Introduction to Quantum Mechanics David J. Griffiths Reed College Prentice Hall Upper Saddle River, New Jersey 07458Fundamental Equations Schrédinger equation: ow nee = HY Time independent Schrédinger equation: Hw = Eu, W = pe tEt/h Standard Hamiltonian: > n=-Lvtiv am ‘Time dependence of an expectation value: d 40) - + 1,0 + (2) Generalized uncertainty principle: 1 2 > I= (A, BI cate >|: (4B) Heisenberg uncertainty principle: o20p > h/2 Canonical commutator: [z, p] = ih Angular momentum: [Le Ly] =#hL,, [Ly,L.]=thbe, [Lz,L2] = ihly Pauli matrices: 01 = \1 0Fundamental Constants Planck’s constant : h 1.05457 x 10-4 Js " Speed of light : 2.99792 x 10® m/s ° " Mass of electron : me = 9.10939 x 10-5! kg Mass of proton : mp = 1.67262 x 10-77 kg Charge of electron: 9 -e = 1.60218 x 10-°C Permittivity of space : 8.85419 x 10-1? C?/Jm Boltzmann constant: kg 1.38066 x 10-75 J/K Hydrogen Atom Fine structure constant: a = e?/Ameolic = 1/137.036 Bohr radius : a = Ameoh?/m.e? = h/am.c = 5.29177 x 10-4 m Bohr energies : E, = Ey/n? (n=1,2,3,..) Ground state energy : -E, = mye4/2(4meo)?h? =a?m,c?/2 = 13.6057 eV Wave function : vo = gase!* Rydberg formula : L =R (4 - >) Rydberg constant : Ro = —E,/2he = 1.09737 x 107 /mLibrary of Congress Cataloging-in-Publication Data Griffiths, David J. (David Jeffrey) Introduction to quantum mechanics / David J. Griffiths. Pp. cm. Includes bibliographical references and index. ISBN 0-13-124405-1 1, Quantum theory. I. Title QC174.12.G75 1994 94-14133 530,1'2—de20 cP Acquisitions Editor: Ray Henderson Assistant Acquisition Editor: Wendy Rivers Editorial Assistant: Pam Holland-Moritz Production Editors: Rose Kernan and Fred Dahl Copy Editor: Rose Kernan Production Coordinator: Trudy Pisciotti (© 1995 by Prentice Hall, Inc. Upper Saddle River, NJ 07458 All rights reserved. No part of this book may be reproduced, in any form or by any means, without permission in writing from the publisher. Printed in the United States of America 10 ISBN O-13-124405-1 Prentice-Hall International (UK) Limited, London Prentice-Hall of Australia Pty. Limited, Sydney ISBN O-13-124405-1 Prentice-Hall of Canada, Inc., Toronto 700.00 Prentice-Hall Hispanoamericana,S. A., Mexico Prentice-Hall of India Private Limited, New Dethi Prentice-Hall of Japan, Inc., Tokyo 9780131 Pearson Education Asia Pte. Ltd., Singapore Editora Prentice-Hall do Brasil, Ltda., Rio de JaneiroCONTENTS PREFACE, vii PARTI THEORY CHAPTER 1 THE WAVE FUNCTION, 1 1.1 The Schrodinger Equation, 1 1.2 The Statistical Interpretation, 2 1.3 Probability, 5 1.4 Normalization, 11 1.5 Momentum, 14 1.6 The Uncertainty Principle, 17 CHAPTER 2 THE TIME-INDEPENDENT SCHRODINGER EQUATION, 20 2.1 Stationary States, 20 2.2 The Infinite Square Well, 24 2.3 The Harmonic Oscillator, 31 2.4 The Free Particle, 44 2.5 The Delta-Function Potential, 50 2.6 The Finite Square Well, 60 iiiiv Contents 2.7 The Scattering Matrix, 66 Further Problems for Chapter 2, 68 CHAPTER 3 FORMALISM, 75 3.1 Linear Algebra, 75 3.2 Function Spaces, 95 3.3 The Generalized Statistical Interpretation, 104 3.4 The Uncertainty Principle, 108 Further Problems for Chapter 3, 116 CHAPTER 4 QUANTUM MECHANICS IN THREE DIMENSIONS, 121 4.1 Schrodinger Equations in Spherical Coordinates, 121 4.2 The Hydrogen Atom, 133 4.3 Angular Momentum, 145 4.4 Spin, 154 Further Problems for Chapter 4, 170 CHAPTER 5 IDENTICAL PARTICLES, 177 5.1 Two-Particle Systems, 177 5.2 Atoms, 186 5.3 Solids, 193 5.4 Quantum Statistical Mechanics, 204 Further Problems for Chapter 5, 218 PART II APPLICATIONS CHAPTER 6 TIME-INDEPENDENT PERTURBATION THEORY, 221 6.1 Nondegenerate Perturbation Theory, 221 6.2 Degenerate Perturbation Theory, 227 6.3 The Fine Structure of Hydrogen, 235 6.4 The Zeeman Effect, 244 6.5 Hyperfined Splitting, 250 Further Problems for Chapter 6, 252CHAPTER 7 THE VARIATIONAL PRINCIPLE, 256 7.1 Theory, 256 7.2 The Ground State of Helium, 261 7.3 The Hydrogen Molecule lon, 266 Further Problems for Chapter 7, 271 CHAPTER 8 THE WKB APPROXIMATION, 274 8.1 The “Classical” Region, 275 8.2 Tunneling, 280 8.3 The Connection Formulas, 284 Further Problems for Chapter 8, 293 CHAPTER 9 TIME-DEPENDENT PERTURBATION THEORY, 298 9.1 Two-Level Systems, 299 9.2 Emission and Absorption of Radiation, 306 9.3 Spontaneous Emission, 311 Further Problems for Chapter 9, 319 CHAPTER 10 THE ADIABATIC APPROXIMATION, 323 10.1 The Adiabatic Theorem, 323 10.2 Berry’s Phase, 333 Further Problems for Chapter 10, 349 CHAPTER 11 SCATTERING, 352 11.1 Introduction, 352 11.2 Partial Wave Analysis, 357 11.3 The Born Approximation, 363 Further Problems for Chapter 11, 373 AFTERWORD, 374 INDEX, 386 Contents vPREFACE Unlike Newton’s mechanics, or Maxwell's electrodynamics, or Einstein's relativity, quantum theory was not created—or even definitively packaged—by one individual, and it retains to this day some of the scars of its exhilirating but traumatic youth. ‘There is no general consensus as to what its fundamental principles are, how it should be taught, or what it really “means.” Every competent physicist can “do” quantum mechanics, but the stories we tell ourselves about what we are doing are as various as the tales of Scheherazade, and almost as implausible. Richard Feynman (one of its greatest practitioners) remarked, “I think I can safely say that nobody understands quantum mechanics.” The purpose of this book is to teach you how to do quantum mechanics. Apart from some essential background in Chapter 1, the deeper quasi-philosophical ques- tions are saved for the end. I do not believe onecan intelligently discuss what quantum mechanics means until one has a firm sense of what quantum mechanics does. But if you absolutely cannot wait, by all means read the Afterword immediately following Chapter 1. Not only is quantum theory conceptually rich, it is also technically difficult, and exact solutions to all but the most artificial textbook examples are few and far between. It is therefore essential to develop special techniques for attacking more realistic problems. Accordingly, this book is divided into two parts'; Part I covers the basic theory, and Part II assembles an arsenal of approximation schemes, with illustrative applications. Although it is important to keep the two parts logically separate, it is not necessary to study the material in the order presented here. Some instructors, for example, may wish to treat time-independent perturbation theory immediately after Chapter 2. ' This structure was inspired by David Park's classic text Introduction to the Quantum Theory, 3rd ed,, (New York: McGraw-Hill, 1992), viiviii Preface This book is intended for a one-semester or one-year course at the junior or senior level. A one-semester course will have to concentrate mainly on Part I; a full-year course should have room for supplementary material beyond Part II. The reader must be familiar with the rudiments of linear algebra, complex numbers, and calculus up through partial derivatives; some acquaintance with Fourier analysis and the Dirac delta function would help. Elementary classical mechanics is essential, of course, and a little electrodynamics would be useful in places. As always, the more physics and math you know the easier it will be, and the more you will get out of your study. But 1 would like to emphasize that quantum mechanics is not, in my view, something that flows smoothly and naturally from earlier theories. On the contrary, it represents an abrupt and revolutionary departure from classical ideas, calling forth a wholly new and radically counterintuitive way of thinking about the world. That, indeed, is what makes it such a fascinating subject. At first glance, this book may strike you as forbiddingly mathematical. We en- counter Legendre, Hermite, and Laguerre polynomials, spherical harmonics, Bessel, Neumann, and Hankel functions, Airy functions, and even the Riemann Zeta function —not to mention Fourier transforms, Hilbert spaces, Hermitian operators, Clebsch- Gordan coefficients, and Lagrange multipliers. Is all this baggage really necessary? Perhaps not, but physics is like carpentry: Using the right tool makes the job easier, not more difficult, and teaching quantum mechanics without the appropriate mathe- matical equipment is like asking the student to dig a foundation with a screwdriver. (On the other hand, it can be tedious and diverting if the instructor feels obliged to give elaborate lessons on the proper use of each tool. My own instinct is to hand the students shovels and tell them to start digging. They may develop blisters at first, but I still think this is the most efficient and exciting way to learn.) At any rate, I can assure you that there is no deep mathematics in this book, and if you run into something unfamiliar, and you don’t find my explanation adequate, by all means ask someone about it, or look it up. There are many good books on mathematical methods—I par- ticularly recommend Mary Boas, Mathematical Methods in the Physical Sciences, 2nd ed., Wiley, New York (1983), and George Arfken, Mathematical Methods for Physicists, 3rd ed., Academic Press, Orlando (1985). But whatever you do, don’t let the mathematics—which, for us, is only a tool—interfere with the physics. Several readers have noted that there are fewer worked examples in this book than is customary, and that some important material is relegated to the problems. This is no accident. I don’t believe you can learn quantum mechanics without doing many exercises for yourself. Instructors should, of course, go over as many problems in class as time allows, but students should be warned that this is not a subject about which anyone has natural intuitions—you’re developing a whole new set of muscles here, and there is simply no substitute for calisthenics. Mark Semon suggested that I offer a “Michelin Guide” to the problems, with varying numbers of stars to indicate the level of difficulty and importance. This seemed like a good idea (though, like the quality of a restaurant, the significance of a problem is partly a matter of taste); Ihave adopted the following rating scheme:Preface ix * —_anessential problem that every reader should study; 4 a somewhat more difficult or more peripheral problem; * + * an unusually challenging problem that may take over an hour, (No stars at all means fast food: OK if you're hungry, but not very nourishing.) Most of the one-star problems appear at the end of the relevant section; most of the three-star problems are at the end of the chapter. A solution manual is available (to instructors only) from the publisher. Thave benefited from the comments and advice of many colleagues, who sug- gested problems, read early drafts, or used a preliminary version in their courses. I would like to thank in particular Burt Brody (Bard College), Ash Carter (Drew Uni- versity), Peter Collings (Swarthmore College), Jeff Dunham (Middlebury College), Greg Elliott (University of Puget Sound), Larry Hunter (Amherst College), Mark Semon (Bates College), Stavros Theodorakis (University of Cyprus), Dan Velleman (Amherst College), and all my colleagues at Reed College. Finally, I wish to thank David Park and John Rasmussen (and their publishers) for permission to reproduce Figure 8.6, which is taken from Park’s Introduction to the Quantum Theory (footnote 1), adapted from I. Perlman and J. O, Rasmussen, “Alpha Radioactivity,” in Encyclopedia of Physics, vol. 42, Springer-Verlag, 1957.PARTI THEORYCHAPTER 1 THE WAVE FUNCTION 1.1 THE SCHRODINGER EQUATION Imagine a particle of mass m, constrained to move along the x-axis, subject to some specified force F(x, t) (Figure 1.1). The program of classical mechanics is to deter- mine the position of the particle at any given time: x(). Once we know that, we can figure out the velocity (v = dx/dt), the momentum (p = mv), the kinetic energy (T = (1/2)mv?), or any other dynamical variable of interest. And how do we go about determining x(1)? We apply Newton's second law: F = ma. (For conservative systems—the only kind we shall consider, and, fortunately, the only kind that occur at the microscopic level—the force can be expressed as the derivative of a potential energy function,’ F = —aV /ax, and Newton's law reads m d?x/di? = —aV /ax.) This, together with appropriate initial conditions (typically the position and velocity at t = 0), determines x(f). Quantum mechanics approaches this same problem quite differently. In this case what we're looking for is the wave function, (x, 1), of the particle, and we get it by solving the Schrddinger equation: t.) "Magnetic forces are an exception, but let's not worry about them just yet. By the way, we shall assume throughout this book that the motion is nonrelativistic (» « ¢).2 Chap. 1 The Wave Function | m x(t) Figure 1.1: A “particle” constrained to move in one dimension under the influ- ence of a specified force, —> F(x,t) oO xy Here i is the square root of —1, and hi is Planck’s constant—or rather, his original constant (/) divided by 27: af 34 h= mn 1.054573 x 10-™J s. [1.2] The Schrédinger equation plays a role logically analogous to Newton's second law: Given suitable initial conditions [typically, W(x, 0)], the Schrédinger equation de- termines W(x, 4) for all future time, just as, in classical mechanics, Newton's law determines x(f) for all future time. 1.2 THE STATISTICAL INTERPRETATION But what exactly is this “wave function”, and what does it do for you once you've got it? After all, a particle, by its nature, is localized at a point, whereas the wave function (as its name suggests) is spread out in space (it’s a function of x, for any given time 1). How can such an object be said to describe the state of a particle? The answer is provided by Born’s statistical interpretation of the wave function, which says that IW(x, OI? gives the probability of finding the particle at point x, at time tor, more precisely,? > probability of finding the particle Mon OF dx { between x and (x + dx), at time ¢. 013) For the wave function in Figure 1.2, you would be quite likely to find the particle in the vicinity of point A, and relatively unlikely to find it near point B. The statistical interpretation introduces a kind of indeterminacy into quantum mechanics, for even if you know everything the theory has to tell you about the ?The wave function itself is complex, but |\W{? = ¥*W (where Y* is the complex conjugate of ) is real and nonnegative—as a probability, of course, must be.Sec. 1.2: The Statistical Interpretation 3 ty? a A BoC x Figure 1.2: A typical wave function. The particle would be relatively likely to be found near 4, and unlikely to be found near 8. The shaded area represents the probability of finding the particle in the range dx. particle (to wit: its wave function), you cannot predict with certainty the outcome of a simple experiment to measure its position—all quantum mechanics has to offer is statistical information about the possible results. This indeterminacy has been profoundly disturbing to physicists and philosophers alike. Is it a peculiarity of nature, a deficiency in the theory, a fault in the measuring apparatus, or what? Suppose I do measure the position of the particle, and I find it to be at the point C. Question: Where was the particle just before I made the measurement? There are three plausible answers to this question, and they serve to characterize the main schools of thought regarding quantum indeterminacy: 1. The realist position: The particle was at C. This certainly seems like a sensible response, and it is the one Einstein advocated. Note, however, that if this is true then quantum mechanics is an incomplete theory, since the particle really was at C, and yet quantum mechanics was unable to tell us so. To the realist, indeterminacy is not a fact of nature, but a reflection of our ignorance. As d’Espagnat put it, “the position of the particle was never indeterminate, but was merely unknown to the experimenter” Evidently W is not the whole story—some additional information (known as a hidden variable) is needed to provide a complete description of the particle. 2. The orthodox position: The particle wasn’t really anywhere. Tt was the act of measurement that forced the particle to “take a stand” (though how and why it decided on the point C we dare not ask). Jordan said it most starkly: “Observations not only disturb what is to be measured, they produce it. ... We compel (the particle] to assume a definite position.” This view (the so-called Copenhagen interpretation) is associated with Bohr and his followers. Among physicists it has always been the Bernard d’Espagnat, The Quantum Theory and Reality, Scientific American, Nov. 1979 (Vol. 241), p. 165 “Quoted in a lovely article by N. David Mermin, Is the moon there when nobody looks?, Physics Today, April 1985, p. 38.4 Chap. 1 The Wave Function most widely accepted position. Note, however, that if it is correct there is something very peculiar about the act of measurement—something that over half a century of debate has done precious little to illuminate. 3. The agnostic position: Refiuse to answer. This is not quite as silly as it sounds—after all, what sense can there be in making assertions about the status of a particle before a measurement, when the only way of knowing whether you were right is precisely to conduct a measurement, in which case what you get is no longer “before the measurement”? It is metaphysics (in the perjorative sense of the word) to worry about something that cannot, by its nature, be tested. Pauli said, “One should no more rack one’s brain about the problem of whether something one cannot know anything about exists all the same, than about the ancient question of how many angels are able to sit on the point of a needle.”* For decades this was the “fall-back” position of most physicists: They'd try to sell you answer 2, but if you were persistent they’d switch to 3 and terminate the conversation, Until fairly recently, all three positions (realist, orthodox, and agnostic) had their partisans, But in 1964 John Beil astonished the physics community by showing that it makes an observable difference if the particle had a precise (though unknown) position prior tothe measurement. Bell's discovery effectively eliminated agnosticism as a viable option, and made it an experimental question whether 1 or 2 is the correct choice. I'll return to this story at the end of the book, when you will be in a better position to appreciate Bell's theorem; for now, suffice it to say that the experiments have confirmed decisively the orthodox interpretation: A particle simply does not have a precise position prior to measurement, any more than the ripples on a pond do; it is the measurement process that insists on one particular number, and thereby in a sense creates the specific result, limited only by the statistical weighting imposed by the wave function. But what if I made a second measurement, immediately after the first? Would I get C again, or does the act of measurement cough up some completely new number each time? On this question everyone is in agreement: A repeated measurement (on the same particle) must return the same value. Indeed, it would be tough to prove that the particle was really found at C in the first instance if this could not be confirmed by immediate repetition of the measurement. How does the orthodox interpretation account for the fact that the second measurement is bound to give the value C? Evidently the first measurement radically alters the wave function, so that it is now sharply peaked about C (Figure 1.3). We say that the wave function collapses upon measurement, to a spike at the point C (W soon spreads out again, in accordance with the Schrédinger equation, so the second measurement must be made quickly). There SQuoted by Mermin (previous footnote), p. 40. Phis statement is alittle too strong: There remain a few theoretical and experimental loopholes, some of which T shall discuss in the Afterword. And there exist other formulations (such as the many worlds interpretation) that do not fit cleanly into any of my three categories. But I think it is wise, at least from a pedagogical point of view, to adopt a clear and coherent platform at this stage, and worry about the alternatives later.Sec. 1.3: Probability 5 bel? Figure 1.3: Collapse of the wave function: graph of |¥/? immediately after a measurement has found the particle at point C. are, then, two entirely distinct kinds of physical processes: “ordinary” ones, in which the wave function evolves in a leisurely fashion under the Schrédinger equation, and “measurements”, in which suddenly and discontinuously collapses.” 1.3 PROBABILITY Because of the statistical interpretation, probability plays a central role in quantum mechanics, so I digress now for a brief discussion of the theory of probability. It is mainly a question of introducing some notation and terminology, and I shall do it in the context of a simple example. Imagine a room containing 14 people, whose ages are as follows: one person aged 14 one person aged 15 three people aged 16 two people aged 22 two people aged 24 five people aged 25. If we let N(j) represent the number of people of age j, then ‘The role of measurement in quantum mechanics is so critical and so bizarre that you may well be wondering what precisely constitutes a measurement. Does it have to do with the interaction between 1 microscopic (quantum) system and a macroscopic (classical) measuring apparatus (as Bohr insisted), or is it characterized by the leaving of @ permanent “record” (as Heisenberg claimed), or does it involve the intervention of a conscious “observer” (as Wigner proposed)? ll return to this thoy issue in the Afterword; for the moment let's take the naive view: A measurement is the kind of thing that a scientist does in the laboratory, with rulers, stopwatches, Geiger counters, and so on.6 Chap. 1 The Wave Function NU) 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 267 Figure 1.4: Histogram showing the number of people, N(/), with age j, for the example in Section 1.3. Nad (15) =1 N(16) =3 N(22) =2 N(24) =2 N(25) =5 while N(17), for instance, is zero. The total number of people in the room is N=SONG). 4] j= In this instance, of course, N = 14.) Figure 1.4 is a histogram of the data. The following are some questions one might ask about this distribution. Question 1. If you selected one individual at random from this group, what is the probability that this person's age would be 15? Answer: One chance in 14, since there are 14 possible choices, all equally likely, of whom only one has this particular age. If P(j) is the probability of getting age j, then P(14) = 1/14, P(IS) = 1/14, P(16) = 3/14, and so on. In general, _ NU) PG) {1.5] Notice that the probability of getting either 14 or 15 is the sum of the individual probabilities (in this case, 1/7). In particular, the sum of all the probabilities is 1— you're certain to get some age: oo Pw =t. 11.6) iaSec. 1.3: Probability 7 Question 2. What is the most probable age? Answer: 25, obviously; five people share this age, whereas at most three have any other age. In general, the most probable j is the j for which P(/) is a maximum, Question 3. What is the median age? Answer: 23, for 7 people are younger than 23, and 7 are older. (In general, the median is that value of j such that the probability of getting a larger result is the same as the probability of getting a smaller result.) Question 4. What is the average (or mean) age? Answer: (14) + (15) + 3(16) + 2(22) + 2(24) + 5125) _ 294 _ 14 4“ In general, the average value of j (which we shall write thus: (/)) is given by voy y= = IND _ jc, (7) j=0 Notice that there need not be anyone with the average age or the median age—in this example nobody happens to be 21 or 23. In quantum mechanics the average is usually the quantity of interest; in that context it has come to be called the expectation value, It’s a misleading term, since it suggests that this is the outcome you would be most likely to get if you made a single measurement (that would be the most probable value, not the average value)—but I’m afraid we're stuck with it. Question 5. What is the average of the squares of the ages? Answer: You could get 14? = 196, with probability 1/14, or 15? = 225, with probability 1/14, or 16? = 256, with probability 3/14, and so on. The average, then, is ie 18] In general, the average value of some function of j is given by (1) =O FDPU)- 11.91 a (Equations 1.6, 1.7, and 1.8 are, if you like, special cases of this formula.) Beware: ‘The average of the squares (({ /)) is not ordinarily equal to the square of the average ((/)?). For instance, if the room contains just two babies, aged 1 and 3, then (x?) = 5, but (x)? = 4, Now, there is a conspicuous difference between the two histograms in Figure 1.5, even though they have the same median, the same average, the same most prob- able value, and the same number of elements: The first is sharply peaked about the average value, whereas the second is broad and flat. (The first might represent the age profile for students in a big-city classroom, and the second the pupils in a one- room schoolhouse.) We need a numerical measure of the amount of “spread” in a8 Ni) Chap. 1 123 The Wave Function NU) A TTT TTT). 456789 0 jf 1234567893 1 j Figure 1.5: Two histograms with the same median, same average, and same most probable value, but different standard deviations. distribution, with respect to the average. The most obvious way to do this would be to find out how far each individual deviates from the average, Aj=i-(/), [1.10] and compute the average of Aj. Trouble is, of course, that you get zero, since, by the nature of the average, A/ is as often negative as positive: (4) = VU - PO = ViPa- VY Pw =()-()=0. (Note that (7) is constant—it does not change as you go from one member of the sample to another—so it can be taken outside the summation.) To avoid this irritating problem, you might decide to average the absolute value of Aj. But absolute values are nasty to work with; instead, we get around the sign problem by squaring before averaging: o = ((As)’). tH] This quantity is known as the variance of the distribution; o itself (the square root of the average of the square of the deviation from the average—gulp!) is called the standard deviation, The latter is the customary measure of the spread about (/). ‘There is a useful little theorem involving standard deviations: a7 = (As) = APL) = OU - (APP) = VU? - 2 + PW) = EPP = AN VIP + UP YL Pw = (7) - 2+ UP, or oF =(P)- (i). [1.12] Equation 1.12 provides a faster method for computing o: Simply calculate (j2) and (j)°, and subtract. Incidentally, I warned you a moment ago that (/*) is not, in general,Sec. 1.3: Probability 9 equal to (/)?. Since o? is plainly nonnegative (from its definition in Equation 1.11), Equation 1.12 implies that Puy, [1.13] and the two are equal only when « = 0, which is to say, for distributions with no spread at all (every member having the same value), So far, I have assumed that we are dealing with a discrete variable—that is, one that can take on only certain isolated values (in the example, / had to be an integer, since I gave ages only in years). But it is simple enough to generalize to continuous distributions. If I select a random person off the street, the probability that her age is precisely 16 years, 4 hours, 27 minutes, and 3.3333 seconds is zero. The only sensible thing to speak about is the probability that her age lies in some interval—say, between 16 years, and 16 years plus one day. If the interval is sufficiently short, this probability is proportional to the length of the interval. For example, the chance that her age is between 16 and 16 plus nvo days is presumably twice the probability that it is between 16 and 16 plus one day. (Unless, I suppose, there was some extraordinary baby boom 16 years ago, on exactly those days—in which case we have chosen an interval too long for the rule to apply. If the baby boom lasted six hours, we'll take intervals of a second or less, to be on the safe side. Technically, we're talking about infinitesimal intervals.) Thus { probability that individual (chosen at random) =pixydx. [1.14] lies between x and (x + dx) The proportionality factor, p(x), is often loosely called “the probability of getting x,” but this is sloppy language; a better term is probability density. The probability that x lies between a and b (a finite interval) is given by the integral of p(x) 6 P= f p(x)dx, [1.15] and the rules we deduced for discrete distributions translate in the obvious way: 400 | pix)dx =1, [1.16] +00 w= f xp(x)dx, (1.17) 400 (FG) -[ S(xdo(x)dx, [1.18] (Axy) = 0?) = (xP. U.19}10 Chap. 1 The Wave Function +Problem 1.1 For the distribution of ages in the example in Section 1.3, (a) Compute (j?) and (j). (b) Determine Aj for each j, and use Equation 1.11 to compute the standard devi- ation. (c) Use your results in (a) and (b) to check Equation 1.12. Problem 1.2 Consider the first 25 digits in the decimal expansion of 7 (3, 1, 4, 1, 5,9,...). (a) If you selected one number at random from this set, what are the probabilities of getting each of the 10 digits? (b) What is the most probable digit? What is the median digit? What is the average value? (c) Find the standard deviation for this distribution. Problem 1.3 The needle ona broken car speedometer is free to swing, and bounces perfectly off the pins at either end, so that if you give it a flick it is equally likely to come to rest at any angle between 0 and 7 (a) What is the probability density, o(@)? [(8) dé is the probability that the needle will come to rest between @ and (@ + d6).] Graph p(8) as a function of 6, from —1/2 to 31/2. (Of course, part of this interval is excluded, so p is zero there.) Make sure that the total probability is 1. (b) Compute (6), (67), and o for this distribution, (©) Compute (sin@), (cos 4), and (cos? 4). Problem 1.4 We consider the same device as the previous problem, but this time we are interested in the x-coordinate of the needle point—that is, the “shadow”, or “projection”, of the needle on the horizontal line. (a) What is the probability density p(x)? [(x) dx is the probability that the pro- jection lies between x and (x + dx).] Graph p(x) as a function of x, from —2r to +2r, where r is the length of the needle. Make sure the total probability is 1. [Hint: You know (from Problem 1.3) the probability that 4 is in a given range; the question is, what interval dx corresponds to the interval d@?] (b) Compute (x}, (x2), and o for this distribution. Explain how you could have obtained these results from part (c) of Problem 1.3.Sec. 1.4: Normalization 11 «Problem 1.5 A needle of length / is dropped at random onto a sheet of paper ruled with parallel lines a distance / apart. What is the probability that the needle will cross a line? (Hint: Refer to Problem 1.4.] +Problem 1.6 Consider the Gaussian distribution pxy= Ae ea? , where A, a, and 2 are constants. (Look up any integrals you need.) (a) Use Equation 1.16 to determine A. (b) Find (x), (x2), anda. (©) Sketch the graph of p(x). 1.4 NORMALIZATION We return now to the statistical interpretation of the wave function (Equation 1.3), which says that |\(x, £)|?is the probability density for finding the particle at point x, at time f. It follows (Equation 1.16) that the integral of ||? must be 1 (the particle's got to be somewhere): [1.20] f. “|w(.)P dx =l. Without this, the statistical interpretation would be nonsense. However, this requirement should disturb you: After all, the wave function is supposed to be determined by the Schrédinger equation—we can’t impose an extrane- ous condition on W without checking that the two are consistent. A glance at Equation 1.1 reveals that if W(x, £) is a solution, so too is A(x, 1), where A is any (complex) constant. What we must do, then, is pick this undetermined multiplicative factor so as to ensure that Equation 1.20 is satisfied. This process is called normalizing the wave function, For some solutions to the Schrédinger equation, the integral is infinite; in that case no multiplicative factor is going to make it 1. The same goes for the trivial solution Y = 0. Such non-normalizable solutions cannot represent particles, and must be rejected. Physically realizable states correspond to the “square-integrable” solutions to Schrédinger’s equation. "Bvidently W(x, ) must go to zer0 faster than 1/\/fx], as [x] + 00. Incidentally, normalization only fixes the modulus of A; the phase remains undetermined, However, as we shall see, the latter carries ro physical significance anyway.12 Chap. 1 The Wave Function But wait a minute! Suppose I have normalized the wave function at time t = 0. How do know that it will stay normalized, as time goes on and W evolves’ (You can't keep renormalizing the wave function, for then 4 becomes a function of f, and you no longer have a solution to the Schrédinger equation.) Fortunately, the Schrédinger equation has the property that it automatically preserves the normalization of the wave function—without this crucial feature the Schrédinger equation would be incompat- ible with the statistical interpretation, and the whole theory would crumble. So we’d better pause for a careful proof of this point: ft weaned ft owned 21) 4 x otde= [ 2iwenPae. at J. es OF [Note that the integral is a function only of f, so T use a total derivative (d/dt) in the first term, but the integrand is a function of x as well as £, so it’s a partial derivative (8/21) in the second one.] By the product rule, a a ow awt |wP = f(y) = ws + 1 gl = 5D =o tae U.22] Now the Schrédinger equation says that ow ih ew et Ot 7 lm ox al 1.23] and hence also (taking the complex conjugate of Equation 1.23) awe in ewe at 2m ax? "ih a th (aw tw a [ih (aw awe all a (¥ ae? Ox? ¥) =x [a (v oe ae) 28 The integral (Equation 1.21) can now be evaluated explicitly: a fr a) ih (.,8¥ AW \ i400 = sO =— + v . . 5 f We, OP dx = 5 (v ao Ge ye [1.26] yur, [1.24] so But W(x, f) must go to Zero as x goes to (+) infinity—otherwise the wave function would not be normalizable. It follows that +50 a [Wo DP dx =0, [1.27] and hence that the integral on the left is constant (independent of time); if W is normalized at = 0, it stays normalized for all future time. QEDSec. 1.4: Normalization 13 Problem 1.7 Attime ¢ = 0 a particle is represented by the wave function Ax/a, if0<2, E (we _ v) dx.2 {129 This expression can be simplified using integration by parts!®: d(x) ih f(y, d¥_ awe a lm I ax Ox ¥) ax. (30) [used the fact that 4x/x = 1, and threw away the boundary term, on the ground that W goes to zero at (-t) infinity.} Performing another integration by parts on the second term, we conclude that dix) ih [aw a | Mae (131) What are we to make of this result? Note that we're talking about the “velocity” of the expectation value of x, which is not the same thing as the velocity of the particle. Nothing we have seen so far would enable us to calculate the velocity of a particle— it’s not even clear what velocity means in quantum mechanics. If the particle doesn’t have a determinate position (prior to measurement), neither does it have a well-defined velocity. All we could reasonably ask for is the probability of getting a particular value, We'll see in Chapter 3 how to construct the probability density for velocity, °To keep things from getting too cluttered, I suppress the limits of integration when they are +00. 10The product rule says that af ax : d; d [ fe--[ geet fe Under the integral sign, then, you can peel a derivative off one factor in a product and slap it onto the other ‘one—it Il cost you a minus sign, and you'll pick up a boundary term. 4 yyy = 48 RO ASR from which it follows that16 Chap. 1 The Wave Function given W; for our present purposes it will suffice to postulate that the expectation value of the velocity is equal to the time derivative of the expectation value of position: d(x) v) = ——. 1.32 == [1.32] Equation 1.31 tells us, then, how to calculate (v) directly from W. ‘Actually, it is customary to work with momentum (p = mv), rather than ve- locity: of (xa ~in f (v =) dx, [1.33] Let me write the expressions for (x} and (p) in a more suggestive way: (x) = fv ewas, [1.34] 1.35] ‘We say that the operator'! x “represents” position, and the operator (/i)(0/8x) “represents” momentum, in quantum mechanics; to calculate expectation values, we “sandwich” the appropriate operator between W* and W, and integrate. That’s cute, but what about other dynamical variables? The fact is, a/l such quantities can be written in terms of position and momentum. Kinetic energy, for example, is and angular momentum is L=rxmv=rxp (the latter, of course, does not occur for motion in one dimension). To calculate the expectation value of such a quantity, we simply replace every p by (/i)(8/8x), insert the resulting operator between * and W, and integrate: ha | so p= f vou, twas. 1136) "1 An operator is an instruction to do something to the function that follows. The position operator tells you to multiply by x; the momentum operator tells you to differentiate with respect to x (and multiply the result by ~if). In this book all operators will be derivatives (d /dt, d? /dt?, # /axay, etc.) or multipliers (2, i, x?, etc.) or combinations of these.Sec. 1.6: The Uncertainty Principle 17 For example, - ay (1) = was [137] Equation 1.36 is a recipe for computing the expectation value of any dynamical quantity for a particle in state W; it subsumes Equations 1.34 and 1.35 as special cases. I have tried in this section to make Equation 1.36 seem plausible, given Born’s statistical interpretation, but the truth is that this equation represents such a radically new way of doing business (as compared with classical mechanics) that it’s a good idea to get some practice using it before we come back (in Chapter 3) and put it on a firmer theoretical foundation. In the meantime, if you prefer to think of it as an axiom, that’s fine with me. Problem 1.11 Why can’t you do integration by parts directly on the middle ex- pression in Equation 1.29—pull the time derivative over onto x, note that 0x/8? = 0, and conclude that d(x) /dt = 0? «Problem 1.12 Calculate d(p)/dt. Answer: dp) _ 8 dt (This is known as Ehrenfest’s theorem; it tells us that expectation values obey Newton's second law.) [1.38] Problem 1.13 Suppose you add aconstant /p to the potential energy (by “constant” I mean independent of x as well as ¢). In classical mechanics this doesn’t change anything, but what about quantum mechanics? Show that the wave function picks up a time-dependent phase factor: exp(—iVot /h). What effect does this have on the expectation value of a dynamical variable? 1.6 THE UNCERTAINTY PRINCIPLE Imagine that you're holding one end of a very long rope, and you generate a wave by shaking it up and down rhythmically (Figure 1.6). If someone asked you, “Precisely where is that wave?” you'd probably think he was a little bit nutty: ‘The wave isn’t precisely anywhere—it’s spread out over 50 feet or so. On the other hand, if he asked you what its wavelength is, you could give him a reasonable answer: It looks like about 6 feet. By contrast, if you gave the rope a sudden jerk (Figure 1.7), you'd get a relatively narrow bump traveling down the line. This time the first question (Where precisely is the wave?) is a sensible one, and the second (What is its wavelength?) seems nutty—it isn’t even vaguely periodic, so how can you assign a wavelength to it?18 Chap. 1 The Wave Function Figure 1.6: A wave with a (fairly) well-defined wavelength but an ill-defined position. Of course, you can draw intermediate cases, in which the wave is fairly well localized and the wavelength is fairly well defined, but there is an inescapable trade-off here: The more precise a wave’s position is, the less precise is its wavelength, and vice versa.” A theorem in Fourier analysis makes all this rigorous, but for the moment I am only concerned with the qualitative argument. This applies, of course, to any wave phenomenon, and hence in particular to the quantum mechanical wave function, Now the wavelength of W is related to the momentum of the particle by the de Broglie formula’*: h xh pss [1.39] Thus a spread in wavelength corresponds to a spread in momentum, and our general observation now says that the more precisely determined a particle’s position is, the less precisely its momentum is determined. Quantitatively, [1.40] where 0; is the standard deviation in x, and gp is the standard deviation in p. This is Heisenberg’s famous uncertainty principle. (We'll prove it in Chapter 3, but I wanted to mention it here so you can test it out on the examples in Chapter 2.) 10 20 30 40 50 x(feet) Figure 1.7: A wave with a (fairly) well-defined position but an ill-defined wave- length 127hat’s why a piccolo player must be right on pitch, whereas a double-bass player can afford to wear garden gloves. For the piccolo, a sixty-fourth note contains many full eycles, and the frequency (we're working in the time domain now, instead of space) is well defined, whereas for the bass, at a much lower register, the sixty-fourth note contains only a few cycles, and all you hear is a general sort of “oomph,” with no very clear pitch. 1371 prove this in due course. Many authors take the de Broglie formula as an axiom, from which they then deduce the association of momentum with the operator (fi/i)(3/8x). Although this is a conceptually cleaner approach, it involves diverting mathematical complications that I would rather save for later.Sec. 1.6: The Uncertainty Principle 19 Please understand what the uncertainty principle means: Like position mea- surements, momentum measurements yield precise answers—the “spread” here refers to the fact that measurements on identical systems do not yield consistent results. You can, if you want, prepare a system such that repeated position measurements will be very close together (by making W a localized “spike”), but you will pay a price: Mo- mentum measurements on this state will be widely scattered. Or you can prepare a system with a reproducible momentum (by making W a long sinusoidal wave), but in that case position measurements will be widely scattered. And, of course, if you're in a really bad mood you can prepare a system in which neither position nor momentum is well defined: Equation 1.40 is an inequality, and there’s no limit on how big o, and gp can be—just make W some long wiggly line with lots of bumps and potholes and no periodic structure. *Problem 1.14 A particle of mass m is in the state Wx, 2) = Aer alms? rsa, where A and a are positive real constants. (a) Find 4. (b) For what potential energy function V(x) does W satisfy the Schrédinger equa- tion? (c) Calculate the expectation values of x, x”, p, and p®. (d) Find 0, and op. Is their product consistent with the uncertainty principle?CHAPTER 2 THE TIME-INDEPENDENT SCHRODINGER EQUATION 2.1 STATIONARY STATES In Chapter 1 we talked a lot about the wave function and how you use it to calculate various quantities of interest. The time has come to stop procrastinating and confront what is, logically, the prior question: How do you get W(x, ¢) in the first place—how do you go about solving the Schrédinger equation? I shall assume for all of this chapter (and most of this book) that the potential,' V, is independent of t. In that case the Schrédinger equation can be solved by the method of separation of variables (the physicist’s first line of attack on any partial differential equation): We look for solutions that are simple products, Wx) = we) SO, 2.1] where w (lowercase) is a function of x alone, and f is a function of alone. On its face, this is an absurd restriction, and we cannot hope to get more than a tiny subset of all solutions in this way. But hang on, because the solutions we do obtain turn out to be of great interest. Moreover, as is typically the case with separation of variables, we will be able at the end to patch together the separable solutions in such a way as to construct the most general solution. ‘tis tiresome to keep saying “potential energy function,” so most people just call V the “potential”, even though this invites occasional confusion with electric potential, which is actually potential energy per unit charge.Sec. 2.1: Stationary States 21 For separable solutions we have av df aw dy ar" dt’ ax? ~ dx? (ordinary derivatives, now), and the Schrédinger equation (Equation 1.1) reads af _ Ray thy ST AVS. Or, dividing through by vf: ldf #1 n-2 --= “VY yy, "Fam y det! [2.2] Now the left side is a function of ¢ alone, and the right side is a function of x alone? The only way this can possibly be true is if both sides are in fact constant—otherwise, by varying , I could change the left side without touching the right side, and the two would no longer be equal. (That's a subtle but crucial argument, so if it’s new to you, be sure to pause and think it through.) For reasons that will appear in a moment, we shall call the separation constant E, Then or (23) and or (2.4) Separation of variables has turned a partial differential equation into two ordi- nary differential equations (Equations 2.3 and 2.4). The first of these is easy to solve (just multiply through by dt and integrate); the general solution is C exp(—iEt/h), but we might as well absorb the constant C into y (since the quantity of interest is the product yf). Then SQ = ete, (2.5) The second (Equation 2.4) is called the time-independent Schrédinger equation: we can go no further with it until the potential V (x) is specified. Note that this would nor be true if V were a function of ¢ as well as x.22 Chap. 2. The Time-Independent Schrédinger Equation The rest of this chapter will be devoted to solving the time-independent Schré- dinger equation, for a variety of simple potentials. But before we get to that I would like to consider further the question: What's so great about separable solutions? After all, most solutions to the (time-dependent) Schrédinger equation do not take the form w(x) f(0). [offer three answers—two of them physical and one mathematical: 1. They are stationary states. Although the wave function itself, Wr, = wae", [2.6] does (obviously) depend on f, the probability density [Wx OF = UW = pretE yet & py (xy? [2.7] does not—the time dependence cancels out.’ The same thing happens in calculating the expectation value of any dynamical variable; Equation 1.36 reduces to hd (QC, p)) -[v Qu, im” dx. [2.8] Every expectation value is constant in time; we might as well drop the factor f(t) altogether, and simply use y in place of V. (Indeed, it is common to refer to y as “the wave function”, but this is sloppy language that can be dangerous, and it is important to remember that the true wave function always carries that exponential time-dependent factor.) In particular, (x) is constant, and hence (Equation 1.33) (p) = 0. Nothing ever happens in a stationary state. 2. They are states of definite total energy. In classical mechanics, the total energy (kinetic plus potential) is called the Hamiltonian: 2 Hes, p= + VC). (2.9) The corresponding Hamiltonian operator, obtained by the canonical substitution p > (h/i)(8/0x), is therefore 2 9 Ta tO) [2.10] Thus the time-independent Schrédinger equation (Equation 2.4) can be written Ay = Ev, [2.11] 3For normalizable solutions, must be real (see Problem 2.1a). Whenever confusion might arise, I'll put a “hat” (*) on the operator to distinguish it from the dynamical variable it represents.Sec, 2.1: Stationary States 23. and the expectation value of the total energy is (y= [viivas=e [Pdr = [2.12] (Note that the normalization of Y entails the normalization of y.) Moreover, Wy = AAW) = WHEW) = EA) = By, and hence UP) = [vives =E fiveax =F. So the standard deviation in H is given by oj, = (H?) — (HP = EB - EB =0. [2.13] But remember, if 7 = 0, then every member of the sample must share the same value (the distribution has zero spread). Conclusion: A separable solution has the property that every measurement of the total energy is certain to return the value E. (That's why I chose that letter for the separation constant.) 3. The general solution is a linear combination of separable solutions. As we're about to discover, the time-independent Schrédinger equation (Equation 2.4) yields an infinite collection of solutions (Wi(x), Yo(x), Wax), ...), each with its associated value of the separation constant (E), E>, E3, ...); thus there is a different wave function for each allowed energy: Wie.) = Wi @e 4, War, 1) = va@eP",.. Now (as youcan easily check for yourself) the (time-dependent) Schrddinger equation (Equation 1.1) has the property that any linear combination’ of solutions is itself a solution. Once we have found the separable solutions, then, we can immediately construct a much more general solution, of the form Wo. = ye nat Vnlxye tt, (2.14) Itso happens that every solution to the (time-dependent) Schrédinger equation can be written in this form—it is simply a matter of finding the right constants (c), ¢2, ...) so as to fit the initial conditions for the problem at hand. You'll see in the following sections how all this works out in practice, and in Chapter 3 we'll put it into more elegant language, but the main point is this: Once you've solved the time-independent 5A linear combination of the functions f(z), fa(2), ... is an expression of the form afi) +erf where cy, €2,... are any (complex) constants +24 Chap. 2. The Time-Independent Schrédinger Equation Schrédinger equation, you’re essentially done; getting from there to the general so- ution of the time-dependent Schrédinger equation is simple and straightforward. «Problem 2.1 Prove the following theorems: (a) For normalizable solutions, the separation constant £ must be real. Hint: Write E (in Equation 2.6) as Ey + iT (with Eo and I real), and show that if Equation 1.20 is to hold for all ¢, F must be zero. (b) ¥ can always be taken to be real (unlike V, which is necessarily complex). Note: This doesn’t mean that every solution to the time-independent Schrédinger equation is teal; what it says is that if you’ ve got one that is not, it can always be expressed as a linear combination of solutions (with the same energy) that are. So in Equation 2.14 you might as well stick to W's that are real. Hint: If w(x) satisfies the time-independent Schrédinger equation for a given E, so too does its complex conjugate, and hence also the real linear combinations (y + ¥*) and i(y — w*)- (©) If V(x) is an even function [ie., V(—x) = V (x)], then W(x) can always be taken to be either even or odd. Hint: If w(x) satisfies the time-independent Schrédinger equation for a given E, so too does (—x), and hence also the even and odd linear combinations y(x) + w(—x). Problem 2.2 Show that £ must exceed the minimum value of V(x) for every normalizable solution to the time-independent Schrédinger equation, What is the classical analog to this statement? Hint: Rewrite Equation 2.4 in the form Pym oY = Sve - EW: ee = EO EW if EZ < Vain. then and its second derivative always have the same sign—argue that such a function cannot be normalized. 2.2 THE INFINITE SQUARE WELL Suppose ifO 0; we know from Problem 2.2 that E < 0 doesn’t work.) Equation 2.17 is the (classical) simple harmonic oscillator equation; the general solution is W(x) = Asinkx + Booskx, [2.18] where A and B are arbitrary constants. Typically, these constants are fixed by the boundary conditions of the problem. What are the appropriate boundary conditions for ¥(x)? Ordinarily, both y and dy/dx are continuous, but where the potential goes to infinity only the first of these applies. (I'll prove these boundary conditions, and account for the exception when V = oo, later on; for now I hope you will trust me.) Continuity of y(x) requires that WO) = va) =0, [2.19] So as to join onto the solution outside the well. What does this tell us about 4 and B? Well, ¥(0) = Asin0 + Beos0 = B, so B = 0, and hence W(x) = Asin kx. [2.20] Then y(a) = Asinka, so either A = 0 [in which case we're left with the trivial— nonnormalizable—solution y(x) = 0], or else sin ka = 0, which means that ka =0, +2, +27, +37,.... [2.21]26 Chap. 2. The Time-Independent Schrédinger Equation But k = 0 is no good [again, that would imply (x) = 0], and the negative solutions give nothing new, since sin(—@) = — sin(®) and we can absorb the minus sign into A. So the distinct solutions are ax withn = 1, 2, 3,.... [2.22] Curiously, the boundary condition at x = a does not determine the constant 4, but rather the constant k, and hence the possible values of E: wah? 2ma> [2.23] In sharp contrast to the classical case, a quantum particle in the infinite square well cannot have just any old energy—only these special allowed values. Well, how do we fix the constant A? Answer: We normalize yp: 2 =1, so [AP == a f [AP sin®ex) dx = [APS A 2 This only determines the magnitude of A, but it is simplest to pick the positive real root: A = J2/a (the phase of A carries no physical significance anyway). Inside the well, then, the solutions are >> Val) = joan (“). (2.24) | aa ‘As promised, the time-independent Schrédinger equation has delivered an infi- nite set of solutions, one for each integer n. The first few of these are plotted in Fig- ure 2.2; they look just like the standing waves on a string of length a. Yi, which car- ries the lowest energy, is called the ground state; the others, whose energies increase in proportion to n?, are called excited states. As a group, the functions q(x) have some interesting and important properties: 1. They are alternately even and odd, with respect to the center of the well. (his even, Yr is odd, Ys is even, and so on.*) 2. As you go up in energy, each successive state has one more node (zero crossing). yy has none (the end points don’t count), 2 has one, 3 has two, and so on. To make this symmetry more apparent, some authors center the well at the origin (so that it runs. from —a/2 to +a/2. The even functions are then cosines, and the odd ones are sines. See Problem 2.4.Sec. 2.2: The Infinite Square Well 27 4 wilX) wel) walX) a -% ax a ~ % Figure 2.2: The first three stationary states of the infinite square well (Equation 2.24), 3. They are mutually orthogonal, in the sense that f vmcorvatn dx =0, [2.25] whenever m ¥ n. Proof J vmtcorvatayar = 2 [sin (x) sin (Ex) a aa ea (ms )I =- cos ‘x | — cos mx )}| dx ahs a a 1 (m= 1 (mtn = }——— sin("—"xzx) - ——— sin mx lam ( a )- (m +a)r ( Me _ 1 f sinf(m —n)x] _ sin[(m +n) =o, ol] (ea) == 2 ta) ~ Note that this argument does nor work if m = n (can you spot the point at which it fails?); in that case normalization tells us that the integral is 1. In fact, we can combine orthogonality and normalization into a single statement’: Jv (x)" Yin (x) dx = bans [2.26] where 5,,,, (the so-called Kronecker delta) is defined in the usual way, _ 40, ifm #n; We say that the y's are orthonormal. 4. They are complete, in the sense that any other function, f(x), can be ex- pressed as a linear combination of them: fx) = eat -\?d sin (“Zx). [2.28] = ra 7In this case the y's are real, so the * on Ym is unnecessary, but for future purposes it’s a good idea to get in the habit of putting it ther.28 Chap. 2. The Time-Independent Schrédinger Equation I'm not about to prove the completeness of the functions /2/a sin(nx/a), but if you've studied advanced calculus you will recognize that Equation 2.28 is nothing but the Fourier series for f(x), and the fact that “any” function can be expanded in this way is sometimes called Dirichlet’s theorem.’ The expansion coefficients (cx) can be evaluated—for a given f(x)—by a method I call Fourier’s trick, which beautifully exploits the orthonormality of {y,): Multiply both sides of Equation 2.28 by Wn (x)*, and integrate. [ato feds = Yo f tnt valoda = extn = em (228) & a (Notice how the Kronecker delta kills every term in the sum except the one for which n =m.) Thus the mth coefficient in the expansion of f(x) is given by om = [ vont seas. [2.30] ‘These four properties are extremely powerful, and they are not peculiar to the infinite square well. The first is true whenever the potential itself is an even function; the second is universal, regardless of the shape of the potential.” Orthogonality is also quite general—T'll show you the proof in Chapter 3. Completeness holds for all the potentials you are likely to encounter, but the proofs tend to be nasty and laborious; I'm afraid most physicists simply assume completeness and hope for the best. The stationary states (Equation 2.6) for the infinite square well are evidently 2 W, (x,t) =f = sin (= +) eter h/matye [2.31] aa I claimed (Equation 2.14) that the most general solution to the (time-dependent) ‘Schrédinger equation is a linear combination of stationary states: Vo.D= Yen? sin (x) eterna (2.32) = If you doubt that this is a solution, by all means check it! It remains only for me to demonstrate that I can fit any prescribed initial wave function, (x, 0), by appropriate choice of the coefficients ¢,. According to Equation 2.32, W(x, 0) = Senne) 8See, for example, Mary Boas, Mathematical Methods in the Physical Sciences, 2nd ed. (New York: John Wiley & Sons, 1983), p. 313; f(x) can even have a finite number of finite discontinuities. See, for example, John L, Powell and Bernd Crasemann, Quantum Mechanics (Reading, MA: Addison-Wesley, 1961), p. 126.Sec. 2.2: The Infinite Square Well 29. The completeness of the y's (confirmed in this case by Dirichlet’s theorem) guarantees that I can always express (x, 0) in this way, and their orthonormality licenses the use of Fourier’s trick to determine the actual coefficients: = ef sin (=) Wa, 0dr. 1 (2.33) aly a That does it: Given the initial wave function, W(x, 0), we first compute the expansion coefficients c,,, using Equation 2.33, and then plug these into Equation 2.32 to obtain (x, 1). Armed with the wave function, we are in a position to compute any dynamical quantities of interest, using the procedures in Chapter 1. And this same ritual applies to any potential—the only things that change are the functional form of. the y's and the equation for the allowed energies. Problem 2.3 Show that there is no acceptable solution to the (time-independent) Schrédinger equation (for the infinite square well) with £ = 0 or E < 0. (This is a special case of the general theorem in Problem 2.2, but this time do it by explicitly solving the Schrédinger equation and showing that you cannot meet the boundary conditions.) Problem 2.4 Solve the time-independent Schrédinger equation with appropriate boundary conditions for an infinite square well centered at the origin [V (x) = 0, for —a/2 x ~ a/2, Problem 2.5 Calculate (x), (x7), (p), (p”), 0x, and o», for the nth stationary state of the infinite square well. Check that the uncertainty principle is satisfied. Which state comes closest to the uncertainty limit? x«Problem 2.6 A particle in the infinite square well has as its initial wave function an even mixture of the first two stationary states: W(x, 0) = Alvi(x) + Yo@)]- (a) Normalize W(x,0). (That is, find A. This is very easy if you exploit the orthonormality of yy and yz. Recall that, having normalized Y at 1 = 0, you can rest assured that it stays normalized—if you doubt this, check it explicitly after doing part b.) (b) Find Y(x, 6) and |Y(x, 0/2. (Express the latter in terms of sinusoidal functions of time, eliminating the exponentials with the help of Euler’s formula: e!? = cos6 +i sin9.) Let w = 2°h/2ma? (c) Compute (x). Notice that it oscillates in time, What is the frequency of the oscillation? What is the amplitude of the oscillation? (If your amplitude is greater than @/2, go directly to jail.)30 Chap. 2. The Time-Independent Schrédinger Equation (d) Compute (p). (As Peter Lorre would say, “Do it ze kveek vay, Johnny!”) (e) Find the expectation value of H. How does it compare with E; and £2? (f) A classical particle in this well would bounce back and forth between the walls. If its energy is equal to the expectation value you found in (e), what is the frequency of the classical motion? How does it compare with the quantum frequency you found in (c)? Problem 2.7 Although the overall phase constant of the wave function is of no physical significance (it cancels out whenever you calculate a measureable quantity), the relative phase of the expansion coefficients in Equation 2.14 does matter. For example, suppose we change the relative phase of yr, and y in Problem 2.6: Wx, 0) = Alyn) +e Ya@)], where ¢ is some constant. Find (x, f), |W(x, )/?, and (x), and compare your results with what you got before. Study the special cases @ = 1/2 and $ =z. +Problem 2.8 A particle in the infinite square well has the initial wave function W(x, 0) = Ax(a — x). (a) Normalize ¥(x,0). Graph it. Which stationary state does it most closely resemble? On that basis, estimate the expectation value of the energy. (b) Compute (x), (p), and (H), at ¢ = 0. (Note: This time you cannot get (p) by differentiating (x), because you only know (x) at one instant of time.) How does (H) compare with your estimate in (a)? «Problem 2.9 Find (x, f) for the initial wave function in Problem 2.8. Evaluate ci, ¢2, and c3 numerically, to five decimal places, and comment on these numbers, (cp tells you, roughly speaking, how much w, is “contained in” Y.) Suppose you measured the energy at time fo > 0, and got the value £3. Knowing that immediate repetition of the measurement must return the same value, what can you say about the coefficients c, after the measurement? (This is an example of the “collapse of the wave function”, which we discussed briefly in Chapter 1.) Problem 2.10 The wave function (Equation 2.14) has got to be normalized; given that the y,'s are orthonormal, what does this tell you about the coefficients c,? Answer: Vie =1 [2.34] nl (In particular, |c,|? is always < 1.) Show that oo (H) = 7 Enlenl*. (2.35] n=lSec. 2.3: The Harmonic Oscillator 31 Incidentally, it follows that (H) is constant in time, which is one manifestation of conservation of energy in quantum mechanics. 2.3. THE HARMONIC OSCILLATOR The paradigm for a classical harmonic oscillator is a mass m attached to a spring of force constant k. The motion is governed by Hooke’s law, ax F=-kx=m— dt (as always, we ignore friction), and the solution is x(t) = Asin(wt) + Beos(wt), ft [2.36] is the (angular) frequency of oscillation. The potential energy is where o V(xy= fk: [2.37] its graph is a parabola. Of course, there’s no such thing as a perfect simple harmonic oscillator—if you stretch it too far the spring is going to break, and typically Hooke’s law fails long before that point is reached. But practically any potential is approximately parabolic, in the neighborhood of a local minimum (Figure 2.3), Formally, if we expand V (x) in a Taylor series about the minimum: 1 V(x) = V0) + Vo) = x0) + 3 V"(x0)( = x0)? +o, subtract V (xp) [you can add a constant to V(x) with impunity, since that doesn’t change the force], recognize that V"(xo) = 0 (since xo is a minimum), and drop the higher-order terms [which are negligible as long as (x — xp) stays small], the potential becomes 1 Vox) & SV" (xo) = x0), which describes simple harmonic oscillation (about the point x9), with an effective spring constant k = "(x).!” That's why the simple harmonic oscillator is so important: Virtually any oscillatory motion is approximately simple harmonic, as Jong as the amplitude is small. ‘Note that ¥”"(xo) = 0, since by assumption xp is a minimum. Only in the rare case V" (xo) = 0 is the oscillation not even approximately simple harmonic.32 Chap. 2. The Time-Independent Schrédinger Equation vx) Figure 2.3: Parabolic approximation (dashed curve) to an arbitrary potential, in the neighborhood of a local minimum. The quantum problem is to solve the Sees equation for the potential Va@y= jm? x? [2.38] (it is customary to eliminate the spring constant in favor of the classical frequency, using Equation 2.36). As we have seen, it suffices to solve the time-independent Schrédinger equation: = +s smaty = Ey. [2.39] In the literature you will find two entirely different approaches to this problem. The first is a straighforward “brute force” solution to the differential equation, using the method of power series expansion; it has the virtue that the same strategy can be applied to many other potentials (in fact, we'll use it in Chapter 4 to treat the Coulomb potential). The second is a diabolically clever algebraic technique, using so-called ladder operators. I'll show you the algebraic method first, because it is quicker and simpler (and more fun); if you want to skip the analytic method for now, that’s fine, but you should certainly plan to study it at some stage. 2.3.1 Algebraic Method To begin with, let's rewrite Equation 2.39 in a more suggestive form: 1[/ad)* al Im (Gz) + (max) \ = Ey. [2.40] The idea is to factor the term in square brackets. If these were numbers, it would be easy: w+? = (uw —iv)(u + iv).Sec. 2.3: The Harmonic Oscillator 33 Here, however, it’s not quite so simple, because u and v are operators, and operators do not, in general, commute (wv is not the same as vu). Still, this does invite us to take a look at the expressions hd = 4 (25 Simos). (241) as, 2m What is their product, a_a,.? Warning: Operators can be slippery to work with in the abstract, and you are bound to make mistakes unless you give them a “test function”, f(x), to act on. At the end you can throw away the test function, and you'll be left with an equation involving the operators alone. In the present case, we have 1 (hd hd (@a,) f(x) = 5 Ga ~imor) (Ga + imax) f(x) ) (?¢ +imoxf) 7 Ls pmo Lief) - trmane E+ (ms)? | _il(nd > = oy (Gz) + (mwx) +imo] f@) [used d(xf)/dx = x(df/dx) + f inthe last step.] Discarding the test function, we conclude that > [jaa ajo! a = = sho. 2.42] aa, = 5 [¢#) + mon] + (2.42] Evidently Equation 2.40 does not factor perfectly—there’s an extra term (1/2)hw. However, if we pull this over to the other side, the Schrédinger equation" becomes (aa, — phony = EW. (2.43] Notice that the ordering of the factors a, and a_ is important here; the same argument, with a., on the left, yields L[/aa\* 1 a4. = 5 (a) + mon - phew. [2.44] a_a, —a,a_ =ho, (2.45] ‘1m getting tired of writing “time-independent Schridinger equation,” so when it’s clear from the context which one I mean, I'll just call it the Schrédinger equation.34 Chap. 2. The Time-Independent Schrédinger Equation and the Schrédinger equation can also be written 1 (asa. + sho) = Ey. [2.46] Now, here comes the crucial step: I claim that if y satisfies the Schrodinger equation, with energy E, then a satisfies the Schrodinger equation with energy (E +he). Proof: 1 1 (aya_+ zhovarw) = (aya_a, + qhoas =a,(a_a, + Stone =a,[(a_a, — drow +hoy] =a,(EW thoy) = (E +hw)(a.). QED [Notice that whereas the ordering of a, and a_ does matter, the ordering of ax. and any constants (such as h, w, and E) does not.] By the same token, a_y is a solution with energy (E — fw): (a_a, — Sraylaw) =a_(a,a_— dow =a_[(a,a_+ show — how) =a(Ey —ha) = (E —ho)(a_y). Here, then, is a wonderful machine for grinding out new solutions, with higher and lower energies—if we can just find one solution, to get started! We call a, ladder operators, because they allow us to climb up and down in energy; a, is called the raising operator, and a_ the lowering operator. The “ladder” of states is illustrated in Figure 2.4, But wait! What if I apply the lowering operator repeatedly? Eventually I'm going to reach a state with energy less than zero, which (according to the general theorem in Problem 2.2) does not exist! At some point the machine must fail. How can that happen? We know that a_y is a new solution to the Schrédinger equation, but there is no guarantee that it will be normalizable—it might be zero, or its square integral might be infinite. Problem 2.11 rules out the latter possibility. Conclusion: There must occur a “lowest rung” (let's call it yo) such that (2.47) That is to say,: The Harmonic Oscillator 35 EsSho ay E+2ho ay Esto ay E v E-ho ay E-2hw ay & Yo Figure 2.4: The ladder of stationary states for the simple harmonic oscillator. or This differential equation for Yo is easy to solve: di mo mo Vo me fxdx = Inyo =- 22x? + constant, Vo h 2h so . ox) = Age. (2.48] To determine the energy of this state, we plug it into the Schrédinger equation (in the form of Equation 2.46), (a,.a_ + (1/2)hi@) Wo = Eowo, and exploit the fact that a_p = 0. Evidently 1 Ey= hw. [2.49] With our foot now securely planted on the bottom rung" (the ground state of the quantum oscillator), we simply apply the raising operator to generate the excited states": Yn) = An(ayye“F", with Ey = (n+ Do. [2.50] !2Note that there can only be one ladder, because the lowest state is uniquely determined by Equation 2.47. Thus we have in fact obtained all the (normalizable) solutions. "3p the case of the harmonic oscillator, it is convenient to depart from our usual custom and number the states starting with n = 0 instead of n = 1. Obviously, the lower limit on the sum in equations such as Equation 2.14 should be altered accordingly.36 Chap. 2. The Time-Independent Schrédinger Equation (This method does not immediately determine the normalization factor A,; I'll let you work that out for yourself in Problem 2.12.) For example, Wi = Alage FP = A, ! ¢ 4 + imo) ore = Sel! (See oY simone ~ Vim Li Oh , which simplifies to Wi(x) = ((AjoV2m) xe" #*” [2.51] 1 wouldn’t want to calculate Wso in this way, but never mind: We have found all the allowed energies, and in principle we have determined the stationary states—the rest is just computation, Problem 2.11 Show that the lowering operator cannot generate a state of infinite norm (ic., f |a-wl’dx < oo, if p itself is a normalized solution to the Schrédinger equation). What does this tell you in the case y = Yo? Hint: Use integration by Parts to show that [ewrendr= freeware. Then invoke the Schrédinger equation (Equation 2.46) to obtain f \a_vPdx = E-— Fhe, where E is the energy of the state y. 4xProblem 2.12 (a) The raising and lowering operators generate new solutions to the Schrédinger equation, but these new solutions are not correctly normalized. Thus a, Wy is proportional to Yn41, and a_y, is proportional to w,—, but we'd like to know the precise proportionality constants. Use integration by parts and the Schrédinger equation (Equations 2.43 and 2.46) to show that . ~ fi tertatde = 00+ Dho, [7 lady! de = he E fs 0 and hence (with i’s to keep the wavefunctions real) ayn =i VO + DRO Vas, (2.52) ay = —iVihO Wn. [2.53]Sec. 2.3: The Harmonic Oscillator 37 (b) Use Equation 2.52 to determine the normalization constant A, in Equation 2.50. (You'll have to normalize Yo “by hand”.) Answer: A =("2)" i" [2.54] "Nan ) Jalioy” . «Problem 2.13 Using the methods and results of this section, Normalize y (Equation 2.51) by direct integration. Check your answer against the general formula (Equation 2.54). (b) Find yz, but don’t bother to normalize it. (©) Sketch Yo, Wi, and Yo (d) Check the orthogonality of yo. Wi, and y2. Note: If you exploit the evenness and oddness of the functions, there is really only one integral left to evaluate explicitly. a) «Problem 2.14 Using the results of Problems 2.12 and 2.13, (a) Compute (x), (p), (x2), and (p?), for the states Yo and yy. Note: In this and most problems involving the harmonic oscillator, it simplifies the notation if you introduce the variable § = ./m@/h x and the constant « = (mw/rh)'4. (b) Check the uncertainty principle for these states. (©) Compute (7) and (V) for these states (no new integration allowed!). Is their sum what you would expect? 2.3.2 Analytic Method We return now to the Schrédinger equation for the harmonic oscillator (Equa- tion 2.39): -—s + dnote = Ey. 2m dx? 2 : Things look a little cleaner if we introduce the dimensionless variable mo 7 [2.55] in terms of &, the Schrédinger equation reads @& 2 i ( — Ky, {2.56] dg?38 Chap. 2. The Time-Independent Schrodinger Equation where K is the energy, in units of (1/2)ha: _2E KaT. [2.57] Our problem is to solve Equation 2.56, and in the process obtain the “allowed” values of K (and hence of £). To begin with, note that at very large & (which is to say, at very large x), &? completely dominates over the constant K, so in this regime a 2 a wey, [2.58] which has the approximate solution (check it!) We) © Ae #2 + Bet, [2.59] The B term is clearly not normalizable (it blows up as |x| > 00); the physically acceptable solutions, then, have the asymptotic form we) > (de®, at large &. [2.60] This suggests that we “peel off” the exponential part, HE) = hGeF?, (2.61) in hopes that what remains [4(E)] has a simpler functional form than y(€) itself." Differentiating Equation 2.61, we have dy = ( - +) eee de \ade and ay (@h dh a _ (Gh a gdh ga “PA ge = (Fe BBETE = Here, so the Schrédinger equation (Equation 2.56) becomes Ph. dh oo 224 (K-Ih=0. 62] ge gg tK—DA=0 [2.62] 1 propose to look for a solution to Equation 2.62 in the form of a power series ing": NE) = a9 + a1§ + an6? + 2 yiae!, 2.63] i '4Note that although we invoked some approximations to motivate Equation 2.61, what follows is exact. The device of sttipping off the asymptotic behavior is the standard first step in the power series method for solving differential equations—see, for example, Boas (cited in footnote 8), Chapter 12. 15 According to Taylor’s theorem, any reasonably well-behaved function can be expr jas power series, so Equation 2.63 involves no real loss of generality. For conditions on the applicability of the series method, see Boas (cited in footnote 8) or George Ariken, Mathematical Methods for Physicists, 3rd ed. (Orlando, FL: Academic Press, 1985), Section 8.5.Sec, 2.3: The Harmonic Oscillator 39 Differentiating the series term by term, dh & = ay t Qarg + 3as6? ++ = Do jaye, =) dé and @h 2 C dei 2 +2 Bas +3 dase? +--+ = OU + DG + Dajy28? ia Putting these into Equation 2.62, we find SG + DG + Dajz2 — 2a + (K ~ Day] 6! = 0. [2.64] = It follows (from the uniqueness of power series expansions') that the coefficient of each power of & must vanish, G+ DU + 2aj42 = 2jaj + (K — Nay and hence that ay a GItI=®), Pe GEDG+D This recursion formula is entirely equivalent to the Schrédinger equation itself. Given ag it enables us (in principle) to generate a2, as, a6, . .. and given a) itgenerates 43,45, a7,.... Let us write [2.65] AE) = Reven(&) + AoaalE), [2.66] where even G) = ao + 428? + 0664 ++ is an even function of & (since it involves only even powers), built on ao, and hoaa(G) = a8 + 038? + a5 +> is an odd function, built on a). Thus Equation 2.65 determines #(&) in terms of two arbitrary constants (ay and a;)—which is just what we would expect, for a second- order differential equation. However, not all the solutions so obtained are normalizable. For at very large j. the recursion formula becomes (approximately) a mei 16See, for example, Ariken (footnote 15), Section 5.7,40 Chap. 2. The Time-Independent Schrodinger Equation with the (approximate) solution ~ © oO” GP for some constant C, and this yields (at large £, where the higher powers dominate) 1 1 2 x ae Cy tet a coh hE) XC > cae Cc me Ce Now, if h goes like exp(é?), then y (remember y?—that’s what we're trying to calcu- late) goes like exp(€?/2) (Equation 2.61), which is precisely the asymptotic behavior we don’t want.'” There is only one way to wiggle out of this: For normalizable solu- tions the power series must terminate. There must occur some “highest” j (call it n) such that the recursion formula spits out a,.2 = 0 (this will truncate either the series even Or the series hoa; the other one must be zero from the start). For physically acceptable solutions, then, we must have K=2n+1, for some positive integer , which is to say (referring to Equation 2.57) that the energy must be of the form E, = (n+ dhe, forn =0,1,2,.... [2.67] Thus we recover, by a completely different method, the fundamental quantization condition we found algebraically in Equation 2.50. For the allowed values of K, the recursion formula reads =An=j)_, G+DG+) 7 If n = 0, there is only one term in the series (we must pick a = 0 to kill haa, and J = 0 in Equation 2.68 yields ay = [2.68] a2 = ho(§) = a0, and hence: “7/2 Wo(5) (which reproduces Equation 2.48). For n = 1 we pick ay = 0,"* and Equation 2.68 with j = 1 yields a3 = 0, so ae Ay) =a, '7Ir’s no surprise that the ill-behaved solutions are still contained in Equation 2.65; this recursion relation is equivalent to the Schrédinger equation, so it’s got to include both the asymptotic forms we found in Equation 2.59. 'SNote that there is a completely different set of coefficients a; for each value of n.Sec. 2.3: The Harmonic Oscillator 41 and hence WG) =agee? (confirming Equation 2.51). For n = 2, j = 0 yields ay = —2a, and j = 2 gives dy = 0, so hy(&) = ao(1 — 267) and ; rl) = aod — 26*)e°F?, and so on. (Compare Problem 2.13, where the same result was obtained by algebraic means.) In general, hy (€) will be a polynomial of degree n in &, involving even powers only, if n is an even integer, and odd powers only, ifn is an odd integer. Apart from the overall factor (ao or ay) they are the so-called Hermite polynomials, H, ().”” The first few of them are listed in Table 2.1. By tradition, the arbitrary multiplicative factor is chosen so that the coefficient of the highest power of § is 2”. With this convention, the normalized” stationary states for the harmonic oscillator are v4 (22) 1 n,@eF?. (2.69) n(x) They are identical (of course) to the ones we obtained algebraically in Equation 2.50. In Figure 2.5a I have plotted Y(x) for the first few n’s. ‘The quantum oscillator is strikingly different from its classical counterpart— not only are the energies quantized, but the position distributions have some bizarre features. For instance, the probability of finding the particle outside the classically allowed range (that is, with x greater than the classical amplitude for the energy in question) is not zero (see Problem 2.15), and in all odd states the probability of Table 2.1: The first few Hermite polynomials, H,(x). Ho =1, Mm =2r, Hp = 4x7 -2, Hy = 8x3 ~ 12%, Ha = 16x4 — 48x? + 12, Hs = 32x5 — 16023 + 120x The Hermite polynomials have been studied extensively in the mathematical literature, and there are many tools and tricks for working with them, A few of these are explored in Problem 2.18. 25] shall not work out the normalization constant here; if you are interested in knowing how it is done, see, for example, Leonard Schiff, Quantum Mechanics, 3rd ed. (New York: MeGraw-Hill, 1968), Section 13,42 Chap. 2. The Time-Independent Schrodinger Equation WolX) wilx) x x wel) 4 Wal) lrool? 0.24 7 | = « Figure 2.5: (a) The first four stationary states of the harmonic oscillator. (b) Graph of |y100/?, with the classical distribution (dashed curve) superimposed.Sec. 2.3: The Harmonic Oscillator 43 finding the particle at the center of the potential well is zero. Only at relatively large n do we begin to see some resemblance to the classical case. In Figure 2.5b I have superimposed the classical position distribution on the quantum one (for n = 100); if you smoothed out the bumps in the latter, the two would fit pretty well (however, in the classical case we are talking about the distribution of positions over time for one oscillator, whereas in the quantum case we are talking about the distribution over an ensemble of identically-prepared systems).”" Problem 2.15 Inthe ground state of the harmonic oscillator, what is the probability (correct to three significant digits) of finding the particle outside the classically allowed region? Hint: Look ina math table under “Normal Distribution” or “Error Functior Problem 2.16 Use the recursion formula (Equation 2.68) to work out Hs(&) and Ho). «Problem 2.17 A particle in the harmonic oscillator potential has the initial wave function W(x, 0) = Alvo(x) + We) for some constant 4. (a) Normalize ¥(x, 0). (b) Find W(x, 1) and |W(x, HP. (c) Find the expectation value of x as a function of time. Notice that it oscillates sinusoidally, What is the amplitude of the oscillation? What is its (angular) frequency? (d) Use your result in (c) to determine (p). Check that Ehrenfest’s theorem holds for this wave function. (e) Referring to Figure 2.5, sketch the graph of || at 1 = 0, x/w, 2x /w, 3/w, and 42r/w. (Your graphs don’t have to be fancy—just a rough picture to show the oscillation.) «Problem 2.18 In this problem we explore some of the more useful theorems (stated without proof) involving Hermite polynomials. (a) The Rodrigues formula states that H,(&) = (-1)"e' [2.70] Use it to derive Hs and Hy. 2\The analogy is perhaps more telling if you interpret the classical distribution as an ensemble of oscillators all with the same energy, but with random starting times.44 Chap. 2. The Time-Independent Schrodinger Equation (b) The following recursion relation gives you H,,,1 in terms of the two preceding Hermite polynomials: Hn.) = 2 Hn (E) — 20 Hy (6). 2.70 Use it, together with your answer to (a), to obtain Hs and Hg. (c) If you differentiate an nth-order polynomial, you get a polynomial of order (n — 1). For the Hermite polynomials, in fact, dH, = 2nHyx(€). 2.72 de in Hn_\(E) [2.72] Check this, by differentiating Hs and Hg. (d) H,) is the nth z-derivative, at z = 0, of the generating function exp(—z? + 2z€); or, to put it another way, it is the coefficient of z"/n! in the Taylor series expansion for this function: ent = = HG). (2.731 n=0 7 Use this to rederive Ho, Hi, and Hy. 2.4 THE FREE PARTICLE We tum next to what should have been the simplest case of all: the free particle IV) = O everywhere}. As you'll see in a moment, the free particle is in fact a surprisingly subtle and tricky example. The time-independent Schrédinger equation reads [2.74] or [2.75] So far, it’s the same as inside the infinite square well (Equation 2.17), where the potential is also zero; this time, however, I prefer to write the gencral solution in exponential form (instead of sines and cosines) for reasons that will appear in due course: W(x) = Ae + Bet, [2.76] Unlike the infinite square well, there are no boundary conditions to restrict the possible values of k (and hence of E); the free particle can carry any (positive) energy. Tacking on the standard time dependence, exp(—i Et /h), Wx t) = Ae 4 Be OTE, (2.771Sec. 2.4: The Free Particle 45 Now, any function of x and f that depends on these variables in the special combination (xv#) (for some constant v) represents a wave of fixed profile, traveling in the -Fx-direction, at speed v. A fixed point on the waveform (for example, a ‘maximum or a minimum) corresponds to a fixed value of the argument, and hence to x and ¢ such that xt vi = constant, or x = uf +constant, which is the formula for motion in the x-direction with constant speed v. Since every point on the waveform is moving along with the same velocity, its shape doesn’t change as it propagates. Thus the first term in Equation 2.77 represents a wave traveling to the right, and the second term represents a wave (of the same energy) going to the left. By the way, since they only differ by the sign in front of k, we might as well write ; Wy(x, 1) = dele, [2.78] and let k run negative to cover the case of waves traveling to the left: VimE k>0=> traveling to the right, keto with 1, 0-5 traveling to the left. (2.79) The speed of these waves (the coefficient of f over the coefficient of x) is Alki om [2.80] Yguantum = On the other hand, the classical speed of a free particle with energy E is given by E = (1/2)mv? (pure kinetic, since V = 0), so ‘3 Evidently the quantum mechanical wave function travels at half the speed of the particle it is supposed to represent! We'll return to this paradox in a moment—there is an even more serious problem we need to confront first: This wave function is not normalizable! For 2euaneum- (2.81] Vetassical f WEY, dx = iar f 1dx = |A|?(0o). [2.82] Inthe case of the free particle, then, the separable solutions do not represent physically realizable states, A free particle cannot exist in a stationary state; or, to put it another way, there is no such thing as a free particle with a definite energy. But that doesn’t mean the separable solutions are of no use to us. For they play a mathematical role that is entirely independent of their physical interpretation: The general solution to the time-dependent Schridinger equation is still a linear46 Chap. 2. The Time-Independent Schrodinger Equation combination of separable solutions (only this time it’s an integral over the continuous variable k, instead of a sum over the discrete index ): Tz fo oapemte dk. [2.83] oo [The quantity 1/./2z is factored out for convenience; what plays the role of the coefficient c,, in Equation 2.14 is the combination (1//2r)(k) dk.] Now this wave function can be normalized [for appropriate #(k)]. But it necessarily carries a range of k’s, and hence a range of energies and speeds. We call it a wave packet. In the generic quantum problem, we are given W(x, 0), and we are to find W(x, 1). For a free particle the solution has the form of Equation 2.83; the only remaining question is how to determine ¢(k) so as to fit the initial wave function: W(x, 0) = m [ode dk. [2.84] 00 This is a classic problem in Fourier analysis; the answer is provided by Plancherel’s theorem (see Problem 2.20): f@)= f rae dk > FD = f pee dx. | (285) Vix Vix F(k) iscalled the Fourier transform of f(x); f(x) is the inverse Fourier transform of F (4) (the only difference is in the sign of the exponent). There is, of course, some restriction on the allowable functions: The integrals have to exist.” For our purposes this is guaranteed by the physical requirement that W(x, 0) itself be normalized. So the solution to the generic quantum problem, for the free particle, is Equation 2.83, with 1 opt V2E Jo ok) = W(x, Oe dx. [2.86] I'd love to work out an example for you—starting with a specific function W(x, 0) for which we could actually calculate @(K), and then doing the integral in Equation 2.83 to obtain W(x, ¢) in closed form, Unfortunately, manageable cases are hard to 2 The necessary and sufficient condition on f(x) is that f. | f(x)|?dx be finite. (In that case [2.\F Pdi is also finite, and in fact the two integrals are equal.) See Arfken (footnote 15), Sec- tion 15.5.Sec. 2.4: The Free Particle 47 come by, and I want to save the best example for you to work out yourself. Be sure, therefore, to study Problem 2.22 with particular care. T return now to the paradox noted earlier—the fact that the separable solution u(x, £) travels at the “wrong” speed for the particle it ostensibly represents. Strictly speaking, the problem evaporated when we discovered that Yj is not a physically achievable state. Nevertheless, it is of interest to discover how information about the particle velocity is carried by the wave function (Equation 2.83). The essential idea is this: A wave packet is a sinusoidal function whose amplitude is modulated by @ (Figure 2.6); it consists of “ripples” contained within an “envelope.” What corresponds to the particle velocity is not the speed of the individual ripples (the so- called phase velocity), but rather the speed of the envelope (the group velocity)— which, depending on the nature of the waves, can be greater than, less than, or equal to the velocity of the ripples that go to make it up. For waves on a string, the group velocity is the same as the phase velocity. For water waves it is one half the phase velocity, as you may have noticed when you toss a rock into a pond: If you concentrate on a particular ripple, you will see it build up from the rear, move forward through the group, and fade away at the front, while the group as a whole propagates out at half the speed. What I need to show is that for the wave function of a free particle in quantum mechanics the group velocity is twice the phase velocity—just right to represent the classical particle speed. The problem, then, is to determine the group velocity of a wave packet with the general form 400 Ya = Fe | o (ke dk, [In our case «w = (ik?/2m), but what I have to say now applies to any kind of wave packet, regardless of its dispersion relation—the formula for « as a function of k.] Let us assume that #(&) is narrowly peaked about some particular value ko. [There is nothing illegal about a broad spread in k, but such wave packets change shape rapidly (since different components travel at different speeds), so the whole notion of a “group.” with a well-defined velocity, loses its meaning. Since the integrand Figure 2.6: A wave packet. The “envelope” travels at the group velocity; the “ripples” travel at the phase velocity,48 — Chap. 2. The Time-Independent Schrodinger Equation is negligible except in the vicinity of ko, we may as well Taylor-expand the function co(k) about that point and keep only the leading terms: ak) = wp + og(k — ko), where 4 is the derivative of w with respect to k, at the point ko. Changing variables from k to s = k — ko, to center the integral at ko, we have +00 xt ky + sellbotss— Comte] ds, Wax,t 0" Tab Att =0, 1 pte vo.0 = f ko + spelor ds, ¢ Viz bw tko + 8) and at later times 1 “oft : won) ei -entthoaie) f lho + syeio#9O-269 as, v2 00 Except for the shift from x to (x — wf), the integral is the same as the one in ¥(x, 0). Thus : W(x, ft) & e Hor bor Wx — wht, 0). [2.87] Apart from the phase factor in front (which won't affect ||? in any event), the wave packet evidently moves along at a speed dw Yaron = Gp [2.88] (evaluated at k = ko), which is to be contrasted with the ordinary phase velocity o Monae = Fs [2.89] In our case, w = (hk?/2m), so o/k = (ik/2m), whereas dw/dk = (hk/m), which is twice as great. This confirms that it is the group velocity of the wave packet, not the phase velocity of the stationary states, that matches the classical particle velocity: Vctassical = Vgroup = 2Uphasc- [2.90] Problem 2.19 Show that the expressions [4e'* + Be], [C coskx + Dsin kx], LF cos(kx+a)], and [G sin(kx+ B)] are equivalent ways of writing the same function of x, and determine the constants C, D, F,G,a, and f in terms of A and B. (In quantum mechanics, with V = 0, the exponentials give rise to fraveling waves, and are most convenient in discussing the free particle, whereas sines and cosinesSec. 2.4: The Free Particle 49 correspond to standing waves, which arise naturally in the case of the infinite square well.) Assume the function is real. «Problem 2.20 This problem is designed to guide you through a “proof” of Plan- cherel’s theorem, by starting with the theory of ordinary Fourier series on a finite interval, and allowing that interval to expand to infinity. (a) Dirichlet’s theorem says that “any” function f(x) on the interval [—a, +a] can be expanded as a Fourier series: £0) = ) Tay sin(nsrx/a) + by costnzx/a). n=0 ‘Show that this can be written equivalently as foy= So cette What is cy, in terms of ay and by? (b) Show (by appropriate modification of Fourier’s trick) that 1 pe a= x/ flae "le dy 2a Ja (©) Eliminate n and ¢, in favor of the new variables k = (n/a) and F(k) = 27m acy. Show that (a) and (b) now become + Saye dx, f@)= Fie Ak; Fk) = 1 2 20 pay where Af is the increment in k from one 7 to the next. (d) Take the limit a + 00 to obtain Plancherel’s theorem. Note: In view of their quite different origins, it is surprising (and delightful) that the two formulas [one for F(X) in terms of f(x), the other for f(x) in terms of F(k)] have such a similar structure in the limit a > oo. Problem 2.21 Suppose a free particle, which is initially localized in the range —a < x ) = ah?, but it may take some algebra to reduce it to this simple form. (@) Does the uncertainty principle hold? At what time ¢ does the system come closest to the uncertainty limit? 2.5 THE DELTA-FUNCTION POTENTIAL ‘We have encountered two very different kinds of solutions to the time-independent Schrédinger equation: For the infinite square well and the harmonic oscillator they are normalizable, and labeled by a discrete index n; for the free paticle they are non-normalizable, and labeled by a continuous variable k. The former represent physically realizable states in their own right, the latter do not; but in both cases the general solution to the time-dependent Schrédinger equation is a linear combina- tion of stationary states—for the first type this combination takes the form of a sumSec. 2.5: The Delta-Function Potential 51 (over n), whereas for the second it is an integral (over k). What is the physical significance of this distinction? In classical mechanics a one-dimensional time-independent potential can give rise to two rather different kinds of motion. If V (x) rises higher than the particle's to- tal energy (E) on either side (Figure 2.7a), then the particle is “stuck” in the potential well—itrocks back and forth between the turning points, but it cannot escape (unless, Ciassical turning points” (a) I ra Classical turning point Classical turning points ©) Figure 2.7: (a) A bound state. (b) Scattering states. (c) A classical bound state, but a quantum scattering state.52 Chap. 2. The Time-Independent Schrédinger Equation of course, you provide it with a source of extra energy, such as a motor, but we're not talking about that). We call this a bound state. If, on the other hand, £ exceeds V(x) on one side (or both), then the particle comes in from “infinity”, slows down or speeds up under the influence of the potential, and returns to infinity (Figure 2.7b). (It can’t get trapped in the potential unless there is some mechanism, such as friction, to dissipate energy, but again, we're not talking about that.) We call this a scattering state. Some potentials admit only bound states (for instance, the harmonic oscillator); some allow only scattering states (a potential hill with no dips in it, for example); some permit both kinds, depending on the energy of the particle. As you have probably guessed, the two kinds of solutions to the Schrédinger equation correspond precisely to bound and scattering states. The distinction is even cleaner in the quantum domain, because the phenomenon of tunneling (which we'll come to shortly) allows the particle to “leak” through any finite potential barrier, so the only thing that matters is the potential at infinity (Figure 2.7c): \é < V(—o0) and V(+oo) = bound state, pol E > V(-oo) or V(+00) => scattering state. In “real life” most potentials go to zero at infinity, in which case the criterion simplifies even further: | E <0 = bound state, (2.92) E>0O = scattering state. Because the infinite square well and harmonic oscillator potentials go to infinity as x —> 00, they admit bound states only; because the free particle potential is zero everywhere, it only allows scattering states.” In this section (and the following one) we Shall explore potentials that give rise to both kinds of states. The Dirac delta function, 5(x), is defined informally as follows: ; oe seo ={ % Zo fe wn f B(x) dx = ifx=0 [2.93] It is an infinitely high, infinitesimally narrow spike at the origin, whose area is 1 (Figure 2.8). Technically, it’s not a function at all, since it is not finite at x = 0 (mathematicians call it a generalized function, or distribution). Nevertheless, it is an extremely useful construct in theoretical physics. (For example, in electrodynam- ics the charge density of a point charge is a delta function.) Notice that 8(x —a) would 231 you are very observant, and awfully fastidious, you may have noticed that the general theorem requiring E > Vin (Problem 2.2) does not really apply to scattering states, since they are not normalizable anyway, If this bothers you, try solving the Schrédinger equation with £ < 0, for the free particle, and note that even linear combinations of these solutions cannot be normalized. The positive energy solutions by themselves constitute a complete set. 4The delta function can be thought of as the limit of a sequence of functions, such as rectangles (or triangles) of ever-increasing height and ever-decreasing width.Sec. 2.5: The Delta-Function Potential 53 80x) Figure 2.8: The Dirac delta function * Equation 2.93). be a spike of area 1 at the point a. If you multiply 8(x — a) by an ordinary function F(x), it’s the same as multiplying by f(a): F()8(x — a) = f(a)8(x — a), [2.94] because the product is zero anyway except at the point a. In particular, +90 +00 [ S(X)8(x — a) dx = fof 8(x-a)dx = f(a). [2.95] That's the most important property of the delta function: Under the integral sign it serves to “pick out” the value of f(x) at the point a. (Of course, the integral need not go from —oo to +00; all that matters is that the domain of integration include the point a, so a ~ € toa + € would do, for any € > 0.) Let’s consider a potential of the form V(x) = —a8(x), [2.96] where a is some constant. This is an artificial potential (so was the infinite square well), but it’s beautifully simple and in some respects closer to reality than any of the potentials we have considered so far. The Schridinger equation reads Bee 2m dx? This potential yields both bound states (E < 0) and scattering states (E> 0); we'll ook first at the bound states. In the region x < 0, V(x) =0, so ay ax? —ad(x)y = Ev. (2.97] [2.98] where [2.99]54 Chap. 2. The Time-Independent Schrodinger Equation (E is negative, by assumption, so « is real and positive.) The general solution to Equation 2.98 is W(x) = Ae + Bet, [2.100] but the first term blows up as x > —oo, so we must choose 4 = W(x) = Be, (x <0) [2.101] In the region x > 0, V(x) is again zero, and the general solution is of the form Fexp(—«x)+ G exp(kx); this time it’s the second term that blows up (as x > +00), so Wx) = Fe, (x > 0). [2.102] It remains only to stitch these two functions together, using the appropriate boundary conditions at x = 0. I quoted earlier the standard boundary conditions for Ww: 2. dis/dx is continuous except at points [2.103] | 1. w is always continuous, and where the potential is infinite. In this case the first boundary condition tells us that F = B, so Be“, (x <0), v@)= en, OS 5. (2.104) [W(x) is plotted in Figure 2.9.] The second boundary condition tells us nothing; this is (like the infinite square well) the exceptional case where V is infinite at the join, and it’s clear from the graph that this function has a kink at x = 0. Moreover, up to this point the delta function has not come into the story at all. Evidently the delta function must determine the discontinuity in the derivative of yy, at x = 0. I'll show you now how this works, and as a byproduct we'll see why dy/dx is ordinarily continuous. The idea is to integrate the Schrddinger equation, from —€ to +€, and then take the limit as € > 0: vx) x Figure 2.9: Bound state wave function for the delta function potential (Equation 2.104).Sec. 2.5: The Delta-Function Potential 55. opt ay 2m Je “dx? +e +e dx +f V(x)w(x)dx = ef w(x)dx. [2.105] The first integral is nothing but dy/dx, evaluated at the two end points; the last integral is zero, in the limit ¢ > 0, since it’s the area of a sliver with vanishing width and finite height. Thus +e A (4) = Frm V(x)Wx) dx. [2.106] Ordinarily, the limit on the right is again zero, and hence dy//dx is continuous. But when V(x) is infinite at the boundary, that argument fails. In particular, if V(x) = —a4(x), Equation 2.95 yields dy) 2ma a(2)-- vO [2.107] For the case at hand (Equation 2.104), dw/dx = —Bee**, for (x > 0), sodw/dx|, = —Be, d/dx =+Bket*, for (x <0), sody/dx|_=+Bk, and hence A(d/dx) = —2Bx. And y/(0) = B. So Equation 2.107 says ma =>. 2.108 7 [2.108] and the allowed energy (Equation 2.99) is We? ma? '- =f [2.109] Finally, we normalize y: $20 ; ~ Be IW(x)P dx = 2B] eM dx = = 1, _ A K so (choosing, for convenience, the positive real root): BaVve=v™. [2.110] h Evidently the delta-function well, regardless of its “strength” a, has exactly one bound state: (2.111)56 Chap. 2 The Time-Independent Schrédinger Equation What about scattering states, with E > 0? For x < 0 the Schrddinger equation reads Py _ mE, Ry, where [2.112] is real and positive. The general solution is w(x) = Ae + Be“, [2.113] and this time we cannot rule out either term, since neither of them blows up. Similarly, forx > 0, w(x) = Fel + Ge, [2.114] The continuity of y(x) at x = 0 requires that F+G=A+B. [2.115] The derivatives are dy/dx = ik (Fe — Ge“), for(x > 0), sody/dx|_, = ik(F - G), dy/dx = ik (Ae — Be), for (x <0), sody/dx|_ = ik(A — B), and hence A(dy/dx) = ik(F — G — A+ B). Meanwhile, y(0) = (A + B), so the second boundary condition (Equation 2.107) says 2 ik(F -G-A+B)=- ard +B), [2.116] or, more compactly, FG = A(1+2if)— BO —2iB), where B Having imposed the boundary conditions, we are left with two equations (Equa- tions 2.115 and 2.117) in four unknowns (A, B, F, and G)—five, if you count k. Nor- malization won’t help—this isn’t a normalizable state. Perhaps we'd better pause, then, and examine the physical significance of these various constants. Recall that exp(ikx) gives rise [when coupled with the time-dependent factor exp(—i Et /h)] to a wave function propagating to the right, and exp(—ikx) leads to a wave propagating to the left, It follows that A (in Equation 2.113) is the amplitude of a wave coming in from the left, B is the amplitude of a wave returning to the left, F (in Equation 2.114) is the amplitude of a wave traveling off to the right, and G is the amplitude of a wave coming in from the right (Figure 2.10). In a typical scattering experiment particles are fired in from one direction—let’s say, from the left. In that case the amplitude of the wave coming in from the right will be zero:Sec. 2.5: The Delta-Function Potential 57 Figure 2.10: Scattering from a delta-function well. G =0 (for scattering from the left). [2.118] A is then the amplitude of the incident wave, B is the amplitude of the reflected wave, and F is the amplitude of the transmitted wave. Solving Equations 2,115 and 2.117 for B and F, we find B= [2.119] =p (If you want to study scattering from the right, set A = 0; then G is the incident amplitude, F is the reflected amplitude, and B is the transmitted amplitude.) Now, the probability of finding the particle at a specified location is given by |W/?, so the relative®’ probability that an incident particle will be reflected back is BP _ BP \4P 1+ 6? Ris called the reflection coefficient. (If you have a beam of particles, it tells you the ‘fraction of the incoming number that will bounce back.) Meanwhile, the probability of transmission is given by the transmission coefficient 2.120} ver 4 (2.121) \aP > 1B " Of course, the sum of these probabilities should be 1—and it is: R4+T=1. (2.122) Notice that R and T are functions of 8, and hence (Equations 2.112 and 2.117) of E: 1 1 R=—_,_, T= —_____.. 2.123] 1+ (WE/ma?) 1+ (ma? /2h? E) t 1 25This is not a normalizable wave function, so the absolute probability of finding the particle at a particular location is not well defined; nevertheless, the ratio of probabilities for two different locations is ‘meaningful, More on this in the next paragraph,58 Chap. 2. The Time-Independent Schrédinger Equation The higher the energy, the greater the probability of transmission (which seems rea- sonable). This is all very tidy, but there is a sticky matter of principle that we cannot al- together ignore: These scattering wave functions are not normalizable, so they don’t actually represent possible particle states. But we know what the resolution to this problem is; We must form normalizable linear combinations of the stationary states, just as we did for the free particle—true physical particles are represented by the resulting wave packets. Though straightforward in principle, this is a messy busi- ness in practice, and at this point it is best to turn the problem over to a computer.”° Meanwhile, since it is impossible to create a normalizable free particle wave function without involving a range of energies, R and T should be interpreted as the approxi- ‘mate reflection and transmission probabilities for particles in a narrow energy range about E. Incidentally, it might strike you as peculiar that we were able to analyse a quintessentially time-dependent problem (particle comes in, scatters off a potential, and flies off to infinity) using stationary states. After all, y (in Equations 2.113 and 2.114) is simply a complex, time-independent, sinusoidal function, extending (with constant amplitude) to infinity in both directions. And yet, by imposing appropriate boundary conditions on this function, we were able to determine the probability that a particle (represented by a localized wave packet) would bounce off, or pass through, the potential, The mathematical miracle behind this is, I suppose, the fact that by taking linear combinations of states spread over alll space, and with essentially triv- ial time dependence, we can construct wave functions that are concentrated about a (moving) point, with quite elaborate behavior in time (see Problem 2.40). As long as we've got the relevant equations on the table, let’s look briefly at the case of a delta-function barrier (Figure 2.11). Formally, all we have to do is change the sign of a. This kills the bound state, of course (see Problem 2.2). On the other hand, the reflection and transmission coefficients, which depend only on @, are unchanged. Strange to say, the particle is just as likely to pass through the barrier as to cross over the well! Classically, of course, the particle could not make itoveran infinitely high barrier, regardless of its energy. In fact, the classical scattering V(x) = a(x) * Figure 2.11: The delta-function barrier. 26There exist some powerful programs for analysing the scattering of a wave packet from a one- dimensional potential; see, for instance, A. Goldberg, H. M. Schey, and J. L. Schwartz, Am. J. Phys. 35, 177 (1967)Sec. 2.5: The Delta-Function Potential 59 problem is pretty dull: If E > Vinx, then T = 1 and R = 0—the particle certainly makes it over; conversely, if E < Vax, then T = 0 and R = 1—it rides “up the hill” until it runs out of energy, and then returns the same way it came. The quantum scattering problem is much richer; the particle has some nonzero probability of passing through the potential even if E < Vinax. We call this phenomenon tunneling; it is the mechanism that makes possible much of modern electronics—not to mention spectacular recent advances in microscopy. Conversely, even if E > Vix, there is a possibility that the particle will bounce back—though I wouldn't advise driving off a cliff in the expectation that quantum mechanics will save you (see Problem 2.41). Problem 2.23 Evaluate the following integrals: (a) [303 - 3x? 42x -— DS + Ddx (b) f5°[cos(3x) + 218(x — 3) dx (©) ft} exp(lx| + 3)5(x — 2) dx. Problem 2.24 Two expressions [D;(x) and D2(x)] involving delta functions are said to be equal if i - f fepnar= [ pondainar, for any (ordinary) function f(x). (a) Show that 8(cx) = aie, [2.124] c where c is a real constant. (b) Let 6(x) be the step function: _fl, ifx>0, 9) = {o: ifx <0. GY {In the rare case where it actually matters, we define 6(0) to be 1/2.] Show that dO /dx = 8(x). «Problem 2.25 What is the Fourier transform of 5(x)? Using Plancherel’s theorem, show that [2.126] Comment: This formula gives any respectable mathematician apoplexy. Although the integral is clearly infinite when x = 0, it doesn’t converge (to zero or anything else) when x + 0, since the integrand oscillates forever. There are ways to patch it up60 Chap. 2. The Time-Independent Schrodinger Equation (for instance, you can integrate from —L to +L, and interpret the integral in Equation 2.126 to mean the average value of the finite integral, as L > 00). The source of the problem is that the delta function doesn’t meet the requirement (square integrability) for Plancherel’s theorem (see footnote 22). In spite of this, Equation 2.126 can be extremely useful, if handled with care. «Problem 2.26 Consider the double delta-function potential V(x) = —af5(x +a) +4(x -a)], where a and a are positive constants. (a) Sketch this potential. (b) How many bound states does it possess? Find the allowed energies, for a = fh? /ma and for « = h?/4ma, and sketch the wave functions. **Problem 2.27 Find the transmission coefficient for the potential in Problem 2.26. 2.6 THE FINITE SQUARE WELL As a last example, consider the finite square well —V, for—a a, [en where Vo is a (positive) constant (Figure 2.12). Like the delta-function well, the finite square well admits both bound states (with E < 0) and scattering states (with E > 0). We'll look first at the bound states. In the region x < —a the potential is zero, so the Schrédinger equation reads fv St TEM or Figure 2.12; The finite square well (Equation 2.127).Sec. 2.6: The Finite Square Well 61 where [2.128] is real and positive. The general solution is yr(x) = Aexp(—Kx) + Bexp(cx), but the first term blows up (as x —* —co), so the physically admissable solution (as before—see Equation 2.101) is W(x) = Be, for (x < -a) [2.129] In the region —a < x < a, V (x) = —Vo, and the Schrodinger equation reads dy ay _ op am det VOW = BY or = where l= [2.130] h Although E is negative, for a bound state, it must be greater than —Vo, by the old theorem E > Vin (Problem 2.2); s0 J is also real and positive. The general solution is W(x) = Csin(lx) + Deos(Ix), for (-a a the potential is again zero; the general solution is (x) = F exp(—«x) + G exp(kx), but the second term blows up (as x —> 00), so we are left with Wa) = Fe, for (x >a). [2.132] ‘The next step is to impose boundary conditions: y and dy/dx continuous at —a and +a. But we can save a little time by noting that this potential is an even function, so we can assume with no loss of generality that the solutions are either even or odd (Problem 2.1c). The advantage of this is that we need only impose the boundary conditions on one side (say, at +a); the other side is then automatic, since w(—x) = £¥(x). I'll work out the even solutions; you get to do the odd ones in Problem 2.28. The cosine is even (and the sine is odd), so I'm looking for solutions of the form Fe, for (x > a), W(x) = 4 Deos(ix), for (0 00; however, for any finite Vo there are only finitely many bound states. 2. Shallow, narrow well. As zo decreases, there are fewer and fewer bound states, until finally (for z) < 2/2, where the lowest odd state disappears) only one remains. It is interesting to note, however, that there is always one bound state, no matter how “weak” the well becomes, You're welcome to normalize yy (Equation 2.133), if you're interested (see Problem 2.29), but I'm going to move on now to the scattering states (E > 0). To the left, where V(x) = 0, we have En + Vo = [2.139] w(x) = Ae™ + Be, for (x < —a), {2.140] where (as usual) (2.1411 Inside the well, where V(x) = —Vo, w(x) =Csin(Ix) + Doos(Ix), for (-a 0. Determine the bound-state energy for the delta-function potential, by treating it as the limit of a finite square well. Check that your answer is consistent with Equation 2.111. Also show that Equation 2.151 reduces to Equation 2.123 in the appropriate limit. «Problem 2.31 Derive Equations 2.149 and 2.150. Hint: Use Equations 2.147 and 2.148 to solve for C and D in terms of F: C= [since + if costa] ep, D= [oor - sina | el F, Plug these back into Equations 2.145 and 2.146. Obtain the transmission coeffi- cient, and confirm Egation 2.151. Work out the reflection coefficient, and check that T+R=1L +#Problem 2.32 Determine the transmission coefficient for a rectangular barrier (same as Equation 2.127, only with +V in the region —a < x < a). Treat separately the three cases E < Vo, E = Vo, and E > Vo (note that the wave function inside the barrier is different in the three cases). Partial answer: For E < Vo,” ve 2a ra 4 i sin (4 =D). '+ tem —5) G m(Vo — E) 4*Problem 2.33 Consider the step function potential: 0, ifx <0, 7@)= | Vo, ifx >0. (a) Calculate the reflection coefficient, for the case E < Vo, and comment on the answer. © (c) For a potential such as this that does not go back to zero to the right of the barrier, the transmission coefficient is not simply |F|?/|A|?, with 4 the incident amplitude and F the transmitted amplitude, because the transmitted wave travels ata different speed. Show that Calculate the reflection coefficient for the case E > Vo. [E=Vo FP T . Vo@ (4p [2.154] for E > Vo. Hint: Youcan figure it out using Equation 2.81, or—more elegantly, but less informatively—from the probability current (Problem 1.9a). What is T for E < Vo? 28This is a good example of tunneling—classically the particle would bounce back66 Chap. 2. The Time-Independent Schrédinger Equation (d) For E > Vo, calculate the transmission coefficient for the step potential, and check that T+ R= 1, 2.7 THE SCATTERING MATRIX The theory of scattering generalizes in a pretty obvious way to arbitrary localized potentials (Figure 2.15). To the left (Region I), V (x) = 0, so V2mE W(x) = Ae + Be, where k= . [2.155] To the right (Region III), V(x) is again zero, so wr) = Fe'™ + Get, [2.156] In between (Region Il), of course, I can't tell you what y is until you specify the potential, but because the Schrédinger equation is a linear, second-order differential equation, the general solution has got to be of the form W(x) = Cf(x) + Dg(x), (2.157] where f(x) and g(x) are any two linearly independent particular solutions.” There will be four boundary conditions (two joining Regions I and Il, and two joining Regions II and III). Two of these can be used to eliminate C and D, and the other two can be “solved” for B and F in terms of A and G: B=Sj\A+SpG, F S214 + SG. [2.158] The four coefficients S;;, which depend on k (and hence on £), constitute a 2 x 2 matrix Vix) a Region I Region II Region IIT Figure 2.15: Scattering from an arbitrary localized potential {V (x) = 0 except in Region II). °See any book on differential equations—for example, J. L. Van Iwaarden, Ordinary Differential Equations with Numerical Techniques (San Diego, CA: Harcourt Brace Jovanovich, 1985). Chapter 3.Sec. 2.7: The Scattering Matrix 67 Su Si (3 =) [2.159] called the scattering matrix (or S-matrix, for short). The S-matrix tells you the outgoing amplitudes (B and F) in terms of the incoming amplitudes (4 and G): (7)-+(8) a In the typical case of scattering from the left, G = 0, so the reflection and transmission coefficients are Ss [2.161] For scattering from the right, A = 0, and LFe 2 = = |S»! 162] iG laco = 2 [2.162] ‘The S-matrix tells you everything there is to know about scattering from a local- ized potential. Surprisingly, it also contains (albeit in a concealed form) information about the bound states (if there are any). For if E < 0, then y(x) has the form Be (Region 1, Wx) = | Cf(e) + Dex) (Region 1D, [2.163] Fe* (Region IID, with k 2m [2.164] h The boundary conditions are the same as before, so the S-matrix has the same structure—only now £ is negative, so k > ix, But this time A and G are nec- essarily zero, whereas B and F are not, and hence (Equation 2.158) at least two elements in the S-matrix must be infinite. To put it the other way around, if you’ve got the S-matrix (for E > 0), and you want to locate the bound states, put in k —> ix, and look for energies at which the S-matrix blows up. For example, in the case of the finite square well, ka ~ cos(2la) — (822K + Py Su (Equation 2.150). Substituting k > ix, we see that 5), blows up whenever cot(2la) = ———68 Chap. 2 The Time-Independent Schrédinger Equation Using the trigonometric identity tan (3) = £V 14 cot) — cots, we obtain tan(la) = i (plus sign), and cot(la) = -5 (minus sign). These are precisely the conditions for bound states of the finite square well (Equation 2.136 and Problem 2.28). «Problem 2.34 Construct the S-matrix for scattering from a delta-function well (Equation 2.96). Use it to obtain the bound state energy, and check your answer against Equation 2.111 Problem 2.35 Find the S-matrix for the finite square well (Equation 2.127). Hint: This requires no new work if you carefully exploit the symmetry of the problem. FURTHER PROBLEMS FOR CHAPTER 2 Problem 2.36 A particle in the infinite square well (Equation 2.15) has the initial wave function W(x, 0) = Asin’ (1x/a). Find (x) as a function of time. «Problem 2.37 Find (x), (p), (x2), (p?), (7), and (V(x)) for the nth stationary state of the harmonic oscillator. Check that the uncertainty principle is satisfied. Hint: Express x and (h/i)(d/dx) in terms of (a, +a_), and use Equations 2.52 and 2.53; you may assume that the states are orthogonal. Problem 2.38 Find the allowed energies of the half-harmonic oscillator va= { (1/2)mo*x?, for (x > 0), 00, for (x < 0). (This represents, for example, a spring that can be stretched, but not compressed.) Hint: This requires some careful thought, but very little actual computation. +«Problem 2.39 Solve the time-independent Schrédinger equation for an infinite square well with a delta-function barrier at the center: a(x), for(-a 0 and a + oo. «Problem 2.40 In Problem 2.22 you analyzed the stationary Gaussian free particle wave packet. Now solve the same problem for the traveling Gaussian wave packet, starting with the initial wave function W(x, 0) = Aen" el, where / is a real constant. Problem 2.41 A particle of mass m and kinetic energy E > 0 approaches an abrupt potential drop Vo (Figure 2.16). (a) What is the probability that it will “reflect” back, if E = Vo/3? (b) I drew the figure so as to make you think of a car approaching a cliff, but obviously the probability of “bouncing back” from the edge of a cliff is far smaller than what you got in (a)—unless you're Bugs Bunny. Explain why this potential does not correctly represent a cliff. Problem 2.42 If two (or more) distinct” solutions to the (time-independent) Schrd- dinger equation have the same energy E, these states are said to be degenerate. For example, the free particle states are doubly degenerate—one solution representing motion to the right, and the other motion to the left. But we have encountered no V(x) Figure 2.16: Scattering from a “cliff” (Problem 2.41). 20]f the two solutions differ only by a multiplicative constant (so that, once normalized, they differ ‘only by a phase factor e'*), they represent the same physical state, and in this case they are not distinct solutions. Technically, by “distinct” I mean “linearly independent.”70 Chap. 2. The Time-Independent Schrodinger Equation normalizable degenerate solutions, and this is not an accident. Prove the following theorem: In one dimension’! there are no degenerate bound states. Hint: Suppose there are two solutions, y; and yz, with the same energy E. Multiply the Schrédinger equation for y by we, and the Schrédinger equation for yy by yr, and subtract, to show that (wd /dx — yidya/dx) is a constant. Use the fact that for normalizable solutions y — 0 at +00 to demonstrate that this constant is in fact zero. Conclude that 2 is a multiple of 1, and hence that the two solutions are not distinct. Problem 2.43 Imagine a bead of mass m that slides frictionlessly around a circular wire ring of circumference a. [This is just like a free particle, except that y(x) = w(x +).] Find the stationary states (with appropriate normalization) and the cor- responding allowed energies. Note that there are fvo independent solutions for each energy £,—corresponding to clockwise and counterclockwise circulation; call them wrt (x) and ys, (x). How do you account for this degeneracy, in view of the theorem in Problem 2.42—that is, why does the theorem fail in this case? x«Problem 2.44 (Avention: This is a strictly qualitative problem—no calculations allowed!) Consider the “double square well” potential (Figure 2.17). Suppose the depth Vo and the width a are fixed, and great enough so that several bound states occur, (a) Sketch the ground-state wave function yy and the first excited state Wo, (i) for the case b = 0, (ii) for b © a, and (iii) for b >> a Qualitatively, how do the corresponding energies (E1 and E) vary, as b goes from 0 to 00? Sketch £y(b) and E(b) on the same graph. Ct) icy The double well is a very primitive one-dimensional model for the potential experienced by an electron in a diatomic molecule (the two wells represent the attractive force of the nuclei). If the nuclei are free to move, they will adopt the configuration of minimum energy. In view of your conclusions in (b), does the electron tend to draw the nuclei together, or push them apart? (Of course, there is also the internuclear repulsion to consider, but that’s a separate problem.) xxxProblem 2.45 (a) Show that mo\'\"* oO Wx) = (=) exp [- 32 (« *"In higher dimensions such degeneracy is very common, as we shall see in Chapter 4. Assume that the potential does not consist of isolated pieces separated by regions where V' = oo—two isolated infinite square wells, for instance, would give rise to degenerate bound states, for which the particle is either in the ‘one of in the other. » ih a per tiaty 4 Ht ~ axe") | 2 mFurther Problems for Chapter 2.71 Figure 2.17: The double square well (Problem 2.44) satisfies the time-dependent Schrédinger equation for the harmonic oscillator potential (Equation 2.38). Here a is any real constant (with the dimensions of lengthy.” (b) Find |W (x, 1)/?, and describe the motion of the wave packet. (c) Compute (x) and (p), and check that Ehrenfest’s theorem (Equation 1.38) is satisfied. Problem 2.46 Consider the potential ox, ifx <0, VO)" |i a), x= 0, where a and a are positive real constants with the appropriate units (see Figure 2.18). A particle starts out in the “well” (0 < x < a), but because of tunneling its wave function gradually “leaks” out through the delta-function barrier. (a) Solve the (time-independent) Schrédinger equation for this potential; impose appropriate boundary conditions, and determine the “energy”, Z. (An implicit equation will do.) (b) I put the word “energy” in quotes because you'll notice that it is a complex number! How do you account for this, in view of the theorem you proved in Problem 2.1a? (c) Writing E = £o + if (with Eo and I real), calculate (in terms of I") the characteristic time it takes the particle to leak out of the well (that is, the time it takes before the probability is | /e that it’s still in the region 0 < x —1 as z > —o0, W(x) © Ae™, for large negative x. This represents, then, a wave coming in from the left with no accompanying reflected wave [i.e., no term exp(—ikx)]. What is the asymptotic form of Yx(x) at large positive x? What are R and T for this potential? Note: sech? is a famous example of a “reflectionless” potential—every incident particle, regardless of its energy, passes right through. See R. E. Crandall and B. R. Litt, Annals of Physics 146, 458 (1983). Construct the S-matrix for this potential, and use it to locate the bound states, How many of them are there? What are their energies? Check that your answer is consistent with part (a). iC x*Problem 2.49 The S-matrix tells you the outgoing amplitudes (B and F) in terms of the incoming amplitudes (A and G): B\_ (Si S2\(A FF) \S1 Sv} \G)> For some purposes it is more convenient to work with the transfer matrix, M, which gives you the amplitudes to the right of the potential (F and G) in terms of those to the left (A and B): F\_(Mu Mn\(A G}]~ \Mn Mnj}\B)" (a) Find the four elements of the M-matrix in terms of the elements of the S-matrix, and vice versa. Express Ri, 7), R,,and T, (Equations 2,161 and 2.162) in terms of elements of the M-matrix. (b) Suppose you have a potential consisting of two isolated pieces (Figure 2.19). Show that the M-matrix for the combination is the product of the two M-matrices for each section separately: M = MoM). (This obviously generalizes to any number of pieces, and accounts for the use- fulness of the M-matrix.) —— v=0 v=0 v=0 Figure 2.19: A potential consisting of two isolated pieces (Problem 2.49).74 Chap. 2. The Time-Independent Schrodinger Equation (©) Construct the M-matrix for scattering from a single delta-function potential at point a: V(x) = —a8(x —a). (d) By the method of part (b), find the Mf-matrix for scattering from the double delta function V(x) = -a[8(x +4) + 5(x — a). What is the transmission coefficient for this potential?CHAPTER 3 FORMALISM 3.1 LINEAR ALGEBRA ‘The purpose of this chapter is to develop the formalism of quantum mechanics— terminology, notation, and mathematical background that illuminate the structure of the theory, facilitate practical calculations, and motivate a fundamental extension of the statistical interpretation. I begin with a brief survey of linear algebra.’ Linear algebra abstracts and generalizes the arithmetic of ordinary vectors, as we encounter them in first-year physics. The generalization is in two directions: (1) We allow the scalars to be complex, and (2) we do not restrict ourselves to three dimensions (indeed, in Section 3.2 we shall be working with vectors that live in spaces of infinite dimension). 3.1.1 Vectors A vector space consists of a set of vectors (\a), |8), |v), ...). together with a set of scalars (a, b, c, ...),? which are subject to two operations—vector addition and scalar multiplication: *1f you have already studied Linear algebra, you should be able to skim this section quickly, but I wouldn't skip it altogether, because some of the notation may be unfamiliar. If, on the other hand, this ‘material is new to you, be warned that I am only summarizing (often without proof) those aspects of the theory we will be needing later. For details, you should refer to atext on linear algebra, such as the classic by P.R. Halmos: Finite Dimensionat Vector Spaces, 2nd ed, (Princeton, NJ: van Nostrand, 1958) >For our purposes, the scalars will be ordinary complex numbers, Mathematicians can tell you about vector spaces over more exotic fields, but such objects play no role in quantum mechanics. 7576 Chap. 3 Formalism Vector addition. The “sum” of any two vectors is another vector: la) + 1B) = ly). [3.1] Vector addition is commutative la) + |B) = |B) + le), [3.2] and associative la) + (1B) + |y)) = (lee) + 1B)) + ly). [3.3] ‘There exists a zero (or null) vector,* |0), with the property that la) + 0) = |e), [3.4] for every vector ja). And for every vector |a) there is an associated inverse vector (| —a)), such that Je) +|—a) =|0). 3.5] Scalar multiplication. The “product” of any scalar with any vector is another vector: ala) = ly) (3.6] Scalar multiplication is distributive with respect to vector addition a(|a) + |B)) = ala) + |B) (3.7] and with respect to scalar addition (a + bya) = ala) + bla). 13.8] It is also associative with respect to the ordinary multiplication of scalars: a(bla)) = (ab)|a). 3.9] Multiplication by the scalars 0 and 1 has the effect you would expect: Olor) = 0); 1 Jer) = Jer). [3.10] Evidently | — a) = (—1lar). There’s a lot less here than meets the eye—all I have done is to write down in abstract language the familiar rules for manipulating vectors. The virtue of such abstraction is that we will be able to apply our knowledge and intuition about the behavior of ordinary vectors to other systems that happen to share the same formal properties. It is customary, where no confusion can arise, to write the null vector without the adorning bracket: 10) +0.Sec. 3.1: Linear Algebra 77 A linear combination of the vectors |), |8), |), ... is an expression of the form ala) + b|B) +cly) +++. 13.11) A vector |A) is said to be linearly independent of the set |), |), |v)... if it cannot be written as a linear combination of them. (For example, in three dimensions the unit vector f is linearly independent of ? and j, but any vector in the xy-plane is linearly dependent oni and j.) By extension, a set of vectors is linearly independent if each one is linearly independent of all the rest. A collection of vectors is said to span the space if every vector can be written as a linear combination of the members of this set.‘ A set of linearly independent vectors that spans the space is called a basis. The number of vectors in any basis is called the dimension of the space. For the moment we shall assume that the dimension (n) is finite, With respect to a prescribed basis lei), lea). «+4 len), (3.12) any given vector |e) =ayler) + asler) + +++ + aqlen) [3.13] is uniquely represented by the (ordered) n-tuple of its components: la) < (a1, @2,....4, [3.14] Its often easier to work with the components than with the abstract vectors them- selves. To add vectors, you add their corresponding components: fox) + 1B) <> (ar + bi, a2 + 2,24 +g): (3.15] to multiply by a scalar you multiply each component: cla) + (car, car, .... Cay); [3.16] the null vector is represented by a string of zeroes: 0) + (0,0,...,0): [3.17] and the components of the inverse vector have their signs reversed: Ja) © (a1, —ax,...,-a,). [3.18] The only disadvantage of working with components is that you have to commit your- self to a particular basis, and the same manipulations will look very different to someone working in a different basis. Problem 3.1 Consider the ordinary vectors in three dimensions (a,7 + a,j +a-k) with complex components. A set of vectors that spans the space is also called complete, though I personally reserve that word for the infinite-dimensional case, where subtle questions of convergence arise.78 Chap. 3 Formalism (a) Does the subset of all vectors with a, is its dimension; if not, why not? = O constitute a vector space? If so, what (b) What about the subset of all vectors whose z component is 1? (c) How about the subset of vectors whose components are all equal? «Problem 3.2 Consider the collection of all polynomials (with complex coefficients) of degree < N in x. Does this set constitute a vector space (with the polynomials as “vectors”)? If 50, suggest a convenient basis, and give the dimension of the space. If not, which of the defining properties does it lack? (a) (b) What if we require that the polynomials be even functions? (c) What if we require that the leading coefficient (ie., the number multiplying xl) be 1? (d) What if we require that the polynomials have the value 0 at x = 1? (@) What if we require that the polynomials have the value | at x = 0? Problem 3.3 Prove that the components of a vector with respect to a given basis are unique. 3.1.2 Inner Products In three dimensions we encounter two kinds of vector products: the dot product and the cross product. The latter does not generalize in any natural way to n-dimensional vector spaces, but the former does—in this context it is usually called the inner product. The inner product of two vectors (|) and |)) is a complex number (which we write as («|8)), with the following properties: (Blor) = (orlB)*, [3.19] (aja) >0, and (ala) = 0 & |a) = |0), [3.20] (eel (BIB) + cly)) = b(@ip) + caly). (3.21] Apart from the generalization to complex numbers, these axioms simply codify the familiar behavior of dot products. A vector space with an inner product is called an inner product space. Because the inner product of any vector with itself is a nonnegative number (Equation 3.20), its square root is rea/—we call this the norm of the vector: vy (arlor); [3.22]Sec. 3.1: Linear Algebra 79 it generalizes the notion of “length”. A “unit” vector, whose norm is 1, is said to be normalized (the word should really be “normal”, but I guess that sounds too anthropomorphic). Two vectors whose inner product is zero are called orthogonal (generalizing the notion of “perpendicular”). A collection of mutually orthogonal normalized vectors, (a;|a)) = 8,;. 13.23] is called an orthonormal set. It is always possible (see Problem 3.4), and almost always convenient, to choose an orthonormal basis; in that case the inner product of two vectors can be written very neatly in terms of their components: (oelB) = afb) + ajbp +--» +a%bn, [3.24] the norm (squared) becomes {ax|az) = Jay)? + Jag? +--+ + lal?, (3.25) and the components themselves are a; = (e;\a). [3.26] (These results generalize the familiar formulas a - b =a} +a? + a2, and a, orthonormal basis 7, j, £.) From now on we shall always work in orthonormal bases unless it is explicitly indicated otherwise. Another geometrical quantity one might wish to generalize is the angle between two vectors. In ordinary vector analysis cos @ = (a-b)/la||b]. But because the inner product is in general a complex number, the analogous formula (in an arbitrary inner product space) does not define a (real) angle 6. Nevertheless, itis still true that the absolute value of this quantity is a number no greater than 1, ayby + dyby + azbz, a+ a - a, for the three-dimensional “aay = faa, = {(ae|B)|° < (ele) (B |B). (3.27) (This important result is known as the Schwarz inequality; the proof is given in Problem 3.5.) So you can, if you like, define the angle between |a) and |B) by the formula (ae|B) (Ble) cos 8 = (3.28) (ala) (B|B) *Problem 3.4 Suppose you start out with a basis (|e;), [e2), ..., |@n)) that is nor orthonormal. The Gram-Schmidt procedure is a systematic ritual for generating from it an orthonormal basis (le), |e)), ... , |@,)). It goes like this: (i) Normalize the first basis vector (divide by its norm): lei) = lier”80 Chap. 3 Formalism (ii) Find the projection of the second vector along the first, and subtract it off: le2) = (ele2)1e1)- This vector is orthogonal to |e); normalize it to get |e). (iii) Subtract from |e3) its projections along |e) and |e’): les) ~ (e{les)le\) — (eales)le2)- This is orthogonal to |e{) and |e); normalize it to get |e,). And so on. Use the Gram-Schmidt procedure to orthonormalize the three-space basis Jer) = ++ I+ OK, ler) = OF+)I+ MA, les) = OF+ 28)F+ OK. Problem 3.5 Prove the Schwarz inequality (Equation 3.27). Hint: Let |y) = 1B) — ((ex|B)/{oele))|@), and use (y|y) > 0. Problem 3.6 Find the angle (in the sense of Equation 3.28) between the vectors Ja) = (1 +8)? + (7 + Wk and |B) = 4 — D7 + OJ + 2 —2i)k. Problem 3.7 Prove the triangle inequality: ||({a) + |8))|l < lll] + IIBIl- 3.1.3 Linear Transformations Suppose you take every vector (in three-space) and multiply it by 17, or you rotate every vector by 39° about the 2-axis, or you reflect every vector in the x y-plane—these are all examples of linear transformations. A linear transformation* (7) takes each vector in a vector space and “transforms” it into some other vector (ja) > |e’) = Fa), with the proviso that the operation is linear: Pala) + 616) = a(F|a)) + b(F IB), (3.29] for any vectors |e), |B) and any scalars a, b. If you know what a particular linear transformation does to a set of basis vectors, you can easily figure out what it does to any vector. For suppose that Tier) Ty le) + Taslez) +++ + Trilen)s Ter) Tiles) + Toole) + +++ + Tr2len). Pen) = Tinlei) + Taylez) +--+ Tanlen)s or, more compactly, Fle) =o Tile), (GH 12m (3.30) a Sim this chapter I'll use a hat (*) to denote linear transformations; this is not inconsistent with my carlier convention (putting hats on operators), for (as we shall see) our operators are linear transformations.Sec. 3.1: Linear Algebra 81 If |) is an arbitrary vector: lax) = alex) +aale2) +--+ + anlen) = > ajles), (3.31) a then Tle) = Do aj(Fle)) = > Pra Tyled = (do Tyale)- (3.321 im jal ist ist j= Evidently 7 takes a vector with components a1, a2, ..., dy into a vector with compo- nents® a= x Ti; [3.33] ja Thus the n? elements 7;; uniquely characterize the linear transformation 7 (with respect to a given basis), just as the n components a; uniquely characterize the vector lat) (with respect to the same basis): P (Ty1, Thay s+ Tan) 3.34] If the basis is orthonormal, it follows from Equation 3.30 that Ty; = (eilP le)). [3.35] Itis convenient to display these complex numbers in the form of a matrix’: Ty Ta. Tin Th Ty ... T; t=| 7" a 3.36] Ty Toa se Tam The study of linear transformations, then, reduces to the theory of matrices. The sum of two linear transformations (§ + 7) is defined in the natural way: ($+ Pla) = Sa) + Fla); 3.37] this matches the usual rule foradding matrices (you add their corresponding elements): U=S+T 6 U;, = 5S, + Ty. [3.38] Notice the reversal of indices between Equations 3.30 and 3.33, This is not a typographical error. Another way of putting it (switching / ++ in Equation 3.30) is that if the components transform with Tj. the basis vectors transform with 7 TT'IL use boldface to denote matrices,82 Chap. 3. Formalism The product of two linear transformations (57) is the net effect of performing them in succession—first 7, then S: Ja) > ja’) = Pla) > |a") = Sia’) = S(P|ay) = S7ja). 3.39] What matrix U represents the combined transformation U = $7? It's not hard to work it out: a= sie , =o5, (5 rua) = y @ sits) a = lUnae. ja Viet fat a Evidently U= T > Un = Yo S/T (3.40) a this is the standard rule for matrix multiplication—to find the ik'* element of the product, yon look at the i row of S and the &® column of , multiply corresponding entries, and add. The same procedure allows you to multiply rectangular matrices, as long as the number of columns in the first matches the number of rows in the second. In particular, if we write the n-tuple of components of [a as ann x 1 column matrix a a: = ° : [3.41] ay the transformation rule (Equation 3.33) can be written a’ = Ta. [3.42] ‘And now, some useful matrix terminology: The transpose of a matrix (which we shall write with a tilde: T) is the same set of elements, but with rows and columns interchanged: Ty Ta +e Ta Tr Tr Tro . ‘ 13.43] Tin Tr +++ Tan Notice that the transpose of a column matrix is a row matrix: (ay ay ss. Gn). [3.44] A square matrix is symmetric if it is equal to its transpose (reflection in the main diagonal—upper left to lower right—leaves it unchanged); it is antisymmetric if this operation reverses the sign: SYMMETRIC: T=T; ANTISYMMETRIC: T= T [3.45]Sec. 3.1: Linear Algebra 83 To construct the (complex) conjugate of a matrix (which we denote, as usual, with an asterisk: T"), you take the complex conjugate of every element: T Th +. Th ay 7. | at ra]? an ee : [3.46] Th Tr os Tin a, A matrix is real if all its elements are real and imaginary if they are all imaginary: REAL: T*=T; IMAGINARY: T* = —T. (3.47) The Hermitian conjugate (or adjoint) of a matrix (indicated by a dagger: T*) is the transposed conjugate: rn Tyo TH 5 To. Th Be ° (at at... at). [3.48] Ty Tho Ty A square matrix is Hermitian (or self-adjoint) ifit is equal to its Hermitian conjugate; if Hermitian conjugation introduces a minus sign, the matrix is skew Hermitian (or anti-Hermitian): HERMITIAN: T' = T; SKEW HERMITIAN: Tt = [3.49] With this notation the inner product of two vectors (with respect to an orthonormal basis—Equation 3.24), can be written very neatly in matrix form: alb. [3.50] (iB) (Notice that each of the three operations discussed in this paragraph, if applied twice, returns you to the original matrix.) Matrix multiplication is not, in general, commutative (ST # TS); the difference between the two orderings is called the commutator: 1S, T] T — TS. 13.51] The transpose of a product is the product of the transposes in reverse order: (St) =T8 [3.52] (see Problem 3.12), and the same goes for Hermitian conjugates: (ST) = TTst, 13.53)84 Chap. 3. Formalism The unit matrix (representing a linear transformation that carries every vector into itself) consists of ones on the main diagonal and zeroes everywhere else: oO... 0 1 Ol... 0 d=]. : if [3.54] In other words, 1y = 64. [3.55] ‘The inverse of a matrix (written T~') is defined in the obvious way: To T=TT'!=1. [3.56] ‘A matrix has an inverse if and only if its determinant! is nonzero; in fact, [3.57] where C is the matrix of cofactors {the cofactor of element Tj; is (—1)'*” times the determinant of the submatrix obtained from T by erasing the i" row and the j column]. A matrix without an inverse is said to be singular. The inverse of a product (assuming it exists) is the product of the inverses in reverse order: (ty! =T Ss". [3.58] ‘A matrix is unitary if its inverse is equal to its Hermitian conjugate: UNITARY : Ut = U"!. [3.59] Assuming the basis is orthonormal, the columns of a unitary matrix constitute an orthonormal set, and so too do its rows (see Problem 3.16). The components of a given vector depend on your (arbitrary) choice of basis, as do the elements in the matrix representing a given linear transformation. We might inquire how these numbers change when we switch to a different basis. The old basis vectors |e;) are—like all vectors—linear combinations of the new ones: fer) = Sulfa) + Sarl fa) +--+ + Sul fads leo) Stal fi) + Spal fa) +++ + Sua fn), len) Sinl fi) + Sonl f2) + +++ + Sunl fn) 8] assume you know how to evaluate determinants. If not, see M. Boas, Mathematical Methods in the Physical Sciences, 2nd ed, (New York: John Wiley, 1983), Section 3.3,Sec, 3.1: Linear Algebra 85 (for some set of complex numbers 5;;), or, more compactly, le) = OSs G12...) [3.60] a This is itself a linear transformation (compare Equation 3.30),? and we know imme- diately how the components transform: > Sia 3.61] in (where the superscript indicates the basis). In matrix form af =Sat. [3.62] What about the matrix representing a given linear transformation 7—how is it modified by a change of basis? In the old basis we had (Equation 3.42) and Equation 3.62—multiplying both sides by S~!—entails"® a® = S~'a/, so al’ = Sa" = S(T‘a‘) = ST‘S-la/. Evidently =sT’s!. (3.63] In general, two matrices (T; and 2) are said to be similar if T; = ST,S~! for some (nonsingular) matrix S. What we have just found is that similar matrices represent the same linear transformation with respect to two different bases. Incidentally, if the first basis is orthonormal, the second will also be orthonormal if and only if the matrix S is unitary (see Problem 3.14). Since we always work in orthonormal bases, we are interested mainly in unitary similarity transformations. While the elements of the matrix representing a given linear transformation may look very different in the new basis, two numbers associated with the matrix are unchanged: the determinant and the trace. For the determinant of a product is the product of the determinants, and hence T det(T/) = det(S T*S~!) = det(S) det(T*) det(S~!) = det T*. [3.64] Notice, however, the radically different perspective: In this case we're talking about one and the same vector, referred to two different bases, whereas before We were thinking of a completely different vector, referred to the same basis. ‘ONote that S~! certainly exists—if S were singular, the |f;)"s would not span the space, so they wouldn't constitute a basis,86 Chap. 3. Formalism And the trace, which is the sum of the diagonal elements, TAT) = DT, (3.65] isl has the property (see Problem 3.15) that Tr(T)T2) = Tr(T2T}), [3.66] (for any two matrices T; and T;), so that Tx(T/) = Tr(ST°S"!) = Tr(T*S"'S) = Te(T*). [3.67] Problem 3.8 Using the standard basis (/, 7, £) for vectors in three dimensions: (a) Construct the matrix representing a rotation through angle @ (counterclockwise, looking down the axis toward the origin) about the z-axis. (b) Construct the matrix representing a rotation by 120° (counterclockwise, looking down the axis) about an axis through the point (1,1,1). (c) Construct the matrix representing reflection in the x y-plane. (d) Are translations (x —> x + xo, y—> y+ 0, 2 2 + 2p, for some constants Xo, Yo. Zo) linear transformations? If so, find the matrix which represents them; if not, explain why not. «Problem 3.9 Given the following two matrices: -1 1 i 20 -i A= ( 2 0 2). B= (0 1 0 ) 2 -2i 2 i3 2 compute (a) A +B, (b) AB, (c) [A, B], (d) A, (e) A*, ( AT, (g) Tr(B), (h) det(B), and (i) B~!. Check that BB~' = 1. Does A have an inverse? «Problem 3.10 Using the square matrices in Problem 3.9 and the column matrices (02) find (a) Aa, (b) atb, (c) Bb, (d) abt. Problem 3.11 By explicit construction of the matrices in question, show that any matrix T can be written (a) as the sum of a symmetric matrix S and an antisymmetric matrix A;Sec. 3.1: Linear Algebra 87 (b) as the sum of a real matrix R and an imaginary matrix I; (©) as the sum of a Hermitian matrix H and a skew-Hermitian matrix K. Problem 3.12 Prove Equations 3.52, 3.53, and 3.58, Show that the product of two unitary matrices is unitary. Under what conditions is the product of two Hermitian matrices Hermitian? Is the sum of two unitary matrices unitary? Is the sum of two Hermitian matrices Hermitian? Problem 3.13 In the usual basis (7, 7, &), construct the matrix T, representing a rotation through angle @ about the x-axis, and the matrix T, representing a rotation through angle 6 about the y-axis, Suppose now we change bases, to?’ = j, j’ = -i, F =k Construct the matrix $ that effects this change of basis, and check that ST,S"! and ST,S~! are what you would expect. Problem 3.14 Show that similarity preserves matrix multiplication (that is if A‘BS = C’, then A/B/ = C/). Similarity does not, in general, preserve symmetry, reality, or Hermiticity; show, however, that if § is unitary, and H¢ is Hermitian, then H/ is Hermitian. Show that $ carries an orthonormal basis into another orthonormal basis if and only if it is unitary. «Problem 3.15 Prove that Tr(T;T) = Tr(T2T;). It follows immediately that ‘THT: T2Ts) = Tr(T2TsT;), but is it the case that T(T;T2T3) = Tr(T2T;T5), in gen- eral? Prove it, or disprove it. Hint: The best disproof is always a counterexample— and the simpler the better! Problem 3.16 Show that the rows and columns of a unitary matrix constitute orthonormal sets. 3.1.4 Eigenvectors and Eigenvalues Consider the linear transformation in three-space consisting of a rotation, about some specified axis, by an angle 4. Most vectors will change in a rather complicated way (they ride around on a cone about the axis), but vectors that happen to lie along the axis have very simple behavior: They don’t change at all (7|a) = |a)). If is 180°, then vectors which lie in the the “equatorial” plane reverse signs (7a) = —|a)). Ina complex vector space,"' every linear transformation has “special” vectors like these, which are transformed into simple multiples of themselves: Pla) = Ala); [3.68] "This is nor always true in a real vector space (where the scalars are restricted to real values). See Problem 3.17.88 Chap. 3 Formalism they are called eigenvectors of the transformation, and the (complex) number A is their eigenvalue. (The null vector doesn’t count, even though, in a trivial sense, it obeys Equation 3.68 for any 7 and any 2; technically, an eigenvector is any nonzero vector satisfying Equation 3.68.) Notice that any (nonzero) multiple of an eigenvector is still an eigenvector with the same eigenvalue, With respect to a particular basis, the eigenvector equation assumes the matrix form Ta=ia (3.69) (for nonzero a), or (T—Alja=0. [3.70] (Here 0 is the zero matrix, whose elements are all zero.) Now, if the matrix (T— 1) had an inverse, we could multiply both sides of Equation 3.70 by (T — A1)~!, and conclude that a = 0, But by assumption a is not zero, so the matrix (T — A) must in fact be singular, which means that its determinant vanishes: (Ti - A) Tr we Tin Tx (Tn-d) Be det(T — Al) = : . : =0. (3.71) Tu Tra se Tin =) Expansion of the determinant yields an algebraic equation for 4: Cad" + Cyl to + CA + Cy = 0, [3.72] where the coefficients C; depend on the elements of T (see Problem 3.19). This is called the characteristic equation for the matrix; its solutions determine the eigen- values. Notice that it’s an nth-order equation, so it has n (complex) roots.'? However, some of these may be duplicates, so all we can say for certain is that an n x n matrix has at least one and at most n distinct eigenvalues. To construct the corresponding eigenvectors it is generally easiest simply to plug each A back into Equation 3.69 and solve “by hand” for the components of a. I’ll show you how it goes by working out an example. Example. Find the eigenvalues and eigenvectors of the following matrix: 2 0 -2 M= (= iw ) 3.73] 1 oO -L The characteristic equation is (2-4) 0 2 21 (i -d) 2 —B4(+4+i)2-i2=0, [3.74] 1 0 (-1-a) "21 is here that the case of real vector spaces becomes more awkward, because the characteristic equation need not have any (real) solutions at all. See footnote 11 and Problem 3.17Sec. 3.1: Linear Algebra 89 and its roots are 0, 1, and i. Call the components of the first eigenvector (a1, a2, a3); 2 0 -2\ /a ay 0 (= i 21) (22) =0(%)=(0), 1 0 -1/ \as ay 0 which yields three equations: 2a, — 2ay —2iay + ian + 2iay | ~ a3 ‘The first determines as (in terms of a1): a3 = a1; the second determines ap: a = 0; and the third is redundant. We may as well pick a, = 1 (since any multiple of an eigenvector is still an eigenvector): 1 0) . for Ay = 1 For the second eigenvector (recycling the same notation for the components) we have 2 0 ~2\ /a a a (2 i 2!) (a) =1(a)=(®). 1 0 -1/ \as a ay which leads to the equations [3.75] 2a, — 2a; = a1, —2iay + faz + 2ia3 = a2, ay ~ a3 = a, with the solution a3 = (1/2)a1, a2 = [(1 — i)/2Ja1; this time we'll pick a) = 2, so that 2 a?) = (« -») 1 Finally, for the third eigenvector, 2 0 -2\ sa a ia, (2 i 21) (#2) =1(=) = (i). 1 Oo -l a a3 ia3 which gives the equations for Ay = I. [3.76] 2a, — 2a3 = ia, —2iay + ia, + ia: jap, a, — a3 =ia3,90 Chap. 3 Formalism whose solution is a; = a; = 0, with a2 undetermined. Choosing az = 1, we conclude 0 #=(1). for As = i. 0 If the eigenvectors span the space (as they do in the preceding example), we are free to use them as a basis: (3.77] TIA) = Alf, Tif) = delhi), Tift) = Anita) The matrix representing 7 takes on a very simple form in this basis, with the eigen- values strung out along the main diagonal and all other elements zero: a 0... 0 Om... 0 T=|. . (3.78] 0 0 1. An The (normalized) eigenvectors are equally simple: 1 0 0 0 1 0 aM=fO], a®=]}0}, ama] 0 [3.79] 0 0 1 A matrix that can be brought to diagonal form (Equation 3.78) by a change of basis is said to be diagonalizable. The similarity matrix that accomplishes the transformation can be constructed by using the eigenvectors (in the old basis) as the columns of S'; Sy =a). (3.80) Example (cont'd). In the example, 1 2 0 s'=(0 a-i :) 1 1 0 so (using Equation 3.57)Sec. 3.1: Linear Algebra 91 and you can check for yourself that 00 0 SMS"! = (0 I 0) 00% ‘There is a great advantage in bringing a matrix to diagonal form—it's much easier to work with, Unfortunately, not every matrix can be diagonalized—the eigenvectors have to span the space. For an example of a matrix that cannot be diagonalized, see Problem 3.18. «Problem 3.17 The 2 x 2 matrix representing a rotation of the xy-plane is pa (089 —sind “(sind cosa }* Show that (except for certain special angles—what are they?) this matrix has no real eigenvalues. (This reflects the geometrical fact that no vector in the plane is carried into itself under such a rotation; contrast rotations in three dimensions.) This matrix does, however, have complex eigenvalues and eigenvectors. Find them. Construct a matrix § which diagonalizes T. Perform the similarity transformation (STS~') explicitly, and show that it reduces T to diagonal form. Problem 3.18 Find the eigenvalues and eigenvectors of the following matrix: 11 =() 1): Problem 3.19 Show that the first, second, and last coefficients in the characteristic equation (Equation 3.72) are Can this matrix be diagonalized? Cx = (-1)", Cyt = (-)""'Te(), and Co = det(T) B81] For a3 x 3 matrix with elements Jj, what is C1? Problem 3.20 Itis pretty obvious that the trace of a diagonal matrix is the sum of its eigenvalues, and its determinant is their product (see Equation 3.78). It follows (from Equations 3.64 and 3.67) that the same holds for any diagonalizable matrix. Prove that det(T) = Arags+An, Te(D) =A baa ter tay (3.82] for any matrix. (The 4's are the n solutions to the characteristic equation—in the case of multiple roots, there may be fewer linearly independent eigenvectors than there92 Chap. 3 Formalism are solutions, but we still count each 4 as many times as it occurs.) Hint: Write the characteristic equation in the form Gr —A)Q2 — A) On = 2) and use the result of Problem 3.19. 3.1.5 Hermitian Transformations In Equation 3.48 I defined the Hermitian conjugate (or “adjoint”) of a matrix as its transpose conjugate: T' = T*. Now I want to give you a more fundamental definition for the Hermitian conjugate of a linear transformation: It is that transformation f# which, when applied to the first member of an inner product, gives the same result as if 7 itself had been applied to the second vector: (Pap) = (al?) [3.83] (for all vectors |w) and |))."° [Ihave to warn you that although everybody uses it, this is lousy notation. For a and £ are not vectors (the vectors are \w) and |B)), they are labels—serial numbers (“F43A-9GT"), or names (“Charlie”), or bar codes—anything you care to use to identify the different vectors. In particular, they are endowed with no mathematical properties at all, and the expression “7 6” is literally nonsense: linear transformations act on vectors, not labels. But it’s pretty clear what the notation means: |T ) means 7'|f), and (7 'a/B) means the inner product of the vector 7 |ar) with the vector |). Notice in particular that (a\eB) = e(alB), [3.84] (ca|B) = c* (ap) [3.85] for any scalar c.] If you're working in an orthonormal basis (as we always shall), the Hermitian conjugate of a linear transformation is represented by the Hermitian conjugate of the corresponding matrix (so the terminology is consistent); for (using Equations 3.50 and 3.53), (a|FB) = ai Tb = (Tia)'b = (7iaip). [3.86] In quantum mechanics, a fundamental role is played by Hermitian transforma- tions (7+ = 7). The eigenvectors and eigenvalues of a Hermitian transformation have three crucial properties: "31 you're wondering whether such a transformation necessarily exists, you should have been a ‘math major. Still, it’s a good question, and the answer is yes. See, for instance, Halmos, (footnote 1), Section 44.Sec. 3.1: Linear Algebra 93 1, The eigenvalues of a Hermitian transformation are real. Proof: Let 4 be an eigenvalue of 7: la) = dja), with ja) # |0). Then (a|Fa) = (ala) = A{aler) Meanwhile, if Hermitian, then (alfa) = (Pala) = (ala) = A* (ala) But (|) 0 (Equation 3.20), so A = , and hence A is real. QED 2. The eigenvectors of a Hermitian transformation belonging to dis- tinct eigenvalues are orthogonal. Proof: Suppose 7 \a) = {a) and 7B) = |B), with A # yx. Then (a1FB) = (clus) = wlalB), and if 7 is Hermitian, (al 7B) = (FalB) = (Aaip) = 2" (a). But 4 = A* (from property 1), and A # 1, by assumption, so (a|B) = 0. QED 3. The eigenvectors of a Hermitian transformation span the space. Comment: If all the n roots of the characteristic equation are distinct, then (by property 2) we have n mutually orthogonal eigenvectors, so they obviously span the space. But what if there are duplicate roots (or, as they are called, in this context, degenerate eigenvalues)? Suppose 4 is m-fold degenerate; any linear combination of two eigenvectors belonging to the same eigenvalue is still an eigenvector (with the same eigenvalue)—what we must show is that there are m linearly independent eigenvectors with eigenvalue 2. The proof is given in most books on linear algebra,"* and I shall not repeat it here. These eigenvectors can be orthogonalized by the Gram- Schmidt procedure (see Problem 3.4), so in fact the eigenvectors of a Hermitian transformation can always be taken to constitute an orthonormal basis. It follows, in Particular, that any Hermitian matrix can be diagonalized by a similarity transfor- mation, with S unitary. This rather technical result is, in a sense, the mathematical support on which much of quantum mechanics leans. As we shall see, it turns out to be a thinner reed than one might have hoped. 4] like the treatment in F. W. Byron, Jr, and R. W. Fuller, Mathematics of Classical and Quantum Physics (Reading, MA: Addison-Wesley, 1969), Vol. I, Section 4.7.94 Chap. 3 Formalism Problem 3.21 A Hermitian linear transformation must satisfy (@|7B) = (Ta|B) for all vectors |w) and |8). Prove that it is (surprisingly) sufficient that (ify) = (Fyly) for all vectors |y). Suppose you could show that (é,|7e,) = (Teq|e,) for every member of an orthonormal basis. Does it necessarily follow that 7 is Hermitian? Hint: First let |y) = |) + |), and then let |v) = |a) + i/). +Problem 3.22 Let (a) Verify that T is Hermitian. (b) Find its eigenvalues (note that they are real). (c) Find and normalize the eigenvectors (note that they are orthogonal). (d) Construct the unitary diagonalizing matrix §, and check explicitly that it diag- onalizes T. (e) Check that det(T) and Tr(T) are the same for T as they are for its diagonalized form. +*Problem 3.23 Consider the following Hermitian matrix: 2 G1 T= (- 2 i) 1 -i 2 (a) Calculate det(T) and Tr(T). (b) Find the eigenvalues of T. Check that their sum and product are consistent with (a), in the sense of Equation 3.82. Write down the diagonalized version of T. (c) Find the eigenvectors of T. Within the degenerate sector, construct two linearly independent eigenvectors (it is this step that is always possible for a Hermitian matrix, but not for an arbitrary matrix—contrast Problem 3.18). Orthogonalize them, and check that both are orthogonal to the third. Normalize all three eigenvectors. (d) Construct the unitary matrix § that diagonalizes T, and show explicitly that the similarity transformation using S reduces T to the appropriate diagonal form, Problem 3.24 A unitary linear transformation is one for which UtU = 1. (a) Show that unitary transformations preserve inner products, in the sense that (Geel B) = (eB), for all vectors |e), |B).Sec. 3.2: Function Spaces 95 (b) Show that the eigenvalues of a unitary transformation have modulus 1. (c) Show that the eigenvectors of a unitary transformation belonging to distinct eigenvalues are orthogonal. 3.2 FUNCTION SPACES We are ready now to apply the machinery of linear algebra to the interesting and important case of function spaces, in which the “vectors” are (complex) functions of x, inner products are integrals, and derivatives appear as linear transformations. 3.2.1 Functions as Vectors Do functions really behave as vectors? Well, is the sum of two functions a function? Sure. Is addition of functions commutative and associative? Indeed. Is there a “null” function? Yes: f(x) = 0. If you multiply a function by a complex number, do you get another function? Of course. Now, the set of all functions is bit unwieldy—we'll be concerned with special classes of functions, such as the set of all polynomials of degree < N (Problem 3.2), or the set of all odd functions that go to zero atx = 1, or the set of all periodic functions with period 2. Of course, when you start imposing conditions like this, you’ve got to make sure that you still meet the requirements for a vector space. For example, the set of all functions whose maximum value is 3 would not constitute a vector space (multiplication by 2 would give you functions with maximum value 6, which are outside the space). The inner product of two functions (f(x) and g(x)] is defined by the integral (fle) = f fesy*ats) ax | 13871 (the limits will depend on the domain of the functions in question). You can check for yourself that it satisfies the three conditions (Equations 3.19, 3.20, and 3.21) for an inner product. Of course, this integral may not converge, so if we want a function space with an inner product, we must restrict the class of functions so as to ensure that (Sg) is always well defined. It is clearly necessary that every admissible function be square integrable: [ireoras <00 [3.88] (otherwise the inner product of f with itse/f wouldn’t even exist). As it turns out,96 Chap. 3 Formalism ‘The first few Legendre polynomials, P, (x). Pp = 3Gx?-1) Ps = 405x535) Ps = 4(35x* — 30x? + 3) Ps = $(63x5 — 70x? + 15x) this restriction is also sufficient—if f and g are both square integrable, then the integral in Equation 3.87 is necessarily finite.'§ For example, consider the set P(N) of all polynomials of degree < N: 1 D(x) = a9 + ax $F agx? +++ -ayx’, [3.89] on the interval —1 < x < 1. They are certainly square integrable, so this is a bona fide inner product space. An obvious basis is the set of powers of x: lei) = 1, |e2) =x, Jes) = 27, ..., lew) =a, 13.90] evidently it’s an N-dimensinal vector space. This is not, however, an orthonormal basis, for 1 {erler) -/ Idx “1 and so on, If you apply the Gram-Schmidt procedure, to orthonormalize this ba- sis (Problem 3.25), you get the famous Legendre polynomials, P,(x) (except that Legendre, who had other things on his mind, didn’t normalize them properly): le.) = Vn = (1/2) Pre), (2 = 1,2,...,.N). [3.91] In Table 3.1 I have listed the first few Legendre polynomials. 1 (erles) -[ xdx = 2/3, “1 +Problem 3.25 Orthonormalize the powers of x, on the interval —1 < x < 1, to obtain the first four Legendre polynomials (Equation 3.91). «Problem 3.26 Let 7(N) be the set of all trigonometric functions of the form wei F(x) = Slay sin(nsex) + by cos(nx)], [3.92] n=0 'SThere is a quick phoney “proof” of this, based on the Schwarz inequality (Equation 3.27). The trouble is, we assumed the existence of the inner product in proving the Schwarz inequality (Problem 3.5), so the logic is circular. For a legitimate proof, see F. Riesz. and B, Sz.-Nagy, Functional Analysis (New ‘York: Unger, 1955), Section 21.Sec. 3.2: Function Spaces 97 on the interval —1 < x < 1. Show that ev, (n =0,41,...,4(N — 1) [3.93] len) = v2 constitutes an orthonormal basis. What is the dimension of this space? Problem 3.27 Consider the set of all functions of the form p(x)e~*"/?, where p(x) is again a polynomial of degree < N in x, on the interval oo for all x, means 0 = hap, ay = hay, a; = har, and soon. If. = 0, thenall the components are zero, and that's nota legal eigenvector: but if 4 4 0, the first equation says ap = 0, so the second gives a, = 0, and the third says a2 = 0, and so on, and we're back in the same bind. This Hermitian operator doesn’t have a complete set of eigenfunctions—in fact it doesn’t have any at all! Not, at any rate, in P(oo). What would an eigenfunction of ¥ look like? If x8(x) = Ag(x), [3.99]Sec. 3.2: Function Spaces 99 where A, remember, is a constant, then everywhere except at the one point x = 4 we must have g(x) = 0. Evidently the eigenfunctions of £ are Dirac delta functions: 8.(x) = B(x — 4), [3.100] and since delta functions are certainly not polynomials, it is no wonder that the operator £ has no eigenfunctions in P(oo). The moral of the story is that whereas the first two theorems in Section 3.1.5 are completely general (the eigenvalues of a Hermitian operator are real, and the eigenvectors belonging to different eigenvalues are orthogonal), the third one (com- pleteness of the eigenvectors) is valid (in general) only for finite-dimensional spaces. In infinite-dimensional spaces some Hermitian operators have complete sets of eigen- vectors (see Problem 3.32d for an example), some have incomplete sets, and some (as we just saw) have no eigenvectors (in the space) at all.'” Unfortunately, the complete- ness property is absolutely essential in quantum mechanical applications. In Section 3.3 I'll show you how we manage this problem. Problem 3.28 Show that exp(—x?/2) is an eigenfunction of the operator 0 = (d?/dx*) — x?, and find its eigenvalue. «Problem 3.29 (a) Construct the matrix D representing the derivative operator D = d/dx with Tespect to the (nonorthonormal) basis (Equation 3.90) in P(N). (b) Construct the matrix representing D with respect to the (orthonormal) basis (Equation 3.93) in the space T(V) of Problem 3.26. (©) Construct the matrix X representing the operator # = x with respect to the basis (Equation 3.90) in P (00). If this is a Hermitian operator (and it is), how come the matrix is not equal to its transpose conjugate? **Problem 3.30 Construct the matrices D and X in the (orthonormal) basis (Equa- tion 3.91) for P(co). You will need to use two recursion formulas for Legendre polynomials: 1 x Pal) = Gant TD) Put (x) + 2 Pai): [3.101] dPy Yen = 4k — DPy-2-100, (3.102 dx "In an n-dimensional vector space, every linear transformation can be represented (with respect to a particular basis) by an n x n matrix, and as long as n is finite, the characteristic Equation 3.71 is ‘guaranteed to deliver atleast one eigenvalue. But ifn is infinite, we can’t take the determinant, there is no characteristic equation, and hence there is no assurance that even a single eigenvector exists100 Chap. 3. Formalism where the sum cuts off at the first term with a negative index. Confirm that X is Hermitian but iD is not. Problem 3.31 Consider the operator D? = d? /dx?. Under what conditions (on the admissable functions) is it a Hermitian operator? Construct the matrix representing DB? in P(N) (with respect to the basis Equation 3.90), and confirm that it is the square of the matrix representing D (Problem 3.29a). Problem 3.32 (a) Show that i is Hermitian in the space T(N) of Problem 3.26. (b) What are its eigenvalues and (normalized) eigenfunctions, in T(N)? (C) Check that your results in (b) satisfy the three theorems in Section 3.1.5. (d) Confirm that iD has a complete set of eigenfunctions in 7 (0) (quote the perti- nent theorem from Fourier analysis). 3.2.3 Hilbert Space To construct the real number system, mathematicians typically begin with the integers, and use them to define the rationals (ratios of integers). They proceed to show that the rational numbers are “dense,” in the sense that between any two of them (no matter how close together they are) you can always find another one (in fact, infinitely many of them). And yet, the set of all rational numbers has “gaps” in it, for you can easily think of infinite sequences of rational numbers whose /imit is not a rational number. For example, [3.103] is a rational number for any finite integer N, but its limit (as N > 00) is In2, which is not a rational number. So the final step in constructing the real numbers is to “fill in the gaps”, or “complete” the set, by including the limits of all convergent sequences of rational numbers. (Of course, some sequences don’t have limits, and those we do not include, For example, if you change the minus signs in Equation 3.103 to plus signs, the sequence does not converge, and it doesn’t correspond to any real number.) The same thing happens with function spaces. For example, the set of all polynomials, P(oo), includes functions of the form x xt x’ In@)= Lx 5 pt gt tw [3.104] (for finite N), but it does not include the limit as N — oo: x xt ox" > lest Sth te =a e [3.105]Sec. 3.2: Function Spaces 101 For e* is not itself a polynomial, although it is the limit of a sequence of polynomials. To complete the space, we would like to include all such functions. Of course, some sequences of polynomials don’t have limits, or have them only for a restricted range of x. For example, the series 1 -x ee converges only for |x| < 1. And even if the sequence does have a limit, the limit function may not be square integrable, so we can’t include it in an inner product space. To complete the space, then, we throw in all square-integrable convergent sequences of functions in the space. Notice that completing a space does not involve the intro- duction of any new basis vectors; it is just that we now allow linear combinations involving an infinite number of terms, la) = Drajle), {3.106} a provided (r|q) is finite—which is to say (if the basis is orthonormal), provided Vai? <0. (3.107) a ‘A complete’ inner product space is called a Hilbert space.” The completion of P(c0) is easy to characterize: It is nothing less than the set of aif square-integrable functions on the interval —1 (—00, +00) (of Lo, for short), because this is where quantum mechanical wave functions live. Indeed, to physicists Ly is practically synonymous with “Hilbert space”. ‘The eigenfunctions of the Hermitian operators i = id/dx and ¢ = x are of particular importance. As we have already found (Equations 3.95 and 3.100), they take the form fils) = Ae, and —_gy(x) = BS(x — 2), respectively. Note that there is no restriction on the eigenvalues—every real number is an eigenvalue of iD, and every real number is an eigenvalue of . The set of all eigenvalues of a given operator is called its spectrum; iD and £ are operators with continuous spectra, in contrast to the discrete spectra we have encountered 1 Note the two entirely different uses of the word “complete”: a set of vectors is complete if it spans the space; an inner product space is complete if it has no “holes” init (ie, it includes all its limits). + Byery finite-dimensional inner product space is trivially complete, sothey'reall technically Hilbert spaces, but the term is usually reserved for infinite-dimensional spaces.102 Chap. 3 Formalism heretofore. Unfortunately, these eigenfunctions do not lie in Hilbert space, and hence, in the strictest sense, do not count as vectors at all. For neither of them is square- integrable: 0 eo f LOY fA@)dx = iar f ee dx = ae f ldx + 00, oe Z -_ and ~ ~ fi scotecods = 1a.7 f° ac — 2000 ay dx = 12450. —2) > ow. oe ~ Nevertheless, they do satisfy a kind of orthogonality condition: f Hoy fdr = Atay f ee" dx = |Aj/?278(A — 1) 00 (see Equation 2.126), and f Bix)" gu(x) dx = 538, f 8(x — A)S(x — x) dx = |B, PP8(A ~ pw). It is customary to “normalize” these (unnormalizable) functions by picking the con- stant so as to leave an unadorned Dirac delta function on the right side (replacing the Kronecker delta in the usual orthonormality condition; Equation 3.23).° Thus Li oe @= Tax with (Ail fu) = 8 - 1), (3.108) are the “normalized” eigenfunctions of iD, and g(x) = 5(x — 2), with (g.1gu) = 60 — 1), [3.109] are the “normalized” eigenfunctions of ¥.7! What if we use the “normalized” eigenfunctions of i D and ¥ as bases for L2?” Because the spectrum is continuous, the linear combination becomes an integral: in=f anh hdd; in=f Palen) da. 3.110) 20 0 207 [ call this “normalization” (in quotes) so you won't confuse it with the real thing. 21 We are engaged here in a dangerous stretching of the rules, pioneered by Dirac (who had a ki of inspired confidence that he could get away with it) and disparaged by von Neumann (who was more sensitive to mathematical niceties), in their rival classics (P. A. M. Dirac, The Principles of Quantum Mechanics, first published in 1930, 4" ed., Oxford (Clarendon Press) 1958, and J. von Neumann, The Mathematical Foundations of Quantum Mechanics, first published in 1932, revised by Princeton Univ. Press, 1955). Dirac notation invites us to apply the language and methods of linear algebra to functions that Tie in the “almost normalizable” suburbs of Hilbert space. It turns out to be powerful and effective beyond any reasonable expectation. That's right: We're going to use, as bases, sets of functions none of which is actually in the space! ‘They may not be normalizable, but they are complete, and that’s all we need.Sec. 3.2: Function Spaces 103 ‘Taking the inner product with | f,), and exploiting the “orthonormality” of the basis (Equation 3.108), we obtain the “components” a;: l= faathulsinda= [~ a8ue—2ya2= So 1 Wed evidently the —A “component” of the vector | f), in the basis of eigenfunctions of iD, is the Fourier transform (Equation 2.85) of the function f(x). Likewise, a= (All) = / e™ f(x)dx = F(-2); 3.411) b= (gilt) =/ 8x -Af@)dx = fR, [3.112] so the A “component” of the vector |) in the position basis is f(4) itself. [If that sounds like double-talk, remember that | f) is an abstract vector, which can be expressed with respect to any basis you like; in this sense the function f(x) is merely the collection of its “components” in the particular basis consisting of eigenvectors of the position operator.] Meanwhile, we can no longer represent operators by matrices because the basis vectors are labeled by a nondenumerable index. Nevertheless, we are still interested in quantities of the form (AIT fu)» which, by force of habit, we shall call the A, matrix element of the operator 7. «Problem 3.33 (a) Show that any linear combination of two functions in L2(a, b) is still in L2(a, 5). If this weren't true, of course, L2(a, b) wouldn’t be a vector space at all. (b) For what range of (real) v is the function f(x) = |x|" in Lo(—1, +1)? (c) For what range of a is the function f(x) = 1—x+x?—x3+-+-in L3(—a, +a)? (d) Show that the function f(x) = e~™" is in Lo, and find its “components” in the basis (Equation 3.108). (e) Find the matrix elements of the operator D? with respect to the basis (Equation 3.108) of Lo Problem 3.34 L(—1, +1) includes discontinuous functions (such as the step func- tion, (x), Equation [2.125], which are not differentiable. But functions expressible as Taylor series (f(x) = ao + aix + apx? + +++) must be infinitely differentiable. How, then, can (x) be the limit of a sequence of polynomials? Note: This is not a difficult problem, once you see the light, but it is very subtle, so don’t waste a lot of, time on it if you're not getting anywhere.104 Chap. 3. Formalism 3.3 THE GENERALIZED STATISTICAL INTERPRETATION My next project is to recast the fundamental principles of quantum mechanics (as developed in Chapters 1 and 2) into the more elegant language of linear algebra. Remember that the state of a particle is represented by its wave function, W(x, ). whose absolute square is the probability density for finding the particle at x, at time r. It follows that Y must be normalized, which is possible (by dividing off a constant) if and only if it is square integrable. Thus 1. The state of a particle is represented by a normalized vector (|W)) in the Hilbert space 2. Classical dynamical quantities (such as position, velocity, momentum and ki- netic energy) can be expressed as functions of the “canonical” variables x and p (and, in rare cases, 1): Q(x, p,1). To each such classical observable we associate a quantum-mechanical operator, Q, obtained from Q by the substitution on 3. PO Tax om The expectation value of Q, in the state W, is ()= f v6.0" O¥e,nds, which we now write as an inner product:” (Q) = (widw). 13.114] Now, the expectation value of an observable quantity has got to be a real number (after all, it corresponds to actual measurements in the laboratory, using rulers and clocks and meters), so (W/OW) = (vlOw)* = (OV|Y), [3.115] for all vectors |‘). It follows (see Problem 3.21) that Q must be a Hermitian operator. Thus 2. Oveervalie quantities, Q(x, p, t), are represented by Hermitian opera- tors, O(x, +, 1); the expectation value of Q, in the state |), is (W| Ow). *5The “lousy notation” I warned you about on page 92 is not so bad in this context, for we are using the function W itself to label the vector |\W), and the expression Ow is perfectly self-explanatory.Sec. 3.3: The Generalized Statistical Interpretation 105 In general, identical measurements on identically prepared systems (all in the same state W/) do not yield reproducible results; however, some states are determi- nate, for a particular observable, in the sense that they always give the same result. TA competent measurement of the total energy of a particle in the ground state of the harmonic oscillator, for example, will always return the value (1/2)fiw.] For a determinate state of observable Q, the standard deviation is zero: 0= 08 = (0 (Q))") = (VIO — (QW) = (0 —(O)WIO — (AN) = (QNIWP. BAG] [I used the fact that the operator (@ — (Q)) is Hermitian to peel it off the second member of the inner product and attach it to the first member.) But the only vector with norm zero is the null vector (Equation 3.20), so (Q — (Q))|'¥) = 0, or OW) = (Q)|¥). [3.117] Evidently determinate states are eigenvectors of Q. Thus 3. A measurement of the observable Q on a particle in the state |) is certain to return the value 2 if and only if |) is an eigenvector of O, with eigenvalue 2. For example, the time-independent Schrédinger equation (Equation 2.11), Aw =Ey. is nothing but an eigenvalue equation for the Hamiltonian operator, and the solutions are states of determinate energy (as we noted long ago). Up to this point I have added nothing new to the statistical interpretation; I have merely explored its implications in the language of linear algebra. But there is a missing part to the story: Although we can calculate the average result of any mea- surement, we still cannot say what the probability of getting a particular result would be if we were to measure a given observable Q on a particle in an arbitrary state |) (except for the special case of position for which the original statistical interpretation supplies the answer). To finish the job, we need the following generalized statistical interpretation, which is inspired by postulate 3 above, and subsumes it as a special case: 3‘. If you measure an observable Q on a particle in the state |W), you are certain to get one of the eigenvalues of O. The probability of getting the particular eigenvalue A is equal to the absolute square of the 4 component of |W), when expressed in the orthonormal basis of eigenvecters.”* 24Notice that we could calculate from this the expectation value of Q, and it that the result is consistent with postulate 2 above. See Problem 3.35(c). important to check106 Chap. 3 Formalism To sustain this postulate, it is essential that the eigenfunctions of Q span the space. As we have seen, in the finite-dimensional case the eigenvectors of a Hermi- tian operator always span the space. But this theorem fails in the infinite-dimensional case—we have encountered examples of Hermitian operators that have no eigenfunc- tions at all, or for which the eigenfunctions lie outside the Hilbert space. We shall take it as a restriction on the subset of Hermitian operators that are observable, that their eigenfunctions constitute a complete set (though they need not fall inside L>)25 Now, there are two kinds of eigenvectors, which we need to treat separately. If the spectrum is discrete (with the distinct eigenvalues separated by finite gaps), we can label the eigenvectors with an integer n: Olen) =Anlén), withn =1,2,3,...: [3.118] the eigenvectors are orthonormal (or rather, they can always be chosen so): (nlm) = Sums [3.119] the completeness relation takes the form of a sum: co IY) =D renlends [3.120] the components are given by “Fourier’s trick”: = (en|¥), (3.120) and the probability of getting the particular eigenvalue 2, is lenl” = Men) {3.122] On the other hand, if the spectrum is continuous, the eigenvectors are labeled by a continuous variable (k): Olex) =Axlex), with — 00 h/2) back in Section 1.6, and you have checked it several times in the problems. But we have never actually proved it. In this section I shall prove a more general version of the uncertainty principle and explore some of its implications. The argument is beautiful, but rather abstract, so watch closely. 3.4.1 Proof of the Generalized Uncertainty Principle For any observable A, we have (quoting Equation 3.116) of = (A= (A) (A = (AN) = (SIP),Sec. 3.4: The Uncertainty Principle 109 where | f) = (A — (A))|). Likewise, for any other observable B, 0% =(glg), where |g) = (B - (B))|). Therefore (invoking the Schwarz inequality, Equation 3.27), 0305 = (/I/iglg) = Ile)? [3.135] Now, for any complex number z, Iz? = (Re(z))? + Um(z))? > (m(z))? = Ine —2)P. [3.136] Therefore, letting z = (fle), 2 1 2 oie} > (Hite) Wein) - 31371 But (fla) = (4 — (4) WB — (BV) = (ICA — (AB (B))Y) ~ A(B) — Bla) + (4)(B))) |ABW) — (B)(W|AW) — (A) (YI BW) + (4)(B) (YT) ~ (B)A) ~ (A)(B) + (A)(B) AB) — (A)(B) Similarly, _. (gif) = (BA) — (A)(B), so _ (fla) — (gif) = (4B) — (BA) = (LA, BD) where {A, B) = AB- BA [3.138] is the commutator of the two operators. Conclusion: 13.139] This is the uncertainty principle in its most general form. (You might think the i makes it trivial—isn’t the right side negative? No, for the commutator carries its own factor of 7, and the two cancel out.) For example, suppose the first observable is position (4 = x), and the second is momentum (B = (h/i)d/dx). To determine the commutator, we use an arbitraryChap. 3 Formalism “test function”, f(x): epfyatsin—ttonat [se 44s] iis, so [3.140] Accordingly, or, since standard deviations are by their nature positive, 0p > x (3.141) ‘That proves the original Heisenberg uncertainty principle, but we now see that it is just one application of a far more general theorem: There will be an “uncertainty principle” for any pair of observables whose corresponding operators do not com- ‘mute. We call them incompatible observables. Evidently, incompatible observables do not have shared eigenvectors—at least, they cannot have a complete set of common. eigenvectors. Matrices representing incompatible observables cannot be simultane ously diagonalized (that is, they cannot both be brought to diagonal form by the same similarity transformation). On the other hand, compatible observables (whose operators do commute) share a complete set of eigenvectors, and the corresponding matrices can be simultaneously diagonalized (see Problem 3.40). «Problem 3.39 Prove the famous “(your name) uncertainty principle,” relating the uncertainty in position (A = x) to the uncertainty in energy (B = p’/2m + V): Ox; > Fy HL KOH PH. For stationary states this doesn’t tell you much—why not? Problem 3.40 Prove the following: (a) If two matrices commute ({A,B] = 0), and you apply the same similarity transformation to both of them (A’ = SAS~', B’ = SBS~'), the resulting matrices also commute ([A’, B'] = 0). (b) Diagonal matrices always commute. (It follows from this that simultaneously diagonalizable matrices must commute. Conversely, if two Hermitian matrices commute, then they are simulatneously diagonalizable—i.e., they have a com- plete set of common eigenvectors. This is not so easy to prove” unless you happen to know that the spectrum of one of them is nondegenerate.) 26See Byron and Fuller (footnote 14), Theorem 4.22,Sec. 3.4: The Uncertainty Principle 111 (c) If matrices A and B commute, and |a) is an eigenvector of A, and the spectrum. of A is nondegenerate, then ja) is also an eigenvector of B. (In that case the matrix § that diagonalizes A also diagonalizes B.) +Problem 3.41 (a) Prove the following commutator identity: (4B, C] = AB, C14 (4, C18. (3.142) (b) Using Equations 3.140 and 3.142, show that [2", p] = itng"|, (©) For any function f(x) that can be expanded in a power series, show that [f@), pl= Hf), where the prime denotes differentiation. 3.4.2 The Minimum-Uncertainty Wave Packet We have twice encountered wave functions that hit the position-momentum uncer- tainty limit (0,0,, = h/2): the ground state of the harmonic oscillator (Problem 2.14) and the Gaussian wave packet for the free particle (Problem 2.22). This raises an interesting question: What is the most general minimum-uncertainty wave packet? Looking back at the proof of the uncertainty principle, we note that there were two points at which inequalities came into the argument: Equation 3.135 and Equation 3.136. Suppose we require that each of these be an equality, and see what this tells us about Y. The Schwarz inequality becomes an equality when the angle between the two vectors (Equation 3.28) is zero—that is, when one is a multiple of the other: |g) = cl,f), for some scalar c. (Study the proof of the Schwarz inequality in Problem 3.5 if you're not convinced.) Meanwhile, in Equation 3.136 I threw away the real part of z; equality results if Re(z) = 0, which is to say, if Re( |g) =Re(c( | f)) = 0. Now {/| /) is certainly real, so this means the constant ¢ must be purely imaginary— let's call it ia. The necessary and sufficient condition for minimum uncertainty, then, is |g) =éalf), where a is real. 13.143] In particular, for the position-momentum uncertainty principle this criterion becomes (25 -w)y =i (x) )y, [3.144] idx112 Chap. 3 Formalism which is a differential equation for W as a function of x, with the general solution (see Problem 3.42) . W(x) = Aen AO lv [3.145] Evidently the minimum-uncertainty wave packet is a Gaussian—and sure enough, the two examples we encountered earlier were Gaussians.” Problem 3.42 Solve Equation 3.144 for W(x). (Note that (x) and (p) are constants, as far as x is concerned.) 3.4.3 The Energy-Time Uncertainty Principle ‘The position-momentum uncertainty principle is usually written in the form h Ax Ap2 5 [3.146] ‘Ax (the “uncertainty” in x) is sloppy notation (and sloppy language) for the standard deviation in the results of repeated measurements on identically prepared systems. Equation 3.146 is often paired with the energy-time uncertainty principle, ALAE> a (3.147) Indced, in the context of special relativity the energy-time form might be thought of as a consequence of the position-momentum version, because x and f (or rather, ct) go together in the position-time four-vector, while p and E (or rather, E /c) go together in the energy-momentum four-vector. So in a relativistic theory Equation 3.147 would be a necessary concomitant to Equation 3.146. But we're not doing relativistic quantum mechanics—the Schrédinger equation is explicitly nonrelativistic: It treats t and x on a very unequal footing (as a differential equation it is first-order in t, but second-order in x), and Equation 3.147 is emphatically not implied by Equa- tion 3.146. My purpose now is to derive the energy-time uncertainty principle, and in the course of that derivation to persuade you that it is really an altogether different beast, whose similarity in appearance to the position-momentum uncertainty principle is quite misleading Afterall, position, momentum, and energy are all dynamical variables—measur- able characteristics of the system, at any given time. But time itself is nor a dynamical variable (not, at any rate, in a nonrelativistic theory): You don’t go out and measure the “time” of a particle, as you might its position or its energy. Time is the indepen- dent variable of which the dynamical quantities are functions. In particular, the At 27 Note that it is only the dependence of W on x that is at issue here—the “constants” 4, a, (x), and (p) may all be functions of time, and as time goes on W may evolve away from the minimal form, All I'm asserting is that if, at some instant, the wave function is Gaussian in x, then (at that instant) the uncertainty product is minimalSec. 3.4: The Uncertainty Principle 113 in the energy-time uncertainty principle is not the standard deviation of a collection of time measurements; roughly speaking (I'll make this more precise in a moment), it is the time it takes the system to change substantially. As a measure of how fast the system is changing, let us compute the time derivative of the expectation value of some observable, O(x, p,1): di dyn av 5 ad aw a= alley) = (Glow) + (I or + (HIOS-) Now the Schrddinger equation says awn ih— = Aw ns (where H = p*/2m + V is the Hamiltonian). So (AY OW) + quid +2). i ar But is Hermitian, so (AW|OW) = (W|A OW), and hence i L This is an interesting and useful result in its own right (see Problems 3.43 and 3.53), In the typical case, where the operator does not depend explicitly on ¢,”* it tells us that the rate of change of the expectation value is determined by the commutator of the operator with the Hamiltonian. In particular, if © commutes with H, then (Q) is constant, and in this sense Q is a conserved quantity, Suppose we pick 4 = H and B = Q, in the generalized uncertainty principle (Equation 3.139), and assume that Q does not depend explicitly on t: 1. ay (lado _ (hr? (dia)\? chop (ztH On) = (5742) = (5) (AZ. Or, more simply, 4g) = iA, oy + (2 60) = i Ob + (GP) [3.148] onog > || HO Fat Let’s define AE = 01 (with A as the usual sloppy notation for standard deviation), and a 22 [3.149] Ars —<2 = ova (3.150] 28 As an example of explicit time dependence, think of the potential energy of a harmonic oscillator ‘whose spring constant is changing (perhaps the temperature is rising, so the spring becomes more flexible): = C/2mloy Px?114 Chap. 3. Formalism Then h AEAI> > (3.151) and that's the energy-time uncertainty principle. But notice what is meant by Af here: Since At represents the amount of time it takes the expectation value of Q to change by one standard deviation. In particular, At depends entirely on what observable (Q) you care to look at—the change might be rapid for one observable and slow for another. But if AE is small, then the rate of change of all observables must be very gradual. and conversely, if any observable changes rapidly, the “uncertainty” in the energy must be large. Example 1. In the extreme case of a stationary state, for which the energy is uniquely determined, all expectation values are constant in time (At = 00)—as, in fact, we noticed some time ago (see Equation 2.8). To make something happen, you must take a linear combination of at least two stationary states—for example, Wx.) = avn wet + bya(we 14". Ifa, b, Wi, and yo are real, £,-E Mean? =a (2)) + PRC? + 2ab va ad¥a 60) 0s ( S tr). The period of oscillation is t = 20vh/(E2 — E\). Roughly, then, AE = Ey — Ey and At = t (for the exact calculation, see Problem 3.44), so h AE At =2nh> z Example 2. How long does it take a free particle wave packet to pass by a particular point (Figure 3.1)? Qualitatively (an exact version is explored in Problem 3.45), At = Ax/v =mAx/p, but E = p*/2m,so AE = pAp/m. Therefore, Ap mA. i AE At = POPME* _ axap> 5. m—D 2 Example 3. The A particle lasts about 10-®* seconds before spontaneously disintegrating. If you make a histogram of all measurements of its mass, you get a kind of bell-shaped curve centered at 1232 MeV/c?, with a width of about 115Sec, 3.4: The Uncertainty Principle 115 4x ——>1 a * Figure 3.1: A free particle wave packet approaches the point 4 (Example 2). MeVic?. Why does the rest energy (mc”) sometimes come out higher than 1232, and sometimes lower? Is this experimental error? No, for 15 AE At = (> MeV)(10~* sec) = 6 x 10-?? MeV sec, whereas fi/2 = 3 x 10- MeV sec. So the spread in m is about as small as the uncertainty principle allows—a particle with so short a lifetime just doesn’t have a very well-defined mass.” Notice the variety of specific meanings attaching to the term At in these exam- ples: In Example 1 it’s a period of oscillation; in Example 2 it’s the time it takes a particle to pass a point; in Example 3 it’s the lifetime of an unstable particle. In every case, however, At is the time it takes for the system to undergo substantial change. It is often said that the uncertainty principle means that energy is not strictly conserved in quantum mechanics—that you're allowed to “borrow” energy AE, as long as you “pay it back” ina time At ~ h/2AE; the greater the violation, the briefer the period over which it can occur. There are many legitimate readings of the energy-time un- certainty principle, but this is not one of them. Nowhere does quantum mechanics license violation of energy conservation, and certainly no such authorization entered into the derivation of Equation 3.151, But the uncertainty principle is extraordinar- ily robust: It can be misused without leading to seriously incorrect results, and as a consequence physicists are in the habit of applying it rather carelessly. +Problem 3.43 Apply Equation 3.148 to the following special cases: (a) Q = 1; (6) Q = H; (©) Q = x; (d) Q = p. In each case, comment on the result, with particular reference to Equations 1.27, 1.33, 1.38, and 2.35. ++Problem 3.44 Test the energy-time uncertainty principle for the wave function in Problem 2.6 and the observable x by calculating oj, 0, and d(x) /dt exactly. 2 Actually, Example 3 is a bit of a fraud. You can’t measure 10-3 sec on a stop-watch, and in practice the lifetime of such a short-lived particle is inferred from the width of the mass plot, using the ‘uncertainty principle as input, However, the point is valid even if the numbers are suspect. Moreover, if you assume the A is about the same size as a proton (~ 10~'* m), then 10 sec is roughly the time it takes light to cross the particle, and it’s hard to imagine that the lifetime could be much fess than that116 Chap. 3 Formalism +*Problem 3.45 Test the energy-time uncertainty principle for the free particle wave packet in Problem 2.40 and the observable x by calculating oy, 0, and d(x)/dt exactly. Problem 3.46 Show that the energy-time uncertainty principle reduces to the “your name” uncertainty principle (Problem 3.39) when the observable in question is x. FURTHER PROBLEMS FOR CHAPTER 3 4*Problem 3.47 Functions of matrices are defined by their Taylor series expansions; for example, LEM + gM + EMP bo 13.152] (a) Find exp(M), if 013 ; . a) @M=(0 0 ‘): cm =( ) (° 00 #0 (b) Show that if M is diagonalizable, then det (eM) [3.153] (This is actually true even if M is not diagonalizable, but it’s harder to prove in the general case.) (c) Show that if the matrices M and N commute, then eMEN _ MN, [3.154] Prove (with the simplest counterexample you can think up) that Equation 3.154 is nor true, in general, for noncommuting matrices. (d) If His Hermitian, show that e is unitary. +Problem 3.48 A particle of mass m is in the ground state of the infinite square well (Equation 2.15). Suddenly the well expands to twice its original size—the right wall moving from a to 2a—leaving the wave function (momentarily) undisturbed. The energy of the particle is now measured. (a) What is the most probable result? What is the probability of getting that result? (b) What is the next most probable result, and what is its probability?Further Problems for Chapter 3. 117 ({c) What is the expectation value of the energy? (If you find yourself confronted with an infinite series, try another method.) Problem 3.49 A harmonic oscillator is in a state such that a measurement of the energy would yield either (1/2)hw or (3/2)hw, with equal probability. What is the largest possible value of (x) in such a state? If it assumes this maximal value at time 1 = 0, what is W(x, 1)? x«Problem 3.50 Find the matrix elements (n|x|n") and (n|p|n’) in the (orthonormal) basis consisting of stationary states for the harmonic oscillator (here |n) refers to the state Y’,,Eq.2.50). [You already calculated the diagonal elements (n = n')in Problem 2.37; use the same technique for the general case.] Construct the corresponding (infinite) matrices, X and P. Show that (1/2m)P* + (ma/2)X? = His diagonal, in this basis. Are its diagonal elements what you would expect? Partial answer: (alxin’) = f (Vi? un 1 + Viiv 1): [3.155] Imo «««Problem 3.51 Show that w= fo (-22) oa, (3.156) 7p where ©(p, t) is the momentum-space wave function. In general, [¥°O(x,44,1)Wdx, imposition space: POS 5 3.157] (cep) (cena in momentum space, 1571 Hint: Notice that x exp(ipx /h) = —ih(d/dp) exp(ipx /h). «Problem 3.52 Find the momentum-space wave function ,(p, 1) for the nth sta- tionary state of the infinite square well. Construct |®,|? (it’s simplest to write separate formulas for odd and even n). Show that |®,,? is finite at p = -tnth/a. «Problem 3.53 Use Equation 3.148 to show that d av (xp) =2(T) — (@ . BPP) = AUT) — (x) (3.158] where T is the kinetic energy (H = T + V). Ina stationary state the left side is zero (why?), so Y av ar) =e). [3.159] dx This is called the virial theorem. Use it to prove that (7) = (V) for stationary states of the harmonic oscillator, and check that this is consistent with the results you got in Problems 2.14 and 2.37.118 Chap. 3 Formalism Problem 3.54 What would it mean for an observable Q to be conserved, in quan- tum mechanics? At a minimum, the expectation value of Q should be constant in time, for any state W. The criterion for this (assuming Q has no explicit time de- pendence) is that © commute with the Hamiltonian (Equation 3.148). But we'd like something more: The probability |c,|? of getting any particular eigenvalue (A,) of Q should be independent of 1. Show that this, too, is guaranteed by the condition [H, 0] = 0. (Assume that the potential energy is independent of r, but do not assume W is a stationary state.) Hint: Q and H are compatible observables, so they have a complete set of simultaneous eigenvalues. «Problem 3.55 * (a) For a function f(x) that can be expanded in a Taylor series, show that f(x +20) = el?!" F(x) (where Xo is any constant distance). For this reason, /h is called the generator of translations in space. (See Problem 3.47 for the meaning of an operator in the exponent.) b) If W(x, 1) satisfies the (time-dependent) Schrédinger equation, show that W(x, t + to) =e HOW (x, 1) (where f is any constant time); —H /h is called the generator of translations in time. ©) Show that the expectation value of a dynamical variable Q(x, p, 1), at time t + to, can be written (Dery = (YCx, Dlel™ GEE, Bt + oe Mo W(x, 1)) Use this to recover Equation 3.148. Hint: Let t = dt, and expand to first order indt. Problem 3.56 In an interesting version of the energy-time uncertainty principle” At = t/m, where T is the time it takes W(x, t) to evolve into a state orthogo- nal to W(x,0). Test this out, using a wave function that is an equal admixture of two (orthonormal) stationary states of some (arbitrary) potential: W(x,0) = C/V2) Wi) + v2). «Problem 3.57 Dirac proposed to peel apart the bracket notation for an inner prod- uct, (c|B), into two pieces, which he called bra ((|) and ket (|)). The latter is a vector, but what exactly is the former? It’s a linear function of vectors, in the sense that when it hits a vector (to its right) it yields a (complex) number—the inner 2°See Lev Vaidman, Am. J. Phys, 60, 182 (1992) for a proof.Further Problems for Chapter 3 119 product."' (When an operator hits a vector, it delivers another vector; when a bra hits a vector, it delivers a number.) Actually, the collection of all bras constitutes another vector space—the so-called dual space. The license to treat bras as separate entities in their own right allows for some powerful and pretty notation (though I shall not exploit it further in this book). For example, if Ja) is a normalized vector, the operator P= la)al [3.160] picks out the component of any other vector that “lies along” |e): PIB) = (alB)|a); we call it the projection operator onto the one-dimensional subspace spanned by fa). (a) Show that P? = P. Determine the eigenvalues of , and characterize its eigenvectors. (b) Suppose je;) is an orthonormal basis for an n-dimensional vector space. Show that h Viele = (3.161) ial This is the tidiest statement of completeness. (c) Let @ be an operator with a complete set of orthonormal eigenvectors: Ole;) =djley) (f= 1,2,3,..-0). Show that @ can be written in terms of its spectral decomposition: O= DV rle es. (3.1621 a Hint: An operator is characterized by its action on all possible vectors, so what you must show is that Ola) = Siena la). a for any vector |a). 3'n a function space, the bra can be thought of as an instruction to integrate unm fir Idx, with the “hole” {---] waiting to be filled by whatever function the bra encounters next.120 Chap. 3 Formalism +Problem 3.58 Imagine a system in which there are just vo linearly independent States: 1 0 n=(4) and m=(?) The most general state is a normalized linear combination: 1W) =all) +512) = (3): with lal? + Jol? Suppose the Hamiltonian matrix is hg He (: A ). where g and h are real constants. The (time-dependent) Schrédinger equation says d HW) = ih—|W). I) ha) (a) Find the eigenvalues and (normalized) eigenvectors of this Hamiltonian. (b) Suppose the system starts out (at = 0) in state |1). What is the state at time 1? Answer: cos(gt /h) ) in(gt /h) J” Note: This is about the simplest nontrivial quantum system conceivable. It is a crude model for (among other things) neutrino oscillations. In that case |1) represents the electron neutrino, and |2) the muon neutrino; if the Hamiltonian has a nonvanishing off-diagonal term g, then in the course of time the electron neutrino will turn into a muon neutrino, and back again. At present this is highly speculative—there is no experimental evidence for neutrino oscillations; however, a very similar phenomenon does occur in the case of neutral K-mesons (K® and K°)CHAPTER 4 QUANTUM MECHANICS IN THREE DIMENSIONS 4.1 SCHRODINGER EQUATION IN SPHERICAL COORDINATES The generalization to three dimensions is straightforward. Schrédinger’s equation says aw ih— = HW; [4.1 ae [4.1] the Hamiltonian operator! H is obtained from the classical energy 1 1 2 smv +V = — (ph + pt P+ 2m 2 by the standard prescription (applied now to y and z, as well as x): ha ha ha a2) x oa Foy Peo ! Pe ae PY Fay PO Tae or - [4.3] ‘Where confusion might otherwise occur, I have been putting “hats” on operators to distinguish them from the corresponding classical observables. don’t think there will be much occasion for ambiguity in this chapter, and the hats get to be cumbersome, so I am going to leave them off from now on.122 Chap. 4 Quantum Mechanics in Three Dimensions for short. Thus [4.4] where Fee [4.5] aye at tag tan is the Laplacian, in Cartesian coordinates. The potential energy V and the wave function W are now functions of r = (x, y, z) and t. The probability of finding the particle in the infinitesimal volume dy = dx dydx is |W(r, 1)? d°x, and the normalization condition reads fisrae : (4.6) with the integral taken over all space. If the potential is independent of time, there will be a complete set of stationary states, W(t. t) = Varo, where the spatial wave function w;, satisfies the time-independent Schridi tion: = SV + Vn = EnV (48) 2m The general solution to the (time-dependent) Schrédinger equation is VED) = Donte, (491 with the constants c, determined by the initial wave function, W(r, 0), in the usual way. (If the potential admits continuum states, then the sum in Equation 4.9 becomes an integral.) «Problem 4.1 (a) Work out all of the canonical commutation relations for components of the operators r and p: [x, yl, [x, py], bx. pel, [py, pz], and so on. Answer: brn p= lpi =i, Ertl = Ep P= 0. (4.10) (b) Show that fu dt d = +o, and —(p) =(-V/). (4.11) m dt (Each of these. of course, stands for three equations—one for each component.) Hint: Note that Equation 3.148 is valid in three dimensions.Sec. 4.1: Schrédinger Equation in Spherical Coordinates 123 (©) Formulate Heisenberg’s uncertainty principle in three dimensions. Answer: O10, Zh/2, oop, Zh/2, 020, >h/2, (4.121 but there is no restriction on, say, 010, 4.1.1 Separation of Variables Typically, the potential is a function only of the distance from the origin. In that case it is natural to adopt spherical coordinates, (r, 6, ) (see Figure 4.1). In spherical coordinates the Laplacian takes the form? 13/,a 1 3 si oo 1 a (4.13) 2 (pe — (4). « ar Var) * sino 36 "7 96) * ante ag? In spherical coordinates, then, the time-independent Schrédinger equation reads | ELSE) rind (8) aba 2m |r? ar or P sind 00 sin’ @ \ ag? +Vy = Ey. (4.14] We begin by looking for solutions that are separable into products: U(r, 8, @) = R(V)YO, 9). [4.15] Putting this into Equation 4.14, we have re + x noe + Pr se sine dr)” Pind 30 00 +VRY = ERY. Figure 4.1; Spherical coordinates: radius r, polar angle 9, and azimuthal angle . ?1n principle, this can be obtained by change of variables from the Cartesian expression (Equation 4.5). However, there are much more efficient ways of getting it; see, for instance, M. Boas, Mathematical Methods in the Physical Sciences, 2nd ed. (New York: John Wiley and Sons, Inc., 1983) Chapter 10, Section 9.124 Chap. 4 Quantum Mechanics in Three Dimensions Dividing by Y R and multiplying by —2mr?/h?: 1d (,dR\ mr? laa (Pa) -Frwo- gf 18 (og), 1 er] ¥ | sind 36 9" " 30) * sin29 Oe? J — ‘The term in the first curly bracket depends only on r, whereas the remainder depends only on 6 and @; accordingly, each must be a constant, For reasons that will appear in due course, I will write this “separation constant” in the form I(/ + 1)? Rar ar 1 loa oat + sta (sno Y [sin@ a6 00 «Problem 4.2 Use separation of variables in Cartesian coordinates to solve the infinite cubical well (or “particle in a box”): 2 if (732) - Bern - alates, (4.16) | =-104. (4.17) Vo.y.z) = {0 fx y, 2 are all between 0 and a; °%2)* | oo, otherwise. (a) Find the stationary state wave functions and the corresponding energies. (b) Call the distinct energies £1, £2, £3, ..., in order of increasing energy. Find E\, Eo, Ex, Es, Es, and Es. Determine the degeneracy of each of these energies (that is, the number of different states that share the same energy). Recall (Problem 2.42) that degenerate bound states do not occur in one dimension, but they are common in three dimensions. (c) What is the degeneracy of E,4, and why is this case interesting? 4.1.2 The Angular Equation Equation 4.17 determines the dependence of yy on 6 and $; multiplying by Y sin? 4, it becomes oy 2 Y sino ( non) +o = —I(1 + 1) sin’ oY. [4.18] Note that there is no loss of generality here—at this stage / could be any complex number. Later (on we'll discover that / must in fact be an integer, and itis in anticipation of that result that T express the separation constant in a way that looks peculiar now.Sec. 4.1: Schrodinger Equation in Spherical Coordinates 125 You may have encountered this equation already—it occurs in the solution to Laplace’s equation in classical electrodynamics. As always, we try separation of variables: YO, G) = 06) 4). [4.19] Plugging this in, and dividing by @, we find 1f. od /(. do 2 12> {i [snes (seS5.) ] +1 + si of toa 7” The first term is a function only of 8, and the second is a function only of ¢, so each must be a constant. This time I’ll call the separation constant m?:* 1 d/. de 6 5 [sme (sno) +10 + Isin? @ = [4.20] 2, ide __ (421) ode The ¢ equation is easy: —mo > Og) = el"? [4.22] (Actually, there are two solutions: exp(im@) and exp(—im@), but we'll cover the latter by allowing m to run negative. There could also be a constant factor in front, but we might as well absorb that into ©. Incidentally, in electrodynamics we would write the azimuthal function () in terms of sines and cosines, instead of exponentials, because electric potentials must be real. In quantum mechanics there is no such constraint, and the exponentials are a lot easier to work with.] Now, when @ advances by 2, we return to the same point in space (see Figure 4.1), so it is natural to require that® Op + 20) = O(9). [4.23] In other words, explim(g + 2x)] = exp(img), or exp(2im) = 1. From this it follows that m must be an integer: ee 8 1, ee a [4.24] 4Again, there is no loss of generality here since at this stage m could be any complex number; in a moment, though, we will discover that m must in fact be an integer. Beware: The letter m is now doing double duty, as mass and as the so-called magnetic quantum number. There is no graceful way to avoid this since both uses are standard. Some authors now switch to M or 4 for mass, but I hate to change rotation in midstream, and I don’t think confusion will arise as long as you are aware of the problem, SThis is a more subtle point than it looks. Afler all, the probability density (||?) is single valued regardless of m. In Section 4.3 we'll obtain the condition on m by an entirely different—and more compelling —argument.Chap. 4 Quantum Mechanics in Three Dimensions The 6 equation, _ dO 2 2 - an _ = 4. sin O55 (sine) +[(1 + Dsin® 6 — m*]@ = 0, [4.25] may not be so familiar. The solution is (0) = AP;" (cos 6), (4.26) where Pi” is the associated Legendre function, defined by° im Pra) syn? (4) Pi(x), 4.27] and P;(x) is the [th Legendre polynomial. We encountered the latter (Equation 3.91) as orthogonal polynomials on the interval (—1, +1); for our present purposes it is more convenient to define them by the Rodrigues formula: 1 x) w-1! [4.28} For example, Po(x) = 1, Pie) = soc" -l)=x, 1 PWD= lz ) Ge? - 1? = 408 -1, and so on. The first few Legendre polynomials were listed in Table 3.1. As the name suggests, P;(x) is a polynomial (of degree /) in x, and is even or odd according to the parity of J. But P(x) is not, in general, a polynomial—if m is odd it carries a factor of V1 — Poa) = 462-1), Pay =U “ [Foe -p]=a0/1 B, #) [Sor] =s0-x, etc, [On the other hand, what we need is Pi" (cos 6), and VT —cos*@ = sind, so PJ" (cos@) is always a polynomial in cos 8, multiplied—if m is odd—by sin@. Some associated Legendre functions of cos 6 are listed in Table 4.1.] P2(x) = (1 — 2x") — SNotice that P-" = P!". Some authors adopt a different sign convention for negative values of 1m; see Boas (footnote 2) p. 505,Sec. 4.1: Schrédinger Equation in Spherical Coordinates 127 Table 4.1 Some associated Legendre functions, P’"(cos#). Pl = sin8 P3 = 15 sin(1 — cos? 4) PP =cos8 =3sin’6 3 sin @(5 cos? @ — 1) P} =3sin6 cos@ 4(5cos? — 3cos6) PS = $(3c0s?6 ~ 1) Notice that / must be a nonnegative integer for the Rodrigues formula to make any sense; moreover, if |m| > 1, then Equation 4.27 says P’" = 0. For any given /, then, there are (2/ + 1) possible values of m: 0, 1,2,. L—-th..., “LO bd Ll. [4.29] But wait! Equation 4.25 is a second-order differential equation: It should have nvo. linearly independent solutions, for any old values of | and m. Where are all the other solutions? Answer: They exist, of course, as mathematical solutions to the equation, but they are physically unacceptable because they blow up at @ = 0 and/or @ = x7, and do not yield normalizable wave functions (see Problem 4.4). Now, the volume element in spherical coordinates’ is ar =r* sind dr dé do, [4.30] so the normalization condition (Equation 4.6) becomes | IWPr? sin6 dr dé de = | iaPrear f IY? sin@ d@ dg = 1. It is convenient to normalize R and Y individually: f IRPr2 dr and f [ IYP sino dé de A Ib Jy ‘The normalized angular wave functions’ are called spherical harmonics: [4.31] 7See, for instance, Boas, (footnote 2), Chapter 5, Section 4. The normaiization factor is derived in Problem 4.47. The ¢ factor is chosen for consistency with the notation we will be using in the theory of angular momentum; itis reasonably standard, though some colder books use other conventions. Notice that yp" = (1 YP128 Chap. 4 Quantum Mechanics in Three Dimensions 1 2 _ (15 ou (s ei (P % (=) % (sx 3\12 of 1\! 2) cos You (a 12 a =) sin @e®!® yelax (&) sin 6(5 cos? @ — 1ye*!# 5 \u2 105 #=(%) Beos?# — 1) w= (se 3 2 5 \ 12 (2) sin@ cos ée*!* Wes(e it [4.32] Y"G,0) = where € = (—1)” for m > 0 and € = | for m <0. As we shall prove later on, they are automatically orthogonal, so an px f f LY", @)ITLYR” @. @)] sin 6 dO do = 813m. [4.33] lo Jo In Table 4.2 I have listed the first few spherical harmonics. xProblem 4.3 Use Equations 4.27, 4.28, and 4.32 to construct Yj) and ¥3. Check that they are normalized and orthogonal. Problem 4.4 Show that ©@) = A lnftan(6/2)] satisfies the @ equation (Equation 4.25) for / = m = 0. This is the unacceptable “second solution” —what’s wrong with it? Problem 4.5 Using Equation 4.32, find Y/(@, 6) and Y}(0, @). Check that they sat- isfy the angular equation (Equation 4.18), for the appropriate values of the parameters Jand m. «Problem 4.6 Starting from the Rodrigues formula, derive the orthonormality con- dition for Legendre polynomials: ' 2 ff Peoreods= (za) by. (4.34) Hint: Use integration by parts.Sec. 4.1: Schrédinger Equation in Spherical Coordinates 129 4.1.3 The Radial Equation Notice that the angular partof the wave function, ¥ (8, ), is the same forall spherically symmetric potentials; the actual shape of the potential, V(r), affects only the radial part of the wave function, R(r), which is determined by Equation 4.16: dR a 2mr? aC #) a (V(r) — EJR=1+ DR. [4.35] This equation simplifies if we change variables: Let u(r) =rRr), [4.36] so that R =u/r, dR/dr = [r(du/dr) — ul/r?, (d/dr)[r2(dR/dr)| = rd?u/dr?, and hence [4.37] © 2m dr? 2 W du v wid+)) 2m This is called the radial equation’, it is identical in form to the one-dimensional Schrddinger equation (Equation 2.4), except that the effective potential, wId+t 2m Vert = V , [4.38] contains an extra piece, the so-called centrifugal term, (h7/2m)[I(1+1)/r?]. Ittends to throw the particle outward (away from the origin). just like the centrifugal (pseudo-) force in classical mechanics. Meanwhile, the normalization condition (Equation 4.31) becomes | lu? dr =1. [4.39] fo ‘We cannot proceed further until a specific potential is provided. Example. Consider the infinite spherical well, ifr a, (4.40) 0, vin ={o o { . Outside the well the wave function is zero; inside the well the radial equation says au (‘Ss Lh =F] 4 (441) dr? P Those m’s are masses, of course—the radial equation makes no reference to the quantum ‘number m.130 Chap. 4 Quantum Mechanics in Three Dimensions where VImE = , [4.42] as usual. Our problem is to solve this equation, subject to the boundary condition u(a) = 0. The case / = 0 is easy: k @u : qe TR ulr) = Asin(kr) + Booster). But remember, the actual radial wave function is R(r) = u(r)/r, and [cos(kr)]/r blows up as r -> 0. So"? we must choose B = 0. The boundary condition then requires sin(ka) = 0, and hence ka = nz, for some integer n. The allowed energies are evidently (a =1,2,3,..), [4.43] the same as for the one-dimensional infinite square well (Equation 2.23). Normalizing u(r) yields A = ,/2/a; inclusion of the angular part (constant, in this instance, since YS, 6) = 1/47), we conclude that 1 sin@ar/a) Yoon = Ta ; [4.44] [Notice that the stationary states are labeled by rhvee quantum numbers, n, /, and m: Vnin (7, 8, $). The energy, however, depends only on and /: E,.] The general solution to Equation 4.41 (for an arbitrary integer /) is not so familiar: u(r) = Arj(kr) + Brny(kr), [4.45] where ji(x) is the spherical Bessel function of order /, and n,(x) is the spherical Neumann function of order /. They are defined as follows: i —: mon =—-v) (4) SOS* 14.46) xdx) x For example, : sinx cos Joe) = ees nox) = -S, x . ld /sinx sinx cosx AG) = (-*)——— (=) -= = xdx x ey Actually, all we require is that the wave function be normalizable, not that itbe finite: R(r) ~ A /r at the origin would be normalizable (because of the +? in Equation 4.31). For a more compelling proof that B = 0, see R. Shankar, Principles of Quantum Mechanics (New York: Plenum, 1980), p. 351Sec. 4,1: Schrédinger Equation in Spherical Coordinates 131 x x x 1d cosx cosx _ sinx mo) = (9a (SS) =- and so on. The first few spherical Bessel and Neumann functions are listed in Table 4.3. Notice that for small x (where sinx © x — x3/3! + x5/5!—--- and cosx &1—2x7/24x4/41—- ++), 1 x AG) *® Zi 1 Joe) © 1; nox) © mG)™ 2: x etc, The point is that the Bessel functions are finite at the origin, but the Neumann functions blow up at the origin. Accordingly, we must have B; = 0, and hence R(r) = Ajilkr). [4.47] There remains the boundary condition, R(a) = 0. Evidently k must be chosen such that jilka) = 0; [4.48] that is, (ka) is a zero of the [""-order spherical Bessel function. Now the Bessel functions are oscillatory (see Figure 4.2); each one has an infinite number of zeros. But (unfortunately, for us) they are not located at nice sensible points (such as m, or nz, or something); they have to be computed numerically!" At any rate, the boundary condition requires that k= 1B, [4.49] a Table 43: The first few spherical Bessel and Neumann functions, jj(x) and m(x). sine cosx cosx _ sinx x Ql-n po for x«1 yr "Abramowitz and Stegun, eds., Handbook of Mathematical Functions (New York: Dover, 1965), Chapter 10, provides an extensive isting.132 Chap. 4 Quantum Mechanics in Three Dimensions Figure 4.2: Graphs of the first four spherical Bessel functions. where fy, is the n'® zero of the /"* spherical Bessel function. The allowed energies, then, are given by Ent = i” Be [4.50] al = Fa Pal? and the wave functions are rtm (1, 9.) = Ant ji(Buit/a)¥/" (8,9), [4.51] with the constant A,, to be determined by normalization. Each energy level is (2/+1)- fold degenerate, since there are (2/ + 1) different values of m for each value of / (see Equation 4.29). Problem 4.7 (a) From the definitions (Equation 4.46), construct jo(x) and n(x). (b) Expand the sines and cosines to obtain approximate formulas for jo(x) and ny(x), valid when x < 1. Confirm that /(x) is finite at the origin but n(x) blows up. Problem 4.8 (a) Check that 4r-/;(kr) satisfies the radial equation (Equation 4.37) with V(r) = 0 and] = 1.Sec. 4.2: The Hydrogen Atom — 133 (b) Determine graphically the allowed energies for the infinite spherical well when T= 1, Show that for large n, Ens © (h?x?/2ma?)(n + 1/2)*. «Problem 4.9 A particle of mass m is placed in a finite spherical well: 0, ifr 0), describing electron-proton scattering, as well as discrete bound states, representing the hydrogen atom, but we shall confine our attention to the latter. (electron) +e, (proton) Figure 4.3: The hydrogen atom.134 Chap. 4 Quantum Mechanics in Three Dimensions 4.2.1 The Radial Wave Function Our first task is to tidy up the notation. Let [4.54] (For bound states, E < 0, so « is real.) Dividing Equation 4.53 by E, we have du _f,__me? 1 d+) «2 dr? 2megh?a (er) (kr)? This suggests that we let 2 me » and py = —o, 4.55; BO eohK (551 so that @ uw po, +1) out) _ 4.56 dp? [ pp dM 1456] Next we examine the asymptotic form of the solutions. Asp -» oo, the constant term in the brackets dominates, so (approximately) au ou iy. dp? The general solution is u(p) = Ae? + Be’, [4.57] but e? blows up (as p > 00), so B = 0. Evidently, u(p) ~ dew? [4.58] for large p. On the other hand, as p > 0 the centrifugal term dominates”; approxi- mately, then, @u_ Ul+1) Sa. dp* p ‘The general solution (check it!) is u(p) = Cp'*! + Dp“, but p~! blows up (as p > 0), so D = 0. Thus u(p) ~ Cp't! [4.59] "This argument does not apply when J = 0 (although the conclusion, Equation 4.59, is in fact valid for that case too). But never mind: All [am trying to do is provide some motivation for a change of variables (Equation 4.60.)Sec. 4.2: The Hydrogen Atom 135 for small p. The next step is to peel off the asymptotic behavior, introducing the new function u(p): u(p) = p'*tePv(p), [4.60] in the hope that v(p) will turn out to be simpler than u(p). The first indications are not auspicious: di di ot x ple’ [a+ 1 ~ p+ 0 dp dq, and @u +0) dv oe _ glee || 2) -2 A lee 2dt 1p) 4055 wn? {{ +p+— vt 2+ Gp taps In terms of v(p), then, the radial equation (Equation 4.56) reads ay d poe 420 41-9) + [oy — 2+ Jv =0. (4.61) ap dp Finally, we assume the solution, v(p), can be expressed as a power series in p: v(p) = > ajp’. [4.62] rad Our problem is to determine the coefficients (ao, a1, a2, ...). Differentiating term by term, } a > iaje = v+ Iaj+1p?- = = [In the second summation I have renamed the “dummy index”: j > j + 1. If this troubles you, write out the first few terms explicitly, and check it. You might say that the sum should now begin at j = —1, but the factor (j + 1) kills that term anyway, so we might as well start at zero.] Differentiating again, ey av & Wet It Daisiel p = Inserting these into Equation 4.61, we have DAG + Daye! +204 D OG + Dayo? i 0 -2 > jayp! +o — 20+ DI Yap! = jo136 Chap. 4 Quantum Mechanics in Three Dimensions Equating the coefficients of like powers yields IG + Vain + 20+ DG + Days — 2fay + [oo - 20+ Dla; = [4.63} _ { 2 +1 +) — po LG 4DG 4242) This recursion formula determines the coefficients, and hence the function v(p): We start with ag = A (this becomes an overall constant, to be fixed eventually by normalization), and Equation 4,63 gives us a1; putting this back in, we obtain az, and so on." Now let’s see what the coefficients look like for large j (this corresponds to large p, where the higher powers dominate). In this regime the recursion formula says ays ay, so [4.64] Suppose for a moment that this were the exact result. Then oo 9 v(p) =A » = pi = Ae’, = i and hence u(p) = Ap'tler, [4.65] which blows up at large p. The positive exponential is precisely the asymptotic behav- ior we didn't want in Equation 4.57. (It’s no accident that it reappears here; after all. it does represent the asymptotic form of some solutions to the radial equation—they just don’t happen to be the ones we're interested in, because they aren’t normalizable.) There is only one way out of this dilemma: The series must terminate. There must occur some maximal integer, jax, such that Gog 41 =O [4.66] (and beyond which all coefficients vanish automatically). Evidently (Equation 4.63) 2Cimax + 1+ 1) — po = 0. You might wonder why I didn’t use the series method directly on u(o)—why factor out the asymptotic behavior before applying this procedure? The reason for peeling off p'*" is largely aesthetic: Without this, the sequence would begin with a long string of zeroes (the first nonzero coefficient being 414.1); by factoring out p!*! we obtain a series that starts out with p®, The e”? factor is more critical—it you don’t pull that out, you get a three-term recursion formula involving aj4.2, aj41, and aj; (try it!), and that is enormously more difficult to work with.Sec, 4.2: The Hydrogen Atom 137 Defining xtl+l [4.67] (the so-called principal quantum number), we have po = 2n. 14.68] But o determines £ (Equations 4.54 and 4.55): E [4.69] so the allowed energies are [4.70] This is the famous Bohr formula—by any measure the most important result in all of quantum mechanics. Bohr obtained it in 1913 by a serendipitous mixture of inapplicable classical physics and premature quantum theory (the Schrédinger equation did not come until 1924). Combining Equations 4.55 and 4.68, we find that 1 «=(,)-= [4.71] 4xeh?) na where dnegh? Seah = 0529 x 10°! m (4.72) is the so-called Bohr radius. It follows (again, from Equation 4.55) that p=2 (4.73) an Evidently the spatial wave functions for hydrogen are labeled by three quantum num- bers (7, /, and m): Ynim(r, 9, &) = Rul) Yi" (8,9), [4.74] where (referring back to Equations 4.36 and 4.60) Rul) = Lple-Pvip), [4.75]138 Chap. 4 Quantum Mechanics in Three Dimensions and v(p) is a polynomial of degree jmax = — 1 — 1 in p, whose coefficients are determined (up to an overall normalization factor) by the recursion formula _ tl +1—-n) OS GANG Fda (476) The ground state (that is, the state of lowest energy) is the case n = 1; putting in the accepted values for the physical constants, we get ne-[ 2, (2) ]=-16ev 477 me" | on? ia) ee ar Evidently the binding energy of hydrogen (the amount of energy you would have to impart to the electron in order to ionize the atom) is 13.6 eV. Equation 4.67 forces 1 =0, whence also m = 0 (see Equation 4.29), so Yioolr, 8,0) = Riolr)¥9@, ). [4.78] ‘The recursion formula truncates after the first term (Equation 4.76 with j = 0 yields a = 0), So v(p) is a constant (ao) and Rio) = we ria. [4.79] Normalizing it, in accordance with Equation 4.31, ~ 2 lao? f° ayia 2@ [Riolr? dr = —5- fe"? dr = lag’ = = 1, 0 @ So 4 50 dy = 2/,/a. Meanwhile, ¥9 = 1/V4zr, so Wroolr 8,6) = [4.80] Ifn =2 the energy is 13. Ey Bey = -34eV; (481) this is the first excited state—or rather, states, since we can have either ] = 0 (in which case m = 0) or / = 1 (with m = —1, 0, or +1), so there are actually four different states that share this energy. If / = 0, the recursion relation (Equation 4.76) gives a = —ao (using j = 0), and ay = 0 (using j = 1),Sec. 4.2: The Hydrogen Atom 139 so u(p) = ao(1 — p), and hence Raotr) = 3 ( (4.82) If / = | the recursion formula terminates the series after a single term, so v(p) is a constant, and we find Ru(r) = aGre [4.83] (In each case the constant dp is to be determined by normalization—see Problem 4.11) For arbitrary n, the possible values of / (consistent with Equation 4.67) are 1=0,1,2,....2-1. [4.84] For each 1, there are (2/ + 1) possible values of m (Equation 4.29), so the total degeneracy of the energy level Ey is [4.85] nt dn) = alt N= i= The polynomial v(p) (defined by the recursion formula, Equation 4.76) is a function well known to applied mathematicians; apart from normalization, it can be written as v(p) = Lit! (2p), [4.86] where ay? Lo p(X) =(-1)? () Lg (x) [4.87] is an associated Laguerre polynomial, and Lge) =e" (4) (ex) (4.88) is the qth Laguerre polynomial." (The first few Laguerre polynomials are listed in Table 4.4; some associated Laguerre polynomials are given in Table 4.5. The first few radial wave functions are listed in Table 4.6 and plotted in Figure 4.4.) The normalized hydrogen wave functions are'* Toy : ; Vaim= (=) poem) ot (Z)are.e. (4391 J na) 2nl(n+D na 4s usual, there are rival normalization conventions in the literature; Lhaye adopted the most nearly standard one. 1SIF you want to see how the normalization factor is calculated, study (for example), L. Schiff, Quantum Mechanies, 2nd ed. (New York: McGraw-Hill, 1968), page 93140 Chap. 4 Quantum Mechanics in Three Dimensions Table 4.4: The first few Laguerre polynomials, L,(x). Lo= Ly=-x41 —4x +2 La La= x34 9x? — 18x +6 — 16x3 4 72x? — 96x + 24 Ls = —x5 +25x* — 200 + 60x? — 600x + 120 Ls = 36x5 + 450x4 — 2400x3 + 5400x? ~ 4320x + 720 Table 4.5: Some associated Laguerre polynomials, £7 _,(x). a2 Li = 6x +18 L} = 12x? — 96x + 144 13 =6 1} = 24x +96 Lb = 3x? — 18x +18 L3 = 60x? — 600x + 1200 They are not pretty, but don’t complain—this is one of the very few realistic systems that can be solved at all, in exact closed form. As we will prove later on, they are mutually orthogonal: f Vim Vnrrm 0? sin dr dO db = SnyS10 8mm» [4.90] xProblem 4.10 Work out the radial wave functions R3o, R31, and R32, using the Tecursion formula (Equation 4.76). Don’t bother to normalize them. +Problem 4.11 (a) Normalize Roo (Equation 4.82), and construct the function ¥299. (b) Normalize Rp; (Equation 4.83), and construct W211, W210, and W2}-1. #Problem 4.12 (a) Using Equation 4.88, work out the first four Laguerre polynomials. (b) Using Equations 4.86, 4.87, and 4.88, find v(p) for the case n = 5,/ = 2.Sec. 4.2: The Hydrogen Atom 141 The first few radial wave functions for hydrogen, R,)(r). Rig = 2a~*? exp(—r/a) f= a? (1-5 Jexperp2a) ta = aE exo onan +5 (2) exp(~r/3a) (0-85) Garcvan Re get (1) ewer +§(5) a (5) Jerre v3 lr agry\r Ry = ca (bE a() P exp (ory =e azo ag (5) |S epcrisay 1 Lry¢ry? Ro= a? ( -s ‘) (5) =r/4a 2= OF T= 5 a) (5) excise 1 (ey) Ro= V2 (2) exp(—r/4a’ 8 T6875 a) oP cri4a) Figure 4.4: Graphs of the first few hydrogen radial wave functions, Ryi(r).142 Chap. 4 Quantum Mechanics in Three Dimensions (©) Again, find v(p) for the case n = 5,/ = 2, but this time get it from the recursion formula (Equation 4.76). xProblem 4.13 (a) Find (r) and (7?) for an electron in the ground state of hydrogen. Express your answers in terms of the Bohr radius a. (b) Find (x) and (x?) for an electron in the ground state of hydrogen. Hint: This requires no new integration—note that r? = x? + y? + 2°, and exploit the symmetry of the ground state. (©) Find (x?) in the state n = 2,/ = 1, m = 1. Hint: This state is not symmetrical inx, y,z. Use x =rsin@ cos. Problem 4.14 What is the probability that an electron in the ground state of hy- drogen will be found inside the nucleus? (a) First calculate the exact answer, assuming that the wave function (Equation 4.80) is correct all the way down to r = 0, Let b be the radius of the nucleus. (b) Expand your result as a power series in the small number ¢ = 2b/a, and show that the lowest-order term is the cubic: P ~ (4/3)(b/a)*. This should be a suitable approximation, provided that b < a (which it is). (c) Alternatively, we might assume that ¥/(r) is essentially constant over the (tiny) volume of the nucleus, so that P ~ (4/3)zrb*|y/(0)|?. Check that you get the same answer this way. (d) Use b © 10-'5m and a * 0.5 x 107!°m to get a numerical estimate for P. Roughly speaking, this represents the “fraction of its time that the electron spends inside the nucleus”. Problem 4.15 (a) Use the recursion formula (Equation 4.76) to confirm that when / = n — 1 the radial wave function takes the form Ruen-t) = Nar beri", and determine the normalization constant N,, by direct integration. (b) Calculate (r) and (72) for states of the form Wann (©) Show that o, = (r)//2n-+ 1 for such states, Note that the fractional spread in r decreases with increasing n (in this sense the system “begins to look classical” for large n). Sketch the radial wave functions for several values of n to illustrate this point.Sec. 4.2: The Hydrogen Atom 143 4.2.2 The Spectrum of Hydrogen In principle, if you put a hydrogen atom into some stationary state W,in, it should stay there forever. However, if you tickle it slightly (by collision with another atom, say, or by shining light on it), then the atom may undergo a transition to some other stationary state—either by absorbing energy and moving up to a higher-energy state, or by giving off energy (typically in the form of electromagnetic radiation) and moving down.'* In practice such perturbations are always present; transitions (or, as they are sometimes called, “quantum jumps”) are constantly occurring, and the result is that a container of hydrogen gives off light (photons), whose energy corresponds to the difference in energy between the initial and final states: [4.91] Now, according to the Planck formula,” the energy of a photon is proportional to its frequency: (4.92) (4.93) where , ro, (£ y 1.097 x 1077 [4.94] = =) 21.097 x 107m, Ach? \4reo R is knownas the Rydberg constant, and Equation 4.93 is the Rydberg formula for the spectrum of hydrogen. It was discovered empirically in the nineteenth century, and the greatest triumph of Boht’s theory was its ability to account for this result— and to calculate R in terms of the fundamental constants of nature. Transitions to the ground state (ny = 1) lie in the ultraviolet; they are known to spectroscopists as the Lyman series. Transitions to the first excited state (n y = 2) fall in the visible region; they constitute the Balmer series, Transitions to n, = 3 (the Paschen series) are in the infrared, and so on (see Figure 4.5). (Atroom temperature, most hydrogen atoms are in the ground state; to obtain the emission spectrum, you must first pump them up into the various excited states; typically this is done by passing an electric spark through the gas.) "By its nature, this involves a time-dependent interaction, and the details will have to wait for Chapter 9; for our present purposes the actual mechanism involved is immater '7The photon is a quantum of electromagnetic radiation; it’s a relativistic object if there ever was one, and therefore outside the scope of nonrelativistic quantum mechanics. It will be useful in afew places to speak of photons and to invoke the Planck formula for their energy, but please bear in mind that this is external tothe theory we are developing,144 Chap. 4 Quantum Mechanics in Three Dimensions 10 tt -20 Paschen series 0 Balmer series Energy (eV) i 2 -90 -100 -110 120 -0F |} ~14.0/ “Lyman series Figure 4.5: Energy levels and transitions in the spectrum of hydrogen. Problem 4.16 Consider the earth-sun system as a gravitational analog to the hy- drogen atom. (a) What is the potential energy function (replacing Equation 4.52)? (Let m be the mass of the earth and M the mass of the sun.) (b) What is the “Bohr radius” for this system? Work out the actual numerical value. (c) Write down the gravitational “Bohr formula”, and, by equating £,, to the clas- sical energy of a planet in a circular orbit of radius 7, show that n = 79/4, From this, estimate the quantum number n of the earth. (A) Suppose the earth made a transition to the next lower level (n — 1). How much energy (in Joules) would be released? What would the wavelength of the emitted photon (or, more likely, graviton) be? xProblem 4.17 A hydrogenic atom consists of a single electron orbiting a nucleus with Z protons. (Z = I would be hydrogen itself, Z = 2 is ionized helium, Z = 3 is doubly ionized lithium, and so on.) Determine the Bohr energies £,(Z), the binding energy E(Z), the Bohr radius a(Z), and the Rydberg constant R(Z) forahydrogenic atom. (Express your answers as appropriate multiples of the hydrogen values.) Where in the electromagnetic spectrum would the Lyman series fall, for Z = 2 and Z = 3?Sec. 4.3: Angular Momentum — 145 4.3, ANGULAR MOMENTUM In classical mechanics, the angular momentum of particle (with respect to the origin) is given by the formula L=rxp, 14.95) which is to say, Lx = ype 2Py, Ly =2px—xpz, and Lz=xpy—ypx. 14.96} The corresponding quantum operators are obtained by the standard prescription (Equation 4.2): [4.97] ee “ay ox)" In the following sections we will deduce the eigenvalues and eigenfunctions of these operators. 4.3.1 Eigenvalues L, and L, do not commute; in fact [providing a test function, f(x, y, z), for them to act upon]: [Le Lf ) ye vf iba vs ox oF azax az” ayax ayaz ey, 2 oF ~*Vaxaz +7 axay *146 Chap. 4 Quantum Mechanics in Three Dimensions All the terms cancel in pairs (by virtue of the equality of cross-derivatives) except two: (4.98: By cyclic permutation of the indices it follows also that [Ly,L.])=ihL, and [L., Ly] =ihLy [4.99} From these fundamental commutation relations the entire theory of angular momen- tum can be deduced. Evidently L,, Ly, and L, are incompatible observables. According to the generalized uncertainty principle (Equation 3.139), \ 2 07,97, > (j0m2.) h 91,01, > 5I(Lz)l- [4.100] It would therefore be futile to look for states that are simultaneously eigenfunctions of Ly and of Ly. On the other hand, the square of the fotal angular momentum, Psel+iV+ [4.101] does commute with Lx: (2b) = 2b) +13, b+ 102, 2 LylLy, elt (ly Lally + belly, Lal + Ue byl Ly(-ihL.) + (iL )Ly + LeGhLy) + (iL y)L; 0. (1 used Equation 3.142 and the fact that any operator commutes with itself.'$) It follows, of course, that L? also commutes with L, and L.: (7, L.J=0, (L,Ly)=0, [L?,L,1=0, [4.102] ounter in quantum mechanics (see footnote 8, Chapter 1) are + Ag, and therefore distributive with respect to addition A, BI +(4, Cl 'SNote that all the operators we er linear. in the sense that A(f + g) = AB + €) = AB + AC. In particular, (4, B + C1Sec. 4.3: Angular Momentum — 147 or, more compactly, [L?,L] =0. [4.103] So L? is compatible with each component of L, and we can hope to find simultaneous eigenstates of L? and (say) Lz: VP faaf and Lif =uf. [4.104] We'll use a “ladder operator” technique, very similar to the one we applied to the harmonic oscillator back in Section 2.3.1. Let Z. =i, 4iL,. [4.105] Its commutator with L, is (Zz, La] = (Lz, Lal 4 i[Lz, Ly] = ihL, ti(-ihL,) = th(Ly £iLy), [Z,, La] = thle. (4.106) And, of course, (L?, La] =0. [4.107] Iclaim that if f is an eigenfunction of L? and L., so also is Lf. For Equation 4.107 says DLs f) = LiL’ f) = LsOf) =MLaf), [4.108] so La f is an eigenfunction of L?, with the same eigenvalue 4, and Equation 4.106 says Liha f) = (Lbs — Lal) f+lsl.f =thlaftlatuf) [4.109] =MHENLsf, , so Lx f is an eigenfunction of L, with the new eigenvalue wh. Ly is called the “raising” operator because it increases the eigenvalue of L, by h, and L is called the “lowering” operator because it Jowers the eigenvalue by h. Fora given value of A, then, we obtain a “ladder” of states, with each “rung” sep- arated from its neighbors by one unit of % in the eigenvalue of L, (see Figure 4.6). To ascend the ladder we apply the raising operator, and to descend, the lowering operator. But this process cannot go on forever: Eventually we're going to reach a state for which the z-component exceeds the total, and that cannot be (see Problem 4.18). So there must exist a “top rung,” f, such that! Lif,=0. [4.110] 'actually, all we can conclude is that Lf; is not normalizable—its norm could be infinite, instead of zero. Problem 4.19 eliminates this alternative.148 Chap. 4 Quantum Mechanics in Three Dimensions Mn h u +n uy n+2h}—————J eh Ls " f oa ug un2h ey u3h as L in y, Figure 4.6: The “ladder” of angular momentum states. Let il be the eigenvalue of L, at this top rung (the appropriateness of the letter 1—sometimes called the azimuthal quantum number—will appear in a moment): Lifi=hlfs Vf = fr (4.111) = (Ly £iLy)(Ly FiLy) = L + LF i(LyLy — LyLx) =L?—L? x i(ihL,), or, putting it the other way around, LP=Lel,+L2 hl. [4.112] It follows that Df =(L-Ly +L +h f= O+WP +H Df, =WIL+ Df and hence A=W +). [4.113] This tells us the eigenvalue of L? in terms of the maximum eigenvalue of L;. Meanwhile, there is also (for the same reason) a bottom rung, fy, such that L_f, =0. [4.114] Let Ail be the eigenvalue of L, at this bottom rung: Lifs=hl fy; L’ fy =dfo- [4.115] Using Equation 4.112, we have L? fy = (L4L-+ 2 -AL) fy = O+ WP - WD) fy =WIT- VD fr.Sec. 4.3: Angular Momentum — 149 and therefore _ A=WIT 1). [4.116} Comparing Equations 4.113 and 4.116, we see that /(/ + 1) = I(Z — 1), so either 1 =1+ 1 (which is absurd—the bottom rung is higher than the top rung!), or else [4.117] Evidently the eigenvalues of L; are mf, where m (the appropriateness of this letter will also be clear in a moment) goes from —I to +! in N integer steps. In particular, it follows that / = —/+ N, and hence 1 = N/2, so ! must be an integer or a half-integer. The eigenfunctions are characterized by the numbers ! and m: Lyf =wIG+ SM; Leff! =hmf, (4.118) where , 1/2, 1, 3/2, For a given value of /, there are 2 + 1 different values of m the “ladder”). Thope you're impressed: By purely algebraic means, starting with the fun- damental commutation relations (Equations 4.98 and 4.99), we have determined the eigenvalues of L? and L,—without ever seeing the eigenfunctions themselves! We tum now to the problem of constructing the eigenfunctions, but I should warn you that this is a much messier business. Just so you know where we’re headed, I'll tell you the punch line before we begin: 7" = ¥/"—the eigenfunctions of L? and L, are nothing but the old spherical harmonics, which we came upon by a quite different route in Section 4.1.2 (that’s why I chose the letters / and m, of course). -Ith.f- LL 4.119) .e., 21 + 1 “rungs” on Problem 4.18 (a) Prove that if f is simultaneously an eigenfunction of L? and of L; (Equa- tion 4.104), the square of the eigenvalue of L, cannot exceed the eigenvalue of L?. Hint: Examine the expectation value of L?. (b) Asittums out (see Equations 4.118 and 4.119), the square of the eigenvalue of L, never even equals the eigenvalue of L? (except in the special case | = m = 0). Comment on the implications of this result. Show that it is enforced by the uncertainty principle (Equation 4.100), and explain how the special case gets away with it. +Problem 4.19 The raising and lowering operators change the value of m by one unit: Laff = (AP) fr", [4.120] where 47" is some constant, Question: What is A", if the eigenfunctions are to be normalized? Hint; First show that Lz is the Hermitian conjugate of Ls (since Ly150 Chap. 4 Quantum Mechanics in Three Dimensions and L, are observables, you may assume they are Hermitian, but prove it if you like then use Equation 4.112. Answer: A” =hyi( +1) —m(m= 1). (4.121: Note what happens at the top and bottom of the ladder. xProblem 4.20 (a) Starting with the canonical commutation relations for position and momentum. Equation 4.10, work out the following commutators: [L.,x] = ihy, [L., y] = —ihx, [L.,z]=0 [Lz Pxl=ihpy, (Le, py) = —ihpe, (Lz, pe] = 0. (b)_ Use these results to obtain (Lz, Lx] = ih Ly directly from Equation 4.96. (c) Evaluate the commutators [L,,r?] and [L., p*] (where, of course, r? = x4 P +2 and p? = p? + p+ p2). (d) Show that the Hamiltonian H = (p?/2m) + V commutes with all three com- ponents of L, provided that V depends only on r. (Thus H, L?, and L; are mutually compatible observables.) [4.122 x«Problem 4.21 (a) Prove that for a particle in a potential V (r) the rate of change of the expectation value of the orbital angular momentum L is equal to the expectation value of the torque: d a =(N), where =rx(—VVP), (This is the rotational analog to Ehrenfest’s theorem.) (b) Show that d(L) /dt = 0 for any spherically symmetric potential. (This is one form of the quantum statement of conservation of angular momentum.) 4.3.2 Eigenfunctions First of all we need to rewrite L,, L,, and L, in spherical coordinates. Now L = (i/i)(r x V), and the gradient, in spherical coordinates, is” 9410, 1 8 vari spt yg 2, ‘ee 123 ar 20 {4.123] George Arfken, Mathematical Methods for Physicists, 3rd ed. (Orlando, FL: Academic Press. 1985), Section 2.5,Sec. 4.3: Angular Momentum 151 meanwhile, r = rf, so h L rex sexdei rex 2 ar 9g * © * Ona og | i But (f x #) = 0, (F x 6) = ¢, and (F x 6) = —6 (see Figure 4.1), and hence h(a eatiee The unit vectors @ and ¢ can be resolved into their Cartesian components: 1a wa) [4.124] 6 = (cos cos $)f + (cos 8 sing) — (sink; [4.125] = —(sin @)i + (cos $)j. (4.126) sin gf + eos) = (cos 8 cos gi + cos 6 sing] — sine 2 Evidently, -sinos = core). (4.127) [4.128] and [4.129] We shall also need the raising and lowering operators: h La = Ly £ily . a oo a [(- sing seo 3 ~ (cong isin year]. But cos ising = e*'*, so La = the** (a tice wa) [4.130] We are now in a position to determine f7"(, @) (I'll drop the subscript and superscript for now). It's an eigenfunction of L., with eigenvalue him: haf L,f= 72 =h S= 79g she so f= gaye". [4.131]152 Chap. 4 Quantum Mechanics in Three Dimensions [Here g(6) is a constant of integration, as far as ¢ is concerned, but it can still depend on 9.) And f is also an eigenfunction of L? (which we'll write in terms of L.. and L,, using Equation 4.112), with eigenvalue #7/(/ + 1): | Pf=(L.Lo+L2-hl yf af af ef war ag? 7 ae a =he® (— =) (ae) (2 he’ (so ico es) ( he Ie ieot are) - =WIL+ Vf. But in view of Equation 4.131, 9//30 = e!™*dg/d0 and 3f/3 = ime'™*g, so _et (5 +i ott 5) (ein (3 +mgcot?) +m? gei"® — mge'™ d 7 em[-4 (4 & g tmecotd + (m ~ 1) cord 28 + mgcotd +m(m ~ 1g] = 10+ Degen Canceling e™*, dg ~~ m= cot) + mgesc” 6 + (m — Deora etm — D0 +cot*6)g — cota + m®gesc?9 =I + 1)g, or, multiplying through by — sin? 6: nto 8 + sin cosa — mg =—I(I + 1)sin’ 6g. This is a differential equation for g(@); it can be written in a more familiar form: sino (snoS2) +04 1) sin? @ — m*]g =0. [4.132] Butthis is precisely the equation for the @-dependent part, @(0), of ¥/" (8, $) (compare Equation 4.25). Meanwhile, the ¢-dependent part of f (to wit, e”*) is identical to &(¢) (Equation 4.22). Conclusion: The spherical harmonics are precisely the (normalized) eigenfunctions of L? and L,. When we solved the Schrédinger equation by separation of variables, in Section 4.1, we were inadvertantly constructing simultaneous eigenfunctions of the three commuting operators H, L*, and L,:Sec. 4.3: Angular Momentum — 153 HY=Ev, Dy =RId+ Dy, Lay =hmy. [4.133] But there is a curious twist to this story, for the algebraic theory of angular momentum permits / (and hence also m) to take on half-integer values (Equation 4.119), whereas the analytic method yielded eigenfunctions only for integer values (Equation 4.29). You might reasonably guess that the half-integer solutions are spurious, but it turns out that they are of profound importance, as we shalll see in the following sections. «Problem 4.22 (a) What is LY}? (No calculation allowed!) (b) Use the result of (a), together with the fact that L.¥/ = hlY}, to determine Y}(@, @), up to a normalization constant. (©) Determine the normalization constant by direct integration. Compare your final answer to what you got in Problem 4.5. Problem 4.23 In Problem 4.3 you showed that ¥3@, 6) = —V/15/8z siné cosbe*, Apply the raising operator to find ¥}(6, @). Use Equation 4.121 to get the normal- ization. Problem 4.24 (a) Prove that the spherical harmonics are orthogonal (Equation 4.33). Hint: This requires no calculation, if you invoke the appropriate theorem. (b) Prove the orthogonality of the hydrogen wave functions Yrim(r, 6, 6) (Equa- tion 4.90). Problem 4.25 Two particles of mass m are attached to the ends of a massless rigid rod of length a. The system is free to rotate in three dimensions about the center (but the center point itself is fixed). (a) Show that the allowed energies of this rigid rotor are _ Wain +1) CI, for n=0,1,2,... m Hint: First express the (classical) energy in terms of the total angular momentum. (b) What are the normalized eigenfunctions for this system? What is the degeneracy of the nth energy level?154 Chap. 4 Quantum Mechanics in Three Dimensions 4.4 SPIN In classical mechanics, a rigid object admits two kinds of angular momentum: orbital (L =r x p), associated with the motion of the center of mass, and spin (S = Iw). associated with motion about the center of mass. For example, the earth has orbital angular momentum attributable to its annual revolution around the sun, and spin angular momentum coming from its daily rotation about the north-south axis. In the classical context this distinction is largely a matter of convenience, for when you come right down to it, $ is nothing but the sum total of the “orbital” angular momenta of all the rocks and dirt clods that go to make up the earth, as they circle around the axis. But an analogous thing happens in quantum mechanics, and here the distinction is absolutely fundamental. In addition to orbital angular momentum. associated (in the case of hydrogen) with the motion of the electron around the nucleus (and described by the spherical harmonics), the electron also carries another form of angular momentum, which has nothing to do with motion in space (and which is not, therefore, described by any function of the position variables r, 0, 6) but which is somewhat analogous to classical spin (and for which, therefore, we use the same word). It doesn’t pay to press this analogy too far: The electron (as far as we know) is a structureless point particle, and its spin angular momentum cannot be decomposed into orbital angular momenta of constituent parts (see Problem 4.26).”! Suffice it to say that elementary particles carry intrinsic angular momentum (S) in addition to their “extrinsic” angular momentum (L). ‘The algebraic theory of spin is a carbon copy of the theory of orbital angular momentum, beginning with the fundamental commutation relations”: [Sx, Sy] =ihS., (Sy, S:] = ihS,, (Sz, S.] = ihS,. [4.134] It follows (as before) that the eigenvectors of S* and S, satisfy” Sism) =h?s(s + ism); Sz|sm) =hm|s_m); [4.135] and Sa|sm) =hys(s + 1) —m(m + 1) |s (m+ 1), [4.136] 21 For a contrary interpretation, see Hans C. Ohanian, “What is Spin?", Am. J. Phys, $4, 500 (1986). 22We shall take these as postulates for the theory of spin; the analogous formulas for orbital angular momentum (Equations 4.98 and 4.99) were derived from the known form of the operators (Equation 4.97) In a mote sophisticated treatment they can both be obtained from the rotational invariance of the three- dimensional world [see, for example, Leslie E. Ballentine, Quantum Mechanics (Englewood Cliffs, NJ Prentice Hall, 1990), Section 3.3]. Indeed, these fundamental commutation relations apply to all forms of angular momentum, whether spin, orbital, or the combined angular momentum of a composite system, which could include some spin and some orbital Because the eigenstates of spin are not functions, 1 revert to the “ket” notation for them. (I could have done the same in Section 4.3, writing |/ m) in place of ¥/", but in that context the function notation seems more natural.) By the way, I'm running out of letters, so I'll use m for the eigenvalue of S:, just ay I did for L, (Some authors write m, and m, at this stage, just to be absolutely clear)Sec. 4.4: Spin 155 where Sy. = S, +7iS,. But this time the eigenvectors are not spherical harmonics (they’re not functions of @ and ¢ at all), and there is no a priori reason to exclude the half-integer values of s and m: 1.3), Pepe It so happens that every elementary particle has a specific and immutable value of s, which we call the spin of that particular species: pi mesons have spin 0; electrons have spin 1/2; photons have spin 1; deltas have spin 3/2; gravitons have spin 2; and so on, By contrast, the orbital angular momentum quantum number (for an electron in a hydrogen atom, say) can take on any (integer) value you please, and will change from one to another when the system is perturbed. But s is fixed, for any given particle, and this makes the theory of spin comparatively simple.”* 0, m=-s,-st+l,...,s-l,5. [4.137] Problem 4.26 If the electron is a classical solid sphere, with radius e =; [4.133 4meomc?” 4.138) Te (the so-called classical electron radius, obtained by assuming that the electron’s mass is attributable to energy stored in its electric field, via the Einstein formula E = mc*), and its angular momentum is (1/2)h, then how fast (in m/s) is a point on the “equator” moving? Does this model for spin make sense? (Actually, the radius of the electron is known experimentally to be much less than r,, but this only makes matters worse.) 4.4.1 Spin 1/2 By far the most important case is s = 1/2, for this is the spin of the particles that make up ordinary matter (protons, neutrons, and electrons), as well as all quarks and all leptons. Moreover, once you understand spin 1/2, it is a simple matter to work out the formalism for any higher spin. There are just rwo eigenstates: we call spin up (informally, +), and |4 (—!)), which we call spin down (|). Using these as basis vectors, the general state of a spin-1/2 particle can be expressed as a two-element column matrix (or spinor): x (j) sexe thx. [4.139] “Indeed, in a mathematical sense, spin 1/2 is the simplest possible nontrivial quantum system, for it admits just ewo possible states. In place of an infinite-dimensional Hilbert space, with all its subtleties and complications, we find ourselves working in an ordinary two-dimensional vector space; in place of unfamiliar differential equations and fancy functions, we are confronted with 2 x 2 matrices and two- ‘component vectors, For this reason, some authors begin quantum mechanics with a treatment of the spin-1/2 system, But the price of mathematical simplicity is conceptual abstraction, and I prefer not to do it that way.156 Chap. 4 Quantum Mechanics in Three Dimensions with x (3) (4.140) 0 () (4.141) for spin down, Meanwhile, the spin operators become 2 x 2 matrices, which we can work out by noting their effect on x, and x_: Equation 4.135 says representing spin up, and 3 3 h h Sx = Ways Sx = xs Sexe = 5x) Sx = (4.142) 4 4 2 and Equation 4.136 gives Spx =hxys Sxp =hxr Spx, = S_x- = 0. [4.143] Now, Sz = S; £i5,, 80 1 3S +S.) and 8, [4.144] and it follows that h hoo. Sexe = 5x5 Sex = 7x45 Sox [4.145] Thus 0 01\. ./00 0 ): Sy =1(5 yi Ss. =1(9 0 [4.146] while 1 n(0 -i\). ,_h(1 0 0) $= 3(% o)i& 3(0 5) 147) It’s a little tidier to divide off the factor of h/2: S = (f/2)o, where «=(? ive? o)ee(g 5): (4.148) These are the famous Pauli spin matrices. Notice that S,, S,, S;, and S? are all Hermitian (as they should be, since they represent observables). On the other hand, S, and S_ are not Hermitian—evidently they are not observable. The eigenspinors of S, are (of course) h ee (6): (eigenvalue +5); x= (?). (eigenvalue — ». [4.149]Sec. 4.4: Spin 157 If you measure S; on a particle in the general state x. (Equation 4.139), you could get +h/2, with probability |a|?, or —f/2, with probability |b|?, Since these are the only possibilities, lal? + 16 (ie., the spinor must be normalized).”5 But what if, instead, you chose to measure S,? What are the possible results, and what are their respective probabilities? According to the generalized statisti- cal interpretation, we need to know the eigenvalues and eigenspinors of S;. The characteristic equation is 1 [4.150] -A h/2 ny? h = wale =+— ap i =o =(4) aa Not surprisingly, the possible values for S, are the same as those for S,. The eigen- spinors are obtained in the usual way: h(0 1\(a\_ hla B\_,(« 2(1 0) (5)=43(5) 9 (2) =#(): so B = a. Evidently the (normalized) eigenspinors of S, are : Oe (#). (cigenvalue +2); yo = ( 4 ) (eigenvalue — 5), 14.151] 4 2 “4 As the eigenvectors of a Hermitian matrix, they span the space; the generic spinor x can be expressed as a linear combination of them: _ (448) (A) @ 4.152 x= (BS) w+ (Se. [4.152] If you measure S,, the probability of getting +h/2 is (1/2)|a + b/?, and the prob- ability of getting —h/2 is (1/2)la — b|?. (You should check for yourself that these probabilities add up to 1.) Example. Suppose a spin 1/2 particle is in the state i (' + ‘) x Ve\ 2 If you measure S., the probability of getting +4/2 is |(1 + i)/V6l? = 1/3, and the probability of getting —f/2 is |2/V6|? = 2/3. If you measure Sy, the probability of 25People often say that Jal? is the “probability that the particle is in the spin-up state", but this is sloppy language: the particle is in state x—not x4—and what the speaker really means is that if you measured S,, |a|? is the probability you'd get h/2, which is an entirely different assertion.158 Chap. 4 Quantum Mechanics in Three Dimensions getting +%/2 is (1/2)|(3 + i)/V6|? = 5/6, and the probability of getting —h/2 is (1/2)\(-1 + i)/6|? = 1/6. Evidently the expectation value of S, is which we could also have obtained more directly: 1 2 0 A/2 d+i/v6 h Si.) =x'Sx = + af (sy =x'sa= (FZ Za) (ar ° )( ave 3 T'd like now to walk you through an imaginary measurement scenario involving spin 1/2, because it serves to illustrate in very concrete terms some of the abstract ideas we discussed back in Chapter I. Let’s say we start out with a particle in the state y. If someone asks, “What is the z-component of that particle’s spin angular momentum?™. we could answer unambiguously: +4/2. For a measurement of S; is certain to return that value. But if our interrogator asks instead, “What is the x-component of that particle’s spin angular momentum?”, we are obliged to equivocate: If you measure Sj, the chances are 50-50 of getting either /2 or —h/2. If the questioner is a classical physicist, or a “realist” (in the sense of Section 1.2), he will regard this as an inadequate—not to say impertinent—response: “Are you telling me that you don'r know the true state of that particle?” On the contrary; I know precisely what the state of the particle is: x. “Well, then, how come you can’t tell me what the x-component of its spin is?” Because it simply does not have a particular x-component of spin. Indeed, it cannot, for if both S, and S, were well defined, the uncertainty principle would be violated. At this point our challenger grabs the test tube and measures the x-component of its spin; let’s say he gets the value +/2. “Aha!” (he shouts in triumph), “You lied! This particle has a perfectly well-defined value of S,: It’s h/2.” Well, sure—it does now, but that doesn’t prove it had that value, prior to your measurement. “You have obviously been reduced to splitting hairs. And anyway, what happened to your uncertainty principle? I now know both S, and S,.” I'm sorry, but you do not: In the course of your measurement, you altered the particle's state; it is now in the state x{?, and whereas you know the value of S,, you no longer know the value of S. “But I was extremely careful not to disturb the particle when I measured S,:"** Very well, if you don’t believe me, check it out: Measure S., and see what you get. (Of course, he may get +h /2, which will be embarrassing to my case—but if we repeat this whole scenario over and over, half the time he will get —fi/2.) To the layperson, the philosopher, or the classical physicist, a statement of the form “this particle doesn’t have a well-defined position” (or momentum, or 26Neils Bohr was anxious to track down the mechanism by which the measurement of 5, inevitably destroys the value of S:, in gedanken experiments of this sort. His famous debates with Einstein include many delightful examples, showing in detail how experimental constraints serve to enforce the uncertainty principle,Sec. 4.4: Spin 159 x-component of spin angular momentum, or whatever) sounds vague, incompetent, or (worst of all) profound. It is none of these. But its precise meaning is, I think, almost impossible to convey to anyone who has not studied quantum mechanics in some depth. If you find your own comprehension slipping, from time to time (if you don't, you probably haven't understood the problem), come back to the spin-1/2 system: It is the simplest and cleanest context for thinking through the conceptual paradoxes of quantum mechanics. Problem 4.27 (a) Check that the spin matrices (Equation 4.147) obey the fundamental commuta- tion relation for angular momentum: [S., Sy] = ihS,. (b) Show that the Pauli spin matrices satisfy 0k =F FED Guo, {4.153} 7 where the indices stand for x, y, or z, and €;4; is the Levi-Cirita symbol: +1 if (jkl = 123,231, or 312; -1 if jk = 132,213, or 321; 0 otherwise. «Problem 4.28 An electron is in the spin state 3i r=4(2) (a) Determine the normalization constant A. (b) Find the expectation values of S,, S,, and S.. (©) Find the “uncertainties” o5,, 03,, and o.. (4) Confirm that your results are consistent with all three uncertainty principles y 'y princip (Equation 4.100 and its cyclic permutations—only with S in place of L, of course). +Problem 4.29 For the most general normalized spinor x (Equation 4.139), com- pute (Sy), (Sy), (S-), (82), ($2), and (S?). Check that ($2) + (5?) + (S2) = (S*). «Problem 4.30 (a) Find the eigenvalues and eigenspinors of S,. (b) If you measured S, on a particle in the general state x (Equation 4.139), what values might you get, and what is the probability of each? Check that the probabilities add up to 1. (©) If you measured $2, what values might you get and with what probability?160 Chap. 4 Quantum Mechanics in Three Dimensions 4#Problem 4.31 Construct the matrix S, representing the component of spin angular momentum along an arbitrary direction , Use spherical coordinates, so that F = sind cos? +sind sing j + cosa k. [4.154] Find the eigenvalues and (normalized) eigenspinors of S,. Answer: OL cos(@/2) \. w_ sin(@/2) Xe (eeNG FX = | 686 cos(8/2) }* (4.155) Problem 4.32 Construct the spin matrices (S,, S,, and S.) for a particle of spin 1 Hint: How many eigenstates of S- are there? Determine the action of S,, 5, and S_ on each of these states. Follow the procedure used in the text for spin 1/2. 4.4.2 Electron in a Magnetic Field A spinning charged particle constitutes a magnetic dipole. Its magnetic dipole moment jz is proportional to its spin angular momentum S: pays: [4.156] the proportionality constant y is called the gyromagnetic ratio” When a magnetic dipole is placed in a magnetic field B, it experiences a torque, #2 x B, which tends to line it up parallel to the field (just like a compass needle). The energy associated with this torque is* H=-p-B, [4.157] so the Hamiltonian of a spinning charged particle, at rest” in a magnetic field B. becomes H=-yB-S, [4.158] where S is the appropriate spin matrix (Equation 4.147, in the case of spin 1/2). Example: Larmor precession. Imagine a particle of spin 1/2 at rest in a uniform magnetic field, which points in the z-direction: B = Bok. [4.159] 27See, for example, D. Griffiths, Introduction to Electrodynamics, 2nd ed. (Englewood Cliffs, NJ: Prentice Hall, 1986), page 239. Classically, the gyromagnetic ratio of a rigid object is g/2m, where q is its charge and m is its mass. For reasons that are fully explained only in relativistic quantum theory, the gyromagnetic ratio of the electron is almost exactly twice the classical value. 28Griffiths, (footnote 27), pages 246 and 268. 2°1f the particle is allowed to move, there will also be kinetic energy to consider; moreover, it will be subject to the Lorentz force (qv x B), which is not derivable from a potential energy funetion and hence does not fit the Schrédinger equation as we have formulated it so far. I'll show you later on how to handle this problem, but for the moment let’s just assume that the particle is free to rotate, but otherwise stationary:Sec. 4.4: Spin 161 ‘The Hamiltonian matrix is yBoh (1 0 = —yms. = 2 (4 4) [4.160] The eigenstates of H are the same as those of Xe, with energy Ey = —(y Boh)/2, X-+ with energy E. = +(y Boh)/2. Evidently, the energy is lowest when the dipole moment is parallel to the field—just as it would be classically. Since the Hamiltonian is time independent, the general solution to the time- dependent Schrédinger equation, [4.161] ax inX = Hx, [4.162] can be expressed in terms of the stationary states: cuban act Boti2 X(t) = axe + by oF = (erm The constants a and b are determined by the initial conditiot a x=(§), where |a|? + ||? = 1. With no essential loss of generality” I'll write @ = cos(a/2) and b = sin(/2), where a is a fixed angle whose physical significance will appear in a moment. Thus Bot 08 (1/2)e17 xO = (ees) (4.163) To geta feel for what is happening here, let’s calculate the expectation value of the spin (S) as a function of time: (Se) = xO!SxO yiy Bot /2 = (cos(/2)er"¥ BH? sine) ® (7 0) (eee ) ; say 0) \ sina /2)e-17 1/2 = 2 sinercosty But. [4.164] Similarly, h (Sy) = xO'S,x() = 5 sina sin(y Bot), [4.165] “This does assume that a and b are real; you can work out the general case if you like, but all it does is add a constant to162 Chap. 4 Quantum Mechanics in Three Dimensions Figure 4.7: Precession of (8) in a uniform magnetic field. and (S:) = xCOTSx(O = * cosa. [4.166] Evidently (S) is tilted at a constant angle o to the z-axis, and precesses about the field at the Larmor frequency @=YBo [4.167] just as it would classically" (see Figure 4.7). No surprise here—Ehrenfest’s theorem (in the form derived in Problem 4.21) guarantees that (S) evolves according to the classical laws. But it’s nice to see how this works out in a specific context. Example: the Stern-Gerlach experiment. In an inhomogeneous magnetic field, there is not only a torque, but also a force, on a magnetic dipole*: F=Viu-B). [4.168] This force can be used to separate out particles with a particular spin orientation. as follows. Imagine a beam of relatively heavy neutral atoms,’ traveling in the y-direction, which passes through a region of inhomogeneous magnetic field (Figure 4,8)—for example, . Bx, yz) = —axi + (By + az)h, [4.169] where Bo is a strong uniform field and the constant a describes a small deviation from homogeneity. (Actually, what we'd like is just the z-component of this field, but See, for instance, The Feynman Lectures on Physics (Reading, MA: Addison-Wesley, 1964) Volume II, Section 34-3, Of course, in the classical case itis the angular momentum vector itself, not just its expectation value, that precesses around the magnetic field. 5°Griffiths, (footnote 27), page 247. Note that F is the negative gradient of the energy (Equation 4.157). “We make them neutral to avoid the large-scale deflection that would otherwise result from the Lorentz force, and heavy so we can construct localized wave packets and treat the motion in terms of classical particle trajectories. In practice, the Stemn-Gerlach experiment doesn’t work, for example, with a beam of free electrons.Sec. 4.4: Spin 163 Spin up = Spin down Figure 4.8: The Stern-Gerlach apparatus. unfortunately that’s impossible—it would violate the electromagnetic law V-B = 0; like it or not, the x-component comes along for the ride.) The force on these atoms is F=ya(-Si + 8.6) But because of the Larmor precession about Bo, 5, oscillates rapidly, and averages to zero; the net force is in the z-direction: F, yoS;, (4.170) and the beam is deflected up or down, in proportion to the z-component of the spin angular momentum. Classically we'd expect a smear, but in fact the beam splits into 2s +1 individual beams, beautifully demonstrating the quantization of S;. (If you use silver atoms, for example, all the inner electrons are paired in such a way that their spin and orbital angular momenta cancel. The net spin is simply that of the outermost—unpaired—electron, so in this case s = 1/2, and the beam splits in two.) That argument was purely classical, up to the final step; “force” has no place in a proper quantum calculation, and you might therefore prefer the following approach to the same problem.” We examine the process from the perspective of a reference frame that moves along with the beam. In this frame the Hamiltonian starts out zero, turns on for a time T (as the particle passes through the magnet), and then turns off again 0, fort < 0, H(t) =} —-y(Bo+az)S;, ford T. (Lignore the pesky x-component of B, which—for reasons indicated above—is irrel- evant to the problem.) Suppose the atom has spin 1/2, and starts out in the state xt) = ay, +by_, fort <0. While the Hamiltonian acts, y(t) evolves in the usual way: x(t) =aye EH 4 by eH forO <1 <7, “This argument follows L. Ballentine, Quantum Mechanics (Englewood Cliffs, NJ: Prentice Hall, 1990), page 172.164 Chap. 4 Quantum Mechanics in Three Dimensions where (from Equation 4.158) h Ex = FY(By +a2);, (4.172 and hence it emerges in the state H(t) = (aeYP2y,) eAYTPE 4 (pe? TB/2y ) eHONTIA, > T), [4.173 The two terms now carry momentum in the z-direction (see Equation 3.131); the spin-up component has momentum _ayTh =, and it moves in the plus-z direction; the spin-down component has the opposite momentum, and it moves in the minus-z direction. Thus the beam splits in two, a» before. (Note that Equation 4.174 is consistent with the earlier result, Equation 4.170. for in this case S; =h/2and p, = FT.) ‘The Stern-Gerlach experiment has played an important role in the philosophy of quantum mechanics, where it serves both as the prototype for the preparation of a quantum state and as an illuminating model for a certain kind of quantum measure- ment. We casually assume that the initial state of a system is known (the Schrédinger equation tells us how it subsequently evolves)—but it is natural to wonder how you get a system into a particular state in the first place. Well, if you want to prepare a beam of atoms in a given spin configuration, you pass an unpolarized beam through a Stern-Gerlach magnet and select the outgoing stream you are interested in (closing off the others with suitable baffles and shutters). Conversely, if you want to measure the z-component of an atom’s spin, you send it through a Stern-Gerlach apparatus and record which bin it lands in. I do not claim that this is always the most practical way to do the job, but it is conceptually very clean and hence a useful context in-which to explore the problems of state preparation and measurement. Pe [4.174 Problem 4.33 In the first example (Larmor precession ina uniform magnetic field): (a) Ifyoumeasured the component of spin angular momentum along the x-direction. at time t, what is the probability that you would get +4/2? (b) Same question, but for the y-component. (c) Same, but for the z-component. 4xProblem 4.34 An electron is at rest in an oscillating magnetic field B = Bocos(wt)k, where Bo and @ are constants.Sec. 4.4: Spin 165 (a) Construct the Hamiltonian matrix for this system. (b) The electron starts out (at ¢ = 0) in the spin-up state with respect to the x-axis [that is, x(0) = x¢?]. Determine x(¢) at any subsequent time. Beware: This is a time-dependent Hamiltonian, so you cannot get x(t) in the usual way from stationary states. Fortunately, in this case you can solve the time-dependent Schrédinger equation (Equation 4.162) directly. (©) Find the probability of getting —h/2 if you measure S,. Answer: sin? (2 sino). (d) What is the minimum field (Bo) required to force a complete flip in S,? 4.4.3 Addition of Angular Momenta ‘Suppose now that we have two spin-1/2 particles—for example, the electron and the proton in the ground state"* of hydrogen. Each can have spin up or spin down, so there are four possibilities in all*: tt Ne dt, ds [4.175] where the first arrow refers to the electron and the second to the proton. Question: What is the foral angular momentum of the atom? Let s=s%+s°. [4.176] Each of the four composite states is an eigenstate of S.—the z-components simply add Sexix2 = (S29 +S) x1 x2 = (SP xadx2 + x1 (S72) = (hy x1) x2 + xia x2) = hm, + m2) x1 x2, [note that $“ acts only on x1, and $® acts only on x2]. Som (the quantum number for the composite system) is just m; + m2: thm 1 thm = 0 4th m 0; dim = -1. 35] put them in the ground state so there won't be any orbital angular momentum to worry about 36More precisely, each particle is in a linear combination of spin up and spin down, and the composite system is ina linear combination of the four states listed.166 Chap.4 Quantum Mechanics in Three Dimensions At first glance, this doesn’t look right: m is supposed to advance in integer steps, from —s to +s, so it appears that s = 1—but there is an extra state with m = 0. One way to untangle this problem is to apply the lowering operator S_ = S“” + 5" to the state +4, using Equation 4.143: San = Myter on = GYTHETAY SAULT + tb. Evidently the three states with s = | are (in the notation |s m)): Wi) = 44 110) ACN + IN) Ps = 1 Criplet). [4.177] -) = 4 (As a check, try applying the lowering operator to |10); what should you get? See Problem 4.35.) This is called the triplet combination, for the obvious reason. Mean- while, the orthogonal state with m = 0 carries s = 0: 1 00) = —y- s = 0 (singlet) 4.178] { yay 4h) (sing! [ (If you apply the raising or lowering operator to this state, you'll get zero. See Prob- Jem 4.35.) Iclaim, then, that the combination of two spin-1/2 particles can carry a total spin of 1 or 0, depending on whether they occupy the triplet or the singlet configuration. To confirm this, I need to prove that the triplet states are eigenvectors of S? with eigenvalue 2h? and the singlet is an eigenvector of S* with eigenvalue 0. Now SF = (SY +8). (SY +8) = (S)? + ($@)? 4.28 .§@, [4.179] Using Equations 4.142 and 4.145, we have SP-SPR) = SL NEP D+ SP NEP D+ NSP H ONG) ENG) ENE ne qeit— th. . Similarly, SYS U1) = Ke N= 41). It follows that ea 42 2 S® $10) = QIit-N+2H-IN= 110), 14.180]Sec. 4.4: Spin 167 and wo 3h S .§100) = ——=(2 It - th -2 =-——|00). [4.181 |00) 4 Va" dt-th-2N+49 a! ). bE J Returning to Equation 4.179 (and again using Equation 4.142), we conclude that an? aa so =(% ©) 110) =29"110), [4.182] 4 4 so |1 0) is indeed an eigenstate of S? with eigenvalue 2h; and 3m? 3h? 3h? $*{00) = (— + —- -2—— } |00) =0, 4.183 100) G + 4 4 Jie [ 1 so |00) is an eigenstate of S? with eigenvalue 0. (I will leave it for you to confirm that {11) and |1—1) are eigenstates of S?, with the appropriate eigenvalue—sce Prob- lem 4,35.) What we have just done (combining spin 1/2 with spin 1/2 to get spin I and spin 0) is the simplest example of a larger problem: If you combine spin s with spin , what total spins s can you get?” The answer is that you get every spin from (51 + 82) down to (s) — s2)—or (sz — 51), if s2 > s;—in integer steps: s = (81 +52), (1 +82 — 1), (81 +52 — 2)... |S — Sal [4.184] (Roughly speaking, the highest total spin occurs when the individual spins are aligned parallel to one another, and the lowest occurs when they are antiparallel.) For example, if you package together a particle of spin 3/2 with a particle of spin 2, you could get a total spin of 7/2, 5/2, 3/2, or 1/2, depending on the configuration. Another example: If a hydrogen atom is in the state nim, the net angular momentum of the electron (spin plus orbital) is / + 1/2 or / — 1/2; if you now throw in the spin of the proton, the atom’s total angular momentum quantum number is / + 1, J, or / — 1 (and J can be achieved in two distinct ways, depending on whether the electron alone is in the 1+ 1/2 configuration or the | — 1/2 configuration). The particular state |s m) with total spin s and z-component m will be some linear combination of the composite states |s1 m1)|s2 m2): |sm) = Y> CHES, 51 my )Is2 m2) [4.185] (because the 2-components add, the only composite states that contribute are those for which m, ++ my = m). Equations 4.177 and 4.178 are special cases of this general 571 say spins for simplicity, bu either one (or both) could just as well be orbital angular momentum (for which, however, we would use the letter) *8For a proof you must look in a more advanced text; see, for instance, Claude Cohen-Tannoudji, Bemard Diu, and Franck Lalo, Quantum Mechanics (New York: John Wiley & Sons, 1977), Vol. 2 Chapter X168 Chap. 4 Quantum Mechanics in Three Dimensions Table 4.7: Clebsch-Gordan coefficients. (A square root sign is understood for every entry; the minus sign, if present, goes outside the radical.) 9 [ee a 38 8 fais eo |ats de Sto St [eis seit ¥ 38 a ie hz “aio oS 2710 ie -210 133% form, with s; = s, = 1/2 (I used the informal notation t = 5 (—3))). ‘The constants C;'85,,, are called Clebsch-Gordan coefficients. A few of the simplest cases are listed in Table 4.7.” For example, the shaded column of the 2 x 1 table tells us that 1 1 1 [21) = Z.122)I1 -1) + Fy2onn. In particular, if two particles (of spin 2 and spin 1) are at rest in a box, and the total spin is 2, and its z-component is 1, then a measurement of S‘ could return the value 2h (with probability 1/3), or h (with probability 1/6), or 0 (with probability 1/2). Notice that the probabilities add up to 1 (the sum of the squares of any column on the Clebsch-Gordan table is 1). These tables also work the other way around: [21)[10) [sim )hs2 m2) = > Co, Ls m). [4.186] For example, the shaded row in the 3/2 x 1 table tells us that BR 1 3 ti Rano=iup+/eap-YnD *°The general formula is derived in Arno Bohm, Quantum Mechanics: Foundations and Applica tions, 2nd ed. (New York: Springer-Verlag, 1986), p. 172.Sec. 4.4: Spin 169 If you put particles of spin 3/2 and spin 1 in the box, and you know that the first has ‘mj = 1/2 and the second has mz = 0 (so m is necessarily 1/2), and you measured the total spin s, you could get 5/2 (with probability 3/5), or 3/2 (with probability 1/15), or 1/2 (with probability 1/3). Again, the sum of the probabilities is 1 (the sum of the squares of each row on the Clebsch-Gordan table is 1). If you think this is starting to sound like mystical numerology, I don’t blame you, We will not be using the Clebsch-Gordan tables much in the rest of the book, but I wanted you to know where they fit into the scheme of things, in case you encounter them later on. In a mathematical sense this is all applied group theory— what we are talking about is the decomposition of the direct product of two irreducible representations of the rotation group into a direct sum of irreducible representations. (You can quote that to impress your friends.) xProblem 4.35 (a) Apply S_ to |10) (Equation 4.177), and confirm that you get /2h|1 —1) (b) Apply Sx. to |00) (Equation 4.178), and confirm that you get zero. (©) Show that {1 1) and |1—1) (Equation 4.177) are eigenstates of S?, with the appropriate eigenvalue. Problem 4.36 Quarks carry spin 1/2. Three quarks bind together to make a baryon (such as the proton or neutron); two quarks (or more precisely a quark and an antiquark) bind together to make a meson (such as the pion or the kaon). Assume the quarks are in the ground state (so the orbital angular momentum is zero). (a) What spins are possible for baryons? (b) What spins are possible for mesons? Problem 4.37 (a) A particle of spin 1 and a particle of spin 2 are at rest in a configuration such that the total spin is 3, and its z-component is 1 (that is, the eigenvalue of S: is i). If you measured the z-component of the angular momentum of the spin-2 particle, what values might you get, and what is the probability of each one? (b) An electron with spin down is in the state 510 of the hydrogen atom. If you could measure the total angular momentum squared of the electron alone (not including the proton spin), what values might you get, and what is the probability of each? Problem 4.38 Determine the commutator of $? with S{” (where S Generalize your result to show that 7 = 2S x $), [4.187]170 Chap. 4 Quantum Mechanics in Three Dimensions Note: Because S‘! does not commute with S?, we cannot hope to find states that are simultaneous eigenvectors of both. To form eigenstates of S*, we need linear comb1 nations of eigenstates of §{. This is precisely what the Clebsch-Gordan coefficient+ (in Equation 4.185) do for us, On the other hand, it follows by obvious inference from Equation 4.187 that the sum $'" + 8° does commute with S?, which only confirm: what we already knew (see Equation 4,103). FURTHER PROBLEMS FOR CHAPTER 4 «Problem 4.39 Consider the three-dimensional harmonic oscillator, for whick the potential is vn= [4.188 (a) Show that separation of variables in Cartesian coordinates turns this into three one-dimensional oscillators, and exploit your knowledge of the latter to deter mine the allowed energies. Answer: Ey, = (n+ 3/2ho. [4.189] (b) Determine the degeneracy d(n) of Ey. «Problem 4.40 Because the three-dimensional harmonic oscillator potential (Equa- tion 4.188) is spherically symmetric, the Schrédinger equation can be handled by separation of variables in spherical coordinates as well as Cartesian coordinates. Use the power series method to solve the radial equation. Find the recursion formula for the coefficients, and determine the allowed energies. Check your answer against Equation 4.189. x«Problem 4.41 (a) Prove the three-dimensional virial theorem 2(T) = (r- VP) [4.190] (for stationary states). Hint: Refer to Problem 3.53 (b) Apply the virial theorem to the case of hydrogen, and show that (T)=-Enx; (V) =2Ey. [4.191] (c) Apply the virial theorem to the three-dimensional harmonic oscillator (Prob- lem 4.39), and show that in this case (T) = (V) = En/2. [4.192]Further Problems for Chapter 4.171 +xxProblem 4.42 The momentum-space wave function in three dimensions is de- fined by the natural generalization of Equation 3.132: 1 : oP) = cape few y(n) dr. [4.193] (a) Find the momentum-space wave function for the ground state of hydrogen (Equation 4.80). Hint: Use spherical coordinates, setting the polar axis along the direction of p. Do the @ integral first. Answer: 6) = + 2)" a 14.194) m= iG 11+ G@piiyPP (b) Check that ¢(p) is normalized. (c) Use $(p) to calculate (p*) (d) What is the expectation value of the kinetic energy in this state? Express your answer as a multiple of £, and check that it is consistent with the virial theorem (Equation 4.191). Problem 4.43 (a) Construct the spatial wave function (y) for hydrogen in the state n = 3,/ = 2, m = 1. Express your answer as a function of r, 6, @, and a (the Bohr radius) only—no other variables (p, 2, etc.), or functions (Y, v, etc.), or constants (4, do, etc.), or derivatives allowed (r is okay, and e, and 2, etc.). (b) Check that this wave function is properly normalized by carrying out the appro- priate integrals over r, 0, and Find the expectation value of r* in this state. For what range of s is the result finite? © +*4Problem 4.44 Suppose two spin-1/2 particles are known to be in the singlet con- figuration (Equation 4.178). Let S‘") be the component of the spin angular momentum of particle number 1 in the direction defined by the unit vector d. Similarly, let Sj? be the component of 2’s angular momentum in the direction 6. Show that ne (SPS) = —~ cos, [4.195] where 6 is the angle between d and b. ++Problem 4.45 Work out the Clebsch-Gordan coefficients for the case s; = 1/2, 5) = anything. Hint: You're looking for the coefficients A and B in sm) = Al} 4)Is (mm — 5)) + BIE (—})ps2 (m + 4)),172 Chap. 4 Quantum Mechanics in Three Dimensions such that |sm) is an eigenstate of S?. Use the method of Equations 4.179 throug 4,182. If you can’t figure out what S (for instance) does to |s2 m2), refer back t Equations 4.136 and 4.144. Use this general result to construct the (1/2) x 1 table of Clebsch-Gordan coefficients, and check it against Table 4.7. Answer: [sym +1/2 3 Fm +1/2 A= |/————:_ B= +,/——_., 241 25141 5271/2. where the signs are determined by s Problem 4.46 Find the matrix representing S, for a particle of spin 3/2 (using the basis of eigenstates of S.). Solve the characteristic equation to determine the eigenvalues of Sy. +x«Problem 4.47 Work out the normalization factor for the spherical harmonics, ay follows. From Section 4.1.2 we know that Yj" = Brei™® P™ (cos6); the problem is to determine the factor BM" (which I quoted, but did not derive, in Equation 4.32). Use Equations 4.120, 4.121, and 4,130 to obtain a recursion relation giving Bi"! in terms of BM", Solve it by induction on m to get Bj" up to an overall constant C(/). Finally, use the result of Problem 4.22 to fix theconstant. You may find the following formula for the derivative of an associated Legendre function useful: aP™ a -x) =V1—x2 Pm! — mx Pm. [4.196] Problem 4.48 The electron in a hydrogen atom occupies the combined spin and position state [vr 2 Roy (\ slixe + (arie. : If you measured the orbital angular momentum squared (L?), what values might you get, and what is the probability of each? fa (b) ©) @d) Same for the z-component of orbital angular momentum (L.). Same for the spin angular momentum squared (S?). Same for the z-component of spin angular momentum (S-). Let If you measured J?, what values might you get, and what is the probability of each? (f) Same for J-. L+S be the total angular momentum. eFarther Problems for Chapter 4. 173 (g) If you measured the position of the particle, what is the probability density for finding it at r, 6,6? (h) Ifyou measured both the z-component of the spin and the distance from the ori- gin (note that these are compatible observables), what is the probability density for finding the particle with spin up and at radius r? +**Problem 4.49 (a) Fora function f(@) that can be expanded in a Taylor series, show that Sp + Go) = e'**/" f() (where ¢p is any constant angle). For this reason, L/h is called the generator of rotations about the z-axis. Hint: Use Equation 4.129, and refer to Problem 3.55. More generally, L - fi/h is the generator of rotations about the direction fi, in the sense that exp(/L - ig/h) effects a rotation through angle ¢ (in the right-hand sense) about the axis fl. In the case of spin, the generator of rotations is S - i/h. In particular, for spin 1/2 iMate x 4.197] tells us how spinors rotate. (b) Construct the (2 x 2) matrix representing rotation by 180° and show that it converts “spin up” (x) into “spin down” (x. expect. (C) Construct the matrix representing rotation by 90° about the y-axis, and check what it does to x. about the x-axis, , as you would (@) Construct the matrix representing rotation by 360° about the z-axis. If the answer is not quite what you expected, discuss its implications. (e) Show that - eI? = cos(y/2) + i(f-@) sin(y/2). [4.198] «Problem 4.50 The fundamental commutation relations for angular momentum (Equations 4.98 and 4,99) allow for half-integer (as well as integer) eigenvalues. But for orbital angular momentum only the integer values occur. There must be some extra constraint in the specific form L = r x p that excludes half-integer values.”® Let a be some convenient constant with the dimensions of length (the Bohr ra say, if we're talking about hydrogen), and define the operators n= Zl+@ Mp) p= J5[P- h/a)y; = - CMe): m= J [a + G/ay]. “This problem is based on an argument in Ballentine, (footnote.34), page 127,174 Chap. 4 Quantum Mechanics in Three Dimensions (a) Verify that [91.92] = [pi. Po] = 0; [41, Pil = [42 Pel (b) « (d ) ih. Thus the q’s and the p’s satisfy the canonical commutation relations for position and momentum. and those of index 1 are compatible with those of index 2. Show that is @& 5ga Gi — 92) + 3 (Pi ~ PD- Check that L, = Hi — Hz, where each H is the Hamiltonian for a harmonic oscillator with mass m = h/a? and frequency w = 1. We know that the eigenvalues of the harmonic oscillator Hamiltonian are (n ~ 1/2)he, where n = 0, 1,2,... (in the algebraic theory of Section 2.3.1, this follows from the form of the Hamiltonian and the canonical commutation rela- tions). Use this to conclude that the eigenvalues of L, must be integers. +#+Problem 4.51 In classical electrodynamics the force on a particle of charge q moving with velocity v through electric and magnetic fields E and B is given by the Lorentz force law: F=qE+vxB). 14.199} This force cannot be expressed as the gradient of a scalar potential energy function. and therefore the Schrédinger equation in its original form (Equation 1.1) cannot accomodate it. But in the more sophisticated form aw ih-— = HW [4.200] there is no problem; the classical Hamiltonian" is where A is the vector potential (B = Vx A) and g is the scalar potential (E = H= aay tay, (4.201) m Ve~ 9A/4r), so the Schrédinger equation (making the canonical substitution p > (i/h)V) becomes ow 1 (h 2 ot =| L( va) +a] w. [4.202] (a) Show that 4 1 it) = Tp ~agAy. (4.203) 7m 41See, for example, H. Goldstein, Classical Mechanics, 2nd ed.. Addison-Wesley, Reading, MA. 1980, page 346,Further Problems for Chapter 4 175 (b) As always (see Equation 1.32) we identify d(r)/dr with (v). Show that ay) q g oe = q(E) + 2 (px B-Bx p))— L(A xB)). [4.2041 mo itt oe xp) (o*®) FE | (©) In particular, if the fields E and B are uniform over the volume of the wave packet, show that ay) m7 = 4 + (¥) XB), [4.205] so the expectation value of (¥) moves according to the Lorentz force law, as we would expect from Ehrenfest’s theorem. x«*Problem 4.52 [Refer to Problem 4,51 for background.] Suppose A= Pe —yi), and g=Kz’, where By and K are constants. (a) Find the fields E and B. (b) Find the allowed energies, for a particle of mass m and charge q, in these fields. «Problem 4.53 [Refer to Problem 4.51 for background,] In classical electrodynam- ics the potentials A and ¢ are not uniquely determined*; the physical quantities are the fields, E and B. (a) Show that the potentials 8 =0- 5 A'=A4VA [4.206] (where A is an arbitrary real function of position and time) yield the same fields as g and A, Equation [4.206] is called a gauge transformation, and the theory is said to be gauge invariant. (6) In quantum mechanics the potentials play a more direct role, and it is of interest to know whether the theory remains gauge invariant. Show that Ws eltthy [4.207] satisfies the Schrédinger equation [4.202] with the gauge-transformed potentials and A’, Since W’ differs from W only by a phase factor, it represents the “See, for example, Griffiths, (Footnote 27), section 10.2.4.176 Chap. 4 Quantum Mechanics in Three Dimensions same physical state“, and the theory is gauge invariant (see Section 10.2.4 for further discussion). That is to say, (r), d(r}/dt, etc, are unchanged. Because A depends on position, (p) (with P represented by the operator (h/i)V) does change, but as we found in Equation [4.203], p does not represent the mechanical momentum (mv) in this context (in Lagrangian mechanics it is so-called canonical momentum).CHAPTER 5 IDENTICAL PARTICLES 5.1 TWO-PARTICLE SYSTEMS Fora single particle, the wave function ‘V(r, t) is a function of the spatial coordinates and the time ¢ (we'll ignore spin for the moment), The wave function for a two- particle system is a function of the coordinates of particle one (r;), the coordinates of particle two (r2), and the time: Wri, 1). [5.1] Its time evolution is determined (as always) by the Schrédinger equation: aw ih— = HW, [5.2] ar where H is the Hamiltonian for the whole system: 2 2 a hi Hat V3+ V(t. 4 2,0) [5.3] 2m 2m, *? (the subscript on V indicates differentiation with respect to the coordinates of particle 1 or particle 2, as the case may be). The statistical interpretation carries over in the obvious way: Morn, OP ar dire (5.4] is the probability of finding particle 1 in the volume dr, and particle 2 in the volume @ ry; evidently W must be normalized in such a way that frmenrorery @ry=1. [5.5]178 Chap. 5 Identical Particles For time-independent potentials, we obtain a complete set of solutions by sep- aration of variables: Wr, rr.) = Wr mete, [5.6 where the spatial wave function (y) satisfies the time-independent Schrédinger equa- tion: viv By + Vw = Ey. [5.7 2m)!" ~ Ima and E is the total energy of the system. x*Problem 5.1 Typically, the interaction potential depends only on the vector r =r; — Pr separating the two particles. In that case the Schridinger equation separates, if we change variables from r), rz tor, R= (myry + mor2)/(m) + m2 (the center of mass). (a) Show that ry = R-+ (j¢/my)r, ty = R—(4t/ma)r, and V, = (u/m2)Ve + V. Vo = (u/my)Ve — V>. where mym> [581 mbm is the reduced mass of the system. (b) Show that the (time-independent) Schrédinger equation becomes x Va Here yy = Ey =. yy Sy ry = Ey. 2m +m) "2p « Solve by separation of variables, letting w(R, r) = Wa(R)yr(r). Note that we satisfies the one-particle Schrédinger equation, with the total mass (m, + m2" in place of m, potential zero, and energy Ep, while , satisfies the one-particle Schrodinger equation with the reduced mass in place of m, potential V (r), and energy E,. The total energy is the sum: E = Ex + E,. Note: What this tells us is that the center of mass moves like a free particle, and the relative motion (that is, the motion of particle 2 with respect to particle |) is the same as if we had a single particle with the reduced mass, subject to the potential V. Exactly the same separation occurs in classical mechanics; it reduces the two-body problem to an equivalent one-body problem. Problem 5.2 In view of Problem 5.1, we can correct for the motion of the nucleus in hydrogen by simply replacing the electron mass with the reduced mass: (a) Find (to two significant digits) the percent error in the binding energy of hydro- gen (Equation 4.77) introduced by our use of m instead of j.. ' See, for example, Jerry Marion, Classical Dynamics, 2nd ed. (New York: Academic Press 1970) Section 8.2Sec. 5.1: Two-Particle Systems 179 (b) Find the separation in wavelength between the red Balmer lines (n =3 — n =2) for hydrogen and deuterium. (c) Find the binding energy of positronium (in which the proton is replaced by a positron—positrons have the same mass as electrons but opposite charge). (d) Suppose you wanted to confirm the existence of muonic hydrogen, in which the electron is replaced by a muon (same charge, but 206.77 times heavier). Where (ic., at what wavelength) would you look for the “Lyman-e” line (n=2>n=1)? 5.1.1 Bosons and Fermions Suppose particle 1 is in the (one-particle) state Y(t), and particle 2 is in the state Wot). In that case y(t), F2) is a simple product: WE, Fa) = Volts) yolr2). [5.9] Of course, this assumes that we can tell the particles apart—otherwise it wouldn't make any sense to claim that number 1 is in state ¥ and number 2 is in state Ws; all we could say is that one of them is in the state Wz and the other is in state Ws, but we wouldn't know which is which. If we were talking about classical mechanics this would be a silly objection: You can always tell the particles apart, in principle—just paint one of them red and the other one blue, or stamp identification numbers on them, or hire private detectives to follow them around. But in quantum mechanics the situation is fundamentally different: You can’t paint an electron red, or pin a label on it, and a detective’s observations will inevitably and unpredictably alter the state, raising doubts as to whether the two had perhaps switched places. The fact is, all electrons are utterly identical, in a way that no two classical objects can ever be. It is not merely that we don’t happen to know which electron is which; God doesn’t know which is which, because there is no such thing as “this” electron, or “that” electron; all we can legitimately speak about is “an” electron. Quantum mechanics neatly accommodates the existence of particles that are indistinguishable in principle: We simply construct a wave function that is noncom- mittal as to which particle is in which state. There are actually two ways to do it: WalF1, 82) = Alva(ti)Wo(F2) = Yat) Walt2)]- [5.10] Thus the theory admits two kinds of identical particles: bosons, for which we use the plus sign, and fermions, for which we use the minus sign. Photons and mesons are bosons; protons and electrons are fermions. It so happens that all particles with integer spin are bosons, and tsa] all particles with half-integer spin are fermions.180 Chap. 5 Identical Particles This connection between spin and “statistics” (as we shall see, bosons and fermions have quite different statistical properties) can be proved in relativistic quantum me- chanics; in the nonrelativistic theory it must be taken as an axiom. It follows, in particular, that two identical fermions (for example, two electrons) cannot occupy the same state. For if vq = Wo, then W- (1, 82) = Ala lt) Ya(t2) — Walt) Walr2)] = 0, and we are left with no wave function at all. This is the famous Pauli exclusion prineiple. It is not (as you may have been led to believe) a bizarre ad hoc assumption applying only to electrons, but rather a consequence of the rules for constructing two-particle wave functions, applying to all identical fermions. T assumed, for the sake of argument, that one particle was in the state y/, and the other in state ys, but there is a more general (and more sophisticated) way to formulate the problem. Let us define the exchange operator P which interchanges the two particles: Pf(ty,2) = ft, 41) [5.12] Clearly, P? = 1, and it follows (prove it for yourself) that the eigenvalues of P are +1. If the two particles are identical, the Hamiltonian must treat them the same: m = m= and V(r, 12) = V(r, 11). It follows that P and H are compatible observables, (P, H]=0, [5.13] and hence we can find a complete set of functions that are simultaneous eigenstates of both. That is to say, we can find solutions to the Schrédinger equation that are either symmetric (eigenvalue +1) or antisymmetric (eigenvalue —1) under exchange: Wri 82) £W(r2, 11) (+ for bosons, — for fermions). [5.14] Moreover, if a system starts out in such a state, it will remain in such a state. The new law (I'll call it the symmetrization requirement) is that for identical particles the wave function is not merely allowed, but required to satisfy Equation 5.14, with the plus sign for bosons and the minus sign for fermions.? This is the general statement. of which Equation 5.10 is a special case. 2It is sometimes suggested that the symmetrization requirement (Equation 5,14) is nothing new— that is forced by the fact that P and H commute. This is false: It is perfectly possible to imagine a system of two distinguishable particles (say, an electron and a positron) for which the Hamiltonian is symmetric. and yet there is no requirement that the wave function be symmetric (or antisymmetric). But identical particles have to occupy symmetric or antisymmetric states, and this is a completely new fundamental Iaw-—on a par, logically, with Schrédinger’s equation and the statistical imerpretation. Of course, there didn’t have to be any such things as identical particles; it could have been that every single particle in nature was clearly distinguishable from every other one. Quantum mechanics allows for the possibility of identical particles, and nature (being lazy) seized the opportunity. (But I’m not complaining —this makes matters enormously simpler!)Sec. 5.1: Two-Particle Systems 181 Example. Suppose we have two noninteracting’ particles, both of mass m, in the infinite square well (Section 2.2). The one-particle states are Vale) = (2 sins), E,=0K (where K = 17h?/2ma?). If the particles are distinguishable, the composite wave functions are simple products: inn ( 1 ¥2) = Vin X)Vng (2), Enns = (+3) K. For example, the ground state is vu= 2 sintexi/a) sin@rx2/a), Ey = 2K; the first excited state is doubly degenerate: 2 wir sin(tx)/a)sin(2tx2/a), Ey, =5K, a Yn = = sincanxs fa) singrxy/a), En = 5K: and so on. If the two particles are identical bosons, the ground state is unchanged, but the first excited state is nondegenerate: 2 [sin(srx, /a) sin(2srx2/a) + sinQ2ax,/a) sin(arx2/a)] (still with energy 5K). And if the particles are identical fermions, there is no state with energy 2K; the ground state is = [sin(rx, /a) sin(2arx2/a) — sin(2xx,/a) sin(rx2/a)]. and its energy is 5K. Problem 5.3 (a) If Yq and Yp are orthogonal, and both normalized, what is the constant 4 in Equation 5.10? (b) If Ya = vs (and it is normalized), what is 4? (This case, of course, occurs only for bosons.) They pass right through one another—never mind how you would set this up in practice! I'll ignore spin—if this bothers you (afterall, a spinless fermion is a contradiction in terms), assume they're in the same spin state.182 Chap. 5 Identical Particles Problem 5.4 (a) Write down the Hamiltonian for two identical noninteracting particles in the infinite square well. Verify that the fermion ground state given in the example is an eigenfunction of H, with the appropriate eigenvalue. (b) Find the next two excited states (beyond the ones given in the example)—wave functions and energies—for each of the three cases (distinguishable, identical bosons, identical fermions). 5.1.2 Exchange Forces To give you some sense of what the symmetrization requirement actually does, I'm going to work out a simple one-dimensional example. Suppose one particle is in state y,(x), and the other is in state W(x), and these two states are orthogonal and normalized. If the two particles are distinguishable, and number 1 is the one in state Wa, then the combined wave function is W(X, x2) = Wales) Wo(e2); [5.15] if they are identical bosons, the composite wave function is (see Problem 5.3 for the normalization) Wa, X2) = Fylde vole2) + ori) Wa (x2)]i [5.16] and if they are identical fermions, it is 1 WW-C, X2) = gle Wola) — Vole )Walx2)]- (5.17] Let’s calculate the expectation value of the square of the separation distance between the two particles, = (xf) + (x3) — 201m). 5.18] (ea ~ 22) Case 1; Distinguishable particles. For the wave function in Equation 5.15. we have ais [stvecnPan f WWo(x2)P? dx2 = (x7)a (the expectation value of x? in the one-particle state ra), 02) = f ivocnian [ xivscn ae = (5,Sec. 5.1: Two-Particle Systems 183 and tan) = f miventan f mivoaitar = wat In this case, then, (01 = a2)" a = (7a + (7 bp — WeralxPa: [5.19] (Incidentally, the answer would—of course—be the same if particle 1 had been in state Ws, and particle 2 in state Ya.) Case 2: Identical particles. For the wave functions in Equations 5.16 and 5.17, sh) = 3Lf stivecsntan f ivstayPare +f divonitan | WetPan + f xivecervotrrdn f yots)"Yoteadare Fs | PVC)" Yon) dy f Yala)" Valea) dra] = Flee + Os £040] = 1 (lee + 024) Similarly, (x3) = ; ((37)6 + 7a) (Naturally, (x3) = (x?), since you can’t tell them apart.) But 1 2 (mim) = s[fnivecmPan [raecotars +f mivsen Pan f ralvotn ae £ fnvacnrvntonan, f sa¥otayvaladdes +f vate) vater) day f se¥ala2y"VaCe) ax]184 Chap. 5 Identical Particles 1 = 5 (Oalx)o + (xo leda + ()an (aba E ()4a (a6) (xha(ade# daal?s where (sons fsa (sy Wald [5.20 Evidently (C81 = 2)? )ae = (27a + (27D = 2B) alo F 2A )aa (5.21 Comparing Equations 5.19 and 5.21, we see that the difference resides in the final term: (Ax) = (Ax) )a Fl )aol?§ (5.22 identical bosons (the upper signs) tend to be somewhat closer together, and identics. fermions (the lower signs) somewhat farther apart, than distinguishable particles ir the same two states. Notice that (x), vanishes unless the two wave functions actually overlap [if Wa(x) is zero wherever W»(x) is nonzero, the integral in Equation 5.20 ix itself zero]. So if Ya represents an electron in an atom in Chicago and ys represent an electron in an atom in Seattle, it’s not going to make any difference whether you antisymmetrize the wave function or not. As a practical matter, therefore, it’s okay to pretend that electrons with nonoverlapping wave functions are distinguishable (Indeed, this is the only thing that allows physicists and chemists to proceed at all for in principle every electron in the universe is linked to every other one via the antisymmetrization of their wave functions, and if this really mattered, you wouldn't be able to talk about any one electron until you were prepared to deal with them all! The interesting case is when there is some overlap of the wave functions. The system behaves as though there were a “force of attraction” between identical bosons. pulling them closer together, and a “force of repulsion” between identical fermions. pushing them apart. We call it an exchange force, although it’s not really a force at all—no physical agency is pushing on the particles; rather, it is a purely geometricai consequence of the symmetrization requirement. It is also a strictly quantum me- chanical phenomenon, with no classical counterpart. Nevertheless, it has profound consequences. Consider, for example, the hydrogen molecule (Hz). Roughly speak- ing, the ground state consists of one electron in the atomic ground state (Equation 4.80) centered on nucleus 1, and one electron in the atomic ground state centered at nucleus 2. If electrons were bosons, the symmetrization requirement (or, if you like the “exchange force”) would tend to concentrate the electrons toward the middle between the two protons (Figure 5.1a), and the resulting accumulation of negative charge would attract the protons inward, accounting for the covalent bond that holdsSec. 5.1: Two-Particle Systems 185 F F & + + F P Pp P P © (a) (b) Figure 5.1: Schematic picture of the covalent bond: (a) Symmetric configura- tion produces attractive force; (b) antisymmetric configuration produces repul- sive force, the molecule together. Unfortunately, electrons aren't bosons, they're fermions, and this means that the concentration of negative charge should actually be shifted to the wings (Figure 5.1b), tearing the molecule apart! But wait. We have been ignoring spin. The complete state of the electron in- cludes not only its position wave function, but also a spinor, describing the orientation of its spin*: wr)x(s). [5.23] When we put together the two-electron state, it is the whole works, not just the spatial part, that has to be antisymmetric with respect to exchange. Now, a glance back at the composite spin states (Equations 4.177 and 4.178) reveals that the singlet combination is antisymmetric (and hence would have to be joined with a symmetric spatial function), whereas the three triplet states are all symmetric (and would require an antisymmetric spatial function). Evidently, then, the singlet state should lead to bonding, and the triplet to antibonding. Sure enough, the chemists tell us that covalent bonding requires the two electrons to occupy the singlet state, with total spin zero.> +Problem 5.5 Imagine two noninteracting particles, each of mass m, in the infinite square well. If one is in the state y, (Equation 2.24) and the other in state Yin orthogonal toy, calculate (x; — x2)"), assuming that (a) they are distinguishable particles, (b) they are identical bosons, and (c) they are identical fermions. Problem 5.6 Suppose you had three particles, one in state y, (x), one in state W(x), and one in state Yie(x). Assuming that ¥r,, ¥, and y- are orthonormal, construct the three-particle states (analogous to Equations 5.15, 5.16, and 5.17) representing (a) distinguishable particles, (b) identical bosons, and (c) identical fermions. Keep in mind that (b) must be completely symmetric under interchange of any pair of particles, and (c) must be completely anti-symmetric in the same sense.) Note: There's a cute “In the absence of coupling between spin and position, we are free to assume that the state is separable in its spin and spatial coordinates. ‘This just says that the probability of getting spin up is independent of the location of the particle. In the presence of coupling, the general state would take the form of a linear combination: yr (r)x4 + W—(F)x-. In casual language, itis often said that the electrons are “oppositely aligned” (one with spin up, and the other with spin down). This is something of an oversimplification, since the same could be said of the m =O triplet state, The precise statement is that they are in the singlet configuration,186 Chap. 5 Identical Particles trick for constructing completely antisymmetric wave functions: Form the Slater determinant, whose first row is yra(x1), Wo(%1), We(%1), etc., whose second row is Wa (x2), Wo(x2), We(x2), etc., and so on (this device works for any number of particles) 5.2 ATOMS A neutral atom, of atomic number Z, consists of aheavy nucleus, with electric charge Ze, surrounded by Z electrons (mass m and charge —e). The Hamiltonian for this system is® 2 ( Ww, 1 \ Ze) 1/1\< le ect lea jal *. The term in curly brackets represents the kinetic plus potential energy of the jth electron in the electric field of the nucleus; the second sum (which runs over all values of j and k except j = k) is the potential energy associated with the mutual repulsion of the electrons (the factor of 1/2 in front corrects for the fact that the summation counts each pair twice). The problem is to solve Schrédinger’s equation. Hy = Ey, [5.25] for the wave function W(t, 19, ...,1z). Because electrons are identical fermions. however, not all solutions are acceptable: only those for which the complete state (position and spin), WT, Ta, +. 648 Z)X(S1, 825+. 58z)s [5.26] is antisymmetric with respect to interchange of any two electrons. In particular, no two electrons can occupy the same state. Unfortunately, the Schrédinger equation with the Hamiltonian in Equation 5.24 cannot be solved exactly (at any rate, it hasn't been) except for the very simplest case. Z = | (hydrogen). In practice, one must resort to elaborate approximation methods. Some of these we shall explore in Part II; for now I plan only to sketch some of the qualitative features of the solutions, obtained by neglecting the electron repulsion term altogether. In section 5.2.1 we'll study the ground state and excited states of helium, and in section 5.2.2 we'll examine the ground states of higher atoms. Pm assuming the nucleus is stationary. The trick of accounting for nuclear motion by using the reduced mass (Problem 5.1) works only for the nwo-body problem—hydrogen; fortunately, the nucleus so much more massive than the electrons that the correction is extremely small even in that case (see Problem 5.2a), and it is smaller still for the heavier atoms. There are more interesting effects, due to magnetic interactions associated with electron spin, relativistic corrections, and the finite size of the nucleus, We'll look into these in later chapters, but all of them are minute corrections to the “purely Coulombic” atom described by Equation 5.24.Sec. 5.2: Atoms 187 Problem 5.7 Suppose you could find a solution y(r,r2,....Fz) to the Schrédinger (Equation 5.25) for the Hamiltonian in Equation 5.24. Describe how you could construct from it a completely symmetric function and a completely an- tisymmetric function, which also satisfy the Schrédinger equation, with the same energy. 5.2.1 Helium After hydrogen, the simplest atom is helium (Z = 2). The Hamiltonian, Wy 1 1 20 1 H=}- Vi- fp + + (5.27 2m ‘- Srey 71 2h [-£ 2 4m 1 jeu Ir — r3] a7 consists of two hydrogenic Hamiltonians (with nuclear charge 2e), one for electron 1 and one for electron 2, together with a final term describing the repulsion of the two electrons. It is this last term that causes all the problems. If we simply ignore it, the Schrédinger equation separates, and the solutions can be written as products of hydrogen wave functions: WEI, 82) = Wnt (Er) nim’ (2s [5.28] only with half the Bohr radius (Equation 4.72), and four times the Bohr energies (Equation 4.70). The total energy would be E=4(E, + Ey), 15.29] where E, = —13.6/n? eV. In particular, the ground state would be 8 WoUri, #2) = vroo(e1)Vreo(e2) = sage te [5.30] (see Equation 4.80), and its energy would be Ey = 8(-13.6eV) = —109 eV. [5.31] Because Wo is a symmetric function, the spin state has to be antisymmetric, so the ground state of helium is a singlet configuration, with the spins “oppositely aligned”. ‘The actual ground state of helium is indeed a singlet, but the experimentally deter- mined energy is —78.975 eV, so the agreement is not very good. But this is hardly surprising: We ignored electron repulsion, which is certainly not a small contribution. Itis clearly positive (see Equation 5.27), which is comforting—evidently it brings the total energy up from —109 to ~79 eV (see Problem 5.10). The excited states of helium consist of one electron in the hydrogenic ground state and the other in an excited state: Ynim Wio0- (5.32]188 Chap. 5 Identical Particles [If you try to put both electrons in excited states, one immediately drops to the ground state, releasing enough energy to knock the other one into the continuum (E > 01 leaving you with a helium ion (He*) and a free electron. This is an interesting sys- tem in its own right—see Problem 5.8—but it is not our present concern] We can construct from this both symmetric and antisymmetric combinations, in the usual way (Equation 5.10); the former go with the antisymmetric spin configuration (the singlet), and they are called parahelium, while the latter require a symmetric spin configuration (the triplet), and they are known as orthohelium. The ground state is necessarily parahelium; the excited states come in both forms. Because the sym- metric spatial state brings the electrons closer together (as we discovered in Section 5.1.2), we expect a higher interaction energy in parahelium, and indeed it is exper- imentally confirmed that the parahelium states have somewhat higher energy than their orthohelium counterparts (see Figure 5.2). Problem 5.8 (a) Suppose you put both electrons in a helium atom into the n = 2 state: what would the energy of the emitted electron be? (b) Describe (quantitatively) the spectrum of the helium ion, He*. Problem 5.9 Discuss (qualitatively) the energy level scheme for helium (a) if elec- trons were identical bosons, and (b) if electrons were distinguishable particles (but still with the same mass and charge). Pretend the electrons still have spin 1/2. *%Problem 5.10 (a) Calculate ((1/jr; ~ r2|)) for the state Yo (Equation 5.30), Hint: Do the d3rz integral first, using spherical coordinates and setting the polar axis along 1, s0 that fa, 3 in = nal = rf +73 — 2rirrcos®. The 4, integral is easy, but be careful to take the positive root. You'll have to break the r integral into two pieces, one ranging from 0 to r1, the other from 7 to 00. Answer: 5/4a. (b) Use your result in (a) to estimate the electron interaction energy in the ground state of helium. Express your answer in electron volts, and add it to Ey (Equa- tion 5.31) to get a corrected estimate of the ground-state energy. Compare the experimental value. Note: Of course, we're still working with an approximate wave function, so don’t expect perfect agreement.Sec. 5.2: Atoms 189 Parahelium Orthohelium e_eee oS PF 0 2S PDF Energy(eV) (1S at 24.46 eV) Figure 5.2: Energy level diagram for helium (the notation is explained in Section 5.2.2). Note that parahelium energies are uniformly higher than their orthohelium counterparts. The numerical values on the vertical scale are relative to the ground state of ionized helium (He*): 4 x (—13.6 eV) = —54.4 eV; to get the total energy of the state, subtract 54.4 eV. 5.2.2 The Periodic Table The ground-state electron configurations for heavier atoms can be pieced together in much the same way. To first approximation (ignoring their mutual repulsion al- together), the individual electrons occupy one-particle hydrogenic states (n,/, m), called orbitals, in the Coulomb potential of a nucleus with charge Ze. If electrons were bosons (or distinguishable particles), they would all shake down to the ground state (1,0,0), and chemistry would be very dull indeed. But electrons are in fact iden- tical fermions, subject to the Pauli exclusion principle, so only two can occupy any given orbital (one with spin up, and one with spin down—or, more precisely, in the190 Chap. 5 Identical Particles singlet configuration), There are n? hydrogenic wave functions (all with the same energy E;,) for a given value of n, so the n = | shell has room for two electrons, the n = 2 shell holds eight, n = 3 takes 18, and in general the nth shell can accomodate 2n? electrons. Qualitatively, the horizontal rows on the Periodic Table correspond to filling out each shelt (if this were the whole story, they would have lengths 2 8, 18, 32, 50, etc., instead of 2, 8, 8, 18, 18, etc.; we'll see in a moment how the electron-electron repulsion throws the counting off). With helium, the = | shell is filled, so the next atom, lithium (Z = 3), has to put one electron into the n = 2 shell. Now, for» = 2 we can have / = 0 or | = 1; which of these will the third electron choose? In the absence of electron- electron interactions, they both have the same energy (the Bohr energies depend on ”, remember, but not on /). But the effect of electron repulsion is to favor the lowest value of /, for the following reason: Angular momentum tends to throw the electron outward (more formally, the expectation value of r increases with increasing /, for « given n), and the farther out an electron gets, the more effectively the inner electrons screen the nucleus (roughly speaking, the innermost electron “sees” the full nuclear charge Ze, but the outermost electron sees an effective charge hardly greater than e) Within a given shell, therefore, the state with lowest energy (which is to say, the most tightly bound electron) is / = 0, and the energy increases with increasing /. Thus the third electron in lithium occupies the orbital (2,0,0). The next atom (beryllium, with Z =4) also fits into this state (only with “opposite spin”), but boron (Z = 5) has to make use of / = 1. Continuing in this way, we reach neon (Z = 10), at which point the n = 2 shell is filled, and we advance to the next row of the periodic table and begin to populate the n = 3 shell, First there are two atoms (sodium and magnesium) with / = 0, and then there are six with / = | (aluminum through argon). Following argon there “should” be 10 atoms with n = 3 and / = 2; however, by this time the screening effect is so strong that it overlaps the next shell, so potassium (Z = 19) and calcium (Z = 201 choose n = 4, / = 0, in preference to n = 3, / = 2. After that we drop back to pick up the n = 3,/ = 2 stragglers (scandium through zinc), followed by n = 4,1 = | (gallium through krypton), at which point we again make a premature jump to the next row (” = 5) and wait until later to slip in the / = 2 and / = 3 orbitals from the n = 4 shell. For details of this intricate counterpoint, refer to any book on atomic physics.” I would be delinquent if I failed to mention the archaic nomenclature for atomic states, because all chemists and most physicists use it (and the people who make up the Graduate Record Exam Jove this kind of thing). For reasons known best to nineteenth- century spectroscopists, | = 0 is called s (for “sharp”), / = 1 is p (for “principal”). 1 = 2isd (“diffuse”), and/ = 3is f (“fundamental”); after that I guess they ran out of 7See, for example, U. Fano and L. Fano, Basic Physics of Atoms and Molecules (New York: John Wiley & Sons, 1959), Chapter 18, or the classic by G. Herzberg, Atomic Spectra and Atomic Structure (New York: Dover, 1944)Sec. 5.2: Atoms 191 imagination, because the list just continues alphabetically (g, h, i, etc.).* The state of a particular electron is represented by the pair n/, with n (the number) giving the shell and I (the letter) specifying the orbital angular momentum; the magnetic quantum number m is not listed, but an exponent is used to indicate the number of electrons that occupy the state in question. Thus the configuration (182s) pyr 5.33] tells us that there are two electrons in the orbital (1,0,0), two in the orbital (2,0,0), and two in some combination of the orbitals (2,1,1), (2,1,0), and (2,1,—1). This happens to be the ground state of carbon. In that example there are two electrons with orbital angular momentum quantum number 1, so the total orbital angular momentum quantum number L (capital L, instead of /, to indicate that this pertains to the total, not to any one particle) could be 2, 1, or 0. Meanwhile, the two (1s) electrons are locked together in the singlet state, with total spin zero, and so are the two (2s) electrons, but the two (2p) electrons could be in the singlet configuration or the triplet configuration. So the total spin quantum number S (capital, again, because it’s the total) could be 1 or 0. Evidently the grand total (orbital plus spin) J could be 3, 2, 1, or 0. There exist rituals (Hund’s rules’) for figuring out what these totals will be, for a particular atom. The result is recorded as the following hieroglyphic: Seth, (5.34] (where S and J are the numbers, and L the letter—capitalized, this time, because we're talking about the totals). The ground state of carbon happens to be * Py: The total spin is | (hence the 3), the total orbital angular momentum is | (hence the P), and the grand total angular momentum is zero (hence the 0). In Table 5.1 the individual configurations and the total angular momenta (in the notation of Equation 5.34) are listed, for the first four rows of the Periodic Table. x#*Problem 5.11 (a) Figure out the electron configurations (in the notation of Equation 5,33) for the first two rows of the Periodic Table (up to neon), and check your results against Table 5.1. (b) Figure out the corresponding total angular momenta, in the notation of Equation [5.34], for the first four elements. List all the possibilities for boron, carbon, and nitrogen. 8The shells themselves are assigned equally arbitrary nicknames, starting (don't ask me why) with K: the K shell isn = 1, the L shell ism = 2, M ism = 3, and soon (atleast they're in alphabetical order). See, for example, Stephen Gasiorowiez, Quantum Physics (New York: John Wiley & Sons, 1974), Chapters 18 and 19192 Chap. 5 Identical Particles Table 5.1: Ground-state electron configurations for the first four rows of the Periodic Table. Element Configuration 1 oH (is) 2. He (is)? 30 Li (He) (2s) 4 Be (He}(25)* 5B (He)(2s)?(2p) 6c (Hey(2sy' (2p)? 7 N (Hey2s)°2p)5 8 0 (Hey(2s)? 2p)" 9 OF (Hey(2s)P?(2p)5 10_Ne (Hey(2s)(2p)6 11 Na (Ney(35) 12) Mg (Ne)(3: 13 AL (Ne}(3s)*(3p) 14 Si (Ne)(35)°3 p)? Is P (Ne)(35773 p)> 16S (Ney(35)7 Bp)" 7 a (Ne)(3s)73p)> 18 Ar (Ne)(35)2(3 p)® 19K (angs) 20 Ca (Ani4s)? 21 Se (An(4s)? Bd) 2. «Th (An(4sP 3d? 23 VY (Ann(4s)?Bd)> 24 Cr (An(4s)(3d)> 25° Mn (An(4sy'Gd)* 2% Fe (An)(4s)?(3d)® 27 Co (Ant4s)?(Ba)7 28 Ni (An(4sy?Bd)® 29° Cu (Any(4s)3d)'? 30 Zn (Antsy? Ba)? 31 Ga (An(4s)?(3d) (4p) 32 Ge (Antds)?3d) apy? 33° As (ants)? Gd)(ap)> 34 Se (An(4s)?(3d)!(4p)* 35 «Br (Any(4s)?(3d)!¢4 p> 360«Kr (Anj(4sy? 3d)!9(4p)® (c) Hund’s first rule says that, all other things being equal, the state with the highest total spin will have the lowest energy. What would this predict in the case of the excited states of helium?Sec. 5.3: Solids 193 (d) Hund’s second rule says that if a subshell (n, !) is no more than half filled, then the lowest energy level has J = |Z — S|; if it is more than half filled, then J = L+ Shas the lowest energy. Use this to resolve the boron ambiguity in (b). (@) Use Hund’s rules and the fact that a symmetric spin state must go with an antisymmetric position state (and vice versa) to resolve the carbon ambiguity in (b), What can you say about nitrogen? Problem 5.12 The ground state of dysprosium (element 66, in the sixth row of the Periodic Table) is listed as °Jg. What are the total spin, total orbital, and grand total angular momentum quantum numbers? Suggest a likely electron configuration for dysprosium. 5.3 SOLIDS In the solid state, a few of the loosely bound outermost valence electrons in each atom become detached and roam around throughout the material, no longer subject only to the Coulomb field of a specific “parent” nucleus, but rather to the combined potential of the entire crystal lattice. In this section we will examine two extremely primitive models: first, the electron gas theory of Sommerfeld, which ignores all forces (except the confining boundaries), treating the wandering electrons as free particles in a box (the three-dimensional analog to an infinite square well); and second, Bloch’s theory, which introduces a periodic potential representing the electrical attraction of the regularly spaced, positively charged, nuclei (but still ignores electron-electron repulsion). These models are no more than the first halting steps toward a quantum theory of solids, butalready they reveal the critical role of the Pauli exclusion principle in accounting for the “solidity” of solids, and provide illuminating insight into the remarkable electrical properties of conductors, semiconductors, and insulators. 5.3.1 The Free Electron Gas ‘Suppose the object in question is a rectangular solid, with dimensions Iy, ly, fz, and imagine that an electron inside experiences no forces at all, except at the impenetrable walls: Vouyye] PO, Es, ..., with degeneracies d), d2, ds, ... (ie., there are d, distinct one-particle states with the same energy E,). Suppose we put N particles (all with the same mass) into this potential; we are interested in the configuration (Nj, N2, N3, ...), for which there are Nj particles with energy £1, N> particles with energy £2, and so on. Question: How many different ways can this be achieved (or, more precisely, how many distinct states correspond to this particular configuration)? The answer, Q(N:, No, Na, ...)s depends on whether the particles are distinguishable, identical fermions, or identical bosons, so we'll treat the three cases separately.” First, assume the particles are distinguishable. How many ways are there to select (from the N available candidates) the N, to be placed in the first “bin”? Answer: the binomial coefficient, “N choose N": N\_ N! (x) “NN = Nyt S72] For there are N ways to pick the first particle, leaving (N — 1) for the second, and so on: M WN) However, this counts separately the Nj! different permutations of the Nj; particles, whereas we don’t care whether number 37 was picked on the first draw, or on the twenty-ninth draw; so we divide by Nj!, confirming Equation 5.72. Now, how many different ways can those N; particles be arranged within the first bin? Well, there are dj states in the bin, so each particle has d, choices; evidently there are (d,)" possibilities in all. Thus the number of ways to put NV; particles, selected from a total population of N, into a bin containing d; distinct options, is N(N = 1)(N = 2). (N= N+) = The same goes for bin 2, of course, except that there are now only (IV — N;) particles left to work with: (N = Niytas? NIN — Ny — Ny) "The presentation here follows closely that of Amnon Yariv, An Introduction to Theory and Appli- cations of Quantum Mechanics (New York: John Wiley & Sons, 1982)208 Chap. 5 Identical Particles and so on, It follows that O(N, Nay Nay ++) Nia" (W- N)las? (N — Ny — No)at © Ni!CN = Ni)! N2!(N — Ni = No)! N3!(N — Ny — No = dvahah ... oo yh =NEL aS =>. (5.78 (You should pause right now and check this result for the Example in Section 5.4.1— see Problem 5.21.) The problem is a lot easier for identical fermions. Because they are indistir. guishable, it doesn’t matter which particles are in which states—the antisymmetrizs tion requirement means that there is just one N-particle state in which a specific +: of one-particle states is occupied. Moreover, only one particle can occupy any give state. There are d, Nn ways to choose the N,, occupied states in the nth bin,"* so d,! « 20% MN) = TDG ND! Tt (Check it for the Example in Section 5.4.1—see Problem 5.21.) ‘The calculation is hardest for the case of identical bosons. Again, the sym: metrization requirement means that there is just one N-particle state in which a spe: cific set of one-particle states is occupied, but this time there is no restriction on the number of particles that can share the same one-particle state. For the nth bin, the question becomes: how many different ways can we assign N,, identical particle~ to d, different slots? There are many ways to solve this combinatorial problem: ar especially clever method involves the following trick: Let dots represent particles ane crosses represent partitions, so that, for example, if d, = 5 and N, = 7, eexexecexex would indicate that there are two particles in the first state, one in the second, three in the third, one in the fourth, and none in the fifth. Note that there are N,, dots anc (dq, — 1) crosses (partitioning the dots into d, groups). If the individual dots ané crosses were labeled, there would be (N, +d, — 1)! different ways to arrange them But for our purposes the dots are all equivalent—permuting them (N,,! ways) does not '8This should be zero, of course, if Ny, > dy, and it is, provided that we consider the factorial ot « negative integer to be infinite,Sec. 5.4: Quantum Statistical Mechanics 209 change the state. Likewise, the crosses are all equivalent—permuting them [(d, ~ 1)! ways] changes nothing. So there are in fact (Na +dn =U)! _ (Ny +dy—1 Nal(dn — 1)! = Mn ) 781 distinct ways of assigning the N,, particles to the d, one-particle states in the nth bin, and we conclude that ry (Nn + dn — 1)! QM, No, Nay. MAG SDE" [5.76] (Check it for the Example in Section 5.4,1—see Problem 5.21.) «Problem 5.21 Check Equations 5.73, 5.74, and 5.76 for the Example in Section SAL. +*Problem 5.22 Obtain Equation 5.75 by induction. The combinatorial question is this: How many different ways can you put V identical balls into d baskets (never mind the subscript n for this problem). You could stick all V of them into the third basket, or all but one in the second basket and one in the fifth, or two in the first and three in the third and all the rest in the seventh, etc. Work it out explicitly for the cases N = 1, N = 2, N =3, and N = 4; by that stage you should be able to deduce the general formula. 5.4.3 The Most Probable Configuration In thermal equilibrium, every state with a given total energy E and a given particle number N is equally likely. So the most probable configuration (Ny, N2, Ns..-.) is the one that can be achieved in the largest number of different ways—it is that particular configuration for which Q(Nj, No, N3, ...) is a maximum, subject to the constraints > N,=N (5.77] m= and > N,E, = E. [5.78] mt ‘The problem of maximizing a function F(x,, x2, x3, ...) of several variables, subject to the constraints fj (x1, *2,*3,...) = 0, Ax, x2,%3,...) = 0, ete., is most con- veniently handled by the method of Lagrange multipliers”: We introduce the new function See, for example, Mary Boas, Mathematical Methods in the Physical Sciences, 2nd ed. (New York: John Wiley & Sons, 1983), Chapter 4, Section 9210 Chap. 5 Identical Particles G(X, 2, X3,0- At Aa, Fra fitiahte [5.74 and set all its derivatives equal to zero: aG aG OXn, Orn In our case it’s a little easier to work with the logarithm of Q, instead of © itself—this turns the products into sums. Since the logarithm is a monotonic functior of its argument, the maxima of Q and In(Q) occur at the same point. So we let imo) +a[v-Som]va[e- Some]. [5.81 n=l nl where « and f are the Lagrange multipliers. Setting the derivatives with respect to u and B equal to zero merely reproduces the constraints (Equations 5.77 and 5.78): 1: remains, then, to set the derivative with respect to N, equal to zero. If the particles are distinguishable, then Q is given by Equation 5.73, and we have (5.80 G G =In(N1) + SNe In(d,) = INN] vale on] eee [5.82 n=l Assuming that the relevant occupation numbers (N,,) are large, we can invoke Stir- ling’s approximation”: In(z!) © zIn(z)—z forz>>1 [5.83 to write G & YON la(dy) — Ny in(Ny) + Ny — Ny ~ BE_Ny] nl +In(N!) +aN + BE. [5.84 It follows that aG aN, = In(dy) — In(Ny) — @ — BEn (5.85) See George Artken, Mathematical Methods for Physicists, 3rd ed. (Orlando, FL: Academ. Press, 1985), Section 10,3. If the relevant occupation numbers are not large—as in the Example of Sectior 5.4.1—then statistical mechanics simply doesn’t apply. ‘The whole point is to deal with such enormous ‘numbers that statistical inference is a reliable predictor. Of course, there will always be one-particle stat. of extremely high energy that are not populated at all; fortunately, Stirling's approximation holds also for 2 =0. [use the word “relevant” to exclude any stray states right at the margin, for which NV, is neither hhuge nor zero.Sec. 5.4: Quantum Statistical Mechanics 211 Setting this equal to zero and solving for N,, we conclude that the most probable occupation numbers for distinguishable particles are Ny = dye(@*PEo), [5.86] If the particles are identical fermions, then Q is given by Equation 5.74, and we have G= Yana.) — In(N,,!) = In{(dy — Nn) I} vel $ars[e-Encl. [5.87] This time we must assume not only that N,, is large, but also that d,, >> N,,7! so that Stirling’s approximation applies to both terms. In that case Ge Y [me — Ny ln(Nn) + Nn — (da — Nn) In(dn — Nn) aI + (de = Ny) AN, ~ BE,Ns | + 0N + BE, 15.88] so = —In(Nn) + Indy — Nx) — a — BE, [5.89] Setting this equal to zero and solving for Nj, we find the most probable occupation numbers for identical fermions: dy N= earBE) 1" 5.90] Finally, if the particles are identical bosons, then Q is given by Equation 5.76, and we have G = SY llnl(V, + de — I! = nC, nat ta[w— Som] ee] e— Sone]. (5.91) = Inf(@, — 11} 21In one dimension the energies are nondegenerate (see Problem 2.42), but in three dimensions dy typically increases rapidly with increasing n (for example, in the case of hydrogen, dy = n2). So itis not unreasonable to assume that for most of the occupied states dy >> 1. On the other hand, dy is certainly not much greater than N, at absolute zero, where all states up to the Fermi level are filled, and hence dy, = Ny Here again We are rescued by the fact that Stirling’s formula holds also for z = 0.212 Chap. 5 Identical Particles Assuming (as always) that N,, > 1, and using Stirling's approximation: oe G®Y[ONn + dy — Vn + dy — WD) = (Ne + dn = 1) — Nn b(n) + Ny = Inf(dy — 1)!] — Nn — BEN} +aN + BE, [5.92 so aG an, = In(Ny +d, — 1) — In(N,) ~ & — BEn- [5.93. Setting this equal to zero and solving for Nj, we find the most probable occupation numbers for identical bosons: dy —1 Nn = SexpBS 1" 15.94; (For consistency with the approximations already invoked, we should really drop the 1 in the numerator, and I shall do so from now on.) Problem 5.23 Use the method of Lagrange multipliers to find the area of the largest rectangle, with sides parallel to the axes, that can be inscribed in the ellipse (x/a)? + (y/b)? = 1. Problem 5.24 (a) Find the percent error in Stirling’s approximation for z = 10. (b) What is the smallest integer z such that the error is less than 1%? 5.4.4 Physical Significance of a and 6 ‘The parameters or and B came into the story as Lagrange multipliers, associated with the total number of particles and the total energy, respectively. Mathematically. they are determined by substituting the occupation numbers (Equations 5.86, 5.90. and 5.94) back into the constraint equations (Equations 5.77 and 5.78). To carry out the summation, however, we need to know the allowed energies (E,) and their degeneracies (d,) for the potential in question. As an example, I'll work out the case of a three-dimensional infinite square well; this will enable us to infer the physical significance of o and B. In Section 5.3.1 we determined the the allowed energies (Equation 5.39): wo, = am* . [5.95] whereSec, 5.4: Quantum Statistical Mechanics 213 my Why WN, k= sa) (7) As before, we convert the sum into an integral, treating k as a continuous variable, with one state (or, for spin s, 2s + 1 states) per volume 13/V of k-space. Taking as our “bins” the spherical shells in the first octant (see Figure 5.4), the “degeneracy” (that is, the number of states in the bin) is _lanedk Vo py ms * 8 G8/V) 2x i ° For distinguishable particles (Equation 5.86), the first constraint (Equation 5.77) be- comes 00 an f PPE Pm dhe = Vem (a y" 0 npr)” so N (2mpn?\*? > : 5.97] V\om The second constraint (Equation 5.78) says Rope ma of, putting in Equation 5.97 for e~*, 3N baa [5.98] (If you include the spin factor, 2s + 1, in Equation 5.96, it cancels out at this point, so Equation 5.98 is correct regardless of spin.) Equation 5.98 is reminiscent of the classical formula for the average kinetic energy of an atom at temperature 7”: E No? where kg isthe Boltzmann constant. This suggests that f is related to the temperature: 1 bre [5.100] aT, (5.99) 22See, for example, David Halliday and Robert Resnick, Fundamentals of Physics, 3rd ed. extended (New York: John Wiley & Sons, 1988), Section 21-5,214 Chap. 5 Identical Particles To prove that this holds in general, and not simply for distinguishable particles in = three-dimensional infinite square well, we would have to demonstrate that differe~ substances in thermal equilibrium with one another have the same value of B. The argument is sketched in many books,” but I shall not reproduce it here—I will simp! adopt equation [5.100] as the definition of T. Itis customary to replace w (which, as is clear from the special case of Equatio- 5.97, is a function of T) by the so-called chemical potential, u(T) = —akgT, (5.101 and rewrite Equations 5.86, 5.90, and 5.94 as formulas for the most probable numbe* of particles in a particular (one-particle) state with energy € (to go from the numbe: of particles with a given energy to the number of particles in a particular state witt that energy, we simply divide by the degeneracy of the state): ne) = 4 GeweT py FERMIDIRAC [5.102 I ewe? 7" BOSE-EINSTEIN Le - ‘The Maxwell-Boltzmann distribution is the classical result for distinguishable par- ticles; the Fermi-Dirac distribution applies to identical fermions, and the Bose- Einstein distribution is for identical bosons. ‘The Fermi-Dirac distribution has a particularly simple behavior as T — 0: ife 2(0), elewkal {% O°, so 1 » ife <4), 3 « 5.103 no {o ife > (0) | All states are filled, up to an energy j1(0), and none are occupied for energies above this (Figure 5.8). Evidently the chemical potential at absolute zero is precisely the Fermi energy: 40) = Ep. [5.104 As the temperature rises, the Fermi-Dirac distribution “softens” the cutoff, as indicated by the rounded curve in Figure 5.8. For distinguishable particles in the three-dimensional infinite square well, we found (Equation 5.98) that the total energy at temperature 7 is 3 SNkaT: [5.108 Bay 23See, for example, Yariv, footnote 17, Section 15.4.Quantum Statistical Mechanics 215 ne) ja T=0 T>0 Er=1(0) . Figure 5.8 Fermi-Dirac distribution for =0 and for T somewhat above zero. from Equation 5.97 it follows that MT) = keT [m (F) +50 (S85). [5.106] I would like to work out the corresponding formulas for identical fermions and bosons, using Equations 5.90and 5.94 in place of Equation 5.86. The first constraint (Equation 5.77) becomes rf 5.107] Qn? Jy Ql Pm—ul/keT EY] " (with the plus sign for fermions and minus for bosons), and the second constraint (Equation 5.78) reads pet i dk. 5.108 = iein f, aeramara ms The first of these determines (7), and the second determines £(7) (from the latter we obtain, for instance, the heat capacity C = 4£/4T). Unfortunately, the integrals cannot be evaluated in terms of elementary functions, and I shall leave it for you to explore the matter further (see Problems 5.25 and 5.26). Problem 5.25 Evaluate the integrals (Equations 5.107 and 5.108) for the case of identical fermions at absolute zero. Compare your results with Equations 5.43 and 5.45. (Note that for electrons there is an extra factor of 2 in Equations 5.107 and 5.108, to account for the spin degeneracy.) ««xProblem 5.26 (a) Show that for bosons the chemical potential must always be less than the mini- mum allowed energy. Hint: n(€) cannot be negative. (b) In particular, for the ideal bose gas (identical bosons in the three-dimensional infinite square well), (7) < 0 for all T. Show that in this case (7) mono- tonically increases as T decreases, assuming that N and V are held constant. Hint: Study Equation 5.107, with the minus sign.216 Chap. 5 Identical Particles (©) A crisis (called bose condensation) occurs when (as we lower 7) p(T) hit+ zero. Evaluate the integral, for y. = 0, and obtain the formula for the critics temperature 7; at which this happens. Note: Below the critical temperature. the particles crowd into the ground state, and the calculational device of replacinz the discrete sum (Equation 5.77) by a continuous integral (Equation 5.107 loses its validity. See F. Mandl, Statistical Physics (London: John Wiley & Sons, 1971), Section 11.5. Hint: co ysl fa = : [5.10 where I’ is Euler’s gamma function and ¢ is the Riemann zeta function. Loos up the appropriate numerical values, (d) Find the critical temperature for ‘He. Its density, at this temperature, is 0.15 gm/cm?, Note: The experimental value of the critical temperature in4He is 2.17 K. The remarkable properties of “He in the neighborhood of T, are discussed 1- the reference cited in (c). 5.4.5 The Blackbody Spectrum Photons (quanta of the electromagnetic field) are identical bosons with spin 1, but the: are a very special case because they are massless particles, and hence intrinsicalls relativistic, We can include them here, if you are prepared to accept four assertior~ that do not really belong to nonrelativistic quantum mechanics: (1) The energy of a photon is related to its frequency by the Planck formule E=hv=ho. (2) The wave number k is related to the frequency by k = 2/2 = w/c, where is the speed of light. (3) Only two spin states occur (the quantum number m can be +1 of I, but ne" 0). (4) The number of photons is not a conserved quantity; when the temperature rise the number of photons (per unit volume) increases. In view of item 4, the first constraint equation (Equation 5.77) does not apply We can take account of this by simply setting « — 0, in Equation 5.81 and everythins that follows. Thus the most probable occupation number for photons is (Equatier 5.94) dy No = Folie 1 [5.110 For free photons in a box of volume V, dy is given by Equation 5.96, multiplied bs 2 for spin (item 3), and expressed in terms of w instead of & (item 2): 24Jn truth, we have no business using this formula, which came from the (nonrelativistic) Schréding= ‘equation; fortunately, the degeneracy is exactly the same for the relativistic case, See Problem 5.3.2.Sec. 5.4: Quantum Statistical Mechanics 217 <— Visible region —>! Frequency [10'* Hz] Figure 5.9: Planck’s formula for the blackbody spectrum, Equation 5.112. (5.111) [5.112] (eer — 1)" This is Planck’s famous blackbody spectrum, giving the energy per unit volume, per unit frequency, in an electromagnetic field at equilibrium at temperature T. It is plotted, for three different temperatures, in Figure 5.9. Problem 5.27 Use Equation 5.112 to determine the energy density in the wave- length range di. Hint: set p(w)dw = p(d)dd, and solve for p(4). Derive the Wien displacement law for the wavelength at which the blackbody energy density is a maximum: 2.90 x 10-3 mK rT : You'll need to solve the transcendental equation (5 — x) = Se~*, using a calculator (or a computer); get the numerical answer accurate to three significant digits. [5.113]218 Chap. 5 Identical Particles Problem 5.28 Derive the Stefan-Boltzmann formula for the total energy density in blackbody radiation: E_ (kb \ ps 16 tm? == = (7: 107'6 Im-™3K~*) T4. 114) 7 (Ss) (7.57 x 10-18 Im ) Tr (5.114) Hint; Use the hint in Problem 5.26(c) to evaluate the integral. Note that¢(4) = 24/90 FURTHER PROBLEMS FOR CHAPTER 5 Problem 5.29 Suppose you have three particles, and three distinct one-particle states (Wa(x), Wo(x), and w-(x)) are available. How many different three-particle states can be constructed (a) if they are distinguishable particles, (b) if they are iden- tical bosons, and (c) if they are identical fermions? [The particles need not be in different states—Wa(X1)Va(X2)Wa(xa) would be one possibility if the particles are distinguishable.] Problem 5.30 Calculate the Fermi energy for electrons in a two-dimensional infi- nite square well. (Let o be the number of free electrons per unit area.) «#*Problem 5.31 Certain cold stars (called white dwarfs) are stabilized against grav- itational collapse by the degeneracy pressure of their electrons (Equation 5.46). As- suming constant density, the radius R of such an object can be calculated as follows: (a) Write the total electron energy (Equation 5.45) in terms of the radius, the number of nucleons (protons and neutrons) NV, the number of electrons per nucleon q and the mass of the electron m (b) Look up, or calculate, the gravitational energy of a uniformly dense sphere Express your answer in terms of G (the constant of universal gravitation), R, . and M (the mass of a nucleon). Note that the gravitational energy is negative. Find the radius for which the total energy, (a) plus (b), is a minimum. Answer: au (% 2 42g5 4 GmM?N1>" © (Note that the radius decreases as the total mass increases!) Put in the actual numbers, for everything except V, using q = 1/2 (actually, q decreases a bit as the atomic number increases, but this is close enough for our purposes). Answer: R=76 x 108N-1, Determine the radius, in kilometers, of a white dwarf with the mass of the sun. Determine the Fermi energy, in electron volts, for the white dwarf in (d), and compare it with the rest energy of an electron. Note that this system is getting dangerously relativistic (see Problem 5.32) @ (e) x+#Problem 5.32 We can extend the theory of a free electron gas (Section 5.3.1) to the relativistic domain by replacing the classical kinetic energy, E = p?/2m, with the relativistic formula, E = ypc? + mc — mc’. Momentum is related to theFurther Problems for Chapter 5 219 wave vector in the usual way: p = fik. In particular, in the extreme relativistic limit, E® pe =hek (a) Replace h?4?/2m in Equation 5.44 by the ultrarelativistic expression, fick, and (b) ©) ) calculate Eo in this regime. Repeat parts (a) and (b) of Problem 5.31 for the ultrarelativistic electron gas. Notice that in this case there is no stable minimum, regardless of R; if the total energy is positive, degeneracy forces exceed gravitational forces and the star will expand, whereas if the total is negative, gravitational forces win out and the star will collapse. Find the critical number of nucleons N, such that gravitational collapse occurs for N > Ne. This is called the Chandrasekhar limit. Answer: 2.0 x 10°". What is the corresponding stellar mass (give your answer as a multiple of the sun’s mass), Stars heavier than this will not form white dwarfs, but collapse further, becoming (if conditions are right) neutron stars. At extremely high density, inverse beta decay, e~ + p* > n + v, converts Virtually all of the protons and electrons into neutrons (liberating neutrinos, which carry off energy, in the process). Eventually neutron degeneracy pressure stabilizes the collapse, just as electron degeneracy does for the white dwarf (see Problem 5.31). Calculate the radius of a neutron star with the mass of the sun, Also calculate the (neutron) Fermi energy, and compare it to the rest energy of a neutron. Is it reasonable to treat such a star nonrelativistically? +«Problem 5.33 (a) Find the chemical potential and the total energy for distinguishable particles in © « the three-dimensional harmonic oscillator potential (Problem 4.39). Hint: The sums in Equations 5.77 and 5.78 can be evaluated exactly in this case—no need to use an integral approximation, as we did for the infinite square well. Note that by differentiating the geometric series, => . B.E3) n= you can get d(x oe , Z (4) = ye + Dx’ and similar results for higher derivatives. Answer: 1+ —) 7 Tat [5.116] 3 B= pt ( Discuss the limiting case kgT Kho. Discuss the classical limit, kg 7 >> hoo, in the light of the equipartition theorem (see, for example, Halliday and Resnick, footnote 22, Section 21-9). How many degrees of freedom does a particle in the three-dimensional harmonic oscillator Possess?PART II APPLICATIONSCHAPTER 6 TIME-INDEPENDENT PERTURBATION THEORY 6.1 NONDEGENERATE PERTURBATION THEORY 6.1.1 General Formulation Suppose we have solved the (time-independent) Schrédinger equation for some po- tential (say, the one-dimensional infinite square well): Hy = Eins 6.1] obtaining a complete set of orthonormal eigenfunctions, 7°, (ein) = Sums [6.2] and the corresponding eigenvalues E?. Now we perturb the potential slightly (say, by putting a little bump in the bottom of the well—Figure 6.1). We'd like to solve for the new eigenfunctions and eigenvalues: Am = Extn [6.3] but unless we are very lucky, we're unlikely to be able to solve the Schrédinger equation exactly, for this more complicated potential. Perturbation theory is a systematic procedure for obtaining approximate solutions to the perturbed problem by building on the known exact solutions to the unperturbed case.222 Chap. 6 Time-Independent Perturbation Theory vay Figure 6.1: Infinite square well with small perturbation. To begin with, we write the new Hamiltonian as the sum of two terms: H=H°+AH', [64 where H’ is the perturbation, For the moment we'll take A to be a small number; later we'll crank it up to 1, and 7 will be the true, exact Hamiltonian. Writing % and E, as power series in A, we have Un = Ve tab teu tes [os Ey, =ED+ABh + VER +. [6.6] Here E} is the first-order correction to the n™ eigenvalue, and 1; is the first-order correction to the n" eigenfunction; E? and y/2 are the second-order corrections, and so on. Plugging Equations 6.4, 6.5, and 6.6 into Equation 6.3, we have CO FAH YIYE + Wy EWG + = (Ey thE, + PEL + DE FAW, FG or (collecting like powers of A): FR +MY, + HW) + OH? + HG) + = Bye + ME y + En Wy) + OE nda + Egiig + Bayt) too To lowest order (A°) this yields H°y;° = E%y°, which is nothing new (just Equation 6.1). To first order (2), HOG] + Hy, = BOY) + Bly? 16.7) To second order (A), HW? + HW) = Eli, + End) + Be, [6.8] and so on. (I’m done with A, now—it was just a device to keep track of the different orders—so crank it up to 1.)Sec. 6.1: Nondegenerate Perturbation Theory 223 Figure 6.2: Constant perturbation over the whole well. 6.1.2 First-Order Theory Taking the inner product of Equation 6.7 with 2 [that is, multiplying by (¥9)* and integrating], [RW + URL Vin) = ED We Wn) ERR But H° is Hermitian, so (Une) = (ARV) = (Evian) = En Vel Vn)« and this cancels the first term on the right. Moreover, (¥/9|W®) = 1, so! EL = WRlH' lp) 16.9] This is the fundamental result of first-order perturbation theory; as a practical matter, it may well be the most important equation in quantum mechanics. It says that the first-order correction to the energy is the expectation value of the perturbation in the unperturbed state. Example. The unperturbed wave functions for the infinite square well are (Equation 2.24) O¢¢) =f 2 gin (™ vies) = y= sin ( = x): ‘Suppose first that we perturb the system by simply raising the “floor” of the well by a constant amount Vp (Figure 6.2). In that case H’ = Vy, and the first-order correction to the energy of the n" state is E, = WAlMolv) = Volvmlva) = Yo. Nin this context it doesn’t matter whether we write (¥i0|H’y) or (WSLH'Ivo) (with the extra vertical bar) because we are using the wave function itself to “label” the state. "But the latter notation is preferable because it frees us from this specific convention.224 Chap. 6 Time-Independent Perturbation Theory woe Figure 6.3: Constant perturbation over half the well. The corrected energy levels, then, are E,, = E° + Vo; they are simply lifted by the amount Vo. Of course! The only surprising thing is that in this case the first-order theory yields the exact answer. Evidently, for a constant perturbation all the higher corrections vanish? If, on the other hand, the perturbation extends only halfway across the well (Figure 6.3), then 2% fa VK EI = f sin? (=s) d=. ado @ 2 Inthis case every energy level is lifted by Vo/2. That's not the exact result, presumably. but it does seem reasonable as a first-order approximation. Equation 6.9 is the first-order correction to the energy; to find the first-order correction to the wave function we first rewrite Equation 6.7: CH — EX = —(H! — ED [6.10] The right side is a known function, so this amounts to an inhomogeneous differential equation for y;|. Now, the unperturbed wave functions constitute a complete set, so ¥} (like any other function) can be expressed as a linear combination of them: t= Vey. (6.11) mn [There is no need to include m =n in the sum, for if y;) satisfies Equation 6.10, so too does (y} + ay), for any constant a, and we can use this freedom to subtract off the 2 term.’ If we could determine the coefficients c, we'd be done. Well, putting Equation 6.11 into Equation 6.10, and using the fact that the y/® satisfies the Incidentally, nothing here depends on the specific nature of the infinite square well—the same result applies for any potential, when the perturbation is constant 3 Alternatively, a glance at Equation 6.5 reveals that any ¥? component in y! might as well be pulled out and combined with the first term. We are only concerned, for the moment, with solving the Schridinger equation (Equation 6.3), and the Wr we get will not, in general, be normalized,Sec. 6.1: Nondegenerate Perturbation Theory 225 unperturbed Schrédinger equation (Equation 6.1), we have LEY = Eel yh =H! - Edn. Foe Taking the inner product with JP, SES = EDL WP LWA) = — UPL) + EA UP nd man If / =n, the left side is zero, and we recover Equation 6.9; if | # n, we get (EP ~ Enyef” = (PH Wes or ouiriye) oi = ‘Bal a Pa [6.121 so Ob erhy) Y Wel Hn) yo. (6.13) | a &. &) Notice that the denominator is safe, since there is no coefficient with m = n, as long as the unperturbed energy spectrum is nondegenerate. But if two different unperturbed states share the same energy, we're in serious trouble (we divided by zero to get Equation 6.12); in that case we need degenerate perturbation theory, which I'll come to in Section 6.2. That completes first-order perturbation theory: E/ is given by Equation 6.9, and y/! is given by Equation 6.13. Ishould warn you that whereas perturbation theory often yields surprisingly accurate energies (that is, E° + E! is quite close to the exact value E,,), the wave functions are notoriously poor. Problem 6.1 Suppose we put a delta-function bump in the center of the infinite square well: H' = a5(x —a/2), where a is aconstant, Find the first-order correction to the allowed energies. Explain why the energies are not perturbed for even n. «Problem 6.2 For the harmonic oscillator [V (x) = (1/2)kx?], the allowed energies are E,=(a+1/2ho, =0,1,2,..0 where @ = /k/m is the classical frequency. Now suppose the spring constant increases slightly: k — (1 +¢)k. (Perhaps we cool the spring, so it becomes less flexible.)226 Chap. 6 Time-Independent Perturbation Theory (a) Find the exact new energies (trivial, in this case). Expand your formula ay - power series in ¢, up to second order. (b) Now calculate the first-order perturbation in the energy, using Equation 6. What is H’ here? Compare your result with part (a). Hint: Itis not necessary in fact, it is not permitted—to calculate a single integral in doing this problem Problem 6.3 Two identical bosons are placed in an infinite square well (Equativ: 2.15). They interact weakly with one another, via the potential V (x1, x2) = —aV8(x) — x2) (where Vp is a constant with the dimensions of energy and a is the width of the well (a) First, ignoring the interaction between the particles, find the ground state anc first excited state—both the wave functions and the associated energies. (b) Use first-order perturbation theory to calculate the effect of the particle-particle interaction on the ground and first excited state energies. 6.1.3 Second-Order Energies Proceeding as before, we take the inner product of the second-order equation (Equa- tion 6.8) with yo: (URE WE) + WELL Vn) = Eg WR a) + En Walrad + ERA IWR)- Again, we exploit the Hermiticity of H®: Ug OUR) = (HOU Wn) = Entre ns sothe first term on the left cancels the first term on the right. Meanwhile, (|) = 1 and we are left with a formula for £?: = lH) — En nln) But (WRIA) = Do oh wl) Ea so (aL WR WRLH Wa) => oe Ex = (WAL Wn) = DO URL mgr msn or, finally, rio) al Wa 16.18Sec. 6.2: Degenerate Perturbation Theory 227 This is the fundamental result of second-order perturbation theory. We could proceed to calculate the second-order correction to the wave function (12), the third-order correction to the energy, and so on, but in practice Equation 6.14 is ordinarily as high as itis useful to pursue this method. «Problem 6.4 (a) Find the second-order correction to the energies (£2) for the potential in Problem 6.1. Note: You can sum the series explicitly to obtain the result —2m(a/zrhin)?, for odd n. (b) Calculate the second-order correction to the ground-state energy (£3) for the potential in Problem 6.2. Check that your result is consistent with the exact solution. «Problem 6.5 Consider a charged particle in the one-dimensional harmonic oscil- lator potential. Suppose we tum on a weak electric field (E) so that the potential energy is shifted by an amount H’ = —qEx. (a) Show that there is no first-order change in the energy levels, and calculate the second-order correction, Hint: See Problem 3.50. (b) The Schrédinger equation can be solved exactly in this case by a change of variables: x’ = x —(qE/mw*). Find the exact energies, and show that they are consistent with the perturbation theory approximation, 6.2 DEGENERATE PERTURBATION THEORY If the unperturbed states are degenerate—that is, if two (or more) distinct states (2 and Wf) share the same energy—then ordinary perturbation theory fails: c\ (Equation 6.12) and E? (Equation 6.14) blow up (unless, possibly, the numerator vanishes, (¥2|H'|W?) = 0—a loophole that will be important to us later on). In the degenerate case, therefore, there is no reason to trust even the first-order correction to the energy (Equation 6.9), and we must look for some other way to handle the problem. 6.2.1 Twofold Degeneracy Suppose that Hy? = Ey, Howe = Ep, and (y Sly?) =0. [6.15] Note that any linear combination of these states, v° =aw) + BYp, [6.16]228 Chap. 6 Time-Independent Perturbation Theory is still an eigenstate of H", with the same eigenvalue E°: Hy? = E’y?. [6.17 Typically, the perturbation (H’) will “break” the degeneracy: As we increase A (from: Oto 1), the common unperturbed energy E° splits into two (Figure 6.4). The essential problem is this: When we turn off the perturbation, the “upper” state reduces down to one linear combination of 9 and w?, and the “lower” state reduces to some other linear combination, but we don’t know a priori what these “good” linear combinations will be. For this reason we can’t even calculate the first-order energy (Equation 6.9) because we don’t know what unperturbed states to use. For the moment, therefore, let’s just write the “good” unperturbed states in the general form (Equation 6.16), keeping a and f adjustable. We want to solve the Schrédinger equation, Hw = Ey, (6.18) with H = H° +AH’ and E=P 44 E+, paw tay ¢eyree. (6.19) Plugging these into Equation 6.18, and collecting like powers of A, as before, we find HOW +H + HOW!) + = EY EEO + By pee, But H°y° = £°y° (Equation 6.17), so the first terms cancel; at order 4! we have Hwy! + Hy? = Boy! + Ely. [6.20] Taking the inner product with y/°: (WelHw') + WELA' YD) = E°yeiy!) + EL Waly). Because H° is Hermitian, the first term on the left cancels the first term on the right. Putting in Equation 6.16 and exploiting the orthonormality condition Equation 6.15. we obtain ayo He) + BglH' Wp) = aE", *t an Figure 6.4: “Lifting” of a degeneracy by a * perturbation.Sec. 6.2: Degenerate Perturbation Theory 229 or, more compactly, oW q+ BWay = OE, (6.211 where Wij = WH WW)), sj = a,b). (6.22) Similarly, the inner product with w? yields Ws + BWop = BE". [6.23] Notice thatthe 7's are (in principle) known—they are just the “matrix elements” of H’, with respect to the unperturbed wave functions ¥/° and yy. Multiplying Equation 6.23 by W.», and using Equation 6.21 to eliminate Wp, we find @[WapWoq — (E! — Waa)(E' — Wp) = 0. [6.24] If @ is nor zero, Equation 6.24 yields an equation for E': (E')? — E'(Waa + Wen) + WaaWoo — WarW oa) = 0. [6.25] Invoking the quadratic formula, and noting (from Equation 6.22) that Wig = Weiss we conclude that a 1 Bhd [Wea + Want VW an — Wie + AW (6.261 This is the fundamental result of degenerate perturbation theory; the two roots corre- spond to the two perturbed energies. But what if @ is zero? In that case 6 = 1, Equation 6.21 says Wa» = 0, and Equation 6.23 gives E! = W»p. This is actually included in the general result (Equation 6.26), with the plus sign (the minus sign corresponds to a = 1, 6 = 0). What's more, the answers, EL = Woy = (WRIA WS), EL = Waa = (Wall We), are precisely what we would have obtained using nondegenerate perturbation theory (Equation 6.9)—we have simply been lucky: The states yi? and Wf were already the “correct” linear combinations. Obviously, it would be greatly to our advantage if we could somehow guess the “good” states right from the start. In practice, we can often do so by exploiting the following theorem: Theorem: Let A be a Hermitian operator that commutes with H’, If yi and Wh are eigenfunctions of 4 with distinct eigenvalues, Ayi=py, Avvsvys, andy dv, then Wp = 0 (and hence 2 and yf are the “good” states to use in perturbation theory).230 Chap. 6 Time-Independent Perturbation Theory Proof: By assumption, [4, H’] = 0, so (Walla, HWS) = 0 (WalAH'Yy) — (RIM AWS) (AWELH Up) — (WRLH vy (HV) URIH' VE) = (ue = v)Wav. Buty # v, $0 Way = 0. QED Moral: If you're faced with degenerate states, look around for some Hermitian operator A that commutes with H’; pick as your unperturbed states ones that are si- multaneously eigenfunctions of H° and 4. Then use ordinary first-order perturbation theory. If you can’t find such an operator, you'll have to resort to Equation 6.26, but in practice this is seldom necessary. (b) (yA WS) = Problem 6.6 Let the two “good” unperturbed states be VA = a.w) + Baw), where crs. and 6. are determined (up to normalization) by Equation 6.21 (or Equation 6.23), with Equation 6.26 for Ex. Show explicitly that (a) ¥2 are orthogonal (yy) = 0); (©) (WAIH'WY = EL. Problem 6.7 Consider a particle of mass m that is free to move in a one-dimensional region of length L that closes on itself (for instance, a bead which slides frictionlessly on a circular wire of circumference L; Problem 2.43). (a) Show that the stationary states can be written in the form Wax) ern (_Lj2 (#/1)V yields the operator (6.44) is — me. [6.45] The first term is the sozal relativistic energy (not counting potential energy, which we aren’t concerned with at the moment), and the second term is the rest energy— the difference is the energy attributable to motion. We need to express T in terms of theSec. 6.3: The Fine Structure of Hydrogen 237 (relativistic) momentum, mv p=, (6.46) Vi- Ger instead of velocity. Notice that dyed a meh . to a ag mvt + mel = (w/o) __ me a2 = oe EF ST ). perme 1 /eF T= @yper = 7 FO T=Vpet+md—me. [6.47] This relativistic equation for kinetic energy reduces (of course) to the classical result (Equation 6.43), in the nonrelativistic limit p < me; expanding in powers of the small number (p/mc), we have rane | fish) —tf=me [ies (Ey ao 2 5 P p a an 6.48] 2m ~ Bmied * (648) The lowest-order® relativistic contribution to the Hamiltonian is evidently a / B Hs 4s = gg [6.49] In first-order perturbation theory, the correction to E, is given by the expectation value of H’ in the unperturbed state (Equation 6.9): El=(H)= py) = Purley). . = (HN) = aa WIP) = ~ a P WLP) {6.50] Now the Schrédinger equation (for the unperturbed states) says Pw =2m(E-V)y, [6.51] and hence® 1 ae E} =-—S((E-V)) =- E°-2EW)+(v?)]. [6.52] " 2me> 2me SThe kinetic energy of the electron in hydrogen is on the order of 10 eV, which is miniscule compared to its rest energy (511,000 eV), So the hydrogen atom is basically nonrelativistic, and we can afford to keep only the lowest-order correction. In Equation 6.48, p isthe relativistic momentum (Equation 16.46), not the classical momentum mv. I is the former that we now associate with the quantum operator iV, in Equation 6.49, There is some sleight-of-hand in this maneuver, which exploits the Hermiticity of p* and of (E— V), In truth, the operator p* is not Hermitian, for states with ! = 0, and the applicability of perturbation theory to Equation 6,49 is therefore called into question. Fortunately, the exact solution is available; it can be obtained by using the (celativistic) Dirac equation in place of the (nonrelativistic) ‘Schrédinger equation, and it confirms the results we obtain here by less rigorous means, (See Problem 617)238 Chap. 6 Time-Independent Perturbation Theory So far, this is entirely general; but we're interested in the case of hydrogen, for whict: V(r) = —(1/4rene2/r: pan feo, ()dys(2) 4), pss r 2met |" "\ dmg) + ae) ‘2’ )" 7 where E,, is the Bohr energy of the state in question. To complete the job, we need the expectation values of 1/r and 1/r? in the (unperturbed) state nim (Equation 4.89). The first is easy (see Problem 6.11): I 1 () = ==>: 6.54 vr’ wa ' where a is the Bohr radius (Equation 4.72). The second is not so simple to derive (see Problem 6.28), but the answer is’ Il = THe [6.55 It follows that 2 1 ey e 1 =a | £3 +28, (— ) = + (——) a Ime [ nt (&) wat (=) T+ 1a or, eliminating a (using Equation 4.72) and expressing everything in terms of E, (using Equation 4.70), EB E}=- Es 3 [6.56] 7 Ome? LT + 1/2 ” Notice that the relativistic correction is smaller than E,, by a factor of E,/me? ~ 2x 10%. You might have noticed that I used nondegenerate perturbation theory in this calculation even though the hydrogen atom is highly degenerate. But the pertur- bation is spherically symmetrical, so it commutes with L? and L;. Moreover, the eigenfunctions of these operators (taken together) have distinct eigenvalues for the n? states with a given E,. Luckily, then, the wave functions Yj, are “good” states for this problem, so as it happens the use of nondegenerate perturbation theory was legitimate. «Problem 6.11 Use the virial theorem (Problem 4.41) to prove Equation 6.54. “The general formula for the expectation value of any power of ris given in Hans A. Bethe and Edwin E. Salpeter, Quanium Mechanics of One- and Two-Electron Atoms, (New York: Plenum, 1977). p. 17.Sec, 6.3: The Fine Structure of Hydrogen 239 Problem 6.12 In Problem 4.43, you calculated the expectation value of r* in the state ys21. Check your answer for the special cases s = 0 (trivial), s = —1 (Equation 6.54), s = —2 (Equation 6.55), and s = —3 (Equation 6.63). Comment on the case s=-7. +*Problem 6.13 Find the (lowest-order) relativistic correction to the energy levels of the one-dimensional harmonic oscillator. Hint; Use the technique of Problem 2.37. 6.3.2 Spin-Orbit Coupling Imagine the electron in orbit around the nucleus; from the electron’s point of view, the proton is circling around it (Figure 6.7). This orbiting positive charge sets up a magnetic field B in the electron frame, which exerts a torque on the spinning electron, tending to align its magnetic moment (j2) along the direction of the field. The Hamiltonian (Equation 4.157) is H —HB. [6.57] ‘The Magnetic Field of the Proton. If we picture the proton (from the elec- tron’s perspective) as a continuous current loop (Figure 6.7), its magnetic field can be calculated from the Biot-Savart law: _ 2r? with an effective current J = e/T, where ¢ is the charge of the proton and T is the period of the orbit. On the other hand, the orbital angular momentum of the electron (in the rest frame of the nucleus) is L = rmv = 2xmr?/T. Moreover, B and L point in the same direction (up, in Figure 6.7), so 1 e manent [6.58] (Lused ¢ = 1/ /éofia to eliminate 19 in favor of €o.) BL. Figure 6.7; Hydrogen atom, from the electron’s perspective.240 Chap. 6 Time-Independent Perturbation Theory Msi Figure 6.8: A ring of charge, rotating about its axis. ‘The Magnetic Dipole Moment of the Electron. The magnetic dipole me- ment of a spinning charge is related to its (spin) angular momentum; the proportional ity factor is the gyromagnetic ratio (which we already encountered in Section 4.4.2 Let’s derive it, using classical electrodynamics. Consider first a charge q smearec out around a ring of radius r, which rotates about the axis with period 7 (Figure 6.8 The magnetic dipole moment of the ring is defined as the current (q/T) times the area (wr): _ gar =a If the mass of the ring is m, its angular momentum is the moment of inertia (mr times the angular velocity (27/7): 2amr? T The gyromagnetic ratio for this configuration is evidently 1/S = q/2m. Notice that it is independent of r (and T). If [had some more complicated object, such a+ a sphere (all I require is that it be a figure of revolution, rotating about its axis). | could calculate p and $ by chopping it into little rings and adding their contributions. As long as the mass and the charge are distributed in the same manner (so that the charge-to-mass ratio is uniform), the gyromagnetic ratio will be the same for each ring, and hence also for the object as a whole. Moreover, the directions of and $ are the same (or opposite, if the charge is negative), so n (Bs ‘That was a purely classical calculation, however; as it turns out, the electron’s mag- netic moment is twice the classical answer: wn. =-> Bin. then the Zeeman Bethe and Salpeter (footnote 7) page 83. \2The gyromagnetic ratio for orbital motion is just the classical value (q /2m)—it is only for spin that there is an “extra” factor of 2.Sec. 6.4 The Zeeman Effect. 245 effect dominates, and fine structure becomes the perturbation. In the intermediate zone, where the two fields are comparable, we need the full machinery of degenerate perturbation theory, and it is necessary to diagonalize the relevant portion of the Hamiltonian “by hand”. In the following sections we shall explore each of these regimes briefly, for the case of hydrogen. Problem 6.18 Use Equation 6.58 to estimate the internal field in hydrogen, and characterize quantitatively a “strong” and “weak” Zeeman field. 6.4.1 Weak-Field Zeeman Effect If Bet X Bin, fine structure dominates (Equation 6.66); the “good” quantum num- bers are n, 1, /, and m, (but not m; and m,, because—in the presence of spin-orbit coupling—L and § are not separately conserved). In first-order perturbation theory, the Zeeman correction to the energy is Ef = (nl jrmj|Hen1 jm;) = Bey (Le + 28). [6.71] Now L + 2S = J +S; unfortunately, we do not immediately know the expectation value of S, But we can figure it out as follows: The total angular momentum J = L+S is constant (Figure 6.10); Land § precess rapidly about this fixed vector. In particular, the (time) average value of S is just its projection along J: S-D J. [6.72] Save = 2 But L = J-S, so L? = P + S* —2J-S, and hence 1 h S.J= - -Dy= ZUG tD+s6+H-1d4+D), [6.73] from which it follows that 429 = (1454) y=[14 GG tM ~— 10+ 1) +3/4 - (6.74) 2+) Jo e741 Figure 6.10: In the presence of spin-orbit i coupling, Land S are not separately conserved; they precess about the fixed total angular momentum, J.246 Chap. 6 Time-Independent Perturbation Theory =13.60V(1 + 02/4) Figure 6.11: Weak-field Zeeman splitting of the ground state; the upper line (m; = 1/2) has slope 1, the lower line (mt; = 1/2) has slope —1 The term in square brackets is known as the Landé g-factor, g. We may as well choose the z-axis to lie along Bex; then E} = pgs Boum), [6.75 where on a= 5.788 x 1075 eV/T [6.76 is the so-called Bohr magneton. The roral energy is the sum of the fine-structure part (Equation 6.66) and the Zeeman contribution (Equation 6.75). For example. the ground state (n = 1,/ = 0, j = 1/2, and therefore g, = 2) splits into two levels: — 13.6eV(1 +07/4) + ua Bex, [67° with the plus sign for my = 1/2, and minus for m; = —1/2. These energies are plotted (as functions of Bex.) in Figure 6.11. Problem 6.19 Consider the (eight) n = 2 states, |2 / j m,). Find the energy o: each state, under weak-field Zeeman splitting, and construct a diagram like Figure 6.11 to show how the energies evolve as Bex: increases. Label each line clearly, anc indicate its slope 6.4.2 Strong-Field Zeeman Effect If Bext >> Bin, the Zeeman effect dominates'*; with Be in the z direction, the “good quantum numbers are now n, /, m;, and m, (but not j and m ; because—in the presence of the external torque—the total angular momentum is not conserved, whereas L; and tn this regime the Zeeman effect is also known as the Paschen-Back effect.Sec. 6.4 The Zeeman Effect 247 S, are). The Zeeman Hamiltonian is Hy = 5 Boy(Le +28), and the “unperturbed” energies are 13.6 eV re Enmin, =~ + Hep Bex + 2m,). (6.78] In first-order perturbation theory, the fine-structure correction to these levels is Ek = (nl my ms\(H) + Hi,)\nl mms) [6.79] The relativistic contribution is the same as before (Equation 6.56); for the spin-orbit term (Equation 6.60) we need (SL) = (S,)(Lx) + (Sy)(Ly) + (SL) = mam, [6.80] (note that (S,) = (Sy) = (Lx) = (Ly) = 0 for eigenstates of S, and L;). Putting all this together (Problem 6.20), we conclude that 13.6eV {3 [10+1)—mm, EL = a? | — — |] 6.81 ts m {2 (itoares oy (The term in square brackets is indeterminate for ] = 0: its correct value in this case is 1—see Problem 6.22.) The ‘otal energy is the sum of the Zeeman part (Equation 6.78) and the fine-structure contribution (Equation 6.81). Problem 6.20 Starting with Equation 6.79 and using Equations 6.56, 6.60, 6.63, and 6.80, derive Equation 6,81 Problem 6.21 Consider the eight n = 2 states, |2 /m;m,). Find the energy of each state, under strong-field Zeeman splitting. (Express your answers as the sum of three terms, as in Equation 6.77: the Bohr energy; the fine structure, proportional to a?; and the Zeeman contribution, proportional to 11 Bex..) If you ignore fine structure altogether, how many distinct levels are there, and what are their degeneracies? Problem 6.22 If / = 0, then j = s, mj = m,, and the “good” states are the same (|n m,)) for weak and strong fields. Determine £' (from Equation 6.71) and the fine structure energies (Equation 6.66), and write down the general result for the 1 = 0 Zeeman effect—regardless of the strength of the field. Show that the strong- field formula (Equation 6.81) reproduces this result, provided that we interpret the indeterminate term in square brackets as 1.248 Chap. 6 Time-Independent Perturbation Theory 6.4.3 Intermediate-Field Zeeman Effect In the intermediate regime, neither H), nor Hf, dominates, and we must treat the twe on an equal footing, as perturbations to the Bohr Hamiltonian (Equation 6.41): = Hi, + Hh. [6.82] T’ll confine my attention here to the case n = 2 and choose as the basis for de- generate perturbation theory the states characterized by /, j, and m,." Using the Clebsch-Gordan coefficients (Problem 4.45 or Table 4.7) to express | mj) asa linear combination of | m;)|s ms}, we have _p} M1 = 134) = 10013 1=of IpZ) = 10013 WS 3 } Db. let 13 VPNs 3) + VTP 1), 13 —VI73ILOZ 5) + Y2PILIE Zs w= VIBN DI 3 + y273|10)| We = 1 -VF7Il 1143) + VT73I10)I In this basis the nonzero matrix elements of Hf, are all on the diagonal, and given by Equation 6.65; H’, has four off-diagonal elements, and the complete matrix —W is (see Problem 6.23) sy-B 0 0 0 0 0 0 0 0 Sy+B 0 0 0 0 0 0 0 0 y-2 0 0 0 0 0 0 0 0 y+ 0 0 0 0 0 0 0 o yi) 4B 0 0 0 0 0 0 2B sy — 4B 0 0 0 0 0 0 0 0 y+ip SB 0 0 0 0 0 0 2p sy +p where y =(@/8)713.6eV and B= ps Bu. The first four eigenvalues are displayed along the diagonal; it remains only to find the eigenvalues of the two 2 x 2 blocks. The characteristic equation for the first is + My — B) + Sy? — 4 You can use /, my, my states if you prefer—this makes the matrix elements of H, easier but those of Hf, more difficult; the 1” -matrix will be more complicated, but its eigenvalues (which are independent of basis) are the same either way.Sec. 6.4 The Zeeman Effect 249 Table 6.2: Energy levels for the n = 2 states of hydrogen, with fine structure and Zeeman splitting. a 2 8 4 6 6 @ & Ey - Sy +B Fy —5y-B B,-y-2B Ex — 3y + B/2+ Ay? + (2/3)¥B + B2/4 Ey ~3y + 8/2~ Jay? + QAP + PA Ex —3y — B/2+ v/4y? ~ 2/378 + B/4 Bx —3y — B/2— v/4y? — 2/3)yB + BIA and the quadratic formula gives the eigenvalues: da = —3y + (B/2) + V4y? + 2/3)7B + (87/4). [6.83] The eigenvalues of the second block are the same, but with the sign of B reversed. The eight energies are listed in Table 6.2, and plotted against Bex. in Figure 6.12. In the zero-field limit (B = 0) they reduce to the fine-structure values; for weak fields (B < y) they reproduce what you got in Problem 6.19; for strong fields (8 > y) we recover the results of Problem 6.21 (note the convergence to five distinct energy levels, at very high fields, as predicted in Problem 6.21). Problem 6.23 Work out the matrix elements of Hi, and Hj, and construct the W-matrix given in the text, for n = 2. +4Problem 6.24 Analyze the Zeeman effect for the n = 3 states of hydrogen in the weak, strong, and intermediate field regimes. Construct a table of energies (analogous vat irre wre 6.12: Zeeman splitting of the n = 2 states of hydrogen in the weak, he8, intermediate, and strong field regimes.250 Chap. 6 Time-Independent Perturbation Theory to Table 6.2), plot them as functions of the external field (as in Figure 6.12), and chech that the intermediate-field results reduce properly in the two limiting cases. 6.5 HYPERFINE SPLITTING The proton itself constitutes a magnetic dipole, though its dipole moment is much smaller than the electron’s because of the mass in the denominator (Equation 6.59): e =—s,, u,=-—S. 6.84 Ho=am,se Hs me (6.84) (The proton is a composite structure, made up of three quarks, and its gyromagnetic ratio is not as simple as the electron’s—hence the g-factor,'* whose measured value is 5.59 as opposed to 2.00 for the electron.) According to classical electrodynamics. adipole ps sets up a magnetic field'® 7, noe ‘“yr- pl +2 Suse [6.85] So the Hamiltonian caution 6.57) of the secon in the magnetic field due to the proton’s magnetic dipole moment, is Hoge [3(Sp FS. 7 8armpme r ) Sp Sel Ho! ys = ht + om pMe —=_S, -S.5°(r). [6.86] According to perturbation theory, the first-order correction to the energy (Equa- tion 6.9) is the expectation value of the perturbing Hamiltonian: Hoge? 3(Sp + F)(Se +) — Sp: Se 1 P= Som pe Pp ) 4 ose? Sp SHWE. [6.87] 3m pm, In the ground state (or any other state for which / = 0) the wave function is spherically symmetrical, and the first expectation value vanishes (see Problem 6.25). Meanwhile. from Equation 4.80 we find that |y00(0)? = 1/Gra°), so [6.88] 'SThe Landé g-factor, in Equation 6.74, plays a similar role in the proportionality between the electron’s total magnetic moment (2; + #2,) and its total angular momentum J. 161 you are unfamiliar with the delta function term in Equation 6.85, you can derive it by treating the dipole as a spinning charged spherical shell, in the limit as the radius goes to zero and the charge goes to infinity (with px held constant). See D. J. Griffiths, Am. J. Phys. 50, 698 (1982).Sec. 6.5: Hyperfine Splitting 251 Triplet Unperturbed a ———s \ ‘ ‘ ‘ ‘ \ \ \ 1 Singlet Figure 6.13: Hyperfine splitting in the ground state of hydrogen, in the ground state. This is called spin-spin coupling because it involves the dot product of two spins (contrast spin-orbit coupling, which involves § - L). In the presence of spin-spin coupling, the individual spin angular momenta are no longer conserved; the “good” states are eigenvectors of the foral spin, S=S,+S,. [6.89] As before, we square this out to get [6.90] (3/4)h?. In the triplet 2h, in the singlet state the But the electron and proton both have spin 1/2, so S2 state (spins “parallel”) the total spin is 1, and hence S° total spin is 0, and S? = 0. Thus 4gn* +1/4, (iplet); 3mpmica® | —3/4, (singled. oon) Eup = Spin-spin coupling breaks the spin degeneracy of the ground state, lifting the triplet configuration and depressing the singlet (see Figure 6.13). The energy gap is evidently gh* AE = —2_ 3mpmaca® = 5.88 x 10-%ev. [6.92] The frequency of the photon emitted in a transition from the triplet to the singlet state is _ AE vs h and the corresponding wavelength is c/v = 21 cm, which falls in the microwave region. This famous “21-centimeter line” is among the most pervasive and ubiquitous forms of radiation in the universe. = 1420 MHz, [6.93]252 Chap. 6 Time-Independent Perturbation Theory Problem 6.25 Leta and b be two constant vectors, Show that fe -A\(b-#) sind dodo = fe b). 16.94 The integration is over the usual range: 0 < @ < 1,0 <@ < 2m. Use this result w demonstrate that 3(Sp -FSe-?) — Sp -Se (eee for states with / = 0. Hint: # = sin8 cos $i + sin sing} + cos ok. Problem 6.26 By appropriate modification of the hydrogen formula, determine the hyperfine splitting in the ground state of (a) muonic hydrogen (in which a muon— same charge and g-factor as the electron, but 207 times the mass—substitutes for the electron), (b) positronium (in which a positron—same mass and g-factor as the electron, but opposite charge—substitutes for the proton), and (c) muonium (in which an antimuon—same mass and g-factor as a muon, but opposite charge—substitutes for the proton). Hint: Don’t forget to use the reduced mass (Problem 5.1) in calculating the “Bohr radius” of these exotic “atoms.”. Incidentally, the answer you get for positronium (4.85 x 10~* eV) is quite far from the experimental value (8.41 x 10-* eV); the large discrepancy is due to pair annihilation (e* +e" —> y + y), which contributes an extra (3/4)A E and does not occur (of course) in ordinary hydrogen. muonic hydrogen, or muonium. See Griffiths (footnote 16) for further details. FURTHER PROBLEMS FOR CHAPTER 6 +#Problem 6.27 Suppose the Hamiltonian H, for a particular quantum system. is a function of some parameter A; let £,(4) and (2) be the eigenvalues and eigenfunctions of H(2). The Feynman-Hellmann theorem states that dE, oH ryan (ala lvn) [6.95] (assuming either that ,, is nondegenerate, or—if degenerate—that the y,’s are the “good” linear combinations of the degenerate eigenfunctions). (a) Prove the Feynman-Hellmann theorem. Hint: Use Equation 6.9. (b) Apply it to the one-dimensional harmonic oscillator, (i) using 4 = o (this yields a formula for the expectation value of V), (ii) using 4 = hi (this yields (7), and (iii) using 4 = m (this yields a relation between (7) and (V’)). Compare your answers to Problem 2.37 and the virial theorem predictions (Problem 3.53)Further Problems for Chapter6 253 «Problem 6.28 The Feynman-Hellmann theorem (Problem 6.27) can be used to determine the expectation values of I/r and 1/r? for hydrogen.!” The effective Hamiltonian for the radial wave functions is (Equation 4.53) Pe Wit) 2 1 and the eigenvalues (expressed in terms of /)'* are (Equation 4.70) met * Cinax + 1+ 12 (a) Use 4 = e in the Feynman-Hellmann theorem to obtain (1/r). Check your result against Equation 6.54. (b) Use 4 = / to obtain (1/r?). Check your answer with Equation 6.55. x«+Problem 6.29 Prove Kramers’ relation: 8 Fhe) 2s + Dale!) + 1Al+ = ]a{r') =0, [6.96] w which relates the expectation values of r to three different powers (s, s — 1, and § — 2), for an electron in the state Ynim of hydrogen. Hint; Rewrite the radial equation (Equation 4.53) in the form ld+1 2 v= [S |. r ar and use it to express f (ur u")dr in terms of (r*), {r°~), and (r*2). Then use integra- tion by parts to reduce the second derivative. Show that f(uru')dr = —(s/2)(r°"}), and f(u'rtu')dr = —[2/(s + 1) f(u’r**!w)dr. Take it from there. Problem 6.30 (a) Plug s = 0, s = 1, s = 2, and s = 3 into Kramers’ relation (Equation 6.96) to obtain formulas for (r~'), (r), {r?), and (r3). Note that you could continue indefinitely, to find any positive power. (b) In the other direction, however, you hit a snag. Put in s = —1, and show that all you get is a relation between (r~) and (r~*). "7G, Sénchez del Rio, Am. J. Phys., 50, 556 (1982); H. S. Valk, Am. J. Phys., 54, 921 (1986), 'SIn part (b) we treat / as a continuous variable; n becomes a function of , according to Equation 4.67, because jmax» Which must be an integer, is fixed, To avoid confusion, J have eliminated n, to reveal the dependence on ! explicitly.254 Chap. 6 Time-Independent Perturbation Theory (c) But if you can get (r~*) by some other means, you can apply the Kramers’ relation to obtain the rest of the negative powers. Use Equation 6.55 (which is derived in Problem 6.28) to determine (r~*}, and check your answer against Equation 6.63. we*Problem 6.31 When an atom is placed in a uniform external electric field Bex. the energy levels are shifted—a phenomenon known as the Stark effect. In this problem we analyze the Stark effect for the n = 1 and n = 2 states of hydrogen. Let the field point in the z direction, so the potential energy of the electron is Hy = —e Boyz = —@ Beat 008 8. ‘Treat this as a perturbation on the Bohr Hamiltonian (Equation 6.41); spinis irrelevant to this problem, so ignore it. (a) Show that the ground-state energy is not affected by this perturbation, in first order. (b) The first excited state is fourfold degenerate: 200, W211, W210, W21-1. Using de- generate perturbation theory, determine the first-order corrections to the energy. Into how many levels does E> split? «c ‘What are the “good” wave functions for part (b)? Find the expectation value of the electric dipole moment (pe = —er), in each of these “good” states. Notice that the results are independent of the applied field—evidently hydrogen in its first excited state can carry a permanent electric dipole moment. Hint. There are a lot of integrals in this problem, but almost all of them are zero. So study each one carefully before you do any calculations: If the ¢ integral vanishes, there’s not much point in doing the r and @ integrals! Partial answer: W\3 = W3, = 3ea Ecxt; all other elements are zero. wxProblem 6.32 Consider the Stark effect (Problem 6.31) for the n = 3 states of hydrogen. There are initially nine degenerate states, Yratm (neglecting spin, of course), and we turn on an electric field in the z direction. (a) Construct the 9 x 9 matrix representing the perturbing Hamiltonian. Partial an- swer: (3.00|z/3 10) = —3V6a, (31 0[z|320) = —3V3a, (31 £1|z|32£1) = —(9/2)a. (b) Find the eigenvalues and their degeneracies. Problem 6.33 Calculate the wavelength, in centimeters, of the photon emitted under a hyperfine transition in the ground state (n = 1) of deuterium. Deuterium is “heavy” hydrogen, with an extra neutron in the nucleus. The proton and neutron bind together to form a deuteron, with spin 1 and magnetic momentFurther Problems for Chapter 6 255 Figure 6.14: Hydrogen atom surrounded by six point charges (crude model for a crystal lattice); Problem 6.34. the deuteron g-factor is 1.71. 4*Problem 6.34 Inacrystal, the electric field of neighboring ions perturbs the energy levels of an atom. As a crude model, imagine that a hydrogen atom is surrounded by three pairs of point charges, as shown in Figure 6.14. (Spin is irrelevant to this problem, so ignore it.) (a) Assuming that r < d),r < dz, and r < d3, show that H! = Vo + 3(Bix? + Boy? + Bsz*) — (Bi + Ba + Bsr, where and Vo = 2(Bid; + Bad} + Bsd3). (b) Find the lowest-order correction to the ground-state energy. (c) Calculate the first-order corrections to the energy of the first excited states (= 2), Into how many levels does this fourfold degenerate system split, (i) in the case of cubic symmetry, 6; = f2 = fs; (ii) in the case of orthorhombic symmetry, 8; = 2 % fs; (iii) in the general case of tetragonal symmetry (all three different)?CHAPTER 7 THE VARIATIONAL PRINCIPLE 7.1 THEORY Suppose you want to calculate the ground-state energy Eg for a system described by the Hamiltonian H, but you are unable to solve the (time-independent) Schrédinger equation. Pick any normalized function whatsoever. Theorem: (lly) (7.1) That is, the expectation value of H in the (presumably incorrect) state y is certain to overestimate the ground-state energy. Of course, if y just happens to be one of the excited states, then obviously (H) exceeds Eg; but the theorem says that the same holds for any whatsoever. Proof: Since the (unknown) eigenfunctions of H form a complete set, we can express Was a linear combination of them!: ye Yoda with HW, = EnVn- ‘If the Hamiltonian admits scattering states, as well as bound states, then we'll need an integral as well as a sum, but the argument is unchanged.Sec. 7.1: Theory 257 Since y is normalized, 1= (WD) = (Demin > endn) =D Deen Winl Vin) = Yo enl? (assuming the eigenfunctions have been orthonormalized: (Yim|Vn) = Sn). Mean- while, (HY = (Sem ml enn) = Yh Enc Vn Yin) = D0 Elen? But the ground-state energy is, by definition, the smallest eigenvalue, so Eg < En. and hence (H) > Be le QED Example 1. Suppose we want to find the ground-state energy for the one- dimensional harmonic oscillator: H i a td ma?x? = + 5 morx?. 2m dx?” 2 Of course, we already know the exact answer, in this case (Equation 2.49): Eq = (1/2yhe; but this makes it a good test of the method, We might pick as our “trial” wave function the gaussian, yx) = de," (7.21 where b is a constant and A is determined by normalization: (73) Now (H) = (T)+(V), (7.4) where, in this case, 17.5] and 50 {7.6}258 Chap. 7 The Variational Principle According to the theorem, this exceeds Ey for any b; to get the tightest bounc let’s minimize (H) with respect to b: d wma? mo =H) == = 0 = b ao! ) 2m 8b? 2h Putting this back into (#7), we find 7 In this case we hit the ground-state energy right on the nose—because (obviously | | “just happened” to pick a trial function with precisely the form of the actual grouné state (Equation 2.48), But the gaussian is very easy to work with, so it’s a popula: trial function even when it bears little resemblance to the true ground state. Example 2. Suppose we’re looking for the ground state energy of the delta: function potential: H= (x). Again, we already know the exact answer (Equation 2.109): Eg = —ma?/2h?. As before, we'll use a gaussian trial function (Equation 7.2). We've already determined the normalization and calculated (7'); all we need is Wy= -aiar f eo 5(x) dx = 20 Evidently, d wa ao) = om ~ Faeb So [7.9] which is indeed somewhat higher than Eg, since 7 > 2. I said you can use any (normalized) trial function y+ whatsoever, and this is true in a sense. However, for discontinuous functions it takes some fancy footwork to assign a sensible interpretation to the second derivative (which you need, in order to calculate (7). Continuous functions with kinks in them are fair game, however: the next example shows how to handle them.Sec. 7.1: Theory 259 Example 3. Find an upper bound on the ground-state energy of the one- dimensional infinite square well (Equation 2.15), using the “triangular” trial wave function (Figure 7.1)?: Ax, if0 72). The variational principle (as Equation 7.1 is called) is extremely powerful. and embarrassingly easy to use. What a chemist does, to find the ground-state energy of some complicated molecule, is write down a trial wave function with a large number of adjustable parameters, calculate (#), and tweak the parameters to get the lowest possible value. Even if y has no relation to the true wave function, one often gets miraculously accurate values for Ey. Naturally, if you have some way of guessing a realistic ys, so much the better. The only trouble with the method is that you never know for sure how close you are to the target—all you can be certain of is that you've got an upper bound. Moreover, the technique applies only to the ground state (see. however, Problem 7.4). «Problem 7.1 Use the gaussian trial function (Equation 7.2) to obtain the low- est upper bound you can on the ground-state energy of (a) the linear potential: V (x) =ax|; (b) the quartic potential: V (x) = ax’. +*Problem 7.2 Find the best bound on E for the one-dimensional harmonic oscil- lator using a trial wave function of the form 4 + BF where A is determined by normalization and 6 is an adjustable parameter. w(x) Problem 7.3 Find the best bound on £g for the delta-function potential —a5 (x — 4/2), using the triangle trail function (Equation 7.10). (This time a is an adjustable parameter.)Sec. 7.2: The Ground State of Helium 261 Problem 7.4 (a) Prove the following corollary to the variational principle: If (y/|¥rg) = 0, then (H) > Ey, where Ey is the energy of the first excited state. ‘Thus, if we can find a trial function that is orthogonal to the exact ground state, we can get an upper bound on the first excited state. In general, it’s difficult to be sure that ¥ is orthogonal to Yg, since (presumably) we don’t know the latter. However, if the potential V (x) is an even function of x, then the ground state is likewise even, and hence any odd trial function will automatically meet the condition for the corollary. (b) Find the best bound on the first excited state of the one-dimensional harmonic oscillator using the trial function W(x) = Axe Problem 7.5 (a) Use the variational principle to prove that first-order nondegenerate perturbation theory always overestimates (or at any rate never underestimates) the ground- state energy. (b) In view of (a), you would expect that the second-order correction to the ground state is always negative. Confirm that this is indeed the case, by examining Equation 6.14. 7.2. THE GROUND STATE OF HELIUM The helium atom (Figure 7.3) consists of two electrons in orbit around a nucleus containing two protons (also some neutrons, which are irrelevant to our purpose). The Hamiltonian for this system (ignoring fine structure and smaller corrections) is ie 2 (22 1 {Faw (2452-——). 7, Im + ~ gre nt a od (na) Our problem is to calculate the ground-state energy, Ey—the amount of energy it would take to strip off the two electrons. (Given Ey it is easy to figure out the “ionization energy” required to remove a single electron—see Problem 7.6.) E, has been measured very accurately in the laboratory: Eg = —78.975 eV (experimental). [7.15] This is the number we would like to reproduce theoretically.262 Chap. 7 The Variational Principle Warf 2 The helium atom. +26 Figure 7. It is curious that such a simple and important problem has no known exact solution.’ The trouble comes from the electron-electron repulsion, e 1 Veo = ——. Ameo |r — Fal [7.16] If we ignore this term altogether, H splits into two independent hydrogen Hamilto- nians (only with a nuclear charge of 2e, instead of e); the exact solution is just the product of hydrogenic wave functions: 8 pansnie, an na and the energy is 8E; = —109 eV (Eq. [5.31]).4 This is a tong way from —79 eV, but it’s a start. ‘To get a better approximation for £,, we'll apply the variational principle, using wo as the trial wave function. This is a particularly convenient choice because it’s an eigenfunction of most of the Hamiltonian: Vol, F2) = Vrooll:)Wroo(r2) Ho = (BE + Vado. (7.18) Thus UH) = 8B, + Wu 17.19] where! to _ 8 etrtnia ee) = (=) (43) [Seren (720) >There do exist exactly soluble three-body problems with many of the qualitative features of helium, but using non-Coulombic potentials (see Problem 7.15). ‘Here a is the ordinary Bohr radius and Ey = —13.6/n? eV is the nth Bohr energy; recall that for a nucleus with atomic number Z, E, + Z?E, anda > a/Z (Problem 4.17). The spin configuration associated with Equation 7.17 will be antisymmetric (the singlet), You can, if you like, interpret Equation 7.19 as first-order perturbation theory, with Vee as H’ However, I regard this as a misuse of the method, since the perturbation is roughly equal in size to the unperturbed potential. I prefer, therefore, to think of it asa variational calculation, in which we are looking for an upper bound on E.Sec. 7.2: The Ground State of Helium 263 % Figure 7.4: Choice of coordinates for the r; integral (Equation 7.20). I'l do the rp integral first; for this purpose r; is fixed, and we may as well orient the Tr» coordinate system so that the polar axis lies along r) (see Figure 7.4). By the law of cosines, [ny rel = yr? +73 — 2rirrcos bs, [7.21] and hence 4rfa ~arzja fu an = f pr sin dred pdf. [7.22] Ir: - rol : |r? + 13 — Urir»cos 6 The @ integral is trivial (277); the 6 integral is [tn 0 1? + 73 — 2ryrycos® nn (Gta = in ave + 42nn- Vite =2nn) an, ifn n. h=4n (Gf ewrtans f ety, dn) ri Jo n = im +r) —In- nil = { 17.23] nr (7.24]264 Chap. 7 The Variational Principle It follows that (V;.) is equal to e 8 2n\ anja] anja. (ex) (ae) fl - ( + ae aril |: 47/47, sin, drid0,dy. The angular integrals are easy (47), and the r; integral becomes id 2r? Sa? —4r/a ar) -tr/a = f [re (r+ oe Jar ee (ve) = 2 (2) = =a (ss) — (H) = -109 eV + 34 eV = —75 eV. [7.26 Not bad (remember, the experimental value is —79 eV). But we can do better. Can we think of a more realistic trial function than Y (which treats the two electrons as though they did not interact at all)? Rather than completely ignoring the influence of the other electron, let us say that, on the average, each electron represents a cloud of negative charge which partially shields the nucleus, so that the other electron actually sees an effective nuclear charge (Z) that is somewhat less than 2 This suggests that we use a trial function of the form Finally, then, (7.28 and therefore wrt: (7.27 We'll treat Z as a variational parameter, picking the value that minimizes (1). This wave function is an eigenstate of the “unperturbed” Hamiltonian (neglect- ing electron repulsion), but with Z, instead of 2, in the Coulomb terms. With this in mind, we rewrite H (Equation 7.14) as follows: e (ZZ H = -—(V} + V3) — —+— 2 nt Are rr [7.28] e (= (Z-2) 1 ) ale + +—— |. Are ni n irr r2] The expectation value of H is evidently 1 (H) = 227 +20 ya + (Vee) [7.29] 469) r Here (1/r) is the expectation value of 1/r in the (one-particle) hydrogenic ground state yoo (but with nuclear charge Z); according to Equation 6.54, [7.30]Sec. 7.2: The Ground State of Helium 265. ‘The expectation value of ze is the same as before (Equation 7.25), except that instead of Z = 2 we now want arbitrary Z—so we multiply a by 2/Z: 2 Veo) = 22 (<5) = 8a \ aren Putting all this together, we find (73) (H) = [227 —42Z(Z -— 2) — (6/4)Z] By = [-22? + 27/4)ZJ Es. 17.32] According to the variational principle, this quantity exceeds of Z. The lowest upper bound occurs when (H) is minimized: d git) = I-AZ + 27/4) Ei = 0. for any value from which it follows that n Z=—=1.69. 7.33. 16 (7.33] This is a reasonable result; it tells us that the other electron partially screens the nucleus, reducing its effective charge from 2 down to 1.69. Putting in this value for Z, we find 3\6 ) E, =-775 eV. [7.34] The ground state of helium has been calculated with great precision in this way, using increasingly complicated trial wave functions with more and more adjustable parameters.’ But we're within 2% of the correct answer, and, frankly, at this point my own interest in the problem begins to fade. Problem 7.6 Using £, = —79.0 eV for the ground-state energy of helium, cal- culate the ionization energy (the energy required to remove just one electron). Hint: First calculate the ground-state energy of the helium ion, He*, with a single electron orbitting the nucleus; then subtract the two energies. +Problem 7.7 Apply the techniques of this Section to the H~ and Li* ions (each has two electrons, like helium, but nuclear charges Z = 1 and Z = 3, respectively). Find the effective (partially shielded) nuclear charge, and determine the best upper bound on E,, for each case. Note: In the case of H~ you should find that (H) > —13.6 eV, which would appear to indicate that there is no bound state at all, since it is energet- ically favorable for one electron to fly off, leaving behind a neutral hydrogen atom. This is not entirely surprising, since the electrons are less strongly attracted to the nucleus than they are in helium, and the electron repulsion tends to break the atom apart. However, it turns out to be incorrect. With a more sophisticated trial wave E. A, Hylleraas, Z Phys. 65, 209 (1930); C. L, Pekeris, Phys, Rev. 145, 1216 (1959).266 Chap. 7 The Variational Principle function (see Problem 7.16) it can be shown that F, < —13.6 eV, and hence tha: a bound state does exist. It’s only barely bound, however, and there are no excited bound states,’ so H~ has no discrete spectrum (all transitions are to and from the continuum). As a result, it is difficult to study in the laboratory, although it exists ir great abundance on the surface of the sun.* 7.3. THE HYDROGEN MOLECULE ION Another classic application of the variational principle is to the hydrogen molecule ion, H},, consisting of a single electron in the Coulomb field of two protons (Figure 7.5). We shail assume for the moment that the protons are fixed in position, a specified distance R apart, although one of the most interesting byproducts of the calculation is going to be the actual value of R. The Hamiltonian is Wy 2711 a 4ne (4 +2). (7.35) where 7; and r2 are the distances to the electron from the respective protons. As always, the strategy will be to guess a reasonable trial wave function, and invoke the variational principle to get a bound on the ground-state energy. (Actually, our main interest is in finding out whether this system bonds at all—that is, whether its energy is less than that of a neutral hydrogen atom plus a free proton. If our trial wave function indicates that there is a bound state, a better trial function can only make the bonding even stronger.) To construct the trial wave function, imagine that the ion is formed by taking a hydrogen atom in its ground state (Equation 4.80), 1 Welt) = fa, 17.36] 9 and then bringing in a proton from far away and nailing it down a distance R away. If B is substantially greater than the Bohr radius, a, the electron’s wave function -e Figure 7.5: The hydrogen molecule a A te ion, Hy. Robert N. Hill, J. Math. Phys. 18, 2316 (1977), For further discussion, see Hans A. Bethe and Edwin E. Salpeter, Quantum Mechanics of One- and Two-Electron Atoms (New York: Plenum 1977), Section 34Sec. 7.3: The Hydrogen Molecule Ion 267 probably isn’t changed very much. But we would like to treat the two protons on an equal footing, so that the electron has the same probability of being associated with cither one. This suggests that we consider a trial function of the form ¥ =A[Welr) + velr)]- 17.37] (Quantum chemists call this the LCAO technique, because we are expressing the molecular wave function as a linear combination of atomic orbitals.) Our first task is to normalize the trial function: 1= fivear=ar[f iverorare + fvetrv? de +2 f vetrrvecraae], {7.38} The first two integrals are 1 (since Wy itself is normalized); the third is more tricky. Let I 15 (Welle) = = | ene By, (739) Picking coordinates so that proton 1 is at the origin and proton 2 is on the z-axis at the point R (Figure 7.6), we have ry = rand rz = vr? + R? — 2r Ros, [7.40] and therefore eG NE FR-2Re0SPay? sing drdodd (741) tg= Vr? + R®-2r Roos 8 x Figure 7.6: Coordinates for the calculation of / (Equation 7.39).268 © Chap.7 The Variational Principle The @ integral is trivial (277). To do the 0 integral, let y= vr2+ R? —2rRcos8, so that d(y*) = 2ydy = 2rR sind do. _ | 1 opree [RP sna de = I [eteydy 0 TR Syn Slee Rte “Rar — RI +a)]. The r integral is now straightforward: 2 -rja [* -2rja nya [* =aRl-¢ (+ Rt aje"rdr te (R—=r-+a)rdr @R 0 0 sete f (Rt ae wy ar}. R Evaluating the integrals, we find (after some algebraic simplification), 2 R\ 1/8 erie : + (2) +5 (2) | [7.42] a) 3\a 1 is called an overlap integral; it measures the amount by which y,(71) overlaps Yelr2) (notice that it goes to 1 as R + 0, and to 0 as R — oo). In terms of J, the normalization factor (Equation 7.38) is 1 2 lAj Pa 7.43] Next we must calculate the expectation value of H in the trial state y. Noting that 5 2 Wy = 11) = Evel ( mY EEK +) Hird = Welr) (where E; = —13.6 eV is the ground-state energy of atomic hydrogen)—and the same with 72 in place of r;—we have Ua zit ay =a|-Ev (1+ A) freer + vet 4€ 1 1 =EW-A G&S -) [fv + twtr} It follows that e (H) = E, — 2142 (se sa) [Wetrni svete + erat ivatr}. [7.44]Sec. 7.3: The Hydrogen Molecule Ion 269 Tl let you calculate the two remaining quantities, the so-called direct integral, 1 D= a(vetrnizivacro), [7.45] and the exchange integral, 1 X=sa (vetroidivtr). {7.46} The results (see Problem 7.8) are aL 2) o-2Rie 4 - (1+ Se (7.47 and X= (1+ 2)¢ Ria 17.48] Putting all this together, and recalling (Equations 4,70 and 4.72) that E, = —(e?/4z€0)(1/2a), we conclude that _ (D+X) (A) = [: +20 ] E, 17.49) According to the variational principle, the ground-state energy is Jess than (H). Of course, this is only the e/ectron’s energy—there is also potential energy associated with the proton-proton repulsion: 2 el 2% p= =-— 5, 7.50) mane R RR (7.50) ‘Thus the fotal energy of the system, in units of —£) and expressed as a function of x = R/a, is less than POST Ty dax PUA ae 751) This function is plotted in Figure 7.7. Evidently bonding does occur, for there exists a region in which the graph goes below —I, indicating that the energy is less than that of a neutral atom plus a free proton (to wit, —13.6 eV). The equilibrium separation of the protons is about 2.4 Bohr radii, or 1.27 A. +Problem 7.8 Evaluate D and X (Equations 7.45 and 7.46). Check your answers against Equations 7.47 and 7.48. xxProblem 7.9 Suppose we used a minus sign in our trial wave function (Equa- tion 7.37): W = Alert) — Ye(ra)1. 17.52]270 Chap. 7 The Variational Principle F(x) 0 Equilibrium A Figure 7.7: Plot of the function F(x), Equation 7.51, showing existence of a bound state. Without doing any new integrals, find F(x) (the analog to Equation 7.51) for this case, and construct the graph. Show that there is no evidence of bonding. (Since the variational principle only gives an upper bound, this doesn’t prove that bonding cannot occur for such a state, but it certainly doesn’t look promising). Note: Actually. any function of the form Ve = Alert) + cP org(r2)] (7.53] has the desired property that the electron is equally likely to be associated with ei- ther proton. However, since the Hamiltonian (Equation 7.35) is invariant under the interchange P:r V. the wave function is of the form W(x) = de, with k = /2m(E-V)/h. The plus sign indicates that the particle is traveling to the right, and the minus sign means it is going to the left (the general solution, of course, is a linear combination of the two). The wave function is oscillatory, with constant wavelength A = 27/k and constant amplitude A. Now suppose that V (x) is not constant, but varies rather slowly in comparison to A, so that over a region containing many full wavelengths the potential is essentially constant. Then it is reasonable to suppose that yv remains practically sinusoidal, except that the wavelength and the amplitude change slowly with x. This is the inspiration behind the WKB approximation. In effect, it identi- fies two different levels of x-dependence: rapid oscillations, modulated by gradual variation in amplitude and wavelength. Mn Holland it’s KWB, in France it’s BWK, and in England it’s JWKB (for Jeffreys)Sec. 8.1: The “Classical” Region 275 By the same token, if £ < V (and V is constant), then W is exponential: W(x) = Ae, with « = /2m(V — B)/h. And if V(x) is nor constant, but varies slowly in comparison with I/«, the solution temains practically exponential, except that A and « are now slowly varying functions of x. Now, there is one place where this whole program is bound to fail, and that is in the immediate vicinity of a classical turning point, where E * V. For here 4 (or 1/k) goes to infinity, and V(x) can hardly be said to vary “slowly” in comparison. As we shall see, a proper handling of the turning points is the most difficult aspect of the WKB approximation, though the final results are simple to state and easy to implement. 8.1 THE “CLASSICAL” REGION. The Schrédinger equation, Pa ES pyre = ey. m dx can be rewritten in the following way: BY [8.1] where P(x) = ¥2m[E — V(x)] [8.2] is the classical formula for the momentum of a particle with total energy E and potential energy V(x). For the moment, I’ll assume that E > V(x), so that p(x) is reall, we call this the “classical” region, for obvious reasons—classically the particle is confined to this range of x (see Figure 8.1). In general, y is some complex function; we can express it in terms of its amplitude, A(x), and its phase, @ (x)—both of which are real: W(x) = A(xel@™, [8.3] Using a prime to denote the derivative with respect to x, we find dy ay _ oy ole oo and276 Chap. 8 The WKB Approximation Vix) Classical region Figure 8.1: Classically, the particle is confined to the region where £ > V(x). ey _ [A" + 214'9' +140" - AY" [8.4] dx? Putting this into Equation 8.1, 2 A" + A'G' + 1Ag" ~ AY =—T 5A. [8.5] This is equivalent to two real equations, one for the real part and one for the imaginary part: 2 r 2 a n P " r AAG =-FA, or A =4[@r-8], [8.6] and 24'g' + Ab" =0, or (479) =0. [8.7] Equations 8.6 and 8.7 are entirely equivalent to the original Schrédinger equa- tion. The second one is easily solved: Lo = 18.8] where C isa (real) constant. The first one (Equation 8.6) cannot be solved in general— so here comes the approximation: We assume that the amplitude A varies slowly, so that the 4” term is negligible. (More precisely, we assume that 4”/A is much less than both (@')? and p?/h?.) In that case we can drop the left side of Equation 8.6, and we are left withSec. 8.1: The “Classical” Region 277 and therefore i om = sz f pods. (8.9) (Ul write this as an indefinite integral, for now—any constant of integration can be absorbed into C, which thereby becomes complex.) It follows, then, that w(x) = [8.10] and the general (approximate) solution will be a linear combination of two such terms, one with each sign. Notice that IcP P(x)” which says that the probability of finding the particle at point x is inversely pro- portional to its (classical) momentum (and hence its velocity) at that point. This is exactly what you would expect—the particle doesn’t spend long in the places where it is moving rapidly, so the probability of getting caught there is small. In fact, the WKB approximation is sometimes derived by starting with this “semiclassical” ob- servation, instead of by dropping the A” term in the differential equation. The latter approach is cleaner mathematically, but the former offers a more plausible physical rationale. IvGyP = (8.11) Example: Potential well with two vertical walls. Suppose we have an infinite square well with a bumpy bottom (Figure 8.2): some specified function, if V (x) throughout] we have 1 Woe C. elo + Ce ib) : Tom ] or, more conveniently, 1 W(x) = == ICising(x) + C2 cos $(x)], [8.13 ¢ Troy ¢ 2 where (exploiting the freedom noted earlier to impose a convenient lower limit on the integral) po g(x) = if p(x’) dx’. [8.14 ho Now W(x) must go to zero at x = 0, so, since $(0) = 0, Cz = 0. Also, W(x) goes to zero atx = a,so gaj=nx (n=1,2,3.. | p(x) dx = nah. (8.16) Conclusion: This quantization condition is our main result; it determines the (approximate) allowed energies. For instance, if the well has a flat bottom [V (x) = 0], then p(x) = V2mE (a constant), and Equation 8.16 says pa = nah, or nah? 2ma? ” which are precisely the energy levels of the original infinite square well (Equa- tion 2.23). In this case the WKB approximation yields the exact answer (the am- plitude of the true wave function is constant, so dropping A” cost us nothing). «Problem 8.1 Use the WKB approximation to find the allowed energies (Z,) of an infinite square well with a “shelf”, of height Vo, extending half-way across (see Figure 6.3):Sec. 8.1: The “Classical” Region 279 Vo. if0 Vo, but do not assume that E,, >> Vo. Compare your result with what we got in Section 6.1.2, using first-order perturbation theory. Note that they are in agreement if either Vo is very small (the perturbation theory regime) or 1 is very large (the semiclassical WKB regime). «Problem 8.2 An illuminating alternative derivation of the WKB formula (Equation 8.10) is based on an expansion in powers of h. Motivated by the free particle wave function, y = A exp(tipx/h), we write Uo) =e", where f (x) is some complex function. (Note that there is no loss of generality here— any nonzero function can be written in this way.) (a) Put this into Schrédinger’s equation (in the form of Equation 8.1), and show that inf’ —(f'Y +p? =0. (b) Write f(x) as a power series in h: FO) = fo) FHA) FW A) to, and, collecting like powers of f, show that (hy = iff = fof. if! = fof + FI, ete (©) Solve for fo(x) and f,(x), and show that—to first order in h—you recover Equation 8.10. Note: The logarithm of a negative number is defined by In(—z) = In(z) +inz, where nis an odd integer. If this formula is new to you, try exponentiating both sides, and you'll see where it comes from.280 © Chap. 8 The WKB Approximation 8.2 TUNNELING So far, Ihave assumed that F > V, so that p(x) is real. But we can easily write down the corresponding result in the nonclassical region (E < )—it’s the same as before (Equation 8.10), only now p(x) is imaginary*: Cc we) Le thf lpooiax [8.17] Consider, for example, the problem of scattering from a rectangular barrier with a bumpy top (Figure 8.3). To the left of the barrier (x < 0), W(x) = Ae™ + Be, [8.18] VimE /h where A is the incident amplitude, B is the reflected amplitude, and k (see Section 2.7). To the right of the barrier (x > a), wx) = Fel; [8.19] F is the transmitted amplitude, and the tunneling probability is LEP T=—,;- \AP [8.20] In the tunneling region (0 < x < a), the WKB approximation gives V(x): Figure 8,3: Scattering from a rectangular barrier with a bumpy top. 2In this case the wave function is real, and the analogs to Equations 8.6 and 8.7 do not follow necessarily from Equation 8.5, although they are still sufficient. If this bothers you, study the alternative derivation in Problem 8.2.Sec. 8.2: Tunneling 281 Figure 8.4: Qualitative structure of the wave function, for scattering from a high, broad barrier. Cpe, D4 fipeniae [; + : 8.21 VO ® Toe Vipool saul But if the barrier is very high and/or very wide (which is to say, if the probability of tunneling is small), then the coefficient of the exponentially increasing term (C) must be small (in fact, it would be zero if the barrier were infinitely broad), and the wave function looks something like? Figure 8.4. The relative amplitudes of the incident and transmitted waves are determined essentially by the total decrease of the exponential over the nonclassical region: IPL 8 ff neniae 14] so that T : [8.22] Y, with y L if Ipoidx. | Example: Gamow’s theory of alpha decay. In 1928, George Gamow (and, independently, Condon and Gurney) used this result to provide the first theoretical account of alpha decay (the spontaneous emission of an alpha particle—two protons and two neutrons—by certain radioactive nuclei). Since the alpha particle carries a positive charge (2e), it will be electrically repelled by the leftover nucleus (charge Ze) as soon asit gets far enough away to escape the nuclear binding force. Gamow pictured the potential energy curve for the alpha particle as a finite square well (representing the attractive nuclear force), extending out to r; (the radius of the nucleus), joined to a repulsive Coulombic tail (Figure 8.5). If E is the energy of the emitted alpha particle, the outer turning point (r)) is determined by 3This heuristic argument can be made more rigorous—see Problem 8.10.282 Chap. 8 The WKB Approximation Mr), Coulomb repulsion Nuclear binding Figure 8.5: Gamow's model for the potential energy of an alpha particle in a radioactive nucleus. 1 2Ze? ame) 72 =E£. [8.23] The exponent y (Equation 8.22) is evidently* if am (hE — 8) ar = SE (Eater J2mE = = [rn cos Jin - Vita — ro] . [8.24] Typically, r1 < ra, and we can simplify this result, The argument of the inverse cosine is close to zero, so the angle itself is close to /2. Call it @ = (#/2)—e: then sin € cos @ = cos(sr/2) cose + sin(zr/2) sine a ftz™_ fr m2 rn ‘In this case the potential does not drop to zero on both sides of the barrier (moreover, this is really 1a three-dimensional problem), but the essential inspiration, contained in Equation 8,22, is all we really need, and henceSec. 8.2: Tunneling 283 Thus ¥ [8.25] where K= (&) 2 980Mev"/?, [8.26] and 2 (6)" in = 1.485 fmc (8.271 (One fermi, fm, is 10-!> m, which is about the size of a typical nucleus.) If we imagine the alpha particle rattling around inside the nucleus, with an average velocity v, the average time between “collisions” with the “wall” is about 2r,/v, and hence the frequency of collisions is v/2r. The probability of escape at each collision is e~*”, so the probability of emission, per unit time, is (v/2r;)e~”, and hence the lifetime of the parent nucleus is about t= 2n 2y, [8.28] v Unfortunately, we don’t know v—but it hardly matters, for the exponential factor varies over a fantastic range (25 orders of magnitude) as we go from one radioac- tive nucleus to another; relative to this the variation in v is pretty insignificant. In particular, if you plot the logarithm of the experimentally measured lifetime against 1/VE (related to y by Equation 8.25), the result is a beautiful straight line (Figure 8.6), confirming that the lifetime of an alpha emitter is governed by the difficulty of penetrating the Coulomb barrier. Problem 8.3. Use Equation 8.22 to calculate the approximate transmission proba- bility for a particle of energy E that encounters a finite square barrier of height Vo > E and width 2a. Compare the exact result (Prob. 2.32) in the WKB regime T < 1. +«Problem 8.4 Calculate the lifetimes of U8 and Po”, using Equation 8.28, with Equation 8.25 for y. Hint: The density of nuclear matter is relatively constant (i.e., the same for all nuclei), so (r;)* is proportional to A (the number of neutrons plus protons). Empirically, r, = (1.07 fmA'?. [8.29] The energy of the emitted alpha particle is determined by Einstein's formula (E = me’):284 Chap. 8 The WKB Approximation 12 238 ‘os1ot ye [years] Pipe tiperiissitipiiiiiiitiris 5 6 7 E [Mev] Figure 8.6: Graph of the logarithm of the lifetime versus 1/¥E, for several alpha emitters. From David Park, Introduction to the Quantum Theory, 3rd ed. (New York: McGraw-Hill, 1992), (See acknowledgment in Preface.) E = m,c? — mac? — mac, [8.30) where m, is the mass of the parent nucleus, m, is the mass of the daughter nucleus and mg is the mass of the alpha particle (which is to say, the He* nucleus). To figure out what the daughter nucleus is, note that the alpha particle carries off two protons and two neutrons, so Z decreases by 2 and A by 4. Look up the relevant nuclear masses. To estimate v, use E = (1/2)mav*; this ignores the (negative) potential energy inside the nucleus, and surely underestimates v, but it’s about the best we can do at this stage. Incidentally, the experimental lifetimes are 6 x 10° years and 0.5 1» respectively. 8.3 THE CONNECTION FORMULAS In the discussion so far I have assumed that the “walls” of the potential well (or the barrier) are vertical, so that the “exterior” solution is simple and the boundary conditions trivial. As it turns out, our main results (Equations 8.16 and 8.22) are reasonably accurate even when the edges are not so abrupt (indeed, in Gamow’s theor they were applied to just such a case). Nevertheless, it is of some interest to study more closely what happens to the wave function at a turning point (E = V), where the “classical” region joins the “nonclassical” region and the WKB approximationSec. 8.3: The Connection Formulas 285 itself breaks down, In this section I'll treat the bound-state problem (Figure 8.1); you get to do the scattering problem for yourself (Problem 8.10).° For simplicity, let's shift the axes over so that the right-hand turning point occurs at x = 0 (Figure 8.7). In the WKB approximation, we have [pit meds Cg hf poe ] ifx <0, Vor all [8.31] gta Le D Denk So leeds’ ifc-0. (pc) [Assuming that V(x) remains greater than E for all x > 0, we can exclude the positive exponent in this region, because it blows up as x — oo.] Our task is to join the two solutions at the boundary. But there is a serious difficulty here: In the WEB approximation, ¥ goes to infinity at the turning point, where p(x) -> 0. The true wave function, of course, has no such wild behavior—as anticipated, the WKB method simply fails in the vicinity of a turning point. And yet, it is precisely the boundary conditions at the turning points that determine the allowed energies. What we need to do, then, is splice the two WKB solutions together, using a “patchin; wave function that straddles the turning point. Since we only need the patching wave function (y/p) in the neighborhood of the origin, we'll approximate the potential by a straight line: V(x) = E+V'O)x, [8.32] and solve the Schrédinger for this linearized V: Linearized potential Turing Classical «0 ~—_-Nonclassical region region Figure 8.7: Enlarged view of the right-hand turning point. 5 Warning: The following argumentis quite technical, and you may wish to skip iton a first reading.286 Chap. 8 The WKB Approximation wa am ea + (E+V'OX)Wp = Ebr or @ a =a°xyy, [8.33] where am, 78 a [Fr o| : [8.34] The a’s can be absorbed into the independent variable by defining [8.35] so that = 2p. 18.36] This is Airy’s equation, and the solutions are called Airy functions.° Since the Airy equation is a second-order differential equation, there are two linearly independent Table 8.1: Some properties of the Airy functions, Differential Equation: 7 Solutions Linear combinations of Airy Functions, Ai(e) and Bi(2) Integral Representation: Ai(2 Asymptotic Forms: Ai(z) ~ BI) ~ =a Classically, a linear potential means a constant force, and hence a constant acceleration—the simplest nontrivial motion possible, and the starting point for elementary mechanics. It is ironic that the same potential in guanrwm mechanics gives rise to unfamiliar transcendental functions, and plays only a peripheral role in the theoryAi(z) os 04 03 02 O41 0.1 -0.2 -03 -0.4 -05 Sec. 8.3: The Connection Formulas 287 Bi(z) 20 18 1.6 14 12 1.0 8 8 -6 -4 Figure 8.8: Graph of the Airy functions. Airy functions, Ai (z) and Bi(z); the general solution is a linear combination of these. ‘Ai and Bi are related to Bessel functions of order 1/3; some of their properties are listed in Table 8.1 and they are plotted in Figure 8.8. Evidently the patching wave function is VUp(x) = adi(ax) + bBi(ax), [8.37] for appropriate constants @ and b. Now Wp is the (approximate) wave function in the neighborhood of the origin; our job is to match it to the WKB solutions in the overlap regions on either side (see Figure 8.9). These overlap zones are close enough to the turning point that the linearized potential is reasonably accurate (so that vp is a good approximation to the true wave function), and yet far enough away from the turning point that the WKB approximation is reliable.” In the overlap regions Equation 8.32 holds, and therefore (in the notation of Equation 8.34) p(x) = 2m(E — E— V'()x) = ha In particular, in overlap region 2, {8.38} 7 This is a delicate double constraint, and itis possible to concoct potentials so pathological that no such overlap region exists. However, in practical applications this seldom occurs. See Problem 8.8,288 Chap. 8 The WKB Approximation ~ x Patching region Figure 8.9: Patching region and the two overlap zones. * n tm 3/2 * 7 dy! 2 3/2, Ipx’ldx' = hed? J Jx' dx! = gheaxy?, 0 0 and therefore the WKB wave function (Equation 8.31) can be written as e D —Haxy? W(x) = Jian’ . [8.39] Meanwhile, using the large-z asymptotic forms? of the Airy functions (from Table 8.1), the patching wave function (Equation 8.37) in overlap region 2 becomes a oe ben ye ex ete”, A YH) * Fane Fran Bol ‘Comparing the two solutions, we see that ‘43 =p, and b=0. [8.41] ah Now we go back and repeat the procedure for overlap region 1, Once again. P(x) is given by Equation 8.38, but this time x is negative, so 0 2 : / pix’) dx' = grax)? [8.42] and the WKB wave function (Equation 8.31) is AL first glance it seems absurd to use a large-z approximation in this region, which after all is supposed to be reasonably close to the tuming point at z = 0 (so that the linear approximation to the potential js valid). But notice that the argument here is a-x, and if you study the matter carefully (see Problem 8.8) you will find that there és (typically) a region in which oxx is large, but at the same time it is reasonable to approximate V (x) by a straight line.