α|2 |α|2 + 1 + |α|2 + 41 so that the average value of the squared Hamiltonian given by Eq. (6.39) becomes   H2 α = (ω)2 |α|4 + 2|α|2 + 41 (6.42) which for 3D oscillators gives     H2 α = (ωx )2 |αx |4 + 2|αx |2 + 41 + (ωy )2 |αy |4 + 2|αy |2 + 41   + (ωz )2 |αz |4 + 2|αz |2 + 41

6.3.3

Energy fluctuations

It is now possible to obtain an expression for the relative energy fluctuation of harmonic oscillators within coherent states. The energy fluctuation Hα is formally given by Hα = H2 α − H2α

CH006.tex

16/3/2011

6.4

13: 23

Page 177

COHERENT STATES AS MINIMIZING HEISENBERG UNCERTAINTY RELATIONS

177

and, comparing Eqs. (6.38) and (6.42), we have     Hα = ω |α|4 + 2|α|2 + 41 − |α|4 + |α|2 + 41 which reduces to Hα = ω|α|

(6.43)

Clearly, the energy fluctuation is not zero because the average value of the Hamiltonian has been performed over a state that is not an eigenstate of this operator. Now, the relative energy fluctuation reads, with the help of Eqs. (6.38) and (6.43),   Hα |α| = Hα |α|2 + 21 which, when |α| becomes important, simplifies to   Hα 1  when |α| >> 1 Hα |α| which, in turn, vanishes when |α| becomes very large:   Hα → 0 when |α| → ∞ Hα This last result, which holds also for 3D oscillators, narrows the behavior of a classical harmonic oscillator for which the energy is always exact, according to classical mechanics.

6.4 COHERENT STATES AS MINIMIZING HEISENBERG UNCERTAINTY RELATIONS Coherent states that present such classical asymptotic behavior also minimize the Heisenberg uncertainty relations, which we shall now prove.

6.4.1

Average values of the first and second moments of Q and P

For this purpose, it is necessary to obtain the mean values of the Q and P operators and of their squares averaged over coherent states. 6.4.1.1 Q and P average values First start from the position operator Q averaged over coherent states: Qα = {α}| Q |{α} which, in view of the expression (5.6) of Q in terms of the ladder operators, becomes

 Qα = {α}|(a† + a)|{α} 2mω

CH006.tex

178

16/3/2011

13: 23

Page 178

COHERENT STATES AND TRANSLATION OPERATORS

and due to the eigenvalue equations, (6.1) and (6.2), transforms to

 {α}|(α∗ + α)|{α} Qα = 2mω Moreover, since α and α∗ are scalars and when coherent states are normalized, this result reduces to

 Qα = (6.44) {α∗ + α} 2mω Observe that, while the Q mean value averaged over Hamiltonian eigenstates is zero, those averaged over the coherent state are not so. Now, Eq. (5.7) allows us to write the P average value over a coherent state, according to

mω Pα = {α}| P |{α} = i {α}|(a† − a)|{α} 2 which, proceeding in the same way as above using Eqs. (6.1) and (6.2), reads

mω ∗ Pα = i (6.45) {α − α} 2 6.4.1.2 Q2 and P2 average values Now, in order to find the dispersion of P and Q within a coherent state, one has first to get the mean values of P2 and Q2 within these states. Then, with the help of Eq. (5.6) it may be written Q2 α = {α}|Q2 |{α} =

 {α}|(a† + a)2 |{α} 2mω

or    {α}| (a† )2 + (a)2 + a† a + aa† |{α} 2mω which, in view of the commutation rule (5.5), transforms to    Q2 α = {α}| a† a† + aa + 2a† a + 1 |{α} 2mω or  Q2 α = [{α}|a† a† |{α} + {α}|aa|{α} + 2{α}|a† a|{α} + {α}|{α}] (6.46) 2mω Next, due to Eq. (6.1) we have Q2 α =

aa |{α} = a α|{α} where aa|{α} = αa |{α} = (α)2 |{α}

(6.47)

the Hermitian conjugate of which is {α}|a† a† = (α∗ )2 {α}|

(6.48)

CH006.tex

16/3/2011

6.4

13: 23

Page 179

COHERENT STATES AS MINIMIZING HEISENBERG UNCERTAINTY RELATIONS

179

Hence, comparing Eqs. (6.47) and (6.48), the average value ( 6.46) becomes Q2 α =

   ∗2 (α ) + (α)2 + 2α∗ α + 1 {α}|{α} 2mω

Finally, rearranging and assuming normalized coherent states, one obtains Q2 α =

 ((α + α∗ )2 + 1) 2mω

(6.49)

Now, the average value P2 α reads, in view of Eq. (5.7), P2 α = {α}|P2 |{α} = −

mω {α}|(a† − a)2 |{α} 2

which transforms after expanding the right-hand-side term into P2 α = −

mω {α}|((a† )2 + (a)2 − 2a† a−1)|{α} 2

so that, proceeding in the same way as for Q2 , we have P2 α = −

6.4.2

mω ∗ ((α − α)2 − 1) 2

(6.50)

Heisenberg uncertainty relations

It is now possible to get the dispersion over Q Qα = Q2 α − Q2α which, comparing Eqs. (6.44) and (6.49), reads

 Qα = 2mω Now, the dispersion over P is Pα =

(6.51)

P2 α − P2α

which, owing to Eqs. (6.45) and (6.50), becomes

mω Pα = 2

(6.52)

Thus, the product of the uncertainties (6.51) and (6.52) yields, for arbitrary α, Qα Pα =

 2

(6.53)

Thus, this 1D uncertainty relation, which may be generalized to three dimensions, is the minimum compatible with the Heisenberg uncertainty relations.

CH006.tex

16/3/2011

13: 23

Page 180

180

COHERENT STATES AND TRANSLATION OPERATORS

6.5

DYNAMICS

Owing to the semiclassical properties of the mean values of operators averaged over coherent states, it would be interesting to find if the dynamics of these mean values behave also semiclassically. Since the time dependence of Boson operators is known, it is convenient to perform the dynamic investigations dealing with coherent states within the time-dependent Heisenberg representation instead of the time-dependent Schrödinger one.

6.5.1

Position and momentum time-dependent average values

6.5.1.1 One-dimensional oscillators First, consider the Heisenberg time dependence of the mean value of Q(t) average over coherent states, which, owing to Eq. (5.153), reads

   {α}|Q(t)|{α} = (6.54) {α}|a† (t)|{α} + {α}|a(t)|{α} 2mω Now, in view of Eqs. (5.151) and (5.152), we have {α}|a(t)|{α} = {α}|a(0)|{α}e−iωt {α}|a† (t)|{α} = {α}|a† (0)|{α}eiωt so that due to (6.1) {α}|a(t)|{α} = αe−iωt {α}|{α} = α(t) {α}|a† (t)|{α} = αeiωt {α}|{α} = α∗ (t) with α(t) = αe−iωt As a consequence, Eq. (6.54) becomes {α}| Q(t)|{α} =

(6.55)

 (α∗ eiωt + αe−iωt ) 2mω

which, if α is real, reduces to

{α}|Q(t)|{α} = 2α

 cos ωt 2mω

(6.56)

6.5.1.2 Two-dimensional oscillators Now pass from 1D to 2D oscillators for which the time-dependent Q(t) operator reads Q(t) = Qx (t) + Qy (t) Moreover, defining the coherent states dealing with the x and y components using

 (a† (t) + ak (t)) ak (t)|{α}k  = αk (t)|{α}k  with Qk (t) = 2mω k

CH006.tex

16/3/2011

13: 23

Page 181

6.5

DYNAMICS

181

where k stands for x or y, the mean values of Qx (t) and Qy (t) averaged on their corresponding coherent states must be given by expressions similar to that of (6.54), that is,

 ({α}k |ak† (t)|{α}k  + {α}k |ak (t)|{α}k ) {α}k |Qk (t)|{α}k  = 2mω so that, using the eigenvalue equation defining the x and y coherent states, it is obtained

   {α}k |α∗k (t)|{α}k  + {α}k |αk (t)|{α}k  (6.57) {α}k |Qk (t)|{α}k  = 2mω Next, if, for some reason, dephasing between αx (t) and αy (t) exists so that αx (t) = αe−iωt

αy (t) = αe±iπ/2 e−iωt

and

the two average values (6.57) read, respectively,

 cos (ωt) 2mω

(6.59)

  π cos ωt ∓ 2mω 2

(6.60)

{α}x |Qx (t)|{α}x  = 2α

{α}y |Q± y (t)|{α}y 

= 2α

(6.58)

Hence, for 2D oscillators obeying Eq. (6.58), the mean value of Q averaged over coherent states, yields {α}y |{α}x |Q± (t)|{α}x |{α}y  = {α}y |{α}x |Qx (t)|{α}x |{α}y  +{α}x |{α}y |Q± y (t)|{α}y |{α}x  or, due to Eqs. (6.59) and (6.60) and after simplification,

 ± (cos (ωt) ∓ sin (ωt)) {α}y |{α}x |Q (t)|{α}x |{α}y  = 2α 2mω

(6.61)

Hence, the two ± equations (6.61) constitute two inverse polarized circular motions. Next, using Eq. (5.159), for the averaged momentum, which may be in correspondence with the average value of Q(t) given by Eq. (6.56), one would obtain

{α}y |{α}x |P(t)|{α}x |{α}y  = −2α

6.5.2

mω sin ωt 2

(6.62)

Kinetic and potential time-dependent average values

Now, using Eqs. (6.56) and (6.62), giving the average values of Q(t) and P(t), it is possible to get the time dependence of the average potential and kinetic energy operators V(t) and T(t). For the first one, which is {α}|V(t)|{α} = 21 mω2 {α}|Q(t)2 |{α}

(6.63)

CH006.tex

182

16/3/2011

13: 23

Page 182

COHERENT STATES AND TRANSLATION OPERATORS

the right-hand-side average value may be expressed in terms of the raising and lowering operators from Eq. (5.153), that is, 2   {α}|Q(t)2 |{α} = {α}| a† eiωt + ae−iωt |{α} (6.64) 2mω which using Eq. (5.166) and the commutation rule (5.5), Eq. (6.64) reads  2   {α}|Q(t)2 |{α} = {α}| a† e2iωt + (a)2 e−2iωt + 2a† a+1 |{α} 2mω so that Eq. (6.63) becomes  2 {α}|V(t)|{α} = 41 ω{α}|( a† e2iωt + (a)2 e−2iωt + 2a† a+1)|{α} Again, owing to Eq. (6.1) and to the Hermitian conjugate, one obtains, respectively, after a double action of a on the right and of a† on the left, {α}| (a)2 |{α} = α2 {α}|{α} = α2

(6.65)

 2 {α}| a† |{α} = α∗2 {α}|{α} = α∗2 Besides, owing to Eq. (6.2) and its Hermitian conjugate, one gets  2 {α}| a† a |{α} = |α|2 {α}|{α} = |α|2

(6.66)

so that, by the aid of these last three equations, the average value of the potential energy becomes {α}|V(t)|{α} = 41 ω(α∗2 e2iωt + α2 e−2iωt + 2|α|2 + 1) which, passing to sine and cosine functions, transforms to {α}|V(t)|{α} = 41 ω((α2 + α∗2 ) cos 2ωt − i(α2 − α∗2 ) sin 2ωt + 2|α|2 + 1) (6.67) which, in turn, if α is real, reduces to

  {α}|V(t)|{α} = 21 ω α2 (cos 2ωt + 1) + 21

(6.68)

On the other hand, the corresponding average value of the kinetic energy, which reads 1 {α}|P(t)2 |{α} (6.69) 2m may be found, proceeding in a similar way as for the potential energy by the aid of Eqs. (5.159) and (5.166): {α}|T(t)|{α} =

{α}|T(t)|{α} = 41 ω{α}|((2a† a+1) − ((a† )2 e2iωt + (a)2 e−2iωt ))|{α} Thus, in view of Eqs. (6.65) and (6.66), we have   {α}|T(t)|{α} = 21 ω α2 (1 − cos 2ωt) + 21

(6.70)

Observe that, owing to Eqs. (6.67) and (6.70), the average value of the Hamiltonian is a constant given by {α}|H|{α} = {α}|T(t)|{α} + {α}|V(t)|{α} = ω(α2 + 21 ) that is, as required, in agreement with Eq. (6.38). Observe also the difference in the behavior of time dependence of the average values of the kinetic and potential

CH006.tex

16/3/2011

13: 23

Page 183

6.6 TRANSLATION OPERATORS

183

operators when passing from the eigenstates of the Hamiltonian to the coherent states. Whereas they are constant when the quantum average is performed on the Hamiltonian eigenstates, they become time dependent when passing to coherent states, although coming back and forth in such a way as the average energy remains constant.

6.6 TRANSLATION OPERATORS The coherent states are deeply connected to the translation operators. As we shall see in this section, the translation operators generate coherent states.

6.6.1

Action of translation operators on ladder operators

To prove this, first seek the action of the translation operator   iQ◦ P A(Q◦ ) = exp − (6.71)  on the raising and lowering operators. For this, consider the unitary operator given by Eq. (2.95) where P is the momentum operator and Q◦ a scalar having the dimension of a length. According to Eq. (2.102) the following canonical transformation holds: A(Q◦ )−1 Q A(Q◦ ) = Q + Q◦

(6.72)

Now, when, expressed in terms of the raising and lowering operators using Eq. (5.7), the translation operator takes the form  

Q◦ mω † ◦ ◦ A(Q ) = A(α ) = exp −i i (a − a) (6.73)  2 or A(α◦ ) = eα

◦ a† −α◦ a

with

= e−α

◦ a+α◦ a†

(6.74)

mω (6.75) 2 The second right-hand-side expression in (6.74) has been written to underline the fact that the order of the operators involved in the exponential is irrelevant. Besides, observe that if α◦ is changed into −α◦ into Eq. (6.74), this equation transforms to α◦ = Q◦

A(−α◦ ) = e−α

◦ a† +α◦ a

= eα

◦ a−α◦ a†

(6.76)

the right-hand side, which is simply the inverse of the translation operator A(α◦ ) given by Eq. (6.74), so that A(−α◦ ) = A(α◦ )−1

(6.77)

Now, use Glauber’s theorem (1.78) in order to transform the translation operator (6.74) and its inverse (6.76), into products of exponential operators involving only a† or a, according to A(α◦ ) = (eα

◦ a†

A(α◦ )−1 = (e−α

◦

)(e−α a )e[α

◦ a†

◦

◦ a† ,α◦ a]/2

)(eα a )e−[α

◦

◦ a†

= (e−α a )(eα

◦ a† ,α◦ a]/2

◦

= (eα a )(e−α

)e[α

◦ a†

◦ a,α◦ a† ]/2

)e−[α

◦ a,α◦ a† ]/2

CH006.tex

16/3/2011

184

13: 23

Page 184

COHERENT STATES AND TRANSLATION OPERATORS

or, since [a† , a] = −1, ◦ a†

A(α◦ ) = (eα

A(α◦ )−1 = (e−α

◦

)(e−α a )e−α

◦ a†

◦

◦2 /2

◦2 /2

)(eα a )eα

◦

= (e−α a )(eα ◦

= (eα a )(e−α

◦ a†

◦ a†

)eα

◦2 /2

)e−α

(6.78)

◦2 /2

(6.79)

Next, return to the situation where α is real and thus equal to α◦ . Then, owing to Eqs. (6.75), (6.74) and (6.77), Eq. (6.72) reads

  −(α◦ a† −α◦ a) † (α◦ a† −α◦ a) )(a + a)(e )= (e ((a† + a) + (α◦ + α◦ )) (6.80) 2mω 2mω or, after simplification and use of Eq. (6.74), A(α◦ )−1 (a† + a) A(α◦ ) = (a† + α◦ ) + (a + α◦ )

(6.81)

which, owing to the first expression of (6.78) and (6.79), appears after simplification to be given by ◦

(eα a )(e−α

◦ a†

)(a† + a)(eα

◦ a†

◦

)(e−α a ) = a† + α◦ + a + α◦

(6.82)

However, the right-hand side of the latter equation may be expressed as ◦

(eα a )(e−α

◦ a†

)(a† + a)(eα

◦ a†

◦

◦

)(e−α a ) = (eα a )(e−α + (e−α

◦ a†

◦ a†

◦ a†

a † eα ◦

◦

)(e−α a ) ◦

)(eα a ae−α a )(eα

◦ a†

)

Hence, after simplifications, because the function of an operator commutes with this operator, Eq. (6.82) yields ◦

(eα a )(e−α

◦ a†

◦ a†

)(a† + a)(eα

◦

◦

◦

)(e−α a ) = (eα a )a† (e−α a ) + (e−α

◦ a†

)a(eα

◦ a†

)

(6.83)

As a consequence, due to Eqs. (6.82) and (6.83), it appears that ◦

◦

(eα a )a† (e−α a ) = a† + α◦ (e−α

◦ a†

)a(eα

◦ a†

) = a + α◦

(6.84) (6.85)

6.6.2 Action of translation operators on Hamiltonian ground states Now, study the action of the translation operators given by Eq. (6.78), on the ground state of the Hamiltonian of the quantum harmonic oscillator, which reads  −|α|2  † ∗ (eαa )(e−α a )|{0} A(α)|{0} = e 2 In order to get the action of the exponential operators on |{0}, expand the exponential operators as  A(α)|{0} =

−|α|2 e 2

 n

   (αa† )n  (−α∗ a)m |{0} n! m! m

(6.86)

CH006.tex

16/3/2011

13: 23

Page 185

6.6 TRANSLATION OPERATORS

Furthermore, due to Eqs. (5.53), that is, √ a|{n} = n|{n − 1}

leading to

a|{0} = 0

185

(6.87)

so that an |{0} = δo n |{0} Eq. (6.86) reduces to 

−|α|2 e 2

A(α)|{0} =

 n

or, keeping in mind Eq. (5.67), that is, (a† )n |{0} = Eq. (6.88) transforms to



A(α)|{0} =

−|α|2 e 2

√ n!|{n}

 n

or, after simplification,



−|α|2 e 2

A(α)|{0} =

 αn † n (a ) |{0} n!

(6.89)

 αn √ n! |{n} n!

 n

(6.88)

 αn |{n} √ n!

Now, since the expansion of coherent states is given by Eq. (6.16), that is,  −|α|2    n  α |{α} = e 2 |{n} √ n! n

(6.90)

(6.91)

Then, by identification of Eqs. (6.90) and (6.91), it follows that A(α)|{0} = |{α}

(6.92)

or, according to Eq. (6.74) eαa

† −α∗ a

|{0} = |{α}

with

a|{α} = α|{α}

and thus, after using Glauber’s theorem (1.79)  −|α|2  † −α∗ a e 2 (eαa )(e )|{0} = |{α}

(6.93)

(6.94)

Moreover, observe that due to the last equation of (6.87), leading to −α∗ a

(e

)|{0} =

n ∞  α∗

n=0

n!

(a)n |{0} = |{0}

Eq. (6.94) simplifies to 

−|α|2 e 2

 †

(eαa )|{0} = |{α}

(6.95)

CH006.tex

16/3/2011

186

13: 23

Page 186

COHERENT STATES AND TRANSLATION OPERATORS

6.6.3

Product of translation operators

Now, consider the two following translation operators where ξ and ζ are c-numbers: A(ξ) = (eξa

† −ξ ∗ a

A(ζ) = (eζa

)

† −ζ ∗ a

)

Their product A(ζ)A(ξ) = (eζa

† −ζ ∗ a

)(eξa

† −ξ ∗ a

)

due to the Glauber theorem (1.78), becomes (eζa

† −ζ ∗ a

)(eξa

† −ξ ∗ a

) = (e(ζa

† −ζ ∗ a)+(ξa† −ξ ∗ a)

)(e[(ξa

† −ξ ∗ a),

(ζa† −ζ ∗ a)]/2

)

(6.96)

Now, since [a, a† ] = 1, the commutator appearing on the last right-hand-side term is [(ξa† − ξ ∗ a), (ζa† − ζ ∗ a)] = −(ξζ ∗ − ξ ∗ ζ) so that Eq. (6.96) yields (eζa

6.7

† −ζ ∗ a

)(eξa

† −ξ ∗ a

) = (e(ξ+ζ)a

† −(ξ ∗ +ζ ∗ )a

)(e−(ξζ

∗ −ξ ∗ ζ)/2

)

(6.97)

COHERENT-STATE WAVEFUNCTIONS

Owing to the quasi-classical behavior of coherent states, it may be of interest to visualize them through their wave mechanics representation, which is the purpose of the present section.

6.7.1

Wavefunctions

According to Eq. (3.43), the wavefunction corresponding to the coherent state is the scalar product α (Q) = {Q}|{α}

(6.98)

α (Q) = {Q}|A(α)|{0}

(6.99)

which, in view of (6.92), reads

with, according to Eq. (6.74), A(α) = eαa

† −α∗ a

(6.100)

Again, in view of Eqs. (5.3) and (5.4), the argument of the translation operator reads

 

mω 1 mω 1 Q − iα P − α∗ Q + iα∗ P αa† − α∗ a = α 2 2mω 2 2mω or, after rearranging,

mω 1 ∗ αa − α a = (α − α ) Q − i(α + α ) P 2 2mω †

∗

∗

(6.101)

CH006.tex

16/3/2011

13: 23

Page 187

6.7

COHERENT-STATE WAVEFUNCTIONS

187

Furthermore, one obtains, respectively, by inversion of Eqs. (6.44) and (6.45)

mω i i ∗ (α − α ) (6.102) = {α}|P|{α} = Pα 2  

∗

(α + α )

1 1 1 = {α}|Q|{α} = Qα 2mω  

(6.103)

Therefore, owing to Eqs. (6.101)–(6.103), the translation operator (6.100) takes the form A(α) = ei(Pα Q−Qα P)/ which by the aid of Glauber’s theorem (1.78) transforms to A(α) = (e−iQα P/ )(eiPα Q/ )(e−ζ )

(6.104)

In the preceding equation, ζ is given by ζ=

1 Pα Qα [Q, P] 22

which due to [Q, P] = i reads i Pα Qα 2 Or, because of Eqs. (6.102) and (6.103) leads by inversion to

 mω ∗ ∗ and Pα = i {α + α} {α − α} Qα = 2mω 2 ζ=

so that we have ζ = − 41 {α∗2 − α2 }

(6.105)

Here, it is possible to get an explicit expression for the coherent state wavefunction (6.99) corresponding to the coherent state, that is, α (Q) = {Q}|{α} = {Q}|A(α)|{0} which due to Eqs. (6.104) and (6.105) takes the form α (Q) = {Q}|(e−iQα P/ )(eiPα Q/ )|{0}(e1/4{α

∗2 −α2 }

)

(6.106)

Now, observe that Eq. (2.119) allows us to write (eiQα P/ )|{Q} = |{Q − Qα } the Hermitian conjugate of this last equation of which is {Q}|(e−iQα P/ ) = {Q − Qα }|

(6.107)

so that the coherent state wavefunction (6.106) takes the form α (Q) = {Q − Qα }|(eiPα Q/ )|{0}(e1/4{α

∗2 −α2 }

)

(6.108)

CH006.tex

188

16/3/2011

13: 23

Page 188

COHERENT STATES AND TRANSLATION OPERATORS

In Eq. (6.108), the exponential operator is acting on the left on an eigenbra of the position operator Q and of all operators functions of Q. Hence, the following eigenvalue equation is verified {Q − Qα }|(eiPα Q/ ) = {Q − Qα }|(eiPα (Q−Qα )/ ) so that Eq. (6.108) transforms to α (Q) = {Q − Qα }|{0}(eiPα (Q−Qα )/ )(e1 /4{α∗2 − α2 })

(6.109)

On the other hand, since the ground-state wavefunction of the quantum harmonic oscillator Hamiltonian is

0 (Q) = {Q}|{0} the scalar product involved in Eq. (6.109) is nothing but the displaced ground-state wavefunction, the origin of which has been displaced by the amount Qα , that is, {Q − Qα }|{0} = 0 (Q − Qα )

(6.110)

Now, keeping in mind that the ground-state wavefunction may be obtained using Eqs. (5.126) and (5.127), leading to  mω 1/4  mω  exp − Q2

0 (Q) = π 2 the translated wavefunction (6.110) becomes   mω 1/4  mω (6.111) exp − (Q − Qα )2

0 (Q − Qα ) = π 2 Finally, in view of Eq. (6.111), and using the expression (6.51) of the uncertainty Q performed over a coherent state, Eq. (6.109) becomes

  mω1/4     Q − Qα 2 exp i Pα (Q − Qα ) exp 41 {α∗2 − α2 } α (Q) = exp − π 2 Qα (6.112) with

 2mω Note that Eq. (6.112) may be shortly written in a narrowing form encountered in wavelet theory, that is, Pα = {α}|P|{α}]

Qα =

and

◦ 2

α (Q) = Ke−(Q/Q ) eiλQ where K, Q◦ , and λ are constants that may be obtained when passing from Eq. (6.112) to this expression.

6.7.2 Time-dependent coherent-state wavefunctions It has been shown above that the wave mechanics representation of the coherent state is given by Eq. (6.112). Now, we require its time dependence, and for this purpose

CH006.tex

16/3/2011

13: 23

Page 189

6.8

FRANCK–CONDON FACTORS

189

we transform Eq. (6.112) involving the average values of Qα and Pα by taking in place of them the corresponding time-dependent average values Q(t)α and P(t)α , leading us to write        mω 1/4 i Q − Q(t)α 2 α (Q, t) = exp P(t)α (Q − Q(t)α ) exp − π 2 Qα  × (exp1/4{α(t)∗2 − α(t)2 })

(6.113)

where α(t) is now given by Eq. (6.55). Next, in the Schrödinger picture, and due to Eqs. (6.56) and (6.62), the time-dependent average values involved in Eq. (6.113) are given by

2 Q(t)α = {α}|Q(t)|{α} = α cos ωt mω √ P(t)α = {α}|P(t)|{α} = −α 2mω sin ωt so that the coherent-state wavefunction (6.113) reads    mω 1/4 mω 2 α (Q, t) = exp − (Q − α mω cos ωt)2 π 2       2 i √ α 2 × exp − α 2mω sin ωtα Q − α mω cos ωt exp i sin 2ωt  2 (6.114) where α stands for the eigenvalue of the coherent state at initial time. Figure 6.1 reports the time dependence of the corresponding modulus | α (Q, t)|2 :    mω 1/2 mω 2 | α (Q, t)|2 = (6.115) exp − cos ωt)2 (Q − α mω π  Inspection of this figure shows that the coherent state initially localized on the right-hand side of the equilibrium position moves back and forth around this position without spreading.

6.8

FRANCK–CONDON FACTORS

One has sometimes to compute the overlap integral (Franck–Condon factors) between the eigenfunctions of two oscillator Hamiltonians, the harmonic potentials of which are displaced, and thus not orthogonal, as illustrated in Fig. 6.2. Franck–Condon factors are met, for instance, in the area of electronic molecular spectroscopy where the subbands of the electronic line shapes correspond to transitions between vibrational states of the ground and first excited electronic states, the latter being displaced. They are also found in theories dealing with IR line shapes of weak H-bonded species. As we have seen above, the energy wavefunctions of the quantum harmonic oscillator are given by the following scalar product:

n (Q) = {Q}|{n}

(6.116)

CH006.tex

190

16/3/2011

13: 23

Page 190

COHERENT STATES AND TRANSLATION OPERATORS

4

4

2

2

V(Q)

|Φ(Q)|2 Q 5

0

5

5

0 1 t— 4ω

t0

4

4

2

2

5

0 2 t— 4ω

5

5

5

0

5

3 t— 4ω

Figure 6.1 Time evolution of the probability density (6.115) of a coherent-state wavefunction, with Q expressed in

 2mω

small units, t small in ω−1 small units, and α = 1,

where |{n} is an eigenket of the number occupation operator. Now, we have shown that the action of the translation operator on an eigenket of the position operator is given by Eq. (2.118), that is, A(Q◦ )|{Q} = |{Q + Q◦ }

(6.117)

Since the translation operator is unitary, so that its inverse is its Hermitian conjugate, the Hermitian conjugate of Eq. (6.117) is {Q}|A(Q◦ )−1 = {Q + Q◦ }|

(6.118)

On the other hand, the wavefunction { m (Q + Q◦ )} displaced by the amount Q◦ with respect to that n (Q) defined by Eq. (6.116) is given by the scalar product { m (Q + Q◦ )} = {Q + Q◦ }|{m} or, in view of Eq. (6.118) { m (Q + Q◦ )} = {Q}|A(Q◦ )−1 |{m}

(6.119)

CH006.tex

16/3/2011

13: 23

Page 191

6.8

FRANCK–CONDON FACTORS

191

Energy

~ |{4}〉 ~ |{3}〉 ~ |{2}〉 ~ |{1}〉 ~ |{0}〉

|{1}〉 |{0}〉 0

Q Q

Figure 6.2

Displaced oscillator wavefunctions generating Franck–Condon factors.

Now, look at the Franck–Condon factors, that is, the following overlap integrals Snm (Q◦ ) =

∞ −∞

{ n (Q)∗ }{ m (Q + Q◦ )}dQ

(6.120)

which, in view of Eqs. (6.116) and (6.119), take the form ◦

∞

Snm (Q ) =

{n}|{Q}{Q}|A(Q◦ )−1 |{m} dQ

−∞

a result that may be simplified using the closure relation involving the eigenstates of the position operator, that is, ∞ |{Q}{Q}| dQ = 1 −∞

Thus Snm (Q◦ ) = {n}|A(Q◦ )−1 |{m}

(6.121)

CH006.tex

192

16/3/2011

13: 23

Page 192

COHERENT STATES AND TRANSLATION OPERATORS

Next, pass to Boson operators for the translation operators appearing in Eq. (6.121). Then, in view of Eq. (6.74), we have Snm (α◦ ) = {n}| (e−α

Snm (α◦ ) ≡ Snm (Q◦ )

◦ (a† −a)

) |{m}

α◦ = Q ◦

with

(6.122) mω 2

(6.123)

In order to calculate the Franck–Condon factors, it is convenient to use for the inverse translation operator appearing in Eq. (6.121), the expression (6.79) leading to ◦2 /2

Snm (α◦ ) = eα

◦

{n}|(eα a )(e−α

◦ a†

) |{m}

(6.124)

or ◦2 /2

Snm (α◦ )= eα

{A}n |{B}m 

(6.125)

with |{B}m  = (e−α

◦ a†

) |{m}

and

◦

{A}n | = {n}|(eα a )

(6.126)

We must now compute the scalar product appearing on the right-hand side of Eq. (6.125), and then find in a first place the expression of the ket defined by Eq. (6.126). To obtain it, first expand the exponential appearing on the right-hand side of Eq. (6.126), according to

  (−1)k α◦k (a† )k |{B}m  = |{m} (6.127) k! k

which, due to Eq. (6.89), is  |{m} =

(a† )m √ m!

 |{0}

(6.128)

Hence, Eq. (6.127) transforms to

  (−1)k α◦k (a† )k+m |{0} |{B}m  = √ k! m! k

Again, using Eq. (6.128), we have (a† )k+m |{0} =

(6.129)

 (k + m)!|{k + m}

so that Eq. (6.129) becomes

 √  (−1)k α◦k (k + m)! |{k + m} |{B}m  = √ k! m! k

(6.130)

Now, save that the minus sign is changed into a plus sign, the bra appearing on the right-hand side of Eq. (6.125) is the Hermitian conjugate of Eq. (6.126), so that it is

CH006.tex

16/3/2011

13: 23

Page 193

6.9

DRIVEN HARMONIC OSCILLATORS

193

given by an expression similar to that of (6.130), except for the presence of the power of (−1). Hence, α◦ being real, this bra appears to be

√   α◦l (l + n)! {A}n | = {l + n}| (6.131) √ l! n! l Thus, as a consequence of Eqs. (6.130) and (6.131), the Franck–Condon factors (6.125) take the form

 √ √   (−1)k α◦k+l (l + n)! (k + m)! ◦ α◦2 /2 Snm (α )= e {l + n}|{k + m} √ √ l! n!k! m! k l (6.132) Finally, due to {l + n}|{k + m} = δl,k+m−n Eq. (6.132) reduces to α◦2 /2

Snm (α◦ )= e

 k=n−m



(−1)k α◦2k+m−n (k + m)! √ √ (k + m − n)! n!k! m!

 with

n≥m

(6.133)

with a similar expression for the situation where m > n in which m is changed into n and vice versa.

6.9

DRIVEN HARMONIC OSCILLATORS

Using the work in the present chapter, it is now possible to find the energy levels of driven harmonic oscillators, the Hamiltonian of which is   2 P 1 2 2 HDr = + Mω Q + bQ 2M 2 Passing to Boson operators by the aid of Eqs. (5.6), (5.7), and (5.9), this Hamiltonian becomes   HDr = ω a† a + 21 + α◦ ω(a† + a) (6.134) with α◦ =

b ω

1 2Mω

6.9.1 Diagonalization of driven Hamiltonians by aid of translation operators 6.9.1.1 Canonical transformations involving translation operators In order to diagonalize the Hamiltonian operator (6.134), consider the matrix elements of this operator in the basis of the eigenstates of the quantum harmonic oscillator:   {n}|HDr |{m} = {n}|{ a† a + 21 + α◦ (a† + a)}|{m}ω (6.135)

CH006.tex

194

16/3/2011

13: 23

Page 194

COHERENT STATES AND TRANSLATION OPERATORS

Next, insert the unity operator built up from the translation operator through 1 = A(α◦ )−1 A(α◦ ) with A(α◦ ) = eα

◦ a† −α◦ a

(6.136)

in such a way as to write {n}|A(α◦ )−1 A(α◦ )HDr A(α◦ )−1 A(α◦ )|{m} = {n}|A(α◦ )−1 A(α◦ )    × a† a + 21 + α◦ (a† + a) A(α◦ )−1 A(α◦ )|{m}ω

(6.137)

Now, observe that the action of the translation operator transforms the eigenstates of the harmonic Hamiltonian into new displaced ones according to A(α◦ )|{n} = |{˜n}

(6.138)

{n}|A(α◦ )−1 = {˜n}|

(6.139)

In order to get the expression of the real oscillator wavefunction corresponding to the transformed ket (6.138) observe that, due to Eq. (6.116), the wavefunction corresponding to the states |{n} is given by

n (Q) = {Q}|{n} = {n}|{Q} whereas the wavefunction corresponding to the bra {˜n}| appearing in Eq. (6.139), and resulting from the action of the translation operator (involving a real α◦ ), is

n˜ (Q) = {Q}|{˜n} or, due to Eq. (6.139),

n˜ (Q) = {Q}|A(α◦ )−1 |{n} and thus, owing to Eqs. (6.75) and (6.77),

n˜ (Q) = {Q}|A(−Q◦ )|{n} with ◦

α =Q

◦

mω 2

Next, due to Eq. (6.117) leading to A(−Q◦ )|{Q} = |{Q − Q◦ } Eq. (6.140) reads

n˜ (Q) = {Q − Q◦ }|{n} or

n˜ (Q) = n (Q − Q◦ )

(6.140)

CH006.tex

16/3/2011

13: 23

Page 195

6.9

DRIVEN HARMONIC OSCILLATORS

195

with

n (Q − Q◦ ) = {Q − Q◦ }|{n} In a similar way, one would obtain

m˜ (Q) = m (Q − Q◦ ) = {Q − Q◦ }|{m} Now, in the context of the transformed states (6.138) and (6.139) corresponding to the wavefunction n (Q − Q◦ ), which is displaced by the amount −Q◦ , let us introduce the following transformed Hamiltonian: ˜ Dr = A(α◦ )HDr A(α◦ )−1 H

(6.141)

6.9.1.2 Hamiltonian diagonalization by the canonical transformation Then, owing to Eqs. (6.137) and (6.141), the matrix elements of the transformed Hamiltonian ˜ Dr take the form H ˜ Dr |{m} ˜ = 21 ω + {˜n}|A(α◦ )a† a A(α◦ )−1 |{m}ω ˜ {˜n}|H + α◦ {˜n}|A(α◦ )(a† + a)A(α◦ )−1 |{m}ω ˜

(6.142)

Moreover, observe that, according to Eq. (6.81), one has A(α◦ )a† A(α◦ )−1 = a† − α◦

(6.143)

A(α◦ )aA(α◦ )−1 = a − α◦

(6.144)

Now, in order to get the result of the canonical transformation on a† a appearing in Eq. (6.142), insert between a† and a the unity operator built up from the unitary translation operator, as follows: A(α◦ )a† aA(α◦ )−1 = A(α◦ )a† A(α◦ )−1 A(α◦ ) a A(α◦ )−1 Then, in view of Eqs. (6.143) and (6.144), we have A(α◦ )a† aA(α◦ )−1 = (a† − α◦ )(a − α◦ ) Hence, owing to this result and to Eqs. (6.143) and (6.144), the sum of the transformed operators appearing on the right-hand side of Eq. (6.142) yields A(α◦ )a† a A(α◦ )−1 + A(α◦ )(a† + a)A(α◦ )−1 = a† a − α◦ (a† + a) + α◦2 + α◦ (a† + a) − 2α◦2 or, after simplification A(α◦ )(a† a + α◦ (a† + a))A(α◦ )−1 = a† a − α◦2 Therefore, according to these results and to Eqs. (6.138) and (6.139), Eq. (6.142) reduces to   ˜ Dr |{m} {˜n}|H ˜ ˜ = {˜n}| a† a+ 21 − α◦2 |{m}ω or, due to Eq. (5.40), ˜ Dr |{m} {˜n}|H ˜ =

   n˜ + 21 − α◦2 ωδm˜ ˜n

(6.145)

CH006.tex

16/3/2011

196

13: 23

Page 196

COHERENT STATES AND TRANSLATION OPERATORS

˜ Dr is diagonal in the Therefore, since, according to Eq. (6.145), the Hamiltonian H basis {|{˜n}} obtained from that {|{n}} through the canonical transformation (6.138), it appears that the following eigenvalue equation has been solved: ˜ Dr |{˜n} = En˜ |{˜n} H with the eigenvalues En˜ =

6.9.2

   n˜ + 21 − α◦2 ω

Diagonalization of the Hamiltonian matrix representation

Besides the above canonical diagonalization of the driven harmonic oscillator Hamiltonian (6.134), it is also possible to diagonalize the matrix representation (6.135) of this Hamiltonian. 6.9.2.1 Matrix elements of the driven Hamiltonian in the basis of the harmonic Hamiltonian Consider Eq. (6.135):   {n}|HDr |{m} = {n}| a† a + α◦ (a† + a) + 21 |{m}ω Next, in view of Eq. (5.40), we have

  {n}|HDr |{ m} = α◦ {n}|(a† + a)|{m}ω + m + 21 {n}|{m}ω

(6.146)

Now, keeping in mind Eq. (5.40), allowing one to write √ √ and {n}|a† = n{n − 1}| a|{m} = m|{m − 1} the matrix elements (6.146) read √ √ {n}|HDr |{m} = α◦ ( n {n − 1}|{m}ω + m {n}|{m − 1}) +(m + 21 ){n}|{m}ω

(6.147)

which are zero because of the orthonormality properties of the eigenstates of the quantum harmonic Hamiltonian, except the following cases:   (6.148) {n}|HDr |{n} = ω n + 21 √ √ {n}|HDr |{n − 1} = α◦ ω( n + n − 1)

(6.149)

with, since the matrix is Hermitian, {n − 1}|HDr |{n} = {n}|HDr |{n − 1}

(6.150)

6.9.2.2 Truncation and diagonalization of the matrix representation The matrix elements involved in the matrix representation of the Hamiltonian (6.134) may be computed using Eqs. (6.148)–(6.150). This Hamiltonian matrix may be built up by starting from the ground-state |{0} and then increasing progressively the quantum number n associated with the kets |{n} and with the bras {n}|. Now, since the kets and bras appearing in Eq. (6.147) belong to a basis that is infinite, the matrix representation must also be infinite. Thus, in order to be numerically diagonalized, the Hamiltonian matrix (6.135) of the Hamiltonian (6.134) must be truncated beyond some value n◦ of

CH006.tex

16/3/2011

13: 23

Page 197

6.10

CONCLUSION

197

12 10

Ek(n)/( ω)

8

Exact energy E7 E6 E5 E4 E3 E2 E1 E0

6 4 2 0

4

6

8 10 12 Number of basis states n

Figure 6.3 Stabilization of the energy of the eight lowest eigenvalues Ek (n◦ )/ω◦ with respect to n◦ . (See color insert.)

the quantum number n, leading, therefore, to a finite square (n◦ + 1) × (n◦ + 1) matrix involving the parameters ω and α◦ . The diagonalization of this truncated matrix leads to approximate solutions Ek (n◦ ) of the exact eigenvalue equation H| k (n◦ ) = Ek | k (n◦ ) Figure 6.3 shows the dependence of the eight lowest eigenvalues Ek (n◦ ) on n◦ when α◦ = 1. Inspection of the figure shows that when n◦ is progressively increased, the lowest eigenvalues Ek (n◦ ) decrease progressively and then stabilize toward their exact values obtained by the aid of Eq. (6.145). Such a result manifests the ability to satisfactorily obtain the eigenvalues of a Hamiltonian by diagonalizing its truncated matrix representations by increasing progressively its dimensions until energy level stabilization occurs. That will be later applied to get the energy levels of anharmonic oscillators for which the direct diagonalization of the Hamiltonian is very difficult or impossible to perform. Now, it may be of interest to observe that, as required from the variation theorem (2.25), the energy of the ground state is lowered when improving the accuracy of the corresponding eigenfunction by increasing the dimension of the truncated basis.

6.10

CONCLUSION

In this chapter, devoted to coherent states assumed to be eigenstates of the lowering operator, the following results have been obtained: (i) the expansion of the coherent states over the eigenstates of the harmonic oscillator Hamiltonian, (ii) the fact that they minimize the Heisenberg uncertainty relations, (iii) the fact that they may

CH006.tex

198

16/3/2011

13: 23

Page 198

COHERENT STATES AND TRANSLATION OPERATORS

be generated by the action of the translation operator on these eigenkets, and (iv) their wave mechanics representation. In addition, using the translation operator, it has been possible to get the overlap (Franck–Condon factors) between two mutually translated eigenstates of the harmonic Hamiltonian. Moreover, it has been shown how to diagonalize the Hamiltonian of a driven harmonic oscillator by aid of a canonical transformation involving translation operators. Finally, this result allows one to verify the accuracy of the energy levels obtained by diagonalizations of truncated matrix representations of the driven harmonic oscillator Hamiltonian, opening therefore a possibility to obtain numerically the energy levels of anharmonic oscillators for which no analytical expression is available. The most important results of this chapter are listed as follows: Definition of coherent states a|{α} = α|{α}

and {α}|a† = α∗ {α}|

Coherent-state expansion in terms of the a† a eigenkets:   αn  2 |{α} = e−|α| /2 |{n} √ n! n Scalar product between two coherent states: {β}|{α} = e−|α−β|

2 /2

Closure relations over coherent states: +∞ +∞ |{α}{α}|d Re(α)d Im(α) = 1 −∞ −∞

Translation operators: A(α◦ ) = eα

◦ a† −α◦ a

= e−α

◦ a+α◦ a†

Generation of coherent states by action of the translation operator: |{α} = A(α◦ )|{0}

BIBLIOGRAPHY P. Carruthers and M. Nieto. Am. J. Phys., 33 (1965): 537. C. Cohen-Tannoudji, B. Diu and F. Laloe, Quantum Mechanics. Wiley-Interscience: Hoboken, NJ, 2006. P. A. M. Dirac. The Principles of Quantum Mechanics, 4th ed. Oxford University Press: Oxford, 1982. S. Koide. Z. Naturforschg., 15a (1960): 123–128. W. H. Louisell. Quantum Statistical Properties of Radiation. Wiley: New York, 1973.

CH007.tex

10/3/2011

CHAPTER

22: 25

Page 199

7

BOSON OPERATOR THEOREMS INTRODUCTION In Chapter 5, some important properties of the ladder operators were found, particularly their action on the eigenkets of the number occupation operator. However, many other theorems dealing with the Boson operators, which are also important when working for not only a single quantum harmonic oscillator but in the context of anharmonic oscillators or sets of coupled harmonic oscillators, exist. The aim of the present chapter is to treat these theorems. This chapter deals with by canonical transformations involving the ladder operators. Then, we consider the normal and antinormal ordering formalism that allows one to pass from equations dealing with noncommuting ladder operators to equations involving only scalars, which are easier to solve and then, after solution, to return to the operator equations that are themselves the solutions of the starting operator equations. A final section illustrates this formalism by applying the procedure to the calculation of time evolution operators of driven quantum harmonic oscillators.

7.1

CANONICAL TRANSFORMATIONS

Here, we shall prove theorems dealing with different canonical transformations on functions of Boson operators, involving operators that are also functions of these Boson operators.

7.1.1 Transformations involving translation operators Start from the Baker–Campbell–Hausdorff relation given by Eq. (1.77): (eξA ){f(B)}(e−ξA ) = {f(eξA Be−ξA )}

(7.1)

where ξ is a c-number, whereas f, A, and B are linear operators. Now, apply this relation to the situation where A is the Boson operator a and where f(B) is a function of both a and its Hermitian conjugate a† , that is, A=a

and

f(B) = f(a, a† )

Then, Eq. (7.1) takes the form (eξa ){f(a, a† )}(e−ξa ) = {f((eξa ae−ξa ), (eξa a† e−ξa ))}

(7.2)

Quantum Oscillators, First Edition. Olivier Henri-Rousseau and Paul Blaise. © 2011 John Wiley & Sons, Inc. Published 2011 by John Wiley & Sons, Inc.

199

CH007.tex

10/3/2011

200

22: 25

Page 200

BOSON OPERATOR THEOREMS

Now, since any function of a single operator commutes with this operator, the following relation is verified: (eξa ) a (e−ξa ) = a

(7.3)

(eξa ) a† (e−ξa ) = a† + ξ

(7.4)

whereas, due to Eq. (6.84)

Thus, owing to Eqs. (7.3) and (7.4), Eq. (7.2) transforms to (eξa ){f(a, a† )}(e−ξa ) = {f(a, a† + ξ)}

(7.5)

Now, apply Eq. (7.1) to another special situation by changing ξ into A = a†

− ξ∗

and taking

{f(B)} = {f(a, a† )}

and

Then, Eq. (7.1) reads (e−ξ

∗ a†

){f(a, a† )}(eξ

∗ a†

) = {f(e−ξ

∗ a†

aeξ

∗ a†

, e−ξ

∗ a†

a † eξ

∗ a†

)}

(7.6)

Of course, for the same reasons as those used to obtain Eq. (7.3), one has (e−ξ

∗ a†

) a† (eξ

∗ a†

) = a†

Besides, according to Eq. (6.85), we have (e−ξ

∗ a†

) a(eξ

∗ a†

) = a + ξ∗

Thus, the canonical transformation (7.6) reads (e−ξ

∗ a†

){f(a, a† )}(eξ

∗ a†

) = {f(a + ξ ∗ , a† )}

(7.7)

Next, consider the general transformation (e−ξ

∗ a†

)(eξa ){f(a, a† )}(e−ξa )(eξ

∗ a†

) ≡ (e−ξ

∗ a†

){(eξa ){f(a, a† )}(e−ξa )}(eξ

∗ a†

)

which, due to Eq. (7.5), reads (e−ξ

∗ a†

)(eξa ){f(a, a† )}(e−ξa )(eξ

∗ a†

) = (e−ξ

∗ a†

){f(a, a† + ξ)}(eξ

∗ a†

)

and owing to Eq. (7.7) transforms to (e−ξ

∗ a†

)(eξa ){f(a, a† )}(e−ξa )(eξ

∗ a†

) = {f(a + ξ ∗ , a† + ξ)}

(7.8)

Hence, using the Glauber theorem (1.79), we have (e[a

† ,a]|ξ|2 /2

)(e−ξ

∗ a† +ξa

){f(a, a† )}(eξ

∗ a† −ξa

)(e−[a

† ,a]|ξ|2 /2

) = {f(a + ξ ∗ , a† + ξ)}

or, after simplification, (e−ξ

∗ a† +ξa

){f(a, a† )}(eξ

∗ a† −ξa

) = {f(a + ξ ∗ , a† + ξ)}

so that, noting the definition of the translation operator (6.74), A(ξ)−1 {f(a, a† )}A(ξ) = {f(a + ξ ∗ , a† + ξ)}

(7.9)

with A(ξ) = (eξ

∗ a† −ξa

)

(7.10)

CH007.tex

10/3/2011

22: 25

Page 201

7.1

CANONICAL TRANSFORMATIONS

201

7.1.2 Transformations involving number occupation operator exponentials 7.1.2.1 Transformations of the ladder operators canonical transformation in which ξ is a c-number:

Now, consider the following

g(ξ) = (eξa a )a(e−ξa a )

(7.11)

g(0) = a

(7.12)

†

†

which for ξ = 0 reads

The derivative of (7.11) with respect to ξ yields       dg(ξ) d ξa† a d −ξa† a −ξa† a ξa† a = a(e ) + (e )a e e dξ dξ dξ or   dg(ξ) † † † † = (a† a eξa a ) a (e−ξa a ) − (eξa a ) a (a† a e−ξa a ) dξ

(7.13)

Again, since a† a commutes with all functions of the product a† a, the first right-handside term of Eq. (7.13) becomes (a† a eξa a )a(e−ξa a ) = (eξa a a† a) a (e−ξa a ) †

†

†

†

so that eq. (7.13) transforms to   dg(ξ) † † = (eξa a )(a† aa−aa† a)(e−ξa a ) dξ or   dg(ξ) † † = (eξa a )[a† , a] a (e−ξa a ) dξ Again, using [a, a† ]= 1, Eq. (7.14) transforms to   dg(ξ) † † = −(eξa a ) a (e−ξa a ) dξ and, in view of Eq. (7.11), into   dg(ξ) = −g(ξ) dξ Next, by derivation of both terms of Eq. (7.15) with respect to ξ, that is,  2    d g(ξ) d g(ξ) =− dξ 2 dξ and, due to Eq. (7.15), it reads  2  d g(ξ) = g(ξ) dξ 2 Again, by recurrence, one obtains  n  d g(ξ) = (−1)n g(ξ) dξ n

(7.14)

(7.15)

(7.16)

CH007.tex

202

10/3/2011

22: 25

Page 202

BOSON OPERATOR THEOREMS

Now, at ξ = 0, and in view of Eq. (7.12), the nth derivative given by Eq. (7.16) reduces to  n  d g(ξ) = (−1)n a (7.17) dξ n ξ=0 Next, write the Taylor expansion of the function (7.11) around ξ = 0, that is,       dg(ξ) 1 d 2 g(ξ) 1 d 3 g(ξ) 2 g(ξ) = g(0) + ξ+ ξ + ξ3 . . . dξ ξ=0 2 dξ 2 ξ=0 3! dξ 3 ξ=0 Hence, comparing Eqs. (7.12) and (7.17), this expansion takes the form   1 2 1 3 g(ξ) = a 1 − ξ + ξ − ξ ..... 2 3! or, passing from the expansion to its corresponding exponential expression, g(ξ) = a(e−ξ ) so that, due to Eq. (7.11), we have (eξa a ) a (e−ξa a ) = a (e−ξ ) †

†

(7.18)

Moreover, by a similar inference as that allowing to pass from Eq. (7.11) to Eq. (7.18), we have (eξa a ) a† (e−ξa a ) = a† (eξ ) †

†

(7.19)

Apply Eqs. (7.18) and (7.19) to reproduce the results of the integration of the Heisenberg equation governing the dynamics of the ladder operators, keeping in mind Eq. (3.88) governing the time dependence of an operator A in the Heisenberg picture, that is, A(t)HP = (eiHt/ )A(e−iHt/ ) which reads for the Boson operator a, and when the Hamiltonian H is that of a harmonic oscillator a(t)HP = (eiωta a ) a (e−iωta a ) †

†

(7.20)

Then, applying Eqs. (7.18) and (7.19) to the situation where ξ = iωt, it yields (eiωta a ) a (e−iωta a ) = a (e−iωt ) †

†

†

†

(eiωta a ) a† (eiωta a ) = a† (eiωt )

(7.21) (7.22)

so that Eq. (7.20) reads a(t)HP = ae−iωt the Hermitian conjugate of which is a† (t)HP = a† eiωt One may verify that these results are equivalent to those of (5.151) and (5.152) obtained by integration of the Heisenberg equation (3.94).

CH007.tex

10/3/2011

22: 25

Page 203

7.1

CANONICAL TRANSFORMATIONS

7.1.2.2 Transformations on functions of the ladder operators the following transformation:

Now, consider

(eξa a ){f(a)}(e−ξa a ) †

203

†

(7.23)

where f(a) is a function of a that may be expanded according to  {f(a)} = {Cn }(a)n

(7.24)

n

where the {Cn } are the scalar coefficients of the expansion. The latter expansion may be transformed using the following unity operator: 1 = (e−ξa a )(eξa a ) †

†

(7.25)

according to {f(a)} =

 {Cn }{a1a · · · 1a}n n

so that Eq. (7.24) reads  † † † † {Cn }{a(e−ξa a )(eξa a )a · · · (e−ξa a )(eξa a )a}n {f(a)} = n

Then, the transformation (7.23) becomes  † † † † † † † † {Cn }{(eξa a ae−ξa a )(eξa a ae−ξa a ) · · · (eξa a ae−ξa a )}n (eξa a ){f(a)}e(−ξa a) = n

Again, using Eq. (7.18), we have (eξa a ){f(a)}(e−ξa a ) = †

†

 {Cn }(ae−ξ )n n

Thus, comparing (7.24), one obtains (eξa a ){f(a)}(e−ξa a ) = {f(ae−ξ )} †

†

(7.26)

In a similar way, one would find (eξa a ){f( a† )}(e−ξa a ) = {f(a† eξ )} †

†

(7.27)

Now, consider a function of both a† and a, defined by the following expansion:  f{(a† , a)} = {Cl,m,...,r,...,s,...,u }{(a† )l (a)m · · · (a)r · · · (a† )s · · · (a† )u } l,m,...,r,...,s,...,u

(7.28) where the {Cl,m,...,r...,s...,u } are the scalar coefficients of the expansion. Then, consider the following transformation over this function: {F(a† , a)} = (eξa a ){f(a† , a)}(e−ξa a ) †

†

(7.29)

CH007.tex

204

10/3/2011

22: 25

Page 204

BOSON OPERATOR THEOREMS

Next, insert in the following way into the right-hand side of Eq. (7.29), the unity operator Eq. (7.25):  {F(a† , a)} = {Cl,m,...,r,...,s,...,u } l,m,...,r,...,s,...,u

× (eξa a )(a† )l (e−ξa a eξa a )(a)m (e−ξa a eξa a ) †

†

†

†

†

· · · (a)r · · · (a† )s · · · (e−ξa a eξa a )(a† )u (e−ξa a ) †

Then, using Eqs. (7.26) and (7.27), we have  {F(a† , a)} =

†

†

{Cl,m,...,r...,s...,u }

l,m,...,r...,s...,u

Hence, comparing Eqs. (7.29) and Eq.(7.28), we have (eξa

†a

){f(a† , a)}(e−ξa a ) = {f(a† eξ , ae−ξ )} †

(7.30)

so that, when ξ = iωt, one obtains (eiωta

†a

){f(a† , a)}(e−iωta a ) = {f(a† eiωt , ae−iωt )} †

(7.31)

7.2 NORMAL AND ANTINORMAL ORDERING FORMALISM We shall now deal with a formalism that allows us to transform an equation involving the noncommuting Boson operators into new scalar ones involving partial derivatives, which may be solved, the obtained solutions being inversely converted into expressions involving the ladder operators, which are the solutions of the above operator equations we want to solve. This formalism concerns what is called, the normal and antinormal ordering.

7.2.1

Normal and antinormal ordering

To introduce this formalism, start from the very simple operator {f(a, a† )} = aa†

(7.32)

which, due to the commutation rule [a, a† ] = 1, and thus to aa† = a† a + 1, reads {f(a, a† )} = a† a + 1 a†

(7.33)

In the latter equation, the raising operator is before the lowering a, at the opposite of the situation given by Eq. (7.32). In case (7.33), the operators a† and a are said to be in normal form, whereas in case (7.32), they are said to be in antinormal form. The following notations are used, respectively, for the two equivalent normal {n} and antinormal {a} expressions: {f {n} (a, a† )} = a† a + 1

and

{f {a} (a, a† )} = aa†

CH007.tex

10/3/2011

22: 25

Page 205

7.2

NORMAL AND ANTINORMAL ORDERING FORMALISM

205

Of course, since the two expressions are equivalent, one must write {f {n} (a, a† )} = {f {a} (a, a† )} Now consider a very general operator f(a, a† ), which is some function of the operators a and a† , susceptible to be expanded according to  {f(a, a† )} = {Cl,m,...,r,...,s,...,u }(a† )l (a)m · · · (a)r · · · (a† )s · · · (a† )u (7.34) l,m,...,r,...,s,...,u

where {Cl,m,...,r,...,s,...,u } are the expansion coefficients. Then, it is possible by systematic aid of aa† = a† a + 1 resulting from the commutation rule [a, a† ] = 1 to write this operator (7.34) either in normal or in antinormal form, according to  {f {n} (a, a† )} = (7.35) { frsn }(a† )r (a)s rs

{f

{a}

 (a, a )} = { fsra }(a)s (a† )r †

(7.36)

sr

Here, frsn and fsra are, respectively, the expansion coefficients of the normal and antinormal form of the operator (7.34). Of course, as above, the three expressions (7.34)– (7.36) being equivalent, one may write {f(a, a† )} = {f {a} (a, a† )} = {f {n} (a, a† })

7.2.2

(7.37)

Normal and antinormal ordering operators

ˆ and A, ˆ the inverses of which are N ˆ −1 and A ˆ −1 . Now, consider two linear operators N ˆ −1 N ˆN ˆ −1 = N ˆ =1 N ˆ =1 ˆ −1 A ˆA ˆ −1 = A A ˆ −1 assume that N

(7.38) (7.39)

ˆ −1 and A

Then, allow one to transform, respectively, the normal and antinormal series expansion (7.35) and (7.36) of Boson operators, to the corresponding series expansion of scalars in which the a and a† operators have been transformed, respectively, into the scalars α and α∗ : ˆ −1 {f {n} (a, a† )} = { f {n} (α, α∗ )} N

(7.40)

ˆ −1 {f {a} (a, a† )} = { f {a} (α, α∗ )} A

(7.41)

with, respectively, { f {n} (α∗ , α)} =

 {{ frsn }(α∗ )r (α)s }

(7.42)

rs

{ f {a} (α∗ , α)} =

 {{ fsra }(α)s (α∗ )r } sr

ˆ and A, ˆ respectively, Premultiply Eqs. (7.40) and (7.41) by N ˆN ˆ −1 {f {n} (a, a† )} = N{ ˆ f {n} (α, α∗ )} N ˆA ˆ −1 {f {a} (a, a† )} = A{ ˆ f {a} (α, α∗ )} A

(7.43)

CH007.tex

206

10/3/2011

22: 25

Page 206

BOSON OPERATOR THEOREMS

Then, using Eqs. (7.38) and (7.39), one obtains ˆ f {n} (α, α∗ )} {f {n} (a, a† )} = N{

and

ˆ f {a} (α, α∗ )} {f {a} (a, a† )} = A{ (7.44)

ˆ and A ˆ transform the scalar functions (7.42) and showing that the linear operators N (7.43) into the corresponding normal and antinormal operators (7.35) and (7.36).

7.2.3 Commutators of Boson operators with functions of Boson operators Now, before continuing the study of normal and antinormal formalism, it is necessary to get some commutators of Boson operators with functions of them, and for this purpose consider the following expression: [(a† )2 , a] = a† a† a − aa† a†

(7.45)

Then, in view of the commutation rule aa† − a† a = 1, the last right-hand side of Eq. (7.45) transforms to (aa† )a† = (a† a + 1)a† so that Eq. (7.45) reads †

[(a† )2 , a] = a† a† a − a† aa − a†

(7.46)

or, factorizing, [(a† )2 , a] = a† (a† a − aa† ) − a† Hence, using in turn the commutation rule of Bosons, leads to [(a† )2 , a] = −2a† Hence, one obtains by derivation

 2a† =

∂(a† )2 ∂a†

(7.47)



so that Eq. (7.47) reads [(a† )2 , a]

 † 2 ∂(a ) =− ∂a†

(7.48)

Next, consider the commutator of a with the third power of its Hermitian conjugate: [(a† )3 , a] = [a† (a† )2 , a]

(7.49)

Again, recall that according to Eq. (1.75), the following relation holds between commutators of operators A, B, and C: [BC, A] = [B, A]C + B[C, A] Then, in order to apply this theorem to Eq. (7.49), take B = a† ,

C = (a† )2

A=a

(7.50)

CH007.tex

10/3/2011

22: 25

Page 207

7.2

NORMAL AND ANTINORMAL ORDERING FORMALISM

207

so that, by application of theorem (7.50), we have [B, A] = [a† , a] = −1 and [C, A] = [(a† )2 , a] which, due to (7.47), yields [C, A] = −2a† Thus, Eq. (7.49), reads [(a† )3 , a] = −(a† )2 − 2(a† )2 = −3(a† )2 The latter result may be also expressed in terms of the derivative of the third power of a† with respect to a† :  † 3 ∂(a ) [(a† )3 , a] = − (7.51) ∂a† Moreover, one obtains by recurrence of Eqs. (7.48) and (7.51)  † n ∂(a ) † n [(a ) , a] = − ∂a†

(7.52)

In order to obtain the Hermitian conjugate of this expression, first write it explicitly according to  † n ∂(a ) (a† )n a − a(a† )n = − ∂a† Next, to get the full Hermitian conjugate of the latter expression, take the Hermitian conjugate of each term and then invert the result, so that   ∂(a)n † n n † a (a) − (a) a = − (7.53) ∂a and thus   ∂(a)n n † [(a) , a ] = (7.54) ∂a Furthermore, consider the following equation involving commutators: [a† , {f(a, a† )}] = [a† , {f {a} (a, a† )}] which holds because of Eq. (7.37) expressing the equivalence between any function of the Boson operators and its antinormal order form. Owing to Eq. (7.36), this commutator takes the form  [a† , {f(a, a† )}] = { frsa }[a† , (a)r (a† )s ] (7.55) rs

which, using Eqs. (7.50) and (7.55), becomes  [a† , {f(a, a† )}] = { frsa }([a† , (a)r ](a† )s + (a)r [a† , (a† )s ]) rs

(7.56)

CH007.tex

208

10/3/2011

22: 25

Page 208

BOSON OPERATOR THEOREMS

Now, we remark that in the latter expression, the second right-hand-side commutator is zero, so that this equation reduces to  { frsa }[a† , (a)r ](a† )s (7.57) [a† , {f(a, a† )}] = rs

Moreover, writing r in place of n, Eq. (7.53), we have [a† , (a)r ] = −r(a)r−1 so that Eq. (7.57) simplifies to [a† , {f(a, a† )}] = −

 { frsa }r(a)r−1 (a† )s

(7.58)

rs

Moreover, observe that, owing to Eq. (7.36) the right-hand side of this last equation is just the partial derivative with respect to a of the antinormal ordered expression of the function of Boson operators. Hence, the commutator (7.58) becomes  {a}  ∂f (a, a† ) [a† , {f(a, a† )}] = − ∂a and, thus, owing to Eq. (7.37),



[a† , {f(a, a† )}] = − In a similar way, one would obtain



[a, {f(a, a )}] = †

7.2.4

∂f(a, a† ) ∂a

∂f(a, a† ) ∂a†

 (7.59)

 (7.60)

Average values over coherent states

Now, consider the following operator written in order form:  { frsn }(a† )r (a)s {f {n} (a, a† )} = rs

Then, according to Eq. (7.35), its average value over a coherent state is  { frsn }{α}|(a† )r (a)s |{α} {α}|{f {n} (a, a† )}|{α} = rs

Next, keeping in mind the definitions (6.1) and (6.2) of coherent states a|{α} = α|{α}

and

{α}|a† = {α}|α∗

and applying them to the above average value, one finds  { frsn }{α}|(α∗ )r (α)s |{α} {α}|{f {n} (a, a† )}|{α} = rs

or {α}|f {n} (a, a† )|{α} =

 { frsn }(α∗ )r (α)s {α}|{α} rs

(7.61)

CH007.tex

10/3/2011

22: 25

Page 209

7.2

209

NORMAL AND ANTINORMAL ORDERING FORMALISM

so that if the coherent state is normalized, {α}|f {n} (a, a† )|{α} =

 { frsn }(α∗ )r (α)s rs

Therefore, owing to Eq. (7.42), we have {α}|f {n} (a, a† )|{α} = f {n} (α∗ , α)

(7.62)

which, due to Eq. (7.37), reads also {α}|f(a, a† )|{α} = f {n} (α∗ , α)

(7.63)

The latter equation shows that the average value of an arbitrary operator function of Boson operators performed on a coherent state is the scalar function defined by Eq. (7.42).

7.2.5

Expression of |{α}{α}|a and of its Hermitian conjugate

Start from the first equation of Eq. (7.61), that is, a|{α} = α|{α}

(7.64)

Then, postmultiply both sides of this equation by the bra {α}|, and one obtains a|{α}{α}| = α|{α}{α}|

(7.65)

Now, the question must be posed: What is the result of the following expression? |{α}{α}|a =? To answer, write the coherent state in terms of the action of the translation operator on the ground state of the Hamiltonian of the harmonic oscillator by the aid of Eq. (6.94), that is, |{α} = (e−|α|

2 /2

∗

)(eαa )(e−α a )|{0} †

(7.66)

the Hermitian conjugate of which is ∗

{α}| = {0}|(e−αa )(eα a )(e−|α| †

2 /2

)

(7.67)

Then, using the two above equations, we have |{α}{α}| = (e−|α|

2 /2

∗

∗

)(eαa )(e−α a )|{0}{0}|(e−αa )(eα a )(e−|α| †

†

2 /2

)

(7.68)

an expression which may be simplified in the following way. First, observe that, by expansion of exp{−α∗ a}, it is possible to write   (α∗ a)3 (α∗ a)2 ∗ (e−α a )|{0} = 1 + α∗ a+ + + · · · |{0} (7.69) 2! 3! Then, due to Eq. (5.35), that is, a|{0} = 0 and thus an |{0} = 0

except if

n=0

CH007.tex

210

10/3/2011

22: 25

Page 210

BOSON OPERATOR THEOREMS

Eq. (7.69) simplifies to ∗

(e−α a )|{0} = |{0}

(7.70)

the Hermitian conjugate of which is {0}|(e−αa ) = {0}| †

(7.71)

As a consequence of Eqs. (7.70) and (7.71), Eq. (7.68) simplifies to |{α}{α}| = (e−|α|

2 /2

∗

)(eαa )|{0}{0}|(eα a )(e−|α| †

2 /2

)

Next, postmultiplying both right- and left-hand-side terms of this last equation by a and after rearranging, using e−|α|

2 /2

e−|α|

2 /2

= e−αα

∗

(7.72)

one obtains ∗

∗

|{α}{α}|a = (e−αα )(eαa )|{0}{0}|(eα a )a †

(7.73)

Moreover, due to the expression of the following partial derivative  ∗  ∂eα a ∗ = (eα a )a ∂α∗ Eq. (7.73) transforms to



−αα∗

|{α}{α}|a = (e

)(e

αa†

∗

∂eα a )|{0}{0}| ∂α∗

 (7.74)

Again, the partial derivative of the exponential operator with respect to α∗ commutes † with the bra, the ket, and the operator eαa , which do not depend on α, thus allowing one to transform Eq. (7.74) into   ∂ ∗ αa† α∗ a |{α}{α}|a = (e−αα ) {(e )|{0}{0}|(e )} (7.75) ∂α∗ Furthermore, denoting ∗

{f(α∗ a, αa† )} = (eαa )|{0}{0}|(eα a ) †

Eq. (7.75) reads −αα∗

|{α}{α}|a = e



∂f(α∗ a, αa† ) ∂α∗

(7.76)

 (7.77)

Now, observe that the following relation is verified:     ∗ ∂e−αα f(α∗ a, αa† ) ∂f(α∗ a, αa† ) −αα∗ ∗ † −αα∗ = −α(e ){f(α a, αa )} + (e ) ∂α∗ ∂α∗ Then, rearranging gives     ∂ ∂f(α∗ a, αa† ) ∗ −αα∗ (e = ) + α {(e−αα ){f(α∗ a, αa† )}} ∗ ∗ ∂α ∂α

(7.78)

CH007.tex

10/3/2011

22: 25

7.2

Page 211

211

NORMAL AND ANTINORMAL ORDERING FORMALISM

As a consequence of Eq. (7.78), Eq. (7.77) transforms with the help of Eq. (7.76) into   ∂ ∗ † ∗ |{α}{α}|a = + α {(e−αα )(eαa )|{0} {0}|(eα a )} ∗ ∂α or, owing to Eq. (7.72)



|{α}{α}|a =

 ∂ 2 † ∗ + α {(e−|α| )(eαa )|{0} {0}|(eα a )} ∗ ∂α

Then after rearranging the exponential involving the scalar |α|2 , we have   ∂ 2 † ∗ 2 |{α}{α}|a = + α {(e−|α| /2 )(eαa )|{0} {0}|(eα a )(e−|α| /2 )} ∗ ∂α Again, in view of Eq. (7.71), it may be written in the more complex form   ∂ 2 † ∗ † ∗ 2 |{α}{α}|a = + α {(e−|α| /2 )(eαa )(e−α a )|{0} {0}|(e−αa )(eα a )(e−|α| /2 )} ∗ ∂α Finally, using Eqs. (7.66) and (7.67), one obtains the result   ∂ |{α}{α}|a = + α {|{α}{α}|} ∂α∗

(7.79)

the Hermitian conjugate of which is  a† |{α}{α}|

=

 ∂ ∗ + α {|{α}{α}|} ∂α

7.2.6 Theorems dealing with normal and antinormal ordering 7.2.6.1 Normal ordering Consider the following normal ordered expansion of any operator function of Boson operators  f(a, a† ) = {Cl,m,...,r,...,s,...,u }(a† )l (a)m · · · (a)r · · · (a† )s · · · (a† )u l,m,...,r,...,s,...,u

 |{α}{α}|a =

 ∂ + α {|{α}{α}|} ∂α∗

(7.80)

Now, consider the average value of this operator over a coherent state allowing one to write via Eqs. (7.62), that is, {α}|f(a, a† )|{α} = { f (α, α∗ )} Moreover, introduce after the operator of the left-hand side of Eq. (7.63), a closure relation over some basis {|k }  f {n} (α, α∗ ) = {α}|f(a, a† )|k k |{α} k

which, on commuting the scalar products with the matrix elements, transforms to  { f {n} (α, α∗ )} = k |{α}{α}|f(a, a† )|k  k

CH007.tex

212

10/3/2011

22: 25

Page 212

BOSON OPERATOR THEOREMS

so that the right-hand side of this last equation may be written formally as the trace over an arbitrary basis {|k } according to { f {n} (α, α∗ )} = tr{|{α}{α}|f(a, a† )} Next, owing to Eq. (7.80), this average value reads { f {n} (α, α∗ )}  =

{Cl,m,...,r,...,s,...,u }tr{|{α}{α}|(a† )l (a)m · · · (a)r · · · (a† )s · · · (a† )u }

l,m,...,r,...,s,...,u

(7.81) Now, the Hermitian conjugate of Eq. (7.65) is |{α}{α}|a† = α∗ |{α}{α}| whereas Eq. (7.79) reads



∂ |{α}{α}|a = α + ∗ ∂α

(7.82)

 |{α}{α}|

(7.83)

so that, by iteration of Eqs. (7.82) and (7.83), we have |{α}{α}|(a† )r = (α∗ )r |{α}{α}|

(7.84)

  ∂ s |{α}{α}|(a)s = α + ∗ |{α}{α}| ∂α

(7.85)

Hence, Eq. (7.81) transforms to  {Cl,m,...,r,...,s,...,u } f {{n} (α, α∗ )} = l,m,...,r,...,s,...,u

      ∂ m ∂ r × tr (α∗ )l α + ∗ · · · α + ∗ · · · (α∗ )s · · · (α∗ )u |{α}{α}| ∂α ∂α

Now, writing explicitly the trace, and using the fact that the bras k | commute with the α and α∗ and the partial derivative with respect to α∗ , that gives   {Cl,m,...,r,...,s,...,u } { f {n} (α, α∗ )} = l,m,...,r,...,s,...,u



∗ l



× (α )

∂ α+ ∗ ∂α

m



k

∂ ··· α + ∗ ∂α

r

∗ u



· · · (α ) · · · (α )

×k |{α}{α}|k  Moreover, since

∗ s

(7.86)

  k |{α}{α}|k  = |k |{α}|2 = 1 k

k

Eq. (7.86) reduces to {f

{n}

∗

(α, α )} = f



∂ α + ∗ , α∗ ∂α

 (7.87)

CH007.tex

10/3/2011

22: 25

Page 213

7.2

with    ∂ ∗ f α + ∗,α = ∂α

NORMAL AND ANTINORMAL ORDERING FORMALISM



{Cl,m,...,r,...,s,...,u }

l,m,...,r,...,s,...,u



∗ l

213



× (α )

∂ α+ ∗ ∂α

m



∂ ··· α + ∗ ∂α

r

∗ s

∗ u



· · · (α ) · · · (α )

ˆ that is, Moreover, premultiply both terms of Eq. (7.87) by the ordering operator N,    ˆ f {n} (α∗ , α)} = N ˆ f α + ∂ α∗ N{ ∂α∗ Then, owing to the first equation of (7.44), one obtains the important result    ∂ {n} † ∗ ˆ {f (a, a )} = N f α + ∗ , α ∂α

(7.88)

7.2.6.2 Theorem dealing with antinormal ordering Consider the following antinormal ordered expansion of a function of Boson operators:  r {f {a} (a, a† )} = (7.89) { fsra }(a)s (a† ) s,r

Then, using the closure relation (6.19) on coherent states, that is, 1 π

+∞ +∞ |{α}{α}|d{Re(α)}d{Im(α)} = 1 −∞ −∞

Eq. (7.89) reads {f

{a}

1 (a, a )} = π

+∞ +∞

†

r

{ fsra }(a)s |{α}{α}| (a† ) d{Re(α)}d{Im(α)}

−∞ −∞ s,r

Again, keeping in mind the properties of the coherent state, that is, (a)|{α} = (α)|{α}

and

{α} |(a† ) = {α}| (α∗ )

and

{α}|(a† )r = {α}|(α∗ )r

and leading by iteration (a)s |{α} = (α)s |{α} the previous equation becomes {f

{a}

1 (a, a )} = π †

+∞ +∞

{ fsra }(α)s (α∗ )r |{α}{α}| d{Re(α)}d{Im(α)}

−∞ −∞ s,r

Finally, due to Eq. (7.43), that is,  { fsra }(α)s (α∗ )r = f {a} (α∗ , α) s,r

(7.90)

CH007.tex

214

10/3/2011

22: 25

Page 214

BOSON OPERATOR THEOREMS

Eq. (7.90) yields {f {a} (a, a† )} =

7.2.7

1 π

+∞ +∞ f {a} (α∗ , α)|{α}{α} |d{Re(α)}d{Im(α)} −∞ −∞

Generalization of theorems dealing with normal ordering

We start from the partial derivative with respect to a† of the normal ordered expansion (7.35), that is,    {n}  ∂f (a, a† ) ∂  n †r s = † { frs }(a ) (a) ∂a† ∂a r,s which reads



∂f {n} (a, a† ) ∂a†

 =

 { frsn }r(a† )r−1 (a)s

(7.91)

r,s

then, owing to the first equation of (7.44), the right-hand side of Eq. (7.91) becomes 

  n † r−1 s n ∗ r−1 s ˆ { frs }r(a ) (a) = N { frs }r(α ) (α) rs

r,s

so that Eq. (7.91) yields

  {n}   ∂f (a, a† ) n ∗ r−1 s ˆ =N { frs }r(α ) (α) ∂a† r,s a result that may also be expressed as 

 {n}  ∂f (a, a† ) ∂  n ∗r s ˆ =N { f }(α ) (α) ∂a† ∂α∗ r,s rs or, owing to Eq. (7.42), as  {n}    ∂f (a, a† ) ∂ n ∗ ˆ = N f (α , α) (7.92) ∂a† ∂α∗ Recall the commutator given by Eq. (7.60), that is,   {n} ∂f (a, a† ) {n} † {a, f (a, a )} = ∂a† allows one to write   {n} ∂f (a, a† ) {af {n} (a, a† )} = {f {n} (a, a† )a} + (7.93) ∂a† Now, observe that the first right-hand-side term of this last equation, which appears to be in normal order, may be viewed as the result of the action of the normal order operator according to {n} ∗ ˆ f {n} (α∗ , α)α} = N{αf ˆ {f {n} (a, a† )a} = N{ (α , α)}

CH007.tex

10/3/2011

22: 25

7.2

Page 215

NORMAL AND ANTINORMAL ORDERING FORMALISM

215

Now, the last right-hand-side term of Eq. (7.93) is given by Eq. (7.92), so that Eq. (7.93) may be written  {n} ∗  ∂f (α , α) {n} † {n} ∗ ˆ ˆ {af (a, a )} = N{αf (α , α)} + N ∂α∗ or    ∂ {n} † {n} ∗ ˆ (7.94) {af (a, a )} = N α + ∗ f (α , α) ∂α By generalization of Eq. (7.94), one now obtains    ∂ m {n} ∗ ˆ α+ ∗ { f (α , α)} (a)m {f {n} (a, a† )} = N ∂α which, due to Eq. (7.37), reads ˆ (a)m {f(a, a† )} = N

 α+

∂ ∂α∗

m

(7.95)

 { f {n} (α∗ , α)}

(7.96)

ˆ −1 , that is, Next, multiplying both the right- and left-hand sides of Eq. (7.96) by N    ∂ m {n} ∗ ˆ −1 (a)m {f(a, a† )} = N ˆ −1 N ˆ N { f (α , α)} α+ ∗ ∂α yields after simplification, with the help of Eq. (7.38)  m ˆ −1 (a)m {f(a, a† )} = α + ∂ N { f {n} (α∗ , α)} ∂α∗

(7.97)

On the other hand, it is clear that the following relation is satisfied since the Boson operator a† is in front of the function of the normal ordered expression f {n} (a, a† ) of the Boson operators: ˆ ∗ )k { f {n} (α∗ , α)}} (a† )k {f {n} (a, a† )} = N{(α

(7.98)

Then, proceeding in the same way as for passing from Eq. (7.96) to Eq. (7.97), one obtains ˆ −1 {(a† )k {f(a, a† )}} = (α∗ )k { f {n} (α∗ , α)} N

(7.99)

Now, consider an operator equation dealing with Boson operators of the form {F(a, a† )} = {(a† )k (a)m f(a, a† )} The question now is what may be in the scalar language an expression of the form ˆ −1 {(a† )k (a)m f(a, a† )} ˆ −1 {F(a, a† )} = N N Owing to Eq. (7.95), it takes the form     ∂ m {n} −1 † −1 † kˆ ∗ ˆ ˆ N {F(a, a )} = N (a ) N α + ∗ { f (α, α )} ∂α

(7.100)

(7.101)

CH007.tex

216

10/3/2011

22: 25

Page 216

BOSON OPERATOR THEOREMS

However, due to Eq. (7.40) and to the equivalence between F(a, a† ) and F{n} (a, a† ), we may write ˆ −1 {F(a, a† )} = N ˆ −1 {F{n} (a, a† )} = {F {n} (α, α∗ )} N it yields ˆ −1 {(a† )k {. . .}} = (α∗ )k N ˆ −1 {. . .} N Hence, since (a† )k is in front of {. . .}, it appears that Eq. (7.101) leads to    ∂ m {n} {n} ∗ ∗ k ˆ −1 ˆ ∗ {F (α, α )} = (α ) N N α + ∗ { f (α, α )} ∂α so that, due to Eq. (7.38), Eq. (7.102) simplifies to   ∂ m {n} {n} ∗ ∗ k {F (α, α )} = (α ) α+ ∗ { f (α, α∗ )} ∂α

7.2.8

(7.102)

(7.103)

Another theorem of interest

Consider the following linear transformation on the ground state of a† a: †

†

(exa a )(eya )|{0} where x and y are complex scalars. Insert between the last operator and the ket the unity operator built up from the first operator, that is, (e−xa a )(exa a ) = 1 †

†

Hence (exa a )(eya )|{0} = (exa a )(eya )(e−xa a )(exa a )|{0} †

†

†

†

†

†

(7.104)

Next, in view of Eq. (7.27) and taking †

f(a† ) = (eya ) Eq. (7.104) reads (exa a )(eya )(e−xa a ) = (eya †

†

†

† ex

)

Thus, Eq. (7.104) becomes †

†

(exa a )(eya )|{0} = (eya

† ex

†

)(exa a )|{0}

However, since |{0} is the ground state of a† a, with corresponding zero eigenvalue, the series expansion of the exponential of a† a on the ground state is zero except for the first term of the expansion, i.e.   x 2 (a† a)2 † (exa a )|{0} = 1 + xa† a+ + · · · |{0} = |{0} (7.105) 2! Hence, Eq. (7.105) leads to the final result †

†

(exa a )(eya )|{0} = (eya

† ex

)|{0}

(7.106)

CH007.tex

10/3/2011

22: 25

Page 217

7.3 TIME EVOLUTION OPERATOR OF DRIVEN HARMONIC OSCILLATORS

217

7.3 TIME EVOLUTION OPERATOR OF DRIVEN HARMONIC OSCILLATORS With the help of the theorems proved above, it is now possible to study the dynamics of driven quantum harmonic oscillators. For this purpose start from their Hamiltonian, which reads  2  P 1 H= + M2 Q2 + bQ 2M 2 Then, the dynamics of this system is governed by the time evolution operator, a solution of the Schrödinger equation   ∂U(t) i = H U(t) with U(0) = 1 ∂t Next, in order to solve this equation, it is suitable to work within the interaction picture and thus to make the following partition: H = H◦ + bQ with H◦ =



P2 1 + M2 Q2 2M 2

 (7.107)

Recall that the time evolution operator U(t) is related to the IP time evolution operator U(t)IP through Eq. (3.122) U(t) = U◦ (t)U(t)IP

(7.108)

with U◦ (t) = (e−iH

◦ t/

)

(7.109)

Hence, according to Eq. (3.114), the IP time evolution operator obeys the IP Schrödinger equation   ∂U(t)IP i = bQ(t)IP U(t)IP (7.110) ∂t with the boundary condition U(0)IP = 1

(7.111)

Next, due to Eq. (3.108), the interaction picture coordinate Q(t)IP appearing in Eq. (7.110) is given by Q(t)IP = U◦ (t)−1 Q(0)U◦ (t)

(7.112)

Moreover, the iteration solution of the integral equation Eq. (7.110) is of the form of (3.124), however, up to infinite order  IP

U(t)

≡1+

b i

 t

 IP



Q(t ) dt + 0



b i

2 t

 IP

Q(t ) dt 0



t  0

Q(t  )IP dt  + · · ·

CH007.tex

218

10/3/2011

22: 25

Page 218

BOSON OPERATOR THEOREMS

a solution that may be written formally as ⎧ ⎛ ⎞⎫ t ⎨ ⎬ P exp⎝−ib Q(t  )IP dt  ⎠ U(t)IP = ⎩ ⎭

(7.113)

0

where P is the Dyson time-ordering operator met in Eq. (3.87). Recall that, owing to Eq. (5.158), the position operators Q(t)IP and Q(t  )IP at different times given by Eq. (7.112) do not commute. Next, passing to Boson operators by the aid of Eqs. (5.6) and (5.7), using Eqs. (5.9) and (7.107), and finally, after neglecting the zero-point energy which is irrelevant for the present purpose, the unperturbed time evolution operator (7.109) reads U◦ (t) = (e−ia

† at

)

(7.114)

whereas the IP time evolution operator (7.110) becomes   ∂U(a, a† , t)IP i = α◦ (a† (t)IP + a(t)IP )U(a, a† , t)IP ∂t

(7.115)

with

  b α =  2M In Eq. (7.115), the IP time-dependent Boson operator is given by ◦

(7.116)

a(t)IP = U◦ (t)−1 aU◦ (t) or in view of Eq. (7.114) a(t)IP = (eia

† at

)a(e−ia

† at

)

and thus, according to Eq. (7.21) a(t)IP = ae−it so that Eq. (7.112) reads

 IP

Q(t)

=

(7.117)

 (a† eit + ae−it ) 2M

Therefore, Eq. (7.115) becomes   ∂U(a, a† , t)IP i = α◦  (a† eit + ae−it ){U(a, a† , t)IP } ∂t

(7.118)

(7.119)

Now, solve the differential equation (7.119) by the aid of the normal ordering procedure according to which it is possible to pass from operators that are functions of the noncommutative Boson operators to scalars. That is possible using the inverse of ˆ operator and of Eqs. (7.97) and (7.99) to get the N ˆ −1 {a† {U(a, a† , t)IP }} = α∗ {U {n} (α, α∗ , t)} N 

ˆ −1 {a{U(a, a† , t)IP }} = α + ∂ N ∂α∗



{U {n} (α, α∗ , t)}

(7.120) (7.121)

CH007.tex

10/3/2011

22: 25

Page 219

7.3 TIME EVOLUTION OPERATOR OF DRIVEN HARMONIC OSCILLATORS

  {n}  † IP  ∂U (α, α∗ , t) ˆ −1 ∂U(a, a , t) N = ∂t ∂t

219

(7.122)

Thus, it is possible to pass from the partial derivative Eq. (7.119) to       {n} ∂ ∂U (α, α∗ , t) ◦ ∗ it −it =α  α e + α+ ∗ e {U {n} (α, α∗ , t)} (7.123) i ∂t ∂α with, corresponding to Eq. (7.111), the boundary condition {U {n} (α, α∗ , 0)} = 1

(7.124)

Now, in order to solve the partial derivative equation (7.123), let us write {U {n} (α, α∗ , t)} = eG(t)

(7.125)

Next, in terms of the new scalar function G(t), the partial derivatives of U {n} (α, α∗ , t) with respect to the scalars t and α∗ are, respectively,     {n} ∂G(t) ∂U (α, α∗ , t) = U {n} (α, α∗ , t) ∂t ∂t 

∂U {n} (α, α∗ , t) ∂α∗



 =

 ∂G(t) U {n} (α, α∗ , t) ∂α∗

Thus, owing to these equations, and after simplification by U {n} (α, α∗ , t), Eq. (7.123) becomes      ∂G(t) ◦ it ∗ −it −it ∂G(t) i (7.126) =α  e α +e α+e ∂t ∂α∗ Again, assume for the intermediate function G(t) appearing in Eq. (7.125), an expression of the form G(t) = A(t) + B(t)α + C(t)α∗

(7.127)

Here, A(t), B(t), and C(t) are unknown functions to be found, which, due to Eqs. (7.124) and (7.125), must satisfy at the initial time A(0) = B(0) = C(0) = 1 Then, in terms of these new functions, the partial derivatives involved in (7.126) are, respectively,   ∂G(t) = C(t) ∂α∗ 

∂G(t) ∂t



 =

∂A(t) ∂t



 +

   ∂B(t) ∂C(t) α+ α∗ ∂t ∂t

so that Eq. (7.126) becomes   ∂A(t) ∂B(t) ∂C(t) ∗ i + α+ α = α◦ {eit α∗ + e−it α + e−it C(t)} ∂t ∂t ∂t

CH007.tex

220

10/3/2011

22: 25

Page 220

BOSON OPERATOR THEOREMS

which, by identification, leads to   ∂A(t) i = α◦ e−it C(t) ∂t  i

∂B(t) ∂t





∂C(t) i ∂t

= α◦ e−it



= α◦ eit

Solving these equations yields, respectively, to C(t) = −α◦ (eit − 1)

(7.128)

B(t) = α◦ (e−it − 1)

(7.129)

A(t) = iα◦2 t + α◦ B(t)

(7.130)

Thus, in view of Eqs. (7.128)–(7.130), Eq. (7.127) becomes G(t) = iα◦2 t + α◦2 (e−it − 1) + α◦ (e−it − 1)α − α◦ (e

it

− 1)α∗

so that Eq. (7.125) reads {U (n) (α, α∗ , t)} = (eiα

◦2 t

)e

◦ (t)α◦2

e−

◦ (t)∗ α∗

e

◦ (t)α

(7.131)

with ◦ (t) ≡ α◦ (e−it − 1)

(7.132)

Now, by the aid of Eq. (7.131), it is possible to return to the time evolution operator, ˆ prompting one to write using the normal ordering operator N ˆ (n) (α, α∗ , t)} = (eiα◦2 t )e ◦ (t)α◦2 N{e ˆ − ◦ (t)∗ α∗ e ◦ (t)α } N{U

(7.133)

Then, according to the first equation of (7.44), we have ˆ {n} (α, α∗ , t)} = {U(a† , a, t)IP } N{U

(7.134)

ˆ − ◦ (t)∗ α∗ e ◦ (t)α } = (e− ◦ (t)∗ a† )(e ◦ (t)a ) N{e

(7.135)

Hence, from Eqs. (7.133)–(7.135), Eq. (7.131) allows us to obtain the IP time evolution operator in the form {U(a, a† , t)IP } = (eiα

◦2 t

◦ ◦ (t)

)eα

(e−

◦ (t)∗ a†

)(e

◦ (t)a

)

(7.136)

Next, use the Glauber–Weyl theorem (1.79) to transform the right-hand-side product of exponential operators (e−

◦ (t)∗ a†

)(e

◦ (t)a

) = e−[

◦ (t)∗ a† , ◦ (t)a]/2

(e−

◦ (t)∗ a† + ◦ (t)a

The commutator appearing on the right-hand side is [ ◦ (t)∗ a† , ◦ (t)a] = | ◦ (t)|2 [a† , a] = −| ◦ (t)|2

)

(7.137)

CH007.tex

10/3/2011

22: 25

Page 221

7.4

CONCLUSION

221

Consequently, Eq. (7.137) becomes (e−

◦ (t)∗ a†

)(e

◦ (t)a

) = e|

◦ (t)|2 /2

(e−

◦ (t)∗ a† + ◦ (t)a

)

Hence, Eq. (7.136) takes the form U(a, a† , t)IP = (eiα

◦2 t

◦ ◦ (t)+| ◦ (t)|2 /2

)eα

(e−

◦ (t)∗ a† + ◦ (t)a

)

(7.138)

Again, owing to Eq. (7.132), it appears that  ◦ ◦  it α (t) + 21 | ◦ (t)|2 = α◦2 ((e−it − 1) + 21 (2 − e − e−it )) or

   it α◦ ◦ (t) + 21 | ◦ (t)|2 = α◦2 e−it − 21 e − 21 e−it



and



 α◦ ◦ (t) + 21 | ◦ (t)|2 = −iα◦2 sin t

so that Eq. (7.138) transforms to U(a, a† , t)IP = (eiα

◦2 t

)(e−iα

◦2

sin t

)(e−

◦ (t)∗ a† + ◦ (t)a

)

(7.139)

As a consequence, owing to Eqs. (7.114) and (7.139), the full time evolution operator (7.108) takes the form U(t) = (eiα

◦2 t

)(e−iα

◦2

sin t )(e−ia† at )(e− ◦ (t)∗ a† + ◦ (t)a )

(7.140)

Now, in view of Eqs. (7.108), (7.113), and (7.114), it appears that this equation may be also written, after simplification, as    t ◦2 ◦2 ◦ ∗ † ◦  IP  P exp −ib = (eiα t )(e−iα sin t )(e− (t) a + (t)a ) Q(t ) dt 0

(7.141) or, due to Eq. (7.118), for the inverse of Eq. (7.141)   t  ◦2 ◦2 ◦ ∗ † ◦ ◦ † it  −it   P exp iα = (e−iα t )(eiα sin t )(e (t) a − (t)a ) s[a e + ae ]dt 0

7.4

(7.142)

CONCLUSION

This chapter dealt with the theoretical properties of the ladder operators, more elaborate than those found in Chapter 5, and has lead to theorems allowing us to make canonical transformations concerning these operators, particularly those involving translation operators and the other time evolution operators. It has also given the most important results concerning normal and antinormal ordering formalism, allowing one to transform quantum equations dealing with noncommuting ladder operators, to

CH007.tex

10/3/2011

222

22: 25

Page 222

BOSON OPERATOR THEOREMS

scalar equivalent ones having the form of partial differential equations. All the results gained in this chapter will be widely used, particularly when studying the reversible or irreversible dynamics of quantum oscillators, more specially the following ones: Canonical transformations on ladder operators Involving the translation operator: (e−ξ

∗ a† +ξa

){f(a, a† )}(eξ

∗ a† −ξa

) = {f(a + ξ ∗ , a† + ξ)}

Involving the Hamiltonian: (eξa

†a

(eiωta

){f(a† , a)}(e−ξa a ) = {f(a† eξ , ae−ξ )} †

†a

){f(a† , a)}(e−iωta a ) = {f(a† eiωt , ae−iωt )} †

Normal ordering formalism Operator to be transformed: {F(a, a† )} = {(a† )k (a)m f(a, a† )} Passage from an operator to its corresponding scalar: ˆ −1 {F(a, a† )} = {F {n} (α, α∗ )} N The corresponding scalar expression:  F {n} (α, α∗ )} = (α∗ )k α +

 ∂ m ∂α∗

{f {n} (α, α∗ )}

BIBLIOGRAPHY W. H. Louisell. Quantum Statistical Properties of Radiation. Wiley: New York, 1973.

CH008.tex

11/3/2011

CHAPTER

14: 40

Page 223

8

PHASE OPERATORS AND SQUEEZED STATES INTRODUCTION In the present chapter, we study new operators, states, and theorems concerning harmonic oscillators that are less usual than those previously studied. Up to now, in the study of quantum oscillators, we have not yet encountered the concept of phase, which is usual in the area of classical oscillators. The corresponding quantum phase may be treated using the phase operator, which is the object of the first section of this chapter, where it will be applied to situations in which the quantum harmonic oscillator is described by coherent states. Now, other quantum states exist that resemble coherent states, although they are more complex. They are the squeezed states. One of their characteristic properties leads to uncertainty relations, which evoke phase properties because they involve time-dependent oscillatory momentum and position uncertainties coming back and forth. These squeezed states will be comprehensively treated in the second section of this chapter, which ends with a study of the Bogolioubov– Valatin transformation allowing one to diagonalize some Hamiltonians involving the product of Boson operators of the same forms as those appearing in squeezed states.

8.1

PHASE OPERATORS

We begin with phase operators evoking the phases appearing in the physics of classical oscillators.

8.1.1 Phase operators in the basis of harmonic Hamiltonian eigenkets For this purpose, define a new operator  through the raising and lowering operators according to the following equations:  a = a† a + 1(ei )  a† = (e−i ) a† a + 1 Quantum Oscillators, First Edition. Olivier Henri-Rousseau and Paul Blaise. © 2011 John Wiley & Sons, Inc. Published 2011 by John Wiley & Sons, Inc.

223

CH008.tex

224

11/3/2011

14: 40

Page 224

PHASE OPERATORS AND SQUEEZED STATES

which read by inversion

 (ei ) =



1

√

a (8.1) a† a + 1   1 (e−i ) = a† √ (8.2) a† a + 1 Now, we prove that the operators appearing on the left-hand side of these two last equations are unitary. Thus, take the product of Eq. (8.1) by Eq. (8.2):     1 1 i −i † (e )(e ) = √ aa √ a† a + 1 a† a + 1 which, owing to the commutation rule (5.5), yields     1 1 (ei )(e−i ) = √ (a† a + 1) √ (8.3) a† a + 1 a† a + 1 Now, one may write the last two terms of the right-hand side of this equation as       1 1 = a† a + 1 a† a + 1 √ (a† a + 1) √ a† a + 1 a† a + 1 or    1 † (a a + 1) √ = a† a + 1 a† a + 1 so that Eq. (8.3) becomes   1 a† a + 1 (ei )(e−i ) = √ a† a + 1 which reduces to (ei )(e−i ) = 1

(8.4)

Moreover, the action of the operator (8.1) on an eigenstate of the harmonic oscillator Hamiltonian yields (ei )|{n} = √

1 a† a

+1

a |{n}

and, owing to Eq. (5.53), (ei )|{n} = √

1

√

n|{n − 1} +1 Again, write the following formal expansion of the square root  1 Ck y k √ = y a† a

(8.5)

(8.6)

k

where an explicit expression of the expansion coefficients Ck need not be given. Applying this expansion to y = a† a + 1 reads  1 Ck (a† a + 1)k with k = 0, 1, . . . = √ a† a + 1 k

CH008.tex

11/3/2011

14: 40

Page 225

8.1

PHASE OPERATORS

225

or, after postmultiplying by |{n} √

1 a† a

+1

|{n} =



Ck (a† a + 1)k |{n}

(8.7)

k

Then, due to Eq. (5.42), this last expression transforms to  1 |{n} = Ck (n + 1)k |{n} √ a† a + 1 k

(8.8)

Furthermore, applying in turn Eq. (8.6) with y = n + 1, that is,  1 Ck (n+1)k = √ n+1 k Eq. (8.8) becomes √

1

|{n} = √

1

n+1 +1 or, changing n + 1 into n, and thus n into n − 1 √

a† a

|{n}

1

1 |{n − 1} = √ |{n − 1} n a† a + 1

(8.9)

Now, recall Eq. (8.5), that is, (ei )|{n} = √

1

√

n|{n − 1} +1 which, in view of Eq. (8.9) transforms after simplification into a† a

(ei )|{n} = |{n − 1}

(8.10)

(8.11)

In a similar way, one obtains for the corresponding operator defined by Eq. (8.2) (e−i )|{n} = |{n + 1}

(8.12)

Hence, as a consequence of Eqs. (8.11) and (8.12), and owing to the orthonormality of the eigenstates of the harmonic oscillator, the matrix elements of the operators (8.1) and (8.2) satisfy {m}|ei |{n} = {m}|{n − 1} = δm,n−1

(8.13)

{m}|e−i |{n} = {m}|{n + 1} = δm,n+1

(8.14)

Now, introduce the following operators cos  = 21 (ei + e−i )

(8.15)

− e−i )

(8.16)

sin  =

1 i 2i (e

Then, in the basis of the eigenstates of the harmonic oscillator, the matrix elements of the first operator read {m}| cos |{n} = 21 {{m}|ei |{n} + {m}|e−i |{n}}

CH008.tex

226

11/3/2011

14: 40

Page 226

PHASE OPERATORS AND SQUEEZED STATES

or, owing to Eqs. (8.13) and (8.14), we have {m}| cos |{n} = 21 (δm,n−1 + δm,n+1 ) In a similar way, we have {m}| sin |{n} =

1 2i (δm,n−1

− δm,n+1 )

As a consequence of these two last equations, it appears that the diagonal elements of these operators are zero, that is, {n}| cos |{n} = {n}| sin |{n} = 0 Moreover, consider the matrix elements of {m}| cos |{n} = 2

2i 1 4 {{m}|(e

cos2 ,

+e

−2i

(8.17)

which, due to Eq. (8.15), read

)|{n} + {m}|2|{n}}

(8.18)

However, since (e2i )|{n} = (ei )(ei )|{n} and, in view of Eq. (8.11), it appears that (e2i )|{n} = (ei )|{n − 1} or, using in turn Eq. (8.11) (e2i )|{n} = |{n − 2}

(8.19)

In a similar way, using Eq. (8.12), we have (e−2i )|{n} = |{n + 2}

(8.20)

Then, Eqs. (8.19) and (8.20) allow to transform Eq. (8.18) into {m}| cos2 |{n} = 41 {m}|(|{n − 2} + |{n + 2}) + 21 δmn or, using the orthonormality properties of the states involved in this expression, it reduces to {m}| cos2 |{n} = 41 (δm,n−2 + δm,n+2 ) + 21 δmn

(8.21)

{m}| sin2 |{n} = − 41 (δm,n−2 + δm,n+2 ) − 21 δmn

(8.22)

In like manner Hence, the diagonal elements of Eqs. (8.21) and (8.22), reduce simply to {n}| cos2 |{n} = {n}| sin2 |{n} =

1 2

(8.23)

Moreover, the dispersions of cos  and sin  in the states |{n}, which are, respectively, given by  ( cos )|n = {n}| cos2 |{n} − {n}| cos |{n}2 ( sin )|n =

 {n}| sin2 |{n} − {n}| sin |{n}2

read, due to Eqs. (8.17) and (8.23), ( cos )|n = ( sin )|n = which indicates a random dispersion of the phase.



1 2

CH008.tex

11/3/2011

14: 40

Page 227

8.1

8.1.2

227

PHASE OPERATORS

Commutation rule involving the phase operators

Now, seek the commutator of a† a with the operator defined by Eq. (8.1), that is, [a† a,(ei )] = a† a √

1 a† a

+1

a− √

1 a† a

+1

aa† a

(8.24)

Because a† a commutes with all its powers, and in view of the expression of the commutator (5.5) of a† and a, Eq. (8.24) reduces to [a† a,(ei )] = √

1 a† a

+1

a† aa − √

1 a† a

+1

(a† a + 1)a

or, factorizing and rearranging, [a† a,(ei )] = − √

1 a† a

+1

a

and after simplification and use of Eq. (8.1), [a† a,(ei )] = −(ei )

(8.25)

[a† a,(e−i )] = (e−i )

(8.26)

Similarly

Hence, due to Eqs. (8.15) and (8.16), Eqs. (8.25) and (8.26) lead to [a† a, cos ] = −i sin 

(8.27)

[a† a, sin ] = i cos 

(8.28)

It is now possible to get the product of the uncertainties over a† a and cos  or sin . Keeping in mind that the product of uncertainties of two operators A and B calculated over kets | is given by Eq. (2.49), that is, (A )2 (B )2 ≥ − 41 |[A, B]|2 and taking A = a† a and B = cos or sin , Eqs. (8.27) and (8.28) allow us to get the uncertainty relations, which read in the present situation (a† a)2 ( cos )2 ≥ 41 | sin |2 (a† a)2 ( sin )2 ≥ 41 | cos |2 or (a† a) ( cos ) ≥ 21 || sin || (a† a) ( sin ) ≥ 21 || cos ||

CH008.tex

228

8.1.3

11/3/2011

14: 40

Page 228

PHASE OPERATORS AND SQUEEZED STATES

Phase operators within coherent-state picture

8.1.3.1 Diagonal matrix elements of cos  Consider now the average value of the operator cos  performed over a coherent-state picture: {α}| cos |{α} = 21 ({α}|ei |{α} + {α}|e−i |{α})

(8.29)

Keeping in mind the expansion of a coherent state on the eigenstates of the harmonic oscillator Hamiltonian given by Eq. (6.16), that is,   (α)n  2 |{α} = e−|α| /2 |{n} (8.30) √ n! n {α}| = e−|α| Eq. (8.29) reads {α}| cos |{α} =

2 /2

  (α∗ )m  {m }| √ m! m

(8.31)

  1 −|α|2   (α∗ )m (α)n ({m}|ei |{n} + {m}|e−i |{n}) e √ √ 2 m! n! n m

so that, passing from the imaginary exponentials to the corresponding cosine function, and due to Eqs. (8.13) and (8.14), we have   1 −|α|2   (α∗ )m (α)n {α}| cos |{α} = e (δm,n−1 + δm,n+1 ) √ √ 2 m! n! n m or   1 −|α|2  (α∗ )n−1 (α)n (α∗ )n+1 (α)n {α}| cos |{α} = e √ √ +√ √ 2 (n − 1)! n! (n + 1)! n! n Since (n − 1)! cannot start from n = 0, the summation must be redefined in the terms where (n − 1)! appears, by changing n into n + 1, leading to  ∗n  1 (α ) (α)n+1 (α∗ )n+1 (α)n 2  {α}| cos |{α} = e−|α| +√ √ √ √ 2 (n + 1)! n! n! (n + 1)! n or  2n  1 |α| (α + α∗ ) 2  {α}| cos |{α} = e−|α| (8.32) √ √ 2 n! (n + 1)! n Next, writing α = |α|eiθ Eq. (8.32) yields −|α|2

{α}| cos |{α} = e

cos θ

(8.33)

 n

or, using (n + 1)! = (n + 1)n! −|α|2

{α}| cos |{α} = e

|α|2n+1 √ √ n! (n + 1)!

  |α|2n+1  cos θ √ n! n + 1 n



CH008.tex

11/3/2011

14: 40

Page 229

8.2

SQUEEZED STATES

229

8.1.3.2 Diagonal matrix element of cos2  Now, consider the average value of cos2  over a coherent state that, owing to Eq. (8.15), reads {α}| cos2 |{α} = 41 {α}|(ei + e−i )2 |{α} or, after expansion of the right-hand-side squared term, and using Eq. (8.4), we have {α}| cos2 |{α} = 41 [{α}|(e2i + e−2i )|{α} + 2{α}|{α}] Next, assuming that the coherent states are normalized, and after using Eqs. (8.30) and (8.31), these matrix elements become   1 1 −|α|2   (α∗ )m (α)n 2 {α}| cos |{α} = + e [{m}|(e2i + e−2i )|{n}] √ √ 2 4 m! n! n m (8.34) or, due to Eqs. (8.19) and (8.20), the equation above becomes  ∗m  1 1 (α ) (α)n 2  {α}| cos2 |{α} = + e−|α| {m}|(|{n − 2} + |{n + 2}) √ √ 2 4 m! n! n m and, thus, after using the orthonormality properties,   1 1 −|α|2   (α∗ )m (α)n 2 {α}| cos |{α} = + e (δm,n−2 + δm,n+2 ) √ √ 2 4 m! n! n m or

  (α∗ )n+2 (α)n 1 1 −|α|2  (α∗ )n−2 (α)n {α}| cos |{α} = + e (8.35) √ √ +√ √ 2 4 (n − 2)! n! (n + 2)! n! n 2

Moreover, as above, shift the index in the sum containing (n − 2)! by changing n into n + 2, so that Eq. (8.35) reads {α}| cos2 |{α} = or, using Eq. (8.33), {α}| cos2 |{α} =

1 1 −|α|2  (α∗ )n (α)n+2 + (α∗ )n+2 (α)n + e √ √ 2 4 n! (n + 2)! n

  |α|2n 1 1 −|α|2 2 |α| cos(2θ) + e √ √ 2 2 n! (n + 2)! n

so that, using finally (n + 2)! = (n + 2)(n + 1)n!, we have     |α|2n 1 1  2 −|α|2 2 2 {α}| cos |{α} = + e |α| cos θ − √ 2 2 n n! (n + 2)(n + 1)

8.2

SQUEEZED STATES

We have often encountered coherent states that may be viewed as the result of the action of the translation operator A(α) = exp{αa† − αa}

CH008.tex

230

11/3/2011

14: 40

Page 230

PHASE OPERATORS AND SQUEEZED STATES

on any eigenstate |{n} of the harmonic oscillator Hamiltonian, and we have found that, whatever they may be, these states minimize the Heisenberg uncertainty relations.

8.2.1 Canonical transformations on ladder operators using squeezing operators However, other interesting states exist that may be considered as generalization of coherent states, the squeezed states, which may be obtained by the action on the kets |{n} of the following operator:   S(z) = exp 21 (za†2 − z∗ a2 )

(8.36)

the Hermitian conjugate of which is     S(z)† = exp 21 (z∗ a2 − za†2 ) = exp − 21 (za†2 − z∗ a2 ) whereas its inverse is

  S(z)−1 = exp − 21 (za† 2 − z∗ a2 ) = S(−z)

(8.37)

that implies that the operator (8.36) is unitary since obeying S(z)† = S(z)−1 Next, consider the following canonical transformation:     S(z)aS(z)−1 = exp 21 (za†2 − z∗ a2 ) a exp − 21 (za†2 − z∗ a2 )

(8.38) (8.39)

To perform this transformation, one may use the Baker–Campbell–Hausdorff formula (1.76): 1 1 eξA Be−ξA = B + [A, B]ξ + [A,[A, B]]ξ 2 + [A,[A,[A, B]]]ξ 3 + · · · 2 3! where A and B are two linear operators and ξ a scalar. If one defines the operator D as D = ξA the Baker–Campbell–Hausdorff relation reads 1 1 eD Be−D = B + [D, B] + [D,[D, B]] + [D,[D,[D, B]]] + · · · 2 3!

(8.40)

Hence, setting in Eq. (8.40) B=a

and

D = 21 (za†2 − z∗ a2 )

Eq. (8.39) takes the form 1 S(z)aS(z)−1 = a + [(za†2 − z∗ a), a] 2 11 [(za†2 − z∗ a), [(za†2 − z∗ a), a]] + 2! 4 11 + [(za†2 − z∗ a)[(za†2 − z∗ a), [(za†2 − z∗ a), a]]] + · · · 3! 8 (8.41)

CH008.tex

11/3/2011

14: 40

Page 231

8.2

231

SQUEEZED STATES

The first commutator appearing in this equation simplifies to [(za†2 − z∗ a2 ), a] = z[a†2 , a]

(8.42)

Now, keeping in mind that the right-hand-side commutator is given by Eq. (7.47), that is, [a†2 , a] = −2a†

(8.43)

the left-hand-side commutator of Eq. (8.42) yields [(za†2 − z∗ a), a] = −2za†

(8.44)

Therefore, the double commutator appearing in Eq. (8.41) reads [(za†2 − z∗ a2 ), [(za†2 − z∗ a2 ), a]] = [(za†2 − z∗ a2 ), −2za† ] = [(−z∗ a2 ), −2za† ] = 2|z|2 [a2 , a† ]

(8.45)

Now, the right-hand-side commutator of Eq. (8.45) is [a2 , a† ] = aaa† − a† aa = a(a† a + 1) − a† aa or [a2 , a† ] = (aa† − a† a)a + a = 2a

(8.46)

so that, the double commutator (8.45) becomes [(za†2 − z∗ a2 ), [(za†2 − z∗ a2 ), a]] = 4|z|2 a

(8.47)

Again, due to Eq. (8.47), the triple commutator appearing in Eq. (8.41) reads [(za†2 − z∗ a), [(za†2 − z∗ a2 ), [(za†2 − z∗ a), a]]] = [(za†2 − z∗ a2 ), 4|z|2 a] = 4|z|2 z[a†2 , a] or, using in turn Eq. (8.43) [(za†2 − z∗ a), [(za†2 − z∗ a2 ), [(za†2 − z∗ a), a]]] = −8z|z|2 a†

(8.48)

Moreover, according to Eq. (8.48), the quadruple commutator of Eq. (8.41) takes the form [(za†2 − z∗ a2 ), [(za†2 − z∗ a2 ), [(za†2 − z∗ a2 ), [(za†2 − z∗ a2 ), a]]]] = [(za†2 − z∗ a2 ), −8z|z|2 a† ] = 8|z|4 [a2 , a† ] so that, owing to Eq. (8.46), it becomes [(za†2 − z∗ a2 ), [(za†2 − z∗ a2 ), [(za†2 − z∗ a2 ), [(za†2 − z∗ a2 ), a]]]] = 16|z|4 a (8.49) At last, collecting the results from (8.44), (8.47), (8.48), and (8.49), the canonical transformation (8.41) appears to be     1 1 2 1 4 1 −1 † 2 4 S(z)aS(z) = a 1+ |z| + |z| − a z + z|z| + z|z| 2! 4! 3! 5!

CH008.tex

232

11/3/2011

14: 40

Page 232

PHASE OPERATORS AND SQUEEZED STATES

or

    1 z 1 1 1 |z| + |z|3 + |z|5 + · · · S(z)aS(z)−1 = a 1+ |z|2 + |z|4 + · · · − a† 2! 4! |z| 3! 5! (8.50)

Next, due to the following expansions of the hyperbolic sine and cosine functions,

where

 cosh z =

cosh z = 1 +

z4 z2 + + ··· 2! 4!

sinh z = z +

z3 z5 + + ··· 3! 5!

ez + e−z 2



 sinh z =

and

ez − e−z 2

 (8.51)

it appears that the canonical transformation (8.50) reduces to the compact expression S(z)aS(z)−1 = a cosh |z| −

z † a sinh |z| |z|

(8.52)

Again, changing z to −z, and using Eqs. (8.36) and (8.38), Eq. (8.52) yields S(z)−1 aS(z) = a cosh |z| +

z † a sinh |z| |z|

(8.53)

Observe that the Hermitian conjugate of this equation is (S(z)−1 aS(z))† = a† cosh |z| + a

z |z| |z|

(8.54)

its left-hand side being (S(z)−1 aS(z))† = (S(z))† a† (S(z)−1 )† or, since the operator S(z) is unitary, (S(z)−1 aS(z))† = S(z)−1 a† S(z)

(8.55)

so that identifying Eqs. (8.54) and (8.55) leads to S(z)−1 a† S(z) = a† cosh |z| +

8.2.2 8.2.2.1

z a sinh |z| |z|

(8.56)

Uncertainty relations for squeezed state Squeezed state

Introduce the squeezed states according to

|{z(t), α(t)} = U◦ (t)−1 A(α◦ )S(|z|)|{0}

(8.57)

Here |{0} is the ground state of the harmonic Hamiltonian, whereas S(|z|) is the squeezing operator defined by Eq. (8.36), and A(α◦ ) the translation operator defined by Eq. (6.74), U◦ (t) being the time evolution operator constructed from a† a, that is, S(|z|) = (e(|z|a

†2 −|z|a2 )/2

)

CH008.tex

11/3/2011

14: 40

Page 233

8.2

A(α◦ ) = (eα

◦ a† −α◦ a

U◦ (t) = e−iωta

SQUEEZED STATES

233

)

†a

Hence, Eq. (8.57) becomes †

|{z(t), α(t)} = (eiωta a ) (eα

◦ a† −α◦ a

) (e(|z|a

†2 −|z|a2 /2)

)|{0}

Now, insert between the translation and the squeezing operators and between the squeezing operator and the ground-state eigenket, the unity operator built up from the time evolution operator, that is, (e−iωta a ) (eiωta a ) = 1 †

†

We have †

|{z(t), α(t)} = (eiωta a ) eα × e(|z|a

◦ a† −α◦ a

†2 −|z|a2 )/2

(e−iωta a ) (eiωta a ) †

†

(e−iωta a ) (eiωta a )|{0} †

†

(8.58)

Next, keeping in mind Eq. (7.31), (eiωta a ) f(a† , a) (e−iωta a ) = {f(a† eiωt , a e−iωt )} †

†

and applying this expression, we have †

◦ a† −α◦ a

(eiωta a ) (eα †

(eiωta a ){e(|z|a

†2 −|z|a2 )/2

∗ (t)a† −α(t)a

) (e−iωta a ) = eα †

} (e−iωta a ) = {e(z †

= {A(α(t))}

∗ (t)a†2 −z(t)a2 )/2

} = {S(z(t))}

(8.59) (8.60)

where α(t) = α◦ e−iωt

and

z(t) = |z|e−2iωt

(8.61)

Now, expansion of the exponential operator appearing in Eq. (8.58) suggests writing the two last terms of the right-hand side of this equation as   (iωta† a)2 iωta† a † (e )|{0} = 1 + iωta a + + · · · |{0} (8.62) 2! so that keeping in mind that the action of a† a on its ground state |{0} is zero, that is, a† a|{0} = 0|{0} Then, it appears that all terms involved in the sum of Eq. (8.62 ) are zero, except that corresponding to n = 0, which acts on the ground state as the unity operator, so that †

(eiωta a )|{0} = |{0}

(8.63)

Hence, owing to Eqs. (8.59), (8.60), and (8.63), Eq. (8.58) becomes |{z(t), α(t)} = A(α(t))S(z(t))|{0}

(8.64)

CH008.tex

234

11/3/2011

14: 40

Page 234

PHASE OPERATORS AND SQUEEZED STATES

8.2.3

Ladder operator functions averaged over squeezed states

8.2.3.1 Average values of a† and a Now, consider the time-dependent mean value of a† averaged over the squeezed state (8.57), that is, a(t)† z,α = {z(t), α(t)}|a† |{z(t), α(t)}

(8.65)

which, due to Eq. (8.64), takes the form a(t)† z,α = {0}|S(z(t))−1 A(α(t))−1 a† A(α(t))S(z(t))|{0}

(8.66)

Next, keeping in mind Eq. (6.81), which implies that the action of the time-dependent translation operator on the Boson operators gives A(α(t))−1 a† A(α(t)) = a† + α(t)∗ A(α(t))−1 aA(α(t)) = a + α(t)

(8.67)

Using also the fact, S(z(t))−1 (a† + α(t)∗ )S(z(t)) = α(t)∗ + S(z(t))−1 a† S(z(t)) and taking into account Eq. (8.56), it appears that S(z(t))−1 A(α(t))−1 a† A(α(t))S(z(t)) = a† cosh |z| + a

z(t) sinh |z| + α(t)∗ |z|

(8.68)

so that Eq. (8.66) becomes a(t)† z,α = {0}|a† cosh |z||{0} + {0}| a (e−2iωt ) sinh |z||{0} + α(t)∗ Finally, in view of Eq. (8.61), and using a|{0} = 0

and

{0}|a† = 0

we have a(t)† z,α = α(t)∗ = α◦ eiωt

(8.69)

the Hermitian conjugate of which is a(t)z,α = α(t) = α◦ e−iωt 8.2.3.2 Average values of (a† )2 and (a)2 square of a(t)† defined by

(8.70)

Moreover, the average value of the

(a(t)† )2 z,α = {z(t), α(t)}|(a† )2 |{z(t), α(t)}

(8.71)

yields, with the help of Eq. (8.64), (a(t)† )2 z,α = {0}|S(z(t))−1 A(α(t))−1 a† a† A(α(t))S(z(t))|{0} Again, insert the following unity operator: 1 = (A(α(t))S(z(t))S(z(t))−1 A(α(t))−1 )

(8.72)

CH008.tex

11/3/2011

14: 40

Page 235

8.2

235

SQUEEZED STATES

between the two raising operators, as follows: (a(t)† )2 z,α = {0}|S(z(t))−1 A(α(t))−1 a† × (A(α(t))S(z(t))S(z(t))−1 A(α(t))−1 )a† A(α(t))S(z(t))|{0} leading to (a(t)† )2 z,α = {0}|(S(z(t))−1 A(α(t))−1 a† A(α(t))S(z(t)))2 |{0} and thus, using Eq. (8.68), (a(t)† )2 z,α = {0}|(a† cosh |z| + (e−2iωt )a sinh |z| + α(t)∗ )2 |{0} Again, performing the square involved on the right-hand-side average value yields (a(t)† )2 z,α = {0}|{(a† )2 cosh2 |z| + (e4iωt )(a)2 sinh2 |z| + (2a† a+1) sinh |z|e2iωt cosh |z|}|{0} + 2α(t)∗ {0}|( cosh |z|a† + 2e2iωt a sinh |z|)|{0} + α(t)∗2 or, after simplifications, due to a† a|{0} = 0 {0}|(a† )2 |{0} = {0}|(a)2 |{0} = 0 (a(t)† )2 z,α = (e−2iωt ) sinh |z| cosh |z| + α◦2 (e2iωt )

(8.73)

the Hermitian conjugate of which reads (a(t))2 z,α = (e2iωt ) sinh |z| cosh |z| + α◦2 (e−2iωt ) 8.2.3.3 Average value of a† a occupation number a† a:

(8.74)

Finally, consider the average value of the

(a† a)z,α = {z(t), α(t)}|a† a|{z(t), α(t)}

(8.75)

Then, insert in the following way, between the two Boson operators, the unity operator (8.72): (a† a)z,α = {0}|(S(z(t))−1 A(α(t))−1 a† A(α(t))S(z(t))) × (S(z(t))−1 A(α(t))−1 aA(α(t))S(z(t)))|{0} Hence, using Eq. (8.68), Eq. (8.76) yields (a† a)z,α = {0}|(a† cosh |z| + a(e−2iωt ) sinh |z| + α◦ (eiωt )) × hc|{0} where hc is the Hermitian conjugate hc = a cosh |z| + a† (e2iωt ) sinh |z| + α◦ (e−iωt )

(8.76)

CH008.tex

236

11/3/2011

14: 40

Page 236

PHASE OPERATORS AND SQUEEZED STATES

The product involved in the right-hand-side term reads (a† cosh |z| + ae−2iωt sinh |z| + α◦ (eiωt )) × hc = |α|2 + α◦ (e−iωt )(a† cosh |z| + a(e−2iωt ) sinh |z|) + a† a cosh2 |z| + (2a† a+1) sinh2 |z| + ((e−2iωt )(a)2 + (e2iωt )(a† )2 ) cosh |z| sinh |z| + α◦ (eiωt )(a cosh (|z|) + a† (e2iωt ) sinh |z|) so that Eq. (8.76) appears to be simply (a† a)z,α = |α|2 + sinh2 |z|

(8.77)

Therefore, the mean value of the oscillator Hamiltonian (5.9) averaged over the squeezed states is given by

Hz,α = ω{z(t), α(t)}| a† a + 21 |{z(t), α(t)} which reads, after using Eq. (8.77),



Hz,α = ω |α|2 + sinh2 |z| + 21

8.2.4

Uncertainty relations for squeezed states

8.2.4.1 Average values of Q and P operators It is now possible, using Eqs. (8.69) and (8.70), to get the expression for the corresponding average value of the position operator Q and of its conjugate momentum P. First, begin with Q leading to Q(t)z,α = {z(t), α(t)}|Q|{z(t), α(t)} which, according to Eq. (5.6), reads   Q(t)z,α = ({z(t), α(t)}|a† |{z(t), α(t)} + {z(t), α(t)}|a|{z(t), α(t)}) (8.78) 2mω or, due to the definition equation (8.65) and to its Hermitian conjugate,   Q(t)z,α = (a(t)† z,α + a(t)z,α ) 2mω and in view of Eqs. (8.69) and (8.70), it yields   ◦ Q(t)z,α = 2α (8.79) cos ωt 2mω On the other hand, the corresponding average value of the momentum P P(t)z,α = {z(t), α(t)}|P|{z(t), α(t)} reads, according to Eqs. (5.7) and (8.65),  mω P(t)z,α = i (a(t)† z,α − a(t)z,α ) 2

CH008.tex

11/3/2011

14: 40

Page 237

8.2

SQUEEZED STATES

so that, in view of Eqs. (8.69) and (8.70), we have  mω ◦ sin ωt P(t)z,α = 2α 2

237

(8.80)

8.2.4.2 Mean value of Q2 and corresponding fluctuation Now consider the following average value of the squared position operator Q(t)2 z,α = {z(t), α(t)}|Q2 |{z(t), α(t)} Then, using Eq. (8.57), and the fact that the squeezing and translation operators are unitary, it becomes  {0}|S(z(t))−1 A(α(t))−1 ((a† )2 +(a)2 +2a† a+1)A(α(t))S(z(t))|{0} 2mω or, due to Eq. (8.71) and its Hermitian conjugate, and also to Eq. (8.75) Q(t)2 z,α =

 ((a† (t))2 z,α + (a(t))2 z,α + 2(a† a(t))z,α + 1) 2mω so that, with Eqs. (8.73), (8.74), and (8.77), we have Q(t)2 z,α =

Q(t)2 z,α =

 ((α(t)∗2 + α(t)2 ) + 2 cos 2ωt sinh |z| cosh |z| 2mω + 2|α(t)|2 + 2 sinh2 |z| + 1)

(8.81)

Next, using the trigonometric formulas cos 2ωt = cos2 ωt − sin2 ωt

(8.82)

the product of hyperbolic sine and cosine functions leads to sinh |z| cosh |z| =

e|z| − e−|z| e|z| + e−|z| e2|z| − e−2|z| = 2 2 4

2 sinh (|z|) cosh (|z|) = sinh (2|z|) 

e|z| − e−|z| 2 sinh (|z|) = 2 2 2

2 =

(8.83)

e2|z| + e−2|z| − 2 2

2 sinh2 |z| = cosh 2|z| − 1

(8.84)

Then, Eq. (8.81) reduces to  {(α∗ (t) + α(t))2 + ( cos2 ωt − sin2 ωt) sinh 2|z| + cosh 2|z|} 2mω which may be also written as Q(t)2 z,α =

Q(t)2 z,α =

 {(α∗ (t) + α(t))2 + ( cos2 ωt − sin2 ωt) sinh 2|z| 2mω + ( cos2 ωt + sin2 ωt) cosh 2|z|}

CH008.tex

238

11/3/2011

14: 40

Page 238

PHASE OPERATORS AND SQUEEZED STATES

or Q(t) z,α 2

 e2|z| − e−2|z| = (α∗ (t) + α(t))2 + ( cos2 ωt − sin2 ωt) 2mω 2

2|z| −2|z| e +e + ( cos2 ωt + sin2 ωt) 2

and thus Q(t)2 z,α =

 {(α∗ (t) + α(t))2 + e2|z| cos2 ωt + e−2|z| sin2 ωt} 2mω

(8.85)

Hence, according to Eqs. (8.79) and (8.85), the fluctuation of the coordinate operator defined by  Qz,α (t) = Q(t)2 z,α − Q(t)2z,α reads

 Qz,α (t) =

  2|z| e cos2 ωt + e−2|z| sin2 ωt 2mω

(8.86)

On the other hand, the following average value of the squared momentum operator P(t)2 z,α = {z(t), α(t)}|P2 |{z(t), α(t)}

(8.87)

reads, in view of Eqs. (5.7), (8.57), (8.71), and (8.75), mω {(a† (t))2 z,α + (a(t))2 z,α − (2a† az,α + 1)} 2 or, using Eqs. (8.73), (8.74), and (8.77), P(t)2 z,α = −

P(t)2 z,α = −

mω {(α(t)∗2 + α(t)2 ) + 2 cos 2ωt sinh |z| cosh |z| 2 − (2|α(t)|2 + 2 sinh2 |z| + 1)}

which can be rearranged by the aid of Eqs. (8.82)–(8.84), according to P(t)2 z,α =

mω ∗ {(α (t) − α(t))2 + e−2|z| cos2 ωt − e2|z| sin2 ωt} 2

(8.88)

Moreover, owing to Eqs. (8.80) and (8.88), the fluctuation of the momentum  Pz,α (t) = P(t)2 z,α − P(t)2z,α takes the form

 Pz,α (t) =

mω  −2|z| e cos2 ωt + e2|z| sin2 ωt 2

(8.89)

so that multiplying this result by Eq. (8.86) leads to the following uncertainty relation: Pz,α (t)Qz,α (t) =

 −2|z| cos2 ωt + e2|z| sin2 ωt) (e 2

(8.90)

CH008.tex

11/3/2011

14: 40

Page 239

8.3

8.3

BOGOLIUBOV–VALATIN TRANSFORMATION

239

BOGOLIUBOV–VALATIN TRANSFORMATION

We shall end the present chapter with the products of Boson operators of the form a† a† and a a, with the Bogolioubov–Valatin transformation allowing one to diagonalize the following Hamiltonian: H = ω1 a1† a1 + ω2 a2† a2 + ω12 (a1† a2† + a1 a2 )

(8.91)

where ω1 and ω2 are the angular frequencies of the two oscillators, ω12 is the energetic coupling parameter, where a1 , a1† and a2 , a2† are ladder operators satisfying the commutation rule [ai , aj† ] = δij

(8.92)

We attempt to diagonalize this Hamiltonian so that it will read H = E ◦ +  1 b†1 b1 +  2 b†2 b2

(8.93)

where bi and b†i are new Boson operators satisfying the commutation relation [bi , b†j ] = δij

(8.94)

and where the i are the angular frequencies of the decoupled oscillators, whereas E ◦ is some reference energy. The diagonalization of the Hamiltonian (8.91) into (8.93) may be performed through the linear Bogolioubov–Valatin transformation b1 = a1 cosh ϕ + a2† sinh ϕ

b†1 = a1† cosh ϕ + a2 sinh ϕ

(8.95)

b2 = a1† sinh ϕ − a2 cosh ϕ

b†2 = a1 sinh ϕ − a2† cosh ϕ

(8.96)

In order to determine the transformation parameter ϕ, suppose that the commutator of b1 with the Hamiltonian (8.91) is equal to that of b†1 b1 with the Hamiltonian (8.93). In this context, observe that, owing to Eq. (8.94), the commutator of b1 with the Hamiltonian (8.93) is simply [b1 , H] =  1 [b1 , b†1 b1 ] =  1 b1 or, using the first equation appearing in (8.95), we have [b1 , H] = (a1 cosh ϕ + a2† sinh ϕ) 1

(8.97)

On the other hand, according to the first equation of (8.95), the commutator of b1 with the Hamiltonian (8.91), reads [b1 , H] = [(a1 cosh ϕ + a2† sinh ϕ), (ω1 a1† a1 + ω2 a2† a2 + ω12 (a1† a2† + a1 a2 ))] or [b1 , H] = [a1 , (ω1 a1† a1 + ω2 a2† a2 + ω12 (a1† a2† + a1 a2 ))] cosh ϕ + [a2† , (ω1 a1† a1 + ω2 a2† a2 ) + ω12 (a1† a2† + a1 a2 )] sinh ϕ

(8.98)

CH008.tex

240

11/3/2011

14: 40

Page 240

PHASE OPERATORS AND SQUEEZED STATES

Again, owing to the commutation rules (8.92), and applying Eqs. (5.15) and (5.16), it appears that [a1 , a1† a1 ] = a1 [a2† , a2† a2 ] = a2†

and and

[a1 , a1† a2† ] = a2† [a2† , a1 a2 ] = −a1

Thereby, Eq. (8.98) transforms to [b1 , H] = a1 (ω1 cosh ϕ − ω12 sinh ϕ) + a2† (ω12 cosh ϕ − ω2 sinh ϕ)

(8.99)

Then, equating the two commutators (8.97) and (8.99)yields  1 (a1 cosh ϕ + a2† sinh ϕ) = a1 (ω1 cosh ϕ − ω12 sinh ϕ) + a2† (ω12 cosh ϕ − ω2 sinh ϕ) or, after identification, since the Boson operators cannot be zero ( 1 − ω1 ) cosh ϕ + ω12 sinh ϕ = 0

(8.100)

−ω12 cosh ϕ + ( 1 + ω2 ) sinh ϕ = 0

(8.101)

Now, since the coefficients cosh ϕ and sinh ϕ are different from zero, the set of equations (8.100) and (8.101) is satisfied if   ( 1 − ω1 ) ω12   =0 −ω12 ( 1 + ω2 ) or, after expansion of the determinant, 2

21 − 1 (ω1 − ω2 ) + (ω12 − ω 1 ω2 ) = 0

In the two solutions of this equation

1 = 21 ((ω1 − ω2 ) ±



2 ) (ω1 + ω2 )2 − 4ω12

the one that must be selected is that allowing 1 in Eq. (8.91) to be equal to ω1 when the coupling ω12 is zero, that is,  2 )

1 = 21 ((ω1 − ω2 ) + (ω1 + ω2 )2 − 4ω12 (8.102) Moreover, according to Eq. (8.100), the ratio of the coefficients sinh ϕ and cosh ϕ reads ω1 − 1 sinh ϕ = tanh ϕ = cosh ϕ ω12 In like manner for the commutator of b2 with the Hamiltonian H given, using Eq. (8.91) or (8.93), and with the help of the first equation of (8.96), one obtains for the second angular frequency appearing in Eq. (8.93)  2 )

2 = 21 ((ω2 − ω1 ) + (ω1 + ω2 )2 − 4ω12 (8.103)

CH008.tex

11/3/2011

14: 40

Page 241

BIBLIOGRAPHY

241

Finally, it is possible to get the expression of E ◦ appearing in Eq. (8.93) by transforming this equation using Eqs. (8.95) and (8.96), that is, H = E ◦ +  1 ((a1† cosh ϕ + a2 sinh ϕ )(a1 cosh ϕ + a2† sinh ϕ)) +  2 ((a1 sinh ϕ − a2† cosh ϕ)(a1† sinh ϕ − a2 cosh ϕ)) or, due to the commutation rules (8.92), one obtains H = E ◦ +  1 (a1† a1 cosh2 ϕ + (a2† a2 + 1) sinh2 ϕ + (a1† a2† + a1 a2 ) sinh ϕ cosh ϕ) +  2 ((a1† a1 + 1) sinh2 ϕ + a2† a2 cosh2 ϕ − (a1† a2† + a1 a2 ) sinh ϕ cosh ϕ) a result that simplifies to H = E ◦ + ( 1 +  2 ) sinh2 ϕ +  1 ( cosh2 ϕ + sinh2 ϕ)a1† a1 +  2 ( cosh2 ϕ + sinh2 ϕ)a2† a2 + ( 1 −  2 ) sinh ϕ cosh ϕ(a1† a2† + a1 a2 ) Finally, equating this last expression to that of H given by Eq. (8.91), leads for the coefficients of a1† a1 , a2† a2 , a1† a2† , and a1 a2 the following results: ωk =  k ( cosh2 ϕ + sinh2 ϕ)

with

k = 1, 2

ω12 = ( 1 −  2 ) sinh ϕ cosh ϕ and also to the conclusion that E ◦ + ( 1 +  2 ) sinh2 ϕ = 0 a result allowing one to get

8.4

E◦

(8.104)

appearing in the diagonal Hamiltonian (8.93).

CONCLUSION

This chapter was devoted to various questions related to the notion of phase for quantum oscillators, which is not without connection with the squeezed states susceptible to be obtained through the action of squeezing operators having the structure of translation operators in which the ladder operators have been replaced by their squared expression. It ended by the Bogoliubov–Valatin transformation allowing one to diagonalize the Hamiltonian of coupled oscillators via terms quadratic in the raising and lowering operators. Even though many results will not be used later, they are nevertheless important in many studies lying beyond the scope of the present work particularly in quantum optics.

BIBLIOGRAPHY A. S. Davydov. Quantum Mechanics, 2nd ed. Pergamon Press: Oxford, New York, 1976. J. R. Klauder and B. Skagerstam. Coherent States. World Scientific: Singapore, 1985. R. Loudon. The Quantum Theory of Light, 3rd ed. Oxford University Press: Oxford, 2000. W. H. Louisell. Quantum Statistical Properties of Radiation. Wiley: New York, 1973.

CH009.tex

PART

16/3/2011

13: 27

Page 243

III

ANHARMONICITY

CH009.tex

16/3/2011

CHAPTER

13: 27

Page 245

9

ANHARMONIC OSCILLATORS

INTRODUCTION In a previous chapter, we studied the general properties of single-degree-of-freedom quantum harmonic oscillators for which it was possible to solve the Hamiltonian eigenvalue equation and thus to get the energy levels. The calculation was also made for the driven harmonic oscillator by diagonalization of its Hamiltonian through canonical transformations, using the translation operators. Then, it was seen that it is possible to reproduce numerically the exact Hamiltonian eigenvalues obtained by this last procedure, after diagonalization of the truncated matrix representation of the driven Hamiltonian in the basis of the eigenkets of the harmonic oscillator Hamiltonian. The aim of the present chapter is to find the energy levels of various anharmonic oscillators of interest for which it is not possible to diagonalize the Hamiltonian, using this numerical procedure for the driven harmonic oscillator. We shall first study the energy levels of oscillators for which the harmonic potential is perturbed by a cubic term. Second, we shall consider oscillators in a Morse potential, a physical model that applies to the vibrational behavior of diatomic molecules. Finally, we shall consider particles in a double-well potential, a model that applies to the inversion of ammonia for which tunneling may proceed through the barrier potential between the two wells. However, before commencing these studies, it may be of interest to find how quantum mechanics predicts the form of the anharmonic potentials in which the nuclei of diatomic molecules move.

9.1

MODEL FOR DIATOMIC MOLECULE POTENTIALS

For this purpose, we shall introduce in this section a very crude model applying to the H+ 2 molecular ion, which is the most simple diatomic molecule, since it involves only one electron and two protons. We shall attempt to simplify the notations, using the centimeter–gram–second (cgs) system. The average kinetic energy T  of the electron belonging to the molecular ion H+ 2 may be approximated by that of the 1D particle-in-a-box model for which, Quantum Oscillators, First Edition. Olivier Henri-Rousseau and Paul Blaise. © 2011 John Wiley & Sons, Inc. Published 2011 by John Wiley & Sons, Inc.

245

CH009.tex

246

16/3/2011

13: 27

Page 246

ANHARMONIC OSCILLATORS

according to Eq. (4.28), the 1D energy levels are given by Enx = nx2

h2 8max2

with

nx = 1, 2, . . .

where m is the mass of the particle and ax the dimension of the 1D box. We shall assume that this kinetic energy is given by the 1D ground state of this model, that is, T  =

h2 8max2

Moreover, in the present situation, the dimension ax of the box may be viewed roughly as the distance R between the two protons. For the electronic ground state, this leads us to write the average kinetic energy T : AT h2 with A = (9.1) T R2 8m where m is now the mass of the single electron. On the other hand, the average Coulombic potential energy V  of the ion is the sum of the average attraction energies V1  and V2  between the single electron and the two protons and of the repulsion energy between the two protons, which is inversely proportional to R, that is, T  =

e2 (9.2) R where e is the elementary electrical charge. The two average attraction energies V1  and V2  must be the same for symmetry reasons. For each of them, one may roughly assume that they are proportional to the inverse of the average distance between the electron and the nuclei, and more crudely that this average distance is proportional to half the distance R, leading us to write V  = V1  + V2  +

V1  = V2  = −

e2 R/2

Hence, the average potential energy (9.2) becomes V  = −4

e2 e2 + R R

or AV with AV = 3e2 (9.3) R Then, for this crude linear model, the electronic energy E of the molecular ion is simply the sum of T  and V , given, respectively, by Eqs. (9.1) and (9.3):   AT AV E = − (9.4) R2 R V  = −

The evolution of E with respect to R given by Eq. (9.4) is reproduced in Fig. 9.1. Inspection of this figure shows, as expected, a minimum of the energy function (9.4), which appears to be the result of a compromise between the positive

CH009.tex

16/3/2011

13: 27

Page 247

9.1

Energy (eV)

10

MODEL FOR DIATOMIC MOLECULE POTENTIALS

247

T

R (Å) 2

4

10

6

8

V E

20 Figure 9.1 Total energy of the molecular ion H+ 2 as a compromise between a repulsive electronic kinetic energy and an attractive potential energy. Energies are in electron volts and distances in Ångström.

kinetic energy, which is decreasing in R, and the negative potential energy, which is correlatively increasing. By Taylor expansion of the energy E, denoted E(R) near its minimum, corresponding to the internuclear distance R = R◦ of the energy curve yields 

   ∂E 1 ∂2 E ◦ (R − R ) + (R − R◦ )2 E(R) = (E)R◦ + ∂R R=R◦ 2! ∂R2 R=R◦     1 ∂3 E 1 ∂4 E ◦ 3 + (R − R ) + (R − R◦ )4 + · · · (9.5) 3! ∂R3 R=R◦ 4! ∂R4 R=R◦ Of course, at the minimum of the energy function, the first derivative is zero, that is, 

∂E ∂R

 =0

(9.6)

Re

Hence, taking as energy reference (E)Re = 0

at

R = Re

and near the minimum, the Taylor expansion (9.5) reads E(R) =

1 1 1 ke (R − Re )2 + ge (R − Re )3 + je (R − Re )4 + · · · 2! 3! 4!

(9.7)

CH009.tex

248

16/3/2011

13: 27

Page 248

ANHARMONIC OSCILLATORS

with, respectively,

 ke =  ge =  je =

∂2 E ∂R2 ∂3 E ∂R3 ∂4 E ∂R4

 (9.8) R=Re



(9.9) R=Re



(9.10) R=Re

In order to get the equilibrium distance Re , we differentiate the energy function E(R) (9.4) with respect to R:   ∂E 2AT AV (9.11) =− 3 + 2 ∂R R R Thus, for the equilibrium distance Re for which Eq. (9.6) is verified, Eq. (9.11) yields 2AT AV − 3 + 2 =0 Re Re so that the equilibrium distance appears to be given by AT Re = 2 (9.12) AV or, owing to Eqs. (9.1) and (9.3),  2  h Re = (9.13) 12me2 and so, inserting numerical values, Re = 1.73 × 10−8 cm = 1.73 Å Now, due to Eq. (9.13) and owing to Eqs. (9.8) and (9.11), the constant ke reads 6AT 2AV ke = 4 − 3 Re Re or, using Eq. (9.12) for Re ,       AV 4 AV 3 ke = 6 AT − AV 2AT 2AT and, after rearranging and simplifying

  1 (AV )4 ke = (9.14) 8 (AT )3 In a similar way, one may obtain for the constants ge and je defined by Eq. (9.9) and (9.10) the following result:   3 (AV )5 (9.15) ge = − 8 (AT )4   9 (AV )6 je = (9.16) 8 (AT )5

CH009.tex

16/3/2011

13: 27

Page 249

9.1

MODEL FOR DIATOMIC MOLECULE POTENTIALS

249

Now, introduce the well-known dimensionless fine structure constant α and the Compton wavelength λc defined, respectively, in the CGS system by e2 1 e2 = 2π = c hc 137   h λc = = 2.43 × 10−10 cm with c, the light velocity mc α=

so that



λc α

 =

h2 h c = mc 2πe2 2πme2

(9.17) (9.18)

(9.19)

Then, due to Eq. (9.19), the equilibrium distance (9.13) may be approximated by   λc π λc  0.5 (9.20) Re = 6 α α Proceeding in a similar way for ke , ge , and je , defined by Eqs. (9.14)–(9.16), we have  2  2 α α 2 2 2 2 ke = 3.33α (mc )  3α (mc ) >0 (9.21) λc λc  3  3 α α 2 2 2 2 ge = 38.11α (mc )  −3 × 12α (mc ) > E1◦

E >> E2◦

and

we see that the differences (ETot − E1◦ ) and (ETot − E2◦ ) are very small, so that the entropies appearing in Eq. (12.80) are very near their values when E1◦ and E2◦ are vanishing. That allows us to truncate up to first order the Taylor expansion of the entropies involved in Eq. (12.80), that is, to write   ◦ ◦ ∂S S(ETot − Ek ) = S(ETot ) − Ek with k = 1, 2 (12.81) ∂E E=ETot Besides, owing to Eq. (12.73), it is possible to relate the partial derivative of the entropy with respect to the energy, to the absolute temperature T , via the thermodynamic relation   ∂S 1 = ∂E E=ETot T Then, Eq. (12.81) reads S(ETot − Ek◦ ) = S(ETot ) −

Ek◦ T

with

k = 1, 2

Hence, the probability ratio (12.80) becomes    S(ETot )/kB  −E ◦ /kB T    W (E1◦ ) g(E1◦ ) e 1 e = ◦ W (E2◦ ) g(E2◦ ) eS(ETot )/kB e−E2 /kB T or, after simplification, 

W (E1◦ ) W (E2◦ )



 =

◦

e−E1 /kB T ◦ e−E2 /kB T



g(E1◦ ) g(E2◦ )

 (12.82)

Of course, when the degeneracies corresponding to the two situations are unity, this last equation reduces to    −E ◦ /kB T  W (E1◦ ) e 1 (12.83) = ◦ ◦ W (E2 ) e−E2 /kB T

CH012.tex

358

11/3/2011

15: 12

Page 358

DENSITY OPERATORS FOR EQUILIBRIUM POPULATIONS OF OSCILLATORS

Moreover, since the probabilities are normalized, that is,  W (Ek◦ ) = 1

(12.84)

k

Eq. (12.82) implies that the probability for the system to have the energy Ei◦ is 1 −E ◦ /kB T )g(Ei◦ ) (e i Z

W (Ei◦ ) = with Z=



◦

(12.85)

(e−Ej /kB T )g(Ej◦ )

(12.86)

1 −E ◦ /kB T ) (e i Z

(12.87)

j

or, when the degeneracy is unity, W (Ei◦ ) = with Z=



◦

(e−Ej /kB T )

(12.88)

j

At last, keeping in mind Eq. (12.66), that is, Wi =

1 −βEi ) (e Z

(12.89)

and, by identification of Eqs. (12.85) and (12.89), it appears that the Lagrange parameter β is given by β=

12.5

1 kB T

(12.90)

CONCLUSION

This chapter has focused attention on the theoretical fact that, at statistical equilibrium, the statistical entropy is maximum. This was approached via the Boltzmann H-theorem, proving that statistical entropy must increase until equilibrium, and numerically verified with the model of Chapter 11 dealing with a large set of weakly coupled harmonic oscillators, which showed that after the statistical entropy has attained its maximum, the energy distribution of the oscillators obeys the Boltzmann law. Finally, using the entropy maximization at statistical equilibrium, it was then possible to get the microcanonical density operator and the Boltzmann canonical density operator, allowing to get the thermal average energy as a function of the partition function whence it is possible to normalize the density operator. This latter canonical density operator will be extensively used in the following chapters in order to study the thermal properties of harmonic oscillators.

CH012.tex

11/3/2011

15: 12

Page 359

BIBLIOGRAPHY

359

BIBLIOGRAPHY P. Blaise, Ph. Durand, and O. Henri-Rousseau. Physica A, 209 (1994): 51. B. Diu, C. Guthmann, D. Lederer, and B. Roulet. Eléments de physique statistique. Hermann: Paris, 1989. Ch. Kittel and H. Kroemer. Thermal Physics, 2nd ed. W. H. Freeman: 1980. H. Louisell. Quantum Statistical Properties of Radiations. Wiley: New York, 1973. F. Reif. Fundamentals of Statistical and Thermal Physics. McGraw-Hill: New York, 1965. F. Reif. Berkeley Physics Course, Vol. 5, Statistical Physics. McGraw-Hill: New York, 1967.

CH013.tex

11/3/2011

15: 31

Page 361

13

CHAPTER

THERMAL PROPERTIES OF HARMONIC OSCILLATORS INTRODUCTION Using the concepts encountered in the previous chapter, Chapter 13 is concerned with the thermal properties of oscillators and specially by thermal average energies, heat capacities, thermal fluctuations of energy, position, and momentum and thermal entropies. It ends by giving the detailed demonstration of the thermal average over Boltzmann density operators for harmonic oscillator, of very general functions of Boson operators, which admits as a special case the Bloch’s theorem dealing with the thermal average of the translation operator.

13.1 BOLTZMANN DISTRIBUTION LAW INSIDE A LARGE POPULATION OF EQUIVALENT OSCILLATORS Consider a set of N equivalent quantum harmonic oscillators with the same Hamiltonian   Hk = ω ak† ak + 21 In the following we shall suppose that N is very large, its magnitude being, for instance, Avogadro’s number. The eigenvalue equation of these Hamiltonians Hk is Hk |{n}k  = Ek◦ |{n}k  with, neglecting the same zero-point energies, Ek◦ = nk ω

(13.1)

Now, assume that this set of oscillators cannot exchange energy with the neigborhood, so that the total energy ETot of the set is constant and suppose that each oscillator may exchange energy with the other ones. In any configuration, among a multitude, the total energy ETot of the set is  Ek◦ Nk (13.2) ETot = k

Quantum Oscillators, First Edition. Olivier Henri-Rousseau and Paul Blaise. © 2011 John Wiley & Sons, Inc. Published 2011 by John Wiley & Sons, Inc.

361

CH013.tex

362

11/3/2011

15: 31

Page 362

THERMAL PROPERTIES OF HARMONIC OSCILLATORS

where Nk is the number of oscillators having the same eigenvalue energy Ek◦ defined by Eq. (13.1). Of course, the total number N of oscillators is the sum over the numbers Nk , that is,  NTot = (13.3) Nk k

Since the number of oscillators and the total energy are constant, one has, respectively, dETot = 0 Thus, Eqs. (13.2) and (13.3) lead to  Ek◦ dNk = 0

and

dNTot = 0 

and

k

dNk = 0

(13.4)

k

The statistical weight of a configuration corresponding to a situation where there are N1 oscillators having the energy E1 , N2 oscillators having the energy E2 , and so on is given by the statistical distribution NTot ! (13.5) W (N1 , N2 , . . . ) =  Nk ! k

where the Nk are constrained to verify simultaneously Eqs. (13.4). Figure 13.1 gives for a set of NTot = 21 oscillators, the values of W (N1 , N2 , . . . ) calculated by Eq. (13.5), subjected to the constraints of Eqs. (13.4), when applied to eight possible distributions of the total energy ETot = 21ω. Inspection of Fig. 13.1 shows that some configurations are more probable than others. The most probable is that corresponding to the situation where there are less and less oscillators when the energy increases. We shall now show that the most probable configuration is that corresponding to the situation where the number of oscillators having a given energy is exponentially decreasing with energy. Thus, we write Eq. (13.5) in logarithm form, that is,  ln W (N1 , N2 , . . . ) = ln (NTot !) − ln (Nk !) (13.6) k

Now, in order to find the most probable configurations, the differential of Eq. (13.6) must be zero, that is,   ∂ ln (Nk !)  dNk = 0 (13.7) d ln W (N1 , N2 , . . . ) = − ∂Nk k

Next, in order to take into account the two constraints (13.4) on the Nk eq , one must use the Lagrange multipliers method described in Section 18.7 leading one to write in place of Eq. (13.7) the following equation:     ∂ ln (Nk eq !)  eq eq dN − β E dN + α dNk eq = 0 − k k k ∂Nk eq k

k

k

Since this last expression must hold for each k, we see that they are as many following equations as they are of k:   ∂ ln (Nk eq !) + βEk − α dNk eq = 0 − (13.8) ∂Nk eq

CH013.tex

11/3/2011

15: 31

13.1

Ek

Page 363

BOLTZMANN DISTRIBUTION LAW INSIDE A LARGE POPULATION

Ek

Ek

363

Ek

{Nk}

{Nk}

{Nk}

{Nk}

0

0

0

1

0

0

1

0

0

2

1

0

1

0

0

0

2

0

2

3

3

4

0

2

5

3

4

1

10 W  9.8  109

12 W  3.7  108

13

14

W  1.7  108

W  4.9  109

Ek

Ek

Ek

Ek

W  1.2

{Nk}

{Nk}

{Nk}

{Nk}

0

3

0

0

0

0

1

0

0

0

0

0

0

0

0

5

7

0

5

0

0

0

0

0

0

0

0

1

14

18

15

15

 105

 103

 105

W  3.3  105

W  1.3

W  3.3

Figure 13.1 Values of W (N1 , N2 , . . . ) calculated by Eqs. (13.5) and for NTot = 21, ETot = 21ω, for eight different configurations verifying Eqs. (13.4). For each configuration, the eight lowest energy levels Ek of the quantum harmonic oscillators are reproduced, with for each of them, as many dark circles as they are (Nk ) of oscillators having the corresponding energy Ek .

In order to calculate the partial derivative of Eq. (13.6) with respect to Nk eq , it is convenient, if the numbers N and Nk eq are very large, to use the Stirling approximation ln (Nk eq !)  Nk eq ln (Nk eq ) − Nk eq Then, the partial derivative of Eq. (13.6) of interest reads   ∂ ln (Nk eq !)  ln (Nk eq ) ∂Nk eq Hence, Eq. (13.8) is (−ln Nk eq − βEk + α)dNk eq = 0 Moreover, since dNk eq  = 0, it yields −ln Nk eq − βEk + α = 0

CH013.tex

364

11/3/2011

15: 31

Page 364

THERMAL PROPERTIES OF HARMONIC OSCILLATORS

so that Nk eq = eα e−βEk

(13.9)

It is this distribution that is the closest to the one described by the configuration of Fig. 13.1 corresponding to the situation leading to W = 9.8 × 109 and where N0 = 10 for E0 = 0, N1 = 5 for E1 = 1, N2 = 3 for E2 = 2, N3 = 2 for E3 = 3, N4 = 1 for E4 = 4, and Nk = 0 for the higher levels. The expression for the Lagrange multiplier α may be obtained by aid of Eqs. (13.3) and (13.9) yielding  NTot = eα e−βEk k

so that eα = where Z is the partition function: Z=

N Tot Z



(13.10)

e−βEk

k

As a consequence, the Lagrange parameter α appears to be   NTot α = ln Z Moreover, with the help of Eq. (13.10), Eq. (13.9) becomes NTot −βEk Nk eq = e Z or Nk eq = NTot W eq (Ek ) where W eq (Ek ) is the Boltzmann probability to find oscillators having the energy Ek , which is given by W eq (Ek ) =

e−βEk Z

(13.11)

Recall that the value of the Lagrange parameter β appearing in the exponential and decreasing with the energy levels Ek has been found above to be given by Eq. (12.90).

13.2 THERMAL PROPERTIES OF HARMONIC OSCILLATORS 13.2.1

Canonical density operators of harmonic oscillators

Consider the canonical density operator ρB of a quantum harmonic oscillator defined by Eq. (12.61), that is, ρB =

1 −βH ) (e Z

(13.12)

CH013.tex

11/3/2011

15: 31

Page 365

13.2 THERMAL PROPERTIES OF HARMONIC OSCILLATORS

365

where Z is the partition function given by Eq. (12.62), that is, Z = tr{(e−βH )}

(13.13)

β is the thermal Lagrange parameter given by Eq. (12.90), that is, β=

1 kB T

(13.14)

and H is the Hamiltonian of the harmonic oscillator given by Eq. (5.9), that is, 

with [a, a† ] = 1 (13.15) H = ω a† a + 21 Owing to Eqs. (13.12) and (13.15), the canonical density operator of the harmonic oscillator reads ρB =

1 −βa† aω −βω/2 e ) (e Z

(13.16)

so that the partition function (13.13) yields Z = (e−βω/2 )tr{(e−βa

† aω

)}

(13.17)

Now, to perform the trace involved in this last equation, it is convenient to use the basis of eigenstates of a† a, that is, a† a|(n) = n|(n)

(n)|(m) = δnm

with

(13.18)

Hence, owing to Eq. (13.15), the partition function (13.13) takes the form  † Z = (e−βω/2 ) (n)|(e−βa aω )|(n) n

Expanding the exponential operator gives Z = (e−βω/2 )

 n

 (n)|

k

(−βω)k (a† a)k k!

 |(n)

Moreover, due to Eq. (13.18) one obtains by recurrence (a† a)k |(n) = nk |(n) so that Eq. (13.19) transforms to −βω/2

Z = (e

)

 n

k



(−βω)k nk (n)| k!



Hence, after coming back to the exponential  (n)|(e−βnω )|(n) Z = (e−βω/2 ) n

and using the normality property of the kets  (e−βnω ) Z = (e−βω/2 ) n

|(n)

(13.19)

CH013.tex

366

11/3/2011

15: 31

Page 366

THERMAL PROPERTIES OF HARMONIC OSCILLATORS

we have Z = (e−βω/2 )



yn

y = (e−βω )

with

(13.20)

n

Now, observe that at temperatures T , which are not very far from the room temperature, the following inequality is generally satisfied for harmonic oscillators describing molecular vibrations: ω > kB T so that, due to Eq. (13.14), βω > 1

e−βω < 1

and thus

In this special situation, the series involved in Eq. (13.20) is convergent and given by  1 yn = with y < 1 1− y n Hence, the partition function (13.20) becomes  −βω/2   −ω/2kB T  e e Z= = −β  ω 1− e 1 − e−ω/kB T

(13.21)

a result that may also be written   1 1 Z= = β  ω/2 −β  ω/2 e −e 2 sinh(ω/2) Moreover, the canonical density operator (13.16) becomes after simplification ρB = (1 − e−βω )(e−βa

† aω

)

(13.22)

a result that may be also written ρB = (1 − e−λ )(e−λa a ) †

(13.23)

and, comparing Eq. (13.14), λ=

ω = βω kB T

(13.24)

13.2.2 Thermal energy Now, consider the mean thermal average energy of a quantum harmonic oscillator that is the average of the Hamiltonian over the canonical density operator, that is, H = tr{ρB H}

(13.25)

which, due to Eqs. (13.12) and (13.23), reads either

 † H = ω(1 − e−λ ) tr (e−λa a ) a† a + 21

(13.26)

or, due to Eq. (13.22), H = ω(1 − e−βω ) tr{(e−βa

† aω

)a† a} +

ω 2

(13.27)

CH013.tex

11/3/2011

15: 31

Page 367

13.2 THERMAL PROPERTIES OF HARMONIC OSCILLATORS

367

However, observe it is unnecessary to separately calculate the partition function and the trace involved in Eq. (13.27), since it has been shown that the thermal average energy (13.25) of a system whatever its Hamiltonian may be, is given by Eq. (12.69), that is,   ∂ ln Z (13.28) H = − ∂β so that it is possible to get the thermal average value of the energy (13.25) using Eq. (13.28). Hence, start from Eq. (13.21) giving ln Z, that is, ω − ln (1 − e−βω ) 2 so that, by differentiation, one obtains     ∂ ln Z ω ω e−βω =− − ∂β 2 1 − e−βω ln (Z) = −β

or, after rearranging,



∂ ln Z ∂β



 =−

ω ω + β  ω e −1 2



Thus, comparing Eq. (13.14), the thermal average energy (13.28) becomes   ω ω H = (13.29) + 2 eω/kB T − 1 which is the Planck expression of the average energy of a quantum oscillator belonging to a population of quantum harmonic oscillators in thermal equilibrium. Of course, the total average energy of a population of N oscillators is HTot  = N H

(13.30)

Moreover, by comparison of Eqs. (13.27) and (13.29), it yields (1 − e−βω )tr{(e−βωa a )a† a} = †

1 eβω − 1

(13.31)

1 eλ − 1

(13.32)

or (1 − e−λ )tr{(e−λa a )a† a } = †

13.2.3

Boltzmann distribution

Now, consider the diagonal matrix elements of the density operator as given by Eq. (12.85), that is, Pn = (n)|ρB |(n) Then, comparing Eq. (13.12), the right-hand-side matrix elements read (n)|ρB |(n) =

1 (n)|(e−βH )|(n) Z

(13.33)

CH013.tex

368

11/3/2011

15: 31

Page 368

THERMAL PROPERTIES OF HARMONIC OSCILLATORS

or, due to Eq. (13.22), (n)|ρB |(n) = (1 − e−βω )(n)|(e−βωa a )|(n) †

Hence, after using the eigenvalue equation a† a|(n) = n|(n) the matrix elements become (n)|ρB |(n) = (1 − e−βω )(n)|e−nβω |(n) or (n)|ρB |(n) = (1 − e−βω )(e−nβω ) Hence, Eq. (13.33) yields Pn = (1 − e−βω )(e−nβω )

(13.34)

This last result, which is the Boltzmann distribution of the energy level of harmonic oscillators, that is, the probability for them to be occupied at any temperature, may be put in correspondence with the result (12.28) obtained in the coarse-grained analysis where an exponential decreasing with energy of the probability occupation appears.

13.2.4 Thermal average of the occupation number Now, observe that, due to Eq. (13.31), and since the occupation number is defined by n ≡ a† a

(13.35)

it appears that its thermal average is n =

1 eβω − 1

(13.36)

Next, comparing Eq. (13.36), 1 + n = 1 +

1 eβω − 1

=

eβω eβω − 1

the ratio of n and 1 + n yields n = e−βω 1 + n Besides, from Eq. (13.37) it reads

(13.37)

n 1 = 1 + n 1 + n while the nth power of (13.37) takes the form  n n (e−nβω ) = 1 + n (1 − e−βω ) = 1 −

so that it results from Eqs. (13.39) and (13.38) that −βω

(1 − e

)(e

−nβω

1 )= 1 + n



n 1 + n

(13.38)

(13.39) n

CH013.tex

11/3/2011

15: 31

Page 369

13.2 THERMAL PROPERTIES OF HARMONIC OSCILLATORS

369

Hence, the Boltzmann distribution function (13.34) becomes Pn =

nn (1 + n)n+1

(13.40)

a result that is widely used in the area of the theory of lasers.

13.2.5

Heat capacity

Now, consider the thermal capacity at constant volume Cv which is, by definition, the time derivative of the total average energy of a population of N oscillators:   ∂HTot (T ) Cv = (13.41) ∂T v where HTot is given by Eq. (13.30) so that   ∂H(T ) Cv = N ∂T v Then, due to Eq. (13.29), Eq. (13.41) reads   ∂ 1 Cv = Nω ∂T eω/kB T − 1 and thus, on differentiation



Cv = Nω



−1 (eω/kB T − 1)2

ω/kB T

e



−ω kB T 2



or  Cv = NkB

ω kB T

2

eω/kB T − 1)2

(eω/kB T

(13.42)

Figure 13.2 discusses the evolution with temperature of the thermal capacity Cv for a mole of oscillators of angular frequency ω = 1000 cm−1 .

13.2.6 Thermal fluctuations 13.2.6.1 Thermal energy fluctuation Now, the thermal fluctuation of the energy of N oscillators is  ETot = HTot 2  − HTot 2 (13.43) with HTot 2  = N 2 H2  Thus Eq. (13.30), becomes

 ETot = N H2  − H2

CH013.tex

15: 31

Page 370

THERMAL PROPERTIES OF HARMONIC OSCILLATORS

CV (T ) (R units)

370

11/3/2011

0

500

1000

1500

2000

T (K) Figure 13.2 Thermal capacity Cv in R units for a mole of oscillators of angular frequency ω = 1000 cm−1 .

Recall that the thermal average of the Hamiltonian may be obtained by Eq. (12.69), that is,   1 ∂Z H = − (13.44) Z ∂β Now, the thermal average of H2 may be found from H2  = tr{ρB H2 } so that, due to Eq. (13.12), we have 1 (13.45) tr{(e−βH )H2 } Z Next, observe that the product of operators appearing under the trace may be written  2 −βH  ∂ e −βH 2 (13.46) )H = (e ∂β2 H2  =

so that Eq. (13.45) reads

 2    1 ∂ −βH tr e Z ∂β2 or, since the partial derivative commutes with the trace operation,   1 ∂2 2 tr{(e−βH )} (13.47) H  = Z ∂β2 Again, since the partition function is given by Eqs. (13.13) and (13.14), that is, 1 β= (13.48) Z = tr{(e−βH )} and kB T Eq. (13.47) reads   2  1 ∂ Z 2 (13.49) H  = Z ∂β2 H2  =

CH013.tex

11/3/2011

15: 31

Page 371

13.2 THERMAL PROPERTIES OF HARMONIC OSCILLATORS

371

Now, observe that the following equation is satisfied:        2    ∂ 1 ∂Z 1 ∂ Z ∂ 1 ∂Z = + ∂β Z ∂β ∂β Z ∂β Z ∂β2 which yields



∂ ∂β



1 ∂Z Z ∂β

 =−



1 Z2

∂Z ∂β



∂Z ∂β



 +

1 Z



∂2 Z ∂β2



Hence, equating the last right-hand side of this last equation and the right-hand side of Eq. (13.49) leads to      ∂ 1 ∂Z 1 ∂Z 2 2 + 2 H  = ∂β Z ∂β Z ∂β or, because of Eq. (13.44), to



H2  = −

∂H ∂β

 + H2

(13.50)

Hence, the thermal energy fluctuation (13.43) is    ∂H ETot = N − ∂β or ETot

    ∂H ∂T =N − ∂T ∂β

and thus, due to the definition (13.41) of the heat capacity at constant volume Cv ,    Cv ∂T (13.51) ETot = N − N ∂β Again, owing to Eq. (13.14) leading to T=

1 kB β

(13.52)

the partial derivative of the absolute temperature with respect to β reads       ∂ 1 1 ∂T = =− ∂β ∂β kB β k B β2 and thus, thanks to (13.52),



∂T ∂β

 = −kB T 2

Thus, owing to this result, and to the expression (13.42) for the heat capacity, Eq. (13.51) leads to    √ eω/kB T ω 2 kB T 2 E Tot = N kB kB T (eω/kB T − 1)2

CH013.tex

372

11/3/2011

15: 31

Page 372

THERMAL PROPERTIES OF HARMONIC OSCILLATORS

or, after simplification, √ Nω

eω/2kB T − 1) Besides, keeping in mind that, due to Eq. (13.29), and when the zero-point energy is ignored, the thermal average (13.30) reduces to ω HTot  = N ω/k T B − 1) (e the relative energy fluctuation becomes E Tot 1 = √ eω/2kB T HTot  N At high temperature, the argument of the exponential being very small, the relative fluctuation reduces to ETot 1 →√ HTot  N E Tot =

(eω/kB T

It must be emphasized that the inverse dependence of the relative fluctuation with respect to the number N of oscillators is the same as that of (12.39) yet encountered in the previous section, dealing with a coarse-grained analysis of a large set of coupled harmonic oscillators. 13.2.6.2 Thermal number occupation fluctuation Starting from Eq. (13.50), that is,   ∂H H2  = − (13.53) + H2 ∂β and passing to Boson operators using Eq. (5.9), reads        1 ∂(a† a + 21 ) 1 2 1 2 † † a a+ =− + a a+ 2 ω ∂β 2 Now, when the zero-point energy is ignored, Eq. (13.53) remains true so that it is possible to write   1 ∂ a† a  (a† a)2  = − + a† a2 ω ∂β or, due to Eq. (13.35), 1 n  = − ω



2

Hence, comparing Eq. (13.36), that is, n = Eq. (13.54) reads 1 n  = − ω 2



∂ ∂β





∂n ∂β

 + n2

(13.54)



1

(13.55)

eβω − 1 1

eβω − 1



 +

1 eβω − 1

2 (13.56)

CH013.tex

11/3/2011

15: 31

Page 373

13.2 THERMAL PROPERTIES OF HARMONIC OSCILLATORS

373

Next, evaluating the partial derivative of the first right-hand-side term leads to      ∂ 1 eβω = −ω ∂β eβω − 1 (eβω − 1)2 so that Eq. (13.56) simplifies to



n2  =

eβω + 1 (eβω − 1)2



or n2  =

eβω − 1 + 2 (eβω − 1)2

and thus n2  =

1 2 + eβω − 1 (eβω − 1)2

Hence, comparing Eq. (13.55), we have n2  = n + 2n2

(13.57)

Now, by definition of the n thermal fluctuation  n = n2  − n2 and with Eq. (13.57) this fluctuation becomes  n = n2 + n

(13.58)

a result that is widely used in the area of the theory of lasers. Equation (13.58) may be also written  1 n = n 1 + n Then, when n >> 1 the argument of the square root may be expanded up to first order in 1/n according to  1 1 1+ 1+ n 2n so that in this limit n = n +

1 2

Hence, in this limit, the relative fluctuations read n 1 1+ 1 n 2n

CH013.tex

11/3/2011

374

15: 31

Page 374

THERMAL PROPERTIES OF HARMONIC OSCILLATORS

13.2.6.3 Thermal average of Q, Q2 , and the potential We now consider the thermal equilibrium value of the position operator Q and of its square Q2 , which are given, respectively, by the following thermal average over the Boltzmann density operator ρB : Q(T ) = tr{ρB Q}

and

Q(T )2  = tr{ρB Q2 }

(13.59)

Recall that within the raising and lowering operators picture of oscillators, Q is given by Eq. (5.6), that is,   Q= (13.60) (a† + a) 2mω whereas the Boltzmann density operator is given by Eqs. (13.23) and (13.24): ρB = (1 − e−λ )(e−λa a ) †

(13.61)

Hence, the thermal average defined by the first equation of (13.59) is, therefore,   † −λ tr{(e−λa a )(a† + a)} Q(T ) = (1 − e ) 2mω Performing the trace over the eigenstates |{n} of a† a gives    † −λ {n}|(e−λa a )(a† + a)|{n} Q(T ) = (1 − e ) 2mω n

(13.62)

Moreover, since a† a is Hermitian a† a|{n} = n|{n} with {m}|{n} = δmn

(13.63)

the two following Hermitian conjugate eigenvalue equations involved in Eq. (13.62) are verified: (e−λa a )|{n} = (e−λn )|{n} †

and

{n}|(e−λa a ) = {n}|(e−λn ) †

(13.64)

Hence, the right-hand side of (13.62) becomes {n}|(e−λa a )(a† + a)|{n} = (e−λn ){n}|(a† + a)|{n} †

(13.65)

Again, owing to Eqs. (5.53) and of its Hermitian conjugate, that is, √ √ a|{n} = n|{n − 1} and thus {n}|a† = n{n − 1}| and due to the orthogonality (13.63), Eq. (13.65) becomes √ √ {n}|(a† + a)|{n} = n{n − 1}|{n} + n{n}|{n − 1} = 0 so that Eq. (13.65) transforms to {n}|(e−λa a )(a† + a)|{n} = 0 †

Hence, the equilibrium thermal average value (13.62) is zero.

(13.66)

CH013.tex

11/3/2011

15: 31

Page 375

13.2 THERMAL PROPERTIES OF HARMONIC OSCILLATORS

375

Next pass to the thermal average value of Q2 , which, according to Eqs. (13.60) and (13.61), is  † Q(T )2  = (1 − e−λ ) tr{(e−λa a )(a† + a)2 } 2mω Expanding the square using [a, a† ] = 1, gives  † † Q(T )2  = (1 − e−λ ) (tr{(e−λa a )(2a† a + 1)} + tr{(e−λa a )((a† )2 + (a)2 )}) 2mω (13.67) Now, observe that, owing to Eq. (13.32) the first right-hand-side term of Eq. (13.67) is (1 − e−λ )tr{(e−λa a )(2a† a + 1)} = (2n + 1) †

(13.68)

with n =

eλ

1 −1

(13.69)

Now, perform trace over the basis of the eigenstates of a† a appearing on the last right-hand-side term of Eq. (13.67)  † † tr{(e−λa a )((a† )2 + (a)2 )} = {n}|(e−λa a )((a† )2 + (a)2 )|{n} n

which, due to the last equation of (13.64), this trace reads  † (e−λn ){n}|((a† )2 + (a)2 )|{n} tr{(e−λa a )((a† )2 + (a)2 )} =

(13.70)

n

Then, using Eqs. (5.71) and its Hermitian conjugate leads to the following Hermitian conjugate linear transformations:   (a)2 |{n} = n(n − 1)|{n − 2} and {n}|(a† )2 = n(n − 1){n − 2}| and by orthogonality of the eigenstates of a† a, Eq. (13.70) gives   † tr{(e−λa a )((a† )2 + (a)2 )} = 2((e−λn ) n(n − 1)δn,n−2 ) n

and thus tr{e−λa a ((a† )2 + (a)2 )} = 0 †

(13.71)

Hence, comparing Eqs. (13.68) and (13.71), Eq. (13.67) becomes simply Q(T )2  =

 (2n + 1) 2mω

or, using Eq. (13.69) Q(T )2  =

 2mω



that is,  Q(T )  = 2mω 2

(13.72) 

2 +1 λ e −1 

1 + eλ eλ − 1



CH013.tex

376

11/3/2011

15: 31

Page 376

THERMAL PROPERTIES OF HARMONIC OSCILLATORS

Again, multiplying both numerator and denominator by the same quantity exp (−λ/2) to get   λ/2  e + e−λ/2 2 Q(T )  = (13.73) 2mω eλ/2 − e−λ/2 with

   λ/2  λ e + e−λ/2 coth = 2 eλ/2 − e−λ/2 Q(T )2  =

(13.74)

   λ coth 2mω 2

to obtain finally, by aid of Eq. (13.24), that is, λ= the following expression: Q(T )2  =

ω kB T

   ω coth 2mω 2kB T

(13.75)

(13.76)

Hence, the thermal average V(T ) of the potential operator V(T ) = 21 mω2 Q(T )2  becomes, comparing Eq. (13.76), V(T ) =

  ω ω coth 4 2kB T

(13.77)

(13.78)

Next, when the absolute temperature is such that kB T >> ω, so that, due to Eq. (13.75), λ ω  coth 2kB T 2ω so that, for this high-temperature limit, Eqs. (13.76) and (13.77) simplify to   2kB T kB T  Q(T )2   = (13.79) 2mω ω mω2 kB T (13.80) 2 In the case of very low temperatures, corresponding to ω >> kB T , due to (13.75), when λ >> 1  λ/2   λ/2  e + e−λ/2 e  1 −λ/2 λ/2 e −e eλ/2 V(T ) =

CH013.tex

11/3/2011

15: 31

Page 377

13.2 THERMAL PROPERTIES OF HARMONIC OSCILLATORS

the coth function reduces to unity, that is,   ω coth  1 when 2kB T

377

ω > kB T

thus, in the very low temperature limit, Eqs. (13.76) and (13.77) reduce to Q(0)2  =

 2mω

ω (13.81) 4 In Eq. (13.81), one may recognize the mean value of the potential of the harmonic oscillator averaged over the ground state |{0} of the harmonic oscillator Hamiltonian. Finally, the fluctuation of the position coordinate at any temperature T , which is defined by  Q(T ) = Q(T )2  − Q(T )2 V(0) =

becomes, in view of Eqs. (13.62), (13.66), and (13.76),      ω Q(T ) = coth 2mω 2kB T

(13.82)

13.2.6.4 Thermal average of P, P2 , and the kinetic operator In like manner as for Q(T ) given by Eq. (13.62), one would obtain for the thermal average of the momentum P(T ) = 0

(13.83)

and for the thermal average of the squared momentum, an expression similar to that (13.72) obtained for Q(T )2 , that is, in the present situation mω (2n + 1) 2

(13.84)

  ω mω coth 2 2kB T

(13.85)

P(T )2  = or, similarly to Eq. (13.76), P(T )2  =

Via this last expression, the thermal average value of the kinetic energy yields, respectively, for very high and very low temperatures is given by   ω ω T(T ) = coth (13.86) 4 2kB T T(T ) =

kB T 2

when

T(0) =

ω 4

kB T > ω

(13.87)

(13.88)

CH013.tex

378

11/3/2011

15: 31

Page 378

THERMAL PROPERTIES OF HARMONIC OSCILLATORS

so that, comparing Eqs. (13.83) and (13.85), the fluctuation of P(T ) is     mω ω P(T ) = coth 2 2kB T

(13.89)

13.2.6.5 Verification of the virial theorem Now, for harmonic oscillators, the thermal average of the kinetic and potential energies obey the virial theorem (2.89), and since we deal with mean values averaged over linear combinations of harmonic oscillator Hamiltonian eigenstates (which are necessarily stationary states), it is not surprising to find that Eqs. (13.78) and (13.86) also obey this theorem (2.89) since   ω ω coth (13.90) T(T ) = V(T ) = 4 2kB T Furthermore, since the thermal averaged Hamiltonian is the sum of the thermal average kinetic and potential operators, it follows from Eq. (13.90) that the form of the virial theorem (2.89) holds also for any temperature   ω ω coth H(T ) = 2T(T ) = 2V(T ) = 2 2kB T whereas, comparing Eqs. (13.80) and (13.87), its high-temperature limit is kB T 2 and, due to Eqs. (13.81) and (13.88), its low temperature yields T(T ) = V(T ) =

(13.91)

ω (13.92) 2 We remark that Eq. (13.92) is in agreement with the results (5.99) and (5.100) found for the average values of the kinetic and potential operators when the harmonic oscillator is in the ground state |{0} of its Hamiltonian. T(0) = V(0) =

13.2.6.6 Equipartition theorem The fact that at high temperatures the thermal average values of the potential and of the kinetic operators are equal and given by Eq. (13.91) is an illustration of the equipartition theorem of classical statistical mechanics, according to which the thermal energy is quadratic with respect to the independent variables and is kB T /2 for each degree of freedom. Now, we prove this theorem in a general way. Suppose that the energy E of the system is quadratic with respect to N classical different continuous independent variables qk , that is, E=

N 

Ek

with

E k = λk qk 2

(13.93)

k=1

For each energy term E k , its thermal average value may be obtained by Eq. (12.69):   ∂ ln Zk (13.94) E k (T ) = − ∂β

CH013.tex

11/3/2011

15: 31

Page 379

13.2 THERMAL PROPERTIES OF HARMONIC OSCILLATORS

379

where Zk is the partition function, which for continuous variable may be got from Eq. (12.67) by passing from the discrete sum to the corresponding integral according to +∞ 2 Zk = e−βλk qk dqk −∞

which yields after integration Zk = so that

1 2



π βλk

   1 1 π − ln β ln Zk = ln 2 λk 2

Hence, the thermal average (13.94) becomes E k (T ) =

1 kB T = 2β 2

that is the equipartition theorem of classical statistical mechanics. As a consequence of this result and due to Eq. (13.93), the thermal average of the total energy E(T ) is the number N of different independent variables times kB T /2: E(T ) = N

kB T RT = 2 2

where R is the ideal gas constant 13.2.6.7 Thermal Heisenberg uncertainty relation Now, we consider the thermal fluctuations of the position and momentum operators. Owing to Eqs. (13.82) and (13.89), the product of the thermal average of the uncertainty relation reads    ω P(T ) Q(T ) = coth 2 2kB T or, in view of the expression of the coth function,    eω/2kB T + e−ω/2kB T P(T ) Q(T ) = 2 eω/2kB T − e−ω/2kB T

(13.95)

When the absolute temperature approaches zero, the arguments of the decreasing exponential also narrow to zero. Thus, after simplification, one obtains the limit  when T → 0 2 As required by Eq. (5.96), this limit corresponds to the lowest Heisenberg uncertainty (5.97) obtained for the ground state of the harmonic oscillator. Also, when the absolute temperature is very large, that is, when kB T > ω, Taylor expansions of the exponentials the arguments of which are very small may be limited to first order, that is, P(T ) Q(T ) →

e±ω/2kB T = 1 ± ω/2kB T

CH013.tex

380

11/3/2011

15: 31

Page 380

THERMAL PROPERTIES OF HARMONIC OSCILLATORS

so that, for this high-temperature limit, Eq. (13.95) reduces to   2  P(T ) Q(T ) = 2 ω/2kB T or   kB T P(T ) Q(T ) = 2 ω

13.2.7

Coherent-state density operator at thermal equilibrium

13.2.7.1 Density operator from the Lagrange multipliers method We now determine the expression for the density operator of a coherent state at thermal equilibrium. Thus, it is convenient to work in the same way as when obtaining the canonical and microcanonical density operators (12.49) and (12.63) using the Lagrange multipliers method. Thus, consider a population of equivalent harmonic oscillators for which one knows the entropy and the average value of the Hamiltonian H of the position operator Q and of its conjugate momentum P. Then, the normalization condition of the density operator ρc , the expression of the statistical entropy S in terms of ρc , and the average values of H , Q, and P lead, respectively, to tr{ρc ln ρc } = S

(13.96)

tr{ρc } = 1

(13.97)

tr{ρc H} = H

(13.98)

tr{ρc Q} = Q

(13.99)

tr{ρc P} = P

(13.100)

Just as for Eq. (12.47), the equation dealing with the maximization dS = 0 of the statistical entropy S is tr{(1 + ln ρc )δρc } = 0 Moreover, due to the constraints linked to Eqs. (13.96)–(13.100), leading to tr{δρc } = 0

(13.101)

tr{Hδρc } = 0

(13.102)

tr{Qδρc } = 0

(13.103)

tr{Pδρc } = 0

(13.104)

one has, according to the Lagrange multipliers method, to multiply each of them by Lagrange multipliers according to λ0 tr{δρc } = 0

(13.105)

CH013.tex

11/3/2011

15: 31

Page 381

13.2 THERMAL PROPERTIES OF HARMONIC OSCILLATORS

381

β tr{Hδρc } = 0

(13.106)

λ1 tr{Qδρc } = 0

(13.107)

λ2 tr{Pδρc } = 0

(13.108)

where λ0 , β, λ1 , and λ2 are, respectively, the Lagrange parameters associated to the constraints (13.101)–(13.104). Next, collecting the constraints multiplied by the corresponding Lagrange multipliers, we have maximizing the statistical entropy tr{(1 + ln ρc + λ0 + βH + λ1 Q + λ2 P)δρc } = 0 Hence, since this last equation must be satisfied irrespective of the basis on which the trace is performed, we have 1 + ln{ρc } + λ0 + βH + λ1 P + λ2 Q = 0 or, by integration, ρc = e−(1+λ0 )−βH+λ1 Q+λ2 P or, since λ0 is a scalar, ρc = e−(1+λ0 ) e−(βH−λ1 Q−λ2 P)

(13.109)

where ω is the angular frequency of the oscillator and m its reduced mass. Again, express the position operator and its momentum conjugate and also the Hamiltonian in which the zero-point energy is ignored, in terms of the Boson operators according to    mω † and P=i (a† + a) (a − a) Q= 2mω 2 H = ω a† a so that the argument of the last exponential of the right-hand side of Eq. (13.109) is     † λ1 PQ + λ2 P = iλ2 mω (a − a) + λ1 (a† + a) 2mω 2mω or   λ1 PQ + λ2 P = {a† (λ1 + iλ2 mω) + a(λ1 − iλ2 mω)} 2mω Now, let λ = ωβ 1 α= λ so that



 (λ1 + iλ2 mω) 2mω

and

1 α = λ ∗



 (λ1 − iλ2 mω) 2mω

βH + λ1 Q + λ2 P = −λ(a† a + α∗ a + αa† )

CH013.tex

382

11/3/2011

15: 31

Page 382

THERMAL PROPERTIES OF HARMONIC OSCILLATORS

13.2.7.2 Some properties In terms of these new scalar and operator variables, the density operator (13.109) takes the form ρc = e−(1+λ0 ) e−λ(a

† a+αa† +α∗ a)

Next, in order to normalize, as required, the density operator, assume that e−(1+λ0 ) = (1 − e−λ )e−λ|α|

2

(13.110)

Then, the density operator, which will appear later to be normalized, reads ρc = (1 − e−λ )e−λ(a

† a+αa† +α∗ a+|α|2 )

or ρc = (1 − e−λ )e−λ(a

† +α)(a+α∗ )

(13.111)

Next, perform the following canonical transformation: A(α)ρc A(α)−1 = (1 − e−λ )A(α)e−λ(a

† +α)(a+α∗ )

A(α)−1

(13.112)

with ∗ a† −αa

A(α) = (eα

)

Next, due to Eqs. (7.9) and (7.10), which read A(α){f(a, a† )}A(α)−1 = {f(a−α∗ , a† − α)} Eq. (13.112) yields A(α)ρc A(α)−1 = (1 − e−λ )(e−λa a ) †

a result that, according to Eq. (13.22), is the Boltzmann density operator, leading to A(α)ρc A(α)−1 = ρB

(13.113)

Observe that, since the Boltzmann density operator is normalized, and since a canonical transformation does not modify the normalization, it appears that ρc has been, indeed, normalized by the assumption. Now, the coherent-state density operator reduces at zero temperature to the pure coherent-state density operator built up from a coherent state. For this purpose, inverse Eq. (13.113), so that ρc = A(α)−1 A(α)ρc A(α)−1 A(α) = A(α)−1 ρB A(α) or, due to Eq. (13.22), ρc = (1 − e−λ )A(α)−1 (e−λa a )A(α) †

Now, insert between the Boltzmann density operator and the translation operator a closure relation over the eigenstates of a† a, that is,  † |{n}{n}|A(α) ρc = (1 − e−λ )A(α)−1 (e−λa a ) n

Then, with the eigenvalue equation of

a† a,

the coherent-state density operator reads  ρc = (1 − e−λ )A(α)−1 (e−λn )|{n}{n}|A(α) n

CH013.tex

11/3/2011

15: 31

Page 383

13.2 THERMAL PROPERTIES OF HARMONIC OSCILLATORS

383

Next, if the temperature vanishes, λ which is given by Eq. (13.24), that is, λ=

ω kB T

becomes infinite, so that e−λ → 0 e−λn = e−nω/kB T → 0

if n  = 0

e−λn = e−nω/kB T = 1

if

n=0

Hence, the sum over the n eigenstates of a† a reduces to the ground state, so that the coherent density operator reduces to {ρc (T = 0)} = A(α)−1 |{0}{0}|A(α) or, comparing Eq. (6.92), {ρc (T = 0)} = |{α}{ ˜ α}| ˜ with a|{α} ˜ = −α|{α} ˜ Hence, when the absolute temperature vanishes, the density operator {ρc (T = 0)} reduces to a coherent state |{α} ˜ of eigenvalue −α, and so is the reason for its name.

13.2.8

Entropy of oscillators at thermal equilibrium

To get now an expression for the classical entropy of a population of oscillators at thermal equilibrium, which is the purpose of the present subsection, one has first to find an expression for the differential of the partition function in terms of the differential changes in the statistical parameter β and of the thermal average differential work dW . Hence, we first calculate dW  and start from the differential expression of the 1D mechanical work along the x abscissa, that is,   ∂E(x) dW = −F(x) dx with F(x) = − dx ∂x and thus, when the energy E(x) is quantized and defined by the energy levels En (x),   ∂En (x) dWn = dx ∂x the thermal average of the differential work is the average over the Boltzmann distribution of the different dWn , that is,    1  −βEn (x) ∂En (x) 1 dW  = e e−βEn (x) and β = dx with Zμ = Zμ n ∂x kB T n

CH013.tex

384

11/3/2011

15: 31

Page 384

THERMAL PROPERTIES OF HARMONIC OSCILLATORS

where Zμ is the partition function of a single oscillator. This expression may be also written   1  ∂e−βEn (x) dW  = − dx Zμ β n ∂x Moreover, since the sum and the partial derivative commute,     ∂Zμ 1 1 ∂  −βEn (x) 1 ∂ ln Zμ e dx = − dW  = − dx = − dx Zμ β ∂x n Zμ β ∂x β ∂x (13.114) Next, the total differential of ln Zμ (x, β) viewed as a function of the independent variables x and β reads     ∂ ln Zμ ∂ ln Zμ dx + dβ (13.115) d{ln Zμ (x, β)} = ∂x ∂β The thermal average Hamiltonian H, that is, the thermal energy, is given by Eq. (12.69):   ∂ ln Zμ (13.116) H = − ∂β Hence, due to Eqs. (13.114) and (13.116), the total differential (13.115) yields d{ln Zμ (x, β)} = −βdW  − Hdβ or d{ln Zμ (x, β)} = −βdW  − d{Hβ} + βdH and thus d{ln Zμ + (Hβ)} = β{dH − dW }

(13.117)

Then, recognizing in the difference between dH and dW  the differential heat exchange dQ, and using for β, Eq. (13.14), Eq. (13.117) reads  H dQ = d ln Zμ + kB T kB T Now, multiplying both terms of this last equation by the Boltzmann constant kB and recognizing on the left-hand side the thermodynamical expression of the differential entropy dS, this expression becomes  H dQ = dS kB d ln Zμ + = T kB T Hence, the canonical entropy takes the form S = k B ln Zμ +

H T

(13.118)

Equation (13.118) holds for one particle at thermal equilibrium, Zμ and H being, respectively, the partition function and the thermal average of this single particle. For N particles and because the partition function is the sum over exponentials,

CH013.tex

11/3/2011

15: 31

Page 385

13.2 THERMAL PROPERTIES OF HARMONIC OSCILLATORS

385

the partition function Z must be the Nth power of Zμ . However, since the particles are indistinguishable, according to Chapter 2 because of the Heisenberg uncertainty relations, this power must be divided by N! in order to avoid redundancies due to indistinguishable situations. Therefore, for N particles, Eq. (13.118) becomes ◦

ln (Zμ )nN nN ◦ H (13.119) + with N = nN ◦ ◦ (nN )! T where N ◦ is the Avogadro number and n the number of moles. Again, after using, respectively, for Zμ and E, Eqs. (13.21) and (13.29), the entropy (13.119) yields   ω/2k B T   R (e ) ω 1 1 ◦ S=n (13.120) ln +N + (nN ◦ )! T eωk B T − 1 2 1 − eω/k B T S = kB

where R is the ideal gas constant.

13.2.9

Oscillator Helmholtz potential

In thermodynamics, the Helmholtz thermodynamic potential is defined by F = U − TS where U is the internal energy. Then, for a population of oscillators, one may assimilate U to the oscillator thermal energy, and thus it is possible to write U = H so that, using for the entropy Eq. (13.118) the thermodynamic potential reads after simplification F = −k B T ln Zμ

(13.121)

where it must be remembered that the partition function Zμ is related to the Boltzmann density operator via Eq. (13.13), that is, 1 Zμ = tr{e−βH } with β = kBT Hence, it appears from Eq. (13.121) that e−βF = Zμ = tr{e−βH }

(13.122)

For a population of harmonic oscillators in thermal equilibrium, Eq. (13.122) reads with the help of Eq. (13.21) e−λ/2 ω with λ = 1 − e−λ kBT so that the thermodynamic potential yields 1 λ F = ln(1 − e−λ ) + β 2β or ω F = k B T ln(1 − e−ω/k B T ) + 2 e−βF =

(13.123)

CH013.tex

386

11/3/2011

15: 31

Page 386

THERMAL PROPERTIES OF HARMONIC OSCILLATORS

13.2.10

Anharmonic oscillators dilatation with temperature

The dilatation of a solid with temperature is a well-known physical observation. This thermal dilatation is a result of anharmonicity we desire to treat here, where the dilation with temperature will be obtained in a numerical way for 1D oscillator. Hence, consider the thermal average value of the Q coordinate of an anharmonic oscillator performed over the Boltzmann density operator. First, the Hamiltonian of an anharmonic oscillator is given by

 H = ω a† a + 21 + λω(a† + a)3 Its eigenvalue equation is H| k  = Ek | k 

(13.124)

with  k | l  = δkl For a given value of λ, this equation may be numerically solved working in the basis where a† a is diagonal. In this basis, the expansion of the eigenkets of H is given by  | k  = Ckn |{n} with a† a |{n} = n|{n} (13.125) n

The thermal average of the Q coordinate is Q(T ) = tr{ρB Q} where the Boltzmann density operator is given by Eq. (13.13) ρB =

1 −βH ) (e Z

with β =

1 kBT

and

Z = tr{e−βH }

and where Q is given in terms of the Boson operators by Eq. (5.6), that is,   (a† + a) Q= 2mω Writing explicitly the thermal average of Q over the basis where the Hamiltonian H is diagonal gives    1  k |(e−βH )(a† + a)| k  Q(T ) = Z 2mω k

Then, according to Eq. (13.124), the action on the bra of the exponential operator gives    1   Ek  k |(a† + a)| k  Q(T ) = exp − Z 2mω kBT k

Next, introduce after (a† + a) the closure relation built on the eigenstates of a† a to get      Ek 1 exp −  k |(a† + a)|{n}{n}| k  Q(T ) = Z 2mω k T B n k

CH013.tex

11/3/2011

15: 31

Page 387

13.2 THERMAL PROPERTIES OF HARMONIC OSCILLATORS

or 1 Q(T ) = Z



387

    Ek exp − Cnk  k |(a† + a)|{n} 2mω k T B n k

with Cn,k = {n}| k 

(13.126)

Again, using the result of the action of a and a† on an eigenket of a† a leads according to Eqs. (5.53) and (5.63) to    1   Ek Q(T ) = exp − Z 2mω kBT n k √  √ × Cnk n + 1 k |{n + 1} + n k |{n − 1} or, using in turn Eq. (13.126), the orthogonality of the eigenkets of a† a and the result of the action of a and a† on an eigenket of a† a leads to    √ √ 1   Ek Q(T ) = exp − Cnk Ck,n+1 n + 1 + Ck,n−1 n Z 2mω kBT n k (13.127) Equation (13.127) allows one to compute the variation with temperature of the average value of the elongation of the anharmonic oscillator from the knowledge of the H eigenvalues Ek and of the expansion coefficients Ckn of the corresponding eigenvectors. Figure 13.3 gives the temperature evolution of Q(T ) calculated in this way by the aid of Eq. (13.127) from the Ek and Ckn computed with the help of Eqs. (9.50) and (9.51). 〈Q(T )〉 2 mω

units 0.01

0.005

0.00

200

400 T (K)

600

800

√ Figure 13.3 Temperature evolution of the elongation Q(T ) (in Q◦◦ = /2mω units) of an anharmonic oscillator. Anharmonic parameter β = 0.017ω; number of basis states 75.

CH013.tex

388

11/3/2011

15: 31

Page 388

THERMAL PROPERTIES OF HARMONIC OSCILLATORS

In three dimensions, the cube of Eq. (13.127) allows one to obtain the temperature dependence of the dilatation of a solid modelized by a 3D anharmonic oscillator. Observe that, according to Eqs. (13.62) and (13.66), which hold when the anharmonicity of the oscillator is missing, the average value of Q is zero for all quantum numbers so that the thermal average Q(T ) vanishes whatever the temperature.

13.3 HELMHOLTZ POTENTIAL FOR ANHARMONIC OSCILLATORS Consider the Hamiltonian of an anharmonic oscillator of the form H = H◦ + V

(13.128)

H◦

is the Hamiltonian of the harmonic oscillator and V the anharmonic where Hamiltonian perturbation. Then, according to Eq. (13.12) the unnormalized Boltzmann density operator of the harmonic and anharmonic oscillators read, respectively, ρ◦ ∝ e−βH

◦

ρ ∝ e−βH

and

(13.129)

Now, the partial differential of these density operators with respect to β read, respectively,    ◦ ∂ρ ∂ρ ◦ −βH◦ ∝ −H e ∝ −He−βH and (13.130) ∂β ∂β Next, in order to express ρ in terms of ρ◦ , first calculate      −βH  ◦ ◦ ∂eβH ∂(eβH e−βH ) −βH βH◦ ∂e = e +e ∂β ∂β ∂β

(13.131)

which, due to (13.130), yields   ◦ ∂(eβH e−βH ) ◦ ◦ = H◦ eβH e−βH − eβH He−βH ∂β Or, since H◦ commutes with the exponential constructed from it, and owing to Eq. (13.128),   ◦ ∂(eβH e−βH ) ◦ ◦ = eβH (H◦ − H)e−βH = −eβH Ve−βH ∂β Due to Eq. (13.129), the latter equation leads to ◦

◦

d{eβH ρ(β)} = −eβH Ve−βH dβ the integration of which from zero to β reads β d{e 0

β  H◦



β

ρ(β )} = − 0



◦



eβ H Ve−β H dβ

(13.132)

CH013.tex

11/3/2011

15: 31

13.3

Page 389

HELMHOLTZ POTENTIAL FOR ANHARMONIC OSCILLATORS

389

Now, observe that when β = 0, it appears from (13.129) that ◦

◦

eβH ρ(β) = eβH e−βH = 1 so that the integration of (13.132) reads βH◦

e

β



◦



e−(β−β )H Ve−β H dβ

ρ(β) − 1 = − 0

or, after premultiplying both members by exp{−βH◦ } and using the last expression of (13.129), ρ(β) = e

−βH◦

β −



◦

e−(β−β )H Vρ(β) dβ

0

The first-order solution of this last integral is −βH◦

ρ(β) = e

β −



◦



◦

e−(β−β )H Ve−β H dβ

0

whereas the second-order solution is ρ(β) = e

−βH◦

β −

e

−(β−β )H◦

Ve

−β H◦

β β



dβ +

0

0

−β H◦

× Ve



dβ dβ

 

◦



e−(β−β )H Ve−(β −β

 )H◦

0



(13.133)

The solution (13.133) is dealing with a density operator that is unnormalized. But that is of no importance if one is interested in the Helmholtz energy F, which is related, via Eq. (13.122), to the Boltzmann density operator through e−βF = tr{e−βH } = tr{ρ(β)}

(13.134)

an expression that is true whatever the normalization of the density operator. Hence, one gets −βF

e

= tr{e

−βH◦

β }−



◦



◦

tr{e−(β−β )H Ve−β H }dβ

0 β

β  + 0



◦



tr{e−(β−β )H Ve−(β −β

 )H◦

Ve−β

 H◦

} dβ dβ

(13.135)

0

Now, due to the invariance of the trace with respect to a circular permutation within it, it appears that 

◦



◦



◦



◦

◦

tr{e−(β−β )H Ve−β H } = tr{e−β H e−(β−β )H V} = tr{e−βH V}

(13.136)

CH013.tex

390

11/3/2011

15: 31

Page 390

THERMAL PROPERTIES OF HARMONIC OSCILLATORS 

◦



tr{e−(β−β )H Ve−(β −β

 )H◦

Ve−β

 H◦

} = tr{e−β

 H◦

−βH◦

= tr{e



◦



e−(β−β )H Ve−(β −β

e

−(β −β )H◦

Ve

 )H◦

−(β −β )H◦

V} V}

(13.137) H◦

to get Moreover, perform the first trace over the eigenstates |(n) of   −βH◦ −βH◦ −nβω V} = (n)|e V|(n) = e (n)|V|(n) tr{e n

(13.138)

n

whereas working in the same way for the trace (13.137) and after inserting a closure relation over the basis {|(n)} after the first operator V yields ◦





◦





◦

tr{e−βH e−(β −β )H Ve−(β −β )H V}  ◦    ◦  ◦ = (n)|e−βH e−(β −β )H V|(m)(m)|e−(β −β )H V|(n) n

m

or ◦





◦





◦

tr{e−βH e−(β −β )H Ve−(β −β )H V}      = e−nβω e−n(β −β )ω (n)|V|(m)e−m(β −β )ω (m)|V|(n) n

m

(13.139) Due to Eqs. (13.136) and (13.137) and to Eqs. (13.138) and (13.139), Eq. (13.135) giving the Helmholtz free energy becomes −βF

e

= tr{e

−βH◦

}−



e

−nβω

β (n)|V|(n)

n

+

 n

dβ

0

e−nβω |(m)|V|(n)2 |

m

 β β

0





e(n−m)ω(β −β ) dβ dβ

0

(13.140) Now, observe the latter integral may be written β β



e 0

(n−m)ω(β −β )

1 dβ dβ = 2 



0

β e 0

(n−m)ωη

β dη

dβ

0

leading to β β 0

 



e(n−m)ω(β −β ) dβ dβ =

0

β e(n−m)βω − 1 2ω n−m

As a consequence, Eq. (13.140) takes the form ◦

e−βF = e−βF −β

 n

e−nβω (n)|V|(n)+

β  e−mβω −e−nβω |(m)|V|(n)2 | 2ω n m (n − m) (13.141)

CH013.tex

11/3/2011

15: 31

Page 391

13.4 THERMAL AVERAGE OF BOSON OPERATOR FUNCTIONS

391

with, in a similar way as in Eq. (13.134), ◦

◦

e−βF = tr{e−βH } From Eq. (13.141), it may be shown1 that ◦

F  F ◦ + V0

with

V0 =

tr{Ve−βH } ◦ tr{e−βH }

a result that allows one to find physical average values by minimization procedure. Besides, Eq. (13.141) may be applied to an anharmonic oscillator in which V is given by    3/2 † V= (a + a)3 2mω with the help of Eqs. (9.41)–(9.48) and using Eq. (13.21) allowing one to write ◦

◦

e−βF = tr{e−βH } = Z =

e−λ/2 1 − e−λ

with

λ=

ω kBT

13.4 THERMAL AVERAGE OF BOSON OPERATOR FUNCTIONS Now, we shall obtain the general expression for the average of any function of Boson operators over the Boltzmann equilibrium density operator. We shall obtain a general expression that reduces to the Bloch theorem when the function of Boson operators is either the position operator or its conjugate momentum. If the demonstration is somewhat tedious, it has the merit of avoiding the mathematical complications required to obtain its simplified form, which is the Bloch theorem.

13.4.1

Calculation of thermal average

In this section we derive the expression of the thermal average of any function of Boson operators over the canonical density operator of an harmonic oscillator, that is, F(a† , a) = tr{{F(a† , a)}ρB (a† , a)} which, in view of Eqs. (13.23) and (13.24), reads F(a† , a) = (1 − e−λ )tr{{F(a† , a)}(e−λa a )} †

λ=

1

ω kBT

(13.142) (13.143)

R. P. Feynman. Statistical Mechanics: A Set of Lectures, 2nd ed. Perseus Books: New York, 1998.

CH013.tex

392

11/3/2011

15: 31

Page 392

THERMAL PROPERTIES OF HARMONIC OSCILLATORS

13.4.1.1 From the basic equation (13.142) to a more tractable one Tracing on the right-hand side, over the eigenstates |(n) of a† a, transforms Eq. (13.142) to F(a† , a)  † (n)|{F(a† , a)}(e−λa a )|(n) = (1 − e−λ ) n

(13.144)

a† a|(n) = n|(n)

(13.145)

with

F(a† , a),

Now, observe that the only terms of which may contribute to the diagonal matrix elements involved on the right-hand side of this last equation, are those having the same power of a† and a. Accordingly, in the trace above, we are free to change a into ka and a† into a† /k where k is some real scalar (that will be defined later). Hence, since the product a† a involved in the density operator is not affected by this change, Eq. (13.144) yields F(a† , a)  † (n)|{F(a† /k, ka)}(e−λa a )|(n) = (1 − e−λ ) n Also we write this last equation in the following more complex form: F(a† , a)   † = (n)|{F(a† /k, ka)}(e−λa a )|(m)δnm (1 − e−λ ) n m

(13.146)

Moreover, due to Eqs. (5.69) and (5.70),  † m  n (a ) (a) |(m) = √ |(0) and (n)| = (0)| √ (13.147) m! n! Equation (13.146) transforms to  n  † m F(a† , a)   (a) (a ) † −λa† a (0)| /k, ka)}(e ) {F(a |(0)δnm = √ √ (1 − e−λ ) n! m! n m (13.148) Now, since the Kronecker symbol δnm appearing on the right-hand side of Eq (13.148) may be viewed as the scalar product of two eigenstates of any operator b† b, of Boson operators that commute with a† and a, that is, δnm = {n}|{m}

with

b† b|{n} = n|{n}

(13.149)

with [a, b] = 0

[a† , b] = 0

[a, b† ] = 0

Next, using for these Boson operators equations similar to those of (13.147),  † m  n (b ) (b) |{m} = √ |{0} and {n}| = {0}| √ m! n! the Kronecker symbol appearing in (13.149) becomes  n † m (b) (b ) δnm = {0}| √ |{0} (13.150) √ n! m!

CH013.tex

11/3/2011

15: 31

Page 393

13.4 THERMAL AVERAGE OF BOSON OPERATOR FUNCTIONS

393

so that Eq. (13.148) reads  n n  F(a† , a) (b) (a) {0}|(0)| = √ √ (1 − e−λ ) n! n! n m  † m † m (b ) (a ) † × {F(a† /k, ka)}(e−λa a ) √ |(0)|{0} √ m! m! Now, one may replace b by μb and b† by b† /μ, where μ is a real scalar, without modifying the right-hand-side average value, so that Eq. (13.150) becomes  n   F(a† , a) (a) (μb)n = {0}|(0)| √ √ −λ (1 − e ) n! n! n m  † m †  (a ) (b /μ)m † −λa† a × {F(a /k, ka)}(e ) √ |(0)|{0} √ m! m! or, rearranging and simplifying the notation for the ket or bra products,    F(a† , a) (μba)n {0}(0)| = (1 − e−λ ) n! n m  † †  (a b /μ)m † × {F(a† /k, ka)}(e−λa a ) |(0){0} m! Then, pass to exponentials F(a† , a) † † † = {0}(0)|(eμba ){F(a† /k, ka)}(e−λa a )(ea b /μ )|(0){0} (1 − e−λ ) Furthermore, introduce after the function of Boson operators the unity operator defined by 1 = (e−μba )(eμba ) leading to F(a† , a) † † † = {0}(0)|(eμba ){F(a† /k, ka)}(e−μba )(eμba )(e−λa a )(ea b /μ )|(0){0} −λ (1 − e ) (13.151) Now, according to Eq. (7.5), it appears that (eμba ){F(a† /k, ka)}(e−μba ) = {F((a† + μb)/k, ka)} and, comparing Eq. (7.106), that is, (e−λa a )(eya )|(0) = (eya †

†

with y=

b† μ

† (e−λ )

)|(0)

(13.152)

CH013.tex

11/3/2011

394

15: 31

Page 394

THERMAL PROPERTIES OF HARMONIC OSCILLATORS

it appears that, since b† is dimensionless and does not act on the ket |(0) because acting on |{0} one finds (e−λa a )(ea †

† b† /μ

)|(0) = {ea

† b† e−λ /μ

}|(0)

(13.153)

Hence, in view of Eqs. (13.152) and (13.153), the average value (13.151) simplifies to F(a† , a) † † = {0}(0)|{F((a† + μb)/k, ka)}(eμba ){eξa b }|(0){0} (1 − e−λ )

(13.154)

with ξ = e−λ /μ

(13.155)

13.4.1.2 Action of the product of exponential operators involved in Eq. (13.154) on |(0){0} It is now required to find the action of the product of the two exponential operators involved on the right-hand side of Eq. (13.154) on the ground state |(0){0} of a† a b† b. Hence, one has to find a function of ladder operators G(μ, a† , b† , a, b) satisfying (eG(μ,a

† ,b† ,a,b)

)|(0){0} = (eμba ){eξa

† b†

}|(0){0}

(13.156)

For this purpose, differentiate both members of Eq. (13.156) with respect to μ, yielding   ∂G † † exp (G) |(0){0} = ba(eμba ){eξa b }|(0){0} ∂μ or   ∂G exp (G) |(0){0} = ba exp{G}|(0){0} ∂μ Then, premultiplying both terms by exp (−G) we have   ∂G |(0){0} = exp{−G}ba exp{G}|(0){0} ∂μ Again, insert between b and a the unity operator built up from exp{−G}, that is,   ∂G |(0){0} = exp{−G} b{G} exp{−G}a exp{G}|(0){0} (13.157) ∂μ and apply Eq. (7.60), that is,



af(a, a ) − f(a, a )a = †

†

∂f(a, a† ) ∂a†



to the function f(a, a† ) = exp{G} Hence

 a exp{G} − exp{G}a =

∂ exp{G} ∂a†



CH013.tex

11/3/2011

15: 31

Page 395

13.4 THERMAL AVERAGE OF BOSON OPERATOR FUNCTIONS

or



∂G a exp (G) = exp (G)a + exp{G} ∂a†

395



Then, premultiplying both terms by exp (−G) we have after simplification   ∂G (13.158) exp{−G}a exp{G} = a + ∂a† In a similar way one would obtain



exp{−G}b exp{G} = b +

∂G ∂b†

 (13.159)

As a consequence of Eqs. (13.158) and (13.159), Eq. (13.157) becomes         ∂G ∂G ∂G |(0){0} = b + a + |(0){0} ∂μ ∂b† ∂a† Then, performing the product involved on the right-hand side gives            ∂G ∂G ∂G ∂G ∂G +b + a |(0){0} |(0){0} = ba + ∂μ ∂b† ∂a† ∂a† ∂b† (13.160) Now, observe that since b|{0} = a|(0) = 0

(13.161)

we have ba|(0){0} = b|{0}a|(0) = 0 so that Eq. (13.160) yields        ∂G ∂G ∂G ∂G +b |(0){0} |(0){0} = ∂μ ∂b† ∂a† ∂a† Again, using in turn Eq. (7.60),



bf(b, b ) − f(b, b )b = †

so that

†



∂G b ∂a† and since, due to Eq. (13.161),

∂f(b, b† ) ∂b†



 =



∂G ∂a†

∂G ∂a†



 with



 b+

(13.162)

f(b, b† ) =

∂2 G ∂b† ∂a†

∂G ∂a†





 b|{0} = 0

Eq. (13.162) reads      2   ∂G ∂G ∂ G ∂G + |(0){0} |(0){0} = ∂μ ∂a† ∂b† ∂b† ∂a†

(13.163)

CH013.tex

396 or 

11/3/2011

15: 31

Page 396

THERMAL PROPERTIES OF HARMONIC OSCILLATORS

    2  ∂G(a† , b† , μ) ∂G ∂G ∂ G(a† , b† , μ) |(0){0} = |(0){0} (13.164) ∂μ ∂a† ∂b† ∂b† ∂a†

Next, in order to solve this partial differential equation involving only a† , b† , and μ, one may seek a solution at an expression of the following form: G(a† , b† , μ) = A(μ) + B(μ)a† b†

(13.165)

where a and b disappear, whereas A(μ) and B(μ) are unknown scalar coefficients. Now, in order to get the expression of the function (13.165) for the special situation where μ = 0, use the fact that in this special situation, Eq. (13.156) reduces to (eG(0,a

† ,b† )

)|(0){0} = {eξa

† b†

}|(0){0}

(13.166)

Thereby, since the arguments of the exponentials appearing on the right- and on the left-hand-side operators of this last equation must be the same, we have G(0, a† , b† ) = ξa† b† Thus, the comparison of this last expression with Eq. (13.165) in which μ = 0 leads, respectively, to A(0) = 0

and

B(0) = ξ

(13.167)

Furthermore, due to Eq. (13.165), it appears that the partial derivative of G with respect to μ reads       ∂A(μ) ∂B(μ) † † ∂G(μ, a† , b† ) (13.168) = + a b ∂μ ∂μ ∂μ while, the crossed second-order partial derivative of G with respect to a† and b† yields  2  ∂ G(μ, a† , b† ) = B(μ) + {B(μ)}2 a† b† (13.169) ∂b† ∂a† At last, due to Eq. (13.165)   ∂G(μ, a† , b† ) = B(μ)b† ∂a†

 and

∂G(μ, a† , b† ) ∂b†

so that, since a† and b† commute,    ∂G(μ, a† , b† ) ∂G(μ, a† , b† ) = {B(μ)}2 a† b† ∂a† ∂b†

 = B(μ)a†

(13.170)

Hence, due to Eqs. (13.168)–(13.170), Eq. (13.164) takes the form      ∂A(μ) ∂B(μ) † † + a b |(0){0} = ({B(μ)} + {B(μ)}2 a† b† )|(0){0} ∂μ ∂μ so that one obtains by identification   ∂B(μ) = {B(μ)}2 ∂μ

(13.171)

CH013.tex

11/3/2011

15: 31

Page 397

13.4 THERMAL AVERAGE OF BOSON OPERATOR FUNCTIONS



By integration Eq. (13.171) yields

∂A(μ) ∂μ

397

 = B(μ)



1 1 μ=− − B(μ) B(0)

(13.172) 

or, in view of the boundary condition appearing in (13.167),   1 1 μ=− − B(μ) ξ and thus, after rearranging, B(μ) =

ξ 1 − ξμ

(13.173)

Next, insert this result into Eq. (13.172) to get   ∂A(μ) ξ = ∂μ 1 − ξμ which by integration yields μ A(μ) − A(0) = ξ 0

dμ 1 − ξμ

and thus, due to the first boundary condition of Eq. (13.167), and after calculation of the integral A(μ) = − ln (1 − ξμ)

(13.174)

Hence, comparing Eq. (13.173), the function (13.165) becomes ξ G(a† , b† , μ) = − ln (1 − ξμ) + a † b† 1 − ξμ so that Eq. (13.156) is  † † ξa b † † (eG(μ,a ,b ) )|(0){0} = exp − ln (1 − ξμ) |(0){0} 1 − ξμ or  † †   ξa b 1 G(μ,a† ,b† ) (e )|(0){0} = exp |(0){0} 1 − ξμ (1 − ξμ) Now, observe that, due to Eq. (13.165), the left-hand side of this last equation is the same as that of (13.156), so that of the identification of the corresponding right-hand sides gives    † † 1 ξa b μba ξa† b† (e ){e }|(0){0} = exp |(0){0} (13.175) (1 − ξμ) 1 − ξμ At last, coming back from ξ to λ by the aid of Eq. (13.155) leading to   1 1 = (1 − ξμ) 1 − e−λ

CH013.tex

398

11/3/2011

15: 31

Page 398

THERMAL PROPERTIES OF HARMONIC OSCILLATORS

Eq. (13.175) transforms to (eμba ){ea

† b† e−λ /μ

}|(0){0} =

1 † † {eξa b /(1−ξμ) }|(0){0} −λ (1 − e )

(13.176)

13.4.1.3 Final step for the thermal average value As a consequence of Eq. (13.176), the thermal average value (13.154) becomes F(a† , a) = {0}(0)|{F((a† + μb)/k, ka)}(ea

† b† (ξ/(1−ξμ))

)|(0){0}

Again, observe that, owing to Eq. (13.155), it yields  −λ  ξ 1 e = 1 − ξμ μ 1 − e−λ

(13.177)

(13.178)

Hence, due to Eqs. (13.14) and (13.36) we have n =

1 e−λ = eλ − 1 1 − e−λ

with

λ=

ω kBT

(13.179)

where n is the thermal average of the occupation number, that is, of a† a or of b† b, that is, n = (1 − e−λ )tr{(e−λa a )a† a} = (1 − e−λ )tr{(e−λb b )b† b} †

†

Equation (13.178) reads ξ n = 1 − ξμ μ

(13.180)

so that the thermal average (13.177) yields F(a† , a) = {0}(0){|F((a† + μb)/k, ka)}(ena

† b† /μ

)|(0){0}

(13.181)

Now, observe that {0}(0)| = {0}(0)|(e−na

† b† /μ

)

that is because, after its expansion, the right-hand side reads    n m (a† )m (b† )m (−1)m −na† b† /μ {0}(0)|(e )= {0}(0)| μ m! m or, after action of each operator within its own subspace,  m  1 −na† b† /μ m n )= (−1) {0}|(b† )m (0)|(a)m {0}(0)|(e μ m! m then, using Eq. (5.73) leading to {0}|(b† )m = (0)|(a† )m = δm,0 it appears, Q.E.D. {0}(0)|(e−na

† b† /μ

) = {0}(0)|1

(13.182)

CH013.tex

11/3/2011

15: 31

Page 399

13.4 THERMAL AVERAGE OF BOSON OPERATOR FUNCTIONS

399

Therefore, Eq. (13.181) becomes F(a† , a) = {0}(0)|(e−na

† b† /μ

){F((a† + μb)/k, ka)}(ena

† b† /μ

)|(0){0} (13.183) Moreover, keeping in mind theorem (1.77) applying to some function F(B) of operator B, that is, eξA F(B)e−ξA = F(eξA Be−ξA ) where ξ is a c-number and A is an operator that does not commute with B, apply it to the canonical transformation appearing on the right-hand side of Eq. (13.183) by taking ξ=

A = a † b†

n μ

and B=

a† + μb k

B = ka

or

Then, this canonical transformation reads (e−na

){F((a† + μb)/k, ka)}(ena b /μ )    †  a + μb † † † † † † † † = F (e−na b /μ ) (ena b /μ ), k(e−na b /μ )a(ena b /μ ) k † b† /μ

† †

(13.184) Besides, since a† commutes with b† , it is clear that (e−na

† b† /μ

) a† (ena

† b† /μ

) = a†

Moreover, applying theorem (7.7), that is, e−ξa F(a, a† )eξa = F(a + ξ, a† ) †

†

with taking n † a μ

ξ=

ξ=

or

n † b μ

and keeping in mind the following commutators [a, b† ] = [a† , b† ] = [a† , b] = [a, b] = 0 one finds, respectively, (e−na

† b† /μ

(e−na

† b† /μ

† b† /μ

)=b+

n † a μ

(13.185)

† b† /μ

)=a+

n † b μ

(13.186)

)b(ena

)a(ena

CH013.tex

400

11/3/2011

15: 31

Page 400

THERMAL PROPERTIES OF HARMONIC OSCILLATORS

As a consequence of Eqs. (13.185) and (13.186), the canonical transformation (13.184) becomes (enab/μ ){F((a† + μb)/k, ka)}(ena b /μ )      (1 + n) † μ n † = F a + b , ka+k b k k μ † †

(13.187)

It is now necessary to find the expressions for the unknown scalars μ and k involved in this equation. For this purpose, we may write Eq. (13.187) in terms of two Boson operators c and c† , which are linear combinations of a and a† and b and b† according to (enab/μ ){F((a† + μb)/k, ka)}(ena

† b† /μ

) = {F(c† , c)}

(13.188)

c† = C1∗ a† + C2∗ b

(13.189)

with c =C1 a+C2 b†

and

so that after identification with the right-hand side of Eq. (13.187), one obtains, respectively, for the coefficients C1 and C2 of Eq. (13.189) C1 = k

C1∗ =

C2 = k

and

(1 + n) k

n μ

C2∗ =

and

(13.190) μ k

(13.191)

Then, since the scalars k, μ, and n appearing in Eqs. (13.190) and (13.191) are real, we have C1 = C1∗

and

C2 = C2∗

so that Eqs. (13.190) and (13.191) read k=

(1 + n) k

k=

 1 + n

and

k

n μ = μ k

leading to and

 μ = k n

Furthermore, introducing these expressions for k and μ into Eq. (13.187), we have (enab/μ ){F((a† + μb)/k, ka)}(ena b /μ )      =F 1 + na† + nb, 1 + na + nb† † †

Hence, using this result allows one to transform the thermal average (13.183) into     F(a† , a) = {0}(0)|{F( 1 + na† + nb, 1 + na + nb† )}|(0){0}

CH013.tex

11/3/2011

15: 31

Page 401

13.4 THERMAL AVERAGE OF BOSON OPERATOR FUNCTIONS

401

which due to the definition of F(a† , a) given by Eq. (13.142) reads (1 − e−λ )tr{(e−λa a ){F(a† , a)}}     = {0}(0)|{F( 1 + na† + nb, 1 + na + nb† )}|(0){0} †

(13.192)

13.4.2 Thermal average of translation operators and Bloch theorem Now, suppose that the operator function to be averaged and given by Eq. (13.192) is a translation operator, that is, A(T ) = (1 − e−λ )tr{(e−λa a ){A(a† , a)}} †

(13.193)

with A(a† , a) = {eαa

† −α∗ a

}

and thus A(T ) = (1 − e−λ )tr{(e−λa a ){eαa †

† −α∗ a

}}

(13.194)

Then, using Glauber’s theorem (1.79), in order to factorize the right-hand-side exponential operators {eαa

† −α∗ a

∗

} = (eαa )(e−α a )e−[αa †

† ,−α∗ a]/2

(13.195)

with e−[αa

† ,−α∗ a]/2

= e|α|

2 [a† ,a]/2

= (e−|α|

2 /2

)

(13.196)

using Eqs. (13.195) and (13.196), the thermal average (13.194) becomes A(T ) = (1 − e−λ )e−|α|

2 /2

∗

tr{(e−λa a )(eαa )(e−α a )} †

†

(13.197)

Now, apply theorem (13.192) to Eq. (13.197) in order to find the expression for its 2 thermal average. Then, ignoring momentously the phase factor e−|α| /2 , we have ∗

(1 − e−λ )tr{(e−λa a )(eαa )(e−α a )} †

= (0)|{0}|(eα(

†

√

√ na† + 1+nb)

)(e−α

∗(

√

√ na+ 1+nb† )

)|{0}2 |{0}1 

Then, factorizing both exponentials, each involving commuting operators, gives ∗

(1 − e−λ )tr{(e−λa a )(eαa )(e−α a )} †

†

√ na†

= (0)|{0}|(eα

√ 1+nb

)(eα

)(e−α

∗

√ na

)(e−α

∗

√ 1+nb†

)|{0}|(0)

Again, working within the two different subspaces leads to (1 − e−λ )tr{(e−λa a )(eα(t)a )(e−α †

= (0)|{(e

√ α na†

†

)(e

−α∗

√ na

∗ (t)a

)}

√ 1+nb

)}|(0){0}|{(eα

)(e−α

∗

√ 1+nb†

)}|{0} (13.198)

CH013.tex

402

11/3/2011

15: 31

Page 402

THERMAL PROPERTIES OF HARMONIC OSCILLATORS

Next, expand the two exponentials of the last right-hand-side matrix element of this last equation to get √

∗

√

{0}|{(eα 1+nb )(e−α 1+nb )}|{0}  αk α∗l  = (−1)l ( 1 + n)k+l (0)|{(b)k (b† )l }|{0} k!l! k

†

(13.199)

l

Then, comparing Eqs. (5.67) and (5.68), that is, √ √ {0}|(b)k = {k}| k! and (b† )l |{0} = l!|{l} it appears that {0}|(b)k (b† )l |(0) =

√ √ k! l!{k}|{l} = l!

so that the double sum of the matrix elements involved on the right-hand side of Eq. (13.199) reduces after simplifications to √ √  |α|2l ∗ † {0}|{(eα 1+nb )(e−α 1+nb )}|{0} = (−1)l (n + 1)l l! l

Therefore, coming back to the exponentials, this last equation becomes √

{0}|{(eα

1+nb

)(e−α

∗

√ 1+nb†

)}|{0} = exp{−|α|2 (n + 1)}

(13.200)

Now, expand the exponential of the first matrix element of the right-hand side of Eq. (13.198), that is, √

∗

√

(0)|{(eα na )(e−α na )}|(0)  αk α∗l  = (−1)l ( n)k+l (0)|{(a† )k (a)l }|(0) k!l! †

k

(13.201)

l

Then, owing to Eq. (5.55), we have (0)|(a† )k = 0

except if k = 0

(a)l |(0) = 0

except if l = 0

This follows that Eq. (13.201) reduces to √

(0)|{(eα

na†

)(e−α

∗

√ na

)}|(0) = 1

(13.202)

As a consequence of Eqs. (13.200) and (13.202), the thermal average (13.198) becomes ∗

(1 − e−λ )tr{(e−λa a )(eαa )(e−α a )} = exp{−|α|2 (n + 1)} †

†

(13.203)

with n given by Eq. (13.179). Again, using Glauber’s theorem yields (1 − e−λ )e−|α|

2 /2

tr{(e−λa a )(eαa †

† −α∗ a

)}e|α| = exp{−|α|2 (n + 1)} 2

so that after simplification A(T ) = (1 − e−λ )tr{(e−λa a )(eαa †

† −α∗ a



 )} = exp −|α|2 n + 21

(13.204)

CH013.tex

11/3/2011

15: 31

Page 403

13.5

13.4.2.1 Bloch theorem Eq. (13.203)

CONCLUSION

403

Of course, one would obtain in a similar way as for

(1 − e−λ )tr{(e−λa a )(eαa †

† +α∗ a



 )} = exp |α|2 n + 21

(13.205)

Next, if we denote and keep in mind Eqs. (5.6) allowing one to pass from the Boson operators to the position operator Q according to   † Q = α(a + a) with α = 2mω it appears that if α(t) is real, the left-hand side of Eq. (13.205) reads (1 − e−λ )tr{(e−λa a )(eα(a †

† +a)

)} = tr{ρB eQ } = eQ 

(13.206)

so that Eq. (13.205) yields 



 e  = exp (n + 21 ) 2mω Q

(13.207)

Now, observe that the thermal average of Q(T )2 defined by Q(T )2  = tr{ρB Q(T )2 } that is, Q(T )2  =

 † (1 − e−λ )tr{(e−λa a )(a† + a)2 } 2mω

is given by Eq. (13.72), that is, Q(T )2  =

 (2n + 1) 2mω

(13.208)

As a consequence of Eqs. (13.207) and (13.208), eQ  = eQ

2 /2

This last result is the Bloch theorem. In a similar way, one would obtain for the momentum eP  = eP

13.5

2 /2

CONCLUSION

Using the canonical operator, it was possible in this chapter to find many thermal properties of quantum harmonic oscillators such as the fundamental Planck law, the thermal average of kinetic and potential energies, the heat capacities, the energy fluctuations, and the part of the Sackur and Tetrode law dealing with entropy. Finally, we

CH013.tex

404

11/3/2011

15: 31

Page 404

THERMAL PROPERTIES OF HARMONIC OSCILLATORS

gave some complex demonstrations of the thermal average energy of ladder operator functions, one consequence of which is the Bloch theorem. The most important results dealing with thermal average of simple operators characterizing harmonic oscillators are reported as follows: Thermal average over Boltzmann density operators Boltzmann density operators: ρB =

1 −βH ) (e Z

with β =

1 kBT

Partition function:  −βω/2  e Z= 1 − e−βω Average Hamiltonian:   ω ω H = + ω/k T B −1 2 e Heat capacity:  Cv = Nk B

ω kBT

2

eω/k B T (eω/k B T − 1)2

Energy fluctuation: E Tot =

√ Nω

eω/2k B T − 1)

(eω/k B T

Average of Q2 : Q(T )2  =

   ω coth 2mω 2k B T

Entropy:     ω/2k B T 1 ) R (e 1 ◦ ω + N S=n ln + (nN ◦ )! T eω/k B T − 1 2 1 − eω/k B T whereas we give hereafter some important theorems dealing with the thermal average of exponential operators involving the ladder operators:

CH013.tex

11/3/2011

15: 31

Page 405

BIBLIOGRAPHY

405

Theorems dealing with thermal averages Thermal average of operators over Boltzmann density operator: (1 − e−λ )tr{(e−λa a ){F(a† , a)}} √ √ √ √ = {0}(0)|{F( 1 + na† + nb, 1 + na + nb† )}|(0){0} †

Thermal average of the translation operator: (1 − e−λ )tr{(e−λa a )(eαa †

† −α∗ a

)} = exp{−|α|2 (n + 21 )}

Bloch’s theorem: (1 − e−λ )tr{(e−λa a )eQ } = exp{(1 − e−λ )tr{(e−λa a )Q2 /2}} †

†

(1 − e−λ )tr{(e−λa a )eP } = exp{(1 − e−λ )tr{(e−λa a )P2 /2}} †

†

BIBLIOGRAPHY B. Diu, C. Guthmann, D. Lederer, and B. Roulet. Physique statistique. Hermann: Paris, 1988. Ch. Kittel and H. Kroemer. Thermal Physics, 2nd ed. W. H. Freeman: New York, 1980. H. Louisell. Quantum Statistical Properties of Radiations. Wiley: New York, 1973. F. Reif. Fundamentals of Statistical and Thermal Physics. McGraw-Hill: New York, 1965.

ch014.tex

11/3/2011

PART

16: 4

Page 407

V

QUANTUM NORMAL MODES OF VIBRATION Part IV was essentially devoted to large sets of weakly coupled harmonic oscillators, allowing one to obtain many thermal properties. However, other kinds of large sets of oscillators exist involving couplings that allow one to separate them so as to get decoupled harmonic oscillators, that is, the normal modes of the oscillator system. The aim of Part V is to treat normal modes.

ch014.tex

11/3/2011

16: 4

Page 409

14

CHAPTER

QUANTUM ELECTROMAGNETIC MODES INTRODUCTION The first chapter of this part, Chapter 14, will treat the quantum modes of the electromagnetic field and the last one the normal modes of molecular systems of 1D solids. The purpose of the present chapter is to study the quantum electromagnetic modes. First, we shall see how to get the classical electromagnetic modes in reciprocal space. Second, we shall show how to pass from these classical modes to the corresponding quantum ones. Then, applying some results encountered in the previous chapters dealing with the properties of quantum harmonic oscillators, it will be possible to introduce the notion of light corpuscles of a given angular frequency, called photons, which are the excitation degrees of the normal modes. Besides, applying the thermal properties of quantum oscillators we have obtained previously, it will be also possible to get different important results dealing with the thermal properties of light such as, for instance, the Planck black-body radiation law or the Stefan–Boltzmann law.

14.1 14.1.1

MAXWELL EQUATIONS Maxwell equations within the geometrical space

We start from the Maxwell equations governing the electric field E(r, t) and the magnetic field B(r, t), that is, ∇ · E(r, t) =

1 ρ(r, t) ε◦

(14.1)

∇ · B(r, t) = 0  ∇ × E(r, t) = −

∂B(r, t) ∂t

(14.2)  (14.3)

Quantum Oscillators, First Edition. Olivier Henri-Rousseau and Paul Blaise. © 2011 John Wiley & Sons, Inc. Published 2011 by John Wiley & Sons, Inc.

409

ch014.tex

410

11/3/2011

16: 4

Page 410

QUANTUM ELECTROMAGNETIC MODES

1 ∇×B(r, t) = 2 c



∂E(r, t) ∂t

 +

1 ε◦ c 2

J(r, t)

(14.4)

where r is the position vector, t the time, ε◦ the vacuum permittivity, c the velocity of light, ρ(r, t) the charge density at position r and time t, and J(r, t) the current density related to ρ(r, t) through the charge conservation law: ∂ρ(r, t) + ∇ · J(r, t) = 0 ∂t

(14.5)

In the absence of charge ρ(r, t) = J(r, t) = 0 so that the four Maxwell equations governing the electric and magnetic fields reduce to ∇ · E(r, t) = 0

(14.6)

∇ · B(r, t) = 0

(14.7)



∂B(r, t) ∇×E(r, t) = − ∂t 1 ∇×B(r, t) = 2 c





∂E(r, t) ∂t

(14.8)  (14.9)

Now, the scalar and vector potentials V (r, t)and A(r, t) may be defined from the electric and magnetic fields via B(r, t) = ∇×A(r, t)

(14.10)

and

  ∂A(r, t) E(r, t) = − − ∇V (r, t) ∂t In the Coulomb gauge V (r, t) and A(r, t) are chosen in such a way as ∇·A(r, t) = 0

and

∇V (r, t) = 0

so that in this gauge Eq. (14.11) simplifies to   ∂A(r, t) E(r, t) = − ∂t

14.1.2

(14.11)

(14.12)

(14.13)

Maxwell equations within reciprocal space

We made Fourier transforms allowing one to pass from geometric to reciprocal spaces:  3/2  1 B(k, t) = B(r, t)e−ik·r d 3 r (14.14) 2π  E(k, t) =

1 2π

3/2 

E(r, t)e−ik·r d 3 r

(14.15)

ch014.tex

11/3/2011

16: 4

Page 411

14.1

 A(k, t) =

1 2π

3/2 

MAXWELL EQUATIONS

A(r, t)e−ik·r d 3 r

411

(14.16)

According to Eqs. (18.47) and (18.53) of Section 18.6, and Eqs. (14.14)–(14.16), Eqs. (14.6)–(14.9) read ik · E(k, t) = 0

(14.17)

ik · B(k, t) = 0

(14.18)



∂B(k, t) ik × E(k, t) = − ∂t 1 ik × B(k, t) = 2 c





∂E(k, t) ∂t

(14.19)  (14.20)

and Eqs. (14.12) and (14.13) yield ik · A(k, t) = 0

(14.21)

  ∂A(k, t) E(k, t) = − ∂t

(14.22)

The passage from geometric space to reciprocal space allows one to transform the Maxwell equations (14.6)–(14.9) and Eqs. (14.12) and (14.13), which are partial differential equations, to the new ones (14.17)–(14.20), which form, for each point k of the reciprocal space, a infinite set of differential equations governing E(k, t) and B(k, t). Now, according to the Helmholtz theorem of vectorial analysis, any vector F(k, t) may be always decomposed according to F(k, t) = F// (k, t) + F⊥ (k, t) with ik × F// (k, t) = 0 ik · F⊥ (k, t) = 0

(14.23)

Thus, due to Eq. (14.17) and Eqs. (14.18)–(14.21), the fields E(k, t), B(k, t), and A(k, t) only involve components perpendicular to the wave vector k, leading one to write E// (k, t) = 0

and

E(k, t) = E⊥ (k, t)

(14.24)

B// (k, t) = 0

and

B(k, t) = B⊥ (k, t)

(14.25)

A// (k, t) = 0

and

A(k, t) = A⊥ (k, t)

(14.26)

ch014.tex

412

11/3/2011

16: 4

Page 412

QUANTUM ELECTROMAGNETIC MODES

14.1.3 Linear combinations of E⊥ (k, t) and B⊥ (k, t) acting as normal modes Now, the direct product of Eq. (14.19) by the vector k, that is,   ∂B⊥ (k, t) k = −ik × (k × E⊥ (k, t)) ∂t reads, since k does not depend on time,   ∂{k×B⊥ (k, t)} = −ik × (k × E⊥ (k, t)) ∂t

(14.27)

Then, applying to the right-hand side of Eq. (14.27) theorem (18.88) of Section 18.8, that is, V × (W × U) = (V · U)W − (V · W)U

(14.28)

where V, W, and U are vectors, yields k × (k × E⊥ (k, t)) = (k · E⊥ (k, t))k−(k · k)E⊥ (k, t)

(14.29)

Thus, keeping in mind that, according to Eqs. (14.23) and (14.24) that k · E⊥ (k, t) = 0

(14.30)

the latter equation and (14.29) and (14.30) allow one to transform Eq. (14.27) into   ∂{k × B⊥ (k, t)} (14.31) = ik 2 E⊥ (k, t) ∂t with k2 = k · k Moreover, introducing the unit vector κˆ through k = k κˆ and, after simplification by k, Eq. (14.31) transforms to   ∂{ˆκ × B⊥ (k, t)} = ikE⊥ (k, t) ∂t

(14.32)

(14.33)

On the other hand, owing to Eqs. (14.24), (14.25), and (14.32), the partial differential equation (14.20) becomes   ∂E⊥ (k, t) = ic2 k{ˆκ × B⊥ (k, t)} (14.34) ∂t Then, adding and subtracting Eqs. (14.33) and (14.34), we have   ∂{E⊥ (k, t) + cκˆ × B⊥ (k, t)} = iω(k){E⊥ (k, t) + cκˆ ×B⊥ (k, t)} ∂t   ∂{E⊥ (k, t) − cκˆ × B⊥ (k, t)} ) = −iω(k){E⊥ (k, t) − cκˆ × B⊥ (k, t)} ∂t

(14.35) (14.36)

where ω(k) = ck

(14.37)

ch014.tex

11/3/2011

16: 4

Page 413

14.1

MAXWELL EQUATIONS

413

Observe that the two equations (14.35) and (14.36), involving linear combinations of the electrical and magnetic fields in reciprocal space act as those governing decoupled normal modes.

14.1.4

Dimensionless normal modes

Now, introduce the two following dimensionless fields defined by:   E⊥ (k, t) − cκˆ × B⊥ (k, t) {α⊥ (k, t)} = −iK(k) 2  {β⊥ (k, t)} = −iK(k)

E⊥ (k, t) + cκˆ × B⊥ (k, t) 2

(14.38)

 (14.39)

where K(k) are real constants allowing dimensionless α⊥ (k, t) and β⊥ (k, t) to be dimensionless. Then, because in the Euclidian space E⊥ (r, t) and B⊥ (r, t) are real, and due to Eq. (18.43) of Section 18.6, it follows from Eqs. (14.14) and (14.15) that E⊥ (k, t)∗ = E⊥ (−k, t)

B⊥ (k, t)∗ = B⊥ (−k, t)

and

(14.40)

so that {α⊥ (k, t)}∗ = {α⊥ (−k, t)}

and

{β⊥ (k, t)}∗ = {β⊥ (−k, t)}

These properties allow one to find the relation between α⊥ (k, t) and β⊥ (k, t) defined by Eqs. (14.38) and (14.39) in the following way: Because K(k) is real, the conjugate complex of α⊥ (k, t) given by Eq. (14.38) reads   E⊥ (k, t)∗ − cκˆ ×B⊥ (k, t)∗ ∗ {α⊥ (k, t)} = iK(k) 2 which transforms, in view of Eq. (14.40), into   E⊥ (−k, t) − cκˆ × B⊥ (−k, t) {α⊥ (k, t)}∗ = iK(k) 2 Hence, changing k into −k , and thus, according to Eq. (14.32), κˆ into −ˆκ , yields   E⊥ (k, t) + cκˆ × B⊥ (k, t) {α⊥ (−k, t)}∗ = iK(k) 2 so that, by comparison of this result with (14.39), we have {β⊥ (k, t)} = −{α⊥ ( − k, t)}∗

(14.41)

Hence, comparing Eqs. (14.38) and (14.39), the partial differential equations (14.35) and (14.36) yield     ∂α⊥ (k, t) ∂α⊥ (k, t)∗ and = −iω(k){α⊥ (k, t)} = iω(k){α⊥ (k, t)}∗ ∂t ∂t (14.42) which, after integration, lead to {α⊥ (k, t)} = {α⊥ (k)}(e−iω(k)t )

and

{α⊥ (k, t)}∗ = {α⊥ (k)}(eiω(k)t ) (14.43)

ch014.tex

414

11/3/2011

16: 4

Page 414

QUANTUM ELECTROMAGNETIC MODES

Thus, from Eqs. (14.42) and (14.43), the dimensionless variables α⊥ (k, t) or α⊥ (k, t)∗ characterized by different values of the wave vector k, are decoupled, so that they may be viewed as the normal modes of the field within the reciprocal space.

14.1.5

Fields in terms of dimensionless normal modes

Next, one obtains by inversion of Eqs. (14.38), (14.39), and (14.41)   {α⊥ (k, t)} − {α⊥ (−k, t)}∗ {E⊥ (k, t)} = i K(k) 

κˆ × α⊥ (k, t) + κˆ × α⊥ (−k, t)∗ {B⊥ (k, t)} = i cK(k) or, in view of Eq. (14.43),





α⊥ (k)e−iω(k)t − α⊥ (−k)∗ eiω(k)t {E⊥ (k, t)} = i K(k)  {B⊥ (k, t)} = i

(14.44)

κˆ × α⊥ (k)e−iω(k)t + κˆ × α⊥ (−k)∗ eiω(k)t cK(k)

(14.45)  (14.46)  (14.47)

On the other hand, to get the expression of the fields within the geometrical space, perform the inverse Fourier transforms of (14.14 ) and (14.15) to get  3/2  1 E⊥ (k, t)eik.r d 3 k E⊥ (r, t) = 2π  B⊥ (r, t) =

1 2π

3/2  B⊥ (k, t)eik.r d 3 k

which, due to Eqs. (14.46) and (14.47), take, respectively, the forms  E⊥ (r, t) = i Eωk (α⊥ (k)(eik.r )e−iω(k)t − α⊥ (−k)∗ (eik.r )eiω(k)t )d 3 k  B⊥ (r, t) = i

(14.48)

Bωk ((ˆκ × α⊥ (k))(eik.r )e−iω(k)t + (ˆκ × α⊥ (−k)∗ )(eik.r )eiω(k)t )d 3 k (14.49)

with, according to Eq. (14.37),  3/2 1 1 Eωk = 2π K(k)

 B ωk =

and

1 2π

3/2

1 cK(k)

(14.50)

Again, let k → −k inside the last part of the integrals (14.48) and (14.49), leading therefore to α⊥ (−k)∗ → α⊥ (k)∗ κ=

k → −κ k

and

eik.r → e−ik.r

ch014.tex

11/3/2011

16: 4

Page 415

14.2

ELECTROMAGNETIC FIELD HAMILTONIAN

(14.48) and (14.49) transform into  E⊥ (r, t) =i Eωk {α⊥ (k)eik.r e−iω(k)t − α⊥ (k)∗ e−ik.r eiω(k)t }d 3 k B⊥ (r, t) =i



415

(14.51)

Bωk {(ˆκ × α⊥ (k))eik.r e−iω(k)t − (ˆκ × α⊥ (k)∗ )e−ik.r eiω(k)t }d 3 k (14.52)

Next, keeping in mind Eq. (14.22), that is,   ∂A⊥ (k, t) E⊥ (k, t) = − ∂t

(14.53)

it appears that, due to Eq. (14.46), and in order to satisfy Eq. ( 14.53), A⊥ (k, t) must obey   α⊥ (k)e−iω(k)t + α⊥ (−k)∗ eiω(k)t A⊥ (k, t) = (14.54) ω(k)K(k) which reads, at an initial time,



A⊥ (k, 0) =

α⊥ (k) + α⊥ (−k)∗ ω(k)K(k)

 (14.55)

Furthermore, due to Eq. (14.54) the potential vector working within the geometrical space, that is, the inverse Fourier transform of A⊥ (k, t), yields  3/2  1 A⊥ (k, t)eik.r d 3 k A⊥ (r, t) = 2π and transforms, after changing as above k into −k inside the last integral, into  A⊥ (r, t) = Aωk (α⊥ (k)eik.r e−iω(k)t + α⊥ (k)∗ e−ik.r eiω(k)t )d 3 k (14.56) with, in view of Eq. (14.37),  Aωk =

14.2

1 2π

3/2

1 ω(k)K(k)

(14.57)

ELECTROMAGNETIC FIELD HAMILTONIAN

Now, consider the classical Hamiltonian H(t) of the electromagnetic field, that is, its energy, which is given in the absence of charge by   ◦ ε E⊥ (r, t)2 + μ◦−1 B⊥ (r, t)2 (14.58) d3r H(t) = 2 where ε◦ and μ◦ are, respectively, the electrical susceptibility and the magnetic permeability of the vacuum related to the velocity of light c through μ ◦ ε◦ c 2 = 1

(14.59)

ch014.tex

11/3/2011

416

16: 4

Page 416

QUANTUM ELECTROMAGNETIC MODES

Now, in the absence of charge, since an isolated electromagnetic field cannot exchange energy, H must remain constant. Hence, t may be omitted in Eq. (14.58), so that   E⊥ (r)2 + c2 B⊥ (r)2 ◦ d3r (14.60) H=ε 2 Moreover, since E⊥ (r) and B⊥ (r) are real, the Parseval–Plancherel identity (18.44) of Section 18.6, allows one to write   2 3 E⊥ (r) d r = |E⊥ (k)|2 d 3 k 

 B⊥ (r)2 d 3 r =

|B⊥ (k)|2 d 3 k

so that, energy (14.60) yields in reciprocal space    |E⊥ (k)|2 + c2 |B⊥ (k)|2 d3k H = ε◦ 2

(14.61)

Next, in view of Eq. (14.46), the squared absolute value of the electric field appearing in Eq. (14.61) reads   (α⊥ (k)∗ − α⊥ (−k)) · (α⊥ (k) − α⊥ (−k)∗ ) |E⊥ (k)|2 = K(k)2 or, performing the product without changing the order of the factors, for reasons that will become obvious when passing to quantum mechanics, |E⊥ (k)|2   α⊥ (k)∗ · α⊥ (k) + α⊥ (−k) · α⊥ (−k)∗ − α⊥ (k)∗ · α⊥ (−k)∗ − α⊥ (−k) · α⊥ (k) = K(k)2 (14.62) so that, after passing from the vectors α⊥ (±k) to their corresponding scalar α(±k)   α⊥ (k)∗ α⊥ (k) + α(−k)α(−k)∗ − α⊥ (k)∗ α(−k)∗ − α(−k)α⊥ (k) 2 |E⊥ (k)| = K(k)2 Now, in view of Eq. (14.47), when ignoring time, the squared absolute value of the magnetic field appearing in Eq. (14.61), reads   1 (ˆκ × α⊥ (k)∗ + κˆ × α⊥ (−k)) · (ˆκ × α⊥ (k) + κˆ × α⊥ (−k)∗ ) 2 |B⊥ (k)| = 2 c K(k)2 yielding c2 |B⊥ (k)|2 = 



(ˆκ × α⊥ (k)∗ ) · (ˆκ × α⊥ (k)) + (ˆκ × α⊥ (k)∗ ) · (ˆκ × α⊥ (−k)∗ ) K(k)2

(ˆκ × α⊥ (−k)) · (ˆκ × α⊥ (k)) + (ˆκ × α⊥ (−k)) · (ˆκ × α⊥ (−k)∗ ) + K(k)2



 (14.63)

ch014.tex

11/3/2011

16: 4

Page 417

14.2

ELECTROMAGNETIC FIELD HAMILTONIAN

417

Now, observe that the cross product of the dimensionless unit vector κˆ = k/k by the vector α⊥ (k) is a new vector α(k), the modulus of which α(k) remains |α⊥ (k)|, but which is orthogonal to the plane κˆ , according to k × α⊥ (k) = α(k) k Hence, the first scalar product involved on the first right-hand side of Eq. (14.63), reads κˆ × α⊥ (k) =

(ˆκ × α⊥ (k)∗ ) · (ˆκ × α⊥ (k)) = α(k)∗ · α(k) or, since the modulus α(k) of α⊥ (k) is the same as that of α(k) (ˆκ × α⊥ (k)∗ ) · (ˆκ × α⊥ (k)) = α(k)∗ α(k)

(14.64)

In like manner (ˆκ × α⊥ (k)∗ ) · (ˆκ × α⊥ (−k)∗ ) = α(k)∗ α(−k)∗ (ˆκ × α⊥ (−k)) · (ˆκ × α⊥ (k)) = α(−k)α(k) (ˆκ × α⊥ (−k)) · (ˆκ × α⊥ (−k)∗ ) = α(−k)α(−k)∗

(14.65)

Therefore, comparing Eqs. (14.64) to (14.65), Eq. (14.63) reads   α(k)∗ α(k) + α(−k) α(−k)∗ + α(k)∗ α(−k)∗ + α(−k) α(k) c2 |B⊥ (k)|2 = K(k)2 (14.66) As a consequence of Eqs. (14.62) and (14.66), the field energy (14.61) becomes after simplification    α(k)∗ α(k) + α(−k) α(−k)∗ ◦ H=ε d3k (14.67) K(k)2 Now, it is convenient to write the classical Hamiltonian as an expression involving Planck’s constant and the factor 21 , which will be of interest when passing to quantum mechanics. Thus, we write Eq. (14.67) as    α(k)∗ α(k) + α(−k) α(−k)∗ (14.68) H = ω(k) d3k 2 with, by identification of (14.67) and (14.68), K(k) reads

2ε◦ K(k) = ω(k)

(14.69)

Finally, changing −k into k, inside the last right-hand side of Eq. (14.68), that does not modify anything since this change concerns a scalar product, the classical Hamiltonian (14.68) takes the final form    α(k)∗ α(k) + α(k) α(k)∗ H = ω(k) (14.70) d3k 2

ch014.tex

418

14.3

11/3/2011

16: 4

Page 418

QUANTUM ELECTROMAGNETIC MODES

POLARIZED NORMAL MODES

Now, in view of Eq. (14.69), the constants defined by Eqs. (14.50) and (14.57) read  3/2   3/2  ω(k) ω(k) 1 1 and B ωk = (14.71) Eωk = ◦ 2π 2ε 2π 2ε◦ c2  Aωk =

1 2π

3/2



(14.72)

2ε◦ ω(k)

Next, recall that according to Eqs. (14.24)–(14.26), the fields E⊥ (k, t), B⊥ (k, t), and A⊥ (k, t) are transverse to the wave vector k, and thus to the corresponding unit vector κˆ = k/k [defined by Eq. (14.32)], so that each field, at any point k of the reciprocal space, may be considered as the sum of two perpendicular combinations both orthogonal to k: E⊥ (k, t) = Eε (k, t) + Eε (k, t)

(14.73)

B⊥ (k, t) = Bε (k, t) + Bε (k, t)

(14.74)

A⊥ (k, t) = Aε (k, t) + Aε (k, t)

(14.75) εˆ k

are the These two perpendicular combinations characterized by εˆ k and two polarized components of the different fields in the reciprocal space, the two polarization vectors εˆ k and εˆ k giving the directions of the polarized components of the fields being perpendicular to the unit vector κˆ characterizing the vector k and thus satisfying εˆ k · εˆ k = εˆ k · κˆ = κˆ · εˆ k = 0 εˆ k · εˆ k = εˆ k · εˆ k = κˆ · κˆ = 1 κˆ =

k k = |k| k

Hence, the polarized modes Eε (k, t), Aε (k, t), and Bε (k, t) are given by  ω(k) Eε (k, t) = i εˆ k (αε⊥ (k, t) − αε⊥ (−k, t)∗ ) 2ε◦

 εˆ k (αε⊥ (k, t) + αε⊥ (−k, t)∗ ) Aε (k, t) = 2ε◦ ω(k)  Bε (k, t) = i

ω(k) (ˆκ × εˆ k )(αε⊥ (k, t) + αε⊥ (−k, t)∗ ) 2ε◦ c2

(14.76)

(14.77)

(14.78)

with αε⊥ (k, t) = εˆ k · α⊥ (k)

(14.79)

ch014.tex

11/3/2011

16: 4

Page 419

14.3

POLARIZED NORMAL MODES

419

while the polarized modes Eε (k, t), Aε (k, t), and Bε (k, t), which are orthogonal to Eε (k, t), Aε (k, t), and Bε (k, t), appear to be given by similar expressions by changing in Eqs. (14.76)–(14.78) the unit vector εˆ k into εˆ k and the corresponding index ε by that ε , that is,  ω(k)  Eε (k, t) = i εˆ (αε ⊥ (k, t) − αε ⊥ (−k, t)∗ ) 2ε◦ k

Aε (k, t) =

 B (k, t) = i ε

 2ε◦ ω(k)

εˆ k (αε ⊥ (k, t) + αε ⊥ (−k, t)∗ )

ω(k) (ˆκ × εˆ k )(αε ⊥ (k, t) + αε ⊥ (−k, t)∗ ) 2ε◦ c2

with αε ⊥ (k, t) = εˆ k · α⊥ (k)

(14.80)

Next, in order to get the electromagnetic field in Euclidian geometric space, take the Fourier transforms of Eqs. (14.73)–(14.75), after changing k into −k within the last integrals [as for the passage from Eqs. (14.48) and (14.49) to Eqs. (14.51) and (14.52)], then, after using Eqs. (14.71) and (14.72), the following description of the electromagnetic field with the geometric space is obtained:  E⊥ (r, t) = i Eωk d 3 k × εˆ k (αε⊥ (k, t)eik·r − αε⊥ (k, t)∗ e−ik·r ) + εˆ k (αε ⊥ (k, t)eik·r − αε ⊥ (k, t)∗ e−ik·r )}

(14.81)

 A⊥ (r, t) =

Aωk d 3 k × {ˆεk (αε⊥ (k, t)eik·r + αε⊥ (k, t)∗ e−ik·r ) + εˆ k (αε ⊥ (k, t)eik·r + αε ⊥ (k, t)∗ e−ik·r )}

(14.82)

 B⊥ (r, t) = i

Bωk d 3 k

× {(ˆκ׈εk ){αε⊥ (k, t)eik·r − αε⊥ (k, t)∗ e−ik·r } + (ˆκ׈εk ){αε ⊥ (k, t)eik·r − αε ⊥ (k, t)∗ e−ik·r }}

(14.83)

ch014.tex

420

11/3/2011

16: 4

Page 420

QUANTUM ELECTROMAGNETIC MODES

14.4 14.4.1

NORMAL MODES OF A CAVITY Fields and corresponding Hamiltonians

Now, suppose that the electromagnetic field is enclosed in a cubic box of length L. Then, as seen above, not all possible values of the k wave vectors are permitted, only certain discrete values depending on the boundary conditions. Then, the continuous wave vectors k appearing in Eqs. (14.81)–(14.83), now transform into discrete wave vectors kn . Moreover, if the length L is so large that L >> λMax where λMax = 2π/|kMin | and where |kMin | is the smallest modulus of the wave vector, it is possible to neglect the effects occurring near the walls of the container and thus to describe the electromagnetic field in terms of a set of discrete components so that the discrete wave vector must satisfy   2π kn = (nx xˆ + ny yˆ + nz zˆ ) L where xˆ , yˆ , and zˆ are the unit vectors along the Cartesian coordinates. In a similar way, the angular frequency ω(k) defined by Eq. (14.37) depending continuously on the modulus k of the wave vector k, transforms to discrete angular frequency ωn according to ω(k) = ck → ωn = ckn with

 kn = |kn | =

 2π 2 nx + ny2 + nz2 L

Hence, the continuous variables αε⊥ (k, t) and αε ⊥ (k, t) defined by Eqs. (14.79) and (14.80) transform into discontinuous variables αε⊥ (t) and αε ⊥ (t): αε⊥ (k, t) →αnε⊥ (t)

and

αε ⊥ (k, t) →αnε ⊥ (t)

Then, after such transformation, the classical Hamiltonian of the electromagnetic field (14.70) involving an integral must transform into the following one involving now a sum, according to  ∗ 

αn αn + αn α∗n H= (14.84) ωn 2 n Now, observe that the change when passing to an infinite space to a finite one of volume V = L 3 must to be compatible with the transformation of Eqs. (14.70) into Eq. (14.84), it is required that the electromagnetic fields (14.81)–(14.83) must be each multiplied by the factor (2π/L)3/2 . Then, one obtains from Eqs. (14.81)–(14.83), respectively, the following expressions: 

ωn E⊥ (r, t) = i 2ε◦ V n × {ˆεkn (αnε⊥ (t)eikn ·r − α∗nε⊥ (t)e−ikn ·r ) + εˆ kn (αnε ⊥ (t)eikn ·r − α∗nε ⊥ (t)e−ikn ·r )}

(14.85)

ch014.tex

11/3/2011

16: 4

Page 421

14.4

A⊥ (r, t) =



n

NORMAL MODES OF A CAVITY

421

 2ε◦ ωn V

× {ˆεkn (αnε⊥ (t)eikn ·r + α∗nε⊥ (t)e−ikn ·r ) + εˆ kn (αnε ⊥ (t)eikn ·r + α∗nε ⊥ (t)e−ikn ·r )} B⊥ (r,t) = i



n

(14.86)

ωn 2ε◦ Vc2

× {(ˆκ × εˆ kn )(αnε⊥ (t)eikn ·r − α∗nε⊥ (t)e−ikn ·r ). + (ˆκ × εˆ kn )(αnε ⊥ (t)eikn ·r − α∗nε ⊥ (t)e−ikn ·r )}

(14.87)

We emphasize that the discrete dimensionless polarized components αnε⊥ (t) and αnε ⊥ (t), just as the components αε⊥ (k, t) and αε ⊥ (k, t) are variables, the magnitudes of which may alter when passing from some wave vector to another one.

14.4.2

Modes density

The total number n of electrical modes inside the cavity is equal to the number of discrete wave vectors kn times 2, because of the two polarization orientations of each wave vector. Its differential reads dn = 2dnx dny dnz

(14.88)

where nx , ny , and nz , which are momentarily considered as continuous variables, are related to the components of the wave vectors through       2πny 2πnx 2πnz kx = ky = kz = (14.89) L L L with kx = k · xˆ

ky = k · yˆ

Then, owing to (14.89), Eq. (14.88) reads  3 1 dn = 2V dkx dky dkz 2π

kz = k · zˆ

with

V = L3

Again, after passing to spherical coordinates, defined in Fig. 14.1, it becomes  3 1 dn = 2V k 2 dk sin θ dθ dφ (14.90) 2π with k=

(kx )2 + (ky )2 + (kz )2

Moreover, passing from the variable k to the corresponding angular frequency ω = kc, Eq. (14.90) yields   1 3 2 dn = 2V ω dω d (14.91) 2πc

ch014.tex

422

11/3/2011

16: 4

Page 422

QUANTUM ELECTROMAGNETIC MODES

z

r θ

y φ x

Figure 14.1 Polar spheric coordinates: x = r sin θ cos φ, y = r sin θ sin φ, and z = r cos θ; and 0 ≤ r < ∞, 0 ≤ θ ≤ π, and 0 ≤ φ ≤ 2π. r is the radial coordinate, θ and φ are respectively the inclination and azimuth angles.

d = sin θ dθ dφ the ω derivative of which being   dn = Vg(ω) dω d

(14.92)

g(ω) =

with

2ω2 (2πc)3

(14.93)

We find the number of modes of the electromagnetic field, which lie in the range between ω and dω. This may be obtained by summing dn given by Eq. (14.91) over the angle variables, allowing one to get the radial density of modes dρ(ω)/dω within the spherical shell lying between ω and ω + dω, that is,      dρ(ω) 1 3 2 d ω = 2V dω 2πc or, due to (14.92) 

dρ(ω) dω



 = 2V

1 2πc

3

π ω

2

2π sin θ dθ

0

dφ 0

Thus, after integration over the angular variables, it yields       dρ(ω) 1 1 3 2 ω2 ω =V = 8πV dω 2πc π 2 c3 or,     dρ(ω) 1 ω2 = Vf (ω) with f (ω) = dω π 2 c3

(14.94)

ch014.tex

11/3/2011

16: 4

Page 423

14.5

14.5

QUANTIZATION OF THE ELECTROMAGNETIC FIELDS

423

QUANTIZATION OF THE ELECTROMAGNETIC FIELDS

14.5.1

Standard Lagrangian within the Coulomb gauge

14.5.1.1 The Lagrangian In order to quantize the electromagnetic field, it is convenient to refer to the standard Lagrangian of a system of charged particles interacting with the electromagnetic field. Thus, differing from the previous sections, it is convenient to take into account the charged particles by considering the standard Lagrangian L(r, t) of such a system within the Coulomb gauge, which is given by 

1 2 L(r, t) = mα r˙ α (t) − VC (t) + LC (r, t) d 3 r (14.95) 2 α with, respectively, LC (r, t) = ε◦



˙ t) − c2 (∇ × A(r, t))2 A(r, 2

+ J(r, r˙ , t) · A(r, t)

(14.96)

Consider the αth particle. In these equations mα is the mass of the particle, r˙ α (t) the time derivative of the position coordinate rα (t):   ∂rα (t) r˙ α (t) = ∂t ˙ t) is the time derivative of the vector potential at the r position, that is, whereas A(r,   ˙ t) = ∂A(r, t) A(r, ∂t and J(r, t) is the current density defined by

qβ r˙ β (t)δ(r − rβ (t)) J(r, r˙ , t) =

(14.97)

β

in which qβ is the electrical charge of the charged β particle, VC is the Coulomb potential defined by

qα q β 1

VC (t) = εCoul α + ◦ 4πε |rα (t) − rβ (t)| α α>β β

with εCoul

α

q2 = α◦ 2ε



1 2π

3 

1 3 d k k2

The Lagrangian (14.95) may be considered as a very general postulate of the electromagnetic theory from which it is possible to deduce the Maxwell equations (14.1)–(14.4) and the Lorentz force law mα r¨ α = qα {E(rα (t), t) + r˙ α (t) × B(rα (t), t)} keeping in mind that all the other symbols have the same meaning as above.

ch014.tex

424

11/3/2011

16: 4

Page 424

QUANTUM ELECTROMAGNETIC MODES

Now, when passing from the geometric space to the reciprocal one, the standard Lagrangian may be shown1 to transform into  

1 |ρ(k, t)|2 3  L(k, t) = mα r˙ α2 (t) − d k + LC (k, t) d 3 k (14.98) ◦k2 2 2ε α with



∗ · A(k,t)) 2 k 2 (A(k, t)∗ · A(k, t)) ˙ ˙ ( A(k, t) − c LC (k, t) = ε◦ + cc 2   J(k, t)∗ · A(k, t) + J(k, t) · A(k, t)∗ + (14.99) 2 ˙ t) its time derivative, and Here A(k, t) is the Fourier transform (14.16) of A(r, t), A(k, J(k, t) is the Fourier transform of J(r, t) defined by  3/2  1 J(k, t) = J(r, t)e−ik·r d 3 r 2π and ρ(k, t) is the Fourier transform of ρ(r, t):  3/2  1 ρ(k, t) = ρ(r, t)e−ik·r d 3 r 2π where ρ(r, t) is the charge density:

ρ(r, t) = qβ δ(r − rβ (t)) β

14.5.1.2 Conjugate momentum of rα (t) and A(k, t) In the Lagrange formalism, the conjugate momentum pα (t) of rα (t) is the partial derivative of the Lagrangian with respect to r˙ α (t):   ∂L(r, t) pα (t) = ∂˙rα (t) Hence, in the present situation where the Lagrangian is given by Eqs. (14.95) and (14.96), the conjugate momentum becomes    ∂J(r, t) pα (t) = mα r˙ α (t) + d3r ∂˙rα (t) or, due to Eq. (14.97), and after commuting the volume integral with the sum over β 

  ∂˙rβ (t)δ(r − rβ (t)) pα (t) = mα r˙ α (t) + qβ ·A(r, t) d 3 r ∂˙rα (t) β

so that

 pα (t) = mα r˙ α (t) + qα

δ(r − rα (t))·A(r, t)d 3 r

1 C. Cohen-Tannoudji, J. Dupont-Roc, and G. Grynberg. Atom-Photon Interactions: Basic Processes and Applications. Wiley Science Paperback Series: New York, 1998.

ch014.tex

11/3/2011

16: 4

Page 425

14.5

QUANTIZATION OF THE ELECTROMAGNETIC FIELDS

425

which simplifies to pα (t) = mα r˙ α (t) + qα A(rα (t), t)

(14.100)

or, more simply, p(t) = mα r˙ α (t) + qA(r(t), t) On the other hand, in the reciprocal space, the conjugate momentum π(k, t)∗ of ˙ A(k, t) is, by definition, the partial derivative of the Lagrangian with respect to A(k, t), that is, the time derivative of A(k, t), yielding    ∂L(k, t) π(k, t)∗ = ˙ ∂A(k, t) where the Lagrangian  L(k, t) involved in the partial derivative is given by Eq. (14.98), so that due to this equation, it becomes    ∂ LC (k , t) d 3 k π(k, t)∗ = ˙  , t) ∂A(k or, in view of Eq. (14.99),    3  ˙  ∗ ˙  ∗ ◦ ∂ A(k , t) · A(k , t)/d k π(k, t) = ε ˙ ∂A(k, t) so that ˙ π(k, t)∗ = ε◦ A(k, t)∗

and thus

˙ π(k, t) = ε◦ A(k, t)

(14.101)

Now, recall that the vector potential A⊥ (k, t) being perpendicular to k, may be decomposed into two polarized vectors according to Eq. (14.75), that is, A⊥ (k, t) = Aε (k, t) + Aε (k, t) which is also true for the time derivative of A⊥ (k, t), that is, ˙ ε (k, t) + A ˙ ε (k, t) ˙ ⊥ (k, t) = A A Hence, the conjugate momentum appearing in (14.101) may be decomposed in the same way according to π⊥ (k, t) = πε (k, t) + πε (k, t)

14.5.2

Quantization in the Schrödinger picture

14.5.2.1 Field quantization in infinite space It is possible to find the quantum operators corresponding to the classical electromagnetic field, keeping in mind that within the Schrödinger picture, the operators do not depend on time. Quantizing the system of material particles, requires, for each α charged particle, to impose on the x, y, and z components (rα )k of rα , and on the x, y, and z components, and also on the corresponding components (pα )j of pα the condition that they are operators obeying the commutation rules: [{(rα )k SP }, {(pα )j SP }] = iδαα δjk

(14.102)

ch014.tex

426

11/3/2011

16: 4

Page 426

QUANTUM ELECTROMAGNETIC MODES

In a similar way, it is convenient to undertake the quantization of the electromagnetic field in the reciprocal space, by assuming that the potential vector Aε (k) and its conjugate momentum πε (k )∗ become, respectively, operators Aε (k) and πε (k )† obeying the commutation rule [{Aε (k)SP }, {πε (k )SP }† ] = i δεε δ(k − k )

(14.103)

Next, since, due to Eq. (14.22), the conjugate momentum πε (k) of Aε (k) defined by Eq. (14.101) is related to the electric field Eε (k) through {πε (k)SP } = −ε◦ {Eε (k)SP }

(14.104)

We have, after changing the classical field Eε (k )∗ , the complex conjugate of Eε (k ) into the quantum operator Eε (k )SP† the Hermitian conjugate of Eε (k )SP , one obtains from Eq. (14.103) the following commutator between Aε (k)SP and its conjugate momentum −ε◦ Eε (k )SP† : 1 δεε δ(k − k ) (14.105) ε◦ Moreover, having obtained the quantum commutation rule dealing with the timeindependent SP quantum operators describing the electromagnetic field in the reciprocal space, it is possible to obtain the corresponding SP operators in the geometric space, by performing the following time-independent transformations, analogous to the time-dependent ones (14.81)–(14.83), applied to the classical fields  SP E⊥ (r) = i Eωk d 3 k [{Aε (k)SP }, {Eε (k )SP }† ] = −i

× {ˆεk (aε (k)eikn ·r − aε (k)† e−ikn ·r ) + εˆ k (aε (k)eikn ·r − aε (k)† e−ikn ·r )} (14.106)  A⊥ (r)SP =

Aωk d 3 k × {ˆεk (aε (k)eikn ·r + aε (k)† e−ikn ·r ) + εˆ k (aε (k)eikn ·r + aε (k)† e−ikn ·r )} (14.107)  B⊥ (r)SP = i

Bωk d 3 k

× {(ˆκ × εˆ k ){aε (k)eikn ·r − aε (k)† e−ikn ·r } +(ˆκ × εˆ k ){aε (k)eikn ·r − aε (k)† e−ikn ·r }}

(14.108)

In these equations, the time-independent operators aε (k) and their Hermitian conjugates aε (k)† replace, respectively, the time-dependent normal modes αε⊥ (k, t) and αε⊥ (k, t)∗ appearing in Eqs. (14.81)–(14.83 ), with, in order to satisfy Eq. (14.105), the following commutation rule between aε (k) and aε (k)† : [aε (k), aε (k )† ] = δεε δ(k − k )

(14.109)

ch014.tex

11/3/2011

16: 4

Page 427

14.5

QUANTIZATION OF THE ELECTROMAGNETIC FIELDS

427

Moreover, due to Eq. (14.70) giving the energy of the electromagnetic field expressed in terms of αε⊥ (k) and of its complex conjugate αε⊥ (k)∗ , the corresponding Hamiltonian operator HSP may be obtained from this last expression by replacing, respectively, in it αε⊥ (k), and αε⊥ (k)∗ by aε (k) and aε (k)† :    aε (k)† aε (k) + aε (k)aε (k)† d3k HSP = ω(k) 2 Then, using the commutator (14.109) the Hamiltonian transforms to    1 (14.110) HSP = H(k)SP d 3 k with H(k)SP = ω(k) aε (k)† aε (k) + 2 14.5.2.2 Field quantization inside a cavity In a cavity of volume V , the operators describing the fields corresponding to the classical fields defined by Eqs. (14.85)–(14.87) may be obtained by proceeding in passing from Eqs. (14.85)–(14.87) to Eqs. (14.106)–(14.108): 

ωn SP E⊥ (r) = i 2ε◦ V n † −ikn ·r × {ˆεkn (anε eikn ·r − anε e ) † −ikn ·r + εˆ kn (anε eikn ·r − anε )} e

SP

A⊥ (r)

=



n

(14.111)

 2ε◦ ω

nV

† −ikn ·r × {ˆεkn (anε eikn ·r + anε e ) † −ikn ·r + εˆ kn (anε eikn ·r + anε )} e SP

B⊥ (r)

=i

n



(14.112)

ωn 2ε◦ Vc2

† −ikn ·r × {(ˆκ × εˆ kn ){anε eikn ·r − anε e } † −ikn ·r + (ˆκ × εˆ kn )(anε eikn ·r − anε )} e

(14.113) α∗nε⊥

have been where the classical variables αnε⊥ and their complex conjugates † replaced, respectively, by the operators anε and their Hermitian conjugates anε required to obey the commutation rules † [anε , amε  ] = δεε δnm

(14.114)

The commutators of the Cartesian components A⊥ (r)l SP and E⊥ (r)l SP with l = x, y, z of operators A⊥ (r)SP and E⊥ (r)SP may be proved to be given by [A⊥ (r)l SP , E⊥ (r)j SP ] = −iε◦−1 δTlj (r − r )

(14.115)

where the last right-hand-side term is the transverse Dirac delta function defined by   3   k i kj 1  ik·(r−r ) T (e ) δlj − 2 d 3 k δlj (r − r ) = 2π k

ch014.tex

428

11/3/2011

16: 4

Page 428

QUANTUM ELECTROMAGNETIC MODES

in which the ki or kj are the Cartesian components of the k vector. Observe that when i = j, the commutator (14.115) reduces to    3  kl2 1 SP SP ◦−1 ik·(r−r ) [A⊥ (r)l , E⊥ (r)l ] = −iε (e ) 1 − 2 d3k 2π k Moreover, the Hamiltonian operator HSP corresponding to the electromagnetic energy (14.84), takes the form  

1 † HSP = Hn SP with Hn SP = ωn anε anε + (14.116) 2 n 14.5.2.3 Eigenvalue equation of the electromagnetic field Hamiltonians We observe that the continuous and discrete sums of Hamiltonians given, respectively, by Eq. (14.110) or (14.116) have the same structure as that (5.9) of the quantum harmonic oscillator, and involve the basic commutation rules (14.109) and ( 14.114), which have also the same structure as that (5.5) dealing with the usual quantum harmonic oscillator. As a consequence, all that has been found for the quantum harmonic oscillator holds also for the Hamiltonians involved in these equations, so that the following eigenvalue equations equivalent to (5.40) read in the present situation     {aε (k)† aε (k)} + 21 |{lε (k)} = lε (k) + 21 |{lε (k)} (14.117) and



† {anε anε } +

1 2



  |{lnε } = lnε + 21 |{lnε }

(14.118)

with for each lε (k) or lnε lnε = 0, 1, 2, . . .

and

lε (k) = 0, 1, 2, . . .

and where the |{lε (k)} and |{lnε } are, respectively, the eigenvectors of {aε (k)† aε (k)} † a } obeying the orthonormality properties and {anε nε

{lnε }|{jnε } = δlnε

jnε

and

{lε (k)}|{jε (k)} = δlε (k) jε (k)

(14.119)

Hence, due to Eqs. (14.110) and (14.116), the eigenvalue equations (14.117) and (14.118) read   H(k)|{lε (k)} = ω(k) lε (k) + 21 |{lε (k)} (14.120)   Hn |{lnε } = ωn lnε + 21 |{lnε }

(14.121)

The quantum numbers lε (k) or lnε are, respectively, the excitation degrees of the continuous mode characterized by the wave vector k and that of the nth discrete mode within the polarization ε. Hence, when the field is in one state |{lε (k)} of the continuous situation, or in one |{lnε } of the discrete case, the corresponding quantum numbers lε (k) or lnε may be viewed through Eqs. (14.120) and (14.121), as the number of energy packets ω(k) or ωn inside the corresponding modes of the electromagnetic field. These energy packets may be considered as light corpuscles, which are called photons.

ch014.tex

11/3/2011

16: 4

Page 429

14.5

QUANTIZATION OF THE ELECTROMAGNETIC FIELDS

429

Now, owing to the correspondence between quantum electromagnetic modes and quantum harmonic oscillators, it is clear that Eqs. (5.53) and (5.63) may be applied to the ladder operators of the electromagnetic field to yield, respectively, for the continuous and discrete case  aε (k)|{lε (k)} = lε (k)|{lε (k) − 1} aε (k)† |{lε (k)} =

 lε (k) + 1|{lε (k) + 1}  lnε |{lnε − 1}

(14.122)

 lnε + 1|{lnε + 1}

(14.123)

anε |{lnε } = † anε |{lnε } =

14.5.3

Heisenberg picture fields

When passing to the Heisenberg picture, the SP time-independent ladder operators † , or a (k) and a† (k), become time dependent, that is, a (t) of the field anε and anε ε nε ε † and anε (t) or aε (k, t) and aε† (k, t). For the situation of electromagnetic fields enclosed in a box, the time dependence of anε (t) is given by the Heisenberg equations (3.94) involving the Hamiltonian Hnε , which read   ∂anε (t) i = [anε (t), Hnε ] ∂t or, due to Eq. (14.116), giving the expression of the total Hamiltonian H of the field, which is the same in the Heisenberg and Schrödinger pictures when an isolated electromagnetic field is considered   ∂anε (t) † i (t)anε (t)] = ωn [anε (t), anε ∂t Again, using the commutator (14.114), which reads † [anε (t), anε (t)] = 1

this equation transforms to



∂anε (t) ∂t

 = −iωn anε (t)

the solution of which is anε (t) = anε (0)e−iωn t On the other hand, for the free space, the Heisenberg equation governing the aε (k, t) is given by   ∂aε (k, t) i = [aε (k, t), H(k)] ∂t where the total Hamiltonian of the field is now given by Eq. (14.110), the solution of this equation being aε (k, t) = aε (k,0)e−iω(k)t

ch014.tex

430

11/3/2011

16: 4

Page 430

QUANTUM ELECTROMAGNETIC MODES

On the other hand, in the free space, the HP operators corresponding to the SP fields (14.111)–(14.113), become  {E⊥ (r, t)HP } = i Eωk d 3 k × {ˆεk (aε (k)eik·r e−iω(k)t − aε (k)† e−ik·r eiω(k)t ) + εˆ k (aε (k)eik·r e−iω(k)t − aε (k )† e−ikn ·r eiω(k)t )} (14.124)  {A⊥ (r, t)HP } =

Aωk d 3 k × {ˆεk (aε (k)eik·r e−iω(k)t + aε (k)† e−ik·r eiω(k)t ) + εˆ k (aε (k)eik·r e−iω(k)t + aε (k)† e−ik·r eiω(k)t )}

 {B⊥ (r, t)

HP

}= i

Bωk d 3 k

× {(ˆκ × εˆ k ){aε (k)eik·r e−iω(k)t − aε (k)† e−ik·r eiω(k)t } + (ˆκ × εˆ k ){aε (k)eik·r e−iω(k)t − aε (k)† e−ik·r eiω(k)t }} whereas within a cavity of volume V the HP operators corresponding to the SP fields (14.106)–(14.108) take the form 

ωn HP {E⊥ (r, t) } = i 2ε◦ V nε † −ikn ·r iωn t × {ˆεkn (anε eikn ·r e−iωn t − anε e e ) † −ikn ·r iωn t + εˆ kn (anε eikn ·r e−iωn t − anε e )} e

{A⊥ (r, t)

HP

}=



nε

(14.125)

 2ε◦ ωn V

† −ikn ·r iωn t × {ˆεkn (anε eikn ·r e−iωn t + anε e e ) † −ikn ·r iωn t + εˆ kn (anε eikn ·r e−iωn t + anε e )} e

{B⊥ (r, t)

HP

}= i

nε



(14.126)

ωn 2ε◦ Vc2

† −ikn ·r iωn t × {(ˆκ × εˆ kn ){anε eikn ·r e−iωn t − anε e e } † −ikn ·r iωn t + (ˆκ × εˆ kn ){anε eikn ·r e−iωn t − anε e }} (14.127) e

14.5.4

Average values of electromagnetic field operators

14.5.4.1 Analogies between A⊥ (r, t)SP and Q and between E⊥ (r, t)SP and P Observe that the SP operators describing the electric and magnetic potential vector

ch014.tex

11/3/2011

16: 4

Page 431

14.5

QUANTIZATION OF THE ELECTROMAGNETIC FIELDS

431

fields defined by Eqs. (14.111) and (14.112) may be written

{E⊥ (r)SP } = {Enε (r)SP +Enε (r)SP } and nε

{A⊥ (r) } = SP

{Anε (r)SP +Anε (r)SP }

nε

with, respectively,



{Anε (r) } = εˆ kn SP

 {Enε (r) } = iεˆ kn SP

 † (e−kn ·r )} {anε (eikn ·r ) + anε 2ε◦ ωn V

ωn † (e−ikn ·r )} {anε (eikn ·r ) − anε 2ε◦ V

(14.128)

and similar expressions for Anε (r)SP and Enε (r)SP by changing εˆ into εˆ  . Besides, keeping in mind the expressions of the operator Q and of its conjugate momentum P given, respectively, by Eqs. (5.6) and (5.7), that is,    Mω † † and P=i Q= (a + a) (a − a) 2Mω 2 where a† and a are, respectively, the lowering and raising operators of the oscillator, whereas M is its reduced mass and ω its angular frequency, it is of interest to remark the analogy between the vector potential operator Anε (r)SP of the electromagnetic field and the coordinate operator Q of the quantum harmonic oscillator and between the electric field operator Enε (r)SP and the momentum operator P conjugate of Q. 14.5.4.2 Mean values performed over Hamiltonian eigenstates Now, write the average value of the electric field operator Enε (r)SP on the eigenstates |{lnε } defined by the eigenvalue equation (14.121), that is,  ωn SP † (e−ikn ·r ))|{lnε }

{lnε }|(anε (eikn ·r ) − anε

{lnε }|{Enε (r) }|{lnε } = iεˆ kn 2ε◦ V Then, due to Eqs. (14.122) and (14.123), it reads   ωn SP {(eikn ·r ) lnε {lnε }|{lnε − 1}

{lnε }|{Enε (r) }|{lnε } = iεˆ kn ◦ 2ε V  −ikn ·r ) lnε + 1 {lnε }|{lnε + 1} } − (e or, due to the orthogonality relations (14.119)

{lnε }|{Enε (r)SP }|{lnε } = 0

(14.129)

In a similar way, one would obtain

{lnε }|{Anε (r)SP }|{lnε } = 0

(14.130)

Results (14.129) and (14.130) are for the electromagnetic fields the equivalent of those (5.85) and (5.93) dealing with harmonic oscillator, that is,

{n}|Q|{n} = 0

and

{n}|P|{n} = 0

ch014.tex

432

11/3/2011

16: 4

Page 432

QUANTUM ELECTROMAGNETIC MODES

14.5.4.3 Mean values performed over coherent states The average values of the electromagnetic fields over coherent states |{αnε } obeying the eigenvalue equation anε |{αnε } = αnε |{αnε } read



{αnε }|{Enε (r) }|{αnε } = iεˆ kn SP

with

{αnε }|{αnε } = 1

ωn {αnε (eikn ·r ) − α∗nε (e−ikn ·r )} 2ε◦ V

(14.131)

where the coherent states are described by the following expansions, which are analogous to that of (6.16), |{αnε } = e−|αnε |

2 /2

(αnε )lnε |{lnε } √ lnε ! l

(14.132)

nε

In a similar way, one would obtain for the magnetic potential vector averaged over coherent states

 SP

{αnε }|{Anε (r) }|{αnε } = εˆ kn {αnε (eikn ·r ) + α∗nε (e−ikn ·r )} (14.133) ◦ 2ε ωn V Of course, when passing to the Heisenberg picture, the Schrödinger picture time-independent operators given by Eqs. (14.131) and (14.133) become time dependent, so that, owing to Eqs. (14.125) and (14.126), they take, respectively, the forms



{αnε }|{Anε (r, t)HP }|{αnε } = εˆ kn {αnε (eikn ·r )(e−iωn t ) 2ε◦ ωn V + α∗nε (e−ikn ·r )(eiωn t )}

(14.134)

and 

{αnε }|{Enε

(r, t)HP }|{α

nε }

= iεˆ kn

ωn {αnε (eikn ·r )(e−iωn t ) − α∗nε (e−ikn ·r )(eiωn t )} 2ε◦ V (14.135)

Observe that the Heisenberg equations (14.134) and (14.135), and those corresponding to the other polarization εˆ kn , have the same structure as the corresponding components (14.85) and (14.86) defined by Eqs. (14.125) and (14.126), appearing in classical electromagnetic theory. Moreover, circular polarized light may be introduced with the help of a 2D coherent state, by aid of an equation similar to (6.61). It must be emphasized that, except for the fact that the commutators [{Anε (r)SP }, {Enε (r)SP }† ]

and

[{Aε (k)SP }, {Eε (k )SP }† ]

are more complicated than those between Q and P, a large part of the relations that have been found for quantum harmonic oscillators hold also for electromagnetic fields. The only differences lie in the presence of the phase factors e−ikn ·r and eikn ·r and also

ch014.tex

11/3/2011

16: 4

Page 433

14.5

QUANTIZATION OF THE ELECTROMAGNETIC FIELDS

433

in the changes occurring when passing from Q to Anε (r)SP and from P to Enε (r)SP , which are

     Mω ωn → εˆ kn and → −ˆεkn 2Mω 2ε◦ ωn V 2 2ε◦ V Owing to the deep analogies between the electric field and the operators P and also between the vector potential and the operator Q, it is possible to apply many results found above for operators Q and P, of quantum harmonic oscillators to the electric field and the vector potential operators. For instance, applying Eq. (6.50), it would appear that, in the Heisenberg picture, the average value of the squared electric field reads

{αnε }|{Enε (r, t)HP }2 |{αnε } ωn = −(ˆεkn )2 ◦ {(αnε e−ikn ·r eiωn t − α∗nε eikn ·r e−iωn t )2 − 1} 2ε V so that the relative dispersion Enε (t)/ Enε (t) of the HP electric field when it is in a coherent state reads 

{αnε }|(Enε (r, t)HP )2 |{αnε } − {αnε }|Enε (r, t)HP |{αnε } 2 Enε (t) = (14.136)

Enε (t)

{αnε }|Enε (r, t)HP |{αnε } Figure 14.2 gives the time dependence of the HP electric field averaged over different coherent states |{αnε } of increasing eigenvalues αnε and also the corresponding relative fluctuations (14.136) indicated by the thickness of the time dependence field function. As expected, the relative fluctuation is lowered when αnε is increasing, so that for αnε = 20, it still vanishes. This example illustrates how the electric field operator averaged over a coherent state approaches the classical electric field when the coherent state parameter becomes very large. It must be also observed that the ondulatory nature of light is described by the quantum linear operators describing the electromagnetic field, whereas the corresponding corpuscular nature of light is under the dependence of the kets (which are related to waves through wave mechanics) over which these operators are averaged. This is summarized in the following tabular expression (14.137): Physical Behaviour Quantum Entities

Examples

Wave

Hermitian operators E(r, t)HP , B(r, t)HP , A(r, t)HP

Corpuscle

Kets

|{αnε } , |{αε (k)} , |{lnε } , |{lε (k)}

Corpuscle

Wavefunctions

{r}|{αε (k)} , {r}|{lnε } (14.137)

More precisely, the number of electromagnetic particles within a given mode, that is, the number of photons of the corresponding angular frequency, may be obtained directly from the quantum number appearing in the eigenvalue equation (14.121) if the mode is an eigenstate of the Hamiltonian corresponding to this mode, or by a number proportional to the transition probability: {lα} } = | {lnε }|{αnε } |2 {Pnε

ch014.tex

434

11/3/2011

16: 4

Page 434

QUANTUM ELECTROMAGNETIC MODES

|αnε|2  4

〈{αnε}|Enε(r, t)HP|{αnε}〉

4

4

t

0

|αnε|2  40

〈{αnε}|Enε(r, t)HP|{αnε}〉 12

t

0 12

|αnε|2  400

〈{αnε}|Enε(r, t)HP|{αnε}〉 40

40

t

0

Figure 14.2 HP electric field averaged over different coherent states of increasing eigenvalue αnε and their corresponding relative dispersion pictured by the thickness of the time dependence field function.

or {lα} {Pnε } = e−|αnε |

2

|αnε |2lnε lnε

lnε !

if the mode is in the coherent state (14.132). Now, of course, it is possible to average the SP or HP electric field operator over squeezed states such as those (8.57) met in Section 8.2. All the results obtained in this section for the mean values of Q, Q2 , and Q averaged over the squeezed states and given, respectively, by Eqs. (8.79), (8.85), and (8.86), can be easily transposed to those of the electric field, its square, and its fluctuation.

14.5.5

Electromagnetic field spectrum

All the results obtained in the previous sections of this chapter hold irrespective of the angular frequency ω or of the corresponding frequency ν = ω/2π of the electromagnetic field. As it may be observed by inspection of Fig. 14.3, they apply to γ rays involved in radioactivity to X and ultraviolet (UV) rays, to visible light,

ch014.tex

11/3/2011

16: 4

Page 435

14.5

QUANTIZATION OF THE ELECTROMAGNETIC FIELDS

ν (Hz) 1019

435

λ (m) Y-rays

1011

X-rays

109

1018 1017

108

1016

UV-rays

1015

Visible light

1014

IR waves

1013

107

105 104

1012 1011

Microwaves

102 101

1010 109

1 Radio, TV

108

10

107

102

106 Figure 14.3

103 Longwaves Electromagnetic field spectrum.

to infrared, and to microwaves and beneath to radar and radio waves. Clearly, by inspection of Fig. 14.3 the frequency ν may vary over a very wide range, since being susceptible to be greater than 1019 Hz, for γ rays and around 106 Hz for radio long waves, whereas the corresponding wavelength λ = c/ν (where c is the velocity of light around 3.108 m s−1 ), may be smaller than 10−11 m for γ rays and around 103 m for long radio waves.

14.5.6

Long wavelength approximation for electric field

We start from Eq. (14.135) in order to obtain the mean value of the polarized electromagnetic field along εˆ k averaged over a coherent state |{αε } :   ωn HP

{αnε }|Enε (r, t) |{αnε } = iεˆ kn {αnε eikn ·r e−iωn t − αnε ∗ e−ikn ·r eiωn t }d 3 k 2ε◦ (14.138)

ch014.tex

11/3/2011

436

16: 4

Page 436

QUANTUM ELECTROMAGNETIC MODES

where this coherent state is defined by anε |{αε } = αnε |{αε }

with

{αnε }|{αnε } = 1

Now, from inspection of Fig. 14.3, the wavelengths of the electromagnetic radiations used in molecular spectroscopy go from 3 × 10−4 for infrared to 3 × 10−7 meters for ultraviolet. Hence, since the modulus kn of the vector k involved in the scalar products k · r appearing in the arguments of the exponentials encountered in Eqs. (14.134) and (14.135), is the inverse of the wavelength kn =

2π λn

it follows that for this range of wavelengths the magnitude of the wave vector kn lies in the interval 2 × 104 ≤ kn ≤ 2 × 107 rad·m−1 Hence, taking |r| as 10 atomic radii, that is, a few angstroms, for example, 2 × 10−9 m, we have 10−5 ≤ kn · r ≤ 10−2 so that e±ikn ·r  1 which is the long wavelength approximation. Then, Eq. (14.138) simplifies to  ωn HP

{αnε }|{Enε (t) }|{αε } = iεˆ kn {αnε (e−iωn t ) − αnε ∗ (eiωn t )} 2ε◦ or, taking αε as real,

{αnε }|{Enε (t)HP }|{αnε } = i{E(ωn )}(e−iωn t − eiωn t ) = 2{E(ωn )} sin ωn t with

 E(ωn ) = εˆ kn

ωn αε 2ε◦

(14.139)

a result that, after introducing a phase −π/2, reads

{αnε }|Enε (t)HP |{αnε } = E(ωn )(eiωn t + e−iωn t )

(14.140)

As it appears, the mean value of the electric field averaged over the coherent state given by Eq. (14.140) appears to be an infinite sum of time-dependent electric fields depending continuously on ω and given by E(ωn , t) = E(ωn )(eiωn t + e−iω t ) n

(14.141)

Of course, the long wavelength approximation holds for microwaves, the wavelengths of which are greater than those of infrared radiations.

ch014.tex

11/3/2011

16: 4

Page 437

14.6

SOME THERMAL PROPERTIES OF THE QUANTUM FIELDS

437

14.6 SOME THERMAL PROPERTIES OF THE QUANTUM FIELDS 14.6.1

Black-body radiation law

Kirchhoff in 1859 asked how does the intensity of the electromagnetic radiation emitted by a black body (a perfect absorber, also known as a cavity radiator) depend on the frequency of the radiation (i.e., the color of the light) and the temperature of the body. The answer was given by Planck who described the experimentally observed black-body spectrum well. We consider the electromagnetic radiation in thermal equilibrium inside an enclosure of volume V whose walls are maintained at absolute temperature T . In this situation, photons corresponding to excitation degrees of the different modes of the electromagnetic radiation, which are continuously absorbed and reemitted by the walls. Thus, due to this mechanism, the radiation inside the container depends on the temperature of the walls. Of course, it is not necessary to investigate the details of the mechanism that brings about thermal equilibrium since general arguments of statistical mechanics suffice by the aid of a coarse-grained analysis to describe the thermal equilibrium situation. If we regard the radiation as a collection of photons, the total number of them inside the enclosure is not fixed but depends on the temperature of the walls. The different modes of the field are specified by equations such as (14.125)–(14.127). Moreover, the radiation field existing in thermal equilibrium inside the enclosure is completely described by the thermal averages of the number occupation of each mode of the field, or, equivalently, by the corresponding thermal energy averages. Hence, the density of energy U(ω, T ) of the electromagnetic field in the range between ω and ω + dω, may be obtained from the expression of the thermal average energies

H(ω, T ) of the electromagnetic modes of angular frequency ω, by multiplying them by the density of modes g(ω) obtained above, that is, U(ω, T ) = g(ω) H(ω, T )

(14.142)

For each mode of the field, and due to similarity between the Hamiltonian (14.116) and that of the usual quantum harmonic oscillator (5.9), it is clear that the thermal average energy is given by Eq. (13.29), that is,

H(ω, T ) =

ω ω + 2 eω/k B T − 1

(14.143)

Now, we already saw that the density g(ω) of modes of the electromagnetic field between ω and (ω + dω) per unit volume is given by (14.93), that is, g(ω) =

2ω2 (2πc)3

(14.144)

Hence, discarding the zero-point energy in Eq. (14.143), the energy density (14.142) of the electromagnetic field becomes   2 ω3 U(ω, T ) = (14.145) (2πc)3 eω/k B T − 1

ch014.tex

438

11/3/2011

16: 4

Page 438

QUANTUM ELECTROMAGNETIC MODES

1

U(ω) normalized

2500 K

2000 K

1500 K 1000 K

0

1.0

3.0

2.0

4.0

ω/1014Hz

Figure 14.4 Energy density U(ω) within a cavity for different temperatures. The U(ω) are normalized with respect to the maximum of the curve at 2500 K.

that is the Planck black-body radiation law, discovered by Max Planck, governing at equilibrium temperature, the energy density of the electromagnetic field enclosed in a cavity at temperature T . This energy density is reproduced in Fig. 14.4 for different temperatures. Planck’s law (14.145) is one of the most fundamental equations in physics and is experimentally well verified. The total electromagnetic energy within the cavity may be obtained by integrating the energy density by unit volume over ω and then multiplying it by the total volume V ◦ , that is, UTot (T ) = V

◦

∞ U(ω, T ) dω 0

so that comparing, Eq. (14.145) 2 UTot (T ) = V (2πc)3 ◦

∞  0



ω3 eω/k B T − 1

dω

(14.146)

Then, changing the variable x=

ω kBT

we have  ω = 3

kBT 

3

 x

3

and

dω =

kBT 

 dx

ch014.tex

11/3/2011

16: 4

Page 439

14.6

SOME THERMAL PROPERTIES OF THE QUANTUM FIELDS

the total energy (14.146) takes the form  4  ∞  3  2 kB x ◦ T4 UTot (T ) = V dx (2πc)3 3 ex − 1

439

(14.147)

0

Moreover, since the integral involves a dimensionless variable, it must lead to a dimensionless number that has to be finite because of the presence of exp x in the denominator of the integrand. Hence, it appears that the total energy is of the form UTot (T ) ∝ T 4 which shows that the total energy of the electromagnetic field inside the cavity is proportional to the fourth power of the absolute temperature. This is the StefanBoltzmann law. To go further in the calculation of UTot (T ), we require the integral involved in Eq. (14.147), which has the value ∞  3  x π4 dx = ex − 1 15 0

so that the Stefan–Boltzmann law reads more precisely UTot (T ) = σT 4 where σ is the Stefan–Boltzmann constant given by π kB4 ◦ V 60 (c)3 We emphasize that all the above results dealing with the black-body radiation hold for the whole electromagnetic spectrum, in particular for the spectrum of the cosmic microwave background2 appearing in Fig. 14.5. σ=

14.6.2 Einstein coefficients The Planck radiation law allows one to find the ratio of the Einstein absorption and emission coefficients. To get this, consider two energy levels of energy E1 and E2 with E1 < E2 , subjected to an electromagnetic field at thermal equilibrium, obeying therefore the black-body radiation law, with this field being able to induce changes in the time-dependent population N1 (t) and N2 (t) (Fig. 14.6). The time dependence of the response to E2 is given by the kinetic equation   dN2 (t) (14.148) = −A21 (ω)N2 (t) − B21 (ω)U(ω)N2 (t) dt Here U(ω) is the energy density of the electromagnetic field at angular frequency ω given by Eq. (14.145) and corresponding to the resonant situation ω= 2

E 2 − E1 

From J. C. Mather, et al., Astrophys. J., 354 (1990): L37–L49.

(14.149)

ch014.tex

440

11/3/2011

16: 4

Page 440

QUANTUM ELECTROMAGNETIC MODES

1

0 3 6

12

24

30

36

42

48

54

60

66

ν/1010 Hz Figure 14.5 Spectrum of the cosmic microwave background (squares) superposed on a 2.735 K black-body emission (full line). The intensities are normalized to the maximum of the curve.

E2

N 2(t)

E2

E2 B Uω A 21

A 21

ω

B12 U ω N 1(t)

E1

E1 Uω

Figure 14.6

E1 Uω

Einstein coefficients for two energy levels.

while A21 (ω) is the spontaneous emission coefficient of Einstein and B21 (ω) the corresponding induced emission coefficient at the angular frequency ω. Now, the depopulation of the ground state E1 is   dN1 (t) (14.150) = −B12 (ω)U(ω)N1 (t) dt and B12 (ω) the induced Einstein absorption coefficient obeying B21 (ω) = B12 (ω)

(14.151)

After some time has occurred, which is large with respect to the characteristic times of the system, that is, t → ∞, a steady state must obtain so that     dN2 (∞) dN1 (∞) = =0 (14.152) dt dt

ch014.tex

11/3/2011

16: 4

Page 441

14.6

SOME THERMAL PROPERTIES OF THE QUANTUM FIELDS

441

Beyond this time and owing to Eqs. (14.148) and (14.150), the steady-state condition (14.152) leads to A21 (ω)N2 (∞) + B21 (ω)U(ω)N2 (∞) = B21 (ω)U(ω, T )N1 (∞) and thus, after rearranging, to   B21 (ω)U(ω, T ) N2 (∞) = N1 (∞) A21 (ω) + B21 (ω)U(ω, T )

(14.153)

However, under the steady conditions, this ratio must obey the equilibrium Boltzmann distributions ratio defined by Eqs. (12.83) so that   e−E2 /k B T N2 (∞) (14.154) = −E /k T N1 (∞) e 1 B or, owing to Eq. (14.149),



N2 (∞) N1 (∞)



= e−ω/k B T

Hence, by identification of Eqs. (14.153) and (14.154) one obtains B21 (ω)U(ω, T ) = e−ω/k B T A21 (ω) + B21 (ω)U(ω, T )

(14.155)

or B21 (ω)U(ω, T ) = e−ω/k B T (A21 (ω) + B21 (ω)U(ω, T )) and B21 (ω)U(ω, T )(1 − e−ω/k B T ) = e−ω/k B T A21 (ω) so that, the ratio of the two Einstein coefficients reads   A21 (ω) (1 − e−ω/k B T ) = U(ω, T ) B21 (ω) e−ω/k B T and the ratio of the induced emission coefficients B21 (ω) times the energy density U(ω) with the spontaneous emission coefficient A21 (ω) yields   1 ω B21 (ω)U(ω, T ) = λ with λB = (14.156) A21 (ω) (e B − 1) kBT or, due to Eq. (13.36),



B21 (ω)U(ω, T ) A21 (ω)

 = n(λB )

where n(λB ) is the thermal average of the occupation number at the absolute temperature T and at the angular frequency ω given by 1 (14.157) (eλB − 1) Now, the energy density of the electromagnetic field is given by Eq. (14.145), that is,   ω3 2 U(ω, T ) = (2πc)3 eω/k B T − 1

n(λB ) =

ch014.tex

442

11/3/2011

16: 4

Page 442

QUANTUM ELECTROMAGNETIC MODES

so that the Einstein coefficients ratio becomes   A21 (ω) 2 (1 − e−ω/k B T ) 3 ω = B21 (ω) (2πc)3 e−ω/k B T (eω/k B T − 1) or, after simplification, 

A21 (ω) B21 (ω)

 =

 3 2 3 = 2 ν ω (2πc)3 c

(14.158)

Hence, the ratio of the spontaneous and induced Einstein coefficients increases with the third power of the frequency.

14.7

CONCLUSION

In this chapter the classical normal modes of the electromagnetic field were obtained by transforming the Maxwell equation from the geometric to reciprocal space. Then, showing that in reciprocal space the conjugate momentum of the vector potential is deeply related to the electric field, it was possible to quantize the electromagnetic field by assuming a commutation rule between the operators corresponding to the vector potential and the electric field of the same kind as that assumed for the position coordinate and its conjugate momentum. It was then possible to find for each electromagnetic mode of the reciprocal space that a Hamiltonian exists that has the same structure in terms of self-conjugate Hermitian ladder operators as that of the usual quantum harmonic oscillator, thereby allowing one to apply all the results obtained for that oscillators to the quantum electromagnetic modes, particularly all those dealing with the Hamiltonian eigenvalue equation and with coherent and squeezed states. It was also shown that the degree of excitation of the Hamiltonian eigenstates of normal modes of a given frequency may be viewed as the number of light corpuscles, that is, the number of photons having this frequency. Moreover, it is apparent that this corpuscular property is related to the electromagnetic field kets, and thus, keeping in mind the link between quantum mechanics and wave mechanics, to quantum wavefunctions describing the field. At the opposite, it became clear that the wave behavior of the electromagnetic field is the reflection of the Hermitian operators describing these fields. Moreover, applying to the electromagnetic normal modes the thermal properties of oscillators, it was possible to find the Planck black-body radiation law and the Stefan–Boltzmann law, and to get the relation between the spontaneous and induced Einstein emission coefficients.

BIBLIOGRAPHY C. Cohen-Tannoudji, J. Dupont-Roc, and G. Grinberg. Photons and Atoms. Wiley: New York, 1997. R. Loudon. The Quantum Theory of Light. Oxford University Press: New York, 1983. H. Louisell. Quantum Statistical Properties of Radiations. Wiley: New York, 1973.

CH015.tex

17/3/2011

11: 24

Page 443

15

CHAPTER

QUANTUM MODES IN MOLECULES AND SOLIDS INTRODUCTION Chapters of Parts II, III, and IV studied the properties of a single harmonic oscillator (Parts II and III) or a large population of such oscillators (Part IV). But, if one wishes to apply these properties to molecules or solids, it is first necessary to extract from these complex systems their normal vibrational modes, where for each of them all the atoms of such extended systems may be classically viewed as oscillating back and forth at the same angular frequency and at the same phase. Hence, as in the last chapter, which dealt with the normal modes of electromagnetic fields, the aim of the present chapter is to describe a method for determining the normal modes of a molecule and those of solids, the last approach leading after quantization of the normal modes to the concept of phonons, that is, to the quantum vibrational energy of a normal mode considered as a quasi-particle in a way that evokes the photons of the electromagnetic field modes.

15.1 15.1.1

MOLECULAR NORMAL MODES Obtainment of the normal modes

Consider a set of N harmonic oscillators of the same reduced masses m that are linearly coupled through the potential V (t) = V ◦ (t) + VInt (t) with, respectively, V ◦ (t) =

1 kii xi2 (t) 2 i

VInt (t) =

1  kij (xi (t) − xj (t))2 2 i

j =i

Quantum Oscillators, First Edition. Olivier Henri-Rousseau and Paul Blaise. © 2011 John Wiley & Sons, Inc. Published 2011 by John Wiley & Sons, Inc.

443

CH015.tex

444

17/3/2011

11: 24

Page 444

QUANTUM MODES IN MOLECULES AND SOLIDS

where xi (t) is the time-dependent elongation of the ith oscillator. The force acting on the ith oscillator is     2 ∂V (t) d xi (t) = − (15.1) m dt 2 ∂xi Besides

 −

or

 −

∂V (t) ∂xi

∂V (t) ∂xi

 = −kii xi (t) −



kij (xi (t) − xj (t))

j=i

 = −(kii +



kij )xi (t) +

j =i

and thus

 −

with Kii =

∂V (t) ∂xi



kij

 =−



kij xj (t)

j=i



Kij xj (t)

(15.2)

j

Kij = −kij

and

j

Hence, owing to (15.2), the dynamics equations (15.1) yield  2   d xi (t) m =− Kij xj (t) 2 dt

(15.3)

j

which may be written in a matrix form according to ¨ M {X(t)} + K {X(t)} = {0}

(15.4)

¨ where {0} is the zero column vector, {X(t)} and {X(t)} are column vectors formed, respectively, by the set of positions xi (t) and accelerations x¨ i (t): ⎛ ⎞ ⎛ ⎞ x¨ 1 (t) x1 (t) ⎜ x¨ 2 (t) ⎟ ⎜ x2 (t) ⎟ ⎜ ⎟ ⎜ ⎟ ¨ {X(t)} =⎜ . ⎟ and {X(t)} = ⎜ . ⎟ (15.5) ⎝ .. ⎠ ⎝ .. ⎠ x¨ N (t)

xN (t)

whereas K is the matrix of the force constants Kij ⎛ K11 K12 … ⎜K21 K22 … K =⎜ ⎝… … … KN1 … … and M is the diagonal matrix M =m 1

⎞ K1N … ⎟ ⎟ … ⎠ KNN

CH015.tex

17/3/2011

11: 24

Page 445

15.1

MOLECULAR NORMAL MODES

445

where 1 is the unity matrix. Premultiply Eq. (15.4) by the inverse of the masses matrix M

−1

−1

¨ M {X(t)} + M

K {X(t)} = {0}

After simplification that gives ¨ {X(t)} + D {X(t)} = {0}

(15.6)

where D = M

−1

K

(15.7)

Next, introduce the diagonalization transformation of the matrix defined by Eq. (15.7) through λ = O

−1

D

O

(15.8)

where O is the eigenvector matrix, whereas λ is the eigenvalue matrix that is diagonal. Introduce within Eq. (15.6) between the matrix and the vector the diagonal unity matrix defined by −1

1 = O O that is, ¨ {X(t)} + D O O

−1

{X(t)} = {0}

Again, premultiply this last equation by the eigenvector matrix O

−1

¨ {X(t)} + O

−1

D O O

−1

{X(t)} = {0}

Then, in view of Eq. (15.8), this expression simplifies to ¨ {Y(t)} + λ {Y(t)} = {0}

(15.9)

with, respectively, ¨ {Y(t)} = O {Y(t)} = O

−1

−1

¨ {X(t)} {X(t)}

(15.10)

and where the λl are the eigenvalues of matrix (15.7). The linear transformation (15.10) reads  yl (t) = alk xk (t) k

where the alk are the components of the transformation matrix, whereas the yl (t) are the components of the column vector {Y(t)}. Equation (15.9) where the transformation matrix is diagonal, summarizes N-decoupled differential equations of the form y¨ l (t) + λl yl (t) = 0

(15.11)

CH015.tex

446

17/3/2011

11: 24

Page 446

QUANTUM MODES IN MOLECULES AND SOLIDS

that may be written



d 2 yl (t) dt 2

 = −2l yl (t)

with

2l = λl

(15.12)

The solutions of the decoupled differential equations are yl (t) = yl (0) sin(l t + l ) where l are phases. The yl (t) are the normal modes of vibration of the oscillator system in which all the parts of the system oscillate at the same angular frequency ll with the same phase l . Next, after multiplying both terms of Eq. (15.9) by the mass matrix involved in Eq. (15.4) ¨ M {Y(t)} + M λ {Y(t)} = {0} one obtains N decoupled equations of the form m y¨ l (t) + m2l yl (t) = 0

(15.13)

˜ which Now, within the normal modes description, the full classical Hamiltonian H, is by definition H˜ = T˜ + V˜ may be written as the sum of decoupled Hamiltonians:    1 1 2 2 2 ˜ ˜ ˜ H= Hl with Hl = pl (t) + ml yl (t) 2m 2

(15.14)

l

with pl (t) = m˙yl (t)

15.1.2

Quantization of the normal modes

In the Schrödinger picture, the operators do not change with time. Hence, in order to pass to quantum mechanics, we have to perform the change: yl (t) → Ql

and

pl (t) → Pl

where Ql and Pl are the time-independent operators corresponding, respectively, to the normal classical variables yl (t) and pl (t), which obey the commutation rule [Ql , Pk ] = iδlk Then, the classical Hamiltonian (15.14) transforms to a Hamiltonian operator H given by  H= Hl l

with Hl =

Pl2 1 + m2l Q2l 2 2m

CH015.tex

17/3/2011

11: 24

Page 447

15.1

MOLECULAR NORMAL MODES

447

The Hamiltonians Hl are those of quantum harmonic oscillators so that all that has been found above for quantum harmonic oscillators may apply to the Hl , in a way similar, for instance, to that used in passing from 1D to 3D harmonic oscillators. Hence, we write  Hl = l al† al + 21 where [ak , al† ] = δkl

Ql =  Pl = i

 (a† + al ) 2Ml l Ml † (al − al ) 2

Of course, all the results obtained for quantum oscillators and for their thermal properties hold for the normal modes of molecules.

15.1.3

Application to a system of two coupled oscillators

Applied to a situation where there are, for instance, only two oscillators, the column vectors (15.5) corresponding to the elongations and to their respective accelerations are given by     x1 (t) x¨ 1 (t) ¨ {X(t)} = and {X(t)} = (15.15) x2 (t) x¨ 2 (t) 

m M = 0

0 m





k + k12 K = 11 −k21

and

−k12 k22 + k21

 (15.16)

Now, let us look at the matrix D given by Eq. (15.7), that is, D = M

−1

K

(15.17)

Observe that since the matrix M is diagonal, its inverse is also diagonal and given by

M

−1

 1/m = 0

0 1/m



Thus, owing to Eq. (15.16), Eq. (15.7) takes the form    −k12 1/m 0 k11 + k12 D = −k21 0 1/m k22 + k21

CH015.tex

448

17/3/2011

11: 24

Page 448

QUANTUM MODES IN MOLECULES AND SOLIDS

Performing the matrix product gives



D = with

 ω11 =

ω12 =

2 −ω12

2 −ω21

2 ω22

 (15.18) 

k11 + k12 m 

2 ω11

and

ω22 = 

k12 m

and

ω21 =

k22 + k21 m k21 m

Then, the diagonalization transformation (15.8), that is, O

−1

D

O − λ =0

Now, when passing to the components Dij of the matrix, it reads  Dij Cj± − λ± Ci± = 0

(15.19)

j

where λ± are the two unknown eigenvalues of the λ diagonal matrix to be found, whereas the Cj± are the unknown components for the matrix   C1+ C1− O = C2+ C2− Since the λ± and Cj± are unknown, Eq. (15.19) corresponds to the following set of simultaneous equations: (D11 − λ)C1 + D12 C2 = 0

(15.20)

D21 C1 + (D22 − λ)C2 = 0

(15.21)

Since the Ci are different from zero, these two last equations are satisfied if the following determinant is zero:   (D − λ) D12   11 =0   D21 (D22 − λ) Expansion of the determinant following the usual rule leads to the second-order equation in λ: λ2 − (D11 + D22 )λ + (D11 D22 − D12 D21 ) = 0 The two solutions for λ are λ± = 21 [(D11 + D22 ) ±



(D11 + D22 )2 − 4(D11 D22 − D12 D21 )]

with, in view of Eq. (15.18), Dij = ωij2

with

i, j = 1, 2

(15.22)

CH015.tex

17/3/2011

11: 24

Page 449

15.1

MOLECULAR NORMAL MODES

449

When the two values of λ± have been obtained by the aid of Eqs. (15.22) and (15.21) in terms of ωij , it is possible with the help of Eq. (15.20) to find the expression of the components of the orthogonal matrix, that is, D12 C2− and (λ− − D11 ) Observe that the orthogonal transformation C1− =

O D O

C1+ =

−1

= λ

takes on, in the present situation, the following form:   2   2 ω11 ω12 C1+ C1− C1+ C2+ C2+

15.1.4

C2−

2 ω21

2 ω22

D12 C2+ (λ+ − D11 )

C1−

C2−

(15.23)  =

λ+ 1 0

0



λ− 1

Identification of symmetric molecules normal modes

When a molecule presents different symmetry elements, it may be of interest to classify its normal modes according to the different irreducible representations of the symmetry point group to which it belongs. That is particularly important in molecular vibrational spectroscopy. Section 18.9 gives some information on the symmetry point groups and on the irreducible representations giving in a compact form how the symmetry operations act. Equation (18.134) in Section 18.9 allows one to analyze the reducible representation  of any molecule belonging to a given point group in terms of the irreducible representation of that point group. This may be seen by studying how the atomic coordinates transform under the different symmetry operations of the point group. To illustrate that, such a procedure is now applied to the H2 O molecule, which admits two symmetry planes σv and σv , one belonging to the plane of the molecule and the other to the plane orthogonal to the first one and separating the molecule into two symmetrical parts, and also a rotational axis of symmetry C2 passing through the intersection of the two planes, as shown in Fig. 15.1. Then, it is shown that the reducible representation of H2 O is given by (18.122) in Section 18.9, that is, C2v ◦

E 9

C2 −1

σv 1

σv 3

where the numbers 9, −1, 1, and 3 are the characters for the four symmetry classes corresponding to the E, C2 , σv , and σv symmetry elements. Because of these symmetry elements and of the identity symmetry element E the H2 O molecule belongs to the C2v point group, the character table of which is given by tabular data in (18.110) in Section 18.9, i.e. C2v

E

C2

σv

σv

 A1  A2  B1 B2

1 1 1 1

1 1 −1 −1

1 −1 1 −1

1 −1 −1 1

Rot and Trans Tz Rz Ry, Tx Rx, Ty

(15.24)

CH015.tex

450

17/3/2011

11: 24

Page 450

QUANTUM MODES IN MOLECULES AND SOLIDS

z C2

σV

y σV

x Figure 15.1

Symmetry elements for a C2v molecule.

where the numbers in (15.24) are the characters {χk (Rr )} of the different irreducible representations k , that is, A1 , A2 , B1 , and B2 for the different symmetry classes Rr , that is, in the present situation E, C2 , σv , and σv . The presence in (15.24), on the lines corresponding, respectively, to A1 , B1 , and B2 , of the notations Tz , Tx , and Ty corresponding to the translations along the three Cartesian coordinates, means that these translations transform, according to these irreducible representations, the explanation being the same for Rz , Ry , and Rx corresponding to the rotations around the z, y, and x axis. Then, by application of Eq. (18.127) of Section 18.9, the reducible representation of the H2 O molecule appears to be ◦ = aA1 A1 ⊕ aA2 A2 ⊕ aB1 B1 ⊕ aB2 B2

(15.25)

where aA1 , aA2 , aB1 , and aB2 are numbers that indicate how often the corresponding irreducible representations k , that is, A1 , A2 , B1 , and B2 appear. Moreover, applying Eq. (18.134) of Section 18.9, that is, a k =

1  k ◦ {χ (Rr )}{χ (Rr )} g r

the components of the reducible representation (15.25) may be obtained using aA1 = a A2 = a B1 = a B2 =

1 4 {(9 × 1) ⊕ (−1 × 1) + (1 × 1) ⊕ (3 × 1)} = 3 1 4 {(9 × 1) ⊕ (−1 × 1) ⊕ (1 × −1) ⊕ (3 × −1)} = 1 4 {(9 × 1) ⊕ (−1 × −1) ⊕ (1 × 1) ⊕ (3 × −1)} = 1 4 {(9 × 1) ⊕ (−1 × −1) ⊕ (1 × −1) ⊕ (3 × 1)} =

1 2 3

CH015.tex

17/3/2011

11: 24

Page 451

15.2

PHONONS AND NORMAL MODES IN SOLIDS

A1

A1

ω1

ω2

451

B2 ω3 Figure 15.2 Three normal modes of a C2V molecule.

Hence, the reducible representation (15.25) becomes  = 3A1 ⊕ A2 ⊕ 2B1 ⊕ 3B2

(15.26)

From inspection of the C2v table of characters, it appears that the representations of the rotations Rx , Ry , and Rz and translations Tx , Ty , and Tz are, respectively, given by Rot = A2 ⊕ B1 ⊕ B2

(15.27)

Tr = A1 ⊕ B1 ⊕ B2

(15.28)

Then, the Vib normal modes representation is the difference between the  reducible representation (15.26) and those Rot and Tr given, respectively, by Eqs. (15.27) and (15.28) so that Vib = 2A1 ⊕ B2 Thus, it appears that one of the three normal modes of H2 O belongs to the irreducible representation B2 is symmetric with respect to the C2 and σv symmetry operations and antisymmetric with respect to σv operation, whereas the two other vibrational modes are fully symmetric since they belong to the irreducible representation A1 . This is shown in Fig 15.2.

15.2

PHONONS AND NORMAL MODES IN SOLIDS

Having determined the normal modes of molecules, we must determine those of solids. This is the aim of the present section. As for molecules, we shall begin by seeking the classical normal modes of the solid and then continue by quantizing them. However, the procedure to get the normal modes of solids will appear to completely differ from that we have used for molecules. The method used for solids will proceed from Fourier

CH015.tex

452

17/3/2011

11: 24

Page 452

QUANTUM MODES IN MOLECULES AND SOLIDS

transforms, allowing one to pass from a geometric space description to a new one in the reciprocal space, so that the quantization rules will be introduced for the normal mode coordinate and momentum components belonging to the reciprocal space.

15.2.1 Determination of the classical normal modes of a long chain of oscillators 15.2.1.1 Basic equations in geometric space We shall limit ourselves to a 1D approach to the solid normal modes first considered from a classical viewpoint. Consider an infinite linear chain of harmonic oscillators of angular frequency ω0 of mass m, coupled to each neighbor via the same force constant mω2 , the distance between two successive oscillators at equilibrium being L. Then, the force acting on the nth oscillator obeys the following equation:   2 d qn (t) = −mω02 qn (t) − mω2 {(qn (t) − qn+1 (t)) + (qn (t) − qn−1 (t))} (15.29) m dt 2 where qn (t) is the time-dependent displacement of the nth oscillator with respect to its equilibrium position. The solutions of this equation are qn (t) = {eiknL e−i(k)t + e−iknL ei(k)t }

(15.30)

where k is a continuous variable having the dimensions of inverse length, whereas (k) is given by  (k) = ω02 + ω2 (2 − eikL − e−ikL ) (15.31) That may be easily verified as follows. First, start from the second time derivative of qn (t) assumed to obey Eq. (15.30), which, due to ∂e±i(k)t = −ω2 e±i(k)t ∂t reads



d 2 qn (t) dt 2



= −(k)2 {eiknL e−i(k)t + e−iknL ei(k)t }

Then, using Eq. (15.31) yields   2 d qn (t) = −{ω02 + ω2 (2 − eikL − e−ikL )}{eiknL e−i(k)t + e−iknL ei(k)t } (15.32) dt 2 and, owing to the fact that e±ikL eiknL = eik(n±1)L

and

e±ikL e−iknL = eik(n∓1)L

it appears, with the help of Eq. (15.30), that (2 − eikL − e−ikL ){eiknL e−i(k)t + e−iknL ei(k)t } = 2qn (t) − (qn+1 (t) + qn−1 (t)) so that, introducing this result into Eq. (15.32) and after simplification and multiplication by m, Eq. (15.29) is obtained.

CH015.tex

17/3/2011

11: 24

Page 453

15.2

PHONONS AND NORMAL MODES IN SOLIDS

453

Next, passing in Eq. (15.31) from the imaginary exponentials to the corresponding trigonometric functions leads to

  kL 2 2 2 (k) = ω0 + 4ω sin (15.33) 2 so that e2in

◦π

= cos(2n◦ π) + i sin(2n◦ π) = 1

when

n◦ = ±1, ±2 . . .

Therefore, if k = k +

2n◦ π L

then 

eik L = eikL e2in Hence, it appears that Eq. (15.31) reads

◦π

= eikL



2n◦ π (k) =  k + L



so that all the information for (k) is confined within the following k interval: π π (15.34) − ≤k≤ L L which is called the first Brillouin zone. Next, from the angular frequency (k), one may get the phase velocity vφ (k) and the group velocity vG (k) defined, respectively, by (k) d(k) and vG (k) = k dk Now, observe that it is possible to write the following infinite sum involving the qn (t) governed by Eq. (15.29) via qn±1 (t) times e−iknL : vφ (k) =

+∞ 

qn±1 (t)e

−iknL

±ikL

=e

n=−∞

+∞ 

qn±1 (t)e−ik(n±1)L

n=−∞

Then, changing in the right-hand-side sum the n ± 1 terms into new ones n does not modify anything since the sum is infinite so that +∞ 

qn±1 (t)e−iknL = e±ikL

n=−∞

+∞ 

qn (t)e−iknL

(15.35)

n=−∞

15.2.1.2 Normal modes within the reciprocal space Now, introduce the following discrete Fourier expansions: ξ(k, t) =

+∞  n=−∞

qn (t)e−iknL

(15.36)

CH015.tex

454

17/3/2011

11: 24

Page 454

QUANTUM MODES IN MOLECULES AND SOLIDS

+∞ 

ζ(k, t) =

pn (t)e−iknL

(15.37)

n=−∞

where k is a continuous variable having the dimension of the inverse length, that is, a 1D wave vector, whereas the pn (t) are the momentum coordinates corresponding to the position coordinates qn (t), that is,   dqn (t) pn (t) = m (15.38) dt Owing to Eq. (15.38), Eq. (15.37) yields ζ(k, t) = m

 +∞   dqn (t) −iknL e dt n=−∞

or, due to Eq. (15.36),  ζ(k, t) = m

dξ(k, t) dt

 (15.39)

Note that, owing to Eqs. (15.36) and (15.37), ξ(−k, t) = ξ(k, t)∗

and

ζ(−k, t) = ζ(k, t)∗

Next, since k is continuous whereas n is discrete, the inverse transformations of Eqs. (15.36) and (15.37) are the following integral Fourier transforms working within the first Brillouin zone (15.34), that is, L qn (t) = 2π

L pn (t) = 2π

π/L ξ(k, t)eiknL dk

(15.40)

ζ(k, t)eiknL dk

(15.41)

−π/L

π/L −π/L

where, according to Eq. (15.34), k runs from −π/L to +π/L. Besides, the second time derivative of Eq. (15.36) reads  +∞  2  d 2 ξ(k, t) d qn (t) −iknL e = dt 2 dt 2 n=−∞ or, in view of Eq. (15.29), 

d 2 ξ(k, t) dt 2

 =−

+∞ 

{ω02 + ω2 {(qn (t) − qn+1 (t)) + (qn (t) − qn−1 (t))}}e−iknL

n=−∞

CH015.tex

17/3/2011

11: 24

Page 455

15.2

and thus 

d 2 ξ(k, t) dt 2



 =

−ω02

−ω

2

PHONONS AND NORMAL MODES IN SOLIDS

+∞ 

2

qn (t)e−iknL −

n=−∞

−

+∞ 

+∞ 

455

qn+1 (t)e−iknL

n=−∞



qn−1 (t)e

−iknL

(15.42)

n=−∞

Next, keeping in mind Eq. (15.35), that is, +∞ 

+∞ 

qn±1 (t)e−iknL = e±ikL

n=−∞

qn (t)e−iknL

(15.43)

n=−∞

It is possible to transform Eq. (15.42) into  +∞  2  +∞   d ξ(k, t) 2 2 −iknL +ikL 2 = −ω − ω q (t)e − e qn (t)e−iknL n 0 dt 2 n=−∞ n=−∞  +∞  −e−ikL qn (t)e−iknL n=−∞

so that, owing to Eq. (15.36), it simplifies to   2 d ξ(k, t) = −{ω02 + ω2 (2 − eikL − e−ikL )}ξ(k, t) dt 2 or, in view of Eq. (15.31),  2  d ξ(k, t) = −(k)2 ξ(k, t) dt 2

(15.44)

Clearly, irrespective of the value of the continuous 1D wave vector k, the second-order time derivative of ξ(k, t) depends on ξ(k, t) for the same value of k in a form that is that of an harmonic oscillator, so that the ξ(k, t) act as normal modes. Now, pass to the conjugate variables of these normal modes. The second time derivative of Eq. (15.39) reads   2 d 3 ξ(k, t) (d ζ(k, t) = m dt 2 dt 3 or  2    d ζ(k, t) d d 2 ξ(k, t) = m dt 2 dt dt 2 and thus, in view of Eq. (15.44),   2 d d ζ(k, t) = −m(k)2 (ξ(k, t)) dt 2 dt so that, owing to Eq. (15.39),  2  d ζ(k, t) = −(k)2 ζ(k, t) dt 2 a result that has the same form as that of (15.44)

(15.45)

CH015.tex

456

17/3/2011

11: 24

Page 456

QUANTUM MODES IN MOLECULES AND SOLIDS

15.2.2

Quantization of the long chain of oscillators

15.2.2.1 Mode quantization Now, within the Schrödinger picture of quantum mechanics, one has to consider ξ(k, t) and ζ(k, t) as operators ξ(k) and ζ(k), which do not depend on time, because of this chosen picture, and which obey the commutation rule [ξ(k), ζ(k)] = i

(15.46)

Then, introduce the two following dimensionless Hermitian self-conjugate operators, analogous to (5.6) and (5.7) used for the quantum oscillators, that is,

 (15.47) (a† (k) + a(k)) ξ(k) = 2m(k)  ζ(k) = i

m(k) † (a (k) − a(k)) 2

(15.48)

with, as a result of Eq. (15.46), the commutator [a(k), a† (k)] = 1

(15.49)

Just as ξ(k, t) and ζ(k, t) have been transformed into time-independent operators ξ(k) and ζ(k), the coordinates and momenta defined by Eqs. (15.40) and (15.41) become time-independent operators qn and pn : L qn = 2π

π/L ξ(k)e

iknL

dk

and

−π/L

L pn = 2π

π/L ζ(k)eiknL dk −π/L

which, by analogy with Eqs. (15.36) and (15.37), take the form ξ(k) =

+∞ 

qn e

−iknL

and

n=−∞

+∞ 

ζ(k) =

pn e−iknL

(15.50)

n=−∞

15.2.2.2 Hamiltonian obtainment The full Hamiltonian operator of the linear set of coupled oscillators related to the classical dynamic equation (15.29) reads HTot =

+∞ 

Hn + HInt

(15.51)

n=−∞

with, respectively,  Hn =

HInt

p2n 1 + mω02 qn2 2m 2



 +∞   1 2 2 mω (qn − qn+1 ) = 2 n=−∞

(15.52)

(15.53)

CH015.tex

17/3/2011

11: 24

Page 457

15.2

PHONONS AND NORMAL MODES IN SOLIDS

457

Even if the operators pn and qn are real, it is convenient, for reasons that will appear later, to write the full Hamiltonian (15.51) using Eq. (15.52) and (15.53) as HTot =

+∞ +∞ +∞ mω02  1  mω2  |pn |2 + |qn |2 + |qn − qn+1 |2 2m n=−∞ 2 n=−∞ 2 n=−∞

(15.54)

Moreover, owing to Eq. (15.43), which holds not only for the scalars qn (t) but also for the operators qn , it reads +∞ 

(qn − qn+1 )e−iknL =

n=−∞

+∞ 

qn e−iknL −

n=−∞

+∞ 

qn+1 e−iknL

n=−∞

a result that transforms, according to Eq. (15.43), into +∞ 

(qn − qn+1 )e−iknL =

n=−∞

+∞ 

qn e−iknL − eikL

n=−∞

+∞ 

qn e−iknL

n=−∞

or +∞ 

(qn − qn+1 )e

−iknL

= (1 − e

n=−∞

ikL

)

+∞ 

qn e−iknL

(15.55)

n=−∞

Then, apply the Bessel–Parseval relation (18.34) of Section 18.6 for a periodic function f (k), where the Cn are the expansion coefficients within the interval −L/2, L/2, that is, +∞ 

L |Cn | = 2π n=−∞

π/L | f (k)|2 dk

2

−π/L

leading to the following functions appearing in (15.56): Eqs.

+∞ 

f (k)

Cn e−iknL

n=−∞ +∞ 

(15.36) ξ(k)

n=−∞ +∞ 

(15.37) ζ(k) (15.55)

(1 − eikL )ξ(k)

n=−∞ +∞ 

qn e−iknL (15.56) pn e−iknL (qn − qn+1 )e−iknL

n=−∞

Then, one obtains, respectively, for the three sums involved in Eq. (15.54), the following relations: +∞ 

L |qn | = 2π n=−∞

π/L |ξ(k)|2 dk

2

−π/L

CH015.tex

458

17/3/2011

11: 24

Page 458

QUANTUM MODES IN MOLECULES AND SOLIDS

+∞ 

L |pn | = 2π n=−∞

π/L |ζ(k)|2 dk

2

+∞ 

L |qn − qn+1 | = 2π n=−∞

−π/L

π/L |(1 − eikL )ξ(k)|2 dk

2

−π/L

As a consequence of these three equations, the Hamiltonian (15.54) reads  π/L  mω02 1 L mω2 2 2 ikL 2 HTot = |ζ(k)| + |ξ(k)| + |(1 − e )ξ(k)| dk (15.57) 2π 2m 2 2 −π/L

Since the last squared modulus involved on the right-hand side of Eq. (15.57) is |(1 − eikL )ξ(k)|2 = 2|ξ(k)|2 (1 − cos kL) or, using the usual trigonometric relations, |(1 − e

ikL



)ξ(k)| = 4|ξ(k)| sin 2

2

2

kL 2



the Hamiltonian (15.57) becomes, HTot

L = 2π

π/L  −π/L

    m 1 2 2 2 kL 2 2 ω0 + 4ω sin |ξ(k)| + |ζ(k)| dk 2 2 2m

or, due to Eq. (15.33), HTot

L = 2π

π/L  −π/L

 m 1 2 2 2 (k) |ξ(k)| + |ζ(k)| dk 2 2m

this latter expression for the total Hamiltonian may also be written as an integral over the Hamiltonian functions of k varying continuously, that is,

HTot

L = 2π

π/L {H(k)} dk

(15.58)

−π/L

with m 1 (k)2 |ξ(k)|2 + |ζ(k)|2 2 2m Finally, the Hamiltonians H(k) may be transformed by the aid of Eqs. (15.47) and (15.48) into   {H(k)} = (k) a(k)† a(k) + 21 (15.59) {H(k)} =

CH015.tex

17/3/2011

11: 24

Page 459

15.2

PHONONS AND NORMAL MODES IN SOLIDS

459

Thus, these Hamiltonians have the same structure as (5.9) of a single quantum harmonic oscillator so that their eigenvalue equations must be of the same kind as that of the Hamiltonian (5.9), that is, as that of Eq. (5.42). Hence   {H(k)}|{n(k)} = n(k) + 21 (k)|{n(k)} where |{n(k)} are the eigenkets of the Hamiltonians H(k), whereas n(k) = 0, 1, 2, . . . are the number of vibrational quanta within the normal mode k, which may be viewed as the excitation degrees of the modes of angular frequency (k). Besides, since each Hamiltonian (15.59) is Hermitian as this (5.9), the eigenkets |{n(k)} form, of course, for each value of the angular frequency, an orthonormalized basis defined by  {m(k)}|{n(k)} = δm(k),n(k) and |{n(k)} {n(k)}| = 1 n(k)

Moreover, as in Eq. (5.12), working for single quantum harmonic oscillator, one may introduce, for each normal mode of the solid, an occupation number operator defined by N(k) = a(k)† a(k) the eigenvalue equation of which is N(k)|{n(k)} = n(k)|{n(k)} with

n(k) = 0, 1, 2, . . .

Since the n(k) may be viewed as the number of vibrational quanta corresponding to the normal mode k, these vibrational quanta are called phonons in solid-state physics. Moreover, in the Heisenberg picture, each lowering operator of the different normal modes obeys the Heisenberg equation   da(k, t)HP i = [a(k, t)HP , H(k)] dt Thereby, using Eqs. (15.49) and (15.59), and proceeding in a similar way as for passing from Eqs. (5.150) to (5.151), one would obtain a(k, t)HP = a(k, 0)HP e−i(k)t Finally, as in the usual quantum harmonic oscillator, one may obtain for each normal mode at any temperature T , the following thermal average (13.32): n(k) = a(k)† a(k) = (1 − e−λ(k) )tr{e−λ(k)a(k)

† a(k)

a(k)† a(k)}

leading to the result n(k) =

1 eλ(k) − 1

with λ(k) =

(k) kB T

which gives the mean number of phonons of k wave vector at temperature T , which is analogous to that (13.36). In a similar way, one would obtain for each normal mode

CH015.tex

460

17/3/2011

11: 24

Page 460

QUANTUM MODES IN MOLECULES AND SOLIDS

of the solid, the thermal average energy having an expression of the same form as that of (13.29), that is,   (k) (k) H(k) = + 2 eω/kB T − 1 and for the heat capacity, an expression of the same form as that of (13.42), that is,   (k) 2 e(k)/kB T Cv (k, T ) = NkB kB T (e(k)/kB T − 1)2

15.3 15.3.1

EINSTEIN AND DEBYE MODELS OF HEAT CAPACITY Einstein model

Before the introduction of quantum ideas, it was not possible to understand why the molar specific heat of solids should fall at low temperature, below the classical equipartition value 3R, with R the ideal gas constant. In 1907, Einstein clarified this mystery using Planck’s hypothesis concerning the quantization of energy oscillators. In his model, Einstein used the rough assumption that all the oscillators of solids as having the same characteristic angular frequency ω◦ . Then, the heat capacity of the solid is equal to the total number of freedom degrees of vibration of the solid, times the heat capacity of each oscillator. If there are N = nN atoms (where n is the number of moles and N the Avogadro number) in the solid, the total number of degrees of freedom is 3N −6  3N since N is very large. Now, the heat capacity of N oscillators is given by Eq. (13.42). Hence, in the Einstein model, the heat capacity of the solid reads  ◦ 2 ◦ eω /kB T ω (15.60) Cv (T ) = 3NkB ◦ kB T (eω /kB T − 1)2 and the molar heat capacity of the solid reads  2 Cv (T ) TE eTE /T (15.61) = 3R C¯ v (T ) = T n T (e E /T − 1)2 where TE is the Einstein temperature defined by ω◦ TE = kB If Eq. (15.61) of the Einstein model reproduces the general sigmoid form of the experimental evolution with the absolute temperature of the heat capacity, however, the experimental specific heat approaches zero slower than that predicted by this equation since it obeys an empirical law of the form C¯ v (T )Exp  T 3 The reason for this discrepancy is the crude assumption that all atoms of the solid vibrate with the same characteristic angular frequency. It is clear that there are always some modes of oscillation corresponding to a sufficiently large group of atoms moving collectively with so small an angular frequency that these modes may contribute more appreciably to the specific heat than that predicted from the Einstein assumption, thus preventing the heat capacity C v (T ) from decreasing quite as rapidly.

CH015.tex

17/3/2011

11: 24

Page 461

15.3

15.3.2

EINSTEIN AND DEBYE MODELS OF HEAT CAPACITY

461

Debye model

To improve the Einstein model, which relies on the approximation that all the oscillators of the solid may be viewed as having the same angular frequency, Debye supposed that the angular frequencies of the vibrational modes vary, the number σ(ω) of these modes lying between ω and ω + dω, being assumed to be those of the normal modes inside a closed cavity of volume V . This number σ(ω) is given by an expression that may be obtained just as that used to pass from Eqs. (14.88) to (14.94), in counting the number of normal modes of the electromagnetic field in a cavity, that is, σ(ω) = 3V

ω2 2π2 cs3

(15.62)

where cs is the effective sonic velocity. Moreover, Debye assumed a cut-off value ωD in such a way as the 3N degrees of freedom of vibration result from ωD 3N =

σ(ω) dω

(15.63)

0

This approximation may be compared to experimental results obtained for a metal from X-ray scattering measurements at 300 K (see Fig. 15.3).

σ(ω)

Arbitrary units

Debye model

0

0.2

0.4

0.6

ω (2π⫻1013Hz)

0.8

1.0

ωD

Figure 15.3 Comparison between the assumed normal mode vibrational frequency distribution σ(ω) given by Eq. (15.62) and an experimental one (solid line) dealing with aluminum at 300 K, deduced from X-ray scattering measurements. [After C. B. Walker. Phys. Rev., 103 (1956): 547–557.]

CH015.tex

17/3/2011

462

11: 24

Page 462

QUANTUM MODES IN MOLECULES AND SOLIDS

Then, using Eq. (15.62), Eq. (15.63) reads 3V 3N = 2π2 cs3

ωD ω2 dω = V 0

3 ωD 2π2 cs3

From this result, one may express the cut-off angular frequency ωD and the volume V using  ωD = cs 6π

2N

1/3 (15.64)

V

and  V = 6π2 N

cs ωD

3

Then, using the Debye cut-off approximation, the heat capacity of the solid reads ωD Cv (T ) =

σ(ω)Cv (ω, T ) dω 0

Hence, using Eq. (15.60) for a single oscillator, we have with N = 1 and Eq. (15.62) 3VkB Cv (T ) = 2π2 cs3



ωD ω2 0

ω kB T

2

eω/kB T dω − 1)2

(eω/kB T

(15.65)

Again, using the notation x=

ω kB T

ω=

and thus

xkB T 

Eq. (15.65) becomes 3VkB Cv (T ) = 2π2



kB T cs 

3 xD 0

ex x 4 dx (ex − 1)2

where xD =

ωD kB T D

(15.66)

or, using for the volume V the last expression of Eq. (15.64), 32 Cv (T ) = nR (xD )3

xD 0

ex x 4 dx (ex − 1)2

(15.67)

CH015.tex

17/3/2011

11: 24

Page 463

15.3

EINSTEIN AND DEBYE MODELS OF HEAT CAPACITY

463

In the high-temperature limit T >> TD and x = ω/kB T TD

0

This result is the Dulong and Petit law governing the heat capacity of solids at high temperature, a limiting result that can also be obtained by the Einstein model. However, the more interesting in Eq. (15.67) is its limiting case of very low temperature, corresponding to x = ω/kB T >> 1. In this low-temperature region, the upper limit xD of the integral appearing in Eq. (15.67) can be replaced by infinity even if xD is maintained in the constant appearing in front of the integral sign: 32 Cv (T ) = nR (xD )3

∞ 0

ex x 4 dx (ex − 1)2

(15.68)

The dimensionless integral is then a constant that does not depend on the temperature and which may be found to be ∞ (ex 0

ex 4 4 x 4 dx = π − 1)2 15

Hence, in the low-temperature limit, and due to Eq. (15.66), Eq. ( 15.68) yields   4π4 T 3 Cv (T ) = nR (15.69) 15 TD where TD is the Debye temperature given by TD =

ωD kB

Observe that the low-temperature limit (15.69) of the Debye heat capacity reproduces satisfactorily the experimental T 3 dependence, as shown in Fig. 15.4.

CH015.tex

17/3/2011

464

Page 464

QUANTUM MODES IN MOLECULES AND SOLIDS

CVD /R

2 CVE /R

1

0

0

50

CVD /R

0.01

100 150 200 250 300

Heat capacity

3 Heat capacity

11: 24

CVE /R

0.05

0

0

5

10

15

T (K)

T (K)

(a)

(b)

20

25

30

Figure 15.4 Temperature dependence of experimental (Handbook of Physics and Chemistry, 72 ed.) heat capacities (dots) of silver as compared to the Einstein (CvE ) and the Debye (CvD ) models as a function of the absolute temperature T . TE = 181 K, TD = 225 K.

15.4

CONCLUSION

In this chapter, we showed how to get the normal modes of a molecule and of 1D solids, by assuming that molecular or solid oscillators involve coupling linear in the elongations of the coupled oscillators. The classical coupled molecular oscillators were decoupled, leading to normal modes having the same properties as harmonic oscillators, which behave on quantization as the harmonic oscillators studied in Parts II, III, and IV. Next, when the molecules involve the symmetry elements of a symmetry group, it was shown using point-group theory, how to determine to what irreducible representation of the symmetry group belong the different molecular normal modes. Then, considering 1D solids, it was shown how, on passing from the geometric to reciprocal space that it is possible to get the solid normal modes acting as usual harmonic oscillators and thus allowing us to apply to them all the results met for single harmonic oscillators and thus, particularly, to find some solid thermal properties such as, for instance, their heat capacities, either in the context of the Einstein model or that of Debye.

BIBLIOGRAPHY C. Cohen-Tannoudji, B. Diu, and F. Laloe. Quantum Mechanics. Wiley-Interscience: Hoboken, NJ, 2006. F. Reif. Fundamentals of Statistical and Thermal Physics. McGraw-Hill: New York, 1965. E. Wilson, J. Decius, and P. Cross. Molecular Vibrations. McGraw-Hill: New York, 1955.

CH016.tex

PART

17/3/2011

9: 54

Page 465

VI

DAMPED HARMONIC OSCILLATORS

CH016.tex

17/3/2011

CHAPTER

9: 54

Page 467

16

DAMPED OSCILLATORS INTRODUCTION In Chapter 5 the energy levels of an isolated quantum harmonic oscillator was found, and in Chapter 13 the properties of a population of quantum harmonic oscillators at thermal equilibrium were derived, a thermal situation that must occur whatever the initial condition. Furthermore, in Chapter 11 we studied a linear chain of quantum harmonic oscillators linearly coupled in the rotating-wave approximation, the initial situation being where one of the oscillators is in a coherent state, whereas the other ones are in the ground state of their Hamiltonian. We found that the system, if it evolves in a deterministic way, leads, however, after a sufficient time to a continuous distribution of the energy between the different oscillators. Then, it was possible using a coarse-grained analysis of the energy distribution between the different oscillators to show in Chapter 13 that the mean statistical entropy of the system increases until it has attained a stable maximum value and that, when this maximum has been attained, the mean distribution of the energy levels of the different oscillators then obeys the thermal equilibrium Boltzmann law. However, the coarse-grained analysis of the irreversible mean evolution of a quantum oscillator toward the thermal equilibrium distribution has not yet been studied. The purpose of the present chapter is to treat this question. This chapter begins with an exposition of the quantum model generally used to treat the irreversible behavior of an oscillator embedded in a thermal bath. Then, second-order time-dependent perturbation theory (i.e., second order with respect to the coupling between the damped oscillator and the bath) is used to calculate the master equation governing the time derivative of the reduced density operator of a driven damped quantum harmonic oscillator. To go beyond the previous perturbative approach, a short subsequent section is devoted, without demonstration, to the results of the Louisell and Walker models, which gives in closed form an expression for the time evolution of the reduced density operator of the driven damped quantum harmonic oscillator, which may be viewed as a result of the integration of a master equation, which would have been obtained up to infinite order of perturbation instead of second order as in the above master equation. In the next section, we transform the master equation to its corresponding antinormal expression (see Chapter 7), which has the form of a Fokker–Planck equation and which may then transform, using the inverse of the antinormal order operator, into a second-order partial differential equation having a structure analogous to the Fokker–Planck equation of Brownian oscillators met in the Quantum Oscillators, First Edition. Olivier Henri-Rousseau and Paul Blaise. © 2011 John Wiley & Sons, Inc. Published 2011 by John Wiley & Sons, Inc.

467

CH016.tex

468

17/3/2011

9: 54

Page 468

DAMPED OSCILLATORS

area of statistical irreversible classical mechanics. In a subsequent section, the quantum Langevin equation governing the irreversible evolution of the average values of Boson operators is derived. Finally, using this Langevin equation, we to get what may be considered as the IP time evolution operator governing the dynamics of a driven damped harmonic oscillator.

16.1 QUANTUM MODEL FOR DAMPED HARMONIC OSCILLATORS 16.1.1

Hamiltonian

The full Hamiltonian of the driven damped quantum harmonic oscillator may be written HTot = H◦ + HDr + V + Hθ

(16.1)

Here, H◦ is the harmonic part of the Hamiltonian of the oscillator of interest, whereas HDr is the driven part of the Hamiltonian of this oscillator. Hθ is the Hamiltonian of the thermal bath and V the coupling Hamiltonian between the thermal bath and the oscillator, which will be damped by this bath. The harmonic part of the Hamiltonian of the oscillator of interest is, of course,   2 P 1 ◦ ◦2 2 (16.2) + Mω Q H = 2M 2 where M is the reduced mass of the oscillator, ω◦ is the corresponding angular frequency, Q is the coordinate operator, and P its conjugate momentum obeying [Q, P] = i The part of the Hamiltonian driving the oscillator is HDr = k ◦ Q

(16.3)

where k ◦ is a constant. Now, the thermal bath may be simulated by an infinite set of quantum harmonic oscillators of reduced masses ml and of angular frequencies ωl , which are varying in a quasi-continuous way. Thus, the Hamiltonian of the bath may be written    p2 1 2 2 l (16.4) + ml ω l q l Hθ = 2ml 2 l

where ql is the position operator of the lth oscillator, whereas pl is the conjugate momentum obeying [qk , pl ] = iδkl The Hamiltonian coupling the driven oscillator to the thermal bath may be assumed to be given by  V= kl Qql (16.5) l

CH016.tex

17/3/2011

9: 54

16.1

Page 469

QUANTUM MODEL FOR DAMPED HARMONIC OSCILLATORS

469

where the kl are the coupling constants between the driven oscillator and the lth oscillator of the bath. Of course, the operators characterizing the driven oscillator and the bath oscillators commute, that is, [Q, pl ] = [P, ql ] = 0 In the following, one will pass from the discrete expression of the thermal bath (16.4) to the following continuous one:   p(ω)2 1 Hθ = g(ω) + m(ω)ω2 q(ω)2 dω 2m(ω) 2  V=

g(ω)k(ω)Qq(ω) dω

In these continuous expressions, g(ω) is the density of modes of angular frequency ω and reduced mass m(ω), whereas k(ω) are the corresponding frequency-dependent coupling constants. Finally, q(ω) and p(ω) are the frequency-dependent position and conjugate momentum obeying [q(ω), p(ω )] = iδ(ω − ω ) Now, we pass from the position and momentum operators of the different oscillators to the corresponding Boson operators by means of the usual transformations (5.6) and (5.7):



 Mω◦ † † Q= (a (a − a) + a) P = i (16.6) 2Mω◦ 2 

 ml ωl † † ql = (b + bl ) pl = i (16.7) (bl − bl ) 2ml ωl l 2 in which a, a† , b†l , and bl are the dimensionless Boson operators obeying the commutation rules of the same kind as that of (5.5) [bk , b†l ] = δkl

[a, a† ] = 1

[al† , b] = [a, b†l ] = [a† , b†l ] = [a, bl ] = 0

(16.8)

In the Boson operator picture and after neglecting the zero-point energy, which is irrelevant the harmonic part of the driven oscillator defined by Eq. (16.2) takes the form (5.9), that is, H = ω◦ a† a

(16.9)

whereas the driven part (16.3) of the Hamiltonian (16.1 ) becomes HDr = α◦ ω◦ (a† + a) with

◦

α =k

◦

 2Mω◦

(16.10)

CH016.tex

470

17/3/2011

9: 54

Page 470

DAMPED OSCILLATORS

Moreover, due to (16.7) and (16.8), the thermal bath Hamiltonian (16.4) yields, after neglecting the zero-point energies of the different oscillators,  ωk b†k bk (16.11) Hθ = k

Thus, owing to Eqs. (16.6) and (16.7), the Hamiltonian (16.5) coupling the driven oscillator to the thermal bath yields 

   V= (a† + a) kk (b† +bk ) ◦ 2Mω 2mk ωk k k

or V=



Kk (a† b†k + abk + a† bk +ab†k )

(16.12)

k

where Kk are dimensionless coupling constants given by  1 kk Kk = 2 Mmk ωk ω◦ In the rotating-wave approximation, it is usual to neglect the double creations and annihilations induced by the terms a† b†k and abk and to single out those interactions in which exist simultaneously an excitation of one of the interacting oscillators and a deexcitation of the last one. Hence, we write in place of Eq. (16.12)  Kk {(a† bk ) + (ab†k )} V= k

which may be generalized to V=

 {Kk (a† bk ) + Kk∗ (ab†k )}

(16.13)

k

Now, consider the various density operators of the system. Since the thermal bath involves a very large number of oscillators, its density operator ρθ (t) may be assumed to be unperturbed by the single driven oscillator to which it is coupled, so that it may be assumed to be constant, leading one to write ρθ (t) = ρθ (t0 ) where t0 is an initial time. This thermal density operator will be viewed as the product of the density operators ρj of the different oscillators forming the bath, each being in thermal equilibrium and thus described by a Boltzmann density operator ρj (16.14) ρθ = j

Moreover, the density operators of the thermal bath oscillators may be assumed to be given at all times by canonical density operators of the form (13.23) †

ρj = (1 − e−λj )(e−λj bj bj )

(16.15)

CH016.tex

17/3/2011

9: 54

16.1

Page 471

QUANTUM MODEL FOR DAMPED HARMONIC OSCILLATORS

471

with, according to Eq. (13.24), ωj (16.16) εj = (1 − e−λj ) kB T where T is the absolute temperature. Now, in the Schrödinger picture, the full density operator of the system at an initial time t = t0 may be considered as the density operator ρ(t0 ) of the driven oscillator at this time, multiplied by that of the bath ρθ (t0 ) at this same time, that is, λj =

ρTot (t0 )SP = ρ(t0 )SP ρθ

(16.17)

where kB is Boltzmann’s constant. Next, in the Schrödinger picture, and according to Eq. (3.170), the dynamics of the full density operator is governed by the Liouville equation   ∂ρTot (t)SP i = [HTot , ρTot (t)SP ] (16.18) ∂t subject to the boundary condition (16.17). Notice that once the expression of ρTot (t)SP has been obtained, it is possible to get the time dependence of the density operator of the driven damped harmonic oscillator by performing the partial trace of the timedependent full density operator over the thermal bath: ρ(t)SP = trθ {ρTot (t)SP }

(16.19)

Due to Eq. (16.1), the Schrödinger–Liouville equation (16.18) reads   ∂ρTot (t)SP = [H◦ , ρTot (t)SP ] + [ HDr , ρTot (t)SP ] i ∂t + [V, ρTot (t)SP ] + [Hθ , ρTot (t)SP ]

16.1.2

Interaction picture

We make the following partition of the Hamiltonian (16.1): ◦ + V+ HDr HTot = HTot

with

◦ HTot = H + Hθ

(16.20)

Within this partition, the operators V and HDr become, respectively, in the interaction ◦ picture with respect to HTot ◦ ◦

V(t)IP = UTot (t)−1 VUTot (t)

(16.21)

Dr (t) = U◦ (t)−1 HDr U◦ (t) H Tot Tot

(16.22)

with ◦ (t) = exp UTot



◦ t −iHTot 

 (16.23)

Next, in view of Eqs. (16.9), (16.11), and (16.20), the time evolution operator Eq. (16.23) transforms to ⎧ ⎛ ⎞⎫ ⎨ ⎬  † ◦ (t) = exp −i ⎝a† a ω◦ t + bj bj ω j t ⎠ (16.24) UTot ⎩ ⎭ j

CH016.tex

472

17/3/2011

9: 54

Page 472

DAMPED OSCILLATORS

Then, since according to Eq. (16.8) the Boson operators of the thermal bath commute with those of the driven harmonic oscillator, this time evolution operator (16.23) factorizes into † ◦ −ib† b ω t ◦ UTot (t) = e−ia a ω t (16.25) e j j j j

or ◦ UTot (t) = U◦ (t)Uθ◦ (t)

(16.26)

with, respectively, U◦ (t) = e−ia

Uθ◦ (t) =

† aω◦ t

(16.27)

Uj◦ (t)

(16.28)

j

and †

Uj◦ (t) = e−ibj bj

ωj t

(16.29)

Moreover, owing to Eqs. (16.13), (16.26), and (16.28), the IP coupling Hamiltonian (16.21) reads ⎧ ⎫   ⎨ ⎬ 

V(t)IP = U◦ (t)−1 Uj◦ (t)−1 (a† bk Kk + ab†k Kk∗ ) Uj◦ (t) U◦ (t) ⎩ ⎭ j

k

or

V(t)

IP

◦

= U (t)

−1 †

◦

a U (t)





 Uj◦ (t)−1



j

 Kk b k

 Uj◦ (t)

+ hc

k

or, writing explicitly the evolution operators dealing with the thermal bath using Eqs. (16.27) and (16.29),      † † † aω◦ t † aω◦ t ib b ω t −ib b ω t IP ia † −ia j j j j

V(t) = (e e j + hc )a (e ) Kk bk e j j

k

so that, following the action of each operator within their respective subspaces, we have ⎧ ⎫ ⎨ ⎬ † † † ◦ † ◦

V(t)IP = (eia aω t )a† (e−ia aω t ) Kj (eibj bj ωj t )bj (e−ibj bj ωj t ) ⎩ ⎭ j

×



(e

ib†k bk

ωk t

)(e

−ib†k bk

ωk t

) + hc

k=j

which reduces to ia† aω◦ t

V(t)IP = e

a† e

−ia† aω◦ t

⎧ ⎨ ⎩

j

Kj e

ib†j bj

ωj t

bj e

−ib†j bj

ωj t

⎫ ⎬ ⎭

+ hc

(16.30)

CH016.tex

17/3/2011

9: 54

16.1

Page 473

QUANTUM MODEL FOR DAMPED HARMONIC OSCILLATORS

since



†

eibk bk

ωk t −ib†k bk ωk t

473

=1

e

j =k

Now, applying theorems (7.21) and (7.22) to the Boson operators of the oscillator of interest and to those of the thermal bath, that is, (eia †

† aω◦ t

(eibj bj

ωj t

)a† (e−ia

† aω◦ t

†

)bj (e−ibj bj

◦

) = a† (eiω t )

ωj t

(16.31)

) = bj (e−iωj t )

it appears that the IP coupling between the driven oscillator and the bath takes the form  ◦ ◦

{Kj a† eiω t bj e−iωj t + Kj∗ ae−iω t b†j eiωj t } (16.32) V(t)IP = j

Now due to Eq. (16.26), the IP expression (16.22) of the driving Hamiltonian (16.10), which depends only on a† and a, reads

Dr (t)IP = α◦ ω◦ U◦ (t)−1 (a + a† )U◦ (t) Uj◦ (t)−1 Uj◦ (t) H Tot Tot j

or, since Uj◦ (t)−1 Uj◦ (t) = 1 we have on simplification

Dr (t)IP = α◦ ω◦ U◦ (t)−1 (a + a† )U◦ (t) H a result that also reads

Dr (t) = U◦ (t)−1 HDr U◦ (t) H

(16.33)

the inverse canonical transformation being

Dr = U◦ (t) HDr (t)U◦ (t)−1 H

(16.34)

Furthermore, according to Eq. (16.10), because the different operators act within their own vector subspace, the driven part of the Hamiltonian (16.22) reads, after simplification,

Dr (t)IP = α◦ ω◦ U◦ (t)−1 (a + a† )U◦ (t) H

(16.35)

so that, due to Eq. (16.27),

Dr (t)IP = α◦ ω◦ (eia† aω◦ t ae−ia† aω◦ t + eia† aω◦ t a† e−ia† aω◦ t ) H which, using (16.31), yields

Dr (t)IP = α◦ ω◦ (ae−iω◦ t + a† eiω◦ t ) H In this same picture, the IP density

operator ρ(t)IP

(16.36)

of the oscillator of interest reads

◦ ◦

(t)−1 ρ(t)SP UTot (t) ρ(t)IP = UTot

(16.37)

CH016.tex

17/3/2011

474

9: 54

Page 474

DAMPED OSCILLATORS

where ρ(t)SP is the Schrödinger picture density operator (16.19). Now, again, due to Eq. (16.26), Eq. (16.37) becomes

ρ(t)IP = U◦ (t)−1 Uθ◦ (t)−1 ρ(t)SP U◦ (t)Uθ◦ (t) so that, since the IP density operator ρ(t)IP is that of the oscillator of interest and not that of the thermal bath, it transforms to

ρ(t)IP = U◦ (t)−1 ρ(t)SP U◦ (t)Uθ◦ (t)−1 Uθ◦ (t) = U◦ (t)−1 ρ(t)SP U◦ (t)

16.1.3

IP Liouville equation

In the chosen interaction picture, the Liouville equation governing the time dependence of the full density operator is, owing to Eq. (3.193),   ∂ρ˜ Tot (t)IP

Dr (t)IP + i V(t)IP ), ρ˜ Tot (t)IP ] = [(H ∂t Now, performing the trace over the thermal bath on this last expression,   ∂ρ˜ Tot (t)IP

Dr (t)IP + V(t)IP ), ρ˜ Tot (t)IP ]} = trθ {[(H i trθ ∂t yields i



∂ρDr (t)IP ∂t



Dr (t)IP , ρ˜ Tot (t)IP ]} + trθ {[ = trθ {[H V(t)IP , ρ˜ Tot (t)IP ]}

(16.38)

with ρDr (t)IP = trθ {ρ˜ Tot (t)IP }

(16.39)

Dr (t)IP does not involve the thermal bath, the first Next, since the IP Hamiltonian H right-hand-side term of Eq. (16.38) reads

Dr (t)IP , ρTot (t)IP ]} = [H

Dr (t)IP, trθ {ρ˜ Tot (t)IP }] trθ {[H And thus, due to Eq. (16.39),

Dr (t)IP , ρTot (t)IP ]} = [H

Dr (t)IP , ρ(t)IP ] trθ {[H Hence, Eq. (16.38) reads   ∂ρDr (t)IP

Dr (t)IP , ρ(t)IP ] = [H i ∂t + trθ {[ V(t)IP , ρTot (t)IP ]} A result that may also be written as     ∂ρDr (t)IP ∂ ρ(t)IP IP IP

i = [HDr (t) , ρDr (t) ] + i ∂t ∂t

(16.40)

with 

∂ ρ(t)IP i ∂t



= trθ {[ V(t)IP , ρTot (t)IP ]}

(16.41)

CH016.tex

17/3/2011

9: 54

Page 475

16.2

16.2

SECOND-ORDER SOLUTION OF EQ. (16.41)

475

SECOND-ORDER SOLUTION OF EQ. (16.41)

Equation (16.41) is the IP Liouville equation governing the dynamics of the undriven oscillator of interest interacting with the thermal bath, the integration of which, up to second order in V(t)IP , is the aim of the present section. The formal integration of Eq. (16.41) leads to the integral equation 

ρ(t)

IP

= ρ(t0 ) + IP

1 i

 t

dt  trθ {[ V(t  − t0 )IP , ρTot (t  )IP ]}

(16.42)

t0

Up to second order in V(t)IP , the integral equation (16.42) reads

ρ(t0 )IP = I1 + I2 ρ(t)IP −

(16.43)

with, respectively,  I1 =

1 i

 t

dt  trθ {[ V(t  − t0 )IP , ρ(t0 )IP ]}

(16.44)

dt  trθ {[ V(t  − t0 )IP , [ V(t  − t0 )IP , ρ(t0 )IP ]]}

(16.45)

t0

 I2 =

1 i

2  t dt t0



t  t0

and where, according to Eq. (3.195), V(τ − t0 )IP = eiH

16.2.1

◦ (τ−t

0 )/

Ve−iH

◦ (τ−t

0 )/

τ = t  or t 

with

Making explicit Eq. (16.43)

To go further, we prove that the integral (16.44) involved in Eq. (16.43) is zero and for this purpose start from the commutator involved in Eq. (16.44): V(t  − t0 )IP , ρTot (t0 )IP ]} trθ {[ = trθ { V(t  − t0 )IP ρTot (t0 )IP } − trθ { V(t  − t0 )IP } ρTot (t0 )IP

(16.46)

Owing to Eq. (16.32) the IP coupling Hamiltonian appearing in Eq. (16.46) is ◦  ◦   ◦

bl Kl e−iωl (t −t ) + hc V(t  − t0 )IP = a† eiω (t −t ) l

or, on simplification by taking t =

t

− t0 , ◦ 

V(t)IP = a† eiω

t

bl Kl e−iωl t + hc

(16.47)

l

Due to Eq. (16.17), the two right-hand sides of Eq. (16.46) yield, respectively, V(t)IP ρTot (t0 )IP } trθ {       † † iω◦ t −iωl t IP −iω◦ t ∗ +iωl t IP

= trθ a e bl Kl e bl Kl e ρ(t0 ) ρθ + trθ ae ρ(t0 ) ρθ l

l

CH016.tex

476

17/3/2011

9: 54

Page 476

DAMPED OSCILLATORS

trθ { ρTot (t0 )IP V(t)IP }  † iω◦ t

= trθ ρθ a e ρ(t0 ) IP



 bl Kl e

−iωl t

 ρθ ae + trθ ρ(t0 ) IP

−iω◦ t



l

 b†l Kl∗ e+iωl t

l

Moreover, since the trace operation over the thermal bath affects neither the lowering and raising operators a and a† of the oscillator of interest nor its IP density operator

ρ(t)IP , these two last equations become, respectively, trθ { V(t)IP ρTot (t0 )IP }       † ◦ † IP iω◦ t −iωl t ∗ +iωl t = trθ a bl + trθ ρθ Kl e ρ(t0 ) e bl ρθ Kl e a ρ(t0 )IP e−iω t l

l

(16.48) trθ { ρTot (t0 )IP V(t)IP } = ρ(t0 )IP a† e

iω◦ t



ρθ trθ







ρ(t0 )IP ae bl Kl e−iωl t +

−iω◦ t

ρθ trθ



l

 b†l Kl∗ e+iωl t

l

(16.49) Next, owing to Eq. (16.15), it appears that   † ρθ bl = εj (e−λj bj bj ) bl j

l

l

so that since each Boson operator of the thermal bath works within its specific state space, this last expression transforms to   † † ρθ bl = εl (e−λl bl bl )bl εj (e−λj bj bj ) l

j=l

l

Moreover, tracing over the thermal bath for this last term may be realized in the basis {|(nl )} defined by the eigenvalue equations dealing with the thermal bath, that is, b†l bl |(nl ) = nl |(nl ) with (nl )|(ml ) = δnl ml

and



(16.50)

|(nl )(nl )| = 1

(16.51)

leading us to write this partial trace according to      † † trθ ρθ bl = εl (nl )|(e−λl bl bl )bl |(nl ) εj (nj )|(e−λj bj bj )|(nj ) l

l

j=l

nl

nj

Again, observe that since the Boltzmann density operators are normalized through the normalization constants εj , one has for each oscillator j  † εj (nj )|(e−λj bj bj )|(nj ) = 1 (16.52) nj

CH016.tex

17/3/2011

9: 54

Page 477

16.2

SECOND-ORDER SOLUTION OF EQ. (16.41)

so that the above equation reduces to      † trθ ρθ bl = εl (nl )|(e−λl bl bl )bl |(nl ) l

477

(16.53)

nl

l

Again, owing to the action of b†l bl and bl on the eigenkets of b†l bl , which obey equations similar to that (5.53), that is, bl |(nl ) =

√ nl |(nl − 1)

(16.54)

the following result is verified: †

(nl )|(e−λl bl bl )bl |(nl ) =

† √ nl (nl )|(e−λl bl bl )|(nl − 1)

so that, using the eigenvalue equation (16.50), it yields †

(nl )|(e−λl bl bl )bl |(nl ) =

√ nl (e−λl (nl −1) )(nl )|(nl − 1)

or, owing to the orthogonality of the kets appearing in (16.51) †

(nl )|(e−λl bl bl )bl |(nl ) =

√ nl (e−λl (nl −1) )δnl ,nl −1 = 0

(16.55)

then Eq. (16.53) reduces to 



 =0

(16.56)

Of course, one would obtain in like manner    † trθ bl ρθ = 0

(16.57)

trθ ρθ

bl

l

l

Thus, as a consequence of Eqs. (16.56) and (16.57), Eqs. (16.48) and (16.49) read trθ { ρTot (t0 )IP } = trθ { V(t)IP } = 0 V(t)IP ρTot (t0 )IP so that Eq. (16.44) yields I1 = 0 The last result implies that Eq. (16.43) reduces to

ρ(t)IP − ρ(t0 )IP = I2  =

1 i

2  t dt t0



t  t0

dt  trθ {[ V(t  − t0 )IP , [ V(t  − t0 )IP , ρTot (t0 )IP ]]}

CH016.tex

478

17/3/2011

9: 54

Page 478

DAMPED OSCILLATORS

which, writing explicitly the double commutator appearing in this last equation, takes the form

ρ(t)IP − ρ(t0 )IP 

1 =+ i

2  t

dt 

t0



1 − i

2

1 − i

dt



1 + i

t 

dt  trθ { V(t  − t0 )IP ρTot (t0 )IP V(t  − t0 )IP }

t0

2  t

dt 

t0



dt  trθ { V(t  − t0 )IP V(t  − t0 )IP ρTot (t0 )IP }

t0

t t0



t 

t 

dt  trθ { V(t  − t0 )IP ρTot (t0 )IP V(t  − t0 )IP }

t0

2  t

dt 

t0

t 

dt  trθ { ρTot (t0 )IP V(t  − t0 )IP V(t  − t0 )IP }

(16.58)

t0

Note that in order to calculate Eq. (16.58) it is not possible to use the invariance of the trace with respect to a circular permutation in order to get some traces in terms of others because trθ is a partial trace over the thermal bath, because this invariance holds only if the trace operation is performed over a basis belonging to the complete space involving both the bath and the oscillator embedded in it.

16.2.2 Calculation of the first average values involved in Eq. (16.58) Now, one has to find the result of the traces involved in Eq. (16.58). For this purpose, begin with the first one of them, which, in view of Eq. (16.47), reads  2 1 trθ { V(t  − t0 )IP ρTot (t0 )IP } V(t  − t0 )IP i       † † iω◦ t  −iωl t  −iω◦ t  ∗ iωk t  IP

a e ae ρ(t0 ) ρθ = +trθ bl K l e bk K k e l

 + trθ

ae

−iω◦ t 

 b†l Kl∗ eiωl t

l

 + trθ

a† eiω

◦  

t

ae

◦  

−iω t

l

† iω◦ t 

a e



bl Kl e−iωl

t

a† e

b k Kk e

+iω◦ t 



ae

◦  

−iω t

k



ρ(t0 ) ρθ IP

 bk Kk e−iωk

k

  b†l Kl∗ eiωl t

 −iωk t 

k



l

 + trθ



k



t 



ρ(t0 )IP ρθ

  b†k Kk∗ eiωk t



ρ(t0 ) ρθ IP

CH016.tex

17/3/2011

9: 54

Page 479

16.2

SECOND-ORDER SOLUTION OF EQ. (16.41)

479

or 

1 i

2

trθ { V(t  − t0 )IP V(t  − t0 )IP ρTot (t0 )IP } 

= +a† ae

iω◦ (t  −t  )

ρ(t0 )IP trθ

 l

k

 † −iω◦ (t  −t  )



ρ(t0 ) trθ

+ aa e

IP

l

 † † iω◦ (t  +t  )

IP

 −iω◦ (t  +t  )

ρ(t0 ) trθ

+ aae

IP

l

 K l K k bl bk e

−iωl t  −iωk t 

e

ρθ

k

 l

   Kl∗ Kk b†l bk e+iωl t e−iωk t ρθ

k



ρ(t0 ) trθ

+a a e

   Kl Kk∗ bl b†k e−iωl t e+iωk t ρθ

   Kl∗ Kk∗ b†l b†k e+iωl t e+iωk t ρθ

(16.59)

k

Next, notice that the trace over the thermal bath is the product of the traces dealing with the different oscillators of the bath, that is, trθ {· · · } = trl {· · · } l

Thus, after separation of the double sums into two parts, one where k  = l and the other where k = l, Eq. (16.59) reads 

1 i

2

trθ { V(t  − t0 ) V(t  − t0 ) ρTot (t0 )} = A(t  − t ◦ ) + B(t  − t ◦ )

(16.60)

with, respectively, A(t  − t ◦ ) ρ(t0 )IP e = +a† a

iω◦ (t  −t  )

ρ(t0 )IP e + aa†

⎩

−iω◦ (t  −t  )

ρ(t0 )IP e + a† a†

+ aa ρ(t0 )IP e

⎧ ⎨ 

iω◦ (t  +t  )

−iω◦ (t  +t  )

t

Kl Kk∗ trl {ρl bl }trk {ρk b†k }e−iωl e+iωk

⎧ ⎨ ⎩

t

Kl∗ Kk trl {ρl b†l }trk {ρk bk}e+iωl e−iωk t

Kl Kk trl {ρl bl }trk {ρk bk }e−iωl e−iωk

l k=l

⎧ ⎨ l k=l

εj trj {ρj }

j=l,k

⎫ ⎬

t 

⎭

l k=l

⎧ ⎨

⎩

⎭

k =l

l

⎩

⎫ ⎬

t 

εj trj {ρj }

j=l,k

⎫ ⎬

t 

⎭

εj trj {ρj }

j=l,k

t

Kl∗ Kk∗ trl {ρl b†l }trk {ρk b†k }eiωl e+iωk

⎫ ⎬

t 

⎭

εj trj {ρj }

j=l,k

(16.61)

CH016.tex

17/3/2011

480

9: 54

Page 480

DAMPED OSCILLATORS

and  

◦

IP iω◦ (t  −t  )

ρ(t0 ) e B(t − t ) = +a a †



 |Kl |

2

trl {ρl bl b†l

}e

−iωl (t  − t  )

l

 ρ(t0 )IP e + aa†



−iω◦ (t  −t  )

+a a ρ(t0 ) e



+ aa ρ(t0 ) e



 |Kl | trl {ρl bl bl }e 2

−iωl (t  + t  )

l

 IP −iω◦ (t  +t  )

trj {ρj }

j=l

trj {ρj }

j=l



IP iω◦ (t  +t  )

|Kl |2 trl {ρl b†l bl }eiωl

l

 † †

 (t  − t  )





 |Kl |

2

  trl {ρl b†l b†l }e+iωl (t + t )

trj {ρj }

j=l



trj {ρj }

j=l

l

(16.62) where †

ρl = εl (e−λl bl bl )

(16.63)

Moreover, observe that the thermal averages involved in Eqs. (16.61) and (16.62) are given by the following equations: †

trl {ρl } = εl trl {(e−λl bl bl )} = 1

(16.64)

Now, owing to Eqs. (16.55), (16.56), and (16.63), the trace over ρl bl is zero whatever l may be, that is, trl {ρl bl } = trl {ρl b†l } = 0

(16.65)

Furthermore, after writing explicitly the trace over the eigenkets of b†l bl , the thermal averages of bl bl read trl {ρl bl bl } = εl



†

(nl )|(e−λl bl bl )bl bl |(nl )

nl

or, using twice Eq. (16.54), trl {ρl bl bl } = εl



†

nl (nl − 1)(nl )|(e−λl bl bl )|(nl − 2)

nl

and thus, due to Eq. (16.50), trl {ρl bl bl } = εl

 nl

nl (nl − 1)(e−λl (nl −2) )(nl )|(nl − 2)

CH016.tex

17/3/2011

9: 54

Page 481

16.2

SECOND-ORDER SOLUTION OF EQ. (16.41)

481

hence, due to Eqs. (1.71) and (1.73), owing to the orthogonality properties (16.51) of the eigenkets of b†l bl , trl {ρl bl bl } = 0

(16.66)

In like manner, using similar reasoning, we have trl {ρl b†l b†l } = 0

(16.67)

Now, in view of Eq. (16.63), the thermal average of the number occupation b†l bl reads †

trl {ρl b†l bl } = εl trl {(e−λl bl bl )b†l bl } or, due to Eq. (13.32), trl {ρl b†l bl } = nl 

(16.68)

where nl  =

1 eωl /kT

(16.69)

−1

Furthermore, the last thermal average of interest is †

trl {ρl bl b†l } = εl trl {(e−λl bl bl )bl b†l } or, using the commutation rule of Boson operators [b, b† ] = 1, †

trl {ρl bl b†l } = εl trl {(e−λl bl bl )(b†l bl + 1)} so that, due to Eq. (16.68), trl {ρl bl b†l } = nl + 1 ≡ nl  + 1

(16.70)

Hence, as a consequence of Eqs. (16.65), (16.66), (16.67), and (16.70), Eq. (16.61) appears to be zero, that is, A(t  − t ◦ ) = 0 Therefore, owing to this result and according to Eqs. (16.64), (16.66), (16.67), (16.68), and (16.70), Eq. (16.60) takes on the simplified form  2 1 V(t  − t0 )IP trθ { V(t  − t0 )IP ρTot (t0 )IP } i    † IP 2 i(ω◦ −ωl )(t  −t  ) = a a ρ(t0 ) |Kl | nl + 1 e l

 + aa ρ(t0 ) †

IP

 l

 |Kl | nl e 2

−i(ω◦ −ωl )(t  −t  )

(16.71)

CH016.tex

482

17/3/2011

9: 54

Page 482

DAMPED OSCILLATORS

16.2.3 Calculation of the other average values involved in Eq. (16.58) Now, we have to get the other average values involved in Eq. (16.58). For instance, in view of Eq. (16.47), the third one reads  2 1 trθ { V(t  − t0 )IP ρTot (t0 )IP V(t  − t0 )IP } i       † † iω◦ t  −iωl t  IP −iω◦ t  ∗ iωk t  a e = +trθ bl K l e

ρ(t0 ) ρθ ae bk K k e l

 + trθ

ae

−iω◦ t 

† iω◦ t 

+ trθ

a e



bl K l e

l

 ae

−iω◦ t 



† +iω◦ t 

ρ(t0 ) ρθ a e IP

 −iωl t 

k



 b†l Kl∗ e+iωl t

l



+ trθ





† +iω◦ t 

ρ(t0 ) ρθ a e 

ρ(t0 ) ρθ ae IP

l

b k Kk e



 bk K k e

−iωk t 

k



 b†l Kl∗ e+iωl t

 −iωk t 

k

 IP



−iω◦ t 



  b†k Kk∗ eiωk t

k

After rearrangement, it transforms to  2 1 V(t  − t0 )IP } trθ { V(t  − t0 )IP ρTot (t0 )IP i    ◦     = +trθ a† Kl Kk∗ ρθ bl b†k e+iωk t e−iωl t ρ(t0 )IP ae−iω (t −t ) l

 IP † iω◦ (t  −t  )

+ trθ a ρ(t0 ) a e

k

l



IP † +iω◦ (t  +t  )

+ trθ a ρ(t0 ) a e †

+ trθ a ρ(t0 ) ae IP

−iω◦ (t  +t  )

  Kl∗ Kk ρθ b†l bk e−iωk t e+iωl t

k

 l







Kl Kk ρθ bl bk e

e

k

 l

 −iωk t  −iωl t 



  Kl∗ Kk∗ ρθ b†l b†k e+iωk t e+iωl t

k

Then, in like manner as passing from Eqs. (16.59) to (16.71), we have  2 1 V(t  − t0 )IP } trθ { V(t  − t0 )IP ρTot (t0 )IP i    ◦   = a† ρ(t0 )IP a |Kl |2 nl + 1e+i(ω −iωl )(t −t ) l

 + a ρ(t0 ) a

IP †

 l

 |Kl | nl e 2

−i(ω◦ −iωl )(t  −t  )

(16.72)

CH016.tex

17/3/2011

9: 54

Page 483

16.2

483

SECOND-ORDER SOLUTION OF EQ. (16.41)

In the same way, one would find for the other two average values of Eq. (16.58)  2 1 trθ { V(t  − t0 )IP V(t  − t0 )IP } ρTot (t0 )IP i    † IP 2 i(ω◦ −iωl )(t  −t  ) ρ(t0 ) a |Kl | nl + 1e =a l





+ a ρ(t0 ) a

IP †

 |Kl | nl e 2

−i(ω◦ −iωl )(t  −t  )

(16.73)

l

and 

1 i

2

trθ { ρTot (t0 )IP V(t  − t0 )IP V(t  − t0 )IP } 



= ρ(t0 ) a a IP †



|Kl | nl + 1 e 2

i(ω◦ −iωl )(t  −t  )

l





+ ρ(t0 )IP aa†

 |Kl |2 nl e

−i(ω◦ −iω

l

)(t  −t  )

(16.74)

l

16.2.4 Time derivative of the IP density operator 16.2.4.1 Basic equation for the variation of the IP density operator owing to Eqs. (16.71)–(16.74), Eq. (16.58) becomes

ρ(t0 + t)IP − ρ(t0 )IP = −a† a ρ(t0 )IP



t0+t

l

− aa ρ(t0 ) †

IP

t0



+a ρ(t0 ) a IP



t0

+ a ρ(t0 ) a

 l

+ a† ρ(t0 )IP a

 l



dt  e−i(ω

|Kl | nl + 1 2

dt



t0

|Kl | nl 

 −t  )

◦ −ω

l )(t

 −t  )

t 

dt  e+i(ω

◦ −ω

l )(t

 −t  )

t0 t

t0+t 2

l )(t

t0 t0+t

l

IP †

◦ −ω

t0

dt 

|Kl | nl 

dt  e+i(ω

t

t0+t 2

l

†

dt 

|Kl |2 nl + 1

t 

dt t0





dt  e−i(ω

◦ −ω

l )(t

 −t  )

t0 t0+t

dt 

|Kl |2 nl + 1 t0

t  t0

dt  e+i(ω

◦ −ω

l )(t

 −t  )

Next,

CH016.tex

484

17/3/2011

9: 54

Page 484

DAMPED OSCILLATORS

+ a ρ(t0 ) a

IP †



t0+t

|Kl | nl  2

l

− ρ(t0 )IP a† a

dt t0



− ρ(t0 ) aa

†

t 

t0+t

dt 

|Kl |2 nl + 1 t0



◦ −ω

l )(t

 −t  )

dt 

|Kl | nl 

t 

dt  e+i(ω

◦ −ω

l )(t

 −t  )

t0 t

t0+t 2

l

dt  e−i(ω

t0

l

IP



t0



dt  e−i(ω

◦ −ω

l )(t

 −t  )

(16.75)

t0

where t = t − t0 In Eq. (16.75), the mean occupation number nl  and the coupling terms Kl implicitly concern the angular frequencies ωl so that it is convenient to rewrite this equation as follows: ρ(t0 )IP

ρ(t0 + t)IP −   t0+t t   † IP   2 +i(ω◦ −ωl )(t  −t  ) = − a a ρ(t0 ) dt dt |Kl | nl + 1e t0 t0+t

− a a ρ(t0 ) †

IP

dt



t0

t0

dt



+a ρ(t0 ) a IP

dt t0

+ a ρ(t0 ) a

dt



t0

− ρ(t0 ) a a IP †

dt



t0

dt 

− ρ(t0 )IP aa† t0

 l

)(t  −t  )

l



dt

 |Kl | nl e 2

−i(ω◦ −ωl )(t  −t  )

l

dt





 |Kl | nl + 1e 2

+i(ω◦ −ωl )(t  −t  )

l



t  dt





 |Kl | nl e 2

−i(ω◦ −ωl )(t  −t  )

l



t  dt t 







t 

t0

|Kl |2 nl + 1e

+i(ω◦ −ω







 |Kl | nl + 1e 2

+i(ω◦ −ωl )(t  −t  )

l

t0

t0+t



dt 

t0

t0+t

|Kl | nl + 1e



t0

t0+t IP †

t 

t0 

 +i(ω◦ −ωl )(t  −t  )

2

l

t

t0 t0+t †

dt





t0

t0+t

+ a ρ(t0 ) a

 t0

dt 

IP †



t

t0+t

+ a† ρ(t0 )IP a

l

t0

 dt 

 l

 |Kl |2 nl e

−i(ω◦ −ω

l

)(t  −t  )

(16.76)

CH016.tex

17/3/2011

9: 54

Page 485

16.2

16.2.4.2 Markov approximation of the form t0+t

dt 

I= t0

t 

SECOND-ORDER SOLUTION OF EQ. (16.41)

485

Observe that in Eq. (16.2.4.1) integrals appear 

dt 

Al e+i(ω

◦ −ω

l )(t

 −t  )

(16.77)

l

t0

where Al is given either by Al = |Kl |2 nl 

(16.78)

Al = |Kl |2 nl + 1

(16.79)

or by

Next, make in the integral (16.77) the following changes of variable τ ≡ t  − t 

ξ ≡ t − t

and

leading to t  = t + t0 − ξ

dt  dt  = dξ dτ

and

Then, the double integrals (16.77) becomes t0+t

I=

t0 +t−ξ 

dξ t0







dτ

Al e

+i(ω◦ −ωl )τ

(16.80)

l

t0

Notice that the infinite sum appearing in this last equation involves imaginary exponentials, the time-independent arguments of which are quasi-continuously varying, so that this sum must vanish when the time τ becomes greater than the correlation time τc :  ◦ Al e+i(ω −ωl )τ  0 if τ > τc (16.81) l

Next, examine in details Eq. (16.80) at the upper limit (t0 + t − ξ) of the integral over the τ variable. Owing to the approximation (16.81), the contribution of the integrand to the integration over τ is negligible for τ > τc , so that this integration limit may be extended from (t0 + t − ξ) to infinity without any sensible changes (see Fig.16.1). Such an approximation, which implies some lack of memory, is known, in the statistical physics of irreversible processes, as the Markov approximation. Hence, we may write the following approximate equation:    t0 +t−ξ  ∞   +i(ω◦ −ωl )τ +i(ω◦ −ωl )τ ≈ dτ dτ Al e Al e t0

l

t0

l

So, the integral (16.80) may be approximated by  t0+t ∞  +i(ω◦ −ωl )τ I≈ dξ dτ Al e t0

t0

l

CH016.tex

486

17/3/2011

9: 54

Page 486

DAMPED OSCILLATORS

t

B

t Δt

τc

A

t 0 t

tτ

t

t  Δt

Integration area over t  and t  .

Figure 16.1

which, after integration over ξ of these integrals, yields  ∞  +i(ω◦ −ωl )τ dτ I ≈ t Al e

(16.82)

l

t0

16.2.4.3 Time variation of the IP density operator of the IP density operator over the time interval t IP :

Now, consider the variation

 ρ(t)IP

ρ(t0 + t)IP − ρ(t0 )IP = t t Due to Eqs. (16.78) and (16.79) and with the help of Eq. (16.82) yielding approximately the value of the integrals (16.77), Eq. (16.2.4.1) reads ⎧ ⎫ ∞ ⎬ ⎨  ρ(t)IP ◦ = −a† a ρ(t0 )IP |Kl |2 nl + 1 ei(ω −ωl )τ dτ ⎩ ⎭ t l

ρ(t0 )IP − aa†

⎩

e−i(ω ∞

|Kl |2 nl + 1

e

⎩

l )τ

dτ

−i(ω◦ −ω

0

⎧ ⎨

∞ |Kl |2  nl 

l

◦ −ω

⎫ ⎬ ⎭

0

⎧ ⎨ l

+ a ρ(t0 )IP a†

∞ |Kl |2 nl 

l

+ a† ρ(t0 )IP a

0

⎧ ⎨

e

⎩

0

i(ω◦ −ω

l )τ

l )τ

⎫ ⎬ dτ

⎭

⎫ ⎬ dτ

⎭

CH016.tex

17/3/2011

9: 54

Page 487

16.2

ρ(t0 )IP a + a†

⎧ ⎨

16.2.5

∞ e

⎩

l )τ

−i(ω◦ −ω

l )τ

∞ |Kl |2 nl + 1

⎩

e−i(ω

dτ

⎭

⎫ ⎬ dτ

◦ −ω

0

⎧ ⎨

∞ |Kl |2 nl 

e

⎩

i(ω◦ −ω

l )τ

487

⎫ ⎬

⎭

0

⎧ ⎨

l

i(ω◦ −ω

0

|Kl |2 nl 

l

− ρ(t0 )IP aa†

e

⎩

⎧ ⎨ l

− ρ(t0 )IP a† a

∞ |Kl |2 nl + 1

l

+ a ρ(t0 )IP a†

SECOND-ORDER SOLUTION OF EQ. (16.41)

l )τ

⎫ ⎬ dτ

⎭

⎫ ⎬ dτ

(16.83)

⎭

0

IP master equation for the density operator

16.2.5.1 Continuous approximation for the thermal bath It is now convenient to make the approximation of considering the set of oscillators of the thermal bath as continuous and thus to pass in Eq. (16.83) from the sums over the thermal bath oscillator to integrals over the continuous angular frequency variable concerning these oscillators, according to 

+∞ |Kl | nl  → g(ω)|K(ω)|2 n(ω)dω 2

l

(16.84)

−∞

Here, g(ω) is the mode density of the thermal bath, K(ω) is the coupling between oscillators of angular frequency ω, whereas n(ω) is the mean number occupation of the oscillator of angular frequency ω which, due to Eq. (16.69), is 1 nl (ωl ) = ω /kT (16.85) l e −1 Owing to this approximation, Eq. (16.83) becomes  ρ(t)IP ρ(t0 ) ∗0 = −a† a ρ(t0 ) 1 − aa† t + a† ρ(t0 )a† 0 ρ(t0 )a ∗1 + a + a† ρ(t0 )a† ∗0 ρ(t0 )a 1 + a − ρ(t0 )a† a ∗1 − ρ(t0 )aa† 0

(16.86)

where +∞

0 ≡

∞ g(ω)|K(ω)| n(ω) 2

−∞

dτ ei(ω

∞ g(ω)|K(ω)| n(ω) + 1 2

−∞

dω

(16.87)

0

+∞

1 ≡

◦ −ω)τ

dτ ei(ω 0

◦ −ω)τ

dω

(16.88)

CH016.tex

488

17/3/2011

9: 54

Page 488

DAMPED OSCILLATORS

16.2.5.2 Calculation of Ω0 and Ω1 One has now to calculate the double integrals (16.87) and (16.88). For this purpose, observe that these integrals are of the form (18.63) +∞

=

⎛∞ ⎞  ◦ f (ω) ⎝ e−i(ω−ω )τ dτ ⎠ dω

−∞

0

with f (ω) = g(ω)|K(ω)|2 n(ω)

g(ω)|K(ω)|2 n(ω) + 1

or

Then, keeping in mind that, as shown in Section 18.6, Eq. (18.63) leads to Eq. (18.71), that is, ⎧ +∞ ⎫  ⎬  +∞ ⎨ 1

= −i dω dω − f (ω)P f (ω)πδ(ω − ω◦ ) dω ⎩ ⎭ ω − ω◦ −∞

−∞

where P denotes the Cauchy principal part, it appears that Eq. (16.87) reads +∞

0 = −∞

⎫ ⎧ +∞  ⎬  ⎨ 1 dω g(ω)|K(ω)|2 n(ω)δ(ω − ω◦ ) dω − i g(ω)|K(ω)|2 P ⎭ ⎩ ω − ω◦ −∞

or

0 = n(ω◦ )

γ 2

+ i ω

 (16.89)

where ω is an angular frequency shift and γ a damping parameter given, respectively, by

ω ≡ −

⎧ +∞ ⎨ ⎩

 g(ω)|K(ω)|2 P

−∞

1 ω − ω◦

γ ≡ 2πg(ω◦ )|K(ω◦ )|2

⎫ ⎬

 dω

⎭

(16.90)

(16.91)

In like manner, Eq. (16.88) is

1 = n(ω◦ ) + 1

γ 2

+ i ω

 (16.92)

CH016.tex

17/3/2011

9: 54

Page 489

16.2

SECOND-ORDER SOLUTION OF EQ. (16.41)

489

16.2.5.3 New expression for time variation of IP density operator Owing to Eq. (16.89) and (16.92), Eq. (16.86) transforms to γ   ρ(t)IP ρ(t0 )IP = −a† a + i ω n + 1 (16.93) t 2 γ  − aa† ρ(t0 )IP − i ω n 2 γ  + a† ρ(t0 )IP a − i ω n + 1 2   IP † γ + a ρ(t0 ) a + i ω n 2 γ  † IP +a ρ(t0 ) a + i ω n + 1 2   IP † γ + a ρ(t0 ) a − i ω n 2 γ  IP † − ρ(t0 ) a a − i ω n + 1 2 γ  − ρ(t0 )IP aa† + i ω n (16.94) 2 with n ≡ n(ω◦ )

(16.95)

16.2.5.4 IP master equation of undriven damped density operator Now, in order to pass from the infinitesimal change in a time interval t of the IP time density operator given by Eq. (16.94) to a partial time derivative, take t = (t  − t0 ) → 0 leading to 

 ρ(t)IP t



 →

∂ ρ(t)IP ∂t



then, according to the transformation Eq. (16.96), Eq. (16.94) yields   γ  ∂ ρ(t)IP = −a† a ρ(t)IP + i ω n + 1 ∂t 2 γ  ρ(t)IP − i ω n − aa† 2  γ † IP +a − i ω n + 1 ρ(t) a 2 γ  + a ρ(t)IP a† + i ω n 2 γ  + a† ρ(t)IP a + i ω n + 1 2 γ  + a ρ(t)IP a† − i ω n 2

(16.96)

CH016.tex

490

17/3/2011

9: 54

Page 490

DAMPED OSCILLATORS

γ

 − i ω n + 1 2 γ  − ρ(t)IP aa† + i ω n 2 which, after rearranging, becomes   ∂ ρ(t)IP = + i[ ρ(t)IP , a† a]ω ∂t γ − (a† a ρ(t)IP + ρ(t)IP a† a − 2 a ρ(t)IP a† ) 2 − nγ(a† a ρ(t)IP + ρ(t)IP aa† − a† ρ(t)IP a† ) (16.97) ρ(t)IP a− a − ρ(t)IP a† a

This last equation represents the coarse-grained time evolution of the reduced IP density operator of the driven damped harmonic oscillator, which is named the IP master equation of this oscillator. 16.2.5.5 IP master equation of driven damped density operator Now, in order to pass from the IP Liouville equation (16.97) dealing with the undriven damped harmonic oscillator to the corresponding one dealing with the driven damped harmonic oscillator, use Eq. (16.40)     ∂ ρDr (t)IP ∂ ρ(t)IP IP IP

i (16.98) = [HDr (t) , ρDr (t) ] + i ∂t ∂t Next, in the independent variations approximation, it may be assumed that the instantaneous action of the damping on the oscillator of interest is the same whether the oscillator is driven or undriven, so that one may write in this spirit   ∂ ρ(t)IP = + i[ ρDr (t)IP , a† a]ω ∂t γ − (a† a ρDr (t)IP + ρDr (t)IP a† a − 2 a ρDr (t)IP a† ) 2 −nγ(a† a ρDr (t)IP + ρDr (t)IP aa† − a† ρDr (t)IP a† ) ρDr (t)IP a− a (16.99) 

Hence, using Eq. (16.36) and due to Eq. (16.99), Eq. ( 16.98) becomes  1 ∂ ρDr (t)IP IP IP = [H Dr (t) , ρDr (t) ] ∂t i + i[ ρDr (t)IP , a† a]ω γ ρDr (t)IP + ρDr (t)IP a† a − 2 a ρDr (t)IP a† ) − (a† a 2 − nγ(a† a ρDr (t)IP + ρDr (t)IP aa† − a† ρDr (t)IP a† ) ρDr (t)IP a− a (16.100)

Equation (16.100) is the IP Liouville equation of the driven damped quantum harmonic oscillator.

CH016.tex

17/3/2011

9: 54

Page 491

16.2

16.2.6

SECOND-ORDER SOLUTION OF EQ. (16.41)

491

Schrödinger picture master equation

Now, to return to the Schrödinger picture, keep in mind that, due to Eq. (3.184) allowing one to pass from the IP density operator ρ(t)IP to the corresponding SP one SP ρ(t) , that is ρ(t)SP = U◦ (t) ρ(t)IP U◦ (t)−1

(16.101)

where U◦ (t) = e−iH

◦ t/

(16.102)

the time derivative of the Schrödinger picture density operator is given by Eq. (3.190), that is,     ∂ρ(t)SP 1 ◦ ∂ ρ(t)IP SP ◦ (16.103) = [H , ρ(t) ] + U (t) U◦ (t)−1 ∂t i ∂t In the present situation, where the Hamiltonian H◦ is given by Eq. (16.9), that is, H◦ = a† aω◦

(16.104)

the time evolution operator appearing in Eqs. (16.101) and (16.103) is U◦ (t) = e−ia

†a

ω◦ t

(16.105)

Hence, returning to the Schrödinger picture, from Eq. (16.100), by using an equation of the form (16.103) we have       ∂ ρDr (t)SP 1 1 ◦ SP

Dr (t)IP , = [H , U◦ (t)[H ρDr (t) ] + ρDr (t)IP ]U◦ (t)−1 ∂t i i − i ωU◦ (t)[a† a, ρDr (t)IP ]U◦ (t)−1 γ − U◦ (t)(a† a ρDr (t)IP + ρDr (t)IP a† a − 2 a ρDr (t)IP a† )U◦ (t)−1 2 − nγU◦ (t)(a† a ρDr (t)IP + ρDr (t)IP aa† − a† ρDr (t)IP a† )U◦ (t)−1 ρDr (t)IP a− a Of course, owing to the expression (16.104) of

H◦ ,

(16.106)

it appears that

[H◦ , ρDr (t)SP ] = [a† a, ρDr (t)SP ]ω◦

(16.107)

Next, inserting the unity operator built up from the evolution operator (16.102), we have

Dr (t)IP U◦ (t)H ρDr (t)IP U◦ (t)−1

Dr (t)IP U◦ (t)−1 U◦ (t) = U◦ (t)[H ρDr (t)IP ]U◦ (t)−1 so that, using Eqs. (16.34) and (16.101), due to Eq. (16.104), we have SP

Dr (t)IP

U◦ (t)H ρDr (t)IP U◦ (t)−1 = HDr ρDr (t)SP

In like manner SP

Dr (t)IP U◦ (t)−1 = U◦ (t) ρDr (t)IP H ρDr (t)SP HDr

CH016.tex

492

17/3/2011

9: 54

Page 492

DAMPED OSCILLATORS

so that the second right-hand-side commutator of Eq. (16.106) reads

Dr (t)IP , U◦ (t)[H ρDr (t)IP ]U◦ (t)−1 = [HDr , ρDr (t)SP ]

(16.108)

Hence, as a consequence of Eqs. (16.107) and (16.108), Eq. (16.106) becomes   ∂ρDr (t)SP i ρDr (t)SP ] − [HDr , ρDr (t)SP ] = −iω◦ [a† a, ∂t  − i ωU◦ (t)[a† a, ρDr (t)IP ]U◦ (t)−1 γ − U◦ (t)(a† a ρDr (t)IP + ρDr (t)IP a† a − 2 a ρDr (t)IP a† )U◦ (t)IP−1 2 − nγU◦ (t)(a† a ρDr (t)IP + ρDr (t)IP aa† − a† ρDr (t)IP a − a ρDr (t)IP a† )U◦ (t)−1

(16.109)

Now, one has to get the result of the canonical transformations involved on the right-hand side of Eq. (16.109). First, consider the canonical transformation over a† a that, according to Eqs. (16.101) and (16.105), is U◦ (t) ρDr (t)IP a† aU◦ (t)−1 = e−ia

† aω◦ t

ρDr (t)IP a† aeia

† aω◦ t

Now, the operator commutes with the exponential operator, hence this expression simplifies to U◦ (t) ρDr (t)IP a† aU◦ (t)−1 = e−ia

† aω◦ t

ρDr (t)IP eia

† aω◦ t

a† a

Then, Eqs. (16.101) and (16.105) allow one to transform this equation into U◦ (t) ρDr (t)IP a† aU◦ (t)−1 = ρDr (t)SP a† a

(16.110)

In like manner, we have the following results for the other canonical transformations of interest: U◦ (t)a† a ρDr (t)IP U◦ (t)−1 = a† a ρDr (t)SP

(16.111)

U◦ (t)aa† ρDr (t)IP U◦ (t)−1 = aa† ρDr (t)SP

(16.112)

U◦ (t) ρDr (t)IP aa† U◦ (t)−1 = ρDr (t)SP aa†

(16.113)

[a, a† ] = 1

where has been used for the two last results. Hence, collecting Eqs. (16.110)–(16.113) and using Eq. (16.10), the master equation (16.109) becomes after simplification   ∂ρDr (t)SP = −iα◦ ω◦ {[a, ρDr (t)SP ] + [a† , ρDr (t)SP ]} ∂t − iω◦ [a† a, ρDr (t)SP ] − i ω [a† a, ρDr (t)SP ] γ − (a† aρDr (t)SP + ρDr (t)SP a† a − 2aρDr (t)SP a† ) 2 − nγ(a† aρDr (t)SP + ρDr (t)SP aa† − a† ρDr (t)SP a− aρDr (t)SP a† ) (16.114)

CH016.tex

17/3/2011

9: 54

Page 493

16.2

SECOND-ORDER SOLUTION OF EQ. (16.41)

493

Recall here that γ is the damping parameter induced by the irreversible influence of the thermal bath, whereas ω is an angular frequency shift induced by the bath, and where n is the thermal average of the occupation number, which, owing to Eq. (16.69), is n =

1 ◦ eω /kT

−1

16.2.7 Matrix representation of master equation (16.114) in basis of harmonic Hamiltonian Now, consider the matrix representation of the master equation (16.114) in the basis {|(n)} of the eigenkets of a† a, obeying a† a|(n) = n|(n) with

(m)|(n) = δmn

In this basis, a matrix element of Eq. (16.114) becomes ∂ (m)|ρDr (t)SP |(n) = −iα◦ (m)|aρDr (t)SP |(n) − (m)|ρDr (t)SP a|(n) ∂t − iα◦ {(m)|a† ρDr (t)SP |(n) − (m)|ρDr (t)SP a† |(n)} − i{(m)|a† aρDr (t)SP |(n) − (m)|ρDr (t)SP a† a|(n)} γ − {(m)|a† aρDr (t)SP |(n) + (m)|ρDr (t)SP a† a|(n)} 2 + γ{(m)|aρDr (t)SP a† |(n)} − nγ{(m)|a† aρDr (t)SP |(n) + (m)|ρDr (t)SP a† a|(n)} − nγ{(m)|a† ρDr (t)SP a|(n) + (m)|aρDr (t)SP a† |(n)} (16.115) Recall that the diagonal elements corresponding to m = n, are called populations, whereas the off-diagonal ones are called coherences. To get expressions for the right-hand-side matrix elements appearing in Eq. (16.115) it is suitable to use Eqs. (5.53) and (5.63) giving the actions of a† and a on |(n) and their Hermitian conjugates, that is, √ √ and a|(n) = n|(n − 1) a† |(n) = n + 1|(n + 1) √ (n)|a = (n + 1)| n + 1

and

√ (n)|a† = (n − 1)| n

Then, in view of these expressions, Eq. (16.115) becomes, after omitting the SP notation for the matrix elements,   ∂ρm,n (t) = −i(m − n){ρm,n (t)} ∂t √ √ − iα◦ { m + 1{ρm+1,n (t)} − n{ρm,n−1 (t)} √ √ + m{ρm−1,n (t)} − n + 1{ρm,n+1 (t)}}

CH016.tex

494

17/3/2011

9: 54

Page 494

DAMPED OSCILLATORS

 γ − {(m + n){ρm,n (t)} − 2 (m + 1)(n + 1){ρm+1,n+1 (t)}} 2  + nγ{ (m + 1)(n + 1){ρm+1,n+1 (t)} √ − (m + n){ρmn (t)} + mn{ρm−1,n−1 (t)}}

(16.116)

with {ρmn (t)} = (m)|ρDr (t)SP |(n) Equation (16.116) may be solved if one knows the initial condition for the various values of these matrix elements ρm,n (0)SP at initial time t = 0, that is, the expression of the density operator ρDr (0)SP of the driven damped harmonic oscillator at this initial time. This may be numerically performed, for instance, by the aid of the Runge– Kutta method. However, that may be avoided since, as we will see, an analytical expression of the reduced time evolution operator of the driven damped harmonic oscillator exists. This may be viewed as the integrated form of a generalization of the master equation (16.114) resulting from an infinite order expansion in the coupling

V(t)IP of Eq. (16.42).

16.3 FOKKER–PLANCK EQUATION CORRESPONDING TO (16.114) However, before seeking such generalization of the master equation, it may be of interest to show how this master equation (16.114) may be transformed into a scalar partial of the same type as the Fokker–Planck equations encountered in the area of classical statistical mechanics treating irreversible processes dealing with Brownian oscillators. With this in mind, it is convenient to convert the SP master equation (16.114) into the antinormal order and thus, for this purpose, to first consider the action of aa† on the density operator in the following way: aa† ρ(t) = a(a† ρ(t) − ρ(t)a† + ρ(t)a† )

(16.117)

which reads aa† ρ(t) = a([a† , ρ(t)] + ρ(t)a† ) Again, applying Eq. (7.59), that is,



[a† , {f(a, a† )}] = −

∂f(a, a† ) ∂a



to the function ρ(t) = ρ(a, a† , t) yields

 [a† , ρ(t)] = −

∂ρ(t) ∂a

 (16.118)

CH016.tex

17/3/2011

9: 54

16.3

Page 495

FOKKER–PLANCK EQUATION CORRESPONDING TO (16.114)

so that Eq. (16.117) transforms to the antinormal order form  a  ∂ρ (t) aa† ρ(t) = −a + aρa (t)a† ∂a

495

(16.119)

Next, pass to the action of a† a on ρ(t), which may be written using the commutation rule [a, a† ] = 1, a† aρ(t) = (aa† − 1)ρ(t) = aa† ρ(t) − ρ(t) Then, using Eq. (16.119), this expression takes the antinormal order form  a  ∂ρ (t) † a aρ(t) = −a (16.120) + aρa (t)a† −ρa (t) ∂a Now, the commutation rule [a, a† ] = 1 of the Boson operators allows one to write ρ(t)a† a = ρ(t)(aa† −1) = ρ(t)aa† − ρ(t) which reads after adding and substracting the same term ρ(t) ρ(t)a† a = (ρ(t)a − aρ(t) + aρ(t))a† − ρ(t) or ρ(t)a† a = ([ρ(t), a]a† + aρ(t))a† − ρ(t) Then observing that Eq. (7.60) allows one to write   ∂ρ(t) [ρ(t), a] = − ∂a† the left-hand side of Eq. (16.121) takes the antinormal form  a  ∂ρ (t) † † a + aρa (t)a† − ρa (t) ρ(t)a a = − ∂a†

(16.121)

(16.122)

(16.123)

Moreover, using the following commutation rule ρ(t)a† a = ρ(t)aa† − ρ(t) on the left-hand side of Eq. (16.123), this expression leads to the antinormal form  a  ∂ρ (t) † † a + aρa (t)a† (16.124) ρ(t)aa = − ∂a† Next, to find the antinormal expression of a† ρ(t)a, write it by adding and subtracting the same term aρ(t) according to a† ρ(t)a = a† (ρ(t)a − aρ(t) + aρ(t)) so that a† ρ(t)a = a† ([ρ(t), a] + aρ(t))

CH016.tex

496

17/3/2011

9: 54

Page 496

DAMPED OSCILLATORS

and thus, owing to Eq. (16.122), a ρ(t)a = −a †

 †

∂ρ(t) ∂a†

 + a† aρ(t)

(16.125)

At last, to come further in the quest of the antinormal form of a† ρ(t)a denote   ∂ρ(t) (16.126) f(t) ≡ ∂a† so that the first right-hand side of Eq. (16.125) reads   † ∂ρ(t) a = a† f(t) ∂a† which after adding and substracting the same term f(t)a† gives   † ∂ρ(t) = a† f(t) + f(t)a† − f(t)a† a ∂a† or   † ∂ρ(t) a = [a† , f(t)] + f(t)a† ∂a† Then, applying Eq. (16.118) with f(t) in place of ρ(t) gives     ∂ρ(t) ∂f(t) a† = − + f(t)a† ∂a† ∂a or, after returning from f(t) to ρ(t) by the aid of (16.126) we have    2    ∂ ρ(t) ∂ρ(t) † † ∂ρ(t) =− + a a ∂a† ∂a∂a† ∂a† which transforms using of Eq. (16.125) into     2 ∂ρ(t) † ∂ ρ(t) † − a + a† a ρ(t) a ρ(t)a = ∂a∂a† ∂a† so that, due to Eq. (16.120) allowing to transform the last right-hand side, we have the final result for the antinormal form of a† ρ(t)a:  2 a   a   a  ∂ρ (t) † ∂ ρ (t) ∂ρ (t) † − a −a a ρ(t)a = + aρa (t)a† −ρa (t) † † ∂a∂a ∂a ∂a (16.127) Hence, collecting Eqs. (16.120) and (16.123) and because aρ(t)a† is yet antinormal, the right-hand-side term involving γ/2 in the master equation (16.114) yields after simplification  a   a  ∂ρ (t) † ∂ρ (t) a + a + 2ρa (t) (16.128) 2aρ(t)a† − a† aρ(t) − ρ(t)a† a = ∂a† ∂a or    a  a∂(ρa (t)) ∂(ρ (t)a† ) † † † + 2aρ(t)a − a aρ(t) − ρ(t)a a = (16.129) ∂a† ∂a

CH016.tex

17/3/2011

9: 54

16.3

Page 497

FOKKER–PLANCK EQUATION CORRESPONDING TO (16.114)

497

Thus, with Eqs. (16.120), (16.124), and (16.127), one obtains after simplification of the last right-hand side of the master equation (16.114) involving nγ, the following antinormal expression: a† aρ(t)+ρ(t)aa† − a† ρ(t)a − aρ(t)a† =

∂2 ρa (t) ∂a∂a†

(16.130)

Now, the commutators multiplying α◦ ω◦ appearing on the right-hand side of the master equation (16.114) may also be transformed into an antinormal form involving partial derivatives, by the aid of Eqs. (16.118) and (16.122), that is,  [a, ρ(t)] + [a†, ρ(t)]

=

  a  ∂ρ (t) ∂ρa (t) − † ∂a ∂a

(16.131)

Finally, the commutator multiplying the term i(ω◦ + ω) appearing on the right-hand side of the master equation (16.114) may be written after adding and subtracting the same term a† ρ(t)a as [a† a, ρ(t)] = a† aρ(t) − ρ(t)a† a + a† ρ(t)a − a† ρ(t)a or [a† a, ρ(t)] = a† [a, ρ(t)] + [a† , ρ(t)]a so that, in view of Eqs. (16.118) and (16.122), it transforms to the antinormal form  [a a, ρ(t)] = a †

†

∂ρa (t) ∂a†



 −

 ∂ρa (t) a ∂a

(16.132)

Thus, collecting Eqs. (16.129)–(16.132), the master equation (16.114) may be put into the following antinormal form: 

∂ρa (t) ∂t





  a  ∂ρa (t) ∂ρ (t) − ∂a† ∂a  a   a  ∂ρ (t) † ∂ρ (t) ◦ a −a − i(ω + ω) † ∂a ∂a  a   a   2 a  † γ ∂ρ (t)a ∂ρ (t)a ∂ ρ (t) + + (16.133) + nγ 2 ∂a† ∂a ∂a ∂a†

= −iα◦ ω◦

Now, it is possible to pass to the scalar representation of this equation using Eq. (7.41), which reads in the present situation ˆ −1 {ρa (a, a† , t)} = {ρa (α, α∗ , t)} A

CH016.tex

498

17/3/2011

9: 54

Page 498

DAMPED OSCILLATORS

so that the above equation (16.133) transforms to the following second-order partial differential equation:  a   a   a  ∂ρ (t) ∂ρ (t) ∂ρ (t) = −iα◦ ω◦ − ∂t ∂α∗ ∂α  a    a  ∂ρ (t) ◦ ∗ ∂ρ (t) −α − i(ω + ω) α ∗ ∂α ∂α   a  a   2 a  ∗ γ ∂ρ (t)α ∂ρ (t)α ∂ ρ (t) + + (16.134) + nγ 2 ∂α∗ ∂α ∂α ∂α∗ This last equation is the Fokker–Planck equation corresponding to the master equation (16.114) governing the dynamics of the driven damped quantum harmonic oscillator to second order in the expansion of the coupling of the oscillator with the thermal bath.

16.4 NONPERTURBATIVE RESULTS FOR DENSITY OPERATOR Recall that the master equation (16.114) from which results the Fokker–Planck equation (16.134) is a partial derivative equation of time, which takes into account, via Eq. (16.42), the irreversible influence of the thermal bath through a second-order expansion of the coupling between the oscillator and the thermal bath. However, a known closed expression for the density operator, at any time t, of driven damped harmonic oscillators, exists, which may be viewed as the result of the integration of a master equation of the same kind as (16.114) but that takes into account the coupling with the thermal bath, to infinite order in place of second order. The demonstration of this closed expression, due to Louisell and Walker1 involves a very complicated treatment that is beyond the level of this book. Hence, in the present book, we shall only give the results of this treatment, leaving for the end of this chapter to show that it is possible to get also their result with the help of another treatment requiring knowledge of the IP time evolution of driven damped quantum harmonic oscillators.

16.4.1

Model

The Hamiltonian for the quantum harmonic oscillator weakly coupled linearly to a bath of oscillators is the same as above, that is,  (a† bj κj + ab†j κj∗ ) H = (a† a+α◦ (a† + a)) + +



j

b†j bj ωj

with

k = 0, 1

(16.135)

j

Just as for the master equation above, the density operator of the thermal bath is considered as the product of the Boltzmann density operators of the bath oscillators, 1

W. Louisell and L. Walker. Phys. Rev., 137 (1965): 204.

CH016.tex

17/3/2011

9: 54

16.4

that is, ρθ =



Page 499

NONPERTURBATIVE RESULTS FOR DENSITY OPERATOR

†

(1 − e−λk )(e−λk bk bk )

with

trθ {ρθ } = 1

499

(16.136)

k

where λk =

ωk kB T

(16.137)

Louisell and Walker considered that, at an initial time, an equilibrium density operator, which is that of a coherent state at temperature T , is thus given by (13.111), so that ρ(0) = (1 − e−λ )e−λ(a

† −α∗ )(a−α ) c c

(16.138)

with λ=

ω kB T

(16.139)

The reader should be aware that the dimensionless scalar parameter αc characterizing the coherent density operator (16.138) has to be clearly distinguished from the dimensionless parameter α◦ reflecting the strength of the driving term in the Hamiltonian (16.135). Furthermore, the full density operator at initial time is taken as the product of the density operator (16.138) times that (16.136) of the thermal bath (16.138), that is, † † ∗ ρTot (0) = (1 − e−λ ){e−λ(a −αc )(a−αc ) } (1 − e−λk )(e−λk bk bk ) (16.140) k

The Liouville equation to be solved was   ∂ρTot (t) i = [H, ρTot (t)] ∂t subject to the boundary condition (16.140), while the density operator of the damped oscillator was obtained from ρTot (t) by performing a partial trace over the thermal bath eigenstates, according to ρ(t) = trθ {ρTot (t)}

16.4.2

Damped density operator at time t

For this model, and using a very long and complicated procedure involving the Markov approximation as for the master equation, Louisell and Walker have found that ∗ ˆ ρ(t) ∼ − φ(t))(α−φ∗ (t))}} = εN{exp{−ε(α

ˆ is the normal ordering operator, and α and where N distinguished from α◦ and αc , whereas φ(t) is given by

α∗

(16.141)

scalar complexes to be

     t γt γt  ◦ ◦ ◦  ∼ φ(t) = αc exp −i(ω + ω)t − − i α exp −i(ω + ω)t − dt  2 2 0

(16.142)

CH016.tex

17/3/2011

500

9: 54

Page 500

DAMPED OSCILLATORS

Then, using the normal ordering operation and after integration of Eq. (16.142), they have obtained an expression for the density operator that is very similar to that of (16.138) of the initial situation, that is, ρ(t) = ε exp{−λ(a† − φ(t))(a− φ∗ (t))}

(16.143)

with

   γt ◦ + β{e−(t/2) e+i(ω +ω)t − 1} φ(t) = αc exp −i(ω + ω)t − 2

(16.144)

and where β=

α◦ (2ω◦2 + iγω◦ ) 2(ω◦2 + γ 2 /4)

(16.145)

Note in the passage from Eq. (16.141) to Eq. (16.143), the change of ε into λ, and that the expressions of the angular frequency shift ω and of the relaxation parameter γ are here the same as those in (16.90) and (16.91) encountered in the calculation of the master equation (16.114). Moreover, the matrix elements of the time-dependent density operator (16.143) in the representation where a† a is diagonal are {n}|ρ(t)|{m} (y − 1)n = (y)n+1 if



φ∗ (t) y

m−n 

n! m!

1/2

    |φ(t)|2 |φ(t)|2 m−n − Ln exp − y y(y − 1)

m ≥ n,

n−m (x)} is the generalized Laguerre polynomial function of the variable x, where {Lm with a similar relation when n ≥ m by permuting everywhere n and m.

16.4.3

Dynamics of averaged damped elongation

Now, as an application of the expression (16.143) of the density operator ρ(t), we study how the average value of the position operator Q evolves with time when the oscillator is driven and damped. 16.4.3.1 Damped translation operator For this purpose, consider the special situation of a driven damped harmonic oscillator, starting at initial time from an undriven situation corresponding therefore to the situation αc = 0 in Eq. (16.138), that is, to ρ0 (0) = (1 − e−λ )(e−λa a ) †

(16.146)

Then, the density operator of the driven damped oscillator will be given by Eq. (16.143), that is, ρ0 (t) = (1 − e−λ ) exp{−λ(a† − φ0 (t))(a− φ0∗ (t))}

(16.147)

which involves a time-dependent argument (16.144) reducing to φ0 (t) = β{e−γt/2 e+i(ω

◦ +ω)t

− 1}

(16.148)

CH016.tex

17/3/2011

9: 54

Page 501

16.4

NONPERTURBATIVE RESULTS FOR DENSITY OPERATOR

501

Observe that in this special situation the density operator (16.147) has the same structure as that (13.11) obtained at any time using the Lagrange multiplier method, so that it appears to be the density operator of a coherent state at some temperature. Now, observe that it is possible to consider the density operator (16.143), which here reads (16.147), as the result of the following canonical transformation: ρ0 (t) = A(φ0 (t))ρ0 (0)A(φ0 (t))−1

(16.149)

with ∗

A(φ0 (t)) = (eφ0 (t)a

† −φ

0 (t)a

)

(16.150)

Now, owing to Eq. (7.9) A(φ0 (t)){f(a† , a)}A(φ0 (t))−1 ∗

= (eφ0 (t)a

† −φ

0 (t)a

∗

){f(a† , a)}(e−φ0 (t)a

† +φ

0 (t)a

) = {f(a† − φ0 (t), a − φ0∗ (t))} (16.151)

so that, as required, A(φ0 (t))(e−λa a )A(φ0 (t))−1 = e−λ(a †

† −φ

∗ 0 (t))(a−φ0 (t))

Hence, owing to Eq. (16.148), the damped translation operator (16.150) allowing to pass from the initial Boltzmann density operator (16.146) to the damped density operator at time t (16.147) is ◦

◦

A(φ0 (t)) = exp{β∗ {e−γt/2 eiω t − 1}a† − β{e−γt/2 e−iω t − 1}a}

(16.152)

16.4.3.2 Damped average elongation Besides, knowledge of the timedependent density operator ρ(t) allows us to get the time dependence of the average value of the position operator Q according to Q(t) = tr{ρ(t)Q} Owing to Eq. (16.149)) and due to Eq. (5.6) giving Q in terms of Boson operators, this equation yields

 Q(t) = tr{A(φo (t))ρo (0)A(φo (t))−1 (a† + a)} 2Mω◦ or, in view of Eq. (16.146),

 † Q(t) = ε tr{A(φo (t))(e−λa a )A(φo (t))−1 (a† + a)} ◦ 2Mω Again, according to the invariance of the trace with respect to a circular permutation, we have

 † Q(t) = ε tr{(e−λa a )A(φ0 (t))−1 (a† + a)A(φ0 (t))} 2Mω◦ Then, theorem (7.9) allows us to transform this expression into

 † Q(t) = ε tr{(e−λa a )(a† + φ0 (t) + a+ φ0∗ (t))} ◦ 2Mω

CH016.tex

502

17/3/2011

9: 54

Page 502

DAMPED OSCILLATORS

Then, performing the trace over the eigenstates of a† a gives

  † Q(t) = ε {n}|(e−λa a )(a† + φ0 (t) + a+ φ0∗ (t))|{n} ◦ 2Mω n

(16.153)

Moreover, keeping in mind Eq. (13.66), that is, {n}|(e−λa a ){a† + a}|{n} = 0 †

and using the orthonormality properties of the basis used for the trace, Eq. (16.153) becomes

  † Q(t) = ε {n}|(e−λa a )|{n}(φ0 (t) + φ0∗ (t)) (16.154) ◦ 2Mω n Next observe that the trace of the Boltzmann density operator, which appears in this last equation, is just unity:  † † ε tr{e−λa a } = 1 = ε {n}|(e−λa a )|{n} n

so that Eq. (16.154) simplifies to



 (φ0 (t) + φ0∗ (t)) 2Mω◦ Besides, in view of Eqs. (16.145) and (16.148), and after incorporating the shift ω into ω◦ , it becomes

 1 ◦ Q(t) = α ◦ ◦2 2Mω 2(ω + γ 2 /4) Q(t) =

◦

◦

× {(2ω◦2 + iγω◦ ){e−γt/2 e+iω t − 1} + (2ω◦2 − iγω◦ ){e−γt/2 e−iω t − 1}} so that

◦

Q(t) = α

 2Mω◦



 2ω◦2 (e−γt/2 cos ω◦ t − 1) ω◦2 + γ 2 /4   γω◦ −γt/2 ◦ − sin ω t e ω◦2 + γ 2 /4

(16.155)

We give in Fig. 16.2, the time evolution of the average position for the driven damped quantum harmonic oscillator. Note that in the very underdamped situation where ω◦ >> γ, Eq. (16.155) reduces to

 ◦ (e−(γt/2) cos ωt − 1) Q(t) = 2α 2Mω◦ Furthermore, if at an initial time, we start from the density operator (16.146), the average value of the elongation reads

 † −λ Q(t) = (1 − e ) tr{(e−λa a )(a† + a)} 2Mω◦

CH016.tex

17/3/2011

9: 54

Page 503

16.5

LANGEVIN EQUATIONS FOR LADDER OPERATORS

503

〈Q(t)〉

0

0

200

400

600

800

t (fs) Figure 16.2 Time evolution of the average position for the driven damped quantum harmonic oscillator.

which is zero Q(t = 0) = 0 while at infinite time, according to Eq. (16.155), we have  

ω◦2  ◦ Q(t = ∞) = −2α ω◦2 + γ 2 /4 2Mω◦

16.5

LANGEVIN EQUATIONS FOR LADDER OPERATORS

16.5.1 Toward Mori’s equation Consider a harmonic oscillator with Hamiltonian H embedded in a thermal bath of Hamiltonian Hθ to which it is coupled through the interaction Hamiltonian V. The full Hamiltonian is H = H◦ + V + Hθ with H◦ , V, and Hθ given, respectively, by Eqs. (16.9), (16.11), and (16.13), that is,  H◦ = a† a  Hθ = ωj b†j bj (16.156) j

 {κj a† bj + κj∗ ab†j } V=

(16.157)

j

with [a, a† ] = 1

[bj , b†j ] = δij

and

The full Hamiltonian is, therefore,   H =  a† a + {κj a† bj + κj∗ ab†j } + ωj b†j bj j

j

with

(16.158)

k = 0, 1 (16.159)

CH016.tex

504

17/3/2011

9: 54

Page 504

DAMPED OSCILLATORS

Now, consider the time-dependent Heisenberg picture of the Boson operator a(t) of the oscillator of interest embedded in the thermal bath. It is given by the Heisenberg equation (3.94), which reads in the present situation   ∂a(t) i = [a, H] (16.160) ∂t According to Eq. (16.159), the commutator involved in this equation is given by  [a, H] =  [a, aa† ] + κj [a, a† ]bj j

Next, due to Eq. (5.15), and keeping in mind the commutation rule of Boson operators, the right-hand-side commutators read [a, a† a] = a

[a, a† ] = 1

so that Eq. (16.160) yields    ∂a(t) κj bj (t) = −i a(t) − i ∂t

(16.161)

j

Again, in order to obtain the right-hand-side unknowns bj (t) appearing in Eq. (16.161), one has to solve the Heisenberg equation (3.94), which here takes on the form   ∂bj (t) i (16.162) = [bj (t), H] ∂t Now, owing to Eq. (16.159), the commutator involved in Eq. (16.162) is  [bj (t), H] = a {κl∗ [bl (t), b†l ] + ωl [bl (t), b†l bl ]} l

while, due to (16.158), the right-hand-side commutators read  [bj (t), b†j ] = 1 and [bj (t), b†j bj ] = bj (t) j

Hence, Eq. (16.162) transforms to   ∂bj (t) = −i(ωj bj (t) + κj∗ a(t)) ∂t

(16.163)

The solution of Eq. (16.163) without the second member κj∗ a(t), that is,   ∂bj (t) = −i(ωj bj (t)) ∂t is simply b◦ j (t) = b◦ j (0)e−iωj t Hence, the solution of the complete equation (16.163) reads bj (t) = bj (0)e

−iωj t

− iκj∗ e−iωj t

t 0



a(t  )eiωj t dt 

(16.164)

CH016.tex

17/3/2011

9: 54

Page 505

16.5

LANGEVIN EQUATIONS FOR LADDER OPERATORS

505

Furthermore, using this result, Eq. (16.161) takes the form 

∂a(t) ∂t

 = −i a(t) − i



bj (0)κj e

−iωj t

−



j

t



a(t  )e−iωj (t−t ) dt 

|κj |

2

j

0

Now, make the change of variable a˜ (t) = a(t)ei t

(16.165)

Then, denoting bj (0) = bj , the above equation transforms to an equation that has the form of the Mori equation of statistical mechanics of irreversible processes: 

∂a˜ (t) ∂t

 = −i



bj κj e

−i(ωj − )t

−



j

t |κj |

j

2



a˜ (t  )e−i(ωj − )(t−t ) dt 

(16.166)

0

16.5.2 Toward the Langevin equation through the Markov approximation Again, perform on the right-hand-side term of the Mori’s equation (16.166) the Markov approximation according to which there is a loss of memory dealing with a˜ (t  ) allowing to take in place of it a˜ (t) and thereby to move it outside the integral. Then, after taking τ = t − t , t a˜ (t + τ)e

−i(ωj − )τ

∞ dτ  a˜ (t)

0

e−i(ωj − )τ dτ

0

Then, within the Markov approximation, the Mori equation (16.166) simplifies to 

∂a˜ (t) ∂t

 = −i



bj κj e

−i(ωj − )t

− a˜ (t)

j

 j

∞ |κj |

2

e−i(ωj − )τ dτ

0

Now, assuming for the last sum a continuous variation of the thermal bath oscillators in the same way as in the above study dealing with the density operator of the driven damped harmonic oscillator [see Eq. (16.84)], that is,  j

∞ |κj | f (ωj ) → 2

g(ω)|κ(ω)|2 f (ω) dω −∞

where f (ωj ) is a function of ωj and g(ω) is the density of modes, this last expression reads ⎛∞ ⎞   ∞   ∂a˜ (t) = −i bj κj e−i(ωj − )t − a˜ (t) g(ω)|κ(ω)|2 ⎝ e−i(ω− )τ dτ ⎠ dω ∂t j

−∞

0

(16.167)

CH016.tex

506

17/3/2011

9: 54

Page 506

DAMPED OSCILLATORS

The last right-hand-side integral has the same form as that (18.63) of Section 18.6, which transforms to Eq. (18.71): ⎫ ⎧ +∞ ⎛∞ ⎞  ⎬  ∞  ⎨ 1 dω dω f (ω) ⎝ e−i(ω− )τ dτ ⎠ dω = −i f (ω)P ⎭ ⎩ ω− −∞

−∞

0

−

+∞ f (ω)πδ(ω − ) dω

−∞

where P denotes the Cauchy principal part      ∞ ∂ a˜ (t) 1 2 = − a˜ (t) g(ω)|κ(ω)| πδ(ωj − ) + iP dω ∂t ωj − −∞

−i



κj bj e−i(ωj − )t

j

This last equation simplifies to   γ   ∂a˜ (t) = −˜a(t) + i( +  ) − i bj κj e−i(ωj − )t ∂t 2 j

with, respectively, ∞  γ = 2πg(ω)|κ(ω)|

2

and

 = −∞

g(ω)|κ(ω)|2 (ω − )

Now, one may incorporate into the Lamb shift  via ˜ = + 

so that the above differential equation takes the simplified form   γ   ∂a˜ (t) ˜ ˜ −i bj κj e−i(ωj − )t = −˜a(t) + i ∂t 2

 dω

(16.168)

(16.169)

(16.170)

j

Then, returning to the initial IP Boson operator by the aid of Eq. (16.165), Eq. (16.170) becomes   γ   ∂a(t) ˜ = −i + a(t) +i bj κj e−iωj t (16.171) ∂t 2 j

This linear first-order differential equation, which is inhomogeneous since involving a second member, may be integrated in the following way by observing that the corresponding homogeneous equation (without second member) reads  ◦  γ  ∂a (t) ˜ =0 + a◦ (t) +i (16.172) ∂t 2 which yields by integration ˜

a◦ (t) = a(0)e−(γ/2)t e−i t

CH016.tex

17/3/2011

9: 54

Page 507

16.5

LANGEVIN EQUATIONS FOR LADDER OPERATORS

507

Then, it is suitable to search for the solution of the complete inhomogeneous differential equation (16.171) an expression of the same form as that in (16.172) but in which the time-dependent function u(t) takes the place of a(0) ˜

a(t) = u(t)e−(γ/2)t e−i t

(16.173)

To find u(t) write the partial time derivative of Eq. (16.173)     γ  ∂u(t) −(γ/2)t −i t ∂a(t) ˜ ˜ ˜ e−(γ/2)t e−i t = e − u(t) e + i ∂t ∂t 2 so that it reads     γ  ∂u(t) (−γ/2)t −i t ∂a(t) ˜ ˜ = e + a(t) +i e ∂t 2 ∂t γ  ˜ ˜ ˜ e(−γ/2)t e−i t − u(t) + i + u(t)e(−γ/2)t e−i t 2 and thus, after simplification,   γ   ∂u(t)  ∂a(t) ˜ ˜ = + a(t) +i e(−γ/2)t e−i t ∂t 2 ∂t so that, by identification with Eq. (16.171), one obtains    ∂u(t) −(γ/2)t −i t ˜ e = −i bj κj e−iωj t e ∂t j

or



∂u(t) ∂t

 = −i



˜

bj κj e−iωj t e(γ/2)t ei t

j

The integration of this last equation reads  t   ˜ ˜  bj κj e−(γ/2+i )t e(γ/2+i )t e−iωj t dt  u(t) = −i o

j

so that Eq. (16.173) becomes a(t) = a(0)e

˜ (−γ/2)t −i t

e

−i



bj κj e

˜ −(γ/2+i )t



t

˜





e(γ/2+i )t e−iωj t dt 

(16.174)

o

j

or, after using Eq. (16.169), a(t) = a(0)(t) +



bj ϕj (t)

(16.175)

j

with, respectively, (t) = e−(γ/2)t e−i( + )t t ϕj (t) = −iκj o

˜



(16.176) 

e−(γ/2+i )(t−t ) e−iωj t dt 

(16.177)

CH016.tex

508

17/3/2011

9: 54

Page 508

DAMPED OSCILLATORS

Now, consider the average value of (16.175) over the Boltzmann density operator ρθ of the thermal bath a(t) = trθ {ρθ a(t)} with, as above, for Eqs. (16.14)–(16.16) † (1 − e−λk )(e−λk bk bk ) ρθ =

with

(16.178) ωk kB T

λk =

k

(16.179)

Owing to (16.179), the time-dependent average value (16.178) takes the form ⎧ ⎛ ⎞⎫ ⎨ ⎬  † a(t) = trθ (1 − e−λk )(e−λk bk bk ) ⎝a(0)(t) + (16.180) ϕj (t)bj ⎠ ⎩ ⎭ j

k

Again, after writing explicitly the trace over the thermal bath using Eq. (16.50) and because each kind of Boson operator acts in its specific space, Eq. (16.180) becomes ⎞ ⎛   † a(t) = (1 − e−λk ) (nk )|(e−λk bk bk ) ⎝a(0)(t) + ϕj (t)bj ⎠ |(nk ) nk

k

j

Moreover, using in turn this specific action of the Boson operators inside their own space, it transforms to  † a(t) = (1 − e−λk ) (nk )|(e−λk bk bk )|(nk )a(0)(t) k

+



ϕj (t)



×

(1 − e−λk )

k=j

†

(1 − e−λj )(nj )|e−λj bj bj bj |(nj )

nj

j



nk

 † (nk )|(e−λk bk bk )|(nk )

(16.181)

nk

Next, due to Eq. (16.52), that is, (1 − e−λj )

 † (nj )|(e−λj bj bj )|(nj ) = 1 nj

and, owing to Eq. (16.65), that is  † (1 − e−λj ) (nj )|{(e−λj bj bj )bj }|(nj ) = 0 nj

Eq. (16.181) simplifies to a(t) = a(0)(t)

(16.182)

a(t) = a(0)e(−γ/2)t e−i( + )t

(16.183)

or, in view of Eq. (16.176),

the Hermitian conjugate of which is a† (t) = a† (0)e(−γ/2)t ei( + )t

(16.184)

CH016.tex

17/3/2011

9: 54

16.6

Page 509

EVOLUTION OPERATORS OF DRIVEN DAMPED OSCILLATORS

509

Equations (16.183) and (16.184), which give the time dependence of the mean values of the Boson operators of the damped oscillator averaged over the thermal bath, are the quantum equivalents of the Langevin equations of classical statistical mechanics dealing with the irreversible behavior of harmonic oscillators. Keep in mind that the partial averaging over the thermal bath, which leads to operators, is denoted by an upper line, allowing one to distinguish it from the complete averaging denoted  , which leads to scalars. Moreover, it is impossible to admit of equations similar to Eqs. (16.183) and (16.184) but that deal with ladder operators in place of the average values, that is, a(t) = a(0)e(−γ/2)t e−i( + )t

a† (t) = a† (0)e(−γ/2)t ei( + )t (16.185) This impossibility follows because if the equations (16.185) were true, they would imply the following commutator: and

[a(t), a† (t)] = [a(0), a† (0)]e−γt which is false since, at infinite time, it would lead erroneously to a zero value and therefore to commutativity of a and a† and thus to commutativity of the position and momentum operators from which a and a† are constructed using Eqs. (5.3) and (5.4).

16.6 EVOLUTION OPERATORS OF DRIVEN DAMPED OSCILLATORS To end the present chapter devoted to the irreversible behavior of quantum oscillators, it may be of interest to find the IP time evolution operator of a driven damped quantum harmonic oscillator using Eqs. (16.183) and (16.184). This interaction picture operator plays a role in the quantum theory of the IR line shapes of weak H-bonded species. Besides, it will be seen that this time evolution operator allows one to get the expression (16.146) of the density operator of the driven damped harmonic oscillator obtained by Louisell and Walker, which has been given without providing their very complicated proof.

16.6.1 Time derivative equation to be solved within normal ordering formalism For this purpose, consider as above the Hamiltonian H of a driven harmonic oscillator of Hamiltonian H embedded within a thermal bath defined by the Hamiltonian Hθ to which it is coupled through the interacting Hamiltonian V. It is given by H = H + V + Hθ with

 H=

 P2 1 + M 2 Q2 + bQ 2M 2

(16.186)

and where V and Hθ are, respectively, given by Eqs. (16.4) and (16.5). In the area of the theory of the IR line shape of H-bonded species, the following expression has to

CH016.tex

510

17/3/2011

9: 54

Page 510

DAMPED OSCILLATORS

be found: ˆ P{exp[−ib



t

Q(t  )dt  ]}

(16.187)

0

Here, Pˆ is the Dyson time-ordering operator, Q(t) is the time-dependent coordinate in the Heisenberg picture governed by Eq. (3.94), that is,   ∂Q(t) 1 = [Q(t), H] (16.188) ∂t i and Q(t) is the Langevin averaged value of Q(t) when damped by the medium, which may be formally given by Q(t) = trθ {ρθ Q(t)} In this last expression, ρθ is, as above, the Boltzmann density operator of the thermal bath defined by (16.179). Of course, the time-dependent coordinate Q(t) is given by integration of the basic commutator equation (16.188), Note that owing to Eqs. (3.86) and (3.87), Eq. (16.187) is the formal solution     t   ˆ U(t) = P exp −ib (16.189) Q(t )dt 0

of the dynamical equation

 i

∂U(t) ∂t

 = bQ(t) U(t)

(16.190)

with U(0) = 1

(16.191)

In order to find the explicit expression of the evolution operator given by the formal Eq. (16.189), we pass to lowering and raising operators a and a† by the aid of Eqs. (5.6) and (5.7). Then, the coefficient b of Hamiltonian (16.186) transforms to the dimensionless coefficient α◦ :

b  ◦ α =

2M Then, within the Boson operator representation, Eq. (16.189) becomes     t † ◦  †   ˆ U(t, a , a) = P exp −iα (16.192) a (t ) + a(t )dt 0

in which a(t) may be obtained using the Heisenberg equation (3.94) reading, in the present situation,   ∂a(t) 1 = [a(t), H] (16.193) ∂t i whereas a† (t) is given by the corresponding Hermitian conjugate. Next, the Langevin averaged values of the ladder operators have to be performed over the Boltzmann density operator ρθ of the thermal bath using the partial trace a† (t) + a(t) = trθ {ρθ (a† (t) + a(t))}

(16.194)

CH016.tex

17/3/2011

9: 54

16.6

Page 511

EVOLUTION OPERATORS OF DRIVEN DAMPED OSCILLATORS

Equation (16.192) is the formal solution of the Schrödinger equation:   ∂U(t, a† , a) i = α◦ {a† (t) + a(t)}U(t, a† , a) ∂t

511

(16.195)

involving, of course, the same boundary condition as (16.191). Now, observe that the average values involved on the right-hand side of Eq. (16.195) and obeying the Heisenberg equation (16.193) are given by the Langevin equations (16.183) and (16.184), which, after incorporating the Lamb shift into the angular frequency, reads a(t) = a(0) e(−γ/2)t e−i( t

and

a† (t) = a† (0)e(−γ/2)t ei( t

As a consequence, the Schrödinger equation (16.195) becomes   ∂U(t, a† , a) = α◦ (a† (t)∗ + a(t))U(t, a† , a) i ∂t

(16.196)

(16.197)

where (t) = e(−γ/2)t e−i t

(16.198)

and with the boundary condition (16.191).

16.6.2

Solution of Eq. (16.197)

To solve this differential equation, it is suitable to use the normal ordering technique, comparing Eqs. (7.20)–(7.122), according to which one has, respectively,   ∂U(t, a† , a) ∂U {n} (t, α∗ , α) −1  N = ∂t ∂t  N−1 {a† U(t, a† , a)} = α∗ {U {n} (t, α∗ , α)}   ∂ −1 †  N {a U(t, a , a)} = α + ∗ {U {n} (t, α∗ , α)} ∂α

(16.199)

where  N is the normal ordering operator and  N−1 its inverse. Then, Eq. (16.197) involving noncommutating operators transforms to the following partial differential equation:  {n}    ∂U (t, α∗ , α) ∂ = −iα◦ α + ∗ (t) + α∗ (t)∗ {U {n} (t, α∗ , α)} ∂t ∂α (16.200) with, due to Eq. (16.191), the boundary condition {U {n} (0, α∗ , α)} = 1

(16.201)

Now, make the change of variable {U {n} (t, α∗ , α)} = (eG(t) )

(16.202)

with, due to Eq. (16.201), the boundary condition G(0) = 0

(16.203)

CH016.tex

512

17/3/2011

9: 54

Page 512

DAMPED OSCILLATORS

and assume for G(t) an expression of the form G(t) = A0 (t) + A1 (t)α + A2 (t)α∗

(16.204)

where A0 (t), A1 (t), and A2 (t) are unknown time-dependent coefficients that, owing to Eqs. (16.202) and (16.203), must satisfy the following boundary conditions: A0 (0) = A1 (0) = A2 (0) = 0

(16.205)

so that Eq. (16.202) becomes ∗

{U {n} (t, α∗ , α)} = (eA0 (t)+A1 (t)α+A2 (t)α )

(16.206)

Then, the partial time derivative of expression (16.202) reads  {n}      ∂U (t, α∗ , α) ∂G(t) ∂G(t) G(t) = (e ) = {U {n} (t, α∗ , α)} ∂t ∂t ∂t while, due to Eq. (16.202), that of Eq. (16.206) yields  {n}        ∂U (t, α∗ , α) ∂A0 (t) ∂A2 (t) ∂A1 (t) = +α + α∗ {U {n} (t, α∗ , α)} ∂t ∂t ∂t ∂t (16.207) Then, by identification of Eqs. (16.200) and (16.207), one obtains, respectively,   ∂A0 (t) (16.208) = −iα◦ (t)A2 (t) ∂t     ∂A2 (t) ∂A1 (t) ◦ = −iα (t) = −iα◦ (t)∗ (16.209) ∂t ∂t so that the time-dependent coefficients involved in Eq. (16.204) appear to be interrelated via the following equations through A1 (t) = −A∗2 (t) ≡ A(t)

(16.210)

16.6.3 Time evolution operator As a consequence, owing to Eqs. (16.204) and (16.210), Eq. (16.202) takes the form {U {n} (t, α∗ , α)} = eA0 (t) (eA(t)α−A

∗ (t)α∗

)

(16.211)

In order to obtain the expressions of A0 (t) and A(t) one has to solve the two coupled equations (16.208) and (16.209), with (t) given by Eq. (16.198), that is,   ∂A0 (t) (16.212) = iα◦ e(−γ/2)t e−i t A(t)∗ ∂t   ∂A(t) (16.213) = −iα◦ e−γt/2 e−i t ∂t subject to the boundary conditions (16.205), that is, A(0) = A0 (0) = 0

CH016.tex

17/3/2011

9: 54

16.6

Page 513

EVOLUTION OPERATORS OF DRIVEN DAMPED OSCILLATORS

513

After integration of Eq. (16.213), and owing to its specific boundary conditions, A(t) is given by A(t) = β◦ (e−γt/2 e−i t − 1) with





+ iγ/2 β =α

2 + (γ/2)2 ◦

◦

(16.214)

(16.215)

Again, inserting this result into Eq. (16.212) gives ◦

A0 (t) = i(α )

 2

+ iγ/2

2 + (γ/2)2

 t









(e−γt /2 e−i t )(e−γt /2 ei t − 1)dt 

0

or, with the ad hoc boundary condition A(0) = 0 ⎛ ⎞   t t

− i(γ/2)   ⎝ e−γt dt  − e−i( −iγ/2)t dt  ⎠ A0 (t) = i(α◦ )2

2 + (γ/2)2 0

(16.216)

0

and thus, after integration, we have   i −γt 1 −γt 1 ◦ 2 (−γ/2)t −γt/2 − 1) − ie sin t − e +e cos t − A0 (t) = |β | − (e γ 2 2 (16.217) with   (α◦ )2 |β◦ |2 = (16.218)

2 + (γ/2)2 Next, observe that, by action of the normal ordering operator  N, on Eq. (16.211), one obtains through Eq. (7.44), the time evolution operator  N{U {n} (t, α∗ , α)} = {U(t, a† , a)} with, due to Eq. (16.211), {U(t, a† , a)} = (eA0 (t) )(e−A(t)

∗ a†

)(eA(t)a )

(16.219)

where A0 (t) and A(t) are, respectively, given by Eqs. (16.217) and (16.214). Then, using the Glauber–Weyl theorem (1.78), we can transform the product of exponential operators appearing in Eq. (16.219) into (e−A(t)

∗ a†

)(eA(t)a ) = {eA(t)a−A(t)

∗ a†

}[e−|A(t)|

2 [a† ,a]/2

]

so that Eq. (16.219) transforms to {U(t, a† , a)} = eξ(t) {eA(t)a−A(t)

∗ a†

} = eξ(t) {e−A(t)

∗ a† +A(t)a

}

(16.220)

with eξ(t) = (eA0 (t) )[e−|A(t)|

2 [a† ,a]/2

] = (eA0 (t)+|A(t)|

2 /2

)

(16.221)

CH016.tex

514

17/3/2011

9: 54

Page 514

DAMPED OSCILLATORS

or, due to Eqs. (16.214) and (16.216),   1 −γt (e − 1) + (e−γt/2 ) sin t ξ(t) = i|β◦ |2 γ

(16.222)

Hence, due to Eq. (16.189), it appears that Eq. (16.220) may be written in the form {U(t, a†, a)} 

= Pˆ exp −ib



t

Q(t  ) dt 

 = eξ(t) {eA(t)a−A(t)

∗ a†

} = eξ(t) {e−A(t)

∗ a† +A(t)a

}

0

(16.223) where the last equality has been written mindful of the fact that in the exponential the position of a† and a is irrelevant. Observe that, due to Eq. (16.214), A(t) = β◦ (e−γt/2 e−i t − 1) the evolution operator reads {U(t, a† , a)} = eξ(t) exp{β◦ (e−γt/2 e−i t − 1)a−β◦ (e−γt/2 ei t − 1)a† }

(16.224)

or, in view of Eqs. (16.148) and (16.150), {U(t, a† , a)} = eξ(t) {A( − φ0 (t))} = eξ(t) {A(φ0 (t))}−1

(16.225)

where {A(φ0 (t))} is the damped translation operator (16.147) extracted from the Louisell and Walker density operator (16.150) through the canonical transformation (16.149). Of course, U(t, a† , a) and A(φ0 (t)) being unitary, the inverse of U(t, a† , a) is {U(t, a† , a)}−1 = e−ξ(t) {A(φ0 (t))}

16.6.4

(16.226)

Calculation of density operator

Now, make the following canonical transformation over the Boltzmann density operator by aid of the time evolution operator expressed by Eq. (16.220) leading to the density operator ρ(t) according to ρ(t) = (1 − e−λ ){U(t, a† , a)}(e−λa a ){U(t, a† , a)}−1 †

(16.227)

where λ is the T temperature function  kB T Equation (16.227) transforms with the help of Eqs. (16.225) and (16.226) into λ=

ρ(t) = (1 − e−λ )e−ξ(t) {A(φ0 (t))}−1 (e−λa a ){A(φ0 (t))}eξ(t) †

or, on simplification, ρ(t) = (1 − e−λ ){A(φ0 (t))}−1 (e−λa a ){A(φ0 (t))} †

Hence, owing to Eq. (16.147), this operator becomes ρ(t) = (1 − e−λ )(e−λ(a

† +φ

∗ 0 (t))(a+φ0 (t))

)

CH016.tex

17/3/2011

9: 54

Page 515

16.7

CONCLUSION

515

which has the same structure as the density operator (16.151) of driven damped harmonic oscillators as obtained by Louisell and Walker. Observe that due to Eq. (16.223)

{U(t, a† , a)}

    t   ˆ = P exp −ib Q(t )dt 0

the expression (16.227) for this driven damped density operator, reads in terms of P and Q,     2      t P 1 ρ(t) = (1 − e−λ )Pˆ exp −ib exp − kB T Q(t  )dt  + M 2 Q2 2M 2 0    t  × Pˆ exp ib Q(t  )dt  0

where Q(t) is the average Langevin value of the damped coordinate Q(t), that is,

Q(t) =

16.7

 {a† (0)ei t + a(0)e−i t }e−γt/2 2M

CONCLUSION

This chapter, devoted to the irreversible behavior of damped oscillators, took into account the influence of the neighborhood by considering it as a thermal bath composed of a very large set of harmonic oscillators linearly coupled to an oscillator embedded in this bath. First, consider second order in the coupling with the bath. An iterative process for the integral equation governing the dynamics of the density operator was set up, yielding a master equation that is an operator equation giving the time derivative of the density operator of the damped oscillator in terms of a complicated combination of products of the ladder operators of this oscillator. We thus derived from this master equation, the corresponding Fokker–Planck equation, which is a time derivative of the scalar corresponding to the antinormal form of the damped oscillator density operator. Such an equation involving only scalars can be integrated (which was not done) in a similar way as the Franck–Condon equation encountered in the classical statistical mechanics of irreversible processes. Moreover, the Langevin equations governing the average values of the damped ladder operators have been found. Using these Langevin equations, we obtained the IP time evolution operator of a damped harmonic oscillator. Finally, using this time evolution operator, we could find for the time-dependent density operator of damped harmonic oscillators, a closed expression equivalent to that obtained by Louisell and Walker, and

CH016.tex

516

17/3/2011

9: 54

Page 516

DAMPED OSCILLATORS

which may be viewed as the solution of the master equation. These important results are summarised as follows: Equations govering the dynamics of driven damped harmonic oscillators Master equation:   ∂ρDr (t)SP = −iα◦ ω◦ {[a, ρDr (t)SP ] + [a† , ρDr (t)SP ]} ∂t − iω◦ [a† a, ρDr (t)SP ] − i ω[a† a, ρDr (t)SP ] γ − (a† aρDr (t)SP + ρDr (t)SP a† a − 2aρDr (t)SP a† ) 2 − nγ(a† aρDr (t)SP + ρDr (t)SP aa† − a† ρDr (t)SP a−aρDr (t)SP a† ) Langevin equations: a(t) = a(0)e(−γ/2)t e−i( + )t

and

a† (t) = a† (0)e(−γ/2)t ei( + )t

Fokker–Planck equation:  a  ! a   a " ∂ρ (t) − ∂ρ∂α(t) = −iα◦ ω◦ ∂ρ∂α(t) ∗ ∂t   a   a  ∂ρ (t) ∂ρ (t) − i(ω◦ + ω) α∗ − α ∂α∗ ∂α  a   a   2 a  ∗ γ ∂ρ (t)α ∂ρ (t)α ∂ ρ (t) + + + nγ ∗ 2 ∂α ∂α∂α∗ ∂α Density operator: ρ(t) = (1 − e−λ )(e−λ(a φ0 (t) = β{e−γt/2 e+i(ω

† +φ

∗ 0 (t))(a+φ0 (t))

◦ +ω)t

)

with

λ=

ω◦ kB T

and

− 1}

BIBLIOGRAPHY C. Cohen-Tannoudji, J. Dupont-Roc, and G. Grinberg. Atom Photon Interactions. Basic Processes and Applications. Wiley: New York, 1992. H. Louisell. Quantum Statistical Properties of Radiations. Wiley: New York, 1973. W. Louisell and L. Walker. Phys. Rev., 137 (1965): 204.

CH017.tex

PART

17/3/2011

9: 57

Page 517

VII

VIBRATIONAL SPECTROSCOPY

CH017.tex

17/3/2011

9: 57

Page 519

17

CHAPTER

APPLICATIONS TO OSCILLATOR SPECTROSCOPY INTRODUCTION All the properties of quantum oscillators obtained in Part III (Chapter 10), Part IV (Chapter 13), and Part VI (Chapter 16) are used in this last part (Chapter 17) to find some important results in vibrational spectroscopy, such as the IR selection rule for quantum harmonic oscillators, Fermi resonances, the tunneling effect through doublewell potentials, and to study, in the linear response theory, the line shape of some physically realistic situations involving anharmonically coupled damped quantum harmonic oscillators met in H bonding such as Fermi resonances or Davydov coupling.

17.1 IR SELECTION RULES FOR MOLECULAR OSCILLATORS 17.1.1

Induced absorption and emission

In the last section of Chapter 4, we studied the Fermi golden rule, giving the probability of transition between two energy levels due to an energetic perturbation. When a molecule has a dipole moment, this dipole operator may interact with an electromagnetic field to induce transitions between two of its energy levels, characterizing it in the absence of a field. In spectroscopy, the electromagnetic field obeys the Maxwell equations. It was found in Chapter 15 that when the electric field is described by a coherent state, its quantum behavior is very nearly that of a classical one where the different modes of the field oscillate at angular frequency ω. Then, the perturbation generated by the interaction between the dipole of the molecule and the electric field is time dependent. Hence, in order to find the transition probability between the energy levels induced by the electromagnetic field, we have to reconsider the analysis of the Fermi golden rule by introducing in it the time dependence of the perturbation. That is the object of the present section. Consider a molecular system having a dipole moment operator  μ and described by a Hamiltonian H◦ that interacts with an electromagnetic field E(ω, t) of angular frequency ω through V(t) =  μE(ω, t)

with E(ω, t) = E(ω)(e−iωt + eiωt )

(17.1)

Quantum Oscillators, First Edition. Olivier Henri-Rousseau and Paul Blaise. © 2011 John Wiley & Sons, Inc. Published 2011 by John Wiley & Sons, Inc.

519

CH017.tex

17/3/2011

520

9: 57

Page 520

APPLICATIONS TO OSCILLATOR SPECTROSCOPY

where E(ω) would be given by Eqs. (14.139) and (14.141). The full Hamiltonian H is the sum of the Hamiltonian H◦ of the molecular system and of theHamiltonian V(t) coupling the molecular system to the electromagnetic field E(ω, t) via the dipole moment operator  μ: H = H◦ + V(t) Now, write the eigenvalue equation of H◦ : H◦ |k  = Ek |k 

(17.2)

k |l  = δkl

(17.3)

with

and seek the transition probability at time t for the system described by H to pass from any eigenstate of H◦ to another because of the presence of V(t), that is, |C(l, t|k, 0)|2 = |k (0)|l (t)|2 In the interaction picture with respect to H◦ and according to Eq. (4.93), the transition probability yields  2   2  t   1  2 iH◦ t  / −iH◦ t  /  k (0)|(e ) V(t) (e )|l (0)dt  |C(l, t|k, 0)| =     0

so that due to Eqs. (17.1) and (17.2)  2   2  t   E(ω)     2 iEk t / −iωt iωt −iEl t /   k (0)|(e ) μ(e + e ) (e )|l (0)dt  |C(l, t|k, 0)| =     0

and, after rearranging  |C(l, t|k, 0)|2 =

E(ω) 

2

 t 2       2 i(ωkl −ω)t i(ωkl +ω)t  |k |μ|l |  (e +e ) dt   

(17.4)

0

with (Ek − El )  Next, the integration of the time-dependent term of Eq. (17.4) reads   t   1 ei(ωkl −ω)t − 1 ei(ωkl +ω)t − 1 i(ωkl −ω)t  i(ωkl +ω)t   (e +e ) dt = + i ωkl − ω ωkl + ω ωkl =

(17.5)

0

Moreover, observe that if ω  ωkl , the first term of the right-hand side of Eq. (17.5) becomes very large with respect to the second one, so that the latter may be neglected, that is,   t  1 ei(ωkl −ω)t − 1 i(ωkl −ω)t  i(ωkl +ω)t   (e +e ) dt  i ωkl − ω 0

CH017.tex

17/3/2011

9: 57

Page 521

17.1

IR SELECTION RULES FOR MOLECULAR OSCILLATORS

Hence, due to this approximation, Eq. (17.4) reads      i(ωkl −ω)t  − 1 2 E(ω) 2  2 2 e |C(l, t|k, 0)| = |k | μ|l |    ωkl − ω  

521

(17.6)

Furthermore, passing to the sine function leads for the time-dependent term of (17.6)    ei(ωkl −ω)t  − 1 2 2   {1 − cos ((ωkl − ω)t)}  =   ωkl − ω  (ωkl − ω)2 so that Eq. (17.6) transforms to



μ|l | |C(l, t|k, 0)| = 4(E(ω)) |k | 2

2

2

sin2 ((ωkl − ω)t/2) 2 (ωkl − ω)2



Then, by a procedure similar to that used in passing from Eq. (4.97) to Eq. (4.102), we have 2π μ|l |2 {δ(Ek − El − ω)}t (17.7) (E(ω))2 |k | |C(l, t|k, 0)|2 =  Hence, the transition probability by unit time defined by   ∂|C(l, t|k, 0)|2 W (l, t|k, 0) = ∂t is W (l, t|k, 0) =

2π (E(ω))2 |k | μ|l |2 δ(Ek − El − ω) 

(17.8)

This result shows that the transition from the eigenstate |l  of the molecular system Hamiltonian H◦ , to another one |k , due to the presence of an electrical field, may occur only if the two following conditions are simultaneously verified: ω=

|Ek − El | 

k | μ|l   = 0

(17.9)

that is, if the angular frequency ω of the radiation satisfies the Bohr condition dealing with the energy levels |k  and |k , and if the corresponding off-diagonal matrix elements of the dipole moment operator are different from zero. That may be summarized in W (l, t|k, 0)  = 0

17.1.2

and

k | μ|l   = 0

Selection rules for harmonic oscillators

The dipole moment operator  μ, which depends on the coordinate Q, may be expanded, up to first order with respect to the equilibrium position to give   ∂ μ  Q μ= μ(Q) = μ(0) + ∂Q Q=0

CH017.tex

522

17/3/2011

9: 57

Page 522

APPLICATIONS TO OSCILLATOR SPECTROSCOPY

where μ(0) is the dipole moment operator at Q = 0, which is a scalar from the viewpoint of the quantum oscillator theory. Then, due to the orthogonality relation (17.3) between the kth and the lth eigenstates leads one to write k |μ(0)|l  = μ(0)k |l  = 0 the off-diagonal matrix elements of the dipole moment operator appearing in Eq. (17.8) and corresponding necessarily to k  = l, according to (17.9), read   ∂ μ k | μ|l  = k |Q|l  (17.10) ∂Q Q=0 Next, when dealing with the vibrations of molecules, the partial derivative in front of the right-hand side of Eq. (17.10), which depends on the electronic part of the molecular system, has to be averaged over the electronic states |El , and thus may be viewed as the scalar. Then, Eq. (17.10) reads   ∂ μ k | μ|l  = El | |El k |Q|l  (17.11) ∂Q Q=0 so that owing to Eq. (17.11), Eq. (17.8) yields 2     2π  ∂ μ  W (l, t|k, 0) = |El  (E(ω))2 |k |Q|l |2 δ(Ek − El − ω) El |    ∂Q Q=0 (17.12) or W (l, t|k, 0) =

2π ◦ 2 (μ ) (E(ω))2 |k |Q|l |2 δ(Ek − El − ω) 

with



∂ μ μ ⇔ El | ∂Q ◦

(17.13)

 |El  Q=0

Apply now the result (17.13) to the special case of harmonic oscillators, the Hamiltonian of which is

H = ω◦ a† a + 21 Then, |k  and |l  are the eigenkets of this Hamiltonian, that is, |k  = |{k} while the energy levels Ek and El are the corresponding eigenvalues verifying H|{k} = Ek |{k} with



Ek = ω◦ k + 21

(17.14)

{k}|{l} = δkl

(17.15)

and

CH017.tex

17/3/2011

9: 57

Page 523

17.1

IR SELECTION RULES FOR MOLECULAR OSCILLATORS

523

where ω◦ is the angular frequency of the harmonic oscillator. Then, the off-diagonal matrix elements involved in Eq. (17.13) read k |Q|l  = {k}|Q|{l}

(17.16)

Moreover, passing to Boson operators using Eq. (5.6), the off-diagonal matrix elements yield   k |Q|l  = {k}|(a† + a)|{l} (17.17) 2Mω◦ Now, due to Eqs. (5.53) and (5.63), that is, √ a† |{k} = k + 1|{k + 1} and

a|{k} =

√

k|{k − 1}

(17.18)

the right-hand side of Eq. (17.17) becomes √ √ k |(a† + a)|l  = {k}|( l + 1|{l + 1} + l|{l − 1}) or, in view of the orthonormality properties of the eigenkets of the harmonic Hamiltonian, √ √ k |(a† + a)|l  = ( l + 1δk,l+1 + lδk,l−1 ) (17.19) so that, after using Eqs. (17.16) and (17.19), Eq. (17.13) reads  2     ∂ μ   W (l, t|k, 0) ∝ El | |El  {δk,l±1 }{δ((k − l)ω◦ − ω)}   ∂Q Q=0 Here it appears that a transition induced by the electromagnetic field occurs if the three following conditions are verified: 1.

If the energy ω of the photon is equal to the difference in energy between Ek and El , that is, satisfies the Bohr condition ω = (k − l)ω◦

2.

If the change in the quantum numbers characterizing the energy levels of the harmonic oscillators obey k =l±1

3.

(17.20)

(17.21)

If the electronic matrix elements over the electronic state |El  of the partial derivative of the dipole moment of the diatomic molecule with respect to Q are different from zero   ∂ μ El | |El   = 0 (17.22) ∂Q Q=0

Note that homonuclear diatomic molecules cannot have, because of symmetry, any dipole moment so that the partial derivative with respect to Q involved in Eq. (17.22)

CH017.tex

524

17/3/2011

9: 57

Page 524

APPLICATIONS TO OSCILLATOR SPECTROSCOPY

must be zero, leading therefore to the possibility of this kind of molecule absorbing or emitting IR radiations in the framework of the mechanism studied here. Of course, the combined conditions (17.20) and (17.21) lead to the resonance condition ω = ω◦

(17.23)

Recall that the angular vibrational frequency of diatomic molecules is roughly given by Eq. (9.31), which is in the range of IR electromagnetic radiations, so that Eq. (17.23) implies that IR radiations may be absorbed due to changes in the vibrational energy levels. Now, since Eq. (17.22) must be verified, the dipole moment operator  μ cannot disappear, which is impossible for homonuclear diatomic molecules, which, being symmetric, cannot have a dipole moment operator. Hence, only heteronuclear diatomic molecules may change in their vibrational energy levels due to interaction with IR radiations. Owing to the selection rule (17.23), it is clear that at zero temperature since the Boltzmann population of the harmonic oscillator energy levels is zero except for the ground state, the only absorption transition is that corresponding to a jump from the ground state |{0} on the first excited state |{1}. Of course, the corresponding emission results from the jump from the excited state |{1} to the ground state |{0}. Figure 17.1 illustrates this with the concomitant transitions of the electromagnetic mode compatible with the energy conservation known as the Bohr condition. On the other hand, at any finite temperature the different energy levels have some probability to be occupied, which is given by the Boltzmann distribution. Hence, different transitions corresponding to the following changes will take place: |{0} → |{1}

|{1} → |{2}

|{2} → |{3}

and so on

However, since the angular frequency of these different transitions is the same in the harmonic approximation, it follows that at a temperature, the spectrum must form a single line of angular frequency ω.

|{1}〉

|(n)〉

ω

|{0}〉 Oscillator mode

|{1}〉

ω

|{0}〉 Electromagnetic mode

Absorption

Oscillator mode

|(n 1)〉 Electromagnetic mode

Emission

Figure 17.1 Absorption or emission by a quantum harmonic oscillator mode resulting from a resonant coupling with an electromagnetic mode of the same angular frequency ω◦.

CH017.tex

17/3/2011

9: 57

Page 525

17.1

IR SELECTION RULES FOR MOLECULAR OSCILLATORS

525

17.1.3 “Forbidden” transitions of weak intensity and anharmonicity The IR transition selection rules [(17.20)–(17.23)], are very simple because they are dealing with harmonic oscillators. However, realistic molecular oscillators are generally anharmonic, their potentials being Morse-like. Hence, we shall examine what changes in the selection rules will be induced by the anharmonicity of the Morse-like potentials. When the eigenvalue equation of the Hamiltonian H of Morse oscillators has been solved, it has been found that, in the basis of eigenkets of the harmonic Hamiltonian, the expansion of the lowest eigenstates of the Hamiltonian are given by (9.88), according to which the three lowest eigenstates of the Morse Hamiltonian may be crudely given by |0   1 − ε2 |{0} + ε|{1} (17.24) |I  = ξ|{0} + 1 − ξ 2 − η2 |{1} + η|{2} (17.25) (17.26) |II  = ζ|{1} + 1 − ζ 2 |{2} with ε