Quantum Mechanics
THIRD EDITION
EUGEN MERZBACHER
University of North Carolina at Chapel Hill
JOHN WILEY & SONS, INC.
New York / Chichester / Weinheim Brisbane / Singapore / Toronto
Acquisitions Editor Marketing Manager Production Editor Designer Illustration Editor
Stuart Johnson Kimberly Manzi Sandra Russell Madelyn Lesure Edward Starr
This book was set in 10/12 Times by University Graphics. Inc. and printed and bo Hamilton Printing Company. The cover was printed by Hamilton Printing Company. This book is printed on acidfree paper.
@
Copyright © 1961, 1970, 1988, 1998 John Wiley & Sons, Inc. All rights reserved. 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 Sections 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 percopy fee to the Copyright Clearance Center, 222 Rosewood Drive, Danvers, MA 01923, (508) 7508400, fax (508) 7504470. Requests to the Publisher for permission should be addressed to the Permissions Department, John Wiley & Sons, Inc., 605 Third Avenue, New York, NY 101580012, (212) 8506011, fax (212) 8506008, EMail: PERMREQ@WILEY.COM.
Library of Congress Cataloging in Publication Data: Merzbacher, Eugen. Quantum mechanics I Eugen Merzbacher.  3rd ed. p. em. Includes bibliographical references and index. ISBN 0471887021 (cloth: alk. paper) 1. Quantum theory. I. Title. QCI74.12.M47 1998 530. I 2dc21 9720756 CIP
Printed in the United States of America 1098765432
530.12 M577 Quantum mechanics 3rd ed. Swanston LCN: 10798500 poi: 200109006/0001 B/C: 31259006333202
Dedicated to our children: Celia, Charles, Matthew, and Mary
Preface
The central role of quantum mechanics, as a unifying principle in contemporary physics, is reflected in the training of physicists who take a common course, whether they expect to specialize in atomic, molecular, nuclear, or particle physics, solid state physics, qmtntum optics, quantum electronics, or quantum chemistry. This book was written for such a course as a comprehensive introduction to the principles of quantum mechanics and to their application in the subfields of physics. The first edition of this book was published in 1961, the second in 1970. At that time there were few graduatelevel texts available to choose from. Now there are many, but I was encouraged by colleagues and students to embark on a further revision of this bo<?k. While this new updated edition differs substantially from its predecessors, the underlying purpose has remained the same: To provide a carefully structured and coherent exposition of quantum mechanics; to illuminate the essential features of the theory without cutting corners, and yet without letting technical details obscure the main storyline; and to exhibit wherever possible the common threads by which the theory links many different phenomena and subfields. The reader of this book is assumed to know the basic facts of atomic and subatomic physics and to have been exposed to elementary quantum mechanics at the undergraduate level. Knowledge of classical mechanics and some familiarity with electromagnetic theory are also presupposed. My intention was to present a selfcontained narrative, limiting the selection of topics to those that could be treated equitably without relying on specialized background knowledge. The material in this book is appropriate for three semesters (or four quarters). The first 19 chapters can make up a standard twosemester (or threequarter) course on nonrelativistic quantum mechanics. Sometimes classified as "Advanced Quantum Mechanics" Chapters 2024 provide the basis for an understanding of manybody theories, quantum electrodynamics, and relativistic particle theory. The pace quickens here, and many mathematical steps are left to the exercises. It would be presumptuous to claim that every section of this book is indispensable for learning the principles and methods of quantum mechanics. Suffice it to say that there is more here than can be comfortably accommodated in most courses, and that the choice of what to omit is best left to the instructor. Although my objectives are the same now as they were in the earlier editions, I have tried to take into account changes in physics and in the preparation of the students. Much of the first twothirds of the book was rewritten and rearranged while I was teaching beginning graduate students and advanced undergraduates. Since most students now reach this course with considerable previous experience in quantum mechanics, the graduated threestage design of the previous editionswave mechanics, followed by spin onehalf quantum mechanics, followed in turn by the fullfledged abstract vector space formulation of quantum mechanicsno longer seemed appropriate. In modifying it, I have attempted to maintain the inductive approach of the book, which builds the theory up from a small number of simple empirical facts and emphasizes explanations and physical connections over pure formalism. Some introductory material was compressed or altogether jettisoned to make room in the early chapters for material that properly belongs in the first half of this course without unduly inflating the book. I have also added several new topics and tried to refresh and improve the presentation throughout.
vii
viii
Preface
As before, the book begins with ordinary wave mechanics and wave packets moving like classical particles. The Schrodinger equation is established, the probability interpretation induced, and the facility for manipulating operators acquired. The principles of quantum mechanics, previously presented in Chapter 8, are now already taken up in Chapter 4. Gauge symmetry, on which much of contemporary quantum field theory rests, is introduced at this stage in its most elementary form. This is followed by practice in the use of fundamental concepts (Chapters 5, 6, and 7), including twobytwo matrices and the construction of a onedimensional version of the scattering matrix from symmetry principles. Since the braket notation is already familiar to all students, it is now used in these early chapters for matrix elements. The easy access to computing has made it possible to beef up Chapter 7 on the WKB method. In order to enable the reader to solve nontrivial problems as soon as possible, the new Chapter 8 is devoted to several important techniques that previously became available only later in the course: Variational calculations, the RayleighRitz method, and elementary timeindependent perturbation theory. A section on the use of nonorthogonal basis functions has been added, and the applications to molecular and condensedmatter systems have been revised and brought together in this chapter. The general principles of quantum mechanics are now the subject of Chapters 9 and 10. Coherent and squeezed harmonic oscillator states are first encountered here in the context of the uncertainty relations. Angular momentum and the nonrelativistic theory of spherical potentials follow in Chapters 11 and 12. Chapter 13 on scattering begins with a new introduction to the concept of cross sections, for colliding and merging beam experiments as well as for stationary targets. Quantum dynamics, with its various "pictures" and representations, has been expanded into Chapters 14 and 15. New features include a short account of Feynman path integration and a longer discussion of density operators, entropy and information, and their relation to notions of measurements in quantum mechanics. All of this is then illustrated in Chapter 16 by the theory of twostate systems, especially spin onehalf (previously Chapters 12 and 13). From there it's a short step to a comprehensive treatment of rotations and other discrete symmetries in Chapter 17, ending on a brief new section on nonAbelian local gauge symmetry. Boundstate and timedependent perturbation theories in Chapters 18 and 19 have been thoroughly revised to clarify and simplify the discussion wherever possible. The structure of the last five chapters is unchanged, except for the merger of the entire relativistic electron theory in the single Chapter 24. In Chapter 20, as a bridge from elementary quantum mechanics to general collision theory, scattering is reconsidered as a transition between free particle states. Those who do not intend to cross this bridge may omit Chapter 20. The quantum mechanics of identical particles, in its "second quantization" operator formulation, is a natural extension of quantum mechanics for distinguishable particles. Chapter 21 spells out the simple assumptions from which the existence of two kinds of statistics (BoseEinstein and FermiDirac) can be inferred. Since the techniques of manybody physics are now accessible in many specialized textbooks, Chapter 22, which treats some sample problems, has been trimmed to focus on a few essentials. Counter to the more usual quantization of the classical Maxwell equations, Chapter 23 starts with photons as fundamental entities that compose the electromagnetic field with its local dynamical properties like energy and momentum. The interaction between matter and radiation fields is treated only in first approximation,
Preface
ix
leaving all higherorder processes to more advanced textbooks on field theory. The introduction to the elements of quantum optics, including coherence, interference, and statistical properties of the field, has been expanded. As a paradigm for many other physical processes and experiments, twoslit interference is discussed repeatedly (Chapters 1, 9, and 23) from different angles and in increasing depth. In Chapter 24, positrons and electrons are taken as the constituents of the relativistic theory of leptons, and the Dirac equation is derived as the quantum field equation for charged spin onehalf fermions moving in an external classical electromagnetic field. The oneparticle Dirac theory of the electron is then obtained as an approximation to the manyelectronpositron field theory. Some important mathematical tools that were previously dispersed through the text (Fourier analysis, delta functions, and the elements of probability theory) have now been collected in the Appendix and supplemented by a section on the use of curvilinear coordinates in wave mechanics and another on units and physical constants. Readers of the second edition of the book should be cautioned about a few notational changes. The most trivial but also most pervasive of these is the replacement of the symbol J.L for particle mass by m, or me when it's specific to an electron or when confusion with the magnetic quantum number lurks. There are now almost seven hundred exercises and problems, which form an integral part of the book. The exercises supplement the text and are woven into it, filling gaps and illustrating the arguments. The problems, which appear at the end of the chapters, are more independent applications of the text and may require more work. It is assumed that students and instructors of quantum mechanics will avail themselves of the rapidly growing (but futile to catalog) arsenal of computer software for solving problems and visualizing the propositions of quantum mechanics. Computer technology (especially MathType@ and Mathematica@) was immensely helpful in preparing this new edition. The quoted references are not intended to be exhaustive, but the footnotes indicate that many sources have contributed to this book and may serve as a guide to further reading. In addition, I draw explicit attention to the wealth of interesting articles on topics in quantum mechanics that have appeared every month, for as long as I can remember, in the American Journal of Physics. The list of friends, students, and colleagues who have helped me generously with suggestions in writing this new edition is long. At the top I acknowledge the major contributions of John P. Hernandez, Paul S. Hubbard, Philip A. Macklin, John D. Morgan, and especially Eric Sheldon. Five seasoned anonymous reviewers gave me valuable advice in the final stages of the project. I am grateful to Mark D. Hannam, Beth A. Kehler, Mary A. Scroggs, and Paul Sigismondi for technical assistance. Over the years I received support and critical comments from Carl Adler, A. Ajay, Andrew Beckwith, Greg L. Bullock, Alan J. Duncan, S. T. Epstein, Heidi Fearn, Colleen Fitzpatrick, Paul H. Frampton, John D. Garrison, Kenneth Hartt, Thomas A. Kaplan, William C. Kerr, Carl Lettenstrom, Don H. Madison, Kirk McVoy, Matthew Merzbacher, Asher Peres, Krishna Myneni, Y. S. T. Rao, Charles Rasco, G. G. Shute, John A. White, Rolf G. Winter, William K. Wootters, and Paul F. Zweifel. I thank all of them, but the remaining shortcomings are my responsibility. Most of the work on this new edition of the book was done at the University of North Carolina at Chapel Hill. Some progress was made while I held a U.S. Senior Scientist Humboldt Award at the University of Frankfurt, during a leave of absence at the University of Stirling in Scotland, and on shorter visits to the Institute of Theoretical Physics at Santa Barbara, the Institute for Nuclear Theory in Seattle,
x
Preface
and TRIFORM Camphill Community in Hudson, New York. The encouragement of colleagues and friends in all of these places is gratefully acknowledged. But this long project, often delayed by other physics activities and commitments, could never have been completed without the unfailing patient support of my wife, Ann.
Eugen Merzbacher
Contents
CHAPTER 1
Introduction to Quantum Mechanics
1
1. Quantum Theory and the Wave Nature of Matter 2. The Wave Function and its Meaning 4 Problems 10
CHAPTER 2
1
Wave Packets, Free Particle Motion, and the Wave Equation 12
1. The Princil'le of Superposition 12 2. Wave Packets and the Uncertainty Relations 14 3. Motion of a Wave Packet 18 4. The Uncertainty Relations and the Spreading of Wave Packets 5. The Wave Equation for Free Particle Motion 22 Problems 24
CHAPTER 3
20
The Schrodinger Equation, the Wave Function, and Operator Algebra 25 1. The Wave Equation and the Interpretation of t/J 25
2. Probabilities in Coordinate and Momentum Space 29 3. Operators and Expectation Values of Dynamical Variables 34 4. Commutators and Operator Algebra 38 5. Stationary States and General Solutions of the Wave Equation 41 6. The Virial Theorem 47 Problems 49
CHAPTER 4
The Principles of Wave Mechanics
51
Hermitian Operators, their Eigenfunctions and Eigenvalues 51 The Superposition and Completeness of Eigenstates 57 The Continuous Spectrum and Closure 60 A Familiar Example: The Momentum Eigenfunctions and the Free Particle 62 5. Unitary Operators. The Displacement Operator 68 6. The Charged Particle in an External Electromagnetic Field and Gauge Invariance 71 7. Galilean Transformation and Gauge Invariance 75 Problems 78
CHAPTER 5
1. 2. 3. 4.
The Linear Harmonic Oscillator
81
79
1. Preliminary Remarks 79 2. Eigenvalues and Eigenfunctions
xii
Contents
3. Study of the Eigenfunctions 84 4. The Motion of Wave Packets 89 Problems 90
CHAPTER 6
Sectionally Constant Potentials in One Dimension
92
1. The Potential Step 92 2. The Rectangular Potential Barrier 97 3. Symmetries and Invariance Properties 99 4. The Square Well 103 Problems 111
CHAPTER 7
The WKB Approximation 113
I. The Method 113 2. The Connection Formulas 116 3. Application to Bound States 121 4. Transmission Through a Barrier 125 5. Motion of a Wave Packet and Exponential Decay Problems 134
CHAPTER 8
131
Variational Methods and Simple Perturbation Theory
135
I. The Calculus of Variations in Quantum Mechanics 135 2. The RayleighRitz Trial Function 139 3. Perturbation Theory of the SchrOdinger Equation 142 4. The RayleighRitz Method with Nonorthogonal Basis Functions 5. The Double Oscillator 149 6. The Molecular Approximation 159 7. The Periodic Potential 165 Problems 176
CHAPTER 9
146
Vector Spaces in Quantum Mechanics
179
I. Probability Amplitudes and Their Composition 179 2. Vectors and Inner Products 186 3. Operators 188 4. The Vector Space of Quantum Mechanics and the BraKet Notation 5. Change of Basis 199 6. Hilbert Space and the Coordinate Representation 202 Problems 206
CHAPTER 10
195
Eigenvalues and Eigenvectors of Operators, the Uncertainty Relations, and the Harmonic Oscillator 207
I. The Eigenvalue Problem for Normal Operators 207 2. The Calculation of Eigenvalues and the Construction of Eigenvectors 209
Contents
xiii
3. Variational Formulation of the Eigenvalue Problem for a Bounded Hermitian Operator 212 4. Commuting Observables and Simultaneous Measurements 214 5. The Heisenberg Uncertainty Relations 217 6. The Harmonic Oscillator 220 7. Coherent States and Squeezed States 225 Problems '231
CHAPTER 11
Angular Momentum in Quantum Mechanics
233
1. Orbital Angular Momentum 233 2. Algebraic Approach to the Angular Momentum Eigenvalue Problem 3. Eigenvalue Problem for L z and L 2 • 242 4. Spherical Harmonics 248 5. Angular Momentum and Kinetic Energy 252 Problems 255
CHAPTER 12
238
Spherically Symmetric Potentials
256
1. Reduction of the CentralForce Problem 256 2. The Free Particle as a CentralForce Problem 257 3. The Spherical Square Well Potential 262 4. The Radial Equation and the Boundary Conditions 263 5. The Coulomb Potential 265 6. The BoundState Energy Eigenfunctions for the Coulomb Potential Problems 275
CHAPTER 13
270
Scattering
278
1. The Cross Section 278 2. The Scattering of a Wave Packet 286 3. Green's Functions in Scattering Theory 290 4. The Born Approximation 295 5. Partial Waves and Phase Shifts 298 6. Determination of the Phase Shifts and Scattering Resonances 7. Phase Shifts and Green's Functions 308 8. Scattering in a Coulomb Field 310 Problems 314
CHAPTER 14
302
The Principles of Quantum Dynamics
315
1. The Evolution of Probability Amplitudes and the Time Development Operator 315 2. The Pictures of Quantum Dynamics 319 3. The Quantization Postulates for a Particle 323 4. Canonical Quantization and Constants of the Motion 326
iv 5. Canonical Quantization in the Heisenberg Picture 6. The Forced Harmonic Oscillator 335 Problems 342
;HAPTER 15
Contents
331
The Quantum Dynamics of a Particle
344
1. 2. 3. 4.
The Coordinate and Momentum Representations 344 The Propagator in the Coordinate Representation 348 Feynman's Path Integral Formulation of Quantum Dynamics 355 Quantum Dynamics in Direct Product Spaces and Multiparticle Systems 358 5. The Density Operator, the Density Matrix, Measurement, and Information 363 Problems 370
CHAPTER 16
The Spin 372
1. Intrinsic Angular Momentum and the Polarization of IjJ waves 372 2. The Quantum Mechanical Description of the Spin 377 3. Spin and Rotations 381 4. The Spin Operators, Pauli Matrices, and Spin Angular Momentum 385 5. Quantum Dynamics of a Spin System 390 6. Density Matrix and Spin Polarization 392 7. Polarization and Scattering 399 8. Measurements, Probabilities, and Information 403 Problems 408
CHAPTER 17
Rotations and Other Symmetry Operations
410
1. The Euclidean Principle of Relativity and State Vector Transformations 410 2. The Rotation Operator, Angular Momentum, and Conservation Laws 413 3. Symmetry Groups and Group Representations 416 4. The Representations of the Rotation Group 421 5. The Addition of Angular Momenta 426 6. The ClebschGordan Series 431 7. Tensor Operators and the WignerEckart Theorem 432 8. Applications of the WignerEckart Theorem 437 9. Reflection Symmetry, Parity, and Time Reversal 439 10. Local Gauge Symmetry 444 Problems 448
CHAPTER 18
BoundState Perturbation Theory
453
451
1. The Perturbation Method 451 2. Inhomogeneous Linear Equations
Conren~
xv
3. Solution of the Perturbation Equations 455 4. Electrostatic Polarization and the Dipole Moment 459 5. Degenerate Perturbation Theory 463 6. Applications to Atoms 467 7. The Variational Method and Perturbation Theory 473 8. The Helium Atom 476 Problems < 480
CHAPTER 19
TimeDependent Perturbation Theory
482
1. The Equation of Motion in the Interaction Picture 482 2. The Perturbation Method 485 3. Coulomb Excitation and Sum Rules 487 4. The Atom in a Radiation Field 491 5.' The Absorption Cross Section 495 6. The Photoelectric Effect 501 7. The Golden Rule for Constant Transition Rates 503 8. Exponential Decay and Zeno's Paradox 510 Problems 515
CHAPTER 20
The Formal Theory of Scattering
517
1. The Equations of Motion, the Transition Matrix, the S Matrix, and the Cross Section 517 2. The Integral Equations of Scattering Theory 521 3. Properties of the Scattering States 525 4. Properties of the Scattering Matrix 527 5. Rotational Invariance, Time Reversal Symmetry, and the S Matrix 530 6. The Optical Theorem 532 Problems 533
CHAPTER 21
Identical Particles
535
1. The Indistinguishability of and the State Vector Space for Identical Particles 535 2. Creation and Annihilation Operators 538 3. The Algebra of Creation and Annihilation Operators 540' 4. Dynamical Variables 544 5. The Continuous OneParticle Spectrum and Quantum Field Operators 546 6. Quantum Dynamics of Identical Particles 549 Problems 552
CHAPTER 22
Applications to ManyBody Systems
555
1. Angular Momentum of a System of Identical Particles 555 2. Angular Momentum and Spin OneHalf Boson Operators 556
7. Coherence. Momentum. 2. 4. Energy. Interference. 2. Interaction with Charged Particles 576 4. Elements of Quantum Optics 580 5. Time Reversal. Fourier Analysis and Delta Functions Review of Probability Concepts 634 Curvilinear Coordinates 638 Units and Physical Constants 640 642 REFERENCES INDEX 647 . Fundamental Notions 569 2. The HartreeFock Method 560 5. and the peT Theorem 608 The OneParticle Approximation 613 Dirac Theory in the Heisenberg picture 617 Dirac Theory in the SchrOdinger Picture and the Nonrelativistic Limit 621 9. 5. and Statistical Properties of the Field 583 Problems 591 CHAPTER 24 573 Relativistic Electron Theory 592 1.xvi 3. 3. 3. Quantum Statistics and Thermodynamics 564 Problems 567 CHAPTER 23 Contents 558 Photons and the Electromagnetic Field 569 1. 4. FirstOrder Perturbation Theory in ManyBody Systems 4. Central Forces and the Hydrogen Atom 623 Problems 629 630 630 APPENDIX 1. and Angular Momentum of the Radiation Field 3. 6. 8. The ElectronPositron Field 592 The Dirac Equation 596 Relativistic Invariance 600 Solutions of the Free Field Dirac Equation 606 Charge Conjugation.
It is equally well known that the hydrogen atom can be excited by absorbing certain discrete amounts of energy and that it can return the excitation energy by emitting light of discrete frequencies. These are empirical facts. By the simple relation (1. we mention Haken and Wolf (1993). atoms and radiation to molecules and condensed matter. and predict the behavior of a vast range of physical systems. h = 6. and Perkins (1982). above all. and somewhat indirectly from the interpretation of spectral lines. atoms. Quantum Theory and the Wave Nature of Matter.6261 X 10.CHAPTER 1 Introduction to Quantum Mechanics Quantum mechanics is the theoretical framework within which it has been found possible to describe. the discrete numbers nn (n = 0.) being the only values which a component of the angular momentum of a system can assume. and the bibliographic information is listed after the Appendix. 3/2. It is essential to have at hand a current summary of the relevant empirical knowledge about systems to which quantum mechanics applies. which consists of a proton and an electron. and nuclei possess discrete energy levels is one of the basic facts of quantum physics. 2. 1. The electron can be removed from the atom and identified by its charge. the constant n = hl2'1T appears as the unit of angular momentum. This introductory chapter sets the stage with a brief review of th~ historical background and a preliminary discussion of some of the essential concepts. and molecules.15 eV sec. spin. molecules. Krane (1987). from particles through nuclei. Niels Bohr discovered that any understanding of the observed discreteness requires. 1 . It comes most directly from observations on inelastic collisions (FranckHertz experiment) and selective absorption of radiation. . charge. Christman (1988). This discreteness of physical properties persists when particles combine to form nuclei. such as mass. All of this is true for systems that are composed of several particles. as well as for the particles themselves. Consider an object as familiar as the hydrogen atom. Divided by 2'1T.1) it links the observed spectral frequency v to the jump AE between discrete energy levels... All of these seem to be of a quantum nature in the sense that they take on only certain discrete values.136 X 10. The notion that atoms. mass. Among many good choices. the introduction of Planck's constant. 112. In the early days. most of which are no more' 'elementary" than atoms and nuclei. bound together by forces of electrostatic attraction. The composite structure of most particles has been unraveled by quantum theoretic 'Many references to the literature on quantum mechanics are found in the footnotes. this constant was often called the quantum of action. and spin. 1 1.34 J sec = 4. and magnetic moment. Matter at the atomic and nuclear or microscopic level reveals the existence of a variety of particles which are identifiable by their distinct properties. correlate. The experimental evidence for this fact is overwhelming and well known.
It is well known that (1.2.2 Chapter 1 lntroduction to Quantum Mechanics analysis of "spectroscopic" information accumulated in highenergy physics experiments. the concept of the photon as a particle with energy E = hv emerged.1) played an important role even in the earliest forms of quantum theory. The constant h connects the wave (v) and particle (E) aspects of light. Louis de Broglie proposed that the waveparticle duality is not a monopoly of light but is a universal characteristic of nature which becomes evident when the magnitude of h cannot be neglected. Heavier objects.2) where the quantum numbers nj are integers.2) gave good results in calculating the energy levels of simple systems but failed when applied to such systems as the helium atom. these behave very much like X rays. Although such effects were first produced with electron beams. The finiteness of Planck's constant is the basic point here. then no . When incident on a crystal. This means that in certain experiments beams of particles with mass give rise to interference and diffraction phenomena and exhibit a behavior very similar to that of light. Exercise 1. such as entire atoms and molecules. From experiments on the interference and diffraction of particles.2) and the Bohr formula (1. can exhibit a particlelike nature and is emitted or absorbed only in quanta of energy. Bohr was able to calculate discrete energy levels of an atom by formulating a set of quantum conditions to which the canonical variables qj and pj of classical mechanics were to be subjected. derive the Balmer formula for the spectrum of hydrogenic atoms from the quantum condition (1. Thus. they are now commonly observed with slow neutrons from a reactor. Although one sometimes speaks of matter waves. which through the nineteenth century had been so firmly established as a wave phenomenon. have also been shown to exhibit wave properties. Einstein used it to explain the photoelectric effect by inferring that light.1. Calculate the quantized energy levels of a linear harmonic oscillator of angular frequency w in the old quantum theory. For our purposes.1). The quantum conditions (1. He thus brought out a second fundamental fact. and each contour integral is taken over the full period of the generalized coordinate qj. it is sufficient to remember that in this "old quantum theory" the classical phase (or action) integrals for a conditionally periodic motion were required to be quantized according to f pjdqj = njh (1. this term is not intended to convey the impression that the particles themselves are oscillating in space. we infer the very simple law that the infinite harmonic plane waves associated with the motion of a free particle of momentum p propagate in the direction of motion and that their (de Broglie) wavelength is given by (1. usually referred to as the wave nature of matter. Exercise 1. Assuming that the electron moves in a circular orbit in a Coulomb field. For if h were zero.3) This relation establishes contact between the wave and the particle pictures.
= A x h « xp I (1. we note that even as light a body as an atom moving with a kinetic energy corresponding to a temperature of T = 10..4) i. This requirement. which serves as a guide in discovering the correct quantum laws. such as the linear dimension of an obstacle in its path. the wave aspect of matter will be hidden from our sight. statistical distributions of variable numbers of photons must playa fundamental role in the theory. which can be regarded as the short wavelength limit of wave mechanics in the same way as geometrical optics is the short wavelength limit of wave optics. to which classical mechanics is applicable. Later we will see that the limiting process which establishes the connection between quantum and classical mechanics can be exploited to give a useful approximation for quantum mechanical problems (see WKB approximation. Reversing the argument that led to de Broglie's proposal. Indeed. except that c is now the speed of sound. just as we expect classically.6 m! We thus expect that classical mechanics is contained in quantum mechanics as a limiting form (h~O). we conclude that (1. satisfy the condition xp » h extremely well. sound) waves are called phonons and behave similarly to photons. the waves of the electromagnetic field are associated with quanta (photons) for which the BohrEinstein relation E = hv holds. even if n is arbitrarily large.3) is valid for photons as well as for material particles. Such a field can be described in classical terms only if the photons can act coherently. as a classical electromagnetic or an elastic wave may. of momentum p = hiA. It is important to remember that such waves are generated in an elastic medium. As will be discussed in detail in Chapter 23. is called the correspondence principle. and the waves can be longitudinal as well as transverse. or quantum. Since photons have no mass. this requirement leads to the peculiar conclusion that a state of exactly n photons cannot represent a classical field. their energy and momentum are according to relativistic mechanics related by E = cp. the gradual transition that we can make conceptually as well as practically from the atomic level with its quantum laws to the macroscopic level at which the classical laws of physics are valid suggests that quantum mechanics must not only be consistent with classical physics but should also be capable of yielding the classical laws in a suitable approximation. If x is a characteristic length involved in describing the motion of a body of momentum p. Evidently. each contributing a very small amount of momentum. in agreement with the connection between energy (density) and momentum (density) in Maxwell's theory of the electromagnetic field. the associated wave would always correspond to A = 0 and would follow the laws of classical mechanics. We may read (1. . The massless quanta corresponding to elastic (e. a wave that propagates in an infinite medium has associated with it a particle.. generally. At macroscopic wavelengths. Chapter 7).3) the other way around and infer that. For example. Macroscopic bodies. corresponding to microwave or radio frequency. a very large number of photons is required to build up a field of macroscopically discernible intensity.6 K still has a wavelength no greater than about a micron or 10. there must be associated with the wave an enormous number of quanta. Let us formulate this a bit more precisely. if . A free particle would then not be diffracted but would go on a straight rectilinear path. To give a numerical example.Quantum Theory and the Wave Nature of Matter 3 matter what momentum a particle had. and not in free space.e.g. if the quantum of action h is negligible compared with xp. If a macroscopic wave is to carry an appreciable amount of momentum.
wavelength. (There is also the phase constant of a wave. it is evident that if our analogy is meaningful at all. an appreciation of Bohr's program is important because. The Wave Function and Its Meaning. exhibits both particle and wave aspects. We speak typically of the (natural) modes of such systems. 2 This duality was described in deliberately vague language by saying that the de Broglie relation "associates" a wavelength with a particle momentum. Quantum mechanics applies equally to protons. whose tireless efforts at interpreting the propositions of quantum mechanics culminated in the formulation of a doctrine of complementarity. Although this epistemological view of the relationship between classical and quantal physics is no longer central to the interpretation of quantum mechanics. A wave is generally described by its velocity of propagation. the size and shape of the enclosure that confines the air column. neutrons. see Wheeler and Zurek (1985). energy levels in an atom or a nucleus. not only in the realm of atoms but at all levels of human experience. e. The electron has been chosen only for definiteness of expression (and historical reasons). it is tempting to see in the discrete properties of atoms the manifestations of bounded wave motion and to connect the discrete energy levels with standing waves. sayan electron. 3For a compilation of original articles on the foundations of quantum mechanics and an extensive bibliography. "quantized. Mathematically. cannot be attained without analyzing the complementary aspects of the phenomena and of the means by which the phenomena are observed. E. and so on. Particle traditionally means an object with a definite position in space.4 Chapter 1 Introduction to Quantum Mechanics 2. Historically. and L. In classical macroscopic physics discrete. for a start." but it is obvious that the concept of an orbit as a trajectory covered with definite speed cannot be maintained. Yet discreteness did not first enter physics with the Bohr atom. are incompatible with each other unless traditional concepts of classical physics are modified to a certain extent. the need for a reconciliation of the two seemingly contradictory concepts of wave and particle was stressed above all by Bohr. through stimulation and provocation. quarks. membranes. According to this body of thought. 3 How a synthesis of the wave and particle concepts might be achieved can." physical quantities appear naturally as the frequencies of vibrating bodies of finite extension. In such a picture. a full description and understanding of natural processes. the clamping of the membrane rim. The vagueness reflects the fact that particle and wave aspects. or air columns. the bounded wave must somehow be related to the confinement of the particle to its "orbit. when they show up in the same thing such as the electron. there must be a correspondence between the energy of an atom and the 2It will be convenient to use the generic term electron frequently when we wish to place equal emphasis on the particle and wave aspects of a constituent of matter. and it is characteristic of a wave that it does not define a location or position sharply. the discrete behavior is enforced by boundary conditions: the fixed ends of the string. Similarly. it has greatly influenced our attitude toward the entire subject. .) Since in a standing wave it is the wavelength (or frequency) that assumes discrete values. and amplitude. for one particle this is undetermined. Ballentine (1987). mesons.. facing us at the outset is the fact that matter. These phenomena have found their simple explanation in terms of interference between incident and reflected waves. such as strings. as we shall see later. Wave means a pattern spread out in space and time.g. As we have seen. perhaps be understood if we recall that the quantum theory must give an account of the discreteness of certain physical properties. Also see the resource letters in the American Journal of Physics: DeWitt and Graham (1971). but.
much as in a simple theory of wave optical phenomena. but the questiOlis can be placed in sharper focus by reexamining the de Broglie relation (1. The connection between wavelength and the mechanical quantities. with their space. Upon its introduction we immediately ask such questions as these: Is l/f a measurable quantity.6) changes the particle momentum to pi = P . For particles with zero spin. nucleons. or the wave number k = 2n1A. t). for many purposes the dynamical effects of the spin are negligible in first approximation. for example.8) . However. mayor may not have directional (Le. Mathematically. such as electrons. and particle momentum: (1.vt (1. For nonrelativistic velocities of the particles and for v«c. we attempt to build up quantum mechanics with mass (different from zero) first by use of scalar waves. polarization) properties. is likely to be much more complicated for an electron bound to a nucleus as in the hydrogen atom.3) between wavelength.3) has already given us a relationship connecting wavelength with energy (or momentum).and time derivatives. appear in the SchrOdinger equation.. or muons. y. We will also see that the spin has profound influence on the behavior of systems comprised of several. pions and K mesons. where we are concerned with interference and diffraction or with the geometrical optics limit.5) Suppose that a beam of particles having momentum p in the x direction is viewed from a frame of reference that moves uniformly with velocity v. The amplitudes or wave fields. the scalar waves are represented by a function l/f(x. the de Broglie formula (1. suitable corrections must be made later. or many.2 The Wave Function and Its Meaning 5 wavelength of the wave associated with the particle motion.mv (1. what feature of the particle aspect of the particle is related to the wave function? We cannot expect entirely satisfactory answers to these questions before we have become familiar with the properties of these waves and with the way in which l/f is used in calculations. which in colorless terminology is called the wave function. one that is not bound in an atom. Hence. momentum or energy. quarks.7) If the particles are in free space. especially if the particles move with nonrelativistic velpcities and are not exposed to magnetic fields. We will see in Chapter 16 that the spin of the particles corresponds to the polarization of the waves. and what precisely does it describe? In particular. the usual Galilean trimsformation x' = x . we must assume that the de Broglie relation is valid also in the new frame of reference and that therefore A' h p' p  =FA mv h (1. Z. the transverse nature of light can often be neglected. which. along the x axis. Erwin SchrOdinger discovered the wave equation that enables us to evaluate the "proper frequencies" or eigenfrequencie's of general quantum mechanical systems. For particles with nonzero spin. this gives an appropriate description. For a free particle. We will neglect the spin for the time being. identical particles. or for a particle moving in any kind of a potential.
A local gauge transformation. there is no compelling reason for it to be a realvalued function. as a measure of the number of amplitude peaks and valleys within a given distance.Vt)/ A. whose amplitude is in principle observable and which are therefore unchanged under the Galilean transformation. What has gone awry? Two explanations come to mind to resolve this puzzle. I/J changes into a transformed wave function.6 Chapter i introduction to Quantum Mechanics We can easily imagine an experimental test of this relation by measuring the spacing of the fringes in a twoslit Youngtype interference apparatus.7 that by allowing I/J to be complexvalued for the description of free particles with momentum p. We will see in Section 4.9) from which we deduce the familiar Doppler shift (1. I/J'.8) and (1.6).(x' A 27T . will then be found to provide a new transformation rule for the phase of the waves. it is in stark conflict with the conclusion (1. Of the outcome of such a test there can hardly be any doubt: The fringe pattern will broaden. An alternative.11) and restoring consistency with the correct quantum relation (1.8) for de Broglie waves. the frequency and wave number must transform relativistically as 'Y(w . (1. Here.11) is to realize that Lorentz. and in the final analysis equivalent. their phase.Vt) = . the main lesson to be learned is that the I/J waves are unlike classical elastic waves. If I/J cannot be a directly measurable amplitude.8) and (1. we must entertain the possibility that. When classical elastic waves. 'Y = (1 _~:) In (1.11) Although the invariance of the wavelength accords with our experience with elastic waves. (1. If the Lorentz transformation x' = 'Y(x . is Galileaninvariant. <P = kx .V't) (1. induced by the Galilean transformation (1. and if Lorentz invariance (instead of Galilean invariance) of the phase of the I/J waves is assumed.6) gives the connection <P = <p' = . under a Galilean transformation.10) and the unsurprising result: A' = A (1.vk) (1. replacing (1.12) is applied. which in its entirety moves at velocity v parallel to the beam. way to avoid the contradiction implied by Eqs.wt = 27T(X .8). the I/J waves of quantum mechanics.11) can be resolved. the conflict between Eqs. and we must ascertain the transformation law for de Broglie waves. In later chapters we will see that both are valid and that they are mutually consistent. transformations may be required in spite of the assumed subluminal particle and frameofreference velocities. rather than Galilei.(x' + A 27T vt . Instead. The transformation (1. which propagate in the "rest" frame of the medium with speed V are viewed from the "moving" frame of reference.13) . corresponding to the increased wavelength.vt) and t' = 'Y(t  ~:).
the transformation law for the wave number contains a relativistic term. For if it were that.7 for a more detailed discussion. is related to the particle mass by (1. our treatment is intended to center on nonrelativistic quantum mechanics. it is consistent with experiment and theory to associate the wave I/J with a single particle or atom or other quantum system as a 1 . Compare the behavior of de Broglie waves for particles of mass m with the changes that the wavelength and frequency of light undergo as we look at a plane electromagnetic wave from a "moving" frame of reference. which was tacitly assumed to be negligible in the nonrelativistic regime. of which we have no direct experimental information. if destructive interference of the I/J waves is to account for the observed dark interference fringes and for the instability of any but the distinguished Bohr orbits in a hydrogen atom. z at time t. If its physical significance remains as yet somewhat obscure to us.15) which greatly exceeds the speed of light and which explains the need for Lorentz transformations under all circumstances. From all that is known to date.) The full meaning of this interpretation of I/J and its internal consistency will be discussed in detail in Chapter 3. we will become accustomed to I/J as an important addition to our arsenal of physical concepts. Here we merely want to advance some general qualitative arguments for this socalled probability interpretation of the quantum wave function for particles with mass. but the intensity of the fringes is measured by E 2 and B2 • In analogy to this situation.14) This result is eminently reasonable in a relativistic quantum mechanics. z. we will retain Galilean transformations and acknowledge the need to transform I/J appropriately. it is impossible for I/J to be nonnegative everywhere. which was introduced into quantum mechanics by Max Born. Yet. (See Section 4. As we progress through quantum mechanics. except for two final chapters on fully relativistic quantum mechanics for photons and electrons. it would have to have a nonnegative value everywhere. z at time t. This relation becomes consistent with (1. in spite of its unusual transformation properties.2 The Wave Function and Its Meaning 7 For low relative velocities (1' = I) the second of these equations again gives the firstorder Doppler frequency shift. It implies that for nonrelativistic particles the phase velocity of the waves is w hw mc v===k p p 2 (1. But the function I/J(x. y.3. we assume that the positive quantity II/J(x. However. t) itself cannot be a direct measure of the likelihood of finding the particle at position x. since. In physical optics. we do not expect to find the particle in those regions of space where I/J = O. (The absolute value has been taken because it will turn out that I/J can have complex values. y. t) 2 measures the probability of finding the particle at position x. However. y.5) and the nonrelativistic equations (1. interference patterns are produced by the superposition of waves of E and B. one thing seems certain: The wave function I/J must in some sense be a measure of the presence of a particle. y. Conversely.) Exercise 1. Thus. in regions of space where the particle may be found.7) only if it is assumed that the frequency w of de Broglie waves. z. I/J must be different from zero.
and that still engenders discomfort in some quarters.8 Chapter i introduction to Quantum Mechanics representative of a statistical ensemble. but that it has an intrinsic probabilistic meaning. Single particles are subject to wave interference effects. Note that the appearance of the interference effects does not require that a whole beam of particles go through the slits at one time.1. in seemingly random fashion. as is customary in speaking about elastic or electromagnetic waves. at a distance Sl = AP and Sz = BP from the slits. the observations can be correlated effortlessly with the mathematical formalism. Bright fringes appear at P if lSI .1. t).t) is shown. depicted by surfaces of equal phase.szl equals an integral multiple of the wavelength. as used in quantum mechanics. generally refers to the spacetimedependent wave function l/J. now there. it has been abundantly established that the interference fringes shown schematically in Figure 1. When many particles have come through. A plane wave.fer. A section of the intensity profile i(r. With the aid of these tools. and 13>. but all scales on this figure are distorted for emphasis. At least if l/J is so interpreted. and not merely to its extreme value.1 are the statistical result of the effect of a very large number of independent particles hitting the screen. this conclusion was already implicit in our earlier discussion regarding an infinite plane wave as representative of a free particle with definite momentum (wavelength) but completely indefinite position. is incident with wavelength A = 2nlk from the left on the narrow slits A and B. In a sense. The conclusion is almost inevitable that l/J describes the behavior of single particles. a regular interference pattern will be seen to have formed. a particle hitting the screen now here.t). Owing to their characteristic properties. The quantity 1l/J1 2 would appear to measure the chance of finding the particle at a certain place. we note that the term amplitude. t) Figure 1. depends on the phase difference a(r. y. The indeterminism that the probabilistic view ascribes to nature. The amplitude and intensity at the spacetime point P(r. As a word of caution. particles can actually be accelerated and observed singly. The interference pattern evolves only after many particles have been deposited on the detection screen. and the interference pattern can be produced over a length of time. Schematic diagram of the geometry in a twoslit experiment.sz) = k (AC). z. In fact. can be illustrated by the idealized experiment shown in Figure 1. such as charge and mass. These considerations motivate the more descriptive name probability amplitude for the wave function l/J(x. .t) = k(SI . particles can be identified singly in the detection devices of experimental physics.
allowing. the most notorious "imperfection" of quantum mechanics being the abandonment of strict classical determinism. just as a light wave goes through both slits in the analogous optical Young interference experiment. If the wave describes the behavior of a single particle. This interpretation thus denies the possibility of a more' 'complete" theory that would encompass the innumerable experimentally verified predictions of quantum mechanics but would be free of its supposed defects. Thus. and evaluation of information are inherent in the theoretical description of physical processes. the singleparticle probability doctrine of quantum mechanics asserts that the indetermination of which we have just given an example is a property inherent in nature and not merely a profession of our temporary ignorance from which we expect to_ be relieved by a future and more complete description. it follows that we cannot decide through which of the two slits the particle has gone. the acquisition. quantum physics can be said to provide a more comprehensive and complete account of the world than that . If we try to avoid this consequence by determining experimentally with some subtle monitoring device through which'slit the particle has passed. The appearance of the interference fringes depends on the passage of the wave through both slits at once. the conditions under which the interference pattern is produced forbid a determination of the slit through which the particle passes. in the twoslit experiment some hidden property of the individual particles would then presumably be responsible for the particular trajectories that the particles follow. but the basic feature should be clear from the present example: Wave aspect and particle aspect in one and the same thing are compatible only if we forego asking questions that have no meaning (such as: "Let us see the interference fringes produced by particles whose paths through an arrangement of slits we have followed!") An alternative view of the probability interpretation maintains that the wave l/J describes only the statistical behavior of a collection or ensemble of particles. and no evidence for any underlying hidden variables has ever been found. for a more detailed and deterministic account of the motion of the single systems that make up the ensemble than is sanctioned by quantum mechanics. there would be the potential for a more refined description of the individual member systems of the ensemble. and again in Chapter 23. and the accumulation of a large number of particles on the screen will result in two wellseparated traces. A single particle now goes definitely through one slit or the other. we know that such a more "complete" descriptionbroadly referred to as realistic or ontologicalcannot succeed unless it allows for something akin to actionatadistance between systems that are widely separated in space. processing. Since the propositions of quantum mechanics are couched in terms of probabilities. From this point of view. Generally. it is reasonable to suppose that l/J fully describes the statistical behavior of single systems.2 The Wave Function and Its Meaning 9 some are found deposited on the screen at locations that they could not reach if they moved along classical paths through either slit. we change the wave drastically and destroy the interference pattern. In the language of the principle of complementarity. at least in principle. For example. On the other hand. From the confrontation between experiments on an important class of quantum systems and a penetrating theoretical analysis that is based on minimal assumptions (first undertaken by John Bell). thereby predetermining the path of a particle. such nonlocal features arise naturally in quantum mechanics. Exactly the same traces are obtained if we block one slit at a time. This impressionistic qualitative description will be put on a firmer footing in Chapters 9 and 10. To summarize.
Amsterdam. The Principles of Quantum Mechanics. New York. D. New York.. Walter Greiner and Berndt Miiller. Amsterdam.. 1977.. Symmetries. Max Jammer. University of Chicago Press. Michael Morrison. N. 1957. including some by the pioneers of the field: L.5 4The following list is a sample of books that will contribute to an understanding of quantum mechanics. Shankar. Kurt Gottfried. Introduction to Quantum Mechanics. Lifshitz. 1961. 1958. 1992. Richard W. John Wiley. Princeton University Press. New York. SpringerVerlag. New York. Mass. Jordan. R. Thomas F. S. PrenticeHall. 1966. NorthHolland. 1986. New York. Reading. Topics in Advanced Quantum Mechanics. James M. 1960. Loose talk about these issues may lead to the impression that quantum mechanics must forsake the classical goal of a wholly rational description of physical processes. Quantum Mechanics. L. E. A. Quantum Mechanics. The Physical Principles of the Quantum Theory.. Hoyt. An Introduction. 4 . Green. B. Feagin. R.. 2nd ed. Elements of Advanced Quantum Theory. AddisonWesley. Clarendon Press. Noordhoff. L. Classical Results. John Wiley. Englewood Cliffs. J. Berlin. Quantum Mechanics. M. Rubin H. Wittke. as this book hopes to demonstrate. Quantum Mechanics II. Dover reprint. Walter Greiner. N. New York. 1989. pp. Reading. Quantum Mechanics. Plenum. Die allgemeinen Prinzipien der Wellenmechanik. 1993. 1996. B. H. Reading. H. Robinett. N. Townsend. 1990. Cambridge University Press. Quantum Mechanics in Simple Matrix Form. SpringerVerlag. A. Reading. New York. Matrix Mechanics. I translated by G. Barry Holstein. Atomic Physics and Human Knowledge.. Dirac.. 1986. New York. and Visualized Examples. Stephen Gasiorowicz. Quantum Physics.10 Chapter I Introduction to Quantum Mechanics aspired to by classical physics. Introduction to Quantum Mechanics. 1965. New York. 1980. II translated by J. Landau and E.J. AddisonWesley. Principles of Quantum Mechanics. Powell and B. A. Mass. A Modern Approach to Quantum Mechanics. Crasemann. 1958. 1992. Modern Systems. 3rd ed. Werner Heisenberg. Berlin. John S. 1995. SpringerVerlag. Park. Ballentine. Berlin. John Wiley. Quantum Mechanics. H. Bell.. J. Eckart and C. translated by C.. Kramers. Bernard Diu. Quantum Mechanics. Vol. Quantum Mechanics. Sakurai. Interscience. Oxford University Press. Mass. 1990. Modern Quantum Mechanics. Quantum Mechanics. Quantum Mechanics. Quantum Mechanics. Roland Omnes. M. Berlin. Vol. P. Griffiths. Ziman. Dicke and J. . Messiah. Sykes and J.. Quantum Mechanics. 1966.. Benjamin. 1990. 1969. 1967. J. Oxford. Vol. Potter. Sources of Quantum Mechanics. J. 1168. Understanding Quantum Mechanics. and Frank Laloe.. van der Waerden. 1958. SpringerVerlag. A. New York. 5/1 of Encyclopedia of Physics.J. McGrawHill. 1930. 1997. John Wiley. Temmer. 1958. with San Fu Tuan. McGrawHill. The Conceptual Development of Quantum Mechanics. AddisonWesley. NorthHolland. Englewood. Nothing could be further from the truth. Niels Bohr. D. John Wiley. P. 1989. Englewood Cliffs. 4th ed. revised ed. Prentice Hall. AddisonWesley. SBooks suitable for introductory study of quantum mechanics include: Claude CohenTannoudji. translated by J. Wolfgang Pauli. 1992. Volume I. 1994.. Introduction to the Quantum Theory. Volumes I and II. M. Groningen. Benjamin/Cummings.J. PrenticeHall. New York. Bohm. 1949. W. L. S. 1994. SpringerVerlag. Quantum Mechanics with Mathematica. A. 1961 and 1962. New York. 2nd ed. Landau. Mass. The Interpretation of Quantum Mechanics. David A. McGrawHill. ed.
the interparticle distance in a gas of noninteracting atoms must be comparable to the de Broglie wavelength. if its wavelength is to be I meter? Are matter waves of macroscopic dimensions a real possibility? 2. or less. For the observation of quantum mechanical BoseEinstein condensation. To what velocity would an electron (neutron) have to be slowed down.Problems 11 Problems 1. How high a particle density is needed to achieve these conditions if the atoms have mass number A = 100 and are at a temperature of 100 nanokelvin? .
t) describe two waves. 1.2).22 MeV sec = 0. Z. Z. I rf/l 12 + 1rf/21 2 by the interference terms rf/lrf/2* + rf/2rf/l*' (An .0545 X 10.5819 X 10. t): If rf/l(X. y.1) and (2.15 eV sec h It is true that diffraction experiments do not give us any direct information about the detailed dependence on space and time of the periodic disturbance that produces the alternately "bright" and "dark" fringes. The Principle of Superposition. t) = rf/l + rf/2 also describes a possible physical situation.34 J sec = 6.1). We have learned that it is reasonable to suppose that a free particle of momentum p is associated with a harmonic plane wave. z. the intensity is determined by I rf/l + rf/21 2 This differs from the sum of .1. we develop the quantum mechanical description of the motion of free particles. their sum rf/(x. Heisenberg's uncertainty relations and the Schrodinger wave equation make their first appearance. The correspondence between quantum and classical motion serves as a guide in the construction.) A careful analysis of the interference experiment would require detailed consideration of the boundary conditions at the slits. on the screen by opening only one slit at a time is Irf/11 2 or Irf/21 2 When both slits are open. the two intensities. of wave packets that propagate like classical particles but exhibit quantum mechanical spreading in space and time.CHAPTER 2 Wave Packets.2) The symbol h is common in quantum physics and denotes the frequently recurring constant h = 2'71' = 1. This assumption is known as the principle of superposition and is illustrated by the interference experiment of Figure 1. t) and rf/2(X. by superposition of harmonic waves. The comparison with optical interference suggests that the fringes come about by linear superposition of two waves. y. Defining a vector k which points in the direction of wave propagation and has the magnitude k = 2'71' A (2. these ideas are formulated in the following fundamental assumption about the wave function rf/(x. asterisk will denote complex conjugation throughout this book. y. The intensity produced .1) we may write the fundamental de Broglie relation as p = hk (2. z. a point of view that has already been stressed in Section 1. Mathematically. Free Particle Motion.2 in the discussion of the simple twoslit interference experiment (Figure 1. and the Wave Equation Building on previous experience with wave motion and simple Fourier analysis. This is similar to the situation 12 . y. but all the evidence points to the correctness of the simple inferences embodied in (2.66 X 10.
which thus far we have not brought into connection with any physically observable phenomenon. the l/J waves of quantum mechanics cannot be observed directly.ikx respectively. . we find that this last equation leads to cos e rot) and of + 8 sin e =a and 8 cos e .wt) = rot) cos(kx + rot)  + rot) (2. We have no physical reason for rejecting complexvalued wave functions because.4) The coordinates y and z can obviously be omitted in describing these onedimensional processes. where the uncritical use of Huygens' principle must be justified by recourse to the wave equation and to Kirchhoff's approximation. Comparing coefficients of cos(kx sin(kx . t) = cos(kx  rot) + 8 sin(kx + 8 sin(8 sin(kx kx  rot) (2. Such phenomena.rot + e) + 8 sin(kx  rot + e) = a(e)[cos(kx  rot) + 8 sin(kx  rot)] for all values of x. and the observed diffraction pattern presumably measures only the intensity 1l/J1 2 • Summarizing our conclusions. nor should the phase constants of these waves have any physical significance. and e. We choose the solution 8 = i. 0) (2. Such a wave is associated with the motion of a free particle moving in the x direction with momentum p = fik. because our purpose in describing the idealized twoslit experiment was merely to show how the principle of superposition accounts for some typical interference and diffraction phenomena. 0) = Ae ikx and = Be. Hence. t. unlike elastic displacements or electric field vectors. it must be required that cos(kx . t) = cos(. a particle moving uniformly in the positive and negative x directions.sin 2 8 = a8 These equations are compatible for all e only if 8 = lor 8 = ± i.3) A plane wave moving in the negative x direction would be written as l/J2(X. and its most general form is l/JI(X.5) and a wave propagating toward decreasing x is = Bei(kx+wt) = Bei(kxwt) l/J2(X. There is no need here for such a thorough treatment. are impressive direct manifestations of the wave nature of matter. respectively.rot). when actually observed as in diffraction of particles by crystals. which describe. The (circular) frequency of oscillation is ro. a = e ie and are thus led to the conclusion that l/J waves describing free particle motion must in general be complex.kx .1 The Principle of Superposition 13 found in wave optics. l/J2(X. An arbitrary displacement of x or t should not alter the physical character of these waves. The simplest type of wave motion to which it will be applied is an infinite harmonic plane wave propagating in the positive x direction with wavelength A = 27Tlk and frequency ro. t) t) = Aei(kxwt) (2.6) The initial values of these waves are l/JI (x. we see that in this scheme an infinite harmonic plane wave propagating toward increasing x is l/Jl (x. Generally. We therefore adopt the principle of superposition to guide us in developing quantum mechanics.
To this end. Free Particle Motion. a certain degree of localization can be achieved. as the fringes on the screen in the diffraction experiment attest. ensures that l/J(x. 0) = cos kx = 'l ·kx + 'I2 e' 2 e' 'kx (2.5) and (2. for all rand t. Strong support for it can be gained by demonstrating that the correspondence with classical mechanics can be established within the framework of this theory. 0) = cos kx.6) are special cases of this with ky = kz = 0 and k x = ±k. thus becomes for arbitrary t: l/J(x. This rule. the temperature. We will refer to wave functions like (2. The plane wave (2.x+kyY+kzZwt) (2. A plane wave propagating in an arbitrary direction has the form l/J(r. instead of describing the field by just one function. 0) = cos kx.7) corresponds to particle motion with momentum.3). . but none of its solutions represent waves propagating without damping in one direction like (2. this must be an acceptable wave function. The heat flow equation is also of the first order in time and describes the behavior of a real quantity. It should be stressed that with the acceptance of complex values for l/J. 0) determines the future behavior of the wave uniquely.2). to which we shall adhere and which we shall generalize. we must find a way of making l/J(r.8) According to the principle of superposition. If this formulation is correct. and the Wave Equation it will be a function of k (see Section 2. at least approximately.14 Chapter 2 Wave Packets. develops independently. but this is accomplished using both E and B in coupled equations. IMaxwell's equations are of the first order even though the functions are real.9) Note that this is not the same as cos(kx . For example. z. t) = l/J(x. By superposition of several different plane waves. we might be able to describe the state of motion of a particle at time t completely by l/J(r.7). the classical motion of a particle that has both reasonably definite position and reasonably definite momentum. of course. t) = ~ ei(kxwt) + ~ ei(kx+wt) (2. into which l/J can be decomposed at t = 0. t). The foregoing discussion points to the possibility that by allowing the wave function to be complex.rwt) = Aei(k.! Wave Packets and the Uncertainty Relations. t) describe. which is precisely defined by (2. How does it develop in time? The principle of superposition suggests a simple answer to this important question: Each of the two or more waves. The real test of this assumption is. we expect that the complex l/J waves may be described by a linear differential equation which is of the first order in time. if the initial wave function is l/J(x. The wave function. but having amplitudes Il/JI = const. y.7) Equations (2. its success in accounting for experimental observations. which at t = 0 satisfies the initial condition l/J(x. the infinite harmonic plane waves (2. t) = Aei(k. this can be written as the sum of two exponential functions: l/J(x. at least at certain times.5) and (2. we are by no means excluding wave functions that are realvalued. 2.7) leave the position of the particle entirely unspecified. as if the other component(s) were not present.wt).6) briefly as plane waves.
kx » 1. On the other hand.2).. = 'For a review of Fourier analysis. 0) be a function that is sectionally continuous..kx /2..kx . the phase goes through many periods as u ranges from tJ. . Hence. 0).kx centered at u = 0.11) show that l/J (k. and Hassani (1991)..g.Xo.10). We now assume that l/J(kx ) is a real positive function and that it has a shape similar to Figure 2..13) For a proof of the uncertainty relation (2. 2 Formulas (2.12) is an even real function of x. The particle is most likely to be found at a position where t/J is appreciable. the phase of e iux in the integrand varies from x tJ. we can write (2. Bradbury (1984).. e.11) to be the inverse of (2. 0) and vfce versa. Let us denote by (xo.e.x the range (.1. no appreciable cancellations in the integrand occur (since l/J is positive definite). a symmetric distribution of kx about a mean value k . it is sufficient that "'(x. i. this happens at the ends of the wave packet.2..2 Wave Packets and the Uncertainty Relations 15 The mathematical tools for such a synthesis of localized compact wave packets by the superposition of plane waves of different wave number are available in the form of Fourier analysis.. AITken (1985). xo) the range of x for which'" is appreciably different from zero (see Figure 2..kx is less than 1. it follows that the widths of It/JI and l/J are effectively related by (2.) determines the initial wave function t/J(x. If Xo tJ.kx for any fixed value of x. x Making the change of variable we may write (2. see any text in mathematical physics.11) under very general conditions to be satisfied by t/J(x. (2.. when Xo tJ.10) and the inverse formula (2. i. Since l/J is appreciably different from zero only in a range tJ.. Denoting by tJ.kx l2 to +tJ. The basic formulas are summarized in Section 1 of the Appendix. where kx = kx . leading to destructive interference. note that the integral in (2. the largest variations that the phase ever undergoes are ±xo Ji. For Eq.e.13). xo).kx l2 to + x tJ. Assuming for simplicity that only one spatial coordinate.10) and (2. need be considered.. by an amount x tJ. It is easy to see for any number of simple examples that the width tJ.12) This is a wave packet whose absolute value is shown in Figure 2.x of the wave packet t/J stand in a reciprocal relationship: (2.kx of the amplitude l/J and the width tJ.13). x...kx /2 and violent oscillations of the e iux term occur..
. 0) = e. . plot I r/!(x. its value will be infinite for the wave packets defined in Exercises 2.72 and y:. .1.. The two functions correspond to a = 0.. and the Wave Equation _ _l""===_ _l._ _ 2 ko 4 =~. 0) 12 for several values of 8.. The plotted amplitude. kx + 8. as in Section 10. 0).13).'''(x) I 1. Normalized wave packet corresponding to the momentum distributions of Figure 2. Calculate </J (kJ and show that the widths of r/! and </J.::::t:o... k x 6 Figure 2._l.1 and 2. modulates the plane wave eikoX • Note the reciprocity of the widths of the corresponding curves in Figures 2. Example of a onedimensional momentum distribution.2. 3Note that if (Ax)2 is taken as the variance of x. and </J(kx ) = x all other values of kx • Calculate r/!(x. The function c/J(kx ) is normalized so that f:~~~ 1 c/J(kx ) 12 dkx = 1. .l.5. satisfy the reciprocal relation (2. for k . (heavy line).l=_x 4 2 0 2 4 Figure 2. Exercise 2. The function ep(kx ) = vaFi [cosh a (kx . and show that (2.2. kx :s.1 and 2.l.AIXI for .13) holds if ax is taken as the width at half maximum.2 ""'=L """" L .8:s.._ _.:) I. Free Particle Motion.3.1.1.16 Chapter 2 Wave Packets. o for Assume </J(kx ) = yI2.ko)]I represents a wave packet moving in the positive x direction with mean wave number ko = 3. Assume r/!(x. II/J(x) I = ji..2.... 3 Exercise 2.00 < x < + 00. reasonably defined.. (COSh ..
it suffices to view a peak in l/! as a crudely localized particle and a peak in </J as a particle moving with an approximately defined velocity. The relation (2.14) The product Llx Llkx assumes a value near its minimum of 1/2 only if the absolute value of </J behaves as in Figure 2. On the contrary. perhaps by making it asymmetric or by noting that in general it is complexvalued. The uncertainty relation (2. has no basis in fact. we will have to make all these statements precise and quantitative as we develop the subject.3. 0) thus describes a particle that is localized within a distance Llx about the coordinate origin.:::: 2 (2. the wave function cannot represent a particle of definite momentum (for if it did l/! would have to be a harmonic plane wave).4. Exercise 2.2 Wave Packets and the Uncertainty Relations 17 The initial wave function l/!(x. although the width of the kxdistribution is the same as in Exercise 2.13) might be averted by choosing </J (kJ more providentially.14). It also suggests that its spreadout momentum distribution is roughly pictured by the behavior of I </J1 2 • If 1</J1 2 has its greatest magnitude at kn the particle is most likely to have the momentum hkx ' Obviously. The fact that in quantum physics both waves and particles appear in the description of the same thing has already forced us to abandon the classical notion that position and momentum can be defined with perfect precision simultaneously. convince yourself that (2.1 and its phase is either constant or a linear function of k x ' Exercise 2. 0) is greater in accordance with the inequality (2. 0) and show that. Equation (2.1.14) together with the equation Px = hkx expresses this characteristic property of quantum mechanics quantitatively: Llx Llpx . By choosing reasonable numerical values for the mass and velocity.15) This uncertainty relation suggests that when l/! is the sum of plane waves with different values of kx .{) ~~ ::5 kx ::5 kx  "2 and kx + "2 ::5 {)  {) kx ::5 kx  + {) and </J = 0 for all other values of kx' Calculate l/!(x.15) limits the precision with which position and momentum . we easily see (and we will rigorously prove in Chapter 10) that in general 1 Llx Llkx . For the present.13) shows that we can make the wave packet define a position more sharply only at the cost of broadening the spectrum of kxvalues which must be admitted.15) does not in practice impose any limitations on the precision with which the position and momentum of a macroscopic body can be determined. Any hope that these consequences of (2. the width of l/!(x. This has been accomplished at the expense of combining waves of wave numbers in a range Llkx about t.V 27T for kx . </J (kJ = v Assume for kx  ~~ 27T "2 ::5 {) kx ::5 kx   + "2 {) = .:::: 2 h (2.
if necessary). . must we admit that the particle described by 4> (or l/J) has no definite momentum (or position) when we merely are unable to determine a single value consistently in making the same measurement on identically prepared systems? Could the statistical uncertainties for the individual systems be reduced below their quantum mechanical values by a greater effort? Is there room in the theory for supplementing the statistical quantum description by the specification of further ("hidden") variables. Since these predictions have been borne out experimentally to high accuracy. Bohr and Heisenberg were the first to show in detail. Generally. No technical or mathematical ingenuity can presumably devise the means of giving a sharper and more accurate account of the physical state of a single system than that permitted by the wave function and the uncertainty relation. Motion of a Wave Packet.px inescapably limits the precision of a simultaneous measurement of the coordinate x to Ji. quarks. we adopt as the central premise of quantum mechanics that For any given state (l/J or 4» the measurement of a particular physical quantity results with calculable probability in a numerical value belonging to a range of possible measured values.x ::::: hl2Ji. The function 1l/J1 2 is proportional to the probability of finding the particle at position x. Rather. Having placed the wave function in the center of our considerations. and the Wave Equation can be simultaneously ascribed to the particle. 14>1 2 is proportional to the probability of finding the particle to have momentum hkx • Under these circumstances. namely. photons. Quantum mechanics contends that the wave function contains the maximum amount of information that nature allows us concerning the behavior of electrons. 3. that the behavior of a material particle is described completely by its wave function l/J or 4> (suitably modified to include the spin and other degrees of freedom. Free Particle Motion. 4 We now know (Bell's theorem) that the most natural kinds of hiddenvariable descriptions are incompatible with some of the subtle predictions of quantum mechanics. how the finite value of h in the uncertainty relation makes the coexistence of wave and particle both possible and necessary. but which is supported by the enormous number of verified consequences derived from it. we must attempt to infer as much as possible about its properties. Similarly. so that two systems that are in the same quantal state (l/J or 4» may be found to be distinct in a more refined characterization? These intriguing questions have been hotly debated since the advent of quantum mechanics. protons. 4See Wheeler and Zurek (1985). Broadly. and the like. these tenets are referred to as the Copenhagen interpretation of quantum mechanics. Yet upon measurement the particle will always be found to have a definite position. These claims constitute a principle which by its very nature cannot be proved. Illuminating as Bohr's thought experiments are. quantum mechanics does not deny that precision measurements of momentum are feasible even when the particle is not represented by a plane wave of sharp momentum.px.18 Chapter 2 Wave Packets. neutrons. they merely illustrate the important discovery that underlies quantum mechanics. We have already discussed in Chapter I what this means in terms of experiments. These idealized experiments demonstrate explicitly how any effort to design a measurement of the momentum component Px with a precision set by Ji. in a number of interesting thought experiments. both quantities are fuzzy and indeterminate (Heisenberg uncertainty principle). and vice versa. but generally not by its precise momentum or position.
3 Motion of a Wave Packet 19 In three dimensions an initial wave packet !/J(r..23) Our interpretation.iw t + ik . .20) Let us consider a wave packet whose Fourier inverse <t> (k) is appreciably different from zero only in a limited range Llk near the mean wave vector hk. In coordinate space. in three dimensions the uncertainty relations are (2. The wave function at time t becomes .. we obtain !/J(r. of course. z.16) and the inverse formula <t>(k) = _1_ (27T)3/2 J !/J(r. (2.. incomplete until we determine and specify the dependence of w on k. then by the rule of Section 2.2). Vi<W t)!/J(r . and d 3 r = dx dy dz is the volume element in (position) coordinate space.16) describes is free. where m is the mass of the particle and nonrelativistic particle motion 5d 3 k = dkx dky dkz denotes the volume element in kspace. 0) is represented by the Fourier integral: 5 !/J(X. y. leads us to identify the group velocity of the !/J wave with the mean particle velocity hk/m. ky in a range Llky . the wave packet !/J(r. t) = exp(. we see that this describes a wave packet performing a uniform translational motion. then <t> is sensibly different from zero only in kx lies in a range Llkx . without any change of shape and with the group velocity (2. This determination will be made on physical grounds. O)e. Vi<W + .(k)ei(k<rwt) d k 3 'I' (2. 0) = !/J(r. Llz.18). If !/J is of appreciable magnitude only in coordinate ranges Llx.16) propagates independently of all the others according to the prescription of (2.7). (VkW)k=i< + k) .22) Ignoring the phase factor in front. 0) = (2:)312 J<t> (k)eikr d 3k 3 < (2. and kz in a range Llkx .21) with obvious abbreviations.17) If the particle which the initial wave function (2. Lly.r d r (2.Vi<W t.. Llp = hLlk. based on the correspondence between classical and quantum mechanics. To see this behavior we expand w(k) about k: w(k) = w(k) = + (k  w + (k k) ." (27T)312 J A. 0) (2.I'(r t) = _1_ . hence. such that (2.18) Formula (2. If we substitute the first two terms of this expansion into (2. t) must move approximately like a classical free particle with mean momentum hk. By (2.ik.1 (the superposition principle) each component plane wave contained in (2.19) These inequalities are generalizations of the onedimensional case.18) is.
for nonrelativistic particles. if quadratic terms are retained.. although obtained in this chapter for nonrelativistic particle mechanics.24) Integrating this. the behavior of the wave packet is given by an exact formula.18). For simplicity. An exponent can be neglected only if it is much less than unity in absolute value.20 Chapter 2 Wave Packets. We may assume that liw represents the particle energy. and the Wave Equation has been assumed. We must examine the conditions under which it is legitimate to neglect the quadratic and higher terms in (2.26) where V is a constant. Without these corrections the wave packet moves uniformly without change of shape. we obtain w(k) =  Ii 2m (k.29) Thus.+ . relativistic case. (2. which may be thought of as an arbitrary constant potential energy.k?  (2.27) A similar equation has already been found to hold for photons. we may omit the averaging bars and require m'Vkw = lik k. let us consider a onedimensional wave packet for which 4J (k) = 0 unless k = kx (ky = kz = 0).21). in which case the neglected term in (2. (2.= . they merely express the proportionality of two fourvectors: (2.k) +   Ii 2m (k . Free Particle Motion. This is not as surprising as it may seem.21). multiplied by t. Thus. The Uncertainty Relations and the Spreading of Wave Packets. 4.2) and (2. + + k. for (2.30) This term generally contributes to the exponent in the integrand of (2. Since this relation must hold for an arbitrary choice of k.) + const. and we should ask what happens in time when the neglect of the quadratic terms is no longer justified. For the nonrelativistic case c~redh~ansion (2. (k 2m 2m m .28) Hence their great universality. is (2. E = liw (2.21) actually terminates because (with V = 0) .. w(k) 1i1C' liP lik = . Indeed.27).25) or (2. the wave packet moves without appreciable change of shape for times t such that . are Lorentzinvariant and thus remain valid in the .
Compute and compare the values of this bound for an electron. and is attributable to the spread in initial velocity. But the latter condition must be satisfied if we are to be allowed to speak of the energy of the particles in a beam at all (rather than a distribution of energies).34) In this derivation. The kinetic energy is E = p. he or she "sees" the wave packet sweeping through his or her location. Consider a wave packet satisfying the relation Llx Llpx = Ii.33) If a monitor is located at a fixed position. a onegram object.31) is violated. so that their motion can be described classically? Exercise 2.31) The product It I Llvx represents an uncertainty in the position at time t. Exercise 2.15) has a companion that relates uncertainties in time and energy.. . If It I grows too large. Make an estimate of the lower bound for the distance Llx. a proton.1 nanometer = lA) be adequately represented by nonspreading wave packets.5. . the wave packet first contracts and eventually spreads. within which an object of mass m can be localized for as long as the universe has existed (= 10 10 years).31)./2m. and the entire universe.00) to the remote future (t 7 + 00).. The determination of the time at which the particle passes the monitor must then be uncertain by an amount: Llx Ii Ii Llt= .7.= Vx V x Llpx LlE Hence. : : : . Can the atoms in liquid helium at 4 K (interatomic distance about 0. Exercise 2. According to (2. this is assured if :5 Ii LlE mli Llt«(LlpJ2 With (2. it was assumed that the wave packet does not spread appreciably in time Llt while it passes through the observer's position. This means that as t increases from the distant past (t 7 . (2. condition (2.33) this condition is equivalent to LlE « E.6. the secondorder contribution to the phase can no 'longer be neglected. Show that if the packet is not to spread appreciably while it passes through a fixed position. the condition Llpx « Px must hold. and the wave packet is broadened so that for very large It I.4 The Uncertainty Relations and the Spreading of Wave Packets 21 or It I Llvx = Llpx m ItI «Ali llPx '= Llx (2. and it is uncertain by an amount LlE = Px Llpx = V x Llpx m (2. The uncertainty relation (2. over and beyond the initial uncertainty Llx.
. will be taken up in later chapters. As expected. (2.. V2 !fJ(r. but considerable care is required to establish valid quantitative statements (see Chapter 19). t) = . t) .. The details of solving Eq. but the presence of i ensures that the quantum mechanical wave equation has solutions with wave character.35). is obviously iii a!fJ~:.rwt). t) (2.7). be complexvalued is again manifest from the appearance of i in the differential equation.2 the units for the abscissa are A1 (for kJ and A(for x). Here we merely draw attention to an interesting alternative form of the wave equation.) + V with constant V. compute their mean velocity (in m/sec).37) . and the Wave Equation The derivation of the timeenergy uncertainty relation given here is narrow and precise. Free Particle Motion.18) gives the most general form of a free particle wave function. for its solution. requiring knowledge of the initial wave function !fJ(r. the mean kinetic energy (in meV). but not its derivative.22 Chapter 2 Wave Packets. and the energy spread (also in meV).g. More generally. + k. the equation { as(r. (2. a similar inequality relates the indeterminancy at in the time of occurrence of an event at the quantum level (e. t) + V } eiS(r. The wave equation for the plane waves ei(k. We must find a linear partial differential equation that admits (2.35). 5. the quantum mechanical wave equation and the diffusion (or heat flow) equation become formally identical.t)/fi = 0 (2.iii "2 at v 2m 2m S(r. Estimate the time scale for the spatial spread of this wave packet.36) If this expression for the wave function is substituted in Eq. t) + V!fJ(r. 0). Exercise 2. respectively. other representations are often useful.1 and 2.35) This quantum mechanical wave equation is also known as the timedependent Schrodinger equation for the free particle. The Wave Equation for Free Particle Motion.26). we need only establish the wave equation for the plane waves (2. using various different coordinate systems for the spatial variables. If the particles described by the wave packets in these figures are neutrons.t)/fi (2.+ [VS(r. t)]2 . That the solutions must.18) as its general solution. the corresponding temperature (in Kelvin). in general. obtained by the transformation !fJ(r. Although the plane wave (Fourier) representation (2.18) satisfies the same equation. To accomplish this. provided that the relation between wand k is given by the dispersion formula (2. with li 2 liw(k) = 2m (k. These are most conveniently obtained as solutions of a wave equation for this motion. For V = 0. + k. t) = eiS(r. Assume that in Figures 2.: . the decay of a nucleus or particle) to the spread aE in the energy associated with the process. it is a firstorder differential equation in t.8. since the linearity of the wave equation ensures that the superposition (2.
at t) :=:: nw = E (2.5 The Wave Equation for Free Particle Motion 23 is derived.44) 6To review the elements of classical Hamiltonian mechanics. to good approximation: as(x. (2. If the term in this equation containing i and h.:=:: S(x. t) is generally complex. the time development of IjJ (x. 0 (2.2 illustrate such a state. t). for times short enough to permit neglect of the changing shape of the wave packet. we assume that both IjJ and c/J are fairly localized functions in coordinate and kspace.42) for the partial derivatives of S(x.. we again consider the motion of a wave packet.+ at [VS(r.in 2m v2 S(r. however. 6 To further explore the connection between classical and quantum mechanics for a free particle. The most popular prototype. modulated by a realvalued Gaussian function of width ax and centered at xo: ljJ(x.41) where x(x) is a smooth realvalued function that has a minimum at Xo and behaves as x( ± 00) ~ + 00 at large distances. .t + ikxX} (2. The function S(r. t)11i is given by .39) would be the HamiltonJacobi equation for the Hamilton Principal Function of classical mechanics. t) + v = .38) Equation (2. t) = nkr .t + k aw k akx akx aw) iX x=.39) must hold.x =.19). t) represents a plane wave. respectively. t) = eiS(x. 0) = C exp{ (X4~:.t + x( x=. as for instance when ljJ(r.22). the hallmarks of quantum mechanics. see Goldstein (1980). In one dimension.nwt = p'r  Et (2.O)11i = exp[ X(x) + ikxx] (2.42) If the localized wave packet is broad enough so that X(x) changes smoothly and slowly over the distance of a mean de Broglie wavelength. but in special cases it may be real.37) informs us that wherever and whenever the wave function ljJ(r. in which case we have S(r. 0) = eiS(x. the nonlinear differential equation • as(r. were negligible.) Somewhat more generally.43) and as(x.t akx ( aw) (2. In the approximation that underlies (2. t) . Subject to the restrictions imposed by the uncertainty relations (2. t)f 2m . t) h wt + . is a state whose IjJ wave function at the initial time t = 0 is a plane wave of mean momentum hk.40) (The dynamics of this particular wave packet is the content of Problem 1 at the end of this chapter. t) does not vanish.1 and 2. we obtain from (2. we consider an initial onedimensional wave packet that can be written in the form ljJ(x. Figures 2. ax t) :=:: hk x = Px (2.
t) with the Hamilton Principal Function sex. (x.(x. t)]2 ax + V = 0 (2. assuming free particle motion. 0) + l/J2(X. As time evolves. t).4(dx)2 + _] ikxx Calculate the corresponding kxdistribution and l/J(x. 4. and the Wave Equation These equations hold for values of x near the minimum of X. A onedimensional initial wave packet with a mean wave number k and a Gaussian x amplitude is given by X2 l/J(x. Express the spreading Gaussian wave function ". the wave packets move and spread. The present chapter shows that our ideas about wave packets can be made quantitative and consistent with the laws of classical mechanics when the motion of free particles is considered. Apply the results to calculate the effect of spreading in some typical microscopic and macroscopic experiments. 3. while it advances according to the classical laws. choosing dx small enough to show that the wave packet spreads in time. Consider a wave function that initially is the superposition of two wellseparated narrow wave packets: l/J. Estimate the length of the wave packets in the direction of motion. hence near the peak Il/JI.45) This shows that for a broad class of wave packets a semiclassical approximation for the wave function may be used by identifying Sex. From them we deduce that in this approximation Sex. t)l1i]. t) = exp[i Sex. 2. 0) = C exp [ . Problems 1. Identify the function sex. t) satisfies the classical HamiltonJacobi equation as(x. We must now turn to an examination of the influence of forces and interactions on particle motion and wave propagation. t) at +~ 2m [as(x.9 . t) obtained in Problem 1 in the form "'(x. 0) chosen so that the absolute value of the overlap integral yeO) = f: "'i(X)"'2(X) dx is very small. A highresolution neutron interferometer narrows the energy spread of thermal neutrons of 20 meV kinetic energy to a wavelength dispersion level of dAtA = 10. t) which corresponds to classical motion of the wave packet. Free Particle Motion. as the wave packets overlap? Justify your answer. Plot I"'(x. t) and show that it satisfies the quantum mechanical HamiltonJacobi equation. t) /2 as a function of x for several values of t. Over what length of time will the wave packets spread appreciably? .24 Chapter 2 Wave Packets. Willi yet) Iincrease in time.
but wave mechanics. however. z.2 V2 t/J(r.35) to (2. Before going into mathematical details. 1 1.1) must.. We will adopt (3.1). the Wave Function.. (3. but of IAn excellent modern textbook on wave mechanics is BialynickiBirula. and Kaminski (1992). y.35) to the case of motion of a particle of mass m in the field of force represented by a potential energy function V(x. t) + V(x. we must examine the properties of this equation. is the equation if/. it would seem wise. The Wave Equation and the Interpretation of t/J. come from agreement between predictions and experiment. As the main tool for extracting information from the wave function.CHAPTER 3 The Schrodinger Equation. We are in the paradoxical situation of having obtained an equation that we believe is satisfied by this quantity. Cieplak.39) for a constant potentialsetting t/J = eiS11i_now implies the equation as [VS]2 if/. More will be said in Chapter 15 about the relation between wave mechanics and classical Hamiltonian mechanics.. It includes an extensive bibliography with many references to the original quantum mechanics literature. remains one of the simplest paradigms of the general theory. A straightforward generalization of the free particle wave equation (2. y. Hermitian operators and their algebraic properties are characterized. this theory will be generalized to encompass many different and more complicated systems. supplemented again by a quantum mechanical term proportional to h. The advanced reader can regard this chapter and the next as a refresher of the concepts and principles of wave mechanics.+ .V2 s + at 2m 2m Vi(x Y z t) = 0 ' . In subsequent chapters.1) Schrodinger advanced this equation on the grounds that the same reasoning that led from (2. Z. t). . of course. at/J(r. which is dependent on the position r and possibly also on the time t. which is the branch of quantum mechanics with a dynamical law in the explicit form (3. t) (3. identifying it as a probability amplitude. The existence of the factor i in this equation is of crucial importance and generally requires that the function S be complexvalued. and Operator Algebra In this chapter we study the properties and behavior'of the Schrodinger wave function as a function of space and time.2) This looks like the HamiltonJacobi equation of classical mechanics in the presence of forces. Hence. 25 . to attempt to say precisely what t/J is. . The ultimate justification for choosing (3. t) at = _ 2mf/.1) as the fundamental equation of nonrelativistic quantum mechanics for particles without spin and call it the wave equation or the timedependent Schrodinger equation. t)l/J(r.
Exercise 3.2.4) has the form of (3. .1. owing to Gauss' integral theorem. if it is possible to define a probability current density j. and the complex conjugate of (3. again for real V. Here C is a constant.(V0/1 *)0/2] =0 (3. as was already explained in Chapter 1.1) by 0/ on the right. A relation of the form (3. Multiply (3. [0/1 *V0/2 . Generalizing (3.6) is made. How can this statement be made precise? Obviously. The consistency of this probabilistic interpretation of the wave function requires that if the probability of finding the particle in some bounded region of space decreases as time goes on.3) if the identification p = j Co/*o/ (V0/*)0/] (3. which together with the probability density p = 0/*0/ satisfies a continuity equation op at+ V • 0 'J= (3. since probabilities are real and positive.3) exactly as in the case of the conservation of matter in hydrodynamics or the conservation of charge in electrodynamics.26 Chapter 3 The Schrodinger Equation.1) on the left by 0/*. and Operator Algebra having so far given only a deliberately vague interpretation of its physical significance.4). we were careless when we used the phrase "probability of finding the particle at position r. The resulting equation 0(0/*0/) at + ~ V . we now try the interpretation that I0/1 2 is proportional to the probability that upon measurement of its position the particle will be found in the given volume element. the quantity 0/ itself cannot be a probability. The probability interpretation of the 0/ waves can be made consistently only if this conservation of probability is guaranteed. Exercise 3. [o/*Vo/ . the Wave Function. the next best guess is that the probability is proportional to 10/1 2 . All hopes we might have entertained in that direction vanished when 0/ became a complex function. The probability of finding the particle in some finite region of space is then proportional to the integral of 0/* 0/ over this region. o/~: 0/2) + 2~i V . a( Derive (3. prove that. Hence. then the probability of finding it outside of this region must increase by the same amount. subtract the two equations and make a simple transformation. In the face of this dilemma. Of course. We have regarded the wave function as a "measure of the probability" of finding the particle at time t at the position r.5) = C ~ [o/*V 0/ 2ml (3. all we can speak of is the probability that the particle is in a volume element d 3 r which contains the point r.1).(Vo/*)o/] 2ml = 0 (3. the square of the amplitude of the wave function.4). This requirement is fulfilled.3) can easily be deduced from the wave equation for realvalued V. Note that it depends on V being real.7) if 0/1 and 0/2 are two solutions of (3." Actually.
9) expresses the simple fact that the probability of finding the particle anywhere in space at all is unity. Functions ljJ for which JljJ*ljJ d r exists are often called quadratically integrable.6a) and normalize the wave function by requiring (3. Is this a reasonable result? So far. 3 . we may set C = 1. but in principle the particle's motion can be tracked even after it is "absorbed. (3. J =  Ii m p = vp  is obtained. Whenever this integral exists. requires generalization of the quantum mechanical formalism (see Chapter 21).5a) j = ~ [ljJ*V ljJ 2ml (V ljJ*)ljJ] (3. we will assume that this normalization has been imposed." The true creation or annihilation of a particle. Since (3. t) is defined by 1/1 probability current density can be expressed as j = eiS'f> as in (2. we have only assumed that p and j are proportional to ljJ and 1/1*.22) the relation . Complexvalued potentials are used to simulate the interaction of a particle (e.36). and the limits of integration will usually be omitted.9) provided that the integral of ljJ*ljJ over all space exists (or is. Calculate p and j for a wave packet of this form and show that for free particle motion in the approximation (2.The Wave Equation and the Interpretation of 1/1 27 We have emphasized that V must be realvalued if conservation of probability is to hold.3.rl/Jo(r) where I/Jo(r) is a real amplitude modulating the plane wave. all spatial integrations will be understood to extend over all space. From here on.8) = p V(Re S)/m Exercise 3. an element of the space L 2 ). Show that if S(r. in mathematical terminology.4. as in nuclear beta decay. Equation (3. 0) = eik.. Assume that the function <f>(k) in expression (2. the (3.18) is real and symmetric with respect to k.g. Exercise 3. Show that I/I(r.1) determines ljJ only to within a multiplicative constant. The imaginary part of such an optical potential serves as a shorthand description of the absorption of the particle by the manybody system. a nucleon) with a complicated manybody system (nucleus). without having to give a detailed account of the dynamic process. unless otherwise stated.
This understanding will save us laborious repetition. Alternative methods for normalizing them are discussed in Sections 4. but very broad wave packet that contains very many wavelengths. of course.3) gives (3. However. It is an idealization of a finite. and state may be used interchangeably. because it represents an infinite amount of energy.12) This equation can be interpreted as stating that probability is similar to a fluid that flows with velocity v = V(Re S)/m. Examples will occur presently. We shall depend on the use of such wave functions. In other words.8) that one of these is simply the continuity equation ap + V . and if the probability current density falls to zero faster than r.6). t) is assumed to contain all information about the state of the physical system at time t. we may use only (3. the infinite plane matter wave is no more objectionable than an infinite electromagnetic wave that is also unphysical.9) holds. which describes a particle that is equally likely to be found anywhere in space at allhence. quadratically integrable.4.9) are as important as normalizable ones. If these limitations are recognized. This irrotational flow (V X v = 0) is known as Madelung flow. the terms wave function l/J. whatever conditions l/J must satisfy for a certain integral to exist will be assumed to be met by the particular wave function used. mathematically this patently unphysical object is extremely convenient.2 as r becomes very large.11) The coupled equations of motion for p and the phase function ReS are easily obtained as the real and imaginary parts of (3. it will remain normalized at all times. [p V(Re S)J at m = 0 (3. the Wave Function. we may'speak without danger of confusion of a normalized state if (3. and we must then speak of relative rather than absolute probabilities.9) does not exist. it is well to emphasize that throughout this book all integrals that are written down will be assumed to exist.10) which simply tells us that the wave equation guarantees the conservation of normalization: If l/J was normalized at t = 0. r .13) .3 and 4. It follows from (3. The second equation of motion is aCRe S) + [V(Re S)]2 + V + h [(Vpf _ V pJ at 2m 4m 2 p 2 p 2 2 =0 (3. One further comment on terminology: Since l/J(r.5a) of the probability density and the definition of the function S to express the wave function as (3. rigid insistence on dealing only with finite wave packets would prevent us from using such simple wave functions as the infinite plane wave e ik.28 Chapter 3 The Schrodinger Equation. Wave functions that are not normalizable to unity in the sense of (3. The former appear when the particle is unconfined. and Operator Algebra If the integral (3. In light of the foregoing remarks.9) exists.2). It is tempting to use the definition (3. Owing to the simple properties of Fourier integrals. with zero absolute probability in any finite volume. then Gauss' theorem applied to (3.5) and (3. and in principle they can always be avoided by the use of finite wave packets. Thus. When the integral (3.
The situation is analogous to the one encountered customarily in the use of probability concepts.12) and (3. In quantum mechanics the term expectation value is preferred when it is desirable to emphasize the predictive nature of the theory and the fact that the behavior of a single particle or physical system is involved.52. Whatever conceptual appeal Eqs'. Z.2 V2 yp yp (3. is defined by (3.13) may have is offset by the abstruseness of the quantum potentials and the complexities of the Bohm approach for manyparticle and relativistic systems. For instance. whereas in the application to quantum mechanics continuous probability distributions such as p = I f/J(x. or the same coin many times in succession. provided that the last term is regarded as a quantum potential that must be added to the ~onventional potential energy V. the elements of probability theory are reviewed in Appendix A. but the experimental verification of the statement "the probability for tossing heads is 48 percent" requires that we toss many identical coins. as we will see in Section 3. Section 2.13) or (3. y. David Bohm" and others advocated a more realistic and deterministic. These probabilities refer to the particular coin and to the individual tossings. (3.9).48 and that of tossing tails 0. There it is assumed that the events are discrete. (3.14) as a classical HamiltonJacobi equation for the dynamics of a particle. 2 Since the interpretation of the wave function is in terms of probabilities.13) is realvalued. or the center of the wave packet. (3. see Cushing (1994). or ontological.15) The expectation value of the position vector r. . The summations in the formulas for expectation (or average or mean) values must then be replaced by integrations. Although the wave function must be regarded as describing an ensemble representing a single system. the predictions for the state can be compared with experimental trials only by the use of a large number of identical and identically prepared systems." 2. interpretation of quantum mechanics than the one adopted in the present text. the average or expectation value of the coordinate x. Taking his cue from these nonlinear equations. it is not unreasonable to think of Eq. which is a random variable.14) Since everything in Eqs.2. and 48 percent of a very large sequence of trials are expected to yield "heads.12) and (3. For a state normalized according to condition (3. the probability of tossing heads with a slightly unsymmetrical coin might be 0. is (3. which are so classical in appearance.2 Probabilities in Coordinate and Momentum Space 29 or + ilt il(Re S) {V(Re S)}2 2m +V=o 2m 11. Probabilities in Coordinate and Momentum Space. t) 12 are common.16) 2Por a fuller and more sympathetic assessment of David Bohm's approach fo quantum mechanics.
] For a general motion in a potential. (2. the discussion in Chapter 2 suggests that knowledge about the momentum of the particle is contained in the Fourier transform cf>(p. t) = (27T~ J If!(r.19) with respect to t: iii acP (p. t).20) and must be worked out from the wave equation or its partner in momentum space.19).2m V li  2 + VCr) If!(r. r ) [.17) The reason for writing (f) in the clumsy form J If! * flf! d r will become clear when 3 less elementary expectation values make their appearance.1).iIi P . r )d3r )3/2 (3. (2. r)d r at 3 2 or iii acf>(p. the time development of cf> (p. t) is more complicated than (3. For the special case of a free particle we know from Chapter 2 how a plane wave of momentum p = lik develops in time and that cf> (p. we rely on the explicit notation cf>(p) and cf> (k) to remind us of the difference between them.18) in kspace are recovered by the substitutions In these equations we have used p rather than k p ~ lik. cf>(k. t) at = (27T1i )312 1 Jexp(ih p . O)e iw(p)t (3.30 Chapter 3 The Schrodinger Equation. At this point. r 3 (3. t) by the reciprocal Fourier transforms (see Appendix.(27T1i )312  Jcf> (p. which is related to If!(r.i p .17).19) = pili as the conjugate Fourier variable. it is not obvious how to set up the calculation of expectation values of momentum or kinetic energy or any other physical quantity that is not expressed simply in terms of the position coordinates. Section 1): 1 If!(r. most easily by differentiating (3. t) d r ] 3 . Since we are assuming that the physical state of a particle at time t is described as fully as possible by the normalized wave function If!(r. and (2. We emphasize once more that the functions If! in coordinate space (or position space) and cf> in momentum space are both equally valid descriptions of the state of the system.20) [In Eqs. which is a parameter. t) = at 1 (27T1i)312 Jiii alf!(r. the other can be calculated from (3. t) exp(i) d p hp . (2. Rather than introducing two different symbols for the functions describing the state in momentum space (p) and in kspace. and Operator Algebra An arbitrary function of r has the expectation value (3.16). t) . 0) was denoted simply as cf>(k).16). and (2. The latter equation is obtained from the wave equation in coordinate space (3. t) exp ( . the Wave Function. but the analogous relations (2. t) = cf> (p. t).18). Given either one of them.18) or (3.18) and cf> (p. Both If! and cf> depend on the time t.17). t) exp (.
t)j2 d r 3 3 (3.21) by c/J*.21) "h "h p' . ~ J c/J*(p.24). t)c/J (P. and the result is a partial differential equation in momentum space: iii. and it is fully equivalent to (3.24) This result should not be surprising because it follows directly from the Fourier transforms (3.5. c/J(p) according to (3.23) Exercise 3. t) 12 d 3p is the probability of finding the momentum of the . t) = p2 c/J (P. and we are left with a delta function. and if we can write exp ( k p' r)v(r) = V(iIi. and the complex conjugate of (3. vanishing sufficiently fast at large distances. This observation. t)d3p = 0 ot (3. If we multiply (3.21) by'c/J. Equation (3. This in turn makes the pintegration a trivial operation.19) are then used again. r This mixed integraldifferential equation for c/J is the wave equation in momentum space. r )V(r) exp(i) c/J(p'. t) (3.2 Probabilities in Coordinate and Momentum Space 31 The first term of this last expression can be transformed by integrating by parts twice and assuming that l/J is subject to boundary conditions.21) and removing the differential operator with respect to p from the integrals. the state tends to be one of rather definite momentum and similar to a plane wave. subtract the tV/O equations. If the Fourier transforms (3. leads us to the assumption that 1 c/J (p.22) Substituting this into (3. Section 1. y.18) and (3. It can be transformed into a differential equation if V is an analytic function of x. oc/J (P. after renaming some variables of the integration.24) then guarantees that this normalization is preserved at all times.1). we can now perform the rintegration. Work through the steps leading to (3. t) ot 2m • 1 + (27T1i. t) + V(ili. so that the surface terms can be neglected. oc/J ~~' t) = f c/J (P.10) or (3.18) and (3. c/J is automatically also so normalized. t)d3r d 3p' (3. Vp)c/J (p.19) that J Ic/J (p. after a little rearranging we obtain iii. in conjunction with the conservation property (3.23). using the Appendix. we obtain.18) is the amplitude with which the momentum p is represented in the wave function l/J(r). If c/J is strongly peaked near a particular value of p.V p ) exp ( k p' r) (3.)3/2 J exp (i p . Since exp(k p' r) represents a state with definite sharp momentum p. and integrate the resulting equation over the entire momentum space (p).25) This equation implies that if l/J is normalized to unity for any t. z. t)j2 d p = J Il/J(r.
g(p). and that 2 I 1> (p. The terms coordinate representation and momentum representation will then be seen to be apt for referring to these two forms of the theory. The probability distributions I t/J(r. The probability interpretation of the momentum wave function 1> allows us to construct the expectation value of the momentum for any state: (3. t) d3rl (3.28) . As Chapters 9 and 10 will show.27) is obtained. t) 12 and 11> (p. r )1>(P. a precise statement and mathematical proof of which will be given in Section 10. any equation can be transcribed into the momentum representation at any desired state of a calculation. t) d r d p 3 3 After liberal interchange of differentiations and integrations. t)g(~ v)t/J(r.26) and write (p) = II II t/J*(r. the desired result (p) = I t/J*(r. the covariance (Appendix. In quantum mechanics. (g(p» = I g(p) I1>(p.32 Chapter 3 The SchriJdinger Equation. but among the infinitely many possible representations no special status will be accorded them. and use of (3. in classical analytical dynamics the symmetry is primarily a formal one.5. we substitute the complex conjugate of (3. the statistical properties of these two physical quantities are determined by the functions t/J and 1>. especially in the Hamiltonian form. t) d 3 r d 3p t/J*(r. More generally. for q and p together determine the state at time t. The symmetry of rand p which the present section exhibits is also evident in classical mechanics. However. and classical physics tacitly assumes that they can both be determined simultaneously and independently with perfect precision. t) ~ Vt/J(r. the Wave Function. t)1 2 d 3p = I t/J*(r.19) into (3.26) In order to derive an expression for this expectation value in terms of t/J. t) 12 are subject to the Heisenberg uncertainty relation.18). the same technique gives for any (analytic) function of momentum. respectively. where (generalized) coordinates q and momenta p stand in a similar reciprocal relationship. t) d r 3 (3. r) p 1>(P. In general. Since these functions are one another's Fourier transforms. it is advantageous to regard both ofthese descriptions as two special representations of a general and abstract formulation of quantum mechanics. Wave mechanics in coordinate space (t/J) and wave mechanics in momentum space (1)) are thus seen to be two equivalent descriptions of the same thing. q and p are correlated and can no longer be measured or chosen entirely independently. Section 2) of the random variables representing position and momentum does not vanish. t) 1 is the probability density in momentum space. t) exp(k p . and Operator Algebra particle in the volume element d 3p in the neighborhood of p at time t. t) = ~ V exp(k p . Although most of the equations in wave mechanics will be written explicitly in the coordinate representation.
PY' z is (f(x.2 Probabilities in Coordinate and Momentum Space 33 where products of V's or iJ/iJx's are to be understood as successive operations of differentiation of the function on the right. Py. in fact. py. can be expressed iIi terms of coordinates and momenta. z). 1X(x. if we wish to calculate the expectation value of a mixed function like f(x. z) 12 dx dpy dz Similar considerations hold for other combinations of components of rand p. the coordinate function F(r. it follows readily that the expectation value of any function of x.29) If the physical system under consideration can be described in the language of classical mecharlics. py. we may conclude that ' (F(r.(p. z) 1 X(x.kPYY)dY (3. From this. p). but indeterminate Y. on which they act. and the virial. The equality of the two expressions on the right side of Eq. 7 v)t/J(r) and . (3. we postulate that (3. can easily be seen to be the probability density of finding the particle to have coordinates x and z. tW d 3 r = f q. not even realvalued). we have not proved (3. t) d 3p (3. PY' z) = f f(x. p = xpx + ypy + zpz (which will be encountered in Section 3. As a result. the expectation value of any (analytic) function f(r) of position is given by (f(r) = f f(r) It/J(r.Heisenberg's uncertainty principle is an articulation of this inability. all quantities associated with it. t) d r 3 I (3. provided that the simultaneous specification of conjugate variables. p). t)F(r. such as kinetic and potential energy. p)q. If products like xpx do appear.6). The trouble here is that our analysis can never give us a wave function the square of whose absolute value would represent the probability density of finding the particle with coordinate x and momentum component Px. angular momentum. z) = (2~)1I2 f t/J(r.*(p. such as x and Px' is avoided. whereas the ycomponent of its momentum is equal to py. we carry out the Fourier analysis only part of the way and define a mixed coordinatemomentum wave function by the integral x(x. because no account has been taken of as simple a function of coordinates and momenta as F = r . The expression for this expectation value can be derived to a certain extent from our previous work. t) exp ( . (p) are Fourier transforms of each other in the sense of (3.31) generally. but we must keep in mind that classically the product xpx can be equally well written as Pxx.(p. whereas the quantum mechanical expectation values of xpx and pxx are generally different (and. For example. Conversely.31) q.31) for any pair of coordinate and momentum wave functions is equivalent to the following theorem: If t/J(r) and q. then for any function F(r. ~ v)t/J(r.18) and (3. t) d 3p However. t)f(ihVp)q.19). The question that then arises is how to compute the expectation value of an arbitrary function F(r. PY' z) 12 . t)F(ihV p. Py.*(p. p) = f {f t/J*(r.31) holds.30) The square of the absolute value of this function.
we have the identity Jo/f(r)o/z(r)d r = JcP f(p)cPz(P) d p 3 3 (3. TV) acts on o/(r).r F(ihVp. 0/1 and cPz in (3.6. Unless it is specifically stated that a given operator is not linear. respectively. as are mere multipliers. Tv)o/(r) = (27T~)3/Z Jei/lip.35) of the wave equation in momentum space. Tv)o/(r) d 3r (3.32) and (3. and Prove that if o/I(r) and o/z(r) are Fourier transforms of cPI(P) cPz(p). [The different subscripts are meant as reminders that. that is. An operator is a rule that maps every function 0/ (or cP) into a function Fo/ (or F cP). p)cP(p) are Fourier transforms of each other in the same sense.37) . p)cP(P) d 3p = (27T~)3/Z (3.33). cP 2(Px. Derivatives are obviously linear operators. An operator is said to be linear if its action on any two functions 0/1 and o/z is such that (3.33) These formulas hold if 0/ and cP go to zero sufficiently fast for large values of their arguments.25).] 3. vanish.32) and F(ihVp. a new function of the coordinates results.31) from the transforms (3. the Wave Function.34 Chapter 3 The Schrodinger Equation. the term operator will henceforth be reserved for linear operators. iot t) = ~ e r vmw 0/1 (Px) . F(r. Show that for a linear harmonic oscillator with V = mClix 212 the wave equations in coordinate and momentum space have the same structure and that every normalized solution o/I(X. so that surface terms. resulting from integration by parts. p)cP(p) J ei/lip. t mw (3.7. generally.35) are not Fourier transforms of each other and do not represent the same state. F is an example of an operator that maps one function into another. Exercise 3.34) which is a generalization of (3. operators important in quantum mechanics is the antilinear variety. Thus. Exercise 3. and Operator Algebra the momentum function F(ihV p.are arbitrary complex numbers.8. Operators and Expectation Values of Dynamical Variables. and similarly F(ihV p. characterized by the property (3.36) where A and JJ. When F( r. t) of the wave equation in coordinate space can be related to a solution Exercise 3. The only other category of. Most of the operators that are relevant to quantum mechanics are linear. Derive the equality of the two expressions on the right side of (3. p) turns cP(p) into a new function of momentum.r F(r.
Antilinear operators are important in quantum mechanics when timereversed physical processes are considered. The explicit form of the op~ erator evidently depends on whether the r/J or cP function is used to represent the state. (3. Construct some examples of linear and antilinear operators and of some operators that are neither linear nor antilinear. The expectation value of the physical quantity F is given by the formula (3.31) we see that with any physical quantity. The reasonableness of the operator assignments in Table 3.. the context will always establish the meaning of the symbols unambiguously.V i h rX. Exercise 3.V I ihV p X P p2 2m V(ihV p) T V E 2 _ h V2 2m V(r) h2 . The student of quantum mechanics must get used to the potentially confusing practice of frequently denoting the mathematical operators.g. it is not difficult to adapt to these conventions.3 Operators and Expectation Values of Dynamical Variables 35 Complex conjugation itself is an example of an antilinear operator.1 Physical Quantity Operator in Coordinate Representation r r Operator in Momentum Representation Position Momentum Angular momentum Kinetic energy Potential energy Total energy ihV p p p L !!:.1 presents some of the most important physical quantities or dynamical variables. when interposed between the wave function and its complex conjugate. as they should be if they are averages of physical quantities that are analogous to the familiar classical quantities of the same name. Table 3.V2 + VCr) 2m L + V(ihV p ) 2m 2 As the discussion in the last section makes clear. \Vhich is a function of coordinates and linear momenta.9. F above) as the physical quantities that they represent. by the same symbols (e. Table 3. and the operators that represent them. even in different representations. From Eq. as in most matters of notation. there is associated a linear operator which.1 is confirmed by showing that their expectation values generally satisfy the laws of classical me . Yet. upon integration gives the expectation value of that physical quantity.. the expectation values of all of these operators are real numbers.38) .
d3 r .1) and its complex conjugate. (3.36 Chapter 3 The Schrodinger Equation.n f = 2 af/J* aljJ in at . and Operator Algebra chanics. the time derivative of the mean x coordinate can be written as or (3. by use of (3. Using (3. We note here that p is the (kinetic) momentum my.6a) and integration by parts. by using the continuity equation.39) where the divergence term has been removed under the assumption that f/J vanishes sufficiently fast at infinity. which is the canonical momentum only if there is no electromagnetic vector potential. . dS = Iv (uV v 2 vV u)J3r 2 where the surface S encloses the volume V. A = O.f/J*V 2 aljJ] d 3 r + ax f [VljJ* aljJ _ ljJ* a(vljJ)] ax ax The integral containing Laplacians can be transformed into a surface integral by Green's theorem: 3 It is assumed that the last integral vanishes when taken over a very large surface S. the Wave Function.41) 3Green's theorem for continuously differentiable functions u and v: Is (uVw  vVu) . more generally.(p ) dt x f =.6. For example.40) which shows that for the expectation values of wave packets the usual relation between velocity and momentum holds. given by d . The time rate of change of (p) is. we finally obtain d m(r)= dt f f/J*:Vf/Jd n z 3 r=(p) (3.in ax 2m f ljJ* ax (aljJ) d a at ax 3 r d3 r [(V 2 ljJ*) aljJ . What remains is d dt (px ) =  f ljJ* av ljJ d r ax 3 or. The appropriate generalization in the case A=#:O will be dealt with in Section 4.
40) and (3.41). can be written as In the last step.44) This formula is of the utmost importance in all facets of quantum mechanics. The time derivative of the expectation value of any operator F. known as Ehrenfest's theorem. If (3. The potential energy is. . In particular. and their commutator. tation. (F) is constant in time for any state. use has again been made of Green's theorem. (3. this condition holds for a sharply peaked wave packet. The physical quantity F is then termed a constant of the motion and is said to be conserved. H. If the state f/J is such that (F(r) "'" F«r»).41) that (r) moves in time very nearly like a classical particle. both acting to their right. we may write the last result as ' iii ~ (F) = (FH . if F commutes with H and does not explicitly depend on t. it follows from (3. Compactly. which may be explicitly time dependent. are applied to a state'" (or c/J) is generally important. We can establish a general formula for the time derivative of the expectation value (F) of any operator F. According to Eq. I[F. of course.44) is also true in the momentum represen The order in which two operators F and H.3 Operators and Expectation Values of Dynamical Variables 37 Equation (3.1) in the form l iii af/J at = Hf/J (3. Exercise 3. written for expectation values.43) for the Hamiltonian (or total energy) operator. (3.HF) + iii\ ¥t) (3. looks like Newton's second law.44). and conforms to the correspondence principle. H] = FH [F.10. Show that Eq. H] = 0 HF I (3. ii 2 2m v + V(r) 2 (3. To this end we write the wave equation (3. assumed to be real.46) the two operators are said to commute.45) is a measure of the difference between the two products. and vanishing boundary surface terms have been omitted.42) by making use of the notation H = .
B] = AB . The commutator [A. as Dirac called the operators that represent physical quantities. The manipulation of commutation relations for various dynamical variables. (3. In commutator bracket notation.BA of two operators plays a prominent role in quantum mechanics. is most safely accomplished by allowing the operators to act on an arbitrary function. in quantum mechanics as in classical mechanics. Px] = [y. Px] = xpx .11. [B. as exemplified in the last section. expressing the quantum version of the law of conservation of energy for a system with a potential that does not explicitly depend on time (conservative forces). B]C + B[A. Prove that the operator Px commutes with a function V(x) of x only if V is constant. All [A + + B. BC] = [A. and generalize this result to three dimensions. Le. it is only necessary to examine whether it also commutes with V(x. for any operators the following elementary rules are easy to verify: [A. C]B [A. xP y = pyX. All three components of momentum commute with V only if V is constant. C] = A[B. the Wave Function.~ ~ X)f = ax ax 1 1 ili[X af . C] + [B.49) All other products formed from Cartesian coordinates and their conjugate momenta are commutative. p y ] = [z.Pxx = yPy . pz] = iii 1 I (3. C] C] = [A.Pxx = iii 1 (3. [A. C]] + [C. (3. which is removed at the end of the calculation. Hence. C] + [A. C] [AB. B] [A. Exercise 3. Bll + [B. C] [A. and so on. linear momentum is conserved only if no external forces act upon the system.z = iii or. such as the pairs x and Px' Y and PY' z and pz. xy = yz. The most important commutators are those involving canonically conjugate variables. [C. f. xpz = pzX.pyy = zPz . z). (H) is constant in time.38 Chapter 3 The Schrodinger Equation. Since p commutes with the kinetic energy operator p 2l2m.50) =0 .P.. B + [B. A] = 0 C] = [A. B] + [A.Pxx)f = (x ~ ~ . in commutator bracket notation. and Operator Algebra Since H commutes with itself. Commutators and Operator Algebra. in the coordinate representation. (xpx . A] = 0 [A. y. Thus.47) The identity operator 1 may be omitted and the fundamental commutation relations for this and the other canonically conjugate variables may be written as XPx . 4. Linear momentum p = mv is a constant of the motion if p commutes with H.a(xf)] = ilif ax ax from which we infer the fundamental commutation relation [x.48) I [x.
M = B + = B. [Lx. + P. dA ] = [A. we get the identify + A 2! [A. [A. Exercise 3. B] = nAnI[A.)] y. f(A)]] and so on.Jrove that.52) if f(x) can be expanded in a power series of the operator x. Bm + .51) Note the similarity of the process with differentiation.g. Apply to the special case A = x. 2 A 1! [A. B] [A. defined by the power series eA Consider the function =l++++·· 1! 2! 3! A A2 A3 of a real number A. (3.12. [A.14. observing that :~ d dAf2 2 = Af(A) . . Show that each component of orbital angular momentum commutes with the kinetic energy operator. =0 Exercise 3. f(x)] n df = i dx 1 (3. B] (3.. Rrove that [Px. Another operator relation that is frequently used involves the function e A .f(A)A = [A.4 Commutators and Operator Algebra 39 Exercise 3. e. Since f(O) eMBe.53) as expected from correspondence with classical mechanics. [A. B]. B]] + A 3! [A. 3 [A.'lP 2~ (p. + P. B] [An.54) If the operators satisfy the relation ({3: constant) [A.13. and make a Taylor series expansion of f(A). [A. B n] = nBnI[A.55) . [A. for a positive integer n. If A and B are two operators that both commute with their commutator. B]] = {3B (3. Prove that d . f(A)] df = [ A. B = Px..(L) dt = (r X VV) = (r X F) (3. 11 = [ypz ..
B] + A [[A. .BJ/2 (3. B]] 2! 2 = G{ A A + B + 1! [A. Exercise 3. B]. [A. B].59) Equations like (3.16. B].40 Chapter 3 The Schrodinger Equation.L 1 (3. and Operator Algebra the righthand side of (3. B] = f.:.57) leading from (3. If [A. In particular.. the Wave Function.63) .. then eAeB = eA+ B +[A.15. the simple law eAeB = e A + B is generally not valid for operators. } G + . If we suppose that AU) is an operator that depends on t.54) caution us against the indiscriminate use of wellknown mathematical rules. f3 f3 (3. Show that if A and B are two operators that both commute with their commutator [A. according to which in the equation the additional term F(A..62) N N N N N lim N_oo k=l f + ~}kl A'(t) N {I + ~}Nk where we use the notation At(t) = dt dA (3.58) Exercise 3.56) to (3.60) dG = { A dA B] + A [A. B] = y B (y: constant) show that eAABe. + ~ smh A~vr. B] 2! . the derivative of eA(t) may be expressed as d dt eA(t) =~ dt = lim N_oo {I + {I A(t)}N = lim N_oo f {I + ~}Nn n=l A'(t) {I + ~}nl N (3.L: constant) [A. B) is expressible as a generally infinite sum of multiple commutators of A and B. } (3. 2 = eAAe AB and prove that + .54) can be summed to give e BeAA AA [A. = B cosh A~vr.61) Another operator identity with many useful applications concerns differentiation of an operator that depends on a parameter t.:. when the variables are noncommuting operators. but the formula is general and no commitment as to the nature of the parameter is required. Consider the product G(A) A + B + 1! [A.56) A special case arises if (f.AA = eAoy B (3. Its place is taken by the CampbeUBakerHausdorffformula. The parameter could be the time t.
..! [A(t)..! 3!" [A(t) A' (t)]] . lim Jo = eA(t) = ( 1 ..y = ti. (3...n 1 or y = N' k ti.. defined as ti.... The operator formalism provides us with the tools to construct formal solutions of the timedependent SchrOdinger equation simply and efficiently.66) The simplicity of this formula is deceptive.iVt11i • It is fair 4For a compilation of many applications of (3. Stationary States and General Solutions ofthe Wave Equation..42) such that at t = 0 the solution agrees with a given initial state "'(r.. e. see Wilcox (1967).A' (t) .62) can be turned into an integral over the real variable x. I ()N r exA(t) A'(t)exA(t) dx 1 + ~ I N Jo r eyA(t) A'(t)eyA(t) dy N.54).. the function F(ll. Differentiate with respect to the variable Aand derive a firstorder differential equation for the function F.42) may be written as "'(r.)N +~ N and finally.!.[A(t) = {:! A'(t) + . the solution of (3. 5.[A(t) A I (t)] dt 11 2!' + 1.5 Stationary States and General Solutions of the Wave Equation 41 The sum in (3. we obtain !!:.. The fundamental mathematical problem is to obtain a solution of Eq.64) with (3..17. t) = exp (  i Ht) "'(r. at = H'" (3. 0)..64).x = Ii = N ti. N N so that d eA(t) = lim • dt = N. .iTtlli e.1. A' (t)] + ..) = :t Prove the identity (3. 0) (3. d eA(t) dt fl 0 exA(t) A'(t)exA(t) dx L I (3. [A(t).. }eA(t) Exercise 3.64) by a second method. Then solve this equation. } (3. and performing the x integrations term by term. A' (t)]] + .. ift a". 4 Combining (3. and eiHtlli cannot be replaced by a product of operators.64) eYA(t) A'(t)eyA(t) dy eA(t) which is the desired result.42). eA(t) = eA(t){l.. The Hamiltonian is usually the sum of a differential operator (kinetic energy T) and a function of position (potential energy V).k = .. If the system is conservative and the Hamiltonian operator H does not depend on t. starting from eM(t).65) [A (t).
69) the functions o/(r.67) or.66) into a more explicit and calculable form for different dynamical systems. This justifies the occasional designation of the spatial factor o/E(r) as a timeindependent wave function.1). the subscript E will be dropped. Here we make the more general observation that any initial state represented by an eigenfunction o/E(r) of the Hamiltonian operator. This equation is also called the timeindependent Schrodinger equation in contrast to the timedependent Schrodinger equation (3. for economy of notation. (3.10). and this is possible only if the eigenvalues E are real numbers. t) and o/E(r) differ only by a timedependent phase factor of constant magnitude.71) . the expectation value of any physical quantity that does not depend on time explicitly is constant and may be calculated from the timeindependent wave function alone: (A) = f o/*(r. corresponding to an eigenvalue E. in the (timeindependent) SchrOdinger equation for a function that is in general entirely different from o/(r. If the initial state is represented by an eigenfunction o/E(r) of H.E*)tJ o/E*(r) o/E (r) must satisfy condition (3. Using a different symbol might spare the student a little confusion now but would make the reading of the literature more difficult later. t) is sanctioned by usage. t) = exp [  k (E .68) known plainly as the Schrodinger equation.66). Use of the symbolo/E(r). o/(r. the timedependent wave function becomes. according to (3.66). If E is real. the probability density P = o/*(r. the timedependent factor in (3. care will be taken to distinguish the two functions by including the variables on which they depend. t) o/(r. t)d 3 r = f o/E*(r)Ao/E(r)d 3 r (3. Hence. (3. in the separable state (3. Whenever necessary. t)Ao/(r.70) .1) can be constructed by superposition of these separable solutions. (Often.69) rives its great significance from the fact that more general solutions of (3.69) are to be physically acceptable.69) of the timedependent SchrOdinger wave equation def(H)o/E(r) = f(E)o/E(r) k Et) o/E(r) (3. more explicitly. they must conserve probability. Hence. (3. If the states represented by (3.69) is purely oscillatory. gives rise to a particularly simple and important class of solutions (3. and Operator Algebra to say that much of the rest of this book will be devoted to turning Eq. The particular solution (3.) For such a state.42 Chapter 3 The Schrodinger Equation. t) = exp ( because for any analytic function f of H. the Wave Function. or merely o/(r). Eigenfunctions o/E(r) and eigenvalues E of H are defined by the eigenvalue equation (3.
7). which.that the particular solutions rf/(r.5 Stationary States and General Solutions of the Wave Equation 43 From (3. Such an interpretation conforms with the uncertainty relation. JiEAt . t) 12 and all other probabilities are constant in time. Since a stationary state (3. We thus see. We have already encountered one special example of a stationary state.69) are said to represent stationary states. j = 0 for such a state. There is no uncertainty in the momentum and in the energy of this plane wave state. This is to be understood in the sense that any determination of the energy of a particle that is in a stationary state always yields a particular value E and only that value. But conservation of energy remains valid. Note that although the state is stationary.73) for any two stationary states with energies E 1 "* E and suitable boundary conditions. t) = exp (  i Et) rf/E(r) (3. The simple regularity conditions that physical considerations lead us to impose on the wave function give rise in many cases to solutions of (3. Exercise 3. i. wave functions of the form (3.18. the plane wave rf/(r. drastically altered. The probability density p = 1 rf/(r. exist only for certain discrete values of E. of course. E being the definite value of its energy and not only its expectation value. For this reason. the properties of momentum are.68) .. This is the state of free particle of momentum p and consequently of energy E = p2/2m.::= Ii. Stationary states are of just such a nature in view of the constancy of Irf/I in time. r) This wave function represents a state that has the definite value E for its energy.e. The observed discrete energy levels of atoms and nuclei are approximated to a high degree of accuracy by stationary states.68). but there is. the particle so described is in motion and not stationary at all. The values of E for which (3.69) have intriguing properties. a complete ignorance of the position of the particle and of the transit time of the particle at a chosen position.69) has the definite (circular) frequency l1J = E/Ii. t) = exp (  i Et) exp(i p . Using Eq. V . prove the orthogonality relation I rf/i(r)rf/2(r)d3 r = 0 2 (3.72) showing that the eigenvalue E is also the expectation value of the energy in this state. Also. it follows that a stationary state is a state of welldefined energy.67) we obtain J (3. instead of depending continuously on the parameter E. (3. correspondingly. When conservative forces are introduced. which implies that a quantum state with a precise energy (JiE = 0) is possible only if unlimited time is available to determine that energy. as does the relation according to which the energy equals h times the frequency.
The energy eigenvalue E is said to be degenerate in this case. The situation is even more pronounced in the case of the socalled unstable particles.74) Here as elsewhere in this book the summation may imply an integration if part or all of the energy spectrum of E is continuous. . the summation in (3. as will be shown in Chapter 19. whereas the lifetime of the state is roughly 10. they live for about 1011 periods. Thus. about 10. during the lifetime of the state some 107 oscillations take place. for the period associated with a typical atomic state is of the order of the period of an electron moving in a Bohr orbit. and their lifetimes :. are excluded. has more than one linearly independent physically acceptable solution for the same eigenvalue E. The point at issue here is that no excited state of an atom is truly discrete.)l2m. Furthermore. More general solutions of (3.67) or (3. The eigenfunctions are exp(i p' r/fi. (3. i. + p. stationary.75) must be regarded as an integral. If there is a lowest energy level. because for any positive value of E.. The corresponding eigenfunctions are linearly independent. t) =~ CE exp ( k Et) exp(k px) (3. The careful phrasing concerning the approximate nature of the connection between energy levels and stationary states is intentional. since plane waves propagating with the same wavelength in different directions are linearly independent.15 sec. (3. + p. if paradoxical.. p can assume two values. 10. It is then not sufficient to designate the eigenfunction by r/JE(r). which correspond to imaginary values of p..68). and Operator Algebra has solutions consistent with the boundary conditions are said to constitute the energy spectrum of the Schrodinger equation. In this case (3. Negative values of E. The exclusion of negative values of E on this ground is an example of the operation of the physically required boundary conditions.. we get the solution.10 sec. or permanent.. t) =~ CE exp (  k Et) r/JE(r) (3. r/J(r. All discrete higher energy levels are classified as excited states. If it were. the corresponding stationary state is called the ground state of the system. Summing over various allowed values of E. This spectrum may consist of isolated points (discrete energy levels) and continuous portions.74) becomes r/J(r.21 sec. Comparing this with the fact that during its history the earth has completed some 109 revolutions around the suna motion that has always been considered one of the stablestthe striking. p = ±Y2mE. Since the energy spectrum is continuous and E can be any positive number.75) There is degeneracy of degree two here. This occurs if the SchrOdinger equation.1) can be constructed by superposition of such particular solutions. we would not see the spectral lines that result from its decay to some other level by photon emission. For a free particle in three dimensions.e. each positive energy E has an infinite degree of degeneracy. 10. feature of the unstable particles is their stability! The stationary state wave functions are particular solutions of the wave equation. Hence. and one or more additional labels must be used to identify a particular eigenfunction. The excited states are nevertheless stable to a high degree of approximation.44 Chapter 3 The Schrodinger Equation. Their natural periods are:.) and the eigenvalues are E = (p. because r/J would become singular for very large values of Ix I. the Wave Function. such as the kaon or the lambda particle. The example of the free particle in one dimension illustrates these features.8 sec.74) may include the same energy value more than once..
00 to + 00. we get by integrating the Schrodinger equation from x = Xo . while the other one diverges.00. In many applications.e) = f "'o+8 Xo' fl. or both.l/J'(xo . The degeneracy is twofold. it represents a bound state. The Wronskian. • Exercise 3. for a periodic potential. oftwo such solutions of (3.2 [V(x)  2m E]l/J(X) dx As long as V(x) is finite.e to x = Xo + 8: l/J'(xo + e) . The potential function V(x) in the SchrOdinger equation is usually smooth and wellbehaved for most values of x. If such a singular potential function can be approximated as the limiting case of a regular one. as is the case for the free particle (V = 0) and for a particle in a periodic potential. l/J itself must also be continuous. In particular. SPor an analysis of the singlevaluedness condition for wave functions. it was assumed that the coordinate x is the usual Cartesian coordinate ranging from . in a forbidden band. for either x ~ + 00 or . if x = Xo is the point of singularity. see Merzbacher (1962). or both. for a periodic potential if the energy E is at a band edge. however. Sometimes we encounter systems that require the imposition of periodic boundary conditions on the solutions of the SchrOdinger equation. we commonly encounter potential functions that are discontinuous or that have discontinuous slopes at some isolated points xo. since the Schrodinger equation is a secondorder differential equation. this is the case for a particle that is constrained to move on a closed loop. the SchrOdinger equation in one dimension for an arbitrary potential V . if the energy is in an allowed band. Neither solution represents a physically acceptable state. only one solution remains finite. the wave function l/J and its first derivative l/J' must be continuous everywhere. l/J~) and l/JCf).2 d2l/JE(x) dx 2 + V(X)l/JE(X) = El/JE(x) (3. (b) One solution and its derivative approach zero as x approaches either + 00 or 00. and the boundary condition to be applied is the singlevaluedness of the wave function. 76). so there is no degeneracy. In the preceding discussion. V(x) is nonanalytic at isolated points. whether analytic or not. This is the case if E lies between discrete energy levels or. l/J~>l/JCf)' .l/J~)' l/JCf>.19.2m fl. The constancy of the Wronskian requires that the other solution diverges at large distance. such as a ring. 5 Generally. (c) Both solutions diverge at large distance. More specifically. . Prove that the Wronskian for two solutions of the onedimensional Schrodinger equation is constant. the allowed stationary states will have twofold degeneracy. Four possibilities can be distinguished: (a) Both solutions and their derivatives are finite at all x. is a nonzero constant for all x. Hence. (d) Finally. If a solution approaches zero for both x ~ ± 00 sufficiently rapidly. this equation implies that l/J' is continuous across the singularity. For instance.5 Stationary States and General Solutions of the Wave Equation 45 As a secondorder linear differential equation.76) has two linearly independent solutions for each value of E.
except for an arbitrary constant complex factor. Hence.e. Most of the analysis in this section can be applied directly to the wave equation (3. and Operator Algebra Thus. t). t) =~ CE exp (  k Et) l/JE(r) (3. l/J(r.1). l/J(r. if the SchrOdinger equation for an arbitrary potential yields solutions l/JE(r) that are complete in the sense that any initial state l/J(r. The probability current density.e. t) is a solution of (3. invariance under time reversal implies that.80) is called its invariance under time reversal.1) for all t: l/J(r. if l/JE(r) is a stationary wave function.20.68). t) = . 0) of physical interest may be expanded in terms of them. i. Hence.1) can be discussed without specifying V in further detail. t) in the form (3. Similarly. is that the wave junction and its slope must be matched continuously.= n . if the complex conjugate of (3.1) with respect to l/J(r.75) is not only a solution of the wave equation Z . It is based on the observation that.. the expansion coefficients in (3. we get iii al/J*~.. t) is also a solution.. the solutions of the SchrOdinger equation with real V may be chosen to be real. the Wave Function. One more important simple property of the wave equation (3. The behavior of the wave equation exhibited by (3. l/J(x.1) is taken and t is replaced by t.: VZl/J*(r. The Fourier transform of the timeindependent Schrodinger equation (3. 0). of particles. The latter is often referred to as the timereversed solution of (3. This equation has the s~me form as (3.1). must also be continuousan evident physical requirement. l/J*(r. 0) = 2: cEl/JE(r) E (3. the joining condition to be assumed at a singular point of the potential. t) (3.80) provided only that V is real. t) is a superposition of stationary states. if l/J(r.21) in momentum space. t) + V(r)l/J*(r. and show that if degeneracy is present.81) . Exercise 3. if V has a finite value.77) but it is the most general solution that can be of physical utility when describing the motion of a free particle. if there are no sources and sinks of probability. Fourier analysis tells us that for any given initial wave packet.aZl/J z at 2m ax (3.46 Chapter 3 The Schrodinger Equation.68) for stationary states of energy E is the integral equation (3. Hence. if the initial wave function can be expanded in terms of solutions of the Schrodinger equation. The generalization to three dimensions is straightforward. it must be real. l/J(x. For a stationary state.79) Thus. Prove the statements of the last paragraph.78) then we automatically know the solution of (3. it follows that if l/JE(r) is a nonaegenerate solution of (3. i. al/J In . l/J~(r) is also one.75) are uniquely defined. which is made up of l/J and l/J'.
EcPE(X) 2: (3.~ + [x.80) reduces to (3. .21. 2m + [xpx. . Furthermore. VJ 2 ~ + x[Px. p. Consider the equation of motion for the operator r .23). the momentum space energy eigenfunctions are then cP(p) = 8(p .85) which is a special case of the result in Exercise 3. According to (3.E }I>E(P) = 0 This equation implies that cPE(P) = 0 for all p. or 1 Z 1 Z dZcPE(X) _ 2m pAE(X) mw dp.8. 2m + 2m + 2m + V] p. Px] . the SchrOdinger equation in momentum space (3.22.~ + x[px. Exercise 3.47).] . In three dimensions. p. (3. iClE cPE(Px) = ~ e c. which may be written as (.21) is (3. VJ 2 .82) For a free particle. the boundary conditions are analogous: cP must approach zero for large values of IPxl. we obtain [xpx' H] = [xpx.81) is simply an algebraic equation. if Po is a fixed momentum with energy E = P6/2m.~ + x[Px. Since in one dimension there is no degeneracy for bound states. t) is the timereversed solution with respect to cP (p.: . p.6 The Virial Theorem 47 The corresponding timedependent momentum wave function that solves the wave equation (3. prove that the wave equation in momentum space is invariant under time reversal and that cP*( P. Exercise 3.] = iTt = pAx. t).Po) 2 2 (3.44). with V(x) = mw x 12. xpx . just as 0/ is required to tend to zero as Ixi ~ 00. A simple example will illustrate the principles of the last two sections. Show that for the stationary states of the harmonic oscillator.PxX = iTt. p. P. VJ = [xpx. H]) By applying several of the rules (3. d zTt dt (r . VJ p. it follows that for the energy eigenstates of the harmonic oscillator the Fourier transforms o/E(X) and cPE(Px) must be simply related to each other.84) which is similar to the SchrOdinger equation in coordinate space. pxl . = [x.83) For a linear h~rmonic oscillator.50) and the fundamental commutation relation (3. 6. except if p Z = 2mE. The Virial Theorem. For real V.o/E (Px ) vmw mw (3. p) = ([r .
. Exercise 3. (r F) T ~ 00.23. 2: i Exercise 3. p)t=o] T I = 2(T) .86) \~)  f l/J*(r· VV)l/Jd r 3 Exercise 3. Az) = AnV(x. Apply the virial theorem for a ~tationary state to the threedimensional anisotropic harmonic oscillator for which 1 VCr) = .87) are constant in time. the Wave Function. Apply the virial theorem to a central potential of the form V = Ar n .48 Chapter 3 The Schrodinger Equation. the same equation holes for (p .87) side tends to zero. VV) = (r· F) I (3.[(r . Show that r . p)t=.m(wTx2 + w~l + w~t) 2 Show that the same result could have been derived from the equality given in Exercise 3. and Operator Algebra Similar relations hold for the y and z components. for a stationary state (T) = n (V) 2 (3.89) Exercise 3.p ..86) does not matter. VV) = (r· F) For a stationary state all expectation values in (3. p) = \~) .25. y.88) are known as the virial theorem. we get ~ (r . Section 34.(r . and 2(T) = (r . r).24.· F.86) for the time average over a time T .. 6Goldstein (1980). using Euler's theorem for homogeneous functions. r = 3ili 1. and it follows that 12(T) = (r .87) and (3. and prove that in spite of the noncommutivity of the operators rand p. and express the expectation value of V for bound states in terms of the energy.6 we obtain from (3. If the expectation values in this expression remain finite as the lefthand (3. .26. p . r. their order on the lefthand side of (3. In a gas of weakly interacting molecules. If the potential has the property V(Ax. VV) = 2(T) + . departures from the ideal gas law can be related to the virial. which explains the terminology. Combining these results. z).88) The results (3.21.f l/J*r· {V(Vl/J) = VV l/J}d r 3 (3. Ay.(r . As in classical mechanics. show that.
f/1* r' . 2[12 ( xpx + PxX 0 ) ()  x 0 (px)] t (!!... = (X)o cos wt + (Px)o sin wt mw (b) Derive a secondorder differential equation of motion for the expectation value (T . .) f/1 ( r' + . the Wigner distribution function is defined as W(r' p') = 1 '(27Tfi)3 f e1p'.x~~ sinz wt mw + [ 2 (XPx + PxX)o 1 ] sin 2wt .. .(x).V).r"'r.z = (ax)o +. Show that z (ax).. Consider a linear harmonic oscillator with Hamiltonian (a) Derive the equation of motion for the expectation value (x).) d 3 r" 2 2 (r ll r ll . without explicit use of the wave function by applying (3.Problems 49 Problems 1. by repeated application of (3. For a free particle in one dimension. == «x .44) and use of the virial theorem. Verify that this reduces to the result of Problem 1 in the limit w ~ O.. show that the probability current density can be expressed without an interference term involving f/11 and f/1z. remembering conservation of energy. If the state f/1(r) is a superposition. (d) Work out the corresponding formula for the variance (apx).. calculate (X Z).Px)z + ~ tz and 3. (ax). 5. 4. and show that it oscillates. For a system described by the wave function f/1(r'). Integrate this equation and. (c) Show that (ax). calculate the variance at time t. f/1(r) = clf/1l(r) + czf/1z(r) where f/11 (r) and f/1z(r) are related to one another by time reversal.(x).)Z). as (x).(x). 2. similarly to the classical oscillator.. = (ax)~ cosz wt + (a. == (X Z).(x)o(PJo . Prove that the probability density and the probability current density at position ro can be expressed in terms of the operators rand p as expectation values of the operators Derive expressions for these densities in the momentum representation.44) repeatedly.. = (X Z).
respectively. Chapter 15. p'). rand p. p') over coordinate and momentum space are probability distributions in momentum and coordinate space. p') d3 r' = I~(p')12 d3~' ~p' and that the expectation value of a function of the operator p is (g(p» = II g(p')W(r'.) (a) Show that W(r'. is not a probability distribution (in phase space). p') d 3 r' d 3p' = 1 (d) Show that the probability density p(ro) at position r o is obtained from the Wigner distribution with 8 p (ro) '> fer) = 5(r  ro) 6. defined over the sixdimensional "phase space" (r'. p') (b) Verify that = (2~)3 ~*(p' I  p")~(p' + p")e2iP'· r 'ln d3p" . the Wave Function.50 Chapter 3 The Schrodinger Equation. 8Because it involves both rand p in the same expression. p') d 3p ' = !1/1(r')\2 and that the expectation value of a function of the operator r in a normalized state is = If f(r')W(r'. and Operator Algebra (In formulas involving the Wigner distribution. it is essential to make a notational distinction between the unprimed operators. which can take negative values.? (b) Prove that I (f(r» W(r'. and the real number variables. p') d 3 r' tfp' (c) Show that the Wigner distribution is normalized as I W(r'. the Wigner distribution is W(r'. the analogous question for the probability current density j is deferred until Problem 2. p'). which carry primes.I W(r'. the function W(r'. (a) Show that if ~(p') is the momentum wave function representing the state. p') 7 Although the integrals of W(r'. p') is a realvalued function. .
which for convenience is denoted by the same letter. their Eigenfunctions and Eigenvalues. generalized to include the continuous spectrum. The momentum operator is related to coordinate displacements. (4.3) for any two possible states !/Jl and !/J2' The momentum p is an example of a Hermitian operator. The formalism of nonrelativistic quantum mechanics is joined with the gaugesymmetric electromagnetic field. (F) must also be real. The central issue is the existence of complete sets of orthonormal eigenfunctions of selfadjoint (Hermitian) operators. " If (F) is the expectation (or average) value of a linear operator representing a physical quantity whose measured values are real. By (4. 1. It was shown in Section 3. Exercise 4. We have learned that every physical quantity F can be represented by a linear operator.CHAPTER 4 The Principles of Wave Mechanics This chapter concludes our account of the general principles on which wave mechanics rests.1. The property of being Hermitian is independent of the choice of representation.2) can be generalized (4. we should be ready for the application of quantum mechanics to spinless particles in one or more dimensions. Hermitian Operators. The expectation value of F is given by the formula (4. F.2) Operators that have this property are called Hermitian. At the end.1) expressed either in the coordinate or the momentum representation.1) this implies that for all !/J which may represent possible states we must demand that (4. to Prove that if F is a Hermitian operator. We now ask what general properties a linear operator must possess in order to be admitted as a candidate for representing a physical quantity. and we encounter the first examples of symmetry operations.3 that for the calculation of the expectation value (p) the momentum may be 51 .
[The momentum p is also represented by the differential operator (Ii/i) V if l/J satisfies periodic boundary conditions (Section 4.s over some mathematical distinctions that are addressed in more thorough presentations. The sum of two Hermitian operators is Hermitian.4).4) with (4. it is useful to define its (Hermitian) adjoint F t by the requirement (4. The identity operator 1. 2. Since the physical interpretation requires that operators representing measurable physical quantities must be selfadjoint.3). is Hermitian. are indicated briefly: 1. see Jordan (1969) and the references cited here. If F is nonHermitian. Their proofs. 3. The existence of the linear operator Ft for operators of the type F(r. condition (4.52 represented by Chapter 4 The Principles of Wave Mechanics ~ V if l/J vanishes l sufficiently fast at large distances. although it glosse. For a compact account. For a more recent discussion. F can be written as a linear combination of two Hermitian operators: (F + G) t (4. we have and hence F is selfadjoint. F + Ft and i(F . Ft =F (4. Hence. and if its adjoint exists and has the same domain as F.Ft) are Hermitian.5) Conversely. not necessarily Hermitian. If A is a real number.] Given an arbitrary. 1 A number of simple theorems about operators can now be given. see Ballentine (1990). . The adjoint of the sum of two operators equals the sum of their adjoints: = Ft + Gt . Comparing (4. But under the very same boundary conditions. we note that F is Hermitian if it is selfadjoint. if an operator F is Hermitian.4) where f and g are arbitrary functions (but subject to the condition that the integrals exist). it has become customary in quantum mechanics to use the terms Hermitian and selfadjoint synonymously. if not trivial. integration by parts gives Hence.6) lMany mathematically unimpeachable treatments have been published since the first appearance of von Neumann (1932). AI is Hermitian. which takes every function into itself. We shall follow this usage.3) is satisfied. p) can be verified by integration by parts. operator F.
The adjoint of a complex number A is its complex conjugate A*.Hermitian Operators. 6. A). how do we find and characterize those particular states of the system in which A may be said to have a definite value? If we make the convention that specific numerical values of dynamical variables (physical quantities) will be denoted by primed symbols (e. the adjoint of their product is given by (4. under what conditions will all these systems yield the same sharp and accurately predictable value A'? Phrased still differently. If F and G are two arbitrary operators. the generic term dynamical variable is often used for such operators. Thus. what kind of f/J corresponds to a probability distribution of the value of A that is peaked at A' and has no spread? .44) for the expectation value of an operator F.47) for x and Px is consistent with the Hermitian nature of the operators. all replicas of each other and each represented by the same wave function f/J. Exercise 4. but their product xpx is not.. Exercise 4. hence (4.3. making use only of the Hermitian property of H but not of its explicit form.7) with an important reversal of the order of the factors. Proof I But f/J*(FG)tf/J d 3 r = I (FGf/J)*f/J d 3 r = I (Gf/J)*Ftf/J d 3 r = I f/J*GtFtf/J d 3 r f/J is arbitrary.1 representing physical quantities are Hermitian. the same question can be phrased in physical terms: If we measure A in a large number of separate systems. (FG)t = GF (4.g. All the operators listed in Table 3.. From the timedependent SchrOdinger equation (3. Since the weak requirement of being Hermitian is sufficIent to establish the most important properties common to the operators that represent physical quantities.8) Corollary. because the two operators fail to commute (see property 4).42).7) follows. prove the equation of motion (3.2. x and Px are dynamical variables. Given any dynamical variable A. If F and G are Hermitian. It is not farfetched to suppose that the failure of x and Px to commute is connected with the uncertainty relation for x and Px and with the incompatibility of precise simultaneous values for the x coordinate and the xcomponent of momentum.g. Their Eigenfunctions and Eigenvalues 53 4. 5. Show that the fundamental commutation relation (3. A') to distinguish them from the operators representing physical quantities (e. The product of two Hermitian operators is Hermitian if and only if they commute.
The eigenvalues are the same in the coordinate or momentum representations. is Hermitian. The only definite values of A which a system can thus have are the eigenvalues A'.11) are sufficient to satisfy condition (4. and no other value. All eigenvalues of A belong to its spectrum.9) for any function of A. which consistency in notation would suggest. the expectation value. when the dynamical variable A is measured if and only if it is in a state represented by the eigenfunction o/A" We often say that the system is in an eigenstate of A. A'.9) for any function f of a dynamical variable. and A' is real. An eigenfunction of A is characterized by the fact that operating with A on it has no effect other than multiplication of the function by the eigenvalue A'. Its eigenfunctions o/E(r) or <f>E(P) represent the states in which measurement of the energy gives the certain and sharp values E. being a dynamical variable. and the real number A' is the corresponding eigenvalue. In particular. in which the outcome of every measurement of A is the same.12) is an eigenfunction of A. (We follow the customary usage of denoting the eigenvalues of H by E.35) in the Appendix. for f(A) = A we demand (A) (4.10) and for f(A) =A 2 we require that (A 2 ) = A. we must demand that.) Our conclusion is that a system will reveal a definite value A'. Hence. The Hamiltonian H representing the energy is an example of a dynamical variable. rather than H'.S4 Chapter 4 The Principles of Wave Mechanics In the particular state 0/. (f(A».11) According to Eq. the corresponding eigenfunctions are Fourier transforms of each other.11) expresses the vanishing of the variance (~A)2 of A: If 0/ is assumed to be quadratically integrable. . It may be noted that (4.2 = (A)Z (4. This follows from the linearity of Fourier transforms.12) A quadratically integrable function 0/ that satisfies (4. this condition implies for the special state in which A has a sharp value A' that The operator A.10) and (4. in such a state. If the system has a classical analogue. (A. Hence. (f(A» = f(A') = A' (4. of any function of A must be equal to f(A '). the operator is obtained from the classical Hamiltonian function of q's and p's by replacing these physical variables by operators. the last equation becomes from which it follows that (4. (4.
15) that if A 2 A.16) We conclude that eigenfunctions belonging to different eigenvalues are orthogonal.A 2 *) f l/J2*l/Jl d 3 r = 0 (4.2 v'f"=lKj2 (4.15) If l/J2 = l/Jh then 42 = A~.* = A. all the n eigenfunctions belonging to an eigenvalue A' must be linearly independent. Consider two eigenfunctions of A. Since the eigenvalues A' are real. Although any two eigenfunctions belonging to the same eigenvalue mayor may not be orthogonal. not considered "different. does. and hence A. We construct a new vector l/J2 as a linear combination of t{11 and t{12 and demand that f l/Jil/J2 If d r 3 =0 l/J2. This means that none of them may be expressed as a linear combination of the others. take the complex conjugate of (4. Proof. Suppose that Al/Jl = A'l/Jl and At{12 = A' l/J2 t{11 and where l/J2 is not a multiple of defined as l/Jb and where the overlap integral of l/J2 is f l/Jit{12 By interpreting d r 3 =K '* 0 l/Jl and l/J2 as vectors in a twodimensional space.1. . Eq.13) (4. their orthogonality. but a most important property of the eigenfnnctions.14) Multiply (4. is l/Jl and l/J2 are assumed to be normalized to unity. the two properties follow from the same simple argument. Owing to the Hermitian property of A. of course. it is always possible to express all of them as linear combinations of n eigenfunctions that are orthogonal and belong to that same eigenvalue (the Schmidt orthogonalization method). (A.15) is trivially satisfied if two different eigenfunctions belong to a particular eigenvalue of A. and multiply it on the right by l/Jl' Then integrate both equations over all space and subtract one from the other. l/J' 2  Kt{11 . we then see from (4. we can depict the Schmidt procedure in Figure 4.17) . Using this' result. and f l/Ji l/J2 d 3 r as their scalar (Hermitian inner) product. the two eigenfunctions are orthogonal in the sense that '* f t{12*t{11 d r = 0 3 (4." Rather..t{11 = A 2l/J2 (4. corresponding to the eigenvalues A. Their Eigenfunctions and Eigenvalues 55 The reality of the eigenvalues of Hermitian operators needs no further proof. to be at all interesting.13) bn the left by t{1i. which demonstrates again that all eigenvalues of A are real. Actually. also normalized.1/1. and A 2: At{11 At{12 = A. An eigenfunction that is obtained from another one merely by multiplication with a constant is. (4. t{11 and t{12.Hermitian Operators.14).
In writing (4. This orthogonal set of functions is by no means unique. any eigenfunction of A with eigenvalue A' will be expressible as a linear combination of 1/J1o I/Jz. the length of the projection. we speak of a repeated eigenvalue. Its existence shows. however. This new eigenfunction of A is orthogonal to I/JI' If there are other eigenfunctions of A which correspond to the same eigenvalue A' and are linearly independent of I/Jl and 1/J2. energy levels that correspond to repeated eigenvalues of the Hamiltonian H are said to be degenerate. and "'~.•• I/J~. In the case of the energy operator. The possible occurrence of repeated eigenvalues is not always made explicit in the formalism but should be taken into account when it arises. When the process is completed. have been normalized to unity. so that I/J~ is orthogonal to I/Jl and I/Jz. and so on. Twodimensional (realvalued) vector analogue of Schmidt orthogonalization. Any other eigenfunction of A can be written as a linear combination of orthogonal ones.19) are said to form an orthonormal set. Henceforth we will therefore always assume that all the eigenfunctions of A are orthogonal and that the eigenfunctions. Since (4. which are quadratically integrable. . therefore. otherwise we say that the eigenvalue is simple. this process of orthogonalization can be continued systematically by demanding that 1/J3 be replaced by I/J~.1. The The unit vectors represent the (generally complexvalued) functions scalar product of the vectors and equals K. that there is no loss of generality if we assume that all eigenfunctions of the operator A are mutually orthogonal. as it frequently does. If this happened. I/J~ ••.19) we are.18) the solutions of the equation (4.56 Chapter 4 The Principles of Wave Mechanics 1/12 Figure 4.18) and (4. a linear combination of 1/J1o I/Jz. and 1/J3. allowing for the possibility that two or more eigenvalues of A may be the same. "'I "'2 "'h "'2.
2:. which determine the A: A: A: '* 2:. What happens if the physical system is not in such an eigenstate? To find out. such that the wave function can be written as (4. even if l/J is not an eigenstate of A.18).22) The expectation value of A in the state l/J is given by (4. is the probability of finding the value At when A is measured. In this particular case.23) are special cases. it is still an assumption that a sharp line can be drawn between the dynamical variables of the system.20) Owing to the orthonormality of the eigenfunctions.21).2 The Superposition and Completeness of Eigenstates 57 2.22) and (4. the probabilities Icil 2 are uniquely determined. hence. the probability of finding it is the restricted sum ICiI2.the coefficients related to l/J by Ci are (4. If an eigenvalue of A is repeated. This conclusion comes about because two probability distributions of the random which produce the same expectation values for any function of must variable be identical. then Ci = 1. let us first look at a state that is a superposition of eigenstates of A. Exercise 4. We have seen the physical significance of the eigenstates of A. and all other Cj = 0 ifj i. . (4.9). From these equations we may infer that the eigenvalues of which are characteristic of the operator A. and (4. Although this interpretation is natural. If l/J happens to be an eigenstate. If this interpretation is made.4. (f(A) = 2: f(ADl cil i 2 (4. showing the consistency of the present interpretation with our earlier conclusions. are in fact the only values of A that can be found in the measurement of this physical quantity.24) of which (4.23). ICiI2. The Superposition and Completeness of Eigenstates.24) agrees with (4. (4. The sums of all the Icl which belong to the same eigenvalue of A will be symbolized by lcil 2 • More generally.22). which depends on the state. Prove (4. .23) It is important to remember that certain eigenvalues of A appearing in this sum may be repeated. l/Ji> of A. for any function of A.21) and the normalization integral is (4. then it follows that for simple eigenvalues. (4.
28) . Cogitation alone cannot provide this assurance. but experience has shown that it is legitimate to assume that any state can be represented to a sufficient approximation by a superposition of eigenstates of those dynamical variables that are actually observed and measured in experiment. the sense in which it represents the wave function 0/ must be clarified. Equation (4. and the state of the system.25) where the sum on the right side includes all physically acceptable eigenfunctions of A.25) is to be treated as if the sum were finite.20) shows how this division of the theory into physical quantities (operators) and particular states (wave functions) with its attendant consequences for the interpretation of A. the superposition appears as a linear combination of the appropriate eigenfunctions. and differentiations may be liberally interchanged. summations.27) corresponds to the choice (4. is known as convergence in the mean. The sum Cio/i may be said to approximate 0/ in the sense of a leastsquare fit at all coordinate points. so that we may generalize (4. until we are assured that the wave function 0/ of an arbitrary state can always be written as a linear combination of the eigenfunction of A and that there are no physical states which cannot be so represented. so that integrations. If a set of n orthonormal functions o/i is given. The interpretation of the formalism is incomplete.26) This kind of convergence. and 1 Ci 12 is related to the possibility of superposing states by constructing linear combinations of eigenfunctions.58 Chapter 4 The Principles of Wave Mechanics possible results of a measurement. Since this infinite sum does not generally converge pointwise. which is far more permissive than pointbypoint uniform convergence. however. show that the superposition L Cio/i which best approximates a state 0/ in the sense of a leastsquare i=1 fit and which minimizes the integral (4. L Exercise 4. which correspond to eigenfunctions o/i represented with nonvanishing contributions in the expansion (4. at least in a probability sense. Mathematically. If (4. The state 0/ is in a certain sense intermediate between its component states o/i' It bears the mark of each of them because a measurement of A in the state 0/ may yield any of the eigenvalues A.21) for the coefficients condition Bessel's inequality Ci' Prove that for this optimal (4.5.20) and write for any state 0/: (4.20) of 0/. which determines the actual outcome of the measurement. it is sufficient to require that the sum Cio/i converge to a L i state 0/ in accordance with the condition (4.
In the last section. but quantum mechanics requires the use of many far more general operators for which the completeness proof is hard to come by. the probability is the restricted sum lcil 2 • The coefficients Ci are called probability amplitudes. and (4.21). is repeated. Such a set of eigenfunctions is said to be complete. 0/2. is a simple eigenvalue.22) and (4.2 The Superposition and Completeness of Eigenstates 59 holds. and one has to proceed heuristically. lcil 2 is the probability of finding the value A. More generally. underlies the general theory of Chapters 9 and 10.2!). which act on vectors to produce other vectors. The integrals in Eqs. are to be sufficient in number to represent an arbitrary state. They are determined by the state according to (4. but the dimensionality of the vector space is no longer finite. . the set of orthonormal 0/ in the sense of (4.21) and conversely determine the state fully [Eq. .18). we sometimes call an Hermitian operator that possesses a complete set of eigenfunctions an observable.. Linear operators are like tensors of rank two. If A. note that the first n coefficients Ci retain their values and that In 2:: I n + 1 • If in the limit n functions represents formula. 2:. As n is increased and a new n + Ist function is included in the superposition of orthonormal functions. and (4.26) and we have for all 0/ Parseval's (4. For a large class of simple operatorsnotably those known to be of the SturmLiouville typethe completeness of the eigenfunctions can be proved rigorously.29) represent the formula for the square of the' 'length." or norm. This geometrically inspired approach of working in a generalized vector (or Hilbert) space. It owes its power and aesthetic appeal to its independence of any particular representation.20)]. we interpreted functions as vectors in order to give a geometric interpretation of the Schmidt orthogonalization procedure. ~ ao the sequence In approaches zero. 4. assuming completeness unless proven otherwise. According to the discussion at the beginning of this section.22) can be thought of as inner products. the expansion of an arbitrary wave function in terms of an orthonormal set of eigenfunctions is reminiscent of the expansion of a. vector in terms of an orthonormal set of basis vectors. (4. o/m and every physical state by a sum Cio/. of the' 'vector" 0/.29) This equation is called the completeness relation and expresses the condition that an operator A must satisfy if its eigenfunctions 0/. The assumption that every physical operator possesses a complete set of orthogonal eigenfunctions is spelled out in the expansion postulate of quantum mechanics: Every physical quantity can be represented by a Hermitian operator with eigenfunctions 0/1. if A. akin to the scalar products for ordinary vectors.. (4. The expansion coefficients are analogous to vector components. Following Dirac. suitably extended to complexvalued vectors.. for the physical quantity A. 2: i where the coefficients are defined by (4. .
depend on the eigenvalue continuously. such as the linear momentum of a particle. although they may be infinite in number. The Continuous Spectrum and Closure. More precisely. If quadratic integrability is too narrow a restriction to impose on the eigenfunctions of A. We thus tailor the mathematics to the requirements of the physics. In the last section. corresponding to the eigenvalue A. Generally.18) of quadratic integrability of eigenfunctions. the discrete eigenfunctions are labeled as l{1. ICA' \2 in the continuum case is the probability density of finding A to have the measured value A'.31) into 2S ee references in footnote 1. all eigenvalues of A must be real. a useful condition that draws the line between those solutions of (4. Thus. The chief obstacle that restricts the eigenvalue spectrum to discrete eigenvalues is the requirement (4. we have assumed that the eigenfunctions of the Hermitian 'operator A representing a physical quantity are countable.60 Chapter 4 The Principles of Wave Mechanics 3. it is best to assume that the expansion postulate holds and to let nature tell us how wide we must make the set of eigenfunctions so that any physically occurring l{1 can be expanded in terms of it. these functions are part of the complete set only if they belong to a continuum of real eigenvalues. but not always. Boundedness of the eigenfunctions everywhere is usually. more indices must be used and summations (or integrations) carried out over these.31) In the notation it is tacitly assumed that for a continuous eigenvalue A' there is only one eigenfunction.25) must be generalized to read • (4.1 2 dA' is the probability that a measurement of A yields values between A' and A'+dA'.30) those eigenfunctions that make up a complete set for all physical states? The subscript A' has been attached to the eigenfunctions in the continuum to label them. Hence.32) Thus. ICA. what should be the criterion for selecting from the solutions of the equation (4. and can be integrated over a finite range of eigenvalues. .. have a continuum of values.24) by requiring that (4. It is essential to remove this limitation. As before. and eigenfunctions that are not quadratically integrable can appear in the expansion of a quadratically integrable wave function l{1 only with infinitesimal amplitude. Substitution of (4. We now extend to the continuum the fundamental assumptions formulated in the expansion postulate and generalize (4. since many physical quantities. the only assumption being that there is some fabric that will suit these requirements? If the probability interpretation of quantum mechanics is to be maintained.30) that must be admitted to the complete set of eigenfunctions and those that must be rejected. If there are several (possibly infinitely many) linearly independent eigenfunctions corresponding to an eigenvalue A'. (4.
we often write the equations of quantum mechanics as if all eigenvalues were discrete.35) should hold for the physically admissible eigenfunctions. With the requirement that the eigenvalues must be real.33) and if the contmuous and discrete eigenfunctions are orthogonal to each other. whether or not the eigenfunctions are quadratically integrable.33) permits evaluation of the expansion coefficients in (4. The imposition of such boundary conditions "quantizes" the continuous spectrum .33) holds.31).34) But the righthand side of this equation vanishes. Moreover. For instance. the summations over the eigenvalues implying integrations over any continuous portions of the spectrum. To keep the notation compact. giving (4. it can ~e merged into the single condition (4. we have the orthogonality condition (4. and the entire spectrum discrete. or periodicity in a large unit cell can be required. it is always possible to make all eigenfunctions quadratically integrable. We say that the eigenfunctions l/JA' are subject to Anormalization if (4. for the continuous spectrum.3 The Continuous Spectrum and Closure 3 61 Jl/J*f(A)l/J d r shows that the righthand side of (4. This convention is suggested by the formal similarity of the expansion equations rP = L i for the discrete. The license for such interchanges is implicit in the extended expansion postulate and is assumed without further notice. since x8(x) = O. This equation expresses orthogonality and deltafunction normalization of the continuum eigenfunctions.36) in close analogy with (4. impenetrable walls at great distances can be erected. by applying boundary conditions that confine the particle to a limited portion of space.38) for the continuous spectrum. The orthonormality condition (4.32) is obtained only if. and CirPi' (4.37) (4. Hence. Jl/J1·Al/JA" d r J(ArPA·)*l/JA" d r 3 = 3 (4. We remark that the derivations in the last paragraph succeed only if the order of the integrations over space and over the range of eigenvalues A I can be interchanged. in the usual sense of equations involving delta functions. the Hermitian property of A.21). we must demand that.
40) are adopted. Many of the features of the continuous spectrum will be illustrated in Section 4.41) relates the discrete Kronecker delta to the "continuous" delta function. 8(A' .38) is achieved if the relations (4.4. Use the closure relation to deduce. if it is to be complete can be derived from the identity l/J(r) = = f {~ .62 Chapter 4 The Principles of Wave Mechanics of A. 3 l/J.(r')l/J. and in the limit of infinitely receding boundaries. The transition is made by introducing a density of (discrete) states.(r) 2 is the probability density in coordinate space if the system is known to be in the ith eigenstate of A.43) .(r) f l/J. In the coordinate representation. But usually p is proportional to the volume of the enclosure. Its eigenvalue equation '7 l Ii Vl/Jp(r) = pl/Jp(r) (4.(r)}l/J(r') d r' • 3 Since this must be true for any function l/J(r). We know that Il/J. the linear momentum p is represented by the operator p ~ liV/i.l/J. A useful condition that a given set of orthonormal functions must satisfy.37) and (4.(r')l/J. (r')l/J(r') d r' . the sum will fail to represent the delta function. it is a simple matter to take proper cognizance of the equally important continuous spectrum. Hence. .N (4.A' I:::t. A Familiar Example: The Momentum Eigenfunctions and the Free Particle. Exercise 4.42) which is known as the closure relation. since then the eigenvalues are truly continuous. 4 c.A") = p(A')8ij (4. the density p becomes infinite. if the equations are written with the notation of the discrete spectrum. per unit eigenvalue interval: . The spacing between the discrete eigenvalues A~ decreases as the confining volume grows.6. p(A ) = I:::t. and it is then possible to speak of a finite density of states per unit volume. This is the number of eigenstates l/J. if the system is known to be at position r. Similarly.(r) = 2: l/J. portions of the discrete spectrum go over into a continuous one. without which the set of eigenfunctions of an operator A would not be complete. we infer that 2: l/J. conversely. p(A '). that the same quantity measures the (relative) probability of finding the system to yield A ~ in a measurement of A.(r) = 8(r  r') (4.39) In the limit of an infinitely large confining region. It is easy to see that consistency between the expansion formulas (4. If the set of orthonormal functions included in the sum is incomplete. 1 4.
p') (4. Py.45) and that the expansion coefficient is given by cf>(p) = l/!(r)e(i1fo)p.16) and (2. the eigenfunctions are said to be given in the knormalization.49) .co < Px. to specify a complete set of eigenfunctions will be discussed in detail later. of course.r d 3r (4.7. we set p = lik and then choose units so that Ii = 1. PY' pz < + co. This equation shows that (4.46) These are. These are the eigenfunctions of the free particle Hamiltonian.4 A Familiar Example: The Momentum Eigenfunctions and the Free Particle 63 has the orthonormal eigenfunctions (4. If. '. just the standard formulas of Fourier analysis (Section 3.44) is given in the pnormalization. Exercise 4. p2 li2d2 H = 2m = .r d p (4. and the three eigenvalues of Px.44) already illustrates the issues involved.44) are simultaneous eigenfupctions of all three components.17) are obtained.ikx corresponding to the two directions in which the particle can move. Each of the three momentum components supports a spectrum of repeated eigenvalues. like the three momentum components.47) which is an integral representation for the delta function (see Appendix. relation (4. Section 1). but the explicit example of (4.2m dx 2 (4. The occurrence of simultaneous eigenfunctions of commuting operators.42). and pz together completely specify the eigenfunction. The expansion postulate implies that an arbitrary wave function can be written in terms of momentum probability amplitudes cf> (p) as l/!(r) = (21T~?/2 (21T~)3/2 f f 3 cf>(p)e(i1fo)p. Show that the eigenfunctions of momentum satisfy the closure In one dimension it is sometimes convenient to use a closely related set of eigenfunctions.3). The nonnegative quantum number k is related to the energy by lik = V2mE The energy normalization is determined by the conditions f +OO 00 A*(E)A(E')ei(k'k)x dx = 8(E .E') (4.44) and eigenvalues .48) They are doubly degenerate for each energy E > 0: l/!E(X) = A(E)e ikx and B(E)e. in the above formulas. but the plane waves (4. and Fourier transforms like (2. The orthogonality of the eigenfunctions is expressed by __ 1_ (21Tfll f e(i1fo)(pp')·r d3 r = 8(p .
52) with different signs in the exponent (representing waves propagating in opposite directions) are orthogonal.E') gives the normalization constant jA(E)j2 = $E ( .52) Any two eigenfunctions (4. The wave functions may be normalized in the basic cube of side length L. thus: f ff I t/Jk(r)j2 dx dy dz = I . the energynormalized eigenfunctions for the free particle in one dimension are (k > 0) 1/4 t/JE(X) = m 87T'2h 2E ) r' "'"kx (4.vE') = vT) 8(E . such that upon translation by a fixed length L in the direction of anyone of the three coordinate axes all eigenfunctions of H shall assume their previous values. Section 1). A similar calculation gives I B(E) 12 • Hence. or zero). 8(vE = 2vE8(E .32) in the Appendix.=  I 2E 27Th m I =hp hv (4.53) where nx . and nz are integers (positive. we can write (4. whether or not they belong to the same energy. This periodicity condition restricts the allowed momenta or wave vectors to (4. In three dimensions.E') (4.50) Using the properties of the delta function (compiled in the Appendix. the identity (A. ny.E') *" 0.64 and Chapter 4 The Principles of Wave Mechanics f If E +'" _'" B*(E)B(E')ei(k'k)x dx = 8(E .49) as 27TIA(E)1 2 • v2m ~ 8(vE . the eigenvalues of the Hamiltonian H = _h 2 V2 2m are defined by imposing certain boundary conditions on the eigenfunctions A particularly useful set is obtained if we require the eigenfunctions to satisfy periodic boundary conditions. negative. if the phase constants are chosen equal to zero. The theory of Fourier series teaches us that there are alternate methods of constructing complete sets of discrete quantized energy eigenfunctions for the free particle Hamiltonians.51) where v = p/m is the particle speed.
provided that the integrations are extended over the periodicity volume.4 A Familiar Example: The Momentum Eigenfunctions and the Free Particle 65 Hence.58) Exercise 4. 2.57) In statistical mechanics.54) and the energy eigenvalues are the discrete numbers !i 2 k? 27T z!i2 E = .58) for the density of states 3 • Obtain a formula for the cumulative number Z of states whose energies are less than or equal to E. 8284.) I (4. + n.(n 2 + n Z 2m mLz x y + nZ) z (4.53) we have L 3 I1px I1py I1pz = h 3 I1n x I1n y I1n z (4. or zero) such that their sum.59) 3Brehm and Mullin (1989). n z to derive the (asymptotic) formula (4. n. negative. . In one dim~nsion..55) Since each allowed energy value can be obtained by a number of different choices of the quantum numbers nx. n z. Exercise 4.9. r (4.8. ny. Compute the number of ways in which one can add the squares of three integers (positive. l/Jk(r) = L 3/2 e ik.. Prove that the free particle energy eigenfunctions.56) n x + n y + n z = 27T z!iz rises rapidly. As the energy E increases. except for E = 0. satisfy the orthogonality conditions appropriate to a discrete eigenvalue spectrum.. successively assumes values no greater than 0. ::!::1. . ny.. I1N/I1E." The number.. Use the Euclidean space in which lattice points are labeled by the integer coordinates n x. this last equation is somewhat loosely interpreted by saying: "Each state occupies a volume h 3 in phase space. ::!::2. . 1. the number of ways in which three perfect squares can be added to yield z Z Z Z mEL (4. considerable degeneracy here. of eigenstates per unit energy interval for any but the lowest energies is to good approximation (4. Compare the numerical results with the asymptotic formula.. pp.= . + n. there is. and according to (4. 100. . free particle energy eigenfunctions that are subject to periodic boundary conditions give rise to an energy spectrum (n = 0. which are subject to periodic boundary conditions. The number of states (4.54) in the volume L 3 I1px I1py I1pz of phase space is I1n x I1n y I1n z.
the degree of degeneracy of any given energy value is less here by a factor of 8 than in the case of periodic boundary conditions. ez are restricted to be positive integers. y.ex'7T sm . The density of states is (4. y. Inside the box.64) produces no new linearly independent eigenfunction.ey + .(L.64) with .ey. the solutions are obtained by separation of variables.58) is that it decreases with energy as lIYE. z) = "'(x. y. y.ex. This different behavior is due to the fact that the degree of degeneracy of each energy value is independent of E. situated in the first coordinate octant with one corner at the origin. for a given energy E there are now more ways of finding suitable . .(~ + e + lC) 2m 2m x y Z = "'(x.ez satisfying the condition 2 .63) and E = . y.4 we will return to this problem as a special case of a onedimensional potential square well of infinite depth. y z) = "'(x. ~ E ~E h The main difference between this density of states and its threedimensional analogue (4. we choose the box to be a cube of length L. since a change of sign in (4.60) = 2!/). we set "'(x. y. Since the expectation value of the potential energy must be finite. . d2 y #' + ~Y = 0. However. .62) d 2Z .56) and (4. whereas in threedimensional space it increases linearly with E. Z ) = C' k sm xX .X = 0.ex.ez = 2 2 2mEL '7T 2 h 2 (4. sm kyy sm k zz = C sm ." familiar from every introductory course in quantum physics. Thus. As a convenient example.61) In a box geometry.2m V2 ". y.ex + .E doubly degenerate. is /).ey . except E approximately /).65) The values of . Hence. z) = ". Closely related to the foregoing formulation of the free particle problem is the problem of the "particle in a box. .= . but here we simply assume zero potential energy inside a finite volume at whose boundaries V jumps to an infinitely high value. and find that the SchrOdinger equation separates into three equations: d 2X dx 2 + k.(x.66 Chapter 4 The Principles a/Wave Mechanics and each eigenvalue. L.N /).1. L) h~ The eigenfunctions that satisfy the boundary conditions "'(0. the wave function must vanish within and on the walls of the box. h2 dl + ~Z = ° (4. . z) = E". z) = ° are 'I'(x. In Section 6. whereas the latter is proportional to YE.(x. z) h2 (4. the SchrOdinger equation is .ez LX • y 2 (4.eL'7T y sm L '7T z • . by virtue of the different righthand sides of (4. and equal to 2. for one dimension. z) = X(x)Y(y)Z(z) (4.65).E = 0.n! ~ ~ = j2. 0. 0) = "'(x. .
2). . These two effects compensate each other.2.''x a 0 a Figure 4.42). and as a result the density of states is the same for the case of periodic boundary conditions and the box.2. the eigenfunctions constitute a complete set. in terms of which an arbitrary wave packet can be expanded by a Fourier series within the basic periodicity cube.65).3) and the corresponding eigenfunctions for a onedimensional box of dimension L = 2a. The three lowest energy eigenvalues (ex = 1. The unit of energy on the ordinate is fl?/8ma 2 • The maximum amplitude of the normalized eigenfunctions is v'17li. For the energy eigenfunctions in the onedimensional box (Figure 4. Show how bett~r 9rT.10. 4Courant and Hilbert (1953).4 A Familiar Example: The Momentum Eigenfunctions and the Free Particle 67 values to satisfy (4. compute the first few terms in the closure relation (4. In either case.~_>i 4r""""""":~ J. 4 Exercise 4. This comforting result is an example of a very general property of the density of states: It is insensitive to the specific details of the boundary conditions and is dependent only' on the size (volume) of the region of integration.
66) From (4.x') are obtained as the number of terms is increased. They are.67) Exercise 4. but unitary operators are of comparable importance. Hermitian operators play a paramount role in quantum mechanics. 5.67) can also be written as 3 "'i(r)Ut U"'2(r) d r = f [UtU"'I(r)]*"'2(r) d r 3 so that a unitary operator is one that satisfies the relation (4. since for two eigenfunctions of a unitary operator. Prove (4. As representatives of physical observables. the equations "'2 "'1 f and (U"'I)*U"'2 d r 3 = U~*U~ f 3 "'i"'2 d r 3 (4.66). like selfadjoint (Hermitian) operators. therefore. corresponding to the eigenvalues U~ and U~. A unitary operator U is defined by the condition that it must have an inverse and preserve the norm of all state vectors: (4. Eq.11. The time development operator.70) . A systematic discussion of symmetry in quantum mechanics is taken up in Chapter 17. Hence. is unitary. Unitary operators also serve to transform quantum mechanical states that describe equivalent physical situations into one another. at the heart of every analysis that seeks to take advantage of the symmetry properties of the physical system under consideration.67) from (4.66) by letting". = choice of the coefficients Cl and C2' + C2"'2' for any By the definition of the adjoint operator Ut. can be assumed to have and complete sets of orthonormal eigenfunctions.68 Chapter 4 The Principles of Wave Mechanics approximations to the delta function 8(x . reasoning analogous to that given for Hermitian operators leads to the conclusion that all eigenvalues of a unitary operator have an absolute value equal to unity: (aj:real) (4.69) (l  U~*U~) f "'i"'2 d r = 0 follow. as formalized in (3. which links the initial state of a system at t = 0 to its state at an arbitrary time t. but it is instructive to illustrate the underlying ideas by some simple examples in the realm of oneparticle wave mechanics.68) Unitary operators.66) it follows that a unitary operator preserves the inner (Hermitian scalar) product of any two states: f f [U"'I(r)]*U"'2(r) d r = 3 f "'i(r)"'2(r) d r Cl"'l 3 (4. (4. Unitary Operators: The Displacement Operator.
= Do = 1 (4. the operator DI.+T/f/!(x) (4. but it should be noted that this is not a general requirement for groups.74) and an inverse of every element: (4. .. An example of a useful unitary operator is the displacement or translation operator DI.f/!(x) = f/!(x ..(.72) Hence. is unitary.=x o Figure 4.:..DT/ = DT/DI.) by a displacement vector g in one dimension. Figure 4. Since f/! (x .76) Finally.5 Unitary Operators: The Displacement Operator 69 and eigenfunctions belonging to different eigenvalues are orthogonal.3.~) is normalized to the same value as f/!(x) . displacement operators have the property DI. Linear displacement or translation of a function l/J(x) ~ Di.3 illustrates the transformation. as for Hermitian operators. Again.DT/f/!(x) = DI. Successive application of two displacement operators DI.~ . If g = vt.+"1' The translation group is said to be Abelian because its elements commute.' defined by its action on any wave function f/!: (4.75) such that D~lDI. linearly independent eigenfunctions belonging to the same eigenvalue may be chosen to be orthogonal.TJ) = DI. since both sides of this equation are equal to D (+1. = DI. :. Such closure under multiplication is the chief characteristic of a group.1.77) which is true for displacement operators.71) where ~ is a real number.. The transformation of the function is active. l/J(x) = l/J(x .73) which expresses the fact that the product of two displacements is again a displacement. and DT/ leads to DT/DI. this figure illustrates an active Galilean transformation (Section 4. the multiplication of the elements of a group must be associative: (4. The additional properties defining a group are the existence of an identity: Do =1 (4. the generalization to higher dimensions is immediate.+T/ (4..7) for a • positive value of t. Although this definition of the displacement operator is for wave functions in one dimension.
Prove that if D~ is to be unitary. we can (4. = D~(l .81). Hermitian. (4. = 1 .. = exp (.82) it is seen that.~) (4.. because the eigenfunctions would then diverge for x ~ + 00 or x ~ . I a k= . such as 7T/~ < k' :5 +7T/~.i ax (4. d~ . the eigenvalues of the displacement operator are expressed as (4.12.iek (4.83) which is proportional to the linear momentum operator Px.79) = D~D. Since D~ is related to the operator k by Eq. For the latter we already know that the eigenvalue equation kf(x) = ~ af(x) = k' f(x) I ax (4.iek) from which we infer the differential equation dD~ = lim D~+. (4. It cannot have an imaginary part.~ ax !/J(x) a) 2! (4. it is necessary that k be ~ Combining an arbitrary displacement write: D~+.70 Chapter 4 The Principles of Wave Mechanics From the basic property of the displacement operator it is seen that for an arbitrarily small displacement e. we can limit ourselves to values of k' in an interval of length 27T/~. which is referred to as the first .80) With Do = 1 for ~ = 0.81) Since D~!/J(x) = !/J(x)  ~ 1i !/J' (x) + e !/J"(x) ± . where n is any integer. The Hermitian operator k is said to be the generator of infinitesimal translations: The eigenvalues and eigenfunctions of the unitary operator D~ can be determined from its representation in terms of the Hermitian operator k.78) where k must be a Hermitian operator to ensure the unitarity of D~. Exercise 4.80) has the solution D~ = eI~ = !/J(x . Thus.00 and not represent possible physical states.84) is satisfied by the eigenfunctions f(x) = eik'X (4.+0  D~ = iD~k e (4. in the coordinate representation.86) from which it is seen that the same eigenvalue results if k' is replaced by k' + 27rn/~. with a small displacement e. to first order: D.85) The eigenvalue k' of the operator k may be any real number.
Eq. = exp{ i(k~g + k. The generalization of the concepts of Brillouin zones and Bloch functions to three dimensions involves no new ideas. Exercise 4.TJ + k~()} (4. The generator of infinitesimal displacements Ik ~vl = (4.14.89) with period g. u(r + R) = u(r). It forms the backbone of every analysis of the physics of periodic lattices.f.6 The Charged Particle in an Electromagnetic Field and Gauge Invariance 71 Brillouin zone.86). TJ. The Hamiltonian operator considered so far and listed as the total energy in Table 3. The degenerate eigenfunctions of D g corresponding to the eigenvalue D'g = exp( ik' g) are f n(x) = ei(k' +Z7Tnlg)x (n: any integer) Hence.91) is proportional to the momentum operator p.88) be an eigenfunction of D g. 6. d 3 are the primitive translation vectors of a threedimensional infinite lattice whose points are at positions R = nIdI + nzd z + n3d3 (ni = integer).90) where the displacement vector 8 is represented by its three Cartesian components (g.R differ by a reciprocal lattice vector G. where n is any integer.e.88) are known as Bloch functions. defined by the condition G .ik .87) is the most general eigenfunction of D g corresponding to the eigenvalue (4. Note that any two values of k I which produce the same eigenvalue eik'. i. with eigenvalue D'g = eik'g. for every R. The Charged Particle in an Electromagnetic Field and Gauge Invariance. such as crystalline solids. By requiring that g(x) defined by (4.1 had the simple form z H = L + VCr) (4.88) are therefore products of plane waves and arbitrary periodic functions u(x) of x: u(x  g) = u(x) (4. it is evident that. The eigenfunctions g(x) = eik'Xu(x) (4. verify the periodicity property (4. If db d z..92) 2m . The coefficients C n are arbitrary.81) is generalized to Df.r. Since displacements in three dimensions commute and may be carried out equivalently in any order. Functions that have the structure (4.13. show that any simultaneous eigenfunction of all the translation operators DR = e. in threedimensional space. = eik. R = 21Tn. Exercise 4.R corresponding to eigenvalues eik'.R is of the form u(r)eik'. C). where u(r) is an arbitrary periodic function of r over the lattice. the linear combination +00 n=oo +00 (4. (4.89) directly from the definition of D g as a displacement operator.
Prove the identity (4. X A) = in {.95) and this relation between the canonical momentum p and the velocity v.B mc (4.93) Here ¢ and A are the scalar and vector potentials of the electromagnetic field. From (4. But classical Hamiltonian mechanics is not limited to such systems. may not commute. v] = v X (v X v) .15.96) and (4.97) we find m d (v) dt = q 2c q (v X B . If a particle of charge q moves in an external electromagnetic field. The validity of this Hamiltonian operator in quantum mechanics is supported by a calculation of the time rate of change of the expectation value of r for the particle: ~ (r) = i~ ([r.c at aA .q(V¢) .(v) dt d 1 = :di ([v.(v X v) X v Exercise 4. In q¢]) .~ A) Evidently.v· v ] ) m 2 1 + :.B X v) . the classical equations of motion may be derived as Hamilton's equations from a Hamiltonian (4. (p . H]) +  (av) at 1 = : In ([ v. . respectively.97) for any operator v whose components Combining (4.(V mc .. v X v = { mc (p X A +A X p) = in {.. in the presence of a vector potential A.97) The righthand side can be evaluated by using the operator identity [v. both of which are now assumed to be given functions of rand t. such as a charged particle in an electrostatic field.95) we infer that the components of the velocity operator generally do not commute and that. this equation defines the velocity operator: (4.q mc (aA) at (4.. and neither is quantum mechanics. v .([v. has the same form in quantum mechanics as in classical mechanics.c (aA) at If we recall that the electric field E is 1 E = V¢ . since the magnetic field B = V X A.96) Newton's second law in quantum mechanical form is . H]) = . which may be timedependent.94) (4.72 Chapter 4 The Principles of Wave Mechanics corresponding to a conservative system with potential energy function V.
Since v X (V f) = 0 and Vat af a = . the cross products of two vectors require careful treatment because noncommuting ractors are involved. tillie I (4. the Hamiltonian is H' = _1 2m (p _q A.(v) dt = !L (v 2c X B .104) If the phase is a constant.2m h 2( )2 "'(r. t ) = h at 2m 2 ( i V .!l.B X v) + q(E) I (4. t) is an arbitrary differentiable function of rand t.102) and deduce that this equation is .104) in (4.. we merely need to substitute "" as given by (4. The new wave function "" is related to the old wave function '" by the unitary phase transformation. To prove that (4.99).99) where fer.102) where the new potentials A' and 4>' are related to the old potentials by (4. A' he )2 ""(r. t) t) (4. (4. global gauge transformations are not interesting here. t) (4. t) at = . t) + q4>"'(r. it is appropriate to discuss the choice of the electromagnetic potentials in quantum mechanics. In quantum field theory. In the new primed gauge.. t) + q4>'''''(r. we speak of a global gauge transformation.100) The timedependent SchrOdinger equation in the old gauge is ih a"'(r.98) In all of this.6 The Charged Particle in an Electromagnetic Field and Gauge Invariance 73 we arrive at the quantum mechanical version of the equation for the acceleration of the particle in terms of the Lorentz force: I m!!. Since multi plication by a constant phase factor e ia is unimportant in ordinary quantum mechanics where '" is a probability amplitude. however.. "" (r.101) and in the new gauge it has the form ih a'" ' (r . At this stage.Vf at the fields E and B are unchanged by the gauge transformation on the potentials: I A 7 A' = A + Vf (4.104) constitutes a local gauge transformation. these transformations have significance.he A (4. The primed and unprimed potentials are said to represent two different but equally valid gauges. t) = U"'(r. where particles can be created and destroyed.)2 + q4>' c iq V .103) with IU = eiq!(r.
and the canonical and kinetic momentum do not necessarily .100) in the two gauges are connected by H' = UHUt + ih  au ut at (4.106) I (4. as distinct from the kinetic momentum mv =p  q A. The representation of the canonical momentum operator p. A = 0 can be chosen. Carry out the details of the gauge transformation explicitly to show that (4.q A = p .17.6) produces a gaugeinvariant current density and that this new j satisfies the continuity equation ap + at V • 'J =0 for the SchrOdinger equation (4. 1 aA c at =0 Under these conditions. and the gauge invariance of the theory finds its expression in the relation This connection ensures the gauge invariance of the expectation value of the kinetic momentum.16.101) if the Hamiltonians (4.105) holds with any unitary transformation of the form (4. we use the definition (4.q A C C C (4. further.107) Exercise 4.104). gauge invariance considerations apply equally if the particle is free (zero E and B fields) and the potentials satisfy the conditions V XA = 0 and V</>+ .101) in the presence of an electromagnetic field. has been kept the same. p = ~ V. The fundamental commutation relations are gaugeinvariant. Show that the substitution V ~ V  ~: A in (3.93) and (4.99) and the commutation relations to verify that the gauge transformation (4.74 Chapter 4 The Principles of Wave Mechanics identical to (4. Exercise 4.104) leaves the operators r and v (but not p) invariant: u(p and that.105) To check the validity of this condition. the trivial gauge </> = 0. =r q A) ut = P Urut C  q Vf . Other gauges are perfectly acceptable. for all choices C I of the gauge. Although we have assumed in the foregoing discussion that an electromagnetic field is actually present.
where Ivl = v. t) and the two Hamiltonians. fixing the most natural gauge. after it has been boosted so that it is moving with velocity v. For the simple example of a free particle state represented by a plane wave with momentum p.rwt) (4. We now turn to transformations that involve uniform motion with constant velocity v. I/J(r. But it is generally assumed that in the absence of electric and magnetic fields. the absolute uniform motion of a closed physical system is undetectable. In this view. or if the particle is neutral. t) = = ei[k.109) This would imply that the boosted plane wave state is I/J(r . t) for all times t. t) of the same system. the state of the system be described by a gauge field represented by A and <p.rwtkvtJ (4.19. In nonrelativistic physics we speak of the Galilean principle of relativity.18. A = 0. show the connection between the two gauges: <p = . t) = ei(p. Exercise 4. .110) This expression is plainly wrong. Galilean Transformations and Gauge Invariance. In particle physics. Its imposition necessarily requires that.10. According to the principle of relativity. For a charged particle in a uniform static electric field E. and <p' = 0. in addition to a wave function I/J. we attempt to relate the state I/J(r. corresponding to a frequency w + k· v + mv 2 /2li. we would expect that the boosted state was just I/J(r . Show that the SchrOdinger equation in the presence of a static magnetic field is not invariant under time reversal. we expect that in the boosted state the momentum is p + mvand the energy E v = E + p' v + mv 2 /2. A' = cEt. ° Exercise 4. as it is in classical physics. Instead. r. There is every reason to believe that the principle of relativity is valid in quantum mechanics. while the associated frequency has been (Doppler) shifted from w to w + k· v.5 were purely spatial displacements and had nothing to do with time. the electromagnetic field is regarded as a manifestation of local gauge symmetry. the choice <p = 0.vt. t)llic] is elevated to a fundamental principle of local gauge symmetry. Derive the gauge transformation function f(r. The main lesson to be learned from this analysis of gauge invariance is that the wave function for a particle is dependent on the explicit or implied choice of a gauge and that the potentials A and <p must be known if the meaning of I/J is to be properly communicated.rEt)/1i = ei[p. A = has been made. The translations considered in Section 4. t) of a system at time t to the state I/Jv(r. If the wave function were itself an observable physical quantity (which it is not). Obtain the equations of motion for the expectation values of rand p in the two gauges and compare their solutions.(rvt)EtJ/Ii ei(k.vt. the invariance of the quantum mechanical description under multiplication of all wave functions by a spacetimedependent phase factor exp [iq f(r. Using the tools developed in Section 4. since it implies that the momentum and the corresponding de Broglie wavelength are the same for the boosted system as for the system at rest. The extension of these ideas to particles with additional (internal) degrees of freedom will be sketched in Section 17. 7.E .5.7 Galilean Transformations and Gauge Invariance 75 have to be equal.
t) are the boosted potentials acting on the boosted quantum system.vt. h 2 2m [i q A(r V ~ nc vt.A(r . viewed from two different frames of reference. We assume that another frame of reference . t) at = . t) at =.vt.ihv .i(mv. z. t) is not the correct wave function for the boosted state. t) satisfies ih al/J(r .vt. t) where A(r . we must multiply l/J(r .vt t) he 'h ' J vt. t) at =+ h2 2m . t)J2 l/J(r. t) 2 (4.vt. and position vector r.112). t) he iq 2 q4J (r . t)l/J(r. we obtain the correct wave equation for the boosted state: ih al/Jv(r.vt. t) characterizing the given state in frameofreference S with coordinates x.76 Chapter 4 The Principles of Wave Mechanics To be in compliance with the Galilean principle of relativity and obtain a correct wave function for the boosted state l/Jv(r.A(r . t)l/Jv(r. t) and substitute this into (4. the boosted state. t) and 4J (r .112) If we invert (4. we find that if l/J(r. Exercise 4. V l/J(r . t) + q4J (r  vt. but not on p and E.101): 2 ih al/J(r. we start again with the wave function l/J(r. t) satisfies the wave equation (4. To show how a passive transformation works.vt. t) J2 l/Jv(r. We may rewrite the last equation as ih al/J(r .vt) by an appropriate spacetimedependent phase factor: (4. t) + q4J(r.3).+ h2 2m [V  . t) (4.111): l/J(r .vt t) + .vt. t) = _ h at 2m [v .vt.rmv t12)1fil/Jv(r.111) The phase factor in (4. t) J l/J(r  vt. It is to be distinguished from a transformation between two descriptions of the same state of one system. t)  iq i 2 [q4J (r  mV2J 2 l/J(r ~ vt.vt. t)l/J(r . Check that (4.101) nC the wave function l/J(r . t) = e.20.111) gives the correct plane wave function for More generally.vt. t) .111) describes the relation between two different but equivalent states of a system or two replicas of a system. y. Such a boost is called an active Galilean transformation (Figure 4.112) depends on v. The extra term on the righthand side shows that this is not the correct wave equation and that l/J(r . t).vt.~q [V  A(r. t) (4. This transformation is called a passive transformation.mv l/J(r .113) The transformation (4. which are in relative motion.
~~ [V' . t') mV2] [qcP' (r'. t) (4. t') = e. t') in S'. t') the wave equation (4.V . trivial."==0+====0+'===========:.114) Here the second. the Galilean principle of relativity appears to be jeopardized.1 . t ) . t) = cP(r' + vt'. the simple gauge transformation that defines the wave function l/J'(r'. z') is in uniform translational motion with velocity v relative to S. t') = cP(r.xrx ' Figure 4. t ' ) (4. Ii 2[ .. t') = e.. S'(x'. t') and cP'(r'.~~ A'(r'.i(mv. A Galilean coordinate transformation (Figure 4.7 Galilean Transformations and Gauge Invarianee 1/1 vt 77 1/1' ':.vt and t' = t (4. t') Since this equation in the primed frame does not have the same form as (4. Passive Galilean transformation.i(mv..115) Hence. V' and V = V' (4.101) in S is transformed into iii al/J(r' + vt'. t) at = ~ at'  v .A (r . . t') at' . From (4. t')l/J'(r ' . x' = x . ]2 .ut. effected by the mapping l/J(x) ~ l{J'(x') = l{J(x' + ut).101) in the unprimed frame. the plane wave is .2 l/J(r' + vt'. of a wave function. in which case the time transforms nontrivially. It is remarkable that if the momentum and energy in the free particle wave function are regarded as components of a relativistic fourvector. equation has been introduced to reduce the likelihood of mathematical errors in carrying out the transformation. if we define the potentials in S' as = = A(r' + vt'.117) Nonrelativistic quantum mechanics is thus seen to be consistent with the Galilean principle of relativity owing to the gauge dependence of the wave function. t') .4. Compare with the active transformation shown in Figure 4. lq .r+mv 2 tl2)I"l/J(r' + vt'.4) is given by r' = r . It also prepares the way for use of Lorentz rather than Galilean transformations.116) restores the expected manifest Galilean form invariance: iii al/J'~::' t') = . . t') A(r.rmv 2tI2)1"l/J(r. l/J'(r'. However.114) we deduce that ~ A'(r'. nrl/J'(r" t') + qcP'(r'. y'.mv l/J(r 2m lie Ii + + vt'.3.
Exercise 4. In frame S. is imposed. (b) Determine the energy spectrum in the presence of the field and show that it coincides with the spectrum for <P = 0 if the flux assumes certain quantized values. If the same state is viewed from a moving frame S'. show that the two counterpropagating waves have different wavelengths.2. This result implies that the de Broglie wavelength is different as viewed from two frames. If t« 1. . 3. determine the energy eigenvalues and the eigenfunctions.109) and a plane wave of equal magnitude and arbitrary phase but opposite momentum. In Chapters 23 and 24 we will return to these issues in the context of relativistic quantum mechanics. which are in relative motion. show that there are stationary state solutions that represent plane waves with exponentially attenuated amplitude. represented by the vector potential A = <Pk X r/[21T(k X r)2]. as in Problem 2. Compare and contrast the Galilean transformation of l/J waves with the analogous transformation of classical waves.) Solve the wave equation for a potential of the form V = .Vo(l + i0. as might be generated by reflection from an impenetrable barrier. and there is no additional phase or gauge transformation. (Work out the appropriate continuity equation.109) describes a free particle state in the frame of reference S. A particle of mass m and charge q is constrained to move in the xy plane on a circular orbit of radius p around the origin 0.) Exercise 4. and a magnetic field. Note that the energy levels depend on the strength of a magnetic field B which differs from zero only in a region into which the particle cannot penetrate (the AharonovBohm effect). If the plane wave (4. show that the passively transformed wave function l/J'(r'.22. where Vo and t are positive constants.21. (a) Show that the magnetic field approximates that of a long thin solenoid with flux <P placed on the z axis. (See also Section 1. consider a standingwave state l/J that is a superposition of a plane wave (4. t') correctly accounts for the momentum of the same particle in frame of reference S'. Calculate the absorption coefficient. Show that the nodes in the wave function l/J are separated by half a wavelength. but is otherwise free. but that the spacing of the nodes nevertheless is the same as in S. Problems 1. Show that the addition of an imaginary part to the potential in the quantal wave equation describes the presence of sources or sinks of probability.78 Chapter 4 The Principles of Wave Mechanics invariant under Lorentz transformations. describing absorption of the waves. If a particle of mass m is constrained to move in the xy plane on a circular orbit of radius p around the origin 0. 2.
1. Preliminary Remarks. such as the vibrations of an elastic medium or the electromagnetic field in a cavity. If a is chosen as the coordinate origin and V(a) as the zero of the energy scale.a)2 + .CHAPTER 5 The Linear Harmonic Oscillator In solving the Schrodinger equation for simple problems. The harmonic oscillator stands out because in many systems it applies to motion near a stable equilibrium. i.3) We call w loosely the (classical) frequency of the harmonic oscillator. and each expansion coefficient oscillates like a simple harmonic oscillator. we are then confronted by the quantum mechanics of many linear harmonic oscillators of various frequencies. an arbitrary potential V(x) may be expanded in a Taylor series: V(x) = V(a) + V'(a)(x .. conservative system in one dimension is . and provides the basis for the description of manyparticle states. It governs the dynamics of most continuous physical systems and fields. then (5. we first assume that rjJ a~d V depend only on the x coordinate. the potential energy term in the SchrOdinger equation is (5. The behavior of most continuous physical systems. in the coordinate representation.1) The eigen"alue equation for this Hamiltonian.2m ~ + V(x)rjJ(x) = ErjJ(x) (5. A familiar example is provided by the oscillations of the atoms in a diatomic molecule. In the neighborhood of x = a. is the ordinary SturmLiouville differential equation h 2 d 2 rjJ(x) .. In this chapter.2) We will solve this SchrOdinger equation for several special forms of the potential V(x). For 79 .. In any such physical system. If x = a is a stable equilibrium position. we discuss the oscillator in the coordinate representation. and V"(a) > O. The linear oscillator is important for another reason. V'(a) = 0. The Hamiltonian for a . V(x) has a minimum at x = a. can be described by the superposition of an infinite number of spatial modes. but harmonic oscillations of various generalized coordinates occur in many different systems. For the linear harmonic oscillator.3) is the first approximation to V(x). has equally spaced energy levels.e. 2 H = 2m Px + V(x) (5. leaving for later (Chapters 10 and 14) the more general representationindependent treatment.a) + 2: V"(a)(x 1 .
3). the Hamiltonian is (5.12) and the energy E is related to e by E = snw (5. In quantum mechanics. Two parameters.4) and Pi.11) The eigenvalue problem for Hg is expressed as (5. Show that Eqs. all quantum field theories make use of the results that we are about to obtain. .Pi.13) . the mass m and the frequency w. = dT (5. The fundamental commutation relation (5. Planck's constant provides a third dimensional number and thus allows us to construct a characteristic energy (nw) and length (Vntmw).1.6) If the Hamiltonian is similarly replaced by the dimensionless quantity H Hi.9) are the correct equations of motion for the linear harmonic oscillator defined by (5.1) and (5.80 Chapter 5 The Linear Harmonic Oscillator this reason.10) is transformed by (5.8) Hamilton's equations for the system reduce to the usual form dpi. we employ these scaling parameters to replace the canonical coordinate x and momentum Px by the dimensionless variables (5.~ = il (5.4) and (5.9) Exercise 5.3).5) into ~Pi. appear in the Hamiltonian (5. (5. = Jh~WPX (5.1) with potential (5. In the interest of accentuating the great generality of the linear harmonic oscillator.5) In terms of the new variables.7) and the time by the dimensionless variable T = wt (5. = nw = 2 pi + 2 e I I (5. They may be used to fix units of mass (m) and time (l/w).
and if rjJ(x) is a solution of the SchrOdinger equation (5. Any linear combination of solutions (5.rjJ( . a threeterm recursion formula is obtained. If a power geries solution of this equation is attempted.18) (5. for if x is changed into x. then it follows that rjJ(x) is also a solution of this equation. if the coordinate origin is chosen suitably.14) and (5.16).15) This yields the eq~ation  J2v dv .17) Exercise 5.x)] These are the even (e) and odd (0) parts of rjJ(x).2g . rjJe(x) = rjJe(x) . Substituting a power series with undetermined coefficients for rjJ and v into (5. Similarly.2 Eigenvalues and Eigenfunctions 81 The SchrOdinger equation for the linear harmonic oscillator is transformed into ~~~ + (28  e)rjJ = 0 (5. if the potential energy satisfies the condition V( . we speak of odd parity if rjJo(x) = rjJo(x). the following two functions must also be solutions: rjJix) = 1 2 [rjJ(x) + rjJ(x)] rjJo(x) = 1 2 [rjJ(x) .2). in constructing the solutions' of (5. if rjJ(x) is a solution of the Schrodinger equation.2). Hence.2) and the transformed equation (5. is said to have even parity. is sanctioned by custom.2. To get a differential equation whose power series solution admits a twoterm recursion relation. The SchrOdinger equation with even V(x) is said to be invariant under reflection.16) where n is defined by the relation 1 e=n+2 or (5.14) By using the same symbol rjJ. Extend the notion of invariance under reflection (of all three Cartesian coordinates) to the Schrodinger equation in three dimensions. we make the substitution (5.x) = V(x) . A state which is represented by a wave function that is even. respectively. Thus. we indulge in an inconsistency that.2) also solves (5. Show that .3. though perhaps deplorable. Eigenvalues and Eigenfunctions. One simple and important property of the harmonic oscillator derives from the fact that V is an even function of x. which is simpler to analyze.+ 2nv de dg = 0 (5.14) for two different functions for the variables x and g. Generally. 2. the equation is unaltered except for the replacement of rjJ(x) by rjJ( x). in both (5. Exercise 5.1) for even V(x). we may confine ourselves to all even and all odd solutions. obtain the recursion relations and compare these.
. unless n is an integer.5. and the ratio of the coefficient of gk to that of gk2 is ~ = ak2 2(k .g. Odd case: vo(O) = 0.6 + ..1) 3! e + 2 (n  1)(n .lIC)el. we need not become involved in complicated estimates. Furthermore.. In order to determine if such a solution is quadratically integrable and can describe a physical state. Since g is proportional to x. because they describe a situation in which it is overwhelmingly probable that the particle is not "here" but at infinity.16) are invariant under reflection g ~ .20) Here k = 2. it is easy to verify that if k > C(n + 2).n .15) that for large values of Igl the eigenfunction If!(g) diverges faster than exp[(112 . . for V e and k = 1.2 + 22 n(n .lIC)el or g exp [(1 .19) v (g) = o g 2(n . This behavior can be avoided only if n is an integer. and we need to look only for solutions of definite.3) 5! e _. it follows from (5. parity. The rule that governs these expansions is evident. .. depending on whether k is even or odd. by imposing the boundary conditions: Even case: viO) = 1. If C is chosen to be greater than 2.1) (5.16) and equating the coefficient of each power of 5.lIC)el or g exp[(112 .20) is always greater than the corresponding ratio of coefficients in the power series expansion of exp[(l .(O) =0 1 v~(O) = or into (5. By substituting V. respectively. it is easy to construct the even and odd solutions. we must consider the asymptotic behavior of V e and V o ' How do these functions behave for large values of Igl1 For the purpose at hand. the ratio (5. If!(g) diverges as I gl ~ 00..6. Eqs.2)(n . (5.. only solutions of definite parity need be considered. For a linear secondorder differential equation like (5. For a fixed value of n.82 Chapter 5 The Linear Harmonic Oscillator if V depends merely on the radial distance r. other than n = even for V e and n = odd for Vo> all terms with k > n + 2 have the same sign.2) k(k .4) 6! ~ /. Unless n is an integer both series are infinite..2). we obtain the power series expansions V (~.16).2) 4! 2 ~ /. Such wave functions are physically not useful.4  23 n(n . where C is an arbitrary constant C > 1. Hence.19) terminates and becomes a polynomial of degree n. V e and vo.4.lIC)el.14) and (5. even or odd. (5. for vo. e l:\ _ g to zero (see Exercise 1   2n 2! ~ /. .3. in which case one of the two series (5.
the operator HI. We have come to a very important conclusion: The number n must be a nonnegative integer (n = 0. + 2: g . 2. we get l{!(g) = ee12 vo(g) (5.26) as defined in (5. .2: 1 1 2 1 2 1 (5. This property of the onedimensional SchrOdinger equation follows from the discussion of Section 3.22) decay rapidly at large distances. familiar to all students of physics since the earliest days of quantum theory. 1. are sometimes referred to as the steps or rungs of a ladder.25) The corresponding solutions are the eigenfunctions.21) and the state has even parity. since for each eigenvalue there exists only one eigenfunction. An alternative and very . The states described by such eigenfunctions are said to be bound.22) and the state has odd parity.23) Classically. 1. a stationary state of the harmonic oscillator can have only one of a discrete set of allowed energies! The energies are thus indeed quantized. .24) and of the Hamiltonian operator (5. hence. the state of excitation n is said to represent the presence of n quanta or particles.. E can assume only the discrete values (5.1/2 is called the number operator. Depending on the physical context to which the harmonic oscillator formalism is applied. and we may speak of a spectrum of discrete energy levels for the harmonic oscillator. In quantum mechanics.. Both (5. Note that the eigenvalues of the linear harmonic oscillator are not degenerate.7).22) are now finite everywhere and quadratically integrable. It measures how many steps of the energy level ladder of the harmonic oscillator have been excited.)..5 [case (b)]. The eigenfunctions (5. 2. so that the particle is confined to the neighborhood of the center of force.21) and (5. That the eigenvalues are equidistant on the energy scale is one of the important peculiar features of the x 2 dependence of the oscillator potential. The nonnegative integers n are the eigenvalues of the operator He  2: = 2: PI.2 Eigenvalues and Eigenfunctions If n is even. .18) (5.21) and (5. The equally spaced energy levels of the harmonic oscillator. 3.. apart from an arbitrary constant factor. . all nonnegative numbers are allowed for the energy of a harmonic oscillator. The numbers En are the eigenvalues of the Schrodinger equation (5. If n is odd. Because its eigenvalues are n = 0. we get 83 (5.
1 and 5. the number n is traditionally and blandly called a quantum number. which makes use only of the algebraic properties of the operators ~ and p!. it is obviously a quantum phenomenon. n = 0. Its minimum value. Since in a stationary state. corresponds to the ground state. will be found in Section 10.30) H2(~) = 2 + 4e H5(~) = 120~  + 32~5 They satisfy the differential equation d2Hn(~) _ 2 c dHn(~) + 2 H (C) de ~ d~ n n~ =0 (5. and their commutator (5. (H) = E. Prove that (5. The energy fiwl2 is called the zeropoint energy of the harmonic oscillator. a few important mathematical properties of the harmonic oscillator eigenfunctions will be derived. It is traditional to define them so that the highest power of ~ appears with the coefficient 2n • Hence. it would follow that both Pxl/J = 0 and xl/J = O. by comparing with (5. but the energy of the oscillator is still fiw/2 and does not vanish as the lowest possible classical energy would.4S~2 + SgJ + 160~3 16~4 (5.19).27) where Hn denotes a Hermite polynomial of degree n.29) Here is a list of the first few Hermite polynomials: Ho(~) = \ 1 H 3W = 12~ Hl(~) = 2~ H4(~) = 12 .16) which can be constructed if n is an integer are known as Hermite polynomials of degree n. As a label of the eigenvalues and eigenfunctions of the ordinary linear harmonic oscillator.) = 0 and (x 2 ) = O.31) The first few harmonic oscillator eigenfunctions are plotted in Figure 5. Exercise 5. But no l/J exists that satisfies these two equations simultaneously.2S) and for odd n. and en is an as yet undetermined normalization constant.1. Being proportional to fi.(1) _ (nl)l2 2(n!) [(n _ 1)12]! vo(~) (5.2. The finite polynomial solutions of (5. (5.4.11). The complete eigenfunctions are of the form l/Jn (x) = CnHn ( if x) exp (  2 ~: x) (5. In this section.32) . a zero energy eigenvalue would imply that both (P.84 Chapter 5 The Linear Harmonic Oscillator general derivation of the eigenvalues and eigenfunctions (or better. we see that for even n. it can be understood on the basis of the uncertainty principle (Chapter 10). Study of the Eigenfunctions. eigenvectors or eigenstates) of the number operator. As in Section 4.6. But first the Hermite polynomials themselves must be normalized. 3. Hn(~) .
l{!3 .6 l{!2 2 4 . Figure 5.33) As a consequence of the relation (5.4 0. S .32). s) = 2 F(I: ) a~ S!:>.1. A particularly simple representation of the Hermite polynomials is obtained by constructing the generating function (5. we see that aF(~.2 0. The energy eigenfunctions of the linear harmonic oscillator. The vertical axis is the energy in units of liw. The functions l/Jn are plotted versus g = Vmwlli x and are normalized as f I l/JnW 12 d~ = 1. for the quantum numbers n = 0 to 5.3 Study of the Eigenfunctions l{!o 85 l{!l 4 2 • 0 0.
2 et. 1.2 and all its derivatives ± 00.s 2 The coefficient F(O. s) = e. s) = (1)" L . = et.2_(St.34) The generating function F(~. For example.33) an.d (5. Proof. s) e2st.28): F(O. 2 (5. This differential equation can be integrated: F(~. The three lowest energy eigenfunctions (n = 0. by Taylor's expansion of (5.2.. s) = F(O. Being dn et.2) for the linear harmonic oscillator in relation to the potential energy.s2n = ~ n=O n! and therefore the generating function has the form F(~. The intercepts of the parabola with the horizontal lines are the positions of the classical turning points.) = (1)" e' n ( ds s=o d~ 2) et.34): H n (~) = dn ~2 ~ ~2 d n n e' (s.86 Chapter 5 The Linear Harmonic Oscillator x 4 YIe/11UJ) 4 o 2 2 Figure 5. the derivative ~ must have at least n real roots. s) can be evaluated from (5. Since tend to zero as ~~ dn1 _t. Vex) = mal x 212 = fUJJ{212.35) a popular alternative form of definition of the Hermite polynomials. s) is useful because it allows us to deduce a number of simple properties of the harmonic oscillator wave functions with ease.s2 + 2s t.2 . Assume that d~n~l has nl real roots. From this definition it follows that all n roots of Hn(~) must be real. e.)2 (5.
39) the eigenfunctions are real. The lefthand side can be integrated explicitly. for p = 0.3 Study of the Eigenfunctions 87 ee times a polynomial of degree n. it can have no more than n such roots. the oscillator eigenfunction (5.37) Comparing the coefficients of equal powers of s nt kAP. we obtain the integral '* In terms of the variable g. The points in coordinate space at which a wave function goes through zero are called its nodes. it equals Hence. although with the particular choice of the arbitrary phase factor embodied in (5. we have for the normalized eigenfunctions (5. The assumption holds for n = I.33) and ($. Evidently.40). the orthonormality of the eigenfunctions is expressed as I r~= 2n n! v 'Tr f+oo H n (g)Hk(g)ee d~ = 5nk (5.36) for nonnegative integers n.34) have been used. k.40) For the sake of generality. .ee(s§)2 e §2_(t§fe 2A §e d~ . To this end we construct the expression f +. we verify that the oscillator energy eigenfunctions for n k are orthogonal. = L L n=O k=O _s n~ f+OO Hn(~)Hk(~)e2A§§2 dg  n!k!  where (5.27) has n nodes. .38) If we recall that g= R x. p. From the generating function we derive conveniently the value of the useful integral (5. For n = k. complex conjugation appears in (5. as expected. whence the assertion follows by induction. (5. we obtain the value of I nkp' In particular.39) and the orthonormality relation J:: f/!~ (X)f/!k(X) dx = 5nk (5.
6.5.42) Exercise 5.00 to + 00.88 Chapter 5 The Linear Harmonic Oscillator In many calculations. For the Hermite polynomials the integral representation (5.43) can be used to verify the formula (5. show that (5.45) This relation shows that the harmonicoscillator eigenfunctions constitute a complete set of orthonormal functions in terms of which an arbitrary function of x can be expanded. and integrate over u from . we record the value of (5. Its name and its significance will be explained in due course. The fundamental expansion postulate of quantum mechanics is thus shown to be valid for the energy eigenfunctions of the harmonic oscillator.36) and (5.?. work out the matrix elements (nlxlk) and (nlx 2 Ik) for the harmonic oscillator. .34) this expression equals n=O i n 1 H n (~)e/. Exercise 5.2/2Hn (11)e. and by (5.42): 2: n=O l/J~ (x' )l/Jn (x) = 8(x  x') (5.43) is valid. If we let s = 11 + iu in the generating function (5. Integral representations of the special functions that we encounter in quantum mechanics are often very useful.33). Validate (5.32) and by checking the initial values Hn(O). 2n n=O n.43) by verifying that it satisfies (5.43) f Consequently. 11 +') lU e u2 d u = 00 ~ r= ~ Hn(~) H n (11) v 11' LJ .T?/2 2 n!V1T = 8(~  11) (5.44) Translated into the x representation. we obtain from (5. For use in the next section. this is the closure relation (4. Alternatively. +OO F(l: !>. From (5. multiply the equation by e.33) for the generating function of Hermite polynomials.41) for p = 1: (5.37). we will need to know the matrix elements of the operator (5.41) the "braket" on the lefthand side of this equation is introduced here merely as a notational shorthand.
t) fi2 a ljJ(x.48) with the expansion coefficients given by C = n f +OO 00 _ ljJ~ (x) ljJ (x.49) then.51) Substituting (5.50) into (5. Accordingly.e. 2 2 t ) (5. = = ~o ~o c~ Ckei(nk)w. 0) dx (5.. t) = exp ( . The timedependent Schrodinger or wave equation In .4 The Motion of Wave Packets 89 Exercise 5. En = the wave packet at time t by the use of (3. it seemed efficient to compile them here in one place. we find (x).C ljJn (x) exp( . tW dx (5. In Chapters 3 and 4 we saw that the solution of this equation can be obtained automatically. Although the physical significance of some of these formulas will become apparent only later.47) determines the time development of the wave. n=O k=O j:: ljJ~ (X)XljJk(X) dx (5.15) (x).x 2 ljJ ( x.51). In the generating function (5.at 2m ax 2 2 mw + . knowing that for the harmonic oscillator.79): fiW( n + i). 4.~ aljJ(x. i.t2 (gz + 71 2 )] In this section we have derived a number of mathematical results pertaining to the solution of the SchrOdinger equation for the linear harmonic oscillator. replace s by s = t (71 and prove Mehler's formula + iu) (5. We now turn our attention to the behavior of a general wave ljJ(x.50) The center of probability of the normalized wave packet..t 2 1 . t~ whose initial form ljJ(x. The Motion of Wave Packets.52) 2: 2: C~Ckei(nk)W' (nlxlk) . if ljJ(x. the expectation value of the position operator x. 0). t) = . 0) = n=O 2: C ljJn (x) n (5. is according to (3.46) ~O Hn(g)Hn(71) 00 n t 2n n! = ~ exp 1 [2 g71 . if the initial wave can be expanded in terms of the timedependent eigenfunctions of the corresponding (timeindependent) Schrodinger equation. The completeness of the orthonormal energy eigenfunctions of the harmonic oscillator was proved in the last section.33). which we assume to be normalized to unity.inwt) n 2 n=O 00 2: (5. So far we have considered only the stationary states of the harmonic oscillator. we can construct iwt) ljJ(x. = j:: xlljJ(x.7. 0) is given. is expanded as ljJ(x.
53) () _j. .42) to calculate the matrix elements of the momentum operator. It shows that the expectation value of the coordinate.42) is substituted in (5.8. . compute the probability that the coordinate of a linear harmonic oscillator in its ground state has a value greater than the amplitude of a classical oscillator of the same energy. 1. n . Verify that (Px)t =m :t (x)t for this wave packet. 4.55) This expression is exact. If (5. Verify that (5. Compare with the results of the virial theorem. obtained in (5. Calculate the matrix elements of p/ with respect to the energy eigenfunctions of the harmonic oscillator and write down the first few rows and columns of the matrix.42). With this result. Exercise 5. (x)t. enter critically. For the energy eigenstates with n = 0.55) is expressible as () t = () 0 cos wt x X + (pJo .54) If we set we can write (5. (5. Exercise 5. and (5. Use (5. and 2. 3. Calculate the expectation value of x 4 for the nth energy eigenstate of the harmonic oscillator.32).9.40).sm wt mw .90 Chapter 5 The Linear Harmonic Oscillator which shows that the matrix elements of x. Can the same result be obtained directly by matrix algebra from a knowledge of the matrix elements of Px? 2. just like the classic coordinate x(t). the selection rule for the matrix elements (nlxlk).52).f ~ ~ x t  2mw n=l LJ r( * vncnCnle iwt + cnlcne) * /Wt"\ (5.50).k = ±l simplifies the summations and gives the result (5. Problems 1.. evaluate (Px)t as a function of time for the wave packet (5. which can also be derived directly from the equation of motion for (x)t (see Problem 3 in Chapter 3). Calculate the expectation values of the potential and kinetic energies in any stationary state of the harmonic oscillator. oscillates harmonically with frequency w.
(a) For this initial wave function.Problems 91 5.. Plot the width of the distribution as a function of temperature.a)2). (d) Calculate the probability density Il/J(x. again using the generating function derive a closedform expression for l/J(x. (b) Plot Pn 'for three typical values of a. governed by the Boltzmann distribution.43) may be used. (c) If the particle moves in the field of the oscillator potential with angular frequency w centered at the coordinate origin. greater than. calculate the probability Pn that the system is found to be in the nth harmonic oscillator eigenstate. the probability per unit length of finding a particle with displacement of x is a Gaussian distribution. Show that if an ensemble of linear harmonic oscillators is in thermal equilibrium. 0) = mw) (h7T 114 exp (mw(x2h.] 6. and check that the P n add up to unity. 1 . t). t) 2 and interpret the result. illustrating the case where a is less than. and equal to (h. Check the results in the classical and the lowtemperature limits. V... Use the generating function for the Hermite polynomials to obtain the energy eigenfunction expansion of an initial wave function that has the same form as the oscillator ground state but that is centered at the coordinate a rather than the coordinate origin: l/J(x. [Hint: Equation (5.
unlike the electrostatic forces that hold an atom together. The mathematics is relatively simple.20): fj. this is obvious. We can prove the theorem by considering the real solutions of Schrodinger's equation (see Exercise 3. There is no physically acceptable solution for E < 0 because of the general theorem that E can never be less than the absolute minimum of V(x). such a schematic potential approximates the real situation and provides a rough orientation with comparatively little mathematical work. for all x. We know from Chapter 2 that the solutions are harmonic plane waves. with wave number k = Y2m(E . in three dimensions as well as one dimension with and without periodic boundary conditions. Classically. such as tunneling and scattering resonances. so that we can concentrate on the physical features. the one for a constant potential is mathematically the simplest.Ejl{1(x) ~ 0 92 .2m If/'(x) + [V(x) . As we will see in Section 8.7. • Often. they extend to some distance and then drop to zero very fast.4. For instance.2 . But as the examples of the harmonic oscillator and the free particle have already shown us. Classically. especially the power of symmetry considerations. but here they are useful to exemplify characteristic quantum properties that arise from the smooth joining of the SchrOdinger wave function (Section 3. the strong nuclear forces acting between protons and neutrons have a short range. The Potential Step. do not occur in nature but serve as convenient models. a sectionally constant periodic potential exhibits some of the important features of any periodic potential seen by an electron in a crystal lattice. they are trivial.CHAPTER 6 Sectionally Constant Potentials in One Dimension Potentials like the rectangular barrier or the square well.5) at the discontinuities of the potential.V) We resume study of the SchrOdinger equation with such a potential because the qualitative features of a real physical potential can often be approximated reasonably well by a potential that is made up of a number of constant portions. Of all SchrOdinger equations.2. Figure 6. was already discussed in Section 4.. which are pieced together from constant zeroforce sections with sharp discontinuities. Next in order of increasing complexity is the potential step V(x) = Vo1](x) as shown in Figure 6. 1. it is also true in quantum mechanics despite the possibility of penetration into classically inaccessible regions. which sees a constant potential V(x) = const. The case of the free particle.1 shows roughly how a rectangular potential wellcommonly called a square wellmight simulate the properties of such an interaction.
2. the wave function has one of the two convex shapes shown in Figure 6.. t/J can never bend down to be finite as x + + 00. Energy eigenfunction for the (Heaviside) step potential function V(x) = Vo "1(x). To avoid these catastrophes.. would move freely until reflected at the potential step.. if it were incident from the left...3a. t/J diverges as x + .J. In Figure 6... everywhere.1. depending on whether the slope is positive or negative. . Hence.. Prove that E must exceed the absolute minimum value of the potential V(x) by noting that E = (H) in the stationary state t/JE(X).J.. there must always be some region where E > V(x) and where the particle can be found classically.. V=Vo . The normalization is arbitrary...The Potential Step V(x) 93 1 ...r1T.. a particle of this energy..... changing the sign of its momentum.. Conservation of energy requires it to turn around.3b.x 2 2 Figure 6. 1 .f.+''===x E="2 Vo o 1 Figure 6...... "/' has the same sign as t/J.l.L..00.. Classically.1.3. Exercise 6... Now we consider the potential step with 0 < E < Yo. corresponding to an energy E = Vo/2. The step function T/(x) is defined in Section 1 of the Appendix... Potential approximating the attractive part of nuclear forces (V = e1Xl/lxl) and a onedimensional square well simulating it...\. ....&.. if t/J is positive at some point x. If V(x) > E for all x... .. In Figure 6. .f .
the reflection is total. It represents a plane wave incident from the left with an amplitude A and a reflected wave that propagates toward the left with an amplitude B. x = 0.K = e. is in conflict with the boundary condition that l/J(x) remain finite as x ~ +00. The second linearly independent solution for x > 0.2).3. except for an arbitrary constant factor. The classical turning point (x = 0) is a point of inflection of the wave function.B) = KC or B _ = ik+K A ik . Hence. (6.a .5 that the wave function in the case of no degeneracy is real.2). el<X. By joining the wave function and its slope smoothly at the discontinuity of the potential.= A ik .E) Since l/J(x) and its derivative l/J'(x) approach zero as x ~ +00. according to case (b) in Section 3. hence. The solution (6. a graph of such a wave function may be drawn (Figure 6.1). According to (6.5 there is no degeneracy for E < Yo.1) Here hk = Y2mE hK = Y2m(Vo . Convex shape of the wave function in the nonclassical region (lfI'hfJ > 0).94 Chapter 6 Sectionally Constant Potentials in One Dimension ''x (a) (b) Figure 6.2) Substituting these values into (6. we have A+B=C ik(A . A wave packet which is a superposition of eigenfunc .K C 2ik . The oscillatory and exponential portions can be joined smoothly at x = 0 for all values of E between 0 and Yo: the energy spectrum is continuous. llX (a: real) =1+ e. we obtain l/J(x) = 1 2Ae "" oos(kx 2 i) (x (x < 0) (6. The SchrOdinger equation has the solution l/J(x)  _{Ae + BeCeI<X ikx ikx (x < 0) (x> 0) (6.1) can be given a straightforward interpretation.3) 2Ae ia/ 2 cos ~ eI<X > 0) in agreement with the remark made in Section 3. IA 12 = IB 12 .
the current density j can be expressed as the sum of an incident and a reflected current density. no matter what value the coefficient C may have. The second particular solution rep .3) reduces to dj/dx = O. Perhaps these remarks can be better understood if we observe that for onedimensional motion the conservation of probability leads to particularly transparent consequences.The Potential Step 95 tions (6. B A or A + B = 0 and C = 0 as Vo .= 1 K""'oo . Classically.ikx . Eq. the current density j is seen to vanish. Hence. or A = O. there is no net current anywhere at all. being reflected at the wall and again giving a vanishing probability of finding the particle in the region of positive x after the wave packet has receded. To the left of the potential step. there is no permanent penetration.4) has the same value at all points x.ikx l/f(x) = { Ceik. Exercise 6. From (6.> 00 or K . but there is also transmission to the right. This packet would move classically. When calculated with the wave functions (6.1) could be constructed to represent a particle incident from the left. 2m i n [ l/f* dx  dl/f dl/f* dx l/f ] (6.> 0 in the region under the barrier. These equations show that at a point where the potential makes an infinite jump the wave function must vanish. The case of an infinitely high potential barrier (Vo . Hence. as discussed in case (a) in Section 3. the joining conditions for the wave function at x = 0 now reduce formally to hm . whereas its slope jumps discontinuously from a finite value (2ikA) to zero.1). (3.2. this particle passes the potential step with altered velocity but no change of direction.3). degenerate eigenfunctions. We next examine the quantum mechanics of a particle that encounters the potential step in one dimension with an energy E > V o.5. The solutions of the Schrodinger equation are now oscillatory in both regions.1) it follows that in this limiting case l/f(x) .> 00. j = jinc + jref' without any interference terms between incident and reflected waves. the current density • } = .2). there is no net momentum in the state (6. For the physical interpretation. the relation IA 12 = 1B 12 ensures that incident and reflected probability currents cancel one another.x + Deik. their explicit construction is best accomplished by specializing the general solution: Ae ikx + Be.5) where Two useful particular solutions are obtained by setting D = 0. to each value of the energy correspond two linearly independent.> (0) deserves special attention. For a stationary state. The particle could be incident from the right or from the left. If there is no current. According to (6. Reflection occurs at the potential step. The first of these represents a wave incident from the left. hence.x (x < 0) (x> 0) (6. Show that for a wave function l/f(x) = Ae ikx + Be. as it does for any essentially real wave function.
. when it arrives at the potential step. there is no contradiction between the splitting up of a wave packet and the probability interpretation of the wave function. split into two parts. the presence of reflection means that the wave packet may. provided that its average energy is close to Vo.3. and the second is the transmission coefficient. R. We have (6. A+B=C k(A . Exercise 6. The equality of these values is assured by (6. Show that.7) In analogy to optics. and this would be very difficult to reconcile with the facts of observation. For a wave packet incident from the left.6) The current density is again constant. the coefficients for reflection and transmission at a potential step are the same for a wave incident from the right as for a wave incident from the left. T. For the splitting up of the wave packet would then imply a physical breakup of the particle. and reflection toward the right.7) ensures that R + T = 1.IBI2) hkllel2 m (x < 0) (x> 0) in agreement with Exercise 6. The remaining constants are related by the condition for smooth joining at x = 0.8) (6. Here we consider only the first case (D = 0).2.96 Chapter 6 Sectionally Constant Potentials in One Dimension resents incidence from the right. This splitting up of the wave packet is a distinctly nonclassical effect that affords an argument against the early attempts to interpret the wave function as measuring the matter (or charge) density of a particle. Note that the relative phase of the reflected to the incident amplitude is zero for reflection from a rising potential step.k[ e 2k + k[ A k + k1 (6. the first term in this sum is called the reflection coefficient.6) and leads to the relation IBI 2 k1 lel 2 IAI 2+ k IAI2 = I (6..B) = k[C from which we solve A .k B k . transmission to the left. electrons and other particles are always found as complete entities with the same distinct properties. but 7T for reflection from a sharp potential drop. but its value is no longer zero.and . j = !~ (IAI2 . After all. for a given energy E. On the other hand.= .9) Equation (6.= . The coefficients R and T depend only on the ratio EIVo. Instead.
2. a case of great interest is that of E < Va. height Vo. A. nK = Y2m(Vo  E). width 2a.+=Io a x Figure 6..a (6..E . In our study of more and more complicated potential forms.. C. We can immediately write down the general solution of the Schrodinger equation for E < Va: (x (a < a) (a < x) x < a) (6.4). The boundary conditions at x = . we now reach a very important case. albeit schematically. the rectangular potential barrier (Figure 6. D are conveniently expressed in terms of matrices: The joining conditions at x = a are similar. For this reason the rectangular potential barrier simulates. There is an advantage in placing the coordinate origin at the center of the barrier so.11) These linear homogeneous relations between the coefficients.I. Owing to the quantum mechanical penetration of a barrier. The Rectangular Potential Barrier. B.4.2 The Rectangular Potential Barrier 97 v a .. . They yield . The particle is free for x < a and x > a..10) < where again require nk = Y2mE.that V(x) is an even function of x. Rectangular potential barrier. the scattering of a free particle from any potential.
sinh 2"" )e"'" 11' . T = IFI = 16e4 kK 2 /(a( ')2 IAI2 ~ + K E 2 (6. e. which transmits poorly.2ika (6.15) Another limiting case is that of a very narrow but high barrier such that Vo K » k.18) and the transmission through this potential barrier is T= E E + ml 2f'l. cosh 2Ka = sinh 2Ka = e2/(a/2 Hence. we obtain the relation between the wave function on both sides of the barrier: A) ((COsh 2"" + i. ( 2/ smh = B 2Ka ( cosh 2Ka .17) the potential may be represented by a delta function positioned at the origin. It assumes an especially simple form for a high and wide barrier. » E. Note that 1'/2 .98 Chapter 6 Sectionally Constant Potentials in One Dimension Combining the last two equations.2lka i. A reflected wave whose amplitude is B is also present. K k 1'/=+ K k (6. In first approximation.4.e 2 = 4.1 (6. A particular solution of interest is obtained from (6.12) by letting G = O.uch that Ka » 1. Under these conditions."2 e2Ka) . s.16) If the "area" under the potential is denoted by g = 2 lim Voa a+O VO~OO (6. V(x) = g8(x) (6. (6.12).14) The square of the absolute value of this quantity is the transmission coefficient for the barrier. This represents a wave incident from the left and transmitted through the barrier to the right. k sinh 2"" )(F) G (6. We calculate easily: =A cosh 2Ka + i(e/2) sinh 2Ka F e.12) where the abbreviated notation e=k K' has been used. I smh .19) . and Ka « 1.13) K Exercise 6. Calculate the determinant of the 2 X 2 matrix in (6. but Voa or K2a is finite.
21) and (6.12) has very simple properties.20) can be written as (A) B (eiV cosh A = i sinh A ) i sinh A e.23). a matrix M exists such that M G (A) = (M M (F) B M ll 12 ) 21 22 (6.18).10) (a The smooth joining conditions at x = a and x = a lead to two linear homogeneous relations between the coefficients A. We recapitulate the form of the general solution of the SchrOdinger equation: (x < < a) (a < x) x < a) (6. rather than pertaining specifically to the rectangularshaped potential. I ai + f3I  f3~ = 1 I (6. Exercise 6. the Schrodinger equation is invariant under time reversal and space reflection. More generally. Using (6. If we regard the wave function on one side of the barrier. Symmetries and Invariance Properties.24) .4 is a real potential and symmetric about the origin. For the deltafunction barrier (6.20).3 Symmetries and Invariance Properties 99 The matrix that connects A and B with F and G in (6.20) and compare this with (6. as we must be. In the next section it will be shown that the conditions (6. B. Since the rectangular barrier of Figure 6. this condition reduces to . identify A and v in terms of g and k. we must add to this an equation expressing the fact that the determinant of the matrix is equal to unity. say for x > a.23) where A and v are two real parameters. then the coefficients A and B must be expressible as linear homogeneous functions of F and G.22) imposed on (6.21) These five equations reduce the number of independent variables on which the matrix depends from eight to three.22).22) Hence. 3.21) and (6. If the matrix elements are constrained by (6. As can be seen from (6. but we want to see how far we can proceed without using the joining conditions explicitly.12) has the form (6. since the matrix depends explicitly on the two independent variables ka and Ka. and G. we are left with two parameters.12).iv cosh A (F) G (6.4. F. Hence. We can exploit these properties to derive the general form of the matrix linking the incident with the transmitted wave. we observe that the eight real numbers ai and f3i in the matrix satis{y the conditions (6. If we write the linear relations as (~) = (:: : ~:: ::: :::)(~) (6. show that (6.12) and Exercise 6.5. verify that (6. as given. are consequences of very general symmetry properties of the physical system at hand.21).
27) Equation (6. As was shown in Section 6. = IS2Z1 ISl1j2 and ISI2I = = 1 Iszd + IS!zlz (6. we IB I can write this as (B* P)(.ih + B*e ikx = C*e. and (6. and S* the complex conjugate. we obtain the condition IAI Z .) = (A* G*)S*S(~) = (A* G*)(~) (6.10).6. the unitarity condition (6. since IA I and IFI 2 measures the probability flow to the right.26) where S denotes the transpose matrix of S.26). the M matrix is more appropriate in onedimensional problems.30) follow from Since the potential is real. (6. according to Section 3. ljJr<x) = 1jJ*(x) A*e.5. while Z and I G 12 measure the flow in the opposite direction. the symmetry properties are best formulated in terms of the S matrix. Applying expression (6..ikx + G*e ih KX KX (x (a < < a) (a < x) x < a) (6. in a onedimensional stationary state. (6. Using matrix notation.10).28) (6.28). the Schrodinger equation has. the timereversed solution.26) implies the following constraints: Is.29) Verify that the conditions (6. It follows that S must obey the condition IS*S = 11 Since the Hermitian conjugate of the matrix S is defined by Sf = S* = (Sf! sfz S~!) S~2 • (6.+ D*e { Pe. For a 2 X 2 matrix S.30) Exercise 6.IGlz or IBl z + IFl z = IAl z + IGl z Z as expected. The Sand M matrices can be simply related if conservation of probability is invoked. the probability current density j must be independent of x.IBl z = IFl z . in addition to (6.100 Chapter 6 Sectionally Constant Potentials in One Dimension An equivalent representation expresses the coefficients Band F of the outgoing waves in terms of the coefficients A and G of the incoming waves by the matrix relation (.) = (~:: ~::)(~) (6.25) Whereas the representation in terms of the S matrix is more readily generalized to threedimensional situations. On the other hand.4) to the wave function (6.31) .29).1.1 and (6. Such a matrix is said to be unitary.26) implies the statement that the inverse of S must be the same as its Hermitian conjugate.
37) For comparison with (6.33) This condition in conjunction with the unitarity relation (6.25) can be combined to yield the condition (6.:) 1 (6.25) we may make the replacements A ~ B* and F ~ G* and obtain an equally valid equation: (~:) = G:: S*8 = ~::)(.24) and (6. (6. another solution is obtained by replacing x in (6.34) If 8 is unitary and symmetric. The wave Ae. in (6. invariance under reflection implies the relations 8 11 = 822 If conservation of probability.10) shows that effectively the directions of motion have been reversed and the coefficient A has been interchanged with B*. time reversal invariance.ikx is the reflected wave.32) Equations C6. Hence.25) we may make the replacements A and obtain (:) = ~ G and B ~ F G:: ~::) (~) G:: ~::)(~) and 8 12 = 8 21 (6. it is easy to verify by comparing Eqs.25) this relation can also be written as (. and F with G*.ikx is incident from KX + Be ikx = "'Cx) = Ce + De~x { Fe'kx + Ge'kx Ae. Be ikx is the corresponding transmitted wave. the matrix M has the structure· det M = 1 (6.10) by x. and Fe. Ge ikx is a wave incident on the barrier from the left. and invariance under space reflection are simultaneously demanded.26) implies that the S matrix must be symmetric as a consequence of time reversal symmetry: S=8 (6.38) .ikx the right.36) "'R(X) (a> x) Now.3 Symmetries and Invariance Properties 101 Comparison of this solution with (6.25) that the M matrix assumes the form: (6.32j and (6.35) subject to the condition: Since the potential is an even function of x.) = Hence. in (6. This substitution gives (x> a) (a> x > a) (6. Hence.
39) with real parameters alo f310 and f32 subject to the additional constraint f3~ ai + f3i  =1 (6.22). Once either one of . Verify the relations (6. can break time reversal symmetry without violating conservation of probability (unitarity). For all such potentials the solution of the Schrodinger equation must be asymptotically of the form "'(x) ~ Ae ikx + Be. if applicable. which do not depend on the structure of the initial wave packet.22) is that the Schrodinger equation admit two linearly independent fundamental solutions that have the asymptotic property (x ~ ±oo) For a real even potential function V(x). and the energy. We thus see that the conditions (6. invariance under reflection. so that V is merely a function of position. Velocitydependent interactions.23) and that they are equivalent to conditions (6. All that is needed to define a matrix M with the properties (6. thus yielding a condition that is equally prescribed by invariance under time reversal together with probability conservation.ikx { Feikx + Geikx (x ~ 00) (x ~ +00) By virtue of the general arguments just advanced.12) and (6. These general properties are shared by all real potentials that are symmetric with respect to the origin and vanish for large values of Ix I. (6. Although the restrictions that various symmetries impose on the S or M matrix usually complement each other. The matrices Sand M depend only on the nature of the dynamical system. the forces.2. The matrix method of this section allows a neat separation between the initial conditions for a particular problem and the matrices Sand M.22) The same concepts can be generalized to' include longrange forces. If no velocitydependent interactions are present. the reality of V ensures invariance under time reversal and implies conservation of probability. guarantees that the S matrix is symmetric [see the second Eq.38).7. these two portions of the eigenfunctions are related by the equation (6.37)].22) can be derived from very general properties without knowledge of the detailed shape of the potential. (6. this can always be accomplished by choosing where "'even and %dd are the realvalued even and odd parity solutions defined in Section 5. as they occur for instance in the presence of magnetic fields. For instance.21) and (6. Check that they are satisfied by the M matrices in Eqs.21) and (6. It should also be noted that the principle of invariance under time reversal is related to probability conservation and therefore to unitarity.39) and (6. in the simple onedimensional problem treated in this section. they are sometimes redundant.102 Chapter 6 Sectionally Constant Potentials in One Dimension Exercise 6.
Eventually. the particle is unconfined and is scattered by the potential. prove that at a fixed energy the value of the transmission coefficient is independent of the direction of incidence. We will encounter other uses of the M and S matrices in the next section and the next chapter. (See also Exercise 6.44) . the transmission coefficient T is given by IFI2/IA 12 if G = 0. show that M = (~ ~) and S = (~ ~) (6. in Chapter 20. and therefore T = 1~12 = IM: 12= IS2112 1 (6. Noting that the wave number k appears in the SchrOdinger equation only quadratically. as a function of k. E < 0. prove the relation "* (6.'s. we treat this last case first and set hk' = \/2m(E + Yo). we must discuss the socalled square (or rectangular) well (Figure 6. The work of this section is Smatrix theory in its most elementary form.41) If for realvalued V 0.9. The Square Well. Finally. If V = 0 for all x (free particle).3.5). Exercise 6. Exercise 6. we will see that similar methods are pertinent in the general theory of collisions. (6. rand t. If E > 0.41) is measured by two complexvalued functions of the energy.4 The Square Well 103 these matrices has been worked out as a function of energy.42) The analogue of this identity in three dimensions is known as an optical theorem (Section 20. if E < 0.12).) 4.10. which are defined by Sll = 2ir and S21 = 1 + 2it. the S matrix has ~e property S(k) S(k) = 1 (6. For example. the departure of the S matrix from the form (6.43) Derive the corresponding properties ofthe matrix M. prove that. and verify them for the example of Eq.Vo for for Ixl ~ a < x < a a (Yo > 0) Depending on whether the energy is positive or negative.40) Exercise 6. Assuming . (6. where the S or scattering matrix plays a central role.6). we distinguish two separate cases. all problems relating to the potential barrier have essentially been solved.Vo :s. it is confined and in a bound state. It is convenient to place the origin of the x axis in the center of the potential well so that V(x) is again an even function of x: V(x) = {0 . Using conservation of probability and invariance under time reversal only.
In units of h 2 12ma2 .63 and E 1 = 3. to require.... It is necessary to join the wave function and its first derivative smoothly at x = a.81.. and the energies for the two discrete states are Eo = 7...8) e~O lim l/J'(a . The Schrodinger equation takes the form d~~X) + k. but the scale is arbitrary. Inside the well we have l/J(x) = A' l/J(x) = cos k'x B' sin k'x for even parity for odd parity (6.104 Chapter 6 Sectionally Constant Potentials in One Dimension t/J(x) ~_.81 1 .::::l:~ Pt~f+IEl = 3. that is. The ground and excited state eigenfunctions are equally normalized...46) Outside the well we have only the decreasing exponential (6...e) 8+0 = lim l/J(a + 8) = lim l/J'(a + 8) 8~O e .63 L... Square well potential of width 2a and depth Yo...45) _ K2l/J(X) = 0 outside the well As for any even potential.L_1_.L_+_L'::::::::""'''....2l/J(X) d~~X) = 0 inside the well (E < 0) (6..5. For the choice of the parameter (3 = 3 there are two bound states...L_+_.~::::.47) since the wave function must not become infinite at large distances...I E1 =7. O ...'v =9 Figure 6... we may restrict the search for eigenfunctions to those of definite parity.J.. lim l/J(a . the well depth is Vo = 9.
11.24).k' tan k' a for the even case and k' cot k' a for the odd case. f3  _J2mvoa li 2 2 ' (Ka)2 + X = f32 In Figure 6.12).50) (a < By requiring invariance under time reversal and imposing the principle of conservation of probability. defined as in (6. (6. The solution then takes the form (x < a) (a < x) x < a) (6.6 we plot F(X) fevenCx) =.10) we need only replace k by iK and K by ik I.51) and this equation yields the energy eigenvalues. evaluated at x = a. we must thus demand that Mu = 0 (6. The general symmetry considerations of Section 6. The transcendental equations k' tan k'a = k' cot k'a = K K (even) (odd) (6. for a square well.51) is equivalent to the eigenvalue conditions (6. By changing Vo into . respectively. the smooth joining conditions are automatically satisfied at x = a for both even and odd eigenfunctions. these two conditions are equivalent to demanding that the logarithmic derivative of "'. Because of the reflection symmetry. Exercise 6.25). X tan2 = f32 . show that. Show that for the square well. We set X = (k' a)2.49) permit us to determine the allowed eigenvalues of the energy E. The boundary conditions at large distances require that A = G = O.4 The Square Well 105 Since an overall constant factor remains arbitrary until determined by normalization. it follows that the boundstate energies are poles of the S matrix. must now be a real matrix with det M = 1. The logarithmic derivative of the outside wave function. reflection symmetry implies that the offdiagonal matrix elements of M are ± 1.K.Vo in (6. This is a very common way of phrasing the smooth joining conditions.49). Exercise 6. In (6. we see that the matrix M. that ofthe inside wave functions is .3 can also be extended to the solutions of the Schrodinger equation with negative values of k 2 and E.48) be continuous at x = a.X and vX (if tan vX ~ 0) fodix) = X cot2 vX (if tan vX :5 0) . giving us the even and odd solutions. A simple graphical method aids in visualizing the roots of (6. In terms of the matrix M. Defining again an S matrix as in (6.49).12. _1_ d"'(x) = d log "'(x) "'(x) dx dx (6. is .
1. The required roots are found by determining the intercepts of the straight line F(X) with the curves f even and f odd' The ordinates (Ka)2 of the intersection points are the scaled values of the boundstate energies. the only pertinent parameter is the value of the dimensionless quantity {3. The curves for tan Vi ~ 0 (feven) alternate with those for tan Vi sO (fodd)' The ordinates of the intercepts are the binding energies in units of h 2l2ma 2 • The dashedline asymptotes intersect the abscissa at the energy eigenvalues (6. again in units h 2l2ma2 .. As Vo is allowed to increase beyond all bounds.6. so that {3 ~ 00. .106 Chapter 6 Sectionally Constant Potentials in One Dimension as functions of the positive independent variable X. we can immediately draw several conclusions: All bound states of the well are nondegenerate. for a particle confined to the box a s x S a. and. By inspection of Figure 6. odd values of n correspond to odd parity. the level spacing increases with increasing n. . the roots of the equations ( Ka) 2 = E 2 1i2/ma 7T2 "4 I 97T2 4"' I 47T2 40 I I I I I I I I I I I 30 I I I I I I I 20 I I I I I I I I I I I I I I I I I I I I 10 I I I O< '_''O<"'. For the square well. the number of bound states is finite and equal to N + 1.~'_>L_ __'_ _.L""""_ X = (k'a)2 10 20 30 40 Figure 6.52). This figure may be used as a template for estimating the boundstate energies for anyonedimensional square well: merely draw a parallel to the diagonal straight line F(X) = /32 . and n denotes the number of nodes. even and odd solutions alternate as the energy increases. if N7T < 2{3 :5 (N + l)7T.. N.. Graphic determination of the energy levels in a square well with /32 = 30. but there can be no odd states unless {3 > 7T/2.X for the desired value of /3.6. even values of n correspond to even parity. For this infinitely deep potential well.L_ _L. if the bound states are labeled in order of increasing energy by a quantum number n = 0. for anyonedimensional square well there is always at least one even state. two special cases merit discussion: (a) Here we let V o ~ 00 while keeping the width of the square well finite.
or vX = Hence. no eigenstate of the infinitely deep well at E + Vo = [corresponding to n = in (4. but instead of being a repulsive barrier the potential is now the attractive delta function well.00 as V o ~ 00.53) In this limit k' 4 00.. but there is one even root given by or (6. There is in this limit no condition involving the slope.54) Thus.55) 8 ~ if we integrate this equation from x = The result is lim l/J' (x)I~: 8 to X = +8 and then take the limit 0.52) has no degeneracy.O + 2~g It l/J(O) = ° (6.g8(x) (6.52) for odd values of n coincide with the energy spectrum for a free particle whose wave function is subject to periodic boundary conditions (see Section 4. As in (6. whereas (6. There are no odd solutions of (6..17). Note that the number of states is essentially the same in either case.52) corresponding to even values of n. it follows that K ~ + 00.g8(x) = El/J(x) 2 2m dx 2 (6. because the corresponding eigenfunction vanishes.6. E + Vo.. e .59) for all but the lowest level. is the distance in energy from the bottom of the well and represents the kinetic energy of the particle in the well.49) in this case. but between any two levels (4. .. the attractive onedimensional delta function well supports only one bound state. however.. ° ° (b) Another interesting special case arises if Vo tends to infinity as a tends to zero. but in such a way that the product Voa remains finite. h d l/J(x) .. therefore.52) The lefthand side of this expression. we denote the area under the potential by g = lim(2Voa). but k' a ~ 0. and k.56) ... (n + 1) 'TT12 (n = 0. Taking into account a shift Vo of the zero of energy and making the identification 2a = L.59) there lies one given by (6...) (6. 1. the wave function itself must vanish outside the well and at the endpoints x = ±a. There is. which for an infinite potential jump can be discontinuous. 2. we see that the energy levels (6. since there is double degeneracy in (4.59)]. Since E ~ .2a ~ 2mVoalh 2 and remains finite. V(x) = .4 The Square Well 107 expressing the boundary conditions are now simply the asymptotes in Figure 6. This conclusion can be verified directly from the SchrOdinger equation.4).
where fik' = V2m(E + Yo).49) can be cast in the alternate form cos ( k'a .n 2 7T) =  k'a {3 for n . ( ) (6.58). Exercise 6.< k'a < (n + 1)2 2 7T 7T (6. The transmission coefficient T is obtained from (6. for E > O. We may carryover the results for the potential barrier.59) defines the matrix M for the square well if the energy is positive.62) .1 = 8 12 = ~.19).56). v Te' MIl (6. Here E > O.. replacing V o by . Exercise 6. Use Eq.14.57) which is equivalent to the energy equation (6.60) Equation (6. and writing F A = .sm ) 2 F • ie' .13. Equation (6. or directly from (6.59) by choosing stationary states with G = 0 (no wave incident from the right).59). we obtain K= mg fi2 (6.. compute approximate values for the boundstate energy levels in units of fi2/ ma 2..58) Devise a simple graphical method for obtaining the roots of (6.. 'q. we find for the relation between the reflected and the incident wave: (6.3. (6. G (coo 2k'a + 2 sin 2k'a).47) we see that 1/1(0) = C' and e_O lim l/J'(x)I~: = 2C'K Inserting these values in (6.108 Chapter 6 Sectionally Constant Potentials in One Dimension From (6.59) where (6. To conclude this chapter we discuss briefly what happens to a particle incident from a great distance when it is scattered by a square well.. this problem has already been solved.V o and K by ik'. Compare with (6..12) becomes iT/'.56) to derive the transmission coefficient T for the deltafunction potential as a function of energy. Show that the energy eigenvalue equations (6.54). Actually. 2k' a .61) This expression defines a phase shift cP between the transmitted and the incident wave. From the properties of the 8 matrix in Section 6. If (3 = 20.
..8 show this behavior for two different values of T 0.L _ .6 0._.7 and 6.. Transmission coefficient T versus EIVo for a square well with 13'17"/4. and approaches the classical value T = 1 at the higher energies. Transmission coefficient T versus EIVo for a deep square well with f3 As E increases.... 8' ~ 2....2 K X 103 5 10 15 20 25 30 35 40 Vo Figure 6....8...= .. the resonances become broader.4 0.2 1 1M11 (6... fluctuates between maxima (T = 1) at 2k' a = n7T and minima near 2k' a = (2n + 1) ~. T = .. \oooO!<:_'_ _'_'_ _'_'_ _"'_.. and T ~ 1. For the square well..._ .63) 1 cos 2k'a 2 + . as expected.. Figures 6........I.7. ' 1 o 1 '_E 2 Vo Figure 6.' ... As a function of energy. The spikes on the left are at the positions of the seven discrete boundstate energy levels. the transmission coefficient rises from zero. .' ...sin 4 8'2 2 2k'a As E ~ 00.....8 0.......4 The Square Well T 109 J._ _'_ o = 315.
When the peaks in the transmission curve are pronounced they are said to represent resonances. If the well is deep and the energy E low (f3 and 8' » 1).110 Chapter 6 Sectionally Constant Potentials in One Dimension f3. The resonances show up as pronounced points of inflection in the function cP (E). Section 6. as in Figures 6. The maxima occur when the distance 4a that a particle covers in traversing the well and back equals an integral number of de Broglie wavelengths. the peaks stand out sharply between comparatively flat minima (see Figure 6. so that the incident wave and the waves that are reflected inside the well are in phase reinforcing each other. ~~ in units of Vo.9 portrays the energy derivative of the phase shift. dq> X OdE 103 2 o o 25 30 35 Yo E X 103 Figure 6. . for the same square well as in Figure 6. From the matrix M for the square well.8 and 6.8.61) and (6. 1 Exercise 6.64) Figure 6. ~:' in units of Vo for a deep square lThe energy derivative of the phase shift can be related to the time delay suffered by a particle at resonance inside the potential well.64) for the phase shift. derive the expression (6.arctan( ~ tan 2k' a) (6.3.9.59). Show that the expressions (6. compute numerically and sketch graphically the energy dependence of the phase shift cP (E) for E > 0 in the resonance domain.8. We find cP = 2ka .15. v: .62) for the elements of the S matrix follow from the general properties derived in Section 6.9.8). as a function of energy. See Merzbacher (1970). For a square well with f3 = 315. The phase shift cP can also be calculated from (6. The energy derivative of the phase shift. well with f3 = 315.
Obtain the transmission coefficient for a rectangular potential barrier of width 2a if the energy exceeds the height Vo of the barrier. however. but the next chapter will deal with potentials for which this condition is well satisfied. 2. Problems 1.Problems 111 Resonance peaks in the transmission of particles are typical quantum features. Resonances in three dimensions will be discussed in Chapter 13. Classically. Plot the transmission coefficient as a function of EIVo (up to EIVo = 3). it is apparently applicable to a particle moving under classical conditions. choosing (2ma 2 V o)1I2 = (37T12) h. Classically.a) . the density of energy levels in a well increases rapidly with increasing energy.8) does not depend on h and is a function of the particle momentum only. as well as from the analogous discussion in Section 4.4 for the particle in a box. the reflection coefficient (6. From formula (6. Yet classically.4. it is different from zero and varies continuously with energy. for such a barrier. T can only be zero or one.V o for a ::: x ::: 0.g8(x + a) If g > 0. To appreciate the distinction. A particle of mass m moves in the onedimensional double well potential 1 Vex) = g8(x . For example. their verification by extending the solutions of the SchrOdinger for discontinuous potentials to the classical limit meets with some obstacles. T jumps from 0 to I at 'E = Yo.52). Thus. in a certain sense quantum mechanics attributes to matter more continuous and less abrupt characteristics than classical mechanics. we know that in one dimension the levels are spaced farther and farther apart with increasing energy. Not only is the square well unrealistic as a representation of the forces. Consider a potential V = 0 for x > a. This paradox is resolved if we recognize that the correct classical limit of quantum equations is obtained only if care is taken to keep the de Broglie wavelength short in comparison with the distance over which the fractional change of the potential is appreciable. In the example of the potential barrier of Figure 6. depending on the available energy. 3. While these observations have general validity.15) is numerically small for E < Yo. whereas in quantum mechanics T changes continuously between these limits. The SchrOdinger equation for the piecewise constant potential patently violates this requirement.e2i8 2 and show that it exhibits maxima (resonances) at certain discrete energies if the potential is sufficiently deep and broad. in three dimensions. Compute and plot the energy levels in units of h 21ma2 as a function of . although the transmission coefficient (6. and V = + 00 for x < O. Show that for x > a the positive energy solutions of the Schrodinger equation have the form Calculate the scattering coefficient 11 . Hence. and the classical picture is not capable of giving a simple account of such strong but smooth energy variations. V = . one only needs to be aware of the totally different energy spectra of the bound states in a square well in one and three dimensions. R is either o or 1. obtain transcendental equations for the boundstate energy eigenvalues of the system. The transmission resonance theory outlined in this section cannot be expected to provide quantitative estimates for phenomena in the atomic and nuclear domain. but also the limitation to one dimension is a gross distortion of the real systems.
E+.) 6. (d) Calculate the transmission coefficients for particles incident from the right and for particles incident from the left. calculate the transmission coefficient and show that it exhibits resonances. 5. if the elements of the S matrix are expressed as SII = 2ir and S21 = I + 2it. (a) Work out the matrix M. (Note the analogy between the system and the FabryPerot etalon in optics. E_. which have the same energy (but different velocities). 7. verify the identity Irl 2 + Itl 2 = 1m t for the complexvalued amplitudes rand t. and the excited (od< parity) energy level. which relates incoming and outgoing amplitudes. (b) Derive the elements of the matrix S. Problem 3 provides a primitive model for a oneelectron linear diatomic molecule with interatomic distance 2a = lXI. between the wells. A particle moves in one dimension with energy E in the field of a potential defined as the sum of a Heaviside step function and a delta function: Vex) = Vo 1J(x) + g8(x) (with Vo and g > 0) The particle is assumed to have energy E > Yo.(lxl).112 Chapter 6 Sectionally Constant Potentials in One Dimensio the dimensionless parameter maglh 2 • Explain the features of this plot. supplemented by a repulsive interaction Ag/!XI between the wells ("atoms"). For the potentials in Problems 5 and 6. Show that. 2a. the system ("molecule") is stable if the particle ("electron") is in the even parity state. if the potential energy of the "molecule" is taken as E". (c) Show that the S matrix is unitary and that the elements Of the S matrix satisfy the properties expected from the applicable symmetry considerations. 4. which relates the amplitudes of the incident and reflected plane waves on the left of the origin (x < 0) to the amplitudes on the right (x> 0). . If the potential in Problem 3 has g < 0 (double barrier). for a sufficiently small value of A. In the limi of large separation. Sketch the total potential energy of the system as a function of Ixi. obtain a simple formula for the splittin1 tlE between the ground state (even parity) energy level.
A. This equation becomes particularly simple if we use the abbreviations k(x) = { ii~ 2 [E ." u(x).1 for the timedependent wave equation in order to make the connection between quantum mechanics and Hamilton's theory of classical mechanics. varies only slowly with x. It can also be applied to threedimensional problems. Wentzel. and L.E] iK(x) if E < V(x) (7. H.2) into (7.otential energy does not have a very simple form. If V = const.1) is named after its proponents in quantum mechanics. The method is useful to advance our understanding of tunneling through a potential barrier. Substitution of (7. 1.V(x)] } 112 if E > V(x) (7. t) in (3. but has its roots in the theory of ordinary differential equations.5) 113 .2) except that the function u(x) now is not simply linear in the variable x. Kramers. while no longer constant. The WKB method for obtaining approximate solutions of the SchrOdinger equation (7. the solution of the Schrodinger equation even in one dimension is usually a complicated mathematical problem that requires the use of approximation methods. (7.2) is related to the function S(x.3) which explains why the WKB method is occasionally referred to as a semiclassical version of quantum mechanics. if the potential is spherically symmetric and a radial differential equation can be established. we'might try a solution of the form I/J(x) = eiu(x) (7.4) and k(x) = 2m } 112 i{ ii 2 [V(x) . The Method. The function u(x) in (7.Et (7. The same idea was already used in Sections 2. Instead of starting with a simplified potential. the WKB approximation assumes that the potential varies slowly as a function of x.CHAPTER 7 The WKB Approximation If the p.1) has the solutions e Ooikx • This suggests that if V. as perturbation theory (Chapter 8) does.1) gives us an equation for the xdependent "phase. t) = iiu(x) . The basic idea is simple.5 and 3.2) as S(x. Brillouin. G. and exponential decay of an unstable system. The solution of the Schrodinger equation is represented as a modulated constantpotential wave function. resonance behavior in the continuum.
6).6) This differential equation is entirely equivalent to (7. Although the Schr6dinger equation is linear. is a nonlinear equation. The corresponding solution of the Schr6dinger equation is (7. like the classical HamiltonJacobi equation.6). Indeed.8) If V is constant.7) or. If we denote these by u+ and u_. . but in this chapter we will take advantage of the nonlinearity to develop a simple approximation method for solving (7. When we omit this term from the equation entirely. If V varies with x.8) is an exact solution. integrating this. uo. we obtain the first crude approximation. We are led to suspect that this second derivative remains relatively small if the potential does not vary too violently.6).11) ± i k'(x) dx + C1 The two different signs in (7.6) is expressible as u(x) = u+  i In[1 = u_ . as befits a secondorder ordinary differential equation.1).10) \lTC(x) \lTC(x) + i uo(x) dx + C1 (7.6) in the form (7.9) If we substitute the nth approximation on the righthand side of this equation. an iteration procedure is suggested by the fact that u" is zero for the free particle. This would usually be regarded as a drawback. (7. (7. we have for n = 0. but the boundary conditions are more easily expressed in terms of "'(x) than u(x). Ul(X) = ± = ± r r r \lTC(x) + i u~(x) dx + Cn+ 1 (7.10). the general solution of (7.i In[A + Aei(uu+)] + B + ei(u+u_)] + B (7.11) give approximations to two particular solutions of (7. Uo = ± r k(x)dx + C (7.8). (7.12) where A and B are arbitrary constants.1).114 We find that u(x) satisfies the equation Chapter 7 The WKB Approximation (7. and (7. to u: (7. a successive approximation can be set up by casting (7.13) which is a simple superposition of two particular approximate solutions of (7. we obtain the (n + l)th approximation by a mere quadrature: Un+l(X) = ± Thus.
If the WKB approximation is valid. if needed at all. is best accomplished after the desired approximations have been made.15) to (7.16) propagate independently in opposite directions. we may expand the integrand and obtain Ul (x) = I ± x [ = r ±k(x) + i 2 kl(X)] k(x) dx + C1 (7.6) is known as WKB approximation.14) In (7. because it only affects the normalization of l/J(x) .14) can then be cast in the form 1:1« A(x) 1 (7. The approximation (7. unless Ilk'(x)1 « I12(x) I I (7. If condition (7. Condition (7.The Method 115 u(x) is baseless unless Our hope that the approximation procedure (7.14) for the validity of the WKB approximation can be formulated in ways that are better suited to physical interpretation. that is.1~) k(x) dx + ~ In k(x) + C1 The constant of integration is of no significance. we may for E > V(x) define an effective wavelength 27T A(x) = k(x) The convergence criterion (7. Condition (7. which. or that the potential energy change over a wavelength is much less than the local kinetic energy. the potential changes so slowly with x that as the waves propagate no reflected (scattered) wave is generated.11) both signs must be chosen the same as in the Uo on which Ul is supposed to be an improvement.16) In a classically accessible region where E > V(x) and k(x) is real. . It leads to the approximate WKB wave function l/J(x) = dw exp [ ±i r k(x)dx] (7.10) will tend toward the correct Ul(X) is close to uo(x).14) can also be written as I: I « « Ip(x) I or A(x) 1:1 [p~f implying that the change of the "local momentum" p(x) = hk(x) over a wavelength must be small by comparison with the momentum itself.14) holds. the two waves (7.17) requiring the wavelength to vary only slowly. If k(x) is regarded as the effective wave number.
. when the classically impenetrable barrier lies to the right of the classical turning point. and D unambiguous. v(x) V(x) ~E E. The Connection Formulas.. where E < V(x).21) The lower limits of the integrals in the exponents have been arbitrarily.19) The socalled connection formulas serve to link WKB solutions of the type (7. (a) Classical turning point at x = a..16) in its more recognizably real form "'(x) = ~ exp [ ± r K(X)dx] (7. (7. to the right of the classically accessible region." . If the WKB approximation can be assumed to be applicable except in the immediate neighborhood of the turning point.1.16) in the classically accessible region of x with solutions of type (7.116 Chapter 7 The WKB Approximation These conditions obviously break down if k(x) vanishes or if k(x) varies very rapidly. chosen to make the meaning of the amplitudes A. it is appropriate to rewrite the WKB wave function (7. we have "'(x) = ~VKW exp [.14) breaks down.18) or whenever V(x) has a very steep behavior. C.20) and "'(x) = ~v~ eXP[i Ja k(X)dx] + ~v~ eXP[i Ja k(X)dx] r r k(x) k(x) for x « a (7.1b).I. The WKB method is not particularly useful unless we find ways to extend the wave function into and through these regions.34) and (7. VKW for x »a (7. but conveniently. to the left of the classically accessible region. . This certainly happens at the classical turning points for which V(x) = E (7.x a (a) ~x (b) Figure 7.. Generally.. The point x = a separates the two regions where E > V and E < V.1 ~ r K(X)dx] + ~ ~ exp [ 1 K(X)dx] r . The results for the two cases will be summarized in Eqs.1a. as shown in Figure 7. Suppose that x = a is a classical turning point for the motion with the given energy. B. In the nonclassical domain. 2. E. A more accurate solution must be used in a region where (7.35).19) in the classically inaccessible region. in these regions we expect that waves propagating in one direction will generate reflected (scattered) waves. (b) Classical turning point at x = b. Analogous considerations hold if the barrier is to the left of the turning point x = b (Figure 7.
4.}) 2mg z.14) implies Iz1 312 » 1/2 Thus. but an analytic approach works if a somewhat special. (7.23) is conveniently transformed by substitution of the dimensionless variable z = ( .24) 113 (x .27) and (7. assumptien is made about the behavior of the potential energy near the turning point. The solutions of the differential equation (7.2m dx 2 + g(x .23). yet often appropriate.26) and that the WKB condition (7. Section 34. (7. and useful historical references. see Schiff (1968). applicable where Izi becomes large. To establish the connection between the two separated regions. we must solve the Schrodinger equation more accurately than by the WKB approximation. we may write V(x) .25) Note that ~ = (2~g)Wz (7.28) where C= 2 lz1312 3 (7. 2Abramowitz and Stegun (1964). Asymptotically.E = g(x .2 The Connection Formulas 117 We now ask the fundamental question: How are the coefficients C and D related to A and B if (7. Section 10.! We suppose that in the neighborhood of x = a.a) (7. the leading terms are: (7.22) where g > O.20) and (7. This can always be done numerically. for large positive z.a)l/J = 0 (7. since k(x) ~ 0 implies an unphysical divergence of If!(x). albeit in different regions? The inadequacy of the WKB approximation near the turning point is evident. fi2 d 2 l/J . The Schrodinger equation for this linear potential.a) into the form (7.25) are the Airy functions 2 Ai(z) and Bi(z).21) are to represent the same state.. the WKB approximation is simply an asymptotic approximation for the solutions of Eq.29) IFor a more general treatment of the WKB approximation at the turning points. .
1I4 cos((  ~) ~) (7.8 0. The approximations diverge at z = O. and (b) the Airy function Bi(z). For large negative values of z: Ai(z) == 1T 1I2 Izl.2.Chapter 7 The WKB Approximation Ai(z) 0. (a) The Airy function Ai(z).114 sin( ( .3 Ri(z) == _1T 1I2 Izl. for realvalued z.30 (7.6 (0) Figure 7. and their asymptotic (WKB) approximations.
(7.5 1. the two functions agree very well. Since for the assumed linear potential (7. (continued) Figures 7.2b are plots of the Airy functions Ai(z) and Bi(z) and their asymptotic (WKB) approximations. Except for z = 0.2a and 7.25 0.25 0. WKB 1.5 (b) Figure 7.g)112 (a  X)3/2 = 2 3"l zI312 = .2 The Connection Formulas Bi{z) 119 1.22).5 1. L a k(x)dx = 3" 2(2.32) .5 Bi(z). where the asymptotic forms diverge. for z < 0.2.
Jx K(X)dx) + ~v~ exp ( Jx K(x)dx) ~ r r K(X) K(X) b b ~ cos( Jb k(x)dx r Vkfx) Exercise 7.1(a) into Figure 7.g) lIz(x  a)3/Z = ~ Iz1 3 2 / =l (7.20) and (7.a ~ 'b . To see this. for instance. Since in this region the decreasing exponential is naturally much smaller than the increasing exponential. !!) .20) and (7. suppose that we know that in the region x » a in Figure 7.33) we verify that the asymptotic forms (7.1(b).20) and (7. are ~V~ exp (. This happens. which separates a classically inaccessible region x < b from the accessible region x > b. show that the connection formula (7.34).Vkfx) 4 _ B sin( Jb r k(x)dx  !!) 4 (7. Neglecting the decreasing exponential could then lead to a gross error in the WKB estimate in the classically allowed region x « a.35) By making the transformation x .Ja K(X)dx) + ~V~ exp ( Ja K(X)dx) r r K(X) K(X) The analogous connection formulas for a classical turning point x = b. the contribution of the decreasing exponential to the wave function may be dwarfed even if the coefficient A is comparable in magnitude to B.1. derive the coefficients C and D in terms of A and B.1(a) the wave function is adequately represented just by the increasing exponential. It is then necessary to resort to more accurate analytical or numerical solutions of the Schrodinger equation. which link WKB wave functions between different regions of real and imaginary k(x).ntribution may dominate the behavior of the wave function in the region x » a.21).120 and for Chapter 7 The WKB Approximation z > 0. (7. when the energy E is close to an extremum of the potential.x which turns Figure 7. but as the coefficient of the increasing exponential component of the wave function. . this "small" co.31) are nothing but WKB wave functions of the type (7. Caution must be exercised in employing these WKB connection formulas. neglecting it entirely is justified in the region x « a.27)(7. f K(x)dx = ~ (2.35) follows from Exercise 7. if Bis finite but IBI « IA I.34) with (7. By comparing the asymptotic expressions for the Airy functions with (7.2. we learn that the WKB wave functions on the two sides of the turning point x = a are connected as follows: ~ cos( fa k(x)dx Vkfx) Jx _ !!) __ Sin( Jxr Vkfx) 4 B_ x k(x)dx  !!) ~ 4 (7. Conversely.21).21). The connection formulas.34) ~V~ exp (. By comparing (7. break down if two neighboring classical turning points are so close to each other that there is no WKB region between them.
~ cos(I: + k(x) dx) sin(f k(x) dx  ~) ~ sin(I: k(x) dx) cos(f k(x) dx  ~) Figure 7. Care is required in continuing the WKB wave function through such singularities. and 3 away from the turning points.) 4 x .3. with B = 0. Thus. 3This proposition is equivalent to the constancy of the Wronskian for the Schrodinger equation (7. to satisfy the boundary condition as x ~ . Simple onedimensional potential well. but also at the singularities of the potential Vex). The basic idea emerges if we choose a simple wellshaped potential with two classical turning points as shown in Figure 7. we observe that the WKB approximation presumes that k(x) is an analytic function.3 Application to Bound States 121 Finally. Classically. (7. Exercise 7. by Eq. the unnormalized WKB wave function in region 1 is l{1 = ~ exp (  J: K(X) dx ) for x < b Hence. and the connection formulas will serve near x = a and x = b. . in region 2. Application to Bound States. This property fails not only at the classical turning points.3. The WKB approximation will be used in regions 1. 3 3. See Bradbury (1984).1).7). The WKB approximation can be applied to derive an equation for the energies of bound states. and thus with conservation of probability. which is of the SturmLiouville type.35).2. even across classical turning points. Section 10. Chapter 7.3. l{1 = ~ cos( ~ Jb r k(x) dx  7!. a particle of energy E is confined to the region between a and b.0 Jb k(x) dx  fa k(x) dx . Show that the WKB approximation is consistent with the generalized continuity equation (3. The usual requirement that l{1 must be finite dictates that the solutions which increase exponentially as one moves outward from the turning points must vanish.00.) 4 for b < x < a This may also be written as l{1 = _2_ cos( r \/k(.7!.
4). In Figure 7. the first term must vanish. The area between two neighboring trajectories is equal to h..4).4. .4... The energy E appears in the integrand as well as in the limits of integration. The expression (7. If we introduce the classical momentum p(x) = ±hk(x) and plot p(x) versus x in phase space. We obtain the condition (7. Phase space representation of the periodic motion of a particle confined between the classical turning points at x = an and x = b n for the boundstate energies En = yean) = V(b n). . As for free particles (Sectiop 4.36) measures the phase change that the oscillatory wave function'" undergoes in region 2 across the well between the two turning points. Hence. 2. which occupied a position intermediate between classical and quantum mechanics. n represents the number of nodes in the wave function. in the WKB approx p x Figure 7. the area of enclosed phase space between the closed curves for n + 1 and n is equal to h.36) may be written as (7.34) only the second of these two terms gives rise to a decreasing exponential in region 3 satisfying the boundary conditions as x .2) in the old quantum theory. If the WKB approximation is used all the way from b to a. It is then evident that condition (7. Dividing this by 21T. we see that according to the WKB approximation nl2 + 1/4 quasiwavelengths fit between b and a. (7. Hence. since the turning points a and b are determined such that V(a) = V(b) = E. a fact that helps to visualize the elusive "'.36) where n = 0.122 Chapter 7 The WKB Approximation By (7.37) equals the area enclosed by the curve representing the periodic motion in phase space and is called the phase integral J in classical terminology. the bounded motion in a potential well can be pictured by a closed curve (Figure 7. This equation determines the possible discrete values of E.37) This equation is very similar to the quantum condition (1. 1.? + 00.
This condition implies that the energy eigenfunction must be the Airy function Ai(z).23).38) is the same as Eq. (7. lIv(x) ex: [E . this is proportional to the length of time (dt) that the particle spends in the interval dx.38) by scaling and displacing the independent variable. and the graph of the solution in Figure 7. we compare the exact solutions of the Schrodinger equation with approximate WKB energies and wave functions for the bound states of a particle of mass m in a potential well defined by Vex) = g Ixl. ) d2 l/J dz2  (2m/)113(x . because for high energies l/J has a very short wavelength in the classically accessible region.g) 1I3(x . It is a rapidly oscillating function of position. though basically different. In statistical mechanics. are nevertheless related to each other in the limit in which the rapid phase fluctuations of quantum mechanics can be legitimately averaged to give the approximate classical behavior.40) produces the differential equation for l/J(z): (7.V(x)r l12 • Classically.38) for x > 0 and characterize the even and odd parity solutions by imposing the boundary conditions at the coordinate origin: (7. VCr) = Cr. if we identify the energy E As in Section 7. Such a Vshaped potential is the onedimensional analogue of the linear centralforce potential. and for all boundstate energies. (7.39) For x> 0. We thus see that the probability concepts used in quantum and classical mechanics.2. The strength of the potential is measured by the positive constant g. to which the confinement of the quarkantiquark constituents of the charmonium "atom" is attributed. zl/J = 0 a) (x> 0) (7. It is sufficient to solve the SchrOdinger equation for the onedimensional linear potential.3 Application to Bound States 123 imation each quantum state may be said to occupy a volume h in twodimensional phase space.41). 1l/J12 dx. The probability.. this rule is useful in the domain where classical mechanics is applicable but some concession must be made to the quantum structure of matter. There are no parameters in (7.41) The boundary condition at large x requires that l/J(z) must vanish asymptotically as z ~ + 00. The WKB method is fittingly called a semiclassical approximation. . g. the substitution = gao z = (2.2 can be used for all values of the potential strength. As an illustrative example. which is related to the original SchrOdinger equation (7. (7.41). E. of finding the particle in an interval dx at x is proportional to the reciprocal of the classical velocity. but its maximum amplitude is modulated slowly by a factor 1IVkfx). It thus represents the relative probability of findinR the particle in the interval dx if a random (in time) determination of its position is made as the particle shuttles back and forth between the turning points. It is interesting to observe the universality of the differential equation (7.
42) are. Except for the ground state energy.g2fi.11. ] =0 (7.[(~gr/3 ~ ] and for the odd eigenfunctions (7.1 n o 1 2 3 4 0.5781 3.8306 If the coordinate x and the length Eo . and are thus made dimensionless.gx) dx = (n + D~ = 32 ~2 n From this condition we obtain the WKB estimate: 9 ( (EWKBi +2 1)2 .:. Table 7.2(a).39) at the origin now require that for the even eigenfunctions l/I~(O) = Ai.124 Chapter 7 The WKB Approximation The unnormalized energy eigenfunctions are l/I(z) original x coordinate (x > 0): = Ai(z).8416 2. expressed as 4Por numerical tables. for x > 0. Table 10. see Abramowitz and Stegun (1964). are scaled in units of!!:.36).1.8558 2.~ ) ] =0 (7. or in terms of the l/In (x) = A{ e~gyl3 (x . the unnormalized energy eigenfunctions functions (7.. which in the present context translates into 2 L Elg ~~ (E .42) The conditions (7.2446 3.( g mg 2)113 . the agreement with the "exact" numerical values is seen to be excellent.2 (7.43) l/In (0) = A{  e~g y/3 E.44) The (negative) zeros of Ai(z) and Ai'(z) may be computed or read off Figure 7.45) The values of E WKB for n = 0 to 4 are entered in the third column of Table 7.8257 0. the approximate energies are extracted from the condition (7.2397 3.5888 3..4 The five lowest energy levels are listed in the second column in Table 7. In the WKB approximation.8853 1. .8086 (ground state) 1.1.
periodic. Le. the WKB. The Hamiltonian of a multidimensional system need not be particularly exotic for the trajectories in phase space to display a far more complicated character than illustrated in Figure 7.2(a).1. it is possible to generalize the WKB method if due attention is paid to the singularities that are the analogues of classical turning points. respectively. (The ground state eigenfunction for the potential V(x) = glxl is shown in Figure 8. The WKB method will now be applied to calculate the transmission coefficient for a barrier on which particles are incident from the left with insufficient energy to pass to the other side classically. Compute and plot the WKB approximation to the eigenfunctions for n = 0 and 1. It is useful to remember that he = 0.) is made.7 GeV. Noting that the two lowest S states have measured rest energies of 3.2 GeV .6. 'Gutzwiller (1990).4 Transmission Through a Barrier 125 These functions can be evaluated from Figure 7. Section 6.. Semiclassical quantum mechanics for complex systems has benefited greatly from recent advances in (nonlinear) classical dynamics and constitutes a subject beyond the scope of this book. and the classical probability densities of finding the particle at position x. Exercise 7.4. For n = 3 and 4 compare the exact eigenfunctions with the WKB wave functions. the S states of the charmonium "atom" are regarded as the energy levels of a charmed quark and antiquark. is relatively straightforward for systems that are described by or reducible to a onedimensional SchrOdinger equation. Transmission Through a Barrier. obtain an estimate for the potential strength parameter g in units of GeV/fm. In a crude model. This problem is very similar to that of the rectangular potential barrier.4.5.1 and 3. For these two energy levels.) Exercise 7. Show that the WKB approximation gives the energy levels of the linear harmonic oscillator correctly. For the potential V(x) = glxl. (Treat this twobody problem as an effective onebody problem with a reduced quark mass.) Exercise 7. fm. if the substitution z = ~(x . half the rest mass of the charmed quark. with its connection formulas. and compare with the exact stationary state wave functions. The WKB method. . sketch the exact quantum mechanical.2.2(a) to obtain) the energy eigenfunctions for n = 0 to 4 and plot them. but no special assumption will be made here concerning the shape of the barrier. Explain why the WKB wave functions have a discontinuous slope at the origin for even values of n. and ignoring relativistic effects on the binding. When the classical system is integrable and its motion multiply . bound by a onedimensional potential that (except near the origin) is represented by V(x) = g Ix I. s 4. compute (or use Figure 7. but in practice one deals frequently with systems exhibiting classically chaotic motion.
7. The result of the calculation is remarkably simple and again is best expressed in terms of a matrix M that connects F and G with A and B.46) V The connection formulas (7.J.~+E _ _ _ _ _ _J' ..34) and (7.46) in much the same way as was done in Chapter 6 for the rectangular barrier. Exercise 7.48) measures the height and thickness of the barrier as a function of energy.. Verify (7.47) where the parameter o= exp(J: K(X)dx) (7.35) can now be used to establish linear relations between the coefficients in (7..5.. The transmission coefficient is defined as T = Il/Jtransl 22 vtrans = Il/Jtransv'k::1 = IFI 2 Il/Jinel Vine Il/JineVk:1 2 IAI 2 2 . Potential barrier..47)...126 Chapter 7 The WKB Approximation .5. the solution of the Schrodinger equation may be written as ~v~exP(i Ja kdx) + ~v~exP(i Ja kdx) ex r k(x) k(x) l/J(x) = (x (a (b < < < a) x x) ~v~ exp (. X a b Figure 7.Ja dx) + ~~ exp ( Ja dx) ex r K(X) v K(X) ~v~ eXP(i Jb kdx) + ~~ eXP(i Jb kdx) (X r k(x) k(x) K K < b) (7. (7. If the WKB approximation is assumed to hold in the three regions indicated in Figure 7.
we calculate 8 for a onedimensional model of a repulsive Coulomb barrier (Figure 7. which repels particles incident from the left._ _l.47) we 1 4 2 ( (7. The essence of the calculation survives the generalization to three dimensions (Section 13...z'2 e2 x (7.L_. let Vbe defined for x < 0 as v= The turning point a is determined by E _ Zz.8). 8 is a measure of the opacity of the barrier.51) = _ Zl~e2 a v 4 v=_l x 3 Ef 2 _ _l .50) Hence...4 Transmission Through a Barrier 127 assuming that there is no wave incident from the right. As an example.J.6).. and (7. From (7.L_~x 4 3 2 1 a o Figure 7. . .. G obtain T = .49) 1 28 +) 2() For a high and broad barrier. .6.. Thus. () » 1.. which a charged particle such as a proton (charge Zle) has to penetrate to reach a nucleus (charge Z2e).2 = I Mil 1 O. Onedimensional analogue of a Coulomb barrier.
52) The barrier inhibits the approach of a positive charged particle to the nucleus. (3) b < x < a.7. and the transmission coefficient is called the penetrability. .8.e. Apply the result to the calculation of the Coulomb barrier penetrability for an alpha particle (ZI = 2) with asymptotic kinetic energy E in the repulsive field of a nucleus (Z2). Express E in MeV and R in Fermis. that b.++\E c x "1'12 'I 3 ·1 4 5 ·1 617 Figure 7. with nuclear radius b = R. (7. Potential barriers surrounding a well are favorable for the occurrence of narrow transmission resonances.128 Chapter 7 The WKB Approximation and we take b = 0. It will be assumed that Vex) is symmetric about the origin. cannot escape unless it penetrates the surrounding Coulomb barrier. callously disregarding the warning signals that caution us about applying the WKB approximation near a singularity of the potential energy. and it is also decisive in the description of nuclear alpha decay. b = sa. i. Hence. (7) c < x. fr. since the alpha particle..7. once it is formed inside the nucleus. more realistically.51) by assuming. A wave packet is seen to be incident from the left. instead of being zero is equal to a fraction of a. As a further application of the WKB method. and that V = 0 outside the interval between c and c. (6) b < x < c. The critical integral is then Ja (0 K dx = V2mE (0 Ii Ja J~ x  I dx = {E fhii ZIZ:ze Ii 2 J! Jo ( 1 du = ZIZ:ze 7T liv 2 u where v = V2E/m is the classical particle velocity at x ~  00. Regions 1 through 7 are defined as: (1) x < c. which is located in the center of the well. let us consider the passage of a particle through a potential well that is bounded by barriers as shown in Figure 7. Exercise 7. Calculate the transmission coefficient for the model Coulomb potential (7. (5) a < x < b. (2) c < x < b. (4) a < x < a. This quantity determines the probability of nuclear fusion.
3.(4~ + 4~) cos L. the following abbreviations have been used: 6 L = fa k(x) dx.36) for bound states.7. whenever cos L L = 0. (7.isin L According to (6. .58) 2 L = This quantity reaches its maximum value. If (»> 1.22). unity. From Eq.i </> = _:. so that penetration 60ur notation is adapted from a thorough discussion of barrier penetration in Bohm (1951).56) We will shortly make use of this property.53) (x> c) vk vk When the WKB method is applied to connect the wave function in regions 1 and 7. .. the transmission coefficient is T =1 2 Mil = 1 1 ( 4&2 + 4&2 1 )2 cos L 2 + 4sm . exp( zkx) A7 • r. that iJL >0 iJE (7.59) (2n + 1) 7T/2 The condition determining the location of the transmission peaks is seen to be the same as the quantum condition (7.kc (7. exp(zkx) B (7. ~ BI r.40). as was pointed out in Section 6.57) 2 1( 4&2 + 4&2 4 1 ) cos L . We are particularly interested in the form of the energy eigenfunctions in regions I and 7: l/J(x) = • j ~ ~ AI r.4 Transmission Through a Barrier 129 In this section. the matrix that links the asymptotic parts of the Schrodinger eigenfunction has the same general form for all potentials that are symmetric about the origin. (7. condition (6. the effect of barrier penetration will be studied for a particle with an energy E below the peak of the barriers. The final matrix relation (7.39) subject to the. or (7.54) we obtain. (x < c) vk vk 7 + ~ r.2 Al M ll e2~ _ (7.55) It follows from the definition of L and from inspection of Figure 7.1 = VTe. p = f k(x) dx . for B 7 = 0.4(f cos L In writing these equations..54) has the form (6. exp(zkx) + . exp( zkx) . This result is expected since. A = . the relation between the coefficients is again most advantageously recorded in matrix notation: At) = ! (e.2i sin L] (B 2 (1 ) 2i { t i 4(f .
The attractive potential inside the nucleus has a mean depth of Vo = 65 MeV. which at t = 0 is localized entirely in region 1 near the coordinate Xo « 0 far to the left of the barrier and moving toward positive x.Eo). and show that near Eo it has the characteristic BreitWigner resonance shape.130 Chapter 7 The WKB Approximation through the barriers is strongly inhibited.8. Motion of a Wave Packet and Exponential Decay.9. is approximately Show that the energy spacing D between neighboring resonances 11" D= iJLliJE 1). Compare with Figure 6.62) .10.7. as shown in Figure 7. It is instructive to consider the motion of a simple broad wave packet incident with well defined positive momentum ftk o = Y2mEo from the left (from region 1 where x < c) onto the well protected by a barrier. Exercise 7.7. T has sharp.57) and evaluating the slowly varying quantity fJ at E = Eo.60) where by definition r = 1 fP(iJLliJE)E=Eo (7.62) and that for low barrier penetration (fJ » well separated. r being its width at half maximum. we get (7. cos L = :+ (aL) aE E=Eo sinL=±1 Substituting these approximations in (7. Apply the resonance approximation to the transmission coefficient T. may be represented by the wave function in kspace: (7.8. Are there other resonances nearby? 5. » r. A graph of T in the resonance region will be similar to Figure 6. narrow resonance peaks at these energies.11. it may usually be assumed that in the vicinity of the resonances in a reasonable approximation.61) Exercise 7. Under conditions favorable for the occurrence of pronounced resonances (fJ » 1). D (7. A nucleon of energy E is incident on a onedimensional toy model of a "nucleus" of radius R = 4 fm (diameter 2R). Approximately rectangular barriers of average 8 fm width and 5 MeV height bound the nucleus symmetrically. so that the resonances are Exercise 7. so that the potential looks qualitatively like Figure 7. Estimate the value of the barrier opacity rP and of L as a function of E. Calculate the energy and width of the lowest nucleon transmission resonance below the barrier. A wave packet. (E .
Elsewhere it vanishes owing to the rapid oscillations of the exponential function.63). . 0) as a superposition of infinite plane waves.k = Y2mE to express !/J(x.k.64) The representation (7.k. t) = I: f(E)YT exp(  ic/J) exp [ik(X  Xo)  k EtJdE (x > c) (7.53). the transmitted wave function at arbitrary times t > 0 is: !/J(x. asymptotically (ixi ~ 00) we can use the freeparticle relation fi. it vanishes in region 1. whose asymptotic form is given by (7. Equation (7. 0) = f: f(E)e.2ko k5 . eik(X+Xo)dE o B2 (7. Le.60) into the integrand of (7. k is not constant. 0) is normalized to unity. 1 (00 dk = y'2. in the presence of the potential V(x)..63) f(E)eik(xXo) dE 0 The amplitude f(E). 0) = y'2.ikxo ( e ikx + ~: e ikx )dE (x < c) (7.E of the packet considerably exceeds the width of the resonance (but is much smaller than the interval between neighboring resonances). In the asymptotic region 7 to the right of the barrier.65) holds because the integral 1 00 B f(E) !.66) differs from zero only when the phase in the integrand is stationary. 0) in terms of the appropriate WKB wave functions. we include only WKB wave functions with B7 = O. If !/J(x. We therefore are entitled to substitute (7.E = voh/i. as exemplified in Figure 2.65) where the kdependent coefficients A l and B l are the same as those that appear in (7. f(E) satisfies the normalization condition (00 Jo 1 f(E) 1 Vo dE 2 = 27Th 1 (7. We also assume that the width /i. Since the wave packet is narrow in k space.. 0) as an integral over E instead of k.63) expresses !/J(x. hvo 1 1 00 I c/J (k) leik(XXo) dE 0 1 00 (7. Though.xo.5 Motion of a Wave Packet and Exponential Decay 131 Here Ic/J (k) I is a smoothly varying function with a fairly sharp peak and a width /i. is a smoothly varying positive function of E with a fairly sharp peak and a width /i. for values of x near x = . defined in (7.1. Jo Ic/J (k) leik(XXo) = . = o + k 2  E livo The initial wave packet can then be written as !/J(x. Since no wave is incident from the right. but we need an expansion of !/J(x. the expansion has the form !/J(x.67) In order to study the behavior of this transmitted wave packet near a very narrow resonance we assume that the mean energy Eo of the incident wave packet corresponds to a resonance.67).k5 k + ko = ko + JC. which vanishes for k < O. the two variables are related approximately as k = k o'+ (k .53). In the asymptotic region 1. Hence.ko) = ko + JC.
.. the probability density increases exponentially as a function of x (up to the front edge of the wave front).68) In (7..00 without appreciable error. the integration has been extended to .t) }XP[iko(X  xo)]eiEOtlli ift>X.t) 2 of a decaying state for three different times..132 Chapter 7 The WKB Approximation Except for uncommonly longrange potential barriers.. (x . the probability 1 density decays exponentially in time... Eq. and equal to Po. t) < t 2 < t 3 • At a fixed position x.) The integral in (7. The result is that in the asymptotic region x > c. over the width of the resonance.L_. assuming that t is not too large. (A.f(Eo)exp [ i ..70) x = Xo Figure 7. t) = exp[~ (x :o.L.. (For extremely long times t . the results to be derived may therefore be suspect.22). after the arrival of the wave front. We may calculate the total probability that at time t the particle has been h/J(x. the phase P may be assumed constant.. . The probability density Il{I(x. t) = +._x Figure 7. At a fixed time.xc Do {o ift<Xxo Do (7.Xo) + 2ipo '2 J~ exp[kE(X :oXo _~ E _ Eo + if/2 t)] dE (7.68).69) This wave function describes a wave packet with a discontinuous front edge at + Do t and an exponentially decreasing tail to the left. ±7Tiff(Eo)e2iPO l{I(x.8 shows the distorted wave packet at various times after it has reached region 7. the wave function in region 7 at positive t becomes ] f k l{J(x.68) is a wellknown Fourier integral that is evaluated in the Appendix.xc . 00. After the pulse arrives at a point x the probability density decays according to the formula (7.. t)12 fo"'==~_.8. With these approximations.
73) The remaining portion of the wave packet is reflected promptly.64) we obtain as a crude estimate Hence. and not a rigorous. Construct and interpret the asymptotic solution of a wave equation (timedependent SchrOdinger equation) which corresponds to one of these complex E values (decaying states). Equation (7.12. Exercise 7. . result of quantum mechanics.exp [ f (t + ::)]} (t> ::) Here we have assumed that c « xo + vot. The study of resonance transmission affords us an example of the familiar exponential decay law. Resonances in the double well (Figure 7. but it is well to point out here that the exponential decay law can be derived only as an approximate.. an order of magnitude estimate for the probability that transmission has occurred is (7. A fraction r IliE of the packet is transmitted according to an exponential time law with a mean lifetime T = !!:. Show that this condition defines poles in the S matrix as a function' of the complex variable E. whereas the imaginary parts are the halfwidths.71) this probability is L Xo vot + lifJ(x. is thus approximately equal to rlf:t. The real parts of these discrete complex values of E are the resonance energies. Show that condition (7.71) implies that the time it takes the incident wave packet to enter the well must be long compared with the classical period of motion and short compared with the lifetime of the decaying state. Decay processes will be encountered again in Chapters 13 and 19. Exercise 7. (6.72) leads to the following simple interpretation: The wave packet reaches the well at time xolvo. so that the lower limit of the integral can be set equal to zero. = hfi2(OL) r oE E=Eo (7. It holds only if the decay process is essentially independent of the manner in which the decaying state was formed and of the particular details of the incident wave packet.7) may also be defined as quasibound states by requiring Al = B 7 = 0 (no incident wave) or M ll = 0 in analogy to truly bound states [see Eq. For a wave packet whose energy spread liE covers a single resonance such that D» liE» r (7. t)~2 dx = 7(2hrvolf(EOW{1 .63).13.51)]. and the well with corresponding barriers can serve as a onedimensional model of nuclear alpha decay. From the normalization (7.72) The total transmission probability for the incident wave packet (7.E. found by letting t + 00.5 Motion of a Wave Packet and Exponential Decay 133 transmitted and is found anywhere in region 7.
and show that the energy eigenvalues are determined by a condition of the form tan(f k dx + a) = ±28 where 8 is the quantity defined in (7. a is a constant dependent on the boundary conditions. c being a positive constant. In this interval the potential energy is V = clxl. 2. Determine the WKB wave functions for positive and negative values of the energy. Compare with the rigorous solutions of this problem. A particle of mass m moves in one dimension between two infinitely high potential walls at x = a and x = a. obtain an equation determining the energy eigenvalues E :S O.134 Problems Chapter 7 The WKB Approximation 1. 3. Apply the WKB approximation to a particle of mass m moving with energy E in the field of an inverted oscillator potential. . V(x) = mw2x 2 /2.48) for the barrier. 4. Estimate the minimum value of C required for an energy level with E :S 0 to exist. and the integral f k dx is to be extended between the clas sical turning points in one of the separate wells. In the WKB approximation. Estimate the limits of the region in which the WKB wave functions are expected to be valid approximations to the exact Schrodinger wave function. Show that at low transmission the energy levels appear in close pairs with a level splitting approximately equal to fiw/'TI'8 where W is the classical frequency of oscillation in one of the single wells. E. Apply the WKB approximation to the energy levels below the top of the barrier in a symmetric double well. Apply the WKB method to a particle that falls with acceleration g in a uniform gravitational field directed along the z axis and that is reflected from a perfectly elastic plane surface at z = O.
2) stationary in the mathematical sense of the term. 1. the Coulomb potential. Simple perturbation theory. The RayleighRitz variational method provides a bridge from wave mechanics to matrix mechanics. will be studied in Chapter 12.CHAPTER 8 Variational Methods and Simple Perturbation Theory Variational methods are as central to quantum mechanics as to classical mechanics. They also serve as a springboard for numerical computation.3) 135 . introduced in this chapter.2) to vanish. The WKB method (Chapter 7) is limited to models that tend to be oversimplified. The timeindependent SchrOdinger equation can be regarded as the EulerLagrange equation for a variational problem that can be formulated as follows: Subject to the constraint (8. we will consider some very general variational methods. The last three chapters dealt with three of these: the harmonic oscillator. the piecewise constant potential. A more systematic study of perturbation methods is the subject of later chapters (18 and 19).1) find the functions l/J(r) and l/J*(r) that cause the variation oJ of the expression (8. and the linear potential. The Calculus of Variations in Quantum Mechanics. More graphically. When quantum mechanics is applied to realistic physical systems. which make it possible to obtain approximate energy eigenvalues and which also shed light on the physical properties of the stationary states. the expression (8. greatly extends our access to interesting applications. If l/J(r) and l/J*(r) are sufficiently smooth and vanish properly at the boundary. The molecular approximation capitalizes on the mass disparity between electrons and nuclei. Explicit analytic solutions of the SchrOdinger equation can be derived only for a limited number of potential energy functions. we must usually employ approximation methods. Another.2) can be transformed by integration by parts into J= J (. This chapter concludes with examples of applications to molecular structure and the band theory of solids.~~ l/J*rpl/J + Vl/J*l/J)d r 3 (8. In the meantime. but somewhat less precisely: Find the functions l/J(r) and l/J*(r) which make the integral (8.
so that I = (H). (8. as if they were two unrelated functions with their conjugate relationship temporarily ignored. this quantity has a lower bound. two differential (Euler) equations are obtained:  aF al/f* dx al/f*' aF d aF al/f . the Lagrangian multiplier.2) has precedence whenever there are ambiguities.4) where E is an as yet undetermined constant.1) and using a Lagrangian multiplier. Since l/f(x) and l/f*(x) satisfy the appropriate regular boundary conditions. As the energy. For an understanding of the variational principle and its implications.. Hence. The equivalence of the variational principle and the SchrOdinger equation can also be demonstrated without constraining l/f(r) and l/f*(r) as in (8. Both of these equations have the form of the timeindependent SchrOdinger equation..2) and (8. all equations are consistent if l/f*(x) is chosen as the complex conjugate of l/f(x). which is not invariant under time reversal and has no simple reality properties. however. but instead by expressing the expectation value of the Hamiltonian as l/f*Hl/f d r (H)= f 3 (8. The variational procedure is easily generalized to a particle in three dimensions and to the Schrodinger equation for a particle in a magnetic field. Eq. According to the method of Lagrangian multipliers.El/f* =0 where l/f' and l/f*' denote derivatives with respect to x.3) are equivalent.2) is more appropriate.136 Chapter 8 Variational Methods and Simple Perturbation Theory which is the formula for the expectation value of the energy. It leads.3) denotes the expectation value of the energy. E is assured to be a real number. : + Vl/f*l/f .6) f l/f*l/f d r 3 If the variation of (H) is defined by 8(H) =  f f (l/f* + 8l/f*)H( l/f + 8l/f)d3 r (8. but (8.7) (l/f* + 8l/f*)(l/f + 8l/f)d 3 r . [The procedure of varying l/f(x) and l/f*(x) independently is equivalent to expressing l/f in terms of Rel/f + i Iml/f and varying Rel/f and Iml/f independently.El/f*l/f )dx =0 (8. If l/f(x) and l/f*(x) are varied independently of each other.+ Vl/f . the constrained problem posed here is equivalent to an unconstrained variational problem for the integral 1.. d = . since (8. to a variational expression with a simple physical significance.El/f = 0 z  liZ dZl/f 2m dx liZ ~l/f* 2m dx z (8. and the energy of the ground state of the system can therefore be determined as the absolztte minimum of (H).] The simultaneous use of l/f(r) and l/f*(r) may seem like an unnecessary complication if V is real. it is sufficient to work with the onedimensional case. For the purpose of establishing the variational principle.5) + Vl/f* .Efl/f*l/f dx: 8 f F(x) dx =8 f (~~ aF dt* . For many applications (8..dx al/f' = .
In practice... differing from an eigenfunction by 5"'. the absolute minimum of (H) obtained by this method gives an upper bound for the lowest eigenvalue of H..9) H'" = E'" where E = (H) (8. and (H) is made stationary by the eigenfunction. f3.9) is the equation whose eigenfunctions and eigenvalues are "'k and E k • To summarize: If a trial function + 8"'.)2] symbolizes all terms of order higher than the first in the generally independent variations 8'" and 5"'*. This bound may even be a fair approximation to the lowest energy eigenvalue if the parametric trial function is flexible enough to . The proof of this proposition follows if we use the Hermitian property of H and choose as the variation of "': with a small real number e.J {J d 8"'*H'" d r + 3 "'*H'" d 3 5"'*'" d r J + J 3 "'*H8'" d r] "'*5'" d r] + 0[(8"'?] (8. such that J ".t. the extrema of (H) are found with the help of the equations "'k "'k iJ(H) iJa = iJ(H) iJf3 = iJ(H) = .t."'k d 3r = 1 and J". 2. then '" must be an eigenfunction of H. from the firstorder (8. E k + 8(H) is obtained. y . altogether erroneous estimates are likely to be obtained unless the trial function simulates the correct ground state wave function reasonably well. upon expanding this expression we obtain: 3 3 5(H) = J"'*'" {J . we can obtain various overestimates of the ground state energy by calculating (H) for suitably chosen trial functions. If '" = "'k is one of the normalized eigenfunctions of H.1 The Calculus of Variations in Quantum Mechanics for trial functions'" + 8'" and "'* 137 (J "'*'" 3 d r r + 8"'*. Obviously.8). . Usually.11) Evidently.8) 3 where 0[(5". = 0 iJy (8. is used to calculate the expectation value of the Hamiltonian.10) Condition (8. Two conclusions can be drawn from (8. Conversely. The variational method is frequently applied to the SchrOdinger equation by using for the calculation of (H) a trial wave function that contains one or more real variable parameters a. then.8): 1. If the expectation value (H) is a differentiable function of these parameters. the equation = 0 then gives. it is necessary to strike a compromise between the desire to improve the estimate of the ground state energy by choosing a "good" wave function and the requirement of ease of calculation.H"'k d3r = Ek then 8(H) = 0 to first order. The condition 5(H) terms in (8. and for small 5'" the change 5(H) is of order (5"'f. if all firstorder variations 5(H) are stipulated to vanish.
extrema of (H) may correspond to excited states of the system.3 by exact methods and by applying the WKB approximation. relative. The other.. This problem was solved in Section 7.2. As expected.Ixl) for Ixl:5 a. How does the result compare with the estimate from the Gaussian trial function? Exercise 8. but is remarkably close to this value.14) (8.2 a 2m + g v'27ra (8. and ljI(x) =0 for Ixl > a Why is it safest to use (8.2r/3 (8.ax 2 (8. IjIt(X) = ( .15) 7T~/3 [2~/3 + 2~13](g~2rl3 = 0. this estimate exceeds the exact ground state energy = 0.2 _ _ g_ 2m 2Y27T a3/2 = 0 or a = When this result is substituted in (H) = (~fj.813(g~2rl3 Eo is obtained.2m dx2 {fj.12) The expectation value of H becomes (H) = f +OO 00 d 1jI.138 Chapter 8 Variational Methods and Simple Perturbation Theory simulate the groundstate eigenfunction ljIo. we consider the Schrodinger equation for the potential V(x) = glxl. the variational estimate (8. Figure 8.13) From this expression we obtain the condition for a minimum: a(H) = iJa fj. Calculate the variational estimate for the ground state energy of the potential V(x) = glxl (with g > 0). By . As an example.3. Exercise 8.809(g~2rl3 .3) here? How good is the estimate? Exercise 8. using the triangular trial function ljI(x) = C(a . Invent other simple trial functions that can be used to estimate the ground state energy of the particle in the potential V(x) = g Ixl (with g > 0).13).) 2a 114 e. Here we make use of the variational method by choosing a normalized Gaussian trial function IjIlx) with adjustable width..(x) .1 shows a comparison of the exact (Airy) ground state wave function and the optimal Gaussian trial function.2 2 + glxl IjIt(x) dx } = fj.1. provided that the trial function is sufficiently adaptable to represent the desired eigenstate of H to a reasonable approximation. use an optimized exponential trial function to estimate the ground state energy.2) rather than (8. g For the Schrodinger equation with a potential V(x) = g Ix I (with > 0).
obtain numerical estimates for the first excited state and compare with the exact value of E 1 in Table 7. If they are not. The RayleighRitz Trial Function. the coefficients Ci and may be varied independently.(x). are to be determined from the variational principle. . we know from Section 4. as in Section 8. linearly independent. The RayleighRitz method will be extended to arbitrary basis functions in Section 8. The l/J(r) can be regarded as vectors in an ndimensional space that is spanned by the basis Xi (r). c. and the thin line represents the optimal normalized Gaussian variational trial function. at the origin. The coefficients c. quadraticall.2 The RayleighRitz Trial Function 139 ~ ~ 2 4 Figure 8. One of the most prominent applications of the variational principle is the RayleighRitz method. as described in Section 7. The x coordinate and the energy are made dimensionless by scaling. or. The meansquare deviation between the two functions is only f:~: Ilfio(x) lfi.1. In this section. the basis functions Xi need not be orthogonal or normalized to unity. This procedure is applicable to the discrete spectrum of an Hermitian operator and consists of using as a trial functio~ l/J a linear combination of n suitably chosen.3. Can you proceed to the second excited state and calculate a variational energy estimate for this state? 2. are referred to as a set of basis functions. lfio(x) = 1.1 integrable functions Xi: l/J(r) = .16) The functions X. The thick line is the normalized ground state energy eigenfunction.l. for the potential Vex) = lxi. that the basis functions are orthonormal. lfi.0. much as nonorthogonal coordinate systems may be employed in geometry. Their real and imaginary parts may be taken as the variational parameters.=1 2: n Ci xJr) (8.4. Generally.1 how to proceed in order to replace them by an equivalent orthonormal set of n basis functions.1.(xW dx = 1O.3 ! using odd trial functions with only one node. we assume . This cumbersome orthogonalization procedure can be avoided by keeping the formalism sufficiently general to accommodate the use of nonorthonormal basis functions.47 Ai[0/2(x . The coefficients c. are the components of the vector.8086)].
21) Exercise 8. j = 1.20) is redundant.17) J 1/1*1/1 d r 3 iJ=l 2: CjCi where the matrix elements of H are defined by (8.. the second set of linear e.24) .21) from the variational principle.140 Chapter 8 Variational Methods and Simple Perturbation Theory The expectation value (H) is expressed in terms of the Ci and c7 as J I/I*HI/I d r 2: 3 n cjvlHli)Ci n (H)= iJ=l (8. . 2.4. (ijHIi) = = J 1/17Hl/lj d r = [J (HI/I)*I/Ii d r 3 3 (J I/IjHl/li r) * = (jiHI i)* d 3 r (8.23) This system of equations for the coefficients Ci has nontrivial solutions only if the characteristic value (or eigenvalue) E is one of the n roots of the determinantal (characteristic or secular) equation: I Dn (E) == det«i/HIi) . Derive (8. the system of equations (8. If we denote the real expectation value of H corresponding to the (initially unknown) optimal trial function by (H) = E.18) Since H is a Hermitian operator. because H is Hermitian.20) produce the n linear homogeneous equations j=l 2: (iIHlj)cj = (H)Ci n (8.21) can be written as j=l 2: (iIHlj)cj = n E Ci (8. a(H) acj = 0 and a(H) aCi = 0 for all i.E5ij) = 0 I (8.. 5(H) = O. Show that.22) or j=l 2: «iIHlj)  n E5ij)cj =0 (8. n (8.19) The variational conditions for making (H) stationary.quations derived from (8.
it is easy to see that i=l j=\ 2: 2: ci*(iIHJj)cj = E 2: Ci*Ci i=\ n n and n n n k= \ j=l k=l 2: 2: cj*(jIHlk)Ck = E 2: Ck*Ck (E  from which by subtraction it follows that E) i=\ 2: Ci*Ci = n 0 Hence. Show that if the n optimal orthonormal trial functions.26) Since the system of equations (8. (8.24) have been ordered so that E.22) and use the property (8. as was shown in Chapter 4.. they may be "orthogonalized" and replaced by an equivalent set of orthogonal functions. obtained by the RayleighRitz method. Hence.2 The RayleighRitz Trial Function 141 If we take the complex conjugate ofEq. Exercise 8. let us assume that the roots of the determinantal equation (8. (8. if the optimal trial functions are not initially orthogonal. corresponding to trial functions t/J and If.22) is linear and homogeneous.27) shows that the determinantal equation takes the simple form E'tE 0 E~E D n (E) == 0 0 0 0 E~E =0 (8. It is important to note that all of these results pertain to the approximate solutions of the Schrodinger equation.. If this orthonormal set is used as a basis.25) If Eqs.22) and (8. ::5 E~l :5 E~ The corresponding n optimal trial functions may be assumed to constitute an orthonormal set. we obtain j=\ 2: c1(jIHlk) = Ec~ n n (8. :5 E~ :5 . whereas we already know them to be true for the exact solutions. are used as basis functions. the matrix elements of H are diagonal: (8. (8.28) 0 .5. E and E..27) In order to help us appreciate the significance and utility of the n characteristic values E produced by the RayleighRitz method.25) are written for two different characteristic values.19). the optimal RayleighRitz trial functions corresponding to different characteristic values are orthogonal: (8. any linear combination of two or more optimal trial functions belonging to the same characteristic value E is also an optimal trial function.
and the bounds are steadily improved by choosing ever larger values of n.142 Chapter 8 Variational Methods and Simple Perturbation Theory Suppose that an (n + I )st basis function is added. E3is above the second excited level. but the present discussion already shows how to link Schrtidinger's wave mechanics with Heisenberg's matrix mechanics. in the sehse of convergence "in the mean. In this new basis. and the two remaining characteristic values lie below and above the old spectrum. Hence. the RayleighRitz method provides a way of estimating upper bounds for the lowest n eigenvalues of the energy operator H. Hence. and one that is 2:. a characteristic value. in practice. it follows that E. it can be seen that there must also be one characteristic value that is ::::. We will return to the matrix representation in Chapter 9.E. since there are exactly n+ 1 roots. In every physical theory two trends are evident. will move down.29) (n+IIHII) (n+l!HI2) From this equation a new set of n + I roots or characteristic values can be calculated. As a new basis function is added. What can you infer about the roots of D 4 (E) = 07 Try to discern a pattern. Thus. Perturbation Theory of the Schrodinger Equation. on the other hand. each interval between E~ and E~+I must contain one new characteristic value. An application of the RayleighRitz is afforded by perturbation theory for degenerate or neardegenerate states in the next section. but generally it need not be a repeated zero of D n +I(E). The values of the determinant D n +1(E) for E = E~ and E = E~+I can be shown to have opposite signs. If E~ = E~+I' a new root of D n +I(E) coincides with these two previous characteristic values. we strive to formulate exact laws and equations that govern the phenomena. . Since the Hamiltonian is an operator with a lower bound for its expectation values.E~. lies above the ground state energy. Figure 8. The determinantal characteristic equation becomes infinitely dimensional. generally it will be slow unless the trial function can be made to resemble the actual eigenfunctions closely. E~ lies above the first excited level. If in the limit as n ~ 00 the basis set approaches a complete set of linearly independent functions for the domain of H. < E3 show that DiED and D 4 (E3) are negative and DiED is positive. and as n is increased its roots approach the eigenvalues of H from above." the optimal trial functions approach the true eigenfunctions of the Hamiltonian operator. we are confronted with the 3. in fact. If the roots of D3 (E) = 0 are distinct and ordered as Ei < E~ . all the new variational estimates E. On the one hand. between E~ and E~+I there must be at least one root.2 illustrates the situation. From the behavior ofthe determinant D n + 1 (E) for E ~ ± 00. the determinantal equation is also simple: EiE o E~E (11Hin + 1) o (2IHln+l) (n+IIHln+l)E =0 (8. which may be chosen to be orthogonal to the previous n basis functions. The equations derived in this section for the approximate determination of energy eigenvalues and the corresponding eigenfunctions are all familiar from linear algebra and can be written in matrix form. provided that E~+I E~. "* Exercise 8.6. There is no assurance that the convergence will be rapid. and so on.
.... ..'"'1"''.. E::+b appears... Given a complex physical system......2...... . .. the degeneracy Ej = E4 is removed and a new approximate eigenvalue.. In quantum mechanics... we choose... .. The complicated actual system may then often be described to good approximation in terms of the cognate system... a similar but simpler comparison system whose quantum equations can be solved exactly.._""...... . ...= E3 Ei..... ...... .. need to obtain more or less approximate solutions to the equations.. These are nevertheless useful.. ... .... ..E3 . E4 ... . + .. and we may regard the motion of an electron in a real many ..... Examples of the uses of perturbation theory are legion.. ." .. For instance.~___.. ". . ... . ..... ......" ... ... because rigorous solutions can usually be found only for oversimplified models of the physical situation. . When the (n+ l)st basis function is added.. E1 Dn(E') =0 D n+ 1(E") = 0 (ground state) Figure 8... Eli . ....E2 .... ... . ......... if possible..:§:n=================:.....:....... because they often serve as a starting point for approximate solutions to the complicated equations of the actual system. ... ~ 2 E" _ Ei. ....... the perturbation theories are examples of this approach..... ......... .... .3 Perturbation Theory of the Schrodinger Equation 143 E+ En+ 1 En n 2 E~___ E~_l :::::::::::::::::::::::::::::==:E~~~'+':l E~l ==========::::. in atomic physics the problem of the motion of an electron in a Coulomb field can be solved rigorously (Chapter 12).:::>.... .. which the approximate eigenvalues approach from above... ... Diagram showing schematically the location of the roots of the characteristic determinantal equations Dn(E') = 0 and Dn+1(E") = 0 in relation to the exact eigenvalues E r of the (rl)st excited energy level of the Hamiltonian H.... ..~< ..... .._  ========= " Ej = E4..
V2 1/1 + VII/I 2m fi2 = EI/I (8.31) is degenerate and corresponds to n linearly independent unperturbed eigenfunctions. including the calculation of higherorder approximations.144 Chapter 8 Variational Methods and Simple Perturbation Theory electron atom as approximated by this simpler motion. estimate of the energy eigenvalue of the SchrOdinger equation (8. Frequently. In a large class of practical problems in the quantum domain. will be given in Chapter 18.30). and often quite accurate. perturbation theory provides at least a first qualitative orientation. If the unperturbed eigenfunction 1/1(0) is used as a trial function and if E(O) is nondegenerate.32) I/I(O)*VI/I(O) d 3 r represents a sensible. and the influence of an anharmonic term in the potential energy of an oscillator on the energy spectrum can be assessed by perturbation theory. This section is our first introduction to perturbation theory for stationary states. and that the corresponding eigenfunctions 1/1(0) will be similar to 1/1. ••• . are merely so close to each other that the perturbation is effective in "mixing" them. even where its quantitative results may be inaccurate. The matrix elements (8. The difference potential V is called a perturbation of the unperturbed potential energy Yo. but that we know the normalized solutions of a similar SchrOdinger equation .34) . We assume that the n unperturbed degenerate or neardegenerate eigenfunctions are orthonormal.yo. rather than being all equal as in the case of degeneracy. I/I~O).33) which is one of the most useful formulas in quantum mechanics.30). A more systematic treatment. The change of the energy levels of an atom in an applied electric field can be calculated by treating the field as an added perturbation. I/I~O). If V is small (in a sense yet to be made precise). and the change in the energy eigenvalue is (8. this method is also useful if the n unperturbed energies.31) with VI differing from Vo by a small amount V = VI .. Suppose that a Schrodinger equation . the variational expression (H) = f 1/1(0)* [  = E(O) + f :~ V 2 + VI] 1/1(0) d 3 r = f I/I(O)*(Ho + V)I/I(O) d 3 r (8.30) with an unmanageable potential energy VI is given.18) of the Hamiltonian H = H o + V now take the form (8. perturbed by the interaction with the other electrons. I/I~O).31) will be reasonably close to the energy eigenvalues E of (8. we intuitively expect that the eigenvalues E(O) of (8. If E(O) in (8. the RayleighRitz method of the last section can be used to calculate a variational estimate of the perturbed energy.V2 1/1(0) 2m fi2 + VoI/I(O) = E(O)I/I(O) (8..
v::. which will serve as basis functions.37) and obtain _ E 1•2  This result shows that for the last term to be negligible and the firstorder estimate (8.. 2+4'1:(1"""'1v"'I. we suppose that two particular eigenvalues E~O) and E&O) of Ho are close to each other. E~O) = E&O) == E(o>.&0). the perturbation V must be weak and the other unperturbed energy levels must be far away on the energy scale..(2IVI2)f + 4!(IIVI2W] (8. Two states .(1IVll) 2 101 v12) 1 E&O) + (2IVI2) + E&O) + (2IVI2) .E~O) . the simultaneous linear equations (8.42) Unless (11 ViI) = (21 V12) and 01 V12) = (21 ViI) = 0.35) The n roots of this (secular) equation are approximate perturbed energy values.33) for the perturbed energy to be reliable.E = 0 (8. As an example..E (2IVll) 1 with roots E 1•2 (1IVI2) I (2IVI2) + E&O) .3 Perturbation Theory of the Schrodinger Equation 145 The determinantal equation for the estimated energy E is det[(E~O) .E)8ij + (il Vlj)] = 0 (8.22) reduce to 11 ViI) ( (2IVll) The amplitudes C1 (11 VI2») (2IVI2) (C1) Cz = ) (E _ E(0»)(C1 Cz (8. the perturbation splits the degenerate levels and the two optimal trial functions are orthogonal.2)E=~0").36) = ~ [E~O) + E&O) + (11 ViI) + (21 V12) (8. and belong to two orthonormal eigenfunctions ".41) and C2 define the optimal trial functions '" = C1 ".39) = E(O) + ~ [(llvll) + (2IVI2):!: Y(I!V!I) . or degenerate (in which case E~O) = E&O»).(1IVll) (0) (I I ) 2 (8. Equations (8.::"2""""W·] We consider two limiting cases in more detail. (b) In the opposite limit of exact degeneracy of the unperturbed levels. and (8. In this case.(:OII.".~O) + C2l/J&0) (8. (a) Here we assume that the unperturbed energies E~O) and E&O) are sufficiently separated so that (8.37) :!: Y (E&O) + (21v=1::.40) Corresponding to these energies.E&O) + (2IVI2) _ E~O) . we can expand (8.~O) and ".37) reduces to 1 1(11 V12) 1 E 1 + 1 VI .35) turnS' for this case into 01V l l) + E~O) .38) The magnitude of the "mixing" matrix element (11 V12) is small compared with the difference between the firstorder perturbed energies.ll:7"":»).
Exercise 8. In Section 8. gij  J X * d r g* iXi . show that leI I = IC21 and that the relative phase of the amplitudes CI and C2 depends on the phase of (11 VI2). en  4. if we define a second Hermitian matrix. The RayleighRitz Method with Nonorthogonal Basis Functions. .8. (8.v (v > 0). Similarly.ii 3 (8. that all diagonal matrix elements (k Ivi k) = 0 (k = I. which are connected by a nonvanishing perturbation matrix element (1\ V12).. while all removed and that the perturbed (ground) state has energy E = the others occur at E = v. Also show that the amplitudes. . we may choose basis functions Xi that are neither orthogonal nor necessarily normalized and again consider trial functions of the form l/J(r) = i=1 2: c n i Xi(r) (8. Show that the degeneracy is partially 1)v.43) The index i = I.7. . 2.. . all have equal magnitudes and phases. It is convenient to introduce a second related basis xi(r) in this ndimensional vector space such that xi(r) = i=1 2: Xi(r)gij n (8..47) . are sometimes said to interact with each other. which define the ground state trial function. A matrix for which (8. and that all offdiagonal matrix elements (kl vi e) (k 1= e) are equal to a real negative value . n on the undetermined coefficients ci has been elevated to a superscript for purposes of this section only to emphasize the geometrical significance of vector representations like (8.46) The gij can be taken to be the elements of an n X n matrix (with the first superscript labeling the rows and the second one the columns). Exercise 8.2.146 Chapter 8 Variational Methods and Simple Perturbation Theory l/J~O) and l/J~o>.44) and (8. If the perturbation affecting two degenerate unperturbed states is such that (11 ViI) = (21 V12).46) holds is said to be Hermitian. the basis for the RayleighRitz trial functions was assumed to be orthonormal.43).45) where 5{ is the Kronecker delta: equal to one if i From (8.45) it follows that =j and zero otherwise.44) and (8.. This is not necessary and not always desirable. n). Instead. 2. Assume that the unperturbed Hamiltonian has an nfold degenerate energy eigenvalue E(O) = 0.
Chapter 18.. 1 Exercise 8. provided only that these span the same space of trial functions (8. All the relations and rules we have established have the same form independent of the choice of equivalent basis functions. Equation (8. The use of the g's to raise and lower indices by summation over repeated adjacent indicesone upper and one lowerand the matching of indices on two sides of an equation are the notational means by which the theory is expressed concisely and conveniently.52) lCoxeter (1969).00 to + 00 the nonorthogonal basis functions Xk(X) = xke[X[ are introduced with k = 1.44) into (8.47) are said to be overlap integrals. For n = 3 and 4. .51) On account of the condition (8.9. On the interval x = .50) provided we identify (8.43).2.44) can now also be inverted: Xi = 2: xjgji i=1 n n (8. (8. The two Hermitian matrices are inverses of each other. For more details.45).46) and (8. we refer to the standard mathematical literature. by substitution of (8. because for i =1= j their values tell us how nonorthogonal the basis functions are. Eq. The calculation of the variational expectation value (H) requires evaluation of the matrix elements (8. the connection k=1 2: n gjkgki = 8{ (8. construct the reciprocal bases.43) can be derived: I/!(r) = 2: 2: CiXjgji = j=1 Cjxj(r) 2: i=1 j=1 n n (8. the new basis X j is called reciprocal to the basis Xi' If the basis Xi is orthonormal. Such a basis is selfreciprocal.49) and an alternative expansion for (8. then the reciprocal basis coincides with it.4 The RayleighRitz Method with Nonorthogonal Basis Functions 147 we obtain. n..45). . • The expressions (8.48) If it is assumed that the determinant of the matrix gki does not vanish.48) can be used to calculate the coefficients gjk. " The formalism developed in this section is isomorphic to the geometry of an ndimensional vector space. . and we have Xi = Xi and gu = gU = 8ij .
Hij = 2: gieHI = e=1 k=1 leHed4 2: 2: e=1 n n n Thus.57) It is a matter of taste which of these sets of homogeneous linear equations one prefers to solve. As in Section 8.10.g. the g's serve to raise and lower indices. However the matrix H i j is generally not a Hermitian matrix because Hj = f Xi*HXj d r = 3 [f i 3 xj(Hx )d r r = H/* (8.E8j]c j j=1 and 2: n =0 (8. We now implement the variational principle for the functional (H) by varying the n components Ci and ci* independently.52) and (8. Show that the overlap integrals are the matrix elements of the identity operator in the nonorthogonal basis. we obtain 2: Hjc j = j=1 n n (Hk (8.58) and this generally does not equal Hji* (except if the basis is orthogonal).54) or their equivalent partners. Derive the n simultaneous equations (8. such as: 2: HkJc j = (H) 2: g/dCi j=1 i=1 n (8.23) in Section 8.56) 2: [Hij j=1 Egij]c j =0 (8.11. Show that anyone of the matrix elements defined in (8. Exercise 8..2.2. Both (8. The system (8.57) give the same characteristic . these systems of equations can also be written [Hj . obtained by raising or lowering indices.53) can be expressed as a linear combination of any of the others.56) is similar to the simultaneous equations (8. e. but the overlap integral matrix g appears with the eigenvalue. The system (8.56) and (8. these equations all reduce to a single system.55) If we again use the notation (H) as = E. where the basis functions were orthonormal.148 Chapter 8 Variational Methods and Simple Perturbation Theory or the related matrix elements Exercise 8.12. Exercise 8. If the basis is orthonormal and thus selfreciprocal.57) contains Hermitian matrices.57) for the variational parameters ci from the condition 8(H) = O.
two masses ml and m2. As a further illustration of approximation methods. Nonorthogonal basis functions are particularly useful and popular in molecular physics and quantum chemistry. There we frequently encounter motion in the neighborhood of a stable equilibrium configuration. Show explicitly that the determinantal equations (8. executed by a fictitious particle of mass m = mlm2/(ml + m2) with a coordinate X = Xl .60)  EBj) =0 (8. Let us consider.62) . important qualitative features can be exhibited with a linear model. 5. as a model. A full discussion of the separation of the relative motion from the motion of the center of mass will be given in Section 15.60) have the same roots.X2 about a fixed center under the action of the elastic force. The potential representing the relative motion of the reduced mass m is Vex) = 1 2mul(lxl . approximated by a harmonic potential. equation: From (8. and some quantitative estimates can also be obtained.Egij) = 0 (8. constrained to move in a straight line and connected with each other by an elastic spring whose length at equilibrium is a. onedimensional models are of limited utility. and PI and P2 their momenta.4.a? (8. This brief introduction to nonorthogonal basis functions draws attention to a variational technique that has wide applicability.61) and prove that th~ optimal trial functions belonging to two different roots of the characteristic equation are orthogonal. besides vibrating along a straight line. An elementary example in the next section will serve as an illustration. pieced together from two harmonic oscillators. Even diatomic molecules rotate in space.5 The Double Oscillator 149 values and the same optimal trial functions.14. we now supplement the discussion of the simple harmonic oscillator (Chapter 5) by study of a more complicated potential.59) and (8.59) Exercise 8. it is shown in classical mechanics that the nonrelativistic twobody problem can be separated into the motion of the center of mass and an equivalent onebody motion. Nevertheless. If Xl and X2 are the coordinates of the two mass points. Only the relative motion of the reduced mass m will be considered in this section. The Double Oscillator. Exercise 8. The correspondence principle suggests that these general dynamical features of any twoparticle system. survive the transition to quantum mechanics. To be sure. The example comes from molecular physics. The characteristic values are the n roots of the determinantal equations det(H i j or det(Hij .13. which is subject only to an interaction potential between the two particles depending on their relative coordinates.55) deduce another convenient form of the variational n j=l 2: H ICj = ECk (8.
1 is to the right of particle 2 (x > 0)... The two parabolas are joined at x = 0 with the common value YeO) = Vo = mu}a 2/2.Il. t) + ..64) For Ixl » a. (8. In quantum mechanics the barrier can be penetrated.64).. the physically acceptable eigenfunctions must be required' to vanish as Ixl ~ 00. t) = .l. As shown in Figure 8.64) approaches the SchrOdinger equation for the simple harmonic oscillator. the potential changes from the limit (I) of a single harmonic oscillator well (of frequency w) to the other limit (II) of two separate oscillator wells (also of .~=. If a 0. for no penetration of the barrier is possible.3. Even if E < Yo.x a a Figure 8. but the particle can tunnel through the barrier. Before attempting to solve (8. hence. but it is not quite that. which measures the probability that particles I and 2 are found in the same place.x. we note that as the parameter a is varied from o to +00. there are also two inner turning points. it is almost the equation of a harmonic oscillator whose equilibrium position is shifted by an amount a. there are two equilibrium positions (x = ±a)._~:..3.'___1f_____L. The wave equation corresponding to the equivalent onebody problem is '* In . ) I 1 (t 2 iJt 2 2m iJx 2 (8.li iJ2r/J(X. (8.::::::.mw2( Ix I . If a = 0. and we have two parabolic potentials.62). we can assume that only one of these potential wells is present..~ iJr/J(x. the wave functions may have a finite value at x = 0..63) and the SchrOdinger equation is (8. and the other when the particles are in reverse order (x < 0)..a )2.__.'f'.62) reduces to the potential for the simple linear harmonic oscillator. The outer classical turning points are at XL and XR' If E < Yo.. one when particle. Classically" if E < Yo. For it is important to note that the absolute value of x appears in the potential energy (8. The double oscillator potential V(x) = mw2(lxla)212. because Hooke's law is assumed to hold for all values of the interparticle coordinate x.150 Chapter 8 Variational Methods and Simple Perturbation Theory v(x) _ _ _ __.
however. V(x). 1. but each is doubly degenerate. particularly in quantum field theory and manybody physics. since the system may occupy an eigenstate at either the harmonic oscillator well on . the extremum of '" between them would also have to coincide with the nodes.) (8. certain features of the eigenfunctions remain unaltered. If an eigenfunction has n nodes. It is customary to call this kind of continuous variation of an external parameter an adiabatic change of the system. is an even function and invariant under x reflection. this would imply that all higher derivatives also vanish at this point. We prove this assertion in two steps: (a) No two adjacent nodes disappear by coalescing as a changes.he left or the one on the right. Such stable features are known as adiabatic invariants. where XL and XR are the smallest and largest roots. because these changes describe the response of the system in time to changes of the external parameter performed infinitesimally slowly. If two nodes did coalesce at x = xo. of the eigenfunctions.65).65) In case II. the two degenerate ground state wave functions are concentrated in the separate wells and do not have definite parity.e. Hence. (b) No node can wander off to or in from infinity as a changes. = 2m [V(x) liZ . Case II serves to illustrate the concept of spontaneous symmetry breaking which arises in many physical systems. The presence of isolated singularities in V (as at x = 0 for the double oscillator) does not affect this conclusion.. the reflection symmetry of the double well is said to be hidden. the energy values are the same as those given by (8. From SchrOdinger's equation we infer that "''' if.2. or nodes. and in these state~ there is equal probability of finding the system in either of the two potential wells. In case I. As a is varied continuously. as the eigenfunction of potential I belonging to the eigenvalue En does. As the limit of case II (a ~ 00) is approached.64).. divided by an infinitely high and broad potential barrier. i.66) . the region between XL and XR in Figure 8.E] (8. To show this. since '" and "" must be continuous. The eigenstates have even or odd parity. without any external influences.. The superposition is generally not stationary. But a function all of whose derivatives vanish at a point can only be '" == O. If the state is to favor one side of the potential over the other. The potential energy for the double well. XR. Classically. As the potential is being distorted continuously. The probability distribution for every energy eigenstate is symmetric about the origin. By the differential equation (8. of V(x) = E. .5 The Double Oscillator 151 frequency w). or broken spontaneously by the ground state energy eigenfunctions. An example of an adiabatic invariant is provided by the number of zeros. we have nondegenerate energy eigenvalues E = liw(n + D (n = 0. we only need to prove that all nodes are confined to the region between the two extreme classical turning points of the motion.3. Thus. this number cannot change in the course of the transition to II. nor can new nodes ever be created by the inverse process. both '" and its first derivative "" would vanish at this point. energies and eigenfunctions change from case I to case II. the coordinate of a particle with energy E is restricted by XL s. we must superpose even (symmetric) and odd (antisymmetric) stationary states. X S. respectively.
:s.J211llr.L.L.L."'I a» the harmonic oscillator eigenfunctions (x ...:::::r::.1l.....__ I2iOO x w w li li V .. the linear combinations + and are approximate eigenfunctions corresponding to the wave function being concentrated at x = +a and at x = a.. (8.J'+L...L. so that "'0 "'1 "'0 "'I "'0 . if a = 0 (case I).. a=O I2iOO x _~::.67) "'n =.... obtained from (5....4 shows this for the two lowest energy eigenvalues (n = 0. Being an adiabatic invariant. When a is very large. Conversely. they become degenerate with the common energy nw/2.3)....00..._ 4 2 0 2 4 1.J..J.....L.Eo = 10.L. In case I (a = 0) we have a simple harmonic oscilator.Eo = ftw.J.a) and (x + a).L. For a ~ 00 (case II)..""'''......... and E I .L.... If a is finite but still large compared with the characteristic amplitude of the two separate oscillators..L. In case II CV2mw/ft a = 4).27) by shifting the coordinate origin....... respectively. If there were a node in the region x :$ XL' it would have to be a point of inflection for the function "'(x).4.+ 2m mw (lxl  a)2 (8. The two lowest energy eigenfunctions 1/10 and 1/11 for the double oscillator.68) lfIl lfIo ..8 6 4 2 0 2 4 8 V .I.L..152 Chapter 8 Variational Methods and Simple Perturbation Theory Then the expression on the right is positive definite in the classically inaccessible region... The two eigenfunctions and correspond to E = nwl2 and 3nw!2.J.J~L.. The number of nodes characterizes the eigenfunctions. may be used as basis functions for a variational calculation of energy eigenvalues of the Hamiltonian p2 1 2 H = "'n 2 [h V... a high barrier separates the wells..... Figure 8. the number of nodes n characterizes the eigenfunctions of the double oscillator for any finite value of a. 1).... The same reasoning holds for the region X:::: XR' It follows that outside the classical region there can be no node and no extremum (see Figure 6... no barrier.3 ftw.I. and E I .. . The existence of such a point is incompatible with the asymptotic condition: "'(x) ~ 0 as x ~ . a point of inflection would also have to be a node.JI. the reflection symmetry is broken.la=4 Ii Figure 8..
.. Since the exact eigenfunctions of H.69) into this formula. Except in the limit of case II (a '» 00).5 The Double Oscillator 153 at least for not too large values of n. a= ~Ta r.. For the normalized trial functions. they overlap. the two components in (8. (8.a)Ho/n (x + a) dx = f f o/n (x + a)Ho/n (x + a) dx o/n (x (8. Since the even operator H does not connect states of opposite parity.a)Ho/n (x . give nothing new. These assignments are reversed for odd n.71) (8.75) we have for n = 0..a) dx " and C n is the overlap integral (8.73) Also 1 N. and for large a: .69) be used.69) are not orthogonal. so that (evenIHlodd) = 0. (8.a) dx = o/n (x . the plus sign gives an even (or symmetric) function. reflection symmetry suggests that the even and odd trial functions (8. for finite a.. a la RayleighRitz. Linear combinations of these. the variational estimate of energies of the double oscillator is then If we substitute (8. and the minus sign an odd (or antisymmetric) function. have definite parity. For even n. Here N± is a normalization constant. = r.:=====  V2(1 ± Cn (8. we obtain for the realvalued oscillator eigenfunctions.70) where An = Bn = f Jo/n (x . it is optimally efficient to choose o/± as trial functions.72) + a)Ho/n (x .74) ) Using the scaled dimensionless quantity..
t) =~ [exp ( iE+t)o/+(X) + exp ( iE_t)o/(X)J = i exp [ .80) implies that at t = 0 particle 1 is definitely to the right of particle 2.X2 is the coordinate difference for two elastically bound particles. breaking the reflection symmetry. the leading terms give A o = nwl2. for the double oscillator with V o » nw. 0/+ and 0/. so that the overlap integral Co is very small.cos 2 t + i 0/+ 0Ws 0/. If x = Xl . B o.2Vo) nw (8.15.79) The physical significance of the frequency W s is best appreciated if we consider a system that is initially represented by a wave function (again assuming n = 0) o/(x. the initial condition (8.76).E = 2WJ 2Vo n nW7T exp (. 0) = 0 1 [o/+(x) + o/(x)] (8.E = 2nw. Bo = nw [  a V7i.. Co = 0. (8.y. the splitting of the two lowest energy levels.:r.80) If the two wells are widely separated (a » 1). as described at the beginning of this section. If the height of the potential barrier between the two wells.a 2 For a » and thus 1. Vo = . exp (2V ) .+ 2 IJ e a2 and Co = e.tend to cancel for x < 0 and the initial wave packet is a single peak around x= +a..77) is introduced. The even state is the ground state.mw2 a2 = . and derive the asymptotic value of (H) as given in (8. and the odd state is the first excited state. is o E_ .. sm W 2s t J(8.nwa2 1 1 2 2 (8.81) This last form shows that the system shuttles back and forth between (0/++0/)/0 and (0/+0/)/0 with the frequency w s l2.68).2n (E+ + E_)t J[0/+ 0+ 0/.e.a 2) (a1V7i) e a2 nw.76) 1 (H) = nw ( 2 =+= showing that the degenerate ground state for a ~ 00 splits symmetrically as a decreases. Bo = a V7i.E+ = !i.78) Exercise 8. and Co for the double oscillator (8. The frequency corresponding to the energy splitting (8. Work out the above expressions for A o. It takes a time T s = 7Tlw s for the system to change from its initial state (0/+ + 0/)/0 to the state .78) is Ws = !i.nw ~ (8..154 Chapter 8 Variational Methods and Simple Perturbation Theory Similarly. The time development of the system is then o/(x.
Since E was assumed to be much less than Vo. The ratio of the time T s during which the exchange takes place and the period 2'1Tlw of the harmonic oscillator is given approximately by WT s = 27T . the lower (even) energy level in each split pair of levels first decreases and eventually increases toward E = hw(2n + 1/2).. the tunneling is strongly inhibited by the presence of the exponential factor. Show explicitly that the expectation value of the parity operator for the wave packet (8. Exercise 8. which means that the particles have exchanged their places in an entirely unclassical fashion.w aYT a /ii@ versus = V mwlh for the four lowest energy eigenstates of the double oscillator.t2 through the classically inaccessible barrier region indicated in Figure 8.16. The energy E 1 2 3 f!.5 The Double Oscillator 155 ("'+ . this exchange requires that the system tunnel with frequency w. As the separation a of the two wells decreases from 00 to 0. [The parity operator changes "'(x) into "'( x).!:!. and the wave function is now peaked around x = a. For comparison with the energy levels.w (v + ~) in units of a a ..a)2/2.w 3.5). = f!. 8.82) If the barrier Vo is high compared to hw. at a ~ 00. Asymptotically."')tv'2.5.81) is zero at all times.3.5 o Figure 8. the doubly degenerate spectrum is E = hw(n + 1/2). the dashed curve shows the barrier height Vo.] The situation is similar for the higher energy levels (Fig.2w s = ! JhW7T exp (2Vo) 4 2Vo hw (8. V(x) = mw2(lx[. E A.
and the two equations become.86) The function iFi is the confluent hypergeometric (or Kummer) function. It is denoted by D.. Abramowitz and Stegun (1964).156 Chapter 8 Variational Methods and Simple Perturbation Theory The upper (odd) level increases monotonically toward E = hw(2n + 3/2).(z) and is defined as (8. Its power series expansion is iFi(a.identical. We obtain for positive x the differential equation d + 1 Z2) .85) which has the same form as (8. c. Instead of proceeding to a detailed treatment of differential equation (8. Thompson (1997).of z is called a parabolic cylinder function. z) = 1 + ~ ~ + a(a + c I! c(c + 1) t + . The boundary condition I{I ~ 0 as x ~ ± 00 implies that we must seek solutions of (8. except that now the substitution z' = ~ {T(x + a) (x < 0) must be used.I{I=O 2 2 dz 2 4 (8. 2 The particular solution of (8.84) which vanishes for very large positive values . which is generally not an integer.64) must be solved accurately.84) For negative x the same equation is obtained. we refer to the standard treatises on mathematical analysis for the solutions. The exact solution of (8. we have z' = z. For a = 0. which is the special case of the simple linear harmonic oscillator. The differential equation valid for negative x is (8.84). Morse and Feshbach (1953).84) and (8. Mathews and Walker (1964)..85) which vanish as z ~ + 00 and z' ~ 00. To obtain the energy eigenvalues for an arbitrary value of a.87) 2Magnus and Oberhettinger (1949).83) in terms of the new quantum number v.64) is facilitated by substituting for positive x the new variable. 1) 2! (8. .1{1 ( v+.84). the SchrOdinger equation (8. (x> 0) and expressing the energy as (8.
3) 4 2·4·z ] (8. The remaining task is to join these two solutions smoothly at x = 0. and (8. J (8.( z) is also a solution of the same equation. 2·.5 The Double Oscillator 157 and z» If z is large and positive (z »1 D v(z) Ivi) ••• == e z?/4 z 1'[ I  v(v .4. Show that if a = 0.89) are useful asymptotic expansions of D. Z4 Although the series'in the brackets all diverge for any finite value of z.91) for even 1/1. even or odd.86) and the eigenfunctions (5. If D.1)(v .1) 2·. It follows that a double oscillator eigenfunction must be proportional to D.91) and (8. (8. ] 2.:' + (v + 1)(v + 2)(v + 3)(v + 4) + .17. As was discussed earlier.90) V7T obtain the eigenvalues of the simple harmonic oscillator from (8.88) and if z is large and negative (z «1 D (z) v and z« 2.85) which vanishes for very large negative values of z' must be D.18. Show the connection between the parabolic cylinder functions (8.(z) is a solution of (8..89).89) y!2.(z) for positive values of x and proportional to D v ( z') for negative values of x.2/ .t Z4 I vi).. Using the identity f(l + v)f(v) = 7T sin (8.. Inspection of the asymptotic behavior shows that the particular solution of (8. These are transcendental equations for v. the point where the two parabolic potentials meet with discontinuous slope.(z)... the roots of (8. the eigenfunctions can be assumed to have definite parity.92) for odd 1/1.22/ 4 zJ 1 _ t v(v . == e.] . Exercise 8. .4 ..2)(v . [Use the properties of the gamma function embodied in the identity (8.~Ta '( ~) = ° (8. The smooth joining condition requires that at x = the even functions have zero slope and the odd functions must vanish.92) give the eigenvalues of the simple harmonic oscillator..3) _ .t + v(v . D.90). .. Exercise 8..e 'e' z fev) V7f (v + 1)(v + 2) 4 _1[ 1 + ''''.2)(v . Matching 1/1 and 1/1' at x = thus leads to ° ° D" ..84). and these two solutions are linearly independent unless v is a nonnegative integer.88) and (8.1)(v .t v + v(v ..27).( z').1) 2.
::: 0) (8. and the lower Exercise 8.4.92) are obtained by numerical computation.94) occur frequently in many branches of physics. Hindered rotations of a molecule.a)2(x + a? = mw2 8a2 (x 2 .94) For large values of the distance between the minima. Quartic doublewell potential. A few of the corresponding eigenfunctions are plotted in Figure 8.6. The unnormalized eigenfunctions are (x. = 0.95) also represent typical features of a doublewell potential.94).62) and (8.a 2)2 (8. vex) a a Figure 8. . The dynamics of the vibrations of atoms in molecules is the most prominent example. estimate the splitting of the two lowest energy levels. A convenient representation of a different anharmonic double oscillator is the potential (Figure 8. described by an angledependent potential such as V(O) = Vo(cos 20 .6): Vex) =  mw2 8a 2 (x . Figure 8. (8.158 Chapter 8 Variational Methods and Simple Perturbation Theory In general.1) (Yo > 0) (8.93) (x:5 0) where the upper sign is to be used if v is a root of D~(V2 a) sign if v is a root of DJV2 a) = O.5 shows how the lowest eigenvalues depend on the parameter a defined in (8. but the boundary conditions for the Schrodinger equation are different from those for a potential that depends on the linear coordinate x.75). the roots v of (8.19. Two (or more) potential wells separated by a potential barrier as in (8.91) and (8. defined by Eq.
In the simplest case of two positively charged atomic cores and just one shared valence electron.5 is approximately 8 X 1011 Hz. which is again in the microwave region. Since it is generally far too difficult to treat such a composite system exactly. corresponding to microwave vibrations of 0. Because of the large (three to four orders of magnitude) disparity in mass between the electrons and atoms.7. if w = 10 12 Hz. r) + Vn(R) (8.+ 2m p2 V(R.82). hence.+ 2m p2 V(r) with a prescribed external potential V(r). On the other hand. ionatom collisions. The Molecular Approximation. Conversely.6 The Molecular Approximation 159 As a numerical example of molecular vibrations. we must resort to various approximation methods. we have Vo/fiw "'" 103 • From (8. The potential V(R. and condensed matter physics are among the important applications. Transitions corresponding to such barrier tunneling oscillations are commonly observed in microwave spectroscopy. and rand p refer to the electron of mass m which moves much more swiftly. In most real physical processes two or more partial systems interact dynamically in significant ways. the value of W s obtained from Figure 8. based on simplifying features of the system and its components. Molecular structure.3 6. We' consider the Hamiltonian for a model of a molecule composed of two (or more) massive atomic cores (nuclei plus electrons in tightly bound inert inner shells) and some loosely bound atomic valence electrons that move in the common field of the cores. we think of the atomic cores as moving relatively slowly under the influence of their mutual interaction and in the average field produced by the fastmoving electrons. Although many physical systems are usefully modeled by simple oneparticle Hamiltonians of the form H =. This is indicated schematically in Figure 8. the shuttling period Ts turns out to be superastronomical (about 10400 years). r) stands for the mutual interaction 3Townes and Schawlow (1955). in addition to their own mutual potential energy.96) where Rand P are the relative canonical coordinates of the slowly moving nuclear framework with reduced mass M. In this section we consider an approximation scheme that yields useful zerothorder energy eigenfunctions for starting perturbation calculations in composite systems where the masses of the interacting subsystems are grossly different.2 cm wavelength. For w = 2 X 10 15 Hz (corresponding to an infrared vibration spectrum common in many molecules). The dynamics of interacting atoms offers a valuable example of a complex composite system whose subsystems (the valence electrons and the atomic cores) can be treated as if they were autonomous though they interact strongly. the electrons respond to the nearly static field of the cores. the Hamiltonian may be written as p2 H =2M +. consider a molecule in which the equilibrium distance between the atoms is a = I A. the exchange "hardly ever" takes place. this fiction cannot always be maintained. and the reduced mass m = 1027 kg. .
'" ~ "..> \1 : " R2 Figure 8.160 Chapter 8 Variational Methods and Simple Perturbation Theory "'rl.'" ".'" I ~R3Irl)  /".97) corresponds to the subsystem Hamiltonian for the electron: He =.100). ". and the electrons move in the static potential field V(R../ I .I IN I~ .98) In principle.. (8.R3 ).c: "".../ : I~ . Hl/J(R.. r) = 71(i)(R)</>W(r) (8. based on the complete Hamiltonian (8.99) Substitution of this expression in the Schrodinger equation for the composite system.//". r) (8. once (8. as a trial function in the variational integral (H)./ / ~ . __ .. The ten interactions between the five centers of force are indicated.:.'"  "'<1~... Two valence electrons (rl and rz) in a molecule moving in the field of three heavy atomic cores (R h R z.101) with yettobedetermined 71(i)(R). )". "". I{' "./ 'l' '" / . ". r) (8..'" .\. between the cores and the electrons..99) and to use the product l/J(R. can be reasonably well represented by single terms in the expansion (8..I l~'\.~ "..+ 2m p2 V(R.7. How does one avoid having to cope with such an unwieldy problem? The simplest idea is to assume that the eigenstates of (8.97) is solved for the complete set of quasistationary states for all R.. which is chiefly due to the Coulomb forces between the charges.. the energy eigenfunction of the complete molecular Hamiltonians (8. In the limit M ~ 00 the nuclear skeleton is frozen in its position. r)... The corecore interaction is explicitly accounted for by the potential Vn (R)."'n. " '" .96). where the external parameter R is fixed. To keep the discussion focused on the .. r) = El/J(R. at least near the ground state.100) yields a set of infinitely many coupled equations for the wave functions 71(i)(R).> ~.". The Schrodinger equation for the (quasi) stationary states of the electron. •• R3 1 I I . ..96) may be expanded as l/J(R. r) = 2: i 71(i)(R)</>W(r) (8. and for the Coulomb repulsion between the electrons.
[VRcPW(r)] d r . on which AR depends.102) the expectation values of the kinetic and potential energies of the heavy structure. (The term adiabatic refers to the semiclassical picture of the parameter R slowly evolving in time. and the trial functions 11(i)(R) are normalized to unity.101) is said to constitute an adiabatic or BornOppenheimer approximation to the exact molecular wave function. (8.~ A~ } 11(i)(R) d R 3 3 (8. the variational principle 8(H) = 0 leads to an effective Schrodinger equation for the trial function 11(i)(R): (8. The normalization ensures that the vector function A R defined as (8. for each value of R.107) .97) as in Section 8. otherwise.6 The Molecular Approximation 161 essentials. but our treatment is fully quantum mechanical. The former is (:~)' =  :~ J11(i)*(R)cPW*(r)V~[71(i)(R)cPW(r)] d R d r 3 3 which can be transformed into (:~) = J 11(i)*(R){2~ (P .ARl + :~ J [VRcPW*(r)] .102) and (8.3. we assume the solution cPW(r) of (8.105). one must work in the subspace of degenerate eigenfunctions of (8.97) to be nondegenerate.105) where the effective potential U(i)(R) is defined as (8.104) is realvalued. using matrices. The approximate eigenfunction (8.103) to the expectation value of Vn(R) to give (8.) The variational procedure requires that we add to the expectation value of the electron Hamiltonian.106) Applied to (8.103) provided that the electron eigenfunctions cPW(r).) The expectation value of H is obtained by adding (8. (We omit the label i.
show that AR is the gradient of a function and therefore can be eliminated by a gauge transformation (8.109).21.108) (corresponding to zero magnetic field in the electromagnetic analogue).6). 4 The success of any calculation in the BornOppenheimer approximation is contingent on the availability of good solutions et>W(r) for the parametric SchrOdinger equation (8. with R symbolizing a set of generalized atom core coordinates.111) as Berry's (geometric) phase. Derive (8. care must be taken to ensure that the wave function t/J(R. More generally. the condition (8. requiring the use of (8. . Exercise 8. the vector A R vanishes. as is the case for simple molecules. If the BornOppenheimer method is applied to more complex nonlinear molecules. The vector AR is known as (Berry's) connection and the integral (8.110) Exercise 8. The notation A R for the vector field defined by (8. not expected from the simple form (8. of which 11(i)(R) is but one factor. The phase integral «) = fAR' dR (8. If.162 Chapter 8 Variational Methods and Simple Perturbation Theory In solving this equation. satisfies the correct boundary (and singlevaluedness) conditions.104).110). the gauge transformation 11(i)(R) ~ exp( i f: et>W(r) ~ exp(i AR~O f: AR • • dR) 11(i)(R) (8.107) rather than (8. r) of the composite system.110). if VR X A R = 0 (8. these solutions are gen4See Shapere and Wilczek (1989).97) for the electron(s). If et>W(r) is a realvalued function for all R.20.109) AR dR )et>W(r) produces no change in the product wave function t/J(R.104) was chosen to emphasize the formal resemblance of Eq. may acquire different values for various classes of topologically distinct closed loops in the parameter space.108) may be violated for isolated singular values of R.103) and verify (8. This nonintegrability may result in the appearance of characteristic spectral features.107) to the simpler form (8. it is possible to choose the electronic eigenfunction et>W(r) realvalued for all R. In quantum chemistry.111) which is gauge invariant.107) to the SchrOdinger equation for a charged particle in the presence of an external electromagnetic field (Section 4. r) for the complete system. (8. but reduces the SchrOdinger equation (8.
with the inclusion or neglect of the usually small correction terms in (8. The Hi molecular ion is the simplest example. which constitute the molecule (linear combination of atomic orbitals. V(x.8). for the lowlying electronic states are eventually derived and. (8. indicating a . we consider the schematic model of a onedimensional homonuclear diatomic molecule. From all of this.6 The Molecular Approximation 163 erally known as molecular orbitals. E(i\R). In effect. or LCAO). Manybody methods (Chapter 22) are used to include the interaction between elections.97) is again treated as a variational problem. X) + ViX) l/!(x. the Schrodinger equation for this example may be written as fi2 [ .2M ax2 iP  2m ax 2 Ii. The potential Vn(X) represents the repulsive interaction between the two atomic cores. Assuming the molecular orbital functions <p~(r) in the expansion (8. X) is not invariant under the nuclear reflection X ~ X. Exercise 8. m is the mass of the electron. their mutual overlap integrals play an important role in the calculations.10 m being the separation of the two atomic wells.2 (Ix l _ ~) 2 (8. the potential energy surfaces U(i)(R) are computed as functions of the nuclear configuration coordinates R. we assume that the interaction V(x.99) to be orthonormal and real. We also assume that X > 0 and thus now neglect the possibility of exchange between the two identical nuclei on the grounds that the barrier tunneling period Ts tends to infinity.22. effective energy eigenvalues. but we crudely simulate it by another double oscillator potential.113) with X = 10. X) ] = El/!(x. The electroncore interaction is an electrostatic attraction. The characteristic frequency of the electron motion in the separated atom is of order w = 5 X 1015 Hz. X) (8. X) = m.110) results if certain offdiagonal matrix elements are neglected. and x is the electron coordinate relative to the center of mass.05 for this case. If the constituent atomic orbitals are nonorthogonal. derive the exact coupled integrodifferential equations for the nuclear configuration wave functions l1(i)(R) and show that Eq. X is the relative coordinate of the two nuclei. (8. In practice.112) provided that the center of mass of the entire molecule is assumed to be at rest.P + V(x. and <p~(r) is represented by trial functions that are linear combinations of energy eigenfunctions of the hypothetically separated atoms. If M is the reduced mass of the molecule.106). which can vibrate but not rotate and in which effectively only one electron is orbiting in the field of both atomic cores (Figure 8. The critical ratio Volliw = (mw/81i) X 2 is about 0. To illustrate the approximation method just described.
(For an even more primitive model.97) for the molecularorbital electron: [ :~ a~2 + m. As we saw in Section 8.) The BornOppenheimer approximation for the wave function consists of assuming that the variational trial solution has the approximate simple product form (8. is thus confirmed.(i)(X) = E".5.. An understanding of the more general covalent chemical binding between dissimilar atoms also relies on the concept of exchange. The electronic SchrOdinger equation (8. besides vibrating along the line joining the nuclei. the potential is U(i)(X) = E(i)(X) + Vn(X) + !!!.115) corresponding to a fixed value of the parameter X. allowing the vibrating atoms to be bound in the molecule. but the mechanism depends on pairs of electrons with spin.112) separates approximately into a stationary state equation (8. The expectation. Confining ourselves to the electronic ground state (i = 0).. their spectra exhibit more complex features.164 Chapter 8 Variational Methods and Simple Perturbation Theory low barrier. rather than on single electrons as considered here. ~ + 2MaX 2 (8. Equation (8. see Problems 3 and 4 in Chapter 6. The electron motion cannot be localized in just one of the two wells and is shared by both of them. J+oo lacf>~)(x)12 eb: 2M 00 ax (8..115) with a potential that is invariant under the reflection x ~ x models the force by which two similar atoms are bound together in a diatomic molecule. and a second Schrodinger equation describing the motion of the nuclei in the presence of the potential E~)(X): [ _!!!. In this simplistic model. the bond in the symmetric eigenstates has its origin in the fact that an electron can be exchanged between the atoms and is shared by them..2 (Ix l . if the system found itself in an antisymmetric state it would promptly dissociate.116). (8. we see that near the minimum EO of E~O)(X) at X = Xo """ 2Yfilmw we may write approximately where C is a number of the order of unity. It follows that the nuclei perform harmonic oscillations with a vibration frequency of the order of YmlM w.117) since the connection AR vanishes for a onedimensional system.114) as in Eq..(i)(X) (8.116) In (8. .5. of relatively slow nuclear motion.101).ir]cf>~)(X) = E~)(X)cf>~)(x) U(i)(X)]". Only in these electronic eigenstates is it possible for the attractive exchange interaction mediated by the shared electron to overcome the corecore repulsion. underlying the BornOppenheimer approximation. we see that the potential function E~i)(X) for the two atoms has a minimum corresponding to a stable equilibrium configuration if the electron is in one of the symmetric eigenstates. but the general nature of the approximations used is the same. Since actual molecules move in three dimensions and can rotate.. By inspection of the scaled Figure 8.
As eigenfunctions of H. onedimensional periodic potential composed of a succession of potential wells. we deal with slowly moving heavy atomic cores and swift valence and conduction electrons. = vex . From Figure 8. The relevant group for the dynamics of a particle in a periodic potential is the group of finite displacements or translations.The Periodic Potential 165 Exercise 8. square well. !/Jl(X) and !/Jz(x). we apply perturbation and matrix techniques to the Schrodinger equation for a particle in the presence of a . with eigenvalue E. Since the kinetic energy. deduce the equilibrium distance between the atoms and the corresponding dissociation energy. 7. two conclusions can be drawn: The translation operator D g commutes with the Hamiltonian. Estimate the vibration frequency of the molecule.) has taught us that in order to derive and understand the energy eigenvalues and eigenfunctions. such that DgV(x) The Periodic Potential. An introduction to the use of group representations in quantum mechanics will be given in Chapter 17. By the standard methods of linear algebra. although in reality the number of atoms in a crystal is large but finite. As a useful idealization of the potential to which an electron in a crystal lattice is exposed. etc. If the potential Vex) is periodic. with eigenvalue E. Our experience with the Schrodinger equation for potentials that are even functions of x (harmonic oscillator. . being a derivative operator.g. it is helpful first to consider the symmetry properties of the Hamiltonian. In this section.5. then Dg!/JE(X) is also an eigenfunction of H. In a solid. As a secondorder linear homogeneous differential equation.5. We may assume these to be orthogonal and suitably normalized. is invariant under arbitrary translations. which was introduced in Section 4. assume that the atoms have mass M = 10 u and that the electronic energy scale of the separate oscillator "atoms" is hw = 10 eV. for all x. corresponding to an energy eigenvalue E.g) = Vex) x the Hamiltonian possesses symmetry under finite translation D g by the displacement ? x . we assume that the lattice of potential wells extends indefinitely in both directions. justifying the use of the adiabatic approximation.HD g = [D g. we may construct those linear .23. the Schrodinger equation for a periodic potential generally admits two linearly independent eigenfunctions. as in a molecule. For a semiquantitative estimate of the properties of the onedimensional model of a homonuclear diatomic molecule with one shared electron. with the same eigenvalue E. the symmetry of the Schrodinger equation and the Hamiltonian is expressed by (8. but the onedimensional example of motions in a periodic potential can motivate and illustrate the group theoretic treatment.118) From this equation. the displaced functions D g!/Jl(X) and Dg!/Jz(x) are linear combinations of !/Jl(X) and !/Jz(x). Group theory is the mathematical discipline that provides the tools for a systematic approach to symmetry in quantum mechanics. H] =0 and if !/J(x) is an eigenfunction of H. DgH .
122) from the SchrOdinger equation for I/Ik(X). we have no need to refer to k as an operator. In this section. Derive (8.7. discussed in Section 4.120) is imposed on the solutions of (8. corresponding to the eigenvalues E and e ikg . J or. the periodic function.ikg .24. (8.E(X). Not surprisingly in view of the connection (8. must satisfy the differential equation 2 m h Z dZukE(x) dx' z . As k is varied continuously over the real axis. this eigenvalue was written as eik'g. except when e ikg = e. respectively. In this equation the kdependence of the eigenfunctions and eigenvalues has been made explicit. ihzk duk. + V(X)UkE(X) . which are eigenfunctions of the translation operator D g• From Section 4. the energy E ranges over continuous "allowed" bands. This property of the solutions of the SchrOdinger equation with a periodic potential is known as Floquet's theorem. (In Section 4. If the function (8. Exercise 8.121) is also an eigenfunction of both Hand Dg.5. the quantum number k (mod 27T/~) labels the irreducible representations of the onedimensional group of finite translations.5 we know that the eigenfunctions of D g are the Bloch functions ikx Il/Ik(X) = e Uk(X) I (8.119) is substituted in the onedimensional Schrodinger equation. The extension of these ideas to threedimensional periodic lattices forms the band theory of solids.119) where Uk(X) is a periodic function: (8. and we therefore omit the prime on the real number k.120) The corresponding eigenvalues have modulus one and can be expressed as e.E(X).ikg or k~ = 7Tn. Since the SchrOdinger equation is real (invariant under time reversal). are certain to be linearly independent. its complex conjugate. which physically correspond to modulated plane waves propagating in opposite directions. if I/Ik(X) is an eigenfunction. which we now more specifically designate by Uk.E(x) m dx + 2m hZ~ UkE(X) . more compactly and intelligibly. In the language of group theory. . a dispersion function E = E(k) between the energy eigenvalues E and the quantum number k is obtained. The two eigenfunctions. • = E(k)UkE(X) . When the periodicity condition (8.119) between I/Ik(X) and Uk.166 Chapter 8 Variational Methods and Simple Perturbation Theory combinations of 1/11 (x) and I/Iz(x). these two functions are related as in a simple gauge transformation.122). and k was reserved for an operator.) From the preceding discussion we conclude that all eigenfunctions of H may be assumed to be Bloch functions.
we assume them to be mutually orthogonal and normalized. we conclude that (8. In an ad hoc fashion. we approximate the degenerate unperturbed state of a single electron by assigning to every lattice point x = n~ an energy value E(O). (8. E(k) is a periodic (but also generally multivalued) function of k withperiod 27T/~.125) From Sections 8. by tunneling through the potential barrier).7 The Periodic Potential 167 Exercise 8.1)7T/~ < k :5 (2n + 1)7T/~ is known as a Brillouin zone.~O)(x) n (8.~O)(x .128) .126) and that the amplitudes Cn are to be determined from the infinite set of simultaneous linear homogeneous equations.E)c n + VCn+l =0 for all n (8. We then relax this condition by introducing a perturbative interaction V that connects the lattice sites and allows the electron to be shared among them (such as. which is expressed by saying that "E(k) has the same symmetry as the reciprocal lattice. Neglecting any overlap between these sharply localized wave functions.13.123) Thus. and show that the new equation has solutions periodic in x. we demonstrate the appearance of allowed (and forbidden) energy bands by applying perturbation theory to two extreme simplified models of an electron in a onedimensional lattice. (An example will be shown in Figure 8.25. corresponding to the old energy E(k). Also show that timereversal invarianc.25." Each interval defined by (2n . A plot of the continuous function E(k) based on these properties is known as the repeatedzone scheme. Summarizing the results of Exercise 8.(O)(x) = 2: Cn ". the energy E(k) is symmetric ("even") with respect to k = n'Tr/g.124) Thus.n~) that is narrowly concentrated at the nth lattice site. We may imagine that in this extreme ("tightbinding") approximation the electron is confined to a lattice site by an infinite potential barrier.~O)(x) = ".) Before solving the SchrOdinger equation explicitly for an example. we postulate that the matrix elements of V are to be nonzero only for the interaction between nearest neighbors: (nlIVln)=v foralln (8.2 and 8. In the first model.e implies the degeneracy E(k) = E(k) (8.122) make the replacement k~ 27T k +n ~ (changing k by a reciprocal lattice vector). In Eq.3 we know that appropriate trial eigenfunctions are of the form ". and an eigenfunction ".127) These equations are easily solved by the substitution (8. VCnl + (E(O) .
are Bloch functions. IjJ(O)(x) = 2: einki.n~) = eikxuk(X) expressible in the form (8.)IjJ~O)(x n . where 2b is the width of the square wells and 2a is the length of the segments between them.129) This simple model shows how the interaction causes the infinitely degenerate unperturbed energy levels to be spread out into a continuous band of allowed energies.130) unless k l 4. Prove that for a periodic potential V(x). Allowed bands may overlap. labeled by the quantum number k and ranging between E(O) + 2v ~ E(k) ~ E(O) . (8. The critical point here is that the periodic potential has nonvanishing matrix elements only between plane wave states for which 11k = 21Tn/ g.40). forbidden energy gaps occur between them. The solution of the SchrOdinger equa . In one dimension.119). Its period is ~ = 2a + 2b. The energy eigenfunctions.V(x) dxl (8.26. The matrix method introduced in Chapter 6 is especially well suited for treating this problem. According to formula (8. to avoid the extreme assumptions of the previous two simple models. the k vectors of the two plane wave states must differ by a reciprocal lattice vector. or nearly so. eikx .168 Chapter 8 Variational Methods and Simple Perturbation Theory leading to the condition: I E(k) = E(O) + 2v cos k~ I (8. Finally. but if the unperturbed levels are sufficiently far apart and the interactions sufficiently weak.2v. which can be ignored if we compensate for it by redefining the zero energy level. (In threedimensional terms.) Exercise 8.  k2 21Tn/~.131) The integral in (8. with period f The unperturbed energy eigenstates are represented by plane waves. This model is appropriate for sufficiently high energy levels. Generalize this result to three dimensions (see Exercise The diagonal matrix elements (k l = k2 ) of the perturbing potential are responsible for a uniform firstorder shift of all unperturbed energies. As a second illustrative modelat the opposite extreme from the tightbinding approximationwe consider a free electron that is perturbed by a weak periodic potential V(x).131) is simply the Nth Fourier component of the periodic potential. If there are other tightly bound unperturbed energy levels. we proceed to an exact calculation of the band structure for the periodic. this occurs whenever k l = k2 = N1T/~ (N: integer). battlementshaped potential of Figure 8.3 shows that significant corrections to the free particle spectrum can be expected when two or more unperturbed energy eigenstates that are connected by a reciprocal lattice vector are degenerate. at and near the point of degeneracy the energy levels split and produce a forbidden gap of magnitude I1E = 21~ f: e27riNXIi. This is known as the KronigPenney potential. and the unperturbed energy is E(O)(k) = h 2 k 2 /2m. The simple perturbation theory of Section 8.9.IjJ~O)(x) n = eikx 2: eik(xn!. there will be a separate band for each of them.14).
137) The eigenvalues of P are roots of the characteristic equation det(P .1)~ < x < n~ .133) (8. The coefficients belonging to successive values of n can be related by a matrix using the procedure and notation of Sections 6.4.132) for a + (n .a. negative. Hence. Equation (8. Periodic potential with rectangular sections (KronigPenney). positive.133) shows that the transfer matrix represents the translation operator D _g. we obtain An) ( Bn (UI ~ i{31 Z{32 i{32 ) (An+leik. as p± = e±ikg (8. tion inside the square wells. the eigenvalues p of P may be expressed.7 The Periodic Potential 169 v I I I g1 I v=o 2a'2b I I I x v=vo Figure 8. Noting that the centers of the plateaus between the potential wells have the coordinates x = n~. we must demand that as n ~ ± 00 the matrix p n should generate an acceptable finite wave function by the rule (8. may be (8. with realvalued k.i{3l)e ik 'g P= ( i{32 e .135) (8.i{31 Bn+1eik'g This may also be written as (~:::) = p(~:) where the transfer matrix P is defined by UI .Vo and lik' = Y2m(E written in the form + Vo). or zero.134) (8. Here n can be any integer.9.pi) = p2 . The period has length g = 2a + 2b. and the well depth is Vo.136). This requirement is most conveniently discussed in terms of the eigenvalue problem of matrix P.P trace P + det P = 0 .26.g) al . where V = .ik 'g subject to the condition detP=ui+f3iM=1 By iteration we have (8.136) Applying these considerations to an infinite periodic lattice.
Since Icosh 2Kal :::::: 1. the eigenvalue condition for the periodic k'z + k" z 2k'k" cos kg = cos 2k"a cos 2k'b  sin 2k"a sin 2k'b (8. Show that if E potential of Figure 8.2(Ul cos k'g + f31 sin k'g)p + 1 = 0 The roots are P± = 1 2 [trace P ± V(trace pi  4] (8. by (8. and a particular energy value is allowed only if == or.9. and we k'z k' sinh 2Ka sin 2k'b cosh 2Ka cos 2k'b + KZ  2K (8.137). (8.142) where hK = V2mE.143) are forbidden or are at edges of allowed bands. or kg "* N7T.144) is plotted as a function of k' for a particular choice of the constants of the potential. The condition Icos kgl :s.138) are unequal. 1 2Itrace pi = lUI cos k'g + /31 sm k'gl . From the continuity of all the functions involved it follows that there must generally be forbidden ranges of energy values in the neighborhood of the discrete values determined by (8. 1 (8.140) This is the desired dispersion relationship between k and E. :s. Since. Exercise 8.12).142) and (8. cos kg = /31 can be read off Eq.142) that all energy values for which 2k'b = N7T (N = integer) (8.143). If the roots (8. using (8. (6.27. 1 separates the allowed bands from the forbidden bands unambiguously in this onedimensional potential model.170 or Chapter 8 Variational Methods and Simple Perturbation Theory pZ .141) P( Ah±») = P± (Ah±») Br) Bh±) For the potential shape of Figure 8. it is readily seen from (8.144) where hk" = V2mE..10 the righthand side of Eqs. From this plot it is then possible to relate the values of the energy E(k) to the parameter k. (8.137)..138) An acceptable solution is obtained. k is defined .9 becomes > 0. Verify that the energies determined by the condition 2k"a 2k' b = N7T are forbidden. UI and obtain for E < 0 the dispersion relationship. + In Figure 8.139) cos kg = UI cos k' g + /31 sin k' g . two linearly independent solutions of the SchrOdinger equation are obtained by identifying the initial values the eigenvectors in: (~:) with (8.
7 The Periodic Potential cos~ 171 ¥f1fL+..142) and (8.+1+'tk'b '''¥''''+11ttt'''''>L''' 2 3 4 1 Figure 8. only the segments of the curve between 1 and + 1 are allowed..ylrJ+I+IJ"'f+11... Since the plotted function equals o cos kg.. Plot of the righthandside of Eqs..10...9) with the choice of 4a == 4b == ~ for the linear dimensions and Y2mb2 V == 7rli.12 for the well depth. The enlarged detail plot shows the allowed bands (heavy black segments on the abscissa) and the forbidden energy gaps between them and illustrates the transition from tight binding to the nearlyfree regime at high energies.. The high peak of the curve at k' == 0 is suppressed.144) as a function of k' for a KronigPenney potential (Figure 8.~+.. (8. .
00 to + 00 and to divide the kspace into zones of length 2'TT'/g. and Dekker (1957).'TT' < kg :5 'TT' and regard the energy E as a multivalued function of the reduced variable k.6). which is the energy at the top of the barrier. Insight can be gained from. it is possible to restrict its range so that .. 7T < kg :5 'TT'.172 Chapter 8 Variational Methods and Simple Perturbation Theory only to within integral multiples of 2'TT'/g. shown in Figure 8. .10. the successive energy bands are assigned to neighboring zones. It is also possible. in the reduced Brillouin zone scheme. to let k range from .12 illustrates the extendedzone scheme and shows that the energy spectrum consists of continuous bands separated by forbidden gaps. with Vo = rrh 2 /8mb 2 • The dashed line indicates the energy E = 1. and to let continuity and 2mb2(E+Vol ~2 12 1. S It is natural to associate the lowest energy with the first (Brillouin) zone. The allowed energies E versus k.. however. 1_ k€ '1T 0 7T Figure 8. Figure 8.1.11. In this extendedzone scheme. is called the reducedzone scheme. This representation of E versus k. 5Christman (1988). for the KronigPenney potential defined in Figures 8. . Weinreich (1965). Ashcroft and Mermin (1976).60h 2I2mb 2 of the single bound level for the isolated well of width 2b and depth Vo (see Figure 6.11. The dotdashed line at rr/4 corresponds to E = 0.9 and 8.
The extendedzone scheme for the same conditions as in Figure 8. The periodicity of the energy in the reciprocal lattice space is emphasized in the repeatedzone scheme._. ~ _ 0 w ~ ~ monotonicity._. The smooth behavior at the zone boundaries is a consequence of the symmetry relation (8.124)._.11. ~ ~ _'_ _'_ _1'_ _'_0 % Figure 8.13 for our KronigPenney model. guide the proper identification of E as a function of k.11. .. as well as physical intuition. the perturbation method applies.. The repeatedzone scheme for the same conditions as in Figure 8. which is shown in Figure 8. 2mb2(E+Vo> fi2 12 L_ L_ L_ ~ L __ __ ._. k then approaches k' and becomes the wave number of the nearlyfree particle. J ' _ _ __ _ _ _ _ _ 1 ~ 2w w 0 w 2w Figure 8.13.7 The Periodic Potential 173 12 ~.. ._ . In the limit E ~ + 00. ~ ~ ~ ': 6 4 / / ._ 2 .12.
Exercise 8. Check the prediction for the gap width in the weak perturbation limit.11.E(x) k~)eik'[x(n1I2)€l} (8. the allowed band for E < 0 degenerates into the discrete energy level Eo.174 Chapter 8 Variational Methods and Simple Perturbation Theory Exercise 8.sin k~ which can be used to construct the eigenfunctions inside the square wells: l/I(+)(x) = = = (l/I()(x)]* eink€{{32eik'[x(n1I2)€l + (al sin k' ~ . The coefficients of the plane waves for two fundamental solutions corresponding to the same energy are given by (A~+») B~+) = eink€(Ab+») Bb+) and (A~») B~) = eink€(Ab») Bb) Because of invariance under time reversal. Discuss the occurrence of allowed bands and forbidden energy gaps.. nKo k~ == cosh K~   KO Sl 'nh K~ l: K k~ == cos k"~  ~.~ sin k"~ == 2m lim(Vob) == Ii V 2mEo . } == eikxuk. emg m of the bound state in the deltafunction well (see Section 6. and cos " E > 0 . for E < 0.28. If in the periodic KronigPenney potential of Figure 8. Show that as ~ ') 00. we may assume the relation Ab) = (Bb+»)* Ab+) Bb) between the two solutions.4). We must now look briefly at the eigenfunctions of the Schrodinger equation.9 the square wells become delta functions in the limit b ') 0 and V o ') 00 such that Vob remains finite. WIthE0 b' th e b' d'mg energy . compute the width of the forbidden gaps in the perturbation approximation as a function of N. 6 Inside the wells they are of the form (8.9. and compare the numerical consequences of this relation with the exact dispersion curve.l)g < x :::::. n~ . prove in the limit of high energies that k~ = 2k"a + 2k'b. we obtain the ratio Ab+) Bb+) {32eik'€ al sin k' ~ . . show that the eigenvalue condition reduces to cos for E < 0.29..sin eikxeik(xn€){ . Using the results of Exercise 8.27 for the KronigPenney potential of Figure 8. the harmonic waves are. the coefficients 6A more detailed discussion of quantum states in simple periodic lattices is found in Liboff(1992). replaced by increasing and decreasing exponentials.{31 cos k' ~ . In this regime.{31 cos k' ~ . Figure 8. In the plateaus where V = 0. From the eigenvalue equation for the matrix P.a.132). Here lor ¥.145) for a + (n .
it has not been necessary to specify the boundary and normalization conditions for the Blochtype energy eigenfunctions "'E(X) = eikxuk.E(X) of the Schrodinger equation.30.tes. except for a phase factor.E(X) defined by Eq.146) where C is a normalization constant.3 for free particles. Exercise 8. show that in a periodic lattice a wave packet that initially has the form of an amplitudemodulated Bloch function. Bloch functions are generally not momentum eigenfunctions.122) for the Hamiltonian operator H k . (8.129). the last expression can be written as (v x) C =h f* Uk(X) ak Uk(X) dx = iJHk 1 / aH h \ iik ) (8. 0) = A(x)eikoXuko(x) with a broad bellshaped realvalued amplitude A(x).E(X) and the eigenfunctions (8. Exercise 8. as anticipated.31. Unlike perfect plane waves. The function Uk. but it is of physical interest to evaluate the expectation value of the velocity in Bloch sta.148) and without change of shape.145) has the periodicity property Uk.a] Uk(X) l ax we have <PJ = m(vJ = C f "'!(X)Px"'k(X) dx =c f U!(X{ H + ~ ~ ]uk (X) dx k (8. Following the derivation and approximation in Section 2.148) This relation shows that (vJ is the group velocity of a wave packet that is a Bloch wave modulated by a broad realvalued amplitude function. Prove the equality dE(k) dk = / aHk ) \ ak (8. periodic boundary conditions may be imposed in a domain L = N~. As for a free particle.7 The Periodic Potential 175 being determined by smoothly joining the eigenfunction at the discontinuities of the potential.14)..I ax e' a· kx Uk(X) = e''10: [ hk h + .149) which is a special case of the HellmannFeynman theorem (Exercise 18. . With the definition (8.145) are Bloch functions. Since h PX"'k(X) = . "'(x. The permissible values of k are discrete for finite N but continuous in the limit N ~ 00. propagates with constant velocity (8. which is a large multiple of the lattice spacing.. Since the idealized periodic lattice is infinite and the emphasis has been on the energy spectrum. we finally deduce the simple result: (8. (8.147) With boundary conditions that ensure that H k is Hermitian. Calculate (v x ) for the dispersion function E(k) in the tightbinding approximation.E(X + g) = Uk.
again vanishing at the endpoints and even in x.150) the corresponding eigenfunctions (8. Ce) Show that the variational procedure produces.150) for kg = N7f. C Compare the results of the different variational calculations with the exact d) ground state energy. and Laloe (1977). an optimal quartic trial function with nodes between the endpoints. it has been evident that the theory can be greatly generalized if it is expressed in terms of the mathematical apparatus of complex vector spaces and their matrix representations. 7 'Adapled from CohenTamoudji.b b::5 Ixl !S a Ixl ::5 b Try the choice b = 0 (triangular trial function) and then improve on this by optimizing the parameter b.dE(k) _ 0 dk (8. interpretable as resulting from Bragg reflection. using normalized wave functions.151) In this chapter.Ixl) t/J.32.(x) = (a 2 X 2 )(ax2 + f3) where the ratio of the adjustable parameters a and f3 is determined variationally. Problems 1. We now turn to this general formulation of quantum mechanics. in addition to the approximation to the ground state.(x) 12 dx for the various cases. as in the preceding ones.t/J. Apply the variational method to estimate the ground state energy of a particle confined in a onedimensional box for which V = 0 for a < x < a. we have formulated quantum mechanics for systems that are best visualized in some kind of (mostly onedimensional) coordinate space or the complementary momentum space. use an unnormalized trapezoidal trial function which vanishes at ::!:a and is symmetric with respect to the center of the well: _ {Ca .145) represent standing waves. evaluate the meansquare deviation fa I t/J(x) . Since it can be shown that for these special values of k. These solutions correspond to the band edges. the condition for the band edges coincides with the equation locating the small forbidden gaps. Exercise 8. (c) Use a quartic trial function of the form t/J. Show that at the band edges. (a) First. At almost every step. (Xl sin k' g . In the nearlyfree particle approximation. group velocity is Verify (8. the (Uk ) .145) are linearly independent and describe uniformly propagating waves. in which case the two solutions become identical. Chapter 11.(x) a . . (b) A more sophisticated trial function is parabolic. Interpret the corresponding stationary energy value.f3l cos k' g = ±f32 (8. unless kg = N7f (N integer). and. Diu. and IfJ(±a) = O.176 Chapter 8 Variational Methods and Simple Perturbation Theory The two solutions IfJ<+)(x) and IfJ<)(x) defined in (8.
(c) Note that the method yields answers for a discrete energy eigenstate even if A is slightly negative. calculate the change in the energy levels of a linear harmonic oscillator that is perturbed by a potential gx 4 • For small values of the coefficient. and derive the virial theorem for stationary states. represented by the Hamiltonian H = . using as a trial function the ground state wave function for the harmonic oscillator Ho(w) = 2" p~ + 2" uie 1 I where w is an adjustable variational parameter. Using a Gaussian trial function. A particle of mass m moves in a potential VCr).5.Problems 177 2. In firstorder perturbation theory. 4. Draw the potential energy curve to judge if this result makes physical sense. (b) Compute the variational estimate of the ground state energy of H for various positive values of'the strength A. By use of the variational principle. Explain. consider the anharmonic oscillator Hamiltonian. Work out an approximation to the energy splitting between the second and third excited levels of the double oscillator defined in Section 8. e. as in Section 5. 3. (a) Estimate the ground state energy by a variational calculation. The nth discrete energy eigenfunction of this system. Show that as inadequate a variational trial function as IxI) 0 a for for Ixl :5 a Ixl > a "'(x) = { yields. where A is a variational (scaling) parameter. Derive an equation that relates w and A. an upper limit to the ground state energy of the linear harmonic oscillator.1. corresponds to the energy eigenvalue En. .Juc2 . where A is a realvalued parameter. In Chapter 6 it was shown that everyonedimensional square well supports at least one bound state. assuming the distance between the wells to be very large compared with the classical amplitude of the zeropoint vibrations. for the optimum value of a. 7. 6. Using scaled variables.2m p2 v:0 e C(l _ ax 2 (Vo and a > 0) 5. make a variational estimate of the ground state energy for a particle in a Gaussian potential well. compare the result with the variational calculation in Problem 2. with an adjustable parameter. which lies within less than 10 percent of the exact value. Apply the variational principle by using as a trial function. "'n(r). prove that the same is true for anyonedimensional potential that is negative for all values of x and that behaves as V i> 0 as x i> ± 00 • 8.
Find the approximate energy change of the ground state and the splitting of the first excited energy level. As an example of Problem 2 in Chapter 7. and calculate the number of eigenstates per unit energy interval for high energies.. and derive an implicit equation for the dispersion function E(k).5). 13. discuss the solutions of the eigenvalue problem for various assumed values of the nonvanishing matrix elements of V. Using a graphic method. 11. Solve the energy eigenvalue problem for a particle that is confined in a twodimensional square box whose sides have length L and are oriented along the x. Apply the WKB approximation to a periodic potential in one dimension. 10. Adapt the HellmannFeynman theorem for the expectation value of a parameterdependent Hamiltonian (Exercise 8. . eigenstates I k) of an unperturbed Hamiltonian H o (with k = I. Assume that n unperturbed. A small perturbation V = Cxy is now introduced. For the given perturbation. construct the optimal superpositions of the unperturbed wave functions in the case of the first excited state. Estimate the width of the valence band of allowed bound energy levels. apply the WKB approximation to the double harmonic oscillator of Section 8. Solve the eigenvalue problem in the ndimensional vector space exactly and derive an implicit equation for the perturbed energies. 2. say 11). and exhibit the nature of the perturbed eigenstates.5. The energy Eo(a) of the lowest eigenstate of a double harmonic oscillator with fixed (d depends on the distance a and has a minimum at a = ao (see Figure 8. . but not necessarily degenerate. .. but not otherwise so that the perturbation matrix elements (kl V\k') =1= 0 only if either k = k' or k = 1 or k' = 1. Find the eigenvalues and eigenfunctions.30) to this problem.and ycoordinate axes with one corner at the origin. and show that if a = ao. 12. n) all interact with one of them. the expectation value of Ix I is equal to ao.178 Chapter 8 Variational Methods and Simple Perturbation Theory 9. and contrast the energy level splitting of the two lowest levels with the results obtained in Section 8.5.
and in Chapters 5 through 8 the timeindependent SchrOdinger equation was applied to specific simple dynamical systems.3) with the expansion coefficients given by (9. The principles of wave mechanics were the subject of Chapter 4. whose admissible eigenfunctions constitute complete orthonormal sets for the expansion of an arbitrary wave function (or state) rfJ(x) of the onedimensional position coordinate.4) The coefficients Cn are the probability amplitudes. o/(x) is a superposition of many eigenstates. or energy operators. in this chapter we begin to formulate the principles of quantum mechanics in terms of the mathematical structure of an abstract vector space (separable Hilbert space). These illustrations have provided us with several examples of Hamiltonians. we summarize the standard procedure: The eigenvalue equation for the Hermitian operator A is Ao/n(x) = A~ o/n (x) " (9.2) An arbitrary state o/(x) is assumed to be expressible as the series. mostly in one dimension. Following Section 4. measurements of A give the result A~ with probability len 2 • If o/(x) happens to be equal to an eigenfunction o/k(X). Here we review and assemble in one place the tools of the state vector and operator formalism of quantum mechanics. Generally. x. Probability Amplitudes and Their Composition.2.1) orthon~rmality where A~ denotes the nth eigenvalue. The eigenfunctions satisfy the condition (9. the measurement of A will yield Ak with certainty and confirm that o/(x) = o/k(X). the determination of which o/n (x) represents the system can be further narrowed down if the eigenstates 1 179 . additional labels are required to characterize the eigenfunctions.CHAPTER 9 Vector Spaces in Quantum Mechanics Generalizing the concepts of wave mechanics. In this very common case. o/(x) = 2: n Cn o/n (x) (9. 1. If the eigenvalue A~ is simple. At the end. If an eigenvalue of A is not simple but corresponds to more than one linearly independent eigenfunction. including the commonly used notational devices. and measurements of A alone do not unambiguously identify all the individual terms in the expansion. we dose the circle by showing how ordinary wave mechanics reemerges from the general theory as a special representation.
is that the operators A and B commute.180 B~. The corresponding probability amplitude (x. (9.4). the nature of the physical system determines how many variables K and its eigenvalues K i represent and what they are. y. X3 interchangeably as dictated by convenience. (x. If after the introduction of B the simultaneous measurements of A and B still do not unambiguously specify all the eigenstates. z). z and Xl> X2. we omit the prime on the discrete eigenvalues of a generic operator like K and rely on the subscript alone to identify K i as denoting a set of eigenvalues. pz 1'1') = l!J(px. or by a denumerably . As we will show in the next chapter. with simple eigenvalues K i • In order to simplify the notation. pz). we symbolize a complete set of operators by a single operator K. Py. The probability 2 of finding K i when K is measured is given by 1 i 1'1') 1 • Applied to the wave(K mechanical formalism encapsulated in Eqs. (j = 1. (We use the notations x. or simultaneously measurable. y. pz). Py. Pz' and the probability amplitude (Px. y. a structureless particle moving in three dimensions is conveniently described by the three commuting observables. For example. pz I'1') is the momentum wave function l!J(px. A very different complete set of commuting observables for the same system consists of the three linear harmonic oscillator Hamiltonians.1)(9. An alternative set of commuting observables is Px' Py.3) with eigenvalues where the quantum numbers nj are nonnegative integers. measuring the particle position. the state may be represented equivalently by an indenumerable continuum of amplitudes. z I'1') = "'(x. are (nl n2 n31 '1'). Since they share a complete set of common eigenfunctions. the physical observables that these operators represent are simultaneously assignable to the system. or to possess a common complete set of eigenstates. z. the probability amplitude Cn can be expressed as C n = (A~ I'1'). Thus.) The probability amplitudes. The fundamental assumption of quantum mechanics is as follows: The maximum information about the outcome of physical measurements on a system is contained in the probability amplitudes that belong to a complete set of observables for the system. z) of the particle. Py. the probability amplitude for finding K i in a measure ment of K is a complex number. Py. which completely specify the state '1'. y. usually denoted by (Kil'l'). With the experience of ordinary Schrodinger wave mechanics as our guide. If the state is denoted by '1'. or (Px. the necessary and sufficient mathematical condition for two observables A and B to be compatible. the procedure is continued until a complete set of compatible Hermitian operators is found. y. with eigenvalues such that eigenstates which correspond to the same eigenvalue of A are distinguished by different eigenvalues of B. we now endeavor to build up a general formalism of quantum mechanics for any dynamical system. Chapter 9 Vector Spaces in Quantum Mechanics "'n (x) of A are also eigenstates of a second Hermitian operator B. to which they correspond. z I'1') is nothing other than the coordinate state function or wave function "'(x.2. y. x. Operators A and B are said to be compatible. For the present. We stress again that the quantity K may stand for several compatible observables.
In particular. discrete and continuous. whenever the need to distinguish them from the operators calls for caution. as long as no confusion is likely to occur. In every instance. In general. these mathematical properties are satisfied by the probability amplitudes (9. Conversely. etc. It is convenient to be able to encompass both kinds of spectra. K. The Stieltjes integral is a mathematical tool and a generalization of the Riemann integral. j Delta functions may be employed as densities in integrations to represent discrete sums as integrals.5) expressing the normalization of the probability amplitudes. we will infer the mathematical properties of the amplitudes from the probability interpretation and the laws of quantum physics.7) An arbitrary state 'I' can be equivalently specified by the probability amplitudes for any complete set of observables. and so on. The. The formula connecting the amplitudes (Ki I'1') and (Lj I'1') is patterned after the expansion postulate. First. 12. IBi) the probability amplitude for finding A. but it is always easy to replace sums by integrals and Kronecker deltas by delta functions.. difference between the two usually arises from a modification of the boundary conditions or from the neglect of a small "interaction" term in the operator. since the values K i constitute all the possible results of measuring the observable K. allowing discrete'sums to be regarded as special cases of integrals. K.6). and some useful physical quantities. To keep the notation uncluttered. since the different measured values of K are supposed to be unambiguously distinguishable and mutually exclusive.6) The amplitude (KiIKi ) must have modulus unity. one can think of the continuous spectrum of an observable as the limiting case of a completely discrete spectrum of a slightly different observable.4) of onedimensional quantum mechanics (wave mechanics). upon measuring A on a system in a state for which a measurement of B yields the value Bi with certainty. by which states describing different outcomes of a measurement are superimposed. in a unified notation. L. when the spectrum is continuous (see Section 9. just like the physical quantities they represent. operators will be denoted simply as A. etc. In this chapter" most equations are written in a manner appropriate for discrete eigenvalues of the observables. and it is therefore possible to make the choice (9. It is customary to denote by (A. such as the energy of an atom. B. Bi. their measured numerical values or eigenvalues will be denoted as A. and there are several ways of doing this. have eigenvalue spectra that are part discrete and part continuous. we must require (9. In the remainder of this section. and represents the distillation of the accumulated empirical evidence underlying quantum lBalientine (1990).1 Probability Amplitudes and Their Composition 181 infinite set of discrete amplitudes like (nj nz n31 '1'). p. (but K i instead of KD. the probabilities must add up to unity: I~ I(Kil'l')I Z = 11 (9. an amplitude may depend on both discrete and continuous variables. ..
If the composition rule (9.11).182 Chapter 9 Vector Spaces in Quantum Mechanics mechanics.ILj )12 = 2: (K.11) For fixed j. therefore.7). (9.) are seen to be subject to the conditions 2: (LjIK. Show the consistency of (9. Expression (9. we get 2: (LjIK. Equation (9.)(K. Generally. the probability amplitudes (LjIK. the probability of finding Lj is (9. j (9. IL n ) are given.9) which differs from the conventional rule for calculating probabilities: 2: I(L IK.ILn) = (LjILn) = 8jn .1. these equations permit the evaluation of the amplitudes (Lj I K.10) does not represent the probability of directly finding Lj for a system that is in state 'It.)(Ln IK.11) with the normalization condition 2: I(K. If conditions (9. (Lj I'It) = 2: (L IK. I'It) .ILj)*(K. the conditional probability of finding K. for K.)1 . with certainty. The link between the probability amplitudes for two representations is established by the composition rule for probability amplitudes.I'It)12 (9.13) According to (9.13). The solution can be inferred by comparing the set of linear equations (9. in a state that is known to "have" the sharp value L j for L is equal to the probability of finding L j in a state that is known to have a sharp value K. that there is an important reciprocal relationship: (9.)* .8) Exercise 9.13) is in accord with all the experimental evidence. )(K. j 2 I(K. so that (9.12) are combined.ILj ) = i i 1 Putting n = j in (9. leaving the system in one of the states in which a remeasurement of K would yield K. we are led to conclude that (9. such an interim measurement of K alters the initial state dramatically and irreversibly. The simple relation (9.) if the amplitudes (K.8) with condition (9.8) is applied to state 1'It) = IL n ). = 8jn (9.14) .11) and (9.12) and. but it is the probability of finding Lj if an actual measurement of K is first carried out.10) by the presence of interference terms.10) is not wrong.8) embraces many of the surprising and peculiar features that set quantum mechanics apart from a classical description. If L is measured on a system in state 'It.
the slits A and B are located at positions r A and r B • At large distance from the slits the amplitudes (riA) and (riB) are to reasonable accuracy represented by (riA) oc eiklrArl and (riB) ex: eiklrrri.16) Exercise 9. Twoslit interference is the paradigm for countless (real and thought) experiments in quantum physics. (kMIA).17) Interference phenomena arise from calculating the probability I(r I '\}f) 12 with this formula. In a twoslit interference experiment with particles of definite wave number (energy) k. interpreted in terms of a statistical distribution of particles on the detecting screen.16) are valid if they are translated into the quantum mechanics of a particle with one position coordinate (wave mechanics in one dimension).5) through (9. to within a constant of proportionality. corresponding to the probability of finding the particle that is known to pass through A. Under these conditions. the two amplitudes (A I'\}f) and (B I'\}f) for particles to appear in the separate slits or channels. Show that (9. corresponds to the complementary outcome that the diffracted particle is not accompanied by a "hit" in the detector. which are determined by the experimental arrangement.1).3. (9. A and B. whose elements are defined as (9.14) expresses the orthonormality of eigenfunctions of observables as well as the closure property.1 Probability Amplitudes and Their Composition 183 expressing the unitary character of the transformation matrix S. Show that Eqs. The twoslit idterference experiment (Figure 1. We extend the discussion to include the effect of an additional monitor that receives a signal alerting it to the passage of the particle through one of the slits. Exercise 9. Show how. (kMIA). subsequently to be diffracted in direction k and to cause the detector to fire. the probability of finding the particle at position r depends on the difference of the distances from A and B to the point of observation. In a typical interference setup. we have an amplitude. the amplitude for finding the particle at r is related to the amplitudes (A I'\}f) and (B I'\}f) by the composition law: (rl'\}f) = (rIA)(AI'\}f) + (rIB)(BI'\}f) (9. and on the relative magnitude and phase of the amplitudes (A I'\}f) and (B I'\}f). The same state can be described in terms of the amplitudes for the outcome of possible measurements in the region beyond the screen with the slits. We characterize the point of observation of the particle (either at great distance r from the slits or on the focal plane of a converging lens) by the direction k of the diffracted wave (Figure 9. If the position r is chosen as the observable for this description. completely determine the state '\}f. There are analogous amplitudes (kMjB) and (kMIB) for passage of the particle through slit B. .15) The unitary property of S implies that sst = 1 (9. A second amplitude.2.1). can be used to illustrate the rules for quantum mechanical amplitudes established in this section.
exhibit interference. Twoslit interference in the presence of a monitor that has the capacity to respond to the passage of a particle diffracted in direction k through slits A or B.19) 2 Generally. the amplitude for diffraction in direction k and no response from the detector is (kMI"I') = (kMIA)(AI"I') + (kMIB)(BI"I') (9. The offaxis placement of the monitor is meant to suggest that it may be more sensitive to particles passing through one slit than the other. Given the state "1'. the probabilities I(kMI "1') 1 and I(kMl"I')IZ. Vector Spaces in Quantum Mechanics B 1'1"> A (a) B 1'1'> A (b) Figure 9. In scenario (a) the monitor registers the passage with probability amplitudes (kMIA) and (kMIB).18) Similarly. while the particle is diffracted.18) and (9. by the composition rule: (kMl "1') = (kMIA)(A I"1') + (kMI B)(B 1"1') (9. In scenario (b) the amplitudes (kMIA) and (kMIB) correspond to the monitor remaining in its latent initial state.1. respectively.184 Chapter 9 . the amplitude for finding the particle to have been diffracted in direction k and the detector to have fired is. calculated from (9. .19).
Chapter 3. however. To prepare for the vector space formulation of quantum mechanics. In general.Probability Amplitudes and Their Composition 185 If. the states that have so far been considered require infinitely many amplitudes for their definition. leads to a typical twoslit interference pattern.17). 2 In Section 9. We say that A is correlated with M. the detector monitors the passage of all particles through either of the two slits perfectly. the probability of observing the particle in the direction k is l(kMlqr>1 2 = l(kMIA>(Alqr> + (kMIB>(BlqrW (9. and indeed many applications require the use of infinitedimensional vector spaces. If. by virtue of its placement near one of the slits or by some other technique. The infinitedimensional vector spaces that are important in quantum mechanics are analogous to finitedimensional vector spaces and can be spanned by a countable basis. III. In order to keep the discussion simple. It may indeed be idle. For descriptions of many fundamental experiments relevant to understanding quantum mechanics. the probability of observing a particle in direction k is l(kMlqrW + l(kMlqr>1 2 = l(kMIA>(Alqr> + (kMIB)(Blqr>1 2 + l(kMIA>(Alqr> + (kMIB>(BlqrW (9. does not need to produce recorded signals in the detector. They are called separable Hilbert spaces. The probability of observing the particle in direction k is then (9. such as the spin of a particle.22) We see that. Two different but equivalent notations will be used in this chapter. we review the fundamentals of the algebra of vector spaces and operators in the next two sections. because the Dirac braket notation alone is not sufficiently flexible to meet all needs. which is correlated to the passage of the particle through the slits. unless we can unambiguously distinguish between the two distinct paths. we have (kMIB> = (kMIA> = O. the intensity exhibits interference effects between the amplitudes pertaining to the two slits. n. but without the capacity to distinguish between the originating slit (perhaps because it is located halfway between A and B). that in Feynman (1965). We emphasize that for interference to be eradicated. . but more general than. for particles diffracted along k. see LevyLeblond and Balibar (1990). important physical systems or degrees of freedom of systems. the monitor M. and B with M. or briefly a state in the abstract vector space. with a monitoring device M in place. vol.4 we will see how the physically motivated rules for probability amplitudes can be fitted into a general mathematical framework.21) This expression. for which a finite number of amplitudes is appropriate. We will consider the probability amplitudes to be components of a vector qr in a space whose' 'coordinate axes" are labeled by the different values K i • We will refer to qr as a state vector. the monitor responds with high efficiency only to passage of the particle through one of the slits. We will be interested only in those properties and theorems for n ~ 00 which are straightforward generalizations of the finitedimensional theory. There are. If. like (9. most of the concepts in this chapter will be developed for complex vector spaces of a finite number of dimensions.20) without an interf~rence term. In general. but in principle it must be able to distinguish path A from path B. on the other hand. say A. for these gen2This discussion is similar to.
denoted as the sum 'Ita + 'Itb . 'It2' . = Ak = O. They are called the components of the vector 'It.t'It. as if it were a number. This is not strictly correct. More detail is found in Reed and Simon (197580) and Riesz and Nagy (1955). any two of which. say 'Ita and 'Itb . 3 2. We also define the multiplication of a vector by an arbitrary complex number A.t(A'It) = (J.186 Chapter 9 Vector Spaces in Quantum Mechanics eralizations to hold. Nevertheless. the interchangeability of several such summations.24) These rules define the addition of vectors. Vectors and Inner Products. + Ak'Itk =0 exists between them. The vector space is said to be ndimensional if there exist n linearly independent vectors. but if no n + I vectors are linearly independent. By confining ourselves to the complex vector space in n dimensions. The components determine the vector completely. but should not be misleading. The k vectors 'It 1. or the legitimacy of certain limiting processes. which they are said to represent.. The components of the sum of two vectors are equal to the sum of the aj'lJ'i and 'Itb = bj'lJ'j.27) for any vector 'It. J. or to form a complete set of vectors. . with properties 'Ita + 'Itb = 'Itb + 'Ita 'Ita + ('Itb + 'Itc ) = ('Ita + 'Itb ) + 'Itc (9. or as basis vectors.26) The vector space contains the null vector. we may choose a set of n linearly independent vectors 'It lo 'It 2 . we avoid questions that concern the convergence of sums over infinitely may terms. can be combined to define a new vector. They are said to span the space. the mathematical conclusions that we draw for the infinitely dimensional space by analogy with ndimensional space can be rigorously justified. J. 'Itk are said to be linearly independent if no relation AI'It i + A2 'It2 + . since an arbitrary vector 'It can be expanded in terms of them: (9.. This is done by associating with any vector 'It a vector A'It.. subject to the rules (A + J.25) (9. The null vector 0 will sometimes be written just plainly as 0. see Jordan (1969). In an ndimensional space.t)'It = A'It + J.29) 3Por a compendium of mathematical results and caveats. . ••• . (A. Our abstract vector space is defined as a collection of vectors.•• . the vectors and operators of the space need to be subjected to certain additional restrictive conditions. we will suppose that these conditions are enforced. denoted by 'It.t: complex numbers) and I· 'It = 'It (9.tA)'It A('Ita + 'Itb ) = A'Ita + A'Itb .23) (9. See also Pano (1971). then components: If 'lJ'a = L 'Ita + 'Itb L = L i (ai + bi)'It j (9. 0. except the trivial one with Al = 12 = . such that 'It+o='It and 0·'It=0 (9. 'Itn • We refer to these vectors as the members of a basis.28) The coefficients aj are complex numbers.
Their order is important. Next we introduce an inner (scalar) product between two vectors. '1') ~ 0 (9. Two vectors.31) where the asterisk denotes complex conjugation. .38) In particular. n) (9. denoted by the symbol (qra. qrb) (9.4. qrb)' This is a complex number with the following properties: (9.2 Vectors and Inner Products 187 and similarly Aqra = 2: (Aa. are said to be orthogonal if their inner product vanishes.)qr. .37) for the components of '1'.31) and (9. and qrb is the postfactor.34) and that ('1'. The scalar product of two arbitrary vectors is expressible as (9. as in Section 8.28) the formula (9. '1') = 11'1'11 2 = 2: la. neither of which is a null vector. we assume that the basis vectors form an orthonormal set. for the square of the norm of a vector ('I'. a nonorthogonal basis is preferred. We further require that (qra' Aqrb) = A(qra. (9. It is possible to construct sets of n vectors that satisfy the orthogonality and normalization conditions (often briefly referred to as the orthonormality property) (i.. qrb) = A*(qra. Generally. By taking the inner product with one of the orthonormal basis vectors. qrb) (9. but occasionally. . In the inner product ('I'm qrb). an orthonormal set can serve as a suitable basis.1 . '1') == 11'1'11 is called the norm or the "length" of the vector '1'. 2 (9.39) .. j = 1. '1') = 1 is called a unit vector (or a normalized vector). we obtain from the expansion (9. The value \/('1'. A vector for which ('I'. qra is called the prefactor.36) Since mutually orthogonal vectors are linearly independent.30) by the above rules for addition and multiplication.33) We also postulate that (9.35) with the equality sign holding if and only if 'I' is the null vector.32) From (9.32) it follows that (Aqra.
We are now in a position to define operators in the vector space.47) (9.43) (9. These share the property (9.4.188 Chapter 9 Vector Spaces in Quantum Mechanics Exercise 9. show that fa can be represented as an inner product. Prove that any linear operator A ha~ the property AO = 0. of course. Operators.41) defines the vector 'lJra uniquely.45) thus stressing that the application of a linear operator is in many ways similar to ordinary multiplication of a vector by a number.5. (9. are equal if A'IJr = B'IJr for every 'IJr. we will assume every operator to be linear and usually omit the adjective' 'linear. But whereas with numbers ab = ba. 'IJr' is a function of 'IJr.43) with linear operators. which assures us that a vector is fully and uniquely specified by its projections on all vectors in the space.48) The last equation states that the operator AB acting on 'IJr produces the same vector that would be obtained if we first let B act on 'IJr and then A on the result of the previous operation.46) Unless it is specifically stated that a particular operator is not linear. although in particular cases two operators may. but (9. For linear operators the parenthesis in (9. 3.42) (9." Two operators. Hence. Thus.44) is most important to us.44) is replaced by A(A'IJr) = A*A'IJr (9. This is the finitedimensional analogue of Riesz's representation theorem. comgeneral AB *" . if 0 On occasion we will also deal with antilinear operators. A and B.42) can be dropped. is the null vector. Just as numbers can be added and multiplied. The special class of operators that satisfy the conditions A('IJra + 'IJrb) = A('IJra) + A('IJrb) A(A'IJr) = AA('IJr) (A: arbitrary complex number) (9. Such operators are called linear. and we may simply write 'IJr I = A 'IJr (9. it is also sensible to define sums and products of operators by the relations: (A + B)'IJr = A'IJr + B'IJr (AB)'IJr = A(B'IJr) (9. An operator A is a prescription or mapping by which every vector 'IJr in the space is associated with a vector 'IJr' in the space. If faCIJr) is a complexvalued (scalar) functional of the variable vector 'IJr with the linearity property (9.41) for every 'IJr in the space and that (9. Exercise 9.40) where A and JL are arbitrary complex numbers. and the notation 'IJr' = A('IJr) is employed. there is no need for operators to yield the same result if they are applied in the reverse order. in BA.
3 Operators 189 mute. denoted by 1. A less trivial example of a linear operator is provided by the equation (9. For instance. for the projections on the basis vectors we have p/'I' = '1'/('1'/. Since Prove that Pa is a linear operator. where 'l'a is a given unit vector. Hence.e. P~'I' = Pa(Pa'l') = Pa'l'a('I'a' '1') = 'l'a('I'a..49) as (9. Pa is termed a projection operator.51) In particular. projection operators are idempotent. this operator may be simply written as A. and we write (9.54) .49) .50) defining the linear operator Pa. '1') = Pa'l' ~= Pa Thus. the projection operators for the basis vectors have the property (9." Exercise 9. The.52) Hence. merely multiplies each vector by the constant factor A. The association is a linear one. This equation associates a vector '1" with every '1'. with the property that 'I' = 1'1' for every '1'. it follows that for any vector '1'. i. A does commute with any function f(A) if f(x) is an analytic function of the variable x. Reasonably. A trivial example of a linear operator is the identity operator. i=1 2: Pi = 1 n (9.6. where A is a number.operator AI. since all '1" are in the "direction" of 'l'a' and the "length" of '1" equals the absolute value of the component ('I'a' '1') of 'I' in that "direction. (9.53) Note also that for every 'I' i=1 2: p/'I' = 2: a/'I'i = 'I' /=1 n n Hence. p/Pj'l' Consequently. '1') = V/ai = ajP/'I'j = 0 if i 1= j (9.
A"I'j can obviously be expanded as A"I'j = i=1 L 'l'iAij n (j = 1.190 Chapter 9 Vector Spaces in Quantum Mechanics When a basis is given. owing to the linearity of A.. The inner product of two vectors can also be written in matrix notation. The column matrices (f) ~ (:) .55) is deliberately chosen to produce the simple relation (9.56) proving the contention that the effect of A on any vector is known if all Aij are known. being again a vector in the space. we have (9. The order of the indices on the righthand side of the defining equation (9. 2. . since "I'b = L bi"l'i' we have i (9.38). . Rather. n) (9.58).57) The possibility of using matrix notation here is not the result of any strange coincidence. Equation (9. an operator A can be characterized by its effect on the basis vectors. we note Hence.55) where the Aij are n2 numbers which..56). To see this explicitly.56) can be written conveniently in matrix form as (9.58) The choice of a particular basis determines the matrices in (9. According to (9. Indeed. the peculiar rule by which matrices ~re multiplied is naturally adapted to the theory of linear transformations and any calculation in which linear quantities playa fundamental role.57) and (9. completely specify the effect of A on any vector.
") A n2 Ann Exercise 9. '1'J show that FA can be represented as an inner product. A 2n A. Show that for a linear operator ('1'b' A'1'a) = 2: 2: biAijaj i j (9. we say that all these matrices constitute a representation of the abstract vector space. For this reason. '1'b) is a complexvalued (scalar) functional of the two variable vectors '1'a and '1'b with the linearity properties FA (A'1'a + /.) The matrix representing the sum of two operators is obtained by adding the matrices representing the two operators: (A + B)ij = Aij + Bij (9. the matrix elements of the projection operator Pa are (9. FA ('1'a.7.LFA ('1'a. we have (AB)ij = ('1'j. '1'b) FA ('1'a. AB'1') = ('1'j..62) defines a linear operator A uniquely..61) and write this equation in matrix form. '1'b) = A*FA ('1'a.L'1'e) = AFA ('1'a.4. As an example.62) for every '1'a and '1'b' and that (9. The matrix elements of A in a given representation with an orthonormal basis can be calculated by the formula (9. '1'b) + /. A = ~ '1'kBkj) ik 2: ('1'j. .59) which follows immediately from (9.. '1'b) + /. If A were antilinear. taken in the same order.36).55) and (9.8.60) C AnI A 21 A I2 A 22 . A'1'k)B19 = 2: A k k (9. .63) For the matrix of the product of two operators.64) B19 This result shows that the matrix of an operator product equals the product of the matrices representing the operators. if'1'a is an arbitrary unit vector with components aj.3 Operators 191 represent the vectors '1'a and '1'b' and the square matrix represents the op~rator A. A'1'b) (9. how would the corresponding equation look? Exercise 9.. '1'b) = ('1'a. A'1'b + /. (Compare with Exercise 9. If FA ('1'a.L*FA ('1'e.L'1'e.
67) for operators or the matrices representing them.69) Hermitian operators thus are generalizations of real numbers (which are identical with their complex conjugates). For a Hermitian operator.55) without assuming the basis to be orthonormal.65) we see that an operator can be moved at will from its position as multiplier of the postfactor in an inner product to a new position as multiplier of the prefactor. provided that the operator's 'adjoint is taken.66) where the symbol A is used for the transpose of matrix A.8 that corresponding to any given linear operator A we may define another linear operator. and vice versa.9. A linear operator which is identical with its adjoint is said to be Hermitian (or selfadjoint). Although there is a potential for confusion. However. It follows from Exercise 9. and Hermitian conjugate of A.64) directly from (9. denoted by At and called the (Hermitian) adjoint of A. represented by a different matrix in every representation.192 Chapter 9 Vector Spaces in Quantum Mechanics Exercise 9. tempting as it may be. that Eq.68) Show. We will return to this point when we consider the connection between different bases.. whereas the operator A is a geometric entity. If A is a linear operator. A"1"b) A* is called the = (A"1"b.65) Specializing to unit vectors "1"a = "1". "1"a)* = ("1"b' At"1"a)* = (At'l'a. because the matrix A depends on the particular choice of basis vectors. (9. . the reader should avoid a complete identification of the two concepts. however. "1"b = 'l'j' we see that Thus the matrix representing the operator At is obtained from the matrix representing the operator A by complex conjugation and transposition: At = A* (9. The proof is left to the reader. Prove the multiplication rule (9. which has the property that for any two vectors. An important theorem concerns the adjoint of the product (9. Exercise 9. we will denote the matrix representing an operator A by the same symbol A. show that it is generally not possible to define a ("transpose") linear operator At which satisfies the equation (9.10.68) defines an antilinear operator At if A is itself antilinear. Note also that ("1"a. (9. "1"b) From this and (9. We thus emphasize the parallelism between linear operators and matrices and rely on the context to establish the intended meaning. "1"a and "1"b' (9.
::: 0 (9. 'JIe)('JIb. A'JIb)* ('JIb' A'JIb) Substituting this value of A into the above inequality. 'JIe) = ('JIa('JIa. For positive definite operators we can derive a useful generalized Schwarz inequality by substituting in (9. A'JI) . 'JIb). An example of' a Hermitian operat~r is afforded by the projection operator Pa' Indeed. In particular. It follows from the Schwarz inequality that the inner product of two vectors ('JIa. the diagonal matrix elements of a Hermitian operator are real.11.70) are also called Hermitian. For a Hermitian operator (9.71) The physical interpretation makes important use of the reality of this scalar product· which is brought into correspondence with the expectation value of a physical quantity represented by the Hermitian operator A. the value of A that accomplishes this is found to be A= ('JIb' A'JIa) ('JIb' A'JIb) ('JIa. ('I'b' Pa'JIe) = = ('JIb' 'JIa('JIa. 'JIe) = (Pa'JIb. 'JIb)*('JIa.75) is derived. 'JIe> » A Hermitian operator A is said to be positive definite if for every vector 'JI. Prove that projection operators are positive definite. The Schwarz inequality may . Matrices that satisfy condition (9.3 Operators 193 If A is a Hermitian operator.72) Exercise 9. the matrix elements that are located symmetrically with respect to the main diagonal are complex conjugates of each other. 'JIa) ('JIa.74) For the choice A = 1. 'JIb) is finite if the norms of 'JIa and 'JIb are finite. we obtain the important result: (9. and A may be any complex number.73) where 'JIa and 'JIb are arbitrary vectors. We obtain The "best" inequality is obtained if A is chosen so as to minimize the lefthand side. the Schwarz inequality (9. ('JI.72) the vector (9. 'JIe = ('JIa.70) That is. By differentiation. the corresponding matrix A satisfies the condition A* = A (9.
78) If an operator A has an inverse. this implies B' . for such an operator. '1" does the same. It is therefore legitimate to speak of the inverse of A and use the notation B = A 1. then A has an inverse.>ther than the null vector) such that A'I' = 0 (9.B) = 0. as 'I' runs through the entire ndimensional vector space. and conclude that AB = 1 and BA =1 (9.81) .45) mayor may not have an inverse. if 'l'a + A'I'b = 0.79) Conversely. '1') = 0. which is defined by '1" = A'I' (9.45) in (9.194 Chapter 9 Vector Spaces in Quantum Mechanics be interpreted geometrically as stating that the cosine of the' 'angle" a between two vectors I('I'a. A linear operator whose inverse and adjoint are identical is called unitary. e ia • For a unitary operator U: *" U. Evidently.45) associates different vectors 'I'~ and '1'. that is.75) the equality sign holds if and only if ('I'. A linear operator A.77) The operator B is unique.'.I = ut or uut = utu = 1 (9.76) exists only if (9.12. An operator B which reverses the action of A. Hence. or if and only if'l'a and 'l'b are multiples of each other. i.79).76) or vice versa.77). A necessary and sufficient condition for the existence of the inverse matrix is that det A O. if there is no nontrivial vector which satisfies (9. We note that in (9. which may be interpreted to imply that they are "parallel" vectors.e. such that 'I' = B'I" (9. Such operators are generalizations of complex numbers of absolute value 1. The matrix that represents A 1 is the inverse of the matrix A. according to the second of the equations (9. (AIAz)I = A2" IA 1 I (9. for if there were another operator B' with the property AB' = 1. Show that a projection operator has no inverse (if the vector space has more than one dimension).B) = 0. 'l'b) I I 'l'a I I 'l'b I has modulus less than or equal to unity. there can be no vector 'I' (<.80) Evidently. Exercise 9. we would have A(B' . with any two different vectors 'l'a and 'l'b or in other words if the operator A preserves linear independence. We may substitute (9. (9.B = O. or BA(B' .
be said to have the sharp value K. condition (9. which we shall also denote by qr.86) from which we see that. emphasizing the formal analogy between unitary operators in the complex vector space and rotations in ordinary space. U(. qr) relative to a complete set of observables. From now on the three expressions (9. An equivalent notation is (K.* = 1 (9. Exercise 9. The operator A is then seen to be normal if and only if (9.85) In dealing with expressions containing operators and their commutators. it is helpful to have recourse to some algebraic rules and identities. the matrix representing U satisfies the condition of unitarity.82) becomes identical with the orthogonality relation in Euclidean space. Prove that products of unitary operators are also unitary.6) as A = A + At A . I qr).2 and 9.4. Since these were already developed and compiled in Section 3.87) demonstrating the identification of the probability amplitude (K. They are as valid in abstract linear vector space as in the coordinate or momentum representation.82) If U is a real matrix.. By taking the inner product of (9.3.(K. there is no need to rederive them here. 1 or (qr. In this sense U can be regarded as defining a "rotation" in the abstract vector space. are the states in which the observable K may.At 2 + i 2i = HI + iHz (9.1 in terms of the mathematics of vector spaces summarized in Sections 9. Iqr) i (9.84) where both HI and Hz are Hermitian operators.86) . The basis vectors qr.. 1qr) = i L: qr. (9. 1 with qr) qr) the inner product (qr.13. The physical state qr of a system is fully characterized by the probability amplitudes (K.. These probability amplitudes may now be taken to be the components of a state vector.83) An arbitrary operator A can be written according to (4. A linear operator A is said to be normal if it commutes with its adjoint: (9. We can now proceed to express the physical principles of Section 9. is certain to be found if K is measured in state qr. 4. The Vector Space of Quantum Mechanics and the BraKet Notation. in the interest of brevity.36). a unitary operator preserves the norms or "lengths" of vectors and the" angles" between any two of them. The state qr is thus expressed as qr) 1 qr = L: qr. qr).(qr. In fact.87) will therefore be used interchangeably with no misgivings. By this we mean only that the particular value K.4 The Vector Space of Quantum Mechanics and the BraKet Notation 195 Hence. 1 == (qr. on account of the orthonormality relation (9. Hermitian and unitary operators are special cases of the class of normal operators.
}S:ets are analogues of what in differential geometry are called vectors (or contravariant vectors).27).93) I'IJf) is normalized to unity if ('IJf I'IJf) = 2: 2: (K.94) i j i consistent with the normalization condition (9.)(K.89) where the arrows indicate the correspondence between the two spaces. Eq.91) ('IJf I = 2: (K. properly ordered.88) This notation has led to the colorful designation of the (a I vector as a bra and the la) vector as a keto To every ket la) there corresponds a bra (ai. The distinction between the two vector spaces arises because the inner product is not commutative. The prefactor 'IJf.92) The components of a bra are complex conjugates of the components of the corresponding ket.8) of probability amplitudes. since (bla) = (alb)* The state (9. testifies to the appropriateness of using vector spaces as the formal structure in which to conduct quantum mechanics.196 Chapter 9 Vector Spaces in Quantum Mechanics with the state in which the observable L has the sharp value Lj .. and vice versa. above all. I i i (9. since generally ('lJfal'lJfb) = ('lJfbl'lJfa)* "* ('lJfbl'lJfa) The same state 'lJfa can be expressed by either a ket vector I'lJfa) or a bra vector ('lJfal. The two spaces are said to be dual to each other. We can also think of the two factors as vectors in two different spaces. (9.)(K. IKj)(K. subject to the conditions la) + Ib) ~ (al and Ala) ~ A*(al (9. I'IJf)*(K. and we use the notation I'IJf) or 'IJf for them interchangeably. each one of the two spaces is a vector space satisfying postulates (9.5). I'IJf)*(K 1IJf) = 2: I(K. Dirac has stylized this notation into la) and (al for the two kinds of vectors. I = 2: ('IJf IK. Equation (9. Instead of thinking of the inner product (9.90) + (bl (9. is a vector in braspace and is denoted by ('lJfJ Their product ('IJf. This circumstance. Taken by itself.86) is seen to be consistent with the composition rule (9.23)(9. II 'IJf) is defined to be the inner product ('IJf. 1IJf) 1 = ' ' 2 j 1 (9. The inner product is written as (9.I'IJf) .86) can be written in ket form as 1'IJf) The corresponding bra equation is = 2: IK. The postfactor 'IJf is said to be a vector in ketspace and is denoted by I'IJf). Bras ('IJf I in this context are oneforms (or covariant vectors). .87) as a product of two vectors 'IJf and 'IJf. (9. and inner products are then also referred to as contractions. I'IJf).
50).95) define a bra (c lA.92). (9. can be expressed in terms of projection operators as 1'1') = L IKi)(Ki I'1') = L Pi I'1') i i (9.97) is established. A Hermitian operator A is characterized in this notation by (aIAlb) = (bIAla)* (9.91) and (9. A linear operator A associates with every ket la) another ket.98) An example of a Hermitian operator is the projection operator P a . We can consider the operator as acting either to the right or to the left.99) Using the projection operators for the basis vectors (9.65) of the Hermitian adjoint operator becomes" in braket ' notation. whichever is convenient. we may allow a linear operator also to act from right to left.95) unambiguously as (cIAla).4 The Vector Space of Quantum Mechanics and the BraKet Notation 197 The braket notation also extends to operators. By letting the equation (cl {Ala)} = {(ciA} la) = (cIAla) (9.96) Hence. or "backwards. the general correspondence Ib) = Ala) ~ (bl = (alAt (9. In braket notation. defined by (9. emphasizing the symmetry of the dual spaces. Ib) = Ala) such that and AAla) = A(Ala») Some unnecessary vertical bars on the left of certain kets have been omitted fo~' reasons of economy in notation." The rules for multiplication to the left are «cil + (czl)A = (ciIA + (czlA «cl A)A and = «cIA)A We thus see that A is a linear operator in braspace as well as in ketspace. and we are justified in writing the expression (9.49) and (9. The definition (9.101) . as given by (9.100) a general state. it is written as P a = la)(al (9.
I =L i ('l'\p.)(K.105) is a Hermitian operator if all numbers K. From this expression we infer that the observable K is represented by the Hermitian operator (9. As in (9.103).55).I'l') i 1 2 = L ('l'IK.109) The matrix representing the special Hermitian operator K.108) which follows from the identity (9.)Ki(Kil'l') i (9.jAIK.107) and the sharp value Kj is an eigenvalue of K.)Kj(K.102) Since I'l') is arbitrary.) = L i IKj)Aij (9. (9. Thus. for an operator A these are defined as the expansion coefficients in the equation: AIK. By comparison with (9.106) Exercise 9.107) make up the chosen basis.\'l') 1 for i j a:e zero. With (9. Just as a state is specified by its components in a given basis. all probabilities /(K. satisfy the closure relation (9. (9.105).)(K.104) of K becomes I(K) = ('l' IKI 'l') I which is one of the fundamental formulas of quantum mechanics. whose eigenvectors according to (9.198 and Chapter 9 Vector Spaces in Quantum Mechanics ('l'1 =L i ('l'1 K. The eigenvalue problem of Hermitian operators in quantum mechanics will receive detailed consideration in Chapter 10.104) where K.105) No confusion is likely to arise from the use of the same symbol K for both the observable and the operator that represents it. has a particularly simple structure. as * KIKj ) =L i IK.oij = KjIK.I(K.) (9.59). and I (Kj I'l') 1 = 1.14. in the state IK) the observable K assumes the sharp and definite value Kj • The state IK) is an eigenvector of the operator K. Prove that K defined by (9.IKj) =L i /K. are the possible results of measuring K. the expectation value (9. it follows that the projection operators P.)K. If I'l') is normalized to unity: (K) = L K.) = L i IK. are real.103) The probability doctrine of quantum mechanics leads to the definition of the expectation value (K) of an observable K in the state I'l'). we deduce that (9. 2 2 If I'l') = IKj ). since (Qllm . an operator is specified by its matrix elements.
Sand R.112).64).. all offdiagonal elements (i j) are zero. For an arbitrary normalized state la) and an operator A. calculate the sum l(aiAIKiW over the entire basis IKi ). performed in this order.15. we must now consider the transformation from one representation to another in the general vector space." Le. A is unitary? 5. Just as we study rotations of coordinate systems in analytic geometry. Change of Basis. The deceptive ease with which it is possible to derive useful relations by employing the identity (9.113) or .103) is exemplified by a calculation of the matrix elements of the product of two operators A and B: * (KiIABI~) = (KiIAIBIKj ) = = (KiIA 2: IKe)(KeIBIK e j) 2: (KiIAIKe)(KeIBI~) e (9. A representation in our vector space corresponds to the introduction of a coordinate system in Euclidean space. is equivalent to a single one whose matrix is simply the product matrix SR. To obtain the new components ak of an arbitrary vector. A succession of two such basis changes.28). Along with an old basis '1'.5 Change of Basis 199 This matrix is "diagonal. Exercise 9. 'l'j. we write 'l'a = 2: ai'l'i = 2: ak'l'k i k ai = Substituting 'l'k from (9. (9. we are guided by the similarity between the geometry of this abstract vector space and geometry in ordinary Euclidean space. The new basis vectors may be expressed in terms of the old ones: (9. and the eigenvalues of K are the diagonal matrix elements.111) which is the matrix element of the product of the two matrices representing A and B. as in Eq. we get 2: Sikak k (9.112) The nonsingular matrix of the transformation coefficients S= (~:: ~:: S~ S~2 1 defines the change of basis. In introducing the notion of the complex vector space for quantum mechanical states. used in (9. we consider a new basis. What value is obtained if 2: i .
. We identify Wi with the unit ket IKi ) and 'l'k with ILk)' The orthonormality of both bases is expressed as (9. . the new matrix elements are defined by A'I'j = But on the other hand 2: 'I'..118) and their closure as 2: IK. both the old and the new basis are orthonormal. matrices).) = Slf(A.117) but this relation does hold if the old basis is orthonormal. we obtain in matrix notation SA or = AS (9. (9. constant. B. we get ILk) = 2: IKi)(K...)S Give examples. by algebraic processes involving numbers (but no other.120) . C.200 The inverse of this relation is Chapter 9 Vector Spaces in Quantum Mechanics (9.)(Kil i = 2: ILj)(Ljl j = 1 (9. .115) We say that A is obtained from A by a similarity transformation. C. (9. the transformation coefficients can generally not be calculated from the formula (9. C.Aij . . . . If.16.119) from the right by ILk). show that f(SlAS. SlBS..119) Multiplying (9. SlCS. i k Comparing the righthand sides of these equations. A.114) We must also determine the connection between the matrices A and A representing the operator A in the old and new representations. . B. Evidently. Exercise 9.Aij = 2: 2: WkSk. as is usually the case.. ILk) .4. the braket notation is convenient. B.) is any function that is obtained from the matrices. If nonorthogonal basis vectors are used. If f(A.116) So far in this section it has not yet been assumed that either the old or the new basis is orthonormal. Nonorthogonal basis vectors were allowed in the RayleighRitz method of variationperturbation theory in Section 8.
.115) for a matrix representing the operator A in the form A = st AS or more explicitly. Using the unitarity condition. for orthonormal bases.112) does. (9. we see that if the matrix S1 = st can be considered as connecting the components of two vectors in the same basis.114) not as a relation between different representations of the components of the same vector ("passive" transformation). This last equation is nothing more than the composition rule (9. the transformation coefficients must satisfy the conditions (9. we may rewrite the similarity transformation equation (9. Hence.121) From (9. Comparing (9.I'I') . The two "rotations.126).122) expressing the new components of the state 'I' in terms of the old components. The operator A which maps the unitary transform U'I' of 'I' into the unitary transform U A'I' of A'I' is defined by the equation A(lJ'Ir) = U(A'I') Hence. (9. (9. For two orthonormal bases.128) which agrees with the matrix equation (9.)(K.117) holds and can be written in the form (9.IA IKe)(KeILk) if (9.5 Change of Basis 201 linking the two representations just as Eq. it represents a unitary operator U that changes every vector 'I' into a vector U'I'.125) where S denotes the transformation matrix with rows labeled by the eigenvalues of L and columns labeled by the eigenvalues of K.124) Hence.114) with (9. that is. (LjIA\Lk) = (9.114). are equivalent but they are performed in opposite "direc!ions". and the other (active) keeping the basis fixed while rotating all vectors and operators.8) " for probability amplitudes.)(K. one is the inverse of the other.127) An alternative interpretation of a unitary transformation consists of regarding (9. but as defining the components of a new vector in a fixed representation ("active" transformation). we have the operator equation A= U Aut (9.126) 2: (Ljl K. we obtain (Ljl'l') = 2: (LjIK.57). similar to (9. by multiplication by ('I'I from the left and subsequent complex conjugation.120).123) and (9." one (passive) affecting only the basis. S must be a unitary matrix (9.
and unitary operators are not necessarily represented by unitary matrices.and infinitedimensional vector spaces in compact form. are indenumerable and form a continuum. we will recover the fundamentals of wave mechanics in the framework of a vector space that supports operators with eigenvalue spectra consisting of continuous portions in addition to discrete points.129) is independent of the choice of the basis IKi ) and that trace AB = trace BA trace AAt (9. i Show that LL j I(Ki IAILj )i2 = = trace AtA (9. because such bases are not selfreciprocal. Yet. but separable. and a normal operator remains normal. Show by the use of the braket notation that trace A = L i (KiIAIKi ) (9. and thus any operator to have at most n distinct eigenvalues.and infinitedimensional.1. Dirac introduced the braket notation in order to make the formal expressions of quantum mechanics more transparent and easier to manipulate. Also show that a symmetric matrix does not in general remain symmetric under such a transformation.3 and 4.4 for discussing the continuous spectrum of such commonplace observables as the position or momentum of a particle. Generally.4. we must work only with a mathematical structure that allows us to expand state vectors in terms of complete sets of orthogonal vectors. demand that if the eigenvalue equation for the observable K. components of vectors are then not simple inner products. a number of important and difficult mathematical questions arise.132) . a unitary operator remains unitary.17. most observables have infinitely many possible values (their eigenvalues). nonorthogonal bases are employed. as in Section 8. the matrix representing a Hermitian operator may not be Hermitian. Show that under an active unitary transformation a Hermitian operator remains Hermitian. Exercise 9.19.130) Exercise 9. 6.4. Hilbert Space and the Coordinate Representation. which motivated our interest 'in the mathematics of linear vector spaces in the first place. n. If the limit n ~ 00 is to have meaning for quantum mechanics with its probability amplitudes and expectation values. The braket notation is not particularly convenient when.131) and that this expression is independent of the bases IKi ) and /Lj ).202 Chapter 9 Vector Spaces in Quantum Mechanics Exercise 9. to which we have alluded on several occasions. KIK') = K'IK') (9.? chanics works.1 and 9. we have so far assumed the vector space to be of a finite number of dimensions. If n is allowed to grow to infinity. As explained at the end of Section 9.6 that it also makes it easy to unify the formalism for finite.18. In this section. This strategy was already employed in Sections 4. in ordinary wave mechanics. The braket notation helps to express the unified theory of finite. vector spaces (n ~ 00) with which quantum me. instead of being discrete. and many of those. The principles spelled out in Sections 9. We will see in Section 9.
4. For an example from atomic physics.137) 1"1') = J !x')dx'(x' 1"1') (x' I"1') I (9. The threedimensional theory is worked out in Section 15. The expansion (9. The corresponding eigenkets must be orthogonal. The coordinate x' serves as the continuously variable label of the component . or a probability density. I"'(x') = (9. denoted here by x'. The results of a measurement can be any real number between 00 and +00. It is even possible for a particular eigenvalue to belong to both the discrete and continuous spectrum. Thus. I(K' I"1')12 is the probability per unit K' interval. The corresponding eigenvectors are written as Ix') xix') = x'lx') (9. and this will be done in Section 10. The application of the formalism to wave mechanics for a point particle in one dimension is straightforward. we assume that there are no repeated eigenvalues of K. and the eigenvalues of x. Thus.138) The components (x' 1"1') constitute a complexvalued function of the real variable x'. an arbitrary state vector can be expanded as (9. form a continuum.134) the sum being over the discrete and the integral over the continuous eigenvalues of the observable K.6 Hilbert Space and the Coordinate Representation 203 has a continuous spectrum. By identifying it with the wave function. all the formulas for the discrete and continuous cases are similar. the prime notation to distinguish eigenvalues (K') from operators (K) has been reinstated.133) With this normalization. except that integrals in the continuous case replace sums in the discrete case.135) showing that 1 1 1 dK' is the probability of finding for the observable K a value (K' "1') 2 between K' and K' + dK' when K' lies in the continuous part of the spectrum.7. see Section 19. only those eigenvectors are admissible for which the eigenvalue K' is real and which can be normalized in terms of delta functions rather than Kronecker deltas: (K' IK") = 8(K' .1.K") (9.136) with the assumed normalization. For continuously variable eigenvalues. For simplicity. The closure relation is now JIx')dx' (x' and the state is expanded as 1 =1 (9. It is easy to generalize the formalism to accommodate repeated eigenvalues.134) gives the expectation value of K as (9.139) we establish the connection between the state vector I"1') and the wave function "'(x'). Since we can measure the particle's position along the x axis. there is a Hermitian operator x for this observable.
the matrix that represents an operator A is now a matrix whose indices. instead of being denumerable. take on a continuum of values.x') ax' = iIi8(x" .x') I ax" (9. Linear momentum in the x direction is another important operator for the system. Exercise 9. The representation used here is spanned by the eigenvectors of the position operator x and is called the coordinate representation.142).130).x')(x"lplx') a 8(x" . The inner product of two states is represented by (9.x") (9. because we are dealing with a onedimensional system. since it vanishes for x" 1= x'.141)) This is said to be diagonal.142) and deduce its consequences. . in agreement with the earlier conventions (Section 4.pxlx') from which we infer that = (x" . Taking the trace on the two sides of the commutation relation (9.143) which.px = iii 1 (9.142) For brevity we omit the subscript x on the momentum p.140) The inner product of two orthogonal states is zero. p.142). is (x"lxp . though highly singular. Although such a matrix can no longer be written down as a square array of numbers. A function f(x) is represented by the matrix element (x"lf(x)lx') = f(x')8(x' . is a welldefined quantity. Here we merely assume the commutation relation (9.142). it is nevertheless appropriate to use the term matrix element for the quantity (x"IAlx'). We know that in the coordinate representation it acts on wave functions as (Ii/i) a/ax'.204 Chapter 9 Vector Spaces in Quantum Mechanics "'(x') of the state vector I'\{f) in an infinitedimensional abstract vector space.1. "'(x') is merely one of many possible ways of representing the state vector. and we can use this information to conjecture that in the abstract vector space it is a Hermitian operator.'  Ii a 8(x" .1).20.5 and will again be taken up in Section 15. The matrix element of the operator equation (9.x') x" .x') (x"lplx') = iii 8(x" . show that a contradiction arises from the application of (9. Resolve the paradox. which satisfies the commutation relation xp . taken between the bra (x"l and the ket Ix'). was already discussed in Section 4. From this point of view. In the coordinate representation. which is at the root of (9.x' = iii  = . The fundamental relationship between linear momentum and linear displacements.
x")t/!(x")dx" = J!x')dx' f(~ ~)t/!(X')dx' ax' I (9. (i fI..)t/!E(X')dx' + mZal JIx')X. is correct. HIE) = EIE) (9. Derive the equations connecting the expressions for components of states and matrix elements of operators in the coordinate and momentum representations. showing that the transformation coefficients are (x'ip') = _ l _ exp Y27rfl.6 Hilbert Space and the Coordinate Representation 205 Exercise 9.Z . we reconsider the linear harmonic oscillator with its Hamiltonian or energy operator. they reassure us that the abstract vector space (braket) formulation of quantum mechanics contains wave mechanics as a representation.!.)l/Jl(X')dx' (9. and that ('I'zlf(p)I'I'I) = J t/!!(x') f(7 a:. As an illustration. Change from the coordinate basis to the momentum basis. pZ mal XZ H=+(9. = JJ Ix')dx' f(7 a:' )S(X' . Exercise 9. p 'x') Represent states and operators in the momentum basis.144) It follows that the action of an operator f(p) on a state is represented as f(p)I'I') = JJ \x')dx'(x'lf(p)!x")dx"(x"I'I') . More generally.Zl/JE(X')dx' ax'z = E J Ix')dx't/!E(X') . Show that for any function f(p) that can be expressed as a power series of its argument.148) appears in the coordinate representation as J.. we see that the eigenvalue equation for the energy operator.(pZ + mZwzxz)IE) 2m = _1 2m J\x')(fI.147) 2m 2 Denoting the eigenvalues of H by E as is customary (instead of H'). (x"lf(p)lx') = f(~ ~)S(X" ax" I x') (9.142) as the momentum.22.146) All of these results confirm that the identification of the operator p which satisfies the commutation relation (9.145) '.21.
) exist for this SchrOdinger equation and that..39): (x' lEn) = o/En(X') = 2~2 n!~/2 (::) 114 exp (  m~:'2)Hn (if x') (9. For an equation like (9.E ) o/E(X') = 0 . Using the momentum representation. corresponding to energy eigenvalue En. En = hW(n + ~) (n = 0. labeled by the quantum number n. It is spanned by a denumerable infinity of basis vectors.152) L (x' IEn)(En Ix") n=O co = 8(x' . calculate the boundstate energy eigenvalue and the corresponding eigenfunction for the potential V(x) = ." The transformation coefficients (x' lEn) are subject to the unitarity conditions (9. show that 1 o/n(r) 12 is not only the probability density for the coordinate vector r.x") (9.2 mw ( (9.150) From Chapter 5 we know that only discrete eigenvalues.iP + 2.206 Chapter 9 Vector Spaces in Quantum Mechanics where o/E(X') = (x' IE). 2 11.149) which is nothing but the Schrodinger equation for the harmonic oscillator. any of these three representations can be equivalently used to expand an arbitrary vector of the same Hilbert space. Although it is qualitatively different from either the coordinate or the momentum representation. by (5.2 .151). we may use the eigenvectors of H as the basis of a representation. this equation requires that for the components of IE). 1. which we designate as the (harmonic oscillator) energy representation.g5(x) (for g > 0).ax. 2. Problems 1. . .X.149). Compare with the results in Section 6. if the system is known to be at position r. Since all the vectors Ix') are linearly independent.4. if the system has energy En' but also conversely the probability of finding the energy to be En. Only those solutions are admissible for which E is real and the normalization (E2 IE 1 ) = f o/'::2(X')o/E1(X')dx' = 512 (0) or 5(E 1  E2) can be achieved. If o/n(r) is the normalized eigenfunction of the timeindependent SchrOdinger equa tion..2 2m . labeled by x'.151) Exploiting the freedom to choose a representation. 2.153) These conditions are satisfied by virtue of the orthonormality and completeness of the eigenfunctions (9. this condition is equivalent to o/E'( + (0) = o/E'( =0 (9.
In Section 10.7 the discussion will be extended to an important nonnorma1 operator. at least in a finitedimensional vector space. (When this happens for the Hamiltonian operator of a system. 1. we examine some further ramifications of this problem. which may be used as basis vectors spanning the space. The chapter concludes with a return to the harmonic oscillator. The normal operator A may have repeated eigenvalues.} The number A ~ which characterizes the eigenvector is called an eigenvalue.) Since any linear combination of eigenvectors belonging to the same eigenvalue is again an eigenvector belonging to the same eigenvalue. as in IA. Squeezed (or stretched) oscillator states make it possible to give an introduction to the concept of quasiparticles. IA. which include Hermitian and unitary operators. Different eigenvalues will be distinguished by their subscripts. of the operator A if AlA. 207 .. names the eigenket belonging to the eigenvalue A. We first give our attention to the eigenvalue problem for normal operators. we speak of degenerate energy eigenvalues. This is accomplished by a successive orthogonalization algorithm (Schmidt orthogonalization method). or eigenket. the Uncertainty Relations. all of which correspond to the same eigenvalue. now in terms of raising (creation) and lowering (annihilation) operators.} is called an (10.1 and illustrated in Figure 4. eigenvector. these eigenvectors form a subspace of the original vector space.). Coherent oscillator states shed light on the connection with classical mechanics and are central to the interpretation of processes in quantum optics. The original linearly independent eigenvectors in this subspace may be replaced by an equal number of orthogonal eigenvectors. preparing for applications in manyparticle theory.} is merely multiplication by a number. every normal operator has a complete set of orthonormal eigenvectors.1.} A ket IA. The physical significance of commutation relations will be discussed and illustrated by the Heisenberg uncertainty relations. An eigenvalue enclosed in a ket I ). The effect of A on IA. In this chapter.CHAPTER 10 Eigenvalues and Eigenvectors of Operators. Our main objective will be to prove that. The Eigenvalue Problem for Normal Operators. and tbe Harmonic Oscillator A thorough understanding of the eigenvalue problem of physically relevant operators and of the corresponding eigenvectors (eigenstates) is essential in quantum mechanics. the essence of which was already described in Section 4. By this we mean the occurrence of more than one linearly independent eigenvector belonging to the same eigenvalue. the annihilation operator.1) = A.
8). A.2) where PC'P I ) is the projection operator in the "direction" of "1'1' The new eigenvector qr~ belongs to the same eigenvalue as '1'1 and "1'2' but it is orthogonal to "1'1: ("I' 1> 'I'~) =0 If there is a third linearly independent eigenvector. AtIA.5) Hence.) (10. i.) = AAtIA. belonging to the eigenvalueA. Show that qr~ in (10. (10.). (A.IJ3> = O. using (10. This procedure is continued until a complete orthonormal basis has been constructed for the subspace of eigenvectors belonging to the repeated eigenvalue.81 or a (10.10) From Eqs.IA = A.e..I (10. At lAD is an eigenket of A.*{A.8IAt If we multiply (10. we replace it by qr~ = [1  p('lr ) 1  PWz )]'I'3 (10. that (.*{A. . I we obtain.) = A~*IA. AtIA.9) as well as {A.8) as well as {.1) that AtAjA. IAt = A.jAD = a{A.(A.) (l0.Ai){A~ IAi) = 0 (10.1. using (10.) = A.7).8) is also an eigenvector of A with eigenvalue A.1) and (10.*{.6) and then multiply on the left by (. for which by definition (lOA) we infer from (10.11) . we find. and may be written as (10.208 Chapter 10 Eigenvalues and Eigenvectors of Operators If '1'1 and '1'2 are two nonorthogonal eigenvectors with repeated eigenvalues. In the braket notation.6) where 1. we construct as a replacement for '1'2 the eigenvector (10. I = A.1) we have also {A.81.) = A. From (10.81. Exercise 10.6) on the left by (A..3) is orthogonal to both "1'1 and qr~. We may assume that IJ3> is orthogonal to IA.8) = 0 and that thus for all eigenvectors. For a normal operator.10) we obtain (A.3) which is orthogonal to both "1'1 and qr~.7) (10. the occurrence of repeated eigenvalues requires that the name of an eigenket be specified by additional labels to supplement the information conveyed by the common eigenvalue A. IA.* If we substitute this result in (10. '1'3' belonging to the same re peated eigenvalue..
. Equation (10.. .IA = A:(A:I Note that the last two relations follow from the first two. the representation of the eigenvalue problem (10...1) and the program for obtaining them. we have not established that any solutions to the eigenvalue problem (10. A = A~.12) Xn Xn The matrix elements Xh X2' .12) is a set of n linear homogeneous equations that possess nontrivial solutions only if Au .13) This equation of the nth degree in the unknown A is called the secular or characteristic equation. IAt = A~*(A~ I AtIA. It is convenient to take these eigenvectors also as being normalized to unity.. . . but that in an ndimensional vector space there are n orthogonal eigenvectors so that a complete basis can be constructed from them..) = A~*IA.. The Calculation of Eigenvalues and the Construction of Eigenvectors. 2. Ann )(XI) = (XI) X2 A ~ . AlA..2 The Calculation of Eigenvalues and the Construction of Eigenvectors 209 showing that any two eigenkets belonging to different eigenvalues of a normal operator are orthogonal.) (A.. .. (10. . Although we have demonstrated the orthogonality of the eigenvectors of normal operators.) (A.A =0 A n2 0 (10. or vice versa.1) takes the form of a matrix equation: . . It may be useful to recapitulate here that the eigenvalue problem for a normal operator can be expressed in anyone of four equivalent ways: . are the eigenvalues of A.A A 12 A 22 . We thus conclude that all linearly independent eigenvectors of a normal operator may be assumed to be orthogonal. ln A 2n . . . Xn are the components of the eigenvector which belongs to the eigenvalue A. . .. . . nor have we yet found a method for calculating the eigenvalues of A and for determining the corresponding eigenvectors. only for normal operators. If a basis is introduced in the space.) = A~IA. Our hope is not only that solutions exist. The roots of (10..A A 21 Ani or det(Aij ABij) = A ln A 2n Ann .. We simultaneously attack both problemsthe existence of solutions of (10.1) actually exist.13).
of a unitary operator have the luW = 1 (10. Exercise 10. be normal. the existence of at least one solution of the set of homogeneous equations is assured. AIAD = A. Because of the properties. property Prove that the eigenvalues U. the eigenvalues have modulus unity.15) . which is orthonormal to lAD.. 0 A~ 0 o o 0 A~ 0 o o o A' n 000 and the ultimately obtained basis vectors are the eigenvectors of A. Prove the converse proposition that an operator A whose eigenvectors span a complete orthonormal basis must commute with its adjoint..12) and solve the set of n homogeneous equations to obtain an eigenvector lAD. We may thus substitute A = A. in (10.2.e. If the operator is Hermitian. 0 0 0 A22 An An2  0 A23 A33 0 A2n A3n Ann  0 The n 1 dimensional matrix of the matrix elements Aij is again a normal matrix.) (10. i.. the matrix representing A will result in the completely diagonal form A. The same procedure as before can then be continued in the nl dimensional subspace.210 Chapter 10 Eigenvalues and Eigenvectors of Operators According to the theorems of linear algebra. If the operator is unitary. nl such steps. Exercise 10.14) In the braket notation. IA = A. proves that for a normal operator n orthonormal eigenvectors can always be found.(KiIA. the normal matrix representing the operator A in this new representation takes the partially diagonalized form: A.(A. Its eigenvalue problem has the same solutions as the original problem except for an eigenvector belonging to the eigenvalue A. After. the eigenvalue problem for the normal operator A appears in the form L j (KiIAI~>(~IA. We then change to a basis that includes lAD as one of its elements.) = A. This procedure. lAD and (A. which is equally applicable whether or not there are repeated eigenvalues. all expectation values and eigenvalues are real.3. I valid for a normal operator.
15): 2: (A~ IKi)(KiIAI~)(KjIA.23) (10. (A reminder: If an eigenvalue is repeated.19) Either (10. More about this is detailed in Section lOA. Since for finite dimensional matrices we know from (9. an additional label is required· to characterize the eigenvectors. if we choose some other basis. Quite different algebraic methods for determining eigenvalues and eigenvectors of certain privileged operators will be discussed later in this chapter..) (10. Hence.1 + . the eigenvectors may all be chosen to be orthonormal: (A.13). and it is unitary.) = Se n (10. as is seen explicitly in the relation obtained from (10.115). basis IK i ) into one spanned by the eigenvectors of A.24) = trace(SlAS) = trace A .Al) = 0 In this proof the property of determinants det AB = det A det B (10. + detA = 0 (10. the eigenvalues of A as defined by (10.Ai) = det[Sl(A .2 The Calculation of Eigenvalues and the Construction of Eigenvectors 211 As we have seen.15) or (10. I is the projection operator for the eigenvector IA D. The resulting form of the matrix representing A is diagonal.Sen ij (10. Indeed.20) has been used.21) the coefficient of each power of A must be independent of the choice of representation. A = SlAS the new characteristic equation is det(A .18) changes the original.130). trace A (10.19) may be used to determine the components of the eigenvectors of A. linked to the previous one by the similarity transformation (9. It is easy to prove these properties for the trace and the determinant of A directly.22) = trace(CAB) = trace(BCA) (10. Although a particular representation was introduced to calculate the eigenvalues of A as roots of (10. I = i=l Pi = 1 2: i=l The transformation matrix S with matrix elements n (10. Consequently. if we expand the characteristic equation explicitly. trace AB = trace BA it follows that trace (ABC) hence.17) where Pi = IAD(A".16) and they form a complete set: 2: IAD(A. the eigenvalues are properties of the operator and cannot depend on the choice of basis.13) are independent of the representation.Al)S] = det(A ..IA. (At + (traceA)(At.) = A.
using (10. consider the matrix e defined as eA = 1+ .30) in the form f(A) = j~ f(Aj) n' n' A~l . we treated the eigenvalue problem for a normal operator as a problem in linear algebra.. + A~ = sum of the eigenvalues of A X A. + A.Aj (i*J) (10. + . we have the useful relation det eA = If eAt = exp( ~ A.25) It is therefore legitimate to consider trace A and det A to be properties of the operator A itself and to attach a representationindependent meaning to them.26) = A. eigenvalues. Prove that a normal ndimensional matrix A satisfies its own characteristic equation (10. Hence.5. Variational Formulation of the Eigenvalue Problem for a Bounded Hermitian Operator.29) Exercise 10. n' and express (10. I! . = product of the eigenvalues of A A (10.20).29) directly from (10... From the completeness relation (10. Exercise 10.212 Similarly. In the last section. . but the task can be prohibitively complicated if the dimensionality n is very large. we take a . from the diagonalized form of the normal operator A we see that trace A and det A = A. we may label these nonrepeating eigenvalues by the subscripts i = 1. A 3 + ..!. A 2 + .. X A~ (10. The characteristic equation provides a means of calculating eigenvalues to any desired approximation.A }] A.30) provided that the singularities of f(z) do not coincide whh any eigenvalue of A.4. Furthermore. ... . (The CayleyHamilton theorem states that this is true for any matrix... X •... A + .!. In this section. Prove (10.) 3.31) showing that any function f(A) can be written as a polynomial in A of degree less than n.27) As an application.28) without recourse to the If f(z) is an analytic function.28) The eigenvalues of e A are eAt..) = exp(trace A) (10.21). Chapter 10 Eigenvalues and Eigenvectors of Operators det 11 = det(SIAS) = det A (10. the function B = f(A) of the normal operator A can be expanded in finite powers of A very simply. 2! 3! = lim N~ (1 + ~) N N (10.17) we see that f(A) = i=1 2: f(ADP n i (10.. and show that An can be expanded as a polynomial in A of order less than n. If n' of the n eigenvalues of A are distinct.!.
Ao)'I}f0 and thus conclude ((A . we obtain the result === Ao ± e[(IP. (A . $ A~ . (A . which it assumes for the vector 'I}f0: Let us calculate A for 'I}f = '1'0 ± eIP.32) a minimum is an eigenvector of A. allowing the identification Ao = AD if AD $ A. especially Hamiltonians..Ao)'l}fo) + ('I}f. 'I}f) (10. Suppose that A has a greatest lower bound Ao. and Ao is the corresponding eigenvalue. A 'I}f) ('I}f." We note immediately that if 'I}f is an eigenvector of A. a vector 'l}fo that makes A of (10. This assumption covers all observables.Ao)lP) eQ $ === 0. (A . Then owing to the Hermitian nature of A.32) and look for the particular 'I}f which minimizes (or maximizes) A['I}f]. Since Ao is the greatest lower bound.Ao)lP o) $ eQ Now let e ~ O.Ao)'I' 0) + ('I}f 0. We employ a" variational principle by defining a realvalued functional A['I}f] = ('I}f. Evidently. we may choose it equal to (A .1 the variational method was already introduced as a useful tool for estimating the lowlying eigenvalues of the Hamiltonian operator. A:w) ('I'. is Hermitian. it must be the least of all eigenvalues. The RayleighRitz method described in Chapter 8 links the two approaches and takes advantage of the variational principle to justify the use of an approximate characteristic equation of lower dimensionality. A'I}f) by ('I}f.. (A . (A . By dividing ('I}f.Ao)IP)] ~ () Since Q == (IP.Ao)lP) ±e(IP. where e is a small positive number and IP is an arbitrary vector. (A . .3 Variational Formulation of the Eigenvalue Problem 213 different tack and assume that A. In Section 8. 'I}f) we have made A independent of the "length" of'l}f and dependent only on its "direction.Ao)'I}f0. Since IP is arbitrary. '1') JL .. We now consider a new variational problem and construct ['I}f] = (W. we have A['I}f0 ± eIP] Upon substitution. (A . we find by applying the above inequality that (IP.Ao)'I' 0) =0 which implies that A'I}f0 = Ao'I}f 0 Thus. such that then A = A. rather than being a general normal operator.
The physical meaning of Hermitian operators as candidates for representing observables motivates us to use basis vectors that are eigenvectors of Hermitian operators. there are n orthogonal eigenvectors. We look for additional observables that share common or simultaneous eigenvectors with the operator A. such as the energy. have only a lower bound. the direction of linear momentum can serve as an observable that complements the characterization of the degenerate energy eigenstates. independent of "1'. Many operators common in quantum mechanics. The presence of repeated eigenvalues for a physical observable. "I' runs through the entire subspace of n . For the free particle Hamiltonian. Hence.I dimensions. 4. The generalization requires only that the operator A be bounded at least from one side. but other choices are possible. such that (A "1'.214 Chapter 10 Eigenvalues and Eigenvectors of Operators where "I' is the orthocomplement of "1'0: As "I' ranges through the ndimensional space. and we will eventually exhaust the entire ndimensional space after n steps. we would like to identify all basis vectors by the eigenvalues of the observable that supports a particular basis. selected because of its relevance to the system under study. In this manner we may continue. While the variational proof of the existence of eigenvalues as given here is limited to Hermitian operators. it has the merit of avoiding the explicit use of a representation. the eigenvalues of a single observable A are usually not enough to characterize a basis unambiguously.4 is a case in point. Also. orthogonal to "1'0. it follows that boundedness of A is assured if. there exists a positive number C. The same argument as before gives for the minimum /Lo of /L: and "I' must be the eigenvector belonging to the second lowest eigenvalue. and we must resort to additional labels or "quantum numbers" to distinguish from each other the different orthonormal basis vectors that correspond to a particular eigenvalue A. The operator A is said to be bounded if A as defined in (10. Because of the occurrence of repeated eigenvalues. Commuting Observables and Simultaneous Measurements. Owing to the translational symmetry of the Hamiltonian. for a given A. since it makes no essential use of the assumption that n is finite. The example of the Hamiltonian of the free particle in threedimensional space in Section 4. /Lo = Ai. "1') for every "1'. and the braket notation was designed with that objective in mind.32) is bounded: +00 > A > 00. it can be generalized to the case n ~ 00. A "1') :5 C("I'. In the next chapter we . can usually be attributed to some symmetry properties of the system. the energy eigenvalue E :::: 0 alone does not suffice to identify an energy eigenstate. Ideally. but that are sufficiently different from A so that their eigenvalues can serve as distinguishing indices for the eigenvectors belonging to repeated eigenvalues of A.. From the Schwarz inequality.
) The commutation relation (10.. or I fA.. of which the free particle is a special case. AB . and L z.35) for A and B is not only necessary. AlA. . a necessary condition for two observables to be simultaneously measurable is that they commute.6. which differentiates between the r orthogonal eigenvectors belonging to the same repeated eigenvalue A~. we obtain by subtraction If this is to be true for all members of the complete set. B] == AB .34) by A. K). In an oftenused but opaque terminology.. which we assume to be r in number.33) and BIA. We emphasize that the commutivity is a necessary condition only if all the eigenvectors of A are also to be eigenvectors of B. the angular momentum is the additional observable of choice.BIA. BIA~.Bj) == 0 (10. K) == ABIA~.34) both hold.33) is multiplied by Band (10. it follows that BAIA~. we require that there be a complete set of states IAiBj ) such that (10.Bj) == BjIA~Bj) (A~ ..BA == 0 I (10. A limited number of eigenvectors may be common to both A and B even if the two operators do not commute. we consider a particular repeated eigenvalue A ~ of A and its associated eigenvectors. AlBIA.. K) (10. To show this. Ly .4 Commuting Observables and Simultaneous Measurements 215 will see that for any system with rotational symmetry. we ask under what conditions two observables A and B possess a complete set of common eigenvectors. letting K serve as an index K == 1 . We denote the corresponding eigenvectors by IA~.35) Hence. A)(A. If (10. K) (10. K) is also an eigenvector of A with eigenvalue A~. Mathematically.37) Exercise 10. B has no nonvanishing matrix element between eigenstates corresponding to different eigenvalues of A. hence. but it is also sufficient for the existence of a common set of eigenvectors for the two operators. A and B are said to be simultaneously measurable or compatible observAbles. Show that if A and B commute.BA)IA. which in Chapter 11 will be shown to be a common eigenstate of the noncommuting operators Lx. K) == A~IA~.. Confining our attention first to a single operator B in addition to A.BA must be equal to the null operator. Such eigenvectors would then represent states in which definite sharp values A~ and Bj can be assigned simultaneously to two observables. r. K) = A=1 2: r IA~.36) If B commutes with A. K) == A. K) This equation shows that BIA~. (An example is the state of zero angular momentum.
C£. K = 1.. A) From (10. .. .. K)  Bj8 AK )Skj = 0 (j fixed.. Bj. .. They are the kets designated as IA.. and B j in common.40) A=1 This system of equations possesses nontrivial solutions only if the determinant of the coefficients vanishes: (10. . If the choice is made intelligently. Bj. . B.. K e = K m means that in (10.. a change of basis may now be effected in order to construct a new set of eigenvectors of A. . B. . . and K e (omitting the prime for simplicity) is a symbol for the set of eigenvalues A. .. If among the r eigenvalues Bj (j = 1. As in Section 10.39) we obtain r linear homogeneous equations for the r unknown transformation coefficients S Kj: ~ «A..Bj) = = ~ ~ IA. K=1A=1 K)SKj r r (10. r) (10. K)SKj (10. .39) Bj IA. C£ = C.Bj) are the required simultaneous eigenvectors of both operators A and B. This simple situation was implicitly assumed to hold when we introduced bras and kets in Section 9... the r roots of this characteristic equation give us the eigenvalues B~.. . r) there remain any that are repeated. In the rare case that all eigenvalues of A are simple (no repeated eigenvalues).Bj) = Bj ~ ~ K=1 A=1 sKj8KA IA.43) The operator K here symbolizes the complete set A.. A)(A. D... which are simultaneously also eigenvectors of B. We assume that (10. .... .. . whose n common eigenvectors can be characterized completely by the eigenvalues A.. A alone constitutes the complete set of operators K = A.40) can then be used to calculate the transformation coefficients SKj' The new vectors IA.. = A.. such that no two eigenvectors have exactly identical eigenvalue addresses. .42) each pair of eigenvalues satisfies the equalities: A. Bj = B~.. One then continues the procedure of choosing additional Hermitian operators C. .. this set of operators is said to be complete.41) B. Equations (10.. AlBIA. all commutipg with A and B as well as with each other. C£...42) but often we write for this simply the compact orthonormality condition (10. it will eventually be possible to characterize all n basis vectors in the space by addresses composed of sets of eigenvalues A.4. Bj.. C.. then a further index may be used to distinguish those eigenvectors that have the same values A... In particular. B~. AlBIA. C.216 Chapter 10 Eigenvalues and Eigenvectors of Operators In the rdimensional subspace of the eigenvectors belonging to A. .Bj) = K~1 ~ !A. C£.2... . If we can find a set of commuting Hermitian operators A.38) Here the coefficients S Kj are defined by the conditions r r BIA. .
The Heisenberg Uncertainty Relations.44) where (10.45) is the projection operator for K f • These equations can be used to reformulate the eigenvalue problem for one of the operators in the set symbolized by K. Equations (10. . .1 5. whereas the eigenvectors belonging to a repeated eigenvalue A.. This 1 For a discussion of functions of operators and matrices. is denoted by (10. A onedimensional projection operator like (10.45) is said to have "rank one" to distinguish it from higher rank projection operators like (10. If the partial sum of all projection operators. If A and B are two Hermitian operators that do not commute.48) extend over all distinct eigenvalues of A. The coefficients in this expansion are the eigenvalues of A. = 1 (10.. contai~ an element of arbitrariness. PA . . a normal) operator A.46) The eigenvalue problem for A can thus be expressed as follows: Given a Hermitian (or more generally. the completeness of the eigenvectors is expressed as the closure relation L Pf f =L f 1 Kf)(Kf 1 = 1 (10. For example. We have seen that only commuting observables can in principle be measured and specified with perfect precision simultaneously. using the closure relation: A = L AIKf)(Kfl f =L f A. decompose the space into a complete set of orthonormal vectors IK f ) such that A is a linear combination of the projection operators IKf)(Kf I.48) The sums in (10. the physical quantities A and B cannot both be sharply defined simultaneously. For example. This form of the problem is convenient because the operators PA . see Merzbacher (1968).48) define what is called the spectral decomposition of the Hermitian operator A. care must be taken to interpret sums over all sets of eigenvalues of K properly.5 The Heisenberg Uncertainty Relations 217 In letting K symbolize the entire complete set of commuting operators.49) Since this sum extends only over distinct eigenvalues of A.47).47) we may write L . which correspond to the same eigen value A. we have and hence. are unique. 1 Kf)(Kf I (10.. For a function f(A) we may write (10.31). this is the same equation as (10.
B] == AB .:::: ~ (C) 1 (10.57) i =F+C 2 .54) Proof Since A is Hermitian. Exercise 10.84).75) by making the identifications '1'a = (A(A»'1' and '1'b = (B(B»'1' and get (LlA?(LlB)2. based on the decomposition (9.:::: G 1 (C) r (10. (10. We define lAB .2. (B(B»'1') We now apply the Schwarz inequality (9. we now define the uncertainty LlA in A as the positive square root of the variance.52) For these quantities we will prove that (LlA?(LlB? . (A(A»'1') if '1' is the state vector of the system. Prove that C is an Hermitian operator. (10.56) Now we use the simple identity.C 2 2 (10. for B.55) The equality sign holds if and only if (B(B»'1' = A(A(A»'1' (10. To make it precise.:::: 1('1'.BA of A and B is different from zero may give us information about the inevitable lack of precision in simultaneously specifying both of these observables.51) and similarly. The uncertainty LlA in an observable A was first introduced qualitatively and approximately in Section 2.50) The imaginary unit has been introduced to ensure that C is a Hermitian operator.53) or LlALlB .7.BA = icl (10. (A(A»(B(B»'1'W (10. Similarly. (LlB)2 = «B(B»'1'.218 Chapter 10 Eigenvalues and Eigenvectors of Operators suggests that the degree to which the commutator [A. (A(A»(B(B» = (A(A»(B(B» + (B(B»(A(A» + !. we can write (LlA)2 = «A(A»'1'.
60) From (10.64) .56) we can obtain two simple relations: (W. Applied to A = x. (10. The Heisenberg uncertainty relation is thus seen to be a direct consequence of the noncommutivity of the position and momentum operators. we can write (10. (IlB)2 A = Subtracting the lefthand sides gives i(C). The righthand side of (10.62) xpx . Since their expectation values are real. B = Px' in the coordinate representation the equation (10.60) we obtain A= (10. (A(A»)(B(B»)W) = A(IlA)2 (W. From (10.63) making specific and quantitative the somewhat vague statements of Chapter 2 and elsewhere. and we conclude that for all states (10.56) and (10.55) as which proves the theorem (10.58) It is of interest to study the particular state W for which" (10. hence by (10. hence A(IlA)2  ±(IlB? = i(C) 2(IlA? i(C) (10. let A = x.!.47).54) is independent of the state W in this case. (B(B»)(A(A»)W) = ±(1lB)2 0 (10.61) As a.58)..53) becomes an equality: IlAIlB = ~ I(C) I Such a state obeys the conditions (10.58) A(IlA)2 + .56) becomes a differential equation for the wave function '" representing the special states W that make (10. as in (3.special case.53). The last equality holds if and only if (F) = 0 (10.63) an equality: (10. B = Px' Then.59) and (10.5 The Heisenberg Uncertainty Relations 219 by which the operator on the left is written as a linear combination of two Hermitian operators F and C.59) Adding the lefthand sides yields 2(F).pxx = inl and hence C=nl.
. We first consider the eigenvalue problem of H.7. these methods exploit the commutation relations among operators and operator identities shown in Chapter 4..56). with particular values of (x) and <PJ.mw2i = 2m 2 2 mw mw  liw 2 (10. qp . the state represented by (10.66). an eigenfunction of the nonHermitian operator Px .66) Known somewhat imprecisely as a minimum uncertainty wave packet.220 Chapter] 0 Eigi!nvalues and Eigenvectors of Operators The equation for l/J is ( Ii d i dx (pJ ) l/J = 2(Ax)2 (x .Ax. Exercise 10. .66) is known as a coherent state.68) Both p and q have continuous spectra extending from ~ 00 to + 00.pq = iii] (10. this state represents a plane wave that is modulated by a Gaussian amplitude function. it is instructive to return to it and to treat it here as an application of algebraic operator techniques. is the ground state wave function.. of which (10.p) H = p2 + .p ) ( q .67) where the Hermitian operators p and q satisfy the canonical commutation relation. Instead of using the more traditional tools of differential equations. 6. according to (10. is given by 1 mal ( q + l. By use of the commutation relation. .(x)i 4(Axi i<Px)X] + Ii (10. and integral representations. Relate Ax to the mass and frequency ofthe harmonic oscillator.69) which is not Hermitian. The Harmonic Oscillator.(x»l/J iii (10. The generalized linear harmonic oscillator is a system whose Hamiltonian.8. Although the harmonic oscillator has been discussed in detail in Chapter 5.65) and has the normalized solution l/J(x) = [27T(Axir 1l4 exp [ (x . As such.66) is.70) where at is the Hermitian adjoint of a: (10.l.71) . with properties that will be discussed in Section 10. {2i (q + i mw L) = liw(aat (10. It is convenient to introduce a new operator a = rr. Since A is imaginary.. expressed in terms of two canonical observables p and q. special functions. because it will give us the important stationary states of the system. we prove easily that H = liW(ata + D ~) (10. the expression (10.
When by successive downward steps an eigenvalue between 0 and 1 has been reached. However. Since by (10.77) Consequently. From (10. we obtain I>Ao~O and alO) = 0 (10. 2.. . Since we have arbitrarily (but conveniently) labeled the lowest step in the ladder by setting n = 0. These properties justify the designation of at as the raising operator and a as the lowering operator.) and the corresponding. we can show that a In) is also an eigenvector of at a with eigenvalue An 1. we prove that An ~ O.76) limits the number of times a lowering operator can be applied. because that would be an eigenvector whose eigenvalue violates the restriction (10. Ao = 0 (10. the eigenvectors of H and of ata are the same. as can be seen from the equation (ata)atln) = at (a t a+l)ln) = (An+l)atln) where the commutation relation (10. First.72) has been used.78) and this is the only eigenvalue below unity.73) we have (10. we conclude that If In) is an eigenvector of ata. 'l'n) (10. condition (10.74) we get ('I'n. Similarly. ata'l'n) = An('I'n.75) This is the equation that we must solve.72) which shows that a is not even normal. and it is sufficient to solve the eigenvalue problem for ata. Expressing the eigenvalues as An (n = 0. Hence.76).70) H is a linear function of ata. a'l'n) = An('I'n.6 The Harmonic Oscillator 221 The commutator of a and at is (10. we can generate from any given eigenvector In) new eigenvectors with different eigenvalues by what is graphically called a ladder method.76) Since (9. 'l'n) or (a'l'n..74) or (10.35) holds for all vectors. . with eigenvalue An + 1. 1. then at In) is also an eigenvector. by applying a again we do not obtain a new eigenvector. By applying these operators repeatedly. eigenvectors by 'l'n = In) (10. atln) is an eigenvector of ata.
Combining (10. The notion of a particle as an excitation of an oscillatorlike dynamical system (such.. and it is then called the number operator. the eigenvalues of the Hamiltonian are (10..79) and (10. excitons. and so on.80).70). leads us to conclude that the matrix elements of a and at connect only neighboring basis states: (n . we obtain all other eigenvectors and eigenvalues by repeated application of the raising operator at.9. (10. (n = 0.222 Chapter 10 Eigenvalues and Eigenvectors of Operators Starting from 10). The ladder property of the lowering and raising operators a and at. is variously known as the ground state. see Chapter 21. the operator N = ata plays a central role when the number of (identical) particles is an observable of a system.75).81) Hence. photons. depending on the physical context in which the system is represented by a harmonic oscillator. The eigenvalues increase in unit steps. as an electromagnetic field or an elastically vibrating body) has been at the core of quantum physics from the beginning. What are the eigenvalues of the kinetic and the potential energy operators ofthe harmonic oscillator? Explain why these don't add up to the (discrete) eigenvalues of the total energy. each carrying energy hw. Exercise 10. the vacuum state. which must not be confused with the null vector 10). For more detail about the quantum mechanics of identical particles.82) in agreement with the discrete energy eigenvalues found in Chapter 5. These particles or quanta are named phonons. The set of eigenvectors obtained is complete.83) (10. Since its eigenvalues are all the nonnegative integers.80) The normalization constant N n must yet be determined. . There is no degeneracy as long as no dynamical variables other than p and q appear to characterize the system.84) (n + llatln) = (nlaln + 1)* = C~ +1 To evaluate Cn we may use the closure relation 00 n'=O 2: In')(n'l = 1 (10. 2. or the noparticle state of the system. and (10. we obtain (10.) (10.85) . An excited energy level of the harmonic oscillator with quantum number n is interpreted as corresponding to the presence of n particles or quasiparticles. The eigenstate 10). 1. and the orthonormality of the states In).1laln) = Cn (10. Hence.
87a) .91) .6 The Harmonic Oscillator 223 as follows: Icn l2 = (n .llaln) = (nlatln = 2: (nlatln')(n' laIn) = (nlataln) = n n' 1)(n .10.I) (IO. The coordinate q of the oscillator can be expressed as q = J 2mw Ii.90a) o o o o o o V2 o o o v'3 (1O. We may therefore write aln) .llaln)*(n .0 0 5 2 (10. Since there is no other restriction on the matrix elements.w H=.90b) Exercise 10.90a) and (l0.Iialn) = vnln . verify the comat a = 1.88) In the representation spanned by the basis vectors In). (a + at) (10.Iialn) Thus.90b). and atl:n) = In I)(n + Ilatln) = vn+1ln + 1) (10.87b) From here it follows that the normalized eigenkets of at a are l'I'n = In) = (n!)1I2(at y'l'o = (n!)1I2(at YI0) I 100 030 li. the matrix H is diagonal and given by (10. mutation relation aa t  Using the matrices (l0. (10.I)(n .86) = In + .89) The matrices representing a and a tare a= 00 o o o 0 V2 o o 0 o v'3 o (10. an = 0 for all n is a possible and consistent choice for the phase.
: q'Z)Hn ( in agreement with Eq..95) by virtue of the recurrence relation.85).. for Hermite polynomials. n + .95).224 Chapter 10 Eigenvalues and Eigenvectors of Operators and its matrix in the In) basis is q= ~~ o Vi o o v'2 ~ o v'2 o Vi (10.r cn\ = J2mw q .93) leads to a set of simultaneous linear equations: Vi C\ v'2 Cz + Vi Co 1 C +\ V = J2~W q' Co = J2~W q'c\ C . and the eigenvalue equation for q appears in the form o Vi 0 Viov'2 o v'2 0 Co Co C\ c\ Cz = q' Cz (10. n vn + n T These simultaneous equations are solved by C n(q') = 2(nI2)(n!)\12Hn ( J¥ q' )co (10.94) Equation (10.93) where the components of the eigenvector of q are the transformation coefficients C n = (nlq') (10.92) An eigenstate of the coordinate q with eigenvalue q' is represented by a column matrix. (5. represented in the form L (q'ln)(nlq") = 8(q' n=O 00 . The result is (nlq') = T(nIZ)(n!)lIZ(::Y/4 exp( ..39) J¥ q') (10.:. The closure condition (10.96) .q") finally determines the constant factor in (10.
12.100) and This result shows that the solution of the eigenvalue problem (10. Using the mathematical tools of Section 5. (10.77) and (10. this eigenket can be expressed as Ila) = eaatOl*alo) = elaj2/2eaatIO) I (10.101) is normalized.103) These eigenstates of the lowering (annihilation) operator a are known as coherent states. this equation is the ground state 10) for a (10.58) we see that (10. (ala) = (010) = 1 (10.66) will be brought out shortly. Transcribe Eqs.3.7 Coherent and Squeezed States 225 Exercise 10. 7. The general state of an harmonic oscillator can be expressed as a superposition of the energy eigenstates 11?).11. Verify the recurrence relation for Hermite polynomials from Exercise 10. Since DOl is unitary. It is interesting to note that the coherent states are normalized to unity. with eigenvalue a I ala) = ala)1 = 0. the eigenket la) in (10. verify Eq.97) as seen from A trivial solution of.99) that DOl has the properties (10. since from Eq.88) in the coordinate (q) representation and calculate (q' In) from these differential relations.98) and (10. (3. the formula (5.102) Using (3.77). A class of states that is of particular importance consists of the eigenstates of the nonHermitian lowering operator a.1).61) and (10. even though the eigenvalue spectrum of the operator a is continuous? 2The first comprehensive treatment of coherent states was given by Glauber (1965).97) may be taken to be (10.96). (10.101) and that all complex numbers are eigenvalues of the operator a. Squeezed States. Coherent and.77).98) causes a shift of the operator a. For some purposes it is instructive to depict the eigenvalue a of a coherent state as a vector in the twodimensional complex plane (see Argand diagram. Figure 10. We deduce from (10. (10. Their relation to the minimum uncertainty wave packets (10. The unitary shifting or displacement operator (10.99) for an arbitrary complex number a.35). .
226 <alpla> Ima= Chapter 10 Eigenvalues and Eigenvectors of Operators Y2mliw o ~Re =~<alqla> a Figure 10. A second useful unitary operator is (10. Twodimensional phase space or Argand diagram representing a coherent state I a) in terms of the eigenvalue a of the lowering or annihilation operator a.108) . Interpret this result in terms of the complex eigenvalue plane (Figure 10. show that for any coherent state D f3 la) = Cia + (3) (10. There is an eigenstate I a) of a for any complex number a.1). Exercise 10.99).1). RAla) = C'leiAa) where I eiAa) is again a coherent state and C' is a phase factor. Since [ata.14.107) = eIe<12/zIf31212+e<f3* (10.106) Exercise 10.13. The inner product of two coherent states I a) and I (3) is (131 a) = and eIe<12/zIf312/Z(0Ief3*aee<atIO) = eIe<12/zIf312/Z(0Iee<a+ef3*aee<f3*10) (10. a] = a.1. but the coherent states do not form an orthogonal set.105) with a realvalued parameter A. Also shown is the effect of a "displacement" Df3 and a "rotation" R). on the coherent state I a). la). Using the property (10. Interpret the meaning of this result in the complex eigenvalue plane (Figure 10. th'e identity (3.104) where I a + 13) is again a coherent state and C is a phase factor. Show that for any coherent state I a).59) gives (10.
110) The mean (expectation) value of n for this distribution is I a 2 1 • Exercise 10.111).88). as . Prove Eq. with eigenvalue a.15.17.109) In)elaI2/2 ~ n ~ In the last step. This is most easily done by expanding the coherent states in terms of the harmonic oscillator eigenstates. Prove that the raising operator at has no normalizable eigenvectors. By requiring that I a) = 2: In)(n I a) is an eigenket of the operator a. Exercise 10.131 in the complex eigenvalue plane (Figure 10. an identity bearing a remarkable similarity to a closure relation can be proved: .) Exercise 10. using (10. n when the operator ata is measured is thus given by the Poisson distribution: 1 (10.5. Even so.88) and the orthonormality of the energy eigenstates have been used. (10. How can this be reconciled with the probability doctrine of quantum mechanics? (See Section 15.16. JIa) d 2a (al = . Eq.2: n=O '" '" (t)k In)e.1) measures the degree to which the two eigenstates are approximately orthogonal. the distance I a . JIa) dRe(a) dIm(a) (al = 1 (10. Exercise 10. and explain the reason.109). Verify (10.111) Here. Evaluate the integral over the entire complex 13 plane. and interpret the result. the integration is extended over the entire a plane with a real element of area. can be written. the latter form an overcomplete set.Ia1 212(n I ~ 10) k=O k! 2: = 2: n=O '" (10. we calculate la) = 2: n In)(nla) = 2: In)(nlelaI2/2eaatIO) n =. which has the coordinate q as its complete set of dynamical variables. To expand the coherent state I a) in terms of the energy or the numberoperator eigenstates In). As we might expect from the lack of restrictions imposed on the eigenvalues and eigenstates of a.7 Coherent and Squeezed States 227 Hence. (10. obtain a recurrence relation for (n I a). on account of (10. The probability Pn(a) of finding the coherent state a) to have the value.18.109) and plane polar coordinates in the complex a plane. An arbitrary state of a system. An arbitrary state can be expanded in terms of them in infinitely many different ways.
112) F(a*) repre . For example: A = a2 a t = a(1 + ata) = a + (1 + ata)a = 2a + a t a2 shows how normal ordering is achieved. Exercise 10. by use of the commutation relations. using the commutation relation (10. a) = A(a t ). in a coherent state a) the expectation value of a product of normally ordered operators.113) For example. Consider an operator A(a t .72).114) and written as a product of expectation values of at and a. In general. a:* )F(a*) represents the state sents the state 1'1'). An operator is normally ordered if. AI '1') in the same sense as. n2. and (An)2 directly from the Poisson distribution (10. like (a tram.115) where iJ/aa t denotes formal differentiation. can be factored: (10. (a» 1 (10. Check the results by computing the expectation'values of n. a)F(at)IO) = A(at. a: t t )F(a ) I0) (10.110).228 Chapter 10 Eigenvalues and Eigenvectors of Operators where en = (nlqr) and F(a t ) is a convergent power series (entire function). a) I a) of a normally ordered operator is (aIA(a t . its square and its variance. What function F( a*) represents the coherent state 1/3)? The action of a normally ordered operator on an arbitrary state can be expressed conveniently in terms of this representation.19. a)la) = A(a*. such factorizations are not permissible in quantum mechanics.20. all lowering operators have been brought to the right of all raising operators. a)Iqr) = A(at . Hence (alqr) = (aIF(at)1 0) = F(a*)(ajO) = e. but coherent states enjoy the unusual property of minimizing certain quantum correlations (or quantum fluctuations). The term coherent reflects their important role in optics and quantum electronics (Section 23. evaluate the expectation value of the number operator ata. This has led to their designation as quasiclassical or semiclassical states. and let it act on an arbitrary ket A(at . we infer that the entire function A(a*. For a coherent state I a). This last equation follows from the commutation relation and the property a I 0) = O. a) which is normally ordered. according to (10.jo'i / 2 F(a*) The entire function F(a*) thus represents the state qr. 2 (10.4). Hence. The expectation value (aIA(a t .112) Exercise 10.
119) so that (10. Exercise 10.120) showing that in the coordinate and momentum language the coherent states are minimum uncertainty (product) states. Choose A = ata.6... The variances are (10. Verify that the eigenvalues must be nonnegative integers.(at . taking advantage of normal ordering. (alqla) = I J!  2mw (a + a*) = ~Ii Re a mw (10. As discussed in Section 9.116) ".(ala 2 + 2mw = Ii a t2 + ata + lla) Ii [(a + a*)2 + 1] 2mw = (alqla)2 +  Ii 2mw and similarly The last terms on the righthand side are proportional to Ii and exhibit the quantum fluctuations in q and p. By definition '(10.a) = ~~ 1m a T2v2limw V Furthermore.21.117) and (alpla) = i ~ (a* . we can calculate (alila) = .7 Coherent and Squeezed States 229 Exercise 10.69).a) liW (10. in the coordinate representation the eigenvalue condition ala) = ala) .22. the number operator. Rederive the function F(a*) which represents the coherent state 113> by letting A = a and requiring aL8) = 13113).(at + a).118) (10. the nonHermitian operators a and at are related to Hermitian canonical coordinate and momentum operators as " q = . and obtain the entire function F( a*) which represents its eigenkets. 2mw The expectation values of q and p in a coherent state Ia) are J! p =i f¥ 2 .
b t can be expressed in terms of a and at by substituting (10.123) and (10. (10. Compute the normalization factors C' and C in (10.i) (10..~)(q I a) 2mw aq = fiK V..127) where A and v are two real numbers that are related by the condition A2  ..(q)2) 4(aq)2 J In this form. btl = 1 fi: fi: 1 P w ( q+i. known to us from (10.l = 1 (10. specified by a given value of mw.124) mw' (10..: ( q  = C exp [ (q . for the same oscillator by defining a new set of raising and lowering operators b= bt = w ' ( q 2ft using an arbitrarily chosen positive parameter w'.126) or b = Aa + vat bt = Aat + va (10.. : (q = ~ a YJ ~ Re a Y i qv'2ftmw 1m a J + +i i (p)q (10.116) into (10.!!:. the uncertainties in q and p are both equal to vJJ2.125) The operators b.230 is transcribed as (q Chapter 10 Eigenvalues and Eigenvectors of Operators + ... coherent states have been familiar since the early days of quantum mechanics.. given in Eqs.. the coherent states are the manifold of those minimum uncertainty states that have definite values for aq and ap. .119). Obviously [b.65) and the solution (qla) = C' exp [ . For a fixed oscillator mode.122) and show how they are related.128) (with A > 1).66).121) This differential equation has the same form as (10.) 2ft mw' p . a(ql a) (10.118) and (10.123) (10.23. socalled squeezed states.) We can construct other minimum uncertainty states with narrower width aq.122) c exp [  . Exercise 10. (If mw = 1.124): (10...
(A + 2 2 v? so that as it should be for a minimum uncertainty state. for instance. subject to the restriction (10.kx wave packets defined by Figures 2.24. is referred to as a quasiparticle transformation.8) verify the values of liq and lip given Prove that the operators a and b are related by a unitary b = Uavt where (10. by choosing very large values w' » w to reduce liq arbitrarily at the expense of letting lip grow correspondingly. [Hint: Use identity (3. we have *" .] Problems 1. these states have been named squeezed states in quantum optics.54).. Since it is possible. in (10..128). and b annihilates a quasiparticle. Carry out numerical integrations to test the uncertainty relation fJ.130) From the relations 00. but the uncertainties of these quantities are determined by w'.123) and (10.26.= .xfJ. If (10.129) The eigenstates of the lowering operator b are defined by (10. a = Ab . which is a superposition of particle states.1 and 2.124) it is apparent that these states are minimum uncertainty states for p and q. Although w is fixed.Problems Exercise 10.131). the transformation (10. In the language of bosons (Chapter 21).131) = . 231 Verify (10.127) is inverted. the uncertainty in either q or p can be controlled in these states by changes in the paraqleter w'. The operator b t creates a quasiparticle.2. Show that U transforms a coherent state into a squeezed state.128)..vb t at = Ab t .v? (10. which generally does not preserve the number of particles (since btb ata). Exercise 10. transformation For a squeezed st~te 1. and not by w: (liq)2 (lip? = h. ~ 1/2 for the .127).vb (10..= h A ( 2mw' hmw' 2mw hmw .25.132) and e l = A + v. Exercise 10.
If a coherent state Ia) (eigenstate of a) of an oscillator i~ transformed into a squeezed state by the unitary operator u = exp[~ (~ (a .a t2 )] calculate the value of !: that will reduce the width of the Hermitian observable + a t )/V2 to 1 percent of its original coherentstate value. Work out the Heisenberg relation for the product of the uncertainties in the current I and the voltage V. how low must the temperature of the device be before quantum fluctuations become comparable to thermal energies? Are the corresponding currentvoltage tolerances in the realm of observability? 7.p)2 subject to the condition Show that the solution if! of this problem satisfies a differential equation which is equivalent to the SchrOdinger equation for the harmonic oscillator. but not the explicit solutions of the eigenvalue problem. The Hamiltonian representing an oscillating LC circuit can be expressed as Q2 <1>2 H = 2C + 2L Establish that Hamilton's equations are the correct dynamical equations for this system. Rederive the onedimensional minimum uncertainty wave packet by using the variational calculus to minimize the expression 1= (Ii. q. 6. 3. Assuming a particle to be in one of the stationary states of an infinitely high onedimensional box. and show that they agree with the Heisenberg uncertainty relation. and show that the charge Q and the magnetic flux <1> can be regarded as canonically conjugate variables. What happens to the width of the conjugate observable (a . p (or the dual pair p. Calculate the value of li. Using the uncertainty relation. If a mesoscopic LC circuit has an effective inductance of L = 1 JLH and an effective capacitance C = 1 pF.X)2(li.a t )/V2i in this transformation? .p for a linear harmonic oscillator in its nth energy eigenstate.xli. and calculate the minimum value of li. q).232 Chapter 10 Eigenvalues and Eigenvectors of Operators 2. Also show that in the limit of very large quantum numbers the uncertainty in x equals the rootmeansquare deviation of the position of a particle moving in the enclosure classically with the same energy. 4. 5. show that the expectation value of the energy of a harmonic oscillator can never be less than the zeropoint energy.p.xli. calculate the uncertainties in position and momentum.
For example [LX' y] [Lx. Orbital Angular Momentum. y] = = [y.ZPy. The operator that represents angular momentum in the coordinate representation is obtained from (11. 1.4) 233 . Beyond its relevance to the classificatio~ of energy levels in centralforce systems. Central forces are derivable from a potential that depends only on the distance r of the moving particle from a fixed point.1) Since central forces produce no torque about the origin.CHAPTER 11 Angular Momentum in Quantum Mechanics We now turn to the motion of a particle in ordinary threedimensional space. we must expect angular momentum to play an equally essential role in quantum mechanics.zpY' y] . The Hamiltonian operator is (11. because only products like XPy. According to the correspondence principle. In classical mechanics this is the statement of Kepler's second law. Px] = ihz ihpz ° (11. the study of orbital angular momentum brings us one step closer to a detailed exposition of symmetry in quantum mechanics (Chapter 17).3) No difficulty arises here with operators that fail to commute. Bohr found the key to the theory of electronic motion in the Coulomb field of the nucleus in the quantization of angular momentum (in units of Planck's constant divided by 21T). py]pz = [Lx. it is well to derive some of the properties of the operator L.2) by replacing p by (h/i)V: (11. x] = [ypz = [ypz = 0. Py] [L z. In view of the great importance of angular momentum as a physical quantity. usually the coordinate origin. . Py] = z[Py. the orbital angular momentum (11. using the basic commutation relations between the components of rand p and the algebraic rules of Section 3. ypz appear.4 for commutators.2) is conserved.
for finite rotations generally 1= .0 X r sin 4J (11. hence. where the displacement vector 3 may itself depend on r. to the first order. E' of(r) = F(r) .fer) = V F(r) = E' V fer) (11. V F(r) = fer) (11.cos 4J) + 0 X . r sin 4J (11.2. If the value fer) of the function at point r is displaced to the new point r + 3. Exercise 11.8) Exercise 11. Suppose fer) is an arbitrary differentiable function in space. Exercise 11. the system cannot in general be assigned definite values for all angular momentum components simultaneously. From these relations we can further deduce the commutation relations between the various components of L: [Lx.1 Because of the rdependence of the displacement. prove that 3 R(r) + 3R.ihypx + ihxpy = ihLz and by cyclic permutation (x ~ y ~ z ~ x) of this result.8) and show that it gives the expected answer for a rotation about the z axis.1. a new function F(r) is obtained by the mapping fer) ~ F(r + 3) = fer) For an infinitesimal displac:ement F(r E.1(r + 3R(r)) = 0 . Insight into the nature of the angular momentum operator is gained by noting its connection with (rigid) rotations.234 Chapter 11 Angular Momentum in Quantum Mechanics Similar relations hold for all other commutators between Land r and between L and p. + E) = F(r) + E . zpx .1) 3R = 0 X (0 X r)(l . If the r dependence of the displacement vector is explicitly indicated as 3R(r). pz] = . (11. equivalently by an angle 4J about the unit vector 0) is described by the displacement vector 3R.5) Since the components of L do not commute.7) A finite rotation R by an angle 4J about an axis that points in the direction of the unit vector 0 through the origin is characterized by the displacement vector (Figure 11.3. The inverse rotation R. 4] = [Lx.xpz] = [Lx> z]Px .3R' The rotation R causes the value of the function f at position r + 3R.x[Lx .1 by an angle 4J about the same axis (or. For an infinitesimal displacement E. Verify (11. show that 8f(r) = or = E.1 = 0 X (0 X r)(1 .9) 3R.6) and the change of the function f is.1 to be displaced to the original position r. applied to the vector function fer) = r.cos 4J) .
(b) illustrates the active rotation of a function or state f(r) about an axis (ft) perpendicular to the plane of the figure: f(r) ~ F(r) = f(r .) + Ii x r sin q. to second order in 6cfJ. (11.10) where o«\» = ro~ation with 8cfJfi is a vector of length 8cfJ pointing in the direction of the axis of an orientation defined by a righthanded screw. (a) (b) Figure 11.8) shows that the displacement can. If the inverse rotation .1. be expressed as E = E 6cfJfi X r + ~ (6cfJ)2 n X (n X r) = o«\» X r + ~ o«\» X (o«\» X r) (11. Rotation about an axis defined by the unit vector fi and the rotation angle cPo (a) shows the displacement aR(r) of the point whose position vector is r.cos q.Orbital Angular Momentum 235 o aR(r) = Ii x (Ii x r)O .aR)' If the rotation angle 6cfJ is infinitesimal.
For an infinitesimal rotation: ('I"IAI'I") . which implies that A' = URAU"k = e(ilfi)4>l\.19) for rotations. we find that this relation leads for a vector operator A to the condition (11.15) For infinitesimal rotations this becomes SA = A' .L f(r) (11.11) (11.12) The operator L/fi is called the generator of infinitesimal rotations. we define a rotationally transformed operator A' such that A 'l/J' (r) = URAl/J(r).16) yields the commutation relation (11.4. verify (11.17) Keeping only terms up to first order in the rotation angle.5).18) The operators r.20) . For any operator A. and rederive the fundamental commutation relations of rand p. and L in (11. For these. p.13) and defines the unitary rotation operator U R = e(ilfi)4>iI.L (11. p. to first order in 8l/J. substitution of (11. For an infinitesimal rotation.('I'IAI'I') = 34> X ('J'IAI'I') (11.5.4) and (11. Sf = 8l/J ft or X r· Vf = 8l/J ft· r X Vf (11.12) can be integrated for a finite rotation R about the fixed axis ft. By letting A = r. Also check that if A = r in (11.A = [A.16) A vector operator A is a set of three component operators whose expectation value in the rotated state l/J'(r) is obtained by rotating the expectation value of A in the original state l/J(r).LAe(ilfi)4>l\.19).18) into (11.14) The rotation operator UR rotates a state represented by the wave function l/J(r) into a new state represented by l/J'(r) = URl/J(r). 34> . and L = r X p are examples of vector operators.236 Chapter 11 Angular Momentum in Quantum Mechanics is infinitesimal. the displacement is simply E. Exercise 11.19) Exercise 11. The result is straightforward: F(r) = e(ilfi)4>iI.L r· V] = i h [A. the correct expression for 8r is obtained. Apply to infinitesimal translations the reasoning that led to (11.18). A scalar operator 8 is an operator whose expectation value is invariant under rotation and which therefore transforms according to the rule 88 =0 (11. Equation (11. 34> X (11. the change of the function f is then. L] (11.
to second order. by using the commutation relations (11. which is usually chosen to be L z • Thus. L of orbital angular momentum. saving the analytic approach for subsequent sections. B) = 8A· B + A· 8B = S«f» x A· B . We give precedence to the algebraic method. Lf owing to the validity of the commutation relations (11. say Lz .7. to specify and measure more than one component fi . and it is therefore possible to require the energy eigenstates of a rotationally invariant system to be also eigenvectors of L 2 and of one component of L. Exercise 11. + L.LS = [S. in preparation for solving the energy eigenvalue problem. it is useful first to derive the common eigenvectors of L z and L 2 • Just as the eigenvalue problem for the harmonic oscillator could be solved by two methods.9. Show that the difference displacement 321 . possible to specify L 2 simultaneously with anyone component of L. commutes with L 2 = L.A· S«f» x B = 0 We note in particular that the orbital angular momentum L commutes with any (potential energy) function V(r) of the scalar r.16) that for a scalar operator S: 237 SL . in that sequence.5). however.Biz! = i 8«f»2 X 8«f»1 . . The Hamiltonian.8.1 Orbital Angular Momentum It follows from (11. H = p 212m + V(r). Verify that any component of L. one analytic based on differential equations. In summary. the other algebraic and starting from the commutation relations. L] = 0 (11. If two rotations S«f»1 and S«f»2 are performed. we conclude that it is not possible. since 8(A . Exercise 11. here also we can proceed by two quite different routes. Does the equation LX L = ihL make sense? (11. Prove that r· Land p' L are null operators.312 = (S«f»2 X S«f»I) X r is effected by the rotation S«f»12 = S«f»2 X S«f»1 and that this rotation correctly induces a secondorder difference i 8~d . It is. in general. + L.23) Exercise 11.6. the total displacement is 321 = 8«f»1 X r + B«f»2 X r + 8«f»2 X (8«f»1 X r) 1 1 + 2 8«f»1 X (8«f»1 X r) + 2 8«f»2 X (8«f»2 X r) The displacement 312 is obtained by interchanging the rotations 1 and 2.23). B of two vector operators is the simplest example of a scalar operator. with the kinetic energy p2/2m and with L 2 : (11.22) Exercise 11.21) The scalar product A . for a particle moving in a centralforce field commutes with L.
Since according to (11. say lz. If we denote the eigenvalues of lz by mfi and those of J2 by Ah2.30)..lzI1+I.is the adjoint of 1+.5) except the names of the operators.33) The eigenvalues m and A.31) Exercise 11. We construct the operators (11. lz commutes with J2.28) (11.10. satisfy the inequality (11.27) (11. From the commutation relations (11. we consider . which are assumed to satisfy the same commutation relations (11. ly. Algebraic Approach to the Angular Momentum Eigenvalue Problem. Let us consider the eigenvalue problem of one of the components of J. Prove (11.5) as the three Cartesian components of orbital angular momentum: (11. we infer further commutation relations: lzl+ .25) and (11. We have replaced L by J in order to emphasize that the eigenvalue problem.31).l+lz Ilz . belonging to the same eigenvector.24) Nothing is different here from (11. lz. it is possible to obtain simultaneous eigenvectors for these two operators.26) Of these three operators only J2 is Hermitian. This option will help us to distinguish between the various independent eigenvectors of lz.34) To prove this inequality. the eigenvalue problem can be written as lzlAm) = mfilAm) j2IAm) = Afi 2 1Am) (11. We start with three Hermitian operators lx. has the capacity of representing a much larger class of physical situations than orbital angular momentum of a single particle.238 Chapter 11 Angular Momentum in Quantum Mechanics 2.29) (11. which will be solved in this section by the algebraic method. The operator I.30) (11.Il+ = = = f1J+ hI2fi lz J2J_JJ2=0 We note the useful identity (11.32) (11.24).
~. Also. In each downward step.1.37) (11. We may therefore write l+IAm) = C+(Am)hIA m+l) L IAm) = C_(Am)h IAmI) (11.39) and (11.. respectively.40) Equations (11. such that (11. there must be a lowest value of m. ·1 (llAl) . J2 1±IAm) = Ah 2 1±IAm) Hence.40) are consistent only if j' = .35) (11.h1z )IAj) = (A l  j)h 2 IAj) = 0 from which the relation between j and A follows: A = j(j + 1) = (11. there must be a greatest value of m. 1+ 1"1) = 0 Multiplying on the left by 1_.32) with 1+ and l_ and apply (11.j or j' = j + 1 The second solution is meaningless because it violates the assumption that j is the greatest and j' the smallest value of m.39) Similarly.38) where C±(Am) are complex numbers yet to be determined. Max(m) = j.. 1. we conclude that from which the inequality (11.l)hJ1 Am) (11. ~. Hence j' = .36). Next we develop again a ladder procedure similar to the method employed in Section 10.j. then 1 ± 1 Am) is also an eigenket of these same operators but with eigenvalues (m ± l)h and Ah 2 . if I Am) is an eigenvector of 1z and J2 with eigenvalues mh and Ah 2.27)" and (11. we obtain L1+I"1) = (J2 . i. Hence.e.2 Algebraic Approach to the Angular Momentum Eigenvalue Problem 239 Since an operator of the form AA t has only nonnegative expectation values. Application of the raising operator 1+ to the eigenket I Aj) should not lead to any new eigenket. we obtain 1zl+IAm) = (m + 1)h1+IAm) 1zL IAm) = (m . the inequality A ::::: m 2 limits the magnitude of m.6 for the harmonic oscillator. If we act on Eq.. 2. it must be possible for a given value of A or j to reach 1"1') = lA.j' = 2j must be a nonnegative integer. Min(m) A = j'(j' . m decreases by unity. j must be either a nonnegative integer or a halfinteger. For a given value of A. j) from IAj) in a sufficient number of steps descending the ladder by repeated application of the lowering operator 1_. (11. hence.34) follows. it follows that j . for any given A. the only possible values for j are Ij = 0. .28).1) j'. Since the eigenvalues of 1z have both upper and lower bounds. . Hence.
37).m 2 . (11. we have the inequality !i.J.m]h 2 (AmIAm) = we conclude that IC+(Am)1 2 = j(j + 1) .44) and similar inequalities.m(m + 1) = (j .37).42) and J_IAm) = Y(j + m)(j . it is of interest to ask what the physical implications of the noncommutivity of such operators as Jx and Jy are.m)(j + m + 1) hlA m+l) (11. the eigenvalues of J z are mh = jh. J y. Since it is impossible to specify two or more components of J simultaneously. . Since 2j + 1 can be any positive number. J y.Jx !i.54). If the commutation relations (11. Note that from (11. since (AmlIJ+IAm) (AmIJ2 .Jy ~ '2 I (Jz) 1 11.Jx = !i.37) and (11. we see that for every dimension it is possible to construct a vector space that is closed under the operations of the algebra of the three operators J x . show that C_(A m) = C~(A ml) We then have J+IAm) = Y(j .38).1)11" (j .jh These are 2j + 1 in number. This is the key to the idea of an irreducible representation of the rotation group (see Chapter 17).42) and (11. A usual choice is to make the phases equal to zero.240 Chapter 11 Angular Momentum in Quantum Mechanics For a given value of j. construct the matrices representing J x .2)11" .43) Exercise 11. which are constrained by the commutation relations.. With the aid of the identity (11. we see that this can be the case only if the expectation values of all components of J vanish: (J) = 0 . and J z in a basis that consists of the common eigenvectors of J z and J2. and there are thus 2j + 1 orthogonal eigenvectors for every value of j. (Am II = (Am + 11 C~(Am)h Multiplying this and (11. (j . we get (AmlIJ+IA m) = IC+(A m)1 2 h 2 (A m+l!A m+l) Let us assume that all eigenvectors are normalized to unity. Using the fact that J_ is the adjoint of J+.31).44) Is there a state for which all components of J can be simultaneously determined.1)11" .hJzlAm) = [j(j + 1) . . (j . Exercise 11.43). .Jy = !i.11. Then.Jz = 01 From (11.m + 1) hlA ml) (11.m)(j +m+ 1) The phases of C+ are not determined and may be chosen arbitrarily. we can now determine the coefficients C± in (11.24) are applied to the Heisenberg uncertainty relation (10. (11. Jz. By the use of Eqs. such that !i.12.
(Jx)2. A cartoon illustrating the angular momentum eigenstates I em) for e = 2. and J z simultaneously.2. then we must also have and similar conditions must hold for the other components. The treatment of the eigenvalue problem given here has been a formal one. We have thus completed the explicit construction of all the operators J which satisfy the commutation relations (11. For all other states. J y." obtained from this model [zero and (6 . respectively] agree with the expectation values of the corresponding quantum operators. In other words. quantum fluctuations make it impossible to specify Jx. Only the commutation relations. no explicit use was made of the con"L. (11. the Hermitian nature of J.) . hence. but nothing else. .34) is a proper inequality and j(j + 1) > m2 The component J z can never be as "long" as the vector J! In the vector model of angular momentum.2 Algebraic Approach to the Angular Momentum Eigenvalue Problem 241 But (aJxf = (f." the states Ijm) were visualized by circular cones centered on the z axis (Figure 11. the desired state has a sharp nonfluctuating angular momentum value of zero.m 2 )fi 212. ± fi. for all states.45) and therefore J 1ljJ) = O. but its z component assumes the possible values "L/' = 0. This is visualized by supposing that the "L" vector is stochastically distributed on one of the circular cones with uniform probability. except the" state 100). The only solution of (l1A5) is the state 100) == Ii = 0. m = 0).' (in units of Ii) +"L/ Figure 11. or JlljJ) 2 = 0 (11. The semiclassical angular momentum vector "L" has length V e(e + l)fi = V6fi. in the "old quantum theory.2). The mean values of "Lx''' and "L. ± 2fi. and certain implicit assumptions about the existence of eigenvectors were utilized. As a consequence.24). if both aJx and (Jx) vanish. In particular.
z=rcosf) The calculations become transparent if we note that the gradient operator can be written in terms of the unit vectors of spherical polar coordinates as V=r~+4'I_~+9!~ or r sin f) o<p r of) (11. The angles a and f3 are the azimuth and the polar angle of the z' axis in the Cartesian xyz coordinate frame.3. It is convenient to express the orbital angular momentum as a differential operator in terms of spherical polar coordinates defined by x = r sin f) cos <p.46) where (Figure 11. Our solution of the eigenvalue problem thus extends to any three operators that satisfy commutation relations like (11. nor is J necessarily r X p.47) cpy f)z 8= cos f) cos <pi + cos f) sin Z' z /ff. e. y = r sin f) sin <p. They are also the first two Euler angles specifying the orientation of the Cartesian coordinate system x' y' z/ with respect to xyz (see Figure 17.::.24). The Eigenvalue Problem for L z and L 2.3) r = sin f) cos <pi + 4' = sin <pi + cos sin cpy f) sin <py + cos sin f)z (11. and the x' and y' axes are not shown. .1). The third Euler angle 'Y is left unspecified here.242 Chapter 11 Angular Momentum in Quantum Mechanics nection between J and spatial rotations. Angles used in the addition theorem of spherical harmonics. We must now return to orbital angular momentum L = r X p and analyze its eigenvalue problem in more detaiL 3.. The projections of the z' axis and the vector r on the xy plane are dashed lines.jTy x Figure 11.g. the isospin operator in the theory of elementary particles and in nuclear physics.
. The angular momentum may now be expressed as L x = . IThis and other arguments for the singlevaluedness of the wave function are discussed in Merzbacher (1962).49) from (11.) I dq. + L2 = z li 2 1 d [ sin2 () dq. . q. q.49) Exercise 11. or q.46). we demand that the wave function be a singlevalued function of the particle's position. cot 0  dq. (fl  T" L = Y i L z li( dO d  cos (fl cot ()  T" dq.50) are simply <I>(q. we obtain L 2 = L2 x + z:.52) to give us physically acceptable wave functions? It is implicit in the fundamental postulates of quantum mechanics that the wave function for a particle without spin must have a definite value at every point in space.i dq.52) What conditions must we impose on the solutions (11.! Hence. The solutions of (11. d) (11.13. do not commute. d) =!!:.. Verify (11.14.sm i fi(.3 The Eigenvalue Problemfor Lz and L2 243 Exercise 11. The conclusions depend strongly on the topology of the space of the coordinates of the system. From the representations (11. unlike its Cartesian components.?. with m being restricted to integers and halfintegers.  d() d sin q. The spherical coordinate representation is particularly advantageous for treating the eigenvalue problem of L z : li d<l> L z <I>(q.) = e imcp (11.sm d() sin )J (11. () are held fixed.46) it is evident that the three spherical polar components of the momentum operator (fiJi) V.) = "'7 = mli<l>(q.48). (11.)@(O) (11.48) cos q. Derive (11. which is done most easily by using the relation df = dr· Vf and considering displacements along the curves on which two of the three spherical polar coordinates r.51) The simultaneous eigenfunctions of L z and L 2 must then be of the form y«(). From (11.. In particular.) = <I>(q. <I> must take on the same value whether the azimuth of a point is given by q.48).2 2 + 1 () (. + 21T.50) where use has been made of the conclusion of the last section that the eigenvalues of any component of angular momentum must have the form mli.. () d .
+2h.. Use the Cartesian representation L z =!!:.. we get the differential equation _1_ . When we substitute from (11..a sin 0 .49). .. but we will see that they are admissible as eigenvalues of different kinds of angular momentum (spin) and also as eigenvalues of other physical quantities that satisfy the angular momentum commutation relations (e.52) and (11.33) for L 2 now can be formulated explicitly as follows: a 1 a] 1 L2 Y(O.58) 'I and known as Legendre's differential equation. a measurement of Lz can yield as its result only the value 0. The term magnetic quantum number is frequently used for the integer m because of the part this number plays in describing the effect of a uniform magnetic field B on a charged particle moving in a central field. isospin). Equation (11. dg d [(1  g2 ) dF] + dg AF = ° (11..55) g= (11.51).. qJ) to be eigenfunctions or'Lz as well. an integer.~ ® + A® ° sin dO 2 0 = (11..57) assumes an especially simple form.53) shows explicitly that the eigenvalues of L z are mh. the eigenvalue problem (11..e.55) is transformed into cos 0. (11. it must be true that the angular momentum about any axis is quantized and can upon measurement reveal only one of these discrete values. Exercise 11.2 + . (11. ± 1. ±h.51).(sin sin 0 dO By a change of variables 0de) . the condition <I>(qJ + 27T) = <I>(qJ) restricts us to those solutions (11. Its examination follows a conventional pattern. qJ) = h2 [ ~ . Y(O.g. Thus.54) We require the functions Y(O. .57) For the particular case m = 0. qJ) = h 2 AY(0. familiar in many problems of mathematical physics. qJ) sm uaqJ sm 0 ao au 2 (11.56) ~ [(1 dg  e) dF] dg m2 F+AF=O 1 e (11. i. . ±2.(x~y~) i ay ax to show that (x ± iy)m is an eigenfunction of L z • With (11...244 Chapter 11 Angular Momentum in Quantum Mechanics Applied to (11.!!. .15. Since the z axis points in an arbitrarily chosen direction.53) for which m = 0. The halfintegral values of m are unacceptable as eigenvalues of a component of orbital angular momentum.
For the same reason. will be reserved for the total orbital angular momentum.58) yields the recursion relation (k + I)(k + 2)ak+2 + [A  k(k + I)]ak = 0 (11. we exclude the second linearly independent solution of (11. and capital letters S.I) . lower case letters s..61) is.. in agreement with the results of Section 11. According to (11. If the series does not terminate at some finite value of k. and in the odd case (aQ = 0) all odd coefficients are proportional to al' As an eigenvalue of a positive operator.. If there are several particles in a central field. . the ratio ak+2/ak . see Bradbury (1984). these functions are symmetric or antisymmetric with respect to the xy plane. P.(J and z . P. p. 2. 171. and the measured values of L 2 can only be 0. 1. will be used to identify the angular momentum state of each particle.2.. 12h 2 . ak = O.58) which are even or odd functions of g.61) These are called Legendre polynomials. 3 The conventional form of the polynomial solutions of (11.59) Equation (11. . . 473.. we need to look only for solutions of (11.59) is satisfied by the coefficients ab and hence that P e( g) indeed 2por a rigorous treatment. The solution of (11. which are familiar in atomic spectroscopy.? . where f is a nonnegative integer. A must be a nonnegative number.. for (J = 0 and 7T.58). 3See Haken and Wolf (1993). Hence. that is.58) is PeW = 2e f! :.59) shows that in the even case (al = 0) all even coefficients are proportional to aQ.59). D. p. 6h 2 . .2 Such singular functions.? 00..g implies (J .58) that is regular at g = 0 can be expanded in a power series. although solutions of the differential equation for almost all values of g.. F. for f + k = even. For f + k = odd.? .z. this will happen if A has the value A = f(f + I) 2 (11. .e ~ (e + 1)£ (11. . .. 2h 2 .3 The Eigenvalue Problem for L. The orbital angular momentum quantum number .f assumes the values 0. and that all higher powers vanish. The coefficient of (11. The series thus behaves like L (1/k)e for even or odd k. and L 2 245 Equation (11. Substitution into (11. implying that it diverges logarithmically for g = ± 1. p. 3. d.58) does not change its form when gis substituted for g. D.? k/(k + 2) as k . (k 2e . f! 2)(k + I) ( + k) where the last factor is a binomial coefficient.. We verify readily that (11. Since g . easily seen to be _ (_I)(ek)/2(f ak  gk in the expansion of f i<f ) (11.62) + k)(f + k . •••• It is customary to designate the corresponding angular momentum states by the symbols S.60) We have thus rederived the law for the eigenvalues of L .? 7T . are not acceptable eigenfunctions of L 2 • We conclude that the power series must terminate at some finite value of k = f.
using the definition (11. Such a generating function is (1 .61).16.68) with the Legendre polynomials defined by (11.30e PsW = i<63~s . we derive a simple recurrence formula by differentiating (11.66) can also be proved directly.2~s + S2)3/2 = 2: n=O co nPn(~)sn1 (11.elf 2.e !)2 1 dee = (I)f (U)1 I+ (e _ l)f d~ ~ _2_ (2e • . The normalization of these orthogonal polynomials can also be obtained easily by .67) 1 1 Exercise 11.el )2 I+ f • 1 1 [de (~2 d~f 1  1)f ][d~f (g2 de ] l)f ] d~ d~ I+ [d .s)(1 .69) 4For pictorial representations of Legendre polynomials and other orbital angular momentum eigenfunctions.65) No complex conjugation is needed because the Legendre polynomials are real functions.efold integration by parts: I+[Pe<~)]2 d~ = 1 1 ( 1 2e .68) with respect to s: (~ . Prove the orthogonality relation (11. Section 9.2~s + S2)112 = 2: n=O co Pn(~)sn (lsi < 1) (11. The orthogonality relation (11. .ly (e . it is helpful to introduce a generating function for Legendre polynomials.64) Since Pe<cos 8) is an eigenfunction of the Hermitian operator L 2 . Only the integration over the polar angle 8 concerns us herenot the entire volume integraland we expect that L 1T Pe<cos 8)Pf' (cos (J) sin (J d (J = 0 if.61) has been adopted because it gives Pe<± 1) = (± 1)f (11.66) directly. see Brandt and Dahmen (1985). it is clear from the general theorems of Chapter 10 that the Legendre polynomials must be orthogonal. using the As usual in the study of special functions.68) To prove the identity of the coefficients Pn(~) in (11.61).63) The first few Legendre polynomials are PO(~) = 4 1 1) P3(~) == l<5~3 . .70e + 3) + 15~) (11. The peculiar constant factor in (11.3~) P4(~) P 1 (g) P 2W =~ = !(3e  = i<35~4 .l)e (e (2 .61) and successive integrations by parts.58).246 Chapter 11 Angular Momentum in Quantum Mechanics solves (11.e (11. definition (11.e + 1 2f = (I)f 1 (11.2.e' =1= .
61) satisfies the recurrence formula (11.g.74) We leave the proof to the interested reader. Legendre functions with different values of m are generally not orthogonal. (11.68) and (11. polynomials is completed by the demonstration that P n (g) as defined by (11. we see that in agreement with (11.71) by defining the associated Legendre functions for m = 0 as e (11..69). The associated Legendre functions are orthogonal in the sense that (11.1(g) (11. Ii is natural to supplement the definition (11.57) with m =1= O. e! ( ~ ~ dgm ~ 1:~)ml2 d(+m (1:2 _ 1)( dge+m!:> (11.3 The Eigenvalue Problem for L z and L 2 247 or. Prove the recurrence relation (11.. Q. (g  s) 2: n=O ~ Pn(g)sn = (1 .70) By substituting s'= 0 in (11.m)! (11.(g) + e(e + I)P.71) are defined for positive integers m ~ e.. If Legendre's equation (11.61).58) is differentiated m times and if the associated Legendre functions 1":'(1:) = (1 _ 1:2)ml2 dmpt<g) = _1_ (1 _ 2( .17.68). by the use of (11.57) for A = e(e + 1).64). P'.75) .npn.34) to verify that the magnetic quantum number cannot exceed the orbital angular momentum quantum number. When g is changed to .2gs + S2) 2: n=O ~ npn (g)sn1 Equating the coefficients of each power of s.1 :2e P.18.72) which is identical with (11. These functions are also called associated Legendre functions of the first kind to distinguish them from the second kind.58). Exercise 11. The equivalence of the two definitions of the Legendre.73) Note that in this relation the two superscripts m are the same. The associated Legendre functions with m ~ are the only nonsingular and physically acceptable solutions of (11.70). Use the inequality (11. For purposes of normalization. depending on whether + m is an even or odd integer. we deduce that :g e [(1  e) d~ig)] .(g) = 0 (11. it is not difficult to obtain the physically acceptable solutions of (11.(g). Exercise 11. we obtain (n + l)Pn+1(g) = (2n + l)gPn(g) . we note that ' J +1 m 2 2 (e I [p(g)] dg = 2e + 1 (e + m)! .70) for Pn(g) defined in Having solved (11.( g) merely t:etains or changes its sign. which is the singular variety.57).
l)f e .5 e32'77' SIn 2 () = f{!.' is the only admissible solution of this equation. y.«().76) The first few associated Legendre functions are P~(g) = ~.. {l5 e±icp cos () sin () = Y ~S.'«().55) is unchanged if m is replaced by m.' (3 ~ cos () ..'(cos ()). cp) as the separable solutions (11. Exercise 11.82) x 2  yg = {If.m '" (11. it follows that the same associated Legendre function must be used for a given absolute value of m.5 (x 32'77' .m)! (_I)m e imCPnm( (e + m)! r e cos ()) (11._ (e + m)'. For m 2:: 0.1) = ~u.. Spherical Harmonics. e±icp sin () 2 jl.... Legendre functions: Justify the following alternative definition of associated df m ( l)m pm(C) = _ . fl5 (x ± iy)z r2 ± iy)2 r2 +2 +2. (1 _ C2)ml2 __ (C2 .79) The spherical harmonics are normalized simultaneous eigenfunctions of L z and L 2 such that L ym z f = ft aym i acp _f = mftym e (11. It is convenient to define the spherical harmonics Y'. are products of e imcp and P'.r 1 y? = ["I cos () = Yfl IT ~ = :+ = :+ jl.54). e '" 2 '" dg f . ~4. (11. x ~ iy 2  (11. Since (11.19..78) Spherical harmonics with negative superscripts (subject to the restriction ::5 m ::5 will be defined by e e) (11.. and since P'.80) L 2 y.m)! .. = e(e + l)ft 2 y. P~(g) = 3~. cp) = 2e 4'77' + 1 (e .52) and (11..248 Chapter 11 Angular Momentum in Quantum Mechanics Returning now to (11. 4.54)... e! (e . which are separable in spherical polar coordinates. :+ {5 2z r2 l Yi l = :+ ~S..81) The first few spherical harmonics are listed below: yO _ 0 V41T ~4.'~' 2: Y = f{!.52) that are normalized with respect to an integration over the entire solid angle. we see that the solutions of (11.
. which is realized by the transformation (j) + (j) + 1T and () + 1T .. (j)) sin () d() d(j) = 8u• 8mm .lJ.85) and it follows that all eigenfunctions of L z and L 2 must have definite parity.1 to the sequence of monomials 1. since both rand p change sign under reflection.83) We have (11. since yr e 2'7T1'7T 1o 0 [Yr«().. Hence.2). ~. requiring that there be a polynomial of every degree and that they be orthogonal in the interval 1 ::5 ~::5 + 1. Hence. The spherical harmonics are thus eigenfunctions of the parity operator Up which changes r into . Since is obtained from r € by repeated application of the raising operator L+ = Lx + iLy and since Up commutes with L+. any wave function that depends on the angles () and (j) can be expanded in the mean in terms of spherical harmonics (see Section 4. for the associated Legendre functions of fixed order m. . (a) The eigenfunctions e imcp of L z are complete in the sense of Fourier series in the range 0 ::5 (j) ::5 21T. though less obviously. Roughly. the azimuthal wave function eimcp is multiplied by (_l)m. Hence. The operator L = r X p is invariant under reflection of all three Cartesian coordinates. L] = 0 (11. when r is changed to r.86) Although no detailed proof will be given here. [Up. or an axial vector. (b) The Legendre polynomials Po(~)..4 Spherical Harmonics 249 Under a coordinate reflection.e. all orbital angular momentum eigenfunctions with the same value of e must have the same parity. and any function that can be expanded in a uniformly converging power series of ~ can also be expanded in terms of Legendre polynomials. e. e Hence. (j)) is multiplied by (1)€. Y. which also form a complete set as varies from 0 to 00.. it is important to note that the spherical harmonics do form a complete set for the expansion of wave functions. The spherical harmonics form an orthonormal set.' has definite parity in consonance with the parity (evenness or oddness) of the angular momentum quantum number e. . a very large class of functions of (j) can be expanded in terms of them. it is an antisymmetric tensor of rank 2. This result is compatible with the reflection properties of orbital angular momentum.'«(). (j))]*Y~'«(). . Pl(~)' P2(~)' . The same is true. Y. are the orthogonal polynomials that are obtained by applying the orthogonalization procedure described in Section 4. Hence. this can be seen from the following facts.r: Uplf!(r) = If!(r) (11. through the origin. or inversion..84) i. and P7(cos lJ) by (_I)e+m. • (11. ~k can be expressed in terms of Legendre polynomials.
94) are merely realizations of (11. The operators L+ and L_ are raising and lowering operators for the magnetic quantum number.48) may be written as L + = fie"l'  L = . + m)(e .m + + I)Y7+ I (O.(a a fieleI' ao .42) and (11.m) y m (2e + 1)(2e .m + 1) pml + ~ (11. cos 0 Y7(O.250 Chapter 11 Angular Momentum in Quantum Mechanics Exercise 11. and Eqs. The addition theorem for spherical harmonics is a useful applicatign.1) yml (2e + 1)(2e .m .2. : ~~~~e++m3)+ 2) Y7:l (e .94) g= cos 0 and the definition (11.20.93) (11.76) dP7 dg With = 1 P':'+I _ ~e !!!L P':' = (e e le e !!!L pm e le (11. To determine aYlaO we note that from the definitions (11.i cot 0 acp ).1) m+1 (2e + 1)(2e .91) (11.78).90) which. according to (11.(a + i cot 0acp a) ao .43). Consider two coordinate systems xyz and x I y I Z I.88) sin 0 ei<p Y7(O. which were derived directly from the commutation relations. cp) I These equations do not come as a surprise.(O.m + l)(2e + 2) y m  I 3) e+1 m)(e + m .. : ~~~~e+m )+ 3 1) Y7+1 + sin 0 ei<p Y7(O.1) eI (11. cp) = (e +(.'(O. (11. cp) = fiy(e fiy(e + m)(e + m LY'!. cp) l)Y7. cp) = (e .87) (e +(2.m)(e .71) and (11. above.1) eI + (11.80). it is then easy to derive the relations L+Y7(O.89) The effect of the operators Lx and Lyon Y7 is conveniently studied by employing the operators (11.1) Ye.e + m)(e .I (11. The complete agreement between these two sets of equations shows that the choice of phase factor made for the spherical harmonics is consistent with the choice of phases for C+(Am) in Section 11.92) The effect of alacp on Y7 is known from (11. Construct P2(g) by the orthogonalization procedure described Some of the most frequently used expansions of angular functions in terms of spherical harmonics may be quoted without proof.m (2e (e + + 1)(e . The addition theorem is the formula express .93) and (11. cp) = m)(e . cp) = (.
93). Since P( is an eigenfunction of L 2 .94) may be used to evaluate Lz'Picos ()').99) it follows that Co 2e + 41T 1 With (11. ~) (11.87). All these requirements can be satisfied only if Plcos 9') = 2: m=e e Cm Yi"m(f3. Hence. ex is equivalent to the transformation ()' ~ . 'P) = fil 41T = e+ 1 Plcos 9) (11. Since. Cm = (_l)m co .96) Since Lz' = sin 13 cos ex Lx 1 _ = sin 13 e.89). ~.4 Spherical Harmonics 251 ing the eigenfunction Picos e') of angular momentum about the z' axis in terms of the eigenfunctions Y~( e. The position vector r has angular coordinates e.~) = ~~ 8mO g[D (11. (11. ex).In a rigid rotation of the figure about the z axis. p(cos 9') can also be expanded in terms of Y~(f3. ex and 'P change by equal amounts. and only Co need be determined.100) .98) and Y~«(). after some calculation. The direction of the z' axis in space is specified by its polar angle 13 and its azimuth ex with respect to the unprimed system. we obtain.ex. For this purpose we specialize to 13 = 0.79). An interchange of e. and e'. or 9 = ()'.()' and must leave the expansion unchanged. ~) of L z ' Figure 11. (11.88). only spherical harmonics with the same subscript e can appear in the expansion. If the linear independence of the spherical harmonics is invoked.97) Eqs. this proves the addition theorem in the form (11. Plcos (J') must be a function of ~ . (11. (11. and (11. Hence. the simple result Thus. 'P.95) The coefficients Cm can be determined by using the condition Lz'Pe(cos 9') = 0 (11. and 9' remains constant.. because Picos ()') is an even function of (J'.3 indicates the various angles.aL_ + cos 13 Lz (11.aL+ + 2 sin 13 sin ex Ly + cos 13 Lz 1 _ + 2 sin 13 e. and 13. Y~(O. from the definitions of ~ and P~. ex)Y~(e. ~' in the two coordinate systems.
Angular Momentum and Kinetic Energy. beca!1se it implies that n is a variable of integration. Since the kinetic energy is represented in the coordinate (or momentum) representations by an operator proportional to V2 .n'). r r A') r e=o . 5.. we infer the further identity: " ~ U + 1 u(r . 8(r . There is no such solid angle variable. a)]*y. but this is misleading. = 8(r . cp.. is expressed by the closure relation: e~o 2: m=e [Y.104) 41T (11. cp) = 2: e 8(1'. 132 (odd permutation of 123) when two or more indices are equal Using the summation convention. it is expedient to relate L 2 to the Laplacian.r') r 2 8(r.103) is obtained. We make the calculation using the concise notation that takes advantage of the summation convention (summing from 1 to 3 over repeated indices) and of the LeviCivita asymmetric (thirdrank tensor) symbol Bijk' The LeviCivita symbol is defined as follows: ijk = 123. a) coincide. It has the property (11. 1") is sometimes written as 8(n .P e(A .100). the identity e=o 2: (U + l)PeCr .s If (11. 231. . 213. The delta function in three dimensions has a representation in spherical polar coordinates. or of the direction of the vector r.. Exercise 11.r ) Hence. we formulate a simple identity.101) is combined with the addition theorem (11.105) This formula will be useful in the theory of scattering from a spherical potential.105) by integrating both sides over all of3space.(e. Check (11.21. (11. cp) and r'(J3.102) for any function fer) of the spatial direction specified bye. 1") (11. 312 (even permutation of 123) ijk = 321. 1") = 41T8(r.r ') = 8(r 2 r') LJ .101) The solidangle delta function on the righthand side is equal to zero unless the two vectors r(e. r ) A A' (11.'(J3.252 Chapter 11 Angular Momentum in Quantum Mechanics The completeness of the spherical harmonics as basis functions for any function of the angles e. cp.106) 5The delta function 8(1'. 1") (11.
(11. hence.r· p = 3ih1 we obtain (11. representationindependent. r'p=ri Ii a or . in the (spher kal) coordinate representation.108). (r X p) = Since = (r X p) .108) Exercise 11. Prove the equality (11. L 2 = (r X p) .109 The component of the gradient Vf in the direction of r is aflar. method for deriving (11.23. Line by line.107) It follows that and we arrive at the important identity (11.108) starts with the operator identity.p2 r] and p . An alternative.Angular Momentum and Kinetic Energy 253 Exercise 11. [p X (p X r)] = r· [p(p • r) .106). It is evident that the components of orbital angular momentum can be written n the form (11. work through the steps leading to Eq. r . (p X r) r .22.
we consider the eigenvalue problem for the last term in (11. With nonnegative integer powers r. r 2 df(r) dr dr This has the solution fer) = Are = Cf(r) (11.113) and the eigenvalue C 2 = e(e + I) (11. s.1 (11. the kinetic energy operator is related to angular momentum by (lUll) Here we see explicitly that L commutes with T. Exercise 11.112) + Br. and by choosing the solution (11..((}. t..254 and consequently Chapter 11 Angular Momentum in Quantum Mechanics in agreement with (11. when converted into functions of x. that (11. are homogeneous multinomials of degree e..81). Since L and therefore also L 2 commutes with any function of r. the functions rey. and not on r. cp). Show that the homogeneous multinomial F = r+s+t=f 2: a rs .((}.e.78) of spherical harmonics and inspection of the formula (11....114) Since L acts only on the variables () and cp. subject to the constraint r + s + t = e and the requirement that they must be solutions of Laplace's equation. these functions can be expressed as 'La rst xrysz'.108) and (11.108): .108). cp) are regular solutions of Laplace's equation. In order to establish the connection between the eigenfunctions of L 2 and the solutions of Laplace's equation. xryz' has (e + 1)( e + 2)/2 coefficients and that the linear relations between them imposed by the requirement V 2 F = 0 leave 2e + I coefficients to be chosen arbitrarily. which accounts for their designation as harmonic functions.. y.24. From the definition (11.(so .113) which is not singular at the origin.115) Thus. we see from (11.71) for associated Legendre functions it follows that the functions rey.!!. since it is patently irrelevant whether differentiation with respect to r is performed before or after a rotation about the origin. z.
Problems 1.(r . 4. Classically.2(x + y)z (a) Determine the normalization constant N as a function of the parameter a. 2.(r . and show that it transforms correctly. For a finite rotation by an angle a about the z axis. we are now prepared to tackle the centralforce problem in quantum mechanics. Show that in Ddimensional Euclidean space the result of Problem 4 generalizes to 1[1 Pr = .Ol ." = Ne.1  .. p) + (p . p) r + (p. apply the rotation operator UR to the function f(r) = ax + by. e equals Aided by the formula connecting the kinetic energy with angular momentum. 3. p). and 312. r) r IJ and that this gives the correct Schrodinger equation with the Hermitian operator Pr = ~ (~+!) ar r 1 whereas (hli)(a/ar) is not Hermitian. 5. For the state represented by the wave function .Problems 255 that the number of linearly independent harmonic multinomials of degree the number of orthogonal spherical harmonics of order e. (b) Calculate the expectation values of Land L 2 • (c) Calculat~ the variances of these quantities. 1. we have for central forces H = 2~ + 2mr + V(r) where Pr = (lfr)(r . Show that for translation into quantum mechanics we must write Pr =  p2 L2 1[1 2 . r)  2 r r IJ h(a + 2rDl) ar = . Explicitly work out the J matrices for j = 112.
and thus dependent only on the distance r from a center of force. L] = 0 (12. and we can therefore require the energy eigenfunctions also to be eigenfunctions of L z and L 2 • These eigenfunctions must then be of the separable form ifJ(r) = RE(r)Y'.3) when spherical polar coordinates are used.CHAPTER 12 Spherically Symmetric Potentials If the potential energy is rotationally invariant. 1.4) 256 . <p) (12. and the attractive Coulomb potential (the oneelectron atom). The equation that is satisfied by the radial factor R(r) is found if we express the Hamiltonian in terms of orbital angular momentum. This constant of the motion enables us to reduce the threedimensional Schrodinger equation to an ordinary differential equation. As examples of central potentials. The operators H. the radial equation. r X p] = [H. we solve the radial SchrOdinger equation for the trivial case of a free particle (V = 0). we may use (11.1) for a particle of mass m moving in a centralforce field commutes with the orbital angular momentum operator.2) angular momentum is a constant of the motion for a particle moving in a rotationally invariant potential. Since the Hamiltonian (12. the spherical square well. chosen as the coordinate origin. orbital angular momentum is conserved.'(fJ. provided that angular momentum conservation is used and the inertial centrifugal force introduced. L z and L 2 all commute with each other. in this case. Reduction of the CentralForce Problem. Since L and therefore also L 2 commutes with any function of r. analogous to the reduction of a centralforce problem in classical mechanics to a dynamical problem for the radial coordinate r alone.111) to write the SchrOdinger equation HifJ == [T + V(r)]ifJ = EifJ for central forces in the form (12. [H.
The energy eigensolutions of Eq.6) we find that u(r) obeys the radial equation (12. u(r) must vanish at r = 0.2] R(r) = ER(r) 2  dr 2m? (12. if l/J is to be finite everywhere. carry through the explicit calculation of V2 in terms of spherical polar coordinates and derive (12. but we emphasize the physical meaning of the method.2 V2l/J = E(~O). Exercise 12.6) From (12. 2. This term is sometimes called the centrifugal potential. (12.8) was treated quite naturally by the method of separation of variables using Cartesian coordinates.9) .7) is similar to the onedimensional SchrOdinger equation.2 The Free Particle as a CentralForce Problem 257 If the separated form (12. If you have never done it before. with energy fl. Our procedure is entirely equiv~lent to the familiar separation of variables of the Laplacian operator in spherical polar coordinates. Although (12. A detailed discussion of these questions requires specific assumptions about the shape of the potential energy.6). For V = 0. (See Appendix. for a general formula for the Laplacian in curvilinear coordinates.7) This equation is identical in form with the onedimensional SchrOdinger equation except for the addition of the term fl. Section 3.4 the SchrOdinger equation for a free particle (V = 0).2mr2 dr fl.4) by comparison with (11.2 d(r d) + e(e + l)fI. this equation can be reduced to the ordinary differential equation for the radial eigenfunction RE(r): fl.2 [ . In Section 4. For instance. 2 e(e + 1)12mr2 to the potential energy. since these coordinates are particularly well suited for describing translations in Euclidean space.3) is substituted. the boundary conditions to be imposed on the solutions are quite different.3). since it represents the potential whose negative gradient is the centrifugal force.49). according to the definition (12. Nevertheless.2 + VCr) ] RE(r) = ERE(r) 2  dr 2mr 2 (12. 2m El/J (12. the function R(r) (omitting the subscript E for brevity) must satisfy the radial equation [ .2 2mr d r d(r d) + e(e + l)fI.5) which is easier to solve than the original partial differential equation.1.) It is sometimes convenient to introduce yet another radial wave function by the substitution u(r) = r R(r) (12.8) can then be assumed to be separable in spherical polar coordinates in the form (12. since r is never negative. The Free Particle as a CentralForce Problem. and in this chapter the radial Schrodinger equation will be solved for several particular cases. it is useful also to look at the free particle problem as a special case of a potential that depends only on the radial coordinate r.
1 l f+1 1 e (1 . is used in the last equations to emphasize that these formulas are valid for all complex values of the variable).12) is easily verified if their integral representation liz)  .(. Chapter 12 Spherically Symmetric Potentials _. (12. By (12.17) We note that the spherical Bessel functions are even or odd functions of their argument.14) The regular solutions of this equation.13) 2 + .!. e! t 1)! z + t+2 O(z ) (12.!!!.12) This differential equation is seen to be related to Bessel's equation if we make the transformation R = J(p) vP and obtain d J dp 2 (12. _ 2 It(z) .15) and known as spherical Bessel functions.11) the radial equation reduces to (12. dJ pdp + [1 _(e +p21/2?]J = 0 (12.s) ds izs 2 t (12. That the latter satisfy Eq.13) they are related to the regular solutions of Eq.16) is used.12). (The variable z.12) and that the radial eigenfunction of the SchrOdinger equation (12.10) If we scale the radial coordinate by introducing the dimensionless variable = J2mE = yr kr. r2 dr (r p dR) dr + e(e + 2 r 1) R = k2R (12.258 or. depending on the parity of e. The first term in the series expansion of (12. hk = Y2m E. which do not have a singularity at p = 0. _ 2t+1 • . We thus see that the spherical Bessel functionjikr) is the regular solution of the radial equation (12. which are defined as (12.16) in powers of z is . (12.10) for the free particle is R(r) = Cit(kr) (12. rather than p. are the Bessel functions J t + 1dp). (12.18) . since.(2e t + .
Since 0/1 and 0/2 correspond to the same energy.20) All other solutions of (12. (12.17) for je to the transformation (12.2 The Free Particle as a CentralForce Problem 259 A useful formula linking Bessel functions with Legendre polynomials is obtained by integrating (12. being asymptotically out of phase by 7T/2 compared to the regular solutions jeCp). is sometimes distinguished as "the" irregular solution.Jp neCp) ] dne(p) djeCp) Je(p) = constant (12. This leads to jeCz) = 2.n_/2.24) The function neCz) is known as the spherical Neumann function.23) This expression is the analogue of the Wronskian for the onedimensional Schrodinger equation.] p (p » e) (12. e'zs PeCs) ds (12.12) is obtained from je by noting that the differential equation is invariant under the substitution e ~ e  1 (12..::.. we obtain a linearly independent solution that can..22) This particular singular solution of the radial equation.20) for jeCp) satisfies the differential equation (12. it is merely necessary to subject the expression (12.p_. (3.12) are singular at the origin and not admissible as energy eigenfunctions for the free particle. e! z eI + O( e+l) z (12.(e Je p = p + 1)7T/2] (p » e) (12. ( ) ~ cos[p . Its coefficient is most easily computed by applying Eq. although any linear combination ofje and n e is also singular at the origin. Using the approximation (12..23) for z=O ne( z) = 2e .. its value must also be unity as p ~ O. of the Schrodinger equation. A particularly simple singular solution of Eq.7) by application of Gauss' divergence theorem that p 2[.23) shows that the constant has value unity for large p. we = 0/1 jeCp)Y'.16) by parts e times and using the definition (11. Exercise 12. . and the leading term is . Substitution of (12.12) to second order in pI. We see that the singularity n e at the origin is contained in the leading term which is proportional to Z eI.19) The asymptotic form of the spherical Bessel functions can be derived from this expression by further integration by parts.61). for large positive p. be written in the form ne(p ) = ~ _si_n[::!. we derive from (12.21).e I 1 J+I I . and infer from (3.22) into (12. Hence.7) to the two solutions 0/2 = neCp)Y'.20) and (12. To exhibit the behavior of n e near the origin.(e_+'1):.::.2.20).21) If this transformation is applied to the asymptotic form (12.17) for je near the origin. dP . (2e)1.. Verify that the asymptotic expression (12.
31) (12. as spherical cylinder functions. Check this for the free particle as r ~ 00 and r ~ O.12) are known.30) cos Z cos z (12. 1 and 2.e = 0.29) h~Z)(p) =. are given below: lo(z) = . 2e . Hankel.5).26) and h~Z)(z) = leCz)  (12. The information we compile in this section about these special functions will be used in Section 12.12) are the spherical Hankel functions of the first and second kind. of the radial equation (12.!) Z3 Z sin z  l (12.e + l)7T/2]) The explicit forms of the spherical Bessel.3 and in Chapter 13 on scattering.exp {i[p P (.e + l)7T/2]) (12. defined by h~I)(Z) = leCz) + ineCz) ineCz) (12.exp{ i[p P 1 z ll(z) (.32) (12. Two other useful singular solutions of Eq. (12. +_l__ .20) and (12. and Neumann functions for.Z)( ) = e Z '(2e) . . Like the Neumann function.27) The generic name for the solutions of Bessel's equation (12. Show that for any two solutions. paradoxically.14) is cylinderfunctions.e! z eI + O(I.22) to be h~I)(p) 1 =. R 1 and R z. sin z =  sin z ZZ ZZ z cos z Jz(z) = no(Z) = (l . the Hankel functions diverge as zeI near the origin: h(I.33) .Z)( e+l) z (12.28) and their asymptotic behavior for large positive p is seen from (12.260 Chapter 12 Spherically Symmetric Potentials Exercise 12.3. the condition r z( R 1  dR dr z   dR 1 R ) z dr = constant (12. and the solutions of (12.25) holds.
these two sets of eigenfunctions are equivalent.e and m.:(O. The regular radial eigenfunctions of the SchrOdinger equation for V = 0 constitute a complete set.:(O. Verify that for E < 0 none of the solutions of the free particle radial equation are physically acceptable owing to their asymptotic behavior. q. while the number of spherical waves is denumerable.Mkr)y'. it is sufficient to specialize this relation to the case where k points along the z axis and consider the expansion e ikz = eikrcos9 = 2: f=O 00 afjikr)Pf(cos 0) From the orthogonality and normalization properties of Legendre polynomials we obtain (with g = cos 0) 2.2 The Free Particle as a CentralForce Problem 261 Exercise 12.12) is an example. r) 'Morse and Feshbach (1953). (12. r) = 2lkr ~ f=O 2: OQ + l)Pik.r and the spherical wavesjeCkr)Y'. q.e + 1 af jikr) = 2 f+1 e . Hence. A useful asymptotic approximation to Eq. and one kind must be capable of expansion in terms of the other. r)  ~ 2lkr  ikr f=O 2: co (12.(f kr + l)1T/2] Pik.36) i fMkr)[Y'. q. p. we have before us two alternative complete sets of eigenfunctions of the free particle Hamiltonian.35) This formula is especially useful in scattering theory.) Actually. r) 47T f=O m=f 2: 2: (12. .4. posing the problem of determining the coefficients in the expansion 00 e ik. 1 1krg PeCg) dg (12.35) by use of the addition theorem for spherical harmonics: e ik. as a consequence of a fundamental theorem concerning SturmLiouville differential equations. but for a given value of the energy.r =f f=O ikr (U + lW (2f cos[kr .20) on the righthand side.37) (2f + 1)(l)fPeCk .m. Nevertheless.:(O. corresponding to the countability of the integer quantum numbers . Both sets exhibit an infinite degree of degeneracy.34) which we compare with (12. For kr » I we get e ik.19) to establish the identity 00 e ikz = f=O 2: (2f + l)ifjf(kr)PeCcos ()) (12.).e 00 + +f lWjeCkr)Pf (k .r = f=O m=f 2: 2: f cf. They are the plane waves e ik .:({3. is obtained from (12. 738.) where a and {3 denote the spherical polar coordinates of the vector k. a)]*Y'. where hk =Y2m E.! of which (12. The more general expansion. the number of plane waves is indenumerable. with k pointing in an arbitrary direction.r = = 2: f=O 00 (2.36) is derived by substituting (12.
3.V = E + Vo. (12.f(.. The condition of regularity at the origin again restricts us to the spherical Bessel function for the solution inside the well. the logarithmic derivative.2 apply provided that we take into account the fact that E must be replaced by the kinetic energy. owing to its finite range and finite depth.r == .262 Chapter 12 Spherically Symmetric Potentials which by use of Eq.42) The interior and exterior solutions must be matched at r = a. Hence.r contributes only in the forward and backward directions. which is a physically reasonable result. Unlike the Coulomb potential. . The Spherical Square Well Potential. If we define k iJ2mE iK = /i. but k is now an imaginary number.38) a very convenient formula in scattering theory...r) ' (12. A square well is a central potential composed of two constant pieces: V = . In this section. Since E < 0 for bound states.2 that only the Hankel function of the first kind decreases exponentially.e is d( 2 dR) /i.5). The radial wave equation for a state of angular momentum .Vo for r < a and V = 0 for r > a (with Vo > 0).40) that would increase exponentially at large distances. The spherically symmetric square well in three dimensions is of interest because it is mathematically straightforward and approximates a number of real physical situations. It is seen that the leading term in the asymptotic expansion of the plane wave e ik . E .2 + Vo) ) r for r< a (12. possesses only a finite number of such levels.39) ~!!. (r 2 dR) + /i.2 = it is easily verified from the asymptotic forms in Section 12. (11. and subject to a force only at the discontinuity at r = a.2mr2 dr r dr + 2mr2 /i. . (1/R) dR/dr or (lIu) du/dr. must be continuous. the emphasis will be on the bound states of a particle in such a potential..!!!. which gives rise to infinitely many discrete energy levels for bound states.103) reduces to e ik. the radial wave function and its derivative are required to be continuous at the discontinuity of the potential.40) has the same solutions as (12.41) Outside the well we must exclude any solution of (12./j(k 2 ikr ikr' r)  . The particle is free inside and outside the well. The eigenfunction outside the well must thus be of the form R(r) = Bh~l)(i J~:E r) for r> 0 (12.2 2 (E + Vo)R for r> a r < a (12. the square well.10). For bound states Vo ~ E ~ O.ikr 2 ikr /j(k .e(f + 1) R = . In conformity with the analogous onedimensional problem (see Section 3. Thus R(r) = Aje ( 2m(E /i. All the results of Section 12..!!!!.f + 1) R= 2 2mr~ dr dr 2mr2 ER for ( 240) 1 . .
41) and (12. p.43) where a 2 = 2m VoIfi 2 The solutions for positive E are asymptotically oscillatory and correspond to scattering states in which the particle can go to infinity with a finite kinetic energy.6.45) The general principles of quantum mechanics require that the eigenfunctions (12. and for odd e we have states of odd parity. If H is the sum of a Hermitian operator H o and a positive definite perturbation V.7.5. We now return to a general discussion of the radial equation for central forces. the eigenfunctions must be normalized in the sense of (4.33).E') (12.o 8m 4. 8. They will be studied in Chapter 13. The radial wave function u(r) must satisfy the equation 2 2 fI? . prove by a variational argument that the ground state energy of H o lies below the ground state energy of H. Compare the energy eigenvalues for S states in the threedimensional square well with the energy eigenvalues of a onedimensional square well of the same depth and width. From Section 12. <p) = Y'. Exercise 12.1 we know that the solutions of the Schrodinger equation can be constructed as u(r) r/J(r. = L~ u*u dr = 1 (12.(l)(Ka) K2 a) h~l)(Ka) (12.44) be normalizable.42) yields an equation for the allowed discrete energy eigenvalues: 2 Ja? 2 K K 2 j. The Radial Equation and the Boundary Conditions.(8.44) Since r does not change under reflection. for even e we have states of even parity. Hence. . these wave functions have the same parity as Y~. applied to (12. the eigenfunctions corresponding to discrete eigenvalues must satisfy the condition ff r/J*r/J r 2 dr dO. Show that a spherical square well has no bound state unless fi2 7T 2 V.47) 2S chiff (1968). or fo~ UE*UE' dr = 8(E . <p) r (12. Exercise 12.u + [fi e(e + d 2 2 1) 2m dr 2mr + V(r) u = Eu ] (12. 86. a2 > .46) If E lies in the continuous part of the spectrum.4 The Radial Equation and the Boundary Conditions 263 This condition. Apply this theorem to prove that in a central potential the ground state of a bound particle is an S state. Since the spherical harmonics are normalized to unity.CVa? je<Ya?  2 K a) __ h. gives useful recurrence relations for spherical cylinder functions and their derivatives. Exercise 12.
+ V(r) a) 2mr L ar 2 2  2 we find. by 3 V(r) = era (12. by integrating by parts. since division by r is involved. This is the condition which consistency of the probability interpretation of quantum mechanics imposes on the eigenfunctions of H. operator must be selfadjoint in the sense of (4. which for r ~ 0 increases as 1/r 2 . Thus.1. .51) for states with O. pp.50) eliminates the second solution.. Applying this requirement to the operator '* H h = 2 2mr 2 a(r ar + .48) with a an integer and a ~ .50) In most cases. Without going into detail. in the relativistic theory of the hydrogen atom.9). O (* dU2 Ul  dr ~ dui dr ) . for any but S states. we note that the appropriate boundary condition is obtained by demanding that the Hamiltonian.49) In applications.48) at small r.45) must have the form (12. are examples of this. including the square well and the Coulomb potential. Furthermore. For the same reason (12.g.53) e '* 'Por potentials that are more singular at the origin..111) is not arepresentation of the kinetic energy at the coordinate origin. 16651667.45) reduces near r = 0 to (12. that any two physically admissible eigensolutions of (12.49)... B = O..0  (12.45) must satisfy the condition · 1 1m r.51) is (12. The general solution of (12. this boundary condition singles out correctly the set of eigenfunctions that pass the test (12. the boundary condition (12. We must not forget that. but mildly singular wave functions are occasionally encountered (e. (11.. we assume that V ~ 0 as r ~ 00. (12.45) is valid only for r 0 and must be supplemented by a boundary condition at r = O.52) Since e ~ 1. see Morse and Peshbach (1953). or energy. Potentials of the form (12. this condition usually may be replaced by the much simpler one requiring that u(r) vanish at the origin: u(O) =0 (12. hence.35). If in the immediate vicinity of the origin V can be neglected in comparison with the centrifugal term.49) or (12. u(r) must be proportional to r e+ 1 at the origin and l/J must behave as reo Hence a power series solution of (12. Section 24.264 Chapter 12 Spherically Symmetric Potentials Most situations of practical interest are covered if we assume that V(r) is finite everywhere except possibly at the origin and that near r = 0 it can be represented .
V(r) = r (12.45) is of the form 'u e = 0.2 u (12. the terms in (12.. For bound states. states.8. the radial equation (12. If E < 0. Even then the form (12. or bound.) for small values of r and infer that for S states again we must require that B = O. and a separate investigation is required to obtain the behavior of the wave function near the origin. The boundary conditions will in general allow only certain discrete energy eigenvalues.55) possesses oscillatory solutions for positive E and exponential solutions for negative E.47). 5. we introduce a new function w(p) by setting I u(p) 2 = p€+l e.45) reduc~s to d2u 2m E _ 0 dr2 + fI. we seek those that satisfy the boundary condition at infinity and at the origin.57) Substituting this expression into (12. ) + B(1 .55) as r ~ 00. with the increasing exponential excluded by the condition that'" must be normalizable in the sense of (12. we obtain d w dp 2 + 2(e + P 1_1) dw dp + [~ _ E 2(e l)]w = 0 (12. Exercise 12.45).56) Sometimes it is also convenient to remove from the unknown dependent variable the portions that describe its behavior at r = 0 and r = 00. the radial equation is conveniently transformed by the introduction of the dimensionless variable p = J2mE r = Kr ~ (12.. For a potential V = Clr and angular momentum that the general solution of (12.53) remains applicable for S states in most cases.5 The Coulomb Potential 265 If e = 0 (S states).Clog r +. The Coulomb Potential. Equation (12. Thus.59) .p w(p) I + P (12. the eigenfunctions have the asymptotic behavior u(r) == exp ( _ V~m E r) representing spatially confined.. Assuming that the potential energy vanishes at great distances.45) containing V and E cannot be neglected.46) or (12.54) = A(r  ~r 2 + . Let us now suppose that V is the potential energy of the Coulomb attraction between a fixed charge Ze and a moving particle of charge e. show (12.58) Of the solutions of this equation.' .
Assume. the series (12.2(f + l)]ak = ° or ak+l ak (k This recursion relation shows that. it follows from (12.:::: 0.. The result is.66) . The energy continuum (E > 0) of a particle in a Coulomb potential is the subject of Section 13.. the ratio ak+/ak is always greater than the corresponding ratio of coefficients in the power series expansion of e(2lIClp...52).62) A simple twoterm recursion relation is found if we expand w(p) in a power series: w(p) = ao + alP + a2p2 + . Hence. it is easy to verify that if k > Cpo + 2(C .2 + . for k .l + .. C> 1/2.f!. If C is chosen to be greater than 1. the coefficients ak are proportional to 2(k + ao oF 0.266 Chapter 12 Spherically Symmetric Potentials For the hydrogen atom. contrary to the conventional procedure. ) P p2 (12. For convenience we introduce a dimensionless parameter Po = lEI = ~ iET h E V . .2(f + 1)]w = ° (12..2k ak + [Po . This power series must terminate at some finite maximum power.8.64) leads to the condition I Po = 2(N +f + 1) I (12. and N = 0. the radial wave function u(r) must behave as r N I near the origin.Po (12..58) can then be written as P 2 d 2w dp + 2(f +1 p)  dw dp + [Po . and w(p) must be a polynomial. 1.1)(f + 1). Let us suppose its degree to be N..62) and equate to zero the coefficient of pk. (k + l)k ak+l + 2(f + 1)(k + l)ak+1 . + f + 1) . The energy levels and eigenfunctions of bound states in the nonrelativistic theory will be discussed in this section and the next. e is the charge of the electron and Z = 1.63) We substitute (12. as shown in Exercise 12. (12.65) where f = 0. 1. that is. This is also true for S states. (12.9.64) l)(k + 2f + 2) for k > 0. 2 .57) that for large values of p the radial eigenfunction u(p) diverges faster than e(llIC)p. Ze 2 K ~ Ze2 (12.60) such that (12. 2. Exercise 12. According to the discussion of the last section and especially Eq. all terms with k > (1/2)po . Furthermore.61) Po For this potential. Equation (12.(f + 1) would have the same sign. aN+I = 0.8 on scattering. Such a strongly divergent wave function is not quadrat ically integrable and is not acceptable to represent the physical system. where C is a constant. the differential equation (12.63) into (12.63) must terminate. If it failed to do so. but aN oF 0. that the radial eigenfunctions for a bound state of the hydrogen atom can be written as k U = peP ( b o b b + .
67) that n must be a positive integer. and the number of linearly independent stationary states of hydrogen having the energy (12.65) is equivalent to the Balmer formula for the energy levels in hydrogenic atoms. n 2:: 1 (12. There are infinitely many discrete energy levels. we see that K= = e 2an2 (12. Its numerical value is a = 0. the degree of degeneracy.6 eV. . U + 1 different eigenfunctions of the same energy are obtained by varying the magnetic quantum number m in integral steps from .5 The Coulomb Potential 267 Obtain the recursion relations for the coefficients. it is obvious from (12.] It is amusing to contemplate that as innocuous an equation as (12.1) consistent with a given value of n. we merely substitute (12.66) in descending powers of r can be useful for calculating approximate radial eigenfunctions.60) into (12..e to .8 cm Z2 2 Using this quantity. . with an energy of approximately 13. ~ ~ (12. we can write the energy simply as En Also. Hence.68) As is well known from the elementary Bohr theory. [A series like (12. is nl e=o 2: (U + 1) = n2 (12. there are n values of . The fact that the energy depends only on the quantum number n implies that there are in general several linearly independent eigenfu~ctions of the Schrodinger equation for hydrogen corresponding to the same value of the energy.67) The result is (12.73) .e = 0. The length a is termed the first Bohr radius of hydrogen if m = me is the mass and e the charge of the electron. In the first place. 1.e+1=Po 2 (12. but not at small.e. To see this. n .e = 0. they have a point of accumulation at Eo = 0 (n ? 00).71) Since N is by its definition a nonnegative integer.69) sets the length scale in the quantum description of the hydrogenic atom. Second.72) The ground state of the hydrogen atom corresponds to n = 1.68).65) and define the principal quantum number n==N+.70) Z na and p =  Z r na (12.e(.2. and show that the boundary conditions give the same eigenvalues and eigenfunctions as usual. all energy levels with n > 1 are degenerate. .529177 X 10.. distances. if V behaves as the Coulomb potential at large.
but n differs from an integer by an edependent quantity. Show that K satisfies the condition (11.~~)Ln (12. For our purposes. The degeneracy of energy eigenstates with different values of e is not a general property of central forces. It occurs for the Coulomb potential because the Hamiltonian of this system is invariant under a less obvious symmetry. The new constant of the motion is the quantummechanical analogue of the RungeLenz vector (apparently first introduced by Laplace):4 K =  1 (L X P .74) This is a vector operator that commutes with the Hamiltonian for the hydrogen atom. (The degeneracy is doubled if the spin is taken into account. Any departure from a strict lIr dependence of the potential removes this degeneracy.68). and that (12. in standard spectroscopic notation (n followed by the symbol for e). the first excited level of hydrogen (n = 2) is fourfold degenerate and consists of the 2S state and three 2P states. which generates an additional constant of the motion. Exercise 12.~~)Lz. (12.10. For instance. Show that the addition of a small lIr 2 term to the Coulomb potential removes the degeneracy of states with different e. the degeneracy with respect to magnetic quantum numbers is clearly present for any central potential. It has its origin in the absence of a preferred spatial direction in such systems.75) and has the properties L·K=K·L=O (12.77) 'See Goldstein (1980).11. . The energy levels are still given by a Balmerlike formula (12. in(.) The occurrence of degeneracy can often be ascribed to some transparent symmetry property of the physical system.75).268 Chapter 12 Spherically Symmetric Potentials For example.76) holds.~~)Ly [Ky . 102. p. the crucial property of the components of K is that they satisfy the commutation relations [Kn K y ] = = [Kz' Kxl in(.19) for a vector operator. the quantum defect in the terminology of oneelectron (alkali) spectra.P X L) 2me 2 + r r (12. that it commutes with the Hamiltonian (12.76) Exercise 12. K z ] = in(. and it reflects their invariance with regard to rotations about the origin.
wherej can be any nonnegative integer or halfinteger."""7 2 2li (2j 4 + 1)2 (12. Check the commutation relations (12. it is convenient to introduce the operators J1 = 2 (L + A) 1 and J 2 = 2 (L 1 .83) Exercise 12. Prove that J 1 and J 2 satisfy the commutation relations (11. as me4 2E (12. The two commuting "angular momentum" operators J 1 and J 2 can be linked to •'rotations" in a fourdimensional Euclidean space.80) Restricting ourselves to the subspace of a definite (negative) boundstate energy E.84) Any state in the subspace spanned by the energy eigenstates corresponding to an eigenvalue E must be an eigenstate of Ji = ~. according to (12.13.81) is then transformed into 2(Ji + J~) + li 2 = 2E (12. we may write this. They have the property (12. From (12.81) Finally.78).77).A) (12. if we identify the positive integer 2j + 1 with the principal quantum number n.2 we know that this operator assumes the eigenvaluesj(j + l)li 2 .5 The Coulomb Potential 269 From this relation it is seen that the vector operator A=J~~K (12. Equation (12.83). An important identity is obtained from the definition (12.84) we therefore deduce that E me = .82) each of which satisfies the angular momentum commutation relations.::.24) and the conditions (12. The invariance of the Ham .12.79) Exercise 12.85) which is identical with the standard form (12. From Section 11.74) through some operator manipulations: (12.78) defined in the subspace of boundstate vectors with energy E < 0 satisfies the commutation relations (12.68) for the energy levels of the nonrelativistic hydrogen atom.
67) is cast in the form  P d 2w 2 dp 2 + (.44) is normalized to unity if Jooo 1 u(r) 1 dr = 1 2 5Bethe and Sa1peter (1957). c. 6. with the radial equation (12. z) = 1 + ~ 1! + a Z + + 1) Z2 1) 2! + ~12. (c .270 Chapter 12 Spherically Symmetric Potentials iltonian for the Coulomb interaction under these "rotations" signals the new symmetry that accounts for the degeneracy of the energy levels of the hydrogen atom with different . the complete eigenfunction o/(r.91) Since the spherical harmonics are normalized to unity. we see that w(p) ce jFj(.62) if the latter by the use of (12. eigenstates in Section 13.e + dw 1 . z) satisfies. .65) for Po is put into (12...e values.e + 2.N.p) dp + Nw = 0 (12. 2p) (12.e + a(a c(c 2) (12.N) (k + l)(k + 2. In the meantime. we return in the next section to the hydrogenic energy eigenfunctions in spherical polar ((oordinates.89) which w = jFj(a.aw = 0 (12. c. c.2a)(a)![(c .86).8.87) Comparing its coefficients with (12. cp) of (12.2a)f(1 . We will take advantage of this property in discussing the positiveenergy.87) as jFj(a.z)  dw dz . 2. Section 3.. It can be shown that5 f OO o e. as can be seen when the value given by (12. unbound or scattering.xxc [ j F j (a. x) f dx = (c .a)[f(c)]2 f(c .64): ak + j ak 2(k .l)! (c .88) This can also be seen by comparing the differential equation z d2 w di + (c . These functions can be expressed in terms of confluent hypergeometric functions.'''= .86) The confluent hypergeometric function has already been defined in (8. The BoundState Energy Eigenfunctions for the Coulomb Potential. This symmetry also explains why it is possible to separate the corresponding Schrodinger equation in parabolic as well as spherical coordinates. e. This section summarizes the most important properties of the radial boundstate wave functions for the attractive Coulomb potential.a .90) The complete normalized wave function can be obtained only if we know the value of the normalization integral for the confluent hypergeometric functions.a) (12.1)!]2 ''''=.
Show that the generating function for the associated Laguerre polynomials (for lsi s.14.1 as a fUJ.e a = N. Section 12.. Ir.95) Exercise 12.93) An elementary definition of the associated Laguerre polynomial is I.15. 2(. The connection is established by the relation IF1(a. 2Kr)P'. 2.6 = The polynomial1F1(N. For good early graphic representations see White (1931). . Exercise 12. (1935).92) to the realvalued eigenfunctions "'n(m ex: " r(e. z)  _ [(1 .S)p+l ~) ~o ~ Ifn(z) (n n + p)! (1296) .a)f(c) cl [f(c _ a)f La (z) (12.. it is customary' to refer instead of (12. normalized to unity. z) of degree N and positive integral argument c is proportional to the associated Laguerre polynomials of classical mathematical physics.92) The radial eigenfunctions Rn ( for the three lowest energy eigenvalues are plotted in Figure 12.'(cos ()){c~s mcp mcp sm (12....6 The BoundState Energy Eigenfunctions for the Coulomb Potential 271 Since p c .e + 1. ca1=n+.e = + 1 + N) = we can write the hydrogenic eigenfunctions.97) 6Brandt and Dahmen (1985).. 2n.s_ 1 (1 . Section 5. s A few comments concerning the properties of the hydrogenic eigenfunctions are appropriate. See also Pauling and Wilson. These are surfaces on which '" = O. For these considerations.94) and the Laguerre polynomial of order q is (12. c. The wave function possesses a number of nodal surfaces.!qp(z) = (1)P dzP Lq(z) dP (12. c=U+2.lction of Zr/a.e + 2..2a = Kr. C. 1) is exp ( _ _ __.21. .Kr IF1(n + . as (12. Calculate the radial hydrogen energy eigenfunctions for n 4 and 5 explicitly. and sketch them graphically.4..
and 3.::I=::::=6".2 0.1 +~~~!::==::7:=~~~:.5 0. 2.l.1 (bl 0.7 0.1. Radial boundstate energy eigenfunctions of the hydrogenic atom for n = 1.15 0.5 0.075 0.1 0.025 L !:LLL. The radial coordinate r is scaled in units of a/Z.6 0. .Zr a 0.5 2 (al 3 RzO 0.Zr 14 a (Cl Figure 12. The wave functions are normalized as L~ [R n ir)fr2 dr = 1.4 0.125 0.3 0.05 0.272 Chapter 12 Spherically Symmetric Potentials 2 1.
04 0.02 (e) 0.3 0.4 0. J'''' 5 (j) 10 15 20 Zr a Figure 12.06 0.1 j<.4 0.3 0.1 Zr 15 (d) 20 a R 31 0:08 0. (continued) .1.6 The BoundState Energy Eigenfunctions for the Coulomb Potential R 30 0.2 0.2 0.
The quantum mechanical significance of the Bohr radius a = fi. )0 exp (2Z) r3 dr = 2: Z (12. (12. which have equal .. The Bohr radius a is inversely proportional to the mass of the particle that moves around the nucleus. the capture by a nucleus of an atomic electron or any other orbiting negatively charged particle can occur ~ith appreciable probability only from a level with e = 0.. the maximum of the function is located at aiZ. this correction can be applied accurately and without difficulty.eparation r from the nucleus. if r = is counted as a surface. The finite size of the nucleus will thus be expected to affect the discrete energy levels of exotic atoms appreciably. Most obviously. r 3a = 4 (.O(r. whereas nuclear size effects are very small for electronic states in the bound states of ordinary atoms. or an antiproton that has been captured by a nucleus.e . Hence.O. be replaced by an effective onebody problem. must be taken into account when comparing the simple Balmer formula (12. the total number of nodal surfaces is n. Hence. we must correct the error made in assuming that the nucleus is infinitely massive and therefore fixed. For e ::1= 0. For instance. composed of an electron and a positron. can be appreciated by observing that the wave function for the Is ground state is ° !/JI. i. of course.1 values of r. Important consequences follow from the fact that the wave function vanishes at the origin except when e = 0 (S states). Similarly. r t has a node at r = 0..99) The maximum of the probability density for finding the particle in the ground state with a radial s. Many other corrections.274 Chapter 12 Spherically Symmetric Potentials There are e . except in S states. a muon.16.m values of efor which P~(cos e) vanishes. ( exp (Z) f. q» 3)1/2 =. because these are the only states for which !/J has a nonzero finite value at the position of the nucleus. Evaluate the width of the probability distribution for the radial coordinate r in the ground state of the hydrogenic atom by calculating the uncertainty Ar.e. e. in the phenomenon of internal conversion an atomic s electron interacts with the nucleus and is imparted enough energy to be eje~ted from the atom in an autoionizing transition. in the nonrelativistic approximation. if we substitute the reduced mass mlm2/(ml + m2) for m. Since for central forces the actual twobody problem can.. or kaon in a mesic atom. cos mq> and sin mq> vanish at m values of the azimuth. is much closer to the nucleus than an electron is in an ordinary atom..68) with the amazingly accurate results of modern atomic spectroscopy. For a positronium atom.98) The expectation value of r in this state is (r) Z)3 (00 .zlme 2 . pion. Exercise 12. It gives rise to small but significant differences between the spectra of hydrogen and deuterium. and the confluent hypergeometric function IFI vanishes at n .
ain from mathematical tables) and plot the 10 lowest energy eigenvalues of a particle in an infinitely deep. if a square well just binds an energy level of angular momentum .) 5. there are small but measurable effects owing to the interactions between the electron and the electromagnetic field (Lamb shift). Further. Given an attractive central potential of the form VCr) =  solve'the Schrodinger equation for the S states by making the substitution g= e. 4. with undetermined parameters a and {3. If the ground state of a particle in a spherical square well is just barely bound. For a detailed account of the spectrum of hydrogenic and heliumlike atoms. estimate the depth of the potential. 3.9). Estimate the value of Va' if the state of the deuteron is to be described with an exponential potential (see Problem 2 for data). Apply the WKB method to an attractive Coulomb potential. . 7Gross (1993). Assuming the eigenfunctions for the hydrogen atom to be of the form rf3 e". 2. If the deuteron is represented by a nucleon. But all are overshadowed in magnitude by the basic gross structure of the spectrum as obtained in this chapter by the application of nonrelativistic quantum mechanics to a charged particle subject to the Coulomb potential. all energy levels are changed by a factor 1/2 compared to hydrogen. and label the states by their appropriate quantum numbers.e (=F 0). solve the Schrodinger equation. hyperfine structure effects arise from the magnetic properties of the nucleus.r1a The deuteron is bound with an energy of 2.5 fermi. which necessitate a relativistic calculation (Section 24. and derive an approximate formula for the Sstate energy levels of hydrogen. Compute (or obt. spherically symmetric square well.r }7. show that the well depth and radius parameters Va and a are related to the binding energy by the expansion where fiK = V2mE Voe. see Bethe and Salpeter (1957). corrections are due to the presence of the electron spin and the high speed of the electron. and finally. and often more significant.Problems 275 masses. its parameters satisfy the condition jel( J2m:~if) = 0 (Use recurrence formulas for Bessel functions from standard texts. with reduced mass. Show that. Are all eigenfunctions and eigenvalues obtained this way? 6.rl2a Obtain an equation for the eigenvalues. Problems 1. moving in a square well with a = 1. others lie outside its scope.226 MeV and has no discrete excited states.? The theory of some of these effects will be discussed later in this book.
Compute the probability that the electron in a hydrogen atom will be found at a distance from the nucleus greater than its energy would permit on the classical theory. compare the density of energy eigenstates in the oscillator and in a cubic box. and establish the correspondence between the two methods. Calculate the probability distribution for the momentum of the electron in the ground state of a hydrogen atom. For the isotropic harmonic oscillator of Problem 9. and verify the uncertainty relation for this state. (j) 2 and show that f(r) can be expressed as an associated Laguerre polynomial (or a confluent hypergeometric function) of the variable mwr2 /h with halfintegral indices. 9. determine the wave function for all times as completely as possible. compare the results with the exact ground state energies.276 Chapter 12 Spherically Symmetric Potentials 7. produces an equation for itO') that. obtain a formula for the degree of degeneracy in terms of the energy. Assuming that the expectation value of the coordinate z has at this time its largest possible value.cillator is known to be an eigenstate of L z with eigenvalue zero and a superposition of the ground and first excited states. Assume that the eigenfunctions are of the form ~: p2 )f(P) exp(±if(j) where p and (j) are plane polar coordinates and e is a nonnegative integer. 10. 11. Exhibit the relation between the energy eigenvalues and the radial quantum numbers for the two systems. Solve the energy eigenvalue problem for the twodimensional isotropic harmonic oscillator. Discuss a few simple eigenfunctions. For n = 1 and +2. show the relation between the eigenfunctions obtained by the two methods. For large energies (large quantum numbers). with appropriate choices of the constants. u = Wit. In the latter case. V = (112)mClh 2 . is equivalent to the radial equation for the isotropic oscillator. 12. assume the eigenfunctions to be of the form Ifr(r. 8. by separation of variables in Cartesian and in spherical polar coordinates. Starting with the radial equation + r= h 2 tfu 2m dr h 2 f(f + 1) Ze 2 uu=Eu 2mr r for the hydrogenic atom. 0. Make the calculation for the n = 1 and 2 levels. For the lowest two energy eigenvalues. show that the transformation ar. . The initial (t = 0) state of an isotropic harmonic os. using R(r) = e/3r as a trial wave function with variational parameter (3. Obtain the expectation value of p~ from this or from the virial theorem. Solve the same problem in Cartesian coordinates. 13. Show t/J = pe exp(  that f(p) can be expressed as an associated Laguerre polynomial of the variable mwp2/h and determine the eigenvalues. Obtain the eigenvalues and establish the correspondence between the two sets of quantum numbers. Also calculate (x 2 ) from the ground state wave function. (j) = re exp (  ~: r )f(r)17(9. Apply the variational method to the ground state (e = 0) of a particle moving in an attractive central potential V(r) = Arn (integer n ~ 1). 14. Solve the Schrodinger equation for the threedimensional isotropic harmonic oscillator.
Under the assumption that the nucleus is much smaller than the atomic radius. show that the energy change is approximately proportional to the nuclear mean square radius. . Using firstorde~ perturbation theory. 17. Obtain the "best" trial wave function of this form and deduce a relation between 'Yand the strength parameter 2rrtVo a2 /h 2 • Evaluate 'Y and calculate an upper bound to the energy for 2mVo a 2 /h 2 = 2. For some representative values of Z = Z'. estimate the limit that must be placed on R so that none of the lowest three energy levels shift by more than 5 percent of the distance between the unperturbed first excited and ground state energy levels. modified by the presence of a uniformly charged spherical shell of charge . Is the level shift due to the finite nuclear size observable? Consider both electronic and muonic atoms. Voa finite) the correct energy and wave function for the hydrogenic atom are obtained. centered at the point charge.Problems 277 15.7. Apply the variational method to the ground state (e = 0) of a particle moving in an attractive (Yukawa or screened Coulomb or Debye) potential e. 16. and 2P energy levels. Perform a firstorder perturbation calculation of the hydrogenic IS. a ~ 00.r1a V(r) = Vo  ria eyr/a (Yo> 0) Use as a trial function R(r) = with an adjustable parameter 'Y. An electron is moving in the Coulomb field of a point charge Ze. Are there any excited bound states? Show that in the limit of the Coulomb potential (Vo ~ 0. 2S. Evaluate the correction for a uniformly charged spherical nucleus of radius R. estimate the correction to the ground state energy of a hydrogenic atom due to the finite size of the nucleus.Z' e and radius R.
Mott and Massey (1965) and Bransden (1983) apply the theory to atomic collision problems. The collision between a projectile and a target particle is sometimes referred to as a binary collision. Electrons of high energy. Usually. as happens in diffraction of electrons. Collisions of nucleons with nuclei at various energies reveal information about the nuclear forces and the structure of the nucleus.CHAPTER 13 Scattering Much of what we know about the forces and interactions in atoms and subatomic particles has been learned from collision experiments. And collisions of hadrons and leptons with protons tell us about interactions of which we have no other direct information. We then establish the connection between calculated quantum mechanical amplitudes and measured cross sections for collisions in which elastic scattering is the dominant process. neutrons. Electrons and heavier projectiles of low energy are scattered from atoms to obtain data that can serve as input information for calculations of kinetic processes in gases where lowenergy collisions predominate. The target particle is frequently at rest (or nearly so) before the collision. Particles that are scattered by the target atoms are subsequently detected by devices that may give us the intensity as a function of the scattering angle (and possibly of the energy of the scattered particles. In merging or colliding beam experiments. Some form of collision experiment is the most common tool for probing atomic and subatomic interactions. The Cross Section. the target atoms are in gas jets. both projectile and target particles initially move IGoldberger and Watson (1964). with controlled initial velocity. In some experiments. in order to distinguish it from the kind of interaction that takes place when an incident beam interacts with a large number of target atoms. to be developed in Chapters 14 and 15. and Taylor (1972) are treatises on collision theory. we know the nature of the particles used as projectiles. and partial wave and phase shift analysis. We begin the discussion with a general introduction to the concept of a cross section. hence short wavelength. their momentum. in Chapter 20. we will take a second comprehensive look at collision theory in the context of the general principles of quantum dynamics. and indeed within nucleons. quantum mechanical integral equations. in which atoms in a target are bombarded with beams of particles. 278 . These are just a few examples of the utility of collisions in studying the internal structure of atoms and nuclei and the interactions that govern elementary particles. but its thermal motion cannot always be neglected. Later. or even entire atoms from a lattice. Newton (1982). 1 1. The stockintrade of scattering theory is developed: incoming and outgoing Green's functions. are particularly well suited to determine the charge distribution in nuclei. if inelastic processes are also involved). usually at right angles to the incident beam. the Born approximation. and perhaps their polarization (defined in Chapter 16). since it constitutes the meeting ground between experimentalists and theoreticians for all varieties of collisions.
it is assumed (unless otherwise specified) that the beam is incident normal to the surface of the target. How~ver." it is desirable to employ beams with high incident intensity. Good resolution in incident velocity (magnitude as well as direction) is prized in all collision experiments. . Sketch of a scattering experiment. if an incident charged or neutral projectile particle is to cause a certain specified effect in a given target particle. in the atomic and subatomic domain the cross section. and the corresponding cross section is Ux = Ix/NTlo. in the interest of securing "good statistics. in physics. and with controlled initial velocity. either per pulse or per unit time. a cross section measurement is similar to the ("Monte Carlo") determination of the size of the bull' s eye on a dart board by recording the successful hits among randomly thrown missiles directed blindly toward the target. a cross section measures the size of the effective area that must be hit. The comparison between measurements and theoretical predictions is made in terms of the cross section for the parti'cular process under consideration. Experimentalists typically measure intensities and yields of certain collision or reaction products that result from the interaction of incident particles or waves with target "atoms. are nevertheless very wide compared with the cross section. The interpretation of most experiments requires accurate knowledge of the physical properties of the target. A pulse of N p = loA projectile particles is incident from the left with velocity Yo. though quite narrow from an experimental point of view. ~A ** *t* Figure 13. which may also be either neutral (stationary target) or charged (colliding beams). Prototypically.1. especially the number of target particles that are exposed to the incident beam. Yet the beam density must be low enough that it can safely be assumed that the incident particles do not interact with one another. either parallel or antiparallel. Broadly speaking. The width of the beam is determined by slits. provided we know the average number of projectiles striking the target per unit area. If the target is solid. although the "atoms" can be any kind of object (particle or wave) that affects the incident beam. The detector D registers Ix outcomes X. which." upon which the projectiles impinge. is essential. The beam intersects an area A of the target (thickness t and particle density n) and encounters NT = ntA target particles.1 involves a collimated homogeneous beam of monoenergetic particles moving in the same sharply defined direction toward the scatterer from a great distance.The Cross Section 279 along the same direction. A precise knowledge of the number of projectiles. Experimentally. The experiment illustrated in Figure 13. The relative velocity (collision energy) of the two interacting particles is usually specified at incidence. generally bears no direct relation to the geometric size of the target particle. or luminosity. which is a physical property of the target particle and its interaction with the projectile.
1). Under these conditions if. Eq. the energetically degraded projectile and the excitation of the target may be observed simultaneously by a coincidence measurement of some outgoing radiation or particle emission.) In the more common geometry of stationary targets exposed to projectile beams. Ix measures the rate or yield of emerging particles with outcome property X per unit time.e.. the quantity Ix is known as the counting rate for the process that is being observed. This direction is defined by the position of the detector. The ratio 10 = NpIA. the outcome X might refer to any of a number of physical processes.1) Since experiments have to be conducted within a reasonable length of time. (In the case of a stationary target. The number NT of target particles exposed to the beam can be expressed as NT = ntA (13. this is just the direction of the incident beam. luminosities are of the order 1032 sec. into a welldefined solid angle tilted at a specified direction with regard to the direction of incidence. or chemical changes in the composition of the target. Sometimes the recoil of the target particle is observed. (In typical highenergy particle accelerators. the emission of radiation or particles. If the product NpNT is interpreted as the number of encounters between projectile and target particles per unit time.1) is usually written in a less symmetric form. the creation of new particles. e." Let A denote the area of overlap of the beam of projectiles and the assemblage of target particles.1) is directly applicable to the colliding beam geometry in particle accelerators. and the quantity L = NpNTIA is generally referred to as the luminosity (in units of sec.) We assume that the experimental arrangement allows us to choose A » (Tx. the .l cm. In measurements. the total number Ix of detectable outcomes with property X is the product: (13. If the collision is not elastic. (13. although differential cross sections may be defined and quoted as "area per unit measure of X. with or without loss of energy. (b) Absorption of the incident projectile by the target. dimensionless ratio (TxlA is the probability of the outcome X in a binary collision of a projectile particle with a target particle. the assemblage of projectiles contains N p particles and the target assemblage contains NT particles. in laboratory practice steady and stable beams are characterized by the number of particles per unit time. within the common overlap area A. All cross sections have the dimension (units) of an area.2 for colliding beams. provided that the presence of the other projectile and target particles does not affect the binary collision.280 Chapter 13 Scattering Let us suppose that we are interested in determining the likelihood of an outcome X in a binary collision.g. If the detector is "ideal.. Expression (13. Let (Tx stand for the cross section that corresponds to the outcome X. the incident intensity. The magnitude of the luminosity is critical if reasonable counting rates are to be achieved.l cm. is the number of particles incident on the target per unit area and unit time." i. measured either through the attenuation of the incident beam or through observation of reaction products. 100 percent efficient.2) . measured at right angles to the relative velocity of the projectiletarget system (Figure 13.2 ). as for example: (a) Scattering.
Similarly. cp that are related to the direction of incidence.The Cross Section 281 where n is the number of target particles per unit volume and t is the thickness of the target. If the binary collision between a projectile particle and any target atom is unaffected by the presence of the other target atoms. (13. such as scattering in a particular spatial direction or with an energy loss that lies in a certain continuous range. Eq. and the velocity (if different from zero) of the target atoms. The initial collision parameters. An example is the differential scattering cross section. including the relative velocity or kinetic energy of the incident particles. or angular distribution. The target is said to be "thin" if t « lInux. a total scattering cross section measures only the occurrence of the scattering process. who interprets the cross section in terms of probabilities calculated from wave functions. are usually determined and quoted with respect to a laboratory frameofreference. the product uxnt is the fraction of the total area A over which beam and target particles interact that is effective in producing the outcome X. du dn I«(). chosen as the polar axis. the outcome X is detection of the particles after scattering at a great distance from the scatterer into the solid angle dn in a direction specified by spherical coordinates (). the relation (13. and not any specific feature of the scattered particle. cp). X may stand for a physical process that is described by one or more variables x with continuous values.2). the intensity of the incident beam. defined as the angle subtended at the target by the detector in a given direction (Figure 13.3) For the cross section Ux we thus obtain the formula Ix ux=ntAIo (13. such as the center .1) can be reexpressed as Ix counting rate for detectable particles with property X ntux = AIo = number of particles incident on the target per unit time I (13. For a thin target. which either does or does not occur. It is then customary to write this cross~ection in the form dux and to define the ratio dux/dx as the differential cross section Jor the outcome X with value x. The outcome X of a collision experiment is defined by the physical arrangement of the measurement.4) becomes the expression for the differential cross section. the theoretical analysis is often carried out in an entirely different frame of reference. If the number of scattered particles per unit solid angle is denoted by I«(). This must be less than unity: NT Ux  A = ntux < 1 setting an upper limit to the thickness t.4) The outcome X in a collision experiment may be a discrete physical characteristic. On the other hand.5) This is the quantity which the experimentalist delivers to the theoretician. corresponding to scattering into a solid angle dn. such as the "yes/no" choice in a nuclear reaction. cp) ntAIo (13. The cross section is a measure of the probability that the outcome lies in the interval between x and x + dx and is usually proportional to dx. du/dn. Yet. Here.
6) This relation cautions us that the two differential cross sections do not have the same value. a distance b from the z axis. parallel to which it approaches from infinity. . a transformation to the corresponding variables.) and the particular variables (scattering angle. in the laboratory frame must be carried out. each incident particle can be assigned an impact parameter. The cross section depends on the collision parameters (incident energy. etc.. In the forward direction. etc. the differential crosssections in the two frames are related by dax = dax dx dx' dx dx' (13. Although this assumption excludes as common a longrange force as the Coulomb field.282 Chapter 13 Scattering t= 0 Figure 13. If these variables are denoted by x in the theoretically preferred frameofreference. In an idealized scattering process. A numerical example at the end of this section illustrates this for the angular distribution. represented by a potential proportional to 1/r and observable as Rutherford scattering. or with the use of relative coordinates appropriate to the interaction between two particles. polarization. In actual fact. this limitation is not severe. a single fixed scattering center or target particle is bombarded by particles incident along the z axis. ofmass frame in which the total momentum of the system is zero. i. x'. energy loss. Formally.2.e. It is assumed that the effect of the scattering center on the particles can be represented by a potential energy V(r) which is appreciably different from zero only within a finite region.) which quantify the outcome X. and for large r the potential falls off faster than 1/r. We now turn to the theoretical calculation of a scattering cross section. the Coulomb field is screened at large distances by the presence of electrons in atoms and by other particles. the wave packet is scattered with equal energy but varying amplitudes in different directions (indicated by the thickness of the arrows). the scattered wave interferes with the advancing incident wave. In elastic scattering. Scattering of a wave packet incident from the left with mean momentum hko. In the classical limit.
.. Obviously. and an azimuth angle q. (13... V = V(r) . with potential energy V(r) = Clr (C constant).8) Hence.\\ +1.. 'P + d'P defines the cross section area du.. this discussion must be refined if the scattering angle is not a continuous smooth function of the impact parameter or of the azimuth angle of the incident particle. we know e. the orbits lie in planes through the center of force. : b .r . q...t .1 In Rutherford scattering of a particle of mass m and incident energy E. du = b db dq.. I'll Figure 13. q. To determine this function from the equation of motion is the problem of classical scattering theory.. the cylinder bounded by b ..z du=bdbdtp \ .3. is traversed by all those orbits that asymptotically end up in the solid angle dO... = bid ~ e!ld cos el dq.. Scattering through an angle 8 of a particle beam incident at impact parameter b and azimuth angle 'P. the functional relationship between the impact parameter b and the scattering angle () is given by b C e = cot2E 2 .. the differential cross section is calculable if...The Cross Section x 283 ... b + db and 'P .. for the given energy.. which together with z define its position in cylindrical coordinates. is determined by these initial coordinates and by the incident energy.. when placed at right angles to the incident beam..r . classically. rainbow scattering)..g. I ..3 illustrates the point.. Classically..7) du db b de bId cos eI = sin e Idbl (13... and the scattering becomes independent of the azimuth angle q. = bid ~ from which we see that dO.. the final asymptotic direction e. around the direction e.i . Figure 13.. The cross section du is simply equal to the size of the area which.. If the scattering potential is spherically symmetric.. but we ignore such singular situations (e. b as a function of the scattering angle Exercise 13. For particles moving on classical orbits. = el dO. Therefore.
48/4 times the incident energy.7 g cm. allowing us to employ the differential cross section (13. Consider the elastic Rutherford scattering of 1OMeV alpha particles from an aluminum foil of t = 1 /Lm thickness.9) where ZI e and Z2e are the nuclear charges of the projectile and target particles. Given that the charges ofthe colliding nuclei are 2e (for . consisting. Equation (13. We conclude this section with a concrete illustrative example. Since the atomic mass of aluminum is about 27 and its density is 2.48 u (instead of approximately 4 u for an a particle colliding with an infinitely massive nucleus). although strictly speaking the waves do not extend to infinity either in width or in length. of particles that were emitted from a source. but with directiondependent amplitudes (Figure 13. radially outgoing harmonic waves. the energy of the relative motion is 3. before they reach the neighborhood of the scatterer" these packets can be described simply and approximately by infinite plane waves ei(kzwt).2). Hence. respectively. As the wavelength increases. say. and the quantum uncertainty begins to limit the simultaneous knowledge of. Pn which for a wellcollimated beam should be close to zero.2 • The reduced mass of the aAI system is 4 X 27/31 = 3. and we would like to describe them again. The Schr6dinger equation for unbound eigenstates must now be solved. and x. We will see in Section 13. and that may have been selected as to energy and momentum by analyzers and slits. corresponding to positive energy eigenvalues. After scattering. by simple. It ceases to be useful when the wave description can no longer be adequately approximated by the geometrical optics limit. Thus we must learn how to calculate cross sections from wave functions. Only comparison between theory and experiment can tell us whether our assumption that all the particles in the beam can be represented by very broad and long wave packets truly reflects the properties of a real beam. as it does. Fortunately. the number of Al atoms per cm3 is n = 6 X 1022 cm 3. Quantum mechanics represents the particles in the beam by wave packets of various shapes and sizes. The number of atoms per cm 2 of area is thus nt = 6 X 10 18 cm.284 Chapter 13 Scattering Derive the differential scattering cross section for Rutherford scattering: du dO ZiZ~e4 1 l6E 2 sin4 (0I2) (13.9) for scattering of alpha particles.8 that quantum mechanics gives essentially the same results for this force law as classical mechanics. the particles are detected at a great distance from the scatterer. asymptotically.8) is a good approximation only if the de Broglie wavelength of the incident particles is much smaller than the dimensions of the scattering region. that have perhaps undergone some acceleration.7 MeV. or 8. which is proportional to the impact parameter. The most significant quantum mechanical aspect of scattering is that we are now concerned with unbound particles and the continuous part of the energy spectrum. and the connection between eigenfunctions and measured intensities must be established. quantum features make their appearance.3 . We would like to suppose that all the particles in the beam with incident intensity 10 can be represented by very broad and very long wave packets propagating in the z direction and that. provided only that the packet is large compared with the scattering region toward which it is directed. the mathematical analysis ofthe next section will show the cross section to be in general independent of the peculiarities of the incident wave packet.
53 barns This is the differential cross section per steradian.4). the target particle is initially at rest. In the centerofmass frame Vp and VT are the velocities of the scattered projectile and the recoiling target particle.3 fm. the projectile has velocity vo. He) and l3e (for AI).The Cross Section 285 VG Figure 13. so that small departures from Rutherford scattering may occur. and the center of mass has constant velocity Va. the kinematic relation between the scattering angle () in the centerofmass or relative coordinate frame and the scattering angle (]I in the laboratory frameofreference is tan ()' sin () cos () + mp mT (13. A mass ratio mp/mT = 3/5 has been assumed in this diagram. The distance of closest approach corresponding to this scattering angle is 9.15 m or 4. In the laboratory frame (J' and (J~ are the scattering and recoil angles.24 cm2 = 1. We then calculate the Rutherford differential cross section in the centerofmass frame for this scattering angle as du dO = 1. and v~ and v~ are the final laboratory velocities. At this scattering angle one finds that d() dO' = 1. respectively. which is close to the nuclear radius of aluminum.53 X 10.28 m2 = 1. In the laboratory frame.3 X 10.10) where mp is the mass of the incident projectile and mT is the mass of the initially stationary target (Figure 13.13 .5 fm. corresponding to ()' = 30° in the laboratory frame. For example. For elastic collisions.4.2°. Kinematics of elastic scattering in the centerofmass and laboratory coordinate frames.53 X 10. the scattering angle in the centerofmass system (as well as in the relative coordinate frame) is () = 34. respectively. their classical distance of closest approach in a headon collision is calculated from conservation of energy as 4.
The normalization constant N is arbitrary. Calculate the probability that in the example of Rutherford scattering just given an alpha particle is scattered through an angle greater than 10.  du dO' = . The initial state I/J(r.1 MeV. We assume that k o is parallel to ro. centered around a mean momentum hko.) It is important to ask if the energy of a lOMeV ex particle is appreciably de'graded as it passes through the 1 micron aluminum foil.I cm. or only 1 percent of the original energy.12) where 4> (k) is a smooth function of narrow width. Since the stopping power of aluminum for protons of the same velocity (2.95 X 10. the calculated cross section for this outcome. as expressed by (13.6 steradian. is du = 1. Let us assume that the motion of the projectiles in the field of a single scatterer (NT = 1) is described by a Hamiltonian 2 H =L + 2m V = T + V = Ho + V (13. for the differential cross section in the laboratory frame. 2. if we succeed in expanding I/J(r.30 cm2 . would move freely toward the coordinate origin. lithe rate of incident ex particles is f o = 10 12 secI.. if we can establish an expansion I/J(r.2 radian in the laboratory frame. Ak. it is expected that about one in 1011 of the ex particles that are incident on the target will be scattered into the detector at this scattering angle. The dynamical problem to be solved is as follows: What is the shape of the wave packet (13. so that the wave packet. (The luminosity in this experiment is font = 10 12 X 6 X 1022 X 104 = 6 X 1038 sec. Finally. 0) = 2: nl C nI/Jn (r) . one has to estimate the likelihood of multiple Rutherford scattering in the target.12).12) at a much later time. of H. The Scattering of a Wave Packet.ro)] d 3 k (13.2). Use the result to estimate the importance of multiple scattering in this example.286 Chapter 13 Scattering and. Indeed.17 X 10. For the foil thickness assumed in our example.5 MeV) is about 250 MeV/em. the answer can be given easily. and the probability that an ex particle is scattered into the detector with ideal detection efficiency is nt du = 1. in its passage through the foil. if unhindered in its motion.95 X 10. is thus an extended wave packet located in the vicinity of ro.11 . I/Jn (r). when the packet has encountered the scatterer and been eventually dispersed by it? In principle. In other words.O) =N f 4>(k) exp[ik . an ex particle (charge 2e) loses about 0.11) where V is appreciably different from zero only within 'a sphere of radius a surrounding the coordinate origin.95 barns = 1. the counting rate at the detector will be about 12 particles per second.24 cm2 If the solid angle subtended by the detector is dO' = 10. At t = 0 a particle is represented by a wave packet of the general form l/J(r. this effect is found to be quite small. 0) in terms of eigenfunctions. but in an opposite direction. and we assume ro to be so large that I/J at t = 0 lies in its entirety well to the left of the scatterer (Figure 13.2. but with fairly sharply defined momentum. 0). Exercise 13.= du sin () dO sin ()' 1.2 . (r .
but these are eigenfunctions pf the operator H o.17) 2The discussion of the time development of wave packets in scattering follows lecture notes by F.12) by (13.Low. (13. However. In order not to interrupt the argument.12).15) is the critical step in this analysis. r . after a time T= 2mro fLk o (13. whether it is expanded in terms of plane waves. Assuming that the scattering detectors are at a macroscopic distance of the order of ro from the scatterer. As we will see in Section 13.~+)(r) bear a considerable resemblance to plane waves. 2 Once the initial wave function "'(r.16) where fLw=2m fL2~ is the energy eigenvalue for the eigenfunction "'~+)(r).E.13) This differs from a plane wave at large r only by an outgoing spherical wave. It will also be shown in Section 13. we assume their existence in this section.3. if "'(r.12)? True. The replacement of (13.14) do indeed exist. 0) has the same expansion coefficients. as in (13. which we will designate as "'~+)(r). the eigenfunctions ".13) asymptotically makes no contJibution in the initial wave packet. 0) is a wave packet of the kind described. Asymptotically.3 that the special wave packet "'(r. and not of H.15) In other words. since they are of the form (r large) (13. we can write at any time t. such solutions of the SchrOdinger equation (13. the plane waves e ik . it is an expansion in terms of a complete orthonormal set. t) = ~ cn"'n(r) exp ( k Ent) How can this program be applied to a wave packet of the form (13.2 The Scattering of a Wave Packet 287 then the wave packet at time t is "'(r. . the outgoing wave in (13. 0) has been expanded in terms of eigenfunctions of H. it will be proved that it is possible to replace the plane wave functions in the expansion by particular eigenfunctions of H. or in terms of the set "'~+)(r): (13.
and assuming that fk(r). should still satisfy the inequality Ii . that in scattering of particles from atoms and nuclei the condition of no spreading can be easily attained by a minimum uncertainty wave packet (liklir "'" 1).17). k o + (k . as in (2.ko) .16) as "'(r. O)e. 1 (likiro « ko 1 This condition. we require that T.19) with (13. m liwo = 2: mvo 1 2 Comparing (13. Since for a wellcollimated beam. unlike 4J(k).2). we can make an expansion about kd: k?.4. (13. t) = f 4J (k) exp[ . we can write "'(r. though large. However. but since the phases have changed with time the outgoing wave may no longer be neglected.18) c In order to be able to neglect the last term. (ro + vot) + iwot] X N( e ik r .koi 2 (13.ik . kr"'" k· kor (13.29).wot .12). + !k(r') e:) dk 3 (13. when k is substituted at expression lik?wT=T 2m t = T into the in Eq. we can write in the exponent effectively. When we examine the pulse at the position of the detector. using (13.19) where lik o vo=. we can thus write (13. 4J (k) is appreciably different from zero only for k "'" k o. "'~+)(r) can again be represented by its asymptotic expansion. is a slowly varying function of k. implies that the wave packet spreads negligibly when it is displaced by the macroscopic distance roo Exercise 13. + f (r)e N ko r iwot f 4J(k) exp{i[kr .288 Chapter 13 Scattering the broad pulse will be traveling through the position of the detectors (see Figure 13.k O T« )2 2m or. t) = "'(r  vot. by numerical example. Show.k· (ro + vot)]} d 3 k with carets denoting unit vectors.= ~ + 2(k . which is large compared with the scattering region but small compared with roo Approximately and asymptotically.3.20) . familiar from Section 2.16).(k .
21) can be expressed as l/J([r .vot]k . 0)12 dt = Ifl<o(i')12 do' f:: Il/J({ko.6 for some discussion of resonances. t) describes a wave packet of finite length. the second term in (13. Sensibly. as if V were absent from the Hamiltonian. fl<o(r) A • (13. the total probability for detecting it at any time is VoIfl<o(i') 2 1 do' f:: Il/J([r .22) The limits of integration in (13.wot r . reduced in amplitude by the factor lIr and modulated by the angular amplitude fl<o(r).21) is a scattered spherical wave packet. all represented by the same general type of wave packet l/Ji (i = 1..l/J(rko . the first term on the righthand side represents the initial wave packet displaced without change of shape. . A detector placed at distance r is reached by this wave packet at time t = (ro + r)/vo. If the ensemble of projectiles contains N p particles.. again by comparison with (13.21) Except for the phase factor e iwot . since at t = 0 the wave packet is entirely in front of the scatterer.vot]ko. since l/J(r..vot. When a resonance is so sharp that the scattering amplitude changes appreciably even over the narrowest tolerable momentum range Ilk.vot = ro~ within a range Ilr that is the same as the spatial width of the initial wave packet. excludes from this treatment those scattering resonances which are characterized by an exceptionally rapid change of the scattering amplitude with the energy of the particle. the scattered wave packet may have a shape that is considerably distorted from the incident wave packet. 0). the number of particles scattered into the solid angle do' is . asymptotically. dt 2 o A r Hence. The assumption. 0) in the second term of (13. 0) 12 r2 dO. (See Section 13. of a slow variation of the scattering amplitude with k. because near such a resonance different momentum components scatter very differently. and represents a radially expanding image of the incident wave packet.vot]ko. Thus. It differs from zero only when r . 0)12 d{ (13.. O)e""ot + .22) may be taken to be ±oo with impunity.21).00 to + 00 either.) The probability of observing the particle at the detector in the time interval between t and t + dt is Vo Ifl<o(r)12 1l/J([r . t) = l/J(r . N p ). O)e. The probability that the incident particle will pass through a unit area located perpendicular to the beam in front of the scatterer is There is no harm in extending this integration from . as if no scattering had occurred. the latter is called the scattering amplitude. There is then strong dispersion.vot. The factor l/J(.ko .12).vot. Le.2 The Scattering of a Wave Packet and consequently. as expected for uniform particle motion. 289 l/J(r. made in deriving (13.
The transformation to an integral equation is performed most efficiently by regarding UI/J on the righthand side of (13.5) and setting NT = ntA = 1 for a single scattering center. Formally.:.r') (13.25) Indeed. by (13.23) establishing the fundamental connection between the scattering amplitude and the differential cross section. the expression  4~ JG(r. it is (a) eigenfunctions of the asymptotic form (13. r') = 41TlS(r . even though it contains the unknown function I/J./J. and (b) will find its answer after we have constructed the solutions I/J~+)(r). Hence.290 whereas 10 = ~ N p Chapter 13 Scattering J+oo I. To this particular solution we may add an arbitrary solution of the homogeneous equation (13.(gko. 0)1 00 2 dg gives the number of particles incident per unit area. Green's Functions in Scattering Theory.24) or = 2mElfi 2 and U = 2mVlfi 2 • It is useful to replace the differential equawhere tion (13.24) is conveniently constructed in terms of the Green's function G(r. r'). The SchrOdinger equation to be solved is ( . and (b) the expansion (13.24) temporarily as a given inhomogeneity. The method of the Green's function by which this may be accomplished is far more general than the immediate problem would suggest. This result is independent of the detailed behavior of the projectile wave packets and the normalization constant N. necessary to show that To complete the discussion.27) .13) exist. These two problems are not unconnected. which is a solution of the equation (V2 e + TC)G(r. V 2 + V) I/J = EI/J (13. ' 3. r')U(r')I/J(r') d 3 r' (13.26) solves (13. then. a particular "solution" of (13. we arrive at the fundamental result: (13.24) by virtue of the properties of the delta function.15) is correct.24) by an integral equation.
Leaving a normalization factor N undetermined.29) are solutions of the inhomogeneous equation (13.3 Green's Functions in Scattering Theory 291 which is the SchrOdinger equation for a free particle (no scattering). the Fourier representation G(r) = f _1_ k 'z 1 2'17"2 1 ITTT = ~ e ik. we obtain.r'I (13. If we introduce the Fourier integral G(r) = f g(k')e ik r '. Nevertheless.. Even if a particular vector k is selected.32) Integrating over the angles. which corresponds physically to the possibility of choosing an arbitrary direction of incidence. we obtain by substitution into (13. and there are infinitely many different ones. the convenient form G(r) = . d 3 k' and the Fourier representation of the delta function.33) Since the integrand has simple poles on the real axis in the complex k' plane. This suggests that our attempt to represent the solutions of (13. r ' ) = G(r . at k ' = ±k.25) as Fourier integrals has failed.25). the integral in (13. (13.25). r d 3 k ' (13. r') imposes definite boundary conditions on the eigenfunctions o/k(r).33) does not really exist. . J+= k'z~ k' dk' eik'r  (13.25).4. The vector k has a definite magnitude.24).31) I g(k ) = 2'17"z k'z _ kZ 1 1 hence.28) is by no means completely defined yet: The Green's function could be any solution of (13.29) Exercise 13.r') (13. _ exp( ± ik Ir . The choice of a particular G(r. fixed by the energy eigenvalue. after a little algebra. but its direction is undetermined. Thus. Verify that the Green's functions (13. Two particularly useful Green's functions are ' G± (r. we thus establish the integral equation (13. A host of Green's functions of the form G(r. the solution exhibits an infinite degree of degeneracy.28) for a particular set of solutions of the SchrOdinger equation (13.30) may be obtained by applying a Fourier transformation to the equation (V 2 + ~)G(r) = 47TC5(r) (13.31) which is a simplified version of (13.r I I) Ir . r )  .
_e_ik::'.L. Alternatively.. The result of the integration is not altered by introducing detours avoiding the two points k' = +k and k' = k.34). we might say that a unique solution of (13. after the integration has been performed.. . Path of integration in the complex k' plane.00 to + 00. I .Re k' X..34) where 1'/ is a small positive number.5.33) can be replaced by another one that does exist. " dk (13.." /"""" .. . because the homogeneous equation (V 2 + e)l{I = 0 does have non+ i1'/) exists and is the trivial solutions. as indicated Imk' . because the integral in (13.. .. thus G+T/(r) = . .. So defined...) The integral e f k' + 00 ::.31) does not exist. Since r is necessarily positive. " ". The poles of the integrand are at = +VJ?... the inverse of (1/47r)(V 2 + Green's function G+T/(r) in (13. I I . I I 'L_:I. . But this operator has no inverse.r. .eik'r k... \ I I I I f \ o \ \ \ . "". . . The trick is to evaluate the expression (13. .. G+T/(r) ~ G+(r).". for if it did G would be the inverse of the operator (1/47r)(V 2 + k 2 ).+  i1'/ "'" +(k  + 2k i1'/) for small 1'/. However. (See also Section 18. It encloses the pole in the right half plane..+r_.... .34) for 1'/ =1= 0 ana to let 1'/ ~ 0. . The path of integration leads along the real axis from . " . I 17rr  f+oo 00 .. ... . Le..2  (.292 Chapter 13 Scattering this approach has the potential to succeed.2 K + 11'/) k ..5). . .. 00 k'2  (e + i1'/) k' dk' is most easily computed by using the complex plane as an auxiliary device (see Figure 13......31). . G+T/(r) exists but is no longer a solution to (13.+.. k.17J 2k Figure 13...2. a closed contour may be used if we complete the path by a semicircle of very large radius through the upper half k' plane.
we obtain a second Green's function =  G_(r) . and standing waves. The Green's functions (13.34) by  1]( 1] > 0). We must show that in the asymptotic limit. Show that the Green's function G 1 (r) = 1 1 2 G+(r) + 2 G_(r) = coskr r (13.r' U(r')l/J~±)(r') d 3 r' (13. If. furthermore. r' in the denominator of the integrand in (13. r  ~ 47rr +ikr f e+ik'.38) may be identified as outgoing.r'l = kYr2  2r· r' + r.33) differently. K k' dk' (13.39) which is the Schrodinger equation rewritten in a form that is particularly convenient for use in scattering theory. Still other Green's functions can be formed by treating the singularities that appear in (13.2  .36). l/J(±)(r) .29). If r is chosen so large that ka 2 /r« 1 then the quadratic term in the exponent can be neglected.Jwo distinct eigensolutions result.36) or (13. They satisfy the integral equation.35) which by the use of the residue at k' = k becomes (13.28).36) If we replace 1] in (13.39) assumes the simple form (13. incoming. When the special forms (13. For large r the integrand can be closely approximated in view of the fact that U =1= 0 only for values of r' < a. 1 = "1+0 G+"I(r) = .2. and (13. the righthand side of equation (13.3 Green's Functions in Scattering Theory in the figure. as r 7 00. If this is done.r  _1 47r f exp(±iklr . Furthermore. k =.5.38) is obtained if the integral in (13.Ne ik . we obtain for large r: l/J~±)(r) == Ne ik. = 11m G_"I(r) "1+0 e ikr r (13. denoted by l/J(+) and l/J().40) .r'l k (13.r'l) U(r')l/J(±)(r') d 3 r' I r .37).2 = kr .37) are substituted in (13. To appreciate this terminology.. the limit and we may write G+(r) 1] 7 293 0 can be taken prior to the integration.13).lim 17rr f eik'r k. we expand in powers of r': klr . we need only multiply them by the time factor exp( iEt/h).kr· r' + k(r X r'f 2r + . (13.33) is replaced by its (Cauchy) principal value. In the exponent. Exercise 13.39) is neglected. the description of scattering in terms of wave packets suggests the designation retarded for G + and advanced for G_.37) in agreement with (13.
294 where we have set Chapter 13 Scattering k' = ki The asymptotic expression (13. corresponding to knormalization.42) 4~N f e+ik'. It can be seen from the integral equation for l/Jl.r + r.41) == e±ikr ) N ( e ik.43) e+ik'. assertion (b) is proved.r o)] d 3 k = N1l/J(kolr . Hence.±)(r) (r large) (13. we may choose N = (2~) 312.20). The outgoing solution is indeed asymptotically of the form (13. completing the discussion begun in Section 13.r' I points to a position behind the scatterer where the wave packet was assumed to vanish at t = O. the lefthand side of (13.r'I. r I) U(r')l/Jl.±)(r') d 3 r' __ m_ 2~h2N f (13. In Chapter 20 it will be shown .r' I = k o . because the vector ko Ir .44) It may usually be assumed that in this integral the variati~n of l/Jl. are called the spherically outgoing and incoming solutions of the SchrOdinger equation: They satisfy the appropriate boundary conditions at infinity.fl. Although the value of N does not affect the results.44) is nearly equal to l/J~)(r') f cf>(k) exp[ik· (kolr . k Ir . Again we note that the wave packet cf> (k) is appreciably different from zero only for vectors k near the direction of k o. It is customary to omit the superscript symbol (+) qualifying the outgoing scattering amplitude and to write fk(r) for fl.±)(r) = = (13.40) can be written as l/Jlt)(r) where fl. To prove assertion (b) we must employ the exact form (13. thus verifying assertion (a) at the beginning of this section.r'U(r')l/Jl.42) shows why l/J<+) and l/JH.O) l/J~)(r') The righthand side of this equation vanishes. Most ar physical situations meet these conditions.39) and demonstrate that for the initial wave packet f cf>(kLexp(ik· r o) d 3 k = f eXPf~kl~ . so that we may approximate as in (13.+) that this is true if the width of the wave packet » a and if we are not at an inordinately narrow resonance for which the scattering amplitude varies extremely rapidly with k. k Ir .+)(r') with k can be neglected.r'V(r')l/J<±)(r') d 3 r' k Equation (13.r' I Hence.r'l .+)(r') d 3 r' = 0 Since U(r') 0 for r' > a.+)(r). it is sufficient to show that for r' < a (13. when supplemented by exp(iEt/h).13).2..
.6.k' is a reciprocal (13. This relation is the Laue condition familiar in condensedmatter physics. R == (k . da dO where the scattering amplitude fk(r) is defined as the coefficient of the outgoing wave in the asymptotic solution l/Jf+)(r) == N ( eik.47) 3See Christman (1988). Interpret this result. prove that l/Jf+l(r) and l/J~J(r) are mutually timereversed scatteringstate solutions of the Schrodinger equation. Section 4. with this normalization. R = 27m k . 3 4. Before we proceed.7.r + .k') . where R is a constant vector. If the potential V is realvalued. . let us summarize the results of the preceding sections.46) where n is an integer. The Born Approximation.Rl/Jf::!:. (a) prove that the scattering solutions l/Jf::!:.4 The Born Approximation 295 that. Exercise 13.14) which satisfies the condition q .fk(r) eikr ) (r large) of the SchrOdinger equation The scattering amplitude for elastic scattering in the direction the formula k' is given by (13.) of the integral form of the SchrOdinger equation are Bloch wave functions.23). If the scattering potential has the translation invariance property VCr + R) = VCr).2. the energy eigenfunctions l/{+)(r) are orthonormal.45) When supplemented by the boundstate energy eigenfunctions.. If particles with an average momentum hk are incident upon a scatterer represented by the potential VCr).)(r + R) = eik. Exercise 13. the differential cross section is given by (13. since (13. since they satisfy the relation l/Jf::!:. they make up a complete set of orthonormal functions .)(r) = and (b) show that the scattering amplitude vanishes unless q lattice vector (Exercise 4.
). neglecting the scattered wave. As the Fourier transform of the potential.49) resembles the field amplitude in Fraunhofer diffraction. k'=k+q k '\\\ \~\ ~v Figure 13. the Born scattering amplitude (13. .296 Chapter 13 Scattering but since l/I<+) appears in the integrand.48). the Born amplitude (13. of the potential between eigenstates of incident momentum k and scattered momentum k' = k + q (in units of h). where the plane wave is substituted for l/I(+) under the integral. It should be evident that the Born approximation corresponds to a first iteration of (13.40). However.) For a centralforce potential. depicting the first Born approximation scattering amplitude (13. Figure 13. this is not an explicit expression.49) where q = k' . V(r). Here the scattering amplitude appears proportional to the matrix element of the scattering potential between the plane waves. (13.3. representing the free particle before and after the scattering.48) which is known as the scattering amplitude in the (first) Born approximation. (See Section 20. (k' I vi k). The scattering angle is O.40).k is known as the 'momentum transfer (in units of fl. In this approximation we obtain (13. it can be used to obtain an estimate of the scattering amplitude if we replace the exact eigenfunction l/I<+) in the integrand by the normalized plane wave term on the righthand side of Eq.6.48) reduces to (13. r . The potential V causes the momentum transfer q. r and e ik '. e ik . Diagram representing the first Born approximation matrix element. The iteration may be continued to obtain higher order Born approximations.6 is a primitive Feynman diagram.
52) where the length 1/a may be considered as the range of the potential.52) for a ~ 0 and Va ~ 0 with Vola = qlq2' Hence.53). Exercise 13.qr' r .4 The Born Approximation 297 The integral over the solid angle in (13.8. however.VOerla (Vo > 0) . and note that k' = k for elastic scattering.ar' sin qr' r . Exercise 13.51) 2 as seen in Figure 13.9. is in exact agreement with both the classical Rutherford cross sections (13. we find f Born(e) =  2m .2 dr . sin .l:2 n Joo V(r ') . in the Born approximation. even though obtained by an approximation method. Apply the Born approximation to the scattering from a square well.l:2 n v. consider the screened Coulomb (or Yukawa) potential e. after a simple integration over angles.a qr (13. . As an application. In the Born approximation. . Calculate the total scattering cross section for the screened Coulomb potential (13. so that " q = 2k sin ~ (13. Obtain the differential scattering cross section in the Born approximation for the potential V(r) = . Note.50) Here we denote scattering angle between k and k' bye. m2qIq~ 4p 4 sin\el2) This result. du dO.ar V(r) = Vo ar (13. Exercise 13.6.53) _ 2m V. Evaluate and plot the differential and total scattering cross sections.0 Joo e.8).2 0 ar qr dr ' (13. 1 h 2a 0 ~ + a 2 = _ 2mVo 1 h 2a 4k2 sin2 (e/2) + a2 The differential scattering cross section is obtained by taking the square of the amplitude (13. that the exact Coulomb scattering amplitude differs from the Born amplitude by a phase factor (see Section 13. The unscreened Coulomb potential between two charges ql and q2 is a limiting case of the potential (13.49) is easily carried out and yields the result f Born(l=]) =  2m .9) and the exact quantum mechanical evaluation of the Coulomb scattering cross sectionone of the many remarkable coincidences peculiar to the pure Coulomb potential.10..52) in the Born approximation and discuss the accuracy of this result.
k(r) e R ek(r)Ye . Let us assume that V is a central potential.:(13.k(r) satisfy the differential equations I d ( r 2 d) [ .56) . and discuss the significance of this fact for the scattering of a wave packet.cp) .11.~ ] R (r) e. Hence. representing an incident and a scattered wave. Show that for a central potential o/k+) is an eigenfunction of the component of L in the direction of k. 5. it has enjoyed great popularity in atomic and nuclear physics. if V is a function of r alone. since the term that is neglected is itself an integral that depends on the potential.ym«().k = 0 (13.39) that o/k+)(r) depends on k and r and on the angle () between k and r only. Its usefulness does not.55) which are common eigenfunctions of H. it is easy to see that the Born approximation is a form of perturbation theory.54) It is desirable to establish the connection between these solutions and the separable solutions (12. m«() _ Ue. should exhibit cylindrical symmetry about the direction of incidence. however. and particularly on the momentum of the scattered particle.cp) . O/k+) is a function of r and the scattering angle (). We must thus look for solutions of the SchrOdinger equation h2 h2~  2m "\720/ + V(r)o/ =  2m 0/ which have the asymptotic form eikT) O/k+) == N ( e ikT cos IJ + fk«() . The Born approximation affords a rapid estimate of scattering cross sections and is valid for reasonably high energies in comparison with the interaction energy. The approximation is thus likely to be valid for weak potentials and high energies.2 r dr dr + e(e r+ 2 1) 2m + h 2 V(r) . Exercise 13. It is to be expected that for a spherically symmetric potential the solution (13. vitiate the need for an exact method of calculating scattering cross sections. What values of n give a meaningful answer? A reliable estimate of the accuracy of the Born approximation is in general not easy to obtain.k(r) and ue.. without loss of generality. we may. Discuss the reasonableness of the result qualitatively. If V = C/r n . and L z • The radial functions Re. the wave function in the region where the potential does not vanish. r (13.1ndeed. we can also see formally from (13. To this task we must now attend.298 Chapter 13 Scattering Exercise 13.3) of the centralforce problem. Because of its simplicity. assume that k points in the positive z direction and that for a given value of k. obtain the functional dependence of the Born scattering amplitude on the scattering angle.39).12. L 2 . in which the kinetic energy operator is the unperturbed Hamiltonian and the potential is the perturbation. Partial Waves and Phase Shifts.. Qualitatively. with eigenvalue zero.
(12. we get for large kr _ Ue. or of k. In general. the magnitude of Be compared with A e is. Each phase shift is. The radial equa~ion for the external region r > a.60) can now be written as Ue.k(r) = 0 (13. The name e scattering phase shift is thus appropriate for ae' particularly if we note that (13.60) In the complete absence of a scattering potential (V = 0 everywhere).k(r) = sin (kr .62) The phase shift a measures the amount by which the phase of the radial wave e function for angular momentum e differs from the noscattering case (Be = 0).58) We will restrict ourselves to potentials that are in accord with this boundary condition. and we would have Be = 0 for all values of e. Since for a realvalued potential Ue.e'TT/2) cos (kr . subject to the boundary condition.k(r) be finite at the origin. this boundary condition depends on the shape of V. the boundary condition at the origin would exclude the irregular solution.5 Partial Waves and Phase Shifts 299 and d2 [ .20) and (12.58).57) respectively.a measure " of the intensity of scattering. To do this. we must know V explicitly and solve (13. a function of the energy. these parameters are realvalued numbers. as we saw in Section 12.22).= A e k .59) Using the asymptotic approximations (12. if necessary.k(r). in most practical cases it reduces tp the requirement that the wave function Re.. and we may set Be A = e tan a e (13.k(r) may be assumed to be real.e'TTl2) R e k(r) . r r kr (kr large) (13.. and by joining the interior solution smoothly onto the exterior solution (13.Be .9) which was solved in the last chapter. is identical with Eq. but. of course.e'TT/2 r kr + ae) (13.4.k(r) _ sin (kr . Hence.57) by numerical methods. The value Be/A e must be determined by solving the Schrodinger equation inside the scattering region (r < a). The general solution of this equation is a linear combination of the regular and irregular solutions and has the form r> a (13. from which it follows that (13. A very useful expression for the cross section can be derived by introducing the ratios Be/A e as parameters. where the scattering potential vanishes. as well as a boundary condition at the origin.59) at r = a. . The value of u~iO) implies a choice of the normalization of ue. (13.dr2 + e(e r2 + 1)' + h2 VCr) 2m  Tr ] Ue.61) where a is a real angle that vanishes for all e if V = 0 everywhere.
12).67) Substitution of these values into (13. Comparing now the coefficients of the outgoing spherical wave e ikr . + l)(1)epe(cos 0) ikr (13.55). (13. (Hint: Use the radial SchrOdinger equation and integration by parts.300 Chapter 13 Scattering Exercise 13.54) and (13.k(r)ue.e2~kikr e=o (U zr + e:' kr 2zk re=o ± (2. The expansion coefficients Ce(k) can be determined by comparing the two asymptotic expressions of the wave function.. peccos 0) (13. (13.(. Show that (13. 0) depends on the angle 0 between k and r and not on the directions of each of these vectors separately (see Exercise 13.13.65) and (13.in (13. 0) == N e=o ± (2. we obtain r fk(O) = L (U e=o ~ + 1) tlille . with which (13.:~/2 + 8e) Pe(cos 0) r (13. l/J~+)(r.65) r ± + l)Pe(cos 0) + fk(O) e ] On the other hand.66). or the scattering amplitude.37). we see that the expansion coefficients must be of the form (13. (12. through the phase shifts we must expand l/J~+)(r. + lWeille sin (kr .64).68) This differs from a plane wave by the presence of the phase shifts and is called a distorted plane wave. which (except for any bound states) are assumed to constitute a complete set of orthonormal eigenfunctions.66) By comparing the incoming spherical waves in (13.k(r) == Ce(k)PeCcos 0) sin (kr .62) implies the normalization Jo (~ ue.66). Thus.65) and (13. (13. we set l/J~+)(r..63) for the radial eigenfunctions. 0) = CeCk)PeCcos 0) ue.64) where use has been made of the fact that l/J~+)(r.(.k') (13.e~/2 e=o r e=o r ± ± + 8e) (13.64) gives the asymptotic expression for the wave function.I 2zk .69) .64) takes the equivalent form l/J~+) == .54) can be written as l/J~+) == N[.. We make use of the asymptotic expansion of the plane wave.) In order to express the differential scattering cross section.~ L CeCk)ieei epeCcos 0) + ~ e=O Ce(k)( _Oeeillepe(cos 0) L 21kr e=O 2zkr 6 ikr 00 ikr 00 .dr) dr = 2~ ~ 8(k . 0) in terms of the separable solutions of the form (13.
If we remeritber that each term (partial wave) in the sum (13.7 for a better estimate of Be. Thus.72). u e. as is the case for scattering at low energies. Such a sum is often said to be a "coherent" superposition of different angular momenta. since Po(cos fJ) = 1. we obtain the total scattering cross section: = f du 0 dO d = . except for Bo(k) corresponding to = O. to scatter more intensely than the high angular momentum components. then we may expect the low angular momentum components. and the scattering amplitude must become independent of the angle fJ.) We conclude that the partial wave sum (13.71). But this argument is of a more general nature. the formula may be seen in a more physical light. which is therefore said to constitute an "incoherent" sum of partial wave contributions. which classically correspond to small impact parameters and therefore close collisions. this semiclassical argument suggests that if the impact parameter b = eli = ~ p k exceeds the range a of the potential.:2 LJ K 47T ~ e=o (2f + . we expect the phase shifts Be for > ka to be vanishingly small. No such interferences occur in the integrated total cross section (13. or angular distribution as it is often called. Hence. if ka » 1. or when e > ka. By (13. all spatial sense of direction is lost. Then the incident waves are long. and at a given instant the phase of the wave changes very little across the scattering region. we may attempt to determine the phase shifts as functions of k empirically by comparing scattering data at various energies with the formula e e :~ = !!k(fJ)j2 = ~ I~o (2e + l)e U 2 i8e (k) sin Bik) Pe(cos fJ)1 (13. e .72) In the differential scattering cross section (13. 1) sm2 Be (13. no appreciable scattering occurs. For suppose that ka « 1. if on physical grounds only a few angular momenta are expected to contribute significantly. More quantitatively. making the classical argument applicable.70) is a particularly useful representation of the scattering amplitude. If the scattering potential is strongest at the origin and decreases in strength as r increases. When this occurs we say that there is only Swave scattering.70) this implies that all phase shifts vanish. (See Section 13. contributions from different partial waves (angular momenta) interfere with each other.70) corresponds to a definite value of angular momentum e. if V is not known beforehand.71) By integrating.5 Partial Waves and Phase Shifts or 301 (13. and this is isotropic.70) This important formula gives the scattering amplitude in terms of the phase shifts by making what is known as a partial wave analysis of the scattering amplitude. In fact. because the scattering amplitude is a sum of terms with different values.
nika) sin 8e (13.n~(ka) sin 8e jika) cos 8e . Indeed.i(kreTrI2).59) and (13. Finally.74) This formula.68). and the incoming' wave.73) except for a factor 4. Show that the scattering amplitude (13. e i(kreTrI2). or I:1b = 11k.6). The theoretical determination of phase shifts requires that we solve the SchrOdinger equation for the given potential and obtain the asymptotic form of the solutions. is the ratio of the coefficients multiplying the outgoing wave. we note that the quantity Sik) = e2ifh(k) (13. Exercise 13. applying (13.69).75) which appears in the scattering amplitude (13.72) are related by the identity = (J" Tim A(O) 47T (13. the phase shifts can be expressed in terms of the logarithmic derivatives at the boundary r = a: f3 e  (~dRe)r=a R dr e Using (13. For large this agrees with (13. e.7) we obtain for the contribution to the total cross section from a range l:1e = I. e.61).14. holds for collisions in general (see Section 20.59) at r = a. Determination of the Phase Shifts and Scattering Resonances.76) .302 Chapter 13 Scattering Each angular momentum value contributes at most a partial cross section (13.70) and the total cross section (13.5 the quantities Se will be identified as eigenvalues of the S matrix. The difference is due to the inevitable presence of diffraction effects for which the wave nature of matter is responsible. known as the optical theorem. if we use the estimate b = fJk for the impact parameter.73) to the total scattering cross section. in the wave function (13. This value is of the same order of magnitude as the maximum classical scattering cross section per unit h of angular momentum. 6. In Section 20. If the interior wave function is calculated and joined smoothly onto the exterior solution (13. we easily find the desired relation between the phase shift and the logarithmic derivative at r = a: f3ik) = ka j~(ka) cos 8e .
. .79) we obtain the simple relations e2i(8 eSe) = f3e ..::. ie . ueCr) .61) that (13. lne = ~e + iSe (13. The first factor ie .80) and (13.. Re(r) = .77) defines the (real) phase angles ge which have a simple and interesting interpretation.ka i. which implies that ReCa) = O. The quantities ge are seen to be the phase shifts if f3e ~ 00..23) we deduce that 1 (13. e ie + ine (13..x.82) which shows that Se is positive for all e.in e le .iSe (13.~e + iSe f3e . The radial wave functions vanish at r = a for all values of e if the potential represents a hard spherical core of radius a. ka. If we introduce the real parameters ~e and Se. and plot their dependence on ka for the lowest few values of e. + in.81) which make explicit how the partial wave contributions to the scattering amplitude depend on f3e.ine ie + ine (13.~n.78) gives the phase shifts for a hard sphere. From (13... Let us apply these results now to the special case of the square well of range a and depth VQ • From (12. so that the wave function cannot penetrate into the interior at all.::.~e .15. Exercise 13.77) where all the sph~rical cylinder functions Cie and ne) and their (primed) derivatives are to be evaluated at the argument ka..."ie + ine f3 _ ka i .= leCk r) r I .78) in the expression (13.41).. f3e .6 Determination of the Phase Shifts and Scattering Resonances 303 and conversely.79) and (12.lne ''. i'e + in' le + . The angles ge are therefore referred to as the hard sphere phase shifts.59) and (13.::... Verify directly from (13.
80).z (13.86) while the quantities ge.e) = . and the partial cross section approaches its maximum value.83) For e= 0. Figures 13. B1> and Bz. J. we get EE . So = ka. f30 = k'a cotan k'a  The logarithmic derivative f30 is a monotonically decreasing function of energyan important property that can be proved more generally for any potential and all f3e.304 where H' = v'2m(E Chapter 13 Scattering + Vo). Since it passes near the value 31T/2 in this energy range. illustrate some important common features of scattering cross sections. and may be regarded as constant. when the particle is effectively free.8. f3iE) = c + bE (13.. calculated for a particular square well.81). but the figures also show that the smooth variation of the phase shifts and cross sections is interrupted by a dramatic change in one of the phase shifts and a corresponding pronounced fluctuation in the partial cross section.84) In the limit E ~ 0. the partial cross sections tend to decrease with increasing angular momentum and increasing energy. z (T1 = 121T/k • It may happen that in a small energy range a rapid change of the logarithmic derivative f3e can be represented by a linear approximation. sin z B1 becomes close to unity. which are determined only to within multiples of 1T..i k  e 'ka . Hence.16.i o 2 r r eZi(Se(. for the particular values of the parameters on which Figures 13. the Pwave phase shift rises rapidly near ka = 0.85) Exercise 13.7. we see that the Swave scattering amplitude is fo Zika 1 . vary slowly and smoothly with energy. and Se. Ll e. we find from (12. If this approximation is substituted in (13.. because the presence of the centrifugal potential makes it improbable for a particle to be found near the center of force.7 and 13.) sin ka (13. which characterize the external wave function. (13. Applying these results to (13.7 and 13. Compute g1> Ll1> S1> f31> and f1 for Pwave scattering from a square well and examine their energy dependence.. k ~ 0.(k'a) f3e = k' a Jik' a) (13. At low energies.. Generally. P waves (and waves of higher angular momentum) are scattered less than S waves.8 are based. this gives the nonvanishing isotropic Swave cross section: (To ~ 41Ta ( ~ z tan kha 1) z for kh = 2mVo ~ .87) EEo +i2 . e= .e'Sa sin Bo = . Thus. The phase shifts Bo.31) the simple expressions Ll o = 1. were normalized so as to go to zero as E ~ 00.30) and (12. (k  k k k ' cotan k' a .
L_ _L_=::::±=:::::::::_..S ka Figure 13..2 = 6.._ _..:::.7... BI> B2 ) for scattering from a square well of radius a with kba = Y2mVoa2/fi.==l. The cross sections are given in units of 'Tra 2 • .6 Determination of the Phase Shifts and Scattering Resonances 305 0 1 ..8._ _ o l. P.....l. 30 20 10 oO... and D phase shifts (Bo.7 S...2..:=.. 2 1 3 4 ka s Figure 13.L..LL.. Momentum dependence of the partial cross sections (ao and al) for Sand P waves corresponding to the phase shifts of Figure 13...
89) reduces to B _ tan er/2 .91) because the conditions under which this formula was derived (ge « 1. with Eo being the resonant energy and r the width of the resonance. They are experimentally resolvable if the width r.E Eo The resonant term in (13.89) Since f3e is a decreasing function of the energy.g.E .8. By its definition. Such sudden variations in the phase shifts are resonances.90) giving a neat separation of the resonant part of the partial wave amplitude from the nonresonant part.e ' e .87) and (13. which is proportional to sin 2 Be. the contribution of the corresponding partial wave to the scattering amplitude can be written according to (13. If the phase shift Be is near resonance.b .and "2 r_~ b (13. which depends only on ~e' If as is the case for low energies or high angular momenta. it can be seen from (13. Since we have discussed phase shifts and resonances in this section entirely in terms of the logarithmic derivatives at the boundary of an interior region within 4The Pwave resonance in Figure 13..87) that the phase Se 2(Be . while narrow.86) is accurate over an energy range large compared with r. r if ~e is also nearly constant in this interval.I rl2 + e'.~e) changes by 21T as E varies from E « Eo  r "2 to E » Eo + "2' Hence.88) as Ie = . the phase shift Be changes by 1T. (13. the hard sphere phase shifts are negligible.306 where Eo = Chapter 13 Scattering il e . The expressions (13.81). is positive.c . b must be negative.88) is positive. and the partial cross section U e. Alo and SI reasonably constant) are not well satisfied. 2e + 1 k 2'g ( r /2 .86) and (13.88) Hence. changes abruptly. Eo . we arrive at the very simple approximate relation r (13. is still wide enough so that the particles in the beam may be represented by spatially broad wave packets with ilE « r. and it thus follows that the quantity r defined in (13.90) predominates then and contributes to the total cross section an amount (13. Under these circumstances. . 4 Profiles with this energy dependence are called BreitWigner cross sections.)Plcos 6) e sm ~e (13.8 is not accurately described by the simple formula (13.91) For a small width r.89) are useful if Eo and r are reasonably constant and if the linear approximation (13. (13.. this represents a sharp maximum centered at Eo with a symmetric shape similar to that of the transmission resonance peaks in Figure 6.
In Sections 6. it interacts in a complicated way with all the nucleons and its reemission may thereby be considerably delayed. so that AE » f. To simplify the discussion.4. we evaluate the integral and obtain = f'7Ti {o 'I'scatt . asymptotically. if some mechanism impedes the escape of the' wave from the box.93) As in (7. fixed potential V. ro = kro J+OO and k· vo = kv o fl2 we find that the scattered wave in Eq. Using k . the probability of finding the particle at that location decreases exponentially with time.Eo U + 1 f/2 .2. when a slow neutron enters a complex nucleus. At low energies the neutron scattering cross section thus exhibits sharp maxima that may be regarded as the remnants of the discrete level structure which exists for E < O. In particular. where transmission resonances were considered. Rather.69) and (A. without affecting any qualitative conclusions. AE. The scattering amplitude is taken to be fee) = . (13.19) is. (See the analogous onedimensional illustration in Figure 7. but in nuclear and particle physics resonances come about through a more complicated mechanism.E .6 Determination of the Phase Shifts and Scattering Resonances 307 which the scattering forces are concentrated.t)] 2ft Do Dot> r + ro (13. our conclusions are independent of the· particular mechanism that operates inside this region. wellcollimated wave packet involving a broad energy range. The analysis of Section 13. exp[~ (~. The interior region is a "black box. A narrow resonance corre .8).k . To see this. proportional to the expression l/Jscatt oc: f oo eik(r+rovot) o r f(e)~ dk oc: ei(EEo)(r+rovo')lfwo 00 E . the scattering may arise from much more complicated interactions within the black box.22) in the Appendix. The reciprocal width of a resonance IIf is a direct measure of the stability of the resonant state or of the delay between the time of absorption of the particle by the black box and its reemission. we saw how a precipitate change· in the potential or a high potential barrier can produce narrow resonances. and these are resonant energies.Eo + l fl2 • dE (13. there is no need for the phase shifts to result from an interaction with a simple." characterized only by the values of the logarithmic derivatives at the surface.1.92) The wave packet is represented by a momentum wave function (jJ (k) that is appreciably different from zero only for k in the direction of incidence but that has an energy spread AE» f. we now assume that we have a very narrow isolated fwave resonance and that all nonresonant contributions to the scattering are negligible. we now consider the extreme case of a very narrow resonance centered at energy Eo and a sharply pulsed.i f/2 Picos 0) (13. once the scattered wave packet has reached a sphere of radius r.94) Dot < r + ro Thus.4 and 7. such metastable states can occur only at certain discrete energies. For ex· ample.68)(7. and resonances may occur as a result of constructive interference at certain discrete frequencies. which assumed a slow energy variation of the scattering amplitude is no longer applicable. Owing to the wave properties of matter.
18. familiar in electrostatics. Since G+(r. r') is a solution of the equations (V 2 + ~)G+ =0 and (V. 7.17.if/2. If the variables E and k are analytically continued into the complex plane.308 Chapter 13 Scattering sponds to a situation in which the incident particle spends a long time. Exercise 13.ifl2 is a simple pole of SeCk) = e2ia. Comparison of the last two equations yields the values of qe(k) and the desired identity: G+(r. Although the relation (13. (13. Eq. in the interaction region before being scattered.2 + ~)G+ =0 if r =1= r'. of the order of the mean lifetime.95) can be determined by letting r """ r' """ 0 (but still r> r').96) 00 . Phase Shifts and Green's Functions. This connection can be elucidated if a partial wave analysis is applied to the integral equation (13. r')je(kr')h~l)(kr) A A ~ (r > r') (13. To this end.20 (2f + ~ qeCk) r.39). Exercise 13. estimate the mean lifetime of the metastable state responsible for the Pwave resonance in units of the period associated with the motion of the particle in the square well.r' I) Ir _ r'l = ik l~O (2f + l)Pe(i' . which appears in the denominator of the scattering amplitude as a complex energy of the resonant state. This has to be compared with the expansion. The remaining unknown coefficients. r') = exp(ikl r . By using the first approximations (12. which comes from the generating function (11.e Ai") l)ik r e+ 1PeCr .68) for Legendre polynomials.r'l . in the expansion (13. (13. Show that for a resonance the quantity fi(d8idE). the outgoing Green's functions must first be expanded in terms of Legendre polynomials. r') = exp(iklr .8.95) where the particular choice of the spherical cylinder functions (Bessel and Hankel functions of the first kind) is dictated by the regular behavior of G+ at r' = 0 and its asymptotic behavior. it is Seen from the separation of these equations in terms of spherical coordinates that the partial wave expansion for r > r' must be of the form G+(r. Using Figure 13. It is possible to think of the quantity Eo . it does not shed any direct light on the dependence of the 8l on the scattering potential.20 qeCk)PeCr .87) shows that Eo .r'l) Ir _ r'l =. qeCk). T = Mf.(k) and that the state with this complex E value has no incoming wave.95) simplifies to 1 Ir . r')je(kr')h~l)(kr) (r > r') (13.76) between phase shifts and logarithmic derivatives of the radial wave functions is a very general and simple one.17) and (12.28) for Bessel and Hankel functions. evaluated at E = Eo is a measure of the lifetime of the metastable state. G+ ~ eikr/r as r ~ 00.
these coupled equations may be solved by successive approximation in an iterative procedure. fJ) into the integral equation (13. From this formula we deduce the rule of thumb that.k(r) = r cos Se je(kr) + kr f: je(kr')ne<kr)ue.100) may be approximated by the first term in their power series expansion [see (12. Se ex: k 2e + 1.47) gives the same estimate as (13.e > ka.39). Exercise 13.97) Letting r ~ 00 al}d replacing ue. The zeroth approximation to the wave function is u~?~(r) = r cos SeJe{kr) When this wave function is substituted in (13.98) by applying a partial wave analysis directly to the scattering amplitude (13.7 Phase Shifts and Green's Functions 309 We now substitute (13.e + 1)1]2 2e ~Hl (00 U(r')r.2H2 dr' Jo (13. For values of.100) if Se « 1. we finally arrive at the simple for!Dula sin Se = k (00 Je{kr')U(r')ua(r')r' dr' Jo ' (13.43) provided that VCr) is a central potential.64) of «/Ji+)(r.ir) satisfies the integral equation ue.k(r')U(r')r' dr' Exercise 13.21.k(r) and the cylinder functions of argument kr by their asymptotic expansions.20.96) and the partial wave expansion (13.98) and (13. .67).101) as an estimate for the phase shifts.k(r')U(r')r' dr' + kr 1 00 (13. Show that a partial wave analysis of the scattering amplitude in the first Born approximation (13.19.'r = Je{kr) . Verify (13. using (13. Show that for all values of r the radial wave function ue.k(r')U(r')r' dr' e (13. For r > a we obtain the radial integral equation Ue k(r) e'"Il € .e!)2 [(2.ike'"Il. Some useful estimates of phase shifts may be based on (13. but this is usually cumbersome. the spherical Bessel functions in the integral (13. For instance. l' 0 je(kr')h(1) (kr)ue.17)].99) je<kr)ne(kr')Ue.98) c This is an explicit expression for the phase shifts in terms of the potential and the radial eigenfunctions.99). we get the approximate phase shift (13. for low energies and high angular momenta. and carry out the integration over the direction of r'. if the potential is not strong enough to produce a resonance. Exercise 13.100) Higher approximations may be obtained by iteration.98). We thus obtain tan Se "'" 2 (.
] '" + . the total scattering cross section is infinite. Eq.c2i2 C2. is scattered.58)]: d h [ 4. proportional to exp(ikr) = exp[ik (~ + 'T/)/2] (13. For example. a particular method of solving the problem will be presented because of its intrinsic interest. d d f2 d'T/ (TJ d'T/ ) e +"4 'T/i2 Cl . Scattering in a Coulomb Field.8. and the constants of separation.) Owing to its great importance in atomic and nuclear physics.109) . Since a Coulomb field.~ + TJ d'T/ ( TJd) d'T/ 1 d 2qlq2 + . Hence. In this section.102) The SchrOdinger equation for the Coulomb potential. = arctan x (13.108) and a radially outgoing part.105) Separation of the parabolic variables ~ and 'T/ is achieved easily. no matter how large its impact parameter. many of the results that we have derived in this chapter under the assumption of a potential with finite range are not immediately applicable without a separate investigation. Exercise 13. exp(ikz) = exp[ik (~ .cdl = 0. This method depends on the observation that the SchrOdinger equation with a Coulomb potential is separable in parabolic coordinates. and that have the form (13. (A. V(r) = qlq2 r (13. (See also Exercise 13.9) diverges. has an infinite range.2 ~ + 'T/ 2 = E'" (13. we must look for solutions that are independent of q. because every incident particle.104) For scattering. and the SchrOdinger equation is replaced by two ordinary differential equations d dil d~ (~ d~ ) where h 2 k 2 condition k?+"4 gil .TJ)/2] (13.'" ~'T/ dq. the concept of a total scattering cross section is meaningless for such a potential.22. V = Clr. hence..107) Scattering is described by a wave function that asymptotically has an incident plane wave part.106) = 2mE.. and are related by the (13. which are defined by the equations 'T/=rz y q. =0 (13.103) describing the electrostatic interaction between two charges ql and q2' becomes in these coordinate$ [see Appendix. which we again choose to be the z axis. a vast amount of work has been done on the continuum (E > 0) eigenstates of the Coulomb potential.2m ~ + 'T/ d~ 2 ( ~d) d~ d + 4. Show that the total cross section obtained from the differential cross section (13.310 Chapter 13 Scattering 8. we are interested in a solution with axial symmetry about the direction of incidence.
We see that the solution of (13. (5.115) For purely imaginary argument leading termS IF1(a.c) 2 7T] (13.89) for the confluent hypergeometric function. quantum number n is defined.3. (13.106). defining an effective velocity v. is (13. z) z.113) to the positive energy unbound scattering states in a Coulomb potential.a) 2 7T) f(c) + f (a) Izlacez exp [ i(a . we see that the Ansatz (13. a generally noninteger.114) which should be compared with Eq. (12.112) For a stationary bound state. the asymptotic expansion of this function has the == f f(c) Izla exp ( ia (c .114). positive or negative.113) linking the principal quantum number n with the effective orbital speed v.110) is indeed a solution if we choose CI = / . which is regular at the origin.51). .116) 5Morse and Feshbach (1953).26) has been used. Eq. we obtain the relation n = _ qlqz hv (13. we know from Chapter 12 that for an attractive Coulomb potential.111) is conveniently transformed if we define a new function g( '1/): (13. c.8 Scattering in a Coulomb Field 311 The behavior of these two portions of the wave function suggests that we look for a particular solution with (13.k 2 The remaining equation for f z.110) Substituting this into (13. If the mean kinetic energy is expressed as mv zl2. If we apply the same defining relation (13. where in the last equality the virial theorem (Exercise 3. This allows us to cast the differential equation for g( '1/) in the form (13.
the scattered wave fails to approach the simple free particle form. Hence.un)] } (13.z = r(1 .u n unn n} (13.121) . cp) = exp[ik(g 7])l2hF1(in. ik'rJ} (13.117) where reI + in) = Ir(l + in)le iUn • Collecting all these results together. However. 4~ n2 sin4 (012) qjq~ 1 16£2 sin4 (012) (13. Applying this approximation to g( 7]).118) of the SchrOdinger equation has the asymptotic form rjJ == exp (. for all finite values of the scattering angle 0. It shows clearly why it is impossible to obtain for a Coulomb potential an asymptotic eigenfunction that has the simple form. Since the asymptotic contributions from the Coulomb potential to the phases are logarithmic functions of r.120) and the differential scattering cross section is du dO.WTT/2) { lrel + in) I exp{i[kz . eikr/r. we obtain for large values of g(7]) == exp (WTTI2) { lrel + in) I exp[i(n In k7] .z) + n In k(r . as r ~ 00. 7]. for which the Coulomb potential is responsible at large distances from the scattering center. it is important to note that the modification. The differential cross section is [see (13. affects only the phase of the incident and scattered waves.cos 8) cos 0)] (13.z) . and hence vary but slowly.119) for large values of r . fk(O) = n exp[in In(1  k(1 . Equation (13. we conclude that the particular solution rjJ(g. Similarly. the analysis of the fate of an incident wave packet in a scattering process can still be carried through. and the result is precisely the same as before.cos 0). 1. where 0 is the scattering angle. provided the incident (modified) plane wave is normalized so that the probability density is unity.23)] where the scattering amplitude fk(O) is again the factor multiplying the radial part of the asymptotically outgoing wave.z) + + n exp{i[kr k(r . The Coulomb force is effective even at very great distances and prevents the incident wave from ever approaching a plane wave. except in the forward direction.n In k(r .119) is valid at large distance from the scattering center. deduced for a potential of finite range.un)] + k7] exp[i(k7] n + n In k7] .312 Chapter 13 Scattering 7].
 e..124) was already discussed in connection with Eq. e2 7Tn = exp (21Tq 1q2 ) == ... Discuss other limiting cases (fast particles. Since IF1(a.52) for a onedimensional WKB model of Coulomb scattering. attractive poten Exercise 13. An angleindependent addition to the phase has been omitted in (13. as in the case of nuclear interactions with charged hadrons. It is then preferable to use spherical polar coordinates and attempt a phase shift analysis. tials) for (13.8 Scattering in a Coulomb Field 313 in exact agreement with the classical Rutherford scattering cross section (13.5 can be extended to include the presence of a longrange lIr potential. Exercise 13. 0) = 1. By expanding the eigenfunction (13. . and .123) The significance of the penetrability. the normalization of the eigenfunction (13.27Tn = 21Tn (13.20 .2 27Tn IljJ(ooW 1Tne (n« 0) (13. the incident wave in the large brace of (13. Calculate the wave function (13. c. the method of this section is not appropriate if the potential is Coulombic only at large distances but. we have as a measure of penetration to the origin. has a different radial dependence near the origin. The exponent G is called the Gamow factor in nuclear physics.122). )12 n7T m e l _ e.90) and the definition of the r function: r(1 + z) = zf(z).122) where in obtaining the last expression use was made of (8. 00 2 = IljJ()1 exp(.23. the penetration probability of finding the particle at the origin relative to the probability of finding it in the incident beam is 2 IljJ(0)1 = 1 ljJ(oo) 12 Ir(l + .119) is normalized to probability density unity far from the scattering center.24. What physical conclusions can you draw from its form? Since the SchrOdinger equation is separable in parabolic coordinates only if the potential behaves strictly as lIr at all distances.118) is such that ljJ = 1 at the scattering center (origin). (13.118) in terms of Legendre polynomials. If the Coulomb potential is repulsive (qlq2 > 0 and n < 0) and strong compared with the kinetic energy of the particle.n1T) 1f(1 + inW' Hence. and the theory of Section 13.. IljJ(OW . phase shifts can be calculated for the Coulomb potential.9) and the Born approximation.120).118) in the forward direction. On the other hand. (7.
. show that in the first Born approximation. Evaluate the total cross section and estimate the accuracy of the result. Also discuss what happens if the wavelength becomes very large compared with the size of the sphere. scattering occurs only when the momentum transfer q (in units of Ii) equals a reciprocal lattice vector G. where R is a constant vector.7). as in the exact formulation (Exercise 13. 2.314 Chapter 13 Scattering Problems 1. If the scattering potential has the translation invariance property VCr + R) = VCr). compute and display on a polar graph the differential cross section for an impenetrable hard sphere when the de Broglie wavelength of the incident particle equals the circumference of the sphere. Using the first three partial waves.
CHAPTER 14 The Principles of Quantum Dynamics If the Ha1lliltonian operator is known. (14. y. We now add the time parameter to the description of quantum states and generalize the fundamental postulate of quantum mechanics (Section 9. 1.1) by asserting that: The maximum information about the outcome of physical measurements on a system at time t is contained in the probability amplitudes (Ki IW(t). We now extend the general principles of quantum mechanics (Chapters 9 and 10) to the laws governing the time evolution of quantum systems. in terms of the amplitudes that specify the state. The only new feature here is that we now recognize formally that the state W is a function of time. if indeed it is so determined? Or. an initial wave function !/J(r. utilizing several equivalent dynamical pictures (SchrOdinger's. which correspond to a complete set of observables K for the system.2) The basic question of quantum dynamics is this: Given an initial state IW(to» of the system.)(K. and we shall assume it. t) according to the timedependent SchrOdinger equation (Chapters 2 and 3). The dynamical law that connects the initial and final amplitudes is contained in the further assumption that the composition rule (9. Heisenberg's. z as observables. to)!K. which is the prototype of a system in interaction with its environment. y. how is the state IW(t) at time t determined from this. For the simple system of a particle with coordinates x.!W(to) . how do the amplitudes (LjIW(t) evolve in time from the initial amplitudes (K i IW(to)? The assertion that !W(to) determines IW(t) is the quantum mechanical form of the principle of causality. The Evolution of Probability Amplitudes and the Time Development Operator. The canonical quantization of systems with classical analogues is discussed and applied to the forced harmonic oscillator.3) where the coefficients (14.4) 315 .8) can be generalized to the timedependent form (LjIW(t) = ~ (LjIT(t. This algorithm for calculating the future behavior of a wave packet from its past history was used in Chapter 13 for scattering calculations. The same state is represented in the momentum representation as <p (p. the amplitude is !/J(r. 0) develops in time into !/J(r. t) = (x.1) which is the timedependent wave function in the coordinate representation. t) = (Px' PY' pz IW(t) (14. z IW(t) (14. Dirac's).
eH(t) i (14. This quantity is called a transition amplitude. t) = T(to. expressing the fundamental linearity of quantum dynamics. to) From (14. to)IMe)(MeIK e p i) (14. Prove the relation (14. t 1 )T(tl> to)I'I'(to = T(t2. Equation (14. This means that if 1'1'aCtO»~ and 1'1'b(tO separately » evolve into 1'1'aCt»~ and 1'1'b (t». if at time to it was known to be in the eigenstate IK i ) of the observables K. to) I'I'(to» I (14. the time development operator has the property IT(t2. to)lqr(to» Hence. we can derive the transformation equation (LjIT(t.4) signifies the probability amplitUde for finding the system at time t in the eigenstate of !L) of the observables symbolized by L. to)IKj ) = 2: 2: (LjINp)(Np!T(t.7) it follows that 1'I'(t2» » = T(t2.7) Since T(t.. They have a direct and simple interpretation: The expression (14. i.9) For small e we may write T(t + e. to)T(to. to).6) Exercise 14.e.10) . because from (14.4) is the matrix element of an operator. to) does not depend on 1'I'(to». With forethought it has been written in the form of a matrix element of an operator T(t.3) without assuming that (14. the principle of superposition applies » to the time development of states.8) we infer that = T(t2' t1)T(tl> to) I (14.4) defines a representationindependent linear operator T(t.8) T(t.316 Chapter 14 The Principles of Quantum Dynamics are independent of the state 1'I'(to».8) if we require that (Lj IT(t. It now follows from the composition rule (14. t) IKi ) = (Lj IKi ) for all times t or T(t. to) relates the initial state 1'I'(to» to the final state 1'I'(t» according to 11'I'(t» = T(t. t) = 1 (14. t)T(t.5) which shows that (14.5) from (14.1. then a superposition CaI'I'aCtO»~ + Cb I'I'b(tO develops into cal'l'a(t» + cblqrb(t». which is valid for any state 1'I'(to».3).6) and (14.3) that the time development or evolution operator T(t. to). each component of the state moves independently of all the others. to) = 1 or (14. From (14. t1 )1'I'(t1» = T(t2. t) = 1  h.3) is consistent with the composition rule (9.
we must select an appropriate Hamiltonian operator. as it must be whenever the Hamiltonian corresponds to the energy operator of the system.42). This is no accident.k eH(t) ] 1'I'(t» Hence. Generally. to) we have with (14. we will see how the laws of wave mechanics derive from the general theory. 1'I'(t» + 8 ~ 1'I'(t» = [1 .10).T(t.) Since.13) Equation (14.1). the differential operator H was generally Hermitian. t)T(t. (14. We will see that it is analogous to the Hamiltonian function in classical mechanics. to) = T(t + 8. In Chapter 15. The form of (14. for we have merely reformulated in abstract language those fundamental assumptions that have already proved their worth in wave mechanics. by (14. will become evident forthwith. to) . apparently capriciously.iii dt ('I'(t) I = ('I'(t) IHt (t) d (14. The linear operator H(t) is characteristic of the physical system under consideration. to) I' T(t = 1m dt 8>0 + 8. we will assume that H is Hermitian. to first order in 8.. We also have 1'I'(t + 8» = T(t + 8. T(t + 8.12) is an equation for the state vector rather than for the wave function (14.1 The Evolution of Probability Amplitudes and the Time Development Operator 317 defining an operator H(t).10) the differential equation for T. to) i ( )l1 ) = . t) I'I'(t» or.12) is reminiscent of the timedependent SchrOdinger equation (3.8). In wave mechanics. but the distinction is one of generality only. iii d "it 1'I'(t» = H(t) I'I'(t» (14. can be useful for describing the . to) = 1. giving the general law of motion for any Hamiltonian system. Of course. even when the system has no classical counterpart. A nonHermitian Hamiltonian operator. This analogy has led to the name Hamiltonian operator for H(t).12) Its bra form is . ''~ dT(t.t e l i ' or (14.H t r to . like a complex index of refraction in optics.12) is the equation of motion for the state vector. It now becomes clear why Planck's constant was introduced in (14. (The reason for introducing the factor i/Ii.11) with the initial condition T(to. To specialize to a particular system.
to)] = 0 dt Since. If 1'I'(to) is normalized to unity.iii . with the initial condition T(to..16) that ('I'(t) I'I'(t) =1 for all times t consistent with the assumption that I (L) 1'I'(t) 12 is the probability of finding the observable L to 'have the value L} at time t. the equation adjoint to (14. The Hamiltonian operator must be found on the basis of experience. to) and subtracting the two equations. the product of the two operators equals the identity at t follows that the time development operator is unitary: Tt(t. the norm of any state vector remains unaltered during the motion. by (14. Often H does not depend on the time. we get d .15) By multiplying (14.('I'(t) I = ('I'(t) IH(t) dt (14. if one is available. to)H(t) (14.16) Hence. then the normalization will be preserved in the course of time.to)H] (14. to) and (14.6). Contact with measurable quantities and classical concepts can be established if we calculate the time derivative of the expectation value of an operator A. through absorption or decay. exchange energy with their environment.: (t .14) and (14. and then T can be obtained for finite time intervals by applying the rule (14.14) on the right by T(t. to ) = i 1. to) = 1 = to.to) H i 1m • n ]n In the limit we get by the definition of the exponential function. to) = Tt(t. to)T(t.: eH 1m 8~O n n. T(t.to)/n.('I'(t)IAI'I'(t) & = ('I'(t)IHAI'I'(t) . by (14. which may itself vary with time: d iii . Hence.~ l' [ 1 . such that ('I'(to) I'I'(to) = 1.11) on the left by rt(t..11) becomes iii dTt~.318 Chapter 14 The Principles of Quantum Dynamics dynamics of dissipative systems which. to) = exp [ k (t . and we have from (14. Physical insight is required to make a judicious choice of operators to be used in the description of the system (such as coordinates. ( 1 . using the clues provided by the classical description. momenta. + ('I'(t)IAHI'I'(t) + iii('I'(t)II'I'(t) ~ M . If H(t) is Hermitian.7) and (14. it (14.[Tt(t. T(t.8) repeatedly to n intervals.'" )n ~~. Quantum dynamics is a general framework and contains no unambiguous prescription for the construction of the operator H whose existence it asserts.10) we have. to)T(t.13) becomes d . to) = 1. each of length 8 = (t .17) It is obvious that T is unitary if H is Hermitian. and spin variables) and to construct the Hamiltonian in terms of these variables.
18) where. Rather. leads to the conclusion: (~) = 0 (14. We see that the commutation relations of H with observables play an important role in the theory. p] (14.20) Hence. but this is not the only way of dealing with dynamics in the framework of vector spaces. The Pictures of Quantum Dynamics. State vectors themselves are not observable or measurable quantities. as usual.22) which.. There we showed that the state vector of a system alone may be made responsible for the time evolution. I'l"(t» completely characterizes the state of a system at time t. equivalent. p]) = ('l"(t) I[H.19) From the equations ~f motion (14.. ) signify expectation values of the operators enclosed.4) are the physically significant objects of quantum mechanics.15).3) and (14. the expectation value of A is constant. (14.20). the brackets ( . We will now discuss other. I'l"(t)('l"(t) I] I'l"(t) = 0 (14. which states' that the probability amplitude for finding the value of Ki of the generic observables K at time t can be written as the inner product (Ki I'l"(t). p = I'l"(t)('l"(t) I (14. . and A is said to be a constant of the motion.2 The Pictures of Quantum Dynamics 319 or ih d (A) = (AH . The mathematical formulation of quantum dynamics given in the last section is not unique. In other words. A special example of a timedependent operator is the density operator for the state I'l"(t).12) and (14.21) which is not surprising in view of the conservation of probability.18) gives the simple result dt (p) = 0 (14. Physically. iii ap = at HI 'l"(t)('l"(t) I . If A is independent of time and commutes with H. this implies that a system represented by the same state vector at two different times has the same statistical distribution at the two times with regard to all observables of the system.HA) dt + iii at (M) = ([A. The observables K and their eigenvectors are regarded as constant.23) These results have sometimes led to the oxymoronic proposition that the density operator is a "timedependent constant of the motion" ! The formal relations derived in this section are all rooted in what we have called the quantum mechanical principle of causality. The definition (14. the eigenvalues of Hermitian operators and probability amplitudes such as (14.19) implies that ([H.I'l"(t)('l"(t) IH d = [H. H]) + iii at (M) (14. which was the fundamental assumption made in Chapter 9. by (14. formulations of quantum dynamics. Comparison with 2.
but the eigenvectors of an observable change from IA ') to U(t) IA '). note that if H is constant and energy is conserved. Expectation values remain unchanged. or pictures. Owing to the unitary property of U(t).25) this transformation has the effect of referring all state vectors back to their values at the initial time O. are equivalent and equally acceptable provided that (a) the operators corresponding to the observables maintain their eigenvalue spectra in the two formulations.320 Chapter 14 The Principles of Quantum Dynamics observation is made in terms of the eigenvalues of observables and of expansion coefficients (probabil. which in the SchrOdinger picture were represented by operators fixed in time (unless they happened to have an explicit time dependence). are represented in the Heisenberg picture by timedependent operators. a new picture is obtained. then T(t. the state vectors are constant in time.28) These equations exhibit the evolution of the Heisenberg operators. as in I'!') and £(t). and (b) the inner products ofphysical states with eigenvectors for the same eigenvalues are equal in the two pictures.26) The observables. I(t) = U(t)LUt(t) = Tt(t. 0) (14. by applying a timedependent unitary transformation U(t) to all states and operators. All the state vectors change from I'!'(t» to U(t) I'!'(t». Elsewhere in this book. t) = Tt (t. As a special case. 0) (14. will be used only in this chapter. Starting from the Schrodinger picture of quantum dynamics. It follows from (a) and (b) that all expectation values remain the same from one picture to the other. It follows that two vector space formulations.27) At the initial time t = 0. the bar notation. which are inner products in the abstract state vector space. and every operator A is transformed into U(t)Aut(t).O) = exp(i Ht) (14. called the Heisenberg picture. when there is no ambiguity. £(0) = L (14. we will simply use I'!') and L(t) for Heisenberg states and operators. and we denote them as II '!'(O» = I'!') I (14. all eigenvalues and all inner products remain invariant. the probability of the particular result being given by I<Lj I '!'(t» 12 if I '!'(t» denotes the state of the system at time t. O)L1'(t. In this formulation. t) I'!'(t» = I'!'(O» (14. To identify the Heisenberg picture. satisfying conditions (a) and (b).ity amplitudes). Measuring an observable L at time t means finding one of its eigenvalues Lj .24) Since U(t) I'!'(t» = T(O. The simplest choice for the transformation U(t) is U(t) = 1'(0.29) .
By differentiating (14. We can consider the body moving in a fixed frame. the Hamiltonian operator does not change in time.2 The Pictures of Quantum Dynamics 321 and H(t) = H. 0) . where the states are constant in time. 0) dt ILj .27) with respect to t and using the equation of motion (14.30) in the Heisenberg picture. Eigenvectors of observables. t)ILj ) (14. To emphasize that aL(t)lat is the Heisenberg form of the derivative of the explicitly timedependent operator L. is the counterpart of Eq. then in the Heisenberg picture the state vectors stand still and the operators with the eigenvectors "rotate" in the opposite direction.18) in the Schrodinger picture. H(t)] + iii d t aL(t) at (14. we obtain the Heisenberg equation of motion for an observable: iii d L(t) = [L(t). Its appearance shows that if in the Schrodinger picture we regard the state vectors as "rotating" in a certain direction in abstract vector space and the operators with their eigenvectors as fixed. except for the allimportant minus sign.32) The expectation value of the operator equation (14. But the mutual relation between state vectors and operators is the same in the two pictures.30) The last term arises from the definition aL(t) at = Tt (t ' 0) aL T(t 0) at ' (14. even in the Heisenberg picture. They are related to each other in much the same way as the two kinematic descriptions of the rotation of a rigid body with respect to a reference frame.34) Differentiating this last equation with respect to t.d ~ ILj .31) in the event that L is explicitly timedependent.11) for the time development operator. the probability amplitudes (inner products) are equal: (14.35) which is very similar to the equation of motion (14. (14.2) in the SchrOdinger picture. O)ILj t) (14. t) + T(t. differ in the SchrOdinger and Heisenberg pictures by the unitary transformation U(t): ILj. with the reference frame rotating backward.t) = T(O. strictly speaking we should write this operator as aL at = aL(t) at (14.11). Since the two pictures are connected by a unitary transformation. corresponding to the same eigenvalues. or the body as being at rest.36) . If H is constant. (14. we obtain dT(t.33) or ILj ) = T(t. t) = 0 or. using (14.
44) where (14. we obtain in ~ l'!r(t» = Hol'!r(t» + U(t)Hut(t)I'!r(t» = U(t)(H .2. = inTt(t.3.21). as in the Schrodinger and Heisenberg pictures.38) and that the expectation value of p is constant as in (14. Instead of attributing the time evolution to either the state vectors or the operators. state I'IJf).Ho)Ut(t)I'lt(t» We define an interaction term V as the difference between the Hamiltonian Hand the model Hamiltonian H o by H giving us the simplelooking formula in = Ho + V (14. 0) = (ii.39) subject to the initial condition U(O) = 1.322 Chapter 14 The Principles of Quantum Dynamics Exercise 14. Show that in the Heisenberg picture the density operator for p = I'IJf(0»('IJf(0) I = I'IJf)('IJf I satisfies the equations in J!.40) and these satisfy the equation of motion in ~ 4t) = I'!r(t» [4t). which is the solution of in dU(t) = . = U(t) I'IJf(t» (14. Generally. and using (14. T(t. The observables are then transformed into new operators (14.37) aat a at Pl and dp dt = 0 (14. Exercise 14. (14. H o may be timedependent. HoJ + in ~ (14.45) . Find the equation of motion for the Heisenberg bra (Lj .43) ~ I'!r(t» = V(t) '!r(t» I (14. We can arrange to let both the state vectors and the observables carry a complementary share of the time development. it is obviously possible to partition the time development operator in infinitely many different ways. If the system and its Hamiltonian are complicated.12).41) More important is the equation for the transformed state vector. it is often sensible to choose for U(t) the adjoint of the time development operator for a suitably chosen model Hamiltonian Ho.U(t)H. 0) J!. d. dt 0 (14.42) By differentiating this equation with respect to t.
3. The interaction picture will be found useful when we consider timedependent perturbation theory in Chapter 19. t) = U(t)IA') and (fill. Let us now apply the general equations of quantum dynamics to the special case of a point particle with a mass m. the SchrOdinger picture is recovered. t 1 ) = U(tz)T(tz.46) and (14. t 1 ) is the time development operator for the state vector in the interaction (or really a generic) picture as defined by I\fret)~ = f(t. Show that the expression (fill.50) we see that the transition amplitude can be expressed equivalently in all pictures as (fill. t 1 )IA') = (B". tzlA'. Exercise 14. let us suppose that at time t 1 the system has the definite sharp value A' for the observable A. t 1 )IA'. t 1 )IA'. tzlf(tz . If H o = 0. tzlf(tz . tl = (Bill vt(t) (14. tzlf(tz. t 1 )U\t1 ) (14.49) Since we infer from (14. to) I\fr(to) (14. t 1 ) = (B"IT(tz . the interaction picture coalesces with the Heisenberg picture. t 1 )IA'. U(t) = exp(i Hot) (14. t 1 ) where f(tz. The resulting formulation of quantum dynamics is known as the interaction (or Dirac) picture. To demonstrate the total equivalence of the various pictures of quantum dynamics. that f(t z . Note that if the model Hamiltonian H o is timeindependent and thus conservative. H o = H. Quantization Postulates for a Particle.47) If H o is chosen to be the full Hamiltonian. t 1 ) for the transition amplitude is quite general and gives the correct answer if the SchrOdinger (Ho = 0) or Heisenberg (Ho = H) pictures are employed.3 Quantization Postulates for a Particle 323 is the transformed version of V. Interaction Schrooinger tzllA'. with (14. or its more sophisticated Lagrangian or Hamil .51) The distinctiveness of the SchrOdinger and Heisenberg pictures is manifested by the important fact that the Hamiltonian H (energy) is the same in both pictures.42). We are concerned with the quantum behavior of this system. but it does have a classical analogueNewtonian mechanics. We ask: "What is the probability that at time t z the system will have value B" if the observable B is measured?" The answer is that the required probability is the square of the absolute value of the amplitude (fill.48) From this it follows.4. t 1 ) = (B".42) that lA:'. t 1 ) Heisenberg picture (14.
for operators (xp) = (1I2)m(d/dt)(x 2 ) is generally wrong. y. but it does not give us an unambiguous prescription of how to construct the quantum form of any given classical theory. y. To make the conversion from classical to quantum mechanics.55) and classically.56) .324 Chapter 14 The Principles of Quantum Dynamics tonian forms. For example. xpx = mx (dx/dt) = (1I2)m(dx 2 /dt) is a valid classical equation if not a particularly useful one. z. y. and the corresponding quantum observables. Px' Py. Pz' which characterize the system. expectation values of operators behave. because classical observables commute and we would have zero on the righthand side. Setting A = x in (14. in the limit h ~ 0. but their ultimate test lies in a comparison between the theoretical predictions and the experimental data.53) where H is the classical Hamiltonian function of x. This principle provides us with a test that the quantum theory of a system with a classical analogue must meet.dt and (1 . although (x)(PJ d(x) = m(x) . The trouble comes from the noncommunitivity of x and Px.(xPx + Pxx) 2 )=  1 m d (x 2 ) d 2 t are both correct. (xH . Certainly. The position operators x. z.52) On the lefthand side there is a velocity. which stand for operators. dpx dt aH ax (14.Hx) ClH lim =ih apx (14.18).54) Ii>O Similarly. which represent numbers. x. we must let h ~ 0 at the same time. Thus. the correspondence principle must be supplemented by a set of quan(ization rules. Hence. x. yet.Hx) ih dt (14. These rules have to be consistent with the correspondence principle. z. we obtain d(x) = (xH . like the corresponding classical quantities. for A = Px' we have in quantum mechanics d(pJ dt (14. The correspondence principle requires that . we cannot expect that every valid classical equation can be turned into a correct quantum equation merely by replacing classical variables by expectation values of operators. we formulate a heuristic version of the correspondence principle as follows: If a quantum system has a classical analogue. z are assumed to form a complete set of commuting observables for this physical system. For the purposes of this chapter it is worthwhile to distinguish in the notation between the classical observables. y. = dx dt ClH Clpx (14. but if we wish to compare this equation with the classical equation for velocities we cannot simply let the operators go over into their classical analogues. We expect that (14.52) is the quantum analogue of one of Hamilton's equations.
x 2 and (xpx + Pxx)2/2 both have the classical limit 2x~~. Prescription (1) must be applied with care if H contains terms such as xpx. pz] = iii] [x.54) and (14. the relation lim (GF .3 Quantization Postulates for a Particle 325 The correspondence principle requires that " (PxH .58) (14. More generally. Pz' also commute.57) can be verified for any H that can be expanded in powers of the coordinates and momenta. Show that the operators x 2 p.x 2 and (xpx + Pxx)212 differ The consistency of conditions (1) and (2) and their agreement with (14. only by terms of order li 2 • + p. z. The symmetrized operator (xpx + Pxx)/2 can then be used instead. Px] = [z. it is usually possible to avoid such ambiguities. pz] = [y.FG) = ao aF _ aF ao Ii+O iii + ao aF ax apx ax apx ay apy aF ao ao aF aF ao +ay apy iJz apz az apz (14.nx by virtue of repeated invocation of the quantum conditions (14.60) . as can be seen if we choose H = Px and H = x. the commutation relation [x. x ]x . x.5. In practice.54) and (14. which can be expanded in a power series. but they are not identical. respectively. commute with each other. + p. because X and Px are noncommuting and would upon translation into operator language give rise to a nonHermitian H. y. Py] = [x. Py] = [z. Px] = iii] agrees with (14. This is in agreement with the classical limit (14.a Hermitian operator identical in form with H but replace all coordinates and momenta by their corresponding operators.57). Let H be. of the coordinates and momenta. F and G. px] = [y.57) because for H = x n . pz] = [y. For instance.57) Similar equations follow for y and z and their conjugate momenta.58). Thus the Hermitian operators x7J. All these conditions can be satisfied if we do the following: 1. Py] = 0 (14. It is Hermitian and leads to the correct classical limit. the three momenta Px' Py. Px] = [z. we can continue this type of reasoning to prove that for any two functions.59) The coordinates.HpJ I1m iii = li>O aH ax (14. The consistency proof can be continued by letting H = x n • Then n 1 n _ x 1 iii [Px. Exercise 14. x ] . 2.ih [Px. Postulate the fundamental commutation relations between the Hermitian operators representing coordinates and momenta: [x. Sometimes there may be several different ways of symmetrizing a term to make it Hermitian. x] + 1 nl _ nl iii [Px.
and denoted by the symbols q and p.3 could not equally well have been applied to more general canonical variables. dt ' +at aA (14. Exercise 14.E.1 equivalently by routes 1. we must prove that the same form of quantum mechanics results whether we use x. To prove that we can pass from the upper left to the lower right corner of Figure 14. t). t). P to make the transition to quantum mechanics. Illustrate the validity of Eq.18). F]/ih.60) is abbreviated as [G.54) and (14. dA =[AH] P.6. (14.57) are special cases of (14. which are by no means restricted to Cartesian coordinates. In classical mechanics. p. Px. which is the same function of its arguments as A(x. in the limit h ~ 0. and the corresponding Poisson bracket on the right. . 397. d(A) = f[A.58). Px. Dirac discovered that this is the classical analogue of the commutator [G.60) is assumed to be valid for any smooth functions of coordinates and momenta.. or q. and is known as the Poisson bracket' (P. these equations are well known to have the same general form for a large class of variables called canonical coordinates and momenta. Px. Equations (14.) ofF and G.62) for the operator A(x. Px. Since in the Hamiltonian form of mechanics the Cartesian coordinates do not occupy a unique position.B. Could we replace x by q and Px by p (assuming for convenience only one degree of freedom). the expression on the righthand side of (14. Canonical Quantization and Constants of the Motion. p. even if a power series expansion cannot be made. All of these arguments can be summarized in the proposition that the classical equation of motion for an arbitrary function A(x.B. F]p. where F and G are the same functions of the ordinary variables as F and G are of the corresponding operators.60) by letting G = x 2 and F = and evaluating both the operator expression on the left. we ask whether the quantization procedure of Section 14. satisfying the more general commutation relations I qp  pq = ih] I (14. 2 or 3. 4. So far we have considered only descriptions of the physical system in terms of Cartesian coordinates for a point particle. 4 we first consider an infinitesimal canonical transfor 'Goldstein (1980).60). dt \ H]) + faA) \ at ih (14. Equation (14. Yet. and could we still apply the same quantization rules? To show that we are indeed at liberty to use canonical variables other than the Cartesian ones. Rather. t).63) instead of (14.326 Chapter 14 The Principles of Quantum Dynamics holds. the connection between classical and quantum mechanics was established by the use of Hamilton's equations of classical mechanics.61) is the correspondence limit of the quantum equation of motion (14.
The commutators are evaluated by applying the quantization rules of the last section for Cartesian coordinates.4 Canonical Quantization and Constants of the Motion canonical transformation X. '" c:r N 4 X. apx (14. e (14.64) The new Hamiltonian is aH aG H'(q. the new Hamiltonian IS constructed by correspondence with (14.65) as H'(q. for an arbitrary function F(x. Px) + . In quantum mechanics. Agreement with (14. G].:.e ax apx apx ax = F(x. p) = F(x. G] e[F.:::.Px unitary transformation q.66) More generally.65) = H(q.65) in the correspondence limit is assured if we define the Hermitian openitors q = x + in e [x. p) + in e [G. p) = H(x. P = Px + in [Px.1. Px) + e in [F. p. a transformation that is generated by an infinitesimal function eG(x.B.69) The Hermitian operator G(x.:::. Px) is constructed from G(x. p) = F(x. H] (14. c: 3 2 :. Px) by letting x and Px become operators.e. p) we find to first order.Px 327 q. H]p. i. Px) = H(q. p) = H(x. '" c: .:::. F(q.ax apx aH aG +e. mation (step 1 in the figure). p) .) = H(q.p 1 c: :. G] (14.p Figure 14. This canonical transformation is paralleled in quantum theory by step 4. Px) from the relations q =x+ aG e.apx ax (14.B. Classical canonical transformations and quantum mechanical unitary transformations..e . . p) + e[G. G]P.67) The corresponding operators satisfy the equation F(q. Px) + e aF aG of aG . (14.64) and (14.:. 0 c: '" c:r .68) again to first order in e. ~ N 0 :.
)0 00 the unitary operator.75) This finite unitary transformation changes the Cartesian variables into (14. in quantum mechanical transformations.68) may be written as F(q. r (14. This is true because two classical canonical transformations made in succession can be replaced by a single direct one.. (14. p) = (1 + ~ G F(q.16 for the conditions under which the last equality in (14.70) showing that the new operators are obtained from the old Cartesian ones by an infinitesimal unitary transformation (step 4): = U.73) [See Exercise 9. (14. p) = H(x. U1qu.) = U1H(q.72) The Hermitian operator G is the generator of this infinitesimal transformation.71) with (14. we obtain the in limit N .76) The commutation relations are also invariant under the finite transformations. as uxutulx') = qUi x') = x' Ulx') Hence. Px)U1 (14. q' = x' and Iq') = Ulx') We see that the quantization of the system can be carried through by the use of the general commutation relations (14. We note that if (14.63) for any pair of canonical variables that . Similarly. U = . Px)( 1.~ G) (14. the eigenvalue spectra of x and q are the same. and apply the same unitary operator N times. Px) = H(U1qU.328 Chapter 14 The Principles of Quantum Dynamics To first order in e. The quantization procedure based on rules (1) and (2) of Section 14. becomes H'(q.69). (14.~ (1 + ~~ G = expC: G) q = UxU t .3 can be extended to new canonical variables that are infinitesimally close to Cartesian.] The commutation relations are invariant under unitary transformations because (14. p)U.63). This completes the proof that the quantization rules of 14. p) )F(X. successive application of unitary operators is equivalent to the application of a single unitary operator.3 can now be immediately generalized to all those canonical variables that can be obtained from the Cartesian ones by a succession of infinitesimal canonical transformations.76) holds.74) and we have arrived at (14. In terms of the new variables.73) holds. the quantum analogue of the classical Hamiltonian.F(x. If we let e = A1N (where A is a finite parameter and N is an integer).
defined by the relation Ulx') = Ix').77) is canonical. (J. cp. z and r.p)ut (14. Clearly. which are observables that commute with the Hamiltonian H (see Section 14.p') = UH'(q. invariant. cp.21. Exercise 14. For more general canonical transformations than these. p) consists in noting that if a (finite or infinitesimal) canonical transformation to new variables q'. (J. y. For the special case e = 7T/2.73) to finite canonical transformations. y.00 to + 00. Px by a continuous succession of infinitesimal transformations. the standard quantization procedure mayor may not be valid. Nor does a unitary transformation between x.close connection with the classical canonical formalism.1). and range from . If the canonical transformation leaves the Hamiltonian invariant. and vice versa. This terminology has asserted itself. p' is made. The correspondence between canonical transformations and unitary operators has led to the frequent designation of unitary operators as quantum mechanical canonical transformations. e Exercise 14. The transformation to these from Cartesian coordinates is canonical. p) (14. because of the discontinuity of the spherical polar coordinates. cp. the new Hamiltonian H' is related to the old one by the equation H(q. calculate the matrix elements of U in the coordinate representation. z.78) which is just an extension of (14.79) then UH . and is a realvalued angle parameter.HU = [U. even though some unitary transformations have no classical analogue.8 and 3.p) = H'(q'.8. H'(q. Section 3. it will be valid whenever the new operators can be obtained from the old ones by a unitary transformation. but it cannot be generated by a succession of infinitesimal tran. Because of its .7. gives rise to a unitary transformation which takes x into x and Px into Px' An important application of canonical transformations concerns the finding of constants of the motion. A simple example of a failure of the standard quantization is provided by the transition to spherical polar coordinates. T. Construct the unitary operator U that effects this transformation. H] =0 . Show that the reflection operator. rederive the result of Exercises 3. if a and b are realvalued constants. p) = H(q. for if it did the eigenvalues of the latter operators would have to be the same as those of x.4 Canonical Quantization and Constants of the Motion 329 can be obtained from x. Show that the transformation Uaxut = ax cos UbPxUt = ax sin e + bpx sin e e + bpx cos e (14. A useful way of obtaining constants of the motion for a timeindependent Hamiltonian operator H(q. so that the new Hamiltonian H' is the same function of the canonical variables as the old one. The general procedure for expressing the Hamiltonian operator and the Schrodinger equation in terms of curvilinear coordinates will be given in the Appendix. the quantization procedure described here is referred to as canonical quantization.formations. (J exist. Noting that this transformation leaves the operator a 2 x 2 + b 2p. contrary to the definition of r.
+ p.330 Hence. such a rotation induces the change 8A == !. Similarly.ScJ> fi .1.ezPz == produces no change in the momenta E . physical observables are obtained which are constants of the motion.85) by (11.82). As an example. fi X A iii' 'I' (14. H is invariant under an infinitesimal transformation U. the generator of infinitesimal translations e. the transformation describes a coordinate translation. + p. The operator A· A == A2 is a rotationally invariant scalar operator. any H that does not depend on the coordinates is invariant under the transformation (14. In this way.eypy .1. [G.ScJ> [A fi· L] == . Hence. H] == 0 and thus G is a constant of the motion. as expected from the connection between momentum and displacement operatbrs.83) Thus.81) According to (14. owing to the fundamental commutation relations. L (for arbitrary vectors fi and rotation angles ScJ» and L itself are .[A G) iii' == . E .80) ie then the (Hermitian) generator G of the transformation commutes with H. in particular.1. changes the coordinates to r I 1 == r .iii [r. If. for infinitesimal rotations we recall the generator eG == 8cJ> fi . p2 H == 2m + V(r) (14.. and we conclude that linear momentum is conserved under these conditions.84) from Section 11.HU) == 0 1 and thus the unitary operator U is a constant of the motion if the transformation leaves H invariant. For a vector operator A.. Hence.83).. eG == E' P (for arbitrary E) and p itself are constants of the motion.19). Chapter 14 The Principles of Quantum Dynamics dt (V) d == iii (VH .66). G (14. Evidently. eG == exPx .86) is invariant under rotations.) P (14. L (14. P] == r  E (14. eG == .82) but. == 1 + r. (14.8. consider a free particle whose Hamiltonian is 1 H == 2m (p. since p2 and r are scalar operators. p' == p (14. as was shown in Section 11. The Hamiltonian for a particle in a centralforce field.
9.. since they are all related to the SchrOdinger picture by a generally timeqependent.)ut formally unchanged (see Exercise 9. they either commute or the commutator i[A. E. verify that the equation of motion (14. unitary transformation U(t). Exercise 14.5 Canonical Quantization in the Heisenberg Picture 331 constants of the motion. Using Eqs.. and orbital angular momentum is conserved for a system with spherical symmetry.88) in ~t = lP(t).. the canonical commutation relations for conjugate variables [q. 2 The canonical quantization procedure can be formulated in any of the pictures of quantum dynamics. 2S ee Fallieros (1995). Prove that if the entire spectrum of H is nondegenerate. C. B] is also a constant of the motion. .) = f(UAUt.87) for expectation values of observables holds in any picture of quantum dynamics. d(t) . . C.. d~) = ([A. B o] and depend on the choice of B o. Such a transformation leaves every algebraic relation between operators f(A. B. p] = qp . ... C.41): l 'f/. =0 = Uf(A. Exercise 14. Illustrate this theorem by constructing an example Hamiltonian for which A = Lx and B = L y are constants of the motion.. (14.. B.16): f(A. dij(t) dt = [(t). then A and B must commute. If the constants of the motion A and B do not commute. UBUt. Canonical Quantization in the Heisenberg Picture.10.41) and (14. The equations of motion for the canonical variables are derived from (14. BJ) (14. there must be degenerate energy eigenvalueS'. Show that if both A and B are constants of the motion.) =0 Hence. R] q 0 (14. the model Hamiltonian for the particular dynamical picture. if/.pq = flU become the same in any picture: The dynamical law.87) for any operator A that does not depend on time explicitly also has the same form in all pictures. UCUt.44). .) 5.
92) and by postulating that if the classical equations of motion are expressed in terms .91) (14. p. t) = ih [F(q. p. q(O)] = (14. ') h [F. t) iJ (14. the equation of motion can be expressed for any dynamical variable F@.54). and not merely their expectation values. In Eq.95) If this commutation relation is applied to (10. t) == F(q(t).m ih t m _ p(O) t (14. In the Heisenberg picture. we have pet) = p(O) and q(t) _ = q(O) + . which otherwise is just one of infinitely many (unitarily) equivalent pictures. confers a measure of distinction on the Heisenberg picture of quantum dynamics. H] + at F@. subject to the commutation relations [qk(t). generally. t).94) hence. (14. i G] (14. p.s. The simple commutation relations (14. qit)] = [Pk(t). p(t)] = 0 (14.91) and (14.93) is to be madebarring. For example. p. p. p. as usual.B. pit)] = ihS kf [qk(t). pet). we find the uncertainty relation (Aq>r(Aq)o = Aq(t)Aq(O) ~ 2m 1 tl h (14. pet)] equal to ihl. this is the quantum analogue of the classical equation of motion. the correspondence [F.30) in summary form as di F@.332 Chapter 14 The Principles of Quantum Dynamics In the Heisenberg picture (Ho = H). its wave packet in the coordinate representation must spread in the long run as 1 tl ') 00. t). the transition from classical to quantum theory for a system that has a classical analogue is made simply by replacing the canonical variables by operators that can change in time.89) By (14.61). t) according to (14. d dt F(q. H]P. complications that may arise from ambiguities in the ordering of operators. + at F(q. p. t) = [F(q.96) which shows that if the particle moves freely.90) The formal similarity between the classical equation of motion for canonical variables and the quantum equations for the corresponding operators. t) d 1 iJ (14. [q(t). q(t) and q(O) do not commute. of Poisson brackets. nor is the commutator [q(0).92) are not valid if the operators are taken at two different times.96). if the system is a free particle in one dimension with H = p 2l2m. G]P. the notation . Thus.
with dq(O)dp = then. the harmonic oscillator offers the simplest nontrivial illustration. From our present general point of view. If the close correspondence between the classical theory and quantum dynamics gives the Heisenberg picture a certain preferred status. or any other picture.98) Verify that for the minimum uncertainty wave packet this result agrees with (14. being expectation values. albeit complexvalued ones. be written as (14. For a free particle in one dimension and an arbitrary initial wave packet.97).99) .96).q t 2" Ii Itldp . the oscillator Hamiltonian may. however. If the initial wave packet at t = 0 is the minimum uncertainty wave packet (10. initially narrow wave packets spread more rapidly than those that initially are broad. that were made in Section 2. it is a matter of taste and convenience whether we navigate in the Heisenberg or the Schrodinger picture.97) Exercise 14.66). Except for denoting the coordinate by q rather than x. and show (as in Problem 2 in Chapter 3) that (14. A() JJ. similar to the Schrodinger picture form.4. Once again. which are more naturally described by moving wave packets. using (14.11. the Schrodinger picture is perhaps a more intuitive form of quantum mechanics. calculate the time development of (dq). this inequality provides a precise formulation of the statements about wave packets contracting long before the present time and spreading in the distant future.5 Canonical Quantization in the Heisenberg Picture 333 on the lefthand side emphasizes that the variances. In the Heisenberg picture. The Schrodinger picture is particularly suitable for a discussion of scattering processes. If the particle is free. than by operators changing in time. are independent of the choice of the quantum dynamical picture.:::m This inequality is consistent with the result of an explicit calculation of the time dependence of the variance of q for a wave packet that has' 'minimum uncertainty" at t = 0: (14. Also compare with the value of the variance (dq)2 as a function of time for a beam of freely moving classical particles whose initial positions and momenta have distributions with variances (dq)6 and (dp)2.
The most convenient one is the same as in Section 10. can be expressed in terms of a and at. If we choose to = 0.12. Exercise 14. the present example is an important exception. Equation (14. a+(t) = at (to) ei".101) Similarly. In the energy representation ofthe Heisenberg picture.72).(tto) (14. They are the raising (creation) and lowering (annihilation) operators in the Schrodinger picture. Thus. H] = f/. the coordinate operator of the harmonic oscillator is explicitly represented by the matrix: (14. which coalesces with the Heisenberg picture at t = 0. We now choose a basis and introduce a fixed representation. a(t)t]a(t) Using the commutation relation (10. d~~t) = [aCt).103) o V2e.72). The canonical variables p and q. the initial values of the operators a(O) = a Work out directly from (14. aCt) = exp(k Ht) a exp ( k Ht) as an application of the identity (3. if/.w[a(t). transcribed into the Heisenberg picture.334 Chapter 14 The Principles of Quantum Dynamics Applying (14. Then determine (jU) and pet).90a) and (10.+ dt da(t) _ iw aCt) =0 (14.100) Although the operator equations of motion are difficult to solve directly in most problems.27). we find the simple differential equation .i"'t o . in principle the equations of motion have all been integrated. The matrices representing aCt) and at(t) are obtained by multiplying the matrices (10.30). necessitating passage to a representation in which these equations become conventional systems of linear differential and integral equations.102) and at(O) = at are two mutually adjoint operators that satisfy the commutation relation (10. This may be called the energy representation.6: A basis supported by the eigenvectors of H.59). and any function of these.90b) representing a and at by ei"'t and ei"'t. we obtain the equation of motion for the lowering (annihilation) operator aCt).100) can be solved immediately. they serve as the constants of integration of the dynamical problem. even though aCt) is an operator: (14. respectively.
w. which is not explicitly timedependent.pP(t) (14. The Forced Harmonic Oscillator. Its elements are harmonic functions of timea general property of the matrix representing any Heisenberg operator in an energy representation that is supported by the fixed eigenvectors of H: HIE) = EIE) Indeed.107) . it is desirable to consider the dynamical effects produced by the addition of a timedependent interaction that is linear in q to the Hamiltonian of the harmonic oscillator: p2 1 H = + . with eigenvalue ae.13. With no additional effort.69) and (10.E")t] The special feature ~f the harmonic oscillator as a perfect periodic system is that all matrix elements oscillate with integral multiples of the same frequency. the Hamiltonian (14. especially in manybody and field theory.106) in either the SchrOdinger or Heisenberg picture.mw2 l 2m 2  qQ(t) where Q(t) is a realvalued function of t.104) giving us the matrix element (E' IA(OIE") = (E'IAIE") exp[k (E' . it will remain in a coherent state. In either the Heisenberg or SchrOdinger picture.105) may be cast in the form H = = hW(ata + ~) + f(t)a + f*(t)a t hW(a\t)a(t) + D f(t)a(t) + f*(t)a\t) + (14. with eigenvalue a.qQ(t) . A(t) = exp(k Ht)A exp ( k Ht) (14. 6. as it should be. for any operator.71).iw " at time t. we may generalize the Hamiltonian even further by introducing a velocitydependent term: p2 H = 2m + 1 2 mw q 2 2 . With the substitutions (10.6 The Forced Harmonic Oscillator 335 This matrix is evidently Hermitian.105) where P(t) is also a real function of t. show that if at t = 0 a linear harmonic oscillator is in a coherent state. provided that we define the complexvalued function f(t): f(t) =  j!  2mw Q(t) + i  f¥ mw P(t) 2 (14. This perturbation corresponds to an external timedependent force that does not depend on the coordinate q (dipole interaction). Exercise 14. For many applications.
dt = _ [a(t). Obviously..336 Chapter 14 The Principles of Quantum Dynamics In most applications. By integrating (14. .112) over an infinitesimal interval from t' .It eiW(tt')f*(t') dt' 0 (14..112) because such a function permits us to write a particular solution of Eq.t') (14. For instance. A Green's function appropriate to Eq.e to t' + e. (14.110) If we choose to = 0.wt dt n which can theI).. at(t)] = 1 (14. this equation simplifies to a(t) = ae.113) This is easily verified by substituting (14.108) we derive the equation of motion .it eiw(tt')f*(t') dt' to (14. i . and the operators are subject to a unitary transformation as they change from the free oscillation regime before T 1 to a free oscillation regime after T2 • Using the equaltime commutation relation..f*(t) n (14. be integrated to produce the general solution a(t) = a(to)eiW(tto) .wt} = . for t 1= t' the Green's function is proportional to eiw(tt').t')f*(t') dt' (14. it can be multiplied by e iwt and cast in the form d . it is instructive to review the solution of (14.{a(t)e..111) Although it calls for unnecessarily heavy artillery. It is therefore reasonable to assume that the disturbance f(t) 1= 0 acts only during the finite time interval T 1 < t < T 2 and that before T 1 and after T2 the Hamiltonian is that of a free oscillator..113) into (14. HJ = _ nw a(t) + f*(t) or . The time development of the system is conveniently studied in the Heisenberg picture. n . we derive the condition lim[G(+e) . we are interested in the lasting rather than the transient changes produced by the timedependent forces in an initially unperturbed harmonic oscillator.109) as a(t) = _i f+oo n 00 G(t . [a(t).109).109) by the use of Green's functions to illustrate a method that has proved useful in many similar but more difficult problems.. .+ dt da(t) i iw a(t) = ..iwt _!:. n .!:.t') dt + iwG(t .109) This inhomogeneous differential equation is easily solved by standard methods.da(t) In .G(e)] = 1 8+0 (14.t') = 8(t . (14.109) is a solution of the equation dG(t . but at t = t' there is a discontinuity.f*(t)e. in which the state vector I qt) is constant.114) for e > O.
GA(t .120) ~ is the Fourier transform of the generalized force f(t).t')f*(t') dt' is the particular solution of (14.6 The Forced Harmonic Oscillator 337 Two particular Green's functions are useful: IOR(t and t') = 11(t . T 2 eiwt'f(t') dt' = f+O:> .t')f*(t') dt' (14.t')eiW(tt') I (14. These two particular solutions of (14. and if we choose to = 0. which vanishes for t aCt) =  k j:: < T r • Similarly. alternatively. We note that aCt) =  k J:: OR(t . aCt) = aout(t)  k j:: GA(t .119) = f. we can write (14.109) which vanishes for t > T2 • If we denote by ain(t) and aout(t) those solutions of the homogeneous equation :it + t da(t) iwa(t) = 0 which coincide with the solution aCt) of the inhomogeneous equation (14. eiwt'f(t') dt' (14.k g*(w) I where g(w) (14.112) are known as retarded and advanced Green's functions.115) (14.111).116) where the Heaviside step function 11(t) = 0 for t < 0 and 11(t) = 1 for t > 0 (see Appendix.117) and (14.118) are equivalent to solution (14.109) for < T 1 and t > T 2 respectively.109).118). we obtain the relation I aout = am .117) and (14.t')f*(t') dt' is the particular solution of (14. By equating the righthand sides of (14..117) or. . Section 1). respectively.118) Both (14.
of an unperturbed Hamiltonian Ho == nw(ata + D (14. a] == a and [ata.118) in the Heisenberg picture can be used to answer all questions about the time development of the system. t I ) (14. _" _ {V(t')V(tll ) T[V(t )V(t )] .338 Chapter 14 The Principles of Quantum Dynamics The solution (14. To implement the interaction picture.126) over the time interval (t I .126) TUb Integration of (14.45). .125) Similarly.124) The equation of motion for the state vector in the interaction picture is (14. (14.)2 jll * rr V(t') dt' jll V(t") dt" ± . f' jll V(t') dt' + (.128) It is sometimes convenient to write this series expansion in a more symmetric form by using the timeordered product of operators. t) and use of the initial condition t I ) = 1 produce an integral equation for the time development operator: (14. the time development operator TUz. at] = at. t I ) in the interaction picture as defined in Eq.122) Timedependent Hamiltonians require more careful treatment than timeindependent ones. because generally the interaction operators at two different times do not commute. According to (14.117) or (14.129) . since [ata. We choose the unperturbed Hamiltonian operator Ho as the model Hamiltonian to define the interaction picture. It is nevertheless instructive also to consider the dynamical problem from the point of view of other pictures. . we regard the Hamiltonian of the forced oscillator as the sum.127) A formal solution of this equation can be constructed by successive iteration: T(t. We define time ordering of two operators as _ . t I ) = 1  * .59). (" (14. We thus obtain V(t) = f(t)aei"" + f*(t)atei"" (14.V(t")V(t') if ttl if t' ::5 ::5 t' ttl (14. H == H o + VU).49) satisfies the equation of motion: in dT~t t I ) = V(t)T(t.121) and an explicitly timedependent "interaction" term. V(t) == f(t)a + f*(t)a t (14. the transformed interaction operator is (14.123) The interaction operator can be evaluated by use of the identity (3.
dtn1T[V(tl)V(t n. V(t~)] .133) where.124). dt~T[V(tDV(t~) ._.. V(t")] are numbers for all t ' and t".. [V(t ' ). . .dt~ .. with the interaction potential (14. t+.h V(t)T(t..126)._.6 The Forced Harmonic Oscillator 339 This convention is easily generalized to products of any number of timedependent operators... t l ) = lim eVNeVNleVN2 . The expression (14. n tl tl tl l. t l ) = T exp{  k{ [f(t')ae. V(t")] = f(t')f*(t")eiw<t'n .~ Jtl Jtl . dt. Z )" _ _ _ .133) can be further reduced if the commutators [V(t'). ( ~. _ T(t + e...134) Expression (14.. e V2 e VI N>oo (14. V(tnI)] or i _ e _ tl ) .8) and write T(t. because the time intervals of length e are infinitesimally short and are not subject to internal time ordering. it dt1dt z ·. This is indeed the case for the forced harmonic oscillator. by definition. since according to (14.126). With it we can prove that if t > t lo the time development operator may be written in the form or formally and compactly as (14.iwt ' + f*(t')a t e iwt '] dt'1 (14.135) . t l ) + e __ n=1 1 2:.. and Ne = t  tl (14.132) Although this is a compact expression for the time development operator.131) To prove that (14.130) is a solution of (14.124).131) is applied to the forced linear harmonic oscillator. To derive a more manageable formula.. t l ) = 1 00 + n=1 2:. n 00 1 ( i )n t+.T(t. Jtl { { ( n. which in the limit e ~ 0 reduces to (14. it is sufficient to write to first order in e. we use the group property (14. t+. we obtain T(t..)n nV(t) it it . If the formula (14.. even if the interaction operators at different times do not commute. ".133) is valid..f*(t')f(t")eiW<t't OO ) (14. because of the presence of the unwieldy time ordering operator it is not yet in a form convenient for calculating transition amplitudes.
t.tt)at. the state at time t is 1'\f1(t» = T(t.124) and (14. t.ttle('a*. is a coherent oscillator state 1 a). T(t. ~ 00 and t ~ +00.142) For the forced harmonic oscillator with an interaction of finite duration during the interval (T.61) can be applied repeatedly to give T(t.136).)at } (14.138) This expression can be connected with the Fourier integral of the applied force given in (14.) = lim exp{i N>ao (V k=' k + 2 1 [Vb n=' vnJ)} ± dt' or..98). we thus obtain the time development operator in the interaction picture in the desired form: T(t. T 2 ) the S operator is i S = e {3 exp [ k g(w)a  k g*(w)at J (14.ttl I a (14. Of particular interest is the limit of the operator t(t. as defined in Section 10. t. like f3. T. V(t"»)} (14. like the operator D in (10.120): g(w) = ift(*(T2. +00) (14. This limiting time development operator is known as the S (or scattering) operator and is defined formally as (14.eaata*alo) = ei{3(t.137) where we have defined (t.140) If the initial state at t = t. t. t.tl l exp{ (*(t.136) For the forced harmonic oscillator. inserting (14.137) is a displacement operator for coherent states.) = e i{3(t.340 Chapter 14 The Principles of Quantum Dynamics The identity (3. t.) =  ki: eiwt'!*(t') dt' (14.) = ~ rr dt' 2ft )tt ( dt"{f(t')!*(t")eiw(t'tOOl .7.tl)a+&(t.tlle&*(t.) = exp {  kL V(t') dt'  2~2 L r dt"[V(t').)a + (t.143) . the real phase f3 appears as the price we must pay for eliminating the time ordering operator. t. a coherent state remains coherent at all times. because the time development operator (14. is a numerical phase. We arrive at the intriguing and important conclusion that.139) In (14. t.141) + () where 'Y.135) into (14.f*(t')f(t")eiw(t't"l} )tt (14. t. under the influence of the (dipole) interaction f(t)a + f*(t)a t . if the limit N ~ 00 and e ~ 0 is carried out. and it stands for: f3(t.137).*al/2/&+alat('*+a*)a I0) = eiy(t.) as t.)1 a) = ei{3(t.) = ift(*(oo.
(14. we use the results from the interaction picture to deduce the time development operator in the SchrOdinger picture. These results can be interpreted in terms of a system of quanta.14. From Eq. we obtain S = e i(3 exp{ k J:: [f(t)ae iwt + f*(t)ate iwt ] dt} (14.148) for the time development operator by applying (14. we obtain Exercise 14.16.15) prove that eAAf(B)e.150) . B] = yB (as in Exercise 3. was first illustrated in Section 6.137) in the interaction picture.146) = (n)n e(n) = ~lg(W)12ne_lg(W)12/1i2 n! h n! with expectation value (n) = I g(w) /h for the oscillator quantum number.110).YB) (14.49) we infer that 1 2 2 (14.145) As the link between the states of the system before the onset and after the cessation of the interaction. n being the number of quanta present.6 The Forced Harmonic Oscillator 341 where we have denoted {3 = {3(+00. what' is the probability that it will be found in the nth excited oscillator energy eigenstate after the interaction terminates? The interaction produces the state S I 0).61). according to (10. Finally. or the S matrix representing it. (0) (14.149) Exercise 14. Exercise 14.147) If we employ the oscillator Hamiltonian (14.120).15. The strength of the interaction determines the average number (n) of quanta present and characterizes the Poisson distribution. which represents the probability that a dipole interaction pulse incident on the vacuum state of our system of quanta leaves after its passage a net number of n quanta behind.AA = f(eA. If [A. the scattering operator.144) Substituting the expression for g(w) defined in (14. If the oscillator is in the ground state before the start of the interaction.3.121) for H o and the time development operator (14. Verify the expression (14. Show that stas = a  i h g*(w) = a + aout  ain (14. These features of the dynamics of the forced or driven linear harmonic oscillator will help us understand the creation and annihilation of photons in Chapter 23. We will encounter the same concepts again in Chapter 20. The interaction term in the Hamiltonian is linear in at and a and creates or annihilates quanta. which is a coherent state with. The transition probability of finding the oscillator in the nth eigenstate after the interaction is.eigenvalue a = (i/h)g*(w).149) and (3. the Poisson distribution (14.
an electron is emitted and the nucleus changes into an 3He nucleus. Using a gauge in which A z = 0. and assuming only that [V(t'). at asymptotically long times before the onset of the interaction. In the nuclear beta decay of a tritium atom CH) in its ground state.\3)112. including the continuum? . (14. For a forced harmonic oscillator with a transient perturbation (14. show that the time development operator is given by T(t + !J. in a time lit short compared with all relevant periods.151) Exercise 14. Compute the probability of atomic excitation to the 2S and 2P states of the helium ion. and compute the probability that the atom is found in the ground state of the helium ion after the emission. A linear harmonic oscillator is subjected to a spatially uniform external force F(t) = CT/(t)'e. V(t")] = 0 during the impulse. A linear harmonic oscillator in its ground state is exposed to a spatially constant force which at t = 0 is suddenly removed.' h J. 3. Use the generating function for Hermite polynomials to obtain a general formula.342 Chapter 14 The Principles of Quantum Dynamics where S is the operator defined in Eq. Problems 1.iwg*(w)(a t ) + wlg(w)j2/fi (14. pz.17.146).152) agrees with a direct calculation based on the Poisson distribution formula (14. aim and a out differ from each other by additive constants. calculate the probability of finding it at time t in an oscillator eigenstate with quantum number n. Show that liE = iwg(w)(a) . 2.. If the oscillator is in the ground state at t < 0.. Derive the energy spectrum and the eigenfunctions in the qrepresentation.152) If I '\f1) is the ground state of the oscillator. Propose alternative methods for solving this eigenvalue problem. derive the change in the unperturbed energy. Assume that the change is sudden. verify that liE given by (14. if I'\f1) is the initial state of the oscillator in the interaction picture.t. w being the natural frequency of the harmonic oscillator. examine the variation of the transition probabilities with n and with the ratio "\/w. show that q = (cpx . How much energy is transferred? 5.qAx)/qB and p = (cPy . Compute the transition probabilities to the excited states of the oscillator.23). deduce that (14. 4. Discuss the remaining degeneracy.At where . Assuming C = (lim.142) and ain and aout are related by (14. How probable is excitation to higher levels. and using the unitarity of S. Noting that the operators a.106).\ is a positive constant and T/(t) the Heaviside step function (A. A particle of charge q moves in a uniform magnetic field B which is directed along the z axis. t) = exp [  i tH. If the term V(t) in the Hamiltonian changes suddenly ("i}1lpulsively") between time t and t + lit.qAy)1c may be used as suitable canonically conjugate coordinate and momentum together with the pair z. V(t') dt' ] Note especially that the state vector remains unchanged during a sudden change of V by a finite amount.119).
t) r . If the oscillator frequency is w. obtain the energy eigenvalues and the stationary states for t > O. is subjected to a timedependent interaction between the ground state 10) and the first excited state: Vet) (n I'It(t». calculate (n I'It(t» for t > O. A linear harmonic oscillator. derive the time dependence of the variances [A(a ~t) r and [A(a0. 7. where ~ is a realvalued parameter. (c) If the system is in the ground state of the oscillator before t = 0. = F(t) I1)(01 + F*(t)!O)(l1 (a) Derive the coupled equations of motion for the probability amplitudes (b) If F(t) = V2fiwll(t). Show that the sudden change generates a squeezed state. with energy eigenstates In). At t < 0 a system is in a coherent state Ia) (eigenstate of a) of an oscillator and subjected to an impulsive interaction .roblems 343 .
we summarize the threedimensional quantum dynamics of a particle of mass m (and charge q) in the coordinate or momentum representations. compactly written as rp . (15. but the derivatfon of the familiar equations of Section 9. Instead. The Coordinate and Momentum Representations. Finally. it is common that our information is less than complete and consists of probabilities for the different pure states that make up a statistical ensemble representing the state of the system.59). We assume that the Hermitian operators rand p each form a complete set of commuting canonically conjugate observables for the system. 1. Generalizing what was done in Section 9. the time evolution of any amplitude is compactly. No really new results will be obtained. The corresponding eigenkets are denoted by Ir'}.1) and that the results of the measurement of these quantities can be any real number between 00 and +00. the bold italic 1 denotes an operator in Hilbert space as well as the identity matrix (dyadic) in three dimensions.pr = iM (15. The eigenvalues of r. In Eq.2) = 8(r' (15. and for many applications conveniently.3) 344 . we show how the formalism is generalized to interacting compound systems and their entangled states. We extend the principles of quantum dynamics to the density operator (or density matrix) as the proper tool for the description of the partial information that is generally available. with appropriate notational changes.1).6 for a particle in one dimension. satisfying the commutation relations (14. From information theory we borrow the concept of the entropy as a quantitative measure of information that resides in the probability distribution. In the coordinate representation. in the quantum domain it is often not possible to prepare a system in a definite (pure) state l'\}f}.3 to the dynamics of a particle in three dimensions. Practice in manipulating the braket formalism can be gained by filling in the detailed proofs. This view of the dynamics of a system leads to Feynman's path integral method. which are omitted here.58) and (14.6 from the general theory will be sketched. and we have rlr'} with the usual normalization (r" Ir') = r'lr'} . Since most quantum systems are composed of more than just one particle in an external field. form a real threedimensional continuum. denoted by r'. derived from the initial state in integral form in terms of a propagator (Green's function).CHAPTER 15 The Quantum Dynamics of a Particle In this chapter. We now apply the quantization procedure of Section 14.r") (15. to which a bareboned introduction is given. we develop quantum dynamics beyond the general framework of Chapter 14.
The Coordinate and Momentum Representations 345 in terms of the threedimensional delta function.6 to obtain (r"IF(r. we avoid this terminology. (15.9) in agreement with Eq.(r') d 3 r" d 3 r' = JJ p")(p"IAI p')(p' 'It) d p" d p' J Ip")(p"IAlp')<p(p') d p" d p' 1 1 3 3 = 3 3 (15.8) and (15. p) is an operator that is a function of position and momentum.1) invariant. Construct the relations that are analogous to (15. Once we are at home in the coordinate representation.4) is the main computational device for deriving useful representation formulas. Since we chose to attach Schrodinger's name to one of the pictures of quantum dynamics. 1 the transformation r ~ p and p ~ r. p) can be expressed as a power series in terms of the momentum p. such as x and Px.31). that may occur.7). ('ltbIF(r. after substituting (15. p)lr') = F( r".2.2) and the commutation relations (15.7) and (15. They can be calculated from the definition (15. which leaves the commutation relations (15.1.8) Equations (15. The closure relation. Exercise 15.1). . lThe coordinate representation employed here is often also referred to as the Schrodinger representation. P)I~a) J "'b(r')F(r" 7V')"'a(r') d r' = 3 (15.8) and performing the r" integration.9) in the momentum representation. (3. The (coordinate) wave function "'(r') is the probability amplitude for a state 1'It): "'(r') = (r'I'It) = (p'I'It) (15. may be used to convert any coordinaterepresentation formula into one that is valid in the momentum representation. For example.8) provide us with the necessary ingredients for rederiving the formulas in Section 3. we can transcribe the derivation in Section 9.5) and the momentum wave function is analogously defined as <p(p') (15.7) The matrix elements of an operator A in the coordinate or momentum representation are continuous rather than discrete functions of the indices. care must be taken to maintain the ordering of noncommuting operators.6) The action of an operator A on an arbitrary state I'It) is expressed by AI'It) = JJ Ir")(r"IAlr')(r' 1'It) d 3 r" d 3 r' = J Ir")(r"IAlr')". If F(r. . we can obtain from (15. If A = F(r. 7v")s(r'  r") (15.
f = f (p"lr')(r'!p') d 3 r' p") e(iIr.346 Exercise 15.)p'·r' (15.10) can be used to translate any algebraic equation of the form F(r. (r'lp')= 1 (27Tfii/2 e (i1r. the phase factor is chosen to be' unity so that finally we arrive at the standard form of the transformation coefficients. The eigenvalues are thus continuous. p) I'1") = 0 (15.p") = g*(p")g(p') Hence.)l/J(r') = pip') F(ifiV p " p')cP(p') = 0 (15.)t/J(r') p')(p'I'I") = F(ifiV p " = F(ifiV p " (15.)P'·r't/J(r') .)p'.12) As an application.13) for the momentumcoordinate transformation coefficients (r' I p '). In (15.11) in the coordinate and momentum representations as F(r" ~ vr.14) there is no restriction on the values of p' other than that they must be real in order to keep the eigenfunctions from diverging in coordinate space. The solution of this differential equation is (r'lp') = g(p')e(ilf. p)I'I") = F(r" (p' !F(r.2..r' (15. = p'lp') in the coordinate representation takes the differential equation form fi .(r' !p') = p'(r'lp') l (15.14) where g(p') is to be determined from a normalization condition.)(P'p")r' d 3r' = (27Tfi)3Ig(p')!Z8(p'  g(p') = (27Tfi)3/2 era 1 . (P') Arbitrarily. and an appropriate normalization condition requires that (p'lp") = 8(p' . but conveniently.10) p')cP(p') Equations (15.16) d3r' e(i1r. the eigenvalue equation for momentum.Vr. p)I'I") ~ Vr. Chapter 15 The Quantum Dynamics of a Particle Prove that (r'!F(r.15) These probability amplitudes linking the coordinate and momentum representations allow us to reestablish the connection between coordinate and momentum wave functions for a given state 11/1): cP(p') = (p'I'I") = f (p' Ir')(r' 1'1") d 3 r' = (27Tfi)3/2 f (15.)(r'I'I") = F(r" ~ Vr.
23) and similarly in the momentum representation.e. When we apply the rules (15.21) in order to cancel the undesirable extra term on the righthand side of (15. V(r) = Vcp(r). of course. (15. t) (15.24) .17) These relations are. The new coordinate basis is spanned by the basis vectors Ir') = ei<P(r)/fil r ') = eirp(r')lfilr') (15.20). p) are (r"IF(r. ~ V' )<r' I'!'(t) = (r'I'!'(t) (15. we are to keep the rule (15. The replacement (15. the matrix elements of an operator F(r.12) to the equation of motion (14. in ..3.(p'I'!'(t) = (p'IHI'!'(t) = H(inV p " p')(p'I'!'(t) a at (15. There is a fundamental arbitrariness in the definition of the basis Ir') of the coordinate representation. substitution p 7 P i. p)eirp(r)lfilr') = (r"!F(r. familiar from Chapter 3.8). which has consequences for the wave function and for the matrix elements of operators.r") If.1) invariant. the wave function for state 1'!') changes from l/!(r') = (r' I '!') to (?I'!') = ~(r') = ei<p(r')lfil/!(r') (15. we must make the transformation F(r. In terms of this new coordinate representation.1) remain valid if the + V(r) is made.6 (except that here we have chosen the passive viewpoint). Prove that the commutation relations (15. as discussed in Section 4. we obtain in the coordinate representation the differential equation in ft (r' I'!'(t) = (r' I HI '!'(t) = H( r'. This ambiguity arises from the option to apply to all basis vectors a unitary transformation that merely changes the "phases by an amount that may depend on r'. since the addition to the momentum p of an arbitrary irrotational vector field.21) is consistent with the commutation relations for rand p. p) 7 F(r.18) where cp(r) is an arbitrary scalar field.Vcp!r') = F(r'" (15.22) for the timedependent wave function l/!(r'. p .1 The Coordinate and Momentum Representations 347 and its Fourier inverse. provided that the vector field V is irrotational.12). Exercise 15. when calculating matrix elements in the new basis. leaves the conditions (15. p)lr') = (r"leirp(r)IIiF(r.19) In the new basis. V X V = O.20) ~ V"  V"cp)8(r' . p + V cp) (15. The ambiguity in the choice of coordinate basis is seen to be a manifestation of the gauge invariance of the theory.
L + 2m VCr) (15.!/J(r.26) we obtain from (15.25) If the Hamiltonian operator H has the common form H = . Exercise 15.1): ih a at !/J(r'. 2. t) + f (p'I V(r) Ip") d 3p" c/J(p".29) must be used for the transcription of the equation of motion into the coordinate (or momentum) representation. the Hamiltonian is timedependent.21). Planck's constant has the same value in all cases. verify the equivalence of (15. the timedependent Schrodinger or wave equation.2m (V  2 iq hc A +V )2] !/J(r. If the system is a particle in an external electromagnetic (gauge vector) field described by the potentials A(r. we obtain in a similar fashion ih a at c/J(p'. the Hamiltonian operator H = _1 2m (p _q A)2 + c V (15.27) In the momentum representation. Generally. t) = '2 ~m c/J(p'. t) (15. a at = h [.30) . t) (15. the cumbersome notation that distinguishes between different pictures and between operators and their eigenvalues (by the use of primes) can be dispensed with. t).4.348 for the momentum wave function Chapter 15 The Quantum Dynamics of a Particle c/J(p'. t) . t) = [ 2 h 2m V.28) The matrix element (p I IVCr) Ip") is a Fourier integral of the interaction potential. The Propagator in the Coordinate Representation. Since we usually choose to work in one particular dynamical picture and one specific representation. t) and c/J(r. t) (15. The equivalence of these various forms of quantum mechanics shows again that the constant h was correctly introduced into the general theory in Chapter 14. t) = (p'l'l'(t) 2 (15.22). By expressing (p' IVCr) Ip") in the coordinate representation. unsurprisingly. the potential energy V includes the fourth component of the electromagnetic potential. Hamiltonians for more complex systems must be constructed appropriately when the need arises.28) and (3. Here.2 + VCr') ] !/J(r ' . the equation of motion is zh . In the SchrOdinger picture and the coordinate representation. (3. We have come full circle and have found wave mechanics in its coordinate or momentum version to be a realization of the general theory of quantum dynamics formulated in ~bstract vector space.
t')K(r. is related to the propagator by GR(r.31) and the composition rule (14. t') = i.. . we define the propagator as the transition amplitude. r'.t') (15.35) can also be read as an integral equation for the wave function. Exercise 15.33) The propagator ca~ also be identified as a (spatial) Green's function for the timedependent SchrOdinger equation. t) (15. t') = (rIT(t. The retarded Green's function. t')lr') I the propagator satisfies the initial condition (for equal time t): K(r. defined as G R = 0 for t < t'..fi2 ( V .6.r')8(t .36) and satisfies the inhomogeneous equation ifi . t.31) of the propagator and the Hermitian character of the Hamiltonian show that K(r'. A  at ' " 2m hc )2 + V]K!(r' r" " t t') (15. t. t.35) This relation justifies the name propagator for K.!1.31) = 8(r .r') (15.32) and is a nonnormalizable solution of the timedependent SchrOdinger equation ifi a K!(r r" t t') i = [ . r'. t') i (15.6. IK(r.a brief introduction will be given in Section 15. t) = J K(r. t') d 3 r' (15. r'. t) and (r'.3. r. r'.34) linking the transition amplitude that reverses the dynamical development between spacetime points (r. t) = K*(r.B. ". Exercise 15.(t . K is its kernel. t') to the original propagator. Equation (15. If a gauge transformation is performed. t. as in Section 4.37) The propagator and the retarded Green's function are at the center of Feynman's formulation of quantum mechanics. to which.3) we see that the propagator relates the Schrodinger wave functions at two different times: t/J(r. aG at = H(t)GR + 8(r . t. From the definition (15.2 The Propagator in the Coordinate Representation 349 In order to transcribe the equation of motion (14. r'. r'. what happens to the propagator? Derive its transformation property. t') (15.5. t'.11) for the time development operator in the coordinate representation. t. t')t/J(r'. From its definition (15.
T(t. t') (15.A K (15.~ ~ 0/: (r')o/n (r) exp [ . K(r. The corresponding advanced Green's function is defined as GA(r. Since the initial condition (15.t) = K(r.t) is also a solution of the timedependent SchrOdinger equation (15. t' . we must have.t)K(r. r'. t. t') = exp [  k t')] H(t  (15. t .5. explicitly. r'. t . t  t') = K(r'. as discussed in Section 3.t') ] lEn )(En Ir') where (15.t') = ~ (rl exp [ i H(t . K*(r. t . t' .31) then depends on the times t and t' only through their difference t .t')] (15. the equation of motion is invariant under time reversal.t') I (15. the time development operator is. r'. owing to time reversal symmetry.kEn (t . K(r.34) of the propagator shows that under these quite common circumstances the propagator is symmetric with respect to the space coordinates: IK(r.37). including the degenerate ones. In this case K*(r.39) we note that the two Green's functions.33).41) and the propagator (15. r'. r'.45) .40) If the Hamiltonian is timeindependent. in terms of the energy eigenfunctions. r'.350 Chapter 15 The Quantum Dynamics of a Particle Exercise 15. (15. If in the Hamiltonian (15.32) selects a unique solution. r'.t') = explicitly.7. r'.38) It satisfies the same equation (15.43) The sum must be extended over the complete set of stationary states. r.42) Hence. t') = h TJ(t' i .37) as the retarded Green's function. t.t'.29) the vector potential A = 0 and VCr) is static (or an even function of t).44) Comparison of this result with the general property (15. t . which solve the inhomogeneous equation (15. are connected with the propagator (which is a solution of the Schrodinger equation without the inhomogeneous term) by i GR G = f t . Since for the Heaviside step function: TJ(t) + TJ(t) = 1 (15. Prove Eq.t') (15.37).
66)] = 0) represented by the minimum uncertainty "'(x. in Problem 1 in Chapter 2 and Problem 2 in Chapter 3. Calculate I "'(x.[21T(. t .8. and in Eq. K(x. this expression reduces to 2 114 "'(x.t')] (15. 0) = [21T(. (14.1X)o] [1 + 2m(.51) All these results for freeparticle dynamics are in agreement with the preliminary conclusions reached in Section 2.1X)~rI/4 exp{ .k ~~ (t . x'.2 The Propagator in the Coordinate Representation 351 The simplest and most important example of a propagator is the one for the free particle.X' )2 2ilit [x' . = (.46) and.1x)~ [ 1 + 4m2(. applying (15.48) solves the timedependent SchrOdinger equation and agrees with the initial condition (15. t .1x)ri li2t2 ] (15.4(~:)~ + ikOX} (15.1x)~ Exercise 15. t') = exp [ .32). Verify that (15. If the system is initially (at t ' wave packet [see (10. T(t. "'(x.47) This Fourier integral can be performed explicitly: K(x. t) .1 4(. 0) = e ikox .4.50) and show that the wave packet moves uniformly and at the same time spreads so that (ax).9.50) [ 2m(.)px(xx') dpx (15.35). In one spatial dimension.48) and (15.t') exp m(x .48) Exercise 15. The generalization of the propagator formalism to .2iko(.1X)~J ilit 1/2 2 X2 exp 4(.49) into (15.1x)~ 0 x )2} dx' 0 If the integration is carried out.1X)~rI/4 x f _ +co co exp {m(x . in its integral form suitable for the continuous momentum variable. If the initial (t' = 0) state of a free particle is represented by "'(x.49) we have.)~tt')/2m e(ilr.98). by substituting (15.43). t) 12 from (15.x'f} 2ili(t . t) = J2.35) produces the usual plane wave.lit [21T(.(.t') (15.1x)o 1 + ikox ilit ik~lit] + 2m (15. verify that (15.1x)~f _ /d.t') = m { 21riIi(t . x'.t') = f +co co e(ilr.
t) = S*(x. the Green's function solution of the equation . The linear harmonic oscillator is a second example. t): K(x. x'.54) where for the linear harmonic oscillator V =mlJix 212.32).54) by a series in powers of x and x'. t) (15.2 + c(t) (15.. t) = S(x'. x'. 0) it follows that for all t. t) at Ij? a2K(x. x'. x'. can be calculated by many different routes. Setting t' = 0 for convenience. as long as Cartesian coordinates are used...52) which satisfies the initial condition (15.56) whe~e the symmetry relations (15.352 Chapter 15 The Quantum Dynamics of a Particle the motion of a free particle in three dimensions is straightforward.. K(x. we may go back to the method developed in Sections 2.:. x'.45) have been employed. For instance.32).2)..5 and 3.. x'. x'.59) for the oscillator reduces to the freeparticle propagator (15. aK(x..x'. x'.59) In the limit w ~ 0. t) = eiS(x. Combining all this information and attempting to solve (15. we conjecture that S must have the form S(x.34) and (15. . x'. x'. 1 b(t) = mw sin wt' c(t) = in .58) All ofthese requirements.58) is solved by a(t) = 2 mw cot wt..x'.53) such that S satisfies the quantum mechanical HamiltonJacobi equation (3.2xx' + X. t) 2 1 (15.ab. 2 m in c =a m (15.1 and write the propagator in terms of a function S(x. t) = (2 '1T'l 'nm~ wt)112 exp{ sm 2z sm wt 'n m~ (x 2 cos wt . t) = K(x.54) yields the coupled ordinary differential equations.m(JiK(x. t) = a(t)x 2 + b(t)xx' + a(t)x. x'.55) = S(x..t)/r. t) ax 2 2m + .2" log (2'1T'in sin wt) giving finally the result K(x.57) Substitution of this Ansatz into (15. at + as 1 2m (as)2 ax  2 in ax2 + V(x) as 2m = 0 (15. can be satisfied only if (15. the propagator (15.2 cos wt} (15. x.48). S(x. including especially the initial condition (15. 0) (15. (15. Since this potential is an even function of the coordinate and since initially. 6= .~ = ..
59) answers all questions about the dynamics of the harmonic oscillator. The expectation values (x) and (p) perform harmonic oscillations. x6 sm (15.122).iwt ) (15. Using (10. Exercise 15. derive If! (x. unforced oscillator Hamiltonian in the SchrOdinger picture (see Exercise 14.59) for the linear harmonic oscillator. it is desirable to express the time development of the oscillator in terms of coherent states.iw (a a+ll2)tl a) = e.63) In words: If we represent a coherent state Ia) by its complex eigenvalue a as a vector in the complex a plane (Figure 10.13. (10.63) again yields the oscillating wave packet (15. verify the result (15.2 The Propagator in the Coordinate Representation 353 Exercise 15. + h XoX sm wt .13) as 1'I'(t) = t e.64) the complex a plane can be interpreted as a quantum mechanical analogue of classical phase space.12. for many applications. in principle. the propagator (15.O) = N exp{  ~: (x  X )2} O (15. show that (15.iwt12 1 ae. This was done in Section 14.59) has the form eiwt/2!(eiW') and deduce the energy eigenvalues of the oscillator by comparing with (15.11.10.61). t) directly from If!{x.122) that the displaced oscillator ground state wave function (15. t) I oscillates without any change of shape. Since (15.6 in the interaction picture for the forced or driven harmonic oscillator. Show that if the initial state of a harmonic oscillator is represented by the displaced ground state wave function If!(x. Exercise 15.62) develops in time under the action of the free.60) is the coordinate representative of a coherent state I a).i (w t 2" mw .60) the state at time t is mw If!(x.43) of the propagator. Conversely. We know from Eq.61) Show that IIf!(x. especially in quantum optics. For the harmonic oscillator. Relate the normalization factors for the two expressions.1). which represents a displaced ground . t) = N exp{ '211. Exercise 15.43). 0) by expanding the initial wave function. show that except for a normalization factor the amplitude (xl'l'(t) calculated from (15.46) to the stationary state expansion (15. the time development of the oscillator is represented by a uniform clockwise rotation of the vector with angular velocity w. as dictated by classical mechanics. Here we revert to the Schrodinger picture. with Xo = Y2h1mw a. (x  Xo cos wtf . An initial state 1'1'(0) = Ia) (15. 2wt)} 411. Although.mw. Applying Mehler's formula (5.
33) to obtain the expansion coefficients and again to sum the expansion. as in the generalized parametric and driven harmonic oscillator. the result (15. The forced or driven harmonic oscillator represents the next stage in complexity of an important dynamical system.f(t'. Exercise 15. it may be obtained by extending the solution of the quantum mechanical HamiltonJacobi equation to the forced oscillator. .67) and g(t.14. It is then possible (but tedious) to show that when the propagator is expressed in terms of the function S(x. t') = .t') (15. the propagator may be evaluated from the time development operator derived. This latter term arises from the presence of the term proportional to ft (occasionally called the quantum potential) in the quantum mechanical HamiltonJacobi equation. Use the generating function (5.6. Z SIn dt" (" ') dt] F(t]) dtz F(tz) sin cb(t1 w t . t)x' + g(t. Rederive (15.58). t) as in (15. t')x . From (15. t') = f(t. except for a purely timedependent term c(t) in (15.mlJix z .60). derive the expectation values (x)t and (p)t in terms of their initial values.xF(t) 2m 2 (15. t') = t It' . The resulting S(x. The propagator formalism and its expression in terms of the action function S provides a natural entree to Feynman' s path integral formulation of quantum dynamics.68) All the quantum mechanical functions S that we have calculated so far are essentially the same as Hamilton's Principal Function in classical mechanics. 2S ee Feynman and Hibbs (1965). adding the timedependent interaction term xF(t) to the potential energy V(x) in (15. It is responsible for the correct normalization of the propagator. x'. in terms of stationary states.61). but it must be augmented by an interactiondependent correction that is linear in the coordinates and has the form: Sinlx.66) We only quote the results of the calculation: z f(t.57) and (15. If the Hamiltonian has the form H =  pZ 1 + . t.57) for the free oscillator can still be used.53). Alternatively.54).354 Chapter 15 The Quantum Dynamics of a Particle state as in (15.64) and its time development. Chapter 3. 1 sm w(t . in Section 14.65) where F(t) stands for an arbitrary driving force. in the interaction picture. t') (15. The addition of anharmonic terms to the interaction complicates matters and makes solving the HamiltonJacobi equation more difficult. This very close connection between classical and quantum mechanics is contingent on the simple form 'of the interaction potential as a polynomial of second degree in x.t') It F(t") sin w(t" t' t') dt" (15.t t' t' It" It"  t') sin w(tz . t) will generally exhibit more distinctive quantum effects.
For a particle moving from spacetime point (x'. t) 1 2 mi 2  V(x. Here we use S (in roman font) to denote the classical function to distinguish it from its quantum mechanical counterpart. which is Hamilton's (classical) Principal Function S for the motion between these two spacetime points.48). reproduces the exponent in the free particle propagator (15. we limit ourselves to the motion of a particle in one dimension. t') to (x. To derive the Feynman path integral expression for the propagator in quantum mechanics.59). t') is the stationary value of the classical action function I(x.73) 3Goldstein (1980).70) is the classical Lagrangian function for the simple onedimensional oneparticle system that we are considering. x'. x'. we first observe that owing to the fundamental group property (14. t) with the classical Lagrangian L = (112)mx 2 . t")lx")(x"IT(t". t) is entirely contained in a realvalued phase function. Hamilton's Principal Function is S(x. nix') dx" (15. nix') = f (xIT(t.69) where L(x(t). t. x" " t t') = = f /(I(x. S (in italicS). t') = Jr' r L(x(t"). x(t"). x(t").3 Feynman's Path Integral Formulation of Quantum Dynamics 355 3. the x and x' dependence of the propagator for a transition from spacetime (x'.1. t') to the final destination (x. t') to (x. Section 10.(1I2)mw 2 x 2 . x'" " t t")K(x" 'x" t" t') dx" " (xIT(t. x(t). Thus. x'. the propagator satisfies the composition rule: K(x . . t') dt" = 0 (15. t") dt" = (15. x(t"). derived from Hamilton's Principle. t') = 8 Jr' r L(x(t"). 8I(x.71) singles out the motion xU) that takes the particle from the initial spacetime point (x'. In this section. t. We saw in the last section that if the potential energy depends on the coordinate x only through terms that are linear or at most quadratic in x. t. t") dt" (15. From classical mechanics 3 it is known that S(x.8) of the time development. For a particle moving from spacetime point (x'. show that S. Hamilton's Principle for the variation of the action. t) with the classical Lagrangian L = (1I2)mx 2 . reproduces the exponent in the harmonicoscillator propagator (15. t). derived from Hamilton's Principle. Exercise 15.16. t. t) (15. Exercise 15.15.72) where it is now understood that xU) is the correct classical motion comlecting the two endpoints. t') = Jr' r L(x(t"). show that S. x'. Feynman's Path Integral Formulation of Quantum Dynamics. t') to (x.
dx K(X2' XI' l K(x.356 Chapter 15 The Quantum Dynamics of a Particle for any value of the time t".74) 2e. When this is done. t). we know that in this approximation the propagator that takes us from (XnI' tnI) to (x n . x'. t. tn) is in the form (15. we partition the time interval (t'. in units of Planck's constant h. t') to the final destination (x.t') to the final spacetime point (x. .74) and take the limit e 7 0 and N 7 00..XN2 x' e t' t'+e 1'+ 2e t2e te t Figure 15. From Section 15. t) now is generally not the actual classical motion x(t) that Hamilton's Principle selects. Equivalently.2. the additivity of the action function shows that each path contributes. x(t").1 Paths linking the initial spacetime point (x' . as indicated schematically in Figure 15. t') The construction of this expression implies that the x integrations are to be performed as soon as any two successive propagators are multiplied. t") dt" Jt' t (15.76) Although all quantities in (15. leaving the integrations to the end.e) . the composition of the N propagators may then be construed as a sum over infinitely many different paths from the initial spacetime point (x'. the path x(t) that takes the particle from the initial spac~time point (x'. In order to utilize a simple approximation. we may first multiply the N propagators. XNI.. t') to the final spacetime point (x. a re..al phase S[x(t)] == r L(x(t"). J dxN. t .. x . t. The curve composed of broken straight segments is a typical path that makes a contribution exp{ kS[X(t)] } to the Feynman path integral.76) are classical functions of coordinates and of time.00 to + 00. t') = J. however.I . the composition rule generalizes to K(x. The propagator for each infinitesimal time interval is now approximated by assuming that the motion of the particle from (x n . t n ) is governed by a potential that is at most a seconddegree polynomial in x.1.t).. The smooth curve represents the classical path x(t) for the particle motion in the potential Vex).75) If we multiply the N elementary propagators for a particular "path" in spacetime together in the integrand of (15.h t n I) to (x n . (15. Since each of the coordinates Xi ranges from . t) into N infinitesimally short intervals of duration e.. t' t' + + e)K(xh x'. t' + e..
which in effect was already used in Chapter 2 for obtaining approximate wave functions. x'. is a useful tool for evaluating the propagator by the Feynman path integral method. corresponds.77) where the factor 'c. 5 For all systems that can be quantized by either method.77).76) into the composition rule (15. Chapter 8. To prove this.1. The actual classical spacetime path x(t) that connects the initial and final spacetime points. but it captures the essence of the argument and produces correct results. even with an arbitrary linear driving term. we finally arrive at Feynman's path integral formula for the propagator. according to Hamilton's Principle. To evaluate such an integral."tlx'. but is exact for a large class of problems. it is equivalent to the canonical form of quantum mechanics. but the Feynman method is quite general. including the free particle and the harmonic oscillator. In the last expression on the righthand side of (15. one can show that those contorted paths that are not accurately represented by the approximation (15. which is the limit of the ~um over all paths sketched in Figure 15.77). developed in Chapter 14. 4S chulman (1981). and a first (semiclassical) approximation for the propagator is therefore I K(x. . In this chapter we have confined ourselves to describing the path integral formulation of quantum dynamics for the simple case of a nonrelativistic particle in one dimension. t') = C(t. (X. t') = C all 2: aths e(illi)S[x(t)] = C Je(illi)S[x(t)]D[x(t)] (15. due to destructive interference caused by extremely rapid phase variations between neighboring paths.2 that this formula is not just an approximation. t') e(illi)S[x(t)] I (15. .77). x'. t') and (x.74). t.78) We saw in Section 15. arises from the product of the timedependent factors C(tnI> t n ) in the propagators "(15. 4 The derivation of (15.77) given here is a bit cavalier. see Townsend (1992). t') = K(x. (x'. to the stationary phase in the path integral (15. the integration over paths reduces in many applications effectively to the sum of contributions from only a few isolated paths. Although it is in general a difficult mathematical problem.75) for the individual path segments contribute negligibly to the sum over all paths in (15. sPor an excellent discussion of interference and diffraction of particle states in relation to path integrals. the differential D[x(t)] is intended to remind us that the propagator is afunctional integral. it is obviously necessary to extend the concepts of mathematical analysis beyond the standard repertoire and define an appropriate measure and a suitable parametrization in the space of possible paths. The stationary phase method. but the path integral approach offers a road to quantum mechanics for systems that are not readily accessible via Hamiltonian mechanics.75) and is attributable to the term proportional to ih in the quantum mechanical HamiltonJacobi equation. which is independent of the coordinates.75) and (15. t). in which the variable of integration is the function x(t).3 Feynman's Path Integral Formulation of Quantum Dynamics 357 Substituting the results from (15. t. The neighboring paths add constructively.
it is usually dispensed with in the physics literature. to make things clear. and we adopt this practice when it is unlikely to lead to misunderstandings. Ex . we confine ourselves in this section to the quantum mechanics of systems with distinguishable particles. is the sum and not the product of the three onedimensional spaces supported by x and y and z. the product space has dimension nl X n2' This idea is easily extended to the construction of direct product spaces from three or more simple spaces. since the three Cartesian coordinates can be chosen independently to specify the location of the particle. prove the identity M) PI ® N2Q2 = (M) ® N2)(P) ® Q2) (15. (On the other hand. outer. y. very important peculiarities require consideration. y. the Euclidean threespace with basis vectors X. M) ® N2 is represented by a matrix that is said to be the direct product of the two matrices representing M) and N 2 separately and that is defined by (15.82) Check this identity for the corresponding matrices.81) Exercise 15. More generally. Furthermore. The direct product space is formed from two independent unrelated vector spaces that are respectively spanned by the basis vectors lAD and IB~) by constructing the basis vectors (15.80) and N2IB~) = we define the direct or tensor product operator M) M) ® N2IAfB~) = Ai·Bz 2: IA~B~)<A~IM)IA~)(B~IN2IB~) Hence. or tensor product of vector spaces for simpler subsystems.358 Chapter 15 The Quantum Dynamics of a Particle 4. Since Chapter 21 deals with these exclusively. z) may be expressed as the direct product Ix) ® Iy) ® Iz) == Ix)ly)lz). If M I and PI are operators in space 1 and N2 and Q2 are operators in space 2. z). Z. y.79) Although the symbol ® is the accepted mathematical notation for the direct product of state vectors. Quantum Dynamics in Direct Product Spaces and Multiparticle Systems. The most immediate example of a direct product space is the state vector space for a particle that is characterized by its position rex. If n) and n2 are the dimensions of the two factor spaces. it is sufficient to restrict the discussion to systems containing just two particles. If the particles are identical. Often the state vector space of a system can be regarded as the direct.) Any operator that pertains to only one of the factor spaces is regarded as acting as an identity operator with respect to the other factor spaces. if M) and N 2 are two linear operators belonging to the vector spaces 1 and 2 such that M)IAD = 2: IA~)<A~IMIIAD Ai 2: IB~)(B~IN2IB~) BZ ® N 2 by the equation (15.17. The basis vector Ir) = lx. We are now prepared to generalize the formalism of oneparticle quantum mechanics unambiguously to systems composed of several particles.
and m r = mlm2/(ml + m2) is the reduced mass of the system. we introduce the twoparticle wave function (15. '" for two particles may be regarded as a wave in a sixdimensional configuration space of the coordinates rl and r2' The Hamiltonian of the twoparticle system (without spin and without external forces) is taken over from classical mechanics and has the general form H = _ I + _2 + 2ml 2~ p2 p2 V(lrl .89) and each of the two subHamiltonians possesses a complete set of eigenfunctions. all the eigenfunctions of (15. including the dynamics of the nucleus. we usually assume the normalization I (15.2ml VI . It is easily verified that the substitutions (15.r21) "'(rh r2) = E"'(rh r2) (15.4 Quantum Dynamics in Direct Product Spaces and Multiparticle Systems 359 amples are the ordinary hydrogenic atom or the muonium atom.~ V~ [ + ] v(lrl . Y. it can no longer be pictured as a wave in the naive sense that we found so fruitful in the early chapters.5). The new Hamiltonian is a sum (15. In analogy to (15. this leads to the SchrOdinger equation 2 h2 h .2mr iP ) (a 2 + al' + at a ax 2 2 2 ( ) '" + V(r)'" = E'" (15. z) and the coordinate of the center of mass R(X. the deuteron composed of a proton and a neutron. M = ml + m2 is the total mass.86) in configuration space.87) transform the SchrOdinger equation to h . Instead. In this equation.84) Since '" is now a function of two different points in space.2M 2 (iP ax2 + ay2 iP + h az2 '" .90) . We denote the two particles by the subscripts 1 and 2. and the positronium (electron and positron) atom. R) = "'(r)<p(R) (15. Z). six spatial coordinates are used to define the basis Irlr2) = Ir l ) Ir2) for the twoparticle system. y.r21) (15.88) where now'" = "'(r.83) The interpretation of this probability amplitude is the usual one: 1 "'(rh r2) 12 d3rl d3r2 is proportional to the probability that particle 1 is found in volume element d3rl centered at rl and simultaneously particle 2 in volume element d3r2 centered at r2' If I/J(rh r2) is quadratically integrable. As long as the spin can be ignored.85) In the coordinate representation. Hence. R) is a function of the relative coordinate r(x.88) can be obtained by assuming that'" is a product "'(r.
93) satisfy the commutation relations for conjugate canonical variables. It is interesting to ask whether the wave function for a twoparticle system is factorable. p and R. or separable. E = E R Chapter 15· The Quantum Dynamics of a Particle + En such that (15.94) and the orbital angular momentum of the system becomes L = rl X PI + r2 X pz If the Hamiltonian is expressed as p2 H =  = r X p +R X P (15. such states are particularly simple to interpret. Exercise 15.91) and (15. deuterium.97) Obviously.6) and the hydrogen atom (Chapter 12).360 and the energy a sum. whose solutions are plane waves. and so on. Equation (15. P defined in (15. This justifies the simple reduced oneparticle treatment of the diatomic molecule (Section 8.96) subsequent quantization and use of the coordinate representation lead again to (15. the relative motion of a system of two particles subject to central forces can be treated like a oneparticle problem if the reduced mass is used. muonium. and can be written as the product of a function that depends only on the coordinates of particle 1 and a function that depends only on particle 2.M + pZ 2mr + V(r) (15. We note that the linear momenta are transformed according to p The kinetic energy takes the form p2 pZ T=+2mr '2.95) '2. The canonical transformation (15. since we can say that in these cases the two particles are described by their own independent probability .87) and (15. Also show that the Jacobian of the transformation from coordinates rI.93) (15.88).92) As anticipated. positronium. the most conspicuous manifestation of the reduced mass is the shift that is observed in a comparison of the spectral lines of hydrogen. As pointed out earlier.18.M = PI + P2 (15. represents the quantum form of Newton's first law: the total momentum of an isolated system is constant.87) could equally well have been made before the quantization. r z and r. Prove that r. such that (15. R is unity.91).
such as a coordinate or a momentum measurement.99) is an eigenstate of the total momentum. R).. Show that the state of two particles with sharp momenta PI and P2' corresponding to the plane wave function I/J(rh r2) = exp(k PI .a) describes the relative motion of the two particles. and cp(X) = Yl/2'Trfl. a oneparticle state of sharp position (delta function in coordinate space. anywhere regardless of the position of particle 2 is constant and independent of the position Xl' Once a measurement shows particle 2 to be located at coordinate X2' then particle 1 is certain to be found at position XI = X2 + a.19. we are led to conclude that particle 1 after such a measurement is represented by. it appears that. An extreme example of a correlated wave function is afforded by the simple model of two particles confined to staying on the x axis and represented by an idealized amplitude: l/J(Xh X2) = l/J(x. corresponding to zero total momentum. If we assume that the quantum mechanical amplitude gives a complete (statistical) account of the behavior of a single system. and nowhere else. This is sometimes expressed by saying that states like (15. centered at XI = X2 + a) or sharp momentum (plane wave with momentum P2)' In their famous 1935 paper.90). X) = l/J(x)cp(X) = (27rli)112 S(XI .X2 . depends on the choice of measurement made on particle 2. which are factorable in relative coordinates. The fundamental coordinate basis states Irlr2) = Ir2)!r2) have this special character.93) into l/J(r.98).99) corresponds to the two particles being separated precisely and invariably by the distance a. These mysterious longrange correlations between the two widely separated particles and the strange dependence of the expected .X2) = 0 we see that (15. but the probability of finding one of the particles.a) = (27rli)112 S(x . and Rosen articulated the distress that many physicists feltand occasionally still feelabout these unequivocal conclusions of quantum mechanics.+ 0) ifr(XI l OXI OX2 .98) is also separable when it is transformed by use of (15. Hence.(O :. fI. Except for the special case (15. rl) exp(k P2 .97). which may be chosen as large as we please. even when the two particles are arbitrarily far apart (large a). after a measurement on particle 2 is undertaken. corresponding to eigenvalue zero. what can be known about the state of particle I. Einstein. wave functions of the type (15. The state represented by (15.99) Here a is a positive constant.4 Quantum Dynamics in Direct Product Spaces and Multiparticle Systems 361 amplitudes.X2). Podolsky. are gellerally not separable with regard to the two particles and are said to be correlated.87) and (15.97) do not exhibit correlations between the two particles. Exercise 15. The wave function l/J(x) = S(x . is a momentum eigenstate of the centerofmass motion. ' Since for any function l/J(XI . say particle 1. r2) (15.a) (15. if the momentum of particle 2 is measured and found to have the value P2' then particle 1 is certain to be found to have the sharp momentum value PI = P2' Thus. Most twoparticle states are not factorable like (15. depending on whether the coordinate or the momentum of particle 2 is measured.
The quest for a return to local realism in physics must thus remain unfulfilled.101) for the basis where for typographic clarity the quantum number labels. Exercise 15. show that the wave function in the mixed relative representation is (x. \ r. R).100) Correlated amplitudes like (15. which cannot be factored with regard to the two subsystems 1 and 2. Furthermore.79).(k)B~(e)I'I") be expressible in terms of nj + n z complex numbers as e. not afflicted with unaccountable actionsatadistance between measuring devices. that is. a term coined by SchrOdinger and illustrated dramatically in his famous cat allegory.362 Chapter 15 The Quantum Dynamics of a Particle behavior of particle 1 on the subjective fickleness of a distant human experimenter. By using the intermediate coordinate basis Ir. benz)) (15. For the correlated state 1'1") represented by (15. .a)S(P) (15. we persist in interpreting the formalism of quantum mechanics as providing complete statistical predictions of the behavior of single systems. Bell showed that such a program cannot be carried out. derive the transformation coefficients (rjrzlr.100). who has no means of interacting with particle 1. seemed to Einstein to signal a violation of the innate sense that the world of physics is governed by local realism. An alternative representation of twoparticle states is given in terms of the' 'mixed" basis states. k and states of the separate subsystems have been placed as arguments in parentheses instead of as the usual subscripts. where r is the relative coordinate vector and P the total momentum.6 6Por a reprint compilation of the key historical papers on entangled states and the puzzling questions they have raised. a more complete theory. In principle. ( :~~. might thus be eventually discovered..f ® IB~(e)(A. consistent with quantum mechanics. A general state may be expanded as 1'1") = 2: \A. are sometimes called entangled.) ® a(nj) (b(l) b(2) .20. and we have to accept the existence of quantum correlations between widely separated subsystems. as long as the theory is required to be local. Any theory built on strict local realism fails to reproduce some predictions of multiparticle quantum mechanicspredictions that have been verified experimentally to a high degree of accuracy.102). P).99) or (15.102) The state is entangled if and only if the amplitudes cannot be expressed in the form (15. pl'I") = S(x . But John S. and provide a correct but incomplete description of physical reality. Einstein tried to resolve this conflict by suggesting that quantum mechanical amplitudes pertain only to ensembles of systems. see Wheeler and Zurek (1983). it is not difficult to construct examples of entangled states for a system composed of two independent subsystems. rather than single systems. P).(k) k.99).(k)B~(e)I'I") (15. Using the basis states (15. The necessary and sufficient condition for the state of the composite system to be factorable with respect to particles 1 and 2 is that the nj X nz dimensional rectangular matrix of the amplitudes (A.
this formula is the normalization condition ('It I'It) = trace p = I (15. the Density Matrix. As a projection operator for a state I'lJ'). and I .pip are associated with the outcomes of measuring the positive operators P. and Information.106) If (15.pI'lJ') = trace(I<I»(<I>lp) = trace(P.p. and lOespecially Sections 4.109) Each projection operator corresponds to a different outcome of the proposed measurement. since owing to normalization.P. The probabilities pip and 1 .105) For A = 1.99) is entangled by making a (Fourier) expansion in terms of momentum eigenfunctions or any other complete set of oneparticle basis functions.103) contains all relevant information about the state.1 and 14. 5. and hence commuting.19). and Information 363 Exercise 15.5 The Density Operator. Check that the amplitude (15. Measurement.21.2. more generally.pip is the probability of finding the system not to be in state 1<1».!A!'lJ') = 2: (KiIAI'It)('ltIKi ) = i i trace(Ap) (15. as can be seen from ('lJ'IAI'It) = 2: ('lJ'IK.107) Since this can also be written as pip = 1(<I>I'lJ')I Z = (<I>I'I')('lJ'I<I» = (<I>lpl<l» (15. Building on the foundations laid in chapters 4. rankone (or onedimensional) projection operators Ph P z.2 and lOAwe characterize a complete orthonormal quantum measurement by considering a set of n mutually orthogonal.p and 1 . and its time development was considered in Sections 14. If this is the . the d'ensity operator p = 1'lJ')(:\If1 (15. space of the system). Measurement. that the diagonal elements of any density matrix are nonnegative.104) Except for an irrelevant phase factor.. the state I'It) can be recovered from the density operator as the eigenvector of p which corresponds to eigenvalue 1. or Hilbert. In a specific application of the formalism.pp) (15.)(K. an expression for a probability in terms of the density operator results: pip = 1(<I>I'It)lZ = ('It I<1»(<1> I '1') = ('ltlp. a complete commuting set of observables) symbolized by K. All expectation values can be expressed in terms of the density operator. the Density Matrix.108) we infer that p is a positive Hermitian operator and. 9. (15. The completeness is expressed by the closure relation (10044): 2: Pe = e=1 n 1 (15. in particular. The density operator for a state !'lJ'(t) was defined in Eq. P n (where n is the dimensionality of the state vector.p. (14. . The density operator is idempotent. The Density Operator. Somewhat imprecisely we say that pip is the probability of finding the system in state I<1». Pe = IKe)(KeI may be the projection operator corresponding to the eigenvalue K e of an observable (or.105) is applied to the projection operator A = I<1»(<1> I = P..
364
Chapter 15 The Quantum Dynamics of a Particle
case, we regard the values K e as the possible outcomes of the measurement. The probabilities of finding these outcomes are
Pe = trace(PeP)
(15.110)
Equation (15.109) guarantees that the probabilities add up to unity. If the state of the system happens to be an eigenstate IKe), the variance (or uncertainty) of K is zero, and K; can be regarded as the sharp value of this observable, akin to a classical observable. Somewhat casually, we call this procedure a "measurement of the set of observable operators K." Although the complete orthonormal measurements just described stand out prominently, it is possible to generalize the notion of a quantum measurement to include nonorthogonal operators. Thus, we assume the existence of a set of r positive Hermitian operators, (15.111) which are positive multiples of rankone projection operators Pj • The projection operators in (15.111) are not necessarily mutually orthogonal, nor do they generally commute. With their weights, Wj (0 :=; Wj :=; 1), they are subject to the completeness, or overcompleteness, relation
j=l
2: A = 2:
j
r
r
WjPj
=1
(15.112)
j=l
If the measurement is to be implemented on the system in any arbitrary state, completeness requires that the number of terms r in this sum must be at least as large as n, the dimensionality of the system's state vector space. The probability that the measurement described by the operators A j yields the jth outcome is given by the formula
Pj
= ('l'IAJ'l') = trace(Ajp)
::5
(15.113)
The condition (15.112) ensures that these probabilities ,add up to unity.
Exercise 15.22.
Using the Schwarz inequality, prove that Pj
wj
•
In this brief discussion of the generalized quantum measurement defined by the set of r operators A j , and technically referred to as a probabilityoperatorvalued measure (or POM), we only emphasize the significance of nonorthogonal terms in (15.113).7 Suppose that the state of the system is an eigenstate I<I>k) of the element A k = WkPk of the POM, corresponding to eigenvalue Wk. The probability of the jth outcome of the measurement defined by the POM is (15.114) This shows that the probability of the kth outcome (j = k) may be less than unity, and that of the other outcomes (j k) may not be zero (as would be expected if the states k and j were orthogonal). There is a quantum mechanical fuzziness inherent in nonorthogonal measurements. In these, unlike complete orthogonal measure
'*
7S ee Peres (1995) for more detail and many references to books and articles on quantum measurement and quantum information. For a clear discussion of entropy in the context of coding theory, see Schumacher (1995).
5
The Density Operator, the Density Matrix, Measurement, and Information
365
ments, Pk can generally not be interpreted as the probability of finding the system to be in state Icl>k), and this state cannot be associated unambiguously with just one of the possible outcomes of such a measurement. Nonorthogonal quantum measurements exhibit more peculiarly quantal features than orthogonal measurements of observables, whose eigenstates can be unambiguously associated with sharp values of the corresponding physical quantities, in a manner reminiscent of classical physics. As an illustration of a POM for a onedimensional system with an infinitedimensional Hilbert space, we draw attention to the closure relation (10.111) for the coherent states of a harmonic oscillator: (15.115) which is precisely in the form (15.112), applied to a continuously variable outcome, identified by the complex number a. If the system is known to be in the kth energy eigenstate of the harmonic oscillator, the probability density (per unit area in the complex a plane of Figure 10.1) for outcome a in a measurement ofthe nonorthogonal POM defined by the coherent states, A", = (1I1T) Ia)(al, is (15.116) Except for the factor w'" = 1I1T, this is the same as (10.110). So far, in this section, the density operator has merely served as an alternative to describing a quantum state by a vector in Hilbert space. It would be possible to formulate all of quantum mechanics exclusively in terms of density operators and their matrix elements, but the required mathematical techniques are generally less familiar than those pertaining to amplitudes and wave functions. (However, effective approximation schemes for complex manyparticle quantum systems have been invented using density operator and density matrix methods. s) In Chapter 16, we will illustrate the use of the density operator and its representative, the density matrix, for the simple case of a spin onehalf system whose spatial coordinate degrees of freedom are irrelevant and can be suppressed. We will find that the full benefit of using the density matrix accrues only when it is applied to a statistical ensemble of imaginary replicas of the system in the tradition of Gibbs, thereby creating a mixture of different quan.tum states. A mixture can be visualized as the set of probabilities, or relative frequencies, Pi' with which N different quantum states l'I'i) or the corresponding density operators Pi = I'l'i)('I'i 1 occur in the ensemble denoted by ~. We must require that
N
LPi = 1
i=1
(15.117)
but the states l'I'i) generally need not be orthonormal. Equations (15.105), (15.107), and (15.108) show that probabilities and expectation values for quantum states depend on the density operator linearly, whereas they depend on the state 1'1') quad
8S ee, for example, Parr and Yang (1989).
366
Chapter 15 The Quantum Dynamics of a Particle
ratically. It follows that all statistical predictions for the ensemble can be expressed in terms of the generalized density operator of the system,
P=
i=1
L PiPi = L
N
N
p;j'l'i)('I'i I
(15.118)
;=1
by the universal formula for the average value of an operator A
(A)
=
L
i=1
N
Pi('I'iI A I'l'i)
=
L Pi trace(Api) = trace(Ap)
i=1
N
(15.119)
The density operator P, like its constituents, Pi, is a positive Hermitian operator. If all probabilities Pi except one vanish, the density operator (15.118) reduces to the idempotent operator (15.103). It is then said to describe a pure state. Otherwise it represents a mixed state. Since in applications one usually employs a particular basis to represent states and operators, the same symbol P is often also used to denote the corresponding density matrix. All density operator relations that we derived for pure states at the beginning of this section carry over to the generalized density operator for a mixture, except for (15.104), which is quadratic in P and characterizes a pure state or onedimensional (rank one) projection operator. Instead, owing to the positive definiteness of the density operator, we have in general, (15.120) We have constructed the density operator P for a mixture from the assumed a priori knowledge of the N pure states Pi representing the ensemble ~ and the corresponding probabilities Pi' It is not possible to reverse this procedure and to infer the composition of a mixture uniquely. A given density operator P is compatible with many (generally, infinitely many) different ways of mixing pure states. We will presently quantify the information loss that is incurred in the mixing process. However, an exceptionally useful decomposition is always provided by a complete set of orthonormal eigenstates liJ) of the Hermitian density operator P and its eigenvalues Pj (15.121) where n is the dimensionality of the Hilbert space, and
j=1
L
n
jiJ)%1
=]
(15.122)
Some of the eigenvalues ~ may be zero, and there is a certain amount of arbitrariness in the choice of eigenvectors, if eigenvalues are repeated. In particular, if all eigenvalues of P are equal to lin, the density operator is proportional to the identity, P = (lin)], and the mixture is as random as possible. Borrowing a term from the physics of spatially orientable systems, a completely mixed state for which p = (lin)] is said to be unpolarized. Any POM composed of positive operators
5
The Density Operator, the Density Matrix, Measurement, and Information
367
Ak
= WkPk, which resolves the identity according to (15.112), can be employed to represent an unpolarized ensemble by writing
I r P= n j=1
L
WjPj
(15.123)
Exercise 15.23. Prove the inequalities (15.120). Hint: Trace inequalities are most easily proved by using the eigenstates of the density operator as a basis. For the second inequality, maximize the probability of finding the system in a superposition state
Exercise 15.24.
If the state of a quantum system is given by a density operator
(15.124) where 1'1'1,2) are two nonorthogonal normalized state vectors, show that the eigenvalues of the density operator are
PI.2 =
I ± VI  4PiP2(1  1('1'1 ~ 2
(15.125)
If a mixed state with density operator P is defined by a given probability distribution of N known pure states Ph P2' ... PN with probabilities Pb P2, ... PN' our incomplete knowledge of the state can be quantified in terms of the information and entropy concepts that are introduced in Section 2 of the Appendix. The Shannon mixing entropy (A.43) for this ensemble ~, denoted by H(~), is
H("f,) = 
L
i=1
N
Pi lnpi
(15.126)
We have chosen to express the entropy in terms of the natural logarithm, so that the nat is the unit of H("f,). The quantity H(~) is a measure of our ignorance of the state, A large mixing entropy H(~) implies a highly randomized ensemble. If the state of the system is pure (p 2 = p), the information is maximal and the mixing entropy is H("f,) = O. The information about the state is complete. If, on the other hand, all Pi are equally probable,
PI
= P2 = , .. = PN = 
1
N
(15.127)
and the mixing entropy is H("f,) = In N nats. In quantum information theory one investigates how, given a set of a priori probabilities about a quantum state, our ignorance and the entropy can be reduced, or information gained, by performing measurements on an ensemble "f,. The decomposition (15.121) of a given density operator in terms of its complete set of orthonormal pure eigenstates occupies a special place among the probability distributions compatible with p. Its mixing entropy is denoted by S(p) and defined as
S(p)
=  L Pi In Pi = i=1
n
i=1
L
n
(pilp In plpi)
= trace(p In p)
(15.128)
368
Chapter 15 The Quantum Dynamics of a Particle
Here, the function In p of the density operator is understood to be defined as in Eq. (10.30). Among all the different entropies that can be usefully defined, S(p) is singled out and referred to as the von Neumann entropy. It can be shown to be the smallest of all mixing entropies (15.126) for a given density operator:
H(fl,) =
2: Pi In Pi 2:::  2: Pj lnpj = S(p)
i=1
j=1
N
n
(15.129)
Thus, the ensemble composed of the orthonormal eigenstates of the density operator, the eigenensemble, is the least random of all the possible decompositions of p. In Chapter 16, this extremal property of the von Neumann entropy will be further demonstrated by several examples.
Exercise 15.25. If an ensemble ~ consists of an equalprobability mixture of two nonorthogonal (but normalized) states 1'1'1) and 1'1'2) with overlap C = ('1'11'1'2), evaluate the Shannon mixing entropy H(~) and the von Neumann entropy, S(p). Compare the latter with the former as Icl varies between 0 and 1. What happens as C ~ O? Exercise 15.26. A given ensemble ~ consists of a mixture of two equiprobable orthonormal states 1'1'1) and 1'1'2) and a third normalized state 1'1'3), which is itself a superposition (not a mixture!) 1'1'3) = CI 1'1'1) + C21 '1'2), so that the density operator is
(0
::5
P
::5
1/2)
(15.130)
Work out the 'eigenprobabilities of p and the Shannon and von Neumann entropies. Discuss their dependence on the mixing probability P and on the amplitudes CI.2' Entropy can be defined for any probability distribution. To gauge the predictability of the outcome of a measurement of an observable K on a system with density operator p, we define the outcome entropy:
H(K) = 
2: p(K
j=1
n
j)
In p(K)
(15.131)
Since the probabilities (15.132) can be calculated directly from the density operator, the value of the outcome entropy, H(K), is independent of the particular ensemble ~ which represents p. Again, the von Neumann entropy stands out, because one can prove that
H(K) = 
2: p(~) In p(~) 2:::
j=1
n

2: Pi In Pi =
i=1
n
S(p)
(15.133)
For the special case of a pure quantum state 1'1'), or p = 1'1')('1' I and S(p) = 0, the relation (15.133) makes the trite but true statement that there is generally an inevitable loss of information, if we know only the probabilities I(Kj 1 1 for measuring '1') 2 the observable K. We are missing the valuable information stored in the relative phases of the amplitudes. The fundamental significance of the von Neumann entropy S(p) should now be apparent. Thermodynamic considerations show that, multiplied by the Boltzmann
5
The Density Operator, the Density Matrix, Measurement, and Information
369
constant, k, the von Neumann entropy is also the form of the entropy whose maximum, subject to certain constraints, yields, according to the second law of thermodynamics, the equilibrium distributions for quantum statistical mechanics. We will implement this principle in Section 22.5, after an introduction to the quantum physics of identical particles. To complete the discussion of the density operator, we must give an account of its time evolution. If the density operator (15.118) for the system, with a Hermitian Hamiltonian H, is' given at some initial time to, each constituent purestate density operator Pi develops according to the equation of motion (14.20), Owing to the linear dependence of p on the components Pi' and the linearity and homogeneity of (14.20), the density operator p develops in the SchrOdinger picture according to the dynamical equation
zli If we assume that
. ap = [H, p] at
(15.134)
a nl in  p  = [H nl] at ' p
for any positive integer n, it follows by induction that
in
L
a n at
=
a in....!!.. pnl at
+
a nl inp  p at
=
[H, p]pnl
+
p[H, pnl]
=
[H, pn]
Hence, (15.134) can be generalized for any analytic function f(p) of the density operator:
in af(p) = [H f(p)] at '
. (15.135)
The equation of motion (14.18) for the expectation value of an operator A, which may be timedependent, can be equally well applied to a mixed state:
d(A) in ;](
=
([A, H])
+
faA) in\
at
=
trace(p[A, H])
+
in trace ( p aA)
at
(15.136)
By substituting A = f(p) in (15.136), and using (15.135), if follows that the expectation value of any function of the density operator is constant in time:
d  <f(p) = 0 dt
(15.137)
Exercise 15.27. Give a direct proof that for a general mixed state, <f(p) = trace(pf) is constant in time, by noting that the density operator evolves in time by a unitary transformation, p(t) = Tp(to)Tt. Exercise 15.28.
Prove that trace(p[f(p), H])
=
trace(f(p)[H, p])
=
0
(15.138)
370
and that consequently,
Chapter 15 The Quantum Dynamics of a Particle
(afa~») =
0
(15.139)
As an important corollary, it follows that the von Neumann entropy, which is the mean of the density operator function In p, remains constant as the system evolves in time:
dS(p)
dt
  trace(p In p) = 0
d dt
(15.140)
This exact conclusion is not inconsistent with the familiar property of entropy in statistical thermodynamics as a quantity that increases in time during the irreversible approach to equilibrium, because it holds only under the precise conditions that we have specified, including the idealization that the probability distribution of the statistical mixture representing the ensemble is fixed in time. The sketchy introduction to the concepts of the quantum theory of measurement and information presented in this section will be supplemented by concrete examples in the next chapter in the context of quantum mechanics in a vector space of only two dimensions. As we apply the results obtained in this section, we should remember that common terms like' 'measurement" and' 'information" are being used here with a specific technical meaning. In particular,this is not the place for a detailed analysis of real experimental measurements and their relation to the theoretical framework. We merely note that, in the information theoretic view of quantum mechanics, the probabilities and the related density operators and entropies, which are employed to assess the properties of quantum states and the outcomes of measurement, provide a coherent and consistent basis for understanding and interpreting the theory.
Problems
1. For a system that is characterized by the coordinate r and the conjugate momentum p, show that the expectation value of an operator F can be expressed in terms of the Wigner distribution W(r', p') as
(F)
=
('I'I FI '1')
=
JJFwCr', p')W(r', p') d r' d p'
3 3
where
Fw(r', p') =
J
e(i1~)p'·r"(r'  2"IFlr'
r"
+
2") d 3 r"
r"
and where the function W(r', p') is defined in Problem 5 in Chapter 3. Show 9 that for the special cases F = fer) and F = g(p) these formulas reduce to those obtained in Problems 5 and 6 in Chapter 3, that is, Fw(r') = fer') and Fw(p') = g(p').
9Recall that in expressions involving the Wigner distribution rand p stand for operators, and the primed variables are realnumber variables.
Problems
371
2. Show that the probability current density at ro is obtained with
jw<ro; r', p') =
~~ B(r'
 ro)
so that the current density at ro is
3. Derive the Wigner distribution function for an isotropic harmonic oscillator in the ground state. . 4. Prove that for a pure state the density operator 1'1')('1' 1 is represented in the Wigner distribution formalism by
pw(r', p') = (2'lTn)3W(r', p')
.
j(ro) =
JW(ro, p') ~ d p'
3
Check that this simple result is in accord with the normalization condition (p) = for the density operator. 5. For a free particle, derive the equation of motion for the Wigner distribution
aW(r', p', t) + p' . V.Wi(r' p' t) = 0
at
m
r
"
from the timedependent Schrodinger equation. What does the equation of motion for W for a particle in a potential VCr) look like? 6. Two particles of equal mass are constrained to move on a straight line in a common harmonic oscillator potential and are coupled by a force that depends only on the distance between the particles. Construct the Schrodinger equation for the system and transform it into a separable equation by using relative coordinates and the coordinates of the center of mass. Show that the same equation is obtained by first constructing a separable classical Hamiltonian and subjecting it to canonical quantization. 7. Assuming that the two particles of the preceding problem are coupled by 'an elastic force (proportional to the displacement), obtain the eigenvalues and eigenfunctions of the SchrOdinger equation and show that the eigenfunctions are either symmetric or antisymmetric with respect to an interchange of the two particles. .
CHAPTER
16
The Spin
The spin (onehalf) of a particle or atom or nucleus provides an opportunity to study quantum dynamics in a state vector space with only two dimensions. All laws and equations can be expressed in terms of two components and 2 X 2 matrices. Moreover, we gain insight into the effect of rotations on quantum states. The lessons learned here are transferable to the general theory of rotations in Chapter 17. Polarization and resonance in static and timevarying fields are characteristic spin features described by the theory and observed in experiments. The spin also lends itself to an explicit and relatively transparent discussion of the interpretation of quantum mechanics and its amplitudes, density matrices, and probabilities. In the quantum mechanics of twodimensional complex vector spaces, it is possible to concentrate on the intriguing features of the theory, untroubled by mathematical complexities.
1. Intrinsic Angular Momentum and the Polarization of t/J Waves. In Chapter 15, we were concerned with the quantum description of a particle as a mass point, and it was assumed that the state of the particle can be completely specified by giving the wave function 1/1 as a function of the spatial coordinates x, y, z, with no other degrees, of freedom. The three dynamical variables were postulated to constitute a complete set. Alternatively and equivalently, the linear momentum components Px, PY' pz also form a complete set of dynamical variables, since ¢ (p) contains just as much information about the state as l/1(r). The Fourier integral links the two equivalent descriptions and allows us to calculate ¢ from 1/1, and vice versa. It is important to stress here that completeness of a set of dynamical variables is to be understood with reference to a model of the physical situation, but it would be presumptuous and quite unsafe to attribute completeness in any other sense to the mathematical description of a physical system. For no matter how complete the description of a state may seem today, the history of physics teaches us that sooner of later new experimental facts will come to light which will require us to improve and extend the model to give a more detailed and usually more complete description. Thus, the wave mechanical description of the preceding chapters is complete with reference to the simple model of a point particle in a given external field, and it is remarkable how many fundamental problems of atomic, molecular, and nuclear physics can be solved with such a gross picture. Yet this achievement must not blind us to the fact that this simple model is incapable of accounting for many of the finer details. In particle physics and in many problems in condensedmatter physics, it is inadequate even for a first orientation. A whole host of quantum properties of matter can be understood on the basis of the discovery that many particles, including electrons, protons, neutrons, quarks, and neutrinos, are not sufficiently described by the model of a point particle whose wave function as a function of position or momentum exhausts its dynamical properties. Rather, all the empirical evidence points to the need for attributing an angular momentum or spin to these particles in addition to their orbital angular momentum, and, associated with this, a magnetic moment. For composite particles like protons
372
1 Intrinsic Angular Momentum and the Polarization of l/J Waves
373
and neutrons, these properties can be understood in terms of their internal quark structure, but leptons like electrons and muons appear to be elementary pointlike constituents of matter, yet nevertheless possess intrinsic angular momentum. What is the most direct evidence for the spin and the intrinsic magnetic moment? Although it was not realized at the time, Stern and Gerlach first measured the intrinsic magnetic moment in experiments 1 whose basic features are interesting here because they illustrate a number of concepts important in interpreting quantum mechanics. The particles, which may be entire atoms or molecules whose magnetic moment fL is to be measured, are sent through a nonuniform magnetic field B. They are deflected by a force which according to classical physics is given by
F = V(fL . B)
(16.1) ,
and they precess around the field under the influence of the torque '1" = fL X B. The arrangement is such that in the region through which the beam passes the direction of B varies only sl~wly, but its magnitude B is strongly dependent on position. Hence, the projection JLB of fL in the direction B remains sensibly unchanged, and we have approximately
(16.2)
By measuring the deflection, through inspection of the trace that the beam deposits on the screen, we can determine this force, hence the component of the magnetic moment in the direction of B. Figure 16.1 shows the outline of such an experiment. The results of these experiments were striking. Classically, we would have expected a single continuous trace, corresponding to values of JLB, ranging from  JL to + JL. Instead, observations showed a number of distinct traces, giving clear proof of the discrete quantum nature of the magnetic moment. Since the vector p. seemed to be capable of assuming only certain directions in space, it became customary to speak of space quantization. Stern and Gerlach also obtained quantitative results. They found that the values of P,B appeared to range in equal steps from a minimum,  p" to a maximum, p,. The value JL of the maximum projection of fL is conventionally regarded as the magnetic moment of a particle. In order to interpret these results, we recall Ampere's hypothesis that the mag
s
L_
Spin up
Spin down
Figure 16.1. Measurement of the vertical component of the magnetic moment of atoms in an inhomogeneous magnetic field (SternGerlach experiment). Silver atoms incident from the left produce two distinct traces corresponding to "spin up" and "spin down."
lSee Cagnac and PebayPeyroula (1971), p. 239.
374
Chapter 16 The Spin
netic properties of matter are attributable to electric currents of one form or another. Thus, the circulating currents due to electrons (of charge e and mass me) in atoms produce an orbital angular momentum L and a magnetic moment f.L connected by the classical relation, (16.3) which, being a simple proportionality of two vectors, is expected to survive in quantum mechanics also. Since any component of L has 2e + 1 eigenvalues, we expect the projection of f.L in a fixed direction, such as B, also to possess U + 1 distinct eigenvalues and to be expressible as
/LB =    m = 130m
ef1,
2me c
(16.4)
e. The Bohr magneton {3o is defined as
where the magnetic quantum number m can assume the values
ef1,
e, e + 1, e 
1,
{3o = 2m e c
(16.5)
and has 2 the value 9.27401 X 10 24 J/T = 9.27401 X 10 21 erg/gauss = 5.78838 X 10 5 eV/T. Since e is an integer, we expect an odd number (2e + 1) of traces in the SternGerlach experiment. It is well known that the classical experiment with a beam of silver atoms, passing through an inhomogeneous magnetic field, yielded instead two traces, i.e., an even number, corresponding to
/LB
=
±2mec
ef1,
=
±{3o
We may ask if the semiclassical arguments used above are valid when we contend with quantum phenomena. Equation (16.2) is p~rely classical, and we may wonder if its application to quantized magnetic moments has not led us astray. The answer to these questions is that like most experiments the SternGerlach experiment has components that are properly and correctly described by the laws of classical physics. For these are the laws that govern the experiences of our senses by which we ultimately, if indirectly, make contact with what happens inside atoms and nuclei. If the particles that the inhomogeneous field in a SternGerlach experiment deflects are sufficiently massive, their motion can be described by wave packets that spread very slowly; hence, this motion can be approximated by a classical description. The correct interpretation was given to the SternGerlach observations only after Goudsmit and Uhlenbeck were led by a wealth of spectroscopic evidence to hypothesize the existence of an electron spin and intrinsic magnetic moment. If one assumes that the electron is in an S state in the Ag atom, there can be no contribution to the magnetic moment from the orbital motion, and /L = eM2m e c measures the maximum value of a component of the intrinsic magnetic moment. Unlike a magnetic moment arising from charged particles moving in spatial orbits, this magnetic moment may be assumed to have only two projections, /La = ± {3o. According to the
2Cohen and Taylor (1996). This useful compilation of fundamental constants is updated and appears annually in the August issue of Physics Today.
Intrinsic Angular Momentum and the Polarization of l/J Waves
375
GoudsmitUhlenbeck hypothesis, we envisage the electron to be a point charge with a finite magnetic dipole moment, the projection of which can take on only two discrete values. It is now known that the electron magnetic moment differs very slightly from the Bohr magneton and has the value 1.001 159652 193 /30' owing to a small quantum electrodynamic correction. The muon magnetic moment similarly differs by a minute amount from its naIvely expected value m e/3o/mmuon. Goudsmit and Uhlenbeck also postulated that the electron has an intrinsic angular momentum (spin), but this quantity is not nearly as easy to measure directly as the magnetic moment. Without appealing to the original justification for the electron spin, which was based on experience with atomic spectra, we can marshal a fundamental argument for the assumption that an electron must have intrinsic angular momentum: From experiment we know that an electron, whether free or bound in an atom, does have a magnetic moment. Unless the atomic electron, moving in the electric field of the nucleus, possesses intrinsic angular momentum, conservation of angular momentum cannot be maintained for an isolated system such as an atom. To elaborate on this point, we note that, just as a moving charge is subject to a force in a magnetic field, so a moving magnetic moment, such as the intrinsic electron moment is envisaged to be, is also acted on by forces in an electric field. The potential energy associated with these forces is
f.t.XE c
v
which, for a central field [E = f(r)r], is proportional to f.t . v X r, or to
f.t·rXv=f.t·L
(16.6)
The factor of proportionality depends only on the radial coordinate r. If the Hamiltonian operator contains, in addition to the central potential, an interaction term like (16.6) proportional to f.t . L, the energy of the electron depends on the relative orientation of the magnetic moment and the orbital angular momentum. It is'apparent that L, whose components do not commute, can then no longer be a constant of the motion. Conservation of angular momentum can be restored only if the electron can participate in the transfer of angular momentum by virtue of an intrinsic spin associated with the intrinsic magnetic moment f.t. We conclude that the magnetic moment of a system must always be associated with an angular momentum (see Section 16.4). For leptons with no internal structure, the relativistic Dirac theory of the electron in Chapter 24 will provide us with a deeper understanding of these properties. However, at a comparatively unsophisticated level in describing interactions that are too weak to disturb the internal structure of the particles appreciably, we may treat mass, charge, intrinsic angular momentum, and magnetic moment as given fixed properties. As the presence of h in the formula JL = eh/2mc shows, the intrinsic spin and the corresponding magnetic moment are quantum effects signaling an orientation in space, and we must now find an appropriate way of including this in the theory. Wave mechanics was developed in Chapter 2 with relative ease on the basis of the correspondence between the momentum of a particle and its wavelength. This suggests that, in our effort to construct a theory that includes the spin, we should be aided by first determining what wave feature corresponds to this physical property. A scattering experiment can be designed to bring out the directional properties of waves. If a homogeneous beam of particles described by a scalar wave function l/J(x, y, z, t), such as alpha particles or pions, is incident on a scatterer, and if the
376
Chapter 16 The Spin
target is composed of spherically symmetric or randomly oriented scattering centers (atoms or nuclei), as discussed in detail in Chapter 13, we expect the scattered intensity to depend on the scattering angle 6 but not on the azimuthal angle cp that defines the orientation of the scattering plane with respect to some fixed reference plane. In actual fact, if the beam in such experiments with electrons, protons, neutrons, or muons is suitably prepared, a marked azimuthal asymmetry is observed, including a rightleft asymmetry between particles scattered at the same angle 6 but on opposite sides of the target. It is empirically found that the scattered intensity can be represented by the simple formula
II = a(6) + b(6) cos cp I
(16.7)
provided that a suitable direction is chosen as the origin of the angle measure cpo The simplest explanation of this observation is that'" representing an electron is not a scalar field, and that '" waves can be polarized. (Here, "electron" is used as a generic term. Polarization experiments are frequently conducted with protons, neutrons, atoms, nuclei, and other particles.) Figure 16.2 shows the essential features of one particular polarization experiment. A beam 10 of unpolarized electrons is incident on an unpolarized scatterer A. The particles, scattered at an angle 61 from the direction of incidence, are scattered again through the angle 62 by a second unpolarized scatterer B, and the intensity of the socalled second scattered particles is measured as a function of the azimuthal angle cp, which is the angle between the first and second planes of scattering. Owing to the axial symmetry with respect to the z axis, the intensities II and Ii are equal, but I R h, and the azimuthal dependence of the second scattered particle beam can' be fitted by an expression of the form (16.7). It is instructive to compare these conclusions with the results of the analogous double scattering experiment for initially unpolarized X rays. With the same basic arrangement as in Figure 16.2, no rightleft asymmetry of X rays is observed, but the polarization manifests itself in a cos 2 cp dependence of the second scattered intensity. Since intensities are calculated as squares of amplitudes, such a behavior suggests that electromagnetic waves may be represented by a vector field that is transverse and whose projection on the scattering plane, when squared, determines the intensity. The presence of a cos cp, instead of a cos 2 cp, term precludes a similar conclusion for the electron waves and shows that, if their polarization can be represented by a vector, the intensity must depend on this vector linearly and not quadratically. Hence, the wave function, whose square is related to the intensity, is not itself a vectorial quantity, and the polarization vector (P) will have to be calculated from it indirectly. In summary, the polarization experiments suggest that the wave must be represented by a wave function, which under spatial rotations transforms neither as a scalar nor as a vector, but in a more complicated way. On the other hand, the interpretation of the SternGerlach experiment requires that, in addition to x, y, z, the wave function must depend on at least one other dynamical variable to permit the description of a magnetic moment and intrinsic angular momentum which the electron possesses. Since both the polarization of the waves and the lining up of the particle spins are aspects of a spatial orientation of the electron, whether it be wave
"*
Geometry of a double scattering experiment. 10. we will see in Chapter 23 that the vector properties of electromagnetic waves are closely related to the intrinsic angular momentum (spin 1) of photons. but is not shown. The first plane of scattering at A is formed by 1o. and the second scattering at B and B I analyzes the degree of polarization. 2.2. is of great generality. and 14. The Quantum Mechanical Description of the Spin. The first scattering polarizes the beam. and h. we have so far implemented it only for the nonrelativistic theory of single particles that have zero spin or whose spin is irrelevant under the given physical circumstances. or particle. we now add a fourth discrete observable that is assumed to be independent of all the coordinate (and momentum) operators and commutes . in the plane of the figure. The angle between the two planes is ~.2 The Quantum Mechanical Description of the Spin 377 Ii 10 Figure 16. y. formed by lIt I R . and Ii. To complement the set of continuously variable fundamental observabIes x. it is not farfetched to suppose that the same extension of the formalism of wave mechanics may account for both observations. Although the formalism of quantum mechanics. which we developed in Chapters 9. Similarly. z for an electron. need not coincide with the first plane of scattering. The second plane of scattering. I.
10) It is instructive to study the behavior of the spin variable separately from the space coordinates and to consider a system whose state is described by ignoring the x. and. y. with two arbitrarily chosen distinct values of u. with one column and two rows. We denote its (eigen)values by u. z) 2 dx dy dz is thus assumed to measure the probability of finding the particle near x.8) Often u = + 1 is referred to as "spin up" and u = 1 as "spin down" (see Figure 16. We assume that the basic rules of quantum mechanics apply to the new independent variable in the same way as to the old ones. the probability amplitude or wave function for an electron now depends on the discrete variable u in addition to x. which is capable of taking on only two distinct values. There is no a priori reason to expect that such a modest generalization of the theory will be adequate. z coordinates and determined. t) = (r. make it reasonable to assume that a variable which can . such as in the normalization integral: f Il/J+(x. t) = (r. The mathematical apparatus of Chapters 9 and 10 can be applied to extend the formalism of wave mechanics without spin to wave mechanics with spin. + 1. In the coordinate representation. l/J+(r. z. z. y. zW dx dy dz =1 (16. we must now introduce an additional summation over the pairs of values which u assumes. tl "!') for "spin down.1). and may be written as l/Ju(r.11) . [l/J±(x. entirely by two spin amplitudes. y. y. t) = (r. Z)12 dx dy dz + f Il/J(x. at least to good approximation. now stands for a twocomponent spin wave function.378 Chapter 16 The Spin with them. u. y. Since space and spin coordinates are assumed to be independent of each other.. the splitting of the spectral lines of oneelectron atoms into narrow doublets. This spin variable. 1. t I"!'). for the projection of the magnetic moment in the direction of the field B. but the appearance of merely two traces in the SternGerlach experiment. tl '\ft) for "spin up" and l/J(r. y." Suppressing the time dependence of the wave function. as measured in the SternGerlach experiment. z) Z)) (16. This can be regarded as a twocomponent object composed of the two complexvalued amplitudes. l/J(x. y.9) where the matrix l/J. Wherever we previously had an integration over the continuously infinitely many values of the position variables.. We designate such a general spin state as X and write it as (16. it is natural to use a twodimensional matrix representation for the specification of the state: 1 l'\ft) ~ l/J = (l/J+(X' y. u= +1 u with ILB = = 1 en 2mc en with ILB = + 2mc  (16. is given a physical meaning by associating the two possible projections of the magnetic moment . take on only two different valuessometimes called a dichotomic variablemay ~ be a sufficiently inclusive addition to the theory. as we will see later. and of revealing the value ILB = =+= 130' respectively..
17) Two spinors are orthogonal if this product is zero. the isospin description of a nucleon. Hence. often referred to as a spinor. the bodily motion of a particle can be ignored or treated classically. X represents a state with spin (~:) = cIa + cz~ . a represents a state with spin "up. Study of the spin formalism in isolation from all other degrees of freedom serves as a paradigm for the behavior of any quantum system whose states can be described as linear superpositions of only two independent states. and only its spin degree of freedom need be considered quantum mechanically. the transverse polarization states of a photon (Chapter 23)." and "down.8) are a = (~) = and ~ = (~) ~ (16. we must require the normalization z z lell + Iczl This can be written as = (ef ei)(~:) =1 (16.12) Thus. Other examples are the coupling of the 2S and 2P states of the hydrogen atom through the Stark effect (Chapter 18). since we can discuss many experiments by assuming that the nuclei are at fixed positions and only their spins are subject to change owing to the interaction with a magnetic field.16) Given two spinors. The analysis of reflection and transmission from a onedimensional potential in Chapters 6 and 7 has already illustrated the convenience of the twodimensional matrix formalism. and the life and death of a neutral kaon (Problem 1 in Chapter 16). In many physical situations. The column matrix (16.15) if we remember that (16. X and X'. The study of nuclear magnetism is an example. Such pairs of orthonormal spinors span the basis of a representation. There are innumerable problems in quantum mechanics where such a twostate formalism is applicable to good approximation.12) are orthogonal and normalized. (16.13) Icll z is the probability of finding the particle with spin up. Such states are more than mathematical idealizations." respectively. + cie~ (16. the magnetic quenching of the triplet state of positronium (Problem 4 in Chapter 17).2 The Quantum Mechanical Description of the Spin 379 The complexvalued matrix elements CI and Cz are the amplitudes for "spin up" and "spin down. The two spinors a and ~ defined in (16. as ata = ~t~ = 1. represents a state vector in an abstract twodimensional complex vector space. The basis states of the representation defined by the assignments (16. but that have nothing to do with spin angular momentum. and Iezl z is the probability of finding it with spin down. .11). the (Hermitian) inner (or complex scalar) product is defined as xtx' = (ci d)(~D = eie.14) (16." In the general state.
u and v. The (Hermitian) adjoint At of At = (A1\ Ail) Ai2 Aiz and the expectation value of A in the state X is (A) =: (16. If we commit ourselves to a specific fixed representation. the same letter may be used for a physical quantity and the matrix (operator) that represents it.21) Z2 C2 Exercise 16. To endow this purely mathematical framework with physical content. if A is a linear operator (perhaps representing a physical quantity). As long as confusion is unlikely to occur. 9. What happens if they are the same? An arbitrary state can be expanded in terms of the orthonormal eigenspinors. For example. it appears as A = (A ll A2l l2 A) A zz (16. by direct calculation. and 10 for complex linear vector (Hilbert) spaces of n dimensions can in this simple case.23) are the probability amplitudes of finding the eigenvalues of A corresponding to the eigenspinors u and v. that the eigenvalues of any Hermitian 2 X 2 matrix are real and its eigenspinors orthogonal if the two eigenvalues are different. the mathematical complexity of the theory is significantly reduced. show that if and only if the expectation value of a physical quantity A is realvalued. where n =: 2.20) (ci c l2 n(A ll A )(Cl) = XtAX A A 2l (16. Prove. we must identify the physical quantities associated with the spin of a particle and link them with the corresponding Hermitian matrices. all equations and theorems for state vectors and linear operators can be interpreted directly as matrix equations.22) The expansion coefficients (utX) = (ui uD(~J and (vtX) = (vi vi)(~J (16. the same symbol can be used for a state and the spinor that represents it. A physical quantity of principal interest is the component of the electron's intrinsic magnetic moment in the direction of the . respectively. of any Hermitian matrix A: (16. is represented as (16.1. Since the state vector space is twodimensional. the spin formalism is much simpler than wave mechanics with its infinitedimensional representations. In the spin matrix formalism. be written out explicitly in terms of twodimensional matrices.19) where the components of A is g are denoted by d l and d z . which produces the new spinor g = AX. In many ways.18) and its action on the spinor X.380 Chapter 16 The Spin All definitions and manipulations introduced in Chapters 4. the matrix A is Hermitian. similarly.
25) where U is a matrix whose elements depend on the three independent parameters of the rotation only. However. here we extend the theory to spin states. the active viewpoint is the appropriate one. Since B can be chosen to point in any direction whatever. we first select this to be the z axis of the spatial coordinate system. We first consider a righthanded rotation of the physical system about the z axis. we suppose that the two spinors are related by X' = Ux (16. this assumption involves no loss of generality. Figure 11. Then the z component of the intrinsic magnetic moment of an electron is evidently represented by the Hermitian matrix /Lz = f30G ~l) (16. 3. which motivated the extension of the theory to dichotomic spin variables. Rotations of systems described by wave functions !/J(x.3 Spin and Rotations 381 magnetic field. which we choose to keep fixed as the rotation is performed. This is an active rotation. /LB. /Lz' and by choosing a different direction for B we can measure any projection. the axis of rotation fi and the angle cPo Since the physical content of the theory should be invariant under rotation. of JL.26) .24) since the eigenvalues of /Lz are to be +. which carries an arbitrary spin state X into a state X'. and the corresponding states may be represented by the basis spinors a and f3.f30 = +. the physical system that we describe by the quantum mechanical state vector is not isolated but is embedded in an environment of external fields or interacts with other systems. How are the other components of JL represented? The magnetic moment JL has three spatial components. such as </LX> = Xt /LxX. the distinction between these two kinds of rotations is purely a matter of taste. any such projection /LB must be represented by a Hermitian matrix with the eigenvalues .. /Lx' /Ly. We will now turn to this task. it follows that (16. The relation between these two spinors may be assumed to be linear. we stipulate that the three components of <JL> must under a rotation transform as the components of an ordinary threevector. y. If our twodimensional formalism is adequate to describe the physical situation. is calculated from matrices and spinors.ehl2m ec. keeping the coordinate axes fixed. Since an expectation value. Spin and Rotations.f30 and +f3o. to be distinguished from a passive rotation. which leaves the physical system fixed and rotates the coordinate system. as is more commonly the case. z) were already considered in Chapter 11.1 (b) pictures an active rotation by an angle cP about the z axis. In order to determine the matrices /Lx and /Ly.1. if. we cannot say how the components of <JL> transform unless we establish the transformation properties of a spinor X under rotation. e. and in any case we will see that a valid linear transformation representing any rotation can be found. and we generally prefer it. we expect that normalization of X implies the same normalization of X': x't X ' = xtutUX = 1 Since X is arbitrary.g. (As will be shown in Section 17. As long as we deal solely with the mutual relation between the physical system under consideration and the coordinate system.) Thus.
and U2 represents a second rotation R 2 also about an axis through the origin. will be considerably facilitated if we consider infinitesimal rotations first.e.27) = ut. A small rotation must correspond to a matrix very near the identity matrix. the imaginary coefficient ensures that J must be Hermitian if U is to be unitary. U for any finite rotation could be constructed from (16.i</J ft. n.31) If the three matrices lx. lz. fLN / J)N = exp(_i </J ft· fL J) (16. UR= lim(l . If U 1 represents a rotation R 1 about an axis through the origin.. then U2 U 1 represents another such rotation R 3 .eft.. the construction of U corresponding to a given rotation R.. and they are called the generators of infinitesimal rotations.382 Chapter 16 The Spin so U must be a unitary matrix. This integration is easily accomplished because any rotation can be regarded as ' the product of N = </Jle successive rotations by a small angle eabout a fixed axis (constant Ii): (16.32) or in the limit N ~ 00." . Our goal. e is the angle of rotation about this axis. i. obtained by performing first R 1 and then R 2 • In this way X is first transformed to X' = U1X. In particular. Their detailed structure is yet to be determined. The exponential function with a matrix in the exponent is defined by (16. according to Euler's famous theorem. and thus for a small rotation we write to a first approximation the first two terms in a Taylor series: U=I"i.33) or by the usual power series expansion. is said to represent this rotation.. ly..30). we could.. have obtained the same physical state directly from X by performing a single rotation R 3 .29) The phase factor has been put in.30) where ft is the axis of rotation. The necessary groundwork for the matrix algebra was laid in Section 3. i. ly. U1 (16. represented by U3 • Hence. The factor iffL has been introduced so that J will have certain desirable properties. J are matrices.J i (16. a unitary matrix has a unique inverse. Alternatively..e.33) as in elementary calculus.28) uut = 1 The unitary matrix U. the unitary rotation matrices are required to have the property (16.4. From this matrix equation we infer that det ut det U = Idet uI 2 = 1 Hence.. where we may read "matrix" for "operator. (16. and (16.30) by successive application of many infinitesimal rotations. and J represents three constant matrices lx. which corresponds to the rotation that takes X into X'. even though UR and Ii . and lz were known. which subsequently is transformed to X" = U 2 X ' = U 2 U 1X. by integration of (16. because all spinors ei<Px represent the same state.
Verify it algebraically for a righthanded rotation about the z axis and express it in 3 X 3 matrix form. It is of decisive importance to note that J itself is a vector operator. In standard 3 X (16. A z is called a vector operator A if the expectation .) are an example of matrices that must satisfy this condition.29). X) ft(n x . Exercise 16. (J) must also transform like a vector. Starting with an infinitesimal rotation about the unit vector R = exp(cf> fi . where Ax.cos cf» + ft X (A) sin cf> (16. The application of the condition (16..36) x. values of Ax. we obtain (16.. Exercise 16. A y. A y . Le. as in Section 11. A z transform under rotation like the components of a vector. As stated at the end of the last section. (A z ) transform as the components of a vector. it is convenient not to attack this problem directly.. The transformation properties of a threevector (A) = XtAX under an active rotation are characterized by the equation (A)' where (A)' = 3 matrix notation.t AX' is the expectation value of A after rotation. This follows from its definition as the generator of the infinitesimal rotation: (16. n z ).33) is actually the solution to our problem.2.3 Spin and Rotations 383 We still have to derive the conditions under which a matrix of the form (16. and thus J is a vector operator.34) Multiplying on the left by Xt . Check the transformation (16. Generally.3. Hence.35) where the expectation value (J) is taken with respect to the state X. A z are three matrices (operators) such that the expectation values (Ax): (A y ). the scalar product ft . J z ' which so far have not been specified at all. but instead to discuss first the rotational transformation properties of a vector (A).38) .37) where R is the usual real orthogonal rotation matrix (with det R = 1) familiar from analytic geometry and corresponding to an active rotation. prove that the rotation matrix R can be represented as (16.36) by visualizing a threedimensional rotation.. Since ft is a vector.29) will lead to severe restrictions on the possible form of the Hermitian matrices J x . ny. the components of the magnetic inoment (. The inner products are invariant under a unitary transformation that represents an arbitrary finite rotation. A y .1. However. J y . represents the rotation R and satisfies the basic requirement (16. this equation appears as = (A) + ft X (ft X (A»)(l . applied simultaneously to both X and X'. (J) is also a rotational invariant. a set of three matrices Ax.
n X (n X A) cos 4> + ft X A sin 4> (16.e problem of the angular momentum operator was taken. p. n . Since X is an arbitrary state.46) for finite rotations. verify that the commutation relations for A and J assure the validity of the condition (16. n . 2.. with values 1. A). .2. J] = inn X A (16..30) on the lefthand side of this equation and equate the terms linear in B on the two sides. Employing the techniques developed in Section 3.!. JA .45) Exercise 16.40) We now substitute the expression (16. except for the replacement of the orbital angular momentum operator L by J. it follows that . Taking care to maintain the order of noncommuting operators. since J itself is a vector operator and must satisfy (16. This generalization was already anticipated in Section 11.A . See also Mathews and Walker (1964). explicitly.41) or An .36) or.36) reduces to (A)' = (A) + ft X (A)B (16.42) which is exactly the same condition as Eq.39) o 0 0 are three antisymmetric matrices. 430. y.43) or. J] = inn X J (16.44) The LeviCivita symbol Bijk was defined in Section 11.4. where an algebraic approach to the eigenvalu. For an infinitesimal rotation.!.384 Chapter 16 The Spin where Xy = 3 0 0 00 ( 1 0 1) 0 0 Xz = ( 1 0 0 0 1 0) (16. k. J .4. Work out their commutation relations an9 compare them with the commutation relations for the components of angular momentum. exp(i ft· J4»)A exp ( i n· J4» n(n .1.5.42): [J.n . (16. (11. JA = [A. z. I[ii'~] = insijkik or sijJiij = inh I (16. n .19) derived in Section 11. 3Biedenharn and Louck (1981). using the subscripts i. we may combine these commutation relations symbolically in the equation JXJ=inJ (16. 3 to denote the Cartesian components x. j. We can make use of the results derived there. J = n X A n n (16. n .
44) imply that the trace of every component of J vanishes. is represented by the diagonal matrix (16. (16.) Although they ~ere prompted by our interest in the twodimensional intrinsic spin of the electron. That a closed matrix algebra satisfying these commutation relations can be constructed for every nonnegative integer n was already proved in Section 11. We now see that 4Biedenharn and Louck (1981). it can be shown that the commutation relations for J are not only necessary but also sufficient for the unitary operator (16. and Spin Angular Momentum: Following the usual convention.48). The commutation relations (11. respectively.4 (For n = 2 a proof will be given in Section 16.4 The Spin Operators.47) so that the matrices UR representing a rotation are unimodular.46). (16. The Spin Operators. It follows that J z must itself be a diagonal matrix.2 that the z component of the vector operator fl.29) takes the more specific form . The basis vectors (or basis spinors or basis kets or basis states) of this representation are the eigenvectors of Jz. we supposed in Section 16. and Spin Angular Momentum 385 Since the trace of a commutator is zero. In the remainder of this chapter. when we deal with angular momentum in more general terms. Exercise 16. implying that the matrix U = exp (  k <f> Jz ) must be diagonal in the representation we have chosen.2. In particular.24) are identical to (16. the commutation relations (16. the commutation relations (16. .4. Pauli Matrices. det UR = 1 (16.24) and that the components Cl and Cz of the spinor X are the probability amplitudes for finding JLz = f3o (spin up) and +f3o (spin down). The states X and X' connected by the unitary matrix U in (16. prove that if the z component of some vector operator is represented by a diagonal matrix. we confine ourselves to the case n = 2.29). The problem of constructing the matrices J in a representation in which J z is diagonal has already been completely solved in Section 11. Section 3.29).48) Applying successive finite rotations to a vector operator and using Eq.5.44) would then have to be satisfied by three n X n matrices.33) to represent rotations and satisfy the requirement (16.2. by (10.25) could have n rows. we then conclude that e icp(R!. and all results would have been essentially the same. Pauli Matrices.5. Hence. A rotation ~bout the z axis can have no effect on these probabilities. J z must also be diagonal (as must be the z component of any vector operator). From the commutation relations.44). the intrinsic magnetic moment. If we evaluate the determinant on both sides of Eq.. however. 4. none of the arguments presented in this section have depended on the dimensionality of the matrices involved.43) or (16. (16.R2) = ± 1 and (16. and we must now explicitly determine the Hermitian 2 X 2 matrices J which satisfy the commutation relations. We will thus be able to use the results of this section in Chapter 17.
49) It is customary and useful to define a vector operator (matrix) 2 X 2 matrix J: proportional to the J=O' Hence. . Prove that (16. this implies .7.6.50) = Ux + IUy .52) Some simple properties of the Pauli matrices are easily derived.56) where k is a constant number. Also show that no matrix exists which anticommutes with all three Pauli matrices. + uyux = 0. The traces of all Pauli matrices vanish: trace Ux = trace u y = trace Uz = 0 (16.8. which correspond to j = 1/2 and m = :!: 1/2. which generally holds only for n = 2. Exercise 16. is the simplest illustration of the WignerEckart theorem which will be derived in Chapter 17. U+ h 2 (16. Prove that the only matrix which commutes with all three Pauli matrices is a multiple of the identity.42) that the trace of any vector operator is zero.42) and (11.43) (Exercise 11. Prove that the Pauli matrices are unitary and that 2 2 2 U x = Uy = U z = (16. = (0 2) 0 0 (16.53) Exercise 16. iJy = h (0 0 1) 0 and J.51) from which we obtain the celebrated Pauli spin matrices. (16.386 Chapter 16 The Spin for the description of the spin of the electron we must use as a basis the two eigenstates of J z and J2. and so Exercise 16. IJy = h (0 0) 1 0 0' (16.that the zcomponent A z of every vector operator is proportional to J z and consequently that all vector operators A are just multiples of J: A=kj=kO' h 2 (16.55) which is a reflection of the general property that the trace of any commutator of two matrices vanishes.= Jx  . (11. The proportionality of A and J. It follows from the commutation relations (16.11) we obtain the matrices J + = Jx + .54) and that any two different Pauli matrices anticommute: uxuy forth. From Eqs. In the twodimensional case (n = 2).
0' sin w) I (16.54) and (16. it can always be U = eiY(1cos w + in . The connection between this group and threedimensional rotations will be made precise in Chapter 17. In the twodimensional spin formalism. The set of all unitary unimodular 2 X 2 matrices constitutes the group SU(2).61) Comparing the last two expressions. 0'. We may now write the rotation matrix (16.60) which is a generalized de Moivre formula. and n is the axis of rotation. cf> UR = exp ( .33) takes the form (16. 0' sin w) (16.58) where "y and ware real angles. all coefficients in (16.. we see that every unitary matrix with "y = 0 represents a rotation.59) Applying the identity (16. A)( 0' . Any unitary 2 X 2 matrix can be written in this form. we see that (16.48). and any 2 X 2 matrix can be represented as A = A0 1 + Al ax + A2 a y + A3 a z = A0 1 + A . B + iO' . by the relation (16. For "y = 0 we have det UR = 1.62) One simple but profound consequence of this equation is that for cf> = 27T we get U = 1. and the matrix URis unimodular.57) must be real. 0' sm 2 2 2 (16. The doublevaluedness of the spin rotation matrices is sanctioned.11. 0' (16. Specialize to the case Ao = o and A = n. A cf>'A . A full rotation by 3600 about a fixed axis.I. this sign change of spinors under rotation is no obstacle to their usefulness. useful identity If A and B are two vectors that commute with prove the I (0' ..58) is the same as I U = exp(i"y + iwn . However.D ...i .10. Take advantage of the properties (16. Exercise 16. which is equivalent to zero rotation (or the identity). since all expectation values and .lD . thus changes the sign of every spinor component. Pauli Matrices. (A X B) I (16. the rotation matrix (16.9. a y.33) in the form .. Vectors (and tensors in general) behave differently: they return to their original values upon rotation. although not required.57) If A is Hermitian.4 The Spin Operators.2w. B) = A . 0' ) = 1 cos . and n is a real unit vector. expressed as Show that if U is a unitary 2 X 2 matrix. 0') = e iY (1 cos w + in .55) of the Pauli matrices to 'work out the eigenvalues and eigenspinors of A in terms of the expansion coefficients Ao and A. where n is a realvalued arbitrary unit vector. Exercise 16. a z are linearly dependent. The angle of rotation is cf> = . ax. Exercise 16.59) to the power series expansion of an exponential. and Spin Angular Momentum 387 The four matrices 1.
. rendering them immune to the sign change. and (16. cP "2 .48). Exercise 16.24) that the intrinsic magnetic moment of an electron is given by f.L = 13oa (16. A vector V in ordinary space must not be confused with a (state) vector such as X in a (complex) vector space. Since the spin rotation matrices are unimodular (det U = 1). equivalent to an overall rotation R3 . we consider the example of a rotation about the x axis by an angle cPo From (16. (16. we are led back to (16. We must now find the matrix representation of the physical observables that are associated with an electron or other spin onehalf particle. In the context of this chapter. the state vector is threedimensional.62). but they are connected.Vz sin V~ = Vy sin cP + Vz cos V~ = cP cP The differences between these two transformati~ns are apparent. the vector operator a is essentially unique.63) if UR is given by (16.lC2 . prove di U"kaUR = fi(fi .64) thus completing the program of determining the components of f.25). but in other situations. we can conclude that utmaU2 U 1 = U.63) is the expression for the rotated form of a vector. the latter is represented by a twodimensional spinor. a) . proving that in the case n = 2 the commutation relations are not only necessary but also sufficient to ensure the validity of the group property (16. on the other hand. rectly that Using the special properties of the Pauli matrices.388 Chapter 16 The Spin matrix elements depend bilinearly on spinors.56). such as when describing the intrinsic degree of freedom of a spinone particle.12. according to (16. we obtain for the spinor components: I Cl = Cl cos = lC I c P . it is evident that if we perform in succession two rotations R 1 and R2 . a] = 0 From Exercise 16. a) cos cP + 0 X a sin cP (16.52). "2 sm C2 C2 . . If A is a vector operator.8 we thus infer that U2 U 101 must be a multiple of the identity. It may be helpful to comment on the use of the term vector that is current in quantum mechanics. the spinor transformation induces the correct transformation among the components of the expectation value (A).fi X (0 . transform according to Vx V. To contrast the different behavior of spinors and ordinary vectors under rotation.48).aU3 or [U2 U 1 01. .62) Since the righthand side of (16. we conclude from (16. Since.L. = Vy cos cP . "2 cP+ sm cos cP "2 The components of a vector V.
Indeed. and Spin Angular Momentum O 389 What about the intrinsic angular momentum of the electron.67) since both Sand L are just different realizations of the generator J. L is applied to 0/.42) for a vector operator twice. • L. The interaction energy (16. we let J = A and S and replace ft by L: [S. which was introduced in its general form in Section 16. S] = inL X S .6) responsible for compromising the spherical symmetry of the central forces is proportional to fA. A term of the form 0' . = 1 . according to (16. like L. when it is added to the orbital angular momentum L. is a vector operator. z (16. y. L is invariant under this rotation. L . As was explained in Section 16. which in turn. Since S.65) of a particle with spin under an infinitesimal threedimensional rotation must be given by UR = l iAS . the spin angular momentum S is nothing other than the generator (16. S .h en . z) of the coordinates. an interaction of this form arises in atoms as a magnetic and relativistic correction to the electrostatic potential. L] = inS x L Next. Pauli Matrices. the total angular momentum. In the presence of a spinorbit interaction. (16.1. It is our hope that an intrinsic angular momentum S can be defined in such a manner that. but 0' couples the two spinor components. y. First. We express the wave function 0/ for the state of the particle in the spinor form (16.64). which is legitimate because S commutes with L.66) is the desired constant of the motion.4 The Spin Operators. It produces a fine structure in atomic spectra. L is no longer a constant of the motion. the operator L acts only on the functions o/±(x.3.66) will again be a constallt of motion. L for a spin onehalf particle. 0/ = (o/+(X' y. is proportional to 0' .68) The scalar operator 0' .65) When the interaction 0' . its spin? It was shown in Section 16. We can verify this identification by employing the commutation relation (16. and (16. In nuclei the spinorbit interaction has its origin in strong interactions and has very conspicuous effects. (16.h en·  i AL i AJ h en . Z») o/(x. it must be proportional to 0'. This yields the equation [L. in addition to its orbital angular momentum.50) of rotations for spinors: S=O' 2 n (16.9). we let J = A = L and replace the vector ft by S. L in the Hamiltonian is often referred to as the spinorbit interaction.1 that conservation of angular momentum is destroyed unless the electron is endowed with an intrinsic angular momentum. The unitary operator that transforms the state (16.
69) Hence. Under these conditions. 5. and we say that the electron has spin 1/2. Quantum Dynamics ofa Spin System. is the subject of Chapter 17. if there are no timedependent exteniatinfluences acting and the system is invariant under translation in time. i'" dx(t) = dt HX(t) (16. it follows that J = L + S commutes with the operator L . + au. No other combination of Land S would have satisfied this requirement.67) has the two eigenvalues + ". corresponding to s = 1/2 if we express 8 2 as s(s + 1)". giving X(t) = exp ( k Ht)X(O) (16. which places the emphasis on symmetries. Show that no operator of the form L L + (1i/2)0' commutes with the scalar 0' .13.70) can be integrated. Obviously.2/4. but since we want to apply the theory to systems that can decay. The essential feature of the equation of motion is its linearity. J = L + 8 is the generator of infinitesimal rotations (multiplied by"'). ' Thus. we note that (16. such as the spin of a particle in isolation from other degrees of freedom. it is convenient to introduce the eigenvalues of H. independent of t. hence. Furthermore. and conservation of angular momentum is merely a consequence of the invariance of the Hamiltonian under rotations. with eigenvalue 3". H must be a constant matrix. we see that when the spin is taken into account. the time development of a twocomponent state or spinor X(t) is governed by the equation of motion. S and is indeed conserved in the presence of a spinorbit interaction.AI) = 0 (16. we will at this stage not assume that H is necessarily Hermitian. Exercise 16. (16. is 1/2. any component of the intrinsic angular momentum S defined by (16. The maximum value of a compone~t of S in units of ".70) where the Hamiltonian H is in this instance a 2 X 2 matrix characteristic of the physical system under consideration. which preserves superpositions.2.72) . Eq.71) in terms of the initial state X(O). which are defined as the roots of the characteristic equation det(H . In the SchrOdinger picture. As usual. L. The general dynamical theory of Chapter 14 is directly applicable to any physical system with two linearly independent states. This broad viewpoint. we have S XL = L X Sand S· L = L· S. other than J = Evidently.390 Chapter 16 The Spin Owing to the commutivity ofL and S. any spinor is an eigenspinor of S2./2 and h/2.
81) . If fez) is a (unction of a complex variable. the function f(H) of the matrix H is a new matrix defined by the relation . its eigenvalues will generally not be real. f(H)x = cJ!(H)XI + c2f(H)X2 = cJ!(E1)XI + c2f(E2)x2 (16.76). we have = E I.80) If the matrix H is Hermitian. the eigenvalues of H are complex numbers and can be expressed as (16.tlfi _ EleiE2ttfi) E2 . If the characteristic equation has only one distinct root.76) may be applied to expand the time development operator f(H) = exp(  k = Ht) in the form 1 (E2eiE.Ht) h + H (eiE2ttfi _ eiElttfi) E2 . (16. E2 Prove Eq.76) is seen to hold. E 2 with E 1 *" HXI = E1XI and HX2 = E2X2 (16.74) If H is not Hermitian. If the Hamiltonian matrix is not Hermitian.73) and an arbitrary twocomponent spinor may be expanded as X = CIXI + C2X2 (16. This must certainly happen if H represents the energy.78) answers all questions about its time development. Xt X is constant and probability is conserved. and derive (16.78) if E 1 E 2 • A system whose Hamiltonian has exactly two distinct eigenvalues may be called a twolevel system.5 Quantum Dynamics of a Spin System If there are two distinct roots . E 1 and E2 are real numbers and the corresponding eigenspinors are orthogonal.14. the equality (16. The formula (16.79) and if H is constant in time this may be integrated to give Xt(t)x(t) = xt(O)e(itfi)Htte(itfi)HtX(O) (16. Xt(t)(H  Ht)X(t) (16.E 1 (16.E 1 exp(i. From (16.77) from it in the limit E2 ~ E 1• Equation (16.75) By substitution into (16..70) it follows in the usual way that dt [Xt(t)x(t)] *" d = i r.75).\ 391 E 2.77) Exercise 16. so that = E h the preceding equation degenerates into (16. If H is Hermitian.
The spinor ('Yb 'Y2 real) (16. respectively. show that the time rate of change of the total probability Xt(t)X(t) is negative for all states X at all times. Exercise 16.3. we have so far characterized the states in terms of twocomponent spinors. but the analysis is more complicated. However.392 Chapter 16 The Spin where the real parts. Verify this conclusion using the results of Exercise 16. if r 1. and the spin formalism can be adapted to all such systems. the two eigenstates are decaying states. E Ob E 02 . As an application. We confine ourselves to the case of twostate systems and use the ordinary electron or nuclear spin 1/2 in a magnetic field as the standard example of the theory. Conversely.85) characterizes a particular state. is applicable in many different areas of physics. If the imaginary parts. show that r 1 and r 2 defined by (16.82) Unless the two decay rates are equal. the dynamics of twostate systems. As was mentioned in Section 16.81) are positive. are positive. we consider some other methods of specifying a state. One gets immediately (f3IT(t. are the energy levels. due to interference between the eigenstates of H.15 as an example. Qualitatively similar effects occur in other fewstate systems.E = E02 . Often a twostate system is prepared or created in a state other than an eigenstate of the Hamiltonian H. and its time development is subsequently observed. O)la) = f3 t e(iln)Hta = (0 The probability obtained from this expression exhibits an interference term. displaying intriguing oscillations in time. In this section. In terms of the energy difference !!. since X depends on four real parameters. r 1 and r 2. Density Matrix and Spin Polarization. Exercise 16. Xl and X2. 6. of H are orthogonal (implying that H is a normal matrix). the state does not generally follow a pure exponential decay law. In many applications. consider the example of the probability amplitude for a transition from an initial state a ("spin up") to a state 13 ("spin down").70) is the superposition (16.16. The general solution of the dynamical equation (16. the same physical state can be described by different spinors. with or without decay. conservation laws and selection rules cause a decaying twolevel system to be prepared in an eigenstate of u z ' say a = (~). In discussing twolevel systems.E01 and the decay rates r 1 and r 20 calculate the probabilities of finding the system at time t in state a or state 13. but measurements can . If the Hermitian matrix r = i(H .15.Ht) is positive definite.2 > 0 and if the two decaying eigenstates.84) where a and b are generally complex constants. and governed by the simple normal Hamiltonian matrix H = al + bux (16.
Since p is Hermitian and its trace equals unity.17. the density matrix of a pure state X is Z * *) _ clc!) _ _ p . p can be expanded in terms of the Pauli matrices u x . (15.86) which requires that trace p = 1 (16. of CI and C2' If X is normalized to unity Icll z + 1c21 z = 1 (16.89) IClizAll + clc!A zl + czciA lz + 1c21 zA zz Z lcll* I Z A AIZ) = trace(pA) CIC!)(All A = trace ( C2CI I Cz ZI ZZ in accord with Eq.88) According to the probability doctrine of quantum mechanics. and discussed in detail in Section 15.5. show that the probability of finding A~ in a measurement of A on the state X is given by l(utxW = trace(pPA \) = trace p ( AI A.92) It is immediately verified that p has eigenvalues 0 and 1. Le. Exercise 16. The eigenspinor that corresponds to the latter eigenvalue is X itself. Pz are three real numbers given by Px = 2 Re(cicz) Py = 2Im(cicz) Pz = ICliz ._A) = A~ trace(puu t ) (16. u z ' and 1. and this is acknowledged by postulating that X and ei"'x (ex: arbitrary. real) represent the same state. (16.57) most generally be represented as (16.6 Density Matrix and Spin Polarization 393 give us only two parameters: the relative probabilities Ic11 2 : Ic21 2 and the relative phase.105). An elegant and useful way of representing the state without the phase arbitrariness is to characterize it by the density operator defined in (14.(CI)( CI C2 . Like any 2 X 2 matrix. it can according to (16.90) where PAl = uu t represents the projection operator for the eigenvalue A~. its knowledge exhausts all that we can find out about the state.XXt .(lcII* I 12 (16. u y. 11 . corresponding to the distinct eigenvalues Ai and A~. Py.87) Cz CZCI C2 subject to the normalization condition (16. In the context of twostate quantum mechanics.86) the only remaining redundancy is the common phase factor of the components of X.19) as p = Iqr)(qr I.91) where Px .Iczl z (16.93) .12.. The expectation value (A) of any operator A is expressed in terms of pas: (A) = XtAX= (ci = c!)(A A 21 ll 1z A ) (CI) A Z2 C2 (16. If A is a Hermitian matrix with eigenspinors u and v.
when p is applied to an arbitrary state cp.lP.97) proving that P transforms like a vector under rotations.91) is required to be idempotent and the identity (16. If (16. . We thus see that p projects cp in the "direction" of X. 0 = . 0) (16. Prove that the probability Po of finding this particle to be in a state Po = represented by the polarization vector 0 is i trace[p(l + 0 . It follows that the density matrix is idempotent: (16. The vector P is also known as the polarization vector of the state. It may be characterized by the two spherical coordinates.96) Hence. the unit vector P may legitimately be said to point in the direction of the particle's spin. We get from this and 1 where use is made of the fact that trace 0' = 0 and a~ analogous equations for a y and a z the simple formula IP = (a> = trace(pO') = trace(0'p) I (16. Exercise 16. The expectation value of ax in the state X is (aX> = trace(pax) = '2 trace ax + '2 P 1 1 . Given a spinor X= (ei~ cos e'f3 sin 8 8) (16.100) and show that this result agrees with expectations for 0 = P.98) Hence. we find that the spinor X is an eigenspinor of the matrix P . 0')] = i (1 +P .99) calculate the polarization vector P and construct the matrix UR which rotates this state into (~). trace(O'ax) = '2 Px . The matrix p applied to it gives zero.P. as it should be.394 Chapter 16 The Spin The other eigenspinor must be orthogonal to X.19. the state is characterized by two independent real parameters. we have (16. which specify a point on the unit sphere.93). Hence. Combining (16.95) Exercise 16. we obtain (16. and 0 .94) since X is assumed to be normalized to unity.91) with (16. trace(a. O'X = X (16.) = Px = 1. the polar and azimuthal angles.59) is employed.87) that the density matrix for a pure spin state is idempotent and has eigenvalues 0 and 1. 0': p. Show directly from (16.18.
5.6 Density Matrix and Spin Polarization 395 Although the language we have used in describing the properties of P refers to spin and rotations in ordinary space.103) . The time evolution of the density matrix p can be obtained from the equation of motion for X. The general elliptic polarization state of a light wave or photon is a superposition of the basis states in a twodimensional complex vector space. Any two "orthogonal" polarization states may be chosen as the basis states for the representation. = Hp . Using the definition (16. and the formalism allows us to define a "polarization vector" corresponding to the state of any twolevel system. but the two (right and left) circular polarization states are usually preferred. The formalism is particularly useful to describe the polarization states of electromagnetic radiati.XXt H iii iii or Iih ~ Exercise 16. di.101) where H is assumed to be a Hermitian 2 X 2 matrix. d(A) = (AH  HA) + iii \ at laA) (16.102) All of these equations are merely concrete matrix realizations of the general formalism of Sections 14. dX (16. stationary state. The elements of the corresponding density matrix are essentially the Stokes parameters of the polarization state.87) of the density matrix.jf . in the context of the polarization of light P is best referred to as the Poincare vector. which in optics is known as the Jones vector space. Examples are the isospin space in nuclear physics and the abstract polarization vector which can be defined to represent two isolated atomic states in interaction with an electromagnetic field as might be generated by laser light. we obtain  ap dX = . The polarization P is then a "vector" in an abstract threedimensional Euclidean space..pH I (16. To combat the almost inevitable confusion caused by the double meaning of the term polarization. in recognition of the mathematician who introduced the twosphere as a convenient tool for representing the elliptic polarization states. and the operator induces "rotations" in this space. Derive the properties of the density matrix that represents a The equation of motion for any expectation value (A) is familiar: Iii .= HX dt . the concepts have more general applicability.1 and 15. The vector P provides yet another representation of the polarization state of light.on.20.Xt at dt + X dXt dt =  1 1 HXX t .
0') (16. Q) Q(Po .105) implies that P precesses about Q with a constant angular velocity (16.63). If Q is a ~onstant vector.107) is the solution of (16. This is merely another way of saying that.103). Since Q . we derive dP dt = d(O') = 1. Their energies are given by the eigenvalues of H. which may be functions of time.108) . Show that if Q is constant. To obtain a simple formula. is of physical interest. (16. and the spin commutation relations summarized in the equation 0' X 0' = 2iO'. P) dt = 2P . Q 1 X (0') or (16. but only the energy difference.Q . 0' has the eigenvalues + 1 and 1. Q)] cos wQt + Q X Po sin wQt + Q X (Po X Q) cos wQt + Q X Po sin wQt (16. Verify that (16..21. it is seen from (16. These two vectors represent the two stationary states of the system. dt ih 2ih = 2ih (Q X (0' X 0') = 0' .104) where Qo and the three components of the vector Q are real numbers.104).106) If P(O) the solution of (16. 'l) sm A A. 0'0') 1 h. it is convenient to represent the Hamiltonian operator H as H = :2 (Q01 + 1 Q . Q . (16. Q) + [Po .=  dP dt 2 P .105) is pet) = = Po and Q/Q =Q = = Q(Po . when the Hamiltonian is Hermitian.] If Q is a constant vector and the initial polarization Po is parallel to Q.105). the normalization of X is conserved during the motion. By (16.107) Po cos wQt + 2'l(Po .105) Since  dP2 dt = d(P .Q(Po . 2 2 +Q A wQt X Po sm wQt • Exercise 16.396 Chapter 16 The Spin It is instructive to derive the equation of motion for the vector P = (0'). dE. (O'H _ HO') = _1 (O'Q.107) and Figure 16.3 that P is constant and equal to Q or Q. . [See also Eq. (16. the eigenvalues of Hare (Qo ± Q)12 and (16. (Q X P) h =0 the vector P maintains a constant length. P and (dPldt)2 are constants of the motion.
L' al2." when tiEo = Q= = 0.3).111) where () is the angle between the polarization vector and Q (Figure 16.Q(Po ' Q) Figure 16.+Q2 t = sin2 () sin2 2:1:.L 2 = + Q. and Laloe (1977).109) can be written as Q2 Q 1 2 v'Q. t . H o = (l + Q= . Precession of the spin polarization vector about Q. a)/2.3. Formula (16.[1 . Q = Q= + Q. 5 5For a full discussion of Rabi oscillations. these transitions between "spin up" and "spin down" eigenstates states of the unperturbed Hamiltonian. and the initial state is totally depleted whenever wQt = (2n + 1)1T. see CohenTannoudji.Po at time t is '1 [1 2 .L) between two energy levels split by the longitudinal field (Q=) by an amount tiEo = Q=.6 Density Matrix and Spin Polarization 397 Q P o. Chapter IV.111) can be interpreted as describing periodic oscillations induced by the transverse field (Q. The probability of P(t) pointing in the direction .L = Po(Q . the maximum amplitude of the Rabi oscillations is unity. . are caused by the constant perturbation H .Po . P(t)] = Q2 . Diu.110) (16. Generically known as Rabi oscillations. and the precession angle 8 is the angle between Po and Q. WQ 2" t = (Po Q2 Q)2 sm2 2" t (16.3.H o = Q. Po) + Po X (Q X Po) (16. with examples.L It (16.L 2 sin 2ft . The initial polarization vector Po and Q define the plane of the figure.109) If. In the special case of "resonance. The Rabi oscillations have the maximum amplitude sin 2 8.L) to the initial polarization Po.. as indicated in Figure 16. we decompose the constant vector Q into two components parallel or longitudinal (Q=) and perpendicular or transverse' (Q. We emphasize that this analysis is exact and does not rely on any perturbation approximation. P(t)] = 2' (Po X Q) sm2 A WQ X .Po .
If Q rotates uniformly about a fixed axis. calculate the polarization vector P as a function of time.liw) X P dt (16. Viewed from the rotating frame of reference.112) The quantity y is the gyromagnetic ratio. Q is a constant vector. A special case of an oscillating field.108) is caused by the entire constant field Q.101) may conveniently be transformed to a frame of reference that rotates similarly. . we have =+wXP as is well known from the kinematics of rigid bodies. and a field Bj. for which a solution of the equation of motion can easily be obtained. the Hamiltonian can be written as H = fJo . dP dt ap dt ap li . Transitions between the two stationary states can be induced if the spin system is exposed to a timedependent oscillating field that has the same or a similar frequency as the spin precession. and the precession motion of P will be changed. the problem has effectively been reduced to the previous one. Exercise 16. and plot the average probability that the particle has "spin 6This is the quantum analogue of the classical Larmor precession described in Goldstein (1980). Exercise 16. and the vector Q is given by Q = liyB A constant field B o causes a precession of P with angular velocity Wo = .23. hence. Derive the new Hamiltonian and show that it corresponds effectively to precession about the constant vector Q . Section 59. and with respect to the rotating system by aPlat. If we denote the time rate of change of P with respect to the fixed system by dPldt. providing an independent derivation of (16. If a constant magnetic field B o. the system will absorb or deliver energy. Equation (16.113) Since in the rotating frame of reference Q .113) can therefore be solved by transcribing the solution of (16.105) appropriately. the equation of motion (16.22. For example.398 Chapter 16 The Spin The energy level splitting (16. Suppose that w is its angular velocity. is placed in a magnetic field B. is that in which the vector Q rotates uniformly about a fixed axis. rotating with angular velocity w in the xy plane. whose degrees of freedom other than the spin can be neglected. Assume P to point in the z direction at t = O. These general principles are at the basis of all the magnetic resonance techniques that are so widely used in basic and applied science. if a spin 1/2 particle.liw is a constant vector.liw.113). pointing along the z axis.yBO•6 If in addition an oscillating magnetic field with the same (or nearly the same) frequency is applied. Calculate the Rabi oscillations in the rotating frame. B = Y"2 li (J' • B (16.= (Q . act in concert on a spin system (gyromagnetic ratio y). It is advantageous to change over to a frame of reference which is rotating with the same angular velocity.
. and that for a proper mixed state. Ipl < 1. q.88).120) holds: o :5 trace p2 :5 p2 = (trace pf = (16. we conclude that generally Ipi :5 1.(P . as required by (16.) is now a twocomponent spinor.1. proper mixed states representing incident particle beams are the rule rather than the exception. almost every statement and equation involving p can be immediately applied to a mixed state. The only exceptions are propositions that assume that p is idempotent or that the polarization vector satisfies Ip I = 1. i.e. Show that a resonance occurs when ro = . A pure or mixed state is represented by a Hermitian density matrix whose eigenvalues are positive and sum to unity. the linearity of the Schrodinger equationallows us to construct the solution (16. (J' (16. Polarization and Scattering.96) are necessary and sufficient for the state to be pure and representable by a spinor x. The superposition principleand more specifically. The theory of scattering was developed in Chapter 13. the forces that cause a beam of particles to be scattered may be spindependent. which are the subject of the next section.114). In spindependent scattering processes. 7. for instance. An unpolarized spin system has p = (1/2)1 and P = O. Spindependent scattering of a particle occurs. illustrating the general density operator theory of Section 15. and it is then necessary to supplement the theory accordingly. X!ncXinc = 1 Following the procedure of Chapter 13."B o. if the Hamiltonian has the form 2 H = L + 2m V(r) + W(r)L .1) 4 If this identity is used in the inequality (16. we must look for asymptotic solutions of the SchrOdinger equation which have the form (16.. These two special cases describe incident beams that are polarized along the direction of the initial momentum and . since the conditions (16. For any density matrix the inequality (15. However. we have 1 P + .5.114) In terms of the polarization vector.116) representing a spinorbit interaction term in addition to a central force. (This arrangement is a model for all magnetic resonance experiments.) Although so far in this section the density matrix p for spin states was assumed to represent a pure spinor state X. The incident particles with spin onehalf are represented by a wave function of the form eikzX_mc = eikz(C I ) C2 . one that is not a pure state. neglecting the spin entirely.7 Polarization and Scattering 399 down" as a function of wlwo for a value of BIlBo = 0.115) from the two particular solutions that correspond to Xinc = a and Xinc = {3.115) but the scattering amplitude f(8.95) or (16. P .
118) by Cz. although its dependence on the scattering angle 0 can be worked out only by a detailed calculation. respectively (Exercise 16.117) and (16.s constants of the motion that the symmetries of the problem generate.119) (16. The polarization is said to be longitudinal. We are thus led to look for two solutions of the asymptotic form (16. depending on the asymptotic form of 1jI. Since the operator J z leaves the radial dependence of the scattered wave unchanged. The state AIjI may represent the same scattering state as 1jI.117) by Cb and (16. such as a phase shift analysis. Show that the incident waves of Jz.116) shows an example of a spindependent Hamiltonian with rotational symmetry. according to Section 16. H is invariant under rotations and. owing to spherical symmetry of the scattering potential.4.24. and adding the two equations. then if IjI is an eigenfunction of H. AIjI is also an eigenfunction of H. If A commutes with the Hamiltonian. we obtain by superposition the more general solution eikz(~J + e~r G:~ ~::)(~J = (e ikz + e: S)XinC Here S stands for 2 X 2 scattering matrix (16. The basic idea is to utilize the obviou. Exercise 16. the solutions (16. commutes with the components of J. Let us assume that.118) must both be eigenfunctions of J z By requiring that and . The form of S can be largely predicted by invariance arguments. The scattering problem is solved if S can be determined as a function of these variables.117) and (16.400 Chapter 16 The Spin opposite to it. Expression (16.120) S depends on the angles 0 and cp. Here we will only deduce the general form of the scattering matrix.118) The quantities in parentheses are the appropriate scattering amplitudes. and on the momentum k.118) are eigenstates of J z with eigenvalues +ftl2 and "'/2. or a different one of the same energy. and both belong to the same energy.24).117) (16. What are the eigenvalues? eikza and eikz(3 are eigenstates Multiplying (16. The incident waves in (16.
In terms of spherical polar coordinates.122) sin cp. see Section 17. cos cp. a set of coupled radial Schrodinger equations must ultimately be solved.icp X function of 0. This is true for the spinorbit interaction in (16. because both Land S or (J' are axial vector operators.117) must go over into (16.5. 0) is normal to the plane of scattering and points The unit vector in the direction of kine X k seatt• We conclude that the scattering matrix has the form fie IS = g(O)l + ih(O)fi .123) The functions g(O) and h(O) are generalizations for the spin 1/2 case of the scattering amplitude f(O) in Chapter 13. For rotationally invariant potentials. but if they are to be computed from the assumed interaction.121) Furthermore.119) is the asymptotic form of the wave function.121) that Sl1 = S22 = g(O). (J' I (16. and their inner product is a scalar operator. Knowing S.118). as behooves an axial (or pseudo) vector. the differential scattering cross section is found to be (16. If (16. while (ux ) remains unchanged. If the density matrix Pine describes the state of spin polarization of the incident beam. and U x has the effect of changing a spin state in such a way that (uy ) and (uz ) change sign. Hence. (For a more general explanation of reflections and parity for systems with spin. we can calculate the intensity of the beam for a given direction. (16.) Since u x /3 = a the reflection in the yz plane changes the incident wave eikza into e ikz /3 and leaves eikrlr invariant.'P.9. S21 = ei'P X function of 0 (16. we may write S = g(O) ( h(O)ei'P h(O)ei'P) g(8) = g(O)l + ih(O)(uy cos cp  Ux sin cp) (16. Consequently. then by a straightforward generalization of the results of Chapter 13. Exercise 16.25. whether the state .116). It follows from this and (16. The operator for reflection in the yz plane is Pxux .124) which is merely the analogue of 1 f(O) 12 for a particle with spin. P x has the effect of changing 'P into 7T . the Hamiltonian H is assumed to be invariant under a reflection with respect to any coordinate plane. where Px simply changes x into x. The terminology "spinflip" amplitude for h(O) and "nonspinflip" for g(O) is selfexplanatory. Show that the same scattering matrix is obtained by requiring reflection symmetry with respect to the xz plane.7 Polarization and Scattering 401 it is easily seen that Sl1 and S22 can be functions of 0 only and that the offdiagonal elements of the scattering matrix have the form S12 = e. they can be parametrized by a generalization of the phase shift analysis of Section 13.
= do trace (PinestS) (16.125) Since SXine is the state of the particles scattered from an incident spin state Xine into the specified direction.IhI 2 O + i(g*h . fi (16.123) to evaluate the trace. Show that if the incident spin state is a pure transverse polarization state. P = Po.gh*)fi + 21hl 2 O • ft. Po = 0.27. or Po = Pofi.402 Chapter 16 The Spin be pure oras is frequently the casemixed.ihl.127) for the incident density matrix.132) Exercise 16. Exercise 16. k kl cos cpo With . fi + gh*)Po X fi (16.130) If the initial beam has transverse polarization and the scattering plane is chosen perpendicular to Po.p.gh*) Po . Po .Po' k k.gh*)Po fi (16.132) satisfies the condition 1 ~ Ipi ~ O. we may write Po' fi = Ipo . If the y axis is chosen to be along the direction of the transverse component of the polarization.129) If we use (16. + (g*h )P Igl2 + Ihl 2+ i(g*h .26. ft. the density matrix corresponding to the scattered part of the wave function is SPincS t SPincS t (16. we obtain P= p (lgl 2. this expression may be generalized to dD.123) for the scattering matrix and Pine = 2" (1 + Po . it follows from (16.130) that P = (lgl2 + IhI 2 )PO + i(g*h . the scattering amplitudes for the initial polarizations Po = ±fi are g ± ih and the scattering leaves the polarization unchanged.131) If the incident beam is unpolarized. a) 1 (16. Show that the magnitude of the polarization given by (16. the scattered beam is polarized normal to the scattering plane: A • P = Pfi = I Igl 2 + Ihl 2 n g*h .128) P = < ) = trace Pseatta = a trace(SpineSt a) doldO (16.gh*) Igl2 + Ihl 2+ i(g*h .gh* A (16.126) Pseatt = trace(SPinest) = da:Id'A H Using the form (16. we obtain the differential cross section in terms of the polarization Po of the incident beam: ~ = Igl2 + Ihj2 + The polarization of the scattered beam is i(g*h .gh*)Po . Hint: Consider Ig .
Illustrating the general concepts of Section 15.1.p_ In p_ (16.133) where the subscripts + and .138) . A = (du/dfl)+ . defined as a measure of the rightleft asymmetry by .7) in Section 16.P)] The given density matrix may be realized by any ensemble of N pure states with Pi = 1. and measurement.137) The Shannon mixing entropy (15. 0') (16. We assume that the spin state of the system is entirely specified by the density matrix p. Exercise 16. such that p =. in agreement with the empirical statement (16. In particle polarization experiments. 0') (16. Measurements. Probabilities. observation. The von Neumann entropy for this density matrix is.126) of this ensemble is H(f/. 8. and Information.135) which correspond to eigenvalues p+ = (1 + P)/2 and p_ = (1 . formula (16. S(p) = p+ = In p_ I . It is again convenient to represent the density matrix by the realvalued polarization threevector P. the asymmetry A equals the degree of polarization P defined in (16.5. we will be able to keep the discussion as general as possible.(du/dfl)_ (duldfl)+ + (du/dfl) (16. and Information 403 these conventions.8 Measurements. 0') and 1 A p. and how it relates to experiment.P)I2. according to (15. this quantity is referred to as the analyzing power.28. and what observables might be measured to determine an unknown p. Probabilities.128) for the differential cross section shows that the scattered intensity depends on the azimuthal angle as I = a(O) + b(O) cos ({).132). By using the 2 X 2 density matrix formalism to represent an arbitrary mixed spin state.P . Show that if Po = ±ft. In this way.128). i=l (16. The spin formalism is so easy to survey that it lends itself particularly well to a demonstration of how quantum mechanics is to be interpreted.) = 2: Pi Inp. such that polarization vectors Pi and probabilities Pi with 2: i N P = i=l 2: Pi Pi N (16.= "2 (l 1 A .(l 1 2 + P . evaluate the asymmetry parameter A. we ask what kinds of ensembles might represent a known p. we find substantiated our original supposition that the rightleft asymmetry in the scattering of polarized beams of particles is a consequence of the particle spin. Assuming that Po is perpendicular to the scattering plane.refer to the sign of the product po· ft.134) Its eigenstates are represented by the purestate density matrices: p+ = "2 (l + P .136) In 2  "2 [(1 + P) In(1 + P) + (1  P) In(1 .
29. a)(l +n. Pta 1 = 4: trace[(l + P . 0: Po = ~ trace[p(l + 0 1 . of course. which generally requires the measurement of the three Stokes parameters for its determination. 0)] (16. This is similar to the description of the polarization state of a beam of light. . according to (15. 0) 1 (16. As an example consider the 2 X 2 density matrix defined by the polarization vector (x P = I" 3 " ") +y+z (16. Figure 16.131). For example.131). 0) = ~ (1 + p. 0 is. H(W. n) A (16.).142) In terms of the polarization vector (Exercise 16. 0)]  ~ (1 . (15. a)] . 0) and 1 A_ = "2 (l . 0) InU (1 + p. which corresponds to projection operators (PaM) A+ = "2 (l + a . and the von Neumann entropy. 0. The outcome entropy for a measurement of a . we can determine the density matrix element Pu = (1 + P z)/2 = Pz' which represents the relative frequency of finding "spin up" in the z direction.143) If three linearly independent observables a . using ensembles with the same density operator (although. all prepared in the same state. 0. Other choices of the direction 0 provide more information about the magnitudes and phases of the density matrix elements. the only possible results of a measurement of U z (or of any other component of a) are + 1 or 1.P22) nz + pdnx + iny) + pt2(nx  iny)] subject to the normalization condition trace p = PI1 + A P22 =1 1 = "2 (1 + P (16. 010 a . By measuring U z for a very large number of replicas of the system.404 Chapter 16 The Spin Exercise 16.p.p. not the same particle). the matrix elements of P can be determined.144) As expected from Eq. 02' (j . compute and compare the Shannon mixing entropy. 03 are measured.a . H(a .4 shows how the Shannon entropy for the outcome of the measurement of a . this entropy reaches its minimum value.136). 0) In[~ (1 . For this mixed state. when the measured spin points in the direction of the polarization vector: 0 = P. the von Neumann entropy (16. a)] (16:141) = "2 [1 + (Pu . We now consider the measurement of the observable a . S(p). 0 depends on P .140) The probability that the system is found with spin in the direction 0 is the expectation value of the projection operator for the eigenstate of a .19).139) and realized by an ensemble ~ of the N = 3 equiprobable pure states that correspond to the spin pointing in the directions of the Cartesian coordinate vectors.
L+L. and compare this with the outcome entropy for a measurement of U'z..LL..139). state Write down the density matrix that represents the pure spin x=~a+~f3 and compare this with the density matrix for the mixed state about which we only know that the probability of "spin up" is onethird.145) and if ft is a unit vector in the yz plane making an angle () with the z axis and 90° . fi..5 1 Figure 16.L..L...J..L.. Compare the answer with the value of the von Neumann entropy of the state.32.120). and Information H(&· ~ 405 . X = Cl a + C2f3 = ICl IeiY1a + Ic21 e iY2 f3 (16. For a mixed state given by the density matrix P = 75 1 (41 7 .J. If p represents the pure state. Exercise 16....J....7i 7 + 7i) ..4. the probability for the value 1 is given by Po = leI 1 sin 2 2~ + + ICll1e21 sin(1'l .. with fi pointing along anyone of the three coordinate axes.1'2) sin () (16..LL. Outcome entropy H(a .l... For the state specified by the polarization vector (16..JL.. p... 34 check the inequalities (15.. Probabilities.() with the y axis.30.. fi as a function of P ..31. . Exercise 16. H(a .. Exercise 16. fi) for a measurement of a . and calculate the eigenvalues and eigenstates. calculate the Shannon entropy.5 0.'l. and the probability of "spin down" is twothirds. fi to yield the value + 1 is Po = leI 2 2 1 cos ~ + 1e21 2 sin2 ~ I Cl11e21 sin(1'1 .147) Exercise 16..... it 1 0. show that the probability for a ..33.146) Similarly.. Evaluate the von Neumann entropy..L. Calculate the von Neumann entropy for these two states.l.'L.. fi). for the outcome of a measurement of a· ft.8 Measurements.1'2) sin () 1e21 2 cos2 ~ (16.
of those particles that respond with "up" or with "down" to this experimental question.148) to (16. In the measurement. Although it can change the state profoundly. The design of the apparatus ensures that they do not overlap and differ from zero only in the region traversed by the upper or lower beam. When.149) In this expression. r') is correlated with the spin state in which 0' . If in the second experiment U z is remeasured. which are lost if only the separated beams are considered. r') (16. fundamental to a proper understanding of quantum mechanics. In this connection. Pup(r.148) to change into a more complicated correlated. initially the state has the simple product form Pinitial = Pspin ® per. r') and Pdown(r. say u y. a particle reveals a spin "up" or "down" with probabilities equal to P+it and Pit. respectively. there is a bodily separation of the particles that the experimenter subjects to the question. Let us assume again that U z is measured in the first experiment. 0') ® pup(r. The upper component pup(r. r') + Pii ~ (l .fi . is measured. "Is the spin up or down?" The beam splits into two components made up. these spatial density matrices can be approximated in terms of wave packets moving along classical trajectories. the preparation of its state is assumed to introduce no correlations between the spin and space degrees of freedom. Neither beam is split any further. Before the particle interacts with the measuring apparatus. we will find that every particle in the upper beam has spin up. the state of the particles can be represented as PfinaJ = Pit ~ (l + fi . or entangled. fi is 1. we will find that each beam is split into two components of equal intensity. and the down component Pdown(r. the z component of the spin is measured. corresponding to the values + 1 and 1 for u y (Figure 16. and every particle in the lower beam has spin down. This example shows the unavoidable effect which a measurement has on the .5). as depicted in Figure 16.406 Chapter 16 The Spin A molecular beam experiment of the SternGerlach type has traditionally been regarded as the prototype of a measurement.1. confirming merely that conceptually the ideal SternGerlach experiment is an exceptionally simple kind of measurement. state. from (16. r') are spatial density matrices that describe the two separated particle beams. this particular measurement does not alter the statistical distribution of the measured quantity (uz )' nor does the spin state change between measurements. The probabilities for "spin up" and "spin down" in this state are P±ii = "2 trace[pspin(l 1 ± fi . r') is said to be correlated with the spin state in which 0" fi is +1. it is interesting to give some thought to a multiple SternGerlach experiment in which two or more spin measurements are carried out in series. r') the purely spatial part of the density matrix. 0') ® Pdown(r. A careful analysis shows that in the region near the magnet where the two beams are well separated. and thus a different component of the spin.149) to account for the phase relations in the spin density matrix. If in the second measurement the inhomogeneous magnetic field has a different direction.148) where Pspin denotes the spin state and per. Usually. additional terms must be included in (16. If by some ingenious manipulation the two separated beams are recombined. 0')] The interaction with the SternGerlach magnet causes the product state (16. r') (16. Thus. respectively.149).
Dt . '* system upon which the measurement is carried out.pi2) (16. rather than U z . If P (short for PsPin) is the spin state before the measurement. and Information 407 P+D. The spin state of the incident beam is represented by the density matrix p. we find anew a splitting of the beam. two observables A and B are termed compatible if for any state of the system the results of a measurement of A are the same. The probability py in (16.134) the probability of finding + 1 is py = '2 trace[p(l + I i i uy)] = '2 + '2 (Pt2 .components of the spin cannot be measured simultaneously on the same sample. In an ideal arrangement of this kind. If fit ±D2' the second beam splitter regenerates the "spin down" polarization state for direction Dt from particles that entered it entirely with "spin up" along direction fit. and u y . a state cannot simultaneously have a definite value of U z and Uyo If we wish to measure U z and uy for a state P. Clearly. is measured in a first experiment. then according to (16. a. the necessary and sufficient condition for this is that the matrices representing A and B commute: ABBA=O Two observables are compatible if and only if the Hermitian matrices representing them commute. two separate copies of the system must be used.5. Probabilities. for they do not commute. . U z and u y are incompatible. the probability of finding u y to be + 1 is simply Ptt12 + P2212 = (Pu + (22)12 = 1/2. showing that the intervening measurement of u y has undone what the first U z measurement had accomplished. a. According to the arguments presented in Section 10. this can happen only if the eigenstates of A are simultaneously also eigenstates of B. = PD. A and B are compatible if measuring B does not destroy the result of the determination of A. 00)] Figure 16. 00)] ~1 . in accordance with the common rule of compounding conditional probabilities. If in a third successive SternGerlach measurement U z is measured again (Figure 16. we may say that the u y measurement has regenerated an amplitude for U z with value 1 in the branch in which the first measurement of U z had produced a pure + 1 spin state. For each beam segment the spin component of the density matrix is specified. Successive StemGerlach measurements of the spin projections DI .5). D2' a. which the intervening U z measurement must wipe out if the probability interpretation of quantum mechanics is to be consistent. whether or not a measurement of B precedes that of A.150) differs from Pu/2 + P22/2 by an (offdiagonal) interference term.8 Measurements.4. Each box symbolizes a beam splitting. producing pure "spin up" and "spin down" states. In other words. = ~ trace ~ trace [p(l+ [p(l ~1 . For example. if we precede this u y measurement by a measurement of u z . The two . In the language of particle physics.150) whereas.
The dynamics is governed by the Hamiltonian matrix H=Mi2 r . A repetition of the measurement of the same quantity in the new state will now yield with certainty this very eigenvalue. inspired guesswork. Rather. or other inferential procedures. in this view there is no reason to draw a line and make a qualitative distinction between a probabilityfree "objective" physical reality and the "subjective" realm of the observer who uses probabilities for interpreting the data. but for an understanding of the physical significance of quantum states it is sufficient to consider the simplest kind. the eigenstates of U'z. we regard the acquisition of information. The acquisition of information provided by the measurement and the subsequent replacement of the original correlated state by selection of one or the other of its components with definite probabilities is conventionally referred to as the reduction of the state. While the idealized SternGerlach experiment illustrates many salient issues in quantum mechanics. acquired through experimental evidence. In particular. namely. A correlated or entangled state like (16. the Stern. representing a particle KO and its antiparticle 'K0: The operator U'x = CP represents the combined parity (P) and charge conjugation (C).ft· 0')/2. respectively. Only with this proviso is it defensible to say that after the reduction has taken place in an ideal measurement. transformation and takes a = IKO) into f3 = I'( 0). Problems 1. the reduction of the statealso known more dramatically as the collapse of the wave packetis not meant to describe a physical process that affects the (probability) amplitudes by actual measurement manipulations. The spinzero neutral kaon is a system with two basis states. is an indispensable part of the specification of a quantum system. is sometimes loosely interpreted by saying that the act of measurement "puts" the system into an eigenstate. Most measurements are more difficult to analyze. In the spirit of the statistical interpretation of quantum mechanics. as an integral part of a full description of the physical system and its evolution. In the persistent debate about the foundations of quantum mechanics and the quantum theory of measurement. in which the various eigenstate projections of the dynamical variable that is being measured are prevented from interfering after the measurement. If we consider the spin properties of the particles in isolation. the system has a definite value of the observable. . the great variety of actual experiments defies any effort to classify all measurements systematically. and its evaluation in terms of probabilities.149) is an example of an ideal measurement (sometimes called a measurement of the first kind) because the spatial separation of the two spin components allows the unambiguous identification of the two spin states and the ("impulsive") measuring interaction leaves the two spatially separated spin states intact. we take the position that'the assignment of probabilities to the outcomes of various possible tests. Gerlach device may be regarded as a spin filter that allocates fractions P+ii and Pii of the particles definitely to the pure spin states represented by the density matrices (l + ft· 0')/2 and (l . or particleantiparticle.408 Chapter 16 The Spin A measurement of the simple kind described by the initial state (16.149).148) and the final correlated state (16. the eigenvalue determined by the measurement.
8 Measurements, Probabilities, and Information
409
where M and f are Hermitian 2 X 2 matrices, representing the massenergy and decay properties of the system, respectively.7 The matrix f is positive definite. A fundamental symmetry (under the combined CP and time reversal transformations) requires that u"M* = Mu" and u"f* = fu". (a) Show that in the expansion of H in terms of the Pauli matrices, the matrix U z is absent. Derive the eigenvalues and eigenstates of H in terms of the matrix elements of M and f. Are the eigenstates orthogonal? (b) Assuming, as is the case to good approximation, that the Hamiltonian also satisfies the CP invariance conditions u"M = Mu" and u"f ,;" fu" show that H is normal, and construct its eigenstates, IK~) and IK~). If the measured lifetimes for these two decaying states are Tl = Mf 1 = 0.9 X 10 10 sec and Tz = Mf z = 0.5 X' 10 7 sec, respectively, and if their mass difference is mz  ml = 3.5 X 10 6 eVlcz, determine the numerical values of the matrix elements of M and f as far as possible. (c) If the kaon is produced in the state K O at t = 0, calculate the probability of finding it still to be a KO at a later time t. What is the probability that it will be found in the KO state? Plot these probabilities, exhibiting particleantiparticle oscillations, as a function of time.
7S ee Perkins (1982) for experimental information on neutral kaons.
CHAPTER
17
Rotations and Other Symmetry Operations.
Although symmetry arguments have already been used in almost every chapter, here we begin a systematic examination of the fundamental symmetries in quantum mechanics. The concepts are best understood by thinking about a concrete example. Rotations exhibit the interesting properties of many symmetry operations, and yet their theory is simple enough to keep the general features from being obscured by too much . technical detail. If the theory of rotations is to be transferrable to other symmetries, it must be seen in the more abstract context of symmetry groups and their matrix representations. Much of the chapter is devoted to the practical problems of adding angular momenta and the extraction of symmetryrelated properties of matrix elements of physical observables. In the last two sections, we deal with discrete symmetries (space reflection and time reversal) and their physical implications, and we return briefly to local gauge symmetries, which are distinctly different from global geometric symmetries.
1. The Euclideq,n Principle of Relativity and State Vector Transformations. The fundamental assumption underlying all applications of quantum mechanics is that ordinary space is subject to the laws of Euclidean geometry and that it is physically homogeneous and isotropic. By this we mean that we can move our entire physical apparatus from one place to another and we can change its orientation in space without affecting the outcome of any experiment. We say that there is no preferred position or orientation in space. The assumption that space is homogeneous and isotropic will be called the Euclidean principle of relativity because it denies that spatial location and orientation have any absolute significance. Gravity seems at first sight to introduce inevitably a preferred direction, the vertical, into any experiment performed on or near the surface of the earth, but in quantum physics we are concerned primarily with atomic, nuclear, and particle processes in which gravitational effects playa negligible role. The apparent anisotropy of space can then usually be ignored, and the isotropy of space for such quantum processes can be tested directly by rotating the system at any desired angle. If gravitation cannot be neglected, as in some extremely sensitive neutron interferometry measurements, 1 there is again no conflict with the Euclidean principle of relativity, because we can imagine the earth to be part of the mechanical system and take its gravitational field into account when a rotation is performed. No violation of the Euclidean principle of relativity has ever been found in any laboratory experiment. On a grander, astronomical and cosmological scale there are legitimate serious questions about the validity of the principle. Understanding the physics of the very early universe may require a fully developed theory that unites gravity with quantum mechanics. The scale at which quantum gravity is expected to be influential, called
'Werner (1994).
410
The Euclidean Principle of Relativity and State Vector Transformations
411
the Planck scale, is characterized, on purely dimensional grounds, by the Planck mass, 1028 eV
=
1016 TeV
The corresponding Planck length is of the order of 10 35 m, and the Planck time is 1043 sec. These ,.estimates make it clear why we will not be concerned with gravity. Here we focus on the remarkable consequences that the Euclidean principle of relativity and its extension to the time dimension have for the structure of quantum mechanics. We will find that this principle severely restricts the possible forms that the quantum description of a given system can take. A transformation that leaves the mutual relations of the physically relevant aspects of a syste~ unaltered is said to be a symmetry operation. The Euclidean principle of relativity amounts to the assumption that geometric translations and rotations are symmetry operations. We first concentrate on rotations about an axis or a point, but in Section 17.9 we will extend the discussion to reflections. Nonrelativistic Galilean symmetry, involving transformations that include motions in time, was discussed in Section 4.7. The symmetry operations associated with the Einstein principle of relativity are based on Lorentz or Poincare transformations and will be taken up in Chapters 23 and 24. When a quantum system with a state vector 'I' is rotated in space to a new orientation, the state vector changes to '1". The Euclidean principle of relativity requires that under rotation all probabilities be invariant, Le., all inner products of two rotated states remain invariant in absolute value. We thus have a mapping of the vector space onto itself, 'I' t'.> '1", such that 1('1", $') 12 = 1('1', $) 12 for every pair of state vectors. Such a mapping is called an isometry. The mapping must be reversible, because we could equally well have started from the new orientation and rotated the system back to its old orientation. In the language of Section 16.3, we are considering active rotations. Generally, we do not require invariance of inner products, which is the hallmark of unitary transformations, but only that the absolute values be invariant. Yet because of a remarkable theorem, we will ultimately be able to confine our attention essentially to unitary and antiunitary transformations. The reasoning given here applies to any symmetry operation and not just to rotations.
Theorem. If a mapping 'I'
t'.>
'1" of the vector space onto itself is given such that
(17.1)
then a second isometric mapping '1" vector,
t'.>
'1''', which is merely a phase change of every
(17.2) can be found such that
is mapped into
412
Chapter 17 Rotations and Other Symmetry Operations
For the proof of this theorem the reader is referred to the literature. 2 The theorem shows that through rephasing of all vectors we can achieve a mapping that has one of the two fundamental properties of a linear operator: The transform of the sum of two vectors is equal to the sum of the transforms of the two vectors [see (9.43)]. It follows from this result and from (17.1) that
I('I'~, 'I'~ + 'I';:W
Hence, by applying (17.1) again,
('I'~,
=
I('I'a' 'I'a + 'I'bW
+
('I'a. 'l'b)*
'1';:)
+
('I'~,
'1';:)* = ('I'a, 'l'b)
or
Re('I'~,
'1';:) = Re('I'a, 'l'b)
Since the absolute value of the inner product ('I' a' 'I'b) is invariant, we must have
Im('I'~,
'1';:) = ±Im('I'a, 'l'b)
The
+ sign implies that
(17.3)
and (A'I')" = A'I''' whereas the  sign implies that
('I'~,
(17.4)
'1';:)
= ('I'a' 'I'b)*
(17.5)
and (A'1')" = A*'1''' (17.6)
Equation (17.4) expresses the second fundamental property of a linear operator [see (9.44)], and from condition (17.3) we infer that in the first case the transformation is unitary. Equation (17.6), on the other hand, characterizes an antilinear operator [see Eq. (9.46)]. It is easy to see the profound implications of this theorem. State vectors that differ by phase factors represent the same state, and a rephasing transformation has no physical significance. It follows that in studying the symmetry operations of a physical system we may confine ourselves to two simple transformationsthose that are linear and those that are antilinear. Any more general symmetry transformation can be supplemented by a phase change and made to fall into one of these two fundamental categories, which are mutually exclusive. Note that the rephasing operation is generally not unitary because different state vectors are generally multiplied by different phase factors. If the symmetry operation is a rotation, the antilinear case is excluded as a possibility because rotations can be generated continuously from the identity operation, which is inconsistent with complex conjugation of a multiplier. Antilinear transformations are important in describing the behavior of a system under time reversal, a topic to which we will return in Section 17.9.
2S ee Wigner (1959), Appendix to Chapter 20, p. 233; see also Bargmann (1964).
2
The Rotation Operator, Angular Momentum, and Conservation Laws
413
2. The Rotation Operator, Angular Momentum, and Conservation Laws. The result of the last section is that, if the Euclidean principle of relativity holds, rotations in quantum mechanics are represented by unitary transformations, validating the assumption made in Section 16.3. Although the discussion in Section 16.4 was phrased in terms of spinors describing the state of a spin onehalf system, the formalism of rotation operators (or matrices) was in no way dependent on the special nature of the system. The unitary operator that in threedimensional space rotates a state 1'It) into l'It l ) = uRI 'It) has the form
UR = exp (
k
it . Jet»
(17.7)
and the Hermitian generators of rotation, J, must satisfy the commutation relations (16.44): (17.8) Since the trace of J j vanishes, the operators (17.7) are unimodular. We know from Chapters 11 and 16 that orbital angular momentum operators L = r X p and spin angular momentum operators S satisfy the commutation relations (17.8). They are realizations of the generic angular momentum operator J. Planck's constant ft was introduced into the definition of the rotation operator in anticipation of the identification of the operators J as angular momentum for the system on whose states these operators act. In Section 11.2, we determined the eigenvalues and eigenvectors for all the Hermitian operators J that satisfy the commutation relations (17.8), as well as the matrices that represent the generalized angular momentum. We now make use of the results obtained there. The eigenvalues of any component of J, such as Jz ' are mft, and the eigenvalues of J2 are j(j + l)ft2. The quantum number j takes on the values j = nonnegative integer or halfinteger, and m is correspondingly integer or halfinteger subject to the restriction j ~ m ~ j. Suppressing any other relevant information that characterizes the state, we denote the eigenvectors by Ijm). Since all nonnegative integers are expressible as 2j + 1, the angular momentum algebra can be realized in a vector subspace of any number of dimensions. In constructing the rotation operator explicitly, we must take into account a further condition that arises because the same rotation R is represented by all operators of the form
UR
= exp [
k
fi .
J(et> + 21Tk)] = exp (
k
fi .
Jet» exp (
k
fi . J21Tk)
where k is an arbitrary integer. The factor exp( 21Tkifi . J/ft) is a unimodular operator whose effect on an eigenstate Ijm) of fi . J is simply to multiply it by exp (im21Tk) = (_1)2km. This is +1 for integer km and 1 for halfinteger km. If a physical state were represented by a superposition of angular momentum eigenvectors with both integral and halfintegral values of j, then since the components with integer j (and m) would remain unchanged while the components with halfintegral j (and m) would change sign, application of the rotation operator exp( 21Tkifi . J/ft) with k = odd would produce an entirely different state vector. Yet, for systems of point particles such a rotation is geometrically and physically equivalent to no rotation at all and behaves like the identity. In other words, the
414
Chapter 17 Rotations and Other Symmetry Operations
mathematical framework allows for state vectors that have no counterpart in physical reality. In ordinary quantum mechanics these states are declared inadmissible by the imposition of a principle. This superselection rule has dramatic physical consequences: for instance, particles of halfintegral spin cannot be created or destroyed singly, or more generally, in odd numbers (because of the way angular momenta add; see Section 17.5 as well as Section 24.5). The general theory of angular momentum presents us with all the possible ways in which state vectors may transform under rotation in three dimensions. It does not, of course tell us which of these possibilities are realized in nature. We have already encountered two actual and important examples: the orbital angular momentum, J = L, and the spin angular momentum, J = S = nu/2, of electrons, protons, neutrons, quarks, and so on. Both of these vector operators satisfy the commutation relations for angular momentum. They correspond to the values j = I = 0, 1, 2, ... and j = s = 1/2, respectively. Generalizing these notions, we now identify as angular momentum any observable that is represented by a generator J (in units of n) of infinitesimal rotations. In order to apply the theory, we must know something about the nature of the particular physical system under consideration. We must know the observables that describe it and how they behave under rotation. Thus, in the case of orbital angular momentum (Chapter 11), we were dealing with the transformation of a function ljJ of the position coordinates x, y, Z, or r, cp, 0, and we were led to the study of spherical harmonics. In the case of the spin (Chapter 16), we deduced the behavior of twocomponent spinors under rotation from the physical connection between the intrinsic angular momentum and magnetic moment, and from the vectorial character of these quantities. Other, more complex examples of angular moment!Jlll will appear shortly. It should be stressed that threedimensional Euclidean space, with its own topology, underlies the theory with which we are concerned here. Quantum systems of lower or higher dimensionality may require qualitatively different treatments. For example, a system that is confined to two dimensions may have structural characteristics that allow a physical distinction between 27Tk rotationsthe integer winding number k being an appropriate index. Even in three dimensions, we can conceive of idealized physical systems other than point particles (e.g., perfect solid or rigid bodies) for which it is meaningful to distinguish between odd and even multiples of 27T rotations. 3 We will return to this point in Section 17.4. A symmetry principle like the Euclidean principle of relativity not only circumscribes the geometric structure of quantum mechanics, but also has important dynamical consequences, notably certain conservation laws, by which the theory can be tested experimentally. Although the same ideas are applicable to almost any symmetry property, to be explicit we will frame the discussion in terms of conservation of angular momentum which is a result of rotational invariance. We assume that the dynamical system under consideration is characterized by a timeindependent Hamiltonian, and in the SchrOdinger picture evolves in time from its initial state I'IJf(0) according to the equation of motion,
d in dt I'IJf(t) = H(a) I'IJf(t)
(17.9)
The Hamiltonian depends on the dynamical variables that describe the system, and by adding the parameter a we have explicitly allowed for the possibility that the
3Merzbacher (1962).
2
The Rotation Operator, Angular Momentum, and Conservation Laws
415
system may be acted upon by external forces, constraints, or other agents that are not part of the system itself. The division of the world into "the system" and "the environment" in which the system finds itself is certainly arbitrary. But if the cut between the system and its surroundings is made suitably, the system may often be described to a highly accurate approximation by neglecting its back action on the "rest of the world." It is in this spirit that the parameter a symbolizes the external fields acting on what, by an arbitrary but appropriate choice, we have delineated as the dynamical sylft:em under consideration. We have seen that a rotation induces a unitary transformation UR of the state vector of the system. If we insist that the external fields and constraints also participate in the same rotation, a new Hamiltonian H(aR) is generated. The Euclidean principle of relativity in conjunction with the principle of causality asserts that, if the dynamical system and the external fields acting on it are rotated together, the two arrangements obtained from each other by rotation must be equivalent at all times, and URI"I"(t) must be a solution of
. d fA dt URI"I"(t) = H(aR)URI"I"(t)
with (17.9) yields the important connection
(17.10)
If the symmetry transformation itself is timeindependent, comparison of (17.10)
(17.11) If the symmetry transformation is timedependent, compatibility of Eqs. (17.9) and (17.10) requires (17.12)
It frequently happens that the effect of the external parameters on the system is invariant under rotation. In mathematical terms, we then have the equality
(17.13) Hence, if the symmetry operator is timeindependent, UR commutes with H. Since, according'to (17.7), URis a function of the Hermitian operator J, the latter becomes a constant of the motion, as defined in Chapter 14. Indeed, the present discussion parallels that of Section 14.4, where the connection between invariance properties and conservation laws was discussed in general terms. Conservation of angular momentum is thus seen to be a direct consequence of invariance under all rotations. As an important special case, the condition (17.13) obviously applies to an isolated system, which does not depend on any external parameters. We thus see that the isotropy of space, as expressed by the Euclidean principle of relativity, requires that the total angular momentum J of an isolated system be a constant of the motion. Frequently, certain parts and variables of a system can be subjected separately and independently to a rotation. For example, the spin of a particle can be rotated independently of its position coordinates. In the formalism, this independence appears as the mutual commutivity of the operators Sand L which describe rotations of the spin and position coordinates, respectively. If the Hamiltonian is such that no correlations are introduced between these two kinds of variables as the system evolves, then they may be regarded as dynamical variables of two separate subsys
416
Chapter 17 Rotations and Other Symmetry Operations
tems. In this case, invariance under rotation implies that both Sand L commute separately with the Hamiltonian and that each is a constant of the motion. The nonrelativistic Hamiltonian of a particle with spin moving in a centralforce field couples L with S, and as we saw in Section 16.4, includes a spinorbit interaction term proportional to L . S:
H =
~+
2m
2
V(r)
+
W(r)L . S
If the L . S term, which correlates spin and orbital motion, can be neglected in a first approximation, the zeroorder Hamiltonian commutes with both Land S, and both of these are thus approximate constants of the motion. However, only the total angular momentum, J = L + S, is rigorously conserved by the full Hamiltonian. We will see in Chapter 24 that in the relativistic theory of the electron even the free particle Hamiltonian does not commute with L or S. Exercise 17.1. Discuss the rotational symmetry properties of a twoparticle system, with its Hamiltonian,
H
=
~+
2m}
2
2
2m2
P2
+
V(lrl  r 2 1)
(See Section 15.4.) Recall the expression (15.95) for the total angular momentum in terms of relative coordinates and the coordinates of the center of mass, and show that if the reduced mass is used, the standard treatment of the centralforce problem in Chapters 11 and 12 properly accounts for the exchange of angular momentum between the two particles. Exercise 17.2. How much rotational symmetry does a system possess, which contains a spinless charged particle moving in a central field and a uniform static magnetic field? What observable is conserved? 3. Symmetry Groups and Group Representations. Because of the paramount importance of rotational symmetry, the preceding sections of this chapter were devoted to a study of rotations in quantum mechanics. However, rotations are but one type of many symmetry operations that playa role in physics. It is worthwhile to introduce the general notion of a group in this section, because symmetry operations are usually elements of certain groups, and group theory classifies and analyzes systematically a multitude of different symmetries that appear in nature. A group is a set of distinct elements a, b, c, ... , subject to the following four postulates: 1. To each ordered pair of elements a, b, of the group belongs a product ba (usually not equal to ab), which is also an element of the group. We say that the law of group multiplication or the multiplication table of the group is given. The product of two symmetry operations, ba, is the symmetry operation that is equivalent to the successive application of a and b, performed in that order. 2. (ab)c = a(bc), i.e., the associative law holds. Since symmetry operations are usually motions or substitutions, this postulate is automatically satisfied. 3. The group contains an identity element e, with the property ea
= ae = a
for every a
3 Symmetry Groups and Group Representations
417
4. Each element has an inverse, denoted by aI, which is also an element of the group and has the property
All symmetry operations are reversible and thus have inverses. For example, rotations form a group in which the product ba of two elements b and a is define~ as the single rotation that is equivalent to the two successive rotations a and b. By a rotation we mean the mapping of a phy~ical system, or of a Cartesian coordinate frame, into a new physical system or coordinate frame obtainable from the old system by actually rotating it. The term rotation is, however, not to be understood as the physical motion that takes the system from one orientation . to another. The intervening orientations that a system assumes during the motion are ignored, and tW9 rotations are identified as equal if they lead from the same initial configuration· to the same final configuration regardless of the way in which the operation is carried out. In the rotation group, generally, ab ba. For instance, two successive rotations by 'TT12one about the x axis and the other about the y axisdo not lead to the same overall rotation when performed in reverse order. The operation "no rotation" is the identity element for the rotation group.
*
Exercise 17.3. Use Eq. (16.62) to calculate the direction of the axis of rotation and the rotation angle for the compound rotation obtained by two successive rotations by 'TT12 about the x and y axes, respectively. Show that the result is different depending on the order in which the two rotations are performed. Convince yourself of the correctness of this result by rotating this book successively by 90° about two perpendicular axes. Exercise 17.4. Show that the three Pauli spin matrices, (Tn (Ty, (Tz, supplemented by the identity 1 do not constitute a group under matrix multiplication, but that if these matrices are multiplied by ± I and ± i a set of 16 matrices is obtained which meets the conditions for a group. Construct the group multiplication table.
A symmetry operation a transforms the state I"1') of a system into a state 1"1' a).
It was shown in Section 17.1 that under quite general conditions this transformation
may be assumed to be either unitary, l"I'a) = Ua I"1'), or antilinear. We assume here that the symmetry operations of interest belong to a group called a symmetry group of the system, which induces unitary linear transformations on the state vectors such that, if a and b are two elements of the group,
(17.14)
When (17.14) is translated into a matrix equation by introducing a complete set of basis vectors in the vector space of I"1'), each element a of the group becomes associated with a matrix D(a) such that
I D(ab)
= D(a)D(b)
I
(17.15)
That is, the matrices have the same multiplication table as the group elements to which they correspond. The set of matrices D(a) is said to constitute a (matrix) representation of the group. Thus far in this book the term representation has been used mainly to describe a basis in an abstract vector space. In this chapter, the same
418
Chapter 17 Rotations and Other Symmetry Operations
term will be used for the more specific group representation. The context usually establishes the intended meaning, and misunderstandings are unlikely to occur. A change of basis changes the matrices of a representation according to the relation
D(a) = SlD(a)S
(17.16)
as discussed in Section 9.5. From a group theoretical point of view, two representations that can be transformed into each other by a similarity transformation S are not really different, because the matrices D(a) obey the same multiplication rule (17.15) as the matrices D(a). The two representations are called equivalent, and a transformation (17.16) is known as an equivalence transformation. Two representations are inequivalent if there is no transformation matrix S that takes one into the other. Since the operators Ua were assumed to be unitary, the representation matrices are also unitary if the basis is orthonormal. In the following, all representations D(a) and all transformations S will be assumed in unitary form. By a judicious choice of the orthonormal basis, we can usually reduce a given group representation to block structure, such that all the matrices D(a) of the representation simultaneously break up into direct sums of smaller matrices arrayed along the diagonal:
D(a)
=
(17.17)
We suppose that each matrix of the representation acquires the same kind of block structure. If n is the dimension of D, each block D 1 , D z, ... is a matrix of dimension nb nz, ... , with nl + nz + ... = n. It is then obvious that the matrices D 1 by themselves constitute an nldimensional representation. Similarly, D z gives an nzdimensional representation, and D i is an njdimensional representation. The original representation has thus been reduced to a number of simpler representations. The state vector (Hilbert) space has been similarly decomposed into a set of subspaces such that the unitary operators U a reduce to a direct sum
Ua
=
U~l)
81
U~z)
81 ...
with each set of operators U~l) satisfying the group property (17.14). If no basis can be found to reduce all D matrices of the representation simultaneously to block structure, the representation is said to be irreducible. Otherwise it is called reducible. Apart from an unspecified equivalence transformation, the decomposition into irreducible representations is unique. (If two essentially different decompositions into irreducible subspaces were possible, the subspace formed by the intersection of two irreducible subspaces would itself have to be irreducible, contrary to the assumption.) There is therefore a definite sense in stating the irreducible representations (or irreps in technical jargon), which make up a given reducible representation. Some of the irreducible matrix representations may occur more than once. It is clearly sufficient to study all inequivalent irreducible representations of a group; all reducible representations are then built up from these. Group theory provides the rules for constructing systematically all irreducible representations from the group multiplication table. Which of these are relevant in the analysis of a par
3
Symmetry Groups and Group Representations
419
ticular physical system depends on the structure of the state vector space of the system. The usefulness of the theory of group representations for quantum mechanics and notably the idea of irreducibility will come into sharper focus if the SchrOdinger equation HI '1') = E 1'1') is considered. A symmetry operation, applied to all eigenstates, must leave the Schrodinger equation invariant so that the energies and transition amplitudes of the system are unaltered. The criterion for the invariance of the Schrodinger equation under the operations of the group is that the Hamiltonian commute with Ua for every element a of the group: (17.18) . In Section 17.2 the same condition was obtained by applying symmetry requirements to the dynamical eq~ations, and the connection between conservation laws and constants of the motion was established. By studying the symmetry group, which gives rise to these constants of the motion, we can learn much about the eigenvalue spectrum of the Hamiltonian and the corresponding eigenfunctions. If E is an nfold degenerate eigenvalue of the Hamiltonian,
Hlk) = Elk)
(k
= 1, 2, ... , n)
(17.19)
the degenerate eigenvectors Ik) span a subspace and, owing to (17.18),
HUalk)
= UaHlk) = EUalk)
Thus, if Ik) is an eigenvector of H corresponding to the eigenvalue E, then Ualk) is also an eigenvector and belongs to the same eigenvalue. Hence, it must be equal to a linear combination of the degenerate eigenvectors,
UaI k) =
2: Ij)Djk(a) j=1
n n
n
(17.20)
where the Djk(a) are complex coefficients that depend on the group element. Repeated application of symmetry operations gives
UbUalk)
= 2:
n
j=1
Ublj)Djk(a)
= 2: 2:
j=1 e=1
If)Deib)Djk(a)
(17.21)
But we also have
Ubalk)
=
2: !f)Dek(ba) e=1
n
n
(17.22)
By the assumption of (17.14), the lefthand sides of (17.21) and (17.22) are identical. Hence, comparing the righthand sides, it follows that
Dek(ba) =
j=l
2: Deib)Djk(a)
(17.23)
This is the central equation ofthe theory. It shows that the coefficients Djk(a) define a unitary representation of the symmetry group. If a vector lies entirely in the ndimensional subspace spanned by the n degenerate eigenvectors of H, the operations of the group transform this vector into another vector lying entirely in the same subspace, Le., the symmetry operations leave the subspace invariant.
420
Chapter 17 Rotations and Other Symmetry Operations
Since any representation D of the symmetry group can be characterized by the irreducible representations that it contains, it is possible to classify the stationary states of a system by the irreducible representations to which the eigenvectors of H belong. A partial determination of these eigenvectors can thereby be achieved. The . labels of the irreducible representations to which an energy eigenvalue belongs are the quantum numbers of the stationary state. These considerations exhibit the mutual relationship between group theory and quantum mechanics: The eigenstates of (17.19) generate representations of the symmetry group of the system described by H. Conversely, knowledge of the appropriate symmetry groups and their irreducible representations can aid considerably in solving the SchrOdinger equation for a complex system. If all symmetries of a system are recognized, much can be inferred about the general character of the eigenvalue spectrum and the nature of the eigenstates. The use of group theoretical methods, applied openly or covertly, is indispensable in the study of the structure and the spectra of complex nuclei, atoms, molecules, and solids. The Schrodinger equation for such manybody systems is almost hopelessly complicated, but its complexity can be reduced and a great deal of information inferred from the various symmetry properties, such as translational and rotational symmetry, reflection symmetry, and symmetry under exchange of identical particles. The observation of symmetric patterns and structures, as in crystals and molecules, suggest the use of finite groups, i.e., transformation groups with a finite number of elements. Often details about forces and interactions are unknown, or the theory is otherwise mathematically intractable, as in the case of strongly interacting elementary particles (quantum chromodynamics). However, the dynamical laws are understood to be. subject to certain general symmetry principles, such as invariance under rotations, Lorentz transformations, charge conjugation, interchange of identical particles, "rotation" in isospin space, and, at least approximately, the operation of the group SU(3) in a threedimensional vector space corresponding to intrinsic degrees of freedom. The irreducible representations of the groups which correspond to these symmetries provide us with the basic quantum numbers and selection rules for the system, allowing classification of states, without requiring us to solve the complete dynamical theory. The utility of group representations in quantum mechanics is not restricted to systems whose Hamiltonian exhibits perfect invariance under certain symmetry transformations. Although the symmetry may only be approximate and the degeneracy of the energy eigenstates can be broken to a considerable degree, the states may still form socalled multiplets, which under the action of the group operations transform according to an irreducible representation. Thus, a set of these states can be characterized by the labels, or "good quantum numbers," of the representation with which they are identified or to which they are said to "belong." An understanding of the pertinent symmetry groups for a given system not only offers considerable physical insight, but as we will see in Sections 17.7 and 17.8, can also simplify the technical calculation of important matrix elements in theory. For example, the orbital angular momentum operator L 2 , whose eigenvalues fee + l)fi? will be seen in the next section to label the irreducible representations of the group of rotations of the position coordinates alone, commutes with the Hamiltonian of an electron in an atom that is exposed to an external uniform magnetic field (but not an electric field) and, also with the spinorbit interaction. Therefore, the quantum numbers f characterize multiplets in atomic spectra.
4 The Representations of the Rotation Group
421
The continuous groups that are of particular interest in quantum mechanics are various groups of linear transformations, conveniently expressible in terms of matrices, whichlike the rotations in Section 16.3can be generated by integration from infinitesimal transformations. The elements of such Lie groups are specified by a minimal set of independent, continuously variable real parametersthree, in the case of ordinary rotationsand the corresponding generators. The algebra of the generators, interpreted as (Hermitian) matrices or operators with their characteristic commutatidn relations, is the mathematical tool for obtaining the irreducible representations of these groups. An important category of continuous groups are the socalled semisimple Lie groups, which are of particular physical relevance and also have attractive mathematical properties (analogous to the richness of the theory of analytic functions). Examples of important semisimple Lie groups are the ndimensional groups O(n) of the real orthogonal matrices; their subgroups SO(n) composed of those IlJ.atrices that have a determinant equal to + I (with the letter S standing for special); and the special unitary groups SU(n). In Chapter 16 we saw that the rotation group R(3), which is isomorphic to 0(3), is intimately related to SU(2). This connection will be developed further in Section 17.4.
Exercise 17.5. An ndimensional proper real orthogonal matrix SO(n), i.e., a matrix whose inverse equals its transpose and which has determinant equal to unity, can be expressed as exp(X), where X is a realvalued skewsymmetric matrix. Show that the group of special orthogonal matrices SO(n) has n(n  1)/2 independent real parameters. (Compare Exercise 16.3.) Similarly, show that the group SU(n) has n 2  1 real parameters.
4.
The Representations of the Rotation Group. The representations of the rotation group R(3), which is our prime example, are generated from the rotation operator (17.7),
UR
= exp(i fi· J4»
(17.7)
The rotations in real threedimensional space are characterized by three independent parameters, and correspondingly there are three Hermitian generators, lx, ly, lz, of infinitesimal rotations. They satisfy the standard commutation relations for the components of angular momentum, (17.8). The eigenvectors of one and only one of them, usually chosen to be lz, can serve as basis vectors of the representation, thus diagonalizing lz. In other Lie groups, the maximum number of generators that commute with each other and can be simultaneously diagonalized is usually greater than one. This number is called the rank of the group. The groups 0(4) and SU(4) both have rank two. The central theorem on group representations is Schur's (second) Lemma:
If the matrices D(a) form an irreducible representation of a group and if a matrix M commutes with all D(a),
[M, D(a)] = 0
for every a
(17.24)
then M is a multiple of the identity matrix.
This result encourages us to look for a normal operator C which commutes with all the generators, and thus with every element, of the given symmetry group. The aim
C2• •••• respectively. Since. J2 = J. 0 0 . any particular irreducible representation may appear repeatedly. more than one operator C is needed so that the eigenvalues will provide a complete characterization of all irreducible representations of a symmetry group..26) with eigenvalues j(j + 1)li? The nonnegative integral or halfintegral angular momentum quantum number j fully characterizes the irreducible representations of the rotation group in three dimensions.g. From Section 11.. all matrices representing the group elements are reduced to block structure as in (17. (17. is chosen to be the familiar square magnitude of the angular momentum operator. the matrix elements of D(a) which connect basis vectors that belong to two distinct eigenvalues of C (e. which commutes with every component of J. + J. we choose the common eigenvectors of J2 and J z as our basis vectors and denote them by Iajm). depending on the nature of the vector (Hilbert) space of the physical system under consideration. For the rotation group R(3) of rank one. Exercise 17. In this basis. In general.25) o where the identity matrices have dimensions corresponding to the multiplicities (degeneracies) of the eigenvalues c[. the eigenvalues Ch C2• •••• of the operator C are convenient numbers (quantum numbers) which may serve as labels classifying the irreducible representations.27) The vector space of the system at hand thus decomposes into a number of disjoint (2j + I)dimensional subspaces whose intersection is the null vector and which are .6. (In four dimensions.422 Chapter 17 Rotations and Other Symmetry Operations is to find an operator C whose eigenvalues can be used to characterize and classify the irreducible representations of the group.2 we copy the fundamental equations: Jzlj m) J+ Ij m) J_lj m) = mtilj m) = V~(J=':. Show that if D(a) commutes with C. the Casimir operator. additional labels a are usually needed to identify the basis vectors completely and uniquely. two Casimir operators and their quantum numbers are needed. Since the quantum numbers a are entirely unaffected by rotation. they may be omitted in some of the subsequent formulas. but they will be reintroduced whenever they are needed.m + l)tilj m  1) 1) (17..) As is customary. the rank of the group is equal to the number of mutually commuting independent Casimir operators sufficient to characterize all irreducible representations. + J~ = 'I2 (J+L + LJ+) + J~ (17. For the important class of the semisimple Lie groups.m)(j+m+l)tilj m + = V(j + m)(j ...• Cl "* C2) are zero.17). If the operator has distinct eigenvalues Cb C2••••• it can by a suitable choice of the basis vectors be represented in diagonal form as C l o 1 c21 0 0) 00 0 0 0 C= ( 0 . If the reduction to block structure produces irreducible representations of the group.
k)n (17. we get 4J « Dm'm(BO) 0) A = 5m.. Hence. Exceptional simplification occurs for the subgroup of twodimensional rotations about the "axis of quantization.m .. J: D U"(R) ~ =. From the point of view of infinitesimal rotations. .32) iB x  V' 2 ( ] + m)( J.28) to first order in E = 4J 0. an important distinction arises. ~ eik.31) to construct the rotation matrices in terms of the first 2j powers of the matrix fi .28) Owing to (9.. J (k' . the representation matrix can be written as exp( i fi . For such special rotations (17.29) where now J stands for the matrix whose elements are (jm'IJljm). 'Y rr j k=j k'=j k'*k kin . Jx would necessarily undo the diagonalization of Jz.30) Exercise 17. and applying (17. when 1. Again we must comment on the halfintegral values of j. For small values of j. 312. and the detailed dependence of the matrix elements of DVl(R) on the quantum numbers and on the rotation parameters 0 and 4J is quite complicated. . 112... Le.7. and 1.m+l ~ Um'.. However.IB z mUm'. since diagonalizing.ml . (17. 1/2.28) and (17. we can make use of the formula (10.m + 1) ~ A further diagonalization of all rotation matrices is clearly impossible. the irreducibility of the representations just obtained is verified. 2 ( ] .64).m . An arbitrary rotation is represented in one of the subspaces by the matrix D~1m(R) = (jm'l URljm) = (jm'l exp ( D(j)(R) = i fi . J4J) lim) (17. In this case. when finite rotations are considered.31) shows that the matrix DV)(21T) = 1 is obtained. for upon rotation about a fixed angle by 4J = 21T.4 The Representations of the Rotation Group 423 invariant under rotation. the system has been restored to its original configuration. Using (17. All continuous unitary irreducible representations of the rotation group are obtained by allowing j to assume the values j = 0.  iBx By + By V . these are on the same footing as the integral values.. and the resulting rotation is equivalent to the operation "no .L. work out the rotation matrices for j = 0. The simplicity of (17. Yet.63) and (9.29) is deceptive.32). . by expanding (17. (17.. as is confirmed by (17. 1.31) The representation matrices are also simple for infinitesimal rotations. for the components of J other than Jz are represented by nondiagonal matrices. J4J) (17. for instance.30) and the explicit form of the angular momentum matrices. for a given value of j.0 .m)(] + m + 1) 8m'.27)." the z axis if J z is chosen to be diagonal.
a doublevalued representation of the group is produced: to any rotation R correspond two distinct matrices differing by a sign. With this understanding. The Euler angles a and f3 are the spherical polar coordinates of the z' axis with respect to the xyz coordinate system (see Figure 11. f3 (rotation about the y" axis or nodal line). .424 Chapter 17 Rotations and Other Symmetry Operations rotation at all. Show that the representation matrices D(II2)(R) are equivalent to the matrices U R in Eq. as we saw in Chapter 16. (16. All intermediate axes are shown as dashed lines. Exercise 17. Hence. and 'Y (rotation about the z' axis). transforming xyz into x'y'z'. defined by the azimuthal angle a and the polar angle {3.33) z· z x Figure 17. they have their place in the theory. these D(j) matrices do not constitute a representation of the rotation group.3). provided that UR contains only doublevalued representations. ex p ( k J) k yi' . Euler angles a (rotation about the z axis). and the second rotation by the angle {3 takes place around this nodal line. J) k J) exp ( ai· (17. Instead. y.62). {3. for halfintegral values ofj. As indicated in Figure 17. a is the angle of a rotation about the z axis of the fixed frameofreference.1. This rotation carries the y axis into the y" direction. a. the complete rotation operator is UR =. rather than in terms of fi and <p.8. exp ( {3y" . They appear naturally in the reduction of the matrix representing the rotation operator UR • Owing to the usual phase ambiguities of quantum mechanics. The final rotation is by the angle y about the new z' axis.1. SU(2)." Hence. This rotation carries the z axis into the z' direction. The general expression for the representation matrices D(R) is most conveniently obtained if the rotation is parametrized in terms of the three Euler angles. we can safely regard them as doublevalued representations of the rotation group. they represent the unitary unimodular group in two dimensions. By the strictest definition. This group is said to be the universal covering group of the rotation group.
See also Feagin (1993).39) is related to time reversal symmetry and can be proved at the end of the chapter.40) A rotation transforms an eigenvector Iem) of L z into an eigenvector Iem)' of L z " with the same values of e and m. Prove the symmetry relation (17.L) Ijm) 2ft We leave the details of working out explicit formulas for the representation matrices for general values of j to the myriad of specialized treatises on angular momentum4 and the rotation group. (17.( .34) and (17. show that the inverse rotation R 1 is generated either by first rotating by . for a computeroriented approach. y.a) f3.J = e(ilh)aJze(ilh){3Jye(ilh)YJz R (17. . we can establish the connection between the rotation matrices and the eigenfunctions of orbital angular momentum.y again around Z.41) 4Rose (1957).1. Using the unitary property of UR.m*(a. y) y) = (l)mm'D~)m'. [The relation (17. and finally by .35) Substitution of (1l34) and (17. 112. Thompson (1994). f3.ame.y. = e. it is easy to see that z. . and finally by a around Z.ym(jm' I exp ( f3 1+ .'1 m(a. Work out the rotation matrices in terms of Euler angles for j = 0. then by .  f3. y) = D~~. and 1.35) into (17. D~. Biedenharn and Louck (1981). and compare the results with the results of Exercise 17.. cp) (17.38) (17.9. Chapters 1619. Hence. z' being the new axis obtained by the rotation from the z axis.] Exercise 17. the eigenvectors of lz(L z) and J2(L 2) can be chosen to be the spherical harmonics: (rlem) = Y7(8. but we note two particularly useful symmetry relations for the matrix elements: D~.10. or equivalently first rotating by a around Z. f3.33) yields the simple result U = e(ilh)az·Je (ilh){3Y·Je (ilh)Yz. By inspection of Figure 17.38).7. Expressed in terms of a (single particle) coordinate basis. f3.y around z'. y) = (jm' le(ilh)aJz e(ilh){3Jy e(ilh)YJzljm) = eiam'eiym(jm'le(ilh){3Jyljm) (17.f3 around y. Brink and Satchler (1968). j = e.f3 around y'._m(a. If j is integral.m*(a.37) . (17. lem)' uRlem) = L m' lem')D~!m(R) (17.39) Exercise 17. then by .4 The Representations of the Rotation Group 425 It is preferable to transform the righthand side into a product of rotations about the fixed axes X. .36) The matrix elements of the irreducible representation characterized by j are then D<.
Using (11. We sometimes say that J I and J 2 are coupled to the resultant J.' ((. a) (17. JI<P) @ exp ( i J fi· J2<P) = exp [ i fi .42) for m = 0. (11.43) Consider in particular a point on the new z' axis. <p)D~!m(R) e (17.100): Pe(cos 8') = 21. The mathematical procedure was outlined in Section 15.82).((. <p' are coordinates of the same physical point. we get Y~(8'.e + 47T y m((3 a) 1 e .*((3. The Addition of Angular Momenta. (15.*(R)Y.42) Here 8.98).~.49) is the total angular momentum of the composite system.47) The operator = JI @ 1 + 1 @ J2 (17.45) By (11.*((3. (JI @ 1 + 1 @ J2)<P] (17. <p)Y. Equation (17.44) where (3.4. we finally retrieve the addition theorem for spherical harmonics. <p. A common rotation of the composite system is represented by the direct product of the rotation operators for each subsystem and may. 8' = 0.((.99).42) can be inverted by using the unitary property of the representation: Y. Y.((. q/) = m'=e 2: yt (8. + 1 m~e f)47T ± Y. we find the connection between the D matrices and spherical harmonics: D<e)*(R) mO = J2. <p') (17.(8. . <p) (17. If this is substituted into (17. and 8'. as an application of Eq.48) often written more simply.I'. <p) = 2: m'=f e Dr:. (17. if less accurately. the states of the composite system are vectors in the direct product space of the two previously separate vector spaces.46) 5.I. which are described in two different vector spaces. are merged. be written as exp ( i fi . if this is multiplied on the left by I r). a)Y. as J = JI + J2 (17. If two distinct physical systems or two distinct sets of dynamical variables of one system.1). <p') = ~:uh m~e ~e Y.I'. a are the spherical polar coordinates of the new z' axis in the old coordinate system (see Figure 17.I.426 Chapter 17 Rotations and Other Symmetry Operations and.
. J lz . JJ = ilieijkh (17. we can thus write the transformation equation Ijlhjm) = 2: mlmZ Ijlj2 ml m2)(iIh ml m2IiIhjm) (17.11. Ji] = [J2. From this basis we propose to construct the eigenvectors of Jz. 1 Prove Eq.4 and again touched upon in Section 17.3. the total angular momentum J also satisfies the angular momentum commutation relations: [Ji .50) The problem of the addition of two angular momenta consists of obtaining the eigenvalues of J z and ~2 and their eigenvectors in terms of the direct product of the eigenvectors of J lz . J~] = [J2. 5 Commas will be inserted between quantum numbers in a ClebschGordan coefficient only if they are needed for clarity. J2z . Ji and of J 2z . can be symbolized by the direct product kets. Thus.51) These constitute a basis in the product space. J~. the problem of adding angular momenta is the problem of determining the transformation coefficients (jlj2 ml m2Ijlhjm) These elements of the transformation matrix are called ClebschGordan or Wigner coefficients or vectoraddition or vectorcoupling coefficients. J~. J2 which form a new basis. (17. [Jz. 5Por a complete discussion of Wigner coefficients and the many notations that are in use for them and their variants. When no confusion is likely. Hence. The summation needs to be carried out only over the two eigenvalues ml and m2. Since each component of J I commutes with each component of J2> which separately satisfy the usual commutation relations for angular momentum. As an abbreviation. see Biedenharn and Louck (1981). (But J2 does not commute with J lz or J 2z !) In the subspace of the simultaneous eigenvectors of Ji and J~ with eigenvalues jl and h. where Land S were combined into J.52) connecting the two sets of normalized eigenvectors. The present section is devoted to the formal and general solution of adding any two commuting angular momenta. J~] = 0 and the eigenvectors of J z and J2 can be required to be simultaneously eigenvectors of Ji and J~ as well. The normalized simultaneous eigenvectors of the four operators Ji. we simplify the notation and write I <ilh mlm2Ijlhjm) == (mlm2Ijm) I for it is understood that jl and j2 are the maximum values of ml and m2. The operators Ji and J~ commute with every component of J.47) and show that J I ® 1 commutes with ® J2 • An important example of adding commuting angular momentum operators was already encountered in Section 16. we will refer to the ClebschGordan coefficients as CG coefficients.5 The Addition of Angular Momenta 427 Exercise 17. (17. Ji] = [Jz. for jl and j2 can be assumed to have fixed values. respectively.
we set ml = iI and m = j in (17.1.j .Ilj. m = j . + m2)(h ± m2 + 1)(m. From the selection rule (17.1.1) = Y(j2 .55) Hence. m2ljm) (17.52) and if the eigenvalue con ditions (JZl + Jz2 )ljlh m l m2) JZ Ijlhjm) = mli lithjm) = (ml + m2)liljlhml~) are used...jlljj) (17. if (j . Continuing in this manner. it can immediately be concluded that the m quantum numbers are subject . we obtain the following recursion relations for the CG coefficients: Y(j ± m)(j + m + 1)(ml~lj. jl +h The last two inequalities together show that the three angular momentum quantum numbers must satisfy the socalled triangular condition (17. ~±lljm) Derive the recursion relation (17.iI . but this time using the lower signs and setting ml = j. and we find that v'2j (jl. we could equally well have expressed all CG coefficients in terms of (iIh.h. h liIhj j) and then have concluded that j must be resticted by the condition j2 .52).. j .jl .1) can be determined.j + jl + 1)(h + j . Owing to the selection rule (17. and j all the CG coefficients in terms of just one of them.428 If the operator J z = JZ1 Chapter 17 Rotations and Other Symmetry Operations + JZ2 is applied to (17. (ii.53) Applying Land J + to (17..jl) (j. j . j . j :5. for nontrivial results m2 can have only the value m2 = j . From these two CG coefficients we can then calculate (iI .53).Ilj.56) The absolute value of this coefficient is determined by normalization (see below).jl . j2 this being the range of values of m2' Hence.54) to give for fixed values of j I' h. j .j .54).j . j .54) + Y(h Exercise 17. to the selection rule: (17.12. j is restricted to the range jl . j .54).jllj. j . we can use the recursion relations (17.j . To appreciate the usefulness of these equations.. as indicated.jl :5.it Ij j) is known. m2 = j .56) is different from zero only if j2 :5.j2 :5.1.54). it + h But sinceiI andh are on a symmetric footing. m:+ 1) = Yjl + ml)(jl ± ml + 1)(ml±l. namely.57) . j :5.53) we deduce that the CG coefficient (17. (17. using the upper signs.1) by again applying (17.it :5.
we note the most useful symmetry relations for CG coefficients: (hhmlmzljljzj m) = (I)jhh(hhmzmllj'lid m) = (j2ih ~. Calculate the CG coefficients needed to couple the two angular momentajl = 3/2 andh = 1 to the possiblej values. m) (17.57) they satisfy the equivalent relations Ij . h.h Ijd2i j). Since m = ml + mz ranges between j and +j.60) The double sums in these two equations can be substantially simplified by the use of the selection rule (17. and that in addition to (17. j are integers.52) gives a complete new basis in the (2jl + 1)(2jz + I)dimensional vector space spanned by the kets IhhmlmZ)' Indeed. If this is done.59) 2: (jlhmlmzljljzjm)(jljzm~m~ljljzjm) jm = 8mlmi8mzmi (17. j . j must be such that they could constitute the three sides of a triangle. is chosen real. and express the coupled states IjJzj m) in terms of the uncoupled states Iilj2mlmZ)' Exercise 17. withj given by the allowed values (17. it follows thatj can assume only the values (17. it is clear from the recursion relations for the CG coefficients that all of them are real numbers if one of them.1) (j + 1)(2j + 3) (17. the CG coefficients satisfy the condition 2: or inversely (jljzml~ljd2i m)(jdzmlmzlhhj'm') = 8mm . j . Show that a symmetry exists among the three quantum numbersjhh.14. It is conventional to choose the sign by demanding that the CG coefficient (hh. (17. Since the old and new bases are both normalized to unity.13. Extensive numerical tables and computer codes are available for CG coefficients. Exercise 17.e.58) Hence. the new kets Ihhj m). Verify the values of two special CG coefficients: (jljO Ijlj j) = JJ' +j 1 and (j2jO Ij2 j j) = j(2j . . either all three quantum numbers h.15. and being eigenkets of Hermitian operators they are also orthogonal. are again (2h + 1)(2h + 1) in number. il..5 The Addition ofAngular Momenta 429 i.59) in conjunction with the recursion relations determines all CG coefficients except for a sign.58). mlljzjd.53).8jj. Furthermore.jZISjl Sj + jz and Ij . Condition (17. 6 Without proof.jllhhj j) be real and positive.61) Exercise 17. jz. or two of them are halfintegral and on~ is an integer.j. the three nonnegative integer or halfintegers jl' h.jllSjz Sj + jl We observe that for fixed values of j land h (17. say (j dz. the CG coefficients must constitute a unitary matrix.62) 6S ee Thompson (1994) for references.
m) = (17.63) from the recursion relations for CG coefficients and the normalization and standard phase conditions.63) (jh 1/2.430 Chapter 17 Rotations and Other Symmetry Operations Determine the value of the trivial CG coefficient (jOmOljOjm). How does it depend on the value of m? When we deal with spin onehalf particles. Using the CG coefficients (17. jl ± 1/2. If the coordinate representation is used for the eigenstates of orbital angular momentum L and if.63) with it = . m) since the allowed j values are j = = '1 :+ m 2jl + 1/2 +1 it ± 1/2. m + 1/2.67) . we have a. From the recursion relations (17.59). the direct product spin space of the complete system is fourdimensional. or coupled. the normalization condition (17.y{m. j.66) If the two particles have spin onehalf.: al ® a2 = al a2' al ® ~2 = al~2' f31 ® a2 = f31 a2' f31 ® ~2 = f31f32 (17. to give the total spin (17.e ± 1/2.64) for the eigenstates with j = .jllit.17. Exercise 17.16. the CG coefficients forJ2 = 1/2 are frequently needed. Apply J2 = L 2 + S2 + 2L .65) to Eq. and a particular basis is spanned by the uncoupled eigenvectors of SI. . 1121jl 1/2. (17.63) are also useful when the commuting spins oftwo particles are added. We will denote the common eigenstates of J z and J2 by a. we obtain the values (it.112. 1/2. the total angular momentum J = L + S and its eigenvectors may be represented in the direct product spin ® coordinate basis. S = L2 + S2 + L+S_ + L_S+ + 2LzSz (17.y{m is an eigenstate of J2. Work out the results (17. as usual. the eigenstates of Sz' a = (~) and ~ = (~) are chosen as a basis to represent the spin.e. jl ± 1/2.yim = aye:i:1I2. Exercise 17. and the convention that (jl' 112.1/2Ijl' 1/2. The formulas (17. m . and 8 2 .64) and verify that a. .e + m +i Y~II2) y~+112 (17. jh j .m = e e V2€ + 1 1 (±v. j) shall be real and positive.54). 1/2.e ±m + i V. These coupled eigenstates can be expected to play an important role in the quantum mechanics of singleelectron and singlenucleon systems.
63). the total spin quantum numbers S = 0 and S = 1 characterize two irreducible representations of R(3) and the eigenstates of 8 2 • The state (17. It follows from the last section that in the reduction each of these irreducible representations of the rotation group appears once and only once. The latter is usually reducible even if the original two representations are irreducible. and show by explicit diagonalization that the singlet and triplet states are the eigenvectors with the appropriate eigenvalues.69) Ii i 1. 1) ® /32 = /31/32 If we write the eigenvalues of 8 2 in the usual form n2 S(S + 1). + 1) = = = al ® a2 = ala2 Ii iI. the integral or halfintegral quantum number j. For the rotation group.68) corresponds to S = 0 and is called a singlet state. Eqs. The three states (17. Formally. this fact is expressed by the equivalence D(h) ® DU2) ~ ( DU~+h) DUt~hl) ~ ~) 000 o 0 0 Dlhhl (17. and the reduction consists of determining the invariant subspaces contained in this space.69) correspond to S = 1 and are said to be the members of a triplet. and no Exercise 17. (17. labels the several representations into which the direct product representation decomposes. successively.48) show that the problem of reducing the direct product representation DUt)(R) ® DUi>(R) is intimately related to the problem of adding two commuting angular momenta. Since the irreducible representations are characterized by the eigenvalues of J 2 . the state vector in the (2iI + 1) X (2h + I)dimensional direct product space transforms according to DUt) ® D(h) .j21 and jl + h in integral steps. For the rotation group in threespace. they are.18.67). work out the 4 X 4 matrices Sz and 8 2 . J = J I + J 2 • Under rotations. The product representation can then be reduced.68) and Ii ii. Starting with the uncoupled basis (17. or decomposed. 6. 0) ~ (al ® /32 + /31 ® a2) = ~ (al/32 + /3l a 2) /31 (17. we obtain the appropriate CG coefficients. it is customary to identify the irreducible representations by their dimensionality. which can assume values between Ijl . The direct products of the matrices of two representations of a group again constitute a representation of the same group. eigenstates of Sz with eigenvalues +n. The ClebschGordan Series. allowing us to write the coupled eigenstates of Sz and 8 2 in the form: (17.6 The ClebschGordan Series 431 By letting h = 1/2 in the expressions (17. 0.70) In particle physics.47) and (17. giving us generally new irreducible representations. which has (except for equivalence transformations produced by changing bases) precisely one irreducible representation of .
72) the linear homogeneous relation for the CG coefficients: 2: (iIizm.fl.70) can then be concisely.f + 1) (.fl.. we call on the identity (17.3/2._+:1 ). in which the matrices and D(h) and D(h) represent rotations... in which D(j) represents rotations.19.20.2(O. This section focuses on the rotational transformation properties of the operators of quantum mechanics and generalizes the concepts introduced in Sections .f2. noting also that the CG coefficients are real. The relation (17.fl. Hence.. e. derive from (17.:..(2_.. <p) (17. ~) = 2: e 1.fl.f2.f.) X 1 41T(2.f2mllnzl.f and appropriate choices for the m's: Y~l(O.(2_. ml From this we find the value of the frequently used integra!. As a useful application of this equation..m~ lili2hm3)D~fm2(R) mi = ml m 2: D<j... $ [2UI . or 2. 1/2..fO)(.70) have the same dimension and that (17.... written as (2jl + 1) ® (2j2 + 1) = [2UI + j2) + 1] $ [2UI + j2 .g..~>"'(R)(id2mlm2Iid2i3m)D~fml *(R) (17.70) immediately and explicitly as (17.73) Exercise 17.j21 + 1] (17.f2001.54) for the CG relations. Show by explicit counting that the matrices on both sides of (17. we may write the equivalence (17..f.71) is correct. rp)y'. Tensor Operators and the WignerEckart Theorem.1) + 1] $ .. the boldface number 2j + 1 thus characterizes an irreducible representation unambiguously. Verify this for a few examples by lettingh andi2 assume some simple values.432 Chapter 17 Rotations and Other Symmetry Operations each dimension.:. So far this chapter has been concerned mainly with the behavior under rotation of state vectors and wave functions.71) Exercise 17. i = . 7. to the basis Ihizjm).. i2 = . The ClebschGordan coefficients furnish the unitary transformation from the basis lid2mlm2).44) to obtain for il = .f2. 1.. if symbolically. 0.f. By use of the unitary condition for the D matrices and the orthogonality of the CG coefficients.:. + Inz)Y~l+"'2(O.f=2_+_1.75) Show that for infinitesimal rotations this equation is identical with the recursion relations (17..72) This expansion is called the ClebschGordan series.
Such information is useful in many applications of quantum mechanics.76) state.81) Exercise 17. since finite rotations can be generated by integration (exponentiation).cos nznx(l .nz sin 4> nxnz<1 . vectors.80) If (17. we find that (1  *eo . according to (16. and tensors.3. From a knowledge of this behavior alone. nxny(1 .77). The operators corresponding to various physical quantities will be characterized by their behavior under rotation as scalars. To first order. = 2: Rijej j=l where 1 .36).cos + n. Again it is helpful to confine the discussion to infinitesimal rotations. Let us suppose that a rotation R takes a state vector W by a unitary transformation UR into the state vector W' = URW.n. the condition for the components of a vector operator is seen to be UR Ap1 = 2: AiRij i (17.n.ny sin 4> nzny(1 .) cos 4> + n.77) Under a righthanded rotation parametrized by the axis 0 and the angle of rotation cfJ. If (17.38).cos 4» . (17. and check that this agrees with the result from formula (16.nx sin 4> 4» . A vector operator A for the system is defined as an operator whose expectation value (A) is a vector that rotates together with the physical system.cos 4» + ny sin 4» 4» + nz sin 4> (1 .1 and 16.cos 4» .eo X A .77).81) is consistent with the properties of the rotation group and its representation by the R matrices.) cos 4> R = ( nynx(l . (17.cos cfJ) 0 X (0 X) sin <f> 0 X] e (17.cos 4» + nx sin 4> (1 . we can.79) e. work out the rotation matrix R for an infinitesimal rotation characterized by the vector E = eo. J) A(1 *eo .7 Tensor Operators and the WignerEckart Theorem 433 11. using a Cartesian basis (x. J) + = A. verify that (17. e' + (17.) cos 4> + n. z). and substituted in (17. Then we require that (WIA!W). 3 y. By carrying out two successive rotations.78) or in matrix form. express the relation between e and e' as e' = [1 + (1 . e where e denotes an arbitrary vector that also rotates with the system into the vector e'. nynz(l . It is convenient to consider the rotationally invariant scalar product (A) .78) is applied to an infinitesimal rotation by an angle cfJ = e. we infer the condition UR AU1 .21.n. much information will be inferred about the structure of the matrix elements. e = A . Since W is an arbitra~y e= (W'IAlw')· e' (17. Exercise 17.22.79) is substituted in (17.
If the system is a particle with spin. consider the case where the coordinate operators x. v'2. This would result in a much more complicated operator J. to the case k = 1 and employ the explicit form of the spherical harmonics Yf.. such as a particle without spin. such a representation is reducible.82) in agreement with the commutation relations (11. p.82). with q = k.83) As a special case (for k = 1). As an example.434 or Chapter 17 Rotations and Other Symmetry Operations I[A. A vector operator A must satisfy (17. as given by Eq. the usual Cartesian tensors induce reducible representations if the rank of the tensor exceeds 1. we may apply (17. On the other hand. as long as E is external to the system and not subject to rotation together with it.83) to the (irreducible) vector operator r and compare the formula with (17. fi . Neither condition (17. k. under rotation induce a representation of the rotation group.81).84) . But E would become a proper vector operator if the system were enlarged so as to include the electromagnetic field in the dynamical description. A/ = URAp'k. This is a set of 2k + 1 operators Tkq. The Cartesian components of A I are linear combinations of the components of A.z. L are all vector operators for this system. In this case. If we apply the rotational transformation of spherical harmonics.42).42).z. (In particular. from (11. we can identify J with L = r X p. the spin wave function must be rotated together with the space wave function. Whether or not a vector A constitutes a vector operator for a system depends on the definition of the physical system and the structure of its angular momentum operator J.81) nor condition (17. The lefthand side of (17. that is.82) would then be satisfied by E. and the commutation relations (17. in analogy to (17. we find ( x' + .x' v'2iy' . If Aj maps a state qr into <1>.82) for arbitrary fi.82) is satisfied by such a field. y.19) and (16. can according to (9. Generally. J] = infi X A I (17. which satisfy the transformation law k URTkqU'k = ~ TlD~k}iR) q'=k (17.81).'y') = (_x + v'2iy .82) for A = J. The theory will be generalized along these lines in Chapter 23.1. an external electric field E acting on the system does not in general make up a vector operator for this systemeven though E is a vector. A vector IS a tensor of rank one.128) be interpreted as the operator into which Aj is transformed by the rotation.82). k . We generalize these ideas and define a tensor operator as a set of operators which.) The reduction decomposes the space of the tensors into rotationally invariant subspaces. z provide a complete description of the dynamical system. k + 1.81). and the expansion coefficients Rij are the matrix elements of the fundamental threedimensional representation of the rotation group in a basis spanned by the • Cartesian unit vectors.. . A/ = URAp'k maps URqr into UR<1>. J itself is a vector operator and satisfies (17. S becomes a vector operator provided that J is now taken to be L + S. The quantities r. Of course. as can be verified by using the fundamental commutation relations between r and p and checking the commutation relations (17. suggesting that we define as a building block of the theory the socalled irreducible spherical tensor operator of nonnegative integer rank k. x v'2iY)D<!)(R) (17. (17.
the labels a and a' signify the totality of all quantum numbers needed to specify the basis states of the system completely. k + 1. we infer the angular momentum selection rules for irreducible spherical tensor operators. The matrix element (a'j'm' ITll ajm) vanishes unless Iq = m' . Hence.87) This important formula embodies the WignerEckart theorem. The constant of proportionality (a'l' II T k II aj) is called the reduced matrix element of the irreducible spherical tensor operator T k.85) Exercise 17. .':~. If we identify j = j 1> k = jz.::. j . and q which specify the orientation of the system. we can now formulate and prove the WignerEckart theorem. how much information about its matrix elements in the angular momentum basis can be inferred? To find the answer.iA y . How does this matrix transform the rotation matrix R into D(l)(R)? Armed with the definition of an irreducible tensor operator.h. We obtain (a'l'm'luRTkquklajm) = q'=k 2: k (a'l'm'ITtlajm)D~~)q(R) Vsing the definition (17.86) must similarly determine the matrix elements of the tensor operator. we convert this equation into the set of simultaneous equations 2: 1'/': D<'. the Cartesian components of any vector operator A define an irreducible spherical tensor operator T 1 of rank 1 by the relations + T1 1   T 1 1  Ax . k) is an irreducible tensor operator.75) shows that the two have exactly the same structure. If the D are known.13.28) of the matrix elements of U R . the theorem also has great practical value. we can conclude that the two solutions of these linear homogeneous recursion relations must be proportional: I (a'l'm' ITkql ajm) = (jkmqljkj'm')(a'l' II Tk II aj) I (17. Construct the 3 X 3 matrix for the transformation (17.(R) (a'l' 1L ' 1 Tli aj1L)D<'.1.7 Tensor Operators and the WignerEckart Theorem 435 where the representation matrix D(l)(R) is the matrix (17.29) for j = 1. because it separates the geometric and symmetryrelated properties of the matrix elements from the other physical properties that are contained in the reduced matrix element.. and j' = 13. and on the "other" quantum numbers a and a'. which answers the following question: If Ii? (q = k.88) . It depends on the nature of the tensor operator. Hence.'~*(R) = 2: (a'l'm' q' I Ttl ajm)D~~)q(R) (17. except for a common factor.53) and (17. let us take the matrix element of the operator equation (17.86) A glance at this equation and (17.85) and show that it is unitary. Since the CG coefficients are readily available.m I (17.83) between the states (a'j'm' I and Iajm). In addition to the angular momentum quantum numbers. The WignerEckart theorem provides us with a fundamental insight. (17.23.75) determine the CG coefficients for given values ofiI. on the angular momentum quantum numbers j and j I. the linear homogeneous relations (17. but not on the quantum numbers m. From the fundamental properties (17. m ' .:  V2 (17.57) of the CG coefficients..
Prove that (a'j and deduce the identity II Sk II aj) = (aj II Tk I a'j)* (17.83) defines a tensor operator.89) In particular. it is much easier and sufficient to check this condition for infinitesimal rotations.91) Exercise 17. qljkjm) Exercise 17. Exercise 17. ± 1 with j = j' = 0 excluded.436 Chapter 17 Rotations and Other Symmetry Operations and the numbers.j == j' .m == m' . j'.24. Tkq] = q'=k 2: k Tkq' (kq' 10 . J.90) where 0 points along the arbitrarily chosen axis of rotation.26.27.91) as a test for an irreducible spherical tensor operator.83) becomes in first order. Derive the angular momentum selection rules for an irreducible secondrank (quadrupole) tensor operator.90) the standard commutation relations (17. and k satisfy the triangular condition: Iljj'I~k~j+j'1 (17. Although (17. to test a given set of 2k + 1 operators Tkq for its rotational behavior and irreducibility. The selection rules for a vector operator (k = 1) are l:i. we derive from (17. it follows that a scalar operator (k = 0) has nonvanishing matrix elements only if m = m' and j = j'.28. prove that the contracted operator q=k 2: +k (l)qSffTk"q (17.27) for the matrix elements of the angular momentum components. (17.m = 0. j. Show that the trace of any irreducible spherical tensor operator vanishes.j = 0.92) (jkmqjjkjm') = (I)q(jkm'.25.90) can be reversed. suggesting (though not proving) the sufficiency of the commutation relations (17. ± 1 and l:i.93) . Exercise 17.83) to (17. For these. except those of rank 0 (scalar operators). If Sf and Tkq are two irreducible spherical tensor operators of rank k. Exercise 17. JI kq) (17. Show that by successive application of infinitesimal rotations (exponentiation) the argument leading from (17. Using the familiar expressions (17. [0 . Show that the tensor operators Sf = (I)qTk"qt and Tff transform in the same way under rotation.90) or (17.
Prove that every irreducible spherical tensor operator of rank k can be expressed as a linear combination of the basic constituent tensor operators Tl(a'j'.62) and the WignerEckart theorem. the electric 2kpole moment. <p) (17. Exercise 17. For a particle in a centralforce potential. Their matrix elements may therefore be evaluated by use of the WignerEckart theorem.97) At each order of k. the field is characterized by 2k + 1 constants At. Exercise 17.29. with energy basis functions separable in radial and spherical polar coordinates. The energy can therefore be expressed as the electric multipole expansion. .30. The static electric moments of a charge distribution.95) to the interaction energy. cf> satisfies Laplace's equation and may be written in the form cf>(r) = k=O q=k 2: 2: +k AirkYl(O. the concepts developed here are of great generality and can be extended to other symmetries and their group representations. Prove that the static 2k pole moment of a charge distribution has zero expectation value in any state with angular momentumj < kl2. provided that the rotations are generated by the operator L = r X p.. show that this Exercise 17. a particle of charge q contributes E = q J l/f*(r)cf> (r)l/J{r) d r 3 (17. In first order. a scalar operator. If the sources of the electric field are at large distances.91) for irreducible spherical tensor operators.94) with the appropriate reduced matrix elements serving as the coefficients in the linear combination.96) the origin being chosen at the center of mass of the atom or nucleus.31. They are best defined in terms of the perturbation energy of the particles in an external electric field E = . E = J p(r)cf> (r) d 3 r = ~ Ai J p(r)rkYl d 3 r (17. are examples of tensor operators. aj) = mm' 2: Ia'j'm')(jkmq Ijkj'm')(ajm I (17. rklkq(r).Eckart Theorem. Although the WignerEckart theorem has been discussed in this section entirely in the context of rotations in ordinary threedimensional space. Verify this property explicitly for the quadrupole moment by use of the formulas (17. = 1. use (17. and there is correspondingly a spherical tensor of rank k. 8.74) to evaluate the reduced matrix elements of the electric multiple moment operators. Applications of the Wigner. the components of these tensors obviously pass the test (17.e. Since spherical harmonics transform irreducibly under rotations in ordinary space. T. For k scalar operator is just the inner product S .8 Applications of the WignerEckart Theorem 437 is a tensor operator of rank zero. i. such as an atom or a nucleus.Vcf>.
A and other known quantities. see Condon and Shortley (1935). Prove from the WignerEckart theorem and formula (17. Since the righthand side of (17. Generally. J Iajm) = j(j + 1)h2 = C(aj I JI aj) Substituting the reduced matrix elements of A and J from the last two equations into (17.? Since both L and S are vector operators with respect to the rotations generated by the total angular momentum J = L + S.102) 7Haken and Wolf (1993). all matrix elements of flo in the angular momentum basis are proportional to each other. It can be used as the starting point of the derivation of the matrix elements of magnetic moment operators that are important in spectroscopy. as an application of the WignerEckart theorem to a vector operator A and to the special vector operator J.32. discussion. Exercise 17. but very thorough.99) can be loosely interpreted as the projection of A on J. (17.99) by which the jdiagonal matrix elements of a vector operator A are expressed in terms of the matrix element of the scalar operator J . a. A: (a'jmIJ' Alajm) = C(a'j II A II aj) According to the results derived in Section 17. it can be evaluated by choosing a' = a and A = J: (ajml J . the magnetic moment operator of an atom may be assumed to have the form (17.438 Chapter 17 Rotations and Other Symmetry Operations Next.101) where m is the electron mass. the constant C must be independent of the nature of the vector operator A and of the quantum numbers m. in so doing. II A I aj) (a'j'm'IJqlajm) . Mizushima (1970). ( a " 'IAql aJm ) . .100) Contributions to the magnetic moment of an atom (or nucleus) arise from the orbital motion of the charged particles and the intrinsic spins in the system. we obtain the important result Alajm) (' 'IJql' ) . According to the WignerEckart theorem. we use the relation (a'j'm'IAqlajm) (a'j' = .62) that the reduced matrix element of J is (aj' II J II aj) = Yj(j + l)h8~. this formula contains the theoretical justification for the vector model of angular momen.(a'jmIJ· + 1) Jm h 2j(j Jm Jm (17. (a'l' I J II aJ) For j' = j this may be further developed by noting that by virtue of the simple properties of J we must have for the scalar operator J . and a'. For an older. Hence. We therefore speak of the magnetic moment of the atom and. flo is also a vector operator. have reference to the expectation value JL = (ajj IJLz Iajj) = (jljO Ijljj)( aj II flo I aj) = Jj i I (aj I flo II aj) (17.tum.98).7.
it is convenient to consider inversions through a fixed origin (r ~ . a reflection with respect to the yz plane changes x into x.101) gives immediately But Hence. Parity.:(j e + 1) (ajjl (gL + gS)J2 + (gL  2 2 gs)(L . such as the electric charge of a particle. The Euclidean principle of relativity may be supplemented by the further assumption that space has no intrinsic chirality. The application of (17. This is no limitation.104) If LS coupling describes the state of the atom. which are symbolized by a. Using the usual notation J. or handedness: Processes take place in the same way in a physical system and its mirror image. should be treated in a reflection if symmetry is to be preserved. and Time Reversal 439 In the classical limit (j ~ 00) the reduced matrix element becomes identical with the magnetic moment Jt. Reflection Symmetry.99) to the magnetic moment vector operator (17. since it is impossible to transform a system into its mirror image without distorting it into some intermediate configurations that are physically very different from the original system. an important difference between rotational symmetry and reflection symmetry must be remembered. obtained from one another by reflection with respect to a plane.8 ) I ajj) (17. Jt = 4mfu. In rotating a system from one orientation to another. since any reflection can be obtained by an inversion followed by a rotation. It is therefore not at all obvious how quantities. I ajj) is an approximate eigenstate of the total orbital and the total spin angular momenta. This is not so in the case of reflection.9 Reflection Symmetry. In developing the mathematical formalism. We may call this assumption the extended Euclidean principle of relativity. and leaves y and z unchanged. Parity. Only experience can tell us whether it is possible to create a suitable mirror image of a complex physical system in accordance with the assumption of reflection symmetry.r) instead of plane reflections.eliJ [1 + J(J + 1)  2mc L(L + 1) + S(S 2J(J + 1) + 1)] = _ eli gJJ 2mc (17. and S for the three angular momentum quantum numbers that characterize the atomic state. an inversion is the same as three successive reflections with respect . we obtain for the magnetic moment of the atom. Conversely.105) The expression in the bracket (gJ) represents the general form for LS coupling of the Lande gfactor. similarly. and Time Reversal. By definition. for which the magnetic moment is to be evaluated. seemingly unrelated to coordinate displacements. and the values gL = 1 and gs = 2 for the electrons. we can proceed gradually and take it through a continuous sequence of rigid displacements. Jt = . all of which are equivalent. Px is transformed into Px andpy andpz remain unchanged. 9. In any test of the extended Euclidean principle of relativity. L.
and there is no other matrix that commutes with every component of u.440 Chapter 17 Rotations and Other Symmetry Operations to perpendicular planes. we may require that U~ = 1 (17. The Euclidean principle of relativity connects physical laws with g