MODERN QUANTUM CHEMISTRY Introduction to Advanced Electronic Structure Theory ATTILA SZABO Laboratory of Chemical Physics National Institutes of Health Bethesda, Maryland NEIL S. OSTLUND Hypercube, Inc. Waterloo, Ontario DOVER PUBLICATIONS, INC. Mineola, New YorkCopyright | Copyright © 1982 by Macmillan Publishing Co., Inc. Copyright © 1989 by McGraw-Hill, Inc. All rights resersd under Pan American and International Copyright Conventions. Published in Canada by General Publishing Company, Ltd.,.30 Lesmilf Road, Don Mills, Toronto, Ontario. Published in the United Kingdom by Constable and Company, Ltd., 3 The Lanchesters, 162-164 Fulham Palace Road, London W6 9ER Bibliographical Note “This Dover edition, first published in 1996, is an unabridged, unaltered republication of the “First Edition, Revised,” originally published by the McGraw-Hill Publishing Com- pany, New York, 1989. The original edition was published by the Maemillan Publishing, Company, New York, 1982. Library of Congress Cataloging-in-Publication Data Seabo, Attila, 1947— Moder quantum chemistry : introduction to advanced cleric structure theory / Attila Szabo, Neil S. Ostlund. pom, Previously published: Ist ed., rev. New York + McGraw-Hill, 1989. Includes bibliographical references and index. ISBN 0-486-69186-1 (pbk.) i 1. Quantum chemistry. I. Ostlund, Neil SIL. Title. 1QD462.S95 1996} 541.2’8—de20 96-1075 erp Manufactured in the United States of America Dover Publications, Inc., 31 East 2nd Street, Mineola, N.Y. 11501TABLE OF CONTENTS Preface to Revised Edition Preface Chapter 1. Mathematical Review 1,1 Linear Algebra 1.1.1 Three-Dimensional Vector Algebra 2 1.1.2. Matrices 5 1.1.3. Determinants 7 1.1.4 N-Dimensional Complex Vector Spaces 9 1.1.5 Change of Basis 13 1.1.6 The Eigenvalue Problem /5 1.1.7 Functions of Matrices 2] 1.2. Orthogonal Functions, Eigenfunctions, and Operators 1.3. The Variation Method 1.3.1. The Variation Principle 3/ 1.3.2. The Linear Variational Problem 33 Notes Further Reading Chapter 2. Many Electron Wave Functions and Operators 2.1 The Electronic Problem 2.1.1 Atomic Units 41 2.1.2 The Born-Oppenheimer Approximation 43 2.1.3. The Antisymmetry or Pauli Exclusion Principle 45 2.2 Orbitals, Slater Determinants, and Basis Functions 2.2.1 Spin Orbitals and Spatial Orbitals 46 2.2.2 Hartree Products 47 ix xi 24 31 38 38 39 46 iiiiv MODERN QUANTUM CHEMISTRY 2.3 24 2.2.3 Slater Determinants 49 2.2.4 The Hartree-Fock Approximation 53 2.2.5 The Minimal Basis H, Model 55 2.2.6 Excited Determinants 58 2.2.7 Form of the Exact Wave Function and Configuration Interaction 60 Operators and Matrix Elements 64 2.3.1 Minimal Basis H, Matrix Elements 64 2.3.2 Notations for One- and Two-Electron Integrals 67 2.3.3 General Rules for Matrix Elements 68 2.3.4 Derivation of the Rules for Matrix Elements 74 2.3.5 Transition from Spin Orbnals to Spatial Orbitals 87 2.3.6 Coulomb and Exchange Integrals 85 2.3.7 Pseudo-Classical Interpretation of Determinantal Energies 87 Second Quantization 89 2.4.1 Creation and Annihilation Operators and Their Anticommutation Relations 89 2.4.2 Second-Quantized Operators and Their Matrix Elements 95 2.5 Spin-Adapted Configurations 97 2.5.1 Spin Operators 97 2.5.2 Restricted Determinants and Spin-Adapted Configurations 100 2.5.3 Unrestricted Determinants 1/04 Notes 107 Further Reading 107 Chapter 3. The Hartree-Fock Approximation 108 3.1 The Hartree-Fock Equations lil 3.2 3.3 3.4 3.1.1 The Coulomb and Exchange Operators 1/2 3.1.2. The Fock Operator 1/4 Derivation of the Hartree-Fock Equations 115 3.2.1 Functional Vanation 115 3.2.2. Minimization of the Energy of a Single Determinant 117 3.2.3. The Canonical Hartree-Fock Equations 120 Interpretation of Solutions to the Hartree-Fock Equations 123 3.3.1 Orbital Energies and Koopmans’ Theorem 123 3.3.2 Brillouin’s Theorem 128 3.3.3. The Hartree-Fock Hamiltoman 130 Restricted Closed-Shell Hartree-Fock: The Roothaan Equations 131 3.4.1 Closed-Shell Hartree-Fock: Restricted Spin Orbitals 132 3.4.2 Introduction of a Basis: The Roothaan Equations 136 3.4.3 The Charge Density 138 3.4.4 Expression for the Fock Matrix 140 3.4.5 Orthogonalization of the Basis 1423.5 3.6 3.7 3.8 TABLE OF CONTENTS V 3.4.6 The SCF Procedure 145 3.4.7. Expectation Values and Population Analysis 149 Model Calculations on H, and HeH* 3.5.1 The 1s Minimal STO-3G Basis Set 153 3.5.2 STO3GH, 159 3.5.3. An SCF Calculation on STO-3G HeH* 168 Polyatomic Basis Sets 3.6.1 Contracted Gaussian Functions 180 3.6.2. Minimal Basis Sets: STO-3G 184 3.6.3 Double Zeta Basis Sets: 4-31G 186 3.6.4 Polarized Basis Sets: 6-31G* and 6-31G"* 189 Some Illustrative Closed-Shell Calculations 3.7.1 Total Energies 191 3.7.2 Ionization Potentials 194 3.7.3 Equilibrium Geometries 200 3.7.4 Population Analysis and Dipole Moments 203 Unrestricted Open-Shell Hartree-Fock: The Pople-Nesbet Equations 3.8.1 Open-Shell Hartree Fock: Unrestricted Spin Orbitals 206 3.8.2 Introduction of a Basis: The Pople-Nesbet Equations 2/0 3.8.3 Unrestricted Density Matrices 2/2 3.8.4 Expression for the Fock Matrices 2/4 3.8.5 Solution of the Unrestricted SCF Equations 2/5 3.8.6 Illustrative Unrestricted Calculations 2/6 3.8.7 The Dissociation Problem and its Unrestricted Solution 221 Notes Further Reading Chapter 4. Configuration Interaction 41 42 43 44 45 46 Multiconfigurational Wave Functions and the Structure of the Full CI Matrix 4.1.1 Intermediate Normalization and an Expression for the Correlation Energy 237 Doubly Excited CI Some Illustrative Calculations Natural Orbitals and the One-Particle Reduced Density Matrix The Multiconfiguration Self-Consistent Field (MCSCF) and Generalized Valence Bond (GVB) Methods Truncated CI and the Size-Consistency Problem 152 180 190 205 229 229 231 233 242 245 252 258 261Vi MODERN QUANTUM CHEMISTRY Notes Further Reading Chapter 5. Pair and Coupled-Pair Theories 5.1 The Independent Electron Pair Approxumation (IEPA) 5.1.1 Invariance under Unitary Transformations: An Example 277 512 Some Illustrative Calculations 284 5.2 Coupled-Pair Theories 5.2.1. The Coupled Cluster Approximation (CCA) 287 5.2.2. The Cluster Expansion of the Wave Function 290 5.2.3 Linear CCA and the Coupled Electron Pair Approximation (CEPA) 292 5.2.4 Some Illustrative Calculations 296 5.3. Many-Electron Theories with Single Particle Hamiltonians 5.3.1 The Relaxation Energy via Cl. IEPA. CCA, and CEPA 303 5.3.2 The Resonance Energy of Polyenes in Hiickel Theory 309 Notes Further Reading Chapter 6. Many-Body Perturbation Theory 6.1 Rayleigh-Schrédinger (RS) Perturbation Theory *62 Diagrammatic Representation of RS Perturbation Theory 6.2.1 Diagrammatic Perturbation Theory for 2 States 327 62.2 Diagrammatic Perturbation Theory for N States 335 6.23 Summation of Diagrams 336 6.3 Orbital Perturbation Theory: One-Particle Perturbations *6.4 Diagrammatic Representation of Orbital Perturbation Theory 6.5 Perturbation Expansion of the Correlation Energy 6.6 The N-Dependence of the RS Perturbation Expansion *6.7 Diagrammatic Representation of the Perturbation Expansion of the Correlation Energy 671 Hugenholtz Diagrams 356 672 Goldstone Diagrams 362 6.73 Summation of Diagrams 368 67.4 What Is the Linked Cluster Theorem? 369 6.8 Some Illustrative Calculations Notes Further Reading 269 269 271 272 286 297 318 319 320 322 327 338 348 350 354 356 370 378 379‘TABLE OF CONTENTS Vii Chapter 7. The One-Particle Many-Body Green’s Function 7.1 Green’s Functions in Single Particle Systems 7.2 The One-Particle Many-Body Green’s Function 7.21 The Self-Energy 389 7.2.2. The Solution of the Dyson Equation 391 7.3. Application of the Formalism to H, and HeH* 7.4 Perturbation Theory and the Green’s Function Method 7.5 Some Illustrative Calculations Notes Further Reading Appendix A, Integral Evaluation with 1s Primitive Gaussians Appendix B. Two-Electron Self-Consistent-Field Program Appendix C. Analytic Derivative Methods and Geometry Optimization by M.C. Zerner Appendix D. Molecular Integrals for H, as a Function of Bond Length Index 380 381 387 392 398 405 410 Al7 437 459 461Szeretettel a sziileimnek AS To my wonderful father and the memory of my dear mother NSOPREFACE TO REVISED EDITION This revised edition differs from its predecessor essentially in three ways. First, we have included an appendix describing the important recent de- velopments that have made the efficient generation of equilibrium geo- metries almost routine. We are fortunate that M. Zerner agreed to write this since our own recent interests have been channeled in other directions. Second, numerous minor but annoying errors have been corrected. For most of these we are indebted to K. Ohno, T. Sakai and Y. Mochizuki who detected them in the course of preparing the Japanese translation which has recently been published by Tokyo University Press. Finally, we have updated the Further Reading sections of all the chapters. We are extremely pleased by the many favorable comments we have received about our book and we hope that the next generation of readers will find this edition useful. ATTILA SZABO NEIL S. OsTLUND.PREFACE The aim of this graduate textbook is to present and explain, at other than a superficial level, modern ab initio approaches to the calculation of the electronic structure and properties of molecules. The first three chapters contain introductory material culminating in a thorough discussion of the Hartree-Fock approximation, The remaining four chapters describe a variety of more sophisticated approaches, which improve upon this approximation. We have strived to make advanced ‘opics understandable to the be- ginning graduate student. Our goal was to teach more than cocktail party jargon; we wanted to give insight into the nature and validity of a variety of approximate formalisms and improve the reader’s problem-solving skills. Needless to say, a book of this size cannot cover all of quantum chemistry; it would be easy to write down a long list of important topics which we have not covered. Nevertheless, we believe that anyone who has mastered the material in this book will have a solid foundation to build upon, and will be able to appreciate most research papers and seminars dealing with electronic structure theory and its applications. The origins of this book go back to our student days when we were trying to understand a variety of quantum chemical formalisms that at first glance appeared to be extremely complicated. We found that by applying such formalisms to simple, analytically tractable, model systems we not only got a feeling as to what was involved in performing actual calculations but also gained insight into the interrelationships among different approximation schemes. The models we used then, and ones we devised or learned sub- sequently, play an important role in the pedagogical approach we adopt in this book. The writing began in 1976 when we jointly taught a special topics course at Indiana University. The manuscript gradually evolved in response to the needs and reactions of students enrolled in a second-semester quantum xixii MODERN QUANTUM CHEMISTRY chemistry course we taught some five times at both Indiana University and the University of Arkansas. Animportant feature of this book is that over 150 exercises are embedded in the body of the text. These problems were designed to help the reader acquire a working knowledge of the material. The level of difficulty has been kept reasonably constant by breaking up the more complicated ones into manageable parts. Much of the value of this book will be missed if the exercises are ignored. In the following, we briefly describe some of the highlights of the seven chapters. Chapter 1 reviews the mathematics (mostly matrix algebra) re- quired for the rest of the book. It is self-contained and suited for self-study. Its presence in this book is dictated by the deficiency in mathematics of most chemistry graduate students. The pedagogical strategy we use here, and in much of the book, is to begin with a simple example that illustrates most of the essential ideas and then gradually generalize the formalism to handle more complicated situations. Chapter 2 introduces the basic techniques, ideas, and notations of quantum chemistry. A preview of Hartree-Fock theory and configuration interaction is used to motivate the study of Slater determinants and the evaluation of matrix elements between such determinants. A simple model system (minimal basis H.) is introduced to illustrate the development. This model and its many-body generalization (N independent H, molecules) reappear in all subsequent chapters to illuminate the formalism. Although not essential for the comprehension of the rest of the book, we also present here a self-contained discussion of second quantization. Chapter 3 contains a thorough discussion of the Hartree-Fock approx- imation. A unique feature of this chapter is a detailed illustration of the computational aspects of the self-consistent-field procedure for minimal basis HeH*. Using the output of a simple computer program, listed in Appendix B, the reader is led iteration-by-iteration through an ab initio calculation. This chapter also describes the results of Hartree-Fock calcula- tions on a standard set of simple molecules using basis sets of increasing sophistication. We performed most of these calculations ourselves, and in later chapters we use these same molecules and basis sets to show how the Hartree-Fock results are altered when more sophisticated approaches are used. Thus we illustrate the influence of both the quality of the one-electron basis set and the sophistication of the quantum chemical method on cal- culated results. In this way we hope to give the reader a feeling for the kind of accuracy that can be expected from a particular calculation. Chapter 4 discusses configuration interaction (CI) and is the first of the four chapters that deal with approaches incorporating electron correlation. One-electron density matrices, natural orbitals. the multiconfiguration self- consistent-field approximation, and the generalized valence bond method arePREFACE xiii discussed from an elementary point of view. The size-consistency problem associated with truncated CI is illustrated using a model consisting of N independent hydrogen molecules, This example highlights the need for so-called many-body approaches, which do not suffer from this deficiency, that are discussed in subsequent chapters. Chapter 5 describes the independent electron pair approximation and a variety of more sophisticated approaches that incorporate coupling be- tween pairs. Since this chapter contains some of the most advanced material in the book, many illustrative examples are included. In the second half of the chapter, as a pedagogical device, we consider the application of many- electron approaches to an N-electron system described by a Hamiltonian containing only single particle interactions. This problem can be solved exactly in an elementary way. However, by seeing how “high-powered” approaches work in such a simple context. the student can gain insight into the nature of these approximations. Chapter 6 considers the perturbative approach to the calculation of the correlation energy of many-electron systems. A novel pedagogical approach allows the reader to acquire a working knowledge of diagrammatic pertur- bation theory surprisingly quickly. Although the chapter is organized so that the sections on diagrams (which are starred) can be skipped without loss of continuity, we find that the diagrammatic approach is fun to teach and is extremely well received by students. Chapter 7 contains a brief introduction to the use of the one-particle many-body Green’s function in quantum chemistry. Our scope is restricted to discussing ionization potentials and electron affinities. The chapter is directed towards a reader having no knowledge of second quantization or Green's functions, even in a simple context. This book is largely self-contained and, in principle, requires no pre- requisite other than a solid undergraduate physical chemistry course. How- ever, exposure to quantum chemistry at the level of the text by I. N. Levine (Quantum Chemistry, Allyn and Bacon) will definitely enhance the student's appreciation of the subject material. We would normally expect the present text to be used for the second semester of a two-semester sequence on quan- tum chemistry. It is also suitable for a special topics course. There is probably too much material in the book to be taught in-depth in a single semester. For students with average preparation, we suggest covering the first four chapters and then discussing any one of the last three, which are essentially independent. Our preferred choice is Chapter 6. For an exceptionally well- prepared class, the major fraction of the semester could be spent on the last four chapters. We have found that a course based on this text can be enriched in a number of ways. For example, it is extremely helpful for students to perform their own numerical calculations using, say, the Gaussian 80 system of programs. In addition, recent papers on the applications ofxiv, MODERN QUANTUM CHEMISTRY electronic structure theory can be assigned at the beginning of the course and the students asked to give short in-class presentations on one or more of such papers at the end-of the course. We have placed special emphasis on using a consistent notation through- out the book, Since quantum chemists use a number of diflerent notations, it is appropriate to define the notation we have adopted. Spatial molecular orbitals (with latin indices i,j,k...) are denoted by w. These are usually expanded in a set of spatial (atomic) basis functions (with greek indices Lv, A...) denoted by ¢. Molecular spin orbitals are denoted by y. Occupied molecular orbitals are specifically labeled by a, b, c,... and unoccupied (virtual) molecular orbitals are specifically labeled by r, s, t,... Many- electron operators are denoted by capital script letters (for example, the Hamiltonian is .#), and one-electron operators are denoted by lower case latin letters (for example, the Fock operator for electron-one is f(1)). The exact many-electron wave function is denoted by ®, and we use ’ to denote approximate many-electron wave functions (i.e. the Hartree-Fock ground state wave function is Yo, while ‘7 is a doubly excited wave function). Exact and approximate energies are denoted by & and E, respectively. All numerical quantities (energies, dipole moments, etc.) are given in atomic units. It is a pleasure to thank those people who helped us with scientific aspects of this book. J, A. Pople has always been a source of inspiration. M. Karplus and W. Reinhardt introduced us to many-body theory. J.-P. Malrieu and K. Schulten independently showed us in the early 1970s how to illustrate the lack of size consistency of doubly excited configuration interaction using the model of N noninteracting H, molecules. This led to our extensive use of this model. M. Zerner constructively criticized the entire manuscript. The following colleagues gave advice, performed cal- culations just for the book, or assisted us in some other useful way: R. Bartlett, J. Binkley, M. Bowen, R. Counts, C. Dykstra, W. Eaton, D. Freeman, W. Goddard II. S. Hagstrom, A. Hernandez, D. Merrifield, J. Neece, I, Shavitt, and R. Whiteside. The following students helped us clarify our presentation and eliminate errors: B. Basinger. G. Caldwell, T. Croxton, R. Farren, R. Feeney, M. Flanagan, J. Freeze. V. Hess. K. Holtzclaw, J. Joens, J. Johnson, R. Jones, J. Kehayias, G. Lipari, D. Lupo, D, McMullen, J. Meek, S. Munchak, N. Peyghambarian, R. Russo, W. Schinzer, D. Shoup, B. Stone, E. Tsang and I. Waight. Of all the secretaries who typed various versions of the manuscript, K. Wagner, M. Mabie, and L. Ferro were the most helpful. J. Hawkins prepared the illustrations. ArriLa SZABO New. S. OstLuNDCHAPTER ONE MATHEMATICAL REVIEW This chapter provides the necessary mathematical background for the rest of the book. The most important mathematical tool used in quantum chem- istry is matrix algebra. We have directed this chapter towards the reader who has some familiarity with matrices but who has not used them in some time and is anxious to acquire a working knowledge of linear algebra. Those with strong mathematical backgrounds can merely skim tke material to acquaint themselves with the various notations we use. Our development is informal and rigour is sacrificed for the sake of simplicity. To help the reader develop those often neglected, but important, manipulative skills we have included carefully selected exercises within the body of the text. The material cannot be mastered without doing these simple problems. Tn Section 1.1 we present the elements of linear algebra by gradually generalizing the ideas encountered in three-dimensional vector algebra. We consider matrices, determinants. linear operators and their matrix repre- sentations, and, most importantly, how to find the eigenvalues and eigen- vectors of certain matrices. We introduce the very clever notation of Dirac, which expresses our results in a concise and elegant manner. This notation is extremely useful because it allows one to manipulate matrices and derive various theorems painlessly. Moreover, it highlights similarities between linear algebra and the theory of complete sets of orthonormal functions as will be seen in Section 1.2. Finally, in Section 1.3 we consider one of the cornerstones of quantum chemistry namely, the variation principle.2 MODERN QUANTUM CHEMISTRY 1.1 LINEAR ALGEBRA We begin our discussion of linear algebra by reviewing three-dimensional vector algebra. The pedagogical strategy we use here, and in much of the book, 1s to start with the simplest example that iilustrates the essential ideas and then gradually generalize the formalism to handle more complicated situations. 1.1.1 Three-Dimensional Vector Algebra A three-dimensional vector can be represented by specifying 1ts components a,,1= 1, 2,3 with respect to a set of three mutually perpendicular unit vectors {é,} as a = bya, + bay + é303 = Dea, (Ll) The vectors é; are said to form a basis, and are called basis vectors. The basis 1s complete m the sense that any three-dimensional vector can be written as a linear combination of the basis vectors. However, a basis is not unique, we could have chosen three different mutually perpendicular unit vectors, é, 1 = 1, 2, 3 and represented d as G= bya, + ba, + bya3 =) Aa (1.2) Given a basis, a vector is completely specified by its three components with respect to that basis. Thus we can represent the vector @ by a column matrix as ay in the basis {2,} (13a) oras ay a =|a,)_ inthe basis {é} (1.3b) a3 The scalar or dot product of two vectors a and b is defined as @ b= ayby + ayb, + ayb3 = ¥ ab, (L4) Note that @a@=a+a3ta ? (1.5) is simply the square of the length (al) of the vector d. Let us evaluate the scalar product & - b using Eq. (1 1) a-b=y) V8 -éab, (1.6) a7MATHEMATICAL REVIEW 3 For this to be identical to the definition of (1.4), we must have 8 =by=3p= 4 where we have introduced the Kronecker delta symbol 6,, This relation is the mathematical way of saying that the basis vectors are mutually per- pendicular (orthogonal) and have unit length (normalized), m other words they are orthonormal. Given a vector a. we can find its component along @, by taking the scalar product of Eq. (1 1) with é, and using the orthonormality relation (1.7) 1 ifi=j 0 otherwise 1.7) @,-a=Y@,-8a,=) 6,4, =a, (1.8) Hence we can rewrite Eq. (1.1) as @=D%é,-a=1-4 (19) where T=) 2g, (1.10) 1s the unit dyadic. A dyadic 1s an entity which when dotted mto a vector gives another vector. The unit dyadic gives the same vector back. Equation (1.10) is called the completeness relation for the basis {é,} since it is an alternate form of Eq. (1.1), which states that any vector @ can be written as a linear combination of the basis vectors {2,}. We now define an operator @ as an entity which when acting on a vector 4 converts it into a vector b Oa=5 (11l) The operator is satd to be linear if for any numbers x and 5 O(xa + yb) = xOG + yOb (1.12) A linear operator 1s completely determined if its effect on every possible vector is known. Because any vector can be expressed in terms of the basis. {2,!, at 1s sufficient to know what @ does to the basis vectors. Now since a 1s a vector, it can be written as a linear combination of the basis vectors éJ. ie, 3 C%= ¥ G0, 1=1,2,3 (1.13) The number O,; is the component of the vector @é; along é;. The nine numbers O,,, 1.) = 1, 2, 3 can be arranged in a two-dimensional array called a matrix as Or, O12 Ors O=|0,,; O02 O23 (1.14) O31 O32 O33,4 MODERN QUANTUM CHEMISTRY We say that O is the matrix representation of the operator @ in the basis {é}. The matrix O completely specifies how the operator © acts on an arbitrary vector since this vector can be expressed as a linear combination of the basis vectors {2,} and we know what © does to each of these basis vectors. Exercise 1.1 a) Show that 0, = b= 0,4, -@2,. b) If @i=6 show that If A and B are the matrix representations of the operators «/ and &, the matrix representation of the operator @, which is the product of «/ and B(@ = AB), can be found as follows: 6, =T2c, 7 = SRE, = 7S eBy § = edaBy (1.15) 4 so that Cy=¥ AuBuy (1.16) rs which is the definition of matrix multiplication and hence C=AB (1.17) Thus if we define matrix multiplication by (1.16), then the matrix repre- sentation of the product of two operators is just the product of their matrix representations. The order in which two operators or two matrices are multiplied is crucial. In general of # Bof or AB #BA. That is, two operators or matrices do not necessarily commute. For future reference, we define the commutator of two operators or matrices as [o, B] = AB - Boil (1.18a) [A, B] = AB- BA (1.18b) and their anticommutator as {of B} = AB + BA (1.194) {A, B} = AB+ BA (1.19b)MATHEMATICAL REVIEW 5 Exercise 1.2 Calculate [A. B] and (A. B} when 1 1 0 1-1 1 1 2 2 B={|-1 0 0 Oo 2 -1 1 0 1 1.1.2 Matrices Now that we have seen how 3 x 3 matrices naturally arise in three-dimen- sional vector algebra and how they are multiplied, we shall generalize these results. A set of numbers {A,;} that are i general complex and have ordered subscripts i= 1,2,...,N andj =1,2,..., M can be considered elements of a rectangular (N x M) matrix A with N rows and M columns Ay Aya Aue Aa{ 42 422 0 Anse (1.20) Ant Av. 77 Anw If N = M the matrix is square. When the number of columns in the N x M matrix A is the same as the number of rows in the M x P matrix B, then A and B can be multiplied to give a N x P matrix C C=AB (1.21) where the elements of C are given by the matrix multiplication rule _% i=1,...,N Pp (1.22) The set of M numbers {a,} i= 1, 2,..., M can similarly be considered elements of a column matrix (ay a=. (1.23) Note that foran N x M matrix A Aa=b (1.24) where b is a column matrix with N elements M b= Aya, 1=1,2,...,N (1.25) aa6 MODERN QUANTUM CHEMISTRY We now introduce some important definitions. The adjoint of an N x M matrix A, denoted by At, is an M x N matrix with elements (At), = Ah (1.26) ie. we take the complex conjugate of each of the matrix elements of A and interchange rows and columns. If the elements of A are real, then the adjoint of A is called the transpose of A. Exercise 1.3 If A is an N x M matrix and B is a M x K matrix show that (AB)' = BtAt. The adjoint of a column matrix is a row matrix containing the complex conjugates of the elements of the column matrix. at = (ata¥- --at) (1.27) Ifa and b are two column matrices with M elements, then by b. “M ath = (afag-+-at)| 2? |= Yat (1.28) i imt Da Note that if the elements of a and b are real and M is 3, this is simply the scalar product of two vectors, cf. Eq, (1.4). Taking the adjoint of Eq. (1.24) and using the result of Exercise 1.3 we have bi =atat (1.29) where bt is a row matrix with the N elements M M * b= an, =(5 aA) i=1,2,...,N (130) at Jat Note that Eq. (1.30) is simply the complex conjugate of Eq. (1.25). We now give certain definitions and properties applicable only to square matrices: 1. A matrix A is diagonal if all its off-diagonal elements are zero Ay = Aubiy- (1.31) 2. The trace of the matrix A is the sum of its diagonal elements trA=) A,. (1.32) 7 3. The unit matrix is defined by IA=AL=A (1.33)MATHEMATICAL REVIEW 7 for any matrix A, and has elements (Dy = 5,- (1.34) 4. The inverse of the matrix A, denoted by A~', is a matrix such that A'A=AA'=1 (1.35) 5. A unitary matrix A is one whose inverse is its adjoint A“l=At. (1.36) A real unitary matrix is called orthogonal. 6. A Hermitian matrix is self-adjoint, ie., AtT=A (1.37a) or Ak = Ay (1.37b) A real Hermitian matrix is called symmetric. Exercise 1.4 Show that b. (AB)~ . c. Jf U is unitary and B= UtAU, then A = UBUt. d. If the product C = AB of two Hermitian matrices is also Hermitian, then A and B commute, e. If A is Hermitian then A~’, if it exists, js also Hermitian. Au an) “1 1 ( An “42) £IfA= |, then A>? = ———_—________ . (a An (Ay1A22— Ar2Aar)\-Anr Ata 1.1.3 Determinants We now define and give some properties of the determinant of a square matrix A. Recall that a permutation of the numbers 1, 2, 3,..., N is simply a way of ordering these numbers, and there are N! distinct permutations of N numbers. The determinant of an N x N matrix A is a number obtained by the prescription Ay 07> Aw det(A) = |A| = : ™ = D(H Y"PAy Ag Ayy (1.38) Any Ann)" where Y, is a permutation operator that permutes the column indices 1, 2, 3,..., N and the sum runs over all N! permutations of the indices; p;8 MODERN QUANTUM CHEMISTRY is the number of transpositions required to restore a given permutation iy, b,..., iy to natural order 1, 2, 3,..., N. Note that it is important only whether p, is an even or odd number. As an illustration we evaluate the determinant of a 2 x 2 matrix A. An 4") A= (2 There are two permutations of the column indices 1 and 2, i.e. 12(p,=0) 21(p2=1) hence |A A TL = (WA aa + (=D! Asada = Agi Az2 ~ Arrday (1.39) Arr Aad] Some important properties of determinants which follow from the above definition are: 1. If each element in a row or in a column is zero the value of the deter- minant is zero, 2. Tf(A)jj = Audip then [A] = [] Aj = 411422 --- Avw. 3. A single interchange of any two rows (or columns) of a determinant changes its sign. 4. [Al = ((At))* 5. |AB| = |4||BI. Exercise 1.5 Verify the above properties for 2 x 2 determinants. Exercise 1.6 Using properties (1)-(5) prove that in general 6. Ifany two rows (or columns) of a determinant are equal, the value of the determinant is zero. (A). 1, then |Aj({A)* = 1. 9. If UtOU = Q and UTU = UUt = I, then [O] = |Q|. Exercise 1.7 Using Eq. (1.39), note that the inverse of a 2 x 2 matrix A obtained in Exercise L.4f can be written as tae ( An “4") Al \-4n AasMATHEMATICAL REVIEW 9 and thus A~' does not exist when |A| = 0. This result holds in general for N x N matrices. Show that the equation Ac=0 where A is an N x N matrix and ¢ is a column matrix with elements c,, 1=1,2,..., N can have a nontrivial solution (c 4 0) only when |A| = 0. For a2 x 2 determinant it is easy to verify by direct calculation that evBis + ¢Bi2 Audl _ fr Av, [Biz Ang] e1Ba1 + ¢2B22 Anal" |Bay Aaa Br2 Aza] This result is a special case of the following property of determinants that we shall use several times in the book. en M Ay Ai 0" Y Bu vt Aww cS M Az Ara 7 Y GBax > Aaw m1 M Ax Ava 7° YS GBye 7° Ann = Au Asa Bu Asyy Az, Are Bae Aaw (1.40) « =Sal Ay Ava *7° Bye °° Ayn A similar result holds for rows. 1.1.4 N-Dimensional Complex Vector Spaces We need to generalize the ideas of three-dimensional vector algebra to an N-dimensional space in which the vectors can be complex. We will use the powerful notation introduced by Dirac, which expresses our results in an exceedingly concise and simple manner. In analogy to the basis {é,} in three dimensions, we consider N basis vectors denoted by the symbol |i), i = 1, 2,..., N, which are called ket vectors or simply kets. We assume this basis is complete so that any ket vector |a) can be written as >= ¥ joa (141) = This is a simple generalization of Eq. (1.1) rewritten in our new notation.10 MODERN QUANTUM CHEMISTRY After we specify a basis, we can completely describe our vector |a) by giving its N components a;, i= 1, 2,..., N with respect to the basis {|i>}. Just as before, we arrange these numbers in a column matrix a as (1.42) ay, and we say that a is the matrix representation of the abstract vector |a) in the basis {|i}. Recall (Eq. (1.27)) that the adjoint of the column matrix a is the row matrix at at = (afay «ay (1.43) Now we introduce an abstract bra vector = ) were chosen because the notation for the scalar product (¢ | >) looks like a bra-c-ket. Note that on the right and obtain CIO =D Da =L dua 4, (1.48a) and = akKil i> = ¥ af, = af (1.486) The expression “multiplying by = al 7>)* = * (1.49) Using these results we can rewrite Eqs. (1.41) and (1.46) as =F ha =F lodlay (1.50a) and [d (1.52) As before, the operator is completely determined if we know what it does to the basis {|i)}: Oi = TINO = LINO) w (1.53) 7 i so that O is the matrix representation of the operator @ in the basis {[i)}. Multiplying (1.53) on the left by (1.55) 7 which upon comparison to Eq. (1.53) yields =O), = On (1.56) As another illustration of the use of the completeness relation and the built-in consistency and simplicity of Dirac notation, let us find the matrix repre- sentation of the operator @ = «7 in terms of the matrix representations of the operators of and # (c.f. Eq. (1.15)) CIE = (Oy, = Gi Ali = =D AjB), 5 We now introduce the adjoint of the operator @, which we denote by ©". If © changes a ket |a) into the ket |b> (cf. Eq. (1.52), then its adjoint changes the bra a] into the bra on the right, we have = = Since = * = OF (1.59) Finally, we say that an operator is Hermitian when it is self-adjoint o=0 (1.60) Thus the elements of the matrix representation of a Hermitian operator satisfy = = } and {[o>} we now wish to find the relationship between them. We use latin letters i, j, k, . .. to specify the bras and kets in the first basis and greek letters a, 8,» to specify the bras and kets of the second basis. Thus we have GP=6, LD<=1 (1.62a) and a] B>= 5. YL faa] = 1 (1.62b) Since the basis {|i} is complete, we can express any ket in the basis {|} as a linear combination of kets in the basis {|i>} and vice versa. That is, a> = tay = YL [D = LDU = LDOa (1.63) where we have defined the elements of a transformation matrix U as = # U,, but rather is given by Eq. (1.66). We now prove that the trans- formation matrix U is unitary. This is a consequence of the orthonormality of the bases: 6, = = 2 = Oil0,; =(UU'), which in matrix notation is just 1=UUt (1.67a) In an analogous way, by starting with }, while Q is its matrix representation in the basis {|a)} Oli) = LID = x [B>Qp. (1.68b) To find the relationship between O and Q we use the, by now, familiar technique of introducing the unit operator Quy = = Y < 7] B> a =L UO (Op (1.69) a Thus Q=U'0U (1.70a) or. multiplying on the left by U and on the right by Ut o=voeut (1.70b) These equations show that the matrices O and Q are related by a unitary transformation. The importance of such transformations lies in the fact that for any Hermitian operator whose matrix representation in the basis {|i)} is not diagonal, it is always possible to find a basis {|} in which the matrix representation of the operator is diagonal, ic., Qap = Oban (1.71) In the next subsection we consider the problem of diagonalizing Hermitian matrices by unitary transformations. Exercise 1.8 Show that the trace of a matrix is invariant under a unitary transformation, ie., if Q=Ut0OU then show that trQ = trO.MATHEMATICAL REVIEW 15 1.1.6 The Eigenvalue Problem When an operator (@ acts on a vector |e, the resulting vector is in general distinct from |). If Ola) is simply a constant times Jay, Le, O)a> = w]e) (1.72) then we say that |«> is an eigenvector of the operator @ with an eigenvalue «,. Without loss of generality we can choose the eigenvectors to be normalized =1 (1.73) In this book we are interested in the eigenvectors and eigenvalues of Hermi- tian operators (© = ©). They have the following properties. 1. The eigenvalues of a Hermitian operator are real. This follows imme- diately from Eq. (1.61), which states that Calla» = =0 77) so that = 0 if @, # wp. Thus orthogonality follows immediately if the two eigenvalucs are not the same (nondegencrate). Two cigenvectors ID and |2> are degenerate if they have the same cigenvalue Oly =a1> —O)2) = @[2> (1.78)16 MODERN QUANTUM CHEMISTRY We now show that degenerate eigenvectors can always be chosen to be orthogonal. We first note that any linear combination of degenerate eigen- vectors is also an eigenvector with the same eigenvalue, ic., (x1) + y[2>) = XC|1> + y€|2> = @X|L> + yf29) (1.79) There are many ways we can find two linear combinations of |1) and |2), which are orthogonal. One such procedure is called Schmidt orthogonali- zation. We assume that |1) and |2) are normalized and let (1|2) = S # 0. We choose |I} = |1) so that ((Ly = S-1]2>) (1.80) Thus the eigenvectors {|x)} of a Hermitian operator can be chosen to form an orthonormal set } is generally not diagonal. However, its matrix representation in the basis formed by its eigenvectors is diagonal. To show this we multiply the eigenvalue equation (1.72) by = bop (1.82) The eigenvalue problem we wish to solve can be posed as follows. Given the matrix representation, O, of a Hermitian operator @ in the orthonormal basis {|/>, i= 1,2,...,.N} we wish to find the orthonormal basis {|)>, a=1,2,...,.N} in which the matrix representation, Q, of € is diagonal, i p = 254g. In other words, we wish to diagonalize the matrix O. We have seen in the last subsection, that the two representations of the operator © are related by a unitary transformation (c.f. Eq. (1.70a)) Q=U'OU Thus the problem of diagonalizing the Hermitian matrix O is equivalent to the problem of finding the unitary matrix U that converts O into a diagonal matrix oy a, 0 U'OU =a = (1.83) Wy, It is clear from this formulation that an N x N Hermitian matrix has N eigenvalues.MATHEMATICAL REVIEW 17 There exist numerous, efficient algorithms for diagonalizing Hermitian matrices.’ For our purposes, computer programs based on such algorithms can be regarded as “black boxes,” which when given O determine U and o. In order to make contact with the discussion of the eigenvalue problem found in most elementary quantum chemistry texts, we now consider a computationally inefficient procedure that is based on finding the roots of the secular determinant. The eigenvalue problem posed above can be reformulated as follows. Given an N x N Hermitian matrix O, we wish to find all distinct column vectors ¢ (the eigenvectors of O) and corresponding numbers « (the eigen- values of O) such that Oc = we (1.84a) This equation can be rewritten as (O—alje=0 (1.84b) As shown in Exercise 1.7, Eq. (1.84b) can have a nontrivial solution (c ¥ 0) only when |o-al|=0 (1.85) This is called the secular determinant. This determinant is a polynomial of degree N in the unknown w. A polynomial of degree N has N roots @,, a =1,2,...,.N, which in this case are called the eigenvalues of the matrix O. Once we have found the eigenvalues, we can find the corresponding eigen- vectors by substituting each w, into Eq. (1.84) and solving the resultmg equations for c*. In this way, ¢* can be found to within a multiplicative constant, which is finally determined by requiring ¢* to be normalized (1.86) In this way we can find N solutions to Eq. (1.84) =e" a=1,2,...,N (187) Since O is Hermitian, the eigenvalues are real and the eigenvectors are orthogonal Leet = bap (1.88) In order to establish the connection with our previous development, let us now construct a matrix U defined as U,, = cj, Le., et ct a o ° od = (cle? «--e) (1.89) Ch oH ch18 MODERN QUANTUM CHEMISTRY Thus the eth column of U is just the column matrix ¢*. Then using (1.87) it can be shown that foo, OU =U =Ue (1.90) Since U,, = c?, the orthonormality relation (1.88) is equivalent to ¥Y ULV ig = YL U)ADig = 6,9 (L.91) which in matrix notation is U'U=1 (1.92) Finally, multiplying both sides of Eq. (1.90) by Ut and using Eq. (1.92) we have U'OU=a (1.93) which is identical to Eq. (1.83). Thus Eq. (1.89) gives the relationship between the unitary transformation (U), which diagonalizes the matrix O, and the eigenvectors (c*) of O. Exercise 1.9 Show that Eq. (1.90) contains Eq. (1.87) forall a=1,2,...,.N. As an illustration of the above formalism, we consider the problem of finding the eigenvalues and eigenvectors of the 2 x 2 symmetric matrix (012 = 03) On. O o- (28 o2) (on On, or, equivalently, solving the eigenvalue problem On O2\(er ce = 1.94) (cn on )(e) (2) me We shall solve this problem in two ways: first via the secular determinant (Eq. (1.85)) and second by directly finding the matrix U that diagonalizes O. For Eq. (1.94) to havea nontrivial solution, the secular determinant must vanish On~o Ow oO. ? — (O22 + O11) + O10 72 ~ 042021 = 0 21 (1.95)MATHEMATICAL REVIEW 19 This quadratic equation has two solutions 1 = MO + O22 — (O22 — O11)? + 40, 2021)"7] (1.96a) 2 = $[041 + O22 + (O22 — O11)? + 4042021)" (1.96b) which are the eigenvalues of the matrix O. To find the eigenvector corre- sponding to a given eigenvalue, say w, we substitute 2 into Eg. (1.94) to obtain Onc} + Oy2c3 = we} (1.97) 02,C32 + O22¢3 = «23 (1.97b) where the superscripts “2” indicate that we are concerned with the second eigenvalue. We then use one of these two equivalent equations and the normalization condition (ci? + (GP =1 (1.98) to solve for c? and c3. As a simple illustration. we consider the case when O41; = O22 = aand 0,, = O2, = b. From Eq. (1.96). the two eigenvalues are @,=a-b (1.99a) w,=atb (1.99b) To find the eigenvector corresponding to 2 we use Eq. (1.97a), which in this case gives ac? + be} = (a + b)c? from which we obtain hah Finally, the normalization condition (1.98) gives dare gare (1.100a) In an entirely analogous way, we find ch=272 ch = 2712 (1.100b) Exercise 1.10 Since the components of an eigenvector can be found from the eigenvalue equation only to within a multiplicative constant, which is later determined by the normalization, one can set c, = 1 and c,=c in Eq. (1.94). If this is done, Eq. (1.94) becomes 01, + On.c=0 02, + O22¢ = we20 MODERN QUANTUM CHEMISTRY After climinating c, find the two roots of the resulting equation and show that they are the same as those given in Eq. (1.96), This technique, which we shall use numerous times in the book for finding the lowest eigenvalue of a matrix, is basically the secular determinant approach without determinants. Thus one can use it to find the lowest eigenvalue of certain N x N matrices without having to evaluate an N x N determinant. Now let us solve the 2 x 2 eigenvalue problem by directly finding the orthogonal matrix U that diagonalizes the symmetric matrix O, ie. U U. 0. Oo, U U @, 0 vou=(Ye &)(u »)( 0 vag A(% 2) uot Uy2 Ur2)\O.2 O22)\U2r U2 0 w ‘ ) The requirement that Uy1U,, + UU. U4, U 42 + UU: 1 0 wu= (Unt t¥nvar Yale t Una) ) use UU + Uy Ui2 + Una o x) Oi) places three constraints (two diagonal and one off-diagonal) on the four elements of the matrix U. Therefore U can be completely specified by only one parameter. Since cos sin8\/cos8 — sin 8\ _[cos? 9+sin® 8 0 1 sin@ —cos@/\sin@ —cos 0) — 0 cos? 6+sin? 0) (1.103) for all values of the parameter @, it is convenient to write cos@ sin 8 v -(35 —cos 0) (1104 This is the most general form of a 2 x 2 orthogonal matrix. Let us now choose @ such that urou = (9 — sin@\ (O11 O12) (cos sind sin6 —cos6J\O,2 O22)\sin@ —cos@, (0, cos? 6 + Oz, sin? 6 4(0,, — O22) sin 26 — 0,2 cos 26 _ + 0,2 s8in20 ~ |4(0,,; — O22) sin20 — 0,2 cos20 0, , sin? @ + O22 cos? 6 — 0,2 sin26 is diagonal. This can be done if we choose @ such that 4011 — O22) sin 20 — 0,2 cos26 = 0 This has the solution I @) == tan! 2012 — 1.105; 2" OO (4.109)MATHEMATICAL REVIEW 21 Thus the two eigenvalues of O are (©, = 0,1 C08? 85 + O22 Sin? Oy + O12 si 20 (1.106a) and (2, = 0, Sin? Oy + O22 cos? Oy — 01 Sin20, (1.106b) Upon comparison of Eqs. (1.104) and (1.89), we find the two eigenvectors to be ct) _ (00800 1107 ch} \sin€y (1.1072) and a . ch sin 0 = -107b) (3) (re) (1.1076) It should be mentioned that the Jacobi method for diagonalizing N x N matrices i§ a generalization of the above procedure. The basic idea of this method is to eliminate iteratively the off-diagonal elements of a matrix by repeated applications of orthogonal transformations, such as the ones we have considered here. Exercise 1.11 Consider the matrices 301 A= (:) 301 s-(c 3) Find numerical values for the eigenvalues and corresponding eigenvectors of these matrices by a) the secular determinant approach: b) the unitary trans- formation approach. You will see that approach (b) is much eas' 1.1.7 Functions of Matrices Given a Hermitian matrix A, we can define a function of A.i.c., f(A), in much the same way we define functions f(x) of a simple variable x. For example, the square root of a matrix A, which we denote by A", is simply that matrix which when multiplied by itself gives A, i.e. AMAA (1.108) The sine or the exponential of a matrix are defined by the Taylor series of the function, e.g., 1 1 2 1 3 wee exp(A)= 14+ A +5 A? +a AS +22 MODERN QUANTUM CHEMISTRY or in general f(A) = ¥ cA" (1.109) no After these definitions, we are still faced with the problem of calculating A’? or exp (A). If A is a diagonal matrix everything is simple, since (Ay = : (1.110) so that % " ee toa 8 n= 0 Yeak (flay) = oo” 7 8 (L411) Say) Similarly, the square root of a diagonal matrix is a2 12 avea{ ° (1.112) ay? What do we do if A is not diagonal? Since A is Hermitian, we can always find a unitary transformation that diagonalizes it, i.e., UtAU=a (1.113a) The reverse transformation that “undiagonalizes” a is A=UaUt (1.113b) Now notice that A? = UaU'UaU' = Ua?UtMATHEMATICAL REVIEW 23 or in general A” = Ua"Ut (1.114) so that S(A) = ¥ 6A" = u(y WE) = Ci (1.127a) or more generally a(x) = |ay a(x) = Cal (1.127b) and define the scalar product of two functions as fax a*(x)b(x) = Calb> (1.128) Thus the orthonormality relation (J.116) becomes. = dy (1.129) In this notation, Eq. (1.118) is Cia) =a; (1.130) and thus Eq. (1.117) becomes l® =D l (1.132b) which is identical to Eq. (1.52). An operator © 1s said to be nonlocal if we calculate b(x) as bya) = Calx) = fae Ol, xJatx) (1.133)28 MODERN QUANTUM CHEMISTRY Thus to find b(x) at the point x = x» we need to know a(x) for all x. Nonlocal operators are sometimes called integral operators. Note that (1.133) is the continuous generalization of b => Oa, (1.134) T and hence we can regard O(x, x’) as a continuous matrix. An operator G is local when we can find b(x) at the point x9 by knowing a(x) only in an infinitesimally small neighborhood of xq. The derivative (d/dx) is an example ofa local operator. If ObAx) = Ax) (1.135a) or in our shorthand notation Oley = wo, |er> (1.135b) then in analogy to matrix algebra, we say that @,(x) is an eigenfunction of ¢ with an eigenvalue ,. We can choose the eigenfunctions to be normalized fax d200648) = Caley =1 (1.136) Multiplying (1.135a) by ##(x), integrating over all x and using (1.136). we have O,= fi dx P8(CbAx) = Cal Olay (1.137) where we have introduced the notation f dx aX(x)Cb(x) = = is the component of |a) along the basis vector |i). Thus we can regard a function b(x) as the x component of the abstract vector |b) in a coordinate system with a continuously infinite number of axes. a. Multiply (2a) on the left by . Using (1b) show that the resulting equation is identical to (1.116) if VEO) = Ci]x> V(x) = Cx). b. Multiply (1a) by = 6( — x’) (2b) This is just the continuous analogue of (1b). c. Multiply (2a) by = Olay = [dx Ofx> = [by Multiply this equation by , show that O(% x) = Y WLIO WF).MATHEMATICAL REVIEW 31 1.3 THE VARIATION METHOD In this section we discuss an important approach to finding approximate solutions to the eigenvalue problem O$(x) = w(x) (1.141), We are interested in ergenvalue problems because the time-independent Schrédinger equation is an eigenvalue equation: H|D) = 6|@y (1.142) where # is a Hermitian operator called the Hamiltonian, |®) is the wave function, and & is the energy. We are interested in finding approximate solutions to eigenvalue equations because the Schrédinger equation cannot be solved exactly, except for the simplest cases. Although the subsequent development is applicable to any eigenvalue problem, we shall use the notation and terminology associated with the Schrédinger equation (1.142). Given the operator #, there exists a set of exact solutions to the Schré- dinger equation, infinite in number, labeled by the index H|0,) = 6|®,) «=0,1,... (1.143) where 6056, 56,5°°° S&S" We have assumed for the sake of simplicity that the set of eigenvalues {6} is discrete. Since # is a Hermitian operator, the eigenvalues &, are real and the corresponding eigenfunctions are orthonormal {®,| 0p) = Sap (1.144) Thus by multiplying Eq. (1.143) by <®,| on the left, we find Kp 3¢|®,) = Sxdap (1.145) Furthermore, we assume that the eigenfunctions of #% form a complete set and hence any function |) that satisfies the same boundary conditions as the set {,>} can be written as a linear combination of the |@®,)’s [®> = J®.de. = ¥ |<, | > (1.146) and a= Laol=¥ (6|0,><0,| (1.147) 1.3.1 The Variation Principle We are now in a position to state and prove an important theorem, called the variation principle: Given a normalized wave function |®) that satisfies32 MODERN QUANTUM CHEMISTRY the appropriate boundary conditions (usually the requirement that the wave function vanishes at infinity), then the expectation value of the Hamiltonian is an upper bound to the exact ground state energy. That is, if <6) =1 (1.148) then Eo (1.149) The equality holds only when |®) is identical to |). The proof of this theorem is simple. First we consider (O|by =1= x (O[0,)<@,|©,>5,,<,| By =F O|O,)<0, |) = Y [¢0,| >]? (1.150) where we have used Eqs. (1.144), (1.146), and (1.147). Then 6] 1B =F HLH |p) Op] B = YS C-|H>/? (1.151) where we have used Eg. (1.145). Finally, since & > 6p for all «, we have BPH = ¥ Sol.|B)/? = FoF K_|OP =F (1.152) @ @ where we have used the normalization condition (1.150). The variation principle for the ground state tells us that the energy of an approximate wave function is always too high. Thus one measure of the quality of a wave function is its energy: The lower the energy, the better the wave function. This is the basis of the variation method in which we take a normalized trial function |®, which depends on certain parameters, and vary these parameters until the expectation value <4 HB) reaches a minimum. This minimum value of = 60» ae Use the variation method with the trial function |®> =New™ to show that —x~' is an upper bound to the exact ground state energy (which is —0.5). You will need the integral -1 (Qm)!n4? 2 2mp-ax? _ JE de xe = TEMATHEMATICAL REVIEW 33 Exercise 1.19 The Schrédinger equation (in atomic units) for the hydro- gen atom is 1 1 —iy2—*)loy = 60». (Leesa Use the variation method with the trial function |6> = Ne" to show that —4/3n = —0.4244 is an upper bound to the exact ground state energy (which is —0.5). You will need the relations d d vynar re (ed)s9 om ae? — (2m) Ire? Jf dere = samt ait m cg om [oder om Exercise 1.20 The variation principle as applied to matrix eigenvalue problems states that if ¢ is a normalized (ce = 1) column vector, then c'Oc is greater or equal to the lowest eigenvalue of O. For the 2 x 2 symmetric matrix (047 = O21) On O 9 = (On On (on 0.) cu (%S e ~ \sin 6 which is normalized for any value of @. Calculate (6) = c'Oc and find the value of 0 (ic., 09) for which «(0) is a minimum. Show that (00) is exactly equal to the lowest cigenvalue of O (sce Eqs. (1.105) and (1.106a)). Why should you have anticipated this result? consider the trial vector 1.3.2 The Linear Variational Problem Given a trial function |®, which depends on a set of parameters, then the expectation value <@|#|®) will be a function of these parameters. In general, it will be such a complicated function that there is no simple way34 MODERN QUANTUM CHEMISTRY of determining the values of the parameters for which <®|3/|®) is a mini- mum. However. if only linear variations of the trial function are allowed, ie, if |® = y cl (1.153) where {Po} 1s a fixed set of N basis functions, then the problem of finding the optimum set of coefficients, {¢,}, can be reduced to a matrix diagonali- zation. To show this we assume that the basis functions are real and orthonormal CHV D = CFV D = 5; (1154) The case where the basis functions are complex and not orthogonal will be considered in Chapter 3. The matrix representation of the Hamiltonian operator in the basis {|'¥,>} is an N x N matrix H with elements (A), = Hy = and ¥ occurs at the same values of the coefficients. If we arbitrarily chose c,,C),. -- ,€y—1 a8 indepen- dent so that cy is determined from the normalization condition (1.156), thenMATHEMATICAL REVIEW 35 we have OL io k=1,2,...,N-1 (1.160) but 0 /dcy is not necessarily zero. However, we still have the undetermined multiplier E at our disposal. We now choose this multiplier so that 0Y/dcy does equal zero and thus 7° k=1,2,...,N-1,N (1.161) so that at 4g, 7 87 py Cy + = CH — 2Ecy (1.162) but since H,, = H,,, we have Hc, — Ee, =0 (1.163) i By introducing a column vector ¢ with elements c,, this set of equations can be written in matrix notation as He = Ec (1.164) which is the standard eigenvalue problem for the matrix H. Since H is symmetric, Eq. (1.164) can be solved to yield N orthonormal eigenvectors c* and corresponding eigenvalues E,, which for convenience are arranged so that Ey < Ey <- + < Ey ,. Thus Het = Ec? a =0.1,...,N-—1 (1.165) with (teh = ciel = bap (1.166) Introducing the diagonal matrix E containing the eigenvalues E, (i¢., (E),p = E,6,9) and the matrix of eigenvectors C defined as C;, = cj, the N relations contained in (1.165) can be written as HC = CE (1.167) Thus instead of finding just one solution for |6) and the expansion coefficients, we actually have found N solutions B= SF avy=¥ Cd) a@=0,1,...,N-1 (1.168) 4 ist which are orthonormal, since <@,|B,> = VPP, |¥,> =D chchSy = Vo ciel = Sup (1.169) 0 a 736 MODERN QUANTUM CHEMISTRY where we have used Egs. (1.154) and (1.166). To discover the significance of the E’s, we consider Bp). |B,> = YP HIP DCF a =) PAG u = (e)"He" = Eel)'e* = Ebay (1.170) where we have used Eqs. (1.165) and (1.166). Thus the eigenvalue E, is the expectation value of the Hamiltonian with respect to |6,). In particular, the lowest eigenvalue E, is the best possible approximation to the ground state energy of % in the space spanned by the basis functions {|¥,>}. Moreover, the variation principle assures us that Eo = (&,| #|Bo> > &o (L171) What is the significance of the remaming E's? It can be shown (see Exercise 1.21) that E, > 6,, «= 1,2,.... Thus E; is an upper bound to the energy , of the first excited state of ¥ and so on. Exercise 1.21 Consider a normalized trial function \o’> that is orthog- onal to the exact ground state wave function, i.c., <@'|o> = 0. a. Generalize the proof of the variation principle of Subsection 1.3.1 to show that 6. ie Consider the function [© = x\o> + y.> where |®>, x = 0. 1 are given by Eq. (1.168). Show that if it is normalized, then +)? c. When x and } are chosen so that |”) is normalized and so that <&'|®> = 0. then from part (2) it follows that (@'|.#|’y > &,. Show that (O]#|8') = E, —)xPE1 — Eo) Since E, > Eo, conclude that E, >. The above argument can be generalized to show that E, > 6,,% = 2,3,...MATHEMATICAL REVIEW 37 In summary, the linear variational method is a procedure for finding the best possible approximate solutions to the eigenvalue problem #|@y = 6|0) (1.172) given a fixed set of orthonormal functions {|¥,>, i= 1,2, ..., N}. The pro- cedure entails forming the matrix representation of the operator ¥ in the finite basis {|¥;>}, ie.. (H),; = ¢Y,|2¢]¥;> and solving the matrix eigenvalue problem He = Ee (1.173) that is, diagonalizing the N x N matrix H. We derived this result by explicitly minimizing the expectation value of the Hamiltonian. However, we can obtain Eq. (1.173) in an alternate way, which we will find useful. In an attempt to solve (1.172), let us approximate |®> as N joy = ¥ cf, (1.174) jet and substitute this expansion into (1.172) YC HV, = 6 Yel) (1.175) i J Multiplying (1.175) by <‘¥,| on the left and replacing 6 by E as a reminder that the expansion (1.174) is approximate, we find YL eKPIH¥) = EL ¢X¥i|¥)> = Ee, J 7 or ¥ Hye, = Ec: (1.176) 7 which in matrix notation is identical to Eq. (1.173). If we had used a complete orthonormal basis {[¥,), 1 = 1, 2,...,.N, N+ 1,...}, we would have ob- tained an equation identical to (1.173) except, of course, H would have been an infinite matrix. The eigenvalues of this matrix are exactly equal to the eigenvalues of the operator #. Thus the linear variational method is equiv- alent to solving the eigenvalue equation, (1.172) in a finite subspace spanned by {M), f= 1,2... Np. Exercise 1.22 The Schrddinger equation (in atomic units) for a hydrogen atom in a uniform electric field F in the z direction is (-3 ve : + Fr cos °) [®> = (% + Fr cos 0)|> = &(F)|®y38 MODERN QUANTUM CHEMISTRY Use the trial function [8> = ei|ts> + ex[2re> where |1s> and |2p, are the normalized eigenfunctions of %, i.c., [sy = 070" P2p.> = (322) *?re“"? cosé to find an upper bound to &(F). In constructing the matrix representation of #, you can avoid a lot of work by noting that H's) = — 31s), Ho|2P.> = — [20> Using (1 + x)'? = 1 + x/2, expand your answer in a Taylor series in F, ic, E(F) = E(0) — 4aF? +--+ Show that the coefficient « which is the approximate dipole polarizability of the system, is equal to 2.96. The exact result is 4.5. NOTES 1 J. H. Wilkinson, The Algebraic Eigenvalue Problem. Oxford University Press. New York, 1965. The classic reference work on the subject FURTHER READIN Acton, F S.. Numerical Methods that Work. Harper & Row. New York. 1970, Chapter 13 of this excellent book contains a nice discussion of a vanety of efficient algorithms for finding. the eigenvalues of matrices Cushing. J T.. Apphed Analytical Mathematics for Physical Sciennsts. Wiley, New York. 1975 The first four chapters contain a more rigorous. yet reasonably accessible. treatment of the mathematics of this chapter Merzbacher, E., Quantum Mechanics, 2nd ed., Wiley. New York, 1970. Most graduate quantum mechanics texts discuss Dirac notation in its full glory, and Chapter 14 of this book contains a good mtroduction. Press, W. H., Flannery, B. P.. Teukolsky, $. A., and Vetterling, W. T . Numerical Recipes, Cambridge University Press, Cambridge. 1986. Chapter 11 of this bible of scientific computing discusses and provides FORTRAN codes for a variety of algorithms for finding the eigenvalues and eigenvectors of matrices.CHAPTER TWO MANY-ELECTRON WAVE FUNCTIONS AND OPERATORS This chapter introduces the basic concepts, techniques, and notations of quantum chemistry. We consider the structure of many-electron operators (e.g. the Hamiltonian) and discuss the form of many-electron wave functions (Slater determinants and linear combinations of these determinants). We describe the procedure for evaluating matrix elements of operators between Slater determinants. We introduce the basic ideas of the Hartree-Fock approximation. This allows us to develop the material of this chapter in a form most useful for subsequent chapters where the Hartree-Fock approxi- mation and a variety of more sophisticated approaches, which use the Hartree-Fock method as a starting point, are considered in detail. In Section 2.1, the electronic problem 1s formulated, i.e., the problem of describing the motion of electrons in the field of fixed nuclear point charges. This is one of the central problems of quantum chemistry and our sole concern in this book. We begin with the full nonrelativistic time-independent Schrédinger equation and introduce the Born-Oppenheimer approximation. We then discuss a general statement of the Pauli exclusion principle called the antisymmetry principle, which requires that many-electron wave functions must be antisymmetric with respect to the interchange of any two electrons. In Section 2.2, we describe one-electron functions (spatial and spin orbitals) and then construct many-electron functions (Hartree products and 3940 MODERN QUANTUM CHEMISTRY Slater determinants) in terms of these one-electron functions. We then con- sider the Hartree-Fock approximation in which the exact wave function of the system is approximated by a single Slater determinant and describe its qualitative features, At this point, we introduce a simple system, the minimal basis (1s orbital on each atom) ab initio model of the hydrogen molecule. We shall use this model throughout the book as a pedagogical tool to illustrate and illuminate the essential features of a variety of formalisms that at first glance appear to be rather formidable. Finally, we discuss the multi- determinantal expansion of the exact wave function of an N-electron system. Section 2.3 is concerned with the form of the one- and two-electron operators of quantum chemistry and the rules for evaluating matrix elements of these operators between Slater determinants. The conversion of expres- sions for matrix elements involving spin orbitals to expressions involving spatial orbitals is discussed. Finally, we describe a mnemonic device for obtaining the expression for the energy of any single determinant. Section 2.4 introduces creation and annihilation operators and the formalism of second quantization. Second quantization is an approach to dealing with many-electron systems. which incorporates the Pauli exclusion principle but avoids the explicit use of Slater determinants. This formalism is widely used in the literature of many-body theory. It is, however, not required for a comprehension of most of the rest of this book, and thus this section can be skipped without loss of continuity. Section 2.5 discusses electron spin and spm operators in many-electron systems and contains a description of restricted and unrestricted spin orbitals and spin-adapted configurations. Spin-adapted configurations, unlike many single determinants derived from restricted spin orbitals, are correct eigen- functions of the total electron spin operator. Singlet, doublet, and triplet spin-adapted configurations as well as unrestricted wave functions, which are not eigenfunctions of the total electron spin operator, are described. 2.1 THE ELECTRONIC PROBLEM Our main interest in this book is finding approximate solutions of the non- relativistic time-independent Schrédinger equation JE|Dy = 6|@y (2.1) where # is the Hamiltonian operator for a system of nuclei and electrons described by position vectors R, and r,, respectively. A molecular coordinate system is shown in Fig. 2.1. The distance between the ith electron and Ath nucleus is r,4 = |r, — R,}; the distance between the ith and jth electron is r,, = |r; — ooh and the distance between the Ath nucleus and the Bth nucleus is Rag = [Ry — Rl. In atomic units, the Hamiltonian for N electronsMANY-ELECTRON WAVE FUNCTIONS AND OPERATORS 41 44) © electrons A,B a nuclei a Figure 2.1. A molecular coordmate system i, j = electrons, 4, B = nucles and M nuclei is Nt. y H=-Y _VP- Pea oe ar," B, ie Tia NN] MM ay +2 Le z. (2.2) In the above equation, M , is the ratio of the mass of nucleus A to the mass of an electron, and Z, is the atomic number of nucleus A. The Laplacian operators V? and V2 involve differentiation with respect to the coordinates of the ith electron and the Ath nucleus. The first term in Eq. (2.2) is the operator for the kinetic energy of the electrons; the second term is the operator for the kinetic energy of the nuclei; the third term represents the coulomb attraction between electrons and nuclei; the fourth and fifth terms represent the repulsion between electrons and between nuclei, respectively. 2.1.1 Atomic Units The units we use throughout this book are called atomic units. To see how these units arise naturally let us consider the Schrédinger equation for the hydrogen atom. In SI units, we have [-# ve -|¢- = 6 23) 2m, tren42 MODERN QUANTUM CHEMISTRY where h is Planck’s constant divided by 2z, m, is the mass of the electron, and —e is the charge on the electron. To cast this equation into dimensionless form we let x, y, 2 Ax’, Ay’, Az’ and obtain co ~ Gnd” The constants in front of the kinetic and potential energy operators can then be factored, provided we choose 4 such that (2.5) where &, is the atomic unit of energy called the Hartree. Solving Eq. (2.5) for 2 we find Ane oh? me (2.6) Thus 4 is just the Bohr radius ag which is the atomic unit of length called a Bohr. Finally, since \ 1 af} ve ho = ea en 2 r if we let &’ = &/&,, we obtain the dimensionless equation Loe "Vy eg (-3° 24 = 6&9 (2.8) which is the Schrédinger equation in atomic units. The solution of this equation for the ground state of the hydrogen atom yields an energy 6” equal to —0,5 atomic units 0.5 Hartrees. Table 2.1 gives the conversion factors X between atomic units and SI units, such that the SI value of any Table 2.1 Conversion of atomic units to SI units Physical quannity Conversion factor X Value of X (SI) Length a 5.2918 x 107"! m Mass m, 91095 x 10°31 kg Charge e 1.6022 x 107°C Energy & 4.3598 x 10718 J Angular momentum h 10546 x 10734 Js 8.4784 x 107° Cm 1.6488 x 10-4! C?m7J-+ $1423 x 10" Vm"! 2.5978 x 10'5 m~3? Electric dipole moment Electric polanzebility Electric field Wave functionMANY-ELECTRON WAVE FUNCTIONS AND OPERATORS 43 quantity Q is related to its value in atomic units Q’ by Q=xg (2.9) Conversion factors for a few other units, which are not related to SI but which are necessary to read the existing literature, are as follows. One atomic unit of length equals 0.52918 Angstréms (A). One atomic unit of dipole moment (two unit charges separated by do) equals 2.5418 Debyes (D), and one atomic unit of energy cquals 27.211 electron volts (eV) or 627.51 kcal/mole. From now on we drop the primes and all our quantities will be in atomic units. 2.1.2 The Born-Oppenheimer Approximation The Born-Oppenheimer approximation is central to quantum chemistry. Our discussion of this approximation is qualitative. The quantitative aspects of this approximation, including the problem of deriving corrections to it, are clearly discussed by Sutcliffe.’ Since nuclei are much heavier than electrons. they move more slowly. Hence, to a good approximation, one can consider the electrons in a molecule to be moving in the field of fixed nuclei. Within this approximation, the second term of (2.2), the kinetic energy of the nuclei, can be neglected and the last term of (2.2), the repulsion between the nuclei, can be considered to be constant. Any constant added to an operator only adds to the operator eigenvalues and has no effect on the operator eigenfunctions. The remaining terms in (2.2) are called the electronic Hamiltonian or Hamiltonian describing the motion of N electrons in the field of M point charges, N NM N Hau =- Ave- 5 ¥ My 1 = miata ¢ N41 Y- (2.10) 1 jet li The solution to a Schrédinger equation involving the electronic Hamiltonian, Here Verec = Ecteceree Q11) is the electronic wave function, Detee = Pore({ts}s (Rai) (2.12) which describes the motion of the electrons and explicitly depends on the electronic coordinates but depends parametrically on the nuclear coordinates, as does the electronic energy, Sctce = Setec({Ra}) (2.13) By a parametric dependence we mean that, for different arrangements of the nuclei, ®,;., is a different function of the electronic coordinates. The nuclear44 MODERN QUANTUM CHEMISTRY coordinates do not appear explicitly in ®,,,,. The total energy for fixed nuclei must also include the constant nuclear repulsion, MoM Bn = Sunt YS Ben (2.4 Az Boa Rap Equations (2.10) to (2.14) constitute the electronic problem, which is our interest in this book. If one has solved the electronic problem, it is subsequently possible to solve for the motion of the nuclei under the same assumptions as used to formulate the electronic problem. As the electrons move much faster than the nuclei, it is a reasonable approximation in (2.2) to replace the electronic coordinates by their average values, averaged over the electronic wave function. This then generates a nuclear Hamiltonian for the motion of the nuclei in the average field of the electrons, M 1 > / N 4 > NM Za N N 4 Hua Dog Vat (-P5v- > yey yt ir Aazt lia iS1 pil iy wy ae “1 % MM ZZ, =- Vi + Serec({Ry}) + one 22M, Mat Facdl Ral) + Dy M 1 -z, ™, Vi + Eo {Ra}) (2.15) The total energy &,.({R4}) provides a potential for nuclear motion. This function constitutes a potential energy surface as shown schematically in Fig. 2.2. Thus the nuclei in the Born-Oppenheimer approximation move on a potential energy surface obtained by solving the electronic problem. Solu- Eu(l Ra ‘) Figure 2.2 Schematic illustration of a potential surface,MANY-ELECTRON WAVE FUNCTIONS AND OPERATORS 45 tions to a nuclear Schrédinger equation, HouvciPauct = Envi (2.16) describe the vibration, rotation, and translation of a molecule, Paci = Prvcil{Ra}) (2.17) and &, which is the Born-Oppenheimer approximation to the total energy of (2.1), includes electronic, vibrational, rotational, and translational energy. The corresponding approximation to the total wave function of (2.1) is, (rs {Rag}) = Perec {t § (Ra Pave (Rah) (2.18) From now on, we will not consider the vibrational-rotational problem but concentrate solely on the electronic problem of (2.11) to (2.14). We thus drop the subscript “elec” and only consider electronic Hamiltonians and electronic wave functions. Where it is convenient or necessary, we will dis- tinguish between the electronic energy of (2.13) and the total energy of (2.14), which includes nuclear-nuclear repulsion, 2.1.3 The Antisymmetry or Pauli Exclusion Principle The electronic Hamiltonian in Eq. (2.10) depends only on the spatial co- ordinates of the electrons. To completely describe an electron it is necessary, however, to specify its spin. We do this in the context of our nonrelativistic theory by introducing two spin functions a(w) and B(w), corresponding to spin up and down, respectively. These are functions of an unspecified spin variable w; from the operational point of view we need only specify that the two spin functions are complete and that they are orthonormal, fides 2*(cpalco) = fideo p*(wyp(o) = 1 (2.19a) = 1 (2.19b) and Jdeo ae(ofleoy = fideo BHwoya(en) = 0 (2.20a) = 0 (2.20b) where the integration has been used in a formal way. In this formalism an electron is described not only by the three spatial coordinates r but also by one spin coordinate «v. We denote these four coordinates collectively by x, x ={1, 0} (2.21) The wave function for an N-electron system is then a function of x,, X,.-., Xy. That is, we write ®(x,,X2,..., Xp). Because the Hamiltonian operator makes no reference to spin, simply making the wave function depend on spin (in the way just described) does46 MODERN QUANTUM CHEMISTRY not lead anywhere. A satisfactory theory can be obtained, however, if we make the following additional requirement on a wave function: A many- electron wave function must be antisymmetric with respect to the interchange of the coordinate x (both space and spin) of any two electrons, DK 4,6 Xp ey Kye ees Ky) = KOK, Kye My eee Ry) (2.22) This requirement, sometimes called the antisymmetry principle, is a very general statement of the familiar Pauli exclusion principle. It is an indepen- dent postulate of quantum mechanics. Thus the exact wave function not only has to satisfy the Schrédinger equation, it also must be antisymmetric in the sense of Eq. (2.22). As we shall see, the requirement of antisymmetry is easily enforced by using Slater determinants. 2.2 ORBITALS, SLATER DETERMIN BASIS FUNCTIONS S, AND In this section we are concerned with the nomenclature, the conventions, and the procedure for writing down the wave functions that we use to describe many-electron systems. We will only consider many-electron wave functions that are either a single Slater determinant or a linear combination of Slater determinants. Sometimes, for very small systems, special functional forms are used for the wave function, but in most cases quantum chemists use Slater determinants. Before considering wave functions for many electrons, however, it is necessary to discuss wave functions for a single electron. 2.2.1 Spin Orbitals and Spatial Orbitals We define an orbital as a wave function for a single particle, an electron. Because we are concerned with molecular electronic structure, we will be using molecular orbitals for the wave functions of the electrons in a molecule. A spatial orbital wy,(r), is a function of the position vector r and describes the spatial distribution of an electron such that [),(9)|? dr is the probability of finding the electron in the small volume element dr surrounding r. Spatial molecular orbitals will usually be assumed to form an orthonormal set fae uri ,in =6,, (223) If the set of spatial orbitals {y,} were complete, then any arbitrary function f(e) could be exactly expanded as am S0= Y ah,0 (2.24) where the a, are constant coefficients. In general, the set would have to be infinite to be complete; however, in practice we will never have available aMANY-ELECTRON WAVE FUNCTIONS AND OPERATORS 47 complete set, but only a finite set {,]i=1,2,..., K} of K such orbitals. This finite set will only span a certain region of the complete space, but we can, however, describe results as being “exact” within the subspace spanned by the finite set of orbitals. To completely describe an electron, it is necessary to specify its spin. A complete set for describing the spin of an electron consists of the two ortho- normal functions o(c) and f(c), i.e, spin up (1) and spin down (|). The wave function for an electron that describes both its spatial distribution and its spin is a spin orbital, (x), where x indicates both space and spin coordinates (see Eq. (2.21)). From each spatial orbital, Y(t), one can form two different spin orbitals—one corresponding to spin up and the other to spin down—by multiplying the spatial orbital by the « or B spin function, respectively, ie. Y@)a(o) x(x) = or (2.25) YOB) Given a set of K spatial orbitals {w,|i= 1, 2,..., K}, one can thus forma set of 2K spin orbitals {y,|i = 1, 2,..., 2K} as Xar—-108) = Ya (Ox() } . =1,2,...,K 2.26) atey= verge) f= b> me If the spatial orbitals are orthonormal, so are the spin orbitals fax xto0x(%) = Gold = 44 (2.27) Exercise 2.1 Given a set of K orthonormal spatial functions, {y7(x)}, and another set of K orthonormal functions, {/(x)}, such that the first set is not orthogonal] to the second set, ie., fac verti where S is an overlap matrix, show that the set {y,} of 2K spin orbitals, formed by multiplying ¥(r) by the « spin function and yf(r) by the B spin function, ie., =S, Xar— 1X) = Wile) . =1,2..., zalx) = VOB) } vk is an orthonormal set. 2.2.2 Hartree Products Having seen that the appropriate wave function describing a single electron isa spin orbital, we now consider wave functions for a collection of electrons,48 MODERN QUANTUM CHEMISTRY = “)'~" i.e. N-electron wave functions. Before considering the form of the exact wave function for a fully interacting system, let us first consider a simpler system containing noninteracting electrons having a Hamiltonian of the form > N ‘ # => Nii) (2.28) 1 where h(i) is the operator describing the kinetic energy and potential energy of electron i. If we neglect electron-electron repulsion, then the full electronic Hamiltonian has this form. Alternatively, h(i) might be an effective one- electron Hamiltonian that includes the effects of electron-electron repulsion in some average way. Now, the operator h(i) will have a set of eigenfunctions that we can take to be a set of spin orbitals {y,}, AO) = 70%) (2.29) We now ask, “What are the corresponding eigenfunctions of #7" Because # 1s. a sum of one-electron Hamiltonians, a wave function which is a simple product of spin orbital wave functions for each electron, PPX Xa oe Xp) = Zal%a)L(%2) aX) (2.30) is an eigenfunction of #, AHP Ep (2.31) with eigenvalue E, which is just the sum of the spin orbital energies of each of the spin orbitals appearing in ‘¥"", 1” E=e,+8 400+ & ‘ (2.32) Such a many-electron wave function is termed a Hartree product, with electron-one being described by the spin orbital y,, electron-two being described by the spin orbital y,, etc. Exercise 2.2 Show that the Hartree product of (2.30) is an eigenfunction Nx of # = >. h(i) with an eigenvalue given by (2.32). = The Hartree product is an uncorrelated or independent-electron wave function because . EPPO, 2 XuMl? dx dey which is the simultaneous probability of finding electron-one in the volume element dx,, centered at x,, electron-two in dx,, etc, is just equal, from (2.30), to the product of probabilities Water? dy e(x2))? dx2--- baGemd?? dxy,MANY-ELECTRON WAVE FUNCTIONS AND OPERATORS 49° that electron-one is in dx,, times the probability that electron-two is in dx, etc. The situation is analogous to the probability of drawing an ace of hearts (1/52) being equal to the probability of drawing an ace (1/13) times the probability of drawing a heart (1/4), since the probability of a particular card being an ace is independent or uncorrelated with the probability that the given card is a heart. The probability of finding electron-one at a given point in space is independent of the position of electron-two when a Hartree product wave function is used. In reality, electron-one and electron-two will be instantaneously repelled by the two-electron coulomb interaction, and electron-one will “avoid” regions of space occupied by electron-two so that the motion of the two electrons will be explicitly correlated. An example of correlated probabilities is provided by 2 hot potatoes and 2 cold apples in a bucket. The probability of obtaining a hot potato upon randomly with- drawing an object from the bucket (1/2) is not equal to the product of the probability of getting a hot object (1/2) times the probability of getting a potato (1/2), since whether the object is hot is perfectly correlated with whether the object is a potato. Assuming independent electrons and a Hamiltonian of the form of Eq. (2.28), there is still a basic deficiency n the Hartree product; it takes no account of the indistinguishability of electrons, but specifically distinguishes electron-one as occupying spin orbital y,, electron-two as occupying x,, etc. The antisymmetry principle does not distinguish between identical electrons and requires that electronic wave functions be antisymmetric (change sign) with respect to the interchange of the space and spin coordinates of any two electrons. 2.2.3 Slater Determinants The Hartree product does not satisfy the antisymmetry principle. However, we can obtain correctly antisymmetrized wave functions as follows. Consider a two-electron case in which we occupy the spin orbitals y, and y,. If we put electron-one in y, and electron-two in y,, we have WUDK a» Xa) = %a(K)ZAX2) (2.33a) On the other hand, if we put electron-one in y, and electron-two in z,, we have PEN Xa) = Yil%2)4,(%1) (2.33b) Each of these Hartree products clearly distinguishes between electrons; how- ever, we can obtain a wave function which does not, and which satisfies the requirement of the antisymmetry principle by taking the appropriate linear combination of these two Hartree products, WX, X2) = 27 G(x )x,(K2) — H0%)xK2)) (2.34) The factor 2~'? is a normalization factor. The minus sign insures that ‘Y(x,, X2) is antisymmetric with respect to the interchange of the coordinates50 MODERN QUANTUM CHEMISTRY of electrons one and two. Clearly, W(X 1, Xo) = — F(X X1) (2.35) From the form of Eq. (2.34), it is evident that the wave function vanishes if both electrons occupy the same spin orbital (ie. if i = j/). Thus the anti- symmetry requirement immediately leads to the usual statement of the Pauli exclusion principle namely, that no more than one electron can occupy a spin orbital. Exercise 2.3 Show that ‘¥(x;, ¥2) of Eq. (2.34) is normalized. Exercise 2.4 Suppose the spin orbitals y; and y, are eigenfunctions of a one-electron operator h with eigenvalues ¢, and ¢, as in Eq. (2.29). Show that the Hartree products in Eqs. (2.33a, b) and the antisymmetrized wave function in Eq. (2.34) are eigenfunctions of the independent-particle Hamil- tonian H% = h(1) + h(2) (c.f. Eq. (2.28)) and have the same eigenvalue namely, 6 +e - The antisymmetric wave function of Eq. (2.34) can be rewritten as a determinant (see Eq, (1.39)) W(X 1, X2) = 271? ils) ube (2.36) aX) X\(%2 and is called a Slater determinant. For an N-electron system the generaliza- tion of Eq. (2.36) is 20K.) XK) el) WX, Xa 6-5 Xx) = (NYE? ala) 1j0%2) (2) (237) Xn) X(Xn) ++» Xl) The factor (N!)~'/? is a normalization factor. This Slater determinant has N electrons occupying N spin orbitals (x, %j.-... 24) Without specifying which electron is in which orbital. Note that the rows of an N-electron Slater determinant are labeled by electrons: first row (x,), second row (x3), etc. and the columns are labeled by spin orbitals: first column (y,), second column (x), etc. Interchanging the coordinates of two electrons corresponds to,inter- changing two rows of the Slater determinant, which changes the sign of the determinant. Thus Slater determinants meet the requirement of the anti- symmetry principle. Having two electrons occupying the same spin orbital corresponds to having two columns of the determinant equal, which makes the determinant zero. Thus no more than one electron can occupy a spin orbital (Pauli exclusion principle). It is convenient to introduce a short-hand notation for a normalized Slater determinant, which includes the normaliza-MANY-ELECTRON WAVE FUNCTIONS AND OPERATORS 51 tion constant and only shows the diagonal elements of the determinant, Ory Kay - 5 Xv) = fea )% 2) °°» Zan) (2.38) If we always choose the electron labels to be in the order X,,X>,-..,%Xy> then Eq. (2.38) can be further shortened to WK 1 Kase XW) = adj Med (2.39) Because the interchange of any two columns changes the sign of a deter- minant, the ordering of spin orbital labels in Eq. (2.39) is important. In our short-hand notation, the antisymmetry property of Slater determinants is [eee Pe nD (2.40) To within a sign, a Slater determinant is completely specified by the spin orbitals from which it is formed (i.e.. the spin orbitals that are occupied). Slater determinants formed from orthonormal spin orbitals are normalized. N-electron Slater determinants that have different orthonormal spin orbitals occupied are orthogonal. Exercise 2.5 Consider the Slater determinants IK) = luz> IL> = beard Show that KK|LY = bu 5p — 605 Note that the overlap is zero unless: 1)k = iand/ = j, in which case|L)=|K> and the overlap is unity and 2) k = j and 1 = iin which case |Z.) = [yx —|K) and the overlap is minus one. We have seen that a Hartree product is truly an independent-electron wave function since the simultaneous probability of finding electron-one in dx, at x,, electron-two in dx, at x,, etc. is simply equal to the product of the probabilities that electron-one is in dx,, electron-two is in dx, etc. Antisymmetrizing a Hartree product to obtain a Slater determinant intro- duces exchange effects, so-called because they arise from the requirement that |‘?|? be invariant to the exchange of the space and spin coordinates of any two electrons. In particular, a Slater determinant incorporates exchange correlation, which means that the motion of two electrons with parallel spins is correlated. Since the motion of electrons with opposite spins remains uncorrelated, it is customary to refer to a single determinantal wave function as an uncorrelated wave function. To see how exchange correlation arises, we now investigate the effect of antisymmetrizing a Hartree product on the electron density. Consider a52 MODERN QUANTUM CHEMISTRY two-electron Slater determinant in which spin orbitals 7, and 7. are occupied MOK, Xo) = [a x21%2)> (2.41) If the two electrons have opposite spins and occupy different spatial orbitals, 1a%)) = Wiltya(@,) (2.42) Xa%1) = Walt )B(@2) (2.43) then by expanding the determinant, one obtains. SEP dy dco = 4 ]b Ors )ee(co Wr €F2)B (C2) — Yelta)o(co)ha(e )B(eo1)]? dx, dx (2.44) for the simultaneous probability of electron-one being in dx, and electron- two being in dx,. Let P(r,, r2) dt,dr, be the probability of finding electron- one in dr, at r, and simultaneously electron-two in dr2 at r2, as shown in Fig. 2.3. This probability is obtained by integrating (averaging) Eq. (2.44) over the spins of the two electrons, Porta) dt dty = [deo, dog [YP dey dey = F[Wite)P Wales)? + Wile)? WoeDP?] drs dra (2.45) The first term in (2.45) is the product of the probability of finding electron-one in dr, at r, times the probability of finding electron-two in dr at 2, if electron-one occupies W, and electron-two occupies >. The second term has electron-one occupying 2 and electron-two occupying W,. Since elec- trons are indistinguishable, the correct probability is the average of the two Pay vt) at dre Figure 2.3 Probability of electron-one being in dr, and electron-two being in dryMANY-ELECTRON WAVE FUNCTIONS AND OPERATORS 53 terms as shown. Thus the motion of the two electrons is uncorrelated. This 1s particularily obvious ify, = 2, for in that case PO.) = Walt Paral? (2.46) Note that P(r,,1r,;) #0 so that there is a finite probability of finding two electrons with opposite spins at the same point in space. If the two electrons have the same spin (say f), we have 710%.) = Wilts )B@,) (2.47) £A%2) = Halt 2)Blo2) (2.48) then, by steps identical to the above, we obtain. Pry, Fa) = AAW sted baleal? + Walradl’Wale |? = [VRC Wale SEW (ea) + HWA Wale We) ] (2.49) where we now have an extra cross term, making the probabilities correlated. This 1s exchange correlation between electrons of parallel spin. Note that P(r,,r,) = 0. and thus the probability of finding two electrons with parallel spins at the same point in space is zero. A Fermi hole is said to exist around an electron. In summary, within the single Slater determinantal description, the motion of electrons with parallel spins 1s correlated but the motion of electrons with opposite spins is not. 2.2.4 The Hartree-Fock Approximation Finding and describing approximate solutions to the electronic Schrodinger equation has been a major preoccupation of quantum chemists since the birth of quantum mechanics. Except for the very simplest cases like H}, quantum chemists are faced with many-electron problems. Central to at- tempts at solving such problems, and central to this book, is the Hartree- Fock approximation. It has played an important role in elucidating modern chemistry. In addition, it usually constitutes the first step towards more accurate approximations. We are now in a position to consider some of the basic ideas which underlie this approximation. A detailed description of the Hartree-Fock method is given in Chapter 3. The simplest antisymmetric wave function, which can be used to describe the ground state of an N-clectron system, 1s a single Slater determinant, [od=|ir2 n> (250) The variation principle states that the best wave function of this functional form is the one which gives the lowest possible energy Eq = <¥ol#|¥o> (251)54 MODERN QUANTUM CHEMISTRY where # is the full electronic Hamiltonian. The variational flexibility in the wave function (2.50) is in the choice of spin orbitals. By minimizing Ey with respect to the choice of spin orbitals, one can derive an equation, called the Hartree-Fock equation, which determines the optimal spin orbitals. We shall show in Chapter 3 that the Hartree-Fock equation is an eigenvalue equation of the form SU) = Ex) (2.52) where (i) is an effective one-electron operator, called the Fock operator, of the form 4 ani tia + vAF(i) (2.53) where v"*(i). which will be explicitly defined in Chapter 3, is the average potential experienced by the ith electron due to the presence of the other electrons. The essence of the Hartree-Fock approximation is to replace the complicated many-electron problem by a one-electron problem in which electron-electron repulsion is treated in an average way. ' ‘The Hartree-Fock potential v*(i), or equivalently the “field” seen by the ith electron, depends on the spin orbitals of the other electrons (i.¢., the Fock operator depends on its eigenfunctions). Thus the Hartree-Fock equa- tion (2.52) is nonlinear and must be solved iteratively. The procedure for solving the Hartree-Fock equation is called the self-consistent-field (SCF) method. The basic idea of the SCF method is simple. By making an initial guess at the spin orbitals, one can calculate the average field (i.e., v“) seen by each electron and then solve the eigenvalue equation (2.52) for a new set of spin orbitals. Using these new spin orbitals, one can obtain new fields and repeat the procedure until self-consistency is reached (i.e., until the fields no longer change and the spin orbitals used to construct the Fock operator are the same as its eigenfunctions). The solution of the Hartree-Fock eigenvalue problem (2.52) yields a set {x} of orthonormal Hartree-Fock spin orbitals with orbital energies {¢,}. The N spin orbitals with the lowest energies are called the occupied or hole spin orbitals. The Slater determinant formed from these orbitals is the Hartree- Fock ground state wave function and is the best variational approximation to the ground state of the system, of the single determinant form. We shall label occupied spin orbitals by the indices a, b,c, ...(1€5 Yas Zp. +). The remaining members of the set {7,} are called virtual, unoccupied, or particle spin orbitals. We shall label virtual spin orbitals by the indices r, s, t,... (1. os Zoo oe In principle, there are an infinite number of solutions to the Hartree- Fock equation (2.52) and an infinite number of virtual spin orbitals. In practice, the Hartree-Fock equation is solved by introducing a finite set ofMANY-ELECTRON WAVE FUNCTIONS AND OPERATORS 55 Xex ° artuat - wspin Xs orbitals x, ne PPP PLPL APPL PPILS ‘occupied —%— x, ‘spin orbitals ——*—_ x —*—- Xe Figure 2.4 The Hartree-Fock ground state kX determinant, |ys%2 °° ake" “Yn spatial basis functions {¢,(r)| = 1, 2,..., K}. The spatial parts of the spin orbitals with the spin function can then be expanded in terms of the known set of functions {¢,}. The spatial parts of the spin orbitals with the B spin can be expanded in the same way and both expansions substituted into the eigenvalue problem (2.52) to obtain matrix eigenvalue equations for the expansion coefficients, These matrix equations (e.g., the Roothaan equations) will be studied in some detail in Chapter 3. It is sufficient for this discussion to realize that using a basis set of K spatial functions {¢,} leads to a set of 2K spin orbitals (K with a spin and K with f spin). This leads to a set of N occupied spin orbitals {7,} and a complementary set of 2K — N unoccupied or virtual spin orbitals {7,}. A single Slater determinant formed from the set {z,} is the variational Hartree-Fock ground state, for which we will use the symbol Vy or |. A pictorial representation of |) is presented in Fig. 2.4. The 2K Hartree-Fock spin orbitals have been ordered according to their energy, and we have neglected possible degeneracies. The occupancy of the N lowest energy spin orbitals—one electron per spin orbital—is indicated by the asterisks. The larger and more complete the set of basis functions {¢,}, the greater is the degree of flexibility in the expansion for the spin orbitals and the lower will be the expectation value Ey = <¥|#/|¥5>. Larger and larger basis sets will keep lowering the Hartree-Fock energy Ey until a limit is reached, called the Hartree-Fock limit. In practice, any finite value of K will lead to an energy somewhat above the Hartree-Fock limit. 2.2.5 The Minimal Basis H, Model At this point we introduce a simple model system, which we will be using throughout this book to illustrate many of the methods and ideas of quantum56 MODERN QUANTUM CHEMISTRY chemistry. The model we use is the familiar minimal basis MO-LCAO description of H. In this model, each hydrogen atom has a 1s atomic orbital and, as the two atoms approach, molecular orbitals (MOs) are formed as a linear com- bination of atomic orbitals (LCAO). The coordinate system is shown in Fig. 2.5. The first atomic orbital. ,, is centered on atom 1 at R,. The value of ¢, at a point in space r is ¢4(r) or, since its value depends on the distance from its origin, we sometimes write ¢, = ¢,(r — R,). The second atomic orbital is centered on atom 2 at Ro, ie, ¢2 = p(t — R2). The exact 1s or- bital of a hydrogen atom centered at R has the form lr — R) = (C/nye-r-F (2.54) where ¢, the orbital exponent, has a value of 1.0. This 1s an example of a Slater orbital. In this book we will be concerned mostly with Gaussian orbitals, which lead to simpler integral evaluations than Slater orbitals. The 1s Gaussian orbital has the form 1 b(t — R) = (2a/nj!te-a-BP (2.55) where « is the Gaussian orbital exponent. For the present, we need not be concerned with the particular form of the 1s atomic orbitals. The two atomic orbitals ¢, and $2 can be assumed to be normalized, but they will not be orthogonal. They will overlap, such that the overlap integral is Sia= far dttryd.t) (2.56) Figure 2.5. Coordinate system for minimal basis H).MANY-ELECTRON WAVE FUNCTIONS AND OPERATORS 57 The overlap will depend on the distance R,, = |R,— R2|, such that Sy2 = 1 when R,2 = O and S,, =0 when Ry, = 0. From the two localized atomic orbitals, 6, and 2, one can form, by Imnear combination, two delocalized molecular orbitals. The symmetric combination leads to a bonding molecular orbital of gerade symmetry (i.¢., symmetric with respect to inversion about the poift centered between the nuclei) Wi = [20 + Si) 71 + 6d (257) whereas, the antisymmetric combination leads to an antibonding molecular orbital of ungerade symmetry (i.¢., antisymmetric with respect to inversion about the point centered between the nuclei) Yo = [20 - Sy)" 71 — 2) (258) Exercise 2.6 Show that y, and , form an orthonormal set. The above procedure is the simplest example of the general technique of expanding a set of spatial molecular orbitals in a set of known spatial basis functions win = ¥ Cubgle) (259) To obtain the exact molecular orbitals of H, one would need an infinite number of terms in such an expansion. Using only two basis functions for H) is an example of a minimal basis set and an obvious choice for the two functions $, and ¢, is the 1s atomic orbitals of the atoms. The correct linear combinations for this simple choice are determined by symmetry, and one need not solve the Hartree-Fock equations; , and 2 of Eqs. (2.57) and (2.58) are the Hartree-Fock spatial orbitals in the space spanned by $, and p>. Given the two spatial orbitals y, and y, we can form four spin orbitals 2Alx) = Wy (Malco) 2X) = WNP) ZalX) = ha(n)a(e) ZalX) = Vo B(O) The orbital energies associated with these spin orbitals can be obtained only by explicitly considering the Hartree-Fock operator. But, as might be ex- Pected, y, and , are degenerate and have the lower energy corresponding toa bonding situation, while 7 and y, are also degenerate having a higher ‘nergy corresponding to an antibonding situation. The Hartree-Fock ground State in this model is the single determinant [od = lane) (261) (2.60)58 MODERN QUANTUM CHEMISTRY X%3————- ———-% Io) = x —*— 8% OH ve +4- 4 Figure 2.6 Three different representations of the Hartree-Fock ground state of minimal basis H,. shown pictorially in Fig. 2.6. Sometimes it is convenient to use a notation that indicates a spin orbital by its spatial part, using a bar or lack of a bar to denote whether it has the f or a spin function. Thus ney Mev 1S mah (262) ts = Vo Ya = V2 In this notation the Hartree-Fock ground state is [o> = Wath = |tD> (2.63) which indicates that both electrons occupy the same spatial orbital y,, but one has an @ spin and one has a f spin. It will be apparent from the context whether yy, denotes a spatial orbital, or a spin orbital made up of the yw, spatial orbital and the @ spin function. 2.2.6 Excited Determinants The Hartree-Fock procedure produces a set {7;} of 2K spin orbitals. The Hartree-Fock ground state, [o> = bata” Lao n> (2.64) is the best (in a variational sense) approximation to the ground state, of the single determinant form. However, it clearly is only one of many determinants that could be formed from the 2K > N spin orbitals. The number of com- binations of 2K objects taken N at a time is the binomial coefficient 2K\___(2K)! N] NIQK—N)!MANY-ELECTRON WAVE FUNCTIONS AND OPERATORS 59 This is the same as the number of different single determinants that one can form from N electrons and 2K spin orbitals; the Hartree-Fock ground state is just one of these. A convenient way of describing these other deter- minants is to consider the Hartree-Fock ground state (2.64) to be a reference state and to classify other possible determinants by how they differ from the reference state, ie., by stating which occupied or hole spin orbitals of the set {74} in (2.64), have been replaced by which virtual or particle spin orbitals of the set {y,}. These other determinants can be taken to represent approxi- mate excited states of the system or, as we shall see shortly, they can be used in linear combination with |W) for a more accurate description of the ground state or any excited state of the system. A singly excited determinant is one in which an electron, which occupied Ya in the Hartree-Fock ground state (2.64), has been promoted to a virtual spin orbital z,, as shown in Fig. 2.7, [a> = leaxa + Lite wd (2.65) A doubly excited determinant, shown in Fig. 2.8, is one in which electrons have been excited from y, and y, to y, and y., [> = bana dete XW> (2.66) 2K All ( a) determinants can thus be classified as either the Hartree-Fock ground state or singly, doubly, triply, quadruply, ... , N-tuply excited states, The importance of these determinants as approximate representations of the true states of the system diminishes, in some sense, in the above order. While the excited determinants are not accurate representations of the Xex Xs x Xnst —K—— x} Figure 2.7 A singly excited determinant,60 MODERN QUANTUM CHEMISTRY Xan Xs Xr Xw Is) Xp Xe 9 HK Figure 2.8 A doubly excited determinant excited states of the system, they are 1mportant as N-electron basis functions for an expansion of the exact N-electron states of the system. 2.2.7 Form of the Exact Wave Function and Configuration Interaction We now consider the use of excited determinants as N-electron basis func- tions. Suppose we have a complete set of functions {7,(x)}. Any function (x,) of a single variable can then be exactly expanded as ®,) = Y aes) (267) where a, is an expansion coefficient. How can we expand a function of two variables ®(x,, x2) in an analogous way? If we think of x, as being held fixed, then we can expand ®(x,, x2) as P(X, 2) = Yi aile2) zi) (2.68) where the expansion coefficients are now functions of x. Since a(x) is a function of a single variable, it can be expanded in the complete set {y;} as ai(x2) = ¥ byxs0%2) (2.69) J Substituting this result in (2.68) gives lx, x2) = Y byrileeaz,(%2) (2.70) ”MANY-ELECTRON WAVE FUNCTIONS AND OPERATORS 61 If, however, we require ® to be antisymmetric, (x1, X2) = —D(xp, 4) (2.71) then b,, = —b, and b; = 0. or Oy, 2) = YY balay) — 40422)] 7 pri = ¥ 2b, kur, (2.72) a Thus an arbitrary antisymmetric function of the two variables can be exactly expanded in terms of all unique determinants formed from a complete set of one-variable functions {7,(x)}. This argument is readily extended to more than two variables, so that the exact wave function for the ground and ex- aited states of our N-electron problem can be written as a linear combination of all possible N-electron Slater determinants formed from a complete set of spin orbitals {7,}. Since all possible determinants can be described by reference to the Hartree-Fock determinant, we can write the exact wave function for any state of the system as [®>=col¥o> +E. cea> + x Gi iN) x Cab ah + + (2.73) 4 ra By summing over a < b, we mean summing over all a and over all b greater than a (i.¢, over all unique pairs of occupied spin orbitals). Similarly, sum- ming over r < s means summing over all unique pairs of virtual spin orbitals. Thus all unique doubly excited configurations are included in the expansion. The situation is analogous for triply and higher excited determinants. Thus the infinite set of N-clectron determinants {]¥;>} = {]¥o), |¥2>, [VR ---} 1s a complete set for the expansion of any N-electron wave function. The exact energies of the ground and excited states of the system are the eigen- values of the Hamiltonian matrix (i.., the matrix with elements <¥,|7[¥,)) formed from the complete set {|;)}. Since every |'¥;> can be defined by specifying a “configuration” of spin orbitals from which it is formed, this procedure is called configuration interaction (C1); CI will be considered in some detail in Chapter 4. The lowest eigenvalue of the Hamiltonian matrix, denoted by &, 1s the exact nonrelativistic ground state energy of the system within the Born-Oppenheimer approximation. The difference between this exact energy, &, and the Hartree-Fock-limit energy, Eo, is called the corre- lation energy Exore = 69 ~ Ey (2.74) since the motion of electrons with opposite spins 1s not correlated with the Hartree-Fock approximation. Unfortunately, the above procedure for the complete solution to the many-electron problem cannot be implemented in practice because one62 MODERN QUANTUM CHEMISTRY cannot handle infinite basis sets. If we work with a finite set of spin orbitals tali orbitals do not form a complete N-electron basis. Nevertheless, diagonalizing the finite Hamiltonian matrix formed from this set of determinants leads to solutions that are exact within the one-electron subspace spanned by the 2K spin orbitals or, equivalently, within the N-electron subspace spanned 1,2,...,2K}, then the (x determinants formed from these spin by the tr determinants. This procedure is called full CI. Even for rela- tively small systems and minimal basis sets, the number of determinants that must be included in a full CI calculation is extremely large. Thus in practice one must truncate the full CI expansion and use only a small fraction 2K y : - . of the ( N ) possible determinants. Figure 2.9 schematically shows how the exact nonrelativistic Born-Oppenheimer wave function is approached as the size of the one-electron and N-electron basis sets increases. Exercise 2.7 A minimal basis set for benzene consists of 72 spin orbitals. Calculate the size of the full CI matrix if it would be formed from deter- minants. How many singly excited determinants are there? How many doubly excited determinants are there? Let us illustrate the above ideas with our minimal basis H, model. Recall (see Eq. (2.60)) that there are four (2K = 4) spin orbitals x3, %2, ¥3> Hartree-Fock Exact Result Limit 50 = 540 3g 50 > nil ci é 20 & 3 3g 10 5 2 1 10 100 1000 19,000 Number of Slater Determinants (2K) Figure 2.9 Dependence of calculations on size of one-electron and N-electron basis setsMANY-ELECTRON WAVE FUNCTIONS AND OPERATORS 63 ! and Z4 in this model. Since N = 2 we can form 2) mn 6 unique determinants. The Hartree-Fock ground state determinant is 7 wu 2 [o> = lated = Wad = [LD 1-1 (2.75) The singly excited determinants are a) MD=PE> (2.16a) u—{—2 WD=PI (2.76b) [2> = [12> ; \ (2.760) u—{—2 2 = |12, (2.76) [WP = |12> ty ) There is only one doubly excited determinant. u—t}-2 [W22> = [22> = waxed = [V7 4 1 (2.77) Within the space spanned by the minimal basis set, the exact wave functions will be linear combinations of these six determinants. The Hartree-Fock ground state has two electrons in a gerade orbital and is of g symmetry (plus times plus equals plus). The doubly excited determinant has two electrons in an ungerade orbital and hence is also of g symmetry (minus times minus equals plus). The singly excited determinants, on the other hand, have one electron in a gerade orbital and one electron in an ungerade orbital and, therefore, are of u symmetry (plus times minus equals minus). The exact ground state wave function of minimal basis H, |®o>, like the Hartree-Fock approximation to it, [Wo is of g symmetry. Therefore, only determinants of g symmetry can appear in the expansion of |® > and thus we have [®o> = colo + eFHVTT> = colo + cf3|7 (2.78) The exact value of the coefficients in (2.78). which describe the wave function |®o) and the value of the exact energy <®o|7|®p). can be found by diag- Onalizing the full CI matrix, ie. the 2 x 2 Hamiltonian matrix in the basis Mo) and |¥72, _( SPl% [o>

. By evaluating such matrix elements, we mean re- ducing them to integrals involving the individual spm orbitals 7, occupied in |K> and |L), and ultimately to integrals mvolving spatial orbitals y;. Before giving general rules for evaluating such matrix elements, we illustrate the procedure with our minimal basis H, model 2.3.1 Minimal Basis H, Matrix Elements Let us evaluate the matrix elements that appear in the full CI matrix of minimal basis H, (see Eq. (2.79)). The exact ground state of this model is a linear combination of the Hartree-Fock ground state {o> = |y1%2) = |IT> and the doubly excited state |“) = |y37.) = |3?> = |22>. We need to evaluate the diagonal elements and <¥7$|#/|Vo). The Hamiltonian for any two-electron system is Z, 1 Z, 1 1 H=(-sVi- 24) +(—703- )ee (a) -E%4)+(ayi-Ezs)e = AL) + A(2) + I (2.80) Ty2 where h(1) is a core-Hamiltonian for electron-one, describing its kinetic energy and potential energy in the field of the nuclei (the “core”). It will be convenient to separate the total Hamiltonian into its one- and two-electron parts ©, =h(l) + h(2) (281) O, =r? (2.82)MANY-ELECTRON WAVE FUNCTIONS AND OPERATORS 65 Let us first consider the matrix element <‘¥|0,|¥ > which, from (2.81), ts a sum of two terms. The first term 1s <¥ olh(L) o> = fer ax [271 (xs0%1)x2(%2) — xol%)01%2)))* X (Es )[27 10%) 20%2) — YalK x2) Ax, Ay {HEX 1) EC% VAC YH aC al 2) + ECT 2) AO Hal) (%2) HE BO VA 12%) %2)— LEO IE )ACE 11% 20%) (2.83) In the above four terms the integration over x, produces either 1 (first two terms) or 0 (last two terms) from the orthonormality of the spin orbitals. Thus 1 I CHolhD o> = 5 fan AOA deaCe)+ 5 Pdr, ZAG. IRE ZACK) (2.84) By exactly the same procedure, one finds that <¥ o|h(2)|¥o> = and thus <#AOs[¥o> = far. 2Fss)htes)eab,) + far, AG hCe Yea.) 285) The integrals m this expression are one-electron integrals, i.c., the integration is over the coordinates of a single electron. The dummy variables of integra- tion are, by convention, chosen to be the coordinates of electron-one. Introducing the following notation for one-electron integrals involving spin orbitals, = + <2Ih|2> (2.87) Exercise 2.8 Show that CHO [PTS = C3|AIBY + <4lhI4> and (PoC 7D = (PI|O,|¥o> = 066 MODERN QUANTUM CHEMISTRY Now, let us evaluate matrix elements of ©. (alCal¥o> = fas de [2- rte rdealee) ~ xalmsdzata)/I* x HL2 One dnabe) — ae 0%2))] faxrdx, {BC Ea) 2 10H a0%2) + XE) 2 a8 MH A%2) HEU AS aI 2 206 aKa) — EOE TZ LAL 2)} (2.88) Since r,2 = rz, We can interchange the dummy variables of integration in the second term of the above expression and show that it is equal to the first term. Similarly, the third and fourth terms are equal. Thus CH oll) = [dns dre 280s )8l2)r i 100s al0) ~ fxs dys ctocietoaritzadnea) — (289) The integrals in this expression are examples of two-electron integrals, the integration is over the eight space and spin coordinates of electron 1 and 2, It is conventional always to choose the dummy variables of integration in a two-electron integral to be the coordinates of electrons 1 and 2. Intro- ducing the following notation for two-electron integrals involving spin orbitals, = (1212) — ¢12]21) (291) and the Hartree-Fock ground state energy is KH ol |Wo> = = CIJh{L> + <2hI2> + <12]12> — C12}21> (2.92) Exercise 2.9 Using the above approach, show that the full CI matrix for minimal basis H, is + ¢2]h|2> <12[34) — ¢12|43> _ + (12]12) — ¢12|21> <34]12> — ¢34]21> <3]h|3> + ¢4]hj4> + (34[34> — ¢(34[43), and that it is Hermitian.MANY-ELECTRON WAVE FUNCTIONS AND OPERATORS 67 2.3.2. Notations for One- and Two-Electron Integrals Before generalizing the above results and presenting general expressions for matrix elements involving N-electron determinants, it is appropriate to summarize the different notations we use in this book for one- and two- electron integrals. The notation for two-electron integrals over spin orbitals that we have introduced in Eq. (2.90), ie., GID = Caras |aatid = fodns dea 80 VeRO idaubK MLK) (2.93) 1s Often referred to as the physicists’ notation. Note that the complex con- Jugated spin orbitals appear side by side on the left and the space-spin coordinate of electron-one appears first. It is clear from this definitic a that Cij|kI> = Cll tk> (2.94) and that * (2.95) Because two-electron integrals often appear in the following combination, we introduce a special symbol for an antisymmetrized two-electron integral Gil |kld = Gi|KLy — = fxr des Pong oavill — Pduone) 296) where ¥,, is an operator which interchanges the coordinates of electron one and two. Note that =0 (2.97) It is an unfortunate fact of life that there is another notation for two- electron integrals over spin orbitals in common use, particularly in the literature of Hartree-Fock theory. This notation. often referred to as the chemists’ notation. is Li|AI] = fax, ds x800,)x,00s)rid xhlwabtte,) (2.98) Note that in this notation spin orbitals, which are functions of the coordi- nate of electron-one, appear side by side on the left and the complex con- jugated spin orbital appears first. By interchanging the dummy variables of integration, one has (alk = [eri] (2.998) In addition, if the spin orbitals are real, as is almost always the case in molecular Hartree-Fock calculations, one has Cail AC] = Cai] 0] = Gai] = [ae] (2.996)68 MODERN QUANTUM CHEMISTRY Table 2.2 Notations for one- and two-electron integrals over spin Gp and spatial (i) orbitals SPIN ORBITALS. LIP] = Calif d> = flax, xPOeyherVg 04) depends on the degree to which the two determinants |K> and |L) differ. We can distinguish three cases. The first, Case 1, is when the two determinants are identical, i.c., the matrix element is a diagonal matrix element . For this case, we choose the deter- minant to be |K>= [o> Yuta > (2.104) The second, Case 2, is when the two determinants differ by one spin orbital, %m in |K being replaced by y, in |L). IL> =] Xen > (2.105) The third, Case 3, is when the two determinants differ by two spin orbitals. %m and x, in |K> being replaced by x, and y,, respectively, in |L), E> =) Xp > (2.106) When the two determinants differ by three or more spin orbitals the matrix element is always zero. Tables 2.3 and 2.4 summarize the rules for the three cases. Note that the larger the difference in the two determinants, the simpler is the matrix element, i.¢., the fewer number of terms it involves. The one-electron matrix elements are zero if the two determinants differ by nwo or more spin orbitals, in the same way that the two-electron matrix elements are zero if the two determinants differ by three or more spin orbitals. In the tables, m and n denote spin orbitals occupied in \K ), so that sums over these indices include all N spin orbitals in that determinant. To use the rules, the two determinants must first be in maximum coin- cidence. Consider, for example, a matrix element between |'¥;> and |?2)