MOLECULAR SPECTROSCOPY IRA N. LEVINE Brooklyn College City University of New York A WILEY-INTERSCIENCE PUBLICATION John Wiley & Sons New York - London - Sydney - TorontoCopyright © 1975, by John Wiley & Sons, Inc. All rights reserved. Published simultaneously in Canada. No part of this book may be reproduced by any means, nor transmitted, nor translated into a machine language with- cout the written permission of the publisher. Library of Congress Cataloging in Publication Data Levine, Ira N date Molecular spectroscopy. “A Wiley-Interscience publication.” Rev. version of the book previously published as v. 2 of the author’s Quantum chemistry. Bibliography: Includes index. 1. Molecular spectra. I. Title. QC454.M6L48 541'.28 74-30477 ISBN 0-471-53128-6 Printed in the United States of America 10987654321PREFACE This book is intended for graduate and advanced undergraduate courses in molecular spectroscopy. It presupposes some knowledge of quantum mechanics, such as is gained in a one-semester introductory quantum- chemistry course. This text is a much-revised version of the book previously published as Volume II of my book Quantum Chemistry. The first chapter of the present book reviews quantum mechanics and atomic and molecular electronic structure, so that this book can be used completely independently of any other book. My aim in writing this book has been to give students a solid under- standing of the theoretical basis of spectroscopic methods and an apprecia- tion of how these methods provide information on molecular structure. Brief descriptions of experimental spectroscopic procedures are also in- cluded. I have tried to present material as clearly and logically as possible, giving sufficient detail in the derivations to make them easy to follow. The text takes into account the limited mathematics and physics background of the average chemistry student. However, in no sense is the treatment superficial or “watered down.” The major changes in the new edition are as follows: There are three new chapters. Chapter | is a review and summary of aspects of quantum mechanics and electronic structure relevant to molecular spectroscopy. This chapter replaces the chapter on electronic structure of polyatomic molecules that was repeated from Volume I of Quantum Chemistry. Chapter 2 is a substantially expanded presentation of matrices. Previously, matrices were covered in the last chapter. The placement of matrices early in the book allows their use throughout the book; in particular, the very tedious and involved treatment of normal vibrations has been replaced by a simpler and clearer treatment using matrices. Chapter 7 covers molecular electronic spectroscopy, and contains two new sections, one on electronic spectra of polyatomic molecules, and one on photoelectron spectroscopy, together with the section on electronic spectra of diatomic molecules from the previous edition. In addition to the new material on matrices, electronic spectra of polyatomic molecules, and photoelectron vvi Preface spectroscopy, the following topics have been added: line widths and shapes (Section 3.5), lasers (Section 3.6), spin relaxation (Section 8.8), Einstein coefficients, oscillator strengths, Fourier-transform NMR, calculation of excited-state wave functions, Koopmans’ theorem, ab initio and semiempirical calculations of ionization potentials, force constants, and magnetic-resonance parameters. I have left group theory as the last chapter. If it is desired to cover group theory earlier in the course, Sections 9.1-9.8 of the group-theory chapter can be studied prior to Chapter 6 on molecular vibrations, and Sections 9,9-9.12 read in conjunction with Chapter 6. Professors James Bolton and Harry King reviewed the manuscript of the previous edition of this book; I have belatedly incorporated many of their excellent suggestions in this edition. I am also grateful to Professors Frank Meeks and Kenneth Sando who reviewed an earlier version of the present edition. Professor Vicki Chuckrow provided helpful advice on some mathematical points. The help of Dr. Theodore Hoffman of Wiley- Interscience is gratefully acknowledged. Any suggestions that readers may have for improvement of the book will be welcomed. TRAN. LevINE Brooklyn, New York September 1974CONTENTS CHAPTER 1 QUANTUM MECHANICS AND iI 1.2 13 1.4 1S 1.6 17 1.8 19 1.10 Ll 1.12 1.13 1.14 1.15 1.16 117 1.18 1.19 1.20 1.21 ELECTRONIC STRUCTURE Quantum Mechanics Mathematics Review Units The Particle in a Box The Harmonic Oscillator Orbital Angular Momentum Angular-Momentum Ladder Operators Parity The Variation Method Perturbation Theory The Central-Force Problem The Two-Particle Problem The Two-Particle Rigid Rotor The Hydrogen Atom Spin The Pauli Principle Many-Electron Atoms Symmetry Point Groups Molecular Electronic Structure The Hartree-Fock Method Semiempirical Methods Roe 27 30 32 35 38 39 41 42 45 47 Saasvill CHAPTER 2 MATRICES 2.1 Matrices 2.2 Eigenvalues and Eigenvectors 2.3. Matrices and Quantum Mechanics CHAPTER 3 TIME-DEPENDENT STATES AND SPECTROSCOPY 3.1 Time-Dependent Perturbation Theory 3.2 Absorption and Emission of Radiation 3.3. Selection Rules 3.4 Spectroscopy 3.5 Line Shapes and Widths 3.6 Lasers CHAPTER 4 VIBRATION AND ROTATION OF DIATOMIC MOLECULES 4.1 | Nuclear Motion in Diatomic Molecules 4.2 Anharmonicity, Vibration—Rotation Interaction, and Centrifugal Distortion 4.3 Potential-Energy Functions for Diatomic Molecules 4.4 Selection Rules for Rotational and Vibrational Transitions 4.5 Rotational Spectra of Diatomic Molecules 4.6 Vibration-Rotation Spectra of Diatomic Molecules 4.7 Parity of Diatomic-Molecule Wave Functions 4.8 Nuclear Spin and the Pauli Principle 4.9 Ortho and Para Modifications of Homonuclear Diatomic Molecules 4.10 The Raman Effect 4.11 Rotational Energies of Non-'= Diatomic Molecules Contents. 78 78 93 110 110 114 123 129 133 135 142 142 149 160 162 165 168 175 178 183 186 188Contents CHAPTER 5 ROTATION OF POLYATOMIC MOLECULES 5.1. Nuclear Motion of Polyatomic Molecules 5.2 Classical Mechanics of a Rigid Rotor 5.3. The Rotational Hamiltonian Operator 5.4 The Spherical Top 5.5 The Symmetric Top 5.6 The Asymmetric Top 5.7 Pure-Rotation Spectra of Polyatomic Molecules 5.8 Microwave Spectroscopy CHAPTER 6 VIBRATION OF POLYATOMIC MOLECULES 6.1 Introduction 6.2 Classical Mechanics of Vibration 6.3. Symmetry and Normal Vibrations 6.4 Quantum Mechanics of Vibration 6.5 Selection Rules for Vibration-Rotation and Pure-Rotation Transitions 6.6 Infrared Spectroscopy 6.7. Raman Spectroscopy 6.8 Degenerate and Near-Degenerate Vibrational Levels 6.9 Parity of Polyatomic-Molecule Wave Functions 6.10 Nuclear Spin and the Pauli Principle CHAPTER 7 ELECTRONIC SPECTROSCOPY 7.1 Electronic Spectroscopy 7.2 Electronic Spectra of Diatomic Molecules 7.3 Electronic Spectra of Polyatomic Molecules 7.4 Photoelectron Spectroscopy 195 195 197 203 210 213 216 219 235 235 235 246 249 255 260 269 271 279 283 296 296 298 306 314x CHAPTER 8 MAGNETIC RESONANCE SPECTROSCOPY 8.1 8.2 8.3 8.4 8.5 8.6 8.7 8.8 8.9 8.10 8.11 Magnetic Resonance Nuclear Magnetic Moments Nuclear Magnetic Resonance NMR Chemical Shifts NMR Spin-Spin Splittings NMR Spectra of Nuclei Other Than Protons Origin of NMR Spin-Spin Coupling Spin Relaxation NMR and Rate Processes Nuclear Quadrupole Moments Electron Spin Resonance CHAPTER 9 GROUP THEORY 9.1 Groups 9.2 Multiplication Tables 9.3 Coordinate Transformations and Matrices 9.4 Representations 9.5 Character Tables 9.6 Basis Functions 9.7 Group Theory and Quantum Mechanics 9.8 Group Theory and Electronic Structure 9.9 Group Theory and Molecular Vibrations 9.10 Group Theory and Molecular Rotation 9.11 The Direct-Product Representation 9.12 Character Tables of Some Point Groups APPENDIX BIBLIOGRAPHY ANSWERS TO SELECTED PROBLEMS INDEX Contents 323 323 323 325 331 337 356 357 361 365 387 389 394 399 402 409 412 417 427 439 458 467 471 475 477MOLECULAR SPECTROSCOPYCHAPTER I QUANTUM MECHANICS AND ELECTRONIC STRUCTURE 1.1 QUANTUM MECHANICS Before beginning our study of molecular spectroscopy, we shall in this chapter review and summarize those aspects of quantum mechanics and atomic and molgcylar electronic structure that are relevant to molecular spectroscopy. sinte it is assumed that the re reader has had a previous course in quantum chemistry, we sith i not zive detailed derivations in this chapter. If the reader is puzzled by any of the discussions in Chapter 1, he can consult Leviné, Gatz, Pilar, or Pauling and Wilson for fuller treatments. (References with the author’s name italicized are to books listed in the Bibliography.) Rather than going through Chapter | in detail, it might be a ood idea to simply skim it, and refer back to it as needed when reading Sieh The state of an n-particle quantum-mechanical system is described by its state function (q,(), where q symbolizes the 3” spatial and n spin coordinates of the particles, and ¢ is the time. The function ¥. must be single- valued and continuous; in_addition, for bound states.’ must be r egrable—by this we mean that [VVdr is finite, where {de denotes integration’ over all space, including a summation over spin variables. (Although {¥*¥dr looks like an indefinite integral, it is under- stood to be a definite integral.) A function that satisfies the preceding conditions is said to be well behaved. To each physical property of a quantum-mechanical system, there cor- responds a quantum-mechanical operator. The operator is formed by 12 Quantum Mechanics and Electronic Structure taking the classical-mechanical expression for the property and replacing each Cartesian coordinate x; by the operator multiplication by x, and each linear momentum component p, by the operator (h/i)(0/0x,). Using a circumflex to denote an operator, we have (1.1) Bs= (1.2) fA i h=h/2x, i=V-1 a Ox, where h being Planck’s constant. If A is the operator corresponding to the physical property A, then any measurement of the property A in any quantum-mechanical system must give a result that is one of the eigenvalues of the operator A. The eigenvalues a, and the eigenfunctions y, of A are defined as those numbers and functions that satisfy Ag,=4,9, (1.3) If we take a large number of identical systems, each with the same state function ¥, and measure A in each system, we shall in general get different results in different systems. (This is one of the striking differences between classical and quantum mechanics.) The average value of A, denoted by experimentally, we average the results of separate observations of A in a very large number N of identical systems, each in the same state ¥: =%,a;/N, where the sum is over the observed values @,,4,...,4y. This last equation is equivalent to (Ad= Z P(a)a @ where the sum is over all possible values of A, and where P(a) is the probability of observing the value a for the property A.] The way in which the state function ¥ changes with time is postulated to1.1 Quantum Mechanics 3 be given by the time-dependent Schrodinger equation: ~ 7 7 AY (1.5) where 7 is the Hamiltonian (or energy) operator for the system; A is obtained by converting the classical-mechanical expression for the system’s energy in terms of Cartesian coordinates and linear momenta to an operator by the procedure described above. Thus for a one-particle, three-dimensional system, the classical-mechanical Hamiltonian function is the sum of the kinetic energy T and the potential energy V: H=T+V=(p2+p}+p?)/2m+ V(x,y,z,1) (1.6) where p,,p,,p, are the linear-momentum components [Equation (1.144)] and the potential energy is defined by -aV/ax=F, OV /ay=E, —V/dz=F, (1.7) where F,,£,,F, are the components of the force on the particle. Replace- ment of classical quantities by operators gives as the Hamiltonian operator a? , a? | a? ste tt dy, 2,0 and aye * ga] tn) yo Fp =~ FE vt4 v(x9,2,0) (1.8) where V? = 07/dx?+0?/dy?+0?/dz? is the Laplacian operator. For an n-particle system, the Hamiltonian is A Sy Re BaD PV erty semen!) (19) i=l t V? = 0?/dx? + 97/ ay? + 97/ dz? (1.10) Integration of (1.5) with respect to time introduces one arbitrary constant. If we know ¥ at the present time fo, we can evaluate this constant and thereby determine ¥ at any future time. Quantum-mechanical operators that correspond to physical properties must be linear and Hermitian. A /inear operator A satisfies A (f+g)= Aft Ag (1.11) A (cf) =cAf (1.12)4 Quantum Mechanics and Electronic Structure where f and g are any functions and c is any constant. A Hermitian operator is a linear operator that satisfies frragdr= [e(Apyt ar (1.13) for all well-behaved functions f and g, where the integration is over all space. The eigenfunctions g, of a Hermitian operator form a complete set. By this we mean that any well-behaved function f that satisfies the same boundary conditions as the g;’s can be expanded as = Deg, (1.14) where the expansion coefficients c, are constants. The preceding statements together with the postulates of spin and the Pauli principle (Sections 1.15 and 1.16) can be taken as the postulates of quantum mechanics. If the potential energy V in the Hamiltonian is independent of time, then there exist solutions of (1.5) that have the form ¥(4q.0) =f(O(4) (1.15) where f is a function of time only and y is a function of the coordinates only. To verify (1.15), we substitute it into (1.5) to get —(h/iS'()Y(q) =f()AY(q), since A does not affect f(t). Division by fy gives i) _ HV) 7 ¥@) Since #1 is independent of t, the right side of (1.16) is a function of the coordinates only, while the left side is a function of ¢ only. By the usual separation-of-variables argument (Section 1.2), each side of (1.16) must equal a constant, which we call E. Therefore, —(h/i)(df/dt)/f= E, which integrates to f(1)= A exp(— iEt/h), where A is an arbitrary constant. Since A can be absorbed into the function ¥(q) in (1.15), we drop it to give foe (1.17) (1.16) Also we have Hy/y=E or Hy = Ey (1.18) which is the time-independent Schrodinger equation or, simply, the Schro- dinger equation. Since the Hamiltonian operator H is formed from the1.1 Quantum Mechanics 5 classical-mechanical expression for the energy, we have E=H. The num- bers E in (1.18) are the eigenvalues of the energy operator, and by the postulate preceding (1.3), the numbers E that satisfy (1.18) are thus the possible energies of the system. States of the form (1.15) are called stationary states; for a stationary state ¥(q,1) =e "*/4(q) (1.19) where E is the energy of the state and y is the wave function. Common notations for the integrals that occur in quantum mechanics are fotdg,dr = = =<@l4lo>* (1.23) Ay=(Aj)* (1.24) Ket notation is sometimes used for functions in quantum mechanics. In this notation, the function f is denoted by the symbol |f>: f=|f>. Ket notation is convenient for denoting eigenfunctions by listing their eigenval- ues. Thus |n/m) denotes the hydrogen-atom stationary-state wave function with quantum numbers a, /, and m. The following properties of Hermitian operators follow from the defini- tion (1.23). The eigenvalues of a Hermitian operator are real. Two eigen- functions of a Hermitian operator that correspond to different eigenvalues are orthogonal; that is, if A is Hermitian and Ag, = 4,9), Aqy= a9 with @,#a,, then = foterdr=0 (1.25)8 Quantum Mechanics and Electronic Structure c= @ilf> equals zero unless g; and f have the same eigenvalue (b,=b,). Thus when f and the g's are eigenfunctions of the same operator, we only include functions g, in the expansion (1.30) that have the same B eigen- value as f. There may be several such functions, due to degeneracy. The commutator [A,B] of two operators is defined as ~BA (1.38) & [4,B=4 Mr [4,8 ]=0 (1.39) we say that A and B commute. If two Hermitian operators A and B commute, then it can be proven that there exists a common complete set of eigenfunctions for them; that is, if (1.39) holds, where A and B are Hermitian, then we can find some complete set of functions g, such that Ag;= 4,9; and By,= b,g;. If one or more of the eigenvalues of A are degenerate, we have an element of choice in that we can take many different linear combinations of the eigenfunctions of a degenerate eigen- value. Not all these linear combinations need be eigenfunctions of B; however, the theorem does show that we can choose linear combinations that will be eigenfunctions of B. The converse of the preceding theorem is also true: if A and B are linear operators that possess a common complete set of eigenfunctions, then A and B commute. For three or more operators, all pairs of operators must commute for there to exist a complete set of eigenfunctions common to all the operators. Thus A, nee and é have a common, complete set of eigenfunctions if and only if [A, B= 0, {A, C= 0, and [B,C]=0. Let a system be in the state ¥ at some instant of time. Let g; be the complete set of orthonormal eigenfunctions of the Hermitian operator G that corresponds to the physical property G: Go, = 81%; Expanding ¥, we have from (1.30) and (1.33) ¥= Daw= Tale (1.40) i i Substitution of (1.40) into (G)=C¥|G|¥> and use of orthonormality gives (G>= Ziel: (1.41)1.1 Quantum Mechanics 9 However. as noted earlier, we have = ZT P( 8g, (1.42) where P(g;) is the probability of observing the value g, for G. Hence we conclude that P(g) =laP=lKel¥>? (1.43) Thus the squares of the absolute values of the coefficients in the expansion (1.40) give the probabilities for observing the various possible values g, of the physical property GJFor the special case where the state function is an eigenfunction of G, ¥=,, we have c,=1 in (1.40), and we are certain to observe g, in a measurement of G. Thus, for a stationary state, we have HY =EY, and an energy measurement is certain to give E. For a property with a continuous range of allowed values (e.g., x, P,)s (1.42) becomes = f P(e)sde (1.44) where P(g) is the probability density for property G; that is, P(g)dg is the probability of observing a value of G between g and g+ dg. Consider a one-particle, one-dimensional system. From (1.4) with A =x, we have (os fvitde= f [WPxdx (1.45) Comparison of (1.44) and (1.45) shows that |¥/ is the probability density for the x coordinate; that is, the probability of finding the particle between xagnd x +dx equals |W? dx. . Suppose that state function ¥ at time ¢, is not an eigenfunction of G. Let the eigenfunctions and eigenvalues of G be g, and g,: Go,=g,9, A measurement of G at time ¢, will give one of the eigenvalues of G. Suppose we find the value g,. For a measurement to have any meaning, it ought to be reproducible; that is, if we immediately remeasure G at a time t,+ dt that differs from ¢, only by an infinitesimal time interval, we ought to get 8, again. But if the second measurement is certain to give g,, then the state function at time 1, +d? must be g,, the eigenfunction of G with eigenvalue 8,- We conclude that the measurement of G at time ¢, changed the state function from ¥ (which is not an eigenfunction of G) to » the eigen- function of G that corresponds to the observed value of G. This sudden change in state function caused by a measurement is called the reduction of the state function. Note that reduction is a probabilistic change, since we) fh) (1.46) For any two Hermitian operators, one can prove that AAAB >4\(4|4.8 |¥)| (1.47) which is the general statement of the uncertainty principle. For the case A=x, B=p,, Eqs. (1.1), (1.2), and (1.47) give (8, 8,] = 5p, -P,2=ih (1.48) Axdp, > 4h (1.49) There is also an uncertainty relation involving energy and time: AEr>$h (1.50) In this relation, AZ is the uncertainty in the energy of a state whose lifetime is 7. Equation (1.50) is not a special case of (1.47); rather, (1.50) must be derived by a special treatment. [See Y. Aharonov and D, Bohm, Phys. Rev., 122, 1649 (1961).] Consider the matrix element . where gy, and g are eigenfunc- tions of a Hermitian operator A that commutes with B: AQ, =4,9; Ag, =4Q)> and 1A, B)= 0. Then, provided the eigenvalues a, and @; are not equal, it follows (Problem 1.3) that <@lBlp> vanishes: Ag, Multiplication by R* followed by integration over all space gives the desired sum rule: D= (1.52) 1.2 MATHEMATICS REVIEW In this section we review some mathematics useful for quantum chem- istry and spectroscopy. Vectors. A physical quantity that has both magnitude and direction can be represented by a directed line segment or vector in three-dimensional space. Let A be a vector quantity. (We use boldface type for vectors.) We set up a Cartesian coordinate system xyz and denote vectors of unit length along the x, y, and z axes by i, j, and k, respectively. If A,, A,, and A, are the projections of A on the x, y, and z axes, then A is given by A=A,i+ A,j+ Ak (1.53) The magnitude of A is A=|Al=(424 42442)? (1.54) Since A is completely specified by the three numbers 4,, A,, A,, we can regard a vector in three-dimensional space as an ordered set of three numbers; thus the vector (2,0,3) has an x component of 2, a y component of 0, and a z component of 3. [More generally, we can regard an ordered set of n numbers (A,,A),...,4,) as a vector in an abstract n-dimensional “space”; an n-dimensional vector space is a mathematical concept rather than a physical one (unless n < 3).] Equality of vectors means equality of their corresponding components. If A=B, then A, = B,, A, = B,, A,=B,. The dot product _A + B of two vectors is given by A+B=A,B,+A,B,+A,B,=B°A (1.55) In particular A+ A=A?+A?+ A?=|AP The dot product can be shown to be a scalar equal to A>B=(A||B]cos@ (1.56)12 Quantum Mechanics and Electronic Structure where @ is the angle between A and B. The cross product A x B is given by AX B= (A,B, — A,B, i+ (A,B, A,B,)j+ (A,B, — A,B, )k (1.57) which is most easily remembered in the form oo A A, A 4 AX B=/4, 4, 4, yy Fak , B, Be B, . B, BB, B, (1.58) (Determinants are discussed below.) The cross product can be shown to be a vector of magnitude |A||B|sin@ (where @ is the angle < 180° between A and B) and direction perpendicular to the plane defined by A and B, such that A, B, and A x B form a right-handed system. The vector operator V (de/) is defined as L V =id/ax+ja/dy +ka/az (1.59) Operators. An operator is a rule for transforming any given function to some other function. If the operator A transforms the function f to the function g, we write g= Af (where the circumflex denotes an operator). The sum and the product of any two operators are defined by (A+B)f = Af+ Bf. and (AB)f= A(Bf). Note that Bf i is some function, and apply- ing A to Bf, we get the function defined as (AB). Two operators A and B are equal if Af= Bf for all functions f. The null operator 0 and the unit operator | are defined by Of =0 and 1f=/. [Note that (1.48) and (1.39) are more precisely written as [%, AJ =iAl (1.60) [4,8 1=0 (1.61) However, we shall usually omit circumflexes over operators that are simply multiplication by a constant.] Repeated application of the definition of operator multiplication shows that the associative law holds for all opera- tors: A(BC)=(AB)C (1.62)1.2. Mathematics Review 13 The commutator [A,B] of two operators is defined by [A,B] = AB— BA. The square of A is defined by 4? = AA, with similar definitions for bighes— Powers. clever ‘A linear operator is one that obeys (1.11) and (1.12) for all functions f and g and all constants c. A Hermitian operator is a linear operator that obeys (1.13) for all well-behaved functions f and g. The eigenvalues and eigenfunctions of an operator are defined by (1.3). Some identities valid for linear operators are (1.63) (1.64) (1.65) ({k a constant) (1.66) (1.67) [4, BC ]=[4, BC+ B[A,C] (1.68) [4B,C ]=[4,€ ]8+4 [BE] (169) T complex numbers. A complex number z has the form z= x + iy, where i=V-—I and x and y are real numbers. We can represent z by a point in the complex xy plane by associating the complex number z= x+ iy with the point whose Cartesian coordinates are (x,y). The distance r of the point z from the origin is the absolute value |z| of z; the angle 6 that the radius vector from the origin to z makes with the positive x axis is the phase of z. We have lel=r=(2+y2)'7, tand=y/x (1.70) x=rcos6, y=rsind (1.71) z=x+iy=r(cosé+isin)=re® (1.72) where we used e =cos6+ isin? (1.73) From (1.70) and (1.72), it follows that |z,z)|=|z,||z2]. The complex conjugate z* of z is defined by z2t=(x+yy)* Sx-y (1.74)14 Quantum Mechanics and Electronic Structure We have z=z* if and only if z is a real number (with y=0). Some identities are gta xt+yt= (1.75) (2y22)*=zhz$, (2,422) = 2h +23, (21/2,)* = 2h /2 Let w"=1, where n is a positive integer. The roots of this equation are the nth roots of unity. The number | has absolute value | and its phase can be taken as 0 or 27 or 4m or 6m or .... Hence 1=e"*, where k =0,1,2,.... If we take the number e?7*/" and raise it to the nth power, we get (e/2%*/"y" = ¢2#* =], Hence the nth roots of unity are wae?" k=0,1,2,..n-1, i= V=1 (1.76) (With k=n, n+1,... we get the same roots as with A=0,1,....) + Determinants. A determinant of order n is a square array of n? elements; the determinant has a value calculated by the procedure described below. Let a, be the element in the ith row and jth column of the nth-order determinant A. The minor of element a, is the determinant of order n—1 obtained by striking out the ‘th row and jth column of A. The cofactor C, of a, is its minor multiplied by (— 1)'*7, When »>1, the value of the determinant A is obtained by taking the 7 elements of any one row (or column), multiplying each such element by its cofactor, and adding the n products. Thus, if we use the elements of the kth row, we have A= Oy Cyt 2 pa 0 + Aen Chen = Sinn (1.77) m= where & can be | or 2 or «+ or”. We have thus defined a determinant of order in terms of determinants of order n—1. A determinant of order | has one element, and the determinant’s value is defined as the value of that element; thus |—5|=—5, where the vertical lines denote a determinant (and not absolute value). For a determinant of order 2, use of (1.77) with k=1 gives a a HPT ay ldga|— @ral4ail = A11422— 412401 (1.78) a, ay where the vertical lines indicate determinants. For a determinant of order1.2 Mathematics Review 1s 3, use of (1.77) with k=1 and (1.78) gives tye Ci2e als) G2 3 ay 43 e a =ay — 42 +413 4, 42 3 432 433 a. a. a1 33 Osi) mez cas) = By 1472433 ~ Ay 432473 — 412421433 + 417431423 F 443491432 — 434342 The principal diagonal of a determinant is the diagonal that runs from the upper left to the lower right. A determinant is in diagonal form when its only nonzero elements lie on the principal diagonal. A determinant is in block-diagonal form when its only nonzero elements lie in square blocks centered along the principal diagonal. Some theorems on determinants are the following: (1) If all the elements in any one row (or any one column) are zero, the determinant equals zero. (2) Interchange of any two rows (or any two columns) multiplies a determinant’s value by — 1. (3) If any two rows (or any two columns) are identical, the determinant equals zero. (4) Multiplication of every element of any one row (or any one column) by the constant k multiplies the determinant’s value by k. (5) Addition of the same constant multiple of the elements of one row to the corresponding elements of another row leaves the determinant unchanged in value. (The corresponding statement holds for columns.) (6) A determinant in diagonal form equals the product of the diagonal elements. (7) A determinant in block-diagonal form equals the product of the block determinants. For example, u vo 0 0 0 ir 0 0 0 eos 0 0 pe sl=]" °lln m + (1.79) 0 0 n m r Bread eos or 4) G24 r OBKEB OCH Le” Simultaneous linear equations. A set of n simultaneous linear equations in the 2 unknowns x,,x2,...,x, has the form Ay X~Hayxy,t +++ +4,,x,= 5, Gy X~ + Ay xX. + +++ + 2,X, = by (1.80) GX + AgyXy+ ++ +4,,%,16 Quantum Mechanics and Electronic Structure where the a’s and 6’s are constants. If b;,=5,=--- =b,=0, the set of equations is homogeneous. If at least one of the b’s is nonzero, the set is inhomogeneous. Let det(a,) be the determinant of the coefficients in (1.80): ay ay aa Fin a. a. det(a,)=|“ “2 0 (1.81) Any Fn aa Qnn Let R, be the determinant obtained by replacement of the elements of column k in det(a,) with b,b,,...b,. Cramer’s rule states that for in- homogeneous equations, the unknown x, is given by 1.82 > Fet(a,) (1.82) A set of n linear homogeneous equations in n unknowns always has the solution x,=x,= ++: =x,=0, which is the ‘rivial solution. Suppose the coefficient determinant det(a,) is not equal to zero for a set of linear homogeneous equations; we can then use Cramer’s rule (1.82). Since the equations are homogeneous, the determinant R, will have a column of zeros, and will equal zero; hence x, = x)= =x,=0, and we have only the trivial solution. Thus for a nontrivial solution of the homogeneous equations to exist, we must have det(a,;)=0. This condition can also be shown to be sufficient to insure the existence of a nontrivial solution. A system of n simultaneous, linear, homogeneous equations in n unknowns has a nontrivial solution if and only if the determinant of the coefficients equals zero. If x,=d,,x,=d,,...,x,=d, is a solution of a set of n linear homo- geneous equations, then x,=cd),x.=cd),...,x,=cd, (where c is any con- stant) is readily seen to be a solution also. Since we have an arbitrary constant in the solution, to solve a set of linear homogeneous equations, we set one of the unknowns (e.g., x,) equal to an arbitrary constant (x,=c), thereby converting the equations to ” equations in n—1 unknowns; we discard one equation and proceed to solve the set of n—1 inhomogeneous equations in n—1 unknowns. (The most efficient way to do this is by successive elimination of unknowns, rather than by Cramer’s rule.) In quantum mechanics, the arbitrary constant c is usually evaluated by normalization.1.2 Mathematics Review ” Linear independence. The set of functions f,,f,.....f, is said to be linearly independent if the only way the equation OS taht o> +6f,=0 (1.83) can be satisfied is with c,=c,=--+- =c,=0. If the functions f,,...f, are linearly dependent, then (1.83) holds with some of the c’s being nonzero. This means we can solve (1.83) for one of the f’s and express it as a linear combination of the other members of the set. For a linearly independent set, no member of the set can be expressed as a linear combination of the other members. Taylor series. Let f(x) be a real function all of whose derivatives exist at the point x =a and within some neighborhood of a. Then f(x) can usually be expanded in an infinite series (called a Taylor series) as follows: 7 we Mass (n), 163) = 1a) +O (60) FO (eae = SOO ay" n=0 . , (1.84) where f(a) is the nth derivative of f(x) evaluated at x=a: qd p(x) (1.85) [There are a few exceptional cases where all the derivatives exist, but the Taylor series does not converge to f(x).] The series (1.84) will converge within some interval a-c 1 Aen (Gn)= EpVa(Gn) (1-107) From (1.106), (1.107), and (1.104), we have E=E\+E,t+-- +E, (1.108) Thus the wave function of a system of noninteracting particles is the product of wave functions for each particle, and the energy is the sum of energies for each particle. The same reasoning applies to a one-particle Hamiltonian that is the sum of terms that each involve only one coordinate: H= H, +H, + H,. Here v= ¥,(0)¥,(9)¥,(2) (1.109) E=E,+E,+E, (1.110) Ha=Ey HYy=B% Ha,=Ew, (1.1) 4 The chain rule. Let f be a function of x, y, and z: f=f(x,y,z). Suppose we carry out the following change of variables: x=x(r,5,0), y=y(r5,0, z=2(r,5,1) Substitution into f converts it to a function of the new independent variables r,s, and 1: L(69.2) = A145, 01 I (50) 263,01 = 8 (75,0) The chain rule relates the partial derivatives of g to those of f: (3), (5) G2)05) Ce), em1.2. Mathematics Review 2 with similar equations for (@g/0s),,, and (dg/ 92), ,. Spherical polar coordinates. The spherical polar coordinates r,0,p are defined by drawing a radius vector r from the origin to point (x,y,z): 0 is the angle between r and the positive z axis; r is the distance of (x,y,z) from the origin; is the angle the projection of r in the xy plane makes with the positive x axis. The ranges of the variables are 0 j’ =i’ > k’'=j’> k’=0, so that Ab_+ mm, + nyn,=0 Al,+ mymy+ nyn,=0 (1.121) L1,+ mym,+ nn,=01.4 The Particle in a Box B 13 UNITS We shall use mainly the cgs Gaussian system of units. This is a mixed system with electrical quantities measured in cgs electrostatic units (esu) and magnetic quantities measured in cgs electromagnetic units (emu). In 1960, the eleventh General Conference on Weights and Measures recommended the International System of Units (Systéme International d’Unités), abbreviated as SI units, for use in science; SI units are essen- tially the rationalized mks system of units. Relations between SI units and Gaussian units are given in Table A.4 of the Appendix. Table A.5 allows one to convert equations from SI to Gaussian units. Quantum chemists often use atomic units, a system of units in which h, the proton charge e, and the electron mass m each have a numerical value of 1. In this system, the unit of length is the bohr, defined by 1 bohr = 4 = h?/ me? =0.52918 A, where ay is the Bohr radius and the angstrom (A) is defined by | A=10 * cm=10~'° m; the unit of energy is the hartree, defined by | hartree = e?/a)= 27.212 eV. . 1.4 THE PARTICLE IN A BOX The one-dimensional particle-in-a-box problem is that of a single par- ticle subject to the following potential-energy function: V(x)=0 for0/ where / is the length of the box. The stationary-state wave functions and energies are found from the time-independent Schrodinger equation. Out- side the box, the wave function must be zero, because V is infinite. Inside the box, where V=0, the Hamiltonian is —(f?/2m)d?/dx?, and the Schrédinger equation is —(# /2m)d*p/ dx? = Ey (1.122) Equation (1.122) can be put in the form (1.97), and (1.99) gives as the general solution $= A cos [(2mE/#)'x] + Bsin[(QmE/A)'?x] (1.123) Since y is zero outside the box, for to be continuous, the function (1.123)u Quantum Mechanics and Electronic Structure must vanish at the ends of the box, x=0 and /. The condition y(0)=0 gives A =O, leaving only the sine term in (1.123). For the condition ¥(/)=0 to hold, the argument of the sine function must equal nz at x=/, where n=1,2,3,.... (Negative values of n do not give linearly independent solutions; n=0 makes y=0 everywhere.) Hence E= nh? /8ml?. The con- stant B is determined by normalization. The allowed wave functions and energies are then = (2/1)? sin(nax/1), OS have the same energy. [We are using ket notation (Section 1.1), denoting each state by listing the quantum numbers a,,n,,1,.] A free particle is a particle subject to no forces, so that V =0 everywhere. For a free particle moving in one dimension, the Schrédinger equation is (1.122) and its solution is [Eq. (1.98)] y=cyexp[i(2mE)'?x/h] +cyexp[ = i(2mE)'?x/h] (1.129) To keep y from going to infinity as |x| goes to infinity, the energy E must be nonnegative: E > 0. 1.5 THE HARMONIC OSCILLATOR The one-dimensional harmonic oscillator has the potential energy V=hkx?=20y?mx? (1.130) where k is the force constant (the force on the particle is F= —~dV/dx= —kx), m is the particle’s mass, and v=(1/2a)(k/m)'? (1.131) is the classical vibration frequency. The Schrédinger equation is ha 2\y= ( Im aa te? = Bb (1.132) Carrying out a power-series solution, one finds that the quadratically integrable normalized solutions of (1.132) are box) =(a/7)'4(2%1) Pe PH (lx) (1.133) a=2avm/h (1.134) v=0,1,2,... (1.135) where the functions H,, are certain polynomials (the Hermite polynomials)6 Quantum Mechanics and Electronic Structure defined by H(z) =(—V'e"(d"e"?/de"), — n=0,1,2,... (1.136) The reader can verify that Ho=1, Hy=2z, Hy, =427-2 (1.137) The Hermite polynomials obey the recursion relation 2H,(z)=aH,_(z)+ $H,4 (2) (1.138) The explicit form of the ground-state (v=0) wave function is Yo=(a/n) tee??? (1.139) The v=0 and v=! wave functions are sketched in Fig. 1.1. For any stationary state of the oscillator, the number of nodes in y equals the quantum number v. For states with » even, the Hermite polynomial H, contains only even powers of x, and y, is an even function; for states with v odd, H, contains only odd powers, and y, is an odd function. The energy levels are found to be E,=(v+ |)hv (1.140) Note that y is nonzero everywhere (except at isolated nodal points). Classically, the oscillator is confined to the region where its total energy is fa) b) Fig. 1.1. Harmonic-oscillator wave functions for (a) 0=0; (b) o= 1.1.6 Orbital Angular Momentum 27 greater than or equal to the potential energy: E > V. Quantum mechani- cally, there is some probability for finding the particle in classically forbidden regions where E< V. For the three-dimensional harmonic oscillator, V = 4k,x? +th,y?+$k,z?, Separation of variables gives the wave function as the product of three one-dimensional harmonic oscillator wave functions, and gives the energy as E=(0, + })hy, +(v, + 4)hv, + (0, + $)Av, (1.141) 2, =0,1,2,... o,=0,1,2,... o,=0,12,... The isotropic three-dimensional harmonic oscillator has kK, =k =k, and », re =). 16 ORBITAL ANGULAR MOMENTUM lvet r be the position vector from the origin to a particle in motion; let v be the particle’s velocity and p its linear momentum: r=xit+yj+zk (1.142) v= dr/ dt =(dx/dt)it (dy/dt)j+(dz/d)k (1.143) p= mv= mo, i+ mv,j+ mo,k (1.144) where m is the particle’s mass, x, y, and z are its coordinates, and i, j, and k are the unit vectors along the axes of a Cartesian coordinate system, (The particle’s acceleration a is a=dv/dt=d’r/dt?.) The orbital angular momentum L of the particle is defined as i j k L=rxp=|x y 2 (1.145) Px Py Pz L.=YP,— Py Ly =2P,— XP ~ L,=xp,—yp, (1.146) The square of the magnitude of the orbital angular momentum is D=L+L=124+13+L? (1.147)B Quantum Mechanics and Electronic Structure Replacement of classical quantities by operators gives L,= —ih(ya/dz— 20/ dy) (1.148) L, =~ ih(z8/ax - x9/22) (1.149) L, = — ih(x8/dy — yd /dx) (1.150) (t= 02+ 02+ 2? (1.151) One finds the following commutation relations to hold: [£.£,]=iné, (1.152) [2.é]=[2.6]-[24]=0 (1.153) Hence there exists a complete set of common eigenfunctions for 2 and any one of its components. The eigenvalue equations for L? and L, are found to be separable in spherical polar coordinates (but not in Cartesian coordinates). Using the chain rule to transform the derivatives, we can find L,, L, and L, in spherical polar coordinates, and then use (1.151) to find 2. One finds L,=— ihd/ dp (1.154) n a? a ee nOn 2=—W#| — +cotd-— + > ll L (33 OM Sint # (1.155) L, and Z have more complicated forms than L, but also involve only @ and q, and not r. Use of (1.8), (1.115), and (1.155) shows that the one-particle kinetic- energy operator can be expressed as (a? 29), 1 pp | ater abe L (1.156) Let Y¥(8,~) be the common eigenfunctions of i? and L,. To solve PY =aY, L, Y =bY (where a and 6 are the eigenvalues), we try a separa- tion of variables, writing Y(8,~)= S(@)T(¢). Solving the differential equa-1.6 Orbital Angular Momentum 2» tions, one finds that the well-behaved, normalized solutions are T(@) = (20) 'eime (1.157) (21+ 1) = |m) 7’? Sim(9)= 2 (alm! P}"(cos@) (1.158) 7=0,1,2,... m=~1,—/+1,...,0,...,/-1,1 where the associated Legendre functions P!"\(w) are defined by Imj2_dt + lm dw!*lnl (w=1), 1=0,1,2,... (1.159) Pw) = (1-2) The eigenvalues of L? and L, are found to be /(/+ 1h? and mh, respec- tively. The angular-momentum eigenfunctions Y,"(6,p)= Sim(9) Tn (Q) are calJed spherical harmonics: (= [mp]? —|my})r ¥"(8,9) = oa (een P!™\(cos8)e”™? (1.160) LY 7(6,9)= K+ IY"8,9) (1.161) L,Y /"(0,9) = mhY/"(0,) (1.162) The eigenfunctions Y,” have been normalized; moreover, being eigenfunc- tions of the Hermitian operators Z? and L,, they are orthogonal; hence 20 po [fren YP O.p)sindd dp=8,,By.q¢ (1.163) 0 oO where the sin@ factor comes from the volume element (1.114). A moving electric charge generates a magnetic field. For a particle with charge q and mass m, its orbital angular momentum L gives rise to a magnetic dipole moment p, given (in Gaussian units) by i, =(q/2me)L (1.164) where c is the speed of light.30 Quantum Mechanics and Electronic Structure 1.7 ANGULAR-MOMENTUM LADDER OPERATORS The last section discussed angular momentum from the viewpoint of solving the differential eigenvalue equations that result from expressing L? and L, as differential operators. The ladder-operator approach uses just the angular-momentum commutation relations, and is also applicable to spin angular momentum (Section 1.15). Rather than using L, we shall use the symbol M to indicate that we are dealing with any kind of angular momentum. AA We are given three linear operators M,,M,,M, that by hypothesis obey the commutation relations [cf. (1.152)]: [m]= inet, [it ihM,, [,,M,]=inM, (1.165) The M? operator is defined as M? = M2+ M+ M? (1.166) It follows from (1.165) and (1.166) that [s,M,]=[?,M,]=[47,17,]=0 (1.167) so that we can have simultaneous eigenfunctions of M? and M,: MPV = 4 Yim MY = By Vy (1.168) where a, and b,, are the eigenvalues, and Y,,, (not necessarily the spherical harmonics) are the common eigenfunctions. The raising operator M, and the lowering operator M_ are defined by =M,+iM,, M_=M,-iM, (1.169) From (1.169) and (1.165), we find that M( MY) = (Bm A) Ms Yo) (1.170) M,(M_ Ym) = (Om ~A)( MY) (L171) so that application of M, to a given eigenfunction Ym converts it to an eigenfunction of M, with “eigenvalue fA higher than that Of Yim’ M_ lowers1.7 Angular-Momentum Ladder Operators wv the M, eigenvalue by h. From (1.169) and (1.167), we find that MMs Vm)=4(M Vm), MM Yq) = 44( M1 Yq) (1-172) so that 47, Y, ‘im and My Y,, are still eigenfunctions of M? with the same eigenvalue a; as Y,,,. From (1.170) and (1.171), the eigenvalues of M, for a given a, are ‘spaced by fh. Using M2+ M2=M?— M2 (1.173) and (1.168), we get (M24 My) Yn = (4)— 85) Yom (1.174) Since M?+ M, is a nonnegative physical quantity, its eigenvalues a,— 63, must be nonnegative; therefore —a}/? , Thus for a given 4, the eigenvalue b,, is bounded above and below. Let Bmax be the highest M, eigenvalue for a given a, and let Y, .,,, be the corresponding eigenfunction. In general, M, acting on any Y,,, converts it to an eigenfunction with M, eigenvalue increased by fh [(1.170)]. However, since Wieaet is the eigenfunc- tion with the highest M, eigenvalue, application of M, to Y, max Cannot give a new eigenfunction, but must give zero: M, Y; max = 9. Applying M_, we get M_ MY, max=0. We now use (1.169), (1. “73, the first equation of (1.165), and (1.168) to get O= MM Yay =(M?— M2—AM,)Y, ran = (4) bog BO max) Yi max = Boag + AD nan (1.175) Similar reasoning starting from M_ Y,,min=0 leads to a, =b2 nin PD cn which, when substituted into (1.175), gives b2,, +hb,., TAD boi, =0. Regarding this as a quadratic Sea in the u unknown a we find the TOOtS: bingy = — min ANG bynax = min — ft. The second root is rejected, since binax Cannot be less than pa We found previously that the M, eigenvalues for a fixed M? eigenvalue are spaced by A; hence 6,,,,= Brain + 1h, where n is a nonnegative integer (0,1,...). Since Duin= — Pray We have bra.= — Byrax tinh, or bmax= 30h, — Pmin=—4nh, — n=0,1,... From (1.175), we have a,=(4 n? + Sn)h?. Setting j = }n, we have Pmax=JAy2 Quantum Mechanics and Electronic Structure Pag = — jh, a= (J? + jh". Thus the eigenvalues of M? are gait, 7=04,19,2,... (1.176) Since, the eigenvalues of M, are spaced by fy, start at jh, and end at —/h, the M, eigenvalues are pene Ay oe PTD lel) eet teal tial tf) (1.177) Note the possibility of half-integral (as well as integral) values for the quantum number j. For orbital angular momentum, only integral quantum numbers occur, but for spin angular momentum, we can have half-integral values. 1.8 PARITY An even function of the 3 Cartesian coordinates is defined as one that obeys Sl Xp Day 2p Xap Yr 7 2n) LOY Zp Xp ParZn) (1-178) An odd function obeys S(= Xp Dy ~ Zp es — Xe Pe Ea) = SMI Ze Xa ParZn) (L179) The parity operator Ii is defined by TRG (Corey sz xyes) =o (Ge X iste 1 l= tt teat sae) etait te 80) {1 is a linear operator. Note that fI(fe)=(f1f)(f1g), where f and g are any two functions. Clearly [?g = g, so that I1?= 1. The eigenvalue equation for TI is figj=k,q, Application of Tl gives Ip,=k,fl9,=k29,=9, since IP? =1. Hence k?=1, and the eigenvalues of the parity operator are 1. For k,= +1, the eigenvalue equation reads I]9,(x,, ...,z,) X4,++-5Z,)5 using (1.180) we have 4 — Xo oy — 2g) = BCE oeeeZa) (1.181) Hence the eigenfunctions of fi that correspond to a +1 eigenvalue are all well-behaved even functions. Similarly, the II eigenfunctions with —1 eigenvalue are all well-behaved odd functions.1.9 The Variation Method 3 For the commutator of fl with the Hamiltonian, we have (1.4) =[IL.7]+[ILV]. Since 7 involves the second derivatives of the Cartesian coordinates, we have I(Tg)= (Ig). so that [I and 7 commute. Hence (i. j=(t, M1 We have TVe)= (iv (fig): if V is an even function, then T= V, and Tl Ve= v lig. so that [I and ¥ commute. Hence when the potential energy V is an even function, the parity operator commutes with the Hamiltonian, and the stationary-state eigenfunctions of H can be chosen to be eigenfunctions of IT. But the eigenfunctions of II are even functions and odd functions. Hence when V is even, each of the wave functions y¥, can be chosen as either even or odd. (Recall, e.g., the harmonic oscillator.) A state whose wave function is even is said to have even parity: similarly for odd wave functions. The following equations hold for even and odd functions: f Moav=0 for f odd (1.182) J foar=2f “Ide for f even (1.183) zZ,) is an odd function of some or all of the 3n its integral over all space vanishes. Similarly, if g(x)... Cartesian coordinates 1.9 THE VARIATION METHOD Because of interelectronic repulsions, the Schrédinger equation for many-electron atoms and molecules cannot be solved exactly. The two main approximation methods used are the variation method and perturba- tion theory. The variation method is based on the following theorem. Given a system with time-independent Hamiltonian H, then if @ is any normalized, well-behaved function that satisfies the boundary conditions of the problem, one can show (by expanding in terms of the eigenfunc- tions of H) that W= [o*Hodr> Ey (1.184) where Ep is the true ground-state energy. If m is not normalized, then (1.184) is replaced by W = (glHlo>/> Eo (1.185)M Quantum Mechanics and Electronic Structure A trial variation function @ thus provides an upper bound for the ground- state energy. One usually includes variational parameters in » and looks for those values of the parameters that minimize the variational integral W,; the function @ then serves as an approximation to the ground-state wave function. A special type of variational function is the linear variation function. Here, @ has the form prnhtaht t6h= 3 (1.186) where the functions f,,...,f, are n linearly independent well-behaved functions, and the parameters c,,...,¢, are numbers chosen to minimize the variational integral W. Substituting (1.186) into (1.185) and setting OW /dc,=0 (j=1,...,n) so as to minimize W, we are led to the following set of n linear, homogeneous equations in the n unknowns ¢,, pen Cna > ((A;-S);W)g]=0, i= 1,2,...,n (1.187) j=l H, = fs iifar, S,= fithar (1.188) For a nontrivial solution of the equations (1.187), the coefficient de- terminant must vanish (Section 1.2). Hence det(H,- S,W)=0 (1.189) Ay-SyWo Ay SpWo Hy SW Hy SyW Hy~SnW Han SW] (1190) ~SyW Hyg SQW Hy Spy H, al The secular equation (1.189) is an algebraic equation of nth degree, which we solve for the n roots: Wo< W, OL WO> jy © & FO — fo Eo *m (1.200) EM = (YAW (1.201) ae 5 lot KYO WP Eo - Eo ( 1.202) ken where the sums are over all states except state n. Because of the difficulty in evaluating the infinite sums needed to find E®,E®),..., one sometimes uses a variation-perturbation approach, in which E®,£®, ... are evaluated by looking for functions that minimize certain integrals involving H’; see Hameka, p. 223, for details. Now consider perturbation theory for an unperturbed energy level that is n-fold degenerate, with the states y, all having energy £. Any linear combination of the n degenerate functions (1.203) is an eigenfunction of H°; as A0, each of the n perturbed wave functions ¥, (j=1,...,m) that correspond to the n-fold- degenerate unperturbed level will go over into some particular linear (1.203)1.10 Perturbation Theory 3 combination of the functions (1.203): sey (1.204) ws 0) @ limy= DoyP=a j= i=l where we used g to denote these n correct zeroth-order wave functions; which are the correct zeroth-order functions depends on H’. (The j subscript in cy indicates the set of coefficients for g{.) Substitution of (1.198) and (1.197) with ¥ replaced by g into (1.195) leads to the following set of linear, homogeneous equations for the coefficients cy in 9: > [(Fai- BPS, )ey]=0, mata (1.205) i=l Hy, = YAW (1.206) [The Kronecker delta occurs in (1.205) because we assumed that the ys were chosen to be orthogonal.] For a nontrivial solution of (1.205), the coefficient determinant vanishes: det (H;,- E'P6,,,)=0 (1.207) This secular equation is an algebraic equation of degree n with n roots Ej, ..., Ef” that give the n first-order energy corrections. Substitution of each £;" in turn into (1.205) allows one to solve for the set of coefficients cy G=1,....n) that go with the root.£. Having found the ” correct zeroth-order wave functions g (j=1,...,n), we can then proceed to find E,, ¥, and so on; the formulas for these corrections turn out to be essentially the same as for the nondegenerate case, provided that g° is used in place of ¥. When the secular determinant in (1.207) is in diagonal form (all off- diagonal elements equal to zero), it follows that the initially chosen unperturbed wave functions Wy of the degenerate level are the correct zeroth-order wave functions g. Conversely, if the initially chosen unper- turbed wave functions are the correct zeroth-order functions, then the secular determinant will be in diagonal form. When the secular determinant is in block-diagonal form, the secular equation (1.207) splits up into several equations. For example, if we have a five-fold degenerate unperturbed level and the secular determinant hap- pens to consist of a 2x2 block and a 3X3 block [as in (1.79)], then the secular equation splits into two separate equations; moreover, two of the38 Quantum Mechanics and Electronic Structure correct zeroth-order wave functions are linear combinations of y\ and y only, and the other three correct zeroth-order functions are linear com- binations of yO, y{, and y only. One method to put the secular determinant in block-diagonal form (and thereby simplify the calculation) is to choose the initial zeroth-order functions y{° to be eigenfunctions of some operator A that commutes with both H° and H’. Since [H’,A]=0, it follows from (1.51) that the integrals H,, vanish when ¥ and y{® correspond to different eigenvalues of A. Therefore, the secular determinant will be in block-diagonal form, with each block corresponding to a different eigenvalue of A. 1.11 THE CENTRAL-FORCE PROBLEM For a one-particle central-force problem, the potential energy is a function only of the particle’s distance from the origin: V=V(r); from (1.156), the Hamiltonian is A= (-f/Im)(a2/ dr? + 2 '8/dr) + (1/2mr?)L? + V(r) (1.208) Because V depends only on r, one finds that this Hamiltonian commutes with the orbital angular: -momentum operators L? and fy. Hence the eigenfunctions of H can also be chosen as eigenfunctions of P and a. Fy= Ey (1.209) Eyat(+ ih, — 1=0,1,2,... (1.210) Ly=mhy, m=—1,...,41 (1.211) The eigenfunctions of [2 and L, are the spherical harmonics Y/"(0,); these eigenfunctions can be multiplied by an arbitrary function of r and still be eigenfunctions of L? and L, (which do not involve r). Hence for a central-force problem, y has the form v= R(r)¥;"(8,9) (1.212) Use of (1.212), (1.208), and (1.210) in (1.209) gives as the differential equation satisfied by the radial function R(r) (=f /2m)[ R"(r) + (2/ A) R()) +L 2/2 + V(r)]R (1) = ER(r) (1.213)1,12 The Two-Particle Problem 39 1.12: THE TWO-PARTICLE PROBLEM Consider a two-particle system with the particles having masses m, and m, and coordinates (x,,¥;,2)) and (x,,y2,2). We define the relative (or internal) coordinates x,y,z of the system by XXX VEV2Vy 2 S22, (1.214) The coordinates X, Y,Z of the center of mass of the system are given by mx, + MX. my, +m. mz, + mz. ya aha y= Ait MY2 zo 222 (1.215) m,+m, mt+m, ” my +m, We shall impose the restriction that the potential energy V depends only on the relative coordinates of the particles: V= V(x,y,z). Substitution of (1.214) and (1.215) into the kinetic-energy expression leads to the following expression for the classical-mechanical Hamiltonian: H=py/2M +[pi/2u+ Vixy.2)] (1.216) where M is the total mass (M = m,+m,), the reduced mass p is defined as B= mym,/(m, + m) (1.217) and the linear momenta py and p, have components defined by Py=M(dX/dt), — py=M(dY/dt), — pz =M(dZ/dt) P.=w(dx/dt), — py=u(dy/dt), pp, = (dz /dt) Since the Hamiltonian (1.216) is the sum of a part dependent only on the center-of-mass coordinates X,Y,Z and a part (the bracketed terms) de- pendent only on the internal coordinates x,y,z, separation of variables allows us to conclude that the (complete) wave function is the product of a function Yyan(X,¥,Z) and a function yi_(x,y,z). Since pz,/2M is a free-particle Hamiltonian, 495 iS a free-particle wave function. The func- tion Pig, iS an eigenfunction of the operator corresponding to the bracketed terms in (1.216). The energy is the sum of an energy of translational motion of the entire system of mass M and an energy of internal (relative) motion of the two particles. We solve for the internal wave functions and energies by setting up a one-particle Schrédinger equation using the reduced mass yu and the relative coordinates x,y,z.4 Quantum Mechanics and Electronic Structure 1.13. THE TWO-PARTICLE RIGID ROTOR The two-particle rigid rotor consists of masses m, and m, attached to the ends of a rigid, massless rod of length d. Since the particles are held at a fixed distance from each other, the potential energy does not change as the rotor moves through space, and V is a constant, which we may take as zero: V=O, Thus the energy is entirely kinetic. V =0 is a special case of V depending only on the relative coordinates of the particles, so we may use the results of Section 1.12 to separate the translational and internal motions; the energy is the sum of translational energy of the system and energy of internal motion. (Ordinarily, the energy of internal motion of a two-particle system is of two types--vibrational and rotational. In this case, the interparticle distance is held fixed, so there is only rotational internal motion.) To find the internal wave functions and energies, we must find the eigenfunctions of the Hamiltonian corresponding to the bracketed terms in (1.216). The potential energy V =0 is a special case of a central force (Section 1.11), so we use spherical polar coordinates to deal with the internal motion. However, since the interparticle distance is constant, the relative spherical polar coordinate r is not a variable; the only variables are the internal angular coordinates @ and » (which define the orientation of the rotor interparticle axis with respect to a coordinate system that translates with the rotor but that does not rotate). The internal wave function y is a function of @ and g; but, since this is a central-force problem, ¥(8,¢) must be a spherical harmonic Y/"(0,p). The Hamiltonian is (1.208) with the r derivatives omitted (since r is not a variable), with r in £2/2mr? set equal to d, with m replaced by the reduced mass Ke mym,/(m,+m,), and with V(r)=0. Thus the Schrédinger equation Aw = Ey for the rotor’s internal motion is (1/2nd?)L?Y (8,9) = EY}"(0,9) Use of (1.161) gives E=I(1+1)h/2pd?, — 1=0,1,2,... (1.218) It is conventional to use J and M, rather than / and m, for the rotational angular-momentum quantum numbers. Also, the moment of inertia J of the rotor about an axis passing through the center of mass and perpen- dicular to the interparticle line is found to equal pd?. (The moment of inertia of a system of particles about an axis is defined by T= Sm, where r, is the perpendicular distance from particle i to the axis.) Hence we write E=J(J+1)/21, J =0,1,2,... (1.219)114 The Hydrogen Atom 4 The two-particle rotor levels are (2J + 1)-fold degenerate, since for each J there are 2/ +1 possible values of the quantum number M (ranging from —J to +J), and E is independent of M. 1.14 THE HYDROGEN ATOM The hydrogenlike atom consists of a nucleus of charge Ze and one electron. (Z=1 for H, Z=2 for He*, etc.) The potential energy is given by Coulomb's law as V=—Ze?/r (1.220) where r is the distance of the electron from the nucleus. Since V depends on only the relative coordinates of the particles, the energy is the sum of translational energy of the atom as a whole and energy of internal motion (Section 1.12). The Hamiltonian for internal motion is A= ~(P/2p)V2- Ze2/r (1.221) where p is the reduced mass [Equation (1.217)] and V? refers to relative coordinates. The Hamiltonian (1.221) is a central-force problem; hence v= RO)Y"(9.9) (1.222) where the radial function R is found by solving (1.213) with V= — Ze?/r. Using a power-series solution, one finds that the bound-state (E <0) radial functions are R,(r) = Nr'e~ 2°/"@ L244 \(2Zr / na) (1.223) where A is the normalization constant, a is defined as a=h*/ pe? (1.224) nis a quantum number such that n=1,2,3,..., l LSS ede m= —1/27~ The symbol fdv will denote integration over the full range of the spatial coordinates only, without summation over spin variables. Analogous to (1.169), the electron-spin raising and lowering operators are S$, =8,4iS,, (1.240) Analogous to the equations M, Y, max =0= M_ Y,,min Of Section 1.7, we have S,a=0, $_B=0 (1.241) Also, since - is a raising operator, we must have B=ca, where c is some constant. We can evaluate c by use of the normalization of the spin functions and the Hermitian property of S, and S,; one finds (Problem 1.8) |c| =A. Choosing the phase of ¢ as zero, we have S$, B=ha (1.242) One finds similarly that S_a=hp (1.243) From (1.240)-(1.243), we get Sa=4hB, §,B= ha (1.244)1.16 The Pauli Principle 45 S,a=4ihB, $,B=—tiha (1.245) Just as orbital angular momentum L gives rise to a magnetic dipole moment p,, spin angular momentum S gives rise to a spin magnetic dipole moment ps. Dirac’s relativistic theory of the electron showed that Bs = —(e/mc)S = — g.(e/2mc)S (1.246) where g, is the electron g factor. Although Dirac’s treatment gave g,=2, subsequent more refined theoretical work and experimental data show that & is not exactly 2; rather Be =U1+a/20+-++)=2.0023 (1.247) where the fine-structure constant a is defined as a = e*/he=1/137.036 (1.248) The dots in (1.247) indicate smaller terms involving a’, a3, and so on. 1.16 THE PAULI PRINCIPLE Consider a system of 7 particles with state function ¥(4),...,9j.-+-54a ++++4n), where g, stands for the four coordinates (three spatial and one spi of particle i. Suppose we interchange the coordinates of particles i and k in W. If the n particles of the system are identical, then experimental evidence and relativistic quantum field theory show that the state function ¥ must be restricted in form so as to be symmetric with respect to exchange if the particles have integral spin (s=0 or | or 2 or...) or antisymmetric with respect to exchange if the particles have half-integral spin (s=4 or 3 or...). This is the Pauli principle. In the nonrelativistic quantum mechanics we are using, the Pauli principle is an additional postulate. By “symmetric with respect to exchange,” we mean that ¥ is unchanged on interchange of the coordinates of identical particles: VG Ghr 1 Fin In) = Gay in 2G In) By “antisymmetric with respect to exchange,” we mean that interchange of the coordinates of identical particles multiplies ¥ by — 1: da) = — VG. V9 de In) arr Quantum Mechanics and Electronic Structure Particles having half-integral spin and requiring antisymmetric wave func- tions are called fermions, particles having integral spin and requiring symmetric wave functions are called bosons. Note that if we put 9, =4; in this last equation, we get ¥ = —¥, so that W=0 when g,=4, that is, when x,=x,, ¥)=Yy, Z3=2,, and m,=my. Hence, there is zero probability for two fermions with the same spin (i.e., the same m, value) to be at the same point in space. The Pauli principle causes electrons of like spin to keep apart from each other. As an example, consider two identical, noninteracting particles, 1 and 2. Since the particles are noninteracting, solution of the Schrédinger equation using separation of variables gives for the spatial part of the system’s wave function: f(x,, ¥1, 21)8(%2, ¥z Zz), which we abbreviate to f(1)g(2). The function f(2)g(1) is degenerate with f(1)g(2), since particles 1 and 2 are identical. The functions f(1)g(2) and f(2)g(1) are each neither symmetric nor antisymmetric with respect to exchange. To satisfy the Pauli principle, instead of f(1)g(2) and f(2)g(1), we must use as the spatial part of the wave function the following normalized linear combinations: 2-7 f(g (2) + g()F(2)] (1.249) 2°71 f(g (2) — 8()FQ)I (1.250) [Of course, if f and g are the same functions, then (1.250) vanishes, and (1.249) is replaced by f(1)f(2).] The spatial functions (1.249) and (1.250) must be multiplied by a spin function. For electrons, we have the following four possible two-electron spin functions, three symmetric and one antisymmetric: So Ms a(1)a(2) l 1 (1251) symmetric: 2-7 a(1)B(2) + B(Ia(2)] ! 0 (1.252) B(I)BQ) el (1.253) antisymmetric: — {27 '/"[«(1)B(2)— B(1)a(2)] 0 0 (1.254) The quantum numbers listed are for the eigenvalues of the total-spin operators S? and S,, where the total spin S is defined as S = S, +S,. Since electrons are fermions, the symmetric two-electron spatial function (1.249) must be multiplied by the antisymmetric spin function (1.254) to give an overall wave function that is antisymmetric; the antisymmetric spatial function (1.250) must be multiplied by one of the symmetric spin functions (1.251)}+(1.253).1.17 Many-Electron Atoms 47 An example is the perturbation treatment of the He atom in which we neglect the interelectronic repulsion e?/r,,. The correct zeroth-order wave function for the ground state is 1s(1)Ls(2)fa(1)B(2)— B(I)ae(2)] /247? (1.255) where Is denotes a hydrogenlike orbital. The correct zeroth-order func- tions for the states arising from the 1s2s configuration are 'S: 2-'7[1s(1)2s(2) + 2s(1)1s(2)]}2- [a (1) B(2)~ B)a(2)] 38: 27 /?[ Ls(1)2s(2) — 28(1)15(2)]ar(1)ae(2) 3§: 27 ![19(1)2s(2) —2s(1)15(2)]27 [a (1) B(2) + BU)a(2)] 3§: 27! 19(1)2s(2) —2s(1)1s(2)] B(1)B(2) (1.256) The term symbols 'S and °S are defined in the next section. For the states of the Is2p configuration, we get zeroth-order wave functions similar to those of (1.256), except that there are 12 (=4 3) zeroth-order functions instead of 4; the factor 3 arises from the spatially degenerate 2p,, 2p,, and 2p, orbitals. For three or more electrons, the wave function cannot be factored into a simple product of a space part and a spin part; see the next section. 1.17 MANY-ELECTRON ATOMS With internal nuclear motion neglected, the Hamiltonian for a many- electron atom with atomic number Z is Hi=- # Sv-zey mal >> fs, (1.257) Gogoi The first term on the right is the operator for the electrons’ kinetic energy; the second term is the operator for the potential energy of attraction between the electrons and the nucleus (r, being the distance between electron i and the nucleus); the third term is potential energy of repulsion between all pairs of electrons (r, being the distance between electrons / and /); the last term is the spin-orbit interaction (discussed below). In addition, there are other relativistic terms besides spin-orbit interaction, which we neglect. If we neglect interelectronic repulsions and spin-orbit interaction, then4a Quantum Mechanics and Electronic Structure the atomic Hamiltonian is approximated by H°, where (70 a 2 Ze? a -2 (- ami, ) (1.258) Since 7° is the sum of hydrogenlike Hamiltonians, the zeroth-order wave function is the product of hydrogenlike functions, one for each electron. We call any one-electron spatial wave function an orbital. To allow for electron spin, each spatial orbital is multiplied by a spin function (either a or B) to give a spin-orbital. To introduce the required antisymmetry into the wave function, we take the zeroth-order wave function as a Slater determinant of spin-orbitals. For example, for the Li ground state, the normalized zeroth-order wave function is Is(e(1) Is(I)BQ) — 2s(1)a(1) Ve | a2) 18(2)B2) _28(2)a2) (1.259) 1s(3)a(3) —15(3)B(3)—-2s(3)a(3) Since interchange of the rows of a determinant multiplies it by — 1, (1.259) is antisymmetric. To fully specify the Slater determinant, it suffices to simply list the spin-orbitals; hence we abbreviate (1.259) to |1sa1sB2sa\. Adopting the convention that a bar indicates spin function B and the absence of a bar indicates spin function a, we further abbreviate (1.259) to [Is Ts 2s] (1.260) Note that if the spin-orbital in column three of (1.259) were the same as that of either column one or column two, the determinant would vanish (since two columns would be identical). Hence antisymmetry of y means that no two electrons can occupy the same spin-orbital in an approximate wave function in which electrons are assigned to orbitals. Use of H° as the approximation to the Hamiltonian gives rise to electron configurations, in which the electrons are assigned to orbitals. Thus for ground-state Li, we have the configuration 1s?2s, with two electrons in the 1s orbital and one in the 2s orbital. The ground-state C configuration is 15?2572p?. Orbitals having the same n and the same / are said to belong to the same subshell; the electron configuration is specified by stating how many electrons are in each subshell. In general, a given electron configuration has several wave functions, each corresponding to the same eigenvalue of the zeroth-order Ham- iltonian A°. This degeneracy is partly removed by inclusion of the electron1.17 Many-Electron Atoms 490 repulsion term H,.,. =>>¢ (1.261) ioy>i thereby splitting each configuration into several ¢erms (see below). Because of H,.p. the operators for orbital angular momenta of individual electrons do not commute with the atomic Hamiltonian. Defining the ‘otal electronic orbital angular momentum L by L=S1L, (1.262) one finds that the total orbital angular momentum operators 2? and L, commute with the Hamiltonian H°+ Hp. Similar to (1.262), one defines the total electronic spin angular momentum S of the atom by S=Z;S;. Since we are interested in the total angular momentum, we need the rule for addition of angular momenta. Let M, and M, be two angular momenta of any type (both orbital, both spin, or one spin and one orbital). Let the eigenvalues of M?, M,,, M?, M,, be j,\(j,+ 1), mh, jC), + 1), mh, respectively. Defining the total angular momentum M as M=M,+M, (1.263) one finds that the components M,, M,, M, of M obey the usual angular- momentum commutation relations; hence by the arguments of Section 1.7, the eigenvalues of M? and M, are J+, J=0,4,1,.. and Mh, My=—J,.0.J (1.264) One can show that the total angular-momentum quantum number J takes on the possible values JS thph tim bes lhl (1.265) (For example, if j,=2 and j,=3, then J has the possible values 5,4, 3,2, 1.) The eigenvalues of the total electronic angular-momentum operators L? and S$? of the atom are L(L+1), = L=0,1,2,... (1.266) S(S+1), $=0,4,1,... (1.267) One finds that states of a given electronic configuration that have the same values of L and S have the same energy eigenvalue for the Hamiltonian A°+ ap: ; such states are said to belong to the same erm. Different terms50 Quantum Mechanics and Electronic Structure arising from the same electronic configuration have different energies. For example, consider the |s2p He configuration. We have s,= 4, s,=} for the two electrons. Hence from (1.265), we have S=1, 0 for the total electronic spin quantum number. We have /,=0, /,=1 for the electrons, and (1.265) gives L=1 for the total electronic orbital angular-momentum quantum number. To designate each term, we specify the L value by a code letter (S,P,D,F,G,...indicate L values of 0,1,2,3,4,...), and we write the mul- tiplicity 2S +1 as a left superscript on the L code letter. Thus the terms arising from the He Is2p configuration are >P and 'P, each having a different energy. Equation (1.256) gives the zeroth-order wave functions of the He ls2s terms >S and 'S. Electrons in filled subshells make no contribution to the total spin or orbital angular momentum. Thus a 1s?2s72p3p configuration gives the terms °S,5P,°D,'S,'P,'D. The terms arising from a configuration with more than one electron in a given partly filled subshell are not so easily found, since the Pauli principle must be taken into account. We omit discussion. Finally, consider H,,. From an electron’s viewpoint, the nucleus is moving around the electron, thereby producing a magnetic field (propor- tional to the electronic orbital angular momentum L); this field interacts with the electronic magnetic moment (which is proportional to S$), giving the spin-orbit interaction. When H,,, is included in the Hamiltgnian, the total electronic orbital and spin angular momentum operators L? and S$? no longer commute with the atomic Hamiltonian, but the Sperator for, the total electronic angular momentum J does commute with H°+ H, + Ae The definition of J is J=L+S. The eigenvalues of J? are JU + Iie. Inclusion of A, , splits each term into several closely spaced /evels, each level corresponding to a different J value. (Different values of J corre- spond to different relative orientations of L and S, and different orienta- tions of these vectors give different amounts of spin-orbit interaction.) The possible values of J are found from the L and S values of the term using the rule (1.265). We designate a level by adding its J value as a right subscript to the term symbol. Thus for a P term, L=] and S=1, soJ can be 2, 1, or 0; hence the levels of the term are *P,, *P,, *Po. Even though 2 and §? don't commute with H when H,,, is included, provided Hi,,, is small, £2 and §? “almost” commute with H, and it is a good approxima- tion to use the quantum numbers L and S to characterize atomic energy levels. Tables of atomic energy levels as derived from atomic spectra are available: C. E. Moore, Atomic Energy Levels, Nat. Bur. Stds. Circular 467, vols. I, II, and II, 1949, 1952, and 1958. Each level is (2/ + 1)-fold degenerate, corresponding to the 2/ + | values of M,, where M,h is the eigenvalue of J,. This degeneracy can be removed1,17 Many-Electron Atoms st by application of an external magnetic field (the Zeeman effect), thereby splitting each level into its component states. The preceding discussion, in which we first add the L,’s to get L and add the S,’s to get S, and then add L and S to get J, is called Russell-Saunders coupling (or L-S coupling). Russell-Saunders coupling is valid when H,,. is small compared to Hep: Spin-orbit interaction increases with increasing atomic number, so that for heavy atoms H,,, exceeds H,,, and Russell- Saunders coupling does not hold. For very heavy atoms, /-/ coupling holds; here, one adds the spin and orbital angular momenta of electron i to get the total angular momentum j, for electron i; the j,’s are then added to give the total electronic angular momentum J. The zeroth-order antisymmetric wave functions of closed-subshell states, states that have only one electron outside a closed-subshell configuration, and states that are one electron short of having a closed-subshell con- figuration can be expressed as a single Slater determinant [e.g. (1.259)]. However, for open-subshell states in general, one has to take an appropriate linear combination of a few Slater determinants to get a state that is an eigenfunction of L? and S*. As noted above, application of an external magnetic field B splits each atomic energy level into states, each state having a different M, value. The electronic orbital angular momentum L gives rise [Equation (1.164)] to an orbital magnetic dipole moment p,=—(e/2mc)L. The spin magnetic dipole moment is given by (1.246) as ps=—(e/mc)S. The interaction energy Eg between an applied magnetic field B and a magnetic dipole moment p is [Halliday and Resnick, Equation (33-12)] E,=—p-B (1.268) Hence the Hamiltonian for the Zeeman interaction is Fig = ~ (fi + fis)" B= Bh '(L+ 28) B= BBA“ (3,48) (1.269) where the Bohr magneton B, is defined as B, = eh/2mc, where J=L+S was used, and where the z direction was taken to be the direction of B. Since H, depends on J,, states with different M, values are split by the magnetic field. The magnetic-field vector B is called the magnetic induction or the magnetic flux density. A more suitable name for B would be the “magnetic field strength”; however, historically, the magnetic-field vector H was given the name magnetic field strength. This is because it was formerly believed that H was the fundamental magnetic-field vector; it is now known that B is the fundamental magnetic-field vector (just as E is the