Atomsin Molecules

A QuantumTheory

RichardF. W. Bader
D epartmentof C hemistry,
McMaster (lniuersity,
Ontario, Canada

CLARENDON PRESS OXFORD

 Oxford Llniversitl, Press, Walton Street, Oxford OX2 6DP
OxJbrd l{ew York
Athens Auckland Bangkok Bombay
Calcutta Cape Tow'n Dar es Salaam Delhi
Florence Hong Kong Istanbul Karachi
Kuala Lumpur ll[adras Madrid Melbourne
Meric'o City Nairobi Paris Singapore
Taipei Tokyo Toronto
and associatedcompanies in
Berlin lbadan

Oxford is u trade mark of Oxford Llniversitlt Press

Published in the Lrnited States by

Orford Llniversity Press Inr'., New' York

C) Richard F. W. Bader, 1990

First published 1990

First published in paperback (w'ith corrections) 1994
Reprinted 1995

All rights reserved.No part o.f this publication ma.v be

reproduced, stored in a retrieval system, or transmilted, in any
.form or by an"vnleans, v"ithout the prior permission in v:riting o.f OrJbrd
lLniversity Press. Within the LIK, erceptions are allowed in respect o.f anlt
fair dealing Jbr the purpose of research rtr private ,studv-,or criticism or
reviev',us permitted under the Copyright, Designsand Patents Act, 1988,or
in the c'u,seof reprogruphit' reproduc'tion in accrtrdancewith the terms of
licences issued by the Copyright Licensing Agenc.v-.Enquiries conL'erning
reproduction outside those terms and in olher counlrie:; should be sent to
the Rights Depurtment, Oxford Universit,- Press, at the address above-

This book i,ssold subiect to the condition that it shall not,

by wa1'-o.f'trade or otherwise, be lent, re-sold, hired out, or otherwise
t'irculated withoul the publisher's prior consent in an-v-form o./ binding
or cover other than that in n'hich it is puhlishetl and w'ithout a similar
condition including thi.st'ondition being imposetl
on the subsequentpurchaser.

British Library' Cataloguing in Publicution Data

Bader. Richard F. W.
Atom.g in molecules: o quantum lheorv.
l. Molecule,s.Strut'ture. Quantum theor,v
I. Title II. Series
s41.28

Lihrary of'Congres.t Cataktging in Publication Datu

Batler, Richard F. W., 1931
Atonts in molecules; a quantutn theor.t'/RicltardF. W. Buder.
(The International serie.son monogruphs on chemistr.v;22)
and inde.r.
Includes bibIiographit'uI re.ferent'e,s
I. Quantunt chentistry. I. Titlc. II. Series.
QD462.2.833 1994 511.2'B-dc20 9-146241
L981/ 0 19 855865I ( Pbkt

Printed in Great Britain bv

Bookcra/t (Buth) Ltd, Midsomer Norttttr. ltttrt

K - 0 2 0 8 5 8 3
 P R E F A CE

Scienceis basedupon experimentand observation.Its purposeis to enableus

to predict the natural eventsthat occur around us and to understandthem as
best we can. The purpose of theory is to provide the conceptualframework in
this questfor the predictionand understandingof observation.Chemistryhas
been in a unique position in this regard. The molecular structure hypoth-
esis-that a moleculeis a collectionof atoms linked by a network of bonds-
was forged in the crucible of nineteenthcentury experimentalchemistry.It
has continued to serveas the principal means of ordering and classifyingthe
observationsof chemistry. There can be no question from the experimental
point of view but that our ability to understandchemicalbehaviouris a result
of the structure imposed on a system and its properties by the presenceof
atoms.The difficulty with the molecularstructurehypothesisis that it is not
related directly to the physicswhich governsthe motions of the nuclei and
electronsthat make up the atoms and the bonds. We have in chemistryan
understandingbased on a classificationschemethat is both powerful and at
the same time, becauseof its empirical nature, limited. What remains is to
show that the atoms in a molecule are real, with propertiespredicted and
defined by the laws of quantum mechanics,and that the structure their
presenceimparts to a molecule is indeed a consequenceof the underlying
physics.By demonstrating that the molecular structure hypothesishas a
theoretical basis, the classificationbased upon the concept of atoms in
moleculesis freed from its empirical constraintsand the full predictive power
of quantum mechanicscan be incorporated into the resulting theory-a
theory of atoms in molecules.
It is the purpose of this book to demonstrateto the reader that such a
theory is obtained when the definition of a quantum subsystem,as derived
from a fundamental principle of physics, is applied to the observed topo-
logical properties of a system'sdistribution of charge. An atom is uniquily
defined as an open system in real space, free to exchange charge and
momentum with its neighbours. Its properties are described by the same
equationsof motion and associatedtheoremsof quantum mechanics,such as
the virial and Ehrenfest theorems, as apply to the total system. The same
topological propertiesof the chargedistribution lead in turn to a definition of
the bonds which link the atoms, and to definitions of structure and structural
stability for molecular systems.The theory makes possiblethe identification
of all possiblestructuresfor a given systemand determinesthe mechanismsof
structural change within the system.
The topology of p, the chargedensity,as displayedin the global properties
of its gradient vector field, yields a faithful mapping of the chemicai .on..pt,
 PREFACE

of atoms, bonds,and structure.The propertiesof this field however,provide

no indication of maxima in p correspondingto the electronpairs of the Lewis
model, a model secondary only to the atomic hypothesis itself in our
interpretation of chemicalreactivity and molecular geometry.The physical
basis of this most important model is one level of abstraction above the
visibletopology of the chargedensityand appearsinsteadin the topology of
the Laplacianof p. This function is the scalarderivativeof the gradientvector
field of the chargedensityand it determineswhereelectronicchargeis locally
concentrated and depleted, the local charge concentrations providing a
mapping of the electron pairs of the Lewis model. The Laplacian of the
chargedensityplays a dominant role throughout the theory,appearingin the
constraint determining the boundary of a quantum subsystemand relating
the spatial propertiesof the chargedensity to the local contributions to the
energy.
The whole of the theory,its descriptionof atoms,bonds,and structureand
its stability,togetherwith the discussionof moleculargeometryand reactivity
in terms of the Laplacian of p, is developedin a continuousmanner from the
initial quantum definition of an atom. At no point in this developmentis the
readeraskedto acceptthe introduction of arbitrary assumptionsor personal
points of view.
The properties of atoms in moleculescan be determinedexperimentally
and quantum mechanics,through the theory of atoms in molecules,predicts
thesepropertiesjust as it predictsthe propertiesof the total system.It comes
as no surpriseto a chemistto know that thereis a carbonyl group, a fragment
of real spacewith a set of definableproperties.The energiesof individual
atoms can be calculatedand the mechanismof energystoragein a biological
moleculecan be determinedand understood at the atomic level.There are
situations when functional groupings of atoms are transferablebetween
systems,carrying with them essentiallyconstant contributions to the energy
and other properties of the systemsin which they occur. This is a known
experimentalresult of chemistry and theseare the groups identified by the
theory of atoms in molecules.
This book presentsan account of the theory of atoms in moleculesto the
scientistswho are responsiblefor performing the experimentsand collecting
the observationson the propertiesof matter at the atomic level.This is done
in the beliefthat the transformationof qualitativeconceptsinto a quantitat-
ive theory will serveto deepenour understandingof chemistry.This consti-
tutesa wide audienceand the book is structuredaccordingly.The material in
each chapter is presentedin such a manner that a reader desiring a general
understandingof its content may do so by reading less than the complete
chapter.The reader desirousof a fuller understandingof the mathematical
and physicalstructureof the theory attainsthis goal by a completereadingof
the chapter.At no point or stageis rigour sacrificedfor easeof presentation,
 PREFACE vii

only the depth of understanding paralleling the chapter depth. Topics

consideredto be important but not essentialto the first readingof the text are
referencedat the appropriate place in the text and placed at the end of the
chapter in which they occur.
Tobermory,Ontario R.F.W.B
Oc t ober 1989.

This book is dedicatedto the membersof my

researchgroup)past and present,who haue
contributedto the deuelopment of the theory
of atoms in molecules.
 CONTENTS

List of Symbols xiv

1. ATOMS IN CHEMISTRY 1
I

1.1 Atoms and the molecular structurehypothesis 1

1.2 Necessarycriteria for a theory of atoms in molecules 2
1.3 The role of the charge density 5
E l.l D e n s i t ym a t r i c e s 11

2. A T O M S A N D T H E T O P O L O G Y O F T H E
CHARGE DENSITY 13
2.1 Introduction 13
2.2 Topological properties of the charge density l4
2.2.1 The dominant form in the chargedensity t4
2.2.2 Critical points and their classification l6
2.2.3 Critical points of molecular chargedistributions I9
2.3 Gradient vector field of the charge density 22
2.3.1 Trajectories of the gradient vector field 22
2.3.2 Phase portraits of the gradient vector field 25
2.4 Elementsof molecular structure 28
2.4-l Equivalenceof the topological and quantum definitions
of an atom 28
2.4.2 Chemical bonds and molecular graphs 32
2.4.3 Rings and cages 35
E2.l Local properties of p and associatedtheorems 40
E2.l.l Local maxima in p 40
82.1.2 Theorems concerning the properties of p 45
E2.2 Mathematical properties of p at a critical point 47
E2.2.1 Eigenvaluesand eigenvectorsof Hessian of p 47
E2.2.2 Analytical expressionfor a trajectory in the vicinity of
a critical point 49

3. M O L E C U L A R S T R U C T U R E A N D I T S C H A N G E 53
3.1 The notion of structurein chemistry 53
3.2 The definition of molecular structure 54
3.2.1 The equivalencerelation-a qualitative discussion 54
3.2.2 Changesin structure 58
3.2.3 Structurediagrams 64
3.2.4 Bonds and structure ^)
3.3 A theory of molecular structure 87
3.3.1 A coming together of mathematicsand chemistr-l 87
3.3.2 The equivalencerelation and structural stabilitv 88
 CONT ENTS

3.3.3 Stable and unstable intersectionsof lines and surfaces

in R3 9l
3.3.4 Criteria for structural stability 94
3.4 Comparison of a structure diagram with other partitionings of Ra 95
3.4.1 Relation between topological and energeticstabilities
of molecular structures 95
3.4.2 Structural homeomorphism between p and the nuclear
potential 97
3.4.3 The mechanicsof p in the one-electroncase 101
3.5 Summary 102
E3.1 Summary of set theoretic language 103
E3.2 Natural coordinate system for an atomic basin 103

4. MATHEMATICAL MODELS OF STRUCTURAL

C H A N GE 110
4.1 Introduction 110
4.2 Catastrophe theory 111
4.2.1 Isolating the unstable piece of a critical point 111
4.2.2 Elementary catastrophes tt2
4.3 Catastrophesin molecular structures 113
4.3.1 Opening a ring structure--the fold catastrophe 113
4.3.2 A general analysis of three-centresystems 115
4.3.3 Formation of a case structure 12l

5. T H E Q U A N T U M ATOM 130
5.1 Chemistry and quantum mechanics 130
5.1.1 From Lewis to quantum mechanicalmodels 130
5.1.2 The role of the charge density in defining structure r32
5.2 Need for a quantum definition of an atom r33
5.2.1 Observational basis for a quantum atom 133
5.2.2 Observablesand their properties for a total system t37
5.3 Need for a subsystemvariation principle 146
5.3.1 Schrodinger'sderivation of wave mechanics 150
5.3.2 The variational definition of a subsystemand its
properties 154
E5.1 The variation of Hamiltonian-based functionals r6l
E5.2 Vectors, tensors,and dyadics 164
E5.3 Divergence of a vector and Gauss'stheorem r66

6. T H E M E C H A N I C S O F A N A T O M I N A
M O L ECULE 169
6.1 An atomic view of the properties of matter r69
6.1.1 The chargeand current densities r69
6.1.2 Variational derivation of the atomic force law 172
 CONTENTS

6.1.3 The atomic virial theorem 175

6.1.4 Correspondencebetweenlocal and subsystemmechanics 178
6. 2 Atomic properties 179
6.2.1 Single-particlebasisfor atomic properties 179
6.2.2 Definition of atomic properties 180
6.3 Energy of an atom in a molecule 185
6.3"1 Physicalconstraintson partitioning the energy 185
6.3.2 The virial and the partitioning of an energyof interaction 185
6.3.3 The electronicenergyof an atom 189
6.3.4 Potential energycontributions to an atomic energy 191
6.4 Propertiesof atoms in molecules 195
6.4.1 Properties determined by the electronic charge density 196
6.4.2 Transferability of atomic properties 209
6.4.3 The origin of strain energy in cyclic hydrocarbon molecules 219
6.4.4 Origin of rotation and inversion barriers 221
6.4.5 Perfect transferability-an unattainable limit 235
6.5 The hypervirial theorem and the definition of bond properties 238
6.5.1 An atomic property expressedas a sum of bond
contributions 239
6.5.2 Surfaceintegrals proportional to the energy 240

7. CHEMICAL MODELS AND THE LAPLACIAN

OF THE CHARGE DENSITY 248
7.1 The physicalbasisof the Lewis electronpair model 248
7.1.1 The electronpair at any cost 248
7.1.2 Are electronslocalizedin pairs? 249
7.1.3 Propertiesof the Laplacian of the chargedensity 252
7.1.4 The valenceshell chargeconcentration(VSCC) 258
7.1.5 Bonded and non-bondedchargeconcentrations 260
7.2 A physical basis for the VSEPR model of molecular geometry 265
7.2.1 Resumeof the VSEPR model 265
7.2.2 Valence-shellcharge concentrations with four maxima 266
7.2.3 Valence-shellcharge concentrationswith five and six
maxima 268
7.2.4 Effects of charge transfer on the properties of the VSCC 274
7.2.5 Summary 275
7.3 The Laplacian of p and chemical reactivity 275
7.3.1 The Laplacian and the local energy of p 275
7.3.2 Lewis acid-base reactions 277
7.4 The characterizationof atomic interactions 288
7.4.1 Definition of atomic interactions 288
7.4.2 Classificationof atomic interactions 290
7.4.3 Hydrogen bonds and van der Waals molecules 302
7.4.4 Intermediate interactions 307
7.5 Atomic interactions in bound and unbound states 315
7.51 The electrostatictheorem and chemical bindins -'il)
xll CONTE NT S

7.5.2 The virial theorem and chemical bondins 322

7.5.3 A summary 331
E7.l The pair density and the localization of electrons 332
87.2 Local charge concentrationsand partial localization of the
Fermi hole 343

8. T H E A C T I O N P R I N C I P L E F O R A Q U A N T U M
SUBSYSTEM 352
8.1 A common basis for classicaland quantum mechanics 352
8.1.1 Introduction 352
8.1.2 State vectors. state functions. and transformation functions 353
8.1.3 Unitary transformations 359
8.1.4 Canonical transformations and classicalmechanics 361
8.2 The quantum action principle 369
8.2.1 The principle of stationary action 369
8.2.2 Applying the principle of stationary action 371
8.3 Atomic action principle 376
8.3.1 Properties of the quantum mechanical Lagrangian 376
8.3.2 Atomic action and Lagrangian integrals 378
8.3.3 Variation of the atomic action integral 380
8.3.4 Principle of stationary action in the Schrodinger
representatlon 382
8.3.5 Atomic statement of the principle of stationary action 384
8.3.6 Examples of unitary transformations 391
8.4 Variational basis for atomic properties 393
8.4.1 Variational derivation of the atomic force law 393
8.4.2 Differential force law 396
8.4.3 Variational derivation of the atomic virial theorem 398
8.4.4 Local virial relationship 401
8.4.5 Summary of the atomic variation principle 402
8.5 Atoms in an electromagneticfield 403
8.5.1 The Lagrangian and Hamiltonian 403
8.5.2 Definition of an atom in an external field 406
8.5.3 Atomic force and virial theorems in the presenceof
external fields 4t0
8.5.4 Atomic contributions to molecular polarizability and
susceptibility 4t6
8. 6 The action principle-in the past and in the future 422

APPENDIX 427
TableAl Units 427
Table A2 Contour valuesfor mapsof the electronicchargedensityand of
the Laplaciandistributionfunction 427
 CONTENTS xiii

Table 43 Properties of p at (3, - 1) critical point together with bonded and

non bonded radii for ground-state diatomic molecules 428
Table 44 Atomic charges,moments, and volumes of ground-state
diatomic molecules 430
Table A5 Radii of isolated ground-state spherical atoms and ions 432

INDEX 433
 LIST OF SYMBOLS

Coordinates

r position vector
rc position vector of critical point in electronic charge density
x the spaceand spin coordinates of a set of electrons
rj position vector of electron j
ro electronic position vector with origin at nucleus of atom Q
dr volume element for a single electron
V gradient with respectto coordinates of an electron
Ra nuclear configuration space
X a point in Ra defining a molecular geometry, the set of nuclear
coordinates
Xo position vector of nucleus e
X" catastrophe point in Ra
Xb bifurcation point in Ra
V* gradient with respectto nuclear coordinates

I dt' a summation over all electronic spin coordinates and the

integration over the spacecoordinates of all electronsbut one,
e q n s( 1 , 3 )a n d ( 1 . 4 )

$ dS1O,r; integral over surfaceof atom Q

Vectors and dyadics

i, j, k unit vectors in x, y, z directions

A vector with componentsA,, Ay, A", eqn (85.11)
A 'B scalar product of A and B, eqn (E5.13)
A x B vector product of A and B
AB dyadic with nine components,eqn (E5.14)

Charge density and related quantities

p(r;X), p(r) electronicchargedensity,eqn (1.3)

pa value of p (r) at bond critical point
p, value of p(r) at a ring critical point
p'(r) density of a spin electrons
Vp(r) gradient vector of p, eqn (2.1)
s path parameter of trajectory of Y p
n(r) unit vector normal to surface
^i eigenvalueof the Hessian matrix of p at a critical point
 LIST OF S}'\,1BOLS xv

ui eigenvectorof the Hessianmatrir of p at a critical point

6 signatureof critical point in p
u) rank of critical point in p
E bond ellipticity
Y tp (r) Laplacian of p, eqn (2.3)
d.b bond path angle
ae geometric bond angle
Rb bond path length
R" geometric equilibrium bond length
n bond order

11.'?: statefunctions
for time-dependent
systems
Y(x,X, r), )J
V (x), rl,(x; X) state functions for stationary states

Functionals

Y ltl,l Schrodinger'senergy functional, eqn (5.54)

Ebll Schrodinger'sconstrained energy functional, eqn (5,58)
I lrltl stationary state energy functional, eqn (5.63)
ELrlt,O] constrainedatomic functional,eqn (5.72)
Slr!,Ol stationarystate atomic energyfunctional,eqn (E5.1)
.f bL,O) Hamiltonian energy functional for an atom, eqn (E5.4)
%'rrlYf action integral, eqn (8.99)
9lY,tf Lagrangeintegral,eqn (8.99)
,trlYl Hamiltonian energy integral
7l't2lY, (l) atomic action integral,eqn (8.111)
9lY,Q,tf a t o m i c L a g r a n g ei n t e g r a l ,e q n ( 8 . 1 1 0 )
{rrlY, A, Of action integral for external fields, eqn (8.210)
glY, A, Q,tf Lagrangeintegral for externalfields,eqn (8.215)

Operators
t,, action integral operator
s ltl Lagrange function operator
q coordinate operator
r electronic position operator
B momentum operator
fr momentum operator in magnetic field
F(r) generatorof infinitesimalunitary transformation.eqn (8.87)
H Hamiltonian operator
fr" electronicHamiltonian, eqn (6.53)
F* force operator for particle rc
XVI LIST OF SYMBOLS

i electronic kinetic energy operator

- f r kinetic energy operator in magnetic field, eqn (8.226)
0 unitary transformation operator. eqn (8.27)
V potential energy operator
i)
' en electron-nuclear potential energy operator, eqn (6.9)
t"" electron-electron potential energy operator, eqn (6.9)
i) nuclear-nuclear potential energy operator, eqn (6.9)

Classical mechanics

q position
Q velocity
p momentum
W' action integral, eqn (8.43)
L(q, q, t) Lagrangian
H (p, q, t) Hamiltonian
G(p, q) generator of infinitesimal canonical transformation, eqn (8.61)
{F, G} Poissonbracket of F and G

Property and energy densities

d (r) polarizabilitydensity,eqn (8.258)

,'ttt 1r,r') first-orderdensity matrix, eqn (E1.1)
f (2)(r,,rr) diagonal elementof second-orderdensity matrix, eqn (E1.3)
p (r r , r z) pair density,eqn (E7.7)
E.(r) electronic energy density, eqn (6.70)
F(r) Ehrenfestforce density, eqn (6.16)
c(r) kinetic energy density, eqn (5.49)
G,t(r.) component of kinetic energy density perpendicular to bond
path at r"
G11(r.) component of kinetic energy density parallel to bond path
at r.
i (r) vector current density, eqn (5.94)
d,r,i(r) infinitesimal change in j(r) caused by variation in r!, eqn (5.95)
io(r) vector current densityfor generatord, eqn (5.98)
i"(r) vector current density in presenceof magnetic field, eqn (8.220)
J'(r) vector current densityfor property F, eqn (8.140)
Ji(r) diamagnetic current density for property F, eqn (8.222)
K(r) kinetic energydensity,eqn (5.49)
l,(Y, VY, \P, r) many-particleLagrangedensity,eqn (8.100)
9o(r,t) Lagrangedensity at point of variation, eqn (8.107)
m(r) magnetizationdensity,eqn (8.272)
 LIST OF SYN{BOLS xvii

PaF) density for property ,4. eqn t6.-15t

F(r) quantum mechanicalstresstensor.eqn (5.28),(6.12)
Tr F (r) trace of the stresstensor,eqns (5.32)and (6.26)
N''(r) virial density,eqn (6.30)

Atomic properties

do(Q) basin contribution to atomic polarizability,eqn (8.260)

d.(O) surfacecontribution to atomic polarizability,eqn (8.259)
,4(OlO') contribution to A.4(O) from interatomic surfacebetweenatoms
Q and f)', the correspondingbond contribution, eqn (6.102)
D2 (O, Q) averagenumber of electron pairs in atom O, eqn (87.17)
E"(O) electronicenergy,eqn (6.73),(6.74)
F(O) Ehrenfestforce,eqn (6.14)
F"(O) Hellmann-Feynman force exerted on electronic charge in atom
Q by nucleusa
f (O, O) electron correlation in Q
G(C)) averagekinetic energy,eqn (5.50)
,K(O) averagekinetic energy,eqn (5.50)
t (O) atomic Lagrangian,eqn (6.88)
/ (C)) percent localization of electrons,eqn (87 .23)
n,(I{, C)) fluctuation in averageelectron population of Q, eqn (E7.9)
M(O) atomic first moment, eqn (6.45)
lV (O) averageelectron population, eqn (6.42)
P,(O) probability of n electronsbeing in O, eqn (E7.1)
Q,"(Q) component of quadrupole moment tensor,eqn (6.48)
q (O) charge on atom O, eqn (6.43)
ru (O) bonded radius
r" (Q) non-bonded radius
S (Q),S (O, r) atomic surface
S(O|O') interatomic surfacebetweenatoms Q and Q'
f(C)) averageelectronic kinetic energy
f,, (O) averageelectronic kinetic energy in magnetic field
2."(Q,O) contribution to V""(O)from electronswithin atom Q
V""(Q,Q') contribution to \,(O) from repulsion of electronsin O with
those in O'
4"(Q), V^(Q) electron-nuclearattractiveenergyfor atom Q. eqn (6.80)
23.(ff)),V?(O) contribution to V"n(O)from interaction of the electronic
chargedensity in the basin of atom Q u'ith its orin nucleus.
eqn(6.81)
4.(O) nuclear-nuclearrepulsionenergy for atom Q. eqn t6.36t
XVIII LIST OF SYMBOLS

u (O) atomic volume

Y'(Q) atomic virial, eqn (6.24)
tr'o(Q) basin virial
t''s$l) surfacevirial, eqn (6.21)
l'^ (Q,8,6\ atomic virial in externalfields
.{t(O) basin contribution to atomic magnetic susceptibility,
eqn (8.277)
,fs(A) surfacecontribution to atomic magnetic susceptibility,
eqn (8.277)

Total properties
d, polarizability tensor
A vector potential of electromagneticfield
B magnetic field vector
I electric field vector
e elementary charge
0 scalar potential of electromagneticfield
D" dissociation energy
E total energy, eqn (6.49)
E" total electronic energy,egn (6.53)
En total nuclear energy,eqn (6.55)
F" Hellmann-Feynman force on nucleus fi, eqn (5.37)
Fo. electronic contribution to force on nucleus I
F"n nuclear contribution to force on nucleus a
F." Hellmann-Feynman force exerted on electrons by nucleus a
m mass of electron
m magnetic dipole moment vector
N total number of electrons
T averagekinetic energy
Ttl component of kinetic energy parallel to chosen axis
TI component of kinetic energy perpendicular to chosen axis
V, electronic potential energy,eqn (6.54)
p dipole moment vector
T magnetic susceptibility,eqn (8.273)
W,, component of energy-momentumtensor,eqn (8.176)
Z nuclear charge
 1

ATOMS IN CHEMISTRY

I have been enabled . . . to exhibit a new view of the first principles or elementsof
bodies and their combinations,which, if established,as I doubt not it will in time, will
produce the most important changesin the system of chemistry.
John Dalton (1807)

1.1 Atoms and the molecular structure hypothesis

Chemistry is the study of the properties of substancesand their trans-

formations. Since substancesare identified and distinguished one from
another by the elementsthey contain and sincethe propertiesof each element
are determinedby the atoms unique to that element,chemistry is the study of
matter at the atomic level. The realization of this fact by Dalton in 1807
marked the birth of chemistry as a branch of science.Dalton's atomic
hypothesis accounted for the existenceof the observed relative combining
weights of the elementsand predicted a new relationship among them, the
law of multiple proportions. Dalton did not simply resurrectthe Greek idea
of atoms; he postulated that atoms retained their identity even when in
chemical combination with other atoms-that atoms exist in molecules.As a
result of the work of van't Hoff and Le Bel in 1874the geometrical aspectof
molecular structure was incorporated into the hypothesisby postulating the
existenceof preferred geometrical arrangementsof bonds terminating at a
single atom. Before the close of his century, there had evolved from Dalton's
atomic hypothesis the concept of molecular structure-that a molecule is a
collection of atoms linked by a network of bonds, the bonds imparting the
structure.
The primary purposein postulating the existenceof atoms in moleculesis a
consequenceof the observation that atoms or groupings of atoms appear to
exhibit characteristic sets of properties (static, reactive, and spectroscopic),
which, in general, vary between relatively narrow limits. In some series of
molecules,the variation in the properties is so slight that a group additivity
scheme for certain properties, including the energy, can be established.
Dalton defined the first of these atomic properties by assigning to each
element a relative combining weight. As the body of descriptive chemistry
grew, it was reahzed that the chemistry of some total system could be
rationalized by assigning separate properties to each type of atom or
functional group in the system.Thus, the knowledge of chemistrf is ordered,
classified,and understood by assigningproperties to atoms and to functional
 ATOMSIN CHEMISTRY r.2
groupings of atoms, propertieswhich can then usedto identify the presenceof
a given group or to understand and predict the behaviour of some total
system.
It was a bold step that Dalton took in assuming that each type of atom
possesseda particular mass whose value was independent of its chemical
state.Why should an atom when in combination with any other atom always
possessthe samemass?The vindication of Dalton's assumption had to await
the results of Rutherford's alpha-scatteringexperimentsand his proposal of
the nuclear atom. In this model, essentiallyall of the atomic mass is found in
the chemically stable nucleus.It also placesall of the positive charge in the
nucleus,thereby stamping each atom with a chemical identity. Becausethe
positive charge is concentrated at a point within the atom, the
nuclear-electron attractive force is the single most dominant force in a
molecular system.As a consequence,a nucleusplays a unique role, that of an
attractor, in determining the topological properties of a molecular charge
distribution-a role which leads directly to the definition of an atom. In this
manner the chemical identity of an atom is imposed on the electroniccharge
distribution.
It is the purpose of this book to demonstrate that the existenceof atoms
with definable properties and the associated concepts of the molecular
structure hypothesisare a consequenceof the quantum description of matter
as applied to the properties of the electroniccharge distribution. In so doing,
the hypothesisis transformed into a theory and the complete description of
matter afforded by quantum mechanicscan be applied to the study of atoms
in molecules.

1.2 Necessary criteria for a theory of atoms in molecules

The definition of a bound atom-an atom in a molecule-must be such that

it enablesone to define all of its averageproperties. For reasonsof physical
continuity, the definition of these properties must reduce to the quantum
mechanicaldefinitions of the correspondingproperties for an isolated atom.
The atomic values for a given property should, when summed over all the
atoms in a molecule, yield the molecular average for that property. The
atomic properties must be additive in the above senseto account for the
observation that, in certain seriesof molecules,the atoms and their properties
are transferablebetweenmolecules,leading to what are known as additivity
schemes.An additivity schemerequires both that the property be additive
over the atoms in a molecule and that the atoms be essentiallytransferable
between molecules.
Two identical piecesof matter possessidentical properties.This elementary
fact should extend down to the atomic domain and requires that atoms be
defined in real spaceso that, if an atom is identical in two different systemsor
1,2 CRITERIAFOR A THEORY OF ATOMS 3

at different siteswithin a given system(e.g.in a solid),then it must contribute

identical amounts to the total properties of the systemsin which it occurs.
Atoms are objects in real space.Theory definesthem through a partitioning
of real space as determined by the topological properties of a molecular
charge distribution, that is, by its form in real space.The constancy in the
properties of an atom of theory, including its contribution to the total energy
of a system, is observed to be directly determined by a corresponding
constancyin its distribution of charge.When the distribution of charge over
an atom is the same in two different molecules,i.e. when the atom or some
functional grouping of atoms is the same in the real space of two systems,
then it makes the same contribution to the total energy and other properties
in both systems (Bader and Beddall 1972). It is because of the direct
relationship between the spatial form of an atom and its properties that we
are able to identify them in different systems.The reader is referredto Fig. 1.t
for an illustration of such transferablefunctional groups.
There is in this discussionno implication that atoms are always essentially
transferable.It is simply that, in the limit of an atom being transferable
between systems,the relationship between the form and properties of an
atom is most evident.Of equal importance, this limit provides a stringent test
of the theory. The properties of atoms in molecules can be experimentally
measured and, when the atoms are transferable, these properties can be
determined with impressive precision. Rossini and co-workers at the
National Bureau of Standardshave determined that the incremental increase
in the standard heat of formation of a normal hydrocarbon per methylene
group is -20.6 + 1.3kJ/mol. The theory of atoms in moleculesnot only
definesa methylenegroup whose distribution of chargein real spaceup to its
atomic boundaries is transferable between hydrocarbon molecules,but, in
addition, defines the energy of this group. This energy is as constant and
transferableas is the group itself and it accountsfor the observedadditivity of
the energy in this seriesof molecules.
Following on this observation, if, as is usually the case, the charge
distribution of a group does changewhen it is transferredto another system,
then its contribution to the total energy will be different and one may relate
its changein energy to this changein its distribution of charge as induced by
the new environment. The ultimate justification for identifying the atoms of
theory with the atoms of chemistry is this very observation-that their
propertieschangein direct proportion to their changein form. Thus, whether
the charge distribution of an atom changesby a little or by a lot, its energy
and other properties change by corresponding amounts. Having found the
transferable methylene group, for example, one may inquire as to how its
properties change when it is introduced into the cyclopropane molecule and
subjectedto geometricstrain.As a result of the geometricconstraintspresentin
the cyclopropanemolecule,a small amount of electronicchargeis transferred
 ATOMSIN CHEMISTRY 1.2

@ffi@@

@ j

Ftc. 1.1. Shapesof moleculesrepresentedby envelopesof constant electronic charge density.

The envelopeshown has the value of 0.001 au. The moleculesare:(a)-(f ) the normal alkanesfrom
methane to hexane;(g)isobutane;(h)neopentane;(i)cyclopropane;(j) cyclobutane;(k)formalde-
hyde, HrC:O; (/) acetone,(CH.)rC:O. The intersectionsof the 'zero-flux' interatomic surfaces
with the envelope are shown in some cases.They define the methyl, methylene,hydrogen, and
carbonyl groups. The isobutane molecule (g), for example,exhibits three methyl groups topped
by a hydrogen atom.

from hydrogen to carbon. This relatively small change in the distribution

of charge of the standard methylene group causes a correspondingly
small increaseof - 38 kJ/mol in its energy,the so-calledstrain energy.The
experimental heat of formation of cyclopropane when compared to three
times the energv of the standard methylene group shows an energy increase
1.3 THE ROLE OF THE CHARGE DENSITY

which equals the theoretically predicted strain energy. This is another

example of the atoms of theory recovering the experimental value of a
chemically important energy.
It is a well-establishedprocedure to use quantum mechanicsto relate a
spectroscopicallymeasuredenergy to the theoretically defined differencein
energiesbetweentwo statesof a system.In a lessdirect but no lessrigorous
manner, the same quantum mechanicsrelates the differencein the heats of
formation of butane and pentane to the theoretically defined energy of the
methylenegroup. It is the purpose of the atoms in moleculesconcept to relate
molecular properties to those of its constituent atoms, whether these are
strictly transferablecontributions or, as is more generally the case,physical
or spectroscopicproperties or modes of reaction that are characteristicof a
givengrouping of atoms within the molecule.What was lacking in chemistry
was the theoretical framework necessaryfor the incorporation of these
chemical observations into the fabric of quantum mechanics so that they
could be treated along with the properties of the total system by a single
theory. There is only one test of the validity of any theory and that is
comparison with observation.The ability of the theory of atoms in molecules
to recover measuredproperties is proper justification that atoms do indeed
exist in molecules as separatedefinable piecesof real space and that their
propertiesare predicted by quantum mechanicsas describedin this treatise.

1.3 The role of the charge density

'are
The question there atoms in molecules?'is equivalent to asking two
equally necessaryquestions of quantum mechanics:does the state function
Y(x, r), which contains all of the information that can be known about a
system,predict a unique partitioning of a moleculeinto subsystemsand, if so,
can one define the observables,their averagevalues,and their equations of
motion for the subsystem?Affirmative answersto both these questions are
what is required to establishthe physicsof an atom in a molecule within the
framework of quantum mechanics.Such questions must be posed within a
completely general development of quantum mechanics and, as demon-
stratedin Chapters 5 and 8, affirmative answersare indeed obtained. A most
important result of this demonstrationis that the quantum subsystemsare
open systemsdefined in real space,their boundaries being determined by a
particular property of the electronicchargedensity (Bader and Beddall 1972).
Thus, while the theory has its origin in quantum mechanics,its vehicleof
expressionis the charge density. It is for this reason that we begin with a
study of the electronicchargedensity.Accepting from theory the statementof
the quantum boundary condition, this initial study will lead directli' to the
definition of atoms. bonds. molecular structure. and the mechanismsof
 ATOMS IN CHEMISTRY 1.3

structural changeand will provide the reader with the knowledge required to
apply the theory to chemical problems.
The state function Y determinesall of the information that can be known
about a quantum system.If one desiresa theory of molecular structure that is
free of arbitrary or subjective assumptions, no information beyond that
contained in Y should be used in its development.The state function for a
molecular systemis a function of the electronic and nuclear coordinatesand
of the time f, Y(x, X, r), where x denotesthe collection of electronicspaceand
spin coordinates and X the collection of nuclear coordinates. While the
generaltheory appliesto the time-dependentcase,we are primarily interested
in systemsin stationary states whose properties do not change with time.
Denoting a solution to Schrodinger'sstationary-stateequation for a fixed
arrangementof the nuclei by /(*; X), the probability of finding each one of l{
electrons in a particular elemental volume dr, - dx,dy,dz, with spin co-
ordinate o, (equal to either the a or B spin coordinate) for a given configura-
tion of the nuclei X is

t*8;X)/(*; X)dx,dr, . . . dr, (1.1)

where dx, : dr,o,. The corresponding probability independent of spin is
obtainedby summingeqn (1.1)over the spin coordinates.If the summingover
all spins is followed by an integration over the spatial coordinates of all the
electrons but one (it matters not which electron is chosen, since r/ is an
antisymmetrizedfunction and hence all electronsare equivalent),the result-
ing expressiongivesthe probability of finding one electron in some elemental
volume, independent of the instantaneouspositions of the remaining elec-
trons,
Probability of one electron being in dr,
: f (spins){Jdr,
Jdt. . . .Jdr"t*8; x)r/(*;X)}dr,. (r.2)
Multiplication of this probability by the number of electrons N gives the
probability of finding any of the electronsin dr, , that is, the total probability
of finding electronic charge in drr. The corresponding probability density,
the probability per unit volume, is called the electrondensity or the electronic
charge density. It is denoted by the symbol p(r; X), the spacecoordinate of a
singleelectronbeing given by the position vector r : ix I jy + kz,
p ( r ; X ) : r VI ( s p i n s ) { J d r r J d t . . . . J d t r / * ( x ; X ) / ( x ; X ) } . (1.3)
The subscript I is suppressedin this equation sincethe result refersto any of
the electrons.The form of integration given in eqn (1.3),which yields the
density of electronicchargein real space,recurs throughout the theory and is
designatedin an abbreviated manner as
p(r;X) - A/J dr't11*(x;
X)/(x; X) (1.4)
1.3 T H E R O L E O F T H E C H A R G ED E N S I T Y

where r' denotes the spin coordinates of all the electronsand the Cartesian
coordinates of all electronsbut one. If one inserts an operator M(r) for some
property between the state functions in eqn (1.4),the result is a real space
density distribution function for that property. All atomic properties are
defined in terms of such density distributions, the atomic average of a
property being obtained by an integration of the correspondingdensity over
the region of spaceassignedto the atom. The charge density is a specialcase
of this more general situation with the operator M equalling the unit
operator and the atomic integration yielding the averagenumber of electrons
for the atom in question.it should be borne in mind that, while one integrates
over the coordinates of the other electrons to obtain the charge density or
property-weighted density, all of the electronscontribute to the form of the
density so obtained. The integration denoted by Jdr' yields a density of one
electron determined by the averagemotion of the remaining electrons.
The analysis of the topological properties of the charge density which
follows can also be applied to the density obtained by an averagingof p over
the motions of the nuclei
p(r;X) : JdXp(r;x)/JdX. (1 . s )
It is the chargedensity averagedover the thermal motions of the nuclei that is
determined in X-ray and electron diffraction experimentsof solids and solid
surfaces.The charge density, whether determined experimentally or calcu-
lated from theory for a single configuration of the nuclei, is a representation
of the time-independent distribution of negative charge throughout three-
dimensional spaceand this is an alternative and equally valid interpretation
of the quantity definedin eqn (1.4)or (1.5).
One can, in principle, perform what are necessarilyinelastic scattering
experimentswhich would determine the probability of finding an electron in
some relatively small region of space,small relative to the dimensionsof the
system to which the electron is bound. This must result in the excitation or
loss of the electron as a consequenceof the uncertainty principle and the
electronic state of the systemis thus changed.Such experiments,leading to a
change in state of the system under observation, are probabilistic in their
outcome and p in this caseis justifiably interpreted as a probability distribu-
tion function.
It is equally possible,however,to measurep in an experiment in which the
electronicstateof the systemis not changedas the scatteringis elasticand the
recoil is to the entire system,the whole of the crystal lattice. for example.
There is no need here for a probabilisticinterpretationof the results.These
experimentsmay be interpreted as providing a measurementof a spatial
distribution of charge. The properties determined directly' b1' the charge
density, such as the electronic moments of a system. are obtained by
averagingthe correspondingoperator over the chargedensitl. an operation
 ATOMS IN CHEMISTRY 1.3

equivalent to the assumption that p provides a description of the static

distribution of electronic charge throughout three-dimensionalspace.This is
the interpretation given to p(r; X) or p(r; X) in this work. Schrodinger(1926)
defined the 'electric density' in his fourth paper as a quantity which, aside
from a summation over the spin coordinates,is identical with the definition of
the electronic charge density given in eqn (1.3).It is interesting to note that
Schrodingerregarded rlr*rltas providing a description of the spreadingout of
electronic charge throughout space.To simplify notation, the symbol p(r)
will normally be used for the charge density, it being understood from the
context that it denotes a charge distribution for a fixed or averagednuclear
configuration.
The charge density p(r\ is the fundamental property measured in a
coherent X-ray scattering experiment, as is commonly employed in the
determination of a crystal structure. This follows since the operator which
describesthe interactions of X-rays with the electrons is, to a very good
approximation (the Born approximation), given by a sum of one-electron
interactions
I;exp ti(k - ko)'r,l (1.6)
where ko and k are the wave vectors of the X-rays incident upon and
scattered by the electron at rj. Because of its one-electron nature, the
averaging of this operator over ry'*ry'reducesto lf times the averageof the
operator for a single electron over the charge density (eqn (l.a)). This
procedure yields the so-calledX-ray scattering factor

, f ( r ) : J p ( r ) e x p[ i ( k - k o) ' r ] d r (r.7)
where s:4nsin?lA. Experimentally,one measuresthe ratio of the intensities
of the scatteredto the incident radiation of wavelength2 as a function of the
scattering angle 0. Such a measurementis proportional to f '. Thus, from
measurementsof scatteredX-ray intensities,one may obtain the scattering
factors and, by their Fourier transform, the charge density, p(r). In X-ray
scattering experiments on crystals, which are the most common sort, the
diffraction pattern is determined by the repeating unit cell, characteristicof
the crystal, and by the particular crystallographic plane (as denoted by the
Miller indices h, k,l) from which the scatteringoccurs.Thus for crystals one
defines a structure factor F (hkl) which is expressedas

F (hkl): L,f, exp(2zih.r;) (1.8)

wheref is the atomic scatteringfactor of theTth atom in the unit cell and r, is
its position. The components of the vector h are hla, klb, and Ilc where a, b,
and c denote the dimensions of the unit cell. The charge distribution at a
point r within a unit cell is then expressedas
p(r) : V -'LoLol,r lntq exp(- 2nih.r) (1.e)
1.3 THE ROLE OF THE CHARGEDENSITY

where Z is the volume of the unit cell.Experimentally,the maxima in p(r) are

found to occur at the positionsof the atomic nuclei and eqn (1.9)can be used
to determine the arrangement of the atoms in a crystal. The mathematical
description of a crystal structure is, therefore,given by a Fourier serieswhich
representsthe electron density distribution. The techniquesneededto over-
come associatedproblems,such as thermal smearingof the nuclear positions,
are being constantly refined and eqn (1.9) is being increasingly used to
determine the charge density p(r) itself from experimental X-ray scattering
results;see,for example,Kappkhan et al. (1989).Stewart (1979)has given
relations that allow for the direct mapping of electrostaticproperties of the
charge density, together with its gradient vector field and its associated
Laplacian density, directly from the experimental X-ray structure factors.
Few chemical observations are dependent upon a nucleus possessinga
small but finite radius and, for questionsof molecular structure and chemical
reactivity, nuclei may be regardedas point charges.Thus, the model of matter
employed here is one wherein the distribution of negative charge is measur-
ably finite over a relatively large region of space and is most highly
concentrated at the positions of the point-like nuclei which are imbedded
in it.
Figure 1.1is a representationof a number of moleculesin terms of an outer
envelopeof their charge distributions. Also indicated are the intersectionsof
the atomic boundaries as determined by the theory of atoms in molecules
with this envelope.These representationslook familiar, resembling models
presently employed to aid in the visualization of molecules.They are also
similar to the pictures of atoms obtained experimentally by the scatteringof
electrons in super microscopes or from the scanning tunnelling electron
microscope(Fig. 1.2).It is the distribution of chargethat scattersthe X-rays
or electrons in these experiments and it is the distribution of charge that
determines the form of matter in real space.Theory enables us to identify
atoms in terms of the morphology of the charge distribution. The functional
groups of chemistry, as representedby the methyl, methylene,and carbonyl
groups shown in Fig. 1.1, are objects in real space and their properties are
predicted by quantum mechanics.
If the state function is approximated by a single Slater determinant, an
antisymm etrrzedsum of products of one-electronspin-orbital states Qr, the
expressionfor the charge density is given by a sum over the products of the
one-electronstates

p(r): L,ff(r)dr(r). (1 r. 0 )

If the state function is determinedbeyond the one-electronapproximation

through someform of configurationinteraction,an expressioncorresponding
to eqn (1.10)still obtains, the spin-orbitalsbeing replacedb1'a set of one-
10 ATOMSIN CHEMISTRY 1.3

(a) (b)

Ftc. 1.2. Iodine atoms absorbed on the (1ll) iace of platinum. The data have been filtered in
order to remove spatial frequenciesbelow 0.25 nm. Image B is a three-dimensional,height
shaded,surfaceplot of the same image data shown in A.
(Courtesy of B. C. Schardt, from Science,(1989),243, 1050.Copyright 1989 by the AAAS.)

electron functions which diagonalize the one-electrondensity matrix, the so-

called natural orbitals 4,,
p(r) : L,A,qf(r)tl,(r): L,l,lrt,(t)l' (1.11)
where the natural orbital occupation numbers 2, possessvaluesbetweenzero
and one (Lciwdin 1955).
There is necessarilya great reduction in the amount of accessibleinforma-
tion in passingfrom the state function ry',a vector in the infinite-dimensional
Hilbert space,via eqn (1.4) to the charge density, a distribution function in
real space.However, from a purely operational point of view there is too
much information in the statefunction, some of it being redundant becauseof
the indistinguishability of the electronsand becauseof the symmetry of their
interactions, and some of it being unnecessaryas a result of the two-body
nature of the Coulombic interaction. In choosing to describe a system in
terms of p rather than rfr, one is not just discarding physically irrelevant
information. Instead one obtains a description of the systemin real spaceand
 REFERENCES t1

through this, as demonstratedin the follou ing chapters,a basis for the
definition of structure in this space.

E1.1 Density matrices

Not all properties of a system in a given state, notably its energies,total,

kinetic,and potential,are determinedby the chargedensity,at leastnot in an
operational form-density functional theory not withstanding. All these
properties are, however, determinedby the one-electrondensity matrix or
o n e- m at r ixf ( t ) ( r ,r' ), w h i c h d i ffe rsfro m th e c h a rgedensi tyonl y i n that one
savesseparatelythe information regarding the coordinatesof one electron
from both r/* and ry'.Thus,

f , t , ( t ,r ' ) : l r l J d r ' / * ( x 7 , x 2 , .. . x " ) / ( x i , x z , xrs). (81. 1)

Expressedin terms of the natural orbitals 4,,

f ,t,(r,r') : L,),nf(r)ry,(r'). (E1.2)

When | : r ' , eqn ( E 1 .2 )b e c o m e se q n (1 .1 1 Ih e n ce,p(r)i s sai dto be a di agonal
e l em entof f ( 1) ( rr', ). W h i l e e q n s(l .l l ) a n d (8 1 .2 )are formal l y al i ke,one can
calculate the kinetic energy from the latter but not from the former, for only
in the latter can one insert the operator betweenthe natural orbitals and let
it act separatelyon 4i or qf . The averagevalue of a two-electronproperty
can be expressedin terms of the diagonal elements of the second-order
( 2 )(rr,rr). As s u mi n ga s u mmati on over el ectronspi ns,i ts
d e ns it y m at r ix f
definition is
t - t z ) ( r, t1z ) : { l g ( l f - \ 1 2 }J d t . J d t o . . . J d r " , l t * r l t . ( E1 . 3 )
(r): p(r) givesthe total density of electrons
Just as the diagonal elementof f
a t r ( it is nor m aliz e dto .A/),s o p tz )(r1 ,rr)g i v e s th e total densi tyof pai rs at r,
and r, (it is normalized to the number of distinct pairs l{(lf - l)12).The
(2)obtainedfrom a determinantalwave function expressedin
expressionfor I-
terms of spin orbitals @,is

t-tz)
(r1'rz): t!'fr',::,'
f ,',1,!ifr',',r,i
I r',r,i"' ( E1 . 4 )
For discussionsof density matrices see McWeeny (1954. 1960),McWeeny
and Sutcliffe(1969,pp. 76-109),and Lowdin (1955).

References

Bader, R. F. W. and Beddall, P. M. (1972).J. Chem. Ph-t's.56. -1310

Kappkhan, M., Tsirel'son,V. G., and Ozerov, R. P. (1989).Dokludr Pltt ' Cft.'nr.303.
1025.
L2 ATOMSIN CHEMISTRY

L o w d i n , P . - O . ( 1 9 5 5 ) .P h y s .R e u . 9 7 , 1 4 7 4 .
McWeeny, R. (1954).Proc. r. Soc. A223,63.
McWeeny, R. (1960).Reu.mod. Phys.32,335.
McWeeny, R. and Sutcliffe, B. T. (1969).Methods oJ'molecularquantum mechanics,
Academic Press,London.
Schardt, B. C., Yau, S.-H., and Rinaldi, F. (1989).Science243, 1050.
Schrodinger,E. (1926).Ann. d. Phys.8l, 109.Referencedin CollectedpapersonwaDe
mechanics,E. Schr'odinger(3rd edn) (1982),together with his four lectureson wave
mechanics.(ChelseaPublishing Co, New York) Seein particular the fourth lecture,
p. 205.
Stewart,R. F. (1979).Chem. Phys.Lett.65,335.
 2
A T O M S A N D T H E T O P O L O G YO F T H E
CHARGEDENSITY

It would be a voluntary and unnecessaryabandonment of most valuable aid, if an

experimentalist,who choosesto consider magnetic power as representedby lines of
magnetic force, were to deny himself the use of iron filings. By their employment he
may make many conditions of the power, even in complicated cases,visible to the eye
at once; may trace the varying direction of the lines of force and determine the relative
polarity; may observein which direction the power is increasingor diminishing; and
in complex systemsmay determine the neutral points or placeswhere there is neither
polarity nor power, even when they occur in the midst of powerful magnets.By their
use probable resultsmay be seenat once, and many a valuable suggestiongained for
future leading experiments.
Michael Faraday (1851)

2.1 Introduction

This chapter describesthe spatial properties of the electronicchargedistribu-

tion and how its dominant morphological feature defines an atom in a
molecule.The chargedensity,p(r), is a physical quantity which has a definite
value at each point in space.It is a scalarfield defined over three-dimensional
space. The topological properties of such a scalar field are conveniently
summarizedin terms of the number and kind of its critical points. These are
points where the first derivativesof p(r) vanish and they thus determine the
positions of extrema in the charge density-maxima, minima, or saddles.
Since the charge density is not an arbitrary field, but one whose form is
dominated by the forcesexertedon it by the nuclei,its topological structure is
relatively simple.This structure is made more evident and more amenableto
analysisthrough the study of the associatedgradient vector field of the charge
density,Vp(r).The propertiesof a vector field are characterizedby associating
a direction as well as a magnitude with each point in space.Such fields are
frequently encountered in physics. A representation of lines of force is a
display of a vector field. Hence the opening quotation from Faraday, who
used iron filings to obtain a display of a magnetic field in terms of its lines of
force.The null points referredto by Faraday are what we have referredto as
critical points, points where the magnetic field or the Vp(r) field vanishes.
While the field we wish to study is different from a magnetic field in that it is
derivable from a scalar field (it is a gradient vector field), the reason for
'lines force' is
studying its of the same as that given by Faraday-'to make
visibleto the eyeat once'its important features,featureswhich we shall relate
back to the scalar field p(r).
t4 ATOMS AND TOPOLOGY OF CHARGEDENSITY 2.2

A display of Vp(r) for a molecule will make visible to the eye, without
further mathematical analysis,the definitions of its atoms and of a particular
set of lines linking certain pairs of nuclei within the molecule-its molecular
graph. There are four possiblekinds of stable critical points in p(r) and each
will be associatedwith a particular element of structure. From this study of
the general topological properties of the charge density, we shall conclude
that atoms exist in moleculesand that they may be linked together to form
structures consisting of chains, rings, and cages (Bader and Beddall 1972;
Bader et al. 1979a,b,1981).

2.2 Topological properties of the charge density

2.2.1 The dominant form in the chargedensity

The form assumedby the distribution of charge in a molecular systemis the

physical manifestation of the forces acting within the system. Dominant
among theseis the attractive force exertedby the nuclei, a consequenceof the
localized nature of the nuclear charge.This interaction is responsiblefor the
single most important topological property exhibited by a molecular charge
distribution of a many-electronsystem-that, in general,p(r; X) exhibits local
maxima only at the positions of the nuclei. This is an observation based on
experimental results obtained from X-ray diffraction studies on crystals and
on the results of theoretical calculations on a large number of systems.A
more detailed discussionof this topic, both theoretical and experimental,is
given in the end-note, Section 82.1.
Becausethe charge density exhibits local maxima at the nuclei, recog-
nizable atomic forms are created within a molecular charge distribution.
These forms are so dominant in determining the structure of the charge
density that their individual properties form recognizablecontributions to
the properties of the total system.Thus, the atomic basisfor the classification
of chemical properties could and did evolve as the working model of
chemistry beforeits underlying physicalbasiswas known. The sphericalatom
model introduced by Bragg to model crystal structuresand the more recent
model of overlapping van der Waals spheres or the space-filling models,
commonly used by chemistsand employed to determine the shapesand sizes
of molecules,are all reflectionsof the basic fact that the charge density is a
maximum at a nucleus and decaysin a nearly spherical manner away from
this point.
The local maxima in a chargedistribution are illustrated in Fig. 2.1. The
chargedensity in the ethylenemolecule,CrHo, is displayedin two ways for
each of three planes: as a projection in the third dimension above the
geometricplane and as a contour map, in which each contour representsa
 2.2 P R O P E R T I E SO F T H E C H A R G E D E N S I T Y 15

at'r''
i
i

Frc. 2.1. Displays of the electronic charge density in three planes of the ethylene molecule,
CrHo. In the diagrams on the left, the value of the density is shown as a projection above the
geometricplane; the diagrams on the right are correspondingcontour plots of p(r). (a) The plane
containing the nuclei. Values of p(r) above an arbitrarily chosen value are not shown. (b) The
plane obtained by a 90' rotation about the C-C axis of the plane shown in (a). (c) The plane
perpendicularto the C-C axis at its midpoint. The valuesof the contours increasein magnitude
from the outer contour inwards with the values given in the Appendix, Table 42. The projected
positions of out-of-plane nuclei are indicated by open crosses'
( From Bader et al (1981).)
16 ATOMS AND TOPOLOGY OF CHARGEDENSITY 2.2

line of constantdensity.The chargedistributionexhibitsa maximumat the

positionof eachnucleusin Fig. 2.1(a),the planeof the nuclei.Figure2.1(b)is
for a planeperpendicular to that shownin Fig. 2.1(a),obtainedby a rotation
aboutthe carbon--carbon axisand containingthesenuclei.The densityagain
exhibitsmaximaat the positionsof the carbonnuclei.The observationof a
maximumin p(r) at a nuclearpositionis true whenthe distributionis viewed
in any planecontainingthe nucleus.It is this propertythat is classifiedas a
localmaximumin p(r).
This behaviourof p(r) is to be contrastedwith that displayedat the
midpoint of the carbon-carbonaxis.In the planesshownin Fig. 2.1(a)and
(b), p(r) has the appearanceof a saddleat this point, while in the plane
perpendicular to and bisectingthe carbon-carbonaxis,p(r)is a maximumat
this samepoint. In this casep(r) is a maximum in one particular plane.
Knowledgeof p(r) in oneor two dimensions is insufficientto characterize
its
three-dimensional form. What is neededis a method of summarizingin a
precisemannerthe principal topologicalfeaturesof a chargedistribution.
This informationis providedby the curvaturesof p(r) at its critical points
(Collardand Hall 1977;Smith er al. 1977\.

2.2.2 Critical points and their classification

Eachtopologicalfeatureof p(r),whetherit be a maximum,a minimum,or a

saddle,hasassociated with it a point in spacecalleda criticalpoint,wherethe
first derivativesof p(r) vanish.Thus,at sucha point denotedby the position
vectorr", Vp(r"): 0 whereVp denotesthe operation
Yp: i 0pl0x + j0pl 0y + kAplAz. (2.r)
Whether a function is a maximum or a minimum at an extremum is, of
course, determined by the sign of its second derivative or curvature at this
point. The secondderivative of a function f (x) at x (illustrated diagrammati-
cally in Fig. 2.2) is the limiting differencebetweenits two first derivatives or
tangent lines which bracket that point

d2f(x)ldxz : lim(flim{[/(x + Ax) - /(x)] lL*]

- lim{ t/(") - f (* - Ax)lI L*)ll Lx). (2.2)
At a point where/(x) is a minimum, the second derivative is the difference
between a positive and a negative slope and is, therefore, greater than
zero, while, at a maximum in f (x), the second derivative is the difference
between a negative and a positive slope and the result is a value less than
zero. For values of x lying between such extrema, both slopes are either
positive or negative and the curvature can be of either sign depending upon
whether x is in the region of a maximum or a minimum. The second
2.2 P R O P E R T I E SO F T H E C H A R G E D E N S I T Y I7

y=f(x)

X1-AX X1 X1*AX X2-AX X2 X2*AX

Frc.2.2. Definition of curvature as the limiting difference(Ax -- 0) in the tangent lines which
bracket a given point; at x, where/(r) is a minimum and the curvature is positive and at xt
where/(x) is a maximum and the curvature is negative.

derivative undergoesa changein sign at the point of inflexion where its value
is zero. It is also clear from its defining equation that, when the curvature is
negativeat x, the value of/(x) is greater than the averageof its values at the
neighbouring points x * dx and x - dx, with the reversebeing true when the
curvature is positive. These considerationscarry over into three dimensions.
It is clear from Fig.2.1 that the critical point at the centre of the ethylene
moleculebetweenthe two carbon nuclei has one positive curvature, along the
internuclear axis, and a negative curvature along each of the two perpen-
dicular axes. Thus the central saddles in the charge density appearing in
Fig. 2.1(a),(b) exhibit the positive curvature along the carbon-carbon axis
and one each of the two negative curvatures,while the maximum shown in
Fig.2.1(c)exhibitsjust the two negativecurvatures.
In general,for an arbitrary choice of coordinate axes,one will encounter
nine second derivatives of the form AzplAxAy in the determination of the
curvatures of p at a point in space.Their ordered 3 x 3 array is called the
Hessianmatrix of the charge density, or simply, the Hessian of p. This is a
real, symmetric matrix and as such it can be diagonalized. This is an
eigenvalueproblem and its solution correspondsto finding a rotation of the
coordinate axes to a new set such that all of the off-diagonal elements-the
mixed second derivatives of p-vanish. This general problem is ubiquitous
throughout chemistryand physicsand its solution is outlined in SectionE2.2
for the reader unfamiliar with it. The new coordinate axes are called the
principal axes of curvature because the magnitudes of the three second
derivatives of p calculated with respect to these axes are extremized. The
18 ATOMSANDTOPOLOGYOFCHARGEDENSITY 2.2

principalaxeswill correspondto symmetryaxesif the criticalpoint is at the

origin of sucha set of axes,as it is for the centralcriticalpoint in ethylene
discussed above.In the caseof symmetricallyequivalentaxes,the corres-
pondingcurvaturesare equaland any linearcombinationof the degenerate
setof axeswill serveasa principalaxisof curvature.The traceof the Hessian
matrix, the sum of its diagonalelements,is invariant to a rotation of the
coordinatesystem.Thus,the valueof the quantityY2p,calledthe Laplacian
of p,
Y ' p : Y ' Y p : 0 2 p | 0 x 2 + a 2 p l 0 y+2 0 2 paf z 2 , (2.3)
is invariant to the choice of coordinate axes. The principal axes and their
corresponding curvatures at a critical point in p are obtained as the eigen-
vectors and correspondingeigenvaluesin the diagonalization of the Hessian
'curvature and eigenvalue'and 'axes
matrix of p(r"). Thus the pairs of names
of curvature and eigenvectors'can be used interchangeablyin describingthe
properties of a critical point in p.
While all of the eigenvaluesof the Hessianmatrix of p at a critical point are
real, they may equal zero.The rank of a critical point, denoted by ,, is equal
to the number of non-zero eigenvaluesor non-zero curvatures of p at the
critical point. The signature,denotedby o, is simply the algebraicsum of the
signs of the eigenvalues,i.e. of the signs of the curvatures of p at the critical
point. The critical point is labelled by giving the duo of values(a, o\. Thus the
central critical point in ethylenewith three non-zero curvatures,one positive
and two negative,is a (3, - 1) critical point.
With relatively few exceptions,the critical points of chargedistributions for
molecules at or in the neighbourhood of energetically stable geometrical
configurations of the nuclei are all of rank three. The near ubiquitous
occurrenceof critical points with a :3 in such casesis another general
observation regarding the topological behaviour of molecular charge dis-
tributions. It is in terms of the propertiesof critical points with o : 3 that the
elementsof molecular structure are defined. A critical point with ar < 3, i.e.
with at least one zero curvature, is said to be degenerate.Such a critical point
is unstablein the sensethat a small changein the chargedensity,as causedby
a displacementof the nuclei, causesit to either vanish or to bifurcate into a
number of non-degenerate or stable(a :3) critical points. Sincestructureis
generic in the sensethat a given structure or arrangement of bonds persists
over a range of nuclear configurations, the observed limited occurrence of
degeneratecritical points is not surprising.One correctly anticipatesthat the
appearanceof a degeneratecritical point in a molecular charge distribution
denotes the onset of structural change,the subject of Chapter 3.
There are just four possible signature values for critical points of rank
three.
(3, - 3) All curvatures are negative and p is a local maximum at rc.
2.2 P R O P E R T I E SO F T H E C H A R G E D E N S I T Y 19

(3, - 1) Two curvaturesare negativeand p is a maximum at r" in the

plane definedby their correspondingaxes.p is a minimum at
r" along the third axis which is perpendicular to this plane.

(3, + 1) Two curvaturesare positive and p is a minimum at r. in the

plane defined by their correspondingaxes.p is a maximum at
r" along the third axis which is perpendicular to this plane.

(3, + 3) All curvatures are positive and p is a local minimum at r".

2.2.3 Critical points of molecular charge distributions

The Coulombic potential becomesinfinitely negativewhen an electron and a

nucleus coalesceand, because of this, the state function for an atom or
moleculemust exhibit a cusp at a nuclear position. That is, as shown by Kato
(1957),the first derivative of the function is discontinuous at the position of a
nucleus.Thus, while the charge density is a maximum at the position of a
nucleus, this point is not a true critical point because Vp, like V r/, is
discontinuous there. However, as discussedin Section 82.1, this is not a
problem of practical import and the nuclear positions behave topologically
as do (3, - 3) critical points in the chargedistribution and hereafterthey will
be referred to as such.
All of the saddle points shown in Fig. 2.1(a)are (3, - 1) critical points.
A (3, - 1) critical point is found between every pair of nuclei which are
considered to be linked by a chemical bond in the ethylene molecule.
Figure 2.3 shows representationsof the charge distribution, similar to those
shown in Fig. 2.1, for the three symmetry planes of the diborane molecule.
While the topology of the charge distribution for the plane of the four
terminal hydrogen nuclei and the two boron nuclei shown in Fig. 2.3(a)is
indistinguishable from that shown in Fig. 2.1(a) for ethylene, the central
critical point in the former molecule is not a (3, - 1) critical point. This is
made clear from Fig. 2.3(b)which shows the chargedensity in the plane of the
bridging hydrogens obtained by a 90" rotation about the boron-boron axis.
The charge density is a minimum at the central point in this plane showing
that it is a (3, + 1) critical point in the charge density. The final view of this
critical point is shown in Fig. 2.3(c), in the plane perpendicular to the
boron-boron axis and containing the bridging protons, where it again
appears as a saddle in p. The axis of the single negative curvature of this
critical point is perpendicularto the plane shown in Fig. 2.3(b).Thus Fig.
2.3(a),(c) exhibits this one negative and one each of the two positive
curvaturesand p has the appearanceof a saddle point in theseplanes.While
(3, - 1) and (3, + 1) critical points can both appear as (2.0) critical points
when viewedin specificplaneswhere every critical point is of rank tw'o.their
full three-dimensionalbehaviour is quite different.The other critical points.
20 ATOMS AND TOPOLOGY OF CHARGE DENSITY 2.2

..::a=:.a.

(@), i'''i'o
..::j-,.,.:

---.'-

'i
iiiii ..-;:)
"l--
r \..
....--
'-'-

FIc. 2.3. Relief and contour maps of the electronic charge density for three planes of the
diborane molecule, BrHu.(a) The plane containing the boron nuclei and the four terminal
protons. (b) The plane obtained by a 90' rotation about the B-B axis of the plane shown in (a).
(c) The plane perpendicularto the B-B axis at its midpoint and containing the bridging hydrogen
atoms. The axis linking the two bridging hydrogenshas been rotated by 90'relative to this same
axis in (b). The projected positions of out-of-plane nuclei are indicated by open crosses.

appearing as saddles connecting neighbouring nuclei in Fig. 2.3(a),(b), are

indeed (3, - 1) critical points in agreement with the chemical structure
usually assignedto this molecule, i.e. terminal BH2 groups linked by two
bridging protons. One notes for future referencethat the minimum of the
2 .2 O F T H E C H A R G ED E N S I T Y
PROPERTIES 21

(3, * 1) central critical point in Fig. 2.3(b) is bounded by a ring of four

(3, - 1) critical points which link the bridging protons to the boron nuclei.
The fourth and final kind of stable critical point is illustrated in Fig.2.4
which gives a representationof the charge distribution in the tetrahedrane
molecule, CoHn. The chemical structure assignedto this molecule is that of a
cage.The critical point in the centre of the molecule appearsas a minimum in
Fig.2.4 and, becauseof the symmetry possessedby this molecule,this critical
point will have the sameappearancewhen viewed in any plane. It is a (3' + 3)
critical point and the charge density is a local minimum at the centre of the
cage structure. Figure 2.4 is interesting as it gives two-dimensional views of
all four kinds of stable critical points. In terms of the structural properties
associatedwith the critical points illustrated above, the reader can satisfy
herself or himself that the (2,0) saddles between the carbon nuclei and
between each carbon and its neighbouring hydrogen are (3, - 1) critical
points, that the (2, - 2) maximum is in the unique plane of the (3' - i)
critical point between the two out-of-plane carbon nuclei, and that the
remaining two (2, 0) saddlesare (3, + 1) critical points in the triangular faces
formed by the carbon nuclei joined by (3, - 1) critical points.
This discussion has shown that the principal topological features of a
charge distribution can be summarized using the rank and signature classi-
fication schemeof its critical points. It has further demonstratedthe existence

Ftc.2.4. Relief map of the electronic charge density in the tetrahedranemolecule.CnHo. The
plane shown is a oo symmetry plane and contains two carbon nuclei and their associated
protons. The charge density at the central critical point is a local minimum *ith a value of
0.165au. The two-dimensional maximum in the foreground is the (2. - 2) marimutn in p in the
interatomic surfaceof the out-of-plane carbon nuclei. The value of p at thrs point is 0l'16 au. A
contour map of the charge density in the same plane is shown in Frg. 1.10.
22 ATOMS AND TOPOLOGY OF CHARGE DENSITY 2.3

of a connection betweenthe number and kind of critical points appearingin a

chargedistribution and its conventional chemical structure.The next stageof
the developmentis to show that thesequalitative associationsof topological
features of p with elements of molecular structure can be replaced with a
complete theory, one which recovers all of the elements of structure in a
manner that is totally independent of any information other than that
contained in the charge density. The topology of the charge density will
define molecular structure.
There is in the charge density an underlying structure which is brought to
the fore in its associatedgradient vector field. We have observedabove that
the only local maxima in a many-electron charge distribution occur at the
positions of the nuclei. This observation confers upon a nucleus the special
role of attractor in the gradient vector field of the charge density. This
identification, which is itself a reflection of the dominance of the nuclear force
in determining the form of p, provides the basisfor the definition of an atom
and the associatedelementsof molecular structure.Theseobservationsaside.
one would still proceedwith a study of the gradient vector field of the charge
density, for the boundary condition of a quantum subsystemis stated in
terms of this field. It is this most remarkable coincidence of reasons for
studying the gradient vector field-one from the quantum definition of a
subsystem,the other from the quite independentdemonstration that its form
yields a mapping of the elements of molecular structure-that gives the
theory of atoms in moleculesits unified structure.

2.3 Gradient vector field of the charge density

2.3.1 Trajectories of the gradient vector field

The gradient vector field of the charge density is representedthrough a

display of the trajectoriestraced out by the vector Yp.A trajectory of Vp, also
called a gradient path, starting at some arbitrary point ro is obtained by
calculating Vp(ro),moving a distanceAr away from this point in the direction
indicated by the vector Vp(ro) and repeatingthis procedure until the path so
generated terminates. This operation is the three-dimensionalanalogue of
approximating a function /(x) in terms of its tangent line at x, f (x + Ax)
: f(x) + (dfldx)Ax, an expressionwhich, as a consequenceof the definition
of a derivative, becomesexact in the limit Ax -r dx.
Figure 2.5 is a contour diagram of the charge density of the sodium
chloride diatomic molecule overlaid with trajectories of Yp to illustrate the
following general properties.

1. Since the gradient vector of a scalar points in the direction of greatest

 trajectory
of Vp

is a unit vector
in the direction
of Vp
M -+0

Frc. 2.5. A contour map for the NaCl molecule overlaid with trajectories of Vp. With the
exception of the four trajectories associatedwith the (3, - l) critical point (denoted by a dot), the
trajectoriesoriginate at infinity and terminate at one of the two nuclei. Two trajectoriesoriginate at
infinity and terminate at the (3, - 1) critical point, while two others orig:nate at this point and
terminate,one each,at the nuclei. The property of zero flux in the gradient vectors of p is illustrated
for the interatomic surfacewhose intersection with this plane is given by the tu o tra.;ectoriesw hich
terminate at the critical point. An arbitrarily drawn surfaceis shown not to hare this propertl of
zero flux. The definition of the derivative dr/dl as the limit of Ar A/ is also sho$n.
24 A T O M S A N D T O P O L O G YO F C H A R G E D E \ S I T \ ' /-. J

increasein the scalar,the trajectoriesof Vp are perpendicularto lines of

constant density--to contour lines of p.

2. The vector Vp(r) is tangent to its trajectory at each point r.

3. Every trajectory must originate or terminate at a point where Vp(r)

vanishes,i.e. at a critical point in p.

4. Trajectoriescannot cross since Vp(r) definesbut one direction at each

point r.

All but two of the trajectoriesshown in the Fig. 2.5 originateat infinity where
Vp vanishes.The remainingtwo originateat the (3, - 1)criticalpoint located
betweenthe nuclei and these,along with all but another two of the trajector-
ies,terminate at one of the two nuclei. The remaining two trajectoriesin this
plane terminate at the (3, - 1) critical point.
An understandingof thesepropertiesof trajectoriesis sufficientfor a reader
to proceedwith the discussionof their usein the definition of structuregiven
in the following section.The presentsectionis concludedwith the derivation
of the differential equation for Vp(r) and the corresponding integral curve
which definesthe points on its trajectory.
Consider the two points defined by the position vectors r, and r2 on a
gradientpath and separatedby the distanceA/ as shown in Fig. 2.5.8y taking
the limit, A/ -- 0, one obtains the dimensionlessderivative

d r l d l : l i m a r - o { ( r z - r t ) l L l } : l i m o ;- o { L r l L ^ l } . (2.4)
This derivativeis the unit vector tangentto the curve definedby a succession
of points ru at the point r, and henceit must give the direction of the vector
Vp(r) at this point. Thus

drldl: Vp(r)/lvp(r)1. (2.s)

What is desiredis a differential equation for Vp(r) itself. Such a result can be
obtained from eqn (2.5)by defining a parameters as a weightedmeasureof
distance along a given path and related to the true path length / by the
property that

l i m a l * s{ A s / A / }: 1l l V p l
or
d s: d / / l V p l . (2.6)
Substitutionof ds for d/ in eqn (2.5)yieldsthe desireddifferentialequation
dr(s)/ds: Vp(r(s)) (2.7)
where the notation r(s) implies that a point r on a given trajectory is
dependentupon the path parameters.
L.) GRADIENT VECTOR FIELD OF CHARGE DENSITY 25

Equation (2.7) represents three first-orderdifferentialequationsand it

yields unique solutionsonly when particularvaluesare assignedto three
constantsof integration.This corresponds to fixing someinitial point on a
trajectory,at s : s1, for example.Then everyother point on the trajectory
that passesthroughthe point r(s,) is obtainedby integratingeqn (2.7)with
the threeconstantsof integrationgivenby the components of r(sr),

r(s): r(sr)+ (2.8)

,[;rr(r(r))dr
A trajectory of the gradient vector field of p(r) is, therefore, a parametrized
integral curve, a solution curve,of the differentialequation for Vp(r). By fixing
a point on a given trajectory all other points which lie on the same trajectory
can be obtained by solving eqn (2.8).It is clear from eqn (2.6)that, since lVp I
becomesincreasingly smaller as a critical point is approached, the value of
the path parameter s must, correspondingly, become increasingly larger,
equalling infinity at the critical point where Vp vanishes.Whether the actual
path is of finite or infinite length, the path parameter variesbetweenthe limits
- o , the origin of the path, and + oo , its terminus. Mathematicians refer to
the origin of a gradient path as its a-limit set, and to its terminus as its
cr.r-limitset.

2.3.2 Phase portraits of the gradient vector field

First-order differential equations, such as eqn (2.7) for Vp, are common
throughout classicalphysics.In the nineteenthcentury, the French mathema-
tician Poincar6 suggestedthat one could discern the qualitative behaviour of
the solutions to such equations in the pattern of the trajectoriesrepresenting
the vector field defined by the differential equation, its so-called phase
portrait. We are also interested in the global arrangement of the gradient
paths rather than in their individual behaviour. Our primary concern is to
look for and to describeany regularity or structure that the gradient paths
exhibit. Since the critical points in a charge distribution serve as the origins
andfor termini of the gradient paths, they emergeas fundamental in the study
of the topology of the charge density, and the elementsof structure are to be
found in the propertiesexhibited by the phaseportraits of the gradient vector
field in the vicinities of the four stable critical points.
An eigenvalue and its associated eigenvector of the Hessian of p (a
principal curvature and its associatedaxis) at a critical point define a one-
dimensional system. If the eigenvalueor curvature is negative, then p is a
maximum at the critical point on this axis and a gradient vector will
approach and terminate at this point from both its left- and right-hand side as
illustrated in Fig. 2.6 for the case(1, - 1), a system of rank 1 and signature
26 ATOMS AND TOPOLOGY OF CHARGEDENSITY 2.3

\\ /
l
\-/
-------+ . <- +-.-+
(1,-1) (1,*1;

\lz-
/1\
ll,z
\

(2 ,-2 1 (2,r21

',fr'
. &':
l . ! !
-" -:i jr
-f' ;'
'ra;drl f
.. -
r. i
' : l

@$,r+-r1""'" :; +"'rril:'!"",

) l l t \=-*
\
*/l II Y -
: i-
It l
1 l | l |
(2' o) (3,-1)

Ftc.2.6. Phaseportraits for maxima and minima in one and two dimensionsand for a saddle.
The final diagram is a display of the trajectories,which terminate at a (3, - l) critical point and
define the interatomic surface,and of the two trajectories which originate at this point and define
the bond path.
2.3 G R A D I E N TV E C T O RF I E L D O F C H A R G ED E N S I T Y 27

- 1.If the eigenvalue is positive,thenp is a minimumat the criticalpoint on

this axisand two gradientvectorswill originateat this point asillustratedfor
the case(1, + 1).In two dimensions, if both eigenvalues are negative,p is a
maximumat the criticalpoint and all trajectories of Vp will terminateat the
critical point as illustratedfor the case (2, - 2). This set of trajectoriesis
definedby all possiblelinearcombinationsof the two associated eigenvectors
which span a two-dimensionalspace.Similarly, if both eigenvalues are
positiveandp is a minimumat the critical point, all trajectories will originate
at the critical point and again define a surfaceas illustrated for the case
(2, + 2).A moreinterestingsituationis obtainedwhenthe eigenvalues are of
oppositesign(thesignatureis zero)and the chargedensity in a plane has the
form of a saddleas illustratedfor the case(2,0).In this situation the two
trajectories associated with the axisof the negativecurvatureterminateat the
criticalpoint while the two associated with the positivecurvatureoriginate
there.The trajectoriesformedby linear combinationsof the two associated
eigenvectors neitherterminatenor originateat the criticalpoint but instead
avoid this point as indicatedfor the caselabelled(2' 0).
In three dimensions,the pairs of eigenvectors associatedwith the two
negativeor two positiveeigenvalues of a (3, - 1) or (3, + 1) criticalpoint,
respectively, will againdefinea surface. Unlike the two-dimensional examples
discussed above,thesesurfaces will be planaronly if the criticalpoint r" liesin
a symmetryplane.If not, the surfacewill be planar only in the immediate
neighbourhoodof r" and will, in general,be curvedbeyondthis regionbut
-
still definedby the uniqueset of trajectorieswhich terminateat a (3' 1)
critical point or originateat a (3, + 1) critical point. The final diagramin
Fig. 2.6 is a displayof the three-dimensional form of the phaseportrait of
a (3, - 1) criticalpoint. The uniquepair of trajectoriesassociated with the
single positive eigenvalue originate at the criticalpoint, asin the (1, + 1)case
illustrated above. The setof trajectories defined by linear combinations of the
pair of eigenvectors associated with the two negativeeigenvalues terminateat
thecriticalpoint and definea surface.The chargedensityis a maximumin the
surfaceat the critical point and a minimum at this samepoint along the
perpendicular axis.The behaviourof the chargedensityat a (3, + 1)critical
pointis just the oppositeof thisandits phaseportraitis obtainedby reversing
all of the arrowsshownfor the (3, - 1) case.
An analyticalexpression for a trajectoryin the vicinityof a criticalpoint is
givenin SectionE2.z.Whilethisexpression is.necessary for numericalwork,a
knowledgeof the phase portraits given above is all that is requiredfor the
definitionof the elements of molecular structure which follows.
 ATOMS AND TOPOLOGY OF CHARGE DENSITY 2.4

2.4 Elementsof molecular structure

2.4.1 Equivalenceof the topological and quantum definitions

of an atom

The gradient vector field of the charge density in the plane containing the
nuclei of the ethylenemolecule is illustrated in Fig. 2.7. A (3, - 3) critical
point, such as occurs at each of the nuclear positions, servesas the terminus,
the co-limit set, of all the paths starting from and contained in some
neighbourhood of the critical point. A (3, - 3) critical point exhibits the
property which defines a point attractor in the gradient vector field of the
charge distribution: there exists an open neighbourhood B of the attractor
which is invariant to the flow of Vp such that any trajectory originating in B
terminates at the attractor. The largest neighbourhood satisfying these
conditions is called the basin of the attractor.
Since (3, - 3) critical points in a many-electron charge distribution are
generally found only at the positions of the nuclei, the nuclei act as the
attractors of the gradient vector field of p(r; X). The result of this identifica-
tion is that the space of a molecular charge distribution, real space, is
partitioned into disjoint regions,the basins,each of which contains one point
attractor or nucleus.This fundamental topological property of a molecular
charge distribution is illustrated in Fig. 2.7(a), which depicts only those
gradient paths of the charge density which terminate at each of the nuclear
attractors in the molecule.While Fig.2.7(a) illustrates this property for only
one plane, it is to be emphasizedthat, becausep is a local maximum at a
nucleus,a (3, - 3) critical point, the basin of an attractor is a region of three-
dimensional spaceand the partitioning so clearly indicated in Fig. 2.7 extends
throughout all of space.An atom,free or bound,is definedas the union of an
ottractor and its associatedbasin.
Alternatively, an atom can be definedin terms of its boundary. The basin of
the single nuclear attractor in an isolated atom covers the entire three-
dimensional spaceR3. For an atom in a moleculethe atomic basin is an open
subset of R3. It is separated from neighbouring atoms by interatomic
surfaces.The existenceof an interatomic surfaceSoudenotesthe presenceof a
(3, - 1) critical point betweenneighbouring nuclei A and B. The presenceof
such a critical point betweencertain pairs of nuclei was noted above as being
a general topological property of molecular charge distributions. Their
presencenow appears as providing the boundaries between the basins of
neighbouring atoms. As discussed above and illustrated in Fig. 2.6, the
trajectories which terminate at a (3, - 1) critical point define a surface,the
interatomic surfoce Snr. The reader will recall Fig. 2.1(c)showing that the
chargedensity is a maximum at a(3, - 1)critical point in this surface.Figure
2.7(b)shows,in addition to the trajectoriesthat terminate at eachof the nuclei
 2.4 E L E M E N T SO F M O L E C U L A R S T R U C T U R E 29

in ethylene,the trajectoriesassociatedwith eachof the (3, - 1)criticalpoints.

Two trajectories terminate at each such critical point in the plane of the
I
I diagram. They denote the intersection of the interatomic surfaceswith this
l plane. It should be borne in mind that each such pair of trajectoriesare but
two of an infinity of gradient paths,all of which terminate at a (3, - 1)critical
point and define a surfacein three dimensions as illustrated in Fig. 2.6. The
atomic sufface So of atom A is defined as the boundary of its basin. In set-
theoretic languageSe 2 UsSnn. (A summary of symbols used in set theory is
given in Section 83.1.) Generally, this boundary comprises the union of a
number of interatomic surfaces,separating two neighbouring basins, and
some portions which may be infinitely distant from the attractor. The atomic
surfaceof a carbon atom in ethylene,as seenin Fig. 2.7 for example,consists
of three interatomic surfaces,two with hydrogens and another with a carbon
atom.
If the topological property which definesan atom is also one of physical
significance,then it should be possibleto obtain from quantum mechanicsan
equivalent mechanicaldefinition. As demonstratedin Chapters 5 and 8, this
can be accomplishedthrough a generalizationof the quantum action princi-
ple to obtain a statement of this principle which applies equally to the total
system or to an atom within the system. The result is a single variational
principle which definesthe observables,their equations of motion, and their
averagevalues for the total system or for an atom within the system.
The generalization of the action principle to a subsystemof some total
system is unique, as it applies only to a region that satisfiesa particular
constraint on the variation of its action integral. The constraint requiresthat
the subsystembe bounded by a surfaceof zero flux in the gradient vectors of
the charge density, i.e.
Vp(r)' n(r) : g for all points on the surfaceS(r). (2.e)
In order for the scalar product of n, the vector normal to the surface(seeFig.
2.5),with Y p to vanish, it is necessarythat the atomic surfacenot be crossed
by any trajectories of Vp and as such it is referred to as a zero-flux surface.
The state function tJrand Y t 'n, where the gradient is taken with respectto
the coordinates of any one of the electrons, vanish on the boundary of a
bound system at infinity. Thus, p and Vp vanish there as well and a total
isolated system is also bounded by a surface satisfying eqn (2.9).Since the
generalizedstatement of the action principle applies to any region bounded
by such a surface,the zero-flux surfacecondition placesthe description of the
total system and the atoms which comprise it on an equal footing.
Becauseof the dominant topological property exhibited by a molecular
charge distribution-that it exhibits local maxima at the positions of the
nuclei-the imposition of the quantum boundary condition of zero flux leads
directly to the topological definition of an atom. Indeed. the interatomic
2.4 E L E M E N T SO F M O L E C U L A R S T R U C T U R E 31

surfaces,along with the surfacesfound at infinity, are the only closedsurfaces

of R3 which satisfy the zero-flux surface condition of eqn (2.9). This is a
natural result of the property of an atomic basin as shown in Fig. 2.7(a\-all
the trajectoriesin the vicinity of a given nucleusterminate at that nucleusand
no trajectoriescrossfrom one basin to another. Sincetrajectoriesof Vp nevef
cross,the zero-flux surfacecondition follows directly from the definition of an
interatomic surface in terms of the set of trajectories which terminate at a
(3, - 1) critical point. In terms of this same definition, it follows that the
vector Vp(r) will be tangent to the surfaceS(r) of an atom at every point r.
It is clear from Fig. 2.5 that a surface,arbitrarily drawn through a charge
distribution, will be crossedby gradient vectors of p and will not satisfy the
zero-flux condition. Any surfaceincluding a nuclear position coordinate is
also excludedas Vp is not definedthere and again the zero-flux condition, eqn
(2.9),will be violated.
The coincidenceof the topological and quantum definitions of an atom
means that the topological atom is an open quantum subsystem,free to
exchange charge and momentum with its environment across boundaries
which are defined in real space and which, in general,change with time. It
should be emphasizedthat the zero-flux surface condition is universal-it
appliesequally to an isolated atom or to an atom bound in a molecule.The
approach of two initially free atoms causesa portion of their surfacesto be
shared in the creation of an interatomic surface.Atomic surfacesundergo
continuous deformations as atoms move relative to one another. They are,
however,not destroyed as atoms separate.
A linked grouping of neighbouring atoms (a concept defined more fully in
the following Section)as well as individual atoms are bounded by a zero-flux
surface.Such a surfacemay be used to define a Wigner-Seitz cell in a solid, a
solute in a solution, or a molecule in a molecular crystal. The zero-flux
surface,eqn (2.9),is the natural boundary condition for a system defined in
real space and such a surface can always be used to define the physically

Frc.2.7. Maps of the gradient vector field of the charge density for the plane containing the
nuclei of the ethylene molecule. Each line representsa trajectory of Vp(r). In (a) only those
lrajectories which terminate at the positions of the nuclei are shown. Each trajectory is
.rrbitrarily terminated at the surface of a small circle about a nucleus. The set of trajectories
* hich terminate at a given nucleusor attractor define the basin of the attractor. (b) The same as
iat. but including the trajectorieswhich terminate and originate at the (3, - 1)critical points in
ihe chargedistribution. The position of a (3, - l) critical point in theseand succeedingdiagrams
rs denoted by a full circle.The pair of trajectorieswhich, in this plane. terminate at each (3, - l)
critical point mark the intersection of an interatomic surfacewith the plane of the figure. The
gradientpaths which originate at the (3, - 1) critical points and definethe bond paths are shown
b1 the heavy lines. (c) A superposition ofthe trajectories associatedriith the t3. - 1) critical
points on a contour map of the charge density. These trajectoriesdefine the boundaries of the
atoms and the molecular graph.
(From Bader et al (1981).)
32 ATOMS AND TOPOLOGY OF CHARGEDENSITY 2.4

relevant pieces of a total system whether they be in intimate contact or

essentially isolated. There is, of course, no such thing as a truly isolated
system.Each is bounded by a zero-flux surface,and systemsso defined differ
only in the degreeof their interaction with neighbouring systems.One should
also anticipate at this point, that the averageproperties of a subsystemand
their equations of motion are determined by the same quantum mechanical
expressionsas apply to an'isolated'system.The major differencebetweenthe
propertiesof a subsystemand those of a systemconsideredto be isolated is to
be found in the fluctuation of their averagevalues.Such fluctuations vanish
for an isolated system,thereby ensuring,for example,that an integer number
of electrons are present.The fluctuation in the population of an interacting
subsystemis, however,not zero and, correspondingly,its averagepopulation
is not necessarilyan integer.
This book is primarily concerned with atoms in molecules but the more
general nature of the theory regarding its ability to provide a description of
the properties of any physically relevant piece of the universe and of its
itrteraction with the remainder of the universe could prove to be of equal if
not greater usefulness.Further discussion of this point is postponed until
Chapter 8 where the proper quantum mechanicalframework is established,
particularly the equation of motion for a subsystem.

2.4.2 Chemical bonds and molecular graphs

Also shown in Fig. 2.7(b)are the pairs of gradient paths which originate at
each (3, - 1) critical point and terminate at the neighbouring attractors. As
previously discussedand illustrated in Fig. 2.6, each such pair of trajectories
is definedby the eigenvectorassociatedwith the unique positive eigbnvalueof
a (3, - 1) critical point. These two unique gradient paths define a line
through the charge distribution linking the neighbouring nuclei along which
p(r) is a maximum with respectto any neighbouring line. Such a line is found
between every pair of nuclei whose atomic basins share a common inter-
atomic surface and, in the general case, it is referred to as an atomic
interactionline (Bader 1975:Runtz et al. 1977:Bader and Essen1984).
The existenceof a (3, - 1) critical point and its associatedatomic inter-
action line indicates that electronic charge density is accumulated between
the nuclei that are so linked. This is made clear by referenceto the displaysof
the charge density for such a critical point, as given in Fig. 2.1 for example,
particularly Fig. 2.1(c)which shows that the charge density is a maximum in
an interatomic surfaceat the position of the critical point. This is the point
where the atomic interaction line intersects the interatomic surface and
chargeis so accumulatedbetweenthe nuclei along the length of this line. Both
theory and observation concur that the accumulation of electronic charge
between a pair of nuclei is a necessarycondition if two atoms are to be
t 2.4 E L E M E N T SO F M O L E C U L A R S T R U C T U R E

bonded to one another. This accumulation of charge is also a sufficient

33

condition when the forceson the nuclei are balancedand the systempossesses
a minimum energyequilibrium internuclear separation.Thus, the presenceof
an atomic interaction line in such an equilibrium geometry satisfiesboth the
necessaryand sufficientconditions that the atoms be bonded to one another.
In this casethe line of maximum charge density linking the nuclei is called a
bond path and the (3, - 1) critical point referred to as a bond critical point
(Bader and Ess6n1984).
For a given configuration X of the nuclei, a moleculargraph is defined as
the union of the closures of the bond paths or atomic interaction lines.
Pictorially, the molecular graph is the network of bond paths linking pairs of
neighbouring nuclear attractors. The molecular graph isolates the pairwise
interactions presentin an assemblyof atoms which dominate and character-
ize the properties of the system be it at equilibrium or in a state of change.
A molecular graph is the direct result of the principal topological proper-
ties of a system'scharge distribution: that the local maxima, (3, - 3) critical
points, occur at the positions of the nuclei thereby defining the atoms, and
that (3, - 1) critical points are found to link certain, but not all pairs of
nuclei in a molecule. The network of bond paths thus obtained is found to
coincide with the network generatedby linking together those pairs of atoms
which are assumedto be bonded to one another on the basis of chemical
considerations.Molecular graphs for a sampling of moleculesin equilibrium
geometrieswith widely different bonding propertiesare illustrated in Fig. 2.8.
The existenceand position of a bond or (3, - 1) critical point in this and
other figures is indicated by a black dot. More examplesof molecular graphs
are given throughout the remainder of the book. The recovery of a chemical
structure in terms of a property of the system'scharge distribution is a most
remarkable and important result. The representationof a chemical structure
by an assumednetwork of lines has evolved through a synthesisof observa-
tions on elemental combination and models of how atoms combine, parti-
cularly models of chemicalvalency-a model which statesthat the ability of a
given type of atom to form bonds, its valency, can be saturated and that
valency is determined by the number of valence electrons.A great deal of
chemical knowledge goes into the formulation of a chemical structure
and, correspondingly, the same information is successfullyand succinctly
summarized by such structures. The demonstration that a molecular
structure can be faithfully mapped on to a molecular graph imparts new
information-that nuclei joined by a line in the structure are linked by a line
through spacealong which electronicchargedensity,the glue of chemistry,is
maximally accumulated.Finding the physicalbasisfor a molecularstructure
also leads to a broadening of the concept that the dominant interactions
between atoms, be they attractive or repulsive,have a common phvsical
representation. This is not an entirely surprisingresult sincethe er er present
34 ATOMS AND TOPOLOGY OF CHARGE DENSITY 2.4

T H l',1

'7t--.^
t \ i / ' ^/f---+-c'
,.,,-t\, L ' ( \

'v-(/" H-+- C------e- C----+-H

- r/\

"r/'Y-1, ,)'i \r/

)<"
/\
,F>'1
i'-",
y v r
\
\ C"--+-C
, /
z

/
/
\
H

v\,/'\y'
JT t
/ \
\
,/T / tH

r \*
i
7t-..----,
H\ .H

H \.----"-'/
\
t-.*--o '\-!/'
,z\,2'f
l / r \
\---t l-------'-o
i
/ r f
I
H i i
A A
FIc, 2.8. Molecular graphs for some molecules in their equilibrium geometries.A bond critical
point is denoted by a black dot.
( From Bader et al. (1981).)

nuclear excursionsfrom an equilibrium separation between a pair of atoms

force a sampling of thesesameportions of a potential surfaceeven though the
atoms are consideredto be bonded to one another. It is the closely related
questions of what is meant by the making and breaking of chemical bonds
that lead one to consider the most important extension of the molecular
structure concept. As discussedin Chapter 3, the dynamic behaviour of the
molecular graphs, as causedby the relative motions of the nuclei, forms the
basisfor the definition of structural stability and the analytical description of
the mechanismsof structural chanse.
 2,4 E L E M E N T SO F M O L E C U L A RS T R U C T U R E 35

that a bond path is not to be understoodasrepresenting

It is to be stressed
a 'bond'.The presence of a bond path linking a pair of nucleiimpliesthat the
I
corresponding atomsare bondedto one another.As demonstrated later,the
interactioncan be characterized and classifiedin termsof the propertiesof
the chargedensityat its associated(3, - l) critical point. The complete
descriptionof the interaction,however,requiresthe evaluationof operators
over the associatedinteratomic surface.These are the topics of later
discussions.At this point we continue with the identificationof the
elementsof molecular structure with the topological propertiesof the
remainingstablecriticalpoints,(3, + 1) and (3, * 3) criticalpoints.

2.4.3 Ringsand cages

The remainingcriticalpointsof rank threeoccurasconsequences of particu-

lar geometrical arrangements of bond paths and they define the remaining
elementsof molecularstructure-rings and cages. If the bond paths are
in
linked so as to form a ring of bondedatoms,as the bridging plane of the
diboranemolecule,for example,whosemolecular graph is shown in Fig. 2.8,
As
thena (3, + 1)criticalpoint is found in the interior of the ring. discussed
aboveand illustratedin Fig. 2.6, the eigenvectors associated with the two
positiveeigenvalues of the Hessianmatrix of p at this criticalpoint generate
an infinitesetof gradientpathswhichoriginateat the criticalpoint anddefine
a surface,calledthe ring surface.This behaviouris illustratedhereby the
gradientpathsin the bridgingplaneof the diboranemoleculeas shownin
Fig.2.9.All of the trajectorieswhich originateat the critical point at the
centreof the ring of nuclei,the (3, * 1)or ring criticalpoint,terminateat the
ring nuclei,but for the set of singletrajectories,eachof which terminatesat
one of the bond criticalpointswhosebond pathsform the perimeterof the
ring. Thesebond paths are noticeablyinwardly curved away from the
geometrical perimeterof the ring,a behaviourcharacteristic of systems which
are electron-deficient. The remainingeigenvector of a ring criticalpoint, its
singlenegativeeigenvector, generatesa pair of gradientpathswhich termin-
ate at the critical point and definea unique axis perpendicularto the ring
surfaceat the criticalpoint.In diboranethisaxisis perpendicular to the plane
shownin Fig. 2.9.\t represents theintersection of the boundaries of the basins
of the hydrogenand boron atomsformingthe ring. A ring,as an elementof
structure,is definedasa part of a molecular graphwhichboundsa ring surfoce.
If the bond pathsare so arrangedas to enclosethe interior of a molecule
with ring surfaces, thena (3, * 3)or cagecriticalpoint is foundin the interior
of the resultingcage.The chargedensityis a localminimumat a cagecritical
point. The phaseportrait in the vicinity of a cagecriticalpoint is shownin
Fig.2.10for the tetrahedrane molecule.Trajectoriesonly originateat sucha
criticalpoint and terminateat nuclei,and at bond and ring critical points,
 = o; o :F
-
c - =

€ 5.E
o t i c )

-^ C S5 o -E
3 5 @

$-=
r t r Q
ar'E c.r
9?X-q
( ) d 0 )

7 o "
; boh
( ) ! F

F:E
7 ; i a
# . r r €

g ;.9 ii
g - H f r

b g P E
o'F i5
E U F c . ,
a * € 9

oo* b 9
- H

b0.i - ()
i v , i

L d L

J ) L . q i r
e c a e

9 F - :
i U i :
. i - q *

-
>'- r - v
o I
! v 9 , r
o

H H 9 -

g o d B
- > ' o . .
-y ( {r ^E. :: : ;
d
d 9 . - =
; c bop
i:.-'- .-
. 9 d ' t
c ! E -
o
ts-
9 ct gr . =
q

o';'i o ()
+ ; ; H
! ., s.=
v F ' ^ -

.- bOh

* ---C

q) (J'=

r t 5
a 5
- ^ a ao
X ()'-
> : o
: Y 6
ua $
d h -
OO c..r b
a . r R o

o >,tr
K F.:
. ! a €
> o 3c !s
'

4'*
o\
' : 9
-
v o r 9
2,4 E L E M E N T SO F M O L E C U L A R S T R U C T U R E 37

thereby defining a bounded region of space.A cage.as the final element of

molecular structure, is a part of a moleculargraph tt'hichcontainsat least two
rings, such that the union of the ring surfacesbounds a region o/ Rt which
containsa(3, + 3) critical point.While it is mathematicallypossiblefor a cage
to be bounded by only two ring surfaces,the minimum number found in an
actual moleculeso far is three,as in bicyclo[1.1.l]pentane,for example.
The representationof the gradient vector field of the tetrahedranemblecule
for the plane shown in Fig. 2.10is of some interest as all four critical points of
rank three are present. As a consequenceof the symmetry present in this
molecule,the bond paths that are shown indicate that each carbon is linked
to a single hydrogen and to three other carbons to form a cage. The
carbon-carbon and carbon-hydrogen bond critical points linking the nuclei
in this plane appear as saddlesin the chargedensity (seeFig. 2.4) and as (2, 0)
critical points in terms of their phase portraits. The two trajectories termin-
ating at each such point and defining the intersection of an interatomic
surfacewith this plane are shown, as are the two trajectoriesthat originate at
each of thesepoints and define a bond path. The critical point at the top of
the diagram is also a bond critical point, the one whose bond path links the
two out-of-plane carbon nuclei.The interatomic surfacefor this pair of nuclei
coincides with the plane of the diagram and hence this bond critical point
appearsas a(2, - 2) critical point, a maximum in p. Indeed,in this particular
plane the phase portrait of this critical point is topologically equivalent to
that for a nucleus.A bond critical point behavesas an attractor in this one
specialsurface,the interatomic surface.The two remaining critical points are
ring critical points and in this plane they also appear as saddlesin p and as
(2, 0) critical points in terms of their phase portraits. Of the two trajectories
which originate at each of these ring critical points, one terminates at a
carbon nucleus, the other at the upper bond critical point. These pairs of
trajectories denote the intersection of two corresponding ring surfaceswith
the plane of the diagram. One notes that the ring surfaces,as well as the
carbon-carbon bond paths which together bound the cage,are all outwardly
curved from the geometricallines and facesof the tetrahedral structure. This
behaviour is, in general,characteristicof organic moleculeswith geometrical
strain. Of the two trajectoriesthat terminate at each of the ring critical points
and define the unique axis perpendicularto the ring surface,one originatesat
the cagecritical point, the other at infinity. Thesetwo ring axes,together with
the two trajectories terminating at the carbon-<arbon bond critical point,
mark the intersections of the interatoriric surfaces with the plane of the
diagram. The two ring axesare themselvesintersectionsof the boundaries of
the basins of three carbon atoms.
The number and type of critical points which can coexistin a systemwith a
finite number of nuclei are governed by the Poincare Hopf relationship.
With the above associationof each type of critical point with an elementof
38 ATOMS AND TOPOLOGY OF CHARGE DENSITY 2.4
2.4 E L E M E N T SO F M O L E C UL A R S T R U C T U R E 39

molecular structure,this relationshipstatesthat

n-blr-c:l (2.10)

where lr is the number of nuclei, b is the number of bond paths (or atomic
interaction lines),r is the number of rings, and c is the number of cages
(Collard and Hall 1917).The collection of numbers (n, b, r, c) is called the
characteristicset of the molecule.
This chapter has demonstratedthat elementsof the molecular structure
hypothesis,atoms, and bonds and the linking togetherof the atoms to form
chains, rings, and cages,as evidencedby the forms of the molecular graphs,
follow directly from the topological properties of the electronic charge
distribution. The experimental development of the structural aspectsof this
hypothesis thus appears as an inevitable consequenceof the form and
properties possessedby the charge distribution, properties which themselves
are a reflection of the forces acting within the system.
The finding that the charge distribution contains the information required
to define a molecular graph for every geometricalarrangement of the nuclei
allows for much more than a pedestrianrecoveryof the notion of structure as
a set of atoms linked by a network of bonds. Instead, the dynamical
behaviour of the molecular graphs, as induced by a motion of the system
through nuclear configuration space,is shown to be describablewithin the
framework of a relatively new field of mathematics,one which demonstrates
that the notions of structure and structural stability are inseparable.The
application of theseideasto the dynamical propertiesof the molecular graphs
defined by the charge density leads to a complete theory of structure and
structural stability, one which predictsthat a changein structure as causedby
the making or breaking of chemicalbonds is an abrupt and discontinuous
process.It is this important aspectof the theory of atoms in moleculesthat we
develop in the next chapter.

Frc. 2.10. (a) A map of the gradient vector field of the electronic charge density for the
tetrahedranemolecule in a oo symmetry plane (seeFig.2.4). (b) A contour map of the electronic
charge density in the same plane showing the bond critical points, and overlaid with the bond
paths and the intersectionsmade with the interatomic surfaces.The diagram shows the phase
portrait for the (3, - 1)critical point betweenthe two out-of-planecarbon nuclei (denotedby the
upper dot) in the plane of its interatomic surfacewhere it behavesas a (2. - 2)criticalpoint (all
trajectories terminate at the critical point). The dot directly beneath this point denotes the
(3, - l)critical point betweenthe in-plane carbon nuclei and the phaseportrait in this caseis for
a p l a n e c o n t a i n i n gt h e b o n d p a t h w h e r et h e b o n d c r i t i c a l p o i n t b e h a v e sa s a ( 1 . 0 ) c r i t r c a lp o i n t .
The same plane is illustrated for the two C-H bond critical points. The central critical point is a
( 3 , + 3 ) o r c a g ec r i t i c a lp o i n t . I n t h i s . p l a n ei t b e h a v e sa s d o e sa ( 2 , + 2 ) c r i t i c a l p o i n t . u i t h a l l t h e
trajectoriesoriginating there.
 ATOMS AND TOPOLOGY OF CHARGE DENSITY Ez.I

Ez.l Local properties of p and associatedtheorems

82.1.1 Local maxima in p

With relatively few exceptions,the electronic charge density exhibits a local

maximum only at the position of a nucleus in both the ground and excited
statesof many-electronsystems,molecules,or solids.The value of the charge
density at a nuclear position, a quantity denoted by p(0), is, with the
exception of a proton, many times larger than its value at any other of its
extrema. The value of p(0) at the nucleus of a free atom in the Hartree-Fock
approximation is roughly proportional to the cube power of the nuclear
charge,the expressionp(0):0.479823'to" uu, having a mean deviation of
2.6 per cent for | < Z < 55. The symbol au denotes atomic units. Atomic
units for a number of properties are defined in Table A1 of the Appendix
which also givesthe factors for their conversioninto other units. On the other
hand the values of p at the saddle or bond critical points found between
certain pairs of nuclei over the range of chemically significant internuclear
separations,lie in the range 0 ( pu < 1.0au. It is the presenceand domi-
nance of the nuclear maxima in a chargedistribution that imposesthe atomic
form on the observed properties of a system.
As mentioned in the text, the local maxima rn p at the positions of nuclei
are not true (3, - 3) critical points becausethe gradient vector of the charge
density is discontinuous at the nuclear cusp that is present in both the state
function and the density (Kato 1957).However, there always existsa function
homeomorphic to p(r; X) which coincideswith p almost everywhereand for
which the nuclear positions are (3, - 3) critical points. In this sense,the
nuclear positions behave topologically as do (3, - 3) critical points in the
charge distribution.
The assumption that the charge distribution possesses local maxima at the
positions of the nuclei is the basis for the determination of crystal structures
by X-ray diffraction techniques.A crystal structure is determined by a least
squares fitting of the amplitudes of calculated structure factors to the
experimentally observedones. The calculated structure factor is based on a
model which assumesa Hartree-Fock spherical free-atom charge density
centred on each nucleus in a unit cell and convoluted with a Debye-Waller
thermal factor. By assigning the nuclear positions to the centroids of such
thermally averaged atomic charge densities in the fitted structures, one in
general obtains good agreementwith the geometrical structures determined
by neutron diffraction techniques. This agreement clearly supports the
assertionthat local maxima in p(r; X) occur at the positions of the nuclei but
it does not preclude the existence of maxima, of considerably smaller
magnitude, at other positions.
E2 .l OF p
LOCALPROPERTIES 4l

The principal justification for the obserration that local maxima in p(r; X)
generally occur only at the positions of the nuclei comes from extensive
studiesof theoreticallydeterminedchargedistributionscalculatedfrom state
functions closeto the Hartree-F-ocklimit. There is well-documentedevidence
that molecular Hartree-Fock charge densities(charge densities obtained
from single determinant SCF calculations carried essentially to the
Hartree-Fock limit) are accurate to within 2 5 per cent of the total charge
density at any point in space,exceptperhapsfor very large distancesfrom the
nuclei (Cade 1972;Bader 1975;Smith et al. 1977).This degree of accuracy,
and the correspondingly good reputation which Hartree-Fock wave func-
tions enjoy in the prediction of related one-electronproperties,are accounted
for in terms of Brillouin's theorem (Brillouin 1933-34; Moller and Plesset
1934).Cade and co-workers generateda set of wave functions for diatomic
moleculesall close to the Hartree-Fock limit using an SCF method with a
carefully selectedset of Slater basis functions together with optimization of
the orbital exponents (Cade and Huo 1973, 1975; Cade and Wahl 1974).
Mclean and Yoshimine (1967) have published wave functions of slightly
lesser quality for a number of linear polyatomic molecules. The charge
densitiesderived from these functions have been analysed using a program
which locates all of their critical points and classifiesthem in terms of the
eigenvaluesof their Hessianmatrices.The resultsof this topological analysis,
which included chargedensitiesof ground and excitedstatesof neutral as well
as of positively and negatively charged species,have been described pre-
viously (Bader et al. 1981).All 50 second-and third-row diatomic hydrides
AH, all 83 heteronucleardiatomicsAB, and all l3linear polyatomicsexhibit
local maxima only at the positions of the nuclei. In all thesemolecuJes,save
two excitedstates,a single(3, - 1)critical point is found on the internuclear
axis between neighbouring nuclear maxima. The only exceptions to these
findings were for a small number of the 47 homonuclear diatomic molecules
which exhibited a small maximum in p at the midpoint of the internuclear
axis. The values of the charge density at theseexceptionalmaxima exceeded
the valuesat the neighbouring minima on the internuclear axis by very small
amounts, within the error of the Hartree-Fock predictions. For example,in
Cr(Xt Ir* ) the value of p at the midpoint of the internuclear axis is 0.3076au,
a value which exceedsthat of the neighbouring minima by 0.0029au, i.e.
approximately I per cent. Thus the maxima that have been found at other
than the nuclear positions are so small as to be within the limits of error
placed on the calculated densities and are trivially small compared to the
maxima found at the nuclear positions. In the above state of Cr, the value of
p at a carbon nucleusis 127.323au. The valuesof p at the (3, - l)critical
points for the ground state chdrge distributions of the diatomics are given in
the Appendix, Table A3.
42 ATOMS AND TOPOLOGY OF CHARGE DENSITY E2.I

The topological properties of the charge distributions of a very large

number of polyatomic moleculesderived from basis setsof varying quality
have now been analysedin many laboratories. These systemsinclude many
different examplesof what are conventionally called bonded pairs of nuclei.
In almost all casesthe only maxima in p(r; X) are found at the nuclei and,
between each such pair of nuclei, a (3, - 1) critical point is observed.
Occasionally, spurious maxrma, again small in magnitude, are observedfor
charge distributions obtained from approximate wave functions expandedin
terms of small or unbalanced basis sets or basis sets without polarizing
functions. These spurious maxima are not present in charge distributions
derived from a calculation using a larger or improved set of basis functions.
Multiple bonds and other bonds containing a diffuse density component are
difficult to describe using Gaussian basis sets.Large Gaussian sets yield a
small maximum in p(r; X) at the midpoint of the C-C axis of acetylene,for
eiample. This spurious maximum is replaced by a proper (3, - 1) critical
point when a singlesand doubles configuration interaction calculation using
the same basis set is performed. This spurious maximum in p(r; X) is not
presentin the density for acetyleneor in substitutedacetylenesobtained from
a single determinant self-consistentfield (SCF) calculation using a large
Slater basis set which includes/functions.
A number of studieson the effectsof Coulomb electron correlation on the
charge density obtained in a singledeterminant SCF calculation have shown
them to be relatively small. This result is not unexpectedsinceHartree-Fock
charge distributions are, becauseof Brillouin's theorem, correct to second-
order. It is essentialthat electron correlation computed by a configuration
interaction (CI) method include single as well as double excitations with
respectto suitablezero-order referencewave functions.Bender and Davidson
(1968)have pointed out that, with the inclusion of only double excitations,
the highly occupied natural spin orbitals will always be simply a trans-
formation of the original occupied SCF orbitals. Thus double excitations,
while accounting for an important fraction of the correlation energy, play
only a minor role in correcting the chargedensity.Only the inclusion of single
excitationsbrings about a significantmixing of the virtual orbitals with the
original occupied SCF set to yield improvements in the SCF charge density.
The one-electroncorrelationfunctionsof Sinanoglucorrespondto the sum of
single excitationsand, as pointed out by Sinanoflu and Tuan (1963),they
representthe corrections to the Hartree-Fock orbitals.
A study, using such CI methods together with the generahzedvalencebond
and multi-configurationSCF methodsfor introducing Coulomb correlation,
demonstratesthat the correlated charge distributions possessthe same
number and kind of critical points in both the p(r) and Y2p(r) fields as are
found for the SCF distributions(Gatti et al. 1988).Thus, the topology of a
E2.l L O C A L P R O P E R T I E SO F P 43

charge distribution and the structure it definesare unaffectedby the addition

of Coulomb correlation. The study covered a wide range of chemical
structures, and the quantitative changes in the properties of the charge
density at the critical points induced by correlation are found to be small in
magnitude and to be more pronounced for shared or covalent interactions
than for systemswith pronounced charge transfer between the atoms. This
study and earlierones(Bader 1975;Smith et al.1977)showsthat, in general,a
Hartree-Fock charge density overestimates,by a few per cent, the value of
p(r)in the neighbourhood of a bond critical point relative to that predicted by
a correlated wave function. The effect is greatest for interactions with a
significant shared density and a relatively small value for the curvature of p
along the bond axis.
Excited statesof one-electronatoms do exhibit non-nuclear maxima as can
excited states of one-electron molecular ions, as demonstrated for Hj by
Bateset al. (1953).There is no doubt that the successof the atomic conceptin
the classificationand prediction of chemical knowledge is a reflection of the
fact that, in most systems,nuclei are the sole attractors of the charge density
and the atomic form is dominant. There is, however, the possibility of
observing non-nuclear maxima that could be responsible for particular
physicaleffects.An apparent example of such behaviour has been found. The
Li, molecule is among the homonuclear diatomics that exhibit a small
relative maximum in the charge density at the bond midpoint at the
Hartree-Fock level.It has now been demonstratedby Gatti et al. (1987)in a
study of charge densitiesof clusters of Li atoms, including electron correla-
tion, that the central non-nuclear maximum in the ground-state charge
distribution of Li, is not an artefact of the singledeterminantal state function
but is instead a property not only of this chargedistribution but of all lithium
clustersthey investigatedup to and including Liu. The same behaviour has
also beenfound by Cao et al. (1987)for clustersof sodium atoms. The value
of p at a non-nuclearattractor is small, 0.0078au, and exceedsthat at the
(3, - 1) critical points to which it is linked by a very small amount-
I 1 x 10-4 au for the sodium clustersand only slightly greaterthan this for
the lithium clusters.The non-nuclearattractorsin p in theseclustersbehave
as pseudoatomsand it is found that the metal atoms are not linked directly to
one another but only indirectly through an intervening pseudoatom
tFig.21 1). It appears that the optimum structure for planar lithiurn and
sodium clusterscorrespondsto a network of linked pseudoatoms,each of
w'hichis bonded to a singly bonded metal atom and to two bridging metal
atoms. The metal atoms bear substantialpositive chargesand the clusters
correspondto positivelychargedmetal atoms with localizedchargedistribu-
tions bound by an intermeshednetwork of negativelychargedpseudoatoms.
The loosely bound and delocalizedelectroniccharge of the pseudoatomsis
44 ATOMS AND TOPOLOGYOF CHARGEDENSITY E2.I

Frc.2.11. Maps of the gradient vector fields of the charge density for Na, and Nan, displayed
in terms of trajectoriesof Vp, and contour maps of the corresponding charge densitiesoverlaid
with bond paths and lines denoting the intersectionsof the interatomic surfaceswith the plane of
the diagrams.The latter trajectoriesof Y p are denoted by heavy lines in both setsof maps. There
are two bonds in Na, and sevenbonds in Nan, The positions of the non-nuclear attractors are
denoted by stars on the gradient vector field maps while the positions of the bond critical points
are denoted by dots. Each (3, * 1) or ring critical point in Nao is located at an intersection of
three gradient paths as occurs immediately above and below the central bond critical point in the
gradient vector field map.

responsiblefor the binding in these systemsand for the conducting proper-

ties,assumingthe samestructure existsin their bulk metals.The properties of
the pseudoatomsare futher discussedin Section 7.4.2.
Another changein the usual topology of p can occur when two nuclei with
disparate charges approach very close to one another. Under these condi-
tions the charge density will continue to exhibit the required cusp at the
position of eachnucleusbut, at somesmall separation,the local maximum at
E2 .l L O C A L P R O P E R T I E SO F p 45

the position of the lesserof the two chargeswill disappear.(This is an example

of what is termed a fold catastrophe,a subjectto be discussedin Chapter 3. In
this case the (3, - 1) critical point initially between the nuclei approaches
and eventually merges with the (3, - 3) nuclear maximum. Both critical
points vanish upon their coalescencebut the cusp condition at the nucleusis
still obeyed.)This behaviour is not observedfor energiesnormally encoun-
tered in chemical reactions for an atom that possesses a core density, as the
value of p(0) is then very great and extremely short internuclear separations
would be required to increasethe value of p at the intervening(3, - 1) critical
point to where it equalled p(0). Hydrogen, however, does not possessa core
and, when in combination with a very electronegativespeciessuch as F+, its
nuclear maximum is absent (Bader et al. 1980).In this molecule the total
charge distribution is dominated by the forces exerted by the partially
unshieldedF nucleusand the propertiesof the molecule FH * approach those
of the correspondingunited atom limit, Ne *. This behaviour is illustrated in
Fig.2.l2.

E2.1.2 Theoremsconcerningthe propertiesof p

Hoffmann-Ostenhofand Morgan (1981)were able to prove that the ground-

state charge distribution of a one-electronhomonuclear diatomic molecule
can exhibit maxima in p only at the positionsof the nuclei.In this proof an
important inequality is used (Hoffmann-Ostenhof and Hoffman-Ostenhof
t977).
- + V tp ' t 2 $ )+ ( e- Z l l r l ) p t r 2 ( r() 0 (E2.r)
where e is the first ionization potential. Unfortunately, only a few theorems
concerning the local behaviour of the exact charge density are known. One,
derived from the nuclear-*lectron cusp condition on the state function (Kato
1957),relatesthe value of the chargedensity at a nucleus,p(0),to the spherical
averageof the derivative of p(r) (Steiner 1963;Bingel 1963)

lim[dp(r)/dr]"": - 22p(0). (F,2.2)

An upper bound on the value of p(0) in an atom, given by Hoffmann-

Ostenhofet al. (1978)is

p(0)< (Zl2n)I r- 2p(r)dr. (F,2.3)

The long-range behaviour of p(r) for both atoms and moleculeshas been
discussedby a number of authors (Ahlrichs 1972;Hoffmann-Ostenhof and
Hoffmann-Ostenhof1977:Tal 1978;Katriel and Davidson 1980).The results
 !o.367

o.277

o.197

-o.3 -o.2 -o.1 o.o o.1 0.2 0.3 \

ry

*
Frc.2.l2. Profile of the chargedensity along the internuclearaxis of the HF molecule and the
gradient vector field of the charge density for a plane containing the two nuclei. A trajectory of p
originates at infinity, passesthrough the proton, and terminates at the F nucleus.
F'2.2 OF P
M A T H E M A T I C A LP R O P E R T I E S 47

of thesestudiesshow that the chargedensity,at a sufficientlylarge distance

from all nuclei,decaysexponentiallyaccordingto
p (r) - e x p l - (2 4 )12 r] (E2.4)

where e is the first ionization potential of the system'

F;2.2 Mathematical propertiesof p at a critical point

E2.2.1 Eigenvaluesand eigenvectorsof Hessianof p

In general, there are nine second derivatives of p(r) with respect to an

arbitrarily chosen set of coordinate axes.Their ordered atray is called the
Hessianmatrix, or simply the Hessianof p. This matrix, when evaluatedat
the position of the critical point r. and denoted by A(r.), is given by

I o" ?'P t'P \

A(r"):
( ff , * ) '-',
(E2.5)

\erer ?rty ?l f
Since it is real and symmetric, A(r.) can always be put in a diagonal
form. This correspondsto finding a rotation of the original coordinate
systemr(x, .y r; *hi.h aligns the new coordinate axes /(x', y', z') with the
principal axes of the critical point. The diago nahzed form of A is denoted
by the matrix A

:(l
;l)
ff+f
A _ (F,2.6)

axis'
in which eachelementis the curvatureof p(r") with respectto a principal
unitary
The matrix which generatesthis coordinate transformation is a
matrix U and. if r and r' denote row vectors,one has
r' : rU. (E1.7)
48 ATOMS AND TOPOLOGY OF CHARGE DENSITY 82.2

The matrix U is obtainedby assumingthat the matrix A satisfiesthree

equationsof the form
eigenvalue
Au' : 2'u' (E2'8)

where u, is a column vector in the matrix U which expressesthe tth

component of the vector r' in terms of the componentsof r, eqn (E2.7).Thus,

lr^ utz,,.\
U:lu^ uzz urrl E2.g)
\"' utz "'l
where g- t - Ur since U is unitary. All three equationsfor the u, can be
expressedin a single matrix equation

AU: UA. (E2.10)

Multiplicationof eqn(E2.10)
from the left by U-' subjectsA to a similarity
transformationwhich placesit in its diagonalizedform (eqn(E2.6)),

U _ 1 A U_ U _ I U A : A . (E2.1)

Thus, if A solvesan eigenvalueequation of the form (E2.8),then a coordinate

transformation can be found which will yield a diagonalized form of the
matrix. The matrix A is itself a tensor (see Section E5.2) which can be
constructedby taking the product of a column and a row vector

V.V,P:
,,zivt:
H(fl
A(r)

where the derivativesare taken at r". Thus, the similarity transformation can
be expressedas

u- lAU - U- lv,v.pu : Y,,Y,,p

: A.,

making clear that the diagonalization results from a transformation of the

coordinates used in the evaluation of the second derivatives.
The eigenvalues7,and eigenvectorsu, of eqn (E2.8)are obtained through
the solution of the correspondingdeterminantalequation

d e t l A - 2 , I :16 (E2.r2)
82.2 OF p
M A T H E M A T I C A LP R O P E R T I E S 49

where I is the unit matrix. The associatedeigenvectorsu, are obtained by

so l vi n gt he equat ionfo r e a c h v a l u e o f ),
(A-2;I)u':Q. (82.13)

82.2.2 Analytical expressionfor a trajectory in the vicinity of

a c r it ic al poi n t

It is possibleto obtain an analyticalexpressionfor the points r(s) on a given

trajectoryof Vprin the neighbourhoodof a critical point. This correspondsto
solving the differentialequation (2.7)and the solution may be expressedin
terms of the eigenvalues ,,1,
and eigenvectorsu, of the Hessianmatrix of p(r) at
r' as
r(s) : ('rule'^r''+ t'rure)" + 6.u.eit' (82.14)

where c1,c2, and c. are arbitrary constants.Equation (E2.14)statesthat the

vectors r(s), which for varying s define the successionof points lying on a
given trajectory,are expressibleas a linearcombination of the three mutually
orthogonal eigenvectors,u,. Because of this simple linear relationship
betweenr(s) and the u;, one may use eqn (E2.14)to predict the form of the
phaseportraits shown in Fig. 2.6.
Equation (82.14)is obtainedby first expressingp(r) in the form of a Taylor
seriesexpansionabout the position of the critical point, r.. In what follows it
i s to be under s t oodth a t a l l d e ri v a ti v e so f p (r" ) a re eval uatedat r: r" . The
componentsof the position vector r, denoting a point in the vicinity of the
critical point rc, are given by (-x,1',z).
Writing out explicitly only the first few terms in the expansion and
recallingthat the first derivativesof p(r) vanish at r", one obtains
p( r ) : p (r.) + l l 2 (r - .x " )(? 2 p l ? x)(,x 2 - -x" )

+ l l 2 ( x - - x " ) ( i 2 p l ? r l - y ) ( ,- 1 - r ' " )

+ l l 2 ( x - , x " () A 2p l t r ? z ) ( z- z , ) +
* remainingsecond-order
termsand third-
and higher-order
terms. (E2.1s)
The differentialexpressionfor Vp(r)in the neighbourhoodof r" (seeeqn (2.7))
i s o b ta inedby t ak ing th e g ra d i e n to f e q u a ti o n(E2 .15),

qq: v p ( r ( s:) )i { ( x - , x " )@ 2 p l ; ' x+2()r , - y . )( ? 2 p l ; ' r e y )

ds

+ (z- z,)(?2pl6x1z)l
+ j { ( x - - x " ) ( 6 2 p l } y C x+) ( - } ' - - y " ) ( O 2 1
p ,l e
2)
-r (z - z")(E2pl1yA) + (E2.16)
50 ATOMS AND TOPOLOGY OF CHARGEDENSITY E2.2

By retaining only the first-order terms in (r - r") eqn (E.2.16)may be

expressedas
c-p
r ) a t
o-p \
cx-
^ ) A
0xcy
^
\
V p ( r ( s -) [) x - x " !-!" z- z , ( 0-p
1
cycx
^',

1
1)
o-p
1
cy-
)
I
I
I
Thus,the differentialeqn(2.7)for a point r in the neighbourhood
of a critical
point is givenby
d r ( s ) / d s :( r ( s -) r " ) ' A (F,2.17)
whereA is the Hessianmatrix definedin eqn (E2.5).
in terms of the diagonalform of A and its
Eqn (82.17)may be re-expressed
eigenvectorsas

- r.)uLtrAU
(r(s)
%P:

dr'(s) I /"r \
"
: [ x. ' -, x i , !'-!'" z'-z'.f I
\
A2
A3
I (E2.18)

Each component on eqn (E2.18)definesa first-order differential equation of

the form
dr'(s)
: /tlx(s'
dt

whose solution is of the form

x ' ( s ): c ' ' e i ' ' '

Hence the general solution to the differential equation for Vp in the neigh-
bourhood of a critical point is as given in eqn (E2.14).
The eigenvectorsu, span the spacein the neighbourhood of a critical point
and, by fixing the valuesof the constantsc;, on€ choosessomeinitial point on
a particular trajectory. One may then follow the successionof points r(s) and
determinethe trajectory by varying the path parameters. If c2: c3:0, for
example, then r(s) lies on the axis defined by u,. If the corresponding
eigenvalueAr 10, then the coefficientei'' decreases as s increases.
Hence the
trajectory of the points r(s) approachesr" along u, and the trajectory
terminatesat r" when s : + :. as illustratedin Fig.2.6 for the case(1, - 1).
 REFERENCES 5l

Correspondingly,if Ar > 0, the points r(s) move away from rc as s increases,

the trajectoryis directedaway from r", and it terminatesat someother critical
pointwhens:*co.
Any trajectory describedby a combination of the eigenvectorsassociated
with both the positive and negativeeigenvalueswill neither originate nor
terminate at the critical point but instead will curve so as to avoid r.. This is
illustratedin Fig. 2.6 for the case(2,0). The two eigenvectorsassociatedwith
the negativeeigenvaluesof a (3, - 1) critical point define a surface,planar in
the immediate neighbourhood of r", and all of the trajectoriesin this surface
terminateat r.; p(r.) is a maximum in this plane.The eigenvectorassociated
with the singlepositive eigenvalueof a (3, - 1) critical point definesa unique
axis, perpendicular to the plane described above, and the two trajectories
lying on this axis originate at r". If the plane defined by the eigenvectors
associatedwith the two negative eigenvaluesis not a symmetry plane of the
molecule,then the linear nature of eqn (E2.10)does not obtain beyond the
immediateneighbourhoodof r.. In this casethe surfaceassociatedwith the
(3, - 1)critical point will be curved,but still well-definedby the unique set of
trajectorieswhich terminate at the critical point as illustrated in Fi5.2.6.
The two eigenvectorsassociatedwith the two positive eigenvaluesof a
(3, +1) critical point also define a unique surfacebut in this caseall of the
trajectoriesin this surfaceoriginateat the critical point and p(r)is a minimum
at r.. The third eigenvectordefinesa unique axis and the two trajectorieson
this axis terminate at r". Thus the phaseportraits of a (3, * 1) critical point
are the reverseof those found for a (3, - 1) critical point.

References

R. (1972).
.A,hlrichs, Chem.Phys.Lett. 15,609.
Bader,R. F. W. (1975). MTP International Reuiewof Science. PhysicalChemisfty,Ser'
2, Vol. l. Theoretical chemistry(ed.A. D. Buckinghamand C. A. Coulson),p- 43.
Butterworth,London.
Bader,R. F. W. and Beddall,P. M. (1972).J. Chem.Phys.56, 3320.
, F . W . and E s s e nH, . (1 9 8 4 ).
Ba d e rR. 1943.
J .C h e m 'Ph y s .80,
Bader,R. F. W., Anderson, S.G.,and Duke,A. J. (1919a). J. Am.chem.Soc'101,1389.
Bader,R. F. W., Nguyen-Dang, T. T., and Tal, Y. (1979b). J. Chem.Phys.70,4316.
Bader,R. F. W., Tal, Y., Anderson, s. G., and Nguyen-Dang, T. T. (1980).IsraelJ.
C h e m . 1 98 ,.
Bader,R. F. W., Nguyen-Dang, T. T., and Tal, Y. (1981).Rep.Prog.Phv-s' 44,893'
Bates. D. R..Ledsham. K., and Stewart, A. L. (1953). Phil.Trans.r. Soc'..London246A,
215.
Bender,C. F. and Davidson,E. R. (1968).J. Chem.Ph,t's.49.1222'
Bingel,W. A. (1963).Z. Naturforsch.Al8,1249.
Brillouin,L. (1933-34). Actualifies Sci.Ind.7l, Sections159.160.
Z6FF-'o;a
Cade,P. E. (1972\. Trans. Am. crystalloqr. Assoc.8, l. /
/ \- at-
"/z
/-
/: i
a
52 ATOMS AND TOPOLOGY OF CHARGE DENSITY

Cade, P. E. and Huo, W. M. (1973).Atom. Data nucl. Duta Tables12,415.

Cade, P. E. and Huo, W. M. (1975).Atom. Data nucl. Data Tables15. 1.
Cade. P. E. and Wahl, A. C. (1974).Atom. Data nucl. Data Tables13, 339.
Cao, W. L., Gatti, C., MacDougall, P. J.,and Bader,R. F. W. (1987).Chem.Phys.Lett.
141,380.
Collard, K. and Hall, G. G. (1977).lnt. J. QuantumChem.12, 623.
Faraday, M. (1855).Experimental researchesin electricity, Vol. III, paragraph 3234,
Dec. 1851,Taylor and Francis.Reprinted in 1965,Dover, New York, N.Y.
Gatti, C., Fantucci, P., and Pacchioni,G. (1987).Theor.Chim. Acta 72, 433.
G a t t i , C . , M a c D o u g a l l ,P . J . , a n d B a d e r ,R . F . W . ( 1 9 8 8 )J. . C h e m . P h y s . 8 8 , 3 7 9 2 .
Hoffmann-Ostenhof, M. and Hoffmann-Ostenhof, T. (1971). Phys. Reu. A16, 1782.
Hoffmann-Ostenhof,T. and Morgan III, J, D. (1981).J. Chem.Phys.75,843.
Hoffmann-Ostenhof,M., Hoffmann-Ostenhof,T., and Thirring, W.(1978).J. Phys.B.
Atom.Mol. Phys.ll, L571.
Kato, W. A. (1957),Commun.pure appl. Math. 10, 151.
Katriel, J. and Davidson, E. R. (1980).Proc. nat. Acad. Sci.,USA 77,4403.
Mclean, A. D. and Yoshimine, M. (1967). Tables of linear moleculewauefunctions,
IBM J. Res.Deu. 12,206.
Moller, C. and Plesset,M. S. (1934).Phys. Reu. 46,618.
Runtz, G. R., Bader, R. F. W., and Messer,R. (1977).Can. J. Chem.55,3040.
Sinanoglu,O. and Tuan, Fu-Tai, D. (1963).J. Chem.Phys.38, 1740.
Smith, V. H., Price, P. F., and Absar,l.(1977). Israel J. Chem.16, l87.
Steiner,E. (1963).J. Chem.Phys.39,2365.
T a l , Y . ( 1 9 7 8 )P. h y s .R e u .A 1 8 , 1 7 8 1 .
I

.|
1

M O L E C U L A R S T R U C T U R EA N D I T S C H A N G E

No one really understands the behaviour of a molecule until he knows its

structure. . .
C . A . C o u l s o n( 1 9 7 2 )

3.1 The notion of structure in chemistry

The essentialunderstandingand original intent associatedwith the notion of

structure in chemistry is that it be a generic property of a system.Structure
originally implied the existenceof a particular network of bonds, which was
presumed to persist over a range of nuclear displacements until some
geometrical parameter attained a critical value at which point bonds were
assumedto be broken and/or formed to yield a new structure. However, the
word 'structure' has over the years acquired a duality of meanings.This has
occurred for two reasons.The first was a result of our inability to unambigu-
ously assigna network of bonds to a given system.The secondwas a result of
our ever-increasingability to experimentally measurethe geometrical para-
meters which characterrzethe minimum energy nuclear configuration of a
system within the Born-Oppenheimer model. Thus, if one referencesthe
'molecular structure' in early books on descriptive chemistry, one
heading
finds a discussion of valency and bonding and of models for assigning a
network of bonds. Referencedunder the same heading in present-dayphys-
ical chemistry or spectroscopy texts are discussions of the experimental
methods used to determine the bond lengths and bond angleswhich are used
to characterrzean equilibrium geometrical structure. Frequently, one en-
countersdiscussionsof molecular structure where the two interpretations of
structure are used interchangeably,which is a most unfortunate practice,
particularly when the resulting confusion of ideas is used to criticize the
physical relevanceof the very notion of molecular structure.
It is most important to distinguish clearly between molecular geometry
and the original intent and use of the notion of structure in the molecular
srructurehypothesis.Geometry is a non-genericproperty sinceany infinitesi-
mal changein a set of nuclearcoordinates,denotedcollectivelyby X, results
in a differentgeometry.Molecular structure,on the other hand, was assumed
to be a genericproperty of a system.Any configurationof the nuclei X' in the
neighbourhoodof a given configurationX, while it has a differentgeometry,
shouldpossessthe samestructure,that is, the samenucleishould be linked by
the same network of bonds in both X and X'. Difficulties ascribed to the
 M O L E C U L A RS T R U C T U R EA N D I T S C H A N G E 3.2

notion of molecularstructureare the inabilitiesto assigna singlegeometrical

structure,averageor otherwise,to rotation- or inversion-related isomers,to a
molecule in an excited vibrational state with geometrical parameters very
differentfrom those for the samemoleculein its ground state,or to a molecule
in a 'floppy' state wherein the nuclear excursions cover a wide range of
geometrical parameters. In reality, these are shortcomings of attempts to
impose the classicalidea of a geometry on a quantum system.The nuclei,like
the electrons,cannot be localized in space and instead are described by a
corresponding distribution function. The definition of structure proposed
here recognizesthis essentialpoint and associatesa molecular structure with
an open neighbourhood of nuclear configuration space,with a corresponding
averagebeing taken over the nuclear distribution function.
It is demonstratedhere that the topological properties of a system'scharge
distribution enable one to assign a molecular graph to each point X in the
nuclear configuration space of a system. This assignment corresponds to
defining a unique network of atomic interaction lines for each molecular
geometry. A molecular structure is then defined as an equivalenceclass of
molecular graphs. This definition associatesa given structure with an open
neighbourhood of the most probable nuclear geometry and removesthe need
of invoking the Born-Oppenheimer approximation for the justification or
ratronahzationof structurein a molecularsystem(Baderet a\.1979,1981).By
defining all possible structures for a given system,the theory shows that a
changein structuremust be an abrupt and discontinuousprocess,one which
is describedin terms of the mathematical theory of dynamical systemsand
their stabilities.

3.2 The definition of molecular structure

3.2.1 The equivalencerelation-a qualitative discussion

We have so far discussedand illustrated the structural elementsof molecules

for a fixed nuclear geometry,that is, at isolated points X in nuclear configura-
tion space.This structural information is summarizedin the molecular graph
which indicates how the nuclei are linked for a given nuclear geometry. As
changes in the nuclear configuration occur, one anticipates changes in
structure and, correspondingly,changesin the molecular graph. A quantitat-
ive, detailed description of changesin the molecular graph is uninteresting.
Instead, we seek to identify in these changes an invariant property of the
system which embodies the notion and properties of molecular structure.
This aim is accomplishedthrough the use of a mathematical deviceknown as
an equivalencerelation as applied to the gradient vector fields of the charge
density. The definition of molecular structure is made precise through the
3,2 THE DEFINITION OF MOLECULAR STRUCTURE 55

languageof mathematics.However, the underl.v-ing ideasare easilygraspedin

their qualitative form and to this end we first illustrate them in terms of the
changesobservedin the topology of the charge density and in its associated
molecular graphs for a simple chemical change in structure. The concise
and more precise statement of structure and structural stability is given in
Section3.3.
It is important to realize from the outset that two different spac€sare
involved in the discussion of structure. First is real space, R3, the space
wherein the behaviour of the systemis determined and observedthrough the
properties of the charge density. This spacemay be referred to as behauiour
space.The topology of the charge density as displayed in behaviour spaceis
determined and controlled by the positions of the nuclei. Changes in the
relative positions of the nuclei causecorresponding changesin the distribu-
tion of electronic charge,and, as a result, structuresare created,they evolve,
and are destroyed.Chemical changesare said to have occurred. The relative
positions of the nuclei are determined by fixing a point in nuclear configura-
tion space,R0, by assigningvaluesto Q internal nuclear coordinates.This is a
to as
Q-dimensionalmathematical spacewhich -&y, alternatively,be referred
the control space.
Atoms exist in real space and they are defined by a partitioning of real
space. The notion of structure is abstract and it is defined through a
partitioning of control spacein a manner determinedby the behaviour of the
systemin real space.In this way structure is associatedwith an open region of
nuclear configuration spaceand necessarilysatisfiesthe condition that it be a
generic property of the system.
Supposea water molecule is at or in the neighbourhood of its equilibrium
geometry denoted by X", to give the first of the molecular graphs shown in
Fig. 3.1.One assumesthat the network of bonds denntedby this molecular
graph will persist as the nuclei execute vibrational motions in the neigh-
bourhood of the equilibrium geometry. In agreementwith this essentialidea
of the meaning of structure, the molecular graph remains unchanged----each
hydrogen nucleusremains linked to the oxygen nucleusby a bond path-for
all configurations X in the neighbourhood of X", even those correspondingto
considerable excursions of the nuclei from their equilibrium positions as
illustrated, for example, by the second of the molecular graphs shown in
Fig. 3.1.All of thesemolecular graphs are said to be equivalent.One obtains
an equivalent molecular graph for any arbitrary X in the vicinity of X",
that is, all nuclear configurations in the open neighbourhood of X" have
associatedwith them charge distributions which yield equivalent molecular
graphs.The structure denoted by any one of thesegraphs is said to be stable'
This is the generalidea behind the definition of a stablestructure-an open
region of nuclear configuration space, all points of which have the same
molecular graph. A molecular graph is determined by the gradient \ ector
56 M O L E C U L A R S T R U C T U R EA N D I T S C H A N G E 3,2

\
S
>__

b
 3.2 THE DEFINITION OF MOLECULARSTRUCTURE 57

o o
I
I Ii
I
__-]__- HI
I
HI
t
Frc. 3.1. Contour plots of p(r) and corresponding gradient vector fields for configurations
along the Cr" dissociativepath of ground-statewater. The molecular graphs and the interatomic
surfacesare superposedon the contour plots. (c) corresponds to a catastrophe point of the
bifurcation type. At this geometry a singularity in p(r) is found betweenthe protons. It is at this
geometry that a bond path is formed between the protons in the dissociation of water into
O(tD) + Hr(1Er*). (e) corresponds to a catastrophe point of the conflict type. The molecular
graphs in (f ) indicate how an asymmetric motion of the systemin this configuration causesthe
bond path from oxygen to switch to one of the protons.
(From Bader et al. (1981).)

fieldand,in the generalcase,theequivalence relationshipmustbeestablished

firstin termsof thisfield.Two vectorfieldsaresaidto beequivalentif onecan
map everytrajectoryof one field on to a correspondingtrajectoryof the
other.By applyingthis definitionto the gradientvectorfieldsof the charge
densitywe obtain an equivalencerelation betweenpoints in the nuclear
configurationspace.For everyarrangement of the nucleiwhichcorresponds
to somepoint X in nuclearconfiguration space, thereis a chargedensity
p(r; X) and an associated gradient
vector field Vp(r; X). If we usethe symbols
Vp(r;X) and Yp(r:X') to denotesuch fields for two configurations X and X',
(b),
saythosedepictedin Fig. 3.1(a), then we can state that the tw'o nuclear
configurations if
areequivalentonly their associated gradient vector fieidsare
58 M O L E C U L A R S T R U C T U R EA N D I T S C H A N G E 3.2

equivalent. The nuclear configuration X is structurally stable if X is an

interior point of its equivalenceclass. That is. one can always find a
neighbourhoodof a structurallystableconfigurationX, such that the neigh-
bourhood is totally containedwithin the region of structuresequivalentto X.
The equivalenceof gradient vector fields for two configurations X and X'
ensuresthat a critical point (a, o) in p(r; X) is mapped on to a critical point of
the sametype in p(r; X')and that a gradientpath connectinga pair of critical
points in p(r; X)is transformedinto a correspondingpath in Vp(r; X'),linking
the imagesof thesetwo critical points. Thus, the molecular graphs associated
with the set of equivalent gradient vector fields consistof the samenumber of
bond paths linking the same nuclei. These molecular graphs represent a
single structure and the region of nuclear configuration space containing
X and all configurations equivalent to it is called the structural region
associatedwith X.
Two molecular graphs are defined to be equivalent if and only if they
belong to the samestructural region. An equiualenceclassof moleculargraphs
is called a molecularstructure.Thus, a unique molecular structure is associ-
ated with a given structural region, and molecular structure,as defined above
through the equivalenceof molecular graphs, necessarilyfulfils the require-
ment of being generic.
Clearly, one could average the electronic charge distribution over the
nuclear distribution function generated by the zero-point vibrational
motions. This latter distribution peaks around the configuration X" and the
nuclear averagedmolecular graph would be equivalent to the one obtained
for this configuration, as it is obtained at all points in the neighbourhood of
X". In this manner one can dissociatethe notion of structure from molecular
geometry and define a molecular graph and a corresponding molecular
structure for the water moleculein its electronicand vibrational ground state.

3.2.2 Changesin structure

There are a number of dissociative pathways for ground-state water and

again, within the notion of structure, one anticipates that, for sufficiently
large nuclear excursions,bonds will be broken andlor formed to yield the
possible dissociation products. How are these anticipated changesin struc-
ture reflectedin the topology of the charge distribution as monitored by the
molecular graph? Consider the particular case of the minimum-energy
adiabaticpathway which maintainsCru symmetryand leadsto the formation
of an oxygen atom and a hydrogen molecule,
H 2 O ( 1 A ') * O ( ' D ) + H 2 ( 1 t s *) .

The gradient vector fields and molecular graphs for points along this
pat hway ar e i l l u s tra te di n F i g .3 .1 . tn F i g .3.1(b) the oxygen nucl eushas
3 .2 THE DEFINITION OF MOLECULARSTRUCTURE 59

recededfrom the protons which in turn have approachedone another.While

the atomic surfacesof the two hydrogen atoms approach one another, they
remain distinct as no (3, - 1) critical point has appeared between the
protons. Thus, as noted above, the gradient vector field and its molecular
graph for the configurationin Fig.3.1(b)are equivalentto thosein Fig.3.1(a)
and they representa unique structure,the open structure associatedwith the
equilibrium geometry of water. This region does not, however, contain the
remaining configurations shown in the figure whose molecular graphs define
other structuresand other structural regions for the system consisting of an
oxygen and two hydrogen atoms. The result of applying the equivalence
relation to the gradient vector fields of the charge density is, therefore, a
partitioning of nuclear configuration spaceinto a number of non-overlapping
structural regions, each of which is characterrzedby a unique molecular
structure. What is demonstrated next is that a change in structure, that is,
passagefrom one structural region to another, is necessarilyan abrupt and
discontinuous process.
The gradient vector field and the structure it definesshown in Fig. 3.1(c)are
not equivalent to those which preceded it. At this point an abrupt change
occursin the topology of the chargedensity.A new critical point in the charge
density appears between the protons. It is at this configuration along the
dissociativepathway that the bond is formed between the hydrogen atoms.
The reader should note that the changein the molecular graph monitors the
important change in the charge distribution which occurs at this critical
nuclear configuration X.-the accumulation of sufficient electronic charge
between the protons to define a line which links them and along which the
charge density is a maximum with respectto any neighbouring line.
The molecular graph for X" is not equivalent to those which precededit
and it representsa new structure. Unlike the structure from which it evolved,
however, it is unstable. A further infinitesimal displacement along the
reaction pathway causesthe new critical point to changethrough its bifurca-
tion into two new critical points creating a new pattern for the gradient
vector field and another structure. This new gradient vector field, shown in
Fig. 3.1(d),defines a ring structure and this one, like the structure which
preceded the unstable structure at X", is stable-it persists for arbitrary
motions of the nuclei in the vicinity of the reaction pathway and for some
further motion of the system along the pathway.
The new critical point formed between the hydrogens for the nuclear
configurationX" is unstablebecauseit has one zero curvature-the curvature
along the C, symmetryaxis.It is a critical point of rank two. Of the two non-
vanishingcurvaturesof the chargedensityat this point, one is positiveand is
associatedwith the line linking the protons,while the secondis negative.lying
along the axis perpendicularto the plane containing the nuclei. When the
systemis extendedpast X", the form of the charge densitl' u'ith respectto
60 M O L E C U L A R S T R U C T U R EA N D I T S C H A N G E 3.2

thesetwo curvaturesdoes not changebut the flat portion-the zero curva-

ture--is replacedby a ripplein p causingthe critical point to split in two. This
behaviour is illustrated in Fig. 3.2 which shows a seriesof profiles of the
charge density along the C, symmetry axis. Each profile is for a particular
value of Ro, the distanceof the oxygen nucleus from the midpoint of the
proton-proton internuclear axis. For configurations preceding the nuclear
configuration X" in which the separation Ro attains its critical value, the
charge density falls off monotonically from the oxygen nucleus along the C,
axis out to infinity. The only critical point on this axis for these configura-
tions is the maximum at the oxygen nucleus. There is an abrupt and
discontinuous change in this behaviour for the nuclear configuration X". A
point of inflexion appears in the profile of the density when Ro attains its
critical value, a point where both the first and second derivatives of p with
respect to the coordinate along the C, axis vanish. An increase in the
separationRo causesthe inflexion point to be replacedby two critical points,
one with a positive curvature and the other, between the protons, with a
negative curvature. This situation is again a stable one for it persists with
further increasein the parameterRo.
In this one particular example, all of the changesin the topology of the
charge density occur along this one axis and the changesoccurring in three
dimensionsare easilyobtainedfrom thoseindicatedin Fig. 3.2by adding one
negativeand one positive curvature to each of the critical points obtained in
the bifurcation. Thus, the new critical point betweenthe protons is a (3, - 1)

FIc. 3.2. Profiles of the charge density along the C, symmetry axis for the dissociation of the
water molecule.In (a) p decreasesmonotonically from a maximum value at the oxygen nucleus.
As Ro is increased,an abrupt change occurs in the form of the curve when a point of inflexion
appears(b). This unstable critical point exists only for this single configuration X" of the nuclei
and a further increasein Ro resultsin its bifurcation to yield a maximum, corresponding to the
formation of a bond critical point and a minimum. corresponding to the formation of a ring
critical point as typified by (c). The svstem has entered the region of the ring structure.
i 3.2 THE DEFINITION OF MOLECULARSTRUCTURE 61
I
I or bond critical point and the secondis a (3. * I ) or ring critical point. If one
I
reversesthe motion of the system starting from the ring structure, the same
bond and ring critical points approach one another as the distance Ro is
decreased,until at its critical value they coalesce,a negativecurvature of one
annihilating a positivecurvature of the other along their line of approach to
produce the inflexion point in p between the protons for the critical config-
uration X", as illustratedin Fig. 3.2.
The above example has illustrated one way of bringing about a change in
structure,through the creation of a critical point in the chargedensity of rank
lessthan three,a so-called degeneratecritical point. Such a critical point is not
stable with respect to the changes induced in the charge distribution by
arbitrary displacementsof the nuclei away from the critical configuration X".
Thus, the structure defined by the molecular graph associated with the
critical nuclear configuration X" is an unstable structure and exists for just
one particular configuration of the nuclei along the dissociativepathway. It
will be referred to as a bifurcation structure and the point in nuclear
configuration spacefor the critical configuration as a bifurcation point.
A further extensionof the parameter Ro leadsto the destruction of the ring
structure and to the formation of another degeneratecritical point. This one
results from the simultaneous coalescenceof the two O-H bond critical
points with the ring critical point. Passage of the system through this
bifurcation point in nuclear configuration spacecausesthe unstable critical
point in the chargedensity to degenerateinto a (3, - 1)critical point, thereby
yielding the new structure illustrated by the molecular graph in Fig. 3.1(e).
This structure, in which the oxygen nucleus is no longer linked to either
proton, persistsfor continued motion along the Cr, pathway and eventually
yields an oxygen atom and a hydrogen molecule.
The structure shown in Fig. 3.1(e)is not a stable structure. It is abruptly
changedinto the structure representedby the secondof the molecular graphs
in Fig. 3.1(f), or its mirror image,for any motion of the nuclei which destroys
the twofold symmetry axis.Thus, as is characteristicof all unstablestructures,
the structure in Fig. 3.1(e)is stable to some but not to all nuclear displace-
ments. It representsthe second and final type of instability that can be
exhibited by a molecular chargedistribution. While all of the critical points in
the molecular graph in Fig. 3.1(e)are of rank three and are themselvesstable,
the graph itselfis unstable.There are two (3, - 1)or bond critical points,the
bond path of one linking the protons, that of the second linking the oxygen
nucleus not to either proton but to the surface manifold of the first bond
critical point. Unlike a nuclear position where p is a local maximum, which
thereforeacts as an attractor of gradient paths in three-dimensionalspace,a
bond critical point is a maximum only in the interatomic surfaceand can act
as an attractor of only those gradient paths which lie in this surface.In this
particular system,the surfaceincludesone of the two gradient paths origin-
62 M O L E C U L A R S T R U C T U R EA N D I T S C H A N G E 3.2

ating at the upper (3, - 1) critical point. Any motion of the nuclei which
causesthis trajectory to leave the surfacemanifold, that is, to move in three
dimensions rather than in just the two-dimensional space of the surface,
causesit to abruptly switch to one or the other of the three-dimensional
attractors whose basinsshare the interatomic surface,the protons.
The ability of a (3, - 1) critical point to act as an attractor in the two-
dimensional manifold of its associatedinteratomic surface can be seen in
Fig. 3.3 which illustrates the gradient vector field for the unstable structure
shown in Fig. 3.1(e)fora plane perpendicularto the plane of the nuclei,along

Frc. 3.3. Gradient vector field of the electronic charge density for the unstable conflict structure
shown in Fig. 3.1(e).The plane shown contains the C, symmetry axis and is perpendicular to the
plane of the nuclei. In the lower portion of the diagram, trajectories terminate at the (3, - l)
critical point between the hydrogen nuclei. In the upper portion, trajectories terminate at the
pseudo (3, - 3) critical point at the oxygen nucleus. These two critical points are linked by the
pair of trajectories which originate at the central (3, - l) critical point indicated by the dot. This
is an unstable intersection of the one-dimensional manifold of this (3, - 1) critical point with the
two-dimensional manifold of the (3, - 1) critical point between the protons.
3 .2 THE DEFINITION OF MOLECULAR STRUCTURE 63

the C, axis. It shows,in the lower portion, the trajectorieswhich terminate

at the H-H bond critical point and define the interatomic surfacebetweenthe
hydrogens.Both this critical point and the one at the oxygen nucleus appear
as (2, - 2) critical points, or two-dimensional attractors in this plane. The
second(3, - 1) critical point appearsas a (2,0) critical point and its bond
path links the oxygen nucleus and the critical point of the H-H bond. The
gradient vector field of this system for the one particular plane shown in
Fig. 3.3 is indistinguishablein form from that for a diatomic molecule, the
critical point of the bond between the hydrogens serving as an attractor.
The phaseportrait in the vicinity of the H-H bond critical point in this case,
however, is unstable and is obtained only when the bond path trajectory
lies within its surfacemanifold, which, in this particular case,is a symmetry
plane.
The type of unstable structure shown in Fig. 3.1(e)is called a conflict
structure-the protons are in conflict as they competefor the terminus of the
line of maximum negativechargedensity linked to the oxygen nucleus.In the
region of the energy surface where the conflict structure is found, which
correspondsto a relatively large separation betweenthe oxygen nucleus and
the midpoint of the hydrogen molecule,there is a small energy barrier along
the length of the exit channel for the separation of the oxygen along the C,
symmetry axis, it being energetically more favourable for the oxygen to
recedefrom the hydrogen moleculewhen it is on either side of this axis.Thus,
the conflict structure in this system is both energeticallyand topologically
unstable,and perpendicularmotions of the exiting oxygen nucleuswill cause
the structure to oscillate from one stable form to the other.
This analysis of the quantum mechanical calculations for various geo-
metries along a reaction pathway illustrates how structural information can
be obtained from the state function. Table 3.1 summarizesthe values of the
chargedensity at the bond critical points, valuesdenoted by the symbol p(r"),
and the total integrated electron populations of the atoms for the various
structuresalong the reaction pathway. (An atomic population, obtained by
integration of the charge density over the basin of the atom, is an obvious
aromic property which is used here in advanceof the formal definition of the
properties of atoms.) It is clear from Fig. 3.1(a) that the oxygen nucleus
dominates the equilibrium charge distribution of the water molecule. The
bond critical points and interatomic surfacesare close to the protons and
there is a substantial transfer of charge to the oxygen atom, each hydrogen
bearing a net chargeof + 0.58e. The value of p(r") for the two O-H bonds in
the equilibrium geometry is typical. This value undergoesa continuous
decreaseas the oxygen nucleus recedesfrom the hydrogens and, accom-
panying this, there is a transfer of charge from the oxygen to the hydrogens.
Correspondingly,the O-H bond critical points and their interatomicsurfaces
move towards the oxygen nucleusas the hydrogensgro\4'in size.Electronic
64 M O L E C U L A R S T R U C T U R EA N D I T S C H A N G E 3.2

Table3.1
at bondcriticalpointsandatomiccharges
Chargedensities for Cz, dissociatiue
pathwao y f HrO X'A,

Nuclear geometry* plr.) Net charge

Ro (au) R, (au) O-o-H H-o-H O-o-H, q(O) q(H)

( a) 1. 109 2 .8 6 8 0 .3 6 7 7 - 1.16 + 0.58

(b) 2.300 2 .1 6 0 0.r428 - 0.32 + 0.16
(c)r 2.390 2 .1 0 0 0 .13 0 8 0.1230 - 0.28 + 0.14
(d) 2.s00 2.000 0.1187 0.l3r2r - 0.24 + 0.12
(e) 2.750 1 .8 0 0 0.1623 0.0930 - 0.14 + 0.07

*These data, all in au, refer to the five configurations shown in Fig. 3.1 and are correspondingly
labelled.
'The value of Ro for the catastrophe point lies
between 2.40 and 2.39 att.
; T h e v a l u e o f p ( r " ) a t t h e ( 3 , + 1 ) o r r i n g c r i t i c a l p o i n t i n t h i s s t r u c t u r ei s 0 . l 1 7 6 a u .

charge transferredto the hydrogensis accumulatedbetweentheir nuclei and,

as indicated by Fig. 3.1(c),a bond is formed betweenthe hydrogensbefore the
O-H bonds are broken, resulting in the formation of a ring structure. The
value of p at the H-H bond critical point at the instant of its formation is only
slightly lessthan the value of p(r") found for the two O-H bonds. The value of
p(r")for the H-H bond undergoesa further continuous increaseas the oxygen
recedes,eventually attaining the value of 0.27 au found in an isolated H,
molecule. The formation of the conflict structure indicates the presenceof
only a weak interaction remaining between the oxygen and the forming
hydrogen molecule. This is reflected in the relatively low value of p at the
corresponding critical point, a value which decays to zero for the eventual
separation of the oxygen atom from the hydrogen molecule.

3.2.3 Structurediagrams

The discussion so far has demonstrated that the definition of structure is

inextricably bound up with the definition of structural stability. We now
show that the two definitions taken together lead to a partitioning of nuclear
configuration spaceinto a finite number of structural regions.
Associatedwith each point in nuclear configuration spaceis a real space
charge distribution and its associatedgradient vector field. By applying the
equivalencerelation to the gradient vector fields defined in real spacewe are
able to defineequivalent points in nuclear configuration space.The result is a
partitioning of nuclear configuration space Ro into a finite number of non-
overlapping regions, each of which is characterizedby a unique molecular
 3 .2 THE DEFINITION OF MOLECULARSTRUCTURE 65

structure.These structurally stable,open regions are separatedby bound-

aries, hypersurfacesin the space RA. A point on a boundary possesses a
structure which is different from that characteristicof either of the structural
regionsit separates. Sincea boundary is of dimensionlessthan Ro, arbitrary
motions of the nuclei will carry a point on the boundary into neighbouring
stable structural regions and its structure will undergo corresponding cha-
nges.The boundaries are the loci of the structurally unstable configurations
I of a system.In general, the trajectory representingthe motion of a system
point in Ro will carry it from one stablestructural region through a boundary
to a neighbouring stable structural region. The result is an abrupt and
discontinuouschangein structure as illustrated in Fig.3.1. In this sensea
change in structure is catastrophic (Thom 1975) and the set of unstable
structuresis called the catastropheset.We shall henceforthrefer to a point in
a structurally stable region of nuclear configuration spaceas a regular point,
and to a point on one of the structurally unstableboundariesas a catastrophe
point.
A knowledge of the stable structural regions and their boundaries as
definedby the catastropheset enablesone to construct a structure diagram, a
diagram which determines all possible structures and all mechanisms of
structural change for a given chemical system. Figure 3.4 is a two-dimen-
sional cross-sectionof the structure diagram for an ABC system.(The reason
for the particular form shown there is justified in Section 4.3.2.)The letters
may stand for atoms or functional groupings of atoms. The full lines,
denoting the catastrophe set, partition nuclear configuration space into
its structural regions. The hypocycloid portion denotes the loci of the
bifurcation catastrophes,of the type illustrated in Fig. 3.1(c),and the three
semiaxes,the loci of the conflict catastrophesas illustratedin Fig.3.1(e).
All of the essentialinformation concerning the possiblestructuresand the
mechanisms of their change for an ABC system are representedin this
structure diagram. The reaction of an oxygen atom with a hydrogen molecule
in the formation of a water molecule along the path representedby the
molecular graphs (e) to (a) in Fig. 3.1 and, with A - O, the oxygen atom, is
representedby motion beginning at the top of the vertical semiaxisin Fig. 3.4
with the conflict structure. The conflict structure representsthe average
structure obtained as the bond path from the oxygen switchesbetween the
hydrogens as depicted by the molecular graphs for the stable structural
regionsseparatedby this axis.At the point where the axis coincideswith the
bifurcation set, the systemundergoesa change in structure and enters the
region of the stable ring structure.The ring critical point w'henfirst formed
will lie immediately betweenthe critical points of the O H bonds. As the
system progressesdown through the ring region. the charge densitv must
changein sucha way that the ring critical point migratesto* ards the critical
point of the H-H bond so that, as previouslydescribed.thel coalescetrr fr)rm
66 M O L E C U L A R S T R U C T U R EA N D I T S C H A N G E 3.2

l
I

I
t
B-fc

A
- - - - - - f
I
t - - - AI

,L';- I B__AC

i \c
'v
s/ i\
Ftc. 3.4. A two-dimensional cross-sectionof the structure diagram for an ABC system.The full
lines, denoting the catastrophe set, partition nuclear configuration space into its structural
regions. The structure associated with each region is indicated by a representative molecular
graph. The two broken lines indicate two possible reaction mechanisms for changing the
structure A-B-C into the structure B-C-A.

a degenerate critical point and a secondbifurcationcatastrophepoint, a

point on the boundaryseparating the ring from the openregionof the water
system.
The structurediagrampredictsthe existence of two possiblemechanisms
for a reactionof the type
A-B-C --+B-C-A.
The conflictmechanism, indicatedby the upperpath in Fig. 3.4,corresponds
to the passage of the systemthrougha conflictcatastrophe point.In this case
the changein structureis obtainedby an abrupt switchingof attractors,the
bond path from A switchingfrom B to C. This changein structureis
illustratedby the sequence of moleculargraphsshownin Fig. 3.5(a)for the
caseof the migration of a methyl group in the carboniumion C.H|. A
structureof the conflicttypeis frequentlyfoundin systemswhereoneinvokes
an interactionbetweena sigmabond of themigratinggroupwith the'z' bond
of the ethylenicfragment.
3.2 THE DEFINITION OF MOLECULARSTRUCTURE 67
- e r l

g E"iF
d U L -

$ " €E :
Ei= iq
€fii;i
X : * c b
e d-o;i.cog
$ - : - . v
, . . r ' A? . 2
g
,t -g' =, H < . 8
p * f
:6: U3= E$ >:
F E ; R ;
# r H J E
E E E E g
.' (E e - i

H E ti is
t'^tr;-q
";;,=E
r - ^ - A
*
ilt X ar

E 3t!:5
i c ! { o
E !.,:I I
&12----
s i U o *
H O .r Y
a - E > L # H , ^

E ;X
q
E€3
U E 9 U :
:€ A€=l
; H€ . = ' FIo
; ; I e ' Fl
6€ & €O oE eF ; €
E _: XE c!

E T Esa :
::=; H5 ' - d
. + a i i O

FlF"s€
. Y v E c H . v
- - - ^

Sr . =. ^ Y^ .e Y : V

-: tr._ I (€
, F€ t e r
*
.
)
- ;
A
P
S
/
\ r ^ b ^ i
? F E ; €
9 uqi E=
Eb
. = 9 9 [ i Y
FP
E ' E = g ;
E S re rX
-g 6E-c
;E;:E
E 5 5 = ;
E d *v 6v ,€
bO6

- :
: 8! - * s E
< 9 r O v
i s d c ' -
C) Q.x.9
- 3 : - b O
> H € =i
tr p.T.E
;- iE-
' : v 3 -

. - + ) -
68 M O L E C U L A R S T R U C T U R EA N D I T S C H A N G E 3.2

The second of the two possible mechanisms,the bifurcation mechanism,

involves the formation of an intermediate ring structure. This mechanismis
again illustrated by the methyl group migration in CrHf, by the second
sequenceof moleculargraphs shown in Fig. 3.5(b).At the catastrophepoint
separatingthe open from the ring structure, a new, degeneratecritical point
appearsin the chargedistribution. Associatedwith this critical point is a pair
of trajectories which link the methyl group to the second methylene group
and a new C-C bond is formed. Further motion of the systempoint resultsin
the bifurcation of this unstablecritical point into two non-degeneratecritical
points, a bond critical point connectingthe methyl with the secondmethylene
group and a ring critical point, in the interior of the newly formed ring
structure. Continued motion of the methyl group across the ethylenic
fragment causesthe ring critical point to migrate towards the bond critical
point which formed the original bond path to the methyl group. When these
two critical points coalesce,the systemis at a secondbifurcation catastrophe
point. At this point, the original bond to the methyl group is broken and the
system enters the structural region of the product molecule.
What have just been describedare the topological changesin the charge
density that must occur in the formation or destruction of a bonded ring of
atoms. As the nuclei are displaced to increase the separation between a
bonded pair of atoms forming a perimeter of the ring, the chargedensity must
undergo a corresponding change so that the minimum in the ring surface
migrates towards the critical point of the extendedbond. The curvatures of
the charge density at both the ring and bond critical points are progressively
softenedin their direction of approach until, at the point of coalescence, their
common curvature vanishes and the bond is ruptured. These are new
windows on the essentialchangesin the charge density which accompany
changes in structure and the views so obtained lead to new questions
concerning the mechanics governing a change in structure. How can one
facilitate or hinder the formation of a degeneratecritical point in a charge
distribution? Critical points in p as well as the nuclei must move to achievea
changein structure.Could one developa liquid model of the chargedensity,a
principal parameter being the viscosity experiencedby a migrating critical
point? As demonstrated in Chapter 6, one can relate the curvatures of the
charge density at a point in space to the local potential and kinetic energy
densitiesof the electrons.While there is a mechanicsof the chargedensity,it is
not complete and the theory of atoms in moleculesoffers opportunities for
the formulation of new models in the searchfor answersto old problems.
One final exampleof a changein structure is illustrated by the successionof
gradient vector fields shown in Fig. 3.6 for the thermal isomerization of HCN
to CNH. The calculated minimum energy path for this reaction exhibits a
singleenergymaximum and, correspondingly,the changein structure occurs
via the conflict mechanism,the geometry of the unique topologically unstable
structure coinciding with that of the transition state. Motion along the
I
I
I
I 3.2 THE DEFINITION OF MOLECULAR STRUCTURE 69

o,o

721

12(),o lgo.o
Frc.3.6. Displaysof the gradientvectorfieldsof the electronicchargedensityfor.configura-
rionsalongtheisomerization reactioncoordinate for HCN. Thelabelrefersto theangle0 iormed
between the C-N axisand the vectorfrom the protonto the CN centreof mass.The bondpath
from the proton switchesattractors,lrom the carbonnucleusto the nitrogennucleusfor some
configuration lying between0 : 72.1"and 72,4"'
70 MOLECULAR STRUCTUREAND ITS CHANGE 3.2

reaction pathway is monitored by the value of the angle 0 formed betweenthe

C-N axis and the vector to the proton with origin at the CN centre of
mass.The molecular geometries,lying on the reaction coordinate for values
of 0 ( 0 < 72.1', all possessequivalent gradient vector fields,hence,they are
all representedby an equivalent molecular graph, one which defines the
stable structure H-C-N. Similarly, all geometries occurring in the range
72.4'< 0 < 180" exhibit a different but again equivalent set of molecular
graphs which define the stable structure C-N-H. Thus, in the range
1Z.t' < 0 <72.4", a switchingof attractors occurs,as is characteristicof the
conflict mechanism.
It would be extremely difficult to determine the exact geometry of the
unstable conflict structure wherein the proton is linked to the surface
manifold of the C-N bond critical point. Unlike the previous examples,the
conflict point is not fixed by symmetry. The conflict structure is very nearly
realized in the molecular graphs for 0 :72.1' and 72.4' and is clearly
bracketed by them. The exact geometry of the transition state is equally
difficult to determine and, within the accuracy of the SCF calculations, its
geometry is bracketed by the same values of the parameter 0. The conflict
mechanism is also found to occur in the corresponding isomerization of
methyl cyanide to methyl isocyanide and again, to within computational
accuracy,the transition state and topologically unstablestructure possessthe
same geometry.Theseare exampleswherein one finds a coincidencebetween
an energetically unstable geometry and a topologically unstable structure.
The general discussion of the extent to which these two definitions of
instability coincideis given in Section3.4.1.
The set of molecular graphs in Fig. 3.6 exemplify the principal ideas of
structure and structural stability. The initial and final structures are each
representedby a unique molecular graph, one which persistsover a range of
nuclear motions. These structures are stable. They are separated by a
structure whose molecular graph is intermediate in form and, since it exists
for only a single nuclear geometry along the reaction path, is unstable.

3.2.4 Bonds and structure

Hydrocarbon moleculesare used to illustrate the definitions of structure and

structural stability introduced above and also the manner in which the
propertiesof the electronicchargedensity at a bond critical point can be used
to summarize the important physical characteristicsof a bonded interaction
betweenatoms (Bader et aI.1983).Figure 3.7 givesthe molecular graphsfor
saturatedhydrocarbon moleculeswith widely varying equilibrium structures,
beginning with acyclic molecules both branched and unbranched, followed
by cyclic and bicyclic molecules, and ending with a number of so-called
).L THE DEFINITION OF MOLECULARSTRUCTURE 7l

propellane molecules and other highly strained systems, tetrahedrane,

cubane,and spiropentane(Wiberg et al. 1987).These structuresare deter-
mined in their entirety by the information contained in the corresponding
statefunctionsas employedin the definition of the electronicchargedensity
( e q n( 1 . 3 ) ) .
With the exception of the propellanes, for which existing models are
inadequate, the structures in this figure are in complete accord with those
that have been assignedto thesemoleculeson the basisof chemical informa-
tion and models of directed valency, augmented with the concept of bent
bonds. The agreement of structures,predicted in this manner by quantum
mechanicsfor a wide variety of systemswith chemically acceptedstructures,
shows that the many models used to rationalize the network of bonds in a
molecule may be replaced with a single theory of molecular structure. The
small ring propellanes,such as [1.1.1]propellane20, possessunusual struc-
tureswith'inverted'geometriesat the bridgeheadcarbon atoms.A bond path
links the bridgehead nuclei in each of the propellane molecules20-23, even
though this resultsin structuresthat have four bond paths, all to one side of a
plane,terminating at each of the bridgeheadnuclei in 20, 2l and 22. While the
model of hybridized orbitals cannot describesuch a situation, even assuming
bent bonds, the molecular graphs for the propellanes demonstrate that the
charge distribution of a carbon atom can be so arranged as to yield bond
paths-lines of maximum charge density-which correspond to an inverted
structure.
The structures of the propellanes consist of three rings, either three- or
four-membered,sharing a common bond, the bridgeheadbond. Addition of a
molecule of hydrogen to a propellane breaks the bridgeheadbond and yields
the correspondingbicyclicmolecule,l6-19. Molecules16 and 19 possess cage
structures,the interior of each molecule being bounded by three curved ring
surfaces and containing a cage critical point. The propellanes and their
bicyclic congeners exemplify the contrasting behaviour exhibited by the
chargedensity betweena pair of nuclei in situations where the atoms are and
are not bonded to one another. Figure 3.8 gives relief maps of the charge
density in the plane which bisects and is perpendicular to the bridgehead
internuclear axis for each propellane and its bicyclic analogue.Such a plane
contains the critical point betweenthe bridgeheadatoms. It also contains the
critical point in the interatomic surfacebetweenthe methyleniccarbon atoms
linked by a peripheralbond and/or the C and H nuclei of a peripheralCHt
group in the caseof a three-membered ring in a propellaneor the correspond-
ing group in the bicyclic molecule.The charge density'is a maximum at a
bond critical point in the interatomic surface.Thus, the diagrams for the
12.2.21systems shown in Fig. 3.8(a)both exhibit three maxima in the charge
density correspondingto the three peripheralbonds betu'eenthe methl'lene
groups. These maxima in bonded density are successivellreplacedbr the
72 M O L E C U L A R S T R U C T U R EA N D I T S C H A N G E 3.2

5 6

', 7
1
V /'
\/
,,\.t \ / y
,/'"..
' \
\ /
\ ,,/\ \\/'7

,{,
--->,(
X A ii /
t\ r\ \--- \
I
l e
. \ l
\
\
\ \
\-)--
\
a

,A
I
a
/
t i r
,44\ *
\ r l \
10
12
\
t

/
A\ \ \ :

.---rq-
/ t \
i-)<.}
\ i \
24
I
II 3 .2 T H E D E F I N I T I O N O F M O L E C U L A RS T R U C T U R E 1 1
IJ

a
I

i r-' t

I I

. \ , ,
\\ \ \ , z' -l-

\2f;-=r1
/ t ' . .\ ,/
/ f c
/ l

14 15

Y
i\
a t
i i
. r)<l ) ,
\ ,"
t-\
-\. -''
Y

16 18
17
\. ,r''
----_.;'-_-- \ \ t '
\ -- ,/ \, -r''
\r- l--l
.t.
.'{
t'..
\t-
Ir
l).,
4i':
i i
t
\ 4 r ,-/ f
."'
r^--'-
\
'\--
------:---
..t'' \
\iX\ /
)-- I Y )
: r f
t l
t 1
19 21
'-
-:-

l .

..'-: .1.-
'.../
\. .. ( .../'
.
'-a--a/'
t.;-a-:-.
..r

r \ l
r i l r
i i

22

Frc. 3.7. Planar projections of molecular graphs of hydrocarbon molecules generated from
:heoreticalchargedistributions. Bond critical points are denoted by black dots. StructuresI to 4
are normal hydrocarbons from methane to butane, 5 is isobutane,6 is pentane,7 is neopentane,
and 8 is hexane.The remaining structuresare identified in Table 3.2.The structuresdepicted in
:nesediagrams are determinedentirely by information contained in the electronicchargedensity.

nuclearmaxima of a CH, group through the seriesto the I L 1.1] systems.The

purposeof the diagram is to contrast the behaviour of the chargedensity at
:he bridgehead critical point for the propellaneswith that found for the
ricyclic molecules.In the propellanes,which possessa bridgeheadbond. the
74 M O L E C U L A R S T R U C T U R EA N D I T S C H A N G E J.Z

Ftc. 3.8. Relief maps of the electronic charge density in the symmetry plane bisecting the
bridgehead bond in the propellanes,shown on the right, and in the corresponding symmetry
plane in each of the related bicyclic structures, shown on the left. (a) 12.2.2]propellaneand
bicyclooctane; (b) [2.2.1]propellane and norborane; (c) [2.1.1]propellane and bicyclohexane,
( d ) [ 1 . 1 . 1 ] p r o p e l l a n ea n d b i c y c l o p e n t a n eC
. o n t r a s t t h e p r e s e n c eo f a c e n t r a l m a x i m u m i n p i n
f2.2.2fpropellanewith its absencein the bicyclic compound. There is a bridgehead bond in the
former. but not in the latter.
)../. THE DEFINITION OF MOLECULARSTRUCTURE 75

ch a rgedens it yis a m a x i mu ma t th i sp o i n t,v ' h i l ei n the bi cl ' t' l i c' mol ecul es,w hi ch
do not possess a bridgeheadbond,the charpletlensitt'is a ninimum at this same
point. In the former moleculesthere is a line of maximum charge density
linking the nuclei, while in the latter such a line is absent and the charge
densityis insteada local minimum at the centralcritical point in the first and
final of the bicyclic molecules.There is, therefore, an essential,qualitative
differencein the manner in which electronicchargeis distributed along a line
linking a pair of bonded nuclei (as between the bridgeheadnuclei in the
propellanes)and along a line linking two nuclei that are not bonded (as
betweenthe nuclei in the correspondingbicyclic structures).
In addition to the fundamental differencein the form of the bridgehead
chargedensity betweenthe two setsof molecules,there are also substantial
quantitativedifferencesin the valuesof p. Thus, its value at the bond critical
point in [.1.1]propellane is 0.203au, four-fifths of the value for a normal
C-C bond, while the value of p at the correspondingcagecritical point in the
bicyclicmoleculeis 0.098au. It is also clearfrom Fig. 3.8(d)that the valuesof
p at the bond critical point and the three adjacent ring critical points are
almost equal, giving rise to a very broad bonded maximum in p in the
bridgeheadinteratomic surfacein [1.1.1]propellane.
The value of the charge density at a bond critical point can be used to
define a bond order (Bader et al. 1983; Cremer and Kraka 1984).The
moleculargraphsfor ethane,ethylene,and acetyleneare shown in Fig. 2.8.In
each case the unique pair of trajectoriesassociatedwith a single (3, - 1)
critical point is found to link the carbon nuclei to one another. Multiple
bonds do not appear as such in the topology of the chargedensity.Instead,
one finds that the extent of chargeaccumulationbetweenthe nuclei increases
u'ith the assumednumber of electronpair bonds and this increaseis faithfully
monitored by the value of p at the bond critical point, a valuelabelledpo.For
carbon-carbonbonds.one can definea bond order n in terms of the valuesof
t)b using a relationshipof the form
n: Q X PIA (p t- B)l

ri'hereA and B areconstantswhich vary slightly with basisset.B is setequal

rcrthe value of pu for ethane and a value for A can be found such that the
relationshipyields bond orders of 1.0, 1.6,2.0, and 3.0 for ethane,benzene,
ethylene,and acetylene,respectively.The bond order po relationship is
distinctfor each pair of atoms A B, sincethe valuesof po for bonds of given
itrrmal order can vary markedly betweendifferentpairs of atoms.Thus, one
cannot estimaterelativebond strengthsfor bonds betu,eendissimilaratoms
h1' comparing their respectivepo values.The orders of nitro-sennitrogen
Lrondsare well describedby a linear relationshipwith po. The original bond
trrd e rs c hem epr opo s e db y C o u l s o n(1 9 5 1 )w a sb a sedon a bond rrrder bond
length relationship.The value of pr is found to increaseas a btrnd lc'nsth
76 M O L E C U L A R S T R U C T U R EA N D I T S C H A N G E 3.2

decreasesand it is possibleto obtain useful bond length-po value relation-

ships(Boyd et al. 1988)for a wide range of A-B systems.While theseschemes
are empirical, they serveto relate familiar models and conceptsto a property
of the chargedensity.However, they also underscorean important point-
that the strength of the bonding between a given pair of atoms increasesas
the amount of charge in the interatomic surface for the equilibrium bond
length,as monitored by its maximum value,pu, increases. This point is well
illustratedby showing that the behaviourof the value of pr on ionization can
be predicted on the basisof the bonding or antibonding nature of the orbital
from which the electron is removed.
Thereis a notion of simplemolecularorbital theory that may be a causefor
concern in employing po as a measureof bond order. One may be led to
conclude that, since a bond critical point can lie in the nodal surfaceof a n
orbital, the value of pv may not properly reflect the zrorbital occupancy.As
stressedby Mulliken, essentialto the orbital theory of electronic structure is
the property of self-consistency-that each orbital be determined by its
interaction with the averageCoulomb and exchangepotential generatedby
electrons in the other occupied orbitals. Thus, the density distributions
derived from o and z orbitals are not independent of one another. This
coupling is very pronouncedand is bestillustratedby consideringthe change
rn pavaluesfor a given bond in vertical (no changein internuclearseparation)
ionization and electronattachment.Table 3.2 lists such valuesfor a number
of diatomic moleculeswhose ground electronicconfigurationinvolvesoccu-
pation of the 2n antrbondingorbital (lz* of homonucleardiatomics).It is a
general observation that the vertical tontzation of a z electron causesan
increasein the o density distribution along the internuclearaxis, while the
addition of a n electron has the reverseeffect(Cadeet al. 1971).As a result,the
value of po increaseswith the removal of, and decreaseswith the addition of, a
2n electron.Just as the progressivefilling of a n bonding orbital causespoand
hencen to increasein the hydrocarbons,the occupation of an antibonding z
orbital leads to a decreasein po. These observationsare in accord with the
predictions of molecular orbital theory regarding the opposite changes in
bond length by a change in the occupation of a bonding or antibonding
orbital (Mulliken 1932,1939).Indeed,the effectsof changingthe occupation
of the 2n orbital are accentuatedwhen the accompanyingchange in equilib-
rium bond length is allowed for. For example, in CF*, po increasesto
0.3636au at R" : 1.229A and in NO * , Pb increasesto 0.7608 au at its
R" : 1.062A.
The bond paths are noticeablycurved in thosestructuresin Fig. 3.7 where
the geometry of the molecule precludes the possibility of tetrahedrally
directed bonds betweenthe carbon nuclei. The differencesbetweenthe bond
path angle ao, the limiting value of the angle subtendedat a nucleusby two
bond paths, and the geometricalor so-calledbond angle a" is important in
1
J.Z
1
THE DEFINITION OF MOLECULARSTRUCTURE 71

v)?\
It- \O ea
.1
- i A

.)T
(,
- d N

c-

\S
t\ - - 6

O\ O\ t--'
\ oo
C-,
\O
Ca
.a
aa
6l
a?
:1.

,iv v .-i
v
-i

A 0)
+
g H *

N ^\./ A\J n U
\./
I

a E
(r)oo\
\o ..l c..l r- -54
- aa ai aa .1
L
A A A F.
i.) I
q)
L
+
Et&t!
( \ 'iP Z Z Z
'1.

c'i (.)
L s
q ) A J
O ca tr)
aA ca ta)
eq L. * O\ t-- tr)
\olr)r) a

- i A - i
-
A + l

q.)
) zzz
\.
U
ci
.o
sf,
c.l F
a
ca O. t--
.a (..l cl bo
() A / i
^
()
C)
+
a ; \

R o t--
U U

;>
} E
R

U a.) €
t-
'f
;/

P V ) + ^ ^ + ()
a.y :
NtrNtrH N O N
-:: v

v v v v v
6 c 0 O * N -
p
\ t
p p
- -
\-/ c..l
H
(.'.1
h
a-l
ts
a'..1 6l ct.l *--
78 M O L E C I J L A R S T R U C T U R EA N D I T S C H A N G E 3.2

quantifyingthe conceptof strain in thesemolecules.Another relatedquantity

is the bond path length, R6, which for a curved bond path exceedsthe
correspondinginternuclear separation or so-calledbond length, R". The
-
values of the bond path angle and the differenceAa d.b- de are listed in
Table 3.3 for the cyclic, bicyclic,and propellane molecules together with their
strain energiesas calculatedusing Franklin's group equivalents. For most of
thesemoleculesAa > 0 and, in thesecases,the bonds are lessstrained than
the geometricalanglesa" would suggest.In general,Aa provides a measure of
the degree of relaxation of the charge density away from the geometrical
constraints imposed by the nuclear framework. The reader is referred to the
original publication of these results for their full discussion.Here we note a
few interesting results for the propellane systems.Becauseof the additional
constraints resulting from the fusion of three small rings to a single bridge-
head bond, the relaxation of the apex bond paths in theselatter moleculesis
inhibited relative to that found for cyclopropane.Thus Aa for the apex angle
in [1.1.l]propellane is actually slightly negative,while the relatively large
values found for the three angles formed by the bond paths of the three-
memberedrings that terminate at a bridgeheadcarbon nucleusresult in bond
path angles that differ by only - 1o from the tetrahedral angle'
The charge density along a bond path attains its minimum value at the
bond critical point and the associatedcurvature or eigenvalueof the Hessian
of p atr", A3,isthus positive.The chargedensity in an interatomic surface,on
the other hand, attains its maximum value at the bond critical point and the
two associatedcurvatures of p at r", i, and ir, those directed along axes
perpendicularto the bond path, are thus negative.In a bond with cylindrical
symmetry,thesetwo negativecurvaturesof p at the bond critical point are of
equal magnitude. However, if electroniccharge is preferentiallyaccumulated
in a given plane along the bond path (as it is for a bond with z-character,for
example),then the rate of fall-off in p is lessalong the axis lying in this plane
than along the one perpendicularto it, and the magnitude of the correspond-
ing curvature of p is smaller. If ,t, is the value of the curvature of smallest
magnitude, then the quantity e :lirlAz - 11, the ellipticity of the bond,
provides a measure of the extent to which charge is preferentially accumu-
lated in a given plane (Bader et al. 1983).The axis of the curvature ir, the
major axis, determines the relative orientation of this plane within the
molecule. The elliptical nature of the charge distribution in the C-C inter-
atomic surfacein ethyleneis illustrated in Fig. 3.9.The ellipticitiesof the C-C
bonds in ethane,benzene,and ethyleneare 0.0,0.23,and 0.45,respectively,for
densities calculated from basis sets containing proper polarizing functions
and the major axis of the ellipticity in each of the latter two molecules is
perpendicular to the plane of the nuclei. The bond ellipticities faithfully
recover the anticipated consequencesof the conjugation and hyperconjug-
ation models of electron delocalization.A classicexample of delocalization
3.2 T H E D E F I N I T I O NO F M O L E C U L A RS T R U C T U R E 79

throughthe conjugative effectis providedby trans-1,3-butadiene. Thecentral

bondin this moleculeexhibitspartialdoublebond characterasevidenced by
n > 1and e > 0. Correspondingly, as a resultof the delocalization of charge,
the n and e valuesof the terminalC-C bondsare decreased from the values
found for ethylene.Of equal importanceis the observationillustratedin
Fig. 3.9,that the major axesof the ellipticityinducedin the centralbond
and of the terminalbondsare parallelto one another.
The increasein ellipticityof the C-C bond from benzeneto ethyleneis
primarilya resultof an increased contractionof the densitytowardsthe bond
path,asseenfrom the magnitudeof 1.,increasing from 0.70to 0.81au,rather
than of a decrease in the magnitudeof ir, the curvatureof p along the
principalaxis,which changesonly by 0.01au from 0.57 to 0.56au. The
magnitudes of both curvaturesareincreased overtheir valuein ethanewhere
theydegenerate to thevalue - 0.48au.As discussed morefully below,it is by
this mechanismthat more chargeis concentratedalong the bond path,
reducingthe associated positivecurvature2, from 0.27to 0.20au,decreasing
Y2p andincreasing the bond orderfrom 1.6to 2.0.Thus,whilethe ellipticity
doesprovidea measureof the n character of a bond,it doesnot followthat an
increase in its valueresultsfrom a simpledecrease in the magnitudeof ,tr, as
one might anticipatefor an increasein a r population.Sucha pedestrian
changein the chargedensity,cannotaccountfor the associated increases in
bond order and in the magnitude of Y2p. Rather, it is the relativepro-
portioningof thechargedensitybetweenthe two planescontainingtheminor
and major axesas measuredby the ratio Atli, that determinese and
correctlysummarizes the topologicalchanges in p whichaccompany what,in
the orbital model,is describedas an increase z in population.
The chemistryof a three-membered ring is verymucha consequence of the
high concentrationof chargein the interiorof the ring relativeto that along
its bond paths, a fact which is reflectedin substantialbond ellipticities
(Cremeret al.1983). Thevaluesof p,, thevalueof thechargedensityat a ring
criticalpoint, is generallyonly slightlylessthan, and in somecasesalmost
equal to, the valuesof h for the peripheralbonds in the caseof a three-
memberedring of carbon atoms.In four-membered and larger rings the
valuesof.p, areconsiderably smaller,asthe geometrical distancebetweenthe
bond and ring critical points is greaterthan in a three-membered ring.
Because electronicchargeis concentrated to an appreciable extentover the
entire surfaceof a three-membered ring, the rate of fall-off in the charge
densityfrom its maximumvaluealongthe bond path toward the interior of
the ring is muchlessthan its rateof declinein directionsperpendicular to the
ring surface.Thus the C-C bonds have substantialellipticities,and their
major axeslie in the plane of the ring. In hydrocarbons,this property is
uniqueto a three-membered ring.The ellipticityof a C-C bond in cyclopro-
paneis actuallyslightlygreaterthan that for the'doublebond'in ethylene,
80 M O L E C U L A R S T R U C T U R EA N D I T S C H A N G E 3.2

t r ) - N

* - \ O Y \ O ( \
+ C \ O ..i Ct.l
A c )

o
$ o.l .a.at $ o\c{\o ol ct'lf 1t-
Ob C- tn3\ra) cO F-* <.tl--O\\Qe.)
IJ
; 6 J3>; - c . i\ o r - ooi,r; \oo
I

.J
l l l l - - - - l

r - '

T+ 'J oo oo.'l> oQQlo!-o'l

$ .. . . 1. . . . c{ .a\r:r-
q € 0.) d r-- cwo o tr'-\oo
: b o ; ; + + 6 - 6 i' o n g . e o p l q
F o. ,-,-= r-'F-o\o\o
. f \ = i== =
R

o i
Fl o. ( ) u * v - ) \ O O O T O - \ O I \ $ \ O "
t a i E q ) o rr,cn6 <t o,trro. ootnoo-O
o
6-r O
=*a 8E -6 ;x Jc c a; s r)\oo' rc)'.)600-
\ J d

' ( s

Eo
- C\ t) * c . . l c . ) - C . l c i S t r )
-
c\

l-r

() v ((^ v
(lo
Lr
a ))- \"P

(.)
()
- d
tr
::
d

= c )
A l -

9
: 9
o
c- r
J o
F
, 5 5
x n - -
u r $ P
i :.. f'''
a)
9 E E ; o o
-J

()
(.) A a
A
, ^
A

o E E
7i >, >'
E E E 5 ; .'9 .9
:' v s , m
O Q '
J.Z T H E D E F I N I T I O N O F M O L E C L ' L . { RS T R U C T U R E 81

c ) ( r )
aa
. a . f , . < r

ra) (r- o\ I--. tr) oo o\ o r/^) r..) o\ \o $ cl oo $ - O \ O \ O O O T n O O :

€ \O t-- c...l \O :f, rO - c.r C ca (^l $ =qqq \qoeqc.] 11\
r o l o * ' O i J - . i i i c d * ' O O o . t c . l O \ t \ O $ O t O
t
I
- r *
I

- Or $ C! t/-) O\ \O * tr) o.l O O\ f-- co \O F- or\ O\ V) - \O r/-' F-

O '.r) \O t'-. CO \O $ c\.l O\ rf, O\ \O \n \O r) . 1 q q 9 q . 1 q n n
F-\o6i +,r; i"a+i r-c.joo.. i i O\ f-- o\\
t / ^ ) O \ O
.a t-- \O f- - tr)
C O O \ O ' O O O O \ O \ - r O - r O
O \ O \ -
- - O O O , O .
-

t - - * . 4 $ O O f - - O l - \ * a n c . r - CO Or l-- c-r O (^..l t-- tt-

- Q n n q c l \ : 9 n v ? n - o q q q q q n \ . ' l v ?
O O ' \ t $ C - - O t - r O c . l $ - o O . a - r a O \ C O - O { F - f - - S
l - - o O O \ O o O c O O \ O O O \O t/') CO O\ - tl^r O\ \O
O r o O - O\

- c . . l c a - C ! - c . l c A $ - C . l c v ) $ - C j - . . . l c O * . . . l . a S t r ) \ O

Q=
c i J = 0 )

x C x
o o a
0 )

' l - - c.l
c..l C.n (r.l
t-l t-J t-J
^ - a
' ) o ( . )
, J O O
a|^ ^/\ Ff\
82 M O L E C U L A R S T R U C T U R EA N D I T S C H A N G E 3.2

)(

N \ O l - - S
aa r - - $ \ c
A o ) $ aa v-, \o 6l

c
o l o o \ e - O i (A $ F S T gnR
TJ
q q q - n q
\ o . . r $ c - r r - - ; : S J r - o , d $
I *J 6l
TJ I I

E; oc.l r.ar-.aF- raF-

q"roq -
\o e! s'f,qj
qoqq
E ( ) < 1 n o ) n\ oh\ o n o\$c.
x o o ..; d\o6i Si\.o J t
" - n = - d I - = d o o o \ - - S
; F
\ <.r i
s
(.) tJ
U = s n s : s :J o8. i 8 8 N 6e
(')
E O
o \ - 6 i o o . i O
E
a 6 t F
6roJ6-.-t o.lo5 6 d\ \n\oc.
?.J v d
q)

6t 0.)
F b0
- C . l c a $ ( n \ O * N c . ) - - a.l ca

c)
-4(- N N€xRn (7
\/- (o
o xN

()
= o c )F
a <
A L
c) t-

O
-a ; U c g c o - )
0) (-.t :d d O
N N g e . =
N
^ i O - = o .
I 3.2 THE DEFINITION OF MOLECLLAR STRUCTURE 83
I

C.Hr*

b d
Frc. 3.9. (a) Contour map of the electroniccharge density for ethylenein the plane perpendicu-
lar to the plane of the nucGi and containing the C-C bond critical point. The charge density is a
maximumat this point in this plane.The magnitude of the curvature of p at the critical point is a
minimum in the direction of the major axis, the axis of the curvature ).r. This axis is
perpendicular to the plane of the nuclei. The diagram illustrates the elliptical nature of the
d i s t r i b u t i o n o fc h a r g e a s s o c i a t e d w i t h t h e p r e s e noc fe a n b o n d . ( b ) T h e m a j o r a x e s o f t h e C - C
bond ellipticities in lrans-butadiene are parallel to one another and perpendicular to the plane
containing the nuclei.This observationis consistentwith the orbital model of a delocalizationof
; chargedensity through conjugation. (c) The molecular graph of the CoH I molecular ion. The
bond orders are given in this and in the following structure. Note the proxrmitl oi the critical
points of the two long bonds to the ring critical point. (d) The resulting open slructure oi C*H --
84 M O L E C U L A R S T R I J C T U R EA N D I T S C H A N G E 3.2

indicatingthat the extentto which chargeis preferentiallyaccumulatedin the

plane of the ring is greaterthan that accumulatedin the z-plane of ethylene.
This property accountsfor the well-documentedability of three-membered
rings to act as an unsaturatedsystemwith the chargedistribution in the plane
of the ring exhibiting properties characteristicof a z-like system,one that is
able to conjugatewith a neighbouringunsaturatedsystem.Suchconjugation
is illustrated by the interaction of the cyclopropyl group with the (formally)
vacant 2p orbital in CHj as depicted in Fig. 3.9(c).The major axis of the
ellipticity induced in the C-CH2 bond has an overlap of 0.97 with the
correspondingaxesof the neighbouring C-C bonds of the cyclopropyl group.
(The overlap is determined simply by taking the scalar product of the
eigenvectorsdefining the major axes of the two bond critical points.) Such
conjugating ability of the cyclopropyl group is rationalized using molecular
orbital models through the choice of a particular set of orbitals, the so-called
'z-like' nature
Walsh orbitals. Theory shows that the of a three-membered
ring is a property of its total charge distribution, one that results from the
proximity of its ring and bond critical points. Understanding the physical
basis of this effectenablesone to predict its appearanceand consequencesin
other systems.
In Section3.2.2it was shown that the opening of a ring structureresulted
from the coalescenceof the ring and a bond critical point, the positive
curvature of the ring point annihilating the in-plane negativecurvature A, of
the bond point to yield a zero curvature characteristic of an unstable or
degeneratecritical point (Fig. 3.2).The decreasein the magnitude of 2, and its
eventual disappearancemeans that the ellipticity of the bond which is to be
broken increasesdramatically and becomesinfinite at the geometry of the
bifurcation point. Thus, a structure possessinga bond with an unusually large
ellipticity is potentially unstable. The two equivalent ring bonds of the
cyclopropylcarbinyl cation, CoHi (Fig. 3.9), provide an example of this
behaviour. The two long bonds of the structure in Fig. 3.9(c)are of order 0.6
and exhibit ellipticities equal to 6.7. Their corresponding paths are very
inwardly curved and their bond path length exceedsthe internuclear separ-
ation by 0.20 A. The structure vergeson instability since either of the bond
critical points of the long bonds can be annihilated by a coalescencewith the
ring critical point. The curvature of p atthe ring critical point, whose position
as indicated in Fig. 3.9 lies almost on the line joining the two bond critical
points, is close to zero and, correspondingly,the associatednegative curva-
ture, ir, of each of the neighbouring bond critical points is equally small in
magnitude. As anticipated for long bonds, the positive curvature of p along
their bond paths is relatively large, as is the second and parallel positive
curvature of p at the ring critical point. The valuesof p at the bond and ring
critical points differ by only 0.001au. Thus, there is a nearly flat-bottomed
trough in the distribution of charge linking these three critical points and
3..!. T H E D E F I N I T I O N O F M O L E C U L A RS T R U C T U R E 85

little energy is required to cause a migratron of the ring point along the
trough to coalescewith a bond point and l ield the open structureshown in
Fig. 3.9(d).It is a general observationthat little energy is required for the
nuclearmotions which result in a migration of a critical point along an axis
associatedwith vanishinglysmall curvature of the chargedensity.Thus, the
energysurfacein the neighbourhoodof the structurein Fig. 3.9(c)is very flat
for such a motion of the nuclei and the structure in Fig. 3.9(d)differsfrom it in
energyby lessthan 1 kcal/mol (Cremer et al' 1983).
Further examplesof potentiallyunstablestructuresbeing revealedthrough
exceptionally high bond ellipticities are provided by the propellanes,parti-
cularly l2.I.llpropellane,2l in Fig. 3.7.The bridgeheadbond critical point
and each of the ring critical points of the two three-memberedrings in this
moleculeare separatedby only 0.07 A and the value of pu : 0'197au exceeds
that of p,by only 0.001au. Consequently,as is evidentfrom the reliefmap of
the chargedensity shown in Fig. 3.8(c),the closeproximity and nearly equal
r aluesfor the bond and ring critical points result in a near zero value for the
curvatureof the densityat the bond critical point in the direction of the three-
memberedring critical points. (Seealso Fig. 3.8(b)for the l2.2.ll casewhich
shows the interaction with a single three-memberedring critical point in
profile more clearly.)The result is a very large ellipticity, equal to 7.21,for the
bridgeheadbond in this molecule.The bridgeheadbonds in both l2.2.lf and
[2.1.1]propellaneare predictedto be the most susceptibleto rupture by the
bifurcation mechanism and both moleculesreadily undergo polymerization
at 50 K. The bifurcation catastrophe undergone by the [1.1'1]propellane
molecule is used to illustrate the mathematical modelling of a structural
instability in Chapter 4.
These properties of a charge distribution have been applied to a study of
the position of the equilibrium dinorcaradiene= [10]-annulene as a function
of the substituentsR. While X-ray diffraction studies yield the geometriesof
the relevant species,and in particular the C1-C6 internuclear separation,
they do not enable one to determine whether or not carbons C1 and C6 are
bonded to one another and hence to determine which of the above two
structuresis the correct one for a given set of substituentsR and R'. Gatti er
,r/. (1985) determined the topological properties of the theoretically deter-
mined charge distributions at the experimentally measured geometriesfor
combinationsof substituentsR and R' : CN, CH., H, and F yieldingC1-C6
86 A \ D I T SC H A N G E
M O L E C U L A RS T R U C T U R E 3.2

separations rangingfrom 1.543Afor R: R': CN to 2.269A for R: R'

: F. The study demonstrated that the dicyanoderivativeand the methyl-
cyano derivative (in the B crystallographic phase)possessa C1-C6 bond,
whereas, of the two crystallographically distinct molecules in a unit cellof the
dimethyl derivative, one possesses a C1-C6 bond, the other, with the
annulenicstructure,does not. The remaining systems also exhibit the annul-
enic structure.The dimethyl derivativewith the shortest C1-C6 separation, ]
equalto 1.770A, is the lastnumberof the seriesto possess a dinorcaradiene I
structure.The bond orderof the C1*C6 bond and the separation betweenits I
critical point and that of the three-membered ring undergo a continuous I
decrease and,correspondingly, the C1-C6 bond ellipticityexhibitsa contin- |
'closed'molecules.The displayof the gradient
uous increasefor the three I
vectorfield for the next memberin the plane of the three-membered ring, I
which possesses a C1-C6 separation only slightly greater, at 1.783A, indi- |
catesthat this separation is pastthe geometryof the bifurcationpoint created I
by the coalescence of the ring and bond criticalpoint, as both criticalpoints I
are absentfrom the disPlaY. I
The sum of the threecurvaturesof p at a bond criticalpoint,the quantity I
Y, po,provides a usefulcharacterizationof the mannerin whichtheelectronic I
chargedensityis distributedin the internuclearregion.As pointed out in I
Section2.2.2, the Laplacian of the chargedensity,the quantity V2p(r). I
determines whereelectronicchargeis locallyconcentrated, whereY'p <0, I
and whereit is locallydepleted, whereY'p > 0. A chemicalbond requiresthe I
-
creationof a (3, 1) or bond criticalpoint in the chargedensitywhich has I
two negativeand one positivecurvature.Thus,mechanisms involvingboth I
chargeconcentrations and chargedepletionsare bought into play in the I
formationof a chemicalbond.The chargedensityis a maximumat the bond I
criticalpoint r. in the interatomicsurface, its two associated curvatures, those I
perpendicularto the bond path, are negative,and chargeis locallyconcen- I
tratedat r" in the interatomicsurface. The chargedensityis a minimumat the I
bond critical point along the bond path; the associated curvatureof p is I
positiveand chargeis locallydepletedat this point relativeto neighbouring I
points on the bond path. Thus, the formation of a chemicalbond is the I
resultantof two competingeffects:the radial contractionof the charge I
densitytowardsthe bond path,leadingto its concentrationat r" and in the I
interatomicsurface,as opposedby a parallelexpansionof p which removes I
chargedensityfrom r. and from the interatomicsurfaceand resultsin its I
separateaccumulationin eachof the atomicbasins.Which of the two effects I
dominatesa giveninteractionis determinedby the signof the Laplacianof p I
at r". Thecompletespectrumof valuesfrom positive,asis characteristic of the I
interaction betweentwo closed-shellsystems,to negative,as found for I
interactionswith a sharingof chargeaccumulated in the internuclearregion, I
,sobserved

I
3.3 A THEORYOF MOLECULARSTRUCTURE 87

All of the C-H bonds and all of the C-C bonds shown in Fig. 3.7,savethe
have negati veval uesfor
b ri d geheadbonds in [1 .1 .1 ] a n d [2 .1 .1 ]p ro p e l l a ne,
Y'po. In these bonds the perpendicularcontractions of p dominate the
interaction and electronic charge is concentrated between the nuclei along
the bond path. The result is a sharing of charge betweenthe atoms similar to
that one associateswith a covalent interaction. The very strained nature of
the bridgeheadbonds in [1.1.1] and [2.1.1]propellaneresultin local concen-
trations of charge in their non-bonded as well as in their bonded regions,a
result which, as discussedby Wiberg et al., 1987,accountsfor their suscepti-
bility to acetolysis. When there is substantial charge transfer across an
interatomic surface,electronicchargeis separatelyaccumulatedin the basins
of the two atoms and, as a result, A3,the parallelcurvatureof p atr",is large
and its sign dominates the value of Y2p6. The presenceof a large positive
'stiffness'to the bond, making it
curvature along the bond path imparts a
resistant to change by changing substituent effects.On the other hand, the
large accumulation of charge in the internuclear region obtained when Y'po
is negative,resultsin a softeningof the parallel curvature and 2. is relatively
small in value. These bonds are more easily perturbed, the shift in the
interatomic surfaceand the associatedchangesin the other bond parameters
increasingwith a decreasein the value of 1". (Slee 1986).
The full usefulnessof the classificationusing Y'po must await the develop-
ment of the quantum mechanicalaspectsof the theory. The Laplacian of the
charge density appearsin the local expressionof the virial theorem and it is
shown that its sign determinesthe relative importance of the local contribu-
tions of the potential and kinetic energiesto the total energy of the system.A
full discussionof this topic is given in Section 7.4.
Theseexamplesdemonstratethat molecular structure and its stability are
predicted by a theory which uses only the information contained in the
quantum mechanicalstate function and that the static and dynamic proper-
ties of a bond can be characterizedin terms of the properties of the charge
density at the bond critical point. The valuesof po, o,6,E, andY2 p6enableone
ro translate the predicted electronic effectsof orbital models into observable
consequencesin the charge distribution.

3.3 A theory of molecular structure

-1.3.1 A coming together of mathematicsand chemistry

The definition of structure and structural stability for a molecular systemas

presentedabove is an example of the application of a general nrathematical
rheory of structural stability. This work has evolved, under the general
headings of differential topology and qualitative dynamics. a theorl of
88 M O L E C U L A R S T R U C T U R EA N D I T S C H A N G E 3.3

dynamical systemsand their stabilities.Contributions of particular import-

ance to the presentwork are the theory of elementarycatastrophesdeveloped
by R. Thom (1975) and the general statement of the theorem of structural
stability given by J. Palis and S. Smale (1970).What must surely strike the
reader during the unfolding of this application is the remarkable easewith
which the abstract mathematical ideas are mapped on to the topological
properties of the charge distribution, a real physical object.
A central problem of catastrophe theory is to model the discontinuous
changesin the observed behaviour of a system that are causedby smooth,
continuous variations in the valuesof the parameterswhich control it. In the
light of the discussion already presented, this is the central problem of
molecular structure-to predict the discontinuous changesin structure that
are causedby a continuous changein the nuclear coordinates,the parameters
which control the behaviour of the charge density as observedin real space.
This aspect of the theory-the mathematical modelling of the structural
changes occurring in the neighbourhood of a bifurcation point in nuclear
configuration space-is developedin Chapter 4. A very good exposition of
catastrophetheory is given by Poston and Stewart(1978).In the next section
we restate the equivalencerelationship and its application to the gradient
vector fields of the charge density in the conciselanguage of set theory and
linear mappings. It is demonstrated that the two mechanismsof structural
instability encountered in a molecular system, the bifurcation and conflict
mechanisms,are predictedas a corollary of the theorem of structural stability
proposed by Palis and Smale (1970).A brief summary of the elementary
notions of set theory appearsin Section 83.1. A reader, disinterestedin the
more mathematical as opposed to the applied aspectsof the theory, may
peruse the examplesand conclusionsand then proceed to Chapter 5.

3.3.2 The equivalencerelationship and structural stability

The idea of studying the properties of the gradient vector field of the charge
density has its basis in quantum mechanics.The boundary condition for the
definition of a quantum subsystem,eqn (2.9),is stated in terms of this field-
that the surfaceof the subsystembe one of zero flux in the gradient vectors of
the charge density. The resulting study of the topology of the charge density
reveals the existencenot only of atoms as defined by the zero-flux surface
condition, but also of a molecular graph as defined through the pairs of
gradient paths associatedwith the positive eigenvectorof the (3, - 1) critical
points. In fact the two manifolds, as defined by the trajectoriesthat terminate
and originate at such a critical point, define, respectively,the interatomic
surface shared by the basins of neighbouring attractors and the bond path
which links them. Thus, the topology of the charge density is such that the
3.3 ATHEORYOFMOLECUL,{RSTRLCTURE 89

quantum definition of an atom is inseparablefrom the definition of a

molecular graph. It is good fortune or design that this is so, for the idea
underlying the elementary theory of structure is that it is to be found in the
dynamical properties of the gradient vector field of the system's scalar
realizatronin behaviour space.It is important to remember that, while the
structural aspectsof the theory of atoms in moleculesare describedby the
dynamical changesin the topology of the chargedensity,the theory is rooted
in quantum mechanics.It is the atom and its properties which are defined by
quantum mechanics.The bond paths and the structure they definejust mirror
and summ arize in a convenient way what the atoms are doing, performing
the same role here as does the assignmentof a set of bonds in the molecular
structure hypothesis.
To define structure and structural stability, use is made of an equivalence
relation of vector fields over R3. The equivalence relation is defined as
follows:two vector fields v and v' over R3 are said to be equivalentif and only
if there existsa homeomorphism,i.e.a bijectiveand bicontinuous mapping of
R3 into R3, which maps the trajectoriesof v into the trajectoriesof v'. By
applying this definition to the gradient vector fields Y p(r; X), X e RQ, one
obtains an equivalencerelation operating in the nuclear configuration space
R0 which states:two nuclear configurations X and X' e Ra are equivalent if
and only if their associatedgradient vector fields Vp(r; X) and Vp(r; X') are
equivalent.We further say that the nuclear configuration X e Ro is structur-
ally stableif X is an interior point of its equivalenceclass.In other words, one
can always find a neighbourhood V of a structurally stable configuration X,
such that Z is totally contained in the equivalenceclassof X. All configura-
rlons in Z possessthe samemolecular graph as does the stable configuration
\.
Indeed for any point X' of V one can find a homeomorphism, h*r,, which
maps the gradient paths of Vp(r; X) into those of Vp(r; X'). It is easily seen
:hat through ftxx, acritical point of type (a,o)in p(r; X) is mapped on to a
;ntical point of the same type in p(r; X'). Moreover, a gradient path
,'onnectinga pair of critical points (r", r!) in p(r; X) is transformed into a
.orresponding path of Vp(r; X'), which connectsthe imagesof r" and r'". Since
rhe distance between X' and X can be made arbitrarily small and since all
nuclear configurations in V are equivalent, one finds that the molecular
graphs associatedwith the points of Z consist of the same number of bond
parhs linking the same nuclei. These molecular graphs represent a single
frructure,and the maximal neighbourhood contained in the equivalenceclass
"''fX is called the structural region associatedwith X.
.{n equivalencerelation for molecular graphs is defined as follows: two
molecular graphs are equivalent if and only if they are associatedr,r'ithtwo
pornts of the same structural region. An equivalenceclass of molecular
gaphs is calleda molecularstructure.It is then seenthat a unique molecular
90 M O L E C U L A R S T R U C T U R EA N D I T S C H A N G E 3.3

structure is associatedwith a given structural region and that molecular

structure, as defined above through the equivalence of molecular graphs,
necessarilyfulfils the requirement of being generic.
Thesedefinitions were previously illustrated in terms of the gradient paths
shown in Fig. 3.1 for successivestagesalong a dissociativepathway of the
water molecule. As discussedin terms of these examples,the application of
the notion of structural stability to the topological study of the molecular
charge distribution leadsto a partitioning of the nuclear configuration space
into a finite number, /, of non-overlapping regions,the structural regions, Wi,
i: !, . . . ,1, each of which is characterizedby a unique molecularstructure.
These structural regions form a denseopen subsetof the nuclear configura-
tion space,i.e.
/ r \
( U W iI : R Q ( 31
.)
\i=1 /

where the symbol ,4 denotesthe closure of the set .4 and [J denotesthe set-
theoretic union. A point which belongs to the union of the I'Il' belongs to
some structural region and is called a regular point. A nuclear configuration
belonging to the complementary of the set of regular points is called a
catastrophepoint. The catastropheset, C, is the collection of all structurally
unstablepoints of nuclear configuration space. Let 0W, denotethe boundary

of the structural region ly',. Since [-J W,is densein Ro, we have

c - u (aw),
I
(3.2)

i.e.the catastrophesetis the union of the boundariesof all / structural regions

W,. Equation (3.2) denotes the catastrophe set C as the loci of structural
changes.Indeed,according to eqn (3.2),a catastrophepoint Y e C belongsto
o f X i n R o t h u s h a sn o n -
A W , , f o r s o m ei e { 1 , . . . , / } . A n y n e i g h b o u r h o o d
empty intersectionswith W,and at leastone structuralregion Wi,withi * i.
Consequently, the slightest displacement of the system from the nuclear
configuration I will cause the molecular graph to change from the graph
associatedwith I to one which representseither the structure over W, or that
over VVr.ThuJ, the molecular graph associatedwith a catastrophe point
denotesa discontinuouschangein structure which resultsfrom a continuous
variation in the set of control parameters,the nuclear coordinates.
As discussedabove and illustrated in Fig. 3.4, the partitioning of nuclear
configuration space obtained as a result of the definition of molecular
structure leads to the concept of a structure diagram. The space RO is
partitioned into a finite number of structural regions with their boundaries,
as defined by the catastrophe set, denoting the configurations of unstable
structures. This information constitutes a system's structure diagraffi, &
I

3.3 A T H E O R YO F M O L E C U L A RS T R U C T U R E 9I

diagramwhich determines all possiblestructuresand all mechanisms of

structuralchangefor a givenchemicalsystem.
The role of a structurediagramis equivalentto that of a phasediagramin
thermodynamics. In both cases,the spaceof the control parameters,the
nuclearcoordinatesin one case,P, V, and f in the other, is partitioned
throughthe assignment of somepropertyto eachpoint in this space.This
propertyis genericsinceit assumes a uniquevalueovereachstructuralregion
or phase. The boundaries between such open regionscorrespondto in-
stabilities,i.e. to discontinuous changes in the genericproperty.Thus, an
important consequence of the definition of molecularstructureis that a
changein structure,like a change in phase, is predictedto be an abrupt,
discontinuousprocessresulting from a continuous changein the control
parameters.
The examplespreviouslydiscussed with reference to the structurediagram
demonstrated theexistence of two kindsof catastrophe points,calledbifurca-
tion and conflictpoints.Both typesof instabilities were illustratedin termsof
the behaviourobservedfor molecularchargedistributions.What we now
showis that the existence of thesetwo kinds of catastrophes and just these
two,is a directconsequence of a theoremof structuralstabilitystatedby Palis
andSmalein 1970.This theorempredictswhat arethe two basicmechanisms
for structuralchangein a chemicalsystem.

in R3
of linesand surfaces
3,3.3 Stableand unstableintersections

In orderto statetheconditionsunderwhichonewill observea catastrophe of

the conflicttype,it is necessaryto introducea numberof definitionsand the
notion of a transversalor stableintersectionof two submanifoldsof three-
dimensional space.
Atomsand bondsaredefined,respectively, by surfacesand linesembedded
in three-dimensional space.A surfaceand a line are submanifolds of dimen-
sionstwo and one,respectively, of R3.It is necessarythat the surfaceor line
beembedded smoothlyin R3,i.e.that it possess a uniquetangenthyperplane
at each point. This condition of smoothnessis certainly satisfiedin the
presentcasesincean interatomicsurfaceanda bondpatharedefinedin terms
of thetrajectoriesof the gradientvectorsof thechargedensityassociated with
a (3, - 1)criticalpoint,asthe surfaceand axisof a ring structurearedefined
b1'thetrajectories associated with a (3, + 1)criticalpoint.One can picture
rhetangentplaneto a point of the curvedinteratomicsurfaceillustratedin
Fig.2.6,a planedefinedby the gradientvectorof p at that point on the
surface.
The problemwe wish to consideris how to judgewhetheror not a given
rntersection of the manifoldsof two criticalpointsis stableor unstableto a
slightjiggling of the nuclei-does the natureof their intersectionremain
92 M O L E C U L A R S T R U C T U R EA N D I T S C H A N G E 3.3

unchangedor is it radically altered?The two kinds of intersectionsare termed

transversal or non-transversal,respectively.As a simple example, consider
the possible interactions between a plane and a curve in R3 as pictured in
Fig. 3.10.When the curve intersectsthe plane at somefinite number of points,
or does not intersect the plane at all, we say that their intersection is
transversal. Such an intersection will obviously not be affected by small
changesin the relative positions of the curve and the plane. If, on the other
hand, the curve happens to be tangent to the plane, the intersection is non-
transversal. A small change in their relative positions will result in a
transition from a non-transversalto a transversalone.
Somenecessarydefinitions are first introduced. Consider a critical point r",
labelled by its rank and signatureas (rr-r,
o). The stablemanifold of r. is defined
to be the (ll2)(a - o)-dimensionalmanifold generatedby the eigenvectorsof
the Hessianmatrix at r" associatedwith its negativeeigenvalues.Its unstable
manifold is the (l l2)(a * o)-dimensionalmanifold generatedby the eigenvec-
tors associatedwith the positive eigenvaluesof the Hessianmatrix. Thus, for
example, the interatomic surface and the bond path are the stable and
unstable manifolds respectively,of a (3, - 1) critical point.
The general statementsregarding the transversalityof the intersectionsof
submanifoldsare stated in terms of the properties of the intersectionsof their
associatedtangent tryperplanes.These are vector subspaces,a plane or line
spanned by a corresponding set of gradient vectors, not necessarilyat the
origin of the coordinate system. Such a vector subspaceis called an affine
subspace.

Ftc. 3.10. The diagram on the left illustratesa transversalintersectionof a line with a surface.A
motion of either manifold relative to the other will not changethe nature of this intersection.The
diagram on the right illustratesa non-transversalintersection.The line is tangent to the surface
at one point. This situation is not stable to arbitrary relative motions of the two manifolds. Such
motions will cause the line to either intersect the surfaceor to not sraze the surface.
I
I

3.3 A THEORYOF MOLECULARSTRUCTURE 93

Let E and F be such affine subspacesin p: of dimensions s and f,

respectively.The intersectionof E and F is said to be transversalif either

(1) EnF:0, t h a t i s , t h e y d o n o t i n t e r s e c to. r

( 2 )s + t ) - 3 a n d d i m ( E n F ) : s * t - 3 .

One can now statethe conditionsfor the intersectionof two submanifoldsM

and ,^{ in R3 to be transversal:
( 1 ) M nl{ : 0, or

(2) the tangent affine hyperplanesof M and l{ intersect transversely.

Examples of non-transversalintersectionsof bond paths, ring axes, and

interatomic and ring surfacesare illustratedin Fig.3.11. They representthe
conflict structures that can be found in the topologies of molecular charge
distributions. The first arises from the intersection of a bond path of one
(3, - 1) critical point (its unstablemanifold) with the interatomic surface(the
stable manifold) of a second such critical point. Examples of the resulting
unstablestructure have beengiven previously in the dissociationof water and
in the isomerization of hydrogen cyanide (Figs 3.1 and 3.6). The example
given in Fig. 3.11is for the minimum-energystructureof the CHi molecular
ion. The molecular graph suggeststhat the structure of this ion corresponds
to a CHf ion interacting with a hydrogen molecule.The secondexample of a
non-transversalintersection results from the intersection of the stable and
unstablemanifolds of (3, + 1) critical points. In this casethe ring axis of one
ring critical point intersects with the surface of the other. The resulting
structure is exemplified by the molecular graph shown for the
[1.1.l]propellanernoleculein which the bond path betweenthe bridgehead
carbons has been previously annihilated in a bifurcation catastrophe.(The

r r +
n
'\- I
,"K
rrAH
I I A \
-4"-*-----l-

(a) (c)
F r c . 3 . 1 1 . E x a m p l e s o f t h e p o s s i b l e n o n - t r a n s v e r s a li n t e r s e c t i o n si n a m o l e c u l a r s y s t e m :
. i r b e t w e e nt h e s t a b l ea n d u n s t a b l em a n i f o l d s o f t w o b o n d c r i t i c a l p o i n t s i n C H . : ( b l b e t * e e n
:ne stable and unstable manifolds of two ring critical points in a dtstorted structure of
- l 1 . 1 ] p r o p e l l a n e(;c ) b e t w e e n t h e u n s t a b l em a n i f o l d o f a b o n d c r i t i c a l p o i n t * r t h t h e s t a b l e
manifold of a ring critical point (the ring axis) in face-protonated*cloprtrpane
 M O L E C U L A R S T R U C T U R EA N D I T S C H A N G E 3.3

structures obtainedfor the [1.1.1]propellane moleculein the neighbourhood

of this bifurcationcatastrophe point are discussed in somedetailin Chapter
4.)The final exampleconcernsthe intersection of the bond path of a (3, - 1)
critical point with the ring axis of a (3, + 1) critical point. The example
shownin Fig.3.11(c) is obtainedfor what is referredto as face-protonated
cyclopropane. This unstablestructureis transformedinto the corner-pro-
tonated specieswhen the bond path from the proton switchesfrom
the one-dimensionalring attractor to any one of the three nuclear
attractors.
The instabilityarisingfrom the intersectionof the stableand unstable
manifoldsof two (3, + 1) critical points doesnot lead to a changein the
networkof bond paths,but only to the formationof a new,stabletwo-ringed
structure.The intersectionof a bond path of a (3, - 1) critical point with
either a ring axis or an interatomicsurface,however,leadsto instabilities
whoseremovalresultsin a changein the bondedstructureof a molecule.
From the examplesof the interactionsthat havebeenfound so far, unstable
structuresof the conflicttype appearto be characteristic of relativelyweak
interactionsassociated with longerthan normal interatomicseparations or
with relativelyhigh-energytransition-state structures.

3.3,4 Criteriafor structuralstability

We are now in a positionto adopt Palisand Smale'stheoremon structural

stability (1970)to describestructuralchangesin a molecularsystem.A
configurationX e R0 is structurallystableif p(r; X) has a finite numberof
criticalpoints suchthat:

(1) eachcriticalpoint is non-degenerate,

and

(2) the stableand unstablemanifoldsof any pair of criticalpointsintersect

transversely.

The immediateconsequence of the theoremis that a structuralinstability

can be established throughone of two mechanisms which correspondto the
bifurcation and conflict catastrophespreviouslydes'cribed. A changein
molecularstructurecan only be causedby the formation of a degenerate
critical point in the electronicchargedistributionor by the creationof an
unstableintersectionof the submanifolds of (3, - 1)and/or (3, + 1)critical
points.The mechanisms of the making and breakingof chemicalbondsare
thusseento be the predictedconsequences of changesin the morphologyof a
molecularchargedistribution.
I

3.4 DIAGRAM
OF A STRUCTURE
COMPARISON 95

of Ra
3.4 Comparisonof a structurediagramwith otherpartitionings

of
stabilities
3.4.1 Relationbetweentopologicaland energetic
molecularstructures

A structurediagrampartitionsnuclearconfigurationspaceinto structural
regions.The stablestructuralregionsare separated by hypersurfaces which
are the loci of the unstablestructures.The Born-Oppenheimerpotential
energyis also definedover nuclearconfigurationspaceto yield an energy
hypersurface E(X) for a molecularsystem.In termsof the criticalpoints of
E(X) one may define and distinguishbetweenenergeticallystable and
unstablegeometries. Thus, there are two definitionsof stabilityassociated
with the nuclear configuration spaceof a system.The first definesregionsof
this space as determined by the topologicalpropertiesof the system's charge
distribution. The other defines points within the same space as determined
by the propertiesof the system'senergyhypersurface. One may inquire as
ro the extentto which these two definitionsof stabilitymay be juxtaposed
(Tal et al. l98l).
The readeris remindedthat it is importantto distinguishbetweenmolecu-
lar geometryand molecularstructure.Geometryis a non-genericproperty
sinceany infinitesimalchangein X resultsin a differentgeometry.Molecular
structure,on the other hand,is a genericproperty.Any configurationX'in
the neighbourhoodof a given regular point X, while it has a different
geometry, possesses the samestructure.Only in this way canoneaccountfor
the invarianceof the networkof bondsto the nuclearmotionsof a system.
The criticalpointsof E(X) enableoneto distinguishbetweenenergetically
stableand unstablegeometries. Energeticallystablegeometriesare points
X e R0 for which E(X) is at a local minimum with respectto all internal
motionsof a system.Energetically unstablegeometries are saddlepoints in
E(X),geometries whichare at a maximumwith respectto oneor moreof the
'reactionco-
s1'stem'sinternal coordinates.One may also define unique
ordinates'(Tachibanaand Fukui 1978)in termsof trajectories of - VxE(X).
A reactioncoordinateconsistsof a pair of trajectoriesof - VXE(X)origin-
aringat a saddlepoint and terminatingat two differentlocal minima.The
associated energeticallyunstablegeometryis the transition state for the
conversionof one stablegeometryinto the other.
Mezey(1981)has proposeda partitioningof nuclearconfigurationspace
Ra into catchmentregions as determinedby the gradient vector field,
- VxE(X).Thislatterpartitioningis similarto the partitioningof a molecu-
iar chargedistributioninto atomicbasins. Corresponding to thedefinitionof
an atomicbasinas the openregionof R3 traversed of
by all the trajectories
fp which terminatea given nucleus,Mezey'scatchmentregion may' be
;rmply definedas the open regionof R0 traversedby all the trajectoriesof
96 M O L E C U L A R S T R U C T U R EA N D I T S C H A N G E 3.4

- VxE(X) which terminate at a local maximum in - E(X) (an energetically

stable geometry). We have found that a molecular graph is, in general,
invariant to the nuclear motions of a molecule in the neighbourhood of an
energeticallystable geometry. Thus, the definition of molecular structure as
an equivalenceclass of molecular graphs associatesa given structure with a
neighbourhood of an energeticallystable geometry.
The boundaries of a catchment region or basin of E(X) are the surfacesof
zero flux in the vector - VyE(X). These boundaries represent the loci of
unstable geometries,as E(X) on a boundary is a maximum with respect to
one or more of the nuclear motions of the system. Such an energetically
unstable geometry, which represents the transition state on a reaction
coordinate, may or may not lie on the boundary between two structural
regions of the structure diagram defined by the topological properties of a
system'schargedistribution. It is not possibleto give a generalanswer to the
question 'Do the geometriesof the energeticallyunstable configurations of a
molecular system correspond to the topologically unstable structures of the
system as defined by its catastrophe set?' since the functional relationship
betweenthe chargedensity and the energyis not known. We have discusseda
number of examples above where this coincidence of energetic and topo-
logical instabilities is indeed found, but it is clear from the same set of
examples that a catastrophe point will not always have an energetically
unstablegeometry associatedwith it. The C2, approach of an oxygen atom to
a hydrogen molecule,describedabove (Fig. 3.1),is energeticallydownhill and
no maxima in E(X) along the reaction coordinate are found at the cata-
strophe points denoting the formation and destruction of the ring structure.
That is, the stable ring structure does not appear as an energy intermediate
along the reaction coordinate. It is also true that not all reactionsinvolving
an energy barrier involve a change in the topological structure of a system.
The transition-state geometriesfor reactions involving internal rotations or
inversionscannot belong to the catastropheset as both reactant and product
possessequivalent molecular graphs in these cases.
The gradient vector field of - E(X) which determinesthe partitioning of
the energy hypersurfaceis determined in turn by the Hellmann-Feynman
force exerted on the nuclei. The electronic contribution to this force is given
by an integralover the chargedensity;thus,the topology of E(X)is not totally
unrelated to the form of the charge density. One can, in fact, re-expressthis
electronic contribution as an integral over the gradient vector field of the
charge density Y p(r; X), weighed by the nuclear-rlectron potentials. It has
been shown (Tal et al. 1981)that one may obtain an understanding of the
correspondencebetween an energetically unstable critical point as deter-
mined by the field - VE(X) and an unstable structure as determined by the
field Vp(r; X) when this correspondenceis indeed observed,as in the iso-
merizations of HCN and its methyl derivative, for example.
3 .4 C O M P A R I S O NO F A S T R U C T LR E D I A G R A M 97

3.4.2 Structural homeomorphism betu'eenp and the nuclear potential

We have frequently commented on the fundamental role played by the

nuclear-electronpotential in the determinationof the topologicalproperties
of the electronicchargedistribution. Becauseof the generalobservationthat
the only local maxima of chargedistributionsof many-electronsystemsoccur
at the positionsof the nuclei,the nucleiare the point attractorsof the gradient
vectorfield of the chargedensity as well as of the gradient vector field derived
from the nuclear potential. To what extent is the form exhibited by the charge
density determined by the properties of the nuclear potential V(r; X)? The
similarity of two fields is bestjudged by comparing their topological proper-
ties. A considerable similarity is shown to exist between the scalar fields
p(r; X) and V(r;X) through a comparison of the structure which is definable
in terms of their associatedgradient vector fields.Indeed, it has been argued
that the structure diagrams generated by these two scalar fields will, in
general,be found to be homeomorphic (Tal et a/. 1980)'
The topology of the nuclear potential

V ( r ; X :) L , Z , ( l r - X "l ) - ' , X : {X"}, (3.3)

is displayed and characterizedby its gradient vector field V V(r;X), where the
gradient is taken with respectto the electronic variable r in a fashion totally
analogousto that followed in the analysisof the topology of p(r; X). The field
of eqn(3.3)becomesinfinite if and only if r : Xo, the coordinatesof one of the
nuclei. At such a point V Z is both discontinuous and infinite. Thus, the
maxima in V(r:X), like the corresponding maxima in p(r; X), are not true
(3, - 3) critical points. However, again as for p(r; X), the phase portrait for
- 3)critical point. This
this point is indistinguishablefrom that for a true (3,
behaviour is made evident in Fig. 3.12which portrays maps of the trajectories
of Y V(r; X) and also of Yp(t; X) for the Cr" approach of an oxygen atom to a
hydrogen molecule. All of the trajectories of Y V in the neighbourhood of a
given nucleus terminate at that nucleus.Thus, a nucleus acts as an attractor
in both V(r;X\ and p(r; X) fields. Moreover, nuclei are the only attractors in
L'(r;X) and they are observed to be attractors in many-electron charge
distributions. Since topological structure and its stability are determined
through the attainment of a state of balance in the competition between the
attractors of a system,one anticipatesa significant degreeof similarity in the
topological properties of V and p. Thus, one finds that the basins of
neighbouring attractors in both fields are separatedby the surfacegenerated
by the trajectories which terminate at a (3, - 1) critical point and the
attractors themselvesare linked by the unique pair of trajectorieswhich
originate at this critical Point.
Figure 3.12 shows that the same sequenceof elementar)'graphs for the
H,O system is obtained for V(r:X) as is obtained for p(r: X). including a
98 M O L E C U L A R S T R U C T U R EA N D I T S C H A N G E 3.4

Frc. 3.12. Gradientmapsof thenuclearpotentialZ (r; X) on theleftandof theelectronic

charge
densityp(r;X) on the right for symmetricstructuresoi the HrO system.The pairs of maps
demonstrate the existence
of equivalentstructures,
3.4 C O M P A R I S O NO F A S T R U C T U R ED I A G R A M

graph correspondingto the bonded ring of nuclei,Fig. 3.12(b).However,as a

consequenceof Poisson'sequation

Y2V1r;X) : -4nL,2,6(r - Xo), (3.4)

the topological properties of Z(r; X) are constrained relative to those for
p(r; X). Since
X) : L,i,,
Y2V(r; (3.s)
the sum of the eigenvalues,1,of the Hessian matrix for a critical point in
V(r;X\ other than at a (3, - 3) critical point must equal zeto. No such
restrictionexistsfor critical points in p(r; X). Thus V(r;X) may possessboth
(3, - 1) and (3, + 1) critical points but with the restriction that IA, + Atl
: l)t.1,where,t, is the eigenvalueof unique sign.It cannot possessa (3, + 3)
critical point and hence it cannot exhibit a graph corresponding to a cage
structure.However,as discussedlater,evenin this caseV(r;X) may possessa
graph which is topologically equivalent to a bonded cage of p(r; X).
A gradient path of V has a simple physical interpretation. It is a line of
force-the path traversedby a test charge moving under the influenceof the
potential V(r:X). At a critical point other than a (3, - 3) critical point, the
force vanishes.Thus a critical point in the field Z(r; X) denotes a point of
electrostaticbalance between the attractors of the system.Since trajectories
'zero-
defining the surface which separatesneighbouring basins satisfy the
fl u x' c ondit ion
YV(r;X)'n(r): Q for all r e S(r), (3.6)
a test charge on such a surface is not drawn to either attractor. On the
contrary, a test charge lying within a basin is drawn to the attractor (the
nucleus)and, thus, the boundary of a given basin definesthe region of space
dominated by the field of its own attractor. With a common set of elementsas
attractors, V(r; X) exhibits the same basic elements of structure as does
p(r; X), with the atoms and bond paths of the latter being replaced by the
topologically equivalent nuclear basinsand nuclear-nuclearinteraction lines
of the former field. The simple and direct physicalinterpretation one may give
to a nuclear basin of Z(r; X) as a region of space dominated by the
electrostaticforce exertedby its nucleuson an elementof electroniccharge is
the embodiment of the idea of what an atomic basin of p(r; X) represents-a
region of spacewhose chemistry is dominated by its contained nucleus.
One finds that the structure diagrams obtained for V(r; X) and p(r; X) for
the HrO systemare homeomorphicin the sensethat both exhibit an identical
partitioning of the control space yielding the same sets of structures,both
stableand, as illustratedin Fig. 3.13,unstable.In addition to finding the same
setsof structuresfor both fields,it has been found that their mechanismsof
structural change are also the same. Thus, the bifurcation and conflict
 ()

ooq
'ic ' A
5.

o.9
II
?9- I

- {.)
{.)-
i:=
6 t
;.tr
€ *
E -

C . ) O
=oo

- 9

()b0

o 6
rr C.)
a d

C) ci)

o =
t s r

a =
a
q ) 0 )
F C L

L X
,rX
ho!

X X
i.l
v v
i.i

* o

EE
q).n
v'F

e*
Li o0
A . A

,i: F

F u^.
C ) O
E a
( ! 9

" 6
. L

a.;

o
IJ.
3 .4 C O M P A R I S O NO F A S T R U C T U R ED I A G R A M 101

mechanismsdescribeall possiblestructural changesin both fields.The formal

statementof this homeomorphismis given in the paper by Tal et al' (1980).
Other systemshave not been studied over such a broad range of nuclear
configurations as has been done for the water system,but comparisonsof the
structures determined by V(r;X) and p(r; X) have been made for many
moleculesat their equilibrium geometries.In general,equivalent structures
are obtained. In some instancesthe graph for V(r; X)must be determinedat a
geometry other than the equilibrium one to obtain an equivalent structure.It
seemsreasonableto conjecture that the structure diagrams fot V(r; X) and
the ground state p(r; X) will, in general,be homeomorphic for a particular
systim. The primary reason for making this conjecture is that p(r; X) is
observedto have the same set of attractors as does V(r;X\; the topological
properties of both fields are controlled and determined by the forces exerted
by the nuclei.
There are a few limiting cases where the mapping between the two
structure diagrams is not possible becauseof constraint imposed on V(r;X)
by Poisson'sequation (eqns (3.4) and (3.5)).Thus Z(r; x) cannot possessa
(3, * 3) critical point, for in such a caseeach A, > 0. Thus, in the interior of a
bonded cage of nuclei, where Vp possessessuch a critical point, the corres-
ponding point in Z must be a degeneratecritical point with three zero roots'
In tetrahedrane,for example,such a critical point resultsfrom the coalescing
of a (3, + 3) critical point with the six (3, - 1) critical points whose bond
paths define the bonds of the cageand with the four ring critical points whose
surfacesenclosethe cagecritical point. This singularity in V behavesas does a
sixfold degeneratebond critical point, and the resulting molecular graph is
characterizedby six bond paths linking the four nuclei as in a regular cage
structure.Thus the physical equivalencein structure observedfor p(r; X) and
t,(r; X) as determined by the equality in the number of their interaction lines
and the nuclei they link is maintained even in the casewhere the constraint of
the Poisson equation prevents a one-to-one mapping of the two structure
diagrams.

1.4.3 The mechanicsof p in the one-electroncase

The potential energyoperator which determinesthe electron-nuclear attract-

iveenergyforasingleelectronist(r;X):-ev(r;X)'Thusforafixed
configuration of the nuclei, an interaction line in an elementary graph of
l'(r; X)representsa minimum-energypath linking two neighbouringnucleiin
the potential energy surface f . If V were the sole potential acting in a
molecular system, one might expect the electronic charge densitl to be
concentratedalong correspondinglines,i.e. the interaction lines of I-would
correspondwith the bond paths of p. In a many-electrons)'stemhtr\\e\ €r.
r02 M O L E C U L A R S T R U C T U R EA N D I T S C H A N G E 3.5

there are electron-electron repulsive forces acting in addition to the force

determinedby the external potential. Thus, the homeomorphism betweenthe
structure diagrams of V and p does not, in general,map a point X e Ra on to
itself,but rather on to another point X' + X. For a given X the critical points
of V do not coincide with those of p. This is true even for a one-electron
system where Z is the sole Coulombic potential as we now show. One may
obtain from Schrodinger'sequation for a one-electroncase the expression
Ep(r): - $14)v2p(r)+ (1/8)tvp(r)'vp(r)l/p(r) + "t'(r)p(r) (3.7)
whereE is the total energyand p(r) -- t*@t(r). This equationmay be used
to obtain an expressionfor VV(r) at r", z critical point in p(r), where
Vp(r) : Q,
Y V(r"\: - Y t(r") : - Ol4)V[V'p(r")] I pF). (3.8)
Unlessdemandedby symmetry,the value of Y2p will not be an extremum at a
critical point in p. Thus, the critical points in V and p will not, in general,
coincide and the distribution of electronic charge, even in a one-electron
system, is not determined entirely by the external force - Y V(r). Bohm
(1952)ascribedthe stability of a stationary state in a quantum systemto the
balancebetweenthe classicalforce - V V and the quantum mechanicalforce
'quantum potential'. For a one-electron
which is given by the gradient of the
systemat a critical point in p(r), Bohm's quantum mechanicalforce is just the
right-hand side of eqn (3.8).The Laplacian of the chargedensity,the quantity
Y2p appearingin the quantum potential, is an important local property of a
system and is the subject of Chapter 7.

3.5 Summary

This chapter has demonstrated that the chemical notion of structure finds
physical expressionin the topological properties exhibited by a molecular
charge distribution. The question as to what is the structure of a given
molecule can be unambiguously answered.Such questionshave been central
to many of the protracted and vigorous discussionsof chemistry in the past
quarter century, the structuresof non-classicalions being one example.As is
typical of a theory replacing a model, however, new questions can be both
asked and answered.Not only can a structure be assigned,but its stability
predicted and the mechanismsof a changein structure understood.Structure
and structural stability are inseparable concepts. The following chapter
demonstrateshow Thom's theory of elementarycatastrophescan be used to
obtain a quantitative description of the structural changesencounteredin the
neighbourhood of a bifurcation catastrophepoint. As will be demonstrated.
an abstract mathematical theory can predict all of the topological changes
and the corresponding structural changesexhibited by a molecular charge
 N A T U R A L C O O R D I N A T ES Y S T E M 103
F3,2

distributionfor motion of a systemin the neighbourhoodof a bifurcation

point.

E3.1 Summary of set theoretic language

x eX x is a member of the set X

X c Y X is a subset of Y if every element of X is an element of I''
)/ contains X.
{xlp(x)} the setof all elementsx for which a specificproperty P(x) holds.
XvY:{xlxeX orxe I} u n i o n o ft w o s e t s Xa n d Y '
XaY:lxlxeX andx€ f) i n t e r s e c t i o nt owfo s e t s X a n d Y -
Rn: {(rr,...,Xn)lx, e R f o r i : l, )n} R' is an n-dimensional
Euclidean space.

Open disc: if x e Rn and r > 0, then the open disc of radius r, centre x is
-
d e f i n e dt o b e t h e s e t { y e R ' l l l x - y l l < r } w h e r el l x y l l i s t h e d i s t a n c e
between x and Y.

Open set: a subsetX of Rn is open if for each x e X there is some open disc
centre x lying inside X.
Neighbourhood of x: a subset Y is a neighbourhoodof x if x lies in some
open set X which is contained in X.

83.2 Natural coordinate system for an atomic basin

The definition of an atom, as the union of an attractor and its basin, servesto
define a narural coordinate systemfor an atom in a molecule (Biegler-Konig
et al. 1981).The new systemis obtained by a mapping of the coordinates
parameter s
lr, y, z) of a point r in an atomic basin into the triple (s, 0, 4). The
position of the point r along the gradient path that
teqn (2.6))determinesthe
is defined by some initial set of angular coordinates 0 and @. This local
coordinate system resemblesthe spherical polar coordinate system with the
path parameter s, and the rays replaced by
radial coordinate r replacedby the
the generally curved gradient paths that traverse the basin of a given atom.
Sincethe limits of s are * oo, the mapping sendsthe bounded spaceof each
atom into a complete spacehomeomorphic to R3. Thus, in terms of this
coordinate system,every atom, whether free or bound, is assigneda complete
'curved' becauseof its interactions
space.The spaceof a bound atom, while
olith it, neighbours,is still complete.The collectionof atomic basinsand the
associatedcoordinate transformations yield an atlas for the molecule'
A gradient path g(r^) through a given point r, e R3 is definedby'

' ^ ) : { r , r ,. ^ ' l T : v p ( r ( sX)),r: ( 0:) ' , " } { E -11)

t l
104 M O L E C U L A R S T R U C T U R EA N D I T S C H A N G E F.3.2

We define the a-limit set of g(:r^),the origin of the path, by

a (r' ) : l i m r( s) (E.3.2a)

and its a;-limit set, the terminus "f ,;p;h, by

:
ar(r,") t(t). (E3.2b)
,1T."
For a given set of initial values r(0) : ro, the differential equation

dr(s)/ds: Vp(r(s)) (E3.3a)

definesa unique gradient path. By virtue of the definition of an attractor A of
Vp, this path will be contained in the basin {2 of A, if ro e Q. Equation (E.3.3a)
is equivalent to the integral equation
fs
v (s )(ro ): r(s ) : ' o * . JVp(r(r))dr
o (E3.3b)

which definesa one-parametergroup of diffeomorphisms {u(t), s e R}.

In the following, eqn (E3.3b) will be regarded as defining a coordinate
transformation within Q. The basic ideas behind the construction of this
transformation are as follows. The parameter s will be one of the new
coordinates;to obtain a proper coordinate transformation from R3 to R3, we
need two additional coordinates to specify the initial value of eqn (E3.3a).
Thus r(s : 0) : ro must belong to a two-dimensionaldifferentiablemanifold,
and the two additional coordinates mentioned above are the local co-
ordinates of ro in the manifold. We make the following general observations:
let M be a two-manifold contained in Q, with an atlas {(U ,, E,)}, where the
open sets Uo cover M, and the one-to-one,continuous maps Eo;Uo--R2
define local coordinate systemsin M.Let lulVobe the imagesof M and(Jo,
respectively,under the group of diffeomorphisms {u(t); s e R}, that is

14: t l v(s)(M),
t ,

seR

4: U- v(s)(U")'

3 y h , ( r ) : ( s , E o ( r . ) ) ,V r e V n , w i t h
F o r e a c ha , d e f i n et h e m a p h o : V o - - t R b
r^ : g(:r)A M, and r : r,' * Ji Vp1r1r;)dr. Then the h, definelocal coordin-
atesystems overM.In particular,letM be a sphereof radiusB centredon A,
so that 14 : Q. An appropriateatlasover the sphereM consistsof the single
chart (M, E\, where
r : ( x o , ! o , z o ) eM , 8(ro):(0,6), 0<0 <n, 0<d <2n
F 3.2 N A T U R A L C O O R D I N A T ES Y S T E M 105

with
Xo: Bsin0 cos@, ,o : Bsin[)sin@. zo : {}cos0. (E3.4)

Correspondingly,we obtaina singlecoordinatesystemoverf), illustratedin

Fig. 83.1,throughthe mappingh definedby
h(r): (s,0d
,) Vr e Q,
(8.3.5)
8- t (10,
Q) : g(r)n M : ro(r)
and
fs
' o * . Jvp
o (r(/))dr.
Equation (E3.5) implies that, through the one-to-one map h, we can
uniquely associatewith a point r contained in the basin O, a triple(s,0,Q)
ri here (0, O) determinesthe path through r and s is the coordinate of r along
that path.
This natural coordinate systemis of particular use in the determination of
aromicproperties.As discussedin Chapter 6, the atomic valueof a property F
rs given by the averageover the atomic basin of an effectivesingle-particle
density/'(r).Thus the value of the property F for atom O is

F(o): i11r;0.. (E3.6)

Jo
The evaluation of F(A) is a non-trivial problem since the atomic surface
u hich bounds the region O does not in generalhave a local definition, much
lessa simple geometrical structure. The integration method, developedhere

i : ( , . E 3 . 1 . T h e c o o r d i n a t es y s t e m( s ,0 , d ) w i t h M b e i n g a s p h e r eo f r a d i u s p . A I : o r h t r \ \ r r S3
. . r g l e t r a j e c t o r yo f V p i n t h e i n t e r a t o m i cs u r f a c ew h i c h t e r m i n a t e sa t t h e ( - 1 . - l t c r t t t c a l p t . r r f lttr r l
the z-axis.
106 M O L E C U L A R S T R U C T U R EA N D I T S C H A N G E 83.2

in terms of the natural coordinate system,avoids the direct and often difficult
determination of the atomic surface by implementing the definition of an
atom as the union of an attractor and its basin. By integrating along the
trajectoriesof Vp(r) which terminate at a given nucleus,one must necessarily
cover the basin of an atom. In addition, becauseof the zero-flux surface
condition, it is impossible to cross an interatomic surfaceinto the basin of a
neighbouring atom.
Since the mapping h is one-to-one,we can use its inverse

h-1:R3--O

/ s\ /x\
/ l t l
Iil.t,l'n
\qt \zt
(83.7a)

defined by

/'\ /"\ llsin0cos@1p

f t - '( e : : . (E3.7b)
) {r) (r""gsin@ } "I 'v p ( r ( r ) ) d r
\ d / \ ' l \ B c o s gI
to rewrite eqn (E3.6) in the form

r(o): l'" or l" ot i.-orr,r(s,0,

oDlyft-'(r,g,d))1.
(E3.8)
Jo Jo J-.o
In eqn (E3.8) | t | = | Y (h- '(s, 9, d))l denotesthe determinantof the
Jacobianmatrix of h- r.

lAxl}s Axl00 0xl0d \ /r, xo xo\

Y(h-', : ( eylasoyloooylodI : | ,, le ,rl. (83.e)
\0zl0s 02100 AzIASI \t, zo zal
where

x ,: 0xl0s, x e : 0 x l A 0, etc.
To evaluate F(O) using eqn (E3.8) we need the values of the Jacobian
det er m inant l Y(h -1 )l fo r a l l s , 0 , Q.E q u a t i on (E 3.7b)reveal sthat a di rect
computation of this determinant would involve products of integrals of
the form
, , ? ,p ( r )
,
J,;#o'
Instead,as will be shown below, we can obtain I Y (h- t ) | at all points (t, 0, Q)
as the solution to an initial-value problem. We first obtain a differential
832 N A T U R A LC O O R D I N A T ES Y S T E M IO7

t to s.
eqn(E3.9)with respect
equationfor thematrixY(h- ) by differentiating
Writing eqn (83.9)in the form
y ( h ' t ) : ( r , ,r s , r 6 ) (E3.10)
we have
r, is the vectot ArlAs,
where,for example,
A Y ( h - t \ l 0 t : ( 0 r , fA sA, r r e
f s,furli i. (E3.11)

From r, : Vp, we get,usingthe chain rule'

ilr,f0s : AF dl)s: ;f (p)'r, (F,3.12)

wheretr(p) denotesthe Hessianmatrix of p. From eqn (E3.2a),we have

lFcos0cosdl A fs
r ". ,: | 0 c o s 0 s i n dI + - | v p a t
I I 0 U "l ""
\ Bsina I
so that

*: - !( ! i'o,0,\
as as\ae
' ' * ' ) :2e
ou\ds
f ypdt):9ry0t
/ oa
Jo Jo

Usingthe chain rule,we obtain

1rrlls:,ff(p)'re. (83.12b)
Similarly,we find
A r r l l s- . f f ( p ) ' r o . (E3.12c)
we obtain
equations(E3.l}a-c)into eqn (E3.11),
Substituting
Ay(h-')l0s: af (p).{(h-'). (E3.13)
problem,with initialvaluesat s : 0,
an initial-value
Equation(E3.13)defines
-
f o,F) Pcos0cos@ Psin0sindl
lo: rr(to) Bcos0sin@ psingcos@ (83'14)
I I
\p,(to) Bsin0 0 I

rrherero is givenby eqn (E3.4).

The systemof differentialequations(E3.13)can be reducedto a single
equationfor thedeterminant
Jifferential of Y(h- 1)throughtheapplication of
x
Davidenko's(1960)trace theorem:let 9(s) be an n n matrix whose
eiements functionsin s. Let l?(s)l be the characteristic
are differentiable
polynomialassociated with 0(s). Then,for all s for which ?(s) +0.
A d
i l E ( s ) l : T r ( ? ( s ) - ' )os
; ( " ( s ) )/t$ t { E 31 5 t
ds
108 REFERENCES

polynomial,/ : / tn- 11is thedeterminant

Sincethecharacteristic l/(h-
r)1,
applicationof the abovetheoremyields,usingeqn (83.13),
a ly ( h - ' ) l / d :s T r ( / ( h - ' ) -r 7 7f u t)( h - ' ) ) ly ( h - ' ) l
: Tr(,tr(pDly (h- ' )1. (83.16)
Noting that Tr(.;f (p)) : Y2p, we obtain the differential equation

a ly ( h - ') l / d s: Y 2p ( r )yl ( h - ') 1 , (E3.17a)

with initial value at s : 0,
'
I y ( h - ) l ( 0 ,0 , 0 ) : | , { o l (E3.17b)

where the matrix lo is given in eqn (E3.14).

Equation (E3.8)togetherwith eqns(E3.17)form the basisof the algorithm
describedby Biegler-Konig et al. (1981)for the evaluation of the atomic
property .F(O).
The author wishesto offer his thanks to ProfessorZeeman for bringing to
his attention the theorem on structural stability by Palis and Smaleand for a
stimulating and most usefuldiscussionof our use of catastrophetheory, all of
which occurred during a visit to the University of Warwick in the spring of
1980.

References

Bader,R. F. W., Nguyen-Dang, T. T., and Tal, Y. (1979).J. Chem.Phys.70,4316.

Bader,R. F. W., Nguyen-Dang, T. T., and Tal, Y. (1981).Rep.Prog.Phys.44,893.
Bader,R. F. W., Slee,T. S.,Cremer,D., and Kraka,E. (1983). J. Am.chem.Soc.105,
5061.
Biegler-K<inig, F. W., Nguyen-Dang, T. T., Tal, Y., Bader,R. F. W., and Duke,A. J.
( 1981)J.. P h y s B:
. At. Mo l . Ph y s .1 4 ,2 7 3 9 .
Bohm,D. (1952).Phys.Reu.85, 166.
Boyd,R., Knop, O., and Choi,S. C. (1988). J. Am.chem.Soc.110,7299.
Cade,P. E., Bader,R. F. W., and Pelletier,J. (1971).J . Chem.Phys.54,3517.
Cade,P. E. and Huo, W. M. (1975). Atom.Data nucl.Data Tables15,1.
Cade,P. E. and Wahl, A. C. (1974).Atom.Data nucl.Data Tables13,339.
Coulson,C. A. (1951). Proc.r. Soc.A207,91.
Coulson,C. A. (1972).From Prefaceof first edition of The shapeand structureof
molecules (2ndedn)(ed.R. McWeeny).Oxford UniversityPress,Oxford.
Cremer,D. and Kraka, E. (1984). CroaticaChem.Acta 57,1259.
Cremer,D., Kraka,8., Slee,T. S.,Bader,R. F. W., Lau, C. D. H., Nguyen-Dang, T. T..
and MacDougall,P. J. (1983). J. Am.chem.Soc.105,5069.
Dav idenk o, D. F . (1 9 6 0 ).
S o uJ. . M a th .1 ,3 1 6 .
Gatti,C'Barzaghi,M., and Simonetta, M. (1985). J. Am.chem.Soc.107,878.
Mezey,P. G. (1981).Theor.Chim.Acta 58,309.
Mulliken,R. S. (1932).Reu.mod.Phvs.4,l.
I

REFERENCES 109

M u l l i k e n , R . S . ( 1 9 3 9 )P
. hys.Reu.56,778.
O'Hare, P. A. G. and Wahl, A. C. (1970).J. Chem'Phls. 53.2469.
Palis,J. and Smale,S. (1970).Pure Math. 14.223.
Poston, T. and Stewart, I. (1978).Catastrophe theorv and its applications.Pitman,
London.
Slee,T. (1986).J. Am. chem.Soc. 108, 606.
Tachibana,A. and Fukui, K. (1978).Theor' Chim. Acta 49,321.
Tal, Y., Bader, R. F. W., and Erkku, J. (1980).Phys. Reu' A2l, l.
Tal, y., Bader, R. F. W., Nguyen-Dang,T. T., Ojha, M., and Anderson,s. G. (1981).
J. Chem. Phys.74, 5162.
Thom, R. (1975). Structural stability and morphogenesis.W.A. Benjamin, Reading,
Massachusetts.
wiberg, K. B., Bader, R. F. W., and Lau, C. D. H. (1987).J. Am. chem.Soc. 109, 985.
 4 l

MATHEMATICAL MODELS OF STRUCTURAL

CHANGE

Next we must concedethat the universewe seeis a ceaseless creation. evolution and
destruction of forms and that the purpose of scienceis to foreseethis change of form
and, if possible,explain it.
Ren6 Thom (1975\

4.1 Introduction

The definitionof structureobtainedthroughthe applicationof the equival-

encerelationto the gradientvectorfieldsof the chargedensityleadsto the
partitioningof nuclearconfigurationspaceinto a finite numberof structural
regions.The boundariesof theseregionsdenotethe configurationsof the
unstablestructures.This informationenablesone to constructa system's
structurediagram,a diagramwhichdetailsall of its stablestructuresand all
of the unstablestructuresthat are transitionalin form betweenthose of
neighbouringstructuralregimes.The collectionof the structurallyunstable
pointsin nuclearconfigurationspaceR0,is calledthe catastrophe setC. As a
corollary of the theorem of structuralstability stated by Palis and Smale
(1970)one finds that only two typesof catastrophepoints are possible:a
configurationX e R0 is structurallyunstableif and only if its associated
chargedensityp(r; X) exhibits(a) at leastone degenerate criticalpoint, also
known as a singularity in the charge density or (b) a non-transversal
intersection of the stableand unstablemanifoldsof a pair of its criticalpoints.
A configurationsatisfyingeither(a) or (b) is calleda bifurcationor conflict
point, respectively.
This is a concisesummaryof a theoryof structurewhich,asdemonstrated
in the precedingchapters,containsthe ideas essentialto the notion of
molecularstructureand, in addition, expandsthat notion to include a
definitionof structuralstability(Baderet al. 1981;Nguyen-Dang and Bader
1982).It is the instabilityof certainstructures and the associated mechanisms
of structuralchangewhich are the subjectof this chapter.It will be demon-
stratedthat Thom's theory of elementarycatastrophes (1975)providesa
quantitativebasisfor the descriptionand predictionof the behaviourof
unstablestructures. The theoryprovidesa mathematical modelfor structural
changesin the neighbourhoodof a bifurcationpoint throughthe analysisof
what are called the universal unfoldings associatedwith singularities
of particularkinds. The possibilityof using the theory of elementary
4.2 C A T A S T R O P H ET H E O R Y 111

catastrophesto describe changes in molecular structure was first pointed

out by Collard and Hall (1977).

theory
4.2 Catastrophe

It is the purposeof catastrophetheory to accountfor and explain the

observationof sudden,discontinuous changesin the behaviourof a system
that arebroughtaboutby smoothcontinuouschanges in the variableswhich
controlit. Changeof this kind is ubiquitousthroughoutnatureandis usually
monitored through a changein form. In biologicalsystems,this is the
problem of morphogenesis. In chemistry,the structural behaviourof a
moleculeis determinedby the form of its distributionof chargein realspace;
the changesin this form as causedby nuclearmotions are conveniently
monitoredandsummarized by corresponding changes in its moleculargraph.
This is the physicalreality and the mathematicalformalismunderlyingthe
notionof the makingand breakingof chemicalbonds,processes which,aswe
now appreciate, are necessarily discontinuous.
One cannot, of course,developor even comprehensively review cata-
strophetheoryin the briefspaceaffordedthe subjectin this book.The theory
hasin anyeventbeenadmirablypresented by Postonand Stewart(1978)and
the readerdesirousof more detailis referredto this text.What we do hereis
illustratethe principalideasof the theoryand its methodof applicationto
obtainusefulresultsin the areaof molecularstructure.The knowledgeof the
theory presentedhere is sufficientto enablethe readerto make similar
applicationsand it will serveas a suitableintroductionto the subjectfor an
interested novice.To one familiarwith the method,the examplesgivenare
furtherevidence of the abilityof abstractmathematics to describeandpredict
the eventswhich occur around us. All readers will observe that the theory
appliesin a direct and natural manner to the study of changes in chemical
srructure.Thereis a behaviourspace, the realspace R3 occupied by theatoms
:n a molecule,and thereis a controlspace, nuclear configuration space Rc, as
ihe interactionsbetweenthe atomsarealtered by the relative motions of their
nuclei.

4.:.1 Isolatingthe unstablepieceof a criticalpoint

\\'e haveobservedin the discussions givenin the two precedingchaptersthat

a non-degenerate criticalpoint in the chargedistributionis stable,i.e.its form
ar summarrzed by its rank and signaturepersistsduring nuclearmotions
* hich causechangesin the chargedensity.Sucha critical point is calleda
\lorse criticalpoint. Thereis a theoremcalledthe Morse theoremwhich
iratesthat everynon-degenerate criticalpoint in an n-dimensional spacema!'
I 1 ] \ I A T H E M A T I C A L M O D E L S O F S T R U C T U R A LC H A N G E 4 . 2

be transformed,by a smoothreversiblecoordinatechange,into a Morse-/

saddlewhich is a function of the form i ,,*?, where the first n - I
t= I
coefficientsequal + 1 and the remaining coefficients equal - 1. In this
convention, when l: ft, the critical point is a local maximum. It can be
proven that a Morse critical point is structurally stable, i.e. its type is
unchanged by the addition of a small perturbation to the function in the
neighbourhood of the critical point.
We have also observed that a singularity in the charge distribution, a
degeneratecritical point, is not stable to a motion of the nuclei. It is a
corollary of the above theorem that a degeneratecritical point is always
structurally unstable. If the rank of a degeneratecritical point in n-dimen-
sional spaceis crr,then its co-rank is defined to be n - c-r.The co-rank is the
number of zero eigenvaluesof the Hessian matrix of the function at the
critical point. A most important theorem regarding the description of the
behaviour of a function in the neighbourhood of a degeneratecritical point is
the splitting theorem. This theorem isolates the degeneratecritical point of
c o - r a n kk : n - a r i n t o a M o r s e p i e c e o ft h e f o r m xf ;, with i: ltot.;o,
T(*
and a degeneratepart which is expressedas a function of the k remaining
variables,called the essentialvariables.

4.2.2 Elementarycatastrophes

Consider a bifurcation point Xo for which the chargedensity p(r; Xu) exhibits
a singularity at r" of rank ot < 3. Then, as a consequenceof the above splitting
theorem, the charge density may be expanded in a Taylor series in a
sufficiently small neighbourhood of r", in which only rr-rcomponents of r
appear up to second-order,
p ( r ;X u ): p ( r . ) + Q + f . (4.1)

In eqn (4.1),Q is a quadratic expressionin ar componentsof r with respectto a

local coordinate system centred at r", and f denotes a polynomial in the
remaining componentsof r, the essentialvariables,starting with a cubic term.
As we considerconfigurationsX I Xu, lying in a small neighbourhoodin
Ra of Xo, the topologicalpropertiesof the chargedistribution in the region of
the original singularity in p(r) vary, thereby inducing global changesin the
entire p field. New structures are generated.To discuss these structural
changesin the neighbourhoodof Xo, we needonly considerthe perturbation
in the termf of eqn (4.1)as causedby nuclear displacementsfrom the point
Xo. The configurationXo is structurallystablewith respectto the Morse part
as describedby the quadratic term Q. This term, therefore,doesnot contrib-
ute to the generation of new structuresand is omitted from any further
4 .3 IN MOLECULARSTRUCTURES
CATASTROPHES 113

consideration.Only the term / leads to changesin structureand Thom has

shown that a sufficiently small but otherwise arbitrary perturbation of f can
always be described in terms of a member of a family of functions which
containsf and which is termed the universal unfolding of I
A function/of the essentialvariables,is said to be of finite co-dimensionif a
small perturbation of/gives rise to only a finite number of topological types.
It is in this case that a map f of co-rank k can be embedded in a family of
deformations (f :Rk x Rq --+R), parametrizedby q variableswhich form the
control spaceRq c Ro. By definition, q, the dimension of the control space,is
the co-dimension of the singularity representedbyl The family of functions
(fu: Rk x Rq -- R) thus defined, describes the universal unfolding of f, in
which the function/itself is the member associatedwith the origin of control
space. In our applications the control space is a subset of the nuclear
configuration space.
The number of elementary catastrophe types dependsupon q. It has been
shown that only for valuesof q ( 5 is the number of catastrophetypes finite.
Thom has classifiedthese types by their co-rank /<and co-dimension q for
values of q ( 4. The concept of an unfolding and the accompanying defini-
tions are illustrated first in terms of the simplest of all catastrophetypes, the
so-called fold catastrophe for which both the co-rank and co-dimension
equal one.

4.3 Catastrophes in molecular structures

4.3.1 Opening a ring structure-the fold catastrophe

It was noted in the discussionof the dissociationof the water moleculethat

the creation of a bond betweenthe hydrogen atoms occurs via the formation
of a singularity in the chargedensity with a singlezero eigenvalue.That is, the
curvature of the chargedensity at the singularity formed betweenthe protons
vanishesonly along the C, symmetryaxis.This behaviouris illustratedby the
seriesof profiles of the chargedensity illustrated in Fig. 3.2.We shall consider
the reverse process,that of an oxygen approaching a hydrogen molecule,
breaking the H-H bond, and forming the open structure of the water
molecule. This is pictured by the gradient vector maps and molecular
structuresshown in the sequence(d), (c),(b) of Fig. 3.1.
The catastrophe associatedwith the breaking of the H-H bond may be
describedby a simple unfolding with one essentialvariable and one control
parameter.The essentialvariable or behaviour coordinate lies along the C,
axis which is taken to be the z-axis.The control parametercorrespondsto the
location of the systemalong the reaction coordinateand this is given by Ro,
the distance of the oxygen nucleus from the H-H midpoint. This so-called
I I 4 M A T H E M A T I C A L M O D E L S O F S T R U C T U R A LC H A N G E 4 . 3

fold catastropheis illustrated again in Fig.4.1. In this diagram the three

characteristiccurves are plotted relative to the value of p at the singularity,
i.e.at the inflexion point in curve c. The labelling of the curvescorrespondsto
that shown for the molecular graphs in Fig. 3.1.Thus, the profile labelled d is
characteristicof the ring structure. The secondarymaximum in this profile is
found at the position of the H-H bond critical point and the minimum at the
ring saddle point. As oxygen approaches Hr, these two extrema in p
approach one another in value and in position. At Ro : 2.39 au, profile c,
they merge yielding a singularity in p and a catastrophe point in configura-
tion (or control) space.This configuration definesa point on the boundary
separating the ring structure from the open structure associatedwith the
equilibrium geometry of water. For values of Ro ( 2.39 au, p exhibits a
simple monotonic decrease(profile b) indicating the loss of the H-H bond
critical point and the absenceof an H-H bond.
The analytical description of this fold catastropheis obtained by replacing
p(r) by a function of the singlebehaviour coordinate ( and parametrizedby a
single control parameter u, denoting displacementsfrom the catastrophe
point along the reaction coordinate. The topological behaviour of p(r) in the

,I
f
.t6

.12

.08

.o4

.00

-1.0 -0.5 Az

Frc. 4.1. Fold catastrophe in HrO. Eachcurveis a profileof p(r)alongthesymmetryaxis,the

z-axis,relativeto the valueat the inflexionpoint z,;.f: p(z)- p(zt)andLz : z - zr.Thevalues
of Roare2.75 au for curved,2.39au for curvec (whe.re thesingularityin p(r)occurs),
and 2.30au
tor curveb.
4 .3 I N M O L E C U L . { RS T R U C T U R E S
CATASTROPHES 115

neighbourhoodof the catastrophepoint is then describedby the expression

f ( < ) : ( l 3 ) ; ,+ r . ; . (4.2)

The behaviour variable ( contains implicitly the third derivative of p with

respect to the actual coordinate z. The germ of this unfolding is the term
Ol3)C3,which is unstable as a critical point: both the first and second
derivatives of this term are zero for ( : 0. If we perturb it through the
addition of a small term u( then its character is changed. This is seen by
settingAflA(: 0 which yields two roots for
(:*ut-u'
to Ro> 2.39au,f exhibitstwo criticalpoints,
For u < 0, corresponding
equally spaced on either side of the position of the original singularity at
; :0, thereby predictinga moleculargraph of the type shown in Fig.3.1(d)
or (e).As lul approacheszero, i.e. as the systemapproachesthe catastrophe
point at Ro : 2.39 au, the two critical points approach one another and
eventuallymergeat u :0 giving rise to the singularity in p. Beyond this point
the two critical points have vanished to yield the open structure of water.
The complete description of the catastrophesinvolved in the dissociation
of the water molecule must clearly involve two behaviour variables and at
leasttwo control parameters.We show in the following sectionthat the most
generaldescription of the formation or destruction of a three-memberedring
structure is describedby such an unfolding, the elliptic umbilic.

1.3.2 A generalanalysisof three-centresystems

The functionf, called the unfolding of the elliptic umbilic,

-(ll3)y3 + wxz- ux - u!, (4'3)
. f( x , y ; F ) : x 2 y F:(w,u,u)eM,
accountsfor the structural changesobservedfor a three-memberedsystem.
Thesechanges,by symmetry, are confined to a plane of the behaviour space.
The behaviour variables (x, y) of eqn (4.3)refer to a local coordinate system
.rn this plane.Variations in the control parametersu, u, and w within a subset
\I of Ro generateall the structuresattainable by a perturbation of the germ
.''f this unfolding defined by

f (*,Y;0): x' Y - 013)Y' (4.4)

The singularitydescribedby eqn (a.4)is exhibitedby the chargedistribution

.'rithe Hl molecular-ion in its equilibrium Drr, geometry and the unfolding of
ihis germ,as given by eqn (4.3)with w : 0, predictsall structuresof this ion
observedin the neighbourhoodof this geometry(Bader et al. 19791.
Figure 4.2 illustrates the sequenceof structures and associatedphase
ptrrtraits found in the Cr, approach of H to Hl'. The initial and final
Ftc. 4.2. Contour plots of p(r) and their corresponding gradient vector fields for the ground
state of the Hl system for configurations in the neighbourhood of the bifurcation catastrophe
point shown in diagram (b). The structuresillustrated by the molecular graphs in (a),(c),and (d).
and the corresponding structures derived from permutations of the protons, are all obtained
f r o m t h e u n s t a b l es t r u c t u r ei n t b t b r s u i t a b l ed i s o l a c e m e n t so f t h e n u c l e i .
4 .3 IN MOLECULARSTRUCTURES
CATASTROPHES n'7

structuresshown, (H-Hr)+ and (H-H-H)*, are the same as those found in

the O * Hz reaction to form H-O-H. There is. however, apparently no
intervening ring formation in the Hi reaction. The molecular graph and
phaseportrait for the equilibrium geometryof Hi is givenby Fig. 4.2(b).This
phase portrait is characteristic of the elliptic umbilic. The reader may
compare this portrait with that given by Berry and Mackley (1977)for the
streamlinesof a fluid flowing in a six-roll mill or with the form of the three-
dimensional pattern decorating the geometrical optics caustic surface ob-
tained by Berry et al. (1979\ to gain some idea of its universality. The (x, y)
coordinate system of eqn (4.3) is in the molecular plane with origin at the
singularity with the y-axis coincident with a Cz axis. The dimensionless
control parametersr,rand u are directly related to the nuclear displacement
coordinates.A displacementfrom the equilibrium configuration which pre-
servesa singleC, axis along y is denotedby (u * 0, u : 0).A motion denoted
by (u + 0, u: 0) is asymmetricand reducesthe symmetry to that of the C,
point group.
Differentiationof eqn (4.3)for w:0 yields,for the critical points inl
Afl\x:2x! - u:0, A f l A y : x 2 - Y 2- u : 0 . (4.5)

In general,for any value of a, theseequations admit two solutions (seeTable

1.1 (p. 128)).The solutions obtained for u:0 correctly predict the number
and relative positions of the critical points observed when the system is
displacedalong the u-axisin control space,i.e.displacementsof Ar symmetry.
Thus, for (u : 0, u < 0) it predicts a structure characterizedby the molecular
graph in Fig. 4.2(a),while for (a : 0, D > 0) it predicts a molecular graph
corresponding to a different structure, as displayed in Fig. 4.2@).For an
asymmetric displacementfrom the bifurcation point, u + 0, the two critical
points no longer lie along either of the behaviour axes,but are located in two
quadrants of the behaviour plane, symmetrically placed with respectto the
original singularity. The molecular graph obtained for such a displacement
correspondsto the samestructure as that obtained for a motion, D > 0, a : 0,
t.e.a structurein which one proton is bonded to the two remaining protons.
These predictions are again in agreement with the observed behaviour as
rllustratedby the phaseportrait and molecular graph given in Fig. 4.2(d).
Sincethe three protons are symmetricallyequivalentin the unique struc-
:ure found at the bifurcationpoint, an equivalentmoleculargraph,and hence
lhe samestructure,is obtainedfor an outward displacementof any one of the
protonsalong its C, axis.Thus, a motion along any one of the threesemiaxes,
,u :0 , t ' < 0) , ( u: * ,/3 r,u > 0 ), i n c o n fi g u ra ti onspaceresul tsi n a struc-
ture of the type representedby the moleculargraph in Fig. 4.2,a).Anl' other
displacementof the systemfrom the bifurcationpoint yieldsa structureof the
iorm (H-H-H)* (Fig. 4.2(c)or (d)).Thus, the unfolding of the elliptic umbilic
* ith w : 0, in the neighbourhood of the catastrophe point. leads to a
1 1 8 M A T H E M A T I C A L M O D E L S O F S T R U C T U R A LC H A N G E 4 . 3 I

partitioning of the control or configuration space of the HjF system into

regions, to each of which a definite structure can be assigned,the structure
I
diagram, Fig. 4.3(b).The three semiaxes,which originate at the catastrophe
point where rr : t): 0, definethe boundariesof three regionsof configuration
space.Within each such region only a single structure is found, of the type
(H-H-H)+. The existenceof three such regions results from the permuta-
tional symmetry of the protons. An exchangeof the form
+ --+
tH (1 F H (2 F H (3 )l [H (1 FH (3FH (2)].
can occur only by passagefrom one such region to another. To do so, the
systemmust cross one of the semiaxespartitioning configuration space,i.e. it
must passthrough the conflict structureof the type shown in Fig. a.2@).Thus
each of the semiaxesis itself a subregion of structure spacealong which the
molecular graphs are of the form [H{Hr)]*. These latter structuresare
necessarilyunstable as they exist in a space of dimension less than the full
dimension of the control space. The union of the three strata of conflict
structureswith the bifurcation point forms the catastropheset describingthe
structural changes in Hl and this union forms a closed subset of the full
configuration space.
Finally, the bifurcation catastrophepoint itself definesa third structure in
which all three protons are bonded to one another via the gradient paths
which originate at the singularity and terminate one at each nucleus (Fig.
4.2(b)).In this electron-deficientmolecule,all three nuclei are bound by the
charge accumulated in the centre of the pseudoring. The molecule is also
partitioned into three atoms by three surfacessymmetrically disposed be-
tween the bond paths and generatedby the gradient vectors which terminate
at the singularity. Thus, all three atoms are defined and linked by gradient
paths associatedwith the singularity in p. In this sense,a singularity in p
resulting from the merger of three (3, - 1) bond critical points and a ring
critical point behavesas a degeneratebond critical point.
In the more generalcase,where the control parameter il, assumesnon-zero
values, the unfolding of the elliptic umbilic predicts a more complex parti-
tioning of control spaceinto regions of different structures.In particular, the
bifurcation set, a singleton for the case w : 0, in which all three nuclei are
bonded to one another in a pseudoringstructure.expandsto definea region
of finite size within which a true ring structure exists. The bifurcation set
predicted by the unfolding of the elliptic umbilic in the (u, u, w) control space
is pictured in Fig. 4.3(a).It is seenthat, for a casein which w +0, this set in
the corresponding u, u-plane defines a hypocycloid-shaped region whose
corners are the points (u, u) : (0, + w2), ( t Ql8).,/3*', - w'l8). Thus, in the
full three-dimensional control space (u, u, w), the bifurcation set comprises
two taperedconeswhich are joined at the umbilic, the origin of the control
space.
4 .3 C A T A S T R O P H E ISN M O L E C U L A RS T R U C T U R E S 119

At any point in configuration spacelying outside the hypocycloid,at a

given w * 0, only two critical points, which are (3, - I ) or bond critical
points may exist.When applied to an A. systemthis unfolding predictsthat
the only structureswhich can exist outside the hypocycloid are of the types
A-(Ar) and (A-A-A). Inside the hypocycloid, the unfolding establishesthe
existenceof four critical points. Three of these are bond critical points, and
the fourth may be eithera (local)minimum or a (local)maximumin the scalar
field/(x, y) depending on whether w > 0 or lt' < 0, respectively.Since p is a
maximum in the (x, y) behaviour plane for any perpendicular displacement
and sincelocal maxima in p generallyoccur only at nuclear positions,we may
exclude negative values of w. The same argument shows that, for a positive
value of w, the unfolding will yield a (3, + 1) critical point in p as required to
obtain a ring structure.
An example of a triatomic system which exhibits a bonded ring structure
over a non-empty set of nuclear configuration space is the water molecule.
We now demonstratethat the elliptic umbilic unfolding eqn (4.3)can be used
to describethe formation and destruction of this ring structure.The origin of
the control plane (u : 0, u : 0, w > 0)is placedat the bifurcation catastrophe
point illustratedin Fig.3.1(c).The control coordinatesu andu denotenuclear
displacementsof B, and A, symmetry, respectively.In the symmetrical
bifurcation configuration, the singularity in p is located on the C, axis
between the protons. The defining equations of the critical points derived
from the unfolding in eqn (4.3)for w > 0 are
\flOx :2x(y+ l v )- u : 0 and A f l A y: ( x 2 - y t ) - u : 0

and, in general,they lead to a quartic equation. For some specialvalues of u

and u, however, the solutions are simple and they sufficeto give an overall
description of the dynamics of a three-atom system. These solutions are
presentedin detail by Bader et al. (1979).Here we summarize the structural
predictionsso obtained.
For a displacement (u : 0, u > 0) from the bifurcation point illustrated in
Fig. 3.1(c),the singularitybetweenthe protons is transformedinto a (3, - 1)
H-H bond critical point and a (3, + 1)critical point. The resulting molecular
graph is characteristicof a ring structural region, the region bounded by the
hy'pocycloid in Fig. 4.3(a). For the opposite displacement from the cata-
strophe point (u:0,, u < 0) corresponding to the oxygen nucleus ap-
proachingthe H, molecule,the singularity in p vanishes.The H-H bond path
rs broken and the molecular graph generatedis typical of the open structure
found for HrO in its equilibrium geometry(Fig. 3.1(a),(b)).The sametype of
singularity occursfor nuclear configurations of the two remaining sidesof the
h1'pocycloidexcept that the singularity occurs at either of the O-H bond
criticalpoints. A motion acrosseither of theseboundariesof the hy'pocy'cloid
correspondsto an opening of the ring structureby the breaking of an O-H
I 2 O M A T H E M A T I C A L M O D E L S O F S T R U C T U R A LC H A N G E 4 . 3

AV

a
I

rll
I
t
I
s3J-p2
4 .3 I N M O L E C UL A R S T R U C T U R E S
CATASTROPHES t2l

bond to yield structuresof the type O-H-H. Three open structuresare thus
found outside of the ring region. Their associatedstructural regions are
separatedfrom one another by three semiaxesdenoting the set of conflict
catastrophepoints characterizedby the structuresO (Ht) and H-(OH).
Thus the unfolding of the elliptic umbilic for positivevaluesof the control
parameter w predicts the structure diagrams for the Hi and HrO systems
and these diagrams agree with the theoretically determined behaviour of
thesetwo systems.

4.3.3 Formation of a cage structure

We have demonstratedthat Thom's theory of elementarycatastrophesfinds

a direct application in the analysis of structural instabilities which corres-
pond to the making andfor opening of a ring structure. The usefulnessof
Thom's classificationtheorem is a consequenceof the fact that all the changes
in Y p that are involved in such a process occur on a two-dimensional
submanifold of the behaviour space of the electronic coordinates. Clearly,
more complex casesof structural changesare to be expected,caseswhose
complete description will necessitatethe use of the full three-dimensional
behaviour space. Such a case is illustrated by the formation of a cage
structure.
The systemwe useto exemplifythis process,the [1.1.1]propellanemolecule
CrHu, belongs to a classof systemswhich chemically are regarded as highly
strained ring systems.In the equilibrium geometry X" of D.n symmetry, the
charge distribution of this molecule exhibits three, three-memberedrings
(Fig. a.a). Each of these rings contains the two bridgehead carbons and a.t
apical carbon atom. Becauseof the symmetry of the system,the associated
ring surfacesare planar. The molecular graph for the equilibrium structure is
shown in Fig. a.a@).
This structure of three fused rings is transformed into a cage structure by
breaking the bond between the bridgehead carbon atoms. The bridgehead
bond length of 1.59A in this moleculi is longer by - 0'06 A than a normal
C-C bond. This increased equilibrium separation has two effects on the
charge distribution: (1) the value of p(r) at the carbon-carbon bond critical
point is reducedin value from that in a normal C-C bond;(2)the (3, + 1) ring
critical points are in the immediate proximity of the carbon-carbon bond
critical point and the value of the charge density at these two points are of
almost equal value. This proximity of the (3, + 1) ring and (3, - 1) bond

Frc. 4.3. (a) A plot in the (u,u,w) control space of the bifurcation set obtained from the
(c)
unfolding of the ittiptic umbilic. The structure diagrams for Hi in (b) and for HtO in
:
.orr..po-nd to sectionsin the (u, u)-planeof the bifurcation set shown in (a) for x' 0 and rr' > 0.
resPectivelY.
1 2 2 M A T H E M A T I C A L M O D E L S O F S T R U C T U R A LC H A N G E 4 . 3

b
Frc. 4.4. (a) A molecular graph illustrating the equilibrium structure of [1.1.1]propellaneand,
in (b), a plot of p(r) in the plane of one of its three-memberedrings. In (b) and in succeeding
diagrams,the position of a (3, + l)critical point is denoted by a triangle. Note the proximity of
the (3, * 1) ring critical point to the (3. - l)bridgehead bond critical point.
 4,3 N M O L E C LL A R S T R U C T U R E S
C A T A S T R O P H EI S 123

criticalpoints portendsthe instabilityof this sl stem.An increasein the

separation between thebridgehead carbonatomsresultsin thecoalescence of
the(3, - 1)bondcriticalpointwith all threeof the(3, + 1)ringcritical points
to createa singularityin p(r) and an unstablestructure(Fig. a.5(b)).The
phaseportrait of this singularityin the planeof the apicalcarbonsis similar
I to that found for the singularityin the HJ system(Fig. 4.2(b)).The
I singularityis thus of the ellipticumbilic type.As illustratedin Fig. 4.5(c),a
further increasein the bridgeheadseparationcausesthis singularityto
bifurcate,to yieldthe elements of a cagestructure,i.e.a (3, * 3) criticalpoint
and three(3, + 1)criticalpoints.This cageis boundedby threering surfaces.
Eachof the ring surfaces containstwo apicalcarbonnuclei,aswell asthe two
bridgeheadcarbonsand henceis characterized by substantialcurvatures. In
Fig.4.5(c)these threering surfaces aredenotedby their intersections with the
oh symmetryplane.
Theidentificationof the singularityin Fig.4.5(b)asbeingoneof theelliptic
umbilic type suggeststhat the structuralchangesdescribedabovecan be
expressed in termsof the unfoldingof the elliptic umbilic catastrophe(eqn
(4.3)).In thefollowing,we shallshowthat thisequationdoesindeedprovidea
model for the transformationof the equilibrium structureinto the cage
structure.Moreover,it will be shown to correctlypredict other structural
changesresultingfrom thosedeformationsof the systemwhich preservethe
oh symmetryplane.
Thephaseportraitof the singularity foundin Fig.4.5(b)indicates that,in a
sufficiently small neighbourhood of this singularity, the charge distribution
canbe represented by a Taylor seriesexpansionof the sameform aseqn(4.1),
wherethebehaviourvariables(x, y) now referto a localcoordinatesystemfor
the on symmetryplane. (In principle,the quadratic term Q(z) must be
replacedby a more generalfunction,Q(*, y, z) which will enableone to
discuss instabilitiesin Vp with respectto generaldeformations in the system's
geometry,e.g.a deformationleadingto a totally asymmetricconfiguration.
However,for the purposeof the presentdiscussion, we shallcontinueto use
(4.
eqn 1)with a necessary change in the sign of the term 0 (z).This signchange
is to ensurethat, within the small neighbourhood considered above,the
chargedistributionincreases in directions perpendicular to the x, y symmetry
plane.)The structuralchanges which are observed in the neighbourhood of
the bifurcation catastrophe point X6, and which result from symmetry-
preserving deformations, canbedescribed in termsof the perturbations of the
functionof eqn (4,4), that is, in terms of the unfolding of eqn (4,3).
We havepreviouslynotedthat the bifurcationsetpredictedby eqn (4.3)
partitionsa givencontrolplanew * 0 into two regions.At any point (u,r')
containedin the regionwhichis boundedby the hypocycloid-shaped cross-
sectionof the bifurcationset,the function/(x, y: tt)of eqn (4.3)exhibits t*'o
saddlepointsand anothercritical point, which is a local maximum if x < 0.
1 2 4 M A T H E M A T I C A L M O D E L S O F S T R U C T U R A LC H A N G E 4 . 3

c
4 .3 I N M O L E C UL A R S T R U C T U R E S
CATASTROPHES 125

and a local minimum if w > 0. In applying eqn (4.3)to the descriptionof the
making or opening of a ring structure, we have discarded the casew < 0 on
the basisthat, in this case,the local maximum in/(x, y;p) would correspond
to an extraneouslocal maximum in the charge distribution at a location
where a (3, + 1) ring critical point is found. In the present application, this
argument does not hold, since,locally,p increasesin directions perpendicular
to the (x, y)-plane. A local maximum in f (*, y;p) now corresponds to a
(3, - 1) bond critical point, and a local minimum to a (3, * 3) cage critical
point. Likewise,a saddlepoint, i.e.a (2,0) critical point, in/corresponds to a
(3, + 1) ring critical point in the charge distribution. It is now apparent that
eqn (4.3) correctly describesthe transformation of the structure associated
with the equilibrium geometry (Fig. a.5(a)) into the cage structure (Fig.
a.5(c)).It is also seenthat the control parameter w appearing in eqn (4.3)is a
signedmeasureof the elongation of the bridgeheadinternuclear distancewith
respectto its value at Xo.
The structural model afforded by eqn (4.3)predicts the existenceof other
structuresfor the [1.1.l]propellane system.Outside the region enclosedby
the bifurcation set in Fig. 4.3(a),two structures,each consisting of two fused
rings, are predicted. The first of these structures consists of two four-
membered rings with one apical and the two bridgehead carbon atoms in
common. This arrangementof the two ring surfacesis stable and must occur
over a dense subset of this region of nuclear configuration space.There are
three equivalent forms of this structure obtained by a permutation of the
apical carbon atom which is common to both rings. The second of these
fused-ring structuresconsistsof a three- and a four-membered ring with the
bridgeheadcarbon atoms common to both. In this type of structure,of which
there are again three equivalent forms, the (3, + 1) ring critical points are
arrangedin such a manner that the stable manifold (the ring axis) of the four-
membered ring intersectsnon-transverselythe unstable manifold (the ring
surface)of the three-memberedring. According to the Palis-Smale theorem

Frc. 4.5. Gradient vector fields in the oo symmetry plane of CrHu and corresponding molecular
graphs (showingjust the carbon nuclear framework) for three different values of the internuclear
separation between the two bridgehead carbon atoms. The crossesdenote the positions of the
protons in this plane. The trajectorieswhich terminate at theseprotons are omitted. The broken
lines in the molecular graphs representprofiles of the ring surfaces.These same lines appear as
trajectoriesin the corresponding Vp(r) plots. (a) The equilibrium geometry. The (3, - l)critrcal
point at the centre of the diagram signifies the existence of a bond path linking the two
bridgehead carbon atoms. (b) This diagram is obtained when Ro. the separation between the
bridgehead carbon atoms, is increasedby 0.4 au from equilibrium value. At this geometn' the
ring critical points merge with the bond critical point to yield the singulant;- in p(r) denoted by-a
full square.This is the geometry of the bifurcation catastrophepoint and the associatedstructure
i s u n s t a b l et o a l l m o t i o n s a w a y f r o m t h i s g e o m e t r y .I f R o i s d e c r e a s e df r o m i t s r a l u e i n ( b ) . o n e
regains the structure shown in (a) with a bond path linking the bridgehead carbon. If Rn is
i n c r e a s e dt,h i s C - C b o n d p a t h v a n i s h e sa n d t h e c a g e s t r u c t u r es h o w n i n t c t r : t r b t a r n e d
(From Bader et al. (1981).)
126 4.3

l
4,3 N M O L E C U L A RS T R U C T U R E S
C A T A S T R O P H EI S T27

this arrangement is unstableand corresponds to a conflictcatastrophe point

in controlspace.A displacement of the system from such a conflictpoint will
resultin a structuralchangeto yield one of the threeequivalent forms of the
stablefused-ringstructures.
The abovepredictionsare confirmedby the structuralchangesobserved
for a displacementof an apical carbon atom from its position on a Cz
symmetryaxisin the catastrophe configurationXr. A displacement alongthis
axisin a directionawayfrom the axisof the bridgehead carbonatomsyields
the conflictstructureshownin Fig. a.6@).This structureillustratesthe non-
transversalintersectionof the stableand unstablemanifoldsof two ring
criticalpoints.A reversalof the abovedisplacement yieldsthe stablearrange-
ment of two fusedfour-memberedrings with the displacedcarbon being
common to both rings, Fig. 4.6(b).A similar structureis obtainedfor a
displacement of the apicalcarbonatom perpendicular to the Cr axisbut still
preservingthe oh symmetryplane(Fig. a.6(c)).
The above observationscan be summarizedin the structurediagram
shownin Fig. 4.7(b).The catastrophe setin this diagramis similarto that for
the HJf system.It consistsof threesemiaxes of conflictconfigurations which
meetat the bifurcationcatastrophe point Xo.Theseaxespartitionthe control
plane(w : 0) into threestructuralregions.Eachof theseregionsis associated
with one of the stable fused-ringstructures.By continuousextension,
structurediagramsfor any plane w * 0 can be constructed.Two such
diagrams,one for w > 0, the other for w < 0 are shownin Fig. 4.7(a) and
(c),respectively.
The precedingdiscussions illustratethe simpleway in whichthe unfolding
of the ellipticumbilic singularity(eqn(4.3))accountsfor structuralchanges
which accompanyspecialsymmetry-preserving deformations.It is to be
realizedthat the model affordedby this equationworks only becausethe
portion of the behaviourspacein whichthesestructuralchanges takeplaceis
of dimension two. This is a consequence of the preservation of the o6
symmetry plane in all the distortions studied so far.
Difficultiesare encountered as more generaldeformationsare considered,
deformations of the [1.1.l]propellane moleculewhichrequirethe useof the
full three-dimensional behaviour space. At the presenttime, Thom'sclassi-
fication theoremdoes not cover situations which involve more than two

Frc. 4.6. Thesediagramsrepresentthree more structuresof CrHu which are obtainedfor

motionsawayfrom theconfiguration of theunstablestructureshownin Fig.4.5(b). Thestructure
in (a)is obtainedby an outwarddisplacement of the topmostapicalcarbonatom.The resulting
structureis itselfunstable,The structurein (b) is obtainedby an inwardsdisplacement oi the
samecarbonatom and the structurein (c)resultsfrom the displacement of this carbonfrom its
symmetryaxis.
(From Baderet al. (1981).)
1 2 8 M A T H E M A T I C A L M O D E L S O F S T R U C T U R A LC H A N G E

A A A
,A "2\

w<o W=O w>o

a b c

Ftc. 4.'1. Cross-sectionsof the structure diagram for CrHu for that portion of its control or
nuclear configuration spacein which the plane of the apical carbon atoms is a symmetry plane.
The structure at the origin of the control space is that given in Fig. 4.5(b) for the bifurction
catastrophe point. An increase or decrease in w corresponds to an increase or decrease,
respectively, in the separation between the bridgehead carbon atoms.
(From Bader et al. (1981).)

Table4.1
Critical pointsin p(r)for H { predictedby unfoldingof etlipticumbilic

Values of control parameters (3, - 1) criticalpoints

Number Location

w:0,rt: oftu,<'0o 2 x:0,.): +\/ - u

x: *,/r,y:0
w:0,u*0 X: l r/r, ! : ul2x
* lu + (u' * u21rrz1tt'

behaviour variables. Possible ways of overcoming these difficulties are

discussedin Bader et al. (1981).

References

Bader,R. F. W., Nguyen-Dang, T. T., and Tal, Y. (1979).J. Chem.Phys.70, 4316.

Bader,R. F. W., Nguyen-Dang, T. T., and Tal, Y. (1981).Rep.Prog.Phys.44,893.
Berry,M. V. and Mackley,M. R. (1977).Phil. Trans.r. Soc.,LondonA287,l.
Berry,M. V., Nye,J. F.,and Wright,F. J.(1979).Phil.Trans.r. Soc.,LondonA291,454.
 REFERE\CES r29

Collard, K. and Hall, G. G. (1977).Int. J. quantum Chem' 12' 623'

Nguyen-Dang,T. T. and Bader, R. F. W. (1982).Phvsicall4A. 68.
Palis,J. and Smale,S. (1970).Pure Math' 14.223'
poston, T. and Stewart, I. (1978).Catasnophe theort' and its applications.Pitman,
London.
W.A. Benjamin, Reading,
Thom, R. (1975). Structural stability and morphogenesis.
Massachusetts.

l
 )

THE QUANTUM ATOM

There can be no doubt but that in quantum mechanicsone has the complete solution
to the problems of chemistry.
G . N . L e w i s( 1 9 3 3 )

5.1 Chemistryand quantummechanics

5.1.1 From Lewisto quantummechanical

models

Lewisintroducedthe conceptof the electronpair into chemistryin 1916.In

this paper(1916)he wasableto rationalizethe knowngeometries of carbon
compoundsby replacingthe prevailingnotion of a cubicalarrangement of an
octet of electronsby a tetrahedralarrangementof four pairs of electrons.
What particularlyservedto laterconvinceLewis(1933)of thepotentialof the
new quantum mechanics, as evidencedby the openingquotation,was its
ability to account for observedmoleculargeometriesand so provide a
theoreticalbasisfor his electronpair model.The early developmentof the
valencebond theoryof directedvalenceby Slater(1931)and Pauling(1931,
1960)providedthe first quantum mechanicalrationalizationof molecular
geometry.The valencebond approximationto the wave function,as ex-
pressedin termsof the overlapof hybridizedatomicorbitalsto describethe
pairingof electrons,coupledwith the notion of electronspin,wasthe natural
theoreticalextensionof the pair conceptof Lewis.The hybrid orbitalswere
spatiallylocalizedin directionscoincidentor nearlycoincidentwith observed
moleculargeometriesand this gavetheoreticalsubstanceto the notion of
picturinga bond asa spatiallylocalizedpair qf electrons. Pauling'semphasis
of the useof ionic-covalent resonance structuresto accountfor the differing
characterof bondedinteractionswasin turn a restatement of Lewis'sideato
representthe bonding betweena pair of atoms in terms of tautomeric
extremesof electron-sharing possibilities.
It wasa principalclaim of Lewis's
first paper(1916)that'Accordingto thetheorywhichI am now presenting, it
is not necessaryto considerthe two extremetypesof chemicalcombination,
correspondingto the very polar and the very nonpolar compounds,as
differentin kind, but only as differentin degree.'
In the descriptionof natureaffordedby quantummechanics, oneclassifies
andcharacterizes the stateof a total systemin termsof theeigenvalues of a set
of commutingobservables actingon an elementof the Hilbert space,the state
vector.Molecularorbital theoryin its canonicalrepresentation as originally
 5.1 CHEMISTRYAND QUANTUM MECHANICS 131

developedby Mulliken (I928a-c, 1932a.1935)and Hund (1933,1937)was,

and is, the theory of electonic stucture for the prediction and classificationof
quantum statesof many-electronsystems.Molecular orbital theory provides
a model of the electronicstructure of a many-electronsystemin terms of a set
of coupled one-electronstatesor orbitals which greatly facilitatesthe predic-
tion and classificationof the electronic states of many-electron atoms and
molecules.In a molecular system,the one-electronstatesthemselvescan be
classifiedas bonding, non-bonding, or antibonding (Mulliken 1932b,1939)
and this classificationcorrelateswith the observedchangesin bond lengths
and dissociation energiesof the new statesgeneratedas a consequenceof a
changein the occupation of a particular orbital. The delocalizednatureof the
canonicalset of molecular orbitals provided a usefulmodel for understanding
i

the properties of unsaturated systems (Coulson and Longuet-Higgins

1947a,b,I948a-c), but was unsatisfactory in providing models of directed
bonds. Lennard-Jones(1949a,b,1952)definedequivalent orbitals (which later
evolved into localized orbitals) to provide the molecular orbital equivalent of
directed valence orbitals and it was only after their introduction that
molecular orbital theory found widespread application to structural prob-
lems in chemistry. Since the advent of suitable computational methods and
machinesin the 1960s.molecular orbitals calculatedwithin the self-consistent
field approximation and expanded in terms of a finite set of basis functions
(Roothaan 1951;Hall 1951)have provided the principal means of obtaining
approximate solutions to the wave functions of many-electron systems of
chemical interest.
The orbital model has, however,been extendedbeyond its intended use of
predicting and providing an understanding of the electronic structure of a
system, by associating the forms of individual orbitals with the assumed
spatially localized pairs of bonded or non-bonded electrons, and by at-
tempting to defineatomic propertiesin terms of coefficientsof atomic-centred
basis functions appearing in the expansion of molecular orbitals. These
steps are admittedly arbitrary, as are attempts to define atoms through a
partitioning of the Hamiltonian operator. This latter step violates the
indistinguishability of the electronsfrom the outset.
One might imagine that, with the advent of quantum mechanicsand its
application to chemistry,Dalton's atomic theory would have been reinforced.
This has not happened.Quantum mechanicshas been shown to account for
the properties of isolated atoms and for the total properties of a molecular
system.The increasedunderstandingthat would result from the discovery of
a firm theoretical basisfor Dalton's theory has not been obtained becauseof
the lack of a quantum definition of an atom in a molecule.This is not to say
that the concepts of atoms and bonds do not appear in the quantum
mechanical treatments of chemical systems.They do, but in the reverse
manner to that describedabove.Rather than finding its quantum basis.the
t32 THE QUANTUM ATOM 5.1

atomic and bond conceptsare built into an approximatetheory to model a

real system. Thus, in valence bond theory a molecular wave function is
approximated in terms of products of atomic wave functions in appropriate
'atoms in moleculesapproach'to the calcu-
valencestates.Moffitt's (1951)
lation of molecular binding energiesis a further example,as is the successof
using atomic functions as a basis for the expansion of molecular orbitals.
Properties of an ionic crystal strongly suggest that it is best regarded as
a collection of interacting ions. Hence, one may successfullyapproximate
the wave functions for such systems using free-ion wave functions with
adjustments for the overlap of non-orthogonal functions as illustrated by
the work of Lowdin (1948, 1956)in an early theoreticalcalculation of the
properties of ionic crystals.
'The
In 1929 Dirac worte: underlying physical laws necessaryfor the
mathematical theory of . . . the whole of chemistry are thus completely
known, and the difficulty is only that the exact application of theselaws leads
to equations much too complicated to be soluble.' The stumbling block to
obtaining useful approximate solutions to the equations of quantum mech-
anicsfor systemsof chemicalinteresthas been largely overcome.Overcoming
this hurdle, however, has not in itself led to a mathematical theory of the
whole of chemistry,at least not to one which is expressiblein the languageof
chemistry. Given a state function, or a good approximation to it, how does
one obtain from it a description of a system'spropertiesin terms of its atoms
and its bonds which, from a chemist'spoint of view, summarizesin a concise
manner its important chemical properties? The varying successesof the
quantum mechanicalmodels of atoms and bonds have servedto demonstrate
the soundnessof the atomic concept but, by building in the idea of atoms,
they have not furthered Dalton's theory of atoms in molecules.The increased
understandingthat would result from the discoveryof a firm theoreticalbasis
for Dalton's theory requires a quantum definition of an atom in a molecule.

5.1.2 The role of the chargedensityin definingstructure

Atoms and bonds have meaningin real spaceand are a reflectionof the
structurepresentin realspace.This structureis not reflectedin the properties
Hilbert spaceof the statefunction.Instead,the
of the infinite-dimensional
physicalbasisfor molecularstructureshouldresidein the quantummechan-
ical functionwhichprovideria descriptionof a systemasit existsin realspace.
distributionof chargeasdefinedin eqn(1.3)for a
This functionis a system's
stationarystateor as definednow for the generaltime-dependent case,

p(r,X,r) : l{I(spins)J
{l],.,drr}Y*(x,X, r)Y(x,X, r) (51
.)
 5.2 QUANTUM DEFINITION OF A\ ATOM 133

solution to the generaltime-depend-

whereY is a properlyantisymmetrrzed
ent Schrodingerequation,
ih0Y l\t : frV and - i ha Y * i r : H Y * (s.2)
As in Chapter 1, x denotes the collection of electronic space and spin
coordinates,X the nuclear coordinates,and r the spacecoordinates(x, y, z) of
a singleelectronand dzr: dxrdy;dz,.It should be noted that the electronic
I

chargedensity contains the information neededto determinethe distribution

of nuclear charge as well, and thus it determines the total distribution of
charge. When r : Xo, a nuclear position coordinate, p exhibits a cusp and
through the cusp condition, eqn (82.2),it determinesthe nuclear chargeZo.
In addition, the positions of the nuclei are evident in the topology of the
electronicchargedensity,as each nuclear cusp behaveslike a local maximum
in the charge distribution with p(X,) x 23"12.
Molecular orbital theory has played the central role in the definition and
understanding of problems of electronic structure. The charge density plays
the corresponding role in the definition and understanding of the concepts
associatedwith molecular structure. The previous chaptershave shown that
atoms, bonds, and structure are indeed consequencesof the dominant
topological property exhibited by a molecular charge distribution. What
remains to be done is to demonstrate that the topological atom and its
properties have a basis in quantum mechanics.

5.2 Need for a quantum definition of an atom

5.2.1 Observationalbasisfor a quantum atom

A theory is only justified by its ability to account for observedbehaviour. It is

important, therefore,to note that the theory of atoms in moleculesis a result
of observationsmade on the properties of the charge density.These obserYa-
tions give rise to the realization that a quantum mechanical description of the
properties of the topological atom is not only possiblebut is also necessary,
for the observations are explicable only if the virial theorem applies to an
atom in a molecule.The original observationsare among the most important
of the properties exhibited by the atoms of theory (Bader and Beddall 1972).
For this reasonand for the purpose of emphasizingthe observationalbasisof
the theory, theseoriginal observationsare now summarized.They provide an
introduction to the consequencesof a quantum mechanicaldescription of an
atom in a molecule.
Figure 5.1 displays,in the form of contour plots, the ground-statecharge
distributions of LiF, LiO, and LiH, each at its equilibrium internuclear
separation.Superimposedon each of these plots is the intersectionof the
134 T H E Q U A N T U I v TA T O M 5.2

LiH
Ftc. 5.1. Contour plots of the ground-state molecular charge distribution of LiF, LiO, and
LiH. The intersectionof the interatomic surfacewith the plane shown in the diagram is indicated
b v a d a s h e dl i n e .
5.2 QUANTUM DEFINITION OF AN ATOM 135

zero-flux interatomic surface,as defined in eqn (2.9),with the plane of the

diagram. The distribution of charge within the Li atom so defined is seento
be remarkably similar in all three of these molecules in spite of the very
differentnaturesof the neighbouringatom. The net chargeof the Li atom (as
determined by an integration of p(r) over the atomic basin to obtain its
average number of electrons followed by its subtraction from the nuclear
charge) is very nearly the same for all three molecules (seeTable 5.1), the
observedvariation following the trend anticipated on the basis of decreasing
electronegativityof the ligand F > O > H. This near constancyin the charge
distribution of Li correctly reflectsthe properties assignedto Li on the basis
of the chemistry observedfor compounds in which Li is bonded to a more
electronegativeelement-those corresponding to a relatively small, singly-
chargedpositive ion with a tightly bound distribution of electroniccharge.
Figure 5.1 illustrates an elementary but important and general observa-
tion: the choice of the zero-flux surfacefor defining an atom maximizes the
possibility of assigningan atomic identity to a given mononuclear region of a
charge distribution. It is clear that any other choice of a partitioning surface
would either include a portion of the neighbouring atom that is very different
in all three cases,or omit a portion of the Li atom chargedensity that changes
by only small amounts through the seriesof molecules.Sincethe partitioning
must exhaust the spaceof a system,the latter possibility would assignthe
very similar omitted portions to the very different neighbouring atoms.
Coupled with the observationof near constancyin the chargedistributions
of the Li atoms in this seriesof moleculeswas the further observationthat the
kinetic energy density distributions exhibited a corresponding degree of
constancy.(As demonstrated later, while there is no unique definition of a
kinetic energy density, all definitions lead to the same averagevalue for the
kinetic energy when integrated over the basin of an atom up to its surface of
zero flux. The averagekinetic energy of an atom is a well-definedquantity as

Table5.1
Somepropertiesof boundLi atoms*

Molecule Net charge Average electronic

of Li atom kinetic energyof Li
atom (au)

LiF(Xrt*) + 0.937 7.354

Lio ( X 2n ) + 0.932 7 .3 5 6
L i H( X '> * ) + 0.913 7.368

*From Bader and Beddall (1972);calculated from state functions

close to the Hartree-Fock limit.
136 T H E Q L J A \ T L ' \ , IA T O M 5.2

a consequence of the zero-fluxsurfacecondition.)Thus,it is an observationin

this set of moleculesand subsequentlyin others, that a constancy in the
distribution of chargefor a topologicallydefinedatom leadsto a correspond-
ing degreeof constancyin the averagekinetic energy of the atom, the total
spreadin valuesfor the Li atoms in the presentexamplesbeing * 32 kJ/mol
( s eeT able 5. 1 ).
According to the virial theorem,the total energyfor a systemwith inverse
squareforcesis equal to minus the averagekinetic energy.If one postulates
the existenceof an atomic statementof the virial theorem, then the above
observationpredictsthat, when the chargedistribution of an atom is identical
in two different systems.the atom will contribute identical amounts to the
total energiesin both systems.Equally importantis the more generalconclu-
sion that the properties of such a topologically defined atom are directly
determinedby its chargedistribution,the propertieschangingin direct response
to changesin the charge density of the atom.
The postulation of an atomic virial theorem for the topologically defined
atoms leadsto a number of important conclusions(Baderand Beddall 1972).

1. The total energyof a moleculeis expressibleas a sum of atomic energies.

2. The averagepotential energy of an atom is defined as the averageof the

virial of the forcesexertedon it-as demonstratedlater, this is the only
non-arbitrary way of partitioning potential energiesof interaction be-
tween systems.

3. A relationshipmust exist betweenthe distribution of chargeand the virial

of the total force exertedon each elementof the chargedensity,the virial
field. The form of the atom must be independent of the individual
contributionsto the forcesexertedon it, sincetheseforceschangeradically
betweenany pair of systems.The chargedensityof an atom must respond
instead only to the sum of the local forces it experiences. The virials of
the individual contributions to the forcesexertedon the density of the Li
ion by the hydride ion differ by thousandsof kJ/mol from those exerted
by the fluoride ion (the one-electronpotential energiesdiffer by l4 x 103
kJ/mol, for example)but the net force in each casecorrespondsto that
emanating from a polarized singly-chargednegative ion. The constancy
in the chargedistribution of Li must be viewed as remarkablewhen one
contemplatesthesevery differentcontributions to the force exertedon it
by its neighbours.One field, however,changesby little-the virial of the
total Ehrenfestforce exertedon the electrons.This relationshipbetween
p and the virial field is the basis of the essentialobservationthat, if the
distribution of charge for an atom is identical in two different systems,
then the atom will contributeidenticalamountsto the total energyin both
systems.
5.2 QUANTUM DEFINITIO\ OF \\ ATOvt 131

It will be shown that eachatom in a sl'stemmakesan additivecontribution

to the averagevalue of everysystemproperti. This is the principleunderlying
the cornerstoneof chemistry-that atoms and functionalgroupingsof atoms
make recognizablecontributions to the total properties of a system. In
practice, we recognize a group and predict its effect upon the static and
reactivepropertiesof a systemin terms of a set of propertiesassignedto the
group. In those limiting situationswherein a group is essentiallythe samein
iwo differentsystems,one obtains a so-calledadditivity schemefor the total
properties,for in this casethe atomic contributions,as well as being additive
over each system,are transferablebetweenthe systems.
Examples are given later of near perfect transferability of groupings of
atoms, caseswhere the group energieschange by lessthan 4 kJ/mol. It is at
this limit that one can determine the properties of atoms in molecules
experimentally,as corresponding additive contributions to heats of forma-
tion. Such transferability is, of course,the exception rather than the rule but
theselimiting examplesprovide the touchstonefor obtaining an understand-
in
ing of the properties of the chemical atom-that these properties change
in its form in real space. When
direct responseto the extent of the changes
the atom
this form remains constant, so too do the contributions which
makes to the properties of the total system in which it is found'
in
It is upon the strength of these observations that the theory of atoms
moleculesrests,for they form the basisfor the identification of the topological
demon-
atom with the chemical atom of a molecular system.This chapter
stratesthat the postulatedexistenceof the atomic virial theorem is confirmed,
as is the complete quantum description of the properties of the topological
requires
atom. The development of the quantum mechanicsof a subsystem
of the derivation of
careful considerationof the properties of observablesand
of the quantum
their equations of motion. These fundamental aspects
the
description of a total system are considerednext, with an emphasis on
changesanticipatedin the descriptionof a subsystem.

5.2.2 Observablesand their properties for a total system

be known
It is a postulate of quantum mechanicsthat everything that can
The value of a physical
about a systemis containedin the statefunction Y.
property is obtained from the state function through the action of a corres-
ponding operator on Y. Thus quantum mechanicsis concerned with obser-
whose action on Y
vables,the name given to the linear,Hermitian operators
character of an
yields the values of a system'sproperties.Tle Hermitian
which intro-
tperator is illustrated for the Hamiltonian fr, an illustration
its role in
ducesthe quantum mechanicalcurrent density and demonstrates
of operators' If
the description of the properties of subsystemaverages
138 THE QUANTUM ATOM 5.2

Y* Ydr is to representa probability, then the norm, ly', of the state function
must remain constantin time. Using Schrodinger'sequation and its complex
conjugate(eqn (5.2)),one has

alyllat: y*\P + \P*y: - (ilh){y*(FY)- (r?Y)*y} (5.3)

where \P : dY/dr. Integration of eqn (5.3)over all configurationspaceyields

d t' tl d t : - (i l h )J - (H Y )* Y } dr. (5.4)

{ Y * (H Y )
In order that the norm remain constantin time it is necessaryand sufficient
that
j v*1riv)dr: J(r?v;.Y dr. (s.5)
An operator with the property exhibitedin eqn (5.5)issaid to be Hermitian
if it satisfiesthis equation for all functions Y defined in the function spacein
which the operator is defined.The mathematical requirementfor Hermiticity
of H expressedin eqn (5.5)placesa correspondingphysical requirementon
the system-that there be a zero flux in the vector current through the surface
S bounding the system.To illustrate this and other propertiesof the total
systemwe shall assume,without loss of generality,a form for H correspond-
ing to a single particle moving under the influence of a scalar potential Z(r)

fr : - ft212m)Y2
+ t$) : pt12^+ t1ry. (s.6)
It is worthwhile mentioning at this point that all propertiesof a subsystem
definedin real space,including its energy,necessarilyrequire the definition of
correspondingthree-dimensionaldensitydistribution functions.Thus, all the
properties of an atom in a molecule are determinedby averagesover effective
single-particledensitiesor'dressedoperators'and the one-electronpicture is
an appropriate one.
The potential energy operator in I? is a real quantity which does not
involve derivativesand, in this case,eqn (5.5)can be rewritten as

J{v*(v' Y)- (v2v;'r,v}dr

: o. (s.7)
The integral in eqn (5.7),likemany to come,can be transformed into a surface
integral using the three-dimensionalanalogueof integration by parts. In one-
dimension,one has

I @*a' Qlax\dx: (O*aLlai l;i - J@Q*

l0x)(aQl0x)dx (s.8)
where x, and x2 representthe limits of integration.In the extensionof this
result to three dimensions,the contributions arising from these limits com-
bine to yield an integral over the surface S bounding the system. When
applied to the complex conjugate term in eqn (5.7) as well, one obtains
Green's theorem
- $Ys\.n
J{o.v' o- ov' f*}dr: fds(d*vd (s.e)
 5.2 QUANTUM DEFINITIO\ OF A\ ATOM 139

where n is the vector normal to the surfaceS. Use of eqn (5.9)to re-express
eqn (5.7)yields
- ( v ' v ) * Y } d r: $ d s ( Y * v Y- Y V Y * ) n
'
J{v*v2y
I
: 1 2 mf it t $ djs' n (5.10)
I
I where the quantum mechanical vector current density j is defined as
I - YVY*).
i: (hl2mixY*vY (5.1 )
in termsof the Hamiltonianoperatorto yield
Eqn (5.10)can be expressed
- -ihfidsj.n. (5.r2a)
JY*(fgdr J(tv)-Ydz:
For a bound system,Y is squareintegrableand,hence,Y and its derivatives
vanishon all elements dS : dSn of the surfacewhenthe surfaceis removedto
infinity. Thus the right-hand sidgof eqn (5.10)vanishesfor the total system
with boundaries at infinity and H is Hermitian.However,if the integrationis
limited to a subsystem Q bounded by a surfaceS(C)),part or all of which
occursfor finite values of the integrationvariables,then one has

- ttivt-Ydr
*-1r?9az : -.{ds(o)i n (s.12b)
f f
whereintegrationover a subsystem is indicatedby the subscriptQ on the
integralsign.In this casethe flux in the currentdensitythroughthe surface
will not, in general,vanishand one cannot assumethat the Hamiltonian
integralis equalto its Hermitianconjugateas is true for the total system.
Finally we note that eqn (5.3)can be re-expressed as

d p l d t* V ' j : g , (5.13)

law for a fluid.In its integratedform

which is a statementof the conservation
t r
dl{(a)/dr: lJ ov . i d r : - QJo s 1 o 1 ; ' n (s.14)

wherel{(O), the averagenumberof electronsin O, equalsthe integralof p

over the regionO. This law statesthat the changein the averagenumberof
particlesin the volumeO is givenby the flux in the vectorcurrentdensity
throughits boundingsurfaceS(O).The final equalitygivenin eqn (5.14)
resultsfrom the useof Gauss's theoremwhichstatesthat the volumeintegral
of the divergence of a vectoris equalto the flux in the vectorthroughthe
surface bounding the system.The divergenceof a vectorandGauss's theorem
in
are reviewed Section E5.3.
Dependingon the nature of the statefunction,someobservables y'ield
'sharp'values-Y is an eigenfunction of the observable-whileotherst'ield
140 T H E Q U A N T U I UA T O M 5.2

'averagevalues'.If two observables,,{ and B commute, i.e.

only

lA,nl: (AB- BA): o, (5.1s)

then they possessa common, complete set of orthonormal eigenfunctions.
For a system in a stationary state for which the state function is an
eigenfunction of the Hamiltonian operator,
HrL : E(t and fr,lt* : E,lt* (s.16)
all of the observableswhich form a pairwise commuting set with H yield a set
of sharp values and this set of numbers is used to classify the state of the
system.The ground state of the water molecule,in addition to possessinga
sharp value for the energy,is labelled rA, as a consequence of the spin and
symmetry operators forming such a commuting set with H. The average
value of an observable that is not a member of this commuting set of
observablesis given by

<A>: 1v,.4Y>I <Y,Y): Jv*,4vorlJY*Ydz (5.17)

and, of course, for a system in a stationary state, the sharp value for a
commuting observableequals its averagevalue.
The equation of motion for the averagevalue of an observableA canbe
obtained directly from Heisenberg'sequation for,4(r) (Messiah1958,p. 319).
The result is

d < A > l d t :( i l h ) < v , WA, l y ) + p A l a t S . (s.18)

Equation (5.18)can also be obtainedby differentiatingthe expressionfor the
averagevalue of ,i (eqn (5.17))assumingY is normahzedto unity,
a q A S 6 t :J { ( a y l*r 4 A v + y * A @ yl 0 t )+ Y * p A l a q v } a r ,
followed by the use of Schrodinger'sequations(5.2),and subsequentlytaking
into account the Hermiticity of the Hamiltonian operator,

a Q S l a t : \ l h ) { ( r i v ,A v >- ( v , A n v ) } + ( 0 A l d t )
: (ilh)<Y,(frA- Aulvl + pAlatS
: (ilh)(Y, [t, A]v ) + <aAI at> . (5.1e)
For an observable^6which commuteswith the Hamiltonian and which does
not possessan explicit time dependenceone has
d ( B ) , , ' d :r 0 (s.20)
and its value is independentof time. In analogy with classicalmechanicsthe
observableF is called a constant of the motion. None of the observables
consideredhere have an explicit time dependenceand the final term on the
r ight - hands id e o f e q n s (5 .1 8 )o r (5 .1 9 )w i l l b e omi tted from thi s poi nt on.
5.2 QUANTUM DEFINITIO\ OF{\ {TOM t4r

With Schrodinger'sequation to describehou' the state function changes

with time and Heisenberg'sequation to determinethe correspondingchange
in the average value of each observable. one has a complete mechanical
description of a system.Heisenberg'sequation is important from another
point of view, as it can be used to obtain relationsh]Psof great general
importance for particular choices of the observable A. Examples are the
Ehrenfesttheoremsgoverning the time rate of changeof the averagevaluesof
- ihY, and the virial
an electronicposition coordinate i and momentum 0 :
-- t'p. These theorems play important
theorem which is obtained when A
roles in the mechanical description of an atom in a molecule.
With ,i set equal to the product of the massand the position coordinate for
an electron, the commutator is
mlfr, if : (hli)( - thv): (hl1\fr (s.2r)
relationsis obtained,
and the first of Ehrenfest's
md(i)/dr : (0). (s.22)
For ,A: p, the commutator
is
tf, 0l : thYt$) (s.23)
and the second Ehrenfest relation is obtained
d(0)/dr : ( - vtlrS) : (F(t)) (s.24)
where,in analogy with classicaldefinitions, the force operator F(! is defined
will be
as the negative of the gradient of the potential. The operator F(r)
force. Equations (5.22) and (5.24) are both quantum
termed the Ehrenfest
hold
analoguesof classicalrelations but, in the quantum case,the relations
values of the mechanical properties and not in a
only ior the average
pointwise manner for each specifictime as they do on a classical trajectory.
The two expressionscan be combined to yield the analogue of Newton's
equation of motion
md2(i) ldt2: <Fttl>. (s.2s)
The left-hand side of eqns (5.22) and (5.24) may be explicitly evaluated
ti' and v*, and expressedin terms of
using Schrodinger'sequations (5.2)for
the quantum vector current density,eqn (5.11).For the time derivativeof (i)
one has
d { Y * i Y } l d t: ( i l h {) ( r ? v * ) i Y - Y * r ( H Y ) i .
d { Y *i Y } l d t : ( - i h l 2 m{)i ( V ' Y * ) V- i Y * V ' Y I
: _ iv.j.
In general,the notation AB will be used to denote a multiplication t-rft$t-t
r..i o. , t o y ield a d y a d i c . A re v i e w o f th e p ro p e rti esof dl adi cs i s sr\en rn
142 THE QUA\TU\1 .{TOM 5.2

Section E5.2.In particular, the following identity holds

v ( j i ): j + i V . j
and hence

d(i)/dr : Jja' - Jv'(ji) dr,

which,usingGauss'stheoremfollowedby multiplicationby *, becomes
m d ( ? ) l d t : m [ j d t - m $d s ( n . j ) i (5.26a)
The surfaceterm vanishesfor a systemwith boundariesat infinity and one
obtainsthe resultthat the time derivativeof the averagepositionvectoris
givenby the integralof the currentdensity,

nrd(i) ldt: mljdr: +{(0y, y) + (y, pv>} : G). (5.26b)

Thecurrenthasthedimensions of a velocityandthus,whenmultipliedby the
mass,it equalsa momentum.Similarly,the time derivativeof a velocityis
an acceleration
and henceone shouldobtain

d ( 0 )l d t : m ! { a j @ l 6 t } d t : ( F ( r ) ) . (s.27)
The derivationof eqn(5.27)alsoentailsthe vanishingof a surfaceterm,one
involvingthe quantumstresstensor.The stresstensorwasfirstintroducedin
its relativisticform in 1927by Schrodinger and its propertieswere later
discussed by Pauli(1933). The derivationof eqn(5.27)followsthat givenby
P a u l i( 1 9 5 8 ) .

m o j l a:t + { H y * ( v y )- v * v ( f v ) + ( v y * ) r ? y- V ( r ? y * ) v } ,
m 0 j l A t: ( h 2l 4 m ){ ( - V 2 V * ) V y + y * V ( V 2 v )
- ( v Y * ) V ' Y * V ( V 2 y * ) Y )+ l l 2 { V ( V Y * ) Y
- Y ( t v * ) Y+ t v * v Y - y * v ( v y ) \ ,
majlat: - (h2 l4m){(vrv*)vv- y*v(v2v)
I ( v Y * ) v z y t- v ( v r Y * ) Y ) - Y V * V / .
Definethe stresstensorF as

{Y*V(VV) + V(VY*)Y - VV*VY - VyVy*}

F : (h214,?r) (5.28)
for the time derivativeof j may be expressed
and the expression as
mAjl7t: +V'A -Y*YVV (5.29\
Hence,
: * (- v Z ) v d t+ f d S F ' . n
m ! { a j @ l e r } dJzw
 .\TO\' l
5 .2 QUANTUM DEFINITIO\ OF .\\ 143

where Gauss'stheorem has been used to transform the volume integral of

V.F into an integral over the surface.an integralrrhich ranishesfor a system
with infinite boundariesto yield the equationslisted under (5.27).While the
current density does not appear in the description of the properties of a
stationarystatein the absenceof a magneticfield. it and the quantum stress
tensor F play important roles in the expressionsobtained for the average
valuesof subspacepropertiesand they are introduced here for that reason.
One further theorem of paramount importance obtainable from
Heisenberg'sequation is the virial theorem as obtained from the operator
i 'p . In t his c as et he c o m m u ta to ri s

( i l h ) l H , i ' 0 :1 ( - h ' l * ) y z- r ' Y t : 2 i - r ' Y t

where f denotes the kinetic energy operator. Hence,

d ( i ' O ) l d t : 2 < i >+ ( - i ' v v > . (5.30)

The usual statement of the virial theorem for a stationary state is

2<i>+(-i.v/):0. (s.31)

In the original derivation of the classicalvirial theorem given by Clausius,an

expressioncorrespondingto eqn (5.30)is also obtained.In the classicalcase
one argues that the time averageof d(i'frildt vanishesover a sufficiently
long period of time or that the motion is periodic to obtain the equivalent
of eqn(5.31).
Alternatively, the virial theorem may be derived directly from eqn (5.29),
which expressesthe forces in terms of the time derivative of the current
density. Dotting the vector r into eqn (5.29)to obtain the virial of these
forces gives
m r ' { l i l A t }: * r'V 'A - Y* Y(r 'V V).
using the identity
The term r'V' F may be re-expressed
V ' ( r ' F ): T r F '+ r ' V ' F '
where TrF denotesthe trace of the tensor F,
T r 3 : ( h 2l 4 m ){ Y * V 2 Y + ( V 2 Y * ) Y
- v Y * . v Y- v Y ' v Y * 1 . (5.32)

Each term in eqn (5.32)gives minus one-half of the averagekinetic energy

when integrated over all space.As demonstratedlater in eqn (5.47).the
averagesof thesetwo ways of expressingthe kinetic energyin eqn (5.32)differ
by a surfaceintegral and one has
- JTt Fdr : 2<i> + (h2l4m)$ds vp'n
144 T H E Q U A N T U N , IA T O M 5.2

Useof theseresultsvields
m [ r .{ a j l Ad} r : 2 < i ) + ( - r . v v S
+ ( h ' l a 0 #d S V p .+n f d S r . F . n
where the final surface term comes from applying Gauss's theorem to the
integral of V'(r' F'). The two surface integrals vanish for a system with
boundariesat infinity. Hence the final result,a statementof the virial theorem
for the general time-dependentcase,

d(i.0) ldt: mft.{f'jlar}dr : 2qi) + ( - i . v v > . (5.33)

The virial of a system, denoted by the symbol {, is defined to be the
average
{ - ( - r ' v v ( r ) ): ( r ' F ( r ) ) .
It is the averageof the virial of the forcesacting on the particleswhich possess
the kinetic energy ( f). When F(r) is the Ehrenfest force acting on the
electrons,t is the potential energy of the electrons.In those caseswhere t,
the potential energyoperator of the system,is a homogeneousfunction of the
coordinates of degree n, one can, by using Euler's theorem, re-expressthe
virial in terms of the averageof the potential energy operator
y. - - n(t$>
and, consequently,for a stationary state one has

2<i>: ,<t). (5.34)

For an isolated atom, the potential energyoperator for the Coulombic forces
is a homogeneousfunction of degree - 1 and in this case one has

2 q i S :- < v > .
For a system in a stationary state, the equation of motion for the mean
value of A, eqn (5.18),assumesthe form

AD : o.
<Ln,
Alternatively,this result can be derived directly by making use of the
Hermitianpropertyof fi and Schrcidinger's equationfor a stationarystate
(eqn(5.16)),
(frA- AnV) : (fi,|t,A,t,>
<lu,Al) : (,1,, - <,!,,
An,!>
A(/>- <r1,,
: E\<r|,, A(/>) : 0. (5.35)
If I{ contains a real parameter s such as a nuclear charge or a nuclear
coordinate and .4 correspondsto the operator Al0s, then the derivation of
eqn (5.35)must be modified to include the term 0El0s,for E contains the
5.2 QI]ANTUM DEFINITION OF AN ATOM 145

same parametersas does H. The result in this caseis

<lH,Al) - (lE, Al) : o

and, when the commutators are evaluated, one obtains the Hellmann-
Feynman theorem
(l/,@Hli s)/):6ElEs. (s.36)
When s is a nuclearcoordinateXo, eqn (5.36)givesthe Hellmann-Feynman
electrostatictheorem
( 1 , , (- V , t ) r l t >: - V" E : F " . (5.37)

Equation (5.35)and the generalizedstatementof the Hellmann-Feynman

theorem can also be derived through a variation of the state function (Epstein
1974b). In this case eqn (5.35) is known as the hyperuirial theorem
(Hirschfelder1960).The variational derivationsare important becausethey
demonstratethat the theoremshold not only for the exact statefunction but
also for approximationsto thesefunctionswhen the latter functions,such as
Hartree-Fock, unrestricted Hartree-Fock, and multiconfigurational
Hartree-Fock, are invariant to the variational parameters.The original such
proof of the Hellmann Feynman theorem was given by Hurley (1954).
It is clearfrom the useof the Hermiticity of H in the derivationof eqn (5.35)
that the hypervirial theorem will assumea different form for a subsystem.A
derivation of the hypervirial theoremfor a subsystemis given at this point to
illustrate the consequenceof the loss of the Hermitian property of H. The
result is generaland appliesto any subsystemregardlessof its definition.To
ensurethat the subsystemaverageof an observableAbe a real number, one
must work with the mean of the averagevalue and its complexconjugate(cc),

;{J"{ , * l H , A f r l td r + i ([F, Alrl,).f o'\ ::

Jo
{(tf, Al)" + cc}.
(5.38)
Workingwith just the first of the averages,
one obtains
Alrlt>": (t, nA,tt)o - |fr11,,
<,1t,\fr, A,lt)"
+ < H , !A, , L n- < , 1At ,W > " (s.3e)
wherethe term |frr1t, Arlt)"hasbeenboth addedand subtractedto maximize
thecorrespondence with eqn(5.35). Makinguseof Schrodinger'sequationas
beforeto get rid of the last two termson the right-handside,
r - {t*V'1Ag}at (5.40)
f i ,l r l , > " : ( h l22 m ) {(:V',1,*)Arlt
<,1,,]A
J,,
where,as in the derivation of eqn (5. t ) . the terms involvins t have cancelled
146 THE QUANTU\I {TO\,I 5.3

out.Usingeqn(5.9)with 6* : {t* andQ : Ar/. gives

I
< r l t , l f rA, ) , l t > n : - i / r 9d S ( A ) j r ' n (5.41)
where jr, the current density for an observablel, is defined in a manner
analogousto that for the vector current j itself,

{{/*v(A,lt)- (v ,lt\4,1,}.
io : (hl2m1) (s.42)
The final statement of the subsystemhypervirial theorem is

+{(tfri,AD" * cc}: - +{1h{ + cc},

ds(o)jr.n (5.43)
a result first obtained by Epstein (1970; (seealso Srebrenik and Bader 1974).
By performing correspondingsubstitutionson the subsystemanalogueof
eqn (5.19),the expressionfor the time derivativeof a subsystemexpectation
value is

+ d { ( Y ,A Y ) n+ < A Y , V ) , r } / d r
, ] ) " + c c ]- + { $d s ( o ) j , . +
: + { ( i l h ) ( [ fA n cc}
+ itf ds(o)(ds/ar)Y*.4v
+ cc). (s.44)
The final term accounts for the dependenceof the surface S bounding the
region O on the time.
The importance of Heisenberg'sequation and of the correspondinghyper-
virial theorem for a stationary state in the description of the properties of the
total system is maintained in the description of a subsystem.Indeed, the
generalized action principle that is employed to establish the quantum
mechanics of a subsystem can be expressedin the form of a variational
statementof the Heisenbergequation of motion. For a stationary state,the
sameprinciple reducesto a variation of the energy and the resulting theorem
is a variational statement of the hypervirial theorem applicable to both the
total system and a subsystem.It is to be emphasized,however, that the
variational derivations of these statementsas obtained through the action
principle apply only to a particular classof subsystems, those bounded by a
surface of zero flux in the gradient vector of the charge density the
definition of a topologicalatom. The useof thesetheoremsin the description
of a subsystemleadsto novel resultsbecauseof the generalnon-vanishingof
the surfaceintegral in eqns (5.43) and (5,44),

5.3 Need for a subsystemvariation principle

An understandingof chemistry requiresa regional descriptionof a system.

The notion that a moleculecan be viewedas a collectionof atoms linked by a
network of bonds, a notion that has already been shown to be rooted in the
topologicalpropertiesof the chargedistribution,is the operationalprinciple
5.3 N E E D F O R A S U B S Y S T E I \U' { R I , \ T I O \ P R I N C I P L E t4l

underlying our classificationand understandingof chemical behaviour. It

would appear,therefore,that to find chemistrl * ithin the framework of
quantum mechanicsone rnust determinethe r aluesof observablesfor pieces
of a total system,i.e.for a subsystem.But hou is one to choosethesepieces?Is
'atoms'and its
therebut one,or are theremany ways of dividing a systeminto
propertiesinto correspondingatomic contributions?If there is an answerto
this problem, then the necessaryinformation must be containedin the state
function, for Y tells us everythingwe can know about a system.
lt is clear from the examplesgiven in the preceding section that one can
write down expressionsfor the expectation values of observablesfor a
subsystem with an arbitrarily defined boundary. One has no guarantee,
however, that the results obtained for such an arbitrarily defined region of
spacewill be of physical significance.Indeed, one can show very simply and
as one important example that the kinetic energy of an arbitrarily defined
region of spaceis not well-defined.
Consider the exPression

V 2 1 Y * Y ) :V ' { ( V Y * ) Y+ Y * ( V Y ) }
: (v2Y* )Y + Y* (V'Y) + 2VY*'VY. (s.4s)
Thus, in the general many-electron case one has

l4m)L{Y*V?v + YViY*} : (h'lZm1l,

- th2 V,Y*'V'Y
- (h'ladL,vi (Y*Y). (s.46)
Sinceelectronsare indistinguishableand Y is antisymmetrized,the averageof
a sum of N one-electronoperators can be replacedby l{ times the averageof
one of the operators. If the sum of the operators is replaced by l{ times a
singleoperator in each term in the above equation and is then summed over
of all
a11the ipin coordinates and integrated over the space coordinates
(1'4)),the
electrons but one (operations denoted by the symbol Jdt', seeeqn
result is
- (h'lam)t{ : (h'l2m)l{
Jdr'{v*V2Y+ YV2Y*tJ Jdz'vY*'vY
- th2l+dl{ Jdz'v2(Y* Y). (5.47)
(1)(r, r') (seeeqn
This is readily restatedin terms of the one-densitymatrix, f
(E1.1)),and its trace, the chargedensity,p(r) (eqn (1'4))
( t )( r ' r ' ) 1 , = , '
- ( h zl 4 m ) { V ' + V ' 2 }f t t r ( r ,r ' ) l , - - , ,: ( h 2l 2 m ) ( V ' V ' ) f
- (h' l 4m\Y 2P (rl . (5' 48)

respectivel)"
This equation or, equivalently,eqn (5.41)can be used to define'
the quantities (Bader and Preston 1969)
K(r):G(r)+L(r)' t5-19t
148 T H E Q U A N T U N {A T O M 5.3

K(r) and G(r)are kinetic energydensitiesand L1r;,a function of the Laplacian

of the chargedensity.Integration of the final coordinater in eqn (5.49)over a
region of spaceQ yields
r t
-(h2r4m) V.Vp(r)dr.l
I xlrydr: I ctrtOr |
Jo Jn Jo
(5.s0)
,K(c)): G(Q)- (h2l4m){ds(Q)vp(r)'n(r)
: G(o) + l.(C))
where Gauss's theorem has been used to replace the final integral on the
right-hand side with an integral over the surfacebounding the region Q. The
surfaceintegral is seento be a measureof the flux in the gradient vectors of
the charge density through the surface bounding the region. From the
discussionof the gradient vector field of the chargedensity given in Chapter
3, it is clear that the surfaceintegral will not vanish for an arbitrary surface.
There are two casesfor which the surfaceintegral will always vanish yielding
an equality in the kinetic energyexpectationvalues:(1) when the integral is
taken over all spaceand one has the result first pointed out by Schrodinger.
that the averagekinetic energy T may be calculatedusing either of the above
forms for the kinetic energy operator,

<i>:T:K:G, ( 5 . 5)1
and (2) when the region Q is bounded by a surfaceof zero flux in the gradient
vector of the chargedensity (eqn (2.9)and repeatedhere as eqn (5.52)),

v p(r)'n(r) : g for all r on the surface S(r). (s.s2

)
Thus, the kinetic energy of the topologically defined atom 7.(C))is a well-
defined quantity and one has

f(O) : K(O) : G(Q) if Q is bounded by a zero-fluxsurface (5.53)

in correspondence with the total systemresult,eqn (5.51).The expressionfor

the kinetic energy
0 lzm)(frrlt, > : (h2I 2m)(v/' V(/>
fr',/,
is the quantum analogueof the classicalexpressionlpl2 l2m. It correspondsto
the use of the first term on the right-hand side of eqn (5.46)or the kinetic
energy density G(r) in eqn (5.49)in the many-electron case.
Cohen (1979)has shown that all kinetic energy densities,when defined in
terms of the generalizationsmade possible through the use of the Wigner
distribution function, yield identical average values when integrated over a
region boundedby a surfaceof zero flux, as in eqn (5.52).It is possibleto have
a surface through which the net integrated flux is zero, the flux being of
opposite sign in different regions of the surface.Such a situation will not, in
general,persistfor changesin the system'ssurface,whether the changeis real.
as causedby nuclear displacements. or virtual, as causedby a mathematical
I

I 5.3 N E E D F O R A S U B S Y S T E M\ . . { R I{ T I O \ PRI\CIPLE 149

variation of the surfaceelements.A surfaceof zero flux in the chargedensity is

maintainedat all times through real motions and through virtual changesas
brought about by a variation of the chargedensitl. It will be demonstrated
later that the virial, as well as the kinetic energ\'.is not uniquelydefinedfor a
subsystemwith an arbitrary boundary when either eqn (5.43)or (5.44)is
evaluatedfor the operator i'f. Thus, the averagetotal energyof the electrons
in a region of spacewith arbitrary boundariesis not defined.These results
have a related and most important consequence:the possible energy func-
tionals, whose variation yields Schrodinger'sequation for a total system
exhibit differentpropertieswith respectto the variations when the functionals
are evaluated over regions with arbitrary boundaries. As demonstrated
below, a unique variational result is obtained only for a subsystemsatisfying
e q n (5 . 52) .
It was statedin the introductory chapterthat the question'arethere atoms
in molecules' is equivalent to asking two equally necessaryquestions of
quantum mechanics:(1) does the state function contain the necessaryin-
formation to predict a unique partitioning of a moleculeinto subsystems,and
(2) does quantum mechanics provide a complete description of the sub-
systemsso defined?The only method at one'sdisposalto answerquestionsof
this nature are extrema principles based upon a variational principle.
Schrodinger (1926) used such a method in his original derivation of the
equation which bears his name. In relatively recent times, two new but
equivalentformulationsof quantum mechanics,one by Feynman (1948),the
other by Schwinger(1951),have appeared,both of them basedupon a single
principle, the quantum analogue of the principle of least action, or more
simply the action principle. The action principle in its generalizedform as
developed by Schwinger offers an alternative variational approach to the
laws of physics,opening the door to new questions and new answersoand
herein lies the power of the method. We shall use it to answer the question
'are
there atoms in molecules?'
For a system in a stationary state, the variation of the action integral
reducesto the variation of an energy integral which is identical to that first
constructedby Schrodingerand used by him to derive the wave equation for
a stationary state.We shall introduce this subject in the following section by
reviewing Schrodinger'sderivation of his wave equation and showing how
this derivation can be generalized to yield a quantum description of a
particular class of quantum subsystems-the topological atom (Srebrenik
and Bader 1975;Bader 1988).This demonstrationis sufficientto provide an
operational understanding of the quantum mechanics of an atom in a
molecule.
Chapter 8 givesa full accountof Schwinger'sprinciple of stationarl'action
and of how, through its generalization,one obtains a definition of a quantum
subsystemand a description of its properties.
 150 THE QUANTUM ATOM 5.3

5.3.1 Schrodinger'sderivation of wave mechanics

Schrodingerconcludeshis first paper (1926)by stating that'the function ry'be

s uc h as t o m a k e th e ' H a m i l to n i n te g ra l '

[d r{ h ' T @ , Agl A q ) + (l ' V }

stationary while fulfilling the normalizing, accessorycondition

I,lt' dt : t)
The quantityT(q,Arlrl}dis the kineticenergyexpressed
in termsof the
momentum as in the first term on the right-hand side of eqn (5.47)that is, as
li'll2*, and V is the potential energy.Schrodingerconsideredspecificallythe
problem of the hydrogen atom for which the explicit form of the functional is

y bl/l: Jdr{(h'lLm)v,lt.vrlt
+ v{/{/} (5.s4)
with V - - e'lr. In his first paper Schrodinger did not allow {t to be
complex. The problem is to find the function {/ such that y l,ltl attains its
minimum value subjectto the constraint that tlt remainnomalized to unity. In
ordinary extrema problems a function is a maximum or a minimum at a
single point in the spaceof its variables and theseproblems are handled by
the methods of differential calculus.In contrast, t [/] is an integral and its
stationarity requires that its value be a minimum with respect to the
averaging of the integrand over all points in configuration space.This is a
problem requiring the calculus of variations.
The integral / [/] is extremizedin the sensethat any arbitrary change in
the function r/(r) from its true value causesno first-order changein the value
of the integral. One can think of varying the integral by assuming the
existenceof the 'correct function r/(r)' and then changing the integral by
changing rlt(r)at each point r into the 'trial function d(r)' by adding to it the
small amount 4(r)

Q$): rlt(r)+ q(r) (s.55)

where ry(r) is assumed to vanish at the boundaries of the system but is
otherwise arbitrary. This varied expressionfor each value of the coordinate r
is substituted into the integral to yield y lfl and all terms of order r12and
higher are discarded. The differencebetween ylfl and t[/] is the first-
order differencein / and is labelled6/l|f, the variation of l.since{t(r)
makes/ lrltl a minimum,6Y [r/] must vanish.Feynman et al. (1964,Section
19.1)givea usefuland engagingintroduction to the calculusof variationsin a
derivation of Newton's equation by a minimizatron of the classicalaction
integral.
We shall allow for the possibility of r/ being complex in our review of
Schrodinger's work. The functions r! and rlt* are treated as independent
 5.3 N E E D F O R A S U B S Y S T E MV A R I A T I O N P R I N C I P L E 15I

variablesand are varied separately.Proceedingwith the variation of ry'as

outlined above, one obtains

d / l r l , l : J t f t 2 l 2 n t ) V , ! * ' Y r+
y V(t*A\dr.
To this rnust be added the effectof the constraint on the normalization of r/
which is written as

!,lt"rl,dr-1:0.
Substitution of the trial function for r! yields
- I : 0
I rlt-,l,dt * [r!*,tdt

i or
(lt,rt):0. (s.s6)
Eqn (5.56)is multiplied by a factor )r and then added to the variation of y.
Requiring the resultant variation, which is label\ed 69 [,/], to be stationary
yields

6Elrl,): 6/ l,l,)+ A<{/,4) : 0

6El,Ll: ! {ft' l2m)v(t*'Yrt }dt : 0.

+ (V + A),1t"4 (5.57)

This method of handling the constraint is known as Lagrange'smethod of

undeterminedmultipliers.One may introduce the constraint in this manner
from the very beginning by defining a new functional 9lltf as

Elrl,l: I bl,l + 1<{t,rlt>

: I {(n'lZm)V{/*'Yr! + (V + i)rlt*rlt}
dr; (5.58)

the variation of Eblll then yields the expression6ElLl.

We desire an expressionfor 6Elrltl in which no derivativesof 4 appear so
that the complete integrand is multiplied by 4(r), the small arbitrary change
we make in /(r). This is easilydone using the equivalentof an integration by
parts. One notes that

v ' ( v { t " r i : Y 2 { t * 4+ Y l t * ' Y 4 or v,lt*'Y4 : Y'(vt*d - Y',lt*4

and, using Gauss'stheoremto transform the volume integralof V'(Vrl*4), we

obtain

\Elrltl: j { - ( h 2l 2 m ) V ' , l t * + ( V + A ) V * } q d t
: 0.
+ (h2l2m)fdsvr2*'n4 (5.59)
The surfaceintesral is over the boundary of the system as found for r : co
t52 THE QUANTUM ATOM 5.3

and 4 is required to vanish on this boundary. This yields

\Elrltl: J{ - @2 + (V - E){/*}rydt
l2m)V',1,*
: [{nrL* - Erlt*}r7dz: 0 (s.60)
wherethe constantl, hasbeenidentifiedwith - E, the negativeof the total
energy.This resultis to be true for all arbitraryfunctions4(r),including,for
example, the casewherertt) :0 for all r but onearbitrarilychosennarrow
rangeof values.The only way in which the integralin eqn (5.60)can vanish
for all suchvariationsryis for the quantity multipliedby 4 to vanish.This
yields
fr,lt* - Et* :0. (5.61a)

A similarvariationof l/* yieldsthe complexconjugateof eqn (5.61)

Hll - ErL:0. (5.61b)
Equation (5.61a or b) is called the Euler-Lagrange or simply the Euler
equation of the variation and, as first demonstratedby Schrodingerin 1926,
variation of t [/] subject to the normalization constraint yields the wave
equation. Becauseof the equivalenceof the two kinetic energydensitieswhen
integratedover all space(eqn (5.5l)) y lrl/f : E and thus the function r/ which
satisfies the wave equation also minimizes the energy of the system.
Schrodingerdemonstratedthat the'quantum conditions'then postulatedfor
the hydrogen atom could be replaced by this variation principle since the
resulting Euler equation yields the quantum numbers r?and / in a natural
way.
One may obtain eqn (5.61aor b) as the Euler equation in the variation of
Ebl without imposing any prescribed boundary conditions on 4(r), the
changeor variation of /(r). This is accomplishedby introducing'the natural
boundary conditions' (Courant and Hilbert 1953).The necessarycondition
for 9 [/] to be stationary is that its first variation 69 [/] as given in eqn (5.59)
vanish. If Eb|l is stationary with respect to variations which do not have
prescribedboundary values,then it is certainly stationary with respectto the
smaller class of variations for which 4(r) vanishes on the boundary which
implies Euler's equation. Thus we need consider only that part of the
variation which depends on the boundary, a step which yields the'natural
boundary conditions',

V{t*'n:0 or V/'n:0 f o r a l l p o i n t so n t h e b o u n d a r y . ( 5 . 6 2 )

This is the necessaryboundary condition to obtain Schrodinger'sequation as

the Euler equation in the variation of E blt). For a bound system,one imposes
the further condition that the state functions rlr and ry'*themselvesvanish on
the boundary at infinity. This is a further, necessaryrestriction on the state
5.3 N E E D F O R A S U B S Y S T E MV A R I A T I O N P R I N C I P L E 153

function if, as stated in the Born postulate,rl,*,lrdr is to representa pro-

bability distribution.This probability must necessarilyvanish on the bound-
aries infinitely far from the attractor which binds the system.
The normalization requirement on rft can be handled differently by using
the functional,
SbLl : J{ (h2l2m)Vrl,*'Y,lt I (,!,,1,>.
+ V{r*,lt}dt (5.63)

A variation of I [r/] through a variation of r/ as given in eqn (5.55)yields

through first-order rn n,

+ (V - Slrl,lrl,*rt}dtl(,lt,rl,>.(5.64)
6Sl,ttl: J{ ft2l2m)Vrl,*'Yq
Theexpansion + <(/,4\- 1 followed
of ((rl,,rlt> by thatfor (1+ (t1,4)l
<{/,{t))-t up to first-order in 11have been used to obtain eqn (5.64).Once
again, ridding the expression of Y4 and demanding that 4 vanish at the
boundaries of the system yields
6Sl,ltl: [ {fi,|t*- Sl,l,]t*}rtdtl(,lt,rl,):0. (5.65)
Since S lrlrl : E at the point of variation, one again obtains the wave
equation as the Euler equation.
One can put the variation problem in a form where the usual ideas of
differential calculus can be used to obtain an extremum in some function.
Simply label all possiblechangesin r/ with a parameter a, such that for some
value of e, say d : 0, the varied function will coincide with the function which
extremizesthe integral. One possible parametric form, for example, is

{t(r, a) : {t(r,O) + ary(r) (5.66)

where,as before,a(r) is any arbitrary function which vanishesat infinity. If we

denote bV f the integrand of some integral llrlt) to be extremized

f : f (lt'Yt),
then the parametrizatron given in eqn (5.66)gives IbL) as a function of a'
Ilt, al -- If tt,tol,Yr!
(a))dr.
The condition for an extremum is now given by the usual condition from
differentialcalculus that (aI lAd)o:o: 0' To obtain the extremum condition
we multiply by dryand evaluateall derivativesat c:0,
do : I{@fla,D(a{tl)a)"du
@Ila$,=o
do)dt : 0.
+ (Aflavl\(Avrltli@o (s.67)
If one makes the following identifications at a : 0,

@I lA$oda : 61, (At lla)odu : 5t, (AYtltlAa)6da : 6Yrlt,

t54 T H E Q U A N T U l v {A T O M 5.3

then one can write down the generalexpressionfor the variation of Ilt!, a] as

6I : [ {(dfI d,l't6{t + GfI ev{/)6v(t } dr. (s.68)

Correspondingly, the expressionfor a trial function is amended to read

d ( . ): / ( r ) + 6 t F ) . (5.6e)
To proceed beyond eqn (5.68)the procedure is always the same:one rids the
expressionof dVr/ (or 6r! for a time-dependentsystem)using an integration
by parts to transform the integrand into a quantity multiplied only by 6rl.
Setting this quantity equal to zero yields the Euler equation.

5.3.2 The variational definition of a subsystemand its properties

It will be demonstrated that the generalizationof Schrodinger'svariation of

the functional9 [/] to a subsystemhas two important consequences: (1) The
variation of the subsystemenergy functional Eblt, Ol yields the hypervirial
theorem for a subsvstem.

6Eb!,Ql : ( - '12)
{(ilh)<\fr,cl )n * cc} (5.70)
where e denotes an infinitesimal and the action of ed on V causes the
variation in r/. Equation (5.70)is the stationary-stateanalogueof Schwinger's
principle of stationary action and it will be shown to be a variational
statement of the hypervirial theorem; (2) the generalizationrepresentedby
eqn (5.70)ispossibleonly if the subsystemis bounded by a surfaceof zero flux
in the gradient vectors of the charge density-the topological atom
(Srebrenikand Bader 1975;Bader 1988).
The extension of Schrodinger's energy functional to the many-electron
case,including theLagrange multiplier A is (in analogy with eqn (5.58))

LV,/* .Y,t + 0 + 0,1,.,1,}

Ebltl: Jdr{(h' l2m) (5.71)
where,for brevity, the symbol Jdr is used here to denote a sum over the spin
coordinatesand an integration over the spatial coordinatesof all lf electrons.
The definition of the corresponding functional for a subsystemQ is

Elll,Ql : i or, for'{(h,l2m)LV,/* .y,,1, (5.72)

+ 0 + ).),1,*,1,}
Jo J
where 2denotes the full many-electronpotential energyoperator. The reader
is reminded that the symbol Jdr' implies a summation over all spins and the
integration over the spatial coordinatesof all electronsbut one. The symbol
J.,dtr implies that the coordinatesof electron'1'are integrated over the
subsystem Q. Since r/ is antisymmetrized with respect to the electronic
coordinates,all electronsare given equivalentdescriptionsand it mattersnot
5.3 NEED FOR A SUBSYSTEMVARIATION PRINCIPLE 155

which set of electroniccoordinatesis integratedover C),just as it mattersnot

which set of electroniccoordinatesis chosento define the chargedensity in
eqn (5.1).There is in this procedureno partitioning of kinetic or potential
energyoperatorsinto setsfor differentregions,a step which does violate the
indistinguishabilityof the electrons.All electronsoccupy the total spaceof
the system,as the limits on the electroniccoordinatesare t .ti for each
degreeof freedom.The symbol <rl,,rl,)r, is correspondinglydefined as

: j or,
<4,,{,)n (s.7
3)
!or'r1,.r1,.
The trial functions / representingvariations rn tJtare given by eqn (5.69)
and substitutionof d(r) for r/(r) rnto I [/, O] yieldsglQ,Ol. At the point of
variation,6 : It andg [d, O] equalsEbL, Al. The variationsbrltand 6rlt* are
not givenprescribedvalueson any of the boundaries,including the boundary
'n,
of the subsystem.Instead only the natural boundary condition, that Y ,V
and V,r/*.n, togetherwith $ and ry'*,vanish on all infinite boundaries,will be
invoked. The functional ElO,Al is to be varied not only with respectto @.
however,but also with respectto the surfacedefining the subsystemQ. Only
by having the surfaceitselfconsideredto be a function of @can the definition
of the subsystembe determined entirely in a non-arbitrary way by the
variational procedure.
The generalexpressionfor the variation of an integralincluding a variation
of its surfacefor variations with respectto r/ is given by
\t nl

U"
6,t, l N/,V / ) d rf : I {(a.f
) Ja
+ (eflev(,)dv/}dr
le,l')6,1,

+ ottn,t)fg/,vrl)drs(',
r). (s.74)
f
where drS denotesa variation of the surfacethrough variations in r/. The
i'ariation of the surfaceappearsin an integral of / over the surfaceand this
term contributes to the first-order change in the functional (Courant and
Hi l b e r t 1953) .
No particular difficulties arise in passing from the one- to the many-
electroncasein the variation of Elrlt,Ol. Referringto eqns (5.72)and (5.74),
o n e h as

6gbll,Ql: J' dt, I dr'{@flArlt)6,1t

+ I' Gf IAVtDAV,'/l
r,) + cc.
+ $ dS(Q,rr1! dr'.f1rlt,Vrl,)ds(Q, (5.75)
In this casef(lt,Vt) =/refers to the integrandof eqn (5.72).Only the integral
over the surfaceof the subsystemO survivesbecauseof the vanishingof r/ and
ry'*on the boundariesat infinity. One again usesan integrationby parts based
t 56 THE QUANTUMATOM 5.3

on t he ident it y
V ,' (V ,{ t* 6 l t) : V ? rl ,* 6 1*1,V ,/* ' 5Y ,rl ,

to rid the expressionof the terms 6Y,rl,: V i6l/. Each of the surfaceintegrals
obtainedin this manner vanishesexceptfor the integralover the surfaceof O.
This step is illustratedin eqn (5.76)which details the variation for one such
term.

0,, r,ovtndvit]
J [0,',,

.ll0,,
'Y,t*
: ( h2l2m )
Io, .6Y,rlt
J o',Io''','t'*6't'
: - th21Zm1

+ ( h 2l 2 m ) 0,,
f Jo,, f |* n,6,t,(s.76)
or(r,)Jo,,o,
The surfaceintegral in eqn (5.76)comes from the application of Gauss's
t heor em t o t h e te rm i n v o l v i n g V ,' (V ,rl ,* 611,).Abefore,
s al l such surface
integrals vanish except for r, : rr becauseof the vanishing of Y,t* on the
boundariesat infinity. From this point on, the coordinater, and the volume
elementdr, will be set equal to r and dr, respectively, and V, and Vf to their
correspondingunscriptedquantities.
T he ex pr ess i o nfo r th e v a ri a ti o n o f 9 l $ ,Ql i ncl udi ng a vari ati on of the
surfaceS (O. r) is. therefore.

6El!, Ol : i' a'Ja'' {rUrlt.+ 1(/*}6ll/

+ osro,
f
,,IO, *.
{(h' I 2m)v rlt n(r)drl

01,V/)] + cc
+ dd,S(A,r)f (s.77)
where

f r : - ( h 2l 2 4 L v i + t . (5.78)
If eqn (5.77)is to be obtainedfor any variations6rltand 6lt* and any regionO.
it appliest o t h e c a s ew h e re f): R 3 , i .e . th e total system.In thi s case,al l
surfaceterms vanish and I [l, Ol and 69 [/, C)] are identical with Illtl and
6ELrlt),respectively, and the Euler equations obtained from the variation are
Schrodinger'sequations

Url,: nl, and fr,!* : nrlt* (s.79)

5.3 N E E D F O R A S U B S Y S T E MV A R I A T I O N P R I N C I P L E 157

with f defrnedas in eqn (5.78)and wherel. : - E. Using eqns(5.79)the

for 6Elrl,,Q] reducesat the point of variationto
expression

6Ebl/,Ql: f ds (C), n(r)6/

r)Jdr'{(h' l2m)v,1,*'
* cc.
+ d,S(Q,r),f(r/.Vr/)) (s.80)

Further progress towards obtaining a general physical result can be made

only by removal of the term involving the variation of the surface of the
subsystem.Consider, towards this goal, an alternative expression for the
integrandf (:t,Yt), one involving the Hamiltonian operator H.Including the
complex conjugate term in eqn (5.80),the integrand f (rlt,V,l,)appears twice
and, using eqn (5.46)which relates two forms of the kinetic energy density,
one has
2f (rl,,Y,lt):{rtt.Hrt,+ (frr/)*rlt}- 2Erl,"rlt
+ 2(h2I 4*)L,V?(,lt*r14. (5.81)

equations
SinceSchrodinger's areassumed to apply,eqn(5.81) to.
simplifies
2f(,l,,Ylt):2(h2
l4m)LVi U,*rl,). (5.82)
Integration of the right-hand side of eqn (5.82)in the manner indicated in eqn
(5.80)transforms it into an integral of the Laplacian of the charge density. A
typical term in this integration can be transformed using Gauss'stheorem to
yield

$ ds(O,r)Jdr, Jdt,Yl(rlt*(t).. . Jdr"

: $ d S ( Qr ),! d t , . . . $ d S ( r ' ) V ' ( {, )/ *' " . . . J d t ' , (5.8
3)

and each such term vanishesfor r, # r sinceY,rl,* and V,r/ vanish along with
the
fu and /* on the infinite boundaries. Thus only the term involving
coordinate r survivesthe integration over the sum of operators Vl. Defining
p'(r) as the chargedensity per electron or p(r)lN,
p,(\\: Idt,,l,*rlt, (5.84)

one obtains for the multiple integrationof the sum of terms appearingin
eqn (5.82)the result
2(h2l4m)$dS(O,r)Y2p'(r).

Dividing this contributionequallybetweenthe appropriateterm appearing

explicitlyin eqn (5.S0)and its complexconjugateyields
5El!,Ql : f ds (o, ) (h'l2m){ J dz'vrl*'n(r)d/
+ +d,rs(Q,r)Yz p'(r)) + cc. (5.85)
158 THE QUANTUM ATOM 5.3

Further progressin obtaining a general principle from this expressionfor

6EbL,A] is not possiblefor an arbitrary region of space.We now demon-
strate that this result is transformed into the atomic statement of the
hypervirial theorem when the subsystemO is restrictedto one which satisfies
a particular variational constraint. We shall introduce the variational con-
straint in terms of a trial function @ and show how the constraint, which
limits the variations to a particular classof subsystems,can be applied from
the beginning of the variation process,beforethe point 6 : t is attained.The
following conditions are to be fulfilled in the constrained variation:

A region O(d) is defined in terms of the trial function @that is bound by a

zero-flux surfacein Y p'^

Y p'o!)'n(r): g for all r in surface S (Q, r) (s.86)

where the trial density is defined as

p ' 6 F:) I d r ' Q *( x ,r ' ) $ ( x , r ' l ' . (s.87)

Recalling that the topological definition of an atom implies the zero-flux
surfacecondition, a region of spacebounded by a surfacesatisfyingeqn (5.86)
at the point of variation is henceforthcalled an atom. It is required that, as @
tends to rlt,O(il is continuously deformable into the region O(/) associated
with the atom. The region O(d) thus representsthe atom in the varied total
system,which is describedby the trial function /, just as O(/) representsthe
atom when the total system is in the state describedby t.
Requiring the fulfilment of these conditions amounts to imposing the
variational constraint that the divergenceof Y p, integrates to zero at all
stagesof the variation, i.e. that

L ( 5 , Q 1:

for all admissible @,which implies

- 1 h 21 4 m ' 1
I",,, V2p' o(r)dr: 0 (5.88a)

d L ( { r ,:od){ - o't+^ll Y'p'(r)or}:0. (5.88b)

( Jo )
Recalling the expressionfor the variation of an integral which includes a
variation of its surface as given in eqn (5.74) with / : Y2p, the variational
constraint given in eqn (5.88)leads immediately to a result enabling one to
eliminate the term involving the variation of the surfacewhich appearsin the
general expressionfor the variation of E l,lt, Ol. From eqn (5.74)one has
( f ) r r
a , J I v ' p ' ( r ) df r: l a , { y 2 p ' ( r l } d $r +O s ( e , r ) d / s ( o , r l y 2 p ' ( r ) .
(Jo ) J J
 5.3 N E E D F O R A S U B S Y S T E M\ ' { R I { T I O \ PRINCIPLE 159
I
I a n d apply ing t he c on s tra i n tth a t 6 L (l r,O) :0 to thi s resul tyi el ds
I
f r '
( h 2l 4 m \ $ a S a * s ( o , r ) Y 2 p ' ( r ) : - ( h ' a n n d , ,[ V t p ' ( r ) \ y d r . ( 5 . 8 9 )
I
J Jn
The variation of Y2p' is easilydeterminedusing the recipegiven in eqn (5.68).
Re-expressingY2p'in terms of r/ and ry'*,one has

Ytp' : ! d|Vz(rlr*r/):Jdr'{(V'V*)l/ + 2Vrl/*'Y,1,+ ,lt*V'rlt} (5.90)

Hence the variation is with respectto ,lt,Vrlt,andY2tlt with the result

dr:
,- {Y'p'(r)} a, a,t,
J i" [a,'{v',t,*
+ 2V{/*' 6V,lt+ {/* 6V' {l }. (s.e1)
The Laplacian of p is what is termed a divergenceexpression-its variation
-'-ieldsonly surfaceterms. The integral in eqn (5.91)can be expressedas the
divergenceof a sum of terms

f f r
' \ v r l , * d+, l {t / w 6 ( / \ ,
I a , |, Y ' p '( r ) ) d r: I o r I d r ' v
Jo Jo J

and by Gauss'stheorem this can be re-expressedas a surfaceintegral. Thus,

the result of imposing the variational constant given in eqn (5.88b)is to
replacethe integral of the surfacevariation appearing in the expressionfor
,tg b!, Al (eqn (5.85))with the result appearing on the right-hand side of eqn
r5.92),
(h2l4m\$ dsdfs(o,r)v2p'(r)
: - (h2l4m)$ OS(O,r)j Ot'{ (V/* )6{/+ ,lt*6v,lt} 'n(t). (s.e2)
It is important that no new volume contributions to the variation be obtained
by the imposition of the constraint given in eqn (5.88b),or one would no
Iongerobtain Schrodinger'sequationsas the Euler-Lagrange equation of the
variation.
Substitution of the identity in eqn (5.92)into eqn (5.85)and combining
terms yields the final result for the variation of the functional E bL, Ql'
6EbL,Ql: (h2l4m)$ dS(O,
r)Jdt' {v,1,.6,1,
- ,lt*6V,lt
)'n(r) * cc. (5.e3)
Equation (5.93)is a statementof a physicalprinciple as it can be restatedin
rermsof the flux in the infinitesimal changein vector current densitl' through
the surfacebounding the atom. In analogy with the definition of the charge
density, the single-particlevector current density, j(r), of a man\'-particle
160 THE QUANTUM ATOM 5.3

systemis definedas I
j(t) : (hl2mi)J dz' {r/*V,1,- (V,lt\,/,}. (s.e4) I
The infinitesimalchangein j(r) causedby a variationin r/ is
d*j(r) : (hI 2mi)[ dr' llt* 6Vl/ - (v,lt\ 6rl,] (5.e5)
or, equivalently,
- (v/* )6,1t
drj(r) : (hI 2mi)! dr'{rlt*Y(6,1,) },
and the expression for the variation of the subsystem energy functional
constrained to a region bounded by a zero-flux surfacebecomes

6EbL,ol : - (ihl2)$ ds(o,r)d,j(r).n(r)+ cc. (5.e6)

To demonstrate that the expression for 6Ellt,O] is equivalent to a
variational derivation of the hypervirial theorem for an atom in a molecule,it
is necessaryto take one more very important step which is to consider the
variations in tlr to be generatedby the action of an operator on tlt. That is,
one makes the identifications

6t : - t(ilh)Gt and 51,*: {ilh)Gfu* (s.e7)

where d is a linear Hermitian operator, an observable,and e denotes an
infinitesimal. At this point we shall use theseexpressionsto substitute for 6tlt
and 6rlt* in the expression for 6E\L,Q] and simply state that their use
correspondsto identifying the variations in the state function with the action
of generatorsof infinitesimal unitary transformations.Such operators,when
acting on a state function, cause changesin the dynamical properties of a
system. It is this identification which enabled Schwinger to transform the
variation of the action integral into the more powerful principle of stationary
action. These ideas are fully developed in Chapter 8 where it will also be
shown that the final result to be obtained for 69L,!,Qf, using eqn (5.97),is a
statement of Schwinger'sprinciple for a stationary state as it applies to an
atom in a molecule.
We proce.d^by defining a current density for the property associatedwith
the operator G in analogy with the current j, defined in eqn (5.42)

Jdr'{r/*VG,tt)- (v/-XG/)}.
jo(r): (hlzmi) (5.e8)
In termsof this vectorcurrentthe variationin the atomicenergyfunctionalis
givenby
6Ebl/,Ql : - (el2){$ds(Q,r)jo(r).n(r)+ cc}. (5.99)

Using the expressionderivedpreviouslyfor a subsystemstatementof the

hypervirialtheorem,eqn (5.43),we arrive at the atomic statementof the
E5.1 H A M I L T O N I A N - B A S E DF U N C T I O N A L S 16r

principle of stationary action for a stationary state

6Ebl/,Ql : - (e12) , lH, Gl,lt>" + cc}.

{(ilh)<,1, (s.100)
Equation (5.100)is a variational result and it applies only to regions of
space bounded by surfaceswhich satisfy the variational constraint of ex-
hibiting a zero-flux in the gradient vector field of the charge density. It
applies, therefore,to an atom, to any linked grouping of atoms, and to the
total system. Equation (5.100)is a generalrzatronof Schrodinger's original
derivation of his equations to yield the same equations together with the
variational statement of the hypervirial theorem. When O equals the total
spaceof the system,one has

6El,lt,Ql: 6El,l : - (el2) g,

{(ilh)<,!,\fr,Gl,! > * cc}: (5.101)
a result which yields not only Schrodinger'sequations but, in addition, the
hypervirial theorem

<,1/,lH,Al{/):0 (5.102)

and hence the variational development of all the mechanical theorems

derivablefrom it. For an atom, the variation in Eb!,91 it proportional to
the flux in the vector current density of the generator G (eqn (5.99))and eqn
(5.100)is consequentlya variational derivation of the subsystemhypervirial
theorem. Thus, for an atom, or any functional grouping of atoms, one
obtains, in analogy with the total system, Schrodinger's equations and a
uariational statemenrof the hypervirial theorem, a statement that applies
uniquely to systemsbounded by zero-flux surfaces,which for any operator
Gis
cc}: {Jds(o,r)ic(r)'n(r)
{(ilh)<rlt,lfr,Glrlt>"+ + cc}. (5.103)
choices
Throughrequisite for the operatord, eqn(5.103) the
determines
force acting on an atom in a molecule and, through the atomic statementof
the virial theorem, its energy. It establishesthe mechanicsof an atom in a
molecule, as is demonstrated in the next chapter.

E5.l The variation of Hamiltonian-based functionals

Schrodinger's use of the expression(ll2m)<frrlt,frrl,>to express the kinetic

energy in his functional Y lrl,l was based on an analogy with the
Hamilton-Jacobi equation of classicalmechanics.This is also the form used
for the kinetic energy as it appearsin the Lagrangian of the action integral, a
consequenceof the Lagrangian, either quantum or classical,being a function
of the field variables and their first derivatives only. Since the Lagrangian
degeneratesinto a functional of the form Ebltl for a stationary state,theseare
termed Lagrangran-based functionals. One -&y, however, construct an
 t62 THE QUANTUM ATOM E5.1 I
energy functional based upon the Hamiltonian operator which differs from
the above by expressing the kinetic energy in terms of the operator
- (h2l2m)V2.We investigatethe variational properties of such a functional
here. In this section the normalization constant will be handled by division
wit h ( { / , , 1t > as i n e q n (5 .6 3 )fo r S frl tl .
In analogy with Sbltl and its many-electron subsystemcounterpart, the
atomic form of the energy functional Sltl/,O] is defined as

: o' o, {(h,l2m)1,Y,,r,*
8l{,,o1 .y,r,
J I
+ Vrlt*(tj lQ!, rl,)" (85.1)
where

(t, t)" : fo,Io r'(/*(/. (8s.2)

The Hamiltonian-based functional E' [/, o] is defined as

8'l(t,Ol : +{tr [/, A] + ,tr*blt,Ofj (85.3)

where
f f
trl,lt,Of : dr dr'| - (h2l2m)L,,lt*v?,1,
J., J
+ Vrlt*,|,\
l(t, t)n. (85.4)
The Lagrangian-based functionalsemploythe form (llzm)<frtr,frrlr> for the
kineticenergyand are,therefore,necessarily real.The functionaltrlrl/,ol
can,however,be complex,asIi is not necessarily Hermitianovera subsystem
as shownin eqn(5.12).Sinceonecannothavea complexquantitystationary
with respectto a variation,onemust vary//' [/, o] togetherwith its complex
conjugateas definedin eqn(E5.3).
The two typesof functionahcan be relatedusingeqn (5.46)
P
8 A L , O l : 8 ' 1 1 , Q+) (h2
l4ml I or Y' p'(r)l(t,t)n (E5.5)
Jo
whereonly the term involvingthe coordinater survivesthe integrationover
the sumof termsL,v?(r!* r/).For a total system,or for a subsystem
satisfying
the zero-flu",.rrfICecondition.one has
8b!,Ol:8'l{/,Qf:E (E5.6a)
whered is the energyeigenvaluefor a stationarystate.Thus the zero-flux
surfaceconditioninsuresthat the real part of tr[/,o] equalssbL,o).
Becauseof the constraintimposedon the variationof Ef(r,ol (eqn(5.88b)),
E5.1 H A M I L T O N I A N - B A S E DF U \ C T I O \ A L S 163

one also has

6Eb!, Ql : 66'lll. Of (E5.6b)

and the variationsof both functionalsyield identical results.Since the

functionL,V?(rl,*r/)is a divergenceexpression,its additionto the functional
Slrlt,Ol to obtain 8'lll,Ql doesnot changethe Euler equationobtained
fromthevariation.However,asshownbelow,whilethevariationoI E'lrL,Q)
yieldsthe sameresultsas does 681,1,,O1, it doesso without placingrestric-
tions on the surfaceof O.
The variationof Elr!, Q] with the impositionof the variationalconstants
givenin eqn(5.88)at everystageof the variationand includinga variationof
the surfaceyields

6EN/,Ql : (h2l4m) $ ds(O,r) tdr'{Yrlt*6{t

- ,lt*6V,lt\' nl (lt , r/ )o + cc (8s.7)

in analogywith the result obtainedfor the variation of 9llr,Ql given in

eqn(5.93).
Thevariationof 8'l(/, Ql involvesvariationwith regardto 6t! and6Y2{t.lt
to performan integrationby partstwicein succession
is necessary to rid the
erpressionof the variationsof gradientsof r/. This procedureis illustrated
belowfor a singleterm in tr [r/, O]
- (h'I2m)
{a(l'*v?,14
Iav?,1'}
6v?,1,
: - (h'l2m)rl,*6V?rl,
: - (h'l2m)(- V/* . 6Vrl,+ V. (,1,*6V,lt))
+ V' (rlt*6V{t- V,l,*6r1,)).
: - (h2l2m)( + V2tL*6,1, (E5.8)

L'singthisresult,the variationof 8'l{/, Ol includinga variationof thesurface

.rf Q is found to be
(.r1,, Q)
V)n6E'Lrl,,
t r r ^
: dr'{H'L*- ar*[r/,o]'l'*]6'l'
;Jrdr J
/r\f f ^
. (;,) t){{/*H(/-
r) dr'd/s(o. Jf*blt,O)(,*rltl
fds(o. J

r,Jo,' {Vrl,*6,Jt
+ (h2l4m){ottn, 'n +
- ,lt*6Yrl,} cc. (E5.9)

{t the point of variation with ,trlrlt,,O] : E, one obtains Schrodinger's

 164 T H E Q U A N T UM A T O M F,5.2

equationsas the Eulerequationsand

,8' bL,Ql : (h2l4m)$dS(Ct. r) J dr' {V,lt*6,1,,

- ,lt*6V,lt|.nl(1,,rlt)a+ cc. (E5.10)

This result, which is obtained without any restrictionson O is identical to

I
that obtained from the variation of 8bl/, Ol by constraining the subsystemto
be bounded by a zero-flux surface.It is clear from eqn (E5.9)that, unlike the
variation of 9llt, Q] or E lrlt, Q], one obtains the sameresult for the variation
of 8'[/, O] whether or not the surfaceis varied.
The Lagrangian-basedfunctional Ebl/,ol or E[r/,o] derives directly
from the Lagrangian as employed in the quantum action principle. For a
total system,both the Lagrangian- and Hamiltonian-basedfunctionals yield
identical variational results. This equivalence in variational behaviour is
maintained for the corresponding subsystem functionals only if the sub-
systemis bounded by a zero-flux surface.Only an atomic region ensuresan
equivalencein both the valuesand the variational properties of the two types
of functionals (eqns(E5.6a,b))thereby preservingthe properties obtained for
a total system.

85.2 Vectors, tensors, and dyadics

A vector in three-dimensional Cartesian space is characterized by three

components
A : i A, * i A , + kA " (E 5.11)

where i, j, and k are unit vectors along x, y, and z, respectively.Equation

(E5.11)can be expressedas a product of a row and a column matrix as

lA,\
A: [i kl
1 l, :,
\A,l
iA*+jAr+kA". (E5.12)

Becauseof the orthogonality of the unit vectorsin a scalaror dot product, the
scalar product of two vectors A and B can be expressedas

/B'\
A.B : [1, Ay A"l
(u,I : A*8, + AyBy + A"8". (85.13)
\B"l
A dot product of two vectors yields a singlenumber, a scalar.It is possibleto
generatea three-by-threematrix by the multiplication of two vectors. This
85.2 V E C T O R S T. E N S O R S A. N D D \ " { D I C S 165

procedureis denoted by the symbol AB which definesthe operation

A,B, I,B,\
//,\ lo,u,
:
A B I , .' Il t-g^, B , B "" ) : l o " B , A , B , A , B- ,l l . (E5.14)
I I
\a,l \1,8.. A.B, A,B=I
'Ihe
result of this operation is called a dyadic or tensor
D: AB.
A dyadic behaveslike a two-headedvector - the scalarproduct of D with a
vectorAyieldsavector
D'A : C,
C,: D r r A , + D r y A y+ D r r A ,
(Es.15)
C,: D*A,+DrrAy+Dr"A=,

C,: D " , A , + D - y A y+ D - - 4 , .

A scalar is obtained only by taking a scalar product of D from both sides

n.D.n: B.c. (E5.16)
A dyadic is required to describethose directed(vectorial)propertiesof a
systemwhich result from the application of a force or field along directions
orthogonal to the observedresultant.The polarizability of a moleculeis, for
example,describedby the polarrzability tensor a. The dipole p induced by an
applied field E is given by
F: a'E. (Es.17)
Even if E is directed along the z-axis, E : kE-, the x- and y-components
of p can differ from zero.Using eqn (85.15),the three componentsof p are
given by
llt: dr"E,

Fy: dy"Ey

: d""Er'
!"

If a dyadic is symmetric,then
A . D :D . A ,
a property possessedby the stresstensor F' (eqn (5.28)).An example of an
unsymmetricaldyadic is provided by the product jr as encounteredin the
derivationof Ehrenfest'sfirst relationship,eqn (5.26).The identity usedin the
derivation of this relation
V'(jr): j + rV'j
 t66 THE QUANTUM ATOM E5.3

is obtained from the matrix multiplication

J,y irt
: [_*ov
v (jr) ot f;:
:,,:1 ir! j.r'l
\

Lox t
\ j,, i"y J"z I
In tensor notation, an elementof the tensorjr is

Tn: i*xt
and
oTo,
A*o:Jxx\*Jxxw'
Two types of terms appear in the stresstensor F in eqn (5.28),terms of the
form ,lr*VV,ltand Vr/*Vrl. The divergenceof this tensor determinesthe force
- V'F. Taking the divergenceof F' yields terms of the form

V ' (/* V Vrl t) : Y rL * ' VV ,l , + {/* V F2{ /)

and

V' (Vr2'tYrl4: V'rl,*Yrlt+ Vlt*' VV,lt.

The subtractionof thesetwo contributions as they appearin V.F' yields the
terms appearing in the equation for the force preceding eqn (5.28).

E5.3 Divergenceof a vector and Gauss'stheorem

Consider an element of surface area dS and define n to be the outwardly

directed unit vector normal to dS. Then j'n is the component of the vector
current density normal to the surfaceelementdS and the current through this
element is j'n dS. The total current out of some region Q bounded by a
surface S(A) is obtained by summing the contributions from all its surface
elements.The resulting surfaceintegral is called 'the flux j through the surface
S(Q)',that is

Flux of j through the surfaceS(O) : $dS(O,r)j(r).n(r). (E5.tg)

This expressionis applied to the correspondingsurfaceintegral of any vector
quantity, evenif there is no physical flow through the surface.Thus references
in the text to a 'flux in the gradient vector of p' do not imply any physical
flow.
one can imagine the region Q bounded by s(o) to be divided into an
arbitrary number of smaller volumes.The total flux through S(O)is still given
by eqn (E5.18)sincethis total flux is equal to the sum of the fluxesout of all
the interior volumes.Consider,for example,a surfaceS,obetweentwo interior
regionsQr and ou. The flux out of Q, through S,ois $dS,oj.n,which must equal
the negativeof the flux out of Qk, {dsiftj.rr1,Sincen, : - nk. Thus the flux
 REFERE\CES 167

out of one region equalsthe flux into the neighbouringregion through their
common surface
Gauss'stheorem is proved by consideringthe flux out of an infinitesimally
small cube. Consider a cube with edgesof lengths A-x,A1',and A,z and denote
the coordinates of the corner nearest the origin as (,x,y, z). The flux of a
vector, say j, through the surfaceof the cube is equal to the fluxes through
each of the six faces. The outward flux through the face A'yA^zwhich is
perpendicular to the x-axis in the negative ,x direction, called face (1), is
- l,(x)A-vAz where .t"(x) is the x-component of j. The flux out of the
The component of
correspondingface at x * Ax, face (2), is/,(x + L,x)A'yA,z.
j,at x * Ax is slightly differentfrom i,at x and to first-order one has

/,(r + Ax):,r(x)+'fi1ix.
The sum of the fluxes through faces(1) and (2) is

Fluxout of (1)and(2):U+L,xL,yL,z.
ox
The total flux through all six facesof the cube is, therefore,
(ai' ai,
Total flux in j out of cube - \ e '
ey *)o*oro,.
-f--l-
0z/

This total flux is also given by the surfaceintegral of j'n and, hence,
'iLV
$ O s j ' n: V
where LV is the volume of the cube. Thus, the divergenceof a vector at a
given point in spaceis the net outflow or flux of i per unit volume in the
neighbourhoodof the point. From the above demonstrationthat the total
flux from a volume is the sum of the fluxes out of each of its parts, we seethat
the flux through a surfaceS(O) bounding a finite volume Q is given by the
sum of the fluxes out of each of its infinitesimalelementsdV, or

": o' i d v (Es.

re)
$ottnr;J
* hich is Gauss's theorem.

References

Bader,R. F. W. (1988).Pure appl. Chem.60, 145'

Bader, R. F. W. and Beddall, P. M. (1972)'J. Chem' Ph,r's. 56. ,1,310.
C h e n t . 3-.fl - .
Bader,R. F. W. and Preston,H. J. T. (1969).Int. J. q u a n t u m
Cohen,L. (1979).J. Chem. Phys.70,788.
Coulson,C. A. and Longuet-Higgins,H. C. Q9a7Q. P r o c . r . S o t ' . . L L t n t Lr r r A l 9 l . -19
 r68 T H E Q L I A N T L : l v{l T O \ 1

Coulson, C. A. and Longuet-Higgins,H. C. (1941h1. Prot'. r. Sor'.,London A192, 16.

Coulson, C. A. and Longuet-Higgins,H. C. (l948ri).Prot'.r. Soc..London A193,447.
Coulson,C. A. and Longuet-Higgins,H. C. (1948b).Prot'.r. Sot'.,London A193,456.
Coulson,C. A. and Longuet-Higgins.H. C. (194t{c).Prot'.r. Soc.,London A195, 188.
Courant, R. and Hilbert, D. (1953).Metfuttlsof'muthemuticalpftusic's, Vol. 1. Wiley I

Interscience,New York.
Dirac, P. A. M. (1929).Proc. r. Sot'.,Lonrlon Al23,l14.
Epstein, S.T.(1974a) The uariutionmethod in quantumchemistry.Academic Press.
New York.
Epstein,S. T. (1974h).J. Chem.Phy,s.60, 3351.
F e y n m a n ,R . P . ( 1 9 4 8 ) .R e t : m . od. Phvs.20,367.
Feynman, R. P., Leighton, R. 8., and Sands, M. (1964).The Feynman lectureson
pftl'sic's, Vol. II. Addison-Wesley,Reading,Massachusetts.
H a l l , G . G . ( 1 9 5 1 )P . r o c . r . S o c . ,L o n d o nA 2 0 5 , 5 1 1 .
Hirschfelder,J. O. (1960).J. Chem. Phys.33, 1462.
H u n d , F . ( 1 9 3 3 )P . h y s i kl , 1 6 3 .
H u n d , F . ( 1 9 3 7 )P . h y s i k5 . l .
Hurfey, A. C. (1954).Proc. r. Soc.,London A226, 179.
Lennard-Jones,J. E. (1949a).Proc. r. Soc.,London A198, 1.
Lennard-Jones,J. E. (1949h).Proc'.r. Sor'.,London A198. 14.
Lennard-Jones,J. E. (1952).J. Chem. Phys.2A,fi24.
L e w i s ,G . N . ( 1 9 1 6 )J. . A m . c ' h e mS. o c . 3 8 , 7 6 2 .
L e w i s ,G . N . ( 1 9 3 3 ) J. . C h e m .P h y s .l , 1 7 .
Lowdin, P.-O. (1948).A theoreticalinuestiglation into somepropertiesof ionit' cr,vstals.
Almqvist and Wiksell, Stockholm.
L o w d i n , P . - O .( 1 9 5 6 ) P . h i l . M a q . S u p p l .5 , l .
Messiah,A. (1958).Quantummechanics, Vol. I. Wiley, New York.
Moflitt, W. (1951).Proc'.r. Soc'.,London A21t1.245.
Mulliken, R. S. (1928a).Phy's.Ret:.32,186.
Mulliken, R. S. (1928b).Phys. Ret;.32, 388.
Mulliken. R. S. (1928c)P . h y s .R e t . 3 2 , 7 6 1 .
M u l l i k e n , R . S . ( 1 9 3 2 a ) .P h v s .R e u . 4 0 ,5 5 .
M u l l i k e n , R . S . ( 1 9 3 2 h ) .R e u .m o d . P h v s . 4 ,L
M u l l i k e n .R S . ( 1 9 3 5 )J. . C h e m .P h y s . 3 , 3 7 5 .
M u l l i k e n ,R . S . ( 1 9 3 9 ) P . h y s .R e t . 5 6 , 1 7 8 .
Pauli, W. (1958)In Encyclopediaofphvsic's, Vol.5, part 1 (ed. S. F. Lrigge).Springer.
Berlin.
P a u l i n g ,L . ( 1 9 3 1 ) J. . A m . c h e m .S o c .5 3 , 1 3 6 7 .
Pauling,L. (1960).The natured'the chemic'al fumd(3rd edn).Cornell University Pre:,.
I t h a c a .N e w Y o r k .
R o o t h a a nC , . C . J . ( 1 9 5 1 ; R e r .m o d . P h 1 ' s . 2 3 , 6 9 .
SchrcidingeE r , . ( 1 9 2 6 1A. n n . t l . P h t ' s . 7 9 . 3 6 1 .
Schrodinger,E. (1921).Ann. tl. Pfir's.82, 265.
S c h w i n g e rJ,. ( 1 9 5 1 )P. h _ r , sR.e r . 8 2 . 9 1 4 ,
S l a t e r ,J . C . ( 1 9 3 1 ) P. l , , r ' rR
. er.37.481.
S r e b r e n i kS , . a n d B a d e r .R . F . W . ( 1 9 7 4 1J . C h e m .P h y ' s 6 . 1,2536.
S r e b r e n i kS. . a n d B a d e r .R . F . W . ( 1 9 7 5 )J. . C h e m .P h ' ; s . 6 3 . 3 9 4 5 .
 6

T H E M E , C H A N I C SO F A N A T O M
IN A MOLECULE

I hope and believethat the present statementswill prove useful in the elucidation of
the magnetic properties of atoms and molecules,and further for explaining the flow of
electricity in solid bodies.
- Erwin Schrodinger (1926).The statementfollowing his introduction of the charge
and current densitiesand the quantum equation of continuity in his fourth paper
'wave mechanics'.
on

6.1 An atomic view of the propertiesof matter

6.1.1 The chargeand current densities

'wave mechanics'to provide a
Schrodinger'sbelief in the ability of his
description of the properties of matter as demonstrated in the opening
quotation was well founded and his mechanicshas been used to understand
phenomenaunknown in 1926.The rolesof the chargeand current densitiesin
quantum mechanicsgo beyond their use in the descriptionof the electrical
and magneticpropertiesof matter.The chargedensityprovidesa description
of the distribution of chargethroughout real spaceand is the bridge between
the concept of state functions in Hilbert space and the physical model of
matter in real space.Schrodingersubscribedto this same view, pointing out
in his fourth paper wherein he definesthe electric density at a point in real
space,'thatthe ry'function itselfcannot be and may not be directly interpreted
in terms of three-dimensionalspace becauseit is in general a function in
configurationspace,and not real space.'Thezero-flux boundary condition,
as stated in terms of the gradient vector field of the charge density, deftnesa
single atom and objects made from collectionsof atoms. As demonstratedin
this chapter,the form and ultimately the properties of an object, microscopic
or macroscopic,are determinedby its distribution of charge.
A subsystemis an open system,free to exchangechargeand momentum
with its environment.Thus the current densityjo for any observabled is of
particularimportancein the mechanicsof a subsystem,sincea non-vanishing
flux in this current impliesa fluctuationin the subsystemaveragevalue of the
property G.
Variation of the energy functional for a subsystemis equal to an infinitesi-
mal times the flux in the current density of the generator G through the
170 T H E M E C H A N I C SO F A N A T O M I N A M O L E C U L E 6.1

surfaceof the subsystem(eqn (5.99)and repeatedhere as eqn (6.1))

6 ' 9 b l l ,Ql : - Gl2) (6.1)

{ $ d S (O ,r )j o(r).n(r) + cc} .
The samesurfaceintegral appearsin the subsystemstatementof the hyper-
virial theorem

Glt)a + cc}: {$dS(C),

{(ilh)<,lt,ln, r)jo(r).n(r)
+ cc} (6.2)

and, becauseof the dependenceof both quantities on the current flux, one
obtains the atomic statement of the principle of stationary action for a
stationarystateas given in eqn (5.100).This principle forms the basisfor the
discussionof the mechanicsof an atom in a molecule and is repeatedhere

SEb!,Ol : - (el2)\(ilh)<,lt,lH,
Gfrl)o+ cc). (6.3)
The non-vanishingof the flux of a quantum mechanicalcurrent in the
absenceof a magneticfield is what distinguishesthe mechanicsof a subsystem
from that of the total system in a stationary state. The flux in the current
density will vanish through any surface on which r/ satisfiesthe natural
boundary condition,Vll ' n : 0 (eqn(5.62)),a condition which is satisfiedby a
system with boundaries at infinity. Thus, for a total system the energy is
stationary in the usual sense,\ElrLl:0, and the usual form of the hyper-
virial theorem is obtained with the vanishing of the commutator average,

cl /) : o.
<,1t,\fr, (6.4)
As noted in the previouschapter,eqn (6.4)is a consequence of the Hermitian
property of H, a property not enjoyed in general,by a subsystem.
To further investigatethe relationship betweenthe variation of an energy
functional and the current consider the time derivative of the density of a
property G. Assume the corresponding operator d to have no explicit time
dependence.Let
Pc:Y*dY

where Y is a solution to Schrodinger'stime-dependentequations (eqn (5.2)).

Then
dpoldt + V ' jG : \lh)V* \fr, Cl Y.

This result is obtained through the use of eqn (5.2)and by the addition and
subtraction of the term Y*tidY, the same procedure followed in the
derivation of the subsystem hypervirial theorem, eqn (5.40). The usual
quantum equation of continuity for the density p : Y*Y as derived in eqn
(5.1e),
dpldt+V'j:0,
is so-calledbecauseof its direct correspondence
with the equationfor
 6.1 A N A T O M I C V I E W O F T H E P R O P E R T I EO
S F M A T T E R 171

a classicalfluid. The presenceof the commutator in the expressionfor

pc( : dpoldt) has no classicalanalogueand is quantum mechanicalin origin.
The expressionfor po is analogousto that for i only when the commutator
va n i s hes .
The magnitudeof the commutator averageis relatedto the product of the
uncertaintiesin the energy and in the observable G (N{essiah1958, pp.
299-301).
LELG> il<tF,Gl)l (6.5)
where AG, the uncertaintyin the value of G, is given by its root-mean-square
d e vi at ion( t he s quarero o t o f th e fl u c tu a ti o ni n th e averageval ue (C ))

A ( G ): " < c ' )- ( ( G ) ) ' l ' , ' (6.6)

The non-vanishingof the subsystemaverageof the commutator implies a
fluctuation in the value of the observableG over the subsystemas measured
by the flux of its vector current density through the surfaceof the subsystem.
Th u s one ant ic ipate sa n d fi n d s n o n -v a n i s h i n gfl uctuati onsi n subsystem
expectationvaluesfor observableswhich do not commute with I?.

I Ev en in a s t at iona rys ta te ,th a t i s , a n e n e rg ye i g enstate,w here both / and

/o vanis h,one f indsth a t a n o n -v a n i s h i n gd i v j n th roughoutsomevol umeO i s
a sso c iat ed wit h an un c e rta i n tyi n th e v a l u eo f G . W hen G i s a constantof the
motion and the valuesE and G aresimultaneouslywell defined,the net flux in
the current through the boundary of O vanishes,as it does when O refersto
th e tot al s y s t em .T h u s , w h e n th e g e n e ra to rd o f an i nfi ni tesi maluni tarl
tra n s f or m at ionpos s e s s eas s h a rp v a l u e th e v a ri a ti on 69l crl t,O] vani she.s
and the energy functional is stationary.
When G doesnot possessa sharp valuein a stationarystate,there is a non-
vanishingcurrent whose net outflow from any infinitesimalregion is deter-
mined by the correspondingcommutator,

V' jc : Glh){/*lU,Glrl,
Th e e ner gyis not s t a ti o n a ryo v e r a v o l u me Q i n s u ch a si tuati on,i ts change
beingdeterminedby the flux of the current of G through the surface(eqn(6 I ))
or, equivalently,by the averageof the commutator (eqn (6 3)) It is clear from
this discussionthat I? retains the property of Hermiticity over a subsystem,
i .e .that
f f
:
| ,!. H:G(i)dr I Ur!*tGrlldt.
Jo Jo

only when H and G commute.

Becauseof the presenceof the surfaceterm in eqn (6.2),the hypervirial
theorem for a subsystemleads to important physical resultswhich have no
counterpartfor the total system.It is our immediategoal to use eqn (6.3)to
172 T H E M E C H A N I C SO F A N A T O M I N A M O L E C U L E 6.1

develop the mechanicsof an atom in a moleculethrough the derivation of the

atomic statementsof the Ehrenfestforce law and virial theorem for which the
surfaceterms describe,respectively,the force acting on the atom in terms of
the pressureexertedover its surfaceand the contribution of this surfaceforce
to the atomic virial.

6.1.2 Variational derivation of the atomic force law

The atomic statementof the principle of stationary action, eqn (6.3),yields a

variational derivation of the hypervirial theorem for any observable G, a
derivation which applies only to a region of spaceQ bounded by a surface
satisfyingthe condition of zero flux in the gradient vector field of the charge
density,

Vp(r)'n(r): 6 for all r on the surfaceof O. (6.7)

One may explicitly evaluatethe variation in the functionalg[/,O] in eqn
(6.3)for a particular observableby applying the correspondinginfinitesimal
unitary transformation to the state function in the functional. This procedure
is illustrated in the general time-dependentcasein Chapter 8. According to
eqn (6.1),the result of this variation is given by the surfaceintegral of the flux
in the current jo. Therefore,an equivalent derivation of the variational result
is obtained by using the subsystemstatementof the hypervirial theorem (eqn
(6.2))and this is the procedure to be followed here.It should be borne in mind
that the use of the hypervirial theorem (eqn (6.2)), which is of general
applicability, in lieu of the atomic statement of the principle of stationary
action (eqn (6.3)),is restricted to regions of space satisfying the zero-flux
boundary condition, since the principle is restricted to such regions.
We shall use the principle of stationary action to obtain a variational
definition of the force acting on an atom in a molecule.This derivation will
illustrate the important point that the definition of an atomic property
follows directly from the atomic statement of stationary action. To obtain
Ehrenfest'ssecond relationship as given in eqn (5.24)for the general time-
dependentcase,the operator d in eqn (6.3)and hencein eqn (6.2)is set equal
to 0r, the momentum operator of the electron whose coordinates are
integrated over the basin of the subsystem Q. The Hamiltonian in the
commutator is taken to be the many-electron.fixed-nucleusHamiltonian

Ii : - (h'l2m)L,v?
- I, L"z,r'6r,- x,l)-' + II e2(lri- .;l)-'
i<j

+ II Z o Z u e 2 ( X-" X p l ) -"
a<B

fr:i+t. (6.8)
 6.1 S F M A T T E R l7_1
A N A T O M I C V I E W O F T H E P R O P E R T I EO

The symbol Z will be usedto denotethe completepotentialenergyoperator,

the sum of the electron nuclear (n, electron-electron V",, and nuclear
nuclear (" potential energy operators,
t : t,n+ 4, + t^^. (6.9)
The commutator of this Hamiltonian and the momentum operator of a single
electron is, as in eqn (5.23),equal to ilNt.
The method of obtaining the subsystemaverageof the commutator and
henceof the force acting on the atom Q, is determinedby the definition of the
fu n ctionalE blr , O l v i a e q n (6 .3 ).It i s d e mo n s tra tedi n S ecti on6.2 that the
mode of integration used in the definition of the subsystemfunctional
'9
b!, Ol, (seeeqn (512) and discussionfollowing)is the only one which leads
to a physicallyrealizableboundary condition. Becauseof eqn (6.3),this same
mode of integration definesthe atomic averageof the commutator and thus
of the atomic force, F(O),
r l
.lH . p' I / )o * c c |: r y 1 d r r l d r ' t l t J t -* (V r V ) r l t
trith)<,1,
\N 12)
'i
Jo J
I : (6.10)
t
F(Q).

The result is multiplied by lt', the total number of electrons,in the definition
of an atomic property. The reader is reminded that the mode of integration
I indicatedby N Idt'rlt* ry'as usedin this definition of an atomic averageis the
sameas that employedin the definition of the electronicchargedensity.p(r)
'l'
(eqns(l .3)and (l .4)) From this point on the subscript will be dropped from
the coordinatesof the electronwhosecoordinatesare integratedonly over Q
and all single-particle, unlabelledcoordinatesand operatorswill referto this
electron.
The correspondingvariation of 9llt,O], subject to the constraint which
givesrise to the zero-fluxboundary condition (eqn (6.7)),may be determined
by evaluatingthe surfaceintegralof the current flux appearingin eqn (6.1)for
the operator G : 0 : -- ihV. Multiplication of the surfaceintegral plus its
complex conjugateby the factor l{ to match the atomic averagedefined in
e q n ( 6 . 1 0 )g i v e s
( N 12){ $dS (O ,r)j o (r)' n (r) + c c } : - ' (6.11)
$ d S (O,r)F (r) n(r)
where F(r) is the quantum mechanicalstresstensor introduced for the one-
electroncasein eqn (5.28).Its many-electronanalogueis defined as

Flry: (Nh' ftm)ldr'{v(vrl*)(/ -v,!*vrl, -v,ltv,lt*\,

+ rl,*Yv,lt (6.12)
a result which may be expressedin terms of the one-electrondensity matrix
f ( 1 ) ( e q n( 8 1 . 1 ) a) s

F1r): (h2l4m){(VV+ V'V')- (VV'+ V'V)}f"'(., r')1,: .,. (6.13)

t74 T H E M E C H A N I C SO F A N A T O M I N A M O L E C U L E 6.1

The stresstensor is a symmetric dyadic and its mathematical properties were

reviewedat the,endof Chapter 5. It has the dimensionsof pressure,force/unit
area,or, equivalently,of an energydensity.The quantum stresstensor plays a
dominant role in the description of the mechanicalproperties of an atom in a
molecule and in the local mechanicsof the charge density.
Combining eqns(6.10)and (6.11)yieldseqn (6.14),the atomic force law for
a stationary state (Bader 1980;Bader and Nguyen-Dang 1981),

F(Q): - (6.14)
$dS(Q,r)b'(r).n(r).
The force may be equivalently expressed using Gauss's theorem as an
integration of the force density - V.7(r) over the basin of the atom,

- vv){r:
F((]): N I dr ldr',1,*( drv.b'(r). (6.1s)
Jo J .|"
Equation (6.14)has a classicalanaloguewhich statesthat the force exertedon
the matter contained in a region O is equal to the negative of the pressure
acting on each elementof the surfacebounding the region. A local form of the
force law is readily obtained in the samemanner as used in the derivation of
the time derivative of the current density in eqn (5.29).No problems arise in
extending the expressionto the many-electroncaseand for a stationary state
the result is

rl,*(- V y'),lt: - V. F'(r)

F(r): .NJdz' (6.16)
which is clearly the differential form of the integrated force law in eqn (6.15).
The integrated and differential force laws have a number of important
consequenceswhich are now explored.
The potential energy operator t aueragedin eqns (6.10)and (6.16)is the
many-particle operator defined in eqns (6.8) and (6.9).The operator - V V
(eqn (6.17))isthe force exertedat the position r of electron 1 by all of the other
electronsand the nuclei in the system,each of the other particles being held
fixed in some arbitrary configuration (V = Vr and r = rr),
- V 1t : L " Z o e ' Y r ( l r ,- X " l ) - t - -
I , , , n ' yr ( r , r ; l ) - t

(6.r7)

The integrationimplied by dr'in eqn (6.16)averages this force on the electron

at r over the motions (i.e. positions) of all of the remaining particles in the
systemand the result rs theforce densityF (r), theforce exertedon the electron
at r by the aueragedistribution of the remainingparticles in the total system.
Integration of this force density over the basin of the atom Q then yields the
averageelectronic or Ehrenfestforce exertedon the atom in the system.Even
I
OF MATTER 175
6.1 A N A T O M I C V I E W O F T H E P R O P E R T I E S

though the force operator - V V invoh'esthe coordinatesof all the particles

in the system,and includestheir mutual interaction.the mode of integration
employed in eqn (6.16)yields a correspondingdensity in real spacewhose
integration over an atom with a boundar.v- defined in real spaceyields the
force acting on the atom.
The direct evaluation of the averagevalue of this operator requiresthe
information contained in the two-electrondensity matrix, yet, accordingto
equations(6.14)and (6.16),this force, in both its differentialand integrated
forms is determined by the stresstensor which requiresonly the one-electron
densitymatrix for its evaluation.One can view eqn (6.16)as a statementthat
the forces acting on a particle arising from the electrostatic interactions
betweenthe particles and describablein terms of the gradient of a potential
energyoperator are balancedby a force - V'F, which is purely quantum in
origin. The virial of the Ehrenfest force, which determines the potential
energyof the electrons,is also describablein terms of the stresstensor F; thus,
the mechanics of a quantum system is determined by the information
containedin the one-electrondensity matrix.
An atomic surface for an atom O is the union of some number of
interatomic surfacesdenotedby S(OlO',r), there being one such surfacefor
each bonded neighbour C)'.Thus, the force acting on an atom as given in eqn
(6.14\can be expressedas a sum of surfaceterms

F(Q): - Ir,,+o$ds(C)1o"
r)F(r).n(r). (6.18)

The sum in this equation runs over the surfacessharedwith atoms bonded to
Q, the atoms linked to O by atomic interaction lines.This expressionfor the
force acting on an atom provides the physical basisfor the model in which a
molecule is viewed as a set of interacting atoms. It isolates, through the
definition of structure, the set of atomic interactions which determines the
force acting on each atom in a molecule for any configuration of the nuclei.

6.1.3 The atomic virial theorem

Following the order of the theoremsderivable from the Heisenbergequation

givenin Chapter 5, we now considerthe useof the virial operator r'p (seeeqn
(5.30))in the statement of the principle of stationary action (eqn (6.3)) to
obtain the atomic statementof the virial theorem.The virial theorem may be
obtained by a scaling of the electroniccoordinates(Lowdin 1959)and, as
shown in Chapter 8, the use of the virial operator as the generator of an
infinitesimal unitary transformation is indeed equivalent to a scaling of the
electroniccoordinater (Srebrenikand Bader 1975;Bader and Nguyen-Dang
1e81).
116 T H E M E C H A N I C SO F A N A T O I V Il N A M O L E C U L E 6.1

in eqn (6.2)by N 12
Multiplication of the commutator averageappearing
f or G : i' p y i e l d sth e re s u l t

Q,,r i.0l r/)" * cc}

12){(ilh)<,/,,U;,
:2t',tJ
o'lor',
- h2l4m){,lt*V'rlt
+ (Vt ,l/.)r!}

- r'vtyt :27(a\+ v b(o).

* rt'i a, far',1,*( (6.1e)
Jo J
The first term is twice the averageelectronic kinetic energy of the atom ?n(O)
expressedin terms of the usual Laplacian operator and equal to the kinetic
energy,K(O)definedin eqns(5.47)and (5.50).Sincethe atom satisfiesthe zero-
flux boundary condition, 7(O) also equals the kinetic energy G(Cl)defined in
the sameequationsas (p 'p)al2m. The two expressions for the kinetic energy
l
differ by I(Q), the averageof the Laplacian of the charge density, which by
Green's theorem reducesto a surfaceintegral of the flux in Vp through the
surface of the atom. Consequently, as previously stressed,the average
electronic kinetic energy of an atom is a well-definedquantity. The second
term arising from the commutator and labelled {o(Q), is the integrated
averageof the virial of the Ehrenfestforce acting on an electron in the basin of
the atom

'4(O): rV o'Jo'
.ll
'rl,*(-r'VV){/.

Substitutionof theoperatori ' p for G in theexpression for thecurrentdensity

and multiplication of the surfaceterm in eqn (6.2) by I,{12yields
(N12){f ds(O,r)j, p. n(r)+ cc} : - (xh2l4ln) .vV)
{$ds(O,r)Jdz'[/* V(r
- v , l t *( r ' V / ) + / V ( r ' V / * )
- v/(. ' vt/ *)l ' n(r)l
: - f d s ( o ,r ) r . F ( r ) . n ( r )
- (h2l4m)f Os1o, r)Vp(r).n(r) (6.20)
where the final line is obtained through the use of the identity V(r'Vr/) :
Vr/ + r' VVr/. The negativeof the first term on the left-hand side of eqn (6.20)
is labelled ltr(Q), and is the virial of the Ehrenfest forces exerted on the
surfaceof the atom. The quantity F'n in eqn (6.21)is the outwardly directed
force per unit area of surfaceand r'7'n is the virial of this force,
{r(a): f ds(Q,r)r.F(r).n(r). (6.21)
The second term in eqn (6.20)is I(O), defined in eqn (5.50),expressedin
6.1 A N A T O M I C V I E W O F T H E P R O P E R T I E SO F M A T T E R 177

its equivalent surfaceform. Equating the commutator and surfaceresults

followed by some rearranging of terms f ields
- 2r@): t,o(Q)+ Y;(Q)+ r(o). (6.22)
Since the atom O is bounded by a surface of zero flux, I(O) : 0 and one
obtains the atomic statementof the virial theorem,
- 2r(q: li(Q) + t;(o) (6.23)

- 2r@): v"(Q). (6.24)

where t (Q), the sum of the surfaceand basin terms, is the total virial for the
atom. While the partitioning of the virial into basin and surfacecontributions
is dependentupon the choice of origin (an origin can always be found which
causesthe surfacevirial to vanish), the value of the total virial { (lO)is, as
evident from its equality with twice the kinetic energy, independent of this
choice.
Equation (6.24) is identical in form to the virial theorem for a total
system-the negative of twice the average kinetic energy of the electrons
equals the virial of the forces exerted on them. It is worthwhile here to
summarize the ways in which this result is dependent upon the zero-flux
boundary condition,€go (6.7).(1)The useof the principleof stationaryaction
to obtain a variational derivation of this theorem is restricted to a region
satisfyingeqn (5.7).(2) Satisfaction of eqn (6.7)ensuresthe vanishing of the
term l,(O) which arisesfrom the surfaceflux of the current density j. p (eqns
(6.20) and (6.22)).(3) The vanishing of l, (O) is also necessaryfor the kinetic
energy I(O) to be well defined.There is no statement corresponding to eqn
variational or otherwise,for a subsystemwith arbitrary boundaries.
16.24),
For a stationary state, a local statement of the virial theorem can be
obtained using the identity
V'(r'F) : Tr F'* r'V'F. (6.25)

As previously noted in eqn (5.32),the trace of the stresstensor is given in

terms of the kinetic energy densitiesdefined in eqn (5.50)by
TrF(r): -I((r) -G(r) (6.26)
or, equivalently,as
Trb'(r): -2G(r)-I(r)' (6.27)
Substitutingthis resultinto eqn (6.25)and rearrangingyields
- 2 G ( r ): - r ' V ' 3 + V ' ( r ' 3 ) - ( h 2 l 4 m ) p
Y(2r ) . ( 62 8 )

Fo r a s t at ionar y s ta te th e l o c a l v i ri a l - r' V' F equal s the vi ri al of the

178 T H E M E C H A N I C SO F A N A T O M I N A M O L E C U L E 6.1

Ehrenfestforce density F(r) as can be seen by taking the virial of eqn (6.16)

r ' F ( r ): N J d z ' , ! *- ( r ' V t 1 t y: - r ' V ' F ' . (6.2e)

Thus, the local statement of the virial theorem is, term for term, the
differential form of the integrated theorem in eqn (6.22). Becauseof this
correspondence,one can define the density corresponding to the total virial
{ (Q) as
{(r): -r'V'7 + V'(r'F) (6.30)
and the local form of the virial theorem can be written as

+ t (r).
th21+m7V2p(r'):2G(r) (6.31)
The kinetic energydensity G(r) is necessarilypositiveand eqn (6.31)demon-
stratesthat, in those regions where electronic charge is locally concentrated,
i.e. where the Laplacian of the charge density is negative, the electronic
potential energydensity {(1) is in local excessover the ratio of 2:I for the
averagevalue of Zto { in the virial theorem. Equation (6.31)is unique in
relating a property of the electronicchargedensity to the local componentsof
the total energy.It will be used extensivelyin the characterizationof bonding
and in the prediction of the mechanisms of generalizedLewis acid-base
reactions.
From eqn (6.30)it is clear that the virial of the electronicforces,which is the
electronicpotential energy,is totally determinedby the stresstensor F'and
henceby the one-electrondensity matrix. The atomic statementof the virial
theorem provides the basis for the definition of the energy of an atom in a
molecule,as is discussedin the sectionsfollowing Section6.2.2.

6.1.4 Correspondencebetweenlocal and subsystemmechanics

It has beenshown (Bader 1980;Bader and Nguyen-Dang 1981;Section8.4.2)

that all of the local expressionsobtained here for a stationary state, can be
obtained in their generaltime-dependentform from relations satisfiedby the
energy-momentum tensor, a quantity which summarizesthe principal prop-
erties of the quantum field, /*/ (Morse and Feshbach 1953).A subsystemis
an open system,free to exchangechargeand momentum with its neighbours.
These local expressions,as shown here for a stationary state, contain the
terms which yield the surfaceintegralsof current flux associatedwith an open
system.The surfaceterms are obtained as a direct result of the variation of the
subsystem energy or action integrals, terms which do not appear in the
variation of the correspondingexpressions for a total systemwith boundaries
at infinity. Thus the mechanicsof a subsystemreflectslocal properties of the
quantum field rlt*rltwhich are absent for a systemwith boundaries at infinity
and thus closedto the exchangeof charse and momentum.
6.2 ATOMIC PROPERTIES t79

6.2 Atomic properties

6.2.1 Single-particlebasisfor atomic properties

The atomic statementsof the Ehrenfest force law and of the virial theorem
establishthe mechanicsof an atom in a molecule.As was stressedin the
derivationsof these statements,the mode of integration used to obtain an
atomic average of an observableis determined by the definition of the
subsystemenergyfunctionalEbll, O]. It is important to demonstratethat the
definition of this functional is not arbitrary, but is determined by the
requirement that the definition of an open system, as obtained from the
principle of stationary action, be stated in terms of a physical property of the
total system.This requirement imposes a single-particlebasis on the defini-
tion of an atom, as expressedin the boundary condition of zero flux in the
gradient vector field of the chargedensity,and on the definition of its average
properties.
It has been shown that the principle of stationary action for a stationary
stateappliesto a systembounded at infinity and to one bounded by a surface
of zero flux in Vp(r). It is demonstratedin Chapter 8, through a variation of
the action integral, that the same boundary conditions are obtained in the
generaltime-dependentcase.One may seekthe most general solution to the
problem of defining an open system by asking for the set of all possible
subsystemsto which the principle of stationary action is applicable.Thus,
one must consider the variation of the energy functional I l{t, tOr} I defined

EbL,{a'}l: o"In,o"f^,
J", dr*{(h2l2r)I,V, rl,*'Y ,rl,

+ (V + 1),1,*,lt.} (6.32)
u here {O,} denotes a set of subspaces,the integration of the coordinates of
electron i being restricted to the region Q,. (In the time-dependentcase,one
performs a correspondingintegration of the many-particle Lagrangian dens-
:r\ over a set of separateregions.)Carrying through the same variational
procedureas followed in the variation of 9ft!,Q], one finds (Srebrenikand
Bader 1975) that the condition for the satisfaction of the principle of
.tationary action is that each subsystemQ, be bounded by a surface S,
satisfyinga zero-flux boundary condition of the form
V i p i (r;)' n (rr): g fo r a l l r , e S , (6.33t
* here

PiF) :.l:,.l:,[, (fl dz;*,)rlt*1,.

180 T H E M E C H A N I C SO F A N A T O M I N A M O L E C U L E 6.2

The quantity p'(rr) is the probability density that one electronis at r, when
each of the remainingelectronsis in one of the subsystemvolumes,Or. Thus
pit) does not, in general,describea physically realizabledistribution of
charge and it requires the diagonal element of the full l{-particle density
matrix for its evaluation. In only one instance does p, assume physical
meaning.This occurs when all the surfacesS, but one are taken at infinity, in
which casepj reducesto (IlN)p, where p is the measurablecharge density.
The defining condition of the subspaceso;, eqn (6.33),then reducesto the
zero-flux surfacecondition on Vp (eqn (6.7)).Thus, out of the complete set of
mathematical solutions to the problem of determining subsystems that
preserve the variational properties of a total system, only one solution is
physically realizable.

6.2.2 Definition of atomic properties

From the precedingdiscussion,the mode of integration used in the definition

of an atomic property is determined by the atomic variation principle and is
the sameas that used in the definition of the chargedensity itself.The atomic
averageof an observable,i is given by

A@)= (.4)n : a, (N {,t. A,t+ 6(')*{'}.

J: [a,, t2)
(6.34)

An atomic property is therefore.determined by the integration of a corres-

ponding property density pArJ over the basin of the atom where

p eF): (N12)
! at' {,lt*A,lt+ (ArtY,t,} (6.35)
and

,4(Q) :
J:dt p ng). (6.36)

one not es t ha t, 1 f A :1 , th e p ro p e rty d e n si tyi n eqn (6.35)reducesto the

electronic charge density. The operator A is either a one-electronoperator, a
function of the electronic coordinate r, Apr\,or elseit is a sum of operators
each of which containsthe coordinater, as illustratedin the averagingof the
commutator [f, 0] in eqn (6.10).The operator ,4 in this ,ur. .orrJrplnds to
- V V, eqn (6.17),wherethe gradientis taken with respectto the coordinater,
and every term in the sums over the nuclear and electronic coordinates
involves the distance of another particle from the electron at r. Thus the
averaging of this operator, as indicated in eqn (6.35), yields a density
representing (l{ times) the force exerted on the density of charge of the
electronat r by the remaining particlesaveragedover all possibleconfigura-
tions, just as the charge density is (.v times) the density of charge of one
 ATOMIC PROPERTIES 181

electron at r as deterrnined by averaging the motion of the remaining

particlesover all possibleconfigurations.Becauseeach contribution to the
operator I involvesthe coordinater, all atomic propertiescan be expressed
u si n geit her t he one -d e n s i tyma tri x f" ' (.,r' ) (e q n (81.1))for a one-el ectron
operator, expressedas A(r,r') in the most generalcase
r ^
A(Ol: , ( r ,r ' ) f ( t )( r ,r ' ) l . : , , ) ,
drtA (6.37)
.J"
(' )(r,,rr) o f th e tw o -d e nsi tymatri x (eqn(E 1.3))for
o r the diagonaleleme n tsf
a tw o- elec t r onoper a to r A (rr,r.,)

,4(C)) :i. dtt

[0,,
A F, ,r r ) I - ( t ) ( r r r, z ) (6.38)

( 2)is nor m a l i z e dto l { (A/ - l )/2 p a i rs .T h e useof the operatori ' p as

wh e r ef
the generator yields the definition of the electronic potential energl' and
a p p e ar ingin t he c o m m u ta to r [Ii ,i ' 0 ] i s a s u m o f tw o-el ectronoperatorsof
t h e f o r m e zl ( l r r - r . r l ) : e z / r , , w i t hi > L T h e y d e f i n et h e e l e c t r o ne l e c t r o n
repulsionenergy of Q as

L:.(o) : i"a''
fo" ( e 'l r r z) f t z t( r , , r r ) . r 6 . i 9t

Th e quant it y 4. ( O) e q u a l s th e re p u l s i o n e n e rgy of the el ectronsi n O.

t'""(O, Q) and one-halfthe repulsionof the electronsin Q with that of thosein
th e r em ainderof t h e s y s te mf)' , 4 ,({ r,Q ' ). S i n ce the averageval ue of the
operator for the total sYstemis

<t,">: v",(Q,a) o) + v""(e'',Q'),(6'40)

+ v,"(Q',
+ v'""(e),Q')
one necessarilyhas 2""(Q, O') : Vr,(O', Q).
The most important consequence of the definition of an atomic property as
givenin eqn (6.34)or (6.36)is that the averagevalue of an observablefor the
to ta l s y s t em< A ) , i s g i v e n b y th e s u m o f i ts a to mi c contri buti ons,4(Q),

<A>: Io/ (ct). (6.41)

Equation (6.41)is true for both one- and two-particle operators.Equation

16.41)statesthat each atom makes an additive contribution to the value of
every property for a total system.This is the principleunderlyingthe corner-
stone of chemistry-that atoms and functional groupings of atoms make
recognizable contributionsto the total propertiesof a system.In practice,we
recognizea group and predict its effectupon the staticand reactiveproperties
of a system in terms of a set of properties assignedto the group. In the
limiting caseof a group being essentiallythe same in two differentsystems.
one obtains a so-calledadditivity schemefor the total properties,for in this
182 T H E M E C H A N I C SO F A N A T O I V Il N A M O L E C U L E 6.2

casethe atomic contributions,as well as being additive in the senseof eqn

(6.41),are transferablebetweenmolecules.
Even a property not representedby a linear Hermitian operator can be
expressedas a sum of atomic contributions, as in eqn (6.41).The polariz-
ability of a molecule, for example, which is determined by the first-order
responseof the chargedensity to an electricfield, is not directly expressibleas
an averageover a correspondingoperator.This is not to say,however,that
the polarizabllity cannot be expressedas an additive atomic property, as is
indeeddone empirically.The atomic contributionsto the molecularpolariz-
ability and magnetic susceptibility are defined and discussedin Chapter 8.
Equation (6.41)is not usually applied directly to vector or tensor moments,
such as the dipole or quadrupole moments of a charge distribution, which
involve the definition of a singleorigin. Molecular moments are more usefully
relatedto a sum of atomic contributions with a local origin and this is done
by usingmomentsin addition to the one in question.This is illustratedfor the
dipole moment which, along with a number of other propertiesdetermined
directly by the chargedensity,is introduced below.
The electronpopulation of an atom in a molecule,its averagenumber of
electronsN (Q),is obtained by setting the operator ,4 equal to 1 in which case
peF) reducesto the electronicchargedensity p(r),

r{(c)) :I" p(r)dr. (6.42)

The net chargeon an atom q(|o-),is given by the sum of its nuclear chargeZne
and its averageelectronic charge - l/(O)e,
q(O) : (Zsr- /V(O))e. (6.43)
Setting A equalto rn, the radial distanceof an electron from the nucleusof
the atom, or some power rcof this distanceyields the correspondingaverage
over the charge density of the atom,

r'(o) :I.'
rirp$)dr. (6.44)

The atomic volume u(O) is defined as a measure of the region of space

enclosedby the intersectionof its interatomic surfacesand an envelopeof the
charge density of some chosenvalue. An atomic surfaceis the union of some
number of interatomic surfaces,there being one such surfacefor each bonded
neighbour and, if the atom is not an interior atom, some portions which may
be infinitely distant from the attractor. It is theselatter, open portions of the
atomic surfacewhich are replaced with an envelopeof the charge density, a
surfaceon which p(r) has a constant value. An envelopeof the chargedensity
can be used to define the 'van der Waals' shape or size of a molecule in
6.2 ATOMIC PROPERTTES 183

relation to its non-bonded interactions with other moleculesin the gaseous

and crystallinephases(Baderet ql.1961;Bader and Preston 1970;Bader et al.
1987aj.The0.001-audensityenvelopehas been shown to yield valuesof the
molecular diametersof methane and other inert monatomic and polyatomic
gasesin good agreementwith the equilibrium diametersof thesemoleculesas
determined by the second virial coefficient or viscosity data fitted with a
Lennard-Jones 6_ 12potential. The 0.001-audensity envelope enclosesover
98 per cent of the electronic charge of a hydrogen atom in a hydrocarbon
molecule and over 99 per cent of the electroniccharge of the atoms carbon to
neon. The 0.002-au contour has proved to provide a useful measure of
molecular size in crYstals.
The first moment of an atom's charge distribution M(Q), is obtained by
averagingthe vector rrr, with origin at the nucleus,over the charge density of
the atom.

M(o): - rlrr,,p(r)dt' (6.4s)

The first moment provides a measure of the extent and direction of the
dipolar polarrzationof the atom's chargedensity by determining the displace-
ment of the atom's centroid of negative charge from the position of its
nucleus.The dipole moment of a neutral molecule is expressedas
-- eLaZnXn - e rp (r)dt (6.46)
F I
where the electronicand nuclear position vectorsr and Xn, respectively,are
measuredfrom a common, arbitrary origin. They are related to the vector rn
with nucleus Q as origin by r : ro + Xo and use of this relationship in eqn
6.46) yields
F :LaQ(o)X" + I'M(C)) : F" * rtu. (6.47)

Thus the total molecular dipole moment can be equated to a sum of atomic
chargesand first moments.The first term in eqn (6.47),1",is the contribution
from the interatomic charge transfer, while the second term, pu, arisesfrom
the polarizations of the individual atomic distributions. In general, both
terms are important in determining p (Bader et al. 1987b)'
Another important atomic polarization is measuredby the axial compon-
ents of the quadrupole moment. The quadrupolar polarization of an atomic
densitymeasuredwith respectto the z-axis is given by

: -,
e""(e) - rfl)pg)dt (6.48)
I rez'n
*'ith correspondingdefinitions for the r- and y-axes.Defined in this manner'
rhe quadrupole moment tensor is traceless-the sum of the three diagonal
c'lementsis zero. Since it corresponds to a real symmetric matrix, it can be
 184 T H E M E C H A N I C SO F A N A T O M I N A M O L E C U L E 6.2

diagonalizedtoobtain threeprincipal axesand their correspondingmoments

of polarization as defined above. Each moment equals zero for a spherically
symmetric charge distribution while, for a sphereflattened at the poles of the
z-axis,an oblate spheroid,0 1Q",(O) : - +Q.,(O) : - iQrn(O). For an
atom with electronic charge depletedin a plane and concentratedalong the
perpendicular(z) axrs,Q""({2)< 0. Thesethree situationsare illustrated by the
correspondingdistortionsof a spherein Fig. 6.1.The quadrupolemoment is
the charge density analogue of a n population of the orbital model. The
physical moment has the advantage,however, of being definable even when
no plane or axis of symmetry exists for the definition of a n system.
The definition of the energy of an atom in a molecule requires detailed
consideration from a number of points of view and the following section is
devoted to that task. The definition is shown to follow directly from the
atomic statementof the virial theorern and, once having establishedthis fact,
the underlying equations are readily put down. We give the final equations
here.The energyof an atom in a molecule,E.(Q), is purely electronicin origin
and is defined as

E"(O):r(Q)+{(O) (6.72)
where the electronic kinetic energy Z(O) and the electronic virial { (Q\ are
defined in the atomic statement of the virial theorem, eqn (6.24).Becauseof
this theorem, E"(Q) satisfiesthe following alternative statementsof the virial
theorem,

E"(Q): -f(O) : ++"1'(O). (6.73)

The potential energy of an atom in a molecule is thus defined to be the

Qzz<o
Qzz<O Qzz:O Qzz>o

F t c . 6 . 1 . P i c t o r i a l d e m o n s t r a t i o no l q u a d r u p o l a r p o l a r i z a t i o n sof an electronic charee dis-

tnDutlon.
I
II 185
6 .3 ENERGYOF AN ATOM IN A MOLECULE

averageof the virial of the forces exerted on its electrons.Why this stepis
essentialto the partitioning of the total energyand what areits consequences,
are the topics of the following section.

6.3 Energy of an atom in a molecule

6.3.1 Physicalconstraintson partitioning the energy

A system which is isolated and characterized by its own Hamiltonian is

separated from other systems to a degree sufficient to exclude, to any
measurableextent, the exchange of identical particles between it and any
other system.The Hamiltonian is unique to a given system and through
Schrodinger'sequation,eqn (5.2)for a time-dependentstateor eqn (5.16)for a
stationarystate,it determinesthe state function. The state function in turn.
determinesall properties of the system,including its decomposition into. and
the properties of, its subsystems.There is no Hamiltonian or state function
for a subsystem of a total system. One cannot partition a many-particle
Hamiltonian operator into distinct sets of operators without violating the
indistinguishability of identical particles. However, all electrons contribute
equally to the charge density at any point in real space in a manner
determined by Schrodinger'sequation and a partitioning of spacetreats all
electronsequivalently.Thus, the definition of an atomic energy proceedsnot
through a spatial partitioning of the Hamiltonian or of the elementsof the
abstract Hilbert spaceon which it acts, but rather through a partitioning of
the Hamiltonian into a sum of effectiveone-electroncontributions. This is
accomplishedthrough the use of the virial operator for a singleelectron.It is
shown that this operator projects out from the total potential energy
r)peratorwhich includesall interparticleinteractions,that part belonging to a
singleelectron.This projection, sinceit is the virial of a force acting at a point
in space,yields an energy density which may be integrated over individual
spatial regions to yield corresponding averagepotential energies.

6.3.2 The virial and the partitioning of an energy of interaction

The discussionwill be confined to the energy of a stationary state with the

srate function expressed for a rigid nuclear framework within the
Born-Oppenheimerapproximation. The Hamiltonian is given in eqn (6'8)
and yields the averageenergy
E : (rlr,Hl, >. (6.4e)
It is the partitioning of the potential energy of interaction that is the
stumbling block in partitioning a total energy. It is shown that the virial
r86 T H E M E C H A N I C SO F A N A T O M I N A M O L E C U L E 6.3

operator can be usedto obtain a spatialpartitioning of the averagepotential

energy.Sincethe potential energyoperator V is a homogeneousfunction of
degree - I with respectto the electronic and nuclearcoordinates,it may be
alternativelyexpressedin terms of the virials of the correspondingforces
exertedon the electronsand the nuclei using Euler's theorem as

I " ( - x " . Y , t y :f i
I,(- r,'v,t1+ (6.s0)
or as

I,t,.F,+ I, x".F, : 1,. (6.51)

A force operator F* fot particleK, - V,.Z is the classicalforceexertedon this
particleby all the particlesin the system.Correspondingly,one may consider
the virial of this force, r,.'F*, to be the potential energy operator for the
particle since,accordingto eqn (6.51),the sum of such operatorsfor all the
particlesin the systemyields the potential energy operator 12.
This idea is further illustratedin terms of the simplestpossibleinteraction,
that betweenjust two particles which interact by an inversesquare force.
Their energyof interactionis describedby d,r. How much of this energyof
interactionbelongsto particle I and how much to particle 2? The answerto
this question is provi9ed by expressingthe potential energy in terms of the
v ir ials of t he fo rc e s ,F r : - V rt' a n d F z : - V rd,r, w hi ch the parti cl es
exert on one another.

r,'F,Irz'F,:trr. (6.s2)
Equation (6.52)expresses the simple but important result that each particle's
shareof the potentialenergyis given by the virial of the force exertedon it by
the other. The virial operator r*'F* is like a projection operator in that it
projects from t, that part of the potential energy operator belonging to
particle rc. In this elementary case,each share of the potential energy is
dependentupon the choice of origin used in the definition of the vectorsr,
and r,. This doesnot turn out to be the casewhen this idea is usedto spatially
partition the potential energy of a many-electronsystem.If one denotesby
- r* v* the complete set of virial operators in eqn (6.50),one has
I.
(I,.- r*'V*)'t: /
and the complete set of virial operators is idempotent as required for
projection operators.
We shall first demonstratethat the identificationof a particle'svirial with
its shareof the potential energyof interactionleadsto the correct physicsfor
the total systemand then show that it forms the basisfor the definition of the
energy of an atom in a molecule.
T he f ir s t s u m m a ti o ni n e q n (6 .5 1 )w h i c hdefi nestheel ectroni cshareof the
potential energyoperator should. togetherwith kinetic energyoperator f of
I

6 .3 E N E R G YO F A N A T O M I N A M O L E C U L E 187

eqn (6.8),define the electronicHamiltonian of the systemand this, in turn, the

electronic energy E.. Thus,
E,: (ri.): ((f * I,r,.F,)). (6.53)

This Hamiltonian definesthe averageelectronicpotential energy 4to be the

averagevirial of the forces exerted on the electrons,the electronic virial,

4:t: (I,r,.F,). (6.s4)

Correspondingly, the nuclear energy of the system is purely potential in
origin and is given by averaging the nuclear share of V given in eqn (6.51).
E n: ( ( 1 , X , . F " )) : - I , X " . V , E . (6.55)

The force - V "E is the Hellmann-Feynman electrostatic force exerted on

nucleusa as definedin eqn (5.37).The total energy E is thus given by
E:E"+E^. ( 65 6 )

When a molecule is in an equilibrium configuration, the external forces

e xe rt edon t he nuc le i ,th e fo rc e s -Y o E, v a n i s ha s does E . and. i n thi s casc-.
E:E".
The identificationof the potential energyof the electronswith the r irial trf
the forces exerted on them as in eqn (6.54) is more than a questit)n t-rf
semanticsas it yields a deeperphysicalinsight into the energyof a molecule
with a rigid nuclearframework.It is becausethe nuclei are clampedthat the
interpretation of the energy is not straightforward. The molecule can be in
mechanicalequilibrium in a given arbitrary configuration X only if each of
the nuclear forces - VoE is balancedby an applied external force.There is an
extendedform of Hamilton's principle in classicalmechanicswhich is used to
handle just this sort of problem-to obtain the equations of motion for a
system subject to forces of constraint and which, in addition, may be non-
conservative.In this situation the classicalaction integral is replacedby the
r ariational integral

,: l2m*I,r,' F,),
at(L,p? (6.s7)
l,',,'
which is to be compared with the expressionfor E" given in eqn (6.53).In
analogy with this classicalresult, it has been shown (Srebrenik et al. 1978)
that the electronicenergy of a molecule,8", can be obtained from a variation
principle if the variation is made subject to the constraints imposed on the
systemvia the external forces that must be applied to the nuclei to maintain
the configuration X.
Thus, it is establishedthat the energy E. is the minimum energy of a set of
electronsmoving in the field of a rigid nuclear framework. The electronic
ririal t (eqn (6.54)) that defines the electronic potential energy can be
I88 T H E M E C H A N I C SO F A N A T O M I N A M O L E C U L E 6.3

expressed
as
'
:-ii"',)l;i'^l'i,h^>
Y.E
+,r"X" (6.s8)
This equation shows that when the external forces on the nuclei vanish, the
electronicenergyE. equalsthe usual total energyE of eqn (6.49).It is more
informative however, to explicitly evaluate the average of the virial of the
forces appearing in eqn (6.54).This yields

f:(4">+<4">-IoXo.Fn" (6.5e)
where Fo" is the electroniccontribution to Fo, the force on nucleusa. That is,

Fo: -YnE: Fon * Fo"

: - z*2Lp*3a#+ * zoe2Jntrlffio, (6.60)

and
-
I"X".Fo. : I"X".yoE + I"x".Fo,,.
The virial of the nuclear-nuclear
repulsiveforcesequalsthe nuclear*nuclear
repulsiveenergy.That is,
: - r'LrLp*oZoZ
I"X".Fon fx"ffls
/r\
- x"l)-' : ( t^^>.
: . rl;) I"I, *nZnZp(lxe
In an equilibriumconfiguration,Fo"is balancedby the nuclearcontribution
Fnnand the virial of the electroniccontributionsFo"appearingin eqn (6.59)
reducesto the nuclear-nuclear repulsiveenergy4^ in agreementwith eqn
(6.58)for the casewhereYoE : 0. Thqs, out of the total potentialenergy
operatort, fi,is purelyelectronic, as (nis purelynuclear,but /." contains
contributionsform both setsof particles.The electronicshareof thispotential
energyof interactionis givenby the projectionof the virial operatorsfor the
electrons,
( ( - l , r , ' Y , 4 ) ) : < 4 " >- I " X " ' F o . :< 4 " >+ I o x o ' F . o ( 6 . 6 1 )
where F.n : - Fo" is the force which nucleus a exerts on the electronsand
I"Xn' F.o is the virial of the forceswhich the nuclei exert on the electrons.The
corresponding nuclear projection is
( ( - I " x o ' Y , t , )) : f " x " ' F o " , (6.62)
which is the virial of the forceswhich the electronsexerton the nuclei.The
additionof this resultto 4^ givesthe nuclearshareof the potentialenergyas
givenin eqn (6.55).
6.3 ENERGYOF AN ATOM IN A MOLECULE 189

Thus the potential energyof the electronsmoving in a fixed-nucleus

frameworkconsistsof the electron-electron repulsionenergy,the electron-
nuclearattractiveenergy,and the virial of theforceswhichthe nucleiexerton
the electrons(eqn (6.59)).For configurationsnot far removedfrom an
equilibriumconfiguration, this latterterm approaches4^in valueand,for a
systemat equilibrium,it equals2".. Since{, theaveragepotentialenergyof
the electrons,is the virial of the forcesexertedon the electrons(eqn(6.54)),
it
satisfiesthe virial theoremderivedpreviouslyin eqn (5.31),
-27:7':V, (6.63)
and the electronic energy E., therefore,satisfiesthe corresponding relation-
ships,
Er: -T:i{' (6.64)
The equationsrepresentedin (6.63)and (6.64)are true for any configuration
of the nuclei.The molecularvirial theoremmay alsobe expressed as
- 2 T: < V >* I " X "' Yo E . (6.65)
the form originally given by Slater (1933).

6.3.3 The electronic energy of an atom

The importanceof definingan electronicenergyand of beingableto express

it in theform of eqn(6.53)isthat it enables andhenceE. as
oneto express,Ff.
a sum of single-particlecontributions,
- r,.v,t}
fr" : L,fi,: I, { - (h,l2m)v? (6.66)
and
E": l{(f,): -r,'v,t}4t.
N J A T Jd t ' { , l t * ( - hl22 m ) V ? (6.67)
It should not escapethe readers'attention that the mode of integration used
in eqn (6.67)is identical to that used in the definition of an atomic average.
Thus the electronicenergy is N rirnesthe auerageenergy of a single electron,
and E" may be expressedas an integral over an effectivesingle-particleenergy
densitv
E. : Jdr {K (r)+ "to(r)} (6.68)
where K(r) is the kinetic energy density defined in eqn (5.50)and {o$) is the
virial field.
t o | ) - A r Jd t '{ , l t * (- r . V t ) r l , }: - r ' V ' F , (6.6e)
the field which determinesthe atomic virial {o(A) (eqns(6.19)and(6.29)).The
virial field foQ) is l{ times the potential energydensity of one electron at r as
190 T H E M E C H A N I C SO F A N A T O M I N A M O L E C U L E 6.3

determined by its averageenergy of interaction with all of the other particles

'averagefield'
in the system.It is an exact prescription of the experiencedby a
single electron in a many-electron system.
Thus the virial operator for a singleelectron definesa single-particleenergy
density required for the partitioning of the potential energy of an interacting
systemand for the definition of the energyof an atom in a molecule.Once the
averageenergy is expressedas an integral over an energy density, the energy
of any region is obtained simply by integrating the density over that region,
assuming the region has a physical significance.It is here that quantum
mechanicsthrough the atomic statementof stationary action plays a crucial
role. The virial theorem for a subsystemis restrictedto an atomic region, one
bounded by a zero-flux surface. In addition, as noted before, the kinetic
energy is not well defined for a region with arbitrary boundaries.Thus, only
the energy of an atomic region satisfies the quantum mechanical virial
theorem, in addition to being well defined.
The above discussionused the total system to illustrate the principle of
using the virial operator to define a potential energy density but the virial
density appearingin the integral for the total systemcannot be taken directly
over to the atomic case.The virial for a quantum subsystemconsists of a
surfaceas well as a basin contribution, eqn (6.23),and the virial density thus
contains in addition to {o$), the term V'(r'o). This definition of a virial
density is in agreementwith that derived from the local statementof the virial
theorem,eqn (6.30).The term V'(r'o) doesnot appearin the integratedvirial
for a total systemsince,in this case,the surfaceterm vanishes.Its presencein
the local expressionfor { (1\ ensuresthat this quantity and its integrated
value for an atom f (O) are independent of the choice of origin.
The subtraction of the kinetic energy density G(r) from each side of the
statementof the local virial theorem given in eqn (6.31)yields a definition of
an energy density E.(r) as

E"(r): G(r)* {(r) : - K(r). (6.70)

The secondequality given in eqn (6.70)follows from the definition of K(r) in

eqn (5.49).The integration of this energy density over a region of space
bounded by a surface of zero flux in Vp yields an energy E"(O) which will
satisfy the various statementsof the atomic virial theorem,

or
I" E" (r)d t II
J O
{ G ( r )+ { ( r ) } d r (6.7r)

E.(O) : 7n(Q)+ {(o). (6.72)

Becauseof the atomic virial theorem. eqn (6.24),the atomic energyE"(O)
6 .3 E N E R G YO F A N A T O M I N A M O L E C U L E l9l

satisfiesthe following relationships,which are the direct analoguesof the all

spaceresults given in eqn (6.64)for 8.,
: - r(o)--+{(o).
E"(O) (6.73\

Becauseof eqn (6.73)and the vanishing of the Laplacian of the chargedensity

over an atomic basin, the following identities hold,
E . ( Q:) - K ( o ) : - G ( o ) : + l T r o ( r ) d r . (6.74)

It is to be emphasizedthat all of the above relationships,together with the

atomic statementsof the virial theorem (eqns (6.72) and (6.74)),remain true
when Q refersto the total system.It is in this sensethat an atom is a quantum
subsystem.
From its definitionin eqn (6.71)interms of an integralof an energydensity,
it is clear that, like other atomic properties (eqn (6,41)).the sum of the
energiesof the atoms in a systemequals the total electronicenerg)'
E": IOE.(C)) t6 75)

and, when there are no forcesacting on any of the nuclei in the s)'st€m.tht:
sum equals the total molecular energy E (eqn (6.49)).
One may also view the possibilityof definingan electronicpotentralencrgl
densityas the result of { (r) being expressiblein terms of the quantum StteS:
tensorb'(r),eqn (6.30).The stresstensor is a function of just the coordinater
becauseit is a functional of the one-electrondensity matrix, one whoseform
implicitly includes the two-electron interactions.The electronic virial also
contains the nuclear-nuclear energy of repulsion (eqn (6.58)).Thus. the
electronic energy E"(O) includes a partitioning of this contribution to the
energy,through the virial of the forceswhich the nuclei exert on the electrons
(eqn (6.61)) even though <t^^> does not directly involve the electronic
coordinate r. The result is a non-arbitrary partitioning of the total energy of a
molecule.Section 5.2.1demonstrated the observational basis for this defini-
tion of the energy of an atom in a molecule by showing that the energy,like
all properties of an atom in a molecule,changesin direct responseto changes
in the atom's distribution of charge.Thus, when the form of an atom is the
samein two different molecules,it makes the same contribution to the total
energy in both molecules.The transferability of atoms and their properties
betweendifferent moleculesis illustrated in Section 6.4.2.

6.3.4 Potential energy contributions to an atomic energy

The individual contributions to the electronicpotential energyof an atom are

obtained by determining the action of the operator - r'V on t in the
expressionfor the basin virial, eqn (6.19).The basin virial is the virial of the
Ehrenfestforce and the force density F(r) of eqn (6.16)is evaluatedas the first
192 T H E M E C H A N I C SO F A N A T O M I N A M O L E C U L E 6.3 i

stepin the determinationof {(A). To keeppropertrack of the subscripts,the

coordinates whichareintegratedoverthe atomicbasinareagainsubscripted
with '1' and so r'V : rr .Vr . Theexpressionfor - V112is givenin eqn(6.17\
and,with thevectorS rrj, from electronT
to electron1,andrrn,from nucleusa
to electron1. definedas
fll:ft-f.l and flo:fr-Xo,

the virial of this force with respectto 11 may be expressedas

- 1 1 . vL t : -LoZor'r r,, - .y r ror)
I,X" r(Zoe2
+ ( l { - t ) { n ' r , ; + r r ' Y , ( e 2r , r ' ) } (6.76)

where the sum over / has been replaced by (l{ - 1) times the interaction
between electrons I and 2. The two-electron terms can be put into a more
useful form by applying the identity (eqn (6.52))
( r r ' V r * r z ' Y r ) r l r r: - r rzr

to (,n/- l)12of the final termsin eqn (6.76)which yields

- \12}
{tlr { e ' rt } + F r ' Yr - 1 1 'Y r ) ( r ' r r r ' ) } . (6.77)
Averagingof this final operatorexpression for the virial of V in the manner
indicatedin eqn (6.69)for the potentialenergydensity{oF), which is the
virial of the Ehrenfestforceeqn (6.29),yields

f o F )- l / I d r 'p * ( - r , . Y, t ) r l , : - 1 1. V , a ( r )
: [ - L*Z rrt r ,or - .Y
I" X" ,(z oe2r rrr)) p(r)
+ ! dr..e'r,r.tf(2)(r,,r,)
+ [ d r r ( r r . Y , - r r . Y r ) ( e ' r y r l \ f . { z t 1 r zr ,) . (6.78)

The nuclear-clectron termsin the operatoron the right-handsidereferto the

electroniccoordinater,, whichis excludedfrom the integrationin eqn(6.78).
The subscript'f is not indicatedon the coordinateappearing in p. The two-
electrontermsinvolvethe coordinates r, and r, and integrationof r/* ry'over
the coordinates of the remainingelectroniccoordinates andmultiplicationby
the factor l{(l{ - I)12,as indicated,yieldsthe two-electrondensitymatrix
;'tz)(r1,rr)(eqn(El.3)).Integrationof the densityin eqn (6.78)for the final
electroniccoordinater1 : r overthe basinof the atom yieldsthe basinvirial
!'o(Q) as indicatedin eqn (6.79)
ro(a): v"n(Q)
+ I"x".F"(o) + v""(Q)
* v(Q,92'1. (6.7e)
The term V""(Q) is the attractive energy of interaction between the
electronic charge density in the basin of atom O and all of the nuclei in the
6.3 E N E R G YO F A N A T O M I N A M O L E C U L E 193

system,
f
V,n({ll: _\L,Z,r' I r;tp(r)dr. (6.80)
Jo

Theprimarycontributionto this energyis from the interactionof the density

where
in the basinof atom O with its own nucleus,a quantitylabelled23"(C))

: - Zae'J " 1p(r)dr.

,'3"(o) (6.81)

The term V,"(Q)is the contribution of the electron-electron repulsion energy

to the energy of atom O as previously defined in eqn (6.39).It equals the
energyof repulsion of the electronsin {r, V,"({t, Q) and one-half of the energy
of repulsion between the electrons in O with those in the remainder of the
V""(l{r,{r'),
system,
V,,(Q): V""(Q,O) + V",(l{r,Q'). (6.82)

The second region O or Q' indicated in V""(Q,O') refers to the region of

'2'.
integration for electron
The quantity F"(O) is the force exerted on the electronic charge density in
the basin of atom O bv nucleusa,
I

F " ( o ): -2"€2 (6.83)

| Fr,lrf";p1r;dt.
Jn

This force is opposite in sign to the force which the charge density exerts on
nucleusa and, thus,the sum of F,(O) over all the atoms in a molecule,i.e.over
rhe total charge distribution, will equal the negative of Fn., the electronic
contribution to the Hellmann-Feynman force exerted on nucleus d, eqn
16.60).Thus, the sum of the virials X"'F"(O) over all the atoms and all the
nucleiin the systemwill equal a sum of the virials Xo'F"" over the nuclei.By
using eqns (6.59)and (6.58),one obtains
- (6.84)
I " I n X " . F " ( Q:) I " X " . F o " :I " X " ' Y o E+ < t ^ ^ > .
\\'hen there are no net forcesexertedon the nuclei and E" : E, the sum of the
r irials of the forces which the nuclei exert on the electrons equals the
nuclear-nuclearrepulsion energy, 2"". Thus, the sum I"X"'F"(O) may be
rnterpretedas the share of the nuclear-nuclear repulsion energy belonging to
arom O. This spatial partitioning of an energy betweenfixed point chargesis
accomplishedby relating it to the virial of a force which is determinedby
an integrationof the chargedensityover the spaceof a molecule,an atom at a
:lme.
The final term in eqn (6.79) arises from^ the projection of the electron-
electron repulsivepotential energy from V by the single-particleoperator
194 T H E M E C H A N I C SO F A N A T O M I N A M O L E C U L E 6.3

- r'V. In the integration of the final term in the virial density given in eqn
(6.78)to obtain V (lo-,O'), there is a cancellationof contributions when the
coordinates of both electronsare in O and the final expressionis

v (Q, Q') :i
JO
dtt
i d r r ( r r ' V 2 - r r ' V r ) r ' r r r t f { 2 ) ( r r ,r r ) (6.85)

where, as before, Q' refers to the space occupied by the remainder of the
system. This term vanishes for the total system and its vanishing for a
partitioning into atomic contributions implies that the energy of repulsion
betweenthe electronsin O with those in O'is equally sharedbetweenthe two
regions. This required symmetry in the interactions between indistinguish-
able particles is contained in the expressionfor V,"(Q), as detailed in eqn
(6.40)and is thus automatically obtained when V (Q, Q') vanishes.
While the individual contributions to the virial {o(A) can be given physical
interpretations,care must be exercisedin this regard. The value of the total
virial {(O) is independentof the choice of origin used in the definition of the
virial operator, but this is not the casefor the basin and surfaceterms treated
separately.The origin-dependent terms appearing in the total virial for an
atom are
I,X".F"(o)+ v (e,o')+ r"(a)
where {t(A) is the surfacevirial defined in eqn (6.21).Since the value of the
total virial is independentof origin, as are the values of the electron-nuclear
and electron-electron contributions, V",(Q) and V""(Q),respectively,the sum
of the three origin-dependent terms is also origin-independent.It has been
shown using numerical examples(Srebrenik and Bader 1975)that the same
choice of origin which causesthe surfaceintegral to vanish (thereby making
the functional Ebl,,Q] stationary with respect to the virial as generator)
causes V(Q,o-') to vanish as well. In this casethe atom's virial reducesto

{(o): v"^({r)
+ v""(Q)
+ 4',(Q), (6.86)
the virial of the nuclear forces equalling V^^({r), the atom's share of the
nuclear-nuclear repulsion energy for a system in an equilibrium configura-
tion. The contributions 2",,(O) and V""(Q) are well-defined as is l'(Q), in
terms of either -27 (O) or the stresstensor,and one may use eqn (6.86)to
determine V""(Q).For a systemin an equilibrium configuration, the sum over
the atomic contributions for each term in eqn (6.86)will yield the value of the
corresponding quantity for the total system,that is,

Lnr @): <t>, : (t,n),

I' v"^(e) I,, 4"(o): (t""),
Ir, 4',(a): ( t,^). (6.87)
The breakdown of the averagepotential energy of an atom into its contribu-
tions as givenin eqn (6.86)isillustratedand discussedin the following section.
6.4 P R O P E R T I E SO F A T O M S I N M O L E C U L E S 195

6.4 Properties of atoms in molecules

It is the operational essenceof the atomic hypothesisthat one can assign

properties to atoms and groupings of atoms in moleculesand on this basis
identify them in a given system or use their properties to predict the
behaviour of the systemin which they are found. The primary purpose of this
section is to demonstrate that the quantum atoms transform this atomic
hypothesis into an atomic theory of matter by identifying the atoms of
chemistry and defining their properties. This section is not a review of
applications, but is rather intended to introduce and illustrate the uses of
various atomic properties.
The properties of atoms in molecules are calculated using the program
with the corresponding acronym PROAIM, which was developed over a
period of years in the author's laboratory and described in the literature
(Biegler-Konig et al. 1982). The first step in this series of programs is
performed by EXTREME which locatesand classifiesthe critical points in a
charge distribution. The information from EXTREME definesthe structure
of a molecule and enablesone to calculatethe trajectorieswhich terminate at
the (3, - 1) critical points and thence define each of the interatomic surfaces
comprising the surfaceof a given atom. An atomic property, eqn (6.36)'is then
determined by numerical integration of the corresponding property density,
eqn (6.35), over the basin of the atom as defined by its stored surface
coordinates. A related program enables one to perform corresponding
property integrations over an interatomic surface to evaluate the surface
integrals appearing in and derived from the atomic hypervirial theorem,
eqn (6.2).
A check on the accuracy of the numerical integrations of the atomic
propertiesis provided by the evaluation of f (O).This quantity, definedin eqn
r5.50),is proportional to the integral of the Laplacian of the charge density
over the basin of the atom or, equivalently,to the flux in the gradient vector
field of p through the surfaceof the atom,

r f
t(Q) : - ( h 2 l 4 m ) o t o ( r ) d r : - ( h 2 l 4 m ) Q o s 1 o , r ) V p ( r ) ' n ( r ) .( 6 . 8 8 )
|
J o J

Becauseof the quantum boundary condition of zero flux, the value of Z(O)
should equal zero and the extent to which this is true in any given caseis a
direct test of how well the atomic surfacehas been approximated in the
rntegrationprocedure.The value of L(O) is a measureof the differencein the
rntegratedvaluesof the two equivalentexpressions, K(Q) and G(C)).for f (Q).
rhe averagekinetic energyof the atom. In favourablecasesthe I'aluesof tlQ)
i a l l i n t h e r a n g e 1 x 1 0 - a t o 1 x 1 0 - t a u , w h i c h r e p r e s e net r r o r s i n t h e
cstimatesof the kinetic energy of less than 0.4 kJ/mol and it is generalil
196 T H E M E C H A N I C SO F A N A T O I V I{ N A M O L E C U L E 6.4 I

possible,using reasonableamounts of computer time, to attain values for

l, ( O ) les st han 3 x 1 0 -3 a u .
There is a practical problem encounteredin the application of the theory.
This concernsthe use of the atomic virial theorem to define the energy of an
atom in a molecule. Charge densitiesare obtained from calculations using
basis sets of finite size and, consequently,they do not properly describethe
Hellmann-Feynman forces acting on the nuclei. Thus, one finds that the
forcesdo not vanish for the configuration of the nuclei which yieldsthe lowest
energy and the requirement of the virial theorem for an equilibrium geo-
metry, that the ratio y : ( V> l<T) : - 2, is not satisfied.Under these
conditions, the average values of ( f) and ( 12) must be multiplied by the
factors (1 + y) and (1 + llil,^respectively, in order to obtain the correct
r es ult s t hatE : - (7 " ) :+ < Z ). T h e re fo re ,each
atomi cki neti cenergy7" (Q)
calculatedfrom the wave function, is multiplied by the factor (1 + y) in order
that their sum, using eqn (6.75),will equal the negativeof the total energy E.
The factor (1 + y) is, in general,little different from - 1, its correct value for
an equilibrium geometry.

6.4.1 Propertiesdeterminedby the electronicchargedensity

'the power
Pauling (1960) defined electronegativity to be of an atom in a
molecule to attract electrons to itself'. This concept has proved to be
extremely useful and it is reflectedin the net chargeson the atoms found in
diatomic molecules.The atomic chargesin a diatomic molecule are a direct
measure of the relative abilities of the two atoms to attract and bind
electronic charge within their basins.The variation in the charge on atom A
in AB where A and B vary acrossthe secondrow of the periodic table, Li -* F
and including hydrogen are displayed in Fig. 6.2. The densitiesused in the
determination of theseatomic chargesare obtained from ground-statewave
functions that are expressedin terms of a large set of Slater-type basis
functions and they are closeto the Hartree-Fock limit (seereferencesto Cade
and Huo (1973,1975)and Cade and Wahl (1974)at the end of Chapter 2).
Other properties of the atoms and bonds in thesemoleculesare given in the
Appendix, as are their internuclearseparationsand state symbols.Each atom
withdraws charge from elementsto the left of it and donates charge to those
on its right, with H appearing between C and N. The orderings are as
anticipated with C and H possessingalmost equal electronegativities. As
illustrated later, the electronegativityof C relative to H increaseswith the
degreeof unsaturationand with the extent of geometricstrain.This result is
anticipated on the basis of the orbital model which predicts the electro-
negativity of C to increaseas the s character of its hybrid bonds to H
increases.Most of the secondaryvariation in chargesacross the table are
explicablein terms of the extent of chargetransferbeing limited by either the
(a)

lTr'-

(b)

LL
L i B e B C N O F
N a M g A l S i P S C I

I 2nd Row Hydride

Ftc. 6.2. (a) Bar graphs of the charges on the atoms in the ground states of the diatomic
nolecules AB where both A and B : Li, Be, B, C, H, N, O, F. This is the ordering of increasing
3iectronegativityas determined by theory-all chargesto the left oi the position of the reference
rrom are negative,all those to its right are positive. (b) Bar graphs of the charge on hy'drogen
;rH). for the second-and third-row diatomic hydrides. The reader is referred to the Appendir.
TablesA2 and A3, for characterizationsof the critical points in thesemoleculesand lor a listrng
..f their atomic properties.The results are obtained from near Hartree-Fock wa\e funcllr)n\
198 T H E M E C H A N I C SO F A N A T O M I N A M O L E C U L E 6.4

number of valenceelectronson the donor or vacancieson the acceptor.The

charges on the third-row elements Na --' Cl are also given relative to H in
their hydrides and, as anticipated, H advancestowards the electronegative
end of the scalein this row relative to its position uis-d-uisthe second-row
elements.Unexpectedly,sodium and magnesiumare slightly lesselectroposi-
tive than their second-rowcongeners.
The chargedistributionsof the second-and third-row hydridesare illustra-
ted in Fig. 6.3in the form of contour maps.The extentand directionof charge
transfer and its effecton the chargedistribution are reflectedin the behaviour
of the interatomic surfaceof zero flux which is indicated for each molecule.In
LiH the surfaceenvelops what is essentiallya Li ion while in HF the total
charge distribution is dominated by the forces exerted by the F nucleus
(Bader and Beddall 1973;Bader and Messer 1974).The chargesq(H) of AH
are also characteristicof the stable polyatomic speciesAH,, the two values
usually differing by lessthan 0.05 e, and they reflectthe chemicalbehaviour of
the hydrides. The hydrides of Li, Be, and B, for example, are all hydridic,
expelling molecular hydrogen from water and, for all of these, q(H) < 0.
There is a sharp break in the value of q(H) for methanefor which q(H) r 0
and this is a faithful reflection of the non-polar nature of this molecule.It has
a low solubility in water and does not dissociate.The remaining hydrides,
NHr, HrO, and HF, are all increasingly polar with q(H) > 0 and the
ordering of the chargesaccounts for the aqueous solution of ammonia being
basic and that of HF being acidic.
The ability to determine the charge on an atom in a molecule removesthe
necessityof defining a numerical electronegativityscale.The concept, how-
ever, remains useful and one may use the atomic populations to demonstrate
that they recover the basic idea underlying electronegativity-to predict the
degree of charge transfer between two atoms. Since hydrogen can either
donate or acceptbut a singleelectron,the electron population of hydrogen in
AH may be used to define an electronegativityper electron of A relative to
hydrogen. This electronegativity is given by X(A) - 1 - l/(H),
where l{(H) is the population of H in AH. A positive or negative value
for X(A) implies that A has a greater or lesserbonding electron affinity than
does hydrogen, respectively.If the X(A) are meaningful, then the difference
lX(A) - X(B)l should determine the charge transfer per valence electron
in AB. Using this concept,the population of A in AB is predictedto be
- tx(B) - X(A)]ul
lf (A)e": [Ar(A)"
where u is the number of valenceelectronson the donor A or the number of
vacancieson the acceptor B, whichever value is limiting. l{(A)" is the electron
population of the isolated A atom. Examplesof predicted and actual atomic
populations for A are: NF, 6.56 (6.56);NO, 6.46 (6.50);CF, 5.21 (5.22);CO,
4.78 (4.65);CN, 4.94 (4.88);LiC, 2.12 (2.12).With the added stipulation thar
6.4 P R O P E R T I E SO F A T O M S I N M O L E C U L E S 199

the chargetransferredper vacancy,lX(A) - X(B)1,cannot exceedu' then all

diatomic fluoride populations are predicted to within a maximum error of
0.08 electrons.This method of predicting the charge transfer betweenatoms
yields significantly larger errors only for some compounds of the elementsBe
and B.
The charge distributions of the diatomic hydrides illustrate a general
phenomenon-that a significant degree of interatomic charge transfer is
accompanied by a polarization of the valence densities of the atoms in a
direction counter to that of the charge transfer. The polarizations are in
responseto the electricfield createdby the chargetransfer,the acceptoratom
polarizing towards the positively charged donor atom which is itself polar-
ized away from the negatively charged acceptor. This polaization of the
donor atom is particularly pronounced when it possesses a greaternumber of
valenceelectronsthan there are vacancies on the acceptor atom, as illustrated
by the first moments (eqn (6.45))for the atoms in the diatomic hydrides given
in Table 6.1. Also listed are the molecular dipole moments and their charge
transfer contributions, eqn (6.47).In general, the magnitude of the molecular
dipole is lessthan that of F. the charge transfer contribution, because of the
opposing atomic polarrzatrons.In some instances, the atomic polarizations
determinethe direction of the molecular moment. The polarizationsof Li and
Na in their hydrides are quite small as they correspond to tightly bound core
densities.For the second row, the atomic polarrzations are largest for the
diffuse valencedensity on Be and B. They are larger still for their third-row
congeners and Si where, becauseof the larger, l0-electron K-L core, the
valencedensity is less tightly bound and more polarizable.
Attempts to assign atomic charges on the basis of measured dipole
moments are unrealistic as such a procedure ignores the polarizations of the
atomic densities.Such an attempt corresponds to assuming the molecular
charge distribution to be composed of a set of spherically symmetric atomic
densities,each centred on its own nucleus,a physically unacceptablemodel
even in the limit of an ionic system.It should be evident from a comparison of
the charge distribution in the non-bonded region of the A atoms that the
reduction in magnitude or reversal in sign of the dipole moment, which
occurs after LiH in the second-row and after Na in the third-row, is a
consequenceof an atomi c polarization and is not indicative of a sudden
increasein the electronegativity of the A atom. The extent of the physical
distortion of those atoms for which the atomic polarizations are greatestis
reflectedin the valuesof their non-bonded radii. A non-bondedradiut r"(Q), is
defined as the axial distancefrom a nucleusto an outer contour of the charge
density on its non-bonded side. The 0.001- or 0.002-aucontour is chosen
since, as discussedin Section 6.4.1, the corresponding density envelopes
provide good approximations to the experimentally determined van der
Waals sizesand corresponding radii for moleculesin the gas or solid phase,
T H E M E C H A N I C SO F A N A T O M I N A M O L E C U L E 6,4

----

5CAIE(oul:
#
0 1 2 3 4 5
 (b)

N, ))

5CAIE(oul:
1#t
0 1 2 3 4 5

Frc. 6.3. Contour maps of the ground-stateelectronicchargedistributions for the second-and

thrid-row diatomic hydrides showing the positions of the interatomic surfaces"The first set of
diagrams (a) also includes a plot for the ground state of the H, molecule. The outer density
contour in these plots is 0.001 au. The remaining contours increase in value according to
Itg-, LiH tt *, BeH 2t*,
the scalegiven in the Appendix (Table A2). (a) The left-hand side H2
1 t + ; 2 n , N H 3 t - , O H 2 n , H F t t * . ( b ) T h e l e f t - h a n ds i d e :N a H ' I * ,
BH r i g h t - h a n ds i d eC H
3t-, sH 2n, HCl lt*.
M g H 2 t + , A I H 1 t * , S i H 2 f l ; r i g h t - h a n ds i d e :P H
202 T H E M E C H A N I C SO F A N A T O M I N A M O L E C U L E 6.4

Table 6.1
Dipole momentsoJ'diatomic hydrides*

q ( A ):
AH(x) z^ - N(A) M(A) M(H) P(CT) dA H ) r.(A )

LiH + 0.911 - 0.001 + 0 .3 8 7 - 2.747 - 2.361 1.88

BeH + 0.868 + 1.520 + 0.571 - 2.203 - 0 . 1l 2 4.68
BH + 0. 7 5 4 + 1.950 + 0.493 - 1.161 + 0.682 4.34
CH + 0.032 + 0.807 - 0 .t2 1 - 0.068 + 0.618 3.86
NH - 0.323 + 0.183 - 0_176 + 0.633 + 0.640 3.51
OH - 0.585 - 0.224 - 0 .1 4 8 + 1.073 + 0.701 3.23
FH - 0.760 - 0.449 - 0 .1 0 4 + 1.317 + 0.764 2.97
NaH + 0.810 + 0.017 + 0 .1 3 .3 - 2.890 - 2.739 2.64
MgH + 0. 7 9 6 + 1.701 + 0.302 - 2.602 - 0.599 5.18
AIH + 0. 8 2 5 + 2.274 + 0.360 - 2.568 + 0.066 4.97
SiH + 0.795 + 1.976 + 0.428 - 2.285 + 0.119 4.49
PH + 0.579 + 1 .4 6 4 + 0.317 - 1.569 + 0.212 4.16
SH + 0.094 + 0 .5 8 7 - 0.009 - 0.239 + 0.339 3.86
CIH - 0.241 - 0.006 - 0.103 + 0.581 + 0.471 3.68
*A negative value forp implies the directron A+H for the dipole. The units are atomic units;
I a u : 2 . 5 4 2 D . T h e n o n b o n d e d r a d i u s r " ( A ) i s t o t h e 0 . 0 0 1a u c o n t o u r o f p .

respectively.A bondedradius,rr(O), of an atom is correspondinglydefined as

the distancefrom its nucleusto the associated(3, - 1) or bond critical point.
The non-bonded radii for Li and Na are close to the values for the
corresponding singly-chargedions while those for the strongly back-polar-
ized atoms are all considerablygreaterthan are their valuesin the free atomic
state(Table 6.1).(The 0.001-and 0.002-auradii of the free atoms are given in
the Appendix (Table A5).)
The presenceof such a large and diffuse(weakly bound) chargedistribution
has important chemical consequences, imparting to the molecule the charac-
teristicsof a strong Lewis base.A classicexample of this behaviour is the
carbon atom in the CO molecule (Fig. 6.a). This molecule has a near-zero
dipole moment because of very pronounced polarizations of the atomic
densities,particularly that of carbon, which oppose the considerablecharge
transfer moment. The charge on oxygen is - 1.33e and the magnitudes of
the opposingatomic dipolesare lM(O)l : 0.98au and lM(C)l : 1.72au with
the non-bondedradius on carbon exceedingits free atomic value by 0.15au.
(Propertiesof the atoms and the bonds in the moleculesshown in Fig. 6.4 are
given in Table 7.8.)The physical importance of the atomic polarization of
carbon is reflectedin the ability of CO to act as a Lewis base,particularly in
6.4 O F A T O I \ { SI \ \ I O L E C U L E S
PROPERTIES 203

the formation of metal carbonyls.The considerabledifferencein the electro-

negativitiesof C and O is reflectedin the relatirelv large dipole moment,
lpl:1.11 au,of the formaldehyde . , C : O ( F i g . 6 . 4 ) .T h e c h a r g e
m o l e c u l eH
transferfrom c to o in formaldehydewhere4(o) : - l'24 e, is only slightly
lessthan it is in CO. Unlike CO, however,the chargetransfercontribution
dominatesthe final moment in formaldehydebecauseof the closeto halving
of the atomic dipole on carbon which resultsfrom the use of its non-bonded
density in the formation of non-polar bonds to the hydrogen atoms
(Ta ble 7. 8) .
The loss of the non-bondedchargeon the carbon atom in CO that occurs
in the formation of HrCO and the accompanyingdramatic reduction in its
atomic moment is illustrated by the charge distributions of Fig. 6.4. Also
illustrated in Fig. 6.4 is the progressionof changesin the atomic charge
distributions which occur through the series OC -- COr-- SCO -- CSz
-- CS. The chargeon the oxygenatom in CO, COr, and OCS variesby only
t 0.03e about the mean value of - 1.30e. When carbon is bonded only to
oxygen, its charge loss per oxygen is correspondinglyconstant. Note the
similarity in the distribution of chargedensity over the oxygen atom and over
the bonded portion of the carbon atom (up to a planar surfacethrough the C
nucleus,perpendicular to the CO axis) in all of the bonds between C and O
shown in Fig. 6.4, even in OCS where the net charge on C is considerably
reduced becauseof a transfer of charge from sulphur. As a consequence,the
C-O interatomic surfaceexhibits the same characteristicshape in all of the
moleculesand the bonded radii changeby only small amounts (Table 7.8).A
C:O bond in thesemoleculesis relatively insensitivein its gross featuresto
the neighbouringbond, or lack of it, on carbon.
The electronegativity differencebetween C and S is less than between O
and C and the distribution of charge in a C-S bond shows a wider variation
in its properties.The net chargeon C in CS, is not twice the magnitude found
in CS, as it is in the oxygen analogues.
The contribution to an atom's population from orbitals of zr symmetry or
pseudo z symmetry found in systemswith axial or planar symmetry, respect-
ively, is easily determined by the separateintegration of their corresponding
densities over an atomic basin (Wiberg and Wendolowski 1981). The
resulting o and z populations are frequently found to undergo opposing
polarizations. Examples of this in a familiar context are provided by the
Hrickel z populations predicted for a system of conjugated double bonds as
compared to the total atomic populations as determinedby theory. The z
populationsof the carbon atoms in the allyl and pentadienylcations,starting
from a terminal carbon atom, are 0.48 and 0.97 in allyl and 0.63, 1.00,and
0.61 for pentadienyl.These values are very similar to the predicted Htckel
populationsof I and 1 for allyl and 213,1,and 213fot pentadienyl.However,
the atoms with the smallest z populations bear the smallest net positive
204 THE MECHANICSOF A\ ATO\I IN A MOLECULE 6.4

:-i

';,,1:/';-'r'
. - /i':a

oco

SCS
6.4 P R O P E R T I E SO F A T O M S I N N I O L E C U L E S 205

chargesbecauseof an opposing polarrzationof the o density.The net charges

on the atoms in the sameorder as the n populationsgiven above are + 0.09
and +0.22 for allyl and + 0.09, + 0.16, and + 0.01 for pentadienyl(see
Fig. 6.8).Thus atomic net chargescannot be assignedon the basisof Htickel z
populationsas is done in models basedon an assumedrelationshipbetween
13C chemical shift observed in
ihr n density at a carbon nucleus and the
nuclear magnetic resonancespectra.
Polarization of the n density is of particular importance when an un-
saturated systemis bonded to a n electron donor or abstracter group (Libit
and Hoffm ann 1974:Sleeer al. 1988).This is illustrated by the polarization of
the z density of the carbonyl group as causedby a substituent X bonded to
carbon of the carbonyl group in the substitutedformyl compounds HXC:O.
A n donating group causesthe z population of the oxygen atom, rather than
that of the carbon atom, to increasewhile a z withdrawing group causesthe n
- H. This is illustrated
population of oxygen to decreaserelative to that for X
in Fig. 6.5 which plots the z population on oxygen versus the resonance
parametero$ of Taft (1956,1960),which provides an empirical measureof the
relative n-donating (oR < 0), z-abstracting (ofl > 0) ability of a group'
Examples of opposing polarizations of the o and z density distributions
which result in an alternation in the corresponding atomic populations are
provided by the substituted benzenes(Bader and Chang 1989a).This is
illustrated in Fig. 6.6 which gives the changesin the n and o chargeson the
carbon atoms of the phenyl group @relative to their values in benzene.The
amino group is z-donating and, as in the substitutedcarbonyls,the polariza-
tion of the z density in the bonds to the carbon atom bearing the substituent
causesits z populations to decreaseand that of the atoms attached to it, the
ortho carbon atoms, to increase, i.e. the z charge q"(C) becomes more
negative. This effect alternates around the ring producing the pattern of
chargeincreaseat ortho and para carbon atoms and its decreaseat the meta
atoms characteristic of ortho-para directing groups. The o populations
changein just the opposite way with the o charge becoming more negativeat
the meta carbon atoms. The nitro group is z electron withdrawing and, thus,
the n population of the carbon bearing the substituent is increasedand the
accompanying alternation causescorresponding increasesat the meta posi-
tion and decreasesat the ortho and para positions,as is characteristicof meta
directing groups. The amino group increasesthe z population of the phenyl
group by 0.084e and activates the ring towards electrophilic aromatic
substitution relative to benzene, while the nitro group decreasesthe 7t
population by 0.099e and deactivatesthe ring.

Frc. 6.4. Contour maps of the ground-state electronic charge distributions for CO. H2CO,
CO2, SCO and CSr, CS, and H2CS showing the positions of the interatomic surfaces.The outer
density contour in theseplots is 0.001au. The remaining contours increasein value according to
the scale given in the Appendix (Table A2).
206 T H E M E C H A N I C SO F A N A T O M I N A M O L E C U L E 6.4

NHz
1. 6 9
1.68
t.o/

t.oo

1.65
1.64
1.63
1.62
1.61
1.60
{ 1.59
1.58
1.57
1.56
1.55
1.54
1.53
t.3z

1.51
-0.5 -0.1 0.1
S i g m aR O

Ftc. 6.5. The contribution to the atomic population on oxygen from the orbitals of r
symmetry, the population lV"(O), plotted versus the Taft's oRo parameter for a series of
substituted formvl derivatives.XHC:O: Me = methyl. Vi = vinyl.

FIc. 6.6. Molecular graphs showing the bond critical points, calculated from theoretically
determined charge densities for aniline and nrtrobenzene.The numbers to the right of each
structure give the changesin the zrcontribution to the chargeon carbon relativeto their valuesin
benzene,Aq,(C), while those to the leit of each structure give the changesin the o contribution to
t h e c h a r g eo n c a r b o n r e l a t i r e t o t h e i r v a l u e si n b e n z e n eA
, q"(C).
 6.4 OF ATOMSIN MOLECULES
PROPERTIES 207

The z-donating/withdrawingability of a substituentX in the substituted

phenyls @-X is found to be the same as that observedin the substituted
formyl derivatives,XHC:O. The effect of most of the same group of
substituentson the chargedistribution and moment of the ethyl group has
also been studied (Slee et al. 1988). The ordering of the total charge
withdrawal by X is found to be the same in all three seriesof molecules.
Information correspondingto the z populations of the orbital model is
recovered in the quadrupole polarization of the atomic charge densities,a
property of the total charge density. A quadrupolar polarization of the
densityalong the z-axis,eqn (6.48),has the form of a d,' orbital, a removal of
chargefrom a plane and its concentration in an axial direction perpendicular
to the plane ( F ig. 6.1 ) fo r Q " " 1 0 . In b e n z e n ea n d ethyl ene,w i th the z-axi s
perpendicularto the plane containing the nuclei, Q,,(C): - 3.34 and :
- 3.38 au, respectively,correspondingto the presenceof a single z electron.
With z taken as the internuclearaxis in acetylene,Q",(C) is large and positive,
equal to + 4.14au correspondingto a torus-likeconcentrationof z density
about the z-axis as reflectedin the negativevalues for Q*,(C): Qrr(C):
- 2.07 au. In the planar methyl cation molecule with a nearly vacant p,,
orbital, Q""(C): + 1.22au and the carbon atom density appears as an
oblate spheroid( Fig 6.1)with Q"" ) 0. This orbital vacancyis partly filled in
the tertiary-butyl cation by inductive and hyperconjugativeelectron release
from the methyl groups and the moment Q""(C) is correspondingly reduced
to + 0.43au in this molecule.
Figure 6.7 illustrates the closecorrespondencebetweenthe polarization of
.1
an atomic densityas measuredby Q""(Q)and its n population by plotting the

o z

o l a

o t 6

o 1 4

o 1 2

o l

ooa

7 \ O 0 6

L O O a

f o.oz
o
-o.o2
*o 04
-o. o6
2 -1 -OA -O.5 -O.4 -O2 0 02

Q=rtCl

poro
FIc. 6.7. A plot of the z population of the para carbon atom in substituted benzenes,N"(C),
v e r s u si t s z c o m p o n e n t of the quadrupole moment, Q""(C).
 T H E M E C H A N I C S O F A N A T O I \ {I N A M O L E C U L E 6.4

z population of the para carbon atom in a seriesof substituted benzenes

versus its quadrupolar polarization along the z-axis. Such a near linear
relationshipis obtainedfor both the ortho and para positions,but the values
for the meta position are not so well correlated(Bader and Chang 1989).A
fragment of the benzenium ion (Fig. 6.8), the protonated o-complex of
benzene and an example of the arenium ion intermediate occurring in
electrophilic aromatic substitution, is thought to resemble the pentadienyl
cation. The benzeniumion intermediateis not planar and does not possess

C1-C2 C1
fl=1.54
€=0.141 Q = 0O tg
Q77;-1a66
V2P=-t.ooa

c2-c3 C2
l''l=1.86 Q=O'1+g
e=o'277 o+t
Qzz=-s
Y2P=-'t1et
C3
Q = 0O t g
Qzz=t'azg

AE(C1)=*eo A E(C 2 )= -3 0 4f(Q$)= -rae

C1-C2
fl=1.54 C1
€=0.136 Q=o'oto
VzP=-t.ozg Q72=-1.308

C2-C3 c
n=1.99 Q=O'16+
€=O'282 Qzz=-s'sa
VzP=-t.tsz
C3
Q = 00 9 9
Qzz=-tzgo

Flc. 6.8. A comparison of the C-C bond and atomic properties of the carbon atoms in the
pentadienylcation with the correspondingfragment (asindicated by the numbering of the atoms)
in the benzenium ion, protonated benzene.The bond properties compared are: bond order n,
bond ellipticity e, and the Laplacian at the bond critical point, V2po. The atomic properties
compared are; the net chargeson the carbons q(C) and their quadrupole moments Q",(C). Also
given are the differencesin energy of rhe carbon aroms, AE(C) : E(C)[CoHi] - E(C)tC5Hil
in kJ mol.
6.4 P R O P E R T I E SO F A T O ] \ { SI \ \ I O L E C U L E S 209

'n' symmetry.However, one ma)' still compare the quadrupolar

orbitals of
polarizations of corresponding carbon atoms in the tu'o molecules.A com-
parison of propertiesof correspondingatoms and bonds in thesetwo cations
is summarrzedin Fig. 6.8 to illustratehow the theory'enablesone to obtain a
quantitativedeterminationof the degreeof similarity in a grouping of atoms
as it occurs in two different moleculesand. in doing so, to provide the
necessarybasefor understandingthe observeddifferences.

6.4.2 Transferabilityof atomic properties

The discussionaccompanyingFig. 5.1 emphasizedthat the use of the zero-

flux surface for the definition of an atom or functional grouping of atoms
maximizesthe extentof the transferabilityof its propertiesbetweensystems,a
characteristicessentialto the role of the atomic concept in chemistry.The
idea illustrated above. that the benzeniumion can be consideredto be the
union of a pentadienyland a methylenegroup by basingthe group recogni-
tion on the comparisonof its propertiesin differentsystems,is an expression
of this concept. By defining a group and its properties and enabling one to
determine the effect of its presenceon the properties of another group, the
theory of atoms in moleculesparallelsthe most important of all chemical
codifiers,the substituenteffect.Studiesreferredto above have illustrated the
ability of the theory to quantify and make understandablethe effectsof a
range of substituents on the properties of the ethyl, formyl, and phenyl
groups.It is the purposeof the presentdiscussionto explorethe limiting case
where a group is transferablewith little or no changein its propertiesand,
correspondingly,its perturbation of the remainder of the systemis minimized.
It is possible to experimentally measure the energy of an atom in a
molecule as an additive contribution to the molecule'sheat of formation in
those instanceswhere a class of moleculesexhibits an additivity schemefor
the energy.Essentialto the theoretical prediction and understandingof this
experimentalobservationis the property of the atoms expressedin eqn (6.75),
that their properties, including their energies,be additive to yield the total
property value for a molecule.It is demonstratedhere that the energiesof the
methyl and methylenegroups, as definedby theory, predict the additiuity and
transferabilityof the group energiesas is observedexperimentally in normal
hydrocarbon molecules.Their properties also predict and account for the
deviations in this additivity scheme that are observed for small cyclic
molecules,deviations which serve to define the strain energy. The ultimate
test of any theory is its ability to predict what can be experimentally
measured.By appealing to the limiting caseof near transferability of atomic
properties, one can demonstrate that the atoms of theory are the atoms of
chemistry(Bader 1986;Bader er al. 1987b;Wiberg et al. 1987).
110 T H E M E C H A N I C SO F A N A T O M I N A M O L E C U L E 6.4

The study of the molar volumesof the normal hydrocarbonsprovided the

earliest example of the additivity of group properties (Kopp 1855).The
experimentally determined heats of formation for the same homologous
seriesof molecules,CH.(CHr).CH., also obey a group additivity scheme
(Franklin 1949;Prosen 1946;Benson 1968;Pittam and Pilcher 1972\.lt is
possible to fit the experimental heats of formation for this series,beginning
with m: 0, with the expression,

A H r o ( 2 9 :82)A * m B , (6.8e)
where ,4 is the contribution from the methyl group and B that from the
methylene group. The generally accepted values for A and B at 25"C are
- 42.34 and - 20.63kJ/mol, respectively.The group enthalpy corrections
from 298 to 0 K are additive for the n-alkanes(Wiberg 1984),as are the group
zero-point energy corrections(Schulmanand Disch 1985).Thus, the calcu-
lated energiesof the vibrationlessmoleculesin their equilibrium geometries
should exhibit the sameadditivity of the energyas representedby eqn (6.89)
and the additivity is indeed mirrored by the single determinantal SCF
energiesat both the 6-316*16-31G* and 6-3lG**16-31G* levelsof approx-
imation. (The notation implies that the SCF calculationswere performed
using the basisset listed first, at the optimized geometry for the basisset listed
second; Hariharan and Pople 1972; Krishnan et al. 1980.)The calculated
molecular energiesE for the n-alkanessatisfythe relationship,

E :28(CH.) + mE(CHr). (6.e0)

The quantity E(CH.) is one-half the energy of ethane, equal to

- 39.61912au, and E(CH') is the energy increment per methylene group
equal to - 39.03779au, using the 6 tt6** 16-3lG* calculatedresults.These
group values fit the calculated energiesto within + 0.00014au, an average
deviation smaller than the experimental one. The calculated results indicate
that the corrections to the energy arising from the correlation of the
electronic motions, a contribution neglectedin a single determinantal calcu-
lation, should also obey a group additivity scheme.This indeed appearsto be
the caseas is demonstrated and discussedlater.
The distributions of charge for molecules in this seriesare illustrated in
Fig. 1.1in terms of an outer envelopeof the chargedensity p and, specifically
for the five- and six-carbon members,in Fig. 6.9 in the form of contour maps
of p. The latter maps show the bond paths linking the nuclei and indicate the
intersection of the interatomic surfaceswith the plane of the diagram. The
intersection of these same surfaceswith the density envelopesare shown in
Fig. 1.1 and they define the methyl and methylene groups as envisagedby
chemistsand as definedby theory.The diagramsshow qualitativelywhat the
atomic properties will demonstrate quantitatively, that the methyl and
 .'; --'-

.:. --

-t\

.. -ti.

,{'.1

- -'...4
g'
...r'

ri;;i
.i:.:-.'

Frc. 6.9. Contour maps of the charge distributions for pentane and hexane.The outer contour
in these plots is 0.001au. The intersectionsof the interatomic surfaceswith the plane of the
diagram are indicated, as are the bond paths. The upper and lower diagrams are in the plane
containing the carbon nuclei and a terminal proton of each methyl group. The middle diagrams
are for thi central methylene group in pentane (on the left) and for one of the two equivalent
central methylenegrorpi in hexane(on the right) in the plane containing the carbon nucleusand
the two protbns. The pbsitions of out-of-plane nuclei are indicated by open crosses.The methyl
groups, ihe methylenegroupr bonded to a methyl group, and the central methl lene groups l ield
superimposablepairs of charge distributtons.
tll T H E M E C H A N I C SO F A N A T O M I N A M O L E C U L E 6.4

Table 6.2
Relutire propertiesoJ methyl and methl'leneqroups in normal hydrocarbons
(Ba tl e re t a l . 1 9 87)

Molecule Methyl group* IVlethylenegroup'

ar\/(cH3) AE(CH,) A rv(cH2

) A E ( C H 2)
(kJr'mol) (kJimol)

Ethane 0.000 0.0

Propane + 0.017 - 45.6 - 0.034 + 90.8
Butane + 0.018 - 45.2 - 0.01 8 + 45.2
Pentane + 0.017 - 4 1 .4 - 0 . 0 l t t + 0.001 + 45.6 + 0.0
Hexane + 0.018 - 41.4 - 0.01 8 + 0.000 + 44.4 - 4.6

* D i f f e r e n c e sr e l a t i v et o m e t h y l g r o u p i n e t h a n e .
r D i f f e r e n c e sr e l a t i v e t o s t a n d a r d m e t h y l e n eg r o u p . T h e s e c o n d s e t o f e n t r i e s b e g i n n i n g w i t h
p e n t a n ea r e f o r m e t h y l e n eg r o u p s b o n d e d o n l y t r : o t h e r m e t h y l e n e s .

methylenegroups in this seriesof moleculesare transferablcwith little change

in their form and, hence,with little change in their properties.
The physical properties of the n-alkanesindicate that the moleculesare
non-polar and this is reflectedin the small magnitudesof the net chargesof
the carbon and hydrogenatoms and of the molecularmoments.Hydrogen is
slightlymore electronegative than carbon in saturatedhydrocarbons,and the
order of group electron-withdrawingability in hydrocarbonswithout geo-
metric strain is H > CH. > CH, > CH > C. In ethane,methyl is bonded to
methyl, while in the other molecules of the n-alkanes it is bonded to
methylene from which it withdraws charge. Table 6.2 lists the populations
and energiesof the methyl groups for m > 0 relative to their valuesin ethane.
To within the accuracyof the numerical integrations of the atomic properties
(which in general are + 0.001e and + 4.0 kJ/mol), the energy and popula-
tion of the methyl group are seen to be constant when it is bonded to
a methylene group. Thus the methyl group is essentially the same in all
the members of the homologous series past ethane. This transferable
methyl group is more stable relative to methyl in ethane by an amount
LE : - 43.4+ 2.0 kJ/mol, and its electron population is greater by an
amount Alf : 0.0175e.
The charge and energy gained by the methyl group is taken from the
methylene group. What is remarkable, and what accounts for the additivity
observed in this seriesof molecules,is that the energy gained by methyl is
equal to the energy lost by methylene.In propane, where the methylene
group transfers charge to two methyl groups, its energy relative to the
6.4 OF ATO\{SI\ \IOLECULES
PROPERTIES 2t3

incrementin eqn (6.90)is E(CHr) - 2LE. and its net chargeis (necessarily)
+ 2L,l\rwhere AE and Al{ are the quantitresdefined above for the methyl
group. Thus the energyas well as the chargeis consert'ed relativeto the group
energiesdefinedin eqn (6.90).In butane, a methl'lene group is bonded to a
singlemethyl group and, correspondingly,its energyis E(CHr) - AE and its
net charge is + AA/. The correspondingmethylene groups in pentaneand
hexane,those bonded to a singlemethyl grup, have the sameenergiesand net
chargesas a methylenegroup in butane.Thus, the chargetranserto methyl is
damped by a single methylene group, and the central methylene group in
pentane and the two such groups in hexane(seeFig. 6.9) should have a zero
net charge and an energy equal to the increment AE(CH,). This is what is
found to within the uncertaintiesin the integrated values,their calculatednet
chargesbeing 0.0005e * 0.0002e and the maximum deviation in the energy
from the standard value being within the integration error of = 4 kJ/mol or
1 kcal/mol. (It should be borne in mind that the total energy of a methylene
group is 10.25x 104kJ/mol.) Therefore,methylenegroups bonded only to
other methylenes,as found in pentane,hexane,and all succeedingmembersof
the series,possessa zero net charge and contribute the standard increment
E(CH) to the total energy of the molecule. The underlying reason for the
observation of additivity in this seriesof moleculesis the fact that the change
in energy for a changein population, the quantity LEILN, is the same for
both the methyl and methylene groups. The small amount of charge shifted
from methylene to methyl makes the same contribution to the total
energy.
It is to be emphasized that the energies assigned to the methyl and
methylene groups are independently determined by the theory of atoms in
molecules(eqn(6.72)).The fact that this assignmentleadsto an energyfor the
transferable methylene group equal to the value E(CH2) in eqn (6.90),an
equation which mirrors the experimental additivity of the energy eqn (6.89),
confirms that the theoretically defined atoms are responsiblefor the experi-
mentally measured increments to the heat of formation, and that quantum
mechanicspredicts the properties of atoms in moleculesjust as it does the
propertiesof the total molecule.It is a straightforward matter to usequantum
mechanicsto relate a spectroscopicallydeterminedenergy to the theoretically
defined differencein energy between two statesof a system.In a less direct,
but no lessrigorous manner, quantum mechanicsalso relatesthe differencein
the experimentallydetermined heats of formation of butane and pentane to
the corresponding theoretical/ydefined energy of the methylene group.
The additivity of the energy in the ll-alkanes is obtained in spite of small
differencesin group properties, differenceswhich necessarilyresult from a
change in the nature of the bonded neighbour. Thus, there are two kinds of
methyl groups: the one unique to ethane and the transferablemethyl group
which is bonded to a methylene group. There are three kinds of methylene
 l 1 ' l
- l a T H E M E C H A N I C SO F A N A T O M I N A M O L E C U L E 6.4

groups:the one unique to propane and two transferableforms, one bonded to

a methyl, the other bonded only to other methylenegroups.Other properties
of these groups exhibit the same pattern of transferablevalues as do their
energies and populations. This is illustrated by the properties listed in
Table 6.3. The first moments for the methyl and methylene groups are
calculated as for an atom, eqn (6.45),with the integration extending over the
basins of all the atoms in the group and with the carbon nucleus as origin.
The magnitude of the moment for the methyl group is unique for ethane but
nearly constant for the remaining molecules.The methylene groups exhibit
two transferablevalues,one for the methyleneattachedto a singlemethyl and
another, slightly smaller,for a methyleneattached only to other methylenes.
In all casesthe hydrogen atoms form the negativeend of the group moment.
The directions of these moments are also transferable.The moment of the
methyl group is parallel to the bond path linking it to the neighbouring
carbon nucleus to within 0.5" through the series.The direction of the
methylenemoment, in those caseswhere it is not dictated by symmetry, lies
0. 05"of f t he ax i sb i s e c ti n gth e H C H a n g l ei n pentaneand 1.1' and 1.2' offthi s
axis in butane and hexane,respectively.Thesegroup moments along with the
group chargescan be used,as are the atomic valuesin eqn (6.46),to determine
the total dipole moment of a hydrocarbon molecule.
The volumes of the methyl and methylenegroups reflect the same pattern
of transferablevalues.The volume of the methyl group in ethane is slightly
larger than that of the methyl groups in the succeedingmembersof the series
whose volumes vary by only * 0.02 au from their averagevalue.The volume
of the methylenegroup bonded to two methyls in propane is slightly greater
than the transferable volume of a methylene bonded to one methyl. The
methylenegroup bonded only to other methylenesalso possesses a character-
istic volume.
Also listed in Table 6.3 are the group values for the electronic correlation
energy calculated from the density functional expressionof Langreth et al.
(1983;Hu and Langreth 1985).It is a complicatedfunctional of the charge
density p(r), involving powers and gradients of p(r). The values of this
quantity for the methyl and methylene groups listed under the heading e
exhibit the same pattern of values as do the other group properties. The
values of e for the transferable methyl and methylene groups exhibit vari-
ations which are remarkably small, particularly when one bears in mind tlle
complicated form of the functional. While no functional for the correlation
energy which yields values of chemical accuracy yet exists, these results
indicate that each of the transferablemethyl and methylene groups should
make a characteristic and essentially constant contribution to the total
correlation energy of a normal hydrocarbon molecule,a result anticipated on
the basis of the ability of SCF calculations to recover the experimental
additivity of the energy.
6.4 OF ATOMSI\ \lOLECULES
PROPERTIES 215

aa r-,

..+
d ^
cc a] cc c\

>ll
+++.+
F..-
oo
o\
oO t--
__i .a aa
R c'l c'.1
ca (.a
b A A
q)

\- tt l
l

tt
R
o
M
oA o- o- o o
- a'l (.'l (.'l
q a-) a') ai a.r L
,{ -i -i .i o
-
tF
t t l l
L
R
U
- tr
L
t oo S O
(.) .a ca
A A
N
!
>|
(-) F t r ) O \ O
L. (-) F =
- A -
aa ca
-
aa c')
? t S Z c
\ O t s
q ) . E oo
0)
-Gl s
- = O

\
O
tr E
l - - c a O - $
\ \ \ i c : €
- - - - -
A A A A A

z 9 q)
L

\O $ \O \O ta)
R Lr) co oo oo oo
$ . f , $ $ $ e Q ?
aa .a ao ri aa
. ^ A A A l : .: c0 :
> . - e
- = > <
+ d
t
t
r
t t
t l
t l
o l ' o )
tr1?>.
= a a

aoo) c)
q
L
-_ x ( ) v
" l c
bo
s
L >'
- Q .a $ .a O\ ta)
^ (.) F :
r a $ $ $
..'t c..l c..l
s
al (^.n ( ) L -
L. . d -
2 A c 6 . { A - i A A

a > q )

L Y 9

0 . ) q . ; ) ( J d d

()
q)
0 0 ) C o . ) c Y
- 6 )
X -
t
. v d
a = g

! : f g E ; o ' Y g

= ; = 6 c )
r - T A A A F | .
g - H B - * + *
216 T H E M E C H A N I C SO F A N A T O \ I I } * A M O L E C U L E 6.4

Table 6.4 lists the individual contributions to the potential energies,as

definedand discussedin eqn (6.86),of the carbon and hydrogenatoms of the
methyl groups in propane to hexane,all relative to their valuesin hexane.
Thesedata illustrate a very important property of an atom-that it responds
only to the total forceexertedon it and not to the individual contributionsto
this force. If it were not for this property of the atom and its energy being
determinedby some net field (the virial of the Ehrenfestforce,eqns (6.30)and
(6.78))rather than by the individual contributions to this field, contributions
which as illustrated here differ markedly even for members of a homologous
series,there would be no chemically recognizableatoms or functional groups.
The remarkable constancy in the total properties of these atoms in the
differentmoleculesis reflectedin the valuesof their populations l{(O), of their
first and second moments r(Q) and r2(O) as defined in eqn (6.44),and their
electronic kinetic energiesf(O). If the distribution of charge is the same for
each type of atom, then the quantity V?,(Q), the potential energy of inter-
action of the nucleus of atom O with its charge distribution (eqn (6.81)),
should also be the same.This is indeed the case,the largest differencebeing
l.TkJlmol for a hydrogen atom and 12.1kJ/mol for a carbon atom. The
interaction of all the nuclei in the moleculewith the chargedensity in atom O,
the quantity V^"(O) defined in eqn (6.80),increasesby very large amounts
with the removal of each CH, group, changing by 6.3 x 103kJ/mol for H
and by 38.5 x 103kJ/mol for C between the first and final members of the
series.The electron--electronand nuclear-nuclear repulsion energies,V""(O)
and Vn,(Q),respectively,decreaseby large amounts with the removal of each
methylenegroup. The sum of the three contributions to give { (O), the total
potential energy of the atom, is, however, the same for each kind of atom
through the series.This must be, since'f (q is equal to twice the total energy
of the atom by the virial theorem, eqn (6.73).It is the virial of all the forces
acting on atom Q. This quantity, along with the kinetic energy of the atom, is
conservedwhen the atom's distribution of chargeremains unchanged.It must
be consideredremarkable that a methyl group with a total energyin excessof
100000kJ/mol, can be transferred between molecules-in reality and in
theory-with changesin its energyof approximately4 kJ/mol. It is still more
remarkablewhen it is recalledthat the individual contributionsto the energy
of a carbon atom change by 8000 to 20 000 kJ/mol between members of the
series.
It has been demonstratedthat atoms are indeed piecesof matter in real
spacewith the propertiesanticipatedin Section 1.2.A consequenceof the
necessityof two identical piecesof matter possessing the samepropertiesas
statedthere,is that one may use symmetryto definethe energyof groupings
of atoms.The energyof an ethyl group is for example,one-halfthe energyof
n-butaneas demonstratedabove.Similarly,the energyof the 2-methylpropyl
group will be given by one-halfthe energyof 2,5-dimethylhexane.Using the
 6.4 P R O P E R T I E SO F A T O I U SI \ \IOLECL'LES 211

lrl ...r x
c.l ci rf', x
V' v
V-)
a.l V) ca ca a I :C :a r,
-
=J -, aC -, \
^ - - c i r , 3 - 3 3 " i v . - - - .
- l
r r r + r
\
. a 3 . .
Vt,.) $
-
rf'' -\a,
"1 oqn-: c.-: n: i.n:*n F
W' \O O ,c V r r- -. t ri 3 .rl
+ t^ C .r F- r, .l r- : --
tt'r t-.-\O, r- -L- :{-r.
co - . , 1 - - -
o\ E
\' l l l
J
*J
R o l O
s t r - t r )
s € , v .
ci>os \o..oc6t .c..f

L
o 5 a = o r s S 5 ? Xl * i
t--\coc t-)c- r9-a 0)
F
r - a l V - -
t , l
l l l l l
X
0)
tf
o s l
tro
- -a-,
OC
It- ca
R V ) O O O \ t t - - @ a l *
. a * . a *
x o\J.do,
O . - *
i- .Or. ;<i ti - od.'ic+
. o \ O c O -
(.)
i + . a € v ) l - - c a - l t \ . i r ,
- . a r / ) l - r , . C C \ ^ l J C
I a.l aa L

(c
-
ta) 6
f l al 6

o 9 S
..1 .a ca
K
\.) ca $ -
ca C-- a- C) $ a,)
Ji-: -ici ..jJ o.-jcc.i
\o aJ
i

-9 .: o ; | | I + I | + T * t l -o
\
I
I L

r- a )L u
lrl
o c v
I \ O F - O
. o a { s
U .]..r--$ \o$$c. I--S$a-
9 - l J o i c > i - i S . r ; i - j E
o + + | "
q
l + l + + E
F-

U
€
L E

(\'
q
()
H
L bo

- ': o
i ) 5< 5
) <5 ) - < = , = = ; < J ^ ^ +
.- - - - c)

L-
: J O i O
o 6 4 d
- j O O i
o 6 5 5
, , " Oi q > O
4,

z l l - T ()
x

q)

()
C) ? t y i : 4 2 ?z ? z i
i c t r ! : s c
a
6 ) ! v
a

$ g ; F; g ; xi E ; ; q ) ^

F ^ -
I t L
2r8 T H E M E C H A N I C SO F A N A T O M I N A M O L E C U L E 6.4

demonstratedadditivity of the zero point energiesand contributions to the

heatsof formation from 0 to 298 K, togetherwith the observationthat minor
perturbationsin saturatedhydrocarbonsare damped by a singlemethylene
group, one can determine the heats of formation of thesegroups in the same
manner and predict that their sum should equal the heat of formation of 2-
methyl pentane.The value of - 173.6kJ/mol predictedin this manner equals
the experimentalvalue (Keith 1990).
Theory predicts what can be observed and performed in the laboratory.
Therefore,it should be possible to use theory to construct a molecule in the
same manner that this is done in the laboratory. Chemists do not begin a
synthesisstarting with separatebeakersof nuclei and electrons,correspond-
ing to the theoretical description of a molecule starting from the
Hamiltonian, but rather from piecesof other molecules.The first requirement
in developing a theoretical parallel to the experimental synthesis is the
identification of the atoms and functional groups of chemistry,coupled with
the definition of their properties, for it is these groups and their properties
that are manipulatedand linked togetherin a synthesis.The secondrequire-
ment is that one be able to predict the changesin the properties of a group
when it is transferredfrom one molecule to another. The first requirement is
fulfilled by the theory of atoms in molecules,the initial step in the develop-
ment of the quantum mechanicsof an open system.The secondrequirement
will be met with the development of the requisite perturbation theory for the
quantum mechanicsof an open system.
The synthesis of a molecule and a prediction of its properties from
theoretically defined atoms in real space is already possible in those cases
where, as exemplified above for the normal hydrocarbons, groups can be
transferred between moleculeswith negligible changesin their distributions
of charge and hence in their properties. One can take advantage of the
observation that functional groups frequently appear to undergo only minor
perturbations when transferred between corresponding environments, as
occurs in the repeatingunits of a peptide,for example.It should be possiblein
such casesto construct a molecule from the atoms of theory and predict its
properties with acceptably small errors. This has been demonstrated by the
synthesis of the simplest of dipeptides glycylglycine,from the appropriate
fragments of formylglycine and glycylamine (Chang and Bader 1990).The
glycyl groups are defined by cutting glycylamine and formylglycine along
their C-N interatomic surfaces as indicated by the vertical bars in the
formulae H2NCH2C(:O)|NH , and HC(:O)INHCH2CO,H, and the dipep-
tide is synthesizedby joining the free sidesof the interatomic surfacesof the C
and the N atoms to yield H2NCH2C(:O)INHCH2CO2H. The surfacesare
joined and the geometry of the dipeptide is determined by the anti-parallel
linking of the bond path vectorsfrom the C and the N nuclei to the (3, - 1)
critical points of the C-N bonds in the reactantgroups.This resultsin a near
6.4 P R O P E R T I E SO F A T O M S I \ \IOLECL'LES 219

perfect matching of the two reactant C N interatomic surfacesand in a

predictedC-N bond length which is in error b1 onll'0.0014. fnit construc-
tion yields a molecular graph that is superimposableon the graph for the
dipeptide, correctly predicting that the geometrical parameters and the
charge distributions of the fragments remain almost unchanged when com-
bined to form the dipeptide. As a consequence,the atomic properties,
including the net charges, moments and energies of the atoms, can be
transferred with only small error in this synthesis.The total energy of the
dipeptide predicted by summing the energiesof the atoms in both fragments
isln-error by only 12kJlmoL In addition, the nonbonded shapeor van der
Waals envelopeof the dipeptide as determinedby the 0.001 au envelopeof the
"reactive surface", a
charge density is also accurately reproduced, as is its
q,runtity determined by the Laplacian of the electronic charge distribution
and discussedin Section7.3.2.
The functional groupings which serve as the building blocks in such a
theoretical synthesiscan be determined using relatively large sets of basis
functions, with or without correlative corrections.Accessto a bank of these
theoretically defined building blocks would enable one to synthesize a
macromoleculeand obtain operationally useful predictions of its properties.
Equally if not more important, is the ability to use the same method of
rynttreils to predict the propertiesof just that portion of a macromolecule
that is of interest by terminating the synthesisat the interatomic surfaces
which define the relevant fragment. With further development of the quan-
tum mechanics of an open system, these building blocks would serve to
provide the zeroth-order approximation to the system of interest, theory
enabling one to take into account the perturbations arising from their
coupling in a new environment. It is the goal of the theory of atoms in
moleculesto do with atoms of theory everything that one does with atoms in
the laboratory.

6.4.3 The origin of strain energy in cyclic hydrocarbon molecules

The hybridization model predicts that the smaller bond angles found in a
moleculewith angular strain should result in an increasein the p characterof
the strained C-C bonds and hence in an increasein the s character of the
associatedC-H bonds (Walsh 1947).Orbital models relate an increasein
electronegativityof a carbon atom relative to that of a bonded hydrogen to
an increasein the s character of its bonding hybrid orbital. Thus, it follows
that the presenceof geometricstrain in a hydrocarbon moleculeshould result
in an increasein the electronegativityof carbon relative to hydrogen. In their
classic study of strain in the cyclopropane molecule, Coulson and Moffitt
(Ig4g) emphasizedthis point by showing that the bond lengths and bond
angles of the methylene group in cyclopropane resemblethose for ethylene.
 220 T H E M E C H A N I C SO F A N A T O M I N A M O L E C U L E 6.4

The argument for an increasein electronegativitywith increasings charcter is

based on energy, an s electron being more tightly bound than a p electron.
The theory of atoms in moleculesshowsthat the electronegativityof a carbon
atom does indeed increaseand its energy decreaseas the extent of geometric
strain increases.
Relative to its population in the standard methylenegroup, each hydrogen
in cyclopropane transfers0.045 e to carbon, reducing the net charge on the
carbon atom from + 0.196 to + 0.106 e (Bader et al. 1987a).While this
charge transfer leads to an increaseof 65.3kJ/mol in the stability of the
carbon atom, it results in a decreaseof 52.3kJ lmol in the stability of each
hydrogen atom. Thus, the methylenegroup in cyclopropaneis calculatedto
be 39.3 kJ/mol less stable than the standard transferablemethylene group.
This yields a total strain energy for the molecule three times this, or
118.0kJ/mol, in good agreementwith the generallyacceptedvalue basedon
the experimentalheatsof formation of 115.1kJ/mol. The methylenegroup in
cyclopropane is more stable than the same group in ethylene by only
8'8 kJ/mol. In terms of the chargetransferwithin the group and its energy,it
resemblesmore closely the ethylenefragment (where q(C) : + 0.090e) than
it does the standard methylenegroup.
In the less strained cyclobutane, the transfer of charge from hydrogen to
carbon relative to the populations in the standard methylene group is
reducedto 0.014e for axial H and 0.012e for the other, and the charge on
carbon is + 0.170e.The hydrogensare destabilizedby 37.7kJlmol and the
carbon stabilized by only 10.5kJ/mol, to give an energy increase of
27.2 kJ lmol for each methylenegroup. This yields a predicted strain energy of
108.8kJ/mol, a value which again is in agreementwith the experimental
value of 110.9kJ/mol. Experimentally, the heat of formation of cyclohexane
is found to be six times the heat of formation of the standard methylenegroup
and to possessno strain energy. An axial hydrogen in this molecule is
calculatedto possess0.007 more electrons,(N(H) : 1.099e) and be more
stable by 6.7 kJ/mol than an equatorial hydrogen.The atomic populations
and energies of a methylene group in cyclohexane differ littls from their
values in the standard group and the energy of the group differs by only
0.0001 au or 0.25 kJ/mol from the standard value. Thus, in agreementwith
experiment, cyclohexane is predicted to be strain-free when its energy is
compared with six times the energy of the standard methylene group.
When a carbon atom is subjected to an increasing amount of geometric
strain, as measuredby the departure of the C-C-C bond path anglesfrom the
tetrahedral value, an increasingamount of charge is transferredfrom H to C
in each CH, group relative to the populations in the standard methylene
group. The same effect occurs to an even greater extent for the bridgehead
carbon atoms in the bicyclic and propellane molecules. The bridgehead
carbon atom in [1.1.0]bicyclobutane,the most strained of the bicvclic
6.4 OF ATOMSI\ \lOLECULES
PROPERTIES 221

molecules,withdraws chargefrom the bridging methy'lenegroups and from

its bonded hydrogen and possesses a net negativechargeof - 0.121e' This
carbon atom is only x 42 kJ/mol lessstablethan a carbon atom in acetylene
which has a net chargeof - 0.177e. The bicyclicsand propellanemolecules
are among the most geometricallystrainedof the hydrocarbonmolecules.As
the number of bridging methylenegroups in thesemoleculesis decreased,the
bridgehead carbon atoms increase in both electronegativity and stability.
There is a physical flow of charge from the peripheral methylene groups to
the bridgehead carbons in the bicyclics and in the propellanes as the
geometrical strain in these molecules is increased.These redistributions of
charge within a methylene group or betweenCHr and a C-H or a C group,
lead to a stabilization of the carbon atom which, in the classicalsense,is most
strained and to an even greater destabilization of the hydrogen or neigh-
bouring methylene groups. The overall result is a strain energy equal to the
increasein the energy of the group relative to that of the standard. Thus,
theory provides an atomic explanation of the origin of strain energy in cyclic
hydrocarbons.The readeris referredto the paper by Wiberg et al. (1987)for a
full discussionof this topic. The few examplesdiscussedhere are introduced
to emphasize that the strain energies calculated for cyclopropane and
cyclobutane and the predicted absenceof strain in cyclohexane,all of which
are in agreementwith experiment,are all predicted by the theory of atoms in
molecules.The energy, as defined in eqn (6.72),of the standard transferable
methylene group, as defined by the zero-flux boundary condition (eqn (6.7))
and as found in the pentane and hexanemolecules,servesas the basisfor the
determination of these results. Not only does theory predict the trans-
ferability of atoms and groupings of atoms without change,it also correctly
predicts the measured changes in their energies when these groups are
perturbed.
While the excellence of the agreement of the relative energies of the
methylene group in the cyclic molecules with the measured strain energies
may be to some extent due to the fortuitous cancellation of errors in the
contributions not specificallyconsidered,namely the correlation energy,the
zero-point energy,and A(A/{r0)between0 and 298 K, the nature of the results
leavesno doubt as to the correctnessof the interpretation that has beengiven,
that the atoms of theory recover the experimentally measuredproperties of
atoms in molecules.

6.4.4 Origin of rotation and inversion barriers

In addition to predicting a total energy,quantum mechanicsenablesone to

calculatethe averagevaluesof its klnetic and potential energy contributions,
the latter consisting of the nuclear-electron attractive potential energy (V""),
222 T H E M E C H A N I C SO F A N A T O ] \ 1I N A M O L E C U L E 6.4

and the electron-electronand the nuclear-nuclearrepulsiveenergies(V", and

Znn,respectively).Theseenergies,along with the contribution to the potential
energy of the electronsarising from the virials of the external forces exerted
on the nuclei (eqn (6.58)),are the only energyquantitiesdefinedin terms of the
usual fixed-nucleus Hamiltonian in the absenceof external fields. Model-
independentpredictions and interpretations of energy changesare, therefore,
restrictedto thesequantities and the atomic contributions to thesequantities.
Since the theory of atoms in molecules enables one to calculate all well-
defined mechanicalproperties at the atomic level,new emphasisis placed on
the study of these energies.
The electron-nuclear interaction energy V^" is the only attractive inter-
action in a molecular system and it is the decreasein the potential energy
resulting from this interaction that is responsiblefor the formation of a bound
molecular state from the separatedatoms. Becauseof the virial theorem and
in the absenceof external forcesacting on the nuclei,the total energyE equals
)V (eqn (6.64)) where V : Vn,+ 4" + 4^ is the total potential energy.
Similarly, the changein energy LE betweentwo statesfree of external forces
equals +LV.Energy changesassociatedwith relatively large reductionsin
internuclear separations, such as those encountered in the formation of
moleculesfrom atoms, lead to a decreasein the attractive potential energy
and to increasesin the electron--electronand nuclear-nuclear energies of
repulsion. The attractive potential energy, however, does not necessarily
decreasewhen the total energydecreasesfor a rearrangementof atoms within
a molecule. Thus, it is possible for the total energy to decreaseand for a
systemto becomemore stable in spite of an increasein the attractive energy,
becauseof an even larger reduction in the energiesof repulsion.Correspond-
ingly, it is possiblefor the energy to increaseand for a systemto becomeless
stable in spite of a decreasein the attractive interactions.
One anticipates that the changes in the electron-electron and
nuclear-nuclear repulsions will in general, parallel one another. One fre-
quently finds, as anticipated on the basis of simple arguments, that the
changesin the repulsiveenergiesV", and \n are nearly equal and that each is
approximately one-half the magnitude of the changein the attractive poten-
tial energy.Consider the formation of AB from neutral atoms A and B with
internuclear separation R. The new interactions without charge transfer are:
nuclear repulsion : Z^ZalR: N"N'/R, which is also the approximate
increasein the electron repulsion energy,where each of theseis one-half the
magnitude of the approximate nuclear-electron energy change of
-(Z"NrlR + ZbN"/R). The same near cancellationoccurs when there is a
transfer of charge A -- B, but the change in the internal energiesto the same
order, - /r * .4" (where 1 and A are the ionization potential electron
affinity, respectively)should be included. Thus, whether a change in state is
accompaniedby an increaseor decreasein the total energy is determined by
6.4 P R O P E R T I E SO F A T O N { S I \ \IOLECL LES 223

the differencein two larger, competing energl changesof approximately

equal magnitude.
Denote the electron-nuclearattractivepotential energyas 4 and the sum
of the electron--electron and nuclear-nuclearrepulsiveenergiesas 2'. We
shall term CaseI the instance where a decreasein total energyE is a result of
a d e cr eas einV ^and a s m a l l e ri n c re a s ei n ( (s e eFi g.6.10).A s noted above,
this is generally the case for a process accompaniedby significantdecreasesin
internuclear separations such as forming a molecule from separatedatoms.
The equivalent of Case I for the reverse process,is an increase in energy
accompanyingan increase in internuclear separations. Table 6.5 lists values
of A,\and A tr(for the dissociation of Nr, BF (which is isoelectronic with N2),
and LiF obtained from CI calculations which give dissociation energies close
to the observedvalues.The magnitudes of A,V^,and A V,"are nearly equal and
eachis approximatelyone-halfof L,Vn", but overall the increase in V^exceeds
the decreasein ( and AE > 0. The value of V^^is larger for N, than for BF
sinceZuZo is a maximum when Zu: Za.

V , V V , V

TI ' I^ r <To If
separationsdecrease separatlonslncrease

V" V V, V

lI^E>0il
scparatlonslncrcase
( b a r r i e ri n e t h a n e )
separationsdecreasc
( b a r r i e ri n N H r )

Frc.6.l0. D i a g r a m a t i c r e p r e s e n t a t i o no f t h e r e l a t i v e c o n t r i b u t i o n s o [ t h e c h a n g e s i n t h e
attractive (4)and repulsive(tr/.)energiesto barriers characterizedby an increaseor decreasein
internuclear separations.
224 THE MECHANICSOF AN ATOl\{IN A MOLECULE

(X ) t=r . a a a ( n e i €
(> t--c.i\o
lr) \o r--
c t . ii + +*
c.{ ca
< g O \ C - t r ) r

ta) O OO a.l O\ - C.l * ca c.l

O \ O \ \ O a n - * $ c A a A c A
N o o O r ) F - $ O d c a c . l
. 1 . 1 - - Q Q - . 1 q -
. . o o - O O O O O O O
( \ l t -
r r l
< g t t l

\OoO\O SoOO F-O\.ic.)

C-l O\ (^.| OO ca O\ OO $ ta) $
- O r t \ O O O S * t f t t \ *
n Q - - Q Q - c l q c t l
I A A

\ t r ^ ' ! -
| | | |
t t t l

4 < g
L

L -OO!f, [\l--- o\\o'\\O(,.)

L c-l oo o\ - tr) C) 6l t-- oo l--
R !t oO rf, a.l \ or\ C.l O\ O c-)
\O$a.l c.)-O c.{ra)-.a
r - r - O O O O O O O O
$ .a c.t
r r l
q
L < g t t t

o<'r* F-ci\o r-\ot r)

iO* a . l c a . a o - a - , 1 $ $ o \ $
.a \O O'
\ di oO O\ O\
= q a \ q c l - Q
O O. O\
q n q c l
a.l

o) r€,1' M M , ^

$ ca C.l
E S < g t t l l
L ' b

q - q q 9 - o Q -
r \O C.l OO $ C.l ca oO oO
trh t-- \o
O\
c.l - c{ $ ca
!
<41 I-* \O \n F

q
q

t\ $eaoo \osrf, oF-o\€

t-- |,'- oO C"l ea (n r) O\ OO O
M A A A A A A
= ; < - - - - a x * -
A A A A A A -
I J c.i c.i c.i c.i c.i J J J c.i

- O \ O \ c a t r - $ \ n N c i \ O
oo rn \o ci \o d o\ o\ o.l a.l
OO oO C{ l'- - \O \O O\ O\ cA
O\ci6l rn(nOO cat\O O\
c.|ncl clclq n9u?c.J
O\ $ t-* 6 tn |i.) \O c.l \O \O
oC\.lo c-o\- rn<'F-l--
t r l t t' - l '
lI l l o r r | l
tqS
F G
| |
:
.r

l
|

(.) .E E :E
'?
O
O H
"j
i -"4,c- - -l ^ .- s I__' 9' - j'\J\J

5 za= E u-- E Za{i

6.4 I\ \IOLECULES
OF ATO]\1S
PROPERTIES 225

Following similar arguments,one would expectCaseII. wherethe decrease

in energyresultsfrom a decreasein the repulsivecontributions to AE, to be
dominatedby increasesin internuclearseparations
characteristicof processes
of bonded atoms or, equivalently,for the reverseprocess,an energyincrease
resulting from an increasein repulsive interactions becauseof dominant
decreasesin internuclearseparations.Consideredfrom the point of view of
LE > 0, CaseI is exemplifiedby barriers to internal rotation and CaseII by
barriers to inversion.Payne and Allen (1971)have discussedvarious combi-
'repulsive'or
nations of energycomponentsin the classificationof barriersas
'attractive'dominant.
We shall discussthe barrier to internal rotation in ethane as an example of
CaseI and the inversionbarrier in NH. and the barrier to bendingin HrO, as
examplesof Case II (Bader et al. 1990).Calculationsat the single-determin-
antal level using the 6-311+ +G(2d,2p) basisset for both geometryoptim-
rzation and energy determination were performed for all three molecules.
Such a calculationis known to recoverthe barrier to rotation in ethane,as
evidenced by a comparison of the calculated and experimental results
reported in Table 6.5. The results reported for ammonia and water are
obtained from configuration interaction calculations using all single and
double configurations generatedfrom the orbital set obtained using the large
set of basisfunctions.
In general,the dominant geometry changeencounteredin the formation of
a rotationally eclipsedconformer from a staggeredone is an increasein the
length of the bond about which rotation occurs. In ethane the C-C separ-
ation increasesby 0.0140 A in forming the eclipsed conformer and this
increaseis more than 10 times larger than the accompanyingdecreasesin the
C-H separationsof 0.0010A. The CCH bond anglesopen by 0.4" and there is
an overall decreaseof 0.0069A in the separationsof the hydrogenson a given
carbon. Referenceto Table 6.5 shows that, as anticipated on the basis of the
general arguments, the rotational barrier is a result of a decreasein the
repulsive potential energy and an even larger increase in the attractive
potential energy (a decreasein its magnitude).
The increasein the C-C separationin ethanedominates the changesin 4^
and V"",both thesequantities undergoing decreasesof correspondingmagni-
tude when the staggered(S) is rotated into the eclipsed(E) conformer (Table
6.5 and Fig. 6.11).The samegeometricalincreaseresultsin a lesseningof the
attractive interactions within the molecule,with the increasein Zuexceeding
the decrease in the repulsive interactions. Thus, the barrier to internal
rotation in ethane results from a reduction in the attractive interactions
between the nuclei and the electrons and in spite of an accompanying
reduction in the electron-clectron and nuclear-nuclear repulsions. The
energy changeswhich give rise to the rotation barrier are characteristic of
CaseI and representa small reversalof the processof C-C bond formation.
226 T H E M E C H A N I C SO F A N A T O M I N A M O L E C UL E 6.4

Eclipsed at StaggeredGeomery

H, ,H Eclipsed
xJ,6-.2x
H
/ \ H
H. ,H
xt6- s2x
H
/
/ L E = 13.8 AE = -0.8
Staggered ,/ a r = 39.3 AT = -49.8
,/ av. = -92.9 AV. = 961.1
AV' = 6 7 .4 AV. - -9t2.1
").-.{"
AE = 13.0
AT = -12.6
A V. = 8 7 1.9
AV, = -846.4
Flc. 6.I 1. Energy changesin kJ/mol for the rotation of the staggeredconformer of ethane (S)
into the eclipsedconformer (E) and into the frozen eclipsedconformer (FE).

The changesin the charge distribution of ethane caused by the internal

rotation are relativelysmall (Fig.6.12).The principal changeis the replace-
ment of the curved C-C interatomic surface in the S conformer with the
planar one in the E conformer, a consequenceof the replacement of a
threefold alternating axis of symmetry with a simple threefold axis. With
referenceto Fig. 6.12, one seesthat the upper arm of the atomic surfaceof an
eclipsed hydrogen atom is moved inwards relative to its form in the S
conformer and its volume, which equals 38.44au in the S geometry,is slightly
decreased,by 0.24 au. The remainder of the basin of a hydrogen atom is
nearly unperturbed by the rotation. The changesin p at the bond critical
points are small, being largest for the C-C bond for which the value of pa
decreasesby 0.0071au from a value of 0.2495au in the S conformer. Its value
for the C-H bond increasesby 0.0008au from its value of 0.2908au. The C-H
bond critical point shifts by 0.0017au towards the proton in the eclipsed
geometry, a harbinger of the small transfer of charge from H to C which
accompaniesthe rotation. The C-H bonds in ethaneexhibit a small ellipticity
with major axesdirected so as to be tangent to the cone obtained by rotation
of the C-H bonds of a methyl group about the C-C axis. This ellipticity
approximatelydoublesin value to 0.014when the group is rotated into the
eclipsedgeometry.Thus, the extent to which electronicchargeis accumulated
in the tangent plane containing the C-H bond as opposed to the one
perpendicular to it, while small, is increasedwhen the molecule attains the
staggeredgeometry.
The atomic contributionsto the energychangesare given in Table 6.6.The
changesin the energiesof the individual atoms are small, a reflectionof the
6.4 P R O P E R T I E SO F A T O \ { S I \ \IOLECULES 227

Frc.6.l2. Contour maps of the electronic charge density for the staggered and eclipsed
conformers of ethane (a) and for the bent and linear conformers of water (b). The outermost
contour has the value 0.001au. The remaining contours increasein value according to the scale
given in the Appendix (Table A2). The bond paths, the interatomic surfaces,and the bond critical
points are also indicated.

small perturbations of the atomic charge distributions which accompany

the internal rotation. The net charge on a H atom in the staggeredform is
q(H) : - 0.061e. On changing to the eclipsedgeometry,there is a transfer of
0.003e from each hydrogen to its bonded carbon atom. Each hydrogen atom
is stabilizedby a small amount in the eclipsedgeometry in spite of the small
decreasein its population. Each carbon atom is destabilizedby an amount
approximately three times the magnitude of the stabilization of the hydrog-
ens bonded to it and this is the source of the barrier at the atomic level. The
transferof chargeto carbon leadsto a decreasein V?(C),the potentialenergy
of interaction of the carbon nucleus with its own charge distribution eqn
(6.81),with A V?(C) : - 35.6kJimol. The energychangefor a carbon atom,
however,is dominated by the increasein the C-C separation.This increase,
 \ \toLECULE 6.-l
228 T H E M E C H A N I C So F A N A T o \ 1 I \

= F 3
e A s a

L J
^
v
9
-
V
-
(t
< g l l l

\O ci t-*
\ . o S
F-- s
r-- ta) ao
& d . i , i ; C)
9
L
bo

3 xs Fx s! EqEE
?--'^
< g
5
i
R S r e -o cs 8
ni ,ri i,.i co
8 E
-c
E ()
q
r* l r l l
a
I
L

rI gn
R !

Hi=g53 !3|
L

Y ^

q
t\
< g
l l t l r r | a

L
c
()

i e l: ; = g3 !E $
O

(.)
a

<i9

= !s ss =3 g+ =3
z ) - 5 5 a e e l - e
a a . = - a a= ==
\ O L < g = = =
6) -$
E O -X
6t
* -
H -
U 3 r . n . l - -
q ! \ . . - o . o . $ t , & a
q
rii -:z r - 6 - r/') C)
t--- \O

t-
- x. ..1 - c) .a o' ra l-- \c o l-:
!rtr rig
g gu *= * | :' $K
; ..i co
-;: ..i : ii
q

:g * - l l l
L
I

:\ ? A e l l I s E Al- *o
= a a a a c c : c
0)

( ) O -
- OJ-l
G -

6ir v
-c >,
\
$IK}a et *gas
F?
(.).^
0 ) e

> 3
tl 9
TfEr i I
T ;T : 0)
u !

z coE f r- UI ZT, CC E t

() F a

q.)

F
. F F
f
6.4 O F A T O \ 1 SI \ \ I O L E C L L E S
PROPERTIES 229

while leading to a reduction in the repulsiveinteractionsbetweenthe two

carbon atoms, gives rise to an even greater reduction in the attractive
interaction of each carbon nucleus with the charge density on the other
carbon atom. (LV^(C) consistsof A,Vo^(C). which is lessthan zero, and the
interactionof each of the other nuclei with the chargedensity of the carbon
atom in question.Sincethe C-H bond lengths have decreasedin the eclipsed
geometry,and only the C-C length has increased, the increasein V"(C) is the
result of the latter effect.)Thus, the carbon atoms are destabilized in the
eclipsedgeometry as a result of a decrease in the attractive interactions of
each carbon nucleus with the charge density of the other. These energy
changes of the carbon atom dominate the change in energy for the total
system and hence the barrier in ethane is an example of Case I.
Orbital arguments concerning the origin of the barrier to rotation in
ethane have been reviewed and discussedby Pitzer (1983).The explanation
favoured by him is based upon a model in which the molecular orbitals for
the staggeredgeometry are first frozen in form and the nuclear framework of
the staggeredgeometry is rotated without changeinto the eclipsedgeometry.
This leads to a small negative barrier, a result which is interpreted to mean
that simple electrostaticinteractionsbetweenthe ends of the moleculeare not
important in the ethane barrier. The freezing of the orbitals destroys their
orthogonality, as required by the Pauli principle, and the change in P
obtained from their re-orthogonalization as carried out in the second step.
resultsin so-calledoverlap or exchangerepulsion contributions to the energy.
that is, to'Pauli repulsions'betweenthe C-H bond orbitals.
While, as shown in Fig. 6.12,the chargedistribution of ethanechangesso
that the basinsof the eclipsedhydrogen atoms do not overlap in forming the
E conformer, thesechangesin p do not correspond to a compressionof the
hydrogen atoms. A pair of eclipsed hydrogens do not share a common
interatomic surface-they do not touch, as occurs in a closed-shellrepul-
sion and, as reflectedin the decreasein the repulsivecontributions to their
energy(Table 6.6),they are not compressed.Instead,the spatial extent of each
is diminished by a transfer of a small portion of its spaceand its contained
chargeto the carbon atoms,each of whosevolumesincreasesby 0.47au. To
demonstrate that the increasedC-C bond length in the E conformer and
hence the barrier are not a result of repulsionsbetweeneclipsedhydrogens,
one can inquire into the origin of the barrier resulting from an internal
rotation of the rigid nuclear framework of ethane from its staggeredto an
eclipsedgeometry without recourseto a model state.Such a configuration of
the nuclei, while physically reaiizable,does not lie on the reaction path and it
possessesan energy slightly in excessof the eclipsed transition-state geo-
metry. Forces act on the nuclei in this constrainedgeometry such as to move
them into the positions they occupy in the eclipsedtransition state.Sincethe
major geometrical change between the staggeredand eclipsed equilibrium
r30 T H E M E C H A N I C SO F A N A T O M I N A M O L E C U L E 6.4

geometriesis a lengthening of the C C bond, one anticipates that the

principal forces operative in the frozen eclipsed geometry are forces of
repulsionacting on the two carbon nuclei.The changesin the potential and
kinetic energiesfor the nuclear motions: staggeredto frozen eclipsed,S --+FE,
frozen eclipsedto eclipsed,FE -r E, and S -- E are summarized in Fig. 6.11.
Because forces act on the nuclei in the FE conformation, the general
statementsof the virial theorem, which include the contributions of the virials
of the nuclear forces to the total electronic energy,must be used (eqn (6.65)).
The contribution of nucleusa to the virial of the forcesacting on the electrons
is equal to - Xo'Fo, where Xo is the position vector of nucleus a and
Fo : - Y oE, the net force acting on it. When the forceson the nuclei vanish,
one obtains the corresponding equilibrium statementsof the virial theorem,
T - - E and E:iV. The generalstatementcorrespondingto T: - E is

T- - E + I,X,'F,. (6.e1)
When the nuclear virial is positive, 7n must exceed lEl to balance the
contribution to the virial arising from the repulsive forces acting on the
nuclei. Sincewe are taking differencesbetweena state where the forcesdo not
vanish and one where they do, the total nuclear virial for the non-equilibrium
state appearsin the difference.Thus, for the reaction S -- FE, the change in
kinetic energy equals minus the changein total energy plus the nuclear virial
(eqn (6.e2)),

L,T: - LE + I"Xo'Fo. (6.e2)

S inc e A i" is p o s i ti v e fo r th e re a c ti o n S-+ FE (Fi g.6.11)and exceedsthe
magnitude of AE, the nuclear virial is positive,showing that this contribution
to the potential energy of the electronsis dominated by repulsiveforcesacting
on the nuclei. The corresponding expression for the change in potential
energy is
L , V: L V u + L V r : 2 L E - I " X , . F o . (6.e3)
In this case Vu decreasesand V, increases,both quantities changing in the
opposite direction to that for the overall reaction S - E, but the increasein
repulsive energy arising from the nuclear-nuclear and electron-<lectron
forces is lessthan the increasein the magnitude of the attractive energy and
overall the change in the potential energy is lessthan zero, i.e. the attractive
contributions dominate. Thus, the increase in energy encountered in the
formation of the frozen eclipsedconformer comesnot from an increasein the
usual potential energy contributions, but from the virials of the repulsive
forces acting on the carbon nuclei in the non-equilibrium geometry. This is
also evident in the expression for the total energy change expressedas
L,E : A,T + A^V.Thefact that AT > AE showsthat LT includesthe contribu-
tion of a repulsivenuclear virial. The atomic contributions to the change
6.4 O F A T O N l SI \ \ I O L E C U L E S
PROPERTIES 23r

$ --+FE, like thosefor S --+E, come primarill from the carbon atoms,AT(C)
: 1 6 . 7 k J / m o l a n d A V ^ ( C ) : - 1 4 9 . 8k J m o l . u . h i l eA f ( H ) : 0 . 8k J / m o l .
-- E result in a
The geometry changes encountered in the reaction FE
decreaseof only 0.4 kJ/mol in the total energy. However, the relaxation in the
geometry, particularly the lengthening of the C-C internuclear separation,
causessizeablechangesin the potential energycontributions,primarily those
of the carbon atoms. The contribution from the nuclear virial vanishes in
reaching this metastablegeometry and the attractive and repulsivepotential
energycontributions changein the direction dictated by CaseI for the partial
unmaking of the C-C bond, with A ( > 0 and Ln < 0. The kinetic energy
undergoesa decreasewhich, asidefrom the small AE value, equals the loss of
the nuclear virial. The value of A?'(H) is only 0.4 kJ/mol for this change
showing that essentiallythe whole of the decreasein f comesfrom the carbon
atom contributions. Thus, the energy increase associated with the rigid
rotation of one methyl with respectto another to give the non-equilibrium
eclipsedgeometry arisesfrom the contribution of the virials of the repulsive
forcesgeneratedon the carbon nuclei and this changein energy overridesthe
decreasein the potential energy V.
One can investigatethe nature of the changesin the chargedensity over the
basins of the carbon atoms to better understand the accompanying changes
in energy. In the S conformation of ethane, each carbon atom is polarized
towards the other with lM(C) I : 0.030au. This value decreases to 0'025au in
the FE conformation. Thus, electronic charge is shifted from the binding to
the antibindingregion of the carbon nuclei(Berlin 1951)whenthe moleculeis
rotated without the possibility of relaxing the C-C separation.Thesecounter-
polarizationscontribute to the forcesof repulsion acting on the carbon nuclei
in the FE conformation. They are destroyedand replacedby an even larger
polarization of each carbon atom into its binding region, lM(C)l : 0.039au,
when the FE conformation is allowed to relax to the force-freeE conforma-
tion. The value of the quadrupole moment Q""(Q), eqn (6.48),for a carbon
atom in the S conformation is + 0.031au, a polarizatron which corresponds
to the transfer of charge from along the C-C axis to a torus-like distribution
about this axis. (For comparison, the same moment for a carbon atom in
acetylenehas the value of 4 au.) This moment is increasedto 0.074au in the
FE conformation and undergoes a further increase to 0.092au in the E
conformation. The effect is a progressive transfer of charge density from
along the C-C internuclear axis to a torus of density about this axis, a
polarizatron which contributes to the lengthening of the C-C bond and to the
shortening of the C-H bonds as occurs in the E conformation.
The principal changesin the charge distribution of ethane upon internal
rotation occur along the C-C bond path and within the basins of the carbon
atoms, with only relatively small changes occurring along the C H bond
paths and within the basinsof the hydrogen atoms. It is difficult to rationalize
t32 T H E M E C H A N I C SO F A N A T O M I N A M O L E C U L E 6.4

theseobservedchangesin the chargedensity with those anticipatedon the

basisof the requirementsof orthogonality of the C-H bond orbitals (Pitzer
1983).One might argue that the greater C-C separation found in the E
conformation is a consequenceof the exchangerepulsionsbetweenthe C-H
bond orbitals. However, the changesin density and energy for the reaction
S -- FE do not correspond to those anticipated on the basisof the overlap of
occupied bond orbitals, as even here the primary changesin density are along
the C-C axis and within the basins of the carbon atoms and the change in
potential energy is attractive rather than repulsive.The barrier in this case,
which is of almost the same size as the equilibrium barrier, results from the
creation of repulsiveforceson the carbon nuclei whosepresenceincreasesthe
total energy of the system by increasingthe electronic kinetic energy of the
carbon atoms. Relaxing the geometry by lengthening the C-C separation to
its value in the E geometry, replacestheseforces of nuclear repulsion with a
barrier arising from an increase in the attractive potential energy of the
carbon atoms.
To understand the origin of the barrier in ethane requires an understand-
ing of why the C-C bond is lengthenedin the eclipsedgeometry.The change
S -- E transformsan alternating axis of symmetry into a simple threefoldaxis.
The alternating field of the protons exertedthe length of the C-C axis in the
Hamiltonian of the S conformation is replaced by a more pronounced field
with a threefold symmetry in the Hamiltonian for the E conformation. This
change in the Hamiltonian has the effect of increasing the quadrupolar
polarization of the electronic charge density along the C-C axis, removing
charge from this axis and concentrating it in a torus-like distribution about
the axis. As a consequence,the electronicchargein the C-C binding region is
lesseffectivein binding these nuclei and the C-C separation increases.This
same quadrupolar polarization slightly increasesthe binding of the protons.
The observationsthat a rotational barrier is characterizedby a lengthening
of the bond about which the rotation occurs and by an increase in the
attractive and a decreasein the repulsive potential energiesas found for
ethane appear to be general. That is, rotational barriers belong to Case I
energy changes.The same observationsare found for the barriers to internal
rotation in methylamine and methanol (Table 6.5),in carboxylic acids and
estersfor rotation about the C-O bond, and for rotation about the C-N bond
in formamide (Wiberg and Laidig 1987).
The inversion barrier is found to be an example of a CaseII energychange
for the moleculesNH., PHr, H.O*, and HrO (Table 6.5).Only the first and
last of these molecules is discussedin detail. The A-H internuclear separ-
ations in ammonia and water are found to decreasewhen the pyramidal or
bent geometry is transformed into its respectiveplanar or linear form, by
0. 0130A in NH . a n d b y 0 .0 1 8 0 A i n H rO . The approach of the protons
towards the A nucleus,while resulting in a lowering of the electron-nuclear
 6 .4 P R O P E R T I E SO F A T O \ { S I \ \IOLEC-L'LES 233

attractivepotentialenergy(Table6.5)upon attainmentof the planar or linear

geometry,leadsto an even greaterincreasein the repulsil'einteractions.The
inversion barrier in thesemoleculesis thus a consequenceof an increasein
the repulsiveinteractionsthat outweighsan accompanyingdecreasein the
attractivepotentialenergy,an exampleof a CaseII energychange.Rauk et al.
(1970),in the first successfulcalculation of the barrier in ammonia, give the
sameexplanationfor its origin. The decreases in internuclearseparationsin
these moleculesresult in an increase, rather than a decrease,in the total
energy of the system,a behaviour opposite to that found for Case I.
The hybridization model predicts an increase in the electronegativityof the
A atom relative to hydrogen in the attainment of the planar or linear
geometries.As discussedin Section 6.4.3,this is the predicted consequenceof
the increasein the s-characterof its bonds to hydrogen, from sp3 to sp2 or
from p to sp, coupled with the fact that s electrons are more tightly bound
than are p electrons.On this basisone predicts a transfer of chargefrom H to
A to accompany the shortening of the A-H bonds in the formation of the
planar or linear geometries.This prediction and the associatedenergetic
consequencesare found to be correct as indicated by the data in Table 6.6.
The molecular charge distributions of pyramidal and bent molecules
undergo larger perturbations during inversion or hnearization, as a conse-
quence of the accompanying change in hybridization, than does a charge
distribution for a molecule undergoing internal rotation. This greater reor-
ganization of the charge is reflected in the larger changes in the atomic
energies,changeswhich are a direct measureof the extent of change in the
distribution of chargeover the basin of each atom. The sizeableredistribution
of charge causedby the change in hybridization of the N atom in attaining
the planar form of ammonia is reflectedin the value of p""(N) changing from
- 2.92au in the pyramidal molecule to - 4.15au in the planar geometry.
This moment is characteristicof a doubly occupiedpz orbital on the N atom.
Table 6.6 also lists the bonded radius of hydrogen in the equilibrium
geometry and the changein this radius when the barrier is attained.The shifts
in the A-H interatomic surfacetowards the proton are much larger for the
inversion and linearization barriers than for the rotation barrier in ethane
(Fi g. 6. 12) .
The changein the relative electronegativitiesof N and H or of O and H and
the associated changes in their stabilities resulting from the change in
hybridization upon inversion or linearization have a direct parallel with the
I
changesin the properties of the carbon and hydrogen atoms of the methylene
group of a normal hydrocarbon when it is transferredto a cyclic systemwith
geometric strain. The electronegativityof the carbon, nitrogen, and oxygen
atoms is increasedrelative to hydrogen in each case.The resulting transfer of
charge from H to A destabilizesthe hydrogensto a greater extent than the A
atom is stabilized and the energy is increased.In terms of the mechanical
_ -1+ T H E M E C H A N I C SO F A N A T O M I N A M O L E C U L E 6.4

properties of atoms, the strain energy of a methylene group in a cyclic

hydrocarbon, the inversion barrier in ammonia, and linearrzation barrier in
water have a common origin.
The transfer of charge from H to A is larger in the hnearization of water
than in the pyramidalization of ammonia and the energybarrier and changes
in atomic energiesare correspondinglygreaterfor Hro than for NH.. The
pattern of energychanges,however,is the samefor the barrier in water as it is
for the barrier in ammonia, with the hydrogensbeing destabilizedto a greater
extent than the oxygen atom is stabilized.The repulsivecontributions to the
energy changesof the hydrogen atoms decreasein both moleculesand the
increasein the repulsive interactions which dominates the barriers in these
moleculesis localizedwithin the A atoms.
The change in Z"o(Ol the change in the attractive interaction of the
nucleusof atom O with its own chargedensity,eqn (6.81!-is stabilizingfor
the A atom and destabilizingfor the hydrogensin each of the molecules.The
observation that the magnitude of the decreasein V: (O) is considerably
greater than is its increasefor all the hydrogens in a given molecule shows
that the binding of the charge transferred to the A atom is significantly
increasedand is one of the principal contributors to the attractive energy
contributions to the barriers in thesemolecules.
The origin of the barrier for the rotation of planar formamide by 90' or
270" investigatedby Wiberg and Laidig (1987)is an interestingcase,as here
the rotation about the C-N bond results in the pyramidalization of the N
atom and its formal hybridization changesfrom sp2 to sp3.Thus, as for the
inversion barriers discussedabove, the N atom is found to gain electronic
charge from its bonded neighbours and to increase in stability when the
amide group is rotated from a non-planar into its planar form. In this
molecule,however, the destabilization of the atoms which donate charge to
the N atom in the planar geometry is lessthan the stabilization acheivedby
the N atom and the equilibrium geometry of this molecule is planar. Here, as
is typical of a Case I energy change and of an internal rotation, the principal
geometricalchangeis a lengtheningof the bond about which rotation occurs
(as in the inversion of ammonia, the bonds to N are shortest when its s
character is greatest, i.e. in its planar form) and the predicted barrier of
66.9kJ lmol is a result of a decreasein the repulsive energiesand of a larger
decreasein the stabilizing nuclear -electronattractive interactions.Becauseof
the transfer of charge from N to C which accompaniesthe loss of planarity,
all of the lengtheningof the C-N bond, which equals0.15au, is taken up by
an increasein the bonded radius of the carbon atom, from 0.83 to 1.04au.
The resonancemodel accounts for the relative stability of the planar geo-
metry by invoking a resonancestructure wherein the N atom donatescharge
to the carbonyl oxygen atom, a resonanceinteraction which is lost upon
rotation. The resonancemodel is, therefore,in direct contradiction with the
6.4 OF ATO\ISI\ \IOLECULES
PROPERTIES 235

hybridization model for these systems * hich predicts the N to be most

electronegativein the planar structure. The theorl of atoms in molecules
showsthe resonancemodel to be wrong in this instance(Wiberg and Laidig
1987).Not only is the predicteddirection of the chargetransferfor nitrogen
incorrect, the properties of the oxygen atom. including its geometrical
parameters,are found to change by only small amounts compared to the
changesundergoneby the C and N atoms.the magnitudeof its energychange
being 20 to 30 times smaller.

6.4.5 Perfect transferability-an unattainable limit

The examplesof the group properties in hydrocarbon moleculesgiven above

demonstrate that the properties of an atom change in direct responseand
proportion to changesin its distribution of charge.Whether the form of an
atom changesby a little or by a lot, its energyand other propertieschangeby
correspondingamounts. It was the observation of a paralleling constancyin
the kinetic energy and the charge distribution of an atom, as discussedin
Section5.2,which led to the developmentof the theory of the quantum atom.
This relationship between the spatial form of an atom and its properties is
most important, for it is becauseof this relationship that we are able to
identify an atom in different systems.
On the basis of the demonstrated relationship between the transferability
of the charge density of an atom and the correspondingconstancy in its
contribution to the energy, one might anticipate that the theorem of
Hohenbergand Kohn (1964)would apply to regionsof spacebounded by the
zero-flux surfacecondition. Their theorem, which forms the basis of density
functional theory, statesthat the ground-statewave function and energy are
unique functionals of the charge density, p(r). The general problem of
extending this theorem to a subsystemhas been considered by Riess and
Mtinch (1981).They have demonstratedthat the theoremis applicableto any
finite, but otherwisearbitrary subsystemof some total, bounded system.As a
corollary to their proof, it is shown that perfect transferability of an atom
betweensystemsis an unattainablelimit (Bader and Becker 1988).
Riessand Mtinch are able to show that the density p(r) associatedwith the
total systemis a unique functional (i.e.totally determined by) of pn(r), where
the latter is the restriction of the density to an arbitrary subdomain Q of the
total system.The only assumption that must be made in obtaining this result
is that the total system, while arbitrarily large, must be bounded. Having
shown that p(r) is a unique functional of pn(r), it follows, together with the
original theorem of Hohenberg and Kohn, that the ground-stateenergy and
the ground-state wave function are unique functionals of the ground-state
density of an arbitrary subdomain of a total, bounded system. The sub-
domain density po(r) thus uniquely determinesthe expectation value of any
J_10 T H E M E C H A N I C SO F A N A T O N { I N A M O L E C U L E 6.4

spin-freeobservableassociatedwith the subdomain O, or with any other

subdomain,or with the total domain of the system.It is a corollary of this
proof that, if the density over a given atom f), or any portion thereof, is
identicalin two moleculesA and B, i.e.poa(r) : pon(r),then po(r1: p"(r)and
the two moleculesare identical. Therefore,the limit of perfect transferability
of atomic propertiesis unattainable.No restriction,however,is placed on
how closelythis limit can be approached.For the examplesof the hydrogen
atoms in the transferablemethyl groups of the n-alkanes given above and
taking into account as well the limits on the experimental accuracy of
measuredheatsof formation, the existenceof the limit may not be of practical
importance.
The work of Riessand Mtinch demonstratesthat the energy of an atom in
a molecule, as well as its other properties, are indeed determined by its
distribution of charge.Thus, this important physicalcharacteristicof an atom
has a basisin theory. The atoms and their propertiesprovide visual proof of
the theoremof Hohenbergand Kohn by showingthat similar distributionsof
charge possesssimilar energies.There is nothing in the original derivation of
the theorem concerning the dependenceof local contributions to the total
energy on the local properties of the charge density. The actual functional
which relatesthe total energy to the chargedensity is unknown and, in terms
of quantum mechanics,a knowledge of the one-densitymatrix, and not just
its diagonal elementswhich define p(r), is required for the determination of
the energy of a system. The quantum stress tensor F'1r; determines the
potential energy density of the electrons,{(r) (eqn (6.30)),and their kinetic
energy density (eqn (6.27)),while it itself is a functional of the one-density
matrix (eqn (6.13)).Thus, the integration of the sum of the two energy
densities or of either one of them separately,together with the use of the
atomic virial theorem (eqn (6.73))will yield the electronic energy of an atom,
or of the total system.If the systemis not in an equilibrium configuration, the
subtraction of the virial of the nuclear forces (as determined by p(r) in the
electrostatictheorem)from E. for the total systemwill yield the total energyE
(eqns(6.55)and (6.56)).Thus, the energiesE" and E are known functionalsof
the one-densitymatrix, ltt)(r, r'). A readerinterestedin a unified account of
density-functional theory and of its application to atoms, molecules, and
chemical conceptsis referred to Parr and Yang, 1989.
While known quantum mechanicalrelationshipsrequire more information
than is contained in just p(r) for the determination of the energy, only a
portion of the information contained in the one-densitymatrix is required for
this purpose (Bader 1980). The one-density matrix is a function of six
variables.The physicalinformation of f (t'(r, r') is however,containedin the
neighbourhood of its diagonal elementsr : r'. To establish this, one ex-
pressesf (1)(r,r') in a new systemof coordinatesdefinedby

X:i(r+r') and *: i(r _ r,) (6.e4)

 '-t
6.4 P R O P E R T I E SO F A T O \ I S I \ \ I O t - F ( ' LL E S /

andconsiders theproperties of I-(t)in theneighbtrurhood trfa pointalongthe

diagonal(X,0).Towardsthis end,w'eperforma Tal lor senesexpansion of
f ( 1 ) i n t h en e i g h b o u r h o o df t h i sp o i n t .I n c l u d i n tge r m su p t o s e c o n d - o r d e r
in dX and dx, this yields
t - r r r ( X+ d x , d x ): f ( t ) ( x , 0 +) [ i l - , ' , ( X . 0 i)X ] r ) X
+ : d x . [ i 2 r ' 1 ) ( x .o ) i X : ] . d X
( 1 ' ( X0. ) , , i x']r ) x
+ [af
+ ] d x . [ a t f t t t ( X . 0 r) ' x t ] ' d x

+ a x . [i 2 f t1 t(x ,o ),,' aixx ] . d x. (6.9s)

Ea ch t er m in eqn ( 6 .9 5 ) h a s a d i re c t p h y s i c a l i nterpretati onw hen the
d e ri va t iv esar e ev alua te da fte r e x p re s s i n gf{ 1 } i n te rms of i ts natural orbi tal
expansion.Retainingthe sameordering of terms as appearsin eqn (6.95),this
yi e l d s

f ( 1 ) ( X+ d X , d x ) : p ( X ) + V p ( X ) ' a X + ] d X ' V V p ( X ) ' d X

+ ( 2 m i l h ) j ( X )' d x * ( 8 m l h 2 ) d x ' F ( X')d x

+ (2milh)dX'Vj'dx. (it.96)

The terms in the expansionderived from derivativeswith respectto X are

identical to those obtained by taking the correspondingderivativesof the
chargedensity p(X) itself.The dyadic VVp is the Hessianmatrix of p. whose
eigenvectorsand eigenvaluesdeterminethe propertiesof the critical points in
the chargedistribution.The trace of this term is the Laplacian of the charge
d e n si ty ,V 2p.
(r)
The derivativeswith respectto x samplethe off-diagonalbehaviourof f
j
and generateterms related to the current density and the quantum stress
tensorF'. The first-orderterm is proportional to the current density,and this
vector field is the x complementof the gradient vector field Vp. The second-
order term is proportional to the stresstensor. Consideredas a real sym-
metric matrix, its eigenvaluesand eigenvectorswill characterrzethe critical
points in the vector field j and its trace determinesthe kinetic energydensities
K(r) and G(r).The cross-termin the expansionis a dyadic whosetrace is the
divergenceof the current density.
The second-orderexpansiongiven in eqn (6.96)recoversall of the physical
quantitiesneededto describea quantum systemand determineits properties:
the charge density p and its gradient vector field Vp define atoms and
determinemany of their propertiesin a stationary state;the current density
determinesthe system'smagnetic propertiesand the changein p in a time-
dependentsystem;and, finally, the stresstensor F determinesthe local and
averagemechanicalpropertiesof the system.Thus, one does not need all the
238 T H E M E C H A N I C SO F A N A T O \ I I \ { \lOLECULE 6.5

information in the statefunction ry'to describea quantum mechanicalsystem

and determineits properties.The expansionin eqn (6.96)demonstratesthat,
even with the existing quantum mechanics,we do not require much more
information than that contained in p itself,the other extremeto the over-
complete knowledge afforded by ,lt* rlr.
Further investigation into the properties of f(1) in the vicinity of its
diagonal elementswould seem to be a worthy endeavour.In a stationary
state,in the absenceof a magneticfield,j and Vj vanishand, in this situation,
eqn (6.96)reducesto

+ jdx vvp(X)'dX
f ( 1 ) ( x+ d x , d x ) : p ( X )+ V p ( X ) ' i ) X
+ (8mlhz)dx'F(r)'dx. (6.e1)

A critical point in the full field t-tt) (X, x) is one for which its first derivatives
with respectto both X and x vanish.Thus, any critical point in p(r) is also a
(1)becauseof the vanishingof the vector current densityin a
critical point in f
stationary state.The gradient vector field of the charge density is thus(r) the
(1)in a stationary state and the topology of f is
gradient vector field of f
susceptibleto study. A critical point in f(1)is characterrzedby the eigenvec-
tors and eigenvalues of two Hessianmatrices,VVp(r)and F(r)' The formation
of a degeneratecritical point in the Hessianof p signalsan instability in the
gradient vector field of p and a change in molecular structure. The stress
tensor F determinesthe local mechanicsof a charge distribution and it is
possiblethat the eigenvectorsand eigenvaluesof this secondHessianat the
iegenerate critical point could summarizethe mechanicalconsequences of
the change in structure. The quantum stress tensor provides a link between
the mechanicaland structural propertiesof a molecular system.

6.5 The hypervirial theorem and the definition of bond properties

The openingsectionof this chapterstressedthe importanceof the presenceof

the surfaceintegral in the hypervirial theorem for an open system,eqn (6'2)'
Unlike the theorem for a total system,eqn (6.4),in which casethe averageof
the commutator of any observabled with the Hamiltonian H vanishes,the
correspondingresult for an atom in a moleculeis proportional to the flux in
the effictive single-particlevector current densityof the property G through
the atomic surface.As a result. the hypervirial theorem plays an important
role in determiningthe propertiesof an atom in a molecule.It also enables
one to relate an atomic propert,vto a sum of bond contributions,as is now
demonstrated.
6 .5 H Y P E R V I R I A LT H E O R E M A \ D B O \ D P R O P E R T I E S 239

6.5.1 An atomic property expressedas a sum of bond contributions

We shall be interestedin observablesthat har e non-r anishingcommutators

with the Hamiltonian fr. Supposethe commutator of H and G is relatedto
the observableA Uv the relation
(ilh)lH,Gl -- A (6.e8)
In this case A and G shall be referred to as associated operators. The
averagingof this commutator and its complexconjugateover the atom Q, as
indicated in eqn (6.2),and multiplication by l{/2 then yield the averagevalue
of A for an atom in a molecule, the quantity A(O). In such a case, the
commutator averageequalsthe changein the value of the property ,4 when a
free atom Q combines to form a molecule,sincethis averagefor the isolated
atom vanishes(eqn (6.4)).If one denotesthe change in the property A by
4,4(Q),then from eqn (6.2)one obtains
4,4(O): (N12){ $dS(O,r) jc(r)'n(r)+ cc}. (6.ee)
Thus the change in the value of the property A that occurs as a result of an
atom entering into chemicalcombination is given by the flux in the vector
current density of its associatedgenerator G through the atomic surface.
An atomic surfaceS(Q, r) is, in general,composed of a number of inter-
atomic surfaces,there being one such surfacefor each atom O' linked to Q by
an atomic interactionline or, in the caseof a bound system,by a bond path.
That is,
S(Q,r) : Ir,,*r,S(QlQ',r) (6.100)

where S(OlO',r) denotesthe interatomic surfacebetweenatom O and Q'.

Thus, the surface integral in eqn (6.99) can be broken up into a sum of
integralsover the interatomic surfacesbounding the atom O and the value of
4,4 (O) is thereby equatedto sum of terms,there being one term for each atom
linked to atom Q by a bond path. If each such surfaceintegral, including the
factor (Nl2), is denotedby A(AlQ'), eqn (6.99)becomes
A / ( C ) ): I r , , + ( ) . 4 ( Q l Q ' ) (6.101)
where
, 4 ( O l C ):' ) ( N1 2 ) { $d S ( O l Q r") j o ( r ) . n ( r +
) cc}. (6.102)

In this manner the atomic averageof the observablesatisfyingeqn (6.98)for

the associatedgenerator G is expressibleas a sum of contributions where
each contribution is determined by an integral over the surface which the
atom shareswith each of its bonded neighbours.This atomic averageis
necessarilythe changein the value of the property relative to its value in the
isolatedatom. An exampleis provided by eqn (6.18),which shows that the
240 T H E M E C H A N I C SO F A N A T O M I N A M O L E C U L E 6.5

Ehrenfestforce acting on an atom in a molecule,.4(O) = F(O), is the vector

sum of the forcesexertedon it by each of its bonded neighbours,forceswhich
vanish when the atom is free. The force exerted on atom O by its bonded
neighbourf)', the quantity F(QlQ'), is determinedby the pressureT. n acting
on each of the surfaceelementsdS(QlQ',r).

6.5.2 Surfaceintegrals proportional to the energy

The dimensions of the property A are those of its associated generator

divided by the time. Thus, when d equalsthe virial operator r.p, which has
the dimensions of action, the property A has the dimensions of energy and
equals twice the electronic kinetic energy plus the virial of the Ehrenfest
forcesacting over the basin of the atom (eqn (6.19)).For a free atom, the sum
2f (q + fb(O) vanisheswhile for a bound atom it equals the negativeof the
virial of the forces acting on the surface of the atom, the term {r(A) (eqns
(6.20) and (6.21)). Thus, - {r(A) equals twice the change in the kinetic
energy of the atom plus the change in its basin virial (the change in the
potential energy of the free atom) when the atom becomesbonded to other
atoms
2Lr@) + A/i(o) : - /;(Cl). (6.103)
According to eqn (6.103),the surface virial measuresthe disparity in the
changesin the valuesof the electronickinetic and potential energiesover the
basin of the atom required for the satisfaction of the virial theorem, as
occasionedby the formation of chemical bonds.
The classicalvirial theorem, when applied to a systemof particlesconfined
to a volume V,equatesthe negativeof twice the kinetic energy of the particles
to the virials of the internal and external forces acting upon them. The virial
of the external forcesis equal and opposite to the virial of the forceswhich the
moleculesexert on an element of surfacearea of the container, the quantity
pr'n dS where p is the pressure.Integration over the surfaceof the container
is transformed using Green's theorem into the integral of 3p d Z over its
volume and the virial of the externalforcesis 3pV.For a systemof N point
particles with kinetic energy(312)NkT,this yields the familiar virial equation
of state, pV : NkT - (a term determined by the ensemble average of the
virial of the intermolecular forces or of the forces between the nuclei and
electrons of the moleculesin the system).
The analogous result is obtained using quantum mechanicswith the pV
operator being defined as (fl): Ql3){i - +r} (Hirschfelderet al. 1967:
these authors define the virial as - it'). The analogy is complete,with the
pressureexertedon an element of surfacearea in the quantum casegiven by
- A ' n dS and its virial by - r?'n dS. The pressurein this case is not
uniform throughout the systemor over the atomic surfaceand, instead,each
6.5 H Y P E R V I R I A L T I . I E O R E \ 1\ \ D I]O\D PROPERTIES 241

i n te rat om ic s ur f ac ema k e s i ts o w n c o n tri b u tro n It' r the surfacevi ri al , a

q u a n t i t yd e n o t e db y / r ( Q l Q ' ) i n e q n ( 6 . 1 0 ' l t .
I , ( O ): I n = ' 7 s { f ) O ' ) (6.104)
where
l ; ( c ) l Q ': ) $ d s ( o 1 Q
r )' r. . F . n .
Fo r a m olec ule in a n e q u i l i b ri u m c o n fi g u ra ti on.A f(O) : - A E (Q) and
one has
A E ( O )- + L l o ( Q :) ] I r , + a / ; ( Q l Q ' ) , (6.r05)
sh o w ingt hat eac h l ;(O l O ' ) me a s u re sth e p o te n ti alenergycontri buti on to
AE(O)arisingfrom the pressureexertedon the interatomicsurfacewith atom
Q'. The virial of the surfaceforce has the dimensionsof force per unit length,
or surfacetensionand 7 ,(OlO') can be interpretedas providing a measureof
the surfaceenergy associatedwith the bond to atom O'.
The separationof the virial into basin and surfacecontributions requires
a choice of origin and a natural and physicallyusefulchoice is to place the
origin of the position vector for atom A, the vector ra, at the nucleusof the
atom in question.If the vectorfrom the nucleusof atom A to that of atom B is
denotedby Ruu,then Ruo- ra - ro and, recallingthat the surfacenormals are
relatedby n^ : - rb, one can write an expressionfor the sum of the surface
virials for the interatomic surfacebetweenatoms A and B as
r ; ( A l B )+ / ; ( B l A ): R " o $ d S ( A l Br,) F . n u . (6.106)
, e i n te g ra lo n th e ri g h t-h andsi deof eqn (6.106)i s a
Fo r a diat om icm ole c u l eth
measureof the contribution of the interatomic surfaceenergy to the total
energychangeAE.
For real d, the statementof the atomic hypervirial theorem yielding the
atomic averageof the commutator is (seeeqn (5.43))

dl)n + cc}: {(Nh2lqd#ds{V/*Grl'+ Grl,*v,l'

(tvl2){<1fr,
- /*v(d,lD- vG,l,*)/) n. (6.107)
An operator, which provides a measure of the distbrtion of the gradient
vector field of an atom from its radial form characteristicof the free state to
that found for a bound atom, is obtainedfrom the commutator of fi and 12,r
b e i n gt he r adial dis t a n c eo f th e e l e c tro nfro m th e n ucl eus.S ubsti tuti onof the
result
1fr,r21: - ft2lm)(l+ 2r' V) (6.r08)
i n to eqn ( 6. 107)f or d : 1 2 y i e l d s

3r\/(Cl) + i ,Y p d r :
JO f
otn,.,rn(r) (6.10e)
1,,11
-a- THE MECHANICS OF AN ATOM IN A MOLECULE 6.5

The first term of the commutator yields three times the averagenumber of
electronsin an atom. The secondintegral representsanother way of counting
electronsover the basin of an atom, a way which reflectsthe distortion of the
gradient vector field of the chargedensity that is causedby the formation of
chemical bonds. Becausethe nucleus acts as an attractor in the gradient
vector field, the vector Y p and the radial vector r are parallel and oppositely
directed for the undistorted gradient vector field of a free atom (Fig. 6.13).
Since the surfaceintegral vanishesin this case,the integral of the product
r 'Y p over the basin of the free atom yields - 3N (O). For the distorted field
of a bound atom the two vectors are no longer antiparallel and their dot
product integratesto a number which is lessthan 3l{(A). The stronger the
interaction with a neighbouring atom, the greater the distortion of the
gradient vector field, the smaller the overlap of r and Vp, and the larger the
differencebetween the two ways of counting electrons in eqn (6.109),a
differencewhich is measuredby the integral over the atomic surface.
The surfaceintegral may, of course,be expressedas a sum, there being an
integral over the interatomic surfacefor each atom bonded to the one in
question.Referenceto Fig. 6.I 3 makesclearthat the distortion of the gradient
vector field away from radial symmetry is a result of the formation of the
interatomic surfaces,that is, as a result of bonds to neighbouringatoms.The
trajectoriesof - Y p, which are radially directedin the neighbourhoodof the
nucleus,bend to avoid a bond critical point and then run parallel to the
associatedinteratomic surface.Each surfaceintegral relatesthe distortion of
the gradient vector field which resultsfrom the formation of the associated
bond path to a correspondingaveragenumber of electrons.A value for this
number of electronsmay be determined for each bond in a molecule by
combining the surfaceintegralsfor each pair of atoms linked by a bond path
in t he m anner i n d i c a te di n e q n (6 .1 1 0 )
$ d s ( A I B , r ) p ( r ) r " ' n " ( r )+ $ d s ( B I A , r ) p ( r ) r o ' n o ( r )
: R u b ' f d s ( A I B , r ) p ( r ) n " ( r ) ( 6 . 11 o )

where,as in eqn (6.106),Ruois the vector linking nucleusA to nucleusB. ln a

polyatomic moleculethere is one such surfaceintegral for every bond path.
The valuesof this integralfor somediatomic moleculesin their ground states
ar e ( B aderand Wi b e rg 1 9 8 7 ):H r, 1 .0 8 0B ; r , 3.097;C r, 4.176;N t, 4.902;N j ,
4.474C ; . t , 3 . 9 3 0O
; H , 1 . 9 1 9O ; ;,4.219.
If one reinstatesthe units of the commutator, h2f m, then the surfaceintegrals
in eqns (6.109) and (6.110) have the dimensions of energy x (length)2.
Thus, division of the integral for the pair of atoms linked by a bond path in
eqn (6.110)by Rlo yieldsa quantity with the dimensionsof energy.Sincethe
value of the surfaceintegral is determined by the changesin the gradient
vector fieldsof the atoms resultingfrom the formation of the bond path and
hence of the bond which links them. the energy it determinesshould be
 o

\\

C
Frc. 6.13. (a) The trajectories of Vp representing the gradient vector field of a lree atom.
(b), (c) The distorted gradient vector fields of bound atoms. The plane in (b) contains H-C-C-H
nuclei of staggeredethane.The plane shown in (c)contains the H-C C-H nuclei representingthe
bridging C-C bond in bicyclo-butane (seeTable 6.7). The C-C bridging bond is rerr curred
(strained)and, correspondingly,its energy is significantly less than that of a C C bond in an
acyclic hydrocarbon.
1 1 i
/.++ T H E M E C H A N I C SO F A N A T O M I N A M O L E C U L E 6.5

proportional to the correspondingchange in the energy of the atoms, the

energy of the bond. The sum of the bond energiesdetermined in this manner
for a given molecule should, therefore,equal the heat of atomization of the
molecule in its vibrationlessstate at 0 K. This procedure works well for
systemswith only small or no interatomic chargetransfer (Bader and Wiberg
1987).Table 6.7 lists the values of the bond energiesdetermined in this
manner for a number of acyclic, cyclic, and bicyclic hydrocarbon molecules,
including molecules with strained C__Cbonds. The charge densities were
obtained from SCF calculationsand the proportionality constantsfor the
C-H and C-C bond energieswere determined by fitting the experimental
energies of atomization of methane and ethane at 0 K. The agreement
between the calculated atomrzatron energy AEi and the experimental value
AE! is satisfactoryand the individual bond energiesexhibit the anticipated
trends, with the C-C bond energy increasing with the degreeof branching
(compare propane, iso-butane, and neo-pentane) and decreasing with a
decreasein ring size in cyclic systems.It is found that the constant of
proportionality linking the value of the integral and the energy approaches
the limiting value of lln as the chargedcnsity approachesthat obtained at the
Hartree-Fock limit.
The purpose of these examples is to illustrate how one can relate a
molecular property to a sum of bond contributions in a rigorous manner. The
method has the further advantageof yielding directly the changein the value
of the property from that observed when the atoms of the molecule are in
their free state. The method is applicable to any property associatedwith a
generatorG via its commutator with the Hamiltonian of the system
A quantity of some physical interest, whose commutator averagewith H
yields the surface integral of the flux in the gradient vector of the charge
dens it y , is gi v e n b y th e o p e ra to r -l n p ' , w here p' : pl N , the densi ty
normalized to one. In the density formulation of quantum mechanics the
entropy of a systemis definedas S[p] : - Tr(plnp) where p is the density
'surprisal' (Alhassid and
operator. The quantity - ln p' has been termed the
Levine 1977)and it, along with SIp], have beenshown to be constantsof the
motion. The commutator averageof Il and - ln p'for an atom in a molecule
also vanishesidentically as a result of the zero-flux surfacecondition. Thus

t{<lH, - ln p'))n+ cc}: (h2l2m)o' p'(r)dr

f

l2m) 6to o'(r).n(r)- 0 . ( 6 . 1 1 1 )

(h2
{ot
Sincethis resultis obtainedfor both stationaryand non-stationarystates,the
energy and Lagrangian functionals are stationary with respectto a variation
with - ln p' as generator.The averaging of this generator over the basin of
6,5 H Y P E R V I R I A LT H E O R E N f{ \ D BO\D PROPERTIES 245

T a b l e 6 .7
Theoretical hond €fi€t'tlit's tn ht rlr,tr'tJfht)tl.\

Molecule Bond energ) Energy at atomization

(kJ/mol) tkJ mol)

CH CC AE; ^EZ

Methane 438.52 1154.1 1754.1

Ethane 440.32 331.00 2972.9 2972.9

Propane 441.83 I )( 339.28 4199.9 4201.0

440.s3(2)
4 3 9r.5 (3 )

n-Butane 440.62(1) 347.s2(4) 5432.5 5431.5

, / ^ . 1
44t.s8(2) 33e.e5(5)
Yf( 4.4\3./ 4 3 9 .8 6 (3 )
f \ t \ _

1 -

n-Pentane 438.52(1) 34e.ee(s) 66s9.4 6661.7

440.03(2) 34r.00(6)
4 4 1 .5 0 (3 )
43e.32(4)

iso-Butane 443.5s(1) 348.44 5441.8 5437.9

431.98(2)
l1 43e.82(3)
V\,,'
r l l
o
z

neo-Pentane 438.99 357.40 6697.6 6671.4

Cyclopropane 444.42 2 9 7 .2 3 3s58.2 3556.0

Cyclohexane 442.50(e) 3 3 5 .6 0 7310.7 1367.2

I 440.03(a)
,"-\,-\
\,d-e
l l
a

B i c y c l o [1 . 1 . 0 ] b u t a n e 4 3 8 .e 4)(l 263.38(4) 4n2.0 4106.2

442.08(2) 3 0 3 .3 0 (5 )
z
436.73(3)
].16 THE MECHANICS OF AN ATO\I I\ { MOLECULE

the atom yields

1 (Q): - p ' tn p' dr,

,[,
a quantity analogous to the missing information function. This quantity has
the property of decreasing in value when a free atom becomes bound (Bader
1980).

References

A l h a s s i d ,Y . a n d L e v i n e ,R . D . ( 1 9 1 7 )J. . C h e m .P h y s . 6 7 , 4 3 2 1 .
B a d e r ,R . F . W . ( 1 9 8 0 ) J. . c h e m .P h y s . 7 3 , 2 8 7 1 .
B a d e r ,R . F . W . ( 1 9 8 6 ) C . a n .J . C h e m . 6 4 ,1 0 3 6 .
B a d e r ,R . F . W . a n d B e c k e r ,P . ( 1 9 8 8 ) .C h e m .P h v s .L e t t . 1 4 8 , 4 5 2 .
B a d e r ,R . F . W . a n d B e d d a l l ,P . M . ( 1 9 1 3 )J. . A m . c ' h e mS . oc.95,305.
B a d e r ,R . F . W . a n d C h a n g ,C . ( 1 9 8 8 a )J. . P h y s .C h e m . 9 3 , 2 9 4 6 .
B a d e r ,R . F . W . a n d C h a n g ,C . ( 1 9 8 9 b ) J. . P h _ v sC. h e m . 9 3 , 5 0 9 5 .
Bader, R. F. W. and Messer,R. R. (1974).Can. J. Chem.52,2268.
B a d e r ,R . F . W . a n d N g u y e n - D a n g ,T . T . ( 1 9 8 1 ) A . d t : .Q u a n t u mC h e m .1 4 , 6 3 .
B a d e r ,R . F . W . a n d P r e s t o n ,H . J . T . ( 1 9 7 0 ) .T h e o r .C h i m .A c t a 1 7 , 3 8 4 .
Bader, R. F. W. and Wiberg, K. B.(1987). Density matices and densityfunctionals
(ed. R. Erdahl and V. H. Smith),671, D. Reidel Pub. Co.
B a d e r ,R . F . W . , H e n n e k e r ,W . H . a n d C a d e ,P . E . ( 1 9 6 7 )J. . C h e m .P h y s . 4 6 , 3 3 4 1 .
Bader,R. F. W., Carroll, M. T., Cheeseman, J. R., and Chang,C. (1987a).J. Am. chem.
S o c .1 0 9 , 7 9 6 8 .
Bader, R. F. W., Larouche, A., Gatti, C., Carroll, M. T., MacDougall, P. J., and
W i b e r g , K . B . ( 1 9 8 7 b )J. . C h e m .P h y s . 8 7 , 1 1 4 2 .
Bader, R. F. W., Cheeseman,J. R., Laidig, K. E., Wiberg, K. B. and Breneman,C.
( 1 9 9 0 ) J. . A m . c h e m .S o c .1 1 2 , 6 5 3 0 .
Benson,S. W. (1968).Thermochemical kinetics.Wiley, New York.
B e r l i n ,T . ( 1 9 5 1 )J. . C h e m .P h y s .1 9 , 2 0 8 .
Biegler-Konig,F. W., Bader, R. F. W., Tang, T. H. (1982).J . comput.Chem.3, 3ll.
Chang, C. and Bader, R. F. W. (1990).Unpublished results.
Coulson, C. A. and Mollitt, W. E. (1949).Phil. Mag.40, 1.
Franklin, J. L. (1949).Ind. Eng1n17 Chem.4l, 1070.
Hariharan, P. C. and Pople,J. A. (1972).Chem. Phys.Lett. 16,211.
Hirschfelder,J. O., Curtiss,C. F., and Bird, R. B. (1961).Molecular theoryof gasesand
liquids.Wiley, New York.
H o h e n b e r g ,P . a n d K o h n , W . ( 1 9 6 4 ) P . h y s .R e r . 8 1 3 6 , 8 6 4 .
H u , C . D . a n d L a n g r e t h ,D . D . ( 1 9 8 5 ) P . f i r ' s .S c r . 3 2 , 3 9 1 .
K e i t h , T . ( 1 9 9 0 )P. e r s o n a cl o m m u n i c a t i o n .
Kopp, H. (1855).As reported in Glasstone.S. (1946).TexthookoJ'physicalchemistry
( 2 n d e d n ) .p . 5 2 5 .V a n N o s t r a n d .N e w 'Y o r k .
K r i s h n a n ,R . , F r i s c h ,M . J . . a n d P o p l e .J . A . ( 1 9 8 0 )J. . C h e m .P h 1 , s . 7 2 , 4 2 4 4 .
L a n g r e t h ,D . D . a n d M e h l , M . J . ( 1 9 8 3 ) P . l r r ' . sR. e r . 8 2 8 , 1 8 0 9 .
 REFERE\CES 247

Libit, L. and Hoffmann, R' (1974).J. Ant. chent'Stic' 96' l-110'

L o w d i n , P . - O . ( 1 9 5 9 )J. . M o l . S p e c t r o s c . 34. 6 '
Messiah,A. (1958).Quantummechanics. Vol. l. \orth-Holland. Amsterdam'
I.
Morse, p. M. and Feshbach, H. (1953). .\/etftodso.l-theoreticalphysics,Vol'
McGraw-Hill, New York.
parr, R. G. and Yang, W. (1989). Densitl'-.lurtt'tionul theort' of atoms and molecules,
Oxford University Press,New York.
pauling, L. (1960).The nature of the chemicalbond l3rd edn).Cornell University Press'
Ithaca, New York.
payne, p. W. and Allen, L. C. (1911). Applicationsof electronic structure theory.
(Modrrn theoreticalchemistry),Vol. IV (ed. H. F. Schaefer),29, Plenum Press,New
York.
Pittam, D. A. and Pilcher, G. (1972).J. chem. Soc.,Faraday Trans. I 68,2224'
Pitzer,R. M. (1983).Acc' chem. Res.16,207.
Prosen,E. J.,Johnson,W. H., and Rossini,F. D' (1946).J. Res.nat' Bur' Stand'37, 5l'
Rauk. A., Allen, L. C., and Clementi, E. (1970)'J ' Chem' Phys' 52, 4133'
Riess,I. and Mtinch, W. (1981).Theor.Chim. Acta 58,295'
Schrodinger,E. (1926).Ann. d. Physik 81, 109.
Schulman,J. M. and Disch, R. L. (1985).Chem. Phys.Lett. 113,291.
S l a t e r ,J . C . ( 1 9 3 3 )J. . C h e m .P h y s '1 , 6 8 7 .
Slee,T., Larouche, A., and Bader, R. F. W. (1988).J . Phys. Chem. 92, 6219.
Srebrenik,S. and Bader, R. F' W. (1975).J. Chem' Phys' 63,3945'
Srebrenik,S.,Bader,R. F. W., and Nguyen-Dang,T. T. (1978).J. Chem' Phys'68,3667'
Taft, R. W. (1956).ln Stericeffectsorganicchemistry(ed.M' S' Newman),Chapter 13'
Wiley, New York.
Taft. R. W. (1960).J. Phys'Chem.64, 1805.
Walsh, A. D. (t947). Nature, London 159,167,7I2'
Wiberg, K. B. (1984).J. comput.Chem.5, 197.
Wiberg, K. B. and Laidig, K' E. (1987).J. Am. chem.Soc' 109, 5935'
Wiberg, K. B. and Wendolowski,J. J. (1981). Proc. nat. Acad. Sci. US/ 78, 6561'
wiberg, K. B., Bader,R. F. W., and Lau, c. D. H. (1987).J. Am. Chem.soc. 109, 1001.
 7

CHEMICAL MODELS AND THE LAPLACIAN

OF THE CHARGEDENSITY

Indeedit seemshardly likely that much progresscan be madein the solutionof the
difficultproblemsrelatingto chemicalcombinationby assigningin advancedefinite
lawsof forcebetweenthe positiveand negativeconstituentsof an atom,and then on
the basisof theselawsbuildingup mechanical modelsof the atom.
G. N . Lew i s(1916)

7.1 The physical basis of the Lewis electron pair model

7.1.1 The electron pair at any cost

The 1916paper of Lewis, introducing the concept of the electron pair, is one
of the most important papersto be publishedin chemistry in this century. It is
a revelation to read this paper and realize the extent to which his ideas
continue to dominate chemical thinking. One must marvel at the intuitive
leap that was involved in advancing the notion of the importance of the
electron pair, coming as it did beforethe developmentof quantum mechanics
or the still later introduction of the concept of electron spin. Indeed the
opening quotation from the paper by Lewis was pennedin responseto Bohr's
then recently advanced mechanical model of the atom. So strong was his
conviction regarding the existenceof a closely coupled pair of electrons(we
would say a localized pair) as being necessaryfor the understanding of
chemical bonding, he felt forced to postulate the failure of Coulomb's law for
small interelectronic separations,and to thus necessarilyreject any model
which made use of this law. It is clear from the quotation given at the
beginning of Chapter 5 that Lewis later (1933) abandoned this extreme
position.
The Lewis model is secondonly to the molecular structure hypothesisitself
in providing a conceptual basis for much of present-daychemical thinking,
particularly with regard to models of molecular geometry and reactivity. It is
essentialthat the physical basis of this model be establishedif one is to place
all of the vital concepts of chemistry on a firm theoretical footing.
It is shown in this chapter that, while the Lewis model of the electron pair
does not correspondto the existenceof spatially localizedpairs of electrons,it
does find a more abstract but no less real physical expression in the
topological properties of the Laplacian of the charge density, the quantity
Ytp(r) as definedin eqn (2.3)(Bader and Essen1984;Bader et al. 1984).The
7.1 T H E L E W I S E L E C T R O \ P A IR \ I O D E L 249

L a p l ac ianof t he c h a rg ed e n s i ty Y 2 p l rl . a l o n g u i t h p(r) and V p(r), serveto

define the conceptualmodels of chemistrl and thel' provide the necessary
basis for the theoreticaldescription of these models.The Laplacian of the
chargedensityY'p(r), as demonstratedin the two precedingchapters,plays a
dominant role throughout the theor,vof atoms in molecules.It is shown here
that the Laplacianprovidesa link betweentheory and the chemicalmodelsof
geometry and reactivity that are basedupon the Lewis model.
What Lewis (1916)consideredto be one of the major accomplishmentsof
his model, was the unified description it provided of chemical bonding.
'However, according to the theory which I am now presenting, it is not
necessary to consider the two extreme types of chemical combination,
correspondingto the polar and the very nonpolar compounds, as different in
kind, but only as different in degree.' The different degrees of polarity
correspondedto the equal or unequal sharing of the electron pair betweenthe
'kernels' of the two atoms, and the two resulting extremesof bonding are to-
day referredto as the covalent and ionic limits. It is shown in SectionT.4 that
the Laplacian of the charge density provides a physical basis for the
classification of the continuous spectrum of atomic interactions and their
mechanicalconsequences, which lie betweenthesetwo extremesof bonding.

7.1.2 Are electronslocalizedin pairs?

The concept of the localized electron pair has been a central theme in the
developmentof bonding theoriessince Lewis first postulated that a chemical
bond was a consequence of a sharedelectronpair. However,the total electron
densitydistribution in a moleculeshowsno indication of discretebonding or
non-bonding electron pairs. The principal topological property of the charge
density of a many-electronsystemis that, in general,it exhibits local maxima
only at the positions of nuclei.When one sums the individual orbital densities
in the determination of the total charge density, all suggestionsof spatially
localized patterns of charge and the associatednodes disappear to yield the
relatively simple topology exhibited by the total chargedensity p. While this
topology, as demonstratedin precedingchapters,provides a faithful mapping
of the concepts of atoms, bonds, and structure, it does not provide any
indication of the maxima in a chargedistribution which would correspond to
the spatially localized pairs of bonded and non-bonded electronsas anticip-
ated on the basis of the Lewis model. The possibility of obtaining a set of
localized orbitals does not imply any physical localization of electrons. A
choice of a particular localized set is arbitrary, becausethe summation of the
orbital densitiesof any set,obtained by a unitary transformation of another,
yields the same total density.Thus, using the forms of individual orbitals to
describethe charge distribution is both incorrect and ambiguous. Does a
'S' 'p' lone pair or two equivalent
keto oxygen for example,poSSeSS an and a
r50 CHEMICAL MODELSAND THE LAPLACIAN 7.1

lone pairs as pictured on the basis of the canonical and localized sets of
orbitals, respectively?The forms exhibited by individual orbitals are not
evidentin the topology of the total chargedistribution.Thesecommentsare,
of course,not criticisms of the orbital model, which is the model for the
prediction and understanding of the electronicstructure of a many-electron
system,but rather of the misuseof this model to describethe spatialstructure
of the electron distribution through illustrations of individual orbitals or
their corresponding densities.
The concept of alocalized electron pair implies that there existsa region of
real space in which there is a high probability of finding two electrons of
opposite spin and for which there is a correspondingly small probability of
exchangeof theseelectronswith the electronsin other regions.Such physical
localization of charge is a direct result of the operation of the Pauli exclusion
principle as it affectsthe pair distribution function for electrons(Bader and
Stephens1975).The charge density p(r), the density of electronsat r,, is
given by N times the probability density of finding one electron at r,. Its
integral over the coordinates of one electron gives lf, the total number of
electrons.Similarly,the pair densityp(rr, rr), definedin eqns(F--'7.7) or (E1.3),
the density of pairs of electronswith coordinates11 and rr, is given by
l{(N - l)/2 times the probability density that one electron is at r, when
the other is at rr. Its integral over the coordinatesof both electronsgives
l{(l{ - l)12,the total number of distinct pairs of electrons.The antisymmetry
requirement of the Pauli principle is imposed on the pair density through the
introduction of the so-calledFermi hole. The theoreticaldevelopmentof this
most important effect which the resulting Fermi correlation has on the pair
distribution function is discussedin Section87.1. We give here a qualitative
discussionof how the operation of the Pauli exclusion principle can indeed
lead to the localization of pairs of electrons,but follow this with a demon-
stration that the requirementsfor achievingsuch localization are, in general,
not met.
As an electron moves through space,it carries with it a doppelganger,its
Fermi hole. The Fermi hole is a distribution function defined relative to an
uncorrelatedpair density,which determinesthe decreasein the probability of
finding an electron with the samespin as some referenceelectron,relative to a
given position of the referenceelectron. The value of the Fermi hole at the
position of an a (or B) referenceelectron equals the total density of a (or p)
electrons at that point, thereby ensuring that there is a zero probability of
finding density of another a (or B) electron at the position of the reference
electron.The Fermi hole, when integrated over all spacefor a fixed position
of the referenceelectron,correspondsto the removal of one electroniccharge
of identical spin, and one may equally interpret the Fermi hole as a
distribution function which describesthe spatial delocalizationof the charge
 7.1 T H E L E W I S E L E C T R O NP A I R M O D E L 251

of the referenceelectron.An electron can go only where its hole goes and, if
the Fermi hole is localized,then so is the electron (Bader and Stephens1975).
I
Consider an electronof a spin in the proximity of a nucleus.Becausethe
electron is tightly bound in a deep potential well, its Fermi hole is strongly
localizedin its immediate vicinity. If the Fermi hole is localizedso as to equal
its maximum value p"-the total density of a electrons-at every point over
this region of space,then all other electronsof a spin are excluded from the
potential well. The same result will be obtained for an electron of p spin and,
as a consequence,a pair of electrons is confined to a region of spacefrom
which all other electrons of both a and p spin are excluded. Thus, Fermi
'pair up' electrons.Rather, since there is
correlation does not act directly to
no Fermi correlation between electrons of opposite spin, an a, B pair is
obtained as a result of all other electronsof both a and B spin being excluded
from a given region of spacein which they are both bound by some attractive
force. Repulsionsbetween the electrons,of course,act in opposition to this
effectand one finds that the long-rangenature of the Coulomb force disrupts
and limits the localization of charge (Sinanoglu 1962).
In general,one finds such extreme localization of the Fermi hole for an
electronpair in the core of an atom (Baderand Stephens1975).In thesecases
it is possible to find a spherical boundary such that the contained Fermi
correlation is maximized, a result which, as shown in Section E7.1, is
equivalent to minimizing the fluctuation in the average population of the
contained charge.It is in this senseof being able to extremize the degreeof
isolation of the electronsto a given region of spacethrough a variation of the
boundary that the electrons are referred to as being physically localized.It
corresponds to finding a region of space such that the exchange of the
electrons within the region with the rest of the system is minimized. Any
change in its boundary increasesthe exchangeand decreasesthe contained
Fermi correlation and the degree of localization. The same spatial localtza-
tion of the Fermi hole is found in a few simple hydridessuch as LiH, BeHr,
and BH. in which, in addition to the core on Li, Be, or B, there are,
respectively,one, two, and three pairs of electrons,each localizedon a proton.
However, in molecules where the bonding is predominantly covalent, and
particularly when there are four or more electron pairs in a valenceshell such
as in CHo, boundedregionsdo not existwhichmaximizethe containedFermi
correlation and the electronsare not physically localized.Instead,the Fermi
hole of each electron, and hence each electron itself, is delocalized to a
significant extent throughout the valence shell of the central atom.
It is shown in Section E7.1 that the orbital requirement for physical
localization of pairs of electronsis not that a set of orbitals be localized,but
that each orbital be localized to its own separate region of space. Such
separatelocalization is apparent even in the canonical set of orbitals for the
1i')
CHEMICAL MODELS AND THE LAPLACIAN 7.1

same molecules,LiH, BeHr, and BH., etc.,which, as noted above,exhibit

localizationof electronpairs. Localizedorbitals are in general,however,not
that localized and they are not localizedto separateregions of space(Daudel
et al.1976).
A physically localized pair of electrons is maximally isolated from the
remainder of the system in that its exchange and all of its correlative
interactions with the remainder of the system are minimized. Hence, the
degreeof physical localization as determined by the properties of the Fermi
correlation parallels the relative importance of the intra- and interpair
correlation energiesas determined by various 'correlated pair' theories. In
LiH, BeH r, and BH3 (Ebbing and Henderson 1965; Kutzelnigg 1973),for
example,intrapair correlation accounts,respectively,for over 90, for 86, and
for 77 per cent of the total correlation energy, in systems for which the
correspondingFermrlocalizations are 96,93, and82 per cent. As pointed out
by Davidson (1972),by Sinanoglu and Skutnik (1968),and by Kutzelnigg
(1973),this is not the general result and, in moleculessuch as CHo and NH,
the intrapair correlation contribution to the total correlation energydrops to
56 and 20 per cent, respectively.
The study of the localizability of the Fermi correlation and the related
fluctuation in the averageelectron population, of the relative importance of
the intra- and interpair contributions to the total correlation energy and the
study of the degreeof the separatelocalizability of orbitals all provide similar
information and all indicate that, in general,the physics of a many-electron
system is not dominated by the behaviour and properties of individual
electron pairs. In addition, localized pairs of electrons, bonded or non-
bonded, are not evident in the topological properties of the charge density
and, moreover, the pair density does not, in general,define regions of space
beyond an atomic core in which pairs of electrons are localized.

7.1.3 Propertiesof the Laplacian of the chargedensity

Where then to look for the Lewis model, a model which in the light of its
ubiquitous and constant use throughout chemistry must most certainly be
rooted in the physics governing a molecular system? If one reads the
introductory chapter on fields in Morse and Feshbach'sbook Methods of
theoretical physics (1953), one finds a statement to the effect that the
Laplacian of a scalar field is a very important property, for it determines
where the field is locally concentrated and depleted.The Laplacian of the
chargedensityat a point r in space,the quantity Y2p(r), is definedin eqn (2.3).
This property of the Laplacian of determining where electronic charge is
locally concentrated and depleted follows from its definition as the limiting
difference between the two first derivatives which bracket the point in
question as definedin eqn (2.2) and illustratedin Fig. 2.2.
 7.1 T H E L E W I S E L E C T R O \ P A IR N { O D E L 253

'local chargeconcentration'as defined

To make clear what is meant by a
by a secondderivative,considerthe function,/'(.x)displayedin Fig. 7.1 along
with its first and secondderivative.Since/(,x)isa steeplyrising function in the
region of xr, the magnitude of the slope at.xr - Ar is greater than that at
xr * Ax and, since both slopesare negative,the curvature of /(x) at x, is
positive.The function exhibits a shoulder in the region of the point xr and the
slopeat x, * Ax is of greatermagnitude than at x, - Ax. Thus, the curvature
of f (x) is negative in the region where/(x) exhibits a shoulder. The function
exhibits a point of inflexion between x, and x, where the curvature is zero.
From eqn (2.2)for the curvature, one seesthat, where the curvature of/(x) is
negative,the value of/(x) at x is greater than the averageof its values at the
neighbouring points, x * dx and x - dx, with the reversebeing true where
the curvature is positive. It is in this sensethat/(x) is said to be concentrated
l in regions where d2f (x)ldx2 < 0 and depleted where d2f (x)ldr' > 0. To
I graphically emphasizethe fact that/(x)is concentratedwhere its curvature is
negative, the negative of the curvature is plotted. Note that the curvature
i

exhibits its largest negative value at x, and f(x) is said to be maximally

concentratedat that point, the point wherethe negatiueof the curuatureexhibits
a maximum.
The slope of /(x) (Fig. 7.1) is always negative.Thus, /(x) itself does not
'(x):0, asidefrom the
exhibit either maxima or minima, i.e. points wheref
maximum at x :0. We seethat the existenceof regionswhere a function is
locally concentrated or depleted as determined by the sign of its second
derivativedoes not imply the correspondingexistenceof maxima or minima
in the function itself.
All of the above considerationscarry over into three dimensions and, in
particular, it follows from eqn (2.2) that the value of p(r) is greater than the
averageof its valuesover an infinitesimal spherecentredon r when the sum of
the three curvaturesof p is negative,that is, when Y'p(r) < 0, and p(r) is less
than this averagewhen Ytp(r) > 0. The charge density decaysexponentially
from a nucleus and, in general,the curvature of p(r) along a radial line from
the nucleus is positive, as for /(x) at xt in Fig. 7.1. Its two curvatures
perpendicular to a radial line are, however, negative. Thus, Y2p(r) may
change sign even in the absenceof a shoulder in the charge distribution, the
presenceof which in any event is of relatively rare occurrence.It is important
to distinguish betweenlocal maxima and minima or shouldersin the charge
density and corresponding extrema in its Laplacian distribution, and this is
best done by first comparing these two distributions for an atom with a
spherically symmetric distribution of charge.
Figure 7.2 displays the value of the charge density and of its associated
Laplacian distribution as a function of position in the y, z-planefor an argon
atom. Sincethe chargedistribution is sphericallysymmetric,the samedisplay
is obtained for any plane containing the nucleus.The charge density exhibits
)<A CH E M I C A L M O D E L S A N D T H E L A P L A C I A N 7.1

f(t) = S e x p (- 7 r ) + e x p ( - 1O ( x - O . 5 ) * * 2 )
I

6
xrl
5
X
Y t

X2
z

'1

0
0.2 0.4 0.6 0.8 1.0
X

First d e rivotive
0

20

x -30

-40

-50

-ou

Second derivotive
20

10

-10
X

I 20

-30

-40

-50
0.2 0.4 0.6 1.0

F t c . 7 . 1 . P l o t s o f a m o n o t o n i c a l l yd e c r e a s i n gf u n c t i o n . f ( x ) a n di t s f i r s t a n d s e c o n dd e r i v a t i v e s .
The negativeof the secondderivative is sho*n to emphasizethat a function is concentratedin a
r e g i o n w h e r e i t s s e c o n dd e r i v a t i v ei s n e g a t i v e .
7.1 T H E L E W I SE L E C T R O \ P . \ I R\ I O D E L 255

-v'P
Ftc.7.2. Relief maps of the charge density and the negative of its Laplacian for a plane
containing the nucleusof an argon atom. Function valuesabove an arbitrary magnitude are not
shown. At the point A, the curvature of p along z (the radial curvature) is positive while the
curvature along the y-axis (a tangential curvature) is negative. In addition to the spike-like
maximum in -Yzp at the nucleus,there are two shellsof charge concentration and three shells
of charge depletion corresponding to the three quantum shells in this atom.

a singlemaximum at the position of the nucleus(the maximum value is not

shown) and decays monotonically for motion away from the nucleus along
any radial line. The shell structure of an atom is not evident in the
distribution of chargethroughout three-dimensionalspace.Shell structureis
made evident in the radial distribution function, a one-dimensionaldistribu-
tion obtained by expressingthe charge density in terms of spherical polar
coordinatesfollowed by its integration over the angular coordinates.The
radial distribution function is the probability of finding electronic charge in
256 CHEMICAL MODELSAND THE LAPLACIAN 7.1

the infinitesimal volume lying betweentwo concentric spheresof radii r and

r * dr. Thus a maximum in a radial distribution function determinesa value
of r at which one is most likely to find electronic charge averaged over all
angles,but the actual distribution of charge in three-dimensionalspacedoes
not exhibit a maximum at the corresponding value of r.
To understand the very different topology exhibited by the Laplacian of
the charge distribution, we consider a spherical surfaceof constant density.
An axis through the origin, say the z-axis, is normal to this surface and
intersectsit at the point (0,0, z). The charge density increaseswith decreasing
z and, in general,the curvature along such a radial line is found to be positive.
Thus Apl}z < 0 and A2plAz2 > 0. Sincethe curvature of p along z is positive,
the value of p(0, 0, z) is less than the average of its values at the points
(0,0, z + dz). The charge density is however, a maximum at (0, 0, z) in the
tangent plane, Apl0x : AplAy: 0, and, hence,the corresponding curvatures
are negative (see Fig. 7.2). Thus p(0, 0, z) is greater than the average of its
valuesat ( * dx, 0, z) and (0, t dy, z). If the sum of the magnitudesof the two
negativecurvatures,those perpendicularto a radial line, exceedsthe value of
the positive or parallel curvature,then Y'p(0,0,2) < 0 and the value of p at
this point is greater than that obtained by averaging p over all points in the
neighbourhoodof the point (0, 0, z).By determininga radius in an atom at
which Y'p(r) is a minimum (or a maximum), one determinesa radius of a
sphere on which electronic charge is maximally concentrated(or depleted).
The observation that p exhibits a maximum at a nuclear position does not
imply that the contraction of p towards the nucleusis the dominant one at all
distancesfrom a nucleus.Indeed this cannot be so, for the integral of Y'p(r)
over an atom free or bound must vanish, as a consequenceof the zero-flux
surface condition. Because of this, in addition to the regions where the
contraction of p towards the nucleus along a radial line is dominant and
Yt p(r) > 0, there must exist a region or regions where the magnitudes of the
curvatures of p perpendicularto a radial line are dominant, yielding negative
valuesfor the Laplacian of p. Thus Y'p(r) exhibits sphericalnodesin an atom
(values of the radius for which Yt p :0) and their number is found to be
related to its shell structure (Fig. 7.2).Thereare pairs of regions(one negative,
one positive) for each principal quantum shell, with the innermost region
being a spike-like maximum in charge concentration (Bader and Beddall
1972;Bader and Essen1984).
The number of shellsdefined in this manner by the Laplacian of the charge
density or in terms of the number of maxima in the radial distribution
function can be lessthan the principal quantum number for elementspast Ca.
The Laplacian distribution, while exhibiting the required number of outer
maxima and minima for elements with z > 40, does not undergo the
correspondingchangesin sign (Sagaret a/. 1988;Shi and Boyd 1988).The
addition of relativisticcorrectionsto a wave function causesthe outer charge
7.1 T H E L E W I SE L E C T R O \ P A I R \ I O D E L 257

distribution to contract inwards for atoms r.rith large Z, Savin (personal

communication)has observedthat even rrith this contraction, the missing
shell structureis not presentin the Laplacian distribution.
The presenceof a shellof chargeconcentrationand one of chargedepletion
in the Laplacian distribution for each quantum shell is demonstratedin Fig.
7.3 for the four quantum shellsof the Kr atom. While the diagram displays
the behaviour of - Y'p along a given radial line. it should be borne in mind
that this is a representationof the structure present in three-dimensional

-v'p

-1.3x10a
-3.2x106

Frc. 7.3. Schematicrepresentationof the profile of - Y2p for the krypton atom along a radial
line. Unlike a radial distribution plot, which is a one-dimensional function. the Laplacian
distribution displays shell structure in three-dimensionalspace.
258 CHEMICALMODELSAND THE LAPLACIAN 7.1

space. The Laplacian distribution, unlike the radial distribution function,

exhibits shell structure in real spacefor a free atom and also for an atom in a
molecule.In this and other displays of the Laplacian, the function plotted is
-Y'p, such that a maximum in the function plotted correspondsto a
maximum in the concentrationof charge.The first sphericalnode inY2 p for a
one-electronatom in its ground state occurs at a radius r : llZ, and this is
very nearly true for a many-electron atom as well, becoming increasinglyso
as the nuclear charge Z increases(Bader and Essen 1984).Thus, the in-
nermost region of the argon atom, where Y2p is negative,in Fig. 7.2 appears
in the display as a spike-like region of charge concentration. There are two
more regions where Y'p < 0, separated and bounded by regions where
Y'p > 0, corresponding to the presenceof a total of three pairs of spherical
shellsof alternating chargeconcentration and chargedepletion in agreement
with the presenceof three quantum shellsin this atom.
The value of the radius r for which -Ytp attains its maximum value in
each succeedingshell defines the radius of a sphere on which electronic
charge is maximally concentrated in each shell. These values are listed in
Table 7.1 and, like the correspondingmaxima in the radial distribution
function whose most probable r valuesare very similar, they exhibit periodic
variations. The radii decreaseacross a row of the periodic table for a given
shell and the valuesfor succeedingshellsincreasedown a family of elements.
Most importantly, the Laplacian of the chargedensity definesthe chemically
important valence-shellcharge concentration, the shell of charge concentra-
tion which, upon chemical combination, is distorted to yield maxima corres-
ponding in number and relative position to the electron pairs anticipated by
the Lewis and related models, such as the VSEPR model of molecular
geometry (Gillespie 1972).The description of thesechangesrequires a study
of the topology of the Laplacian distribution.

7.1.4 The valenceshell chargeconcentration(VSCC)

Extrema in the Laplacian of p areclassifiedby rank and signaturein the same

way as are critical points in the charge density. The critical points in the
Laplacianoccur whereY (Yt p): 0 and the eigenvalues of the Hessianof Y2p
are the principal curvatures of V2p at the critical point. The topological
discussionswill always refer to the negative of the Laplacian, the quantity
-Ytp. Since charge is concentratedwhere Ytp <0, a local maximum in
- V'p is synonymous with a maximum in the concentration of electronic
charge. Thus a local maximum in -Y'p, a (3, - 3) critical point with
Ytp <0, will denote a local concentrationin electroniccharge and a local
minimum in - Y'p, u (3, + 3) critical point, and, with Y, p > 0, will denotea
local depletion in electronic charge.
 7.1 THE LEWISELECTRO\ PAIR \{ODEL 259

\O C.l
- O C C. ra)
O
n
$ o o
NO _.-,t.r) ..\ r r:- nq
O c i - i c i v - , S - - - O
c. \o \o
t--. O\ ct)
(.) L ti ! L L A I
z Z

F
ta)
- \ O O
(\.l
- - V '
-
r
l+ \O
crl
A
tr) cA
A
a'l

la) -E +
C'.1
C.l
S c.;i ra) oo v tl
V.) _ 0 0 I
V.) \O a.l I

U
H
tr. C fis O
L

\O CO -f, Cj .a .a c , . r * t/-)
\ o N . a - $ o o $ .a a.t a
-l

xx?
a.]
..i ,r; i
F- J
ca -ct.lv
(.) C ) l
a \-/ a A . i a
c{

l
U)
c- c\|
v

co Oc.l Ntr)oo $ ta)

l - * C . ) l a ) + S * $ r f , - t-- y co
i dJ ioc.j J r.-- - cnO,O
a $ O Or-- | an
c\
x.x
a a - l
(d

U) , 7 . \ o
F

r-:
=
z il
o d
\o oo
? L I s ca \o Ct.l l--
ca F- - n tr)
o
$O
r-o
c\
.v ^ g - i
! ) N $
() ts
(.):-
d
U a r') U 6 oi-i
L
-
0)
c) L
- O \
() O\ \O oO ca O\ s .r)Q
. a O - R - i
- 9
. ; ;
-
C) J o o j 9 V l g S O '
oo A A -
l-
d d a s *
J\ r l ,..i co
C) v F A ?
s
o o l
X l-- c\ - f r l

O\ C)tr) *Ot.- N * O
rr) t- Ctl or\ F- \ ^ A ^ €6 c - t
$ 9 V V v

J ic.i OO".; T\ i d i v
b O d ooV) + l
c) 5
, . 1 ! - l
a c a z
Lr
N
I
tr)

a I
I c..l A a
d)
c.l o - fTl

o\ $$ ..)oos A A c { -
$ $S c{O\O\ A -
: N ;
L ^i ;; A-i .j It -i ^'A l-)

a c'.1
lf q) d I
J 7 W J Z v =

L
.
L
l N
: \
6
L
N
L
m
L
{
a.l ca a..l a-, $ *
260 CHEMICAL MODELSAND THE LAPLACIAN 7.1

The outer quantum shell of an atom is divided into an inner region over
which Y'p < 0 and an outer one over which Yt p > 0. The portion of the shell
over whichY'p < 0, is called the ualence-shell chargeconcentrationor VSCC.
Within this shell is the sphere over whose surface the valence electronic
charge is maximally and uniformly concentrated.
Each point on the surfaceof the sphereof maximum concentration in the
valence shell of charge concentration is a (1, - 1) critical point: the two
curvatures tangent to the surface are zero reducing the rank to one and the
curvature perpendicularto the surfaceis negativeSince - Y'p is a maximum
on the surfaceof the sphere.In generalthis surfacepersistswhen the atom is
in chemical combination (the derivative of - Yt p normal to the surface
remains equal to zero and its corresponding curvature is negative),but the
sphereis distorted and the surfaceis no longer one of uniform concentration,
as the two tangential curvatures assumeeither positive or negative values.
The topology of the Laplacian of p on this surface is equivalent to the
description of the hills and valleys of a surfaceterrain. If a local maximum is
formed on the surface,then the two tangential curvatures are negative and
the resultis a (3, - 3) critical point in - Y2p.lf the two tangentialcurvatures
assumepositive values, a local minimum is formed on the surface and the
result is a (3, + 1) critical point in - Y'p. If the two curvatures assume
valuesof opposite sign,there is a saddlepoint in the surfaceand the result is a
(3, - 1) critical point in - Y2p.The local maxima that are createdwithin the
valence shell of charge concentration provide a one-to-one mapping of the
electron pairs of the Lewis model, both bonded and non-bonded.
Just as the energychangesof interestto chemistry are but small fractions of
the total energy of a system,so the changesinduced in the charge density by
the interactionsbetweenatoms are only small'ripples' in the total density.
These small changesare magnified and made evident by the Laplacian of the
charge density. What is remarkable is that the description of the changesso
obtained correspondsto the Lewis model of spatially localizedelectron pairs.

7.1.5 Bonded and non-bonded chargeconcentrations

The structure exhibited by the Laplacian through its collection of valence-

shell critical points is characteristic of a given atom in a given classical
valencestate.In methaneand methyl fluoride there is a local maximum inihe
VSCC of the carbon along the C-F and each of the C-H bond paths (Fig.
7.4a). This figure displays the zero envelope of the Laplacian distribution
which separatesthe VSCC from regions of charge depletion. This surface
encompassesthe four local charge concentrationsor (3, - 3) critical point in
- Y'p in the VSCC of carbon and is contiguouswith the VSCC's on the
protons in methane. The greatly reduced charge concentration on the C-F
7.1 T H E L E W I SE L E C T R O \ P A I R \ I O D E L 261

(a) (b)

The
Ftc. 7.4. Representationsof the Laplacian distributions of methane and methylfluoride.
graphs. The
figures in (a) are displays of the zero invelope oi v2p(r), those in (b) of the atomic
"iuelope "n.o-puriing tt. inner shell charge concentration on carbon appears as a small
protons
sphere.The enveiop"r oi th" bonded maxima in the VSCC of carbon also encompassthe
in CUo and CH.F. There is a transfer of chargefrom C to F in CH3F and the bonded maximum
along ihe C-F axis is reduced to the small region lying betweenthe C nucleusand the envelope
on F. An atomic graph displays the connectivity of the critical points in a VSCC. The carbon
nucleusis denoted by-a solid cross,the positions of the remaining nuclei by open crosses.There is
-Ytp, at each of the four vertices.
a bonded maximum, a (3, - 3) ciitical pornt in

axis in methyl fluoride is separate from the remainder of the VSCC on

carbon. Each of the bonded maxima in the VSCC of carbon is linked to the
-Ytp, which
other three by unique pairs of trajectoriesof the gradient of
originate at intervening saddle points on the surfaceof charge concentration
(fiE. Z.+U).They are (3, - 1) critical points in - Y'p. These lines are the
unulogu.r of the bond paths defined by the correspondingtrajectoriesof the
gradient of the charge density. This network of lines, in the case of the
 262 CHEMICAL MODELS AND THE LAPLACIAN 7.1

Laplacian distribution called the atomic graph, partitions the surface of

charge concentration into four segmentsand the basic structure is that of a
tetrahedron with curved faces. In the centre of each face there is a local
minimum in the surface of charge concentration, a (3, + 1) critical point in
- Yt p. Thus, the sphericalsurfaceof uniform charge concentration of a free
carbon atom is transformed, upon the formation of methane, into one with
four tetrahedrally arranged concentrationsof charge,together with centresof
I
local charge depletion in each of the resulting triangular faces.The surfaceof
I
charge concentration within the VSCC of the carbon atom in the molecule is I
no longer spherical, the different critical points lying at slightly different I
distancesfrom the nucleus, but all within a range of 0.010A about the
I
averagevalue of 0.545A. fnis value representsa slight increaseover the free-
|
atom value of 0.500A. rrrr replacementof a hydrogen in CHo by a fluorine I
leavesthe basic structure of the atomic graph for carbon unchanged,but it
I
does causemajor changesin the extents of charge concentration and charge I
depletion (Fig. 7.4a).Of chemical significanceis the increasein the extent of
I
charge depletion in the triangular face opposite the fluorine atom, a point I
illustrated and discussedin Section7.3.2.
I
The manner in which the uniform sphere of charge concentration for an
I
oxygen atom is distorted upon formation of the water molecule to yield local I
charge concentrations in its valence shell is illustrated in some detail. Relief I
maps of the charge density and the negative of its Laplacian are shown in I
Fig. 7.5 for two planes of the water molecule. In the plane perpendicular to I
the one containing the nuclei, p exhibits a single maximum at the position of
I
the oxygen nucleus as anticipated on the basis of its general topological
I
behaviour. Thus, as previously noted, the topology of p itself offers no direct
I
evidenceto support a model that assumesthe existenceof two localized.non-
I
bonded electron pairs. There are, however, small departures from circular I
symmetry in p giving rise to variations in its local curvatures and the I
corresponding display of the Laplacian of p shows, in a rather dramatic I
fashion, the existenceof two local concentrationsof electronic charge in the
I
valence shell of the oxygen atom in this non-bonded plane. In such a two-
I
dimensional display, the uniform sphere of maximum charge concentration I
appearsas a circular ring in the free atom (seeFig. 7.2), while here it appears I
with two maxima and two minima. The third curvature at each of the I
maxima is negative, showing that the two-dimensional maxima are indeed I
(3, - 3) critical points in -Ytp which indicate the presenceof two local
I
concentrationsof non-bonded electroniccharge. I
In the plane of the nuclei, p again exhibits maxima only at the positions of
t
the nucleiand ridgeslinking the oxygennucleusto the protons,indicatingthe I
presenceof two bond paths. The negative of the Laplacian distribution, on
I
the other hand, in addition to the maxima at the positions of the nuclei. I
exhibits two equivalent bonded maxima in the valence-shellcharge concen-

|
7.1 T H E L E W I SE L E C T R O \ P A I R \ l O D E L 263

P -v"P

Frc. 7.5. Relief maps for p and -Ytp for the two symmetry planes of the water molecule:
lower maps are for the bonded plane (plane containing the nuclei) and upper ones for the non-
bonded plane.There is a spike-likeconcentration of charge at the position of the oxygen nucleus
(which tn this and remaining diagrams is terminated at an arbitrary value), surrounded by the
inner shell of chargedepletion.The valenceshell oi chargeconcentration,the VSCC, of oxygen is
not uniform but exhibits maxima and saddlesin thesetwo-dimensional relief maps. In the non-
bonded plane p exhibits a single local maximum at the position of the oxygen nucleus.In this
-Yzp)
same pline the VSCC of oxygen exhibits two local maxima ((3, - 3) critical points in
and two saddle-points in - Yzp.In the bonded plane,p exhibits a local maximum at each of the
nuclei. as does - Yzp.In addition, the VSCC of oxygen also exhibits two local maxima, one
along each bond path to a proton. What appearsas a third maximum in the VSCC of oxygen is
not a local maximum but another view of the (3, - 1) saddle point betweenthe two non-bonded
maxima appearingin the upper diagram. The value of - Y2p at this saddle point is lessthan its
value at a non-bonded maximum. Thus, the lower diagram makes clear the general observation
that bonded charge concentrations are smaller than non-bonded concentrations. The two
diagrams show that the uniform sphereof chargeconcentrationof an oxygen atom is distorted to
yield four local maxima or local charge concentrationsand four intervening saddle points.

tration of the oxygen atom which prove to be (3, - 3) critical points. The
third, even larger maximum in the two-dimensional display of the VSCC of
oxygen in the bonded plane is actually another view of the (3, - 1) critical
point linking the two non-bonded maxima. This same critical point appears
in the non-bonded plane as a saddle point indicating that it does indeed
264 CHEMICAL MODELSAND THE LAPLACIAN 7.1

possessone positive curvature.One notes that the magnitude of Y2p at this

saddlepoint is lessthan that at the positionsof the non-bondedmaxima but
greater than that at the positions of the bonded maxima. Clearly, the non-
bonded charge concentrations are of considerably greater magnitude than
are the bonded concentrations.The secondsaddlepoint in - V2p,illustrated
in these two diagrams and lying between the two bonded maxima, is a
(3, + 1)criticalpoint. The relativelylargeangleformed betweenthe two non-
bonded chargeconcentrationscausesa changein the atomic graph from that
found for carbon in methane.The saddlepoint betweenthe bonded maxima
and two of the local minima in neighbouring facescoalesceto yield a single
(3, * 1) critical point between the bonded maxima and one edge of the
tetrahedronis missing in the atomic graph for oxygen.
Thus the VSCC of the oxygen atom in the water molecule exhibits four
local maxima in - Yt p, i.e. four local concentrationsof electronic charge-
two bonded concentrations(eachlies on a bond path linking the oxygento a
proton) and two larger and equivalent non-bonded concentrations.Orbital
models of non-bonded electron pairs offer alternative descriptions for an
oxygen atom in water or in a ketone: two distinct pairs, one s-like and the
other p-like when two orbitals of the canonical set are invoked, or two
equivalent pairs when described within some transformed, localized set of
orbitals. The total charge density has a unique set of properties for each
quantum state, properties that are independent of the choice of orbital set
and of the consequences observedon excitation or ionization.
The four local charge concentrations form the corners of a distorted
tetrahedron,with the angle betweenthe bonded maxima of 103' being less
than the angle of 138" found between the non-bonded maxima. Lewis, of
course, assumed a tetrahedral arrangement for four valence electron pairs
and one has here the beginningsof a theory of molecular geometry,a subject
pursued in the following section.
There is no suggestion in this demonstration of the correspondence
betweenthe number and relative positions of local chargeconcentrationsand
the electron pairs assumedin the Lewis model that the former proves the
existenceof the latter. As discussedabove,there is, in general,no evidencefor
the existenceof localized pairs of electronsbeyond the core regions of atoms,
free or bound. Thus the presenceof a (3, - 3) critical point in - V2p does
not imply the existenceof a locabzed pair of electrons.It does show that
electroniccharge is locally concentratedat that point. However, the faithful-
ness of the mapping between the assumedpairs of the model and the local
maxima in - Yt p, as will be further illustrated, leaves no doubt that the
number of local chargeconcentrationspresentin the valenceshell of an atom
is related to an electron count and that the Lewis model has a validity which
extends beyond its statement in terms of the model of localized electron
pairs.
7 .2 A P H Y S I C A L B A S I SF O R T H E \ - S E P RN l O D E L 265

7.2 A physical basis for the VSEPR model of molecular geometry

7 ,2 .1 Res um 6of t h e V SE PR mo d e l

The valence-shell electronpair repulsionor VSEPR model of Gillespie(1972)

is a natural extensionof the localized electron pair model of Lewis. It has
become the most successfuland widely used model for the prediction of
geometriesof closed-shellmolecules.We give here only a brief review of the
basictenetsof this model as regardsthe number,relativeorientation,and size
of the localized electron pairs assumedby the model for differing numbers of
valence electrons. Sufficient examples are given to demonstrate that these
extensionsto the assumedpropertiesof the Lewis pair are recoveredin the
correspondingproperties of the local chargeconcentrationsas definedby the
Laplacian of the electronicchargedensity.More detailsof the comparisonof
the VSEPR model and the propertiesof the VSCC of the central atom are
g i ve n by B ader et a l . (1 9 8 4 ,1 9 8 8 ).
The VSEPR model has two basic assumptions:(1) the valence charge
density is spatially localized into pairs of electrons;and (2) the geometrical
arrangement of the ligands about an atom is that which maximizes the
interpair separation,for both the bonded and non-bondedpairs.Initially, one
does not distinguish between bonded and non-bonded pairs but simply
determinesthe most probable arrangementof all the pairs on the surfaceof a
sphereby maximizing the smallestdistancebetween any two pairs. If all the
pairs are bonded and equivalent,this most probablearrangementdetermines
the geometry.In general,however,not all the pairs in the valenceshell are
equivalentand in terms of the electron-pairspheremodel this meansthat the
electron pair domains are not all the same size.This inequivalenceleads to
the following subsidiary postulates.

1. Non-bonding or lone pairs have larger domains than bonding pairs in the
same valence shell.
2. Bonding pair domains in the valence shell of a central atom decreasein
size with increasing electronegativity of the ligand, and increasein size
with increasingelectronegativityof the central atom'
3. Double and triple bond domains which are composed of two and three
electron pairs, respectively, are larger than single-bond electron pair
domains.

The spherical surface, on which the electron pairs are assumed to be

localized in Gillespie's model, is identified with the sphere of maximum
charge concentration in the VSCC of the central A atom, and the localized
pairs of electrons are identified with the local concentrations of electronic
chargepresenton the sphereof maximum charge concentration.
 CHEMICAL MODELSAND THE LAPLACIAN 7.2
266

7.2.2 Valence-shellchargeconcentrationswith four maxima

The hydrides AHn, A : C, N, and O, are assumedto possessfour pairs of

electronsin the valenceshell of A. The propertiesof the bonded and non-
bonded chargeconcentrationsin the VSCC of the A atom are listedin Table
7.2.lneach case,as illustratedabovefor C and O, the surfaceof the sphereof
- Y'p, but it
maximum chargeconcentrationexhibitsfour local maxima in
is no longer completely spherical, as it was in the free atom' The radial
distanceof a bonded chargeconcentrationis increasedby 0.02to 0.04A ouer
its value in the free atom, while the non-bonded charge concentrationsare
is
drawn slightly closer to the nucleus (Table 7.2). The magnitude of V2p
considerably greater for a non-bonded maximum than it is for a bonded
maximum and electronic charge is concentrated to a greater extent in the
-Ytp, the curvature
former than in the latter. The radial curvature of
perpendicularto the surfaceof the sphereof chargeconcentration listed ?SFr
in Table i.2,isalways greaterfor the non-bondedthan for the bondedcharge
'thinner' in radial extent
concentrations and the former concentrationsare
than are the bonded ones.
Also of importance to the VSEPR model are the relative sizes of the
bonding and non-bondingelectronpairs.The estimateof the sizeof the local
maximum in - Y, p is given in terms of its area on the sphere of charge
-
concentration.This area is given by the expression2nr2n(1 cos al2)where
the angle a is the averageof the anglessubtendedat the nucleusby the lines
linking the nucleus to the edges of the charge concentration in question
(Badei et al. 1984).The boundary of a charge concentration is defined by the
position of the neighbouring saddle points in Y2p, or points where V2p
.hu.rg., sign if Y,p> 0 at the saddlepoint. In agreementwith the VSEPR
model, the area of a non-bonded chargeconcentrationis considerablygreater
than that of a bonded concentration in the same molecule. These observa-
tions regarding the propertiesof the bonded and non-bonded chargeconcen-
trations are found to be of general applicability, regardlessof the number of
electron pairs presenton the A atom. They are understandablein terms of the
VSEpR postulate that a non-bonded charge concentration is attracted by a
single nucleus,while a bonded one is attracted by, and shared between,two
nuclei.
The transformation of a bonded into a non-bonded charge concentration
on passagefrom CHn to NH. results in a decreasein the angle between the
Uonaedchargeconcentrations,the angle bAb, and, hence,to an angle greater
than tetrahedralbetweenthe non-bonded and bonded chargeconcentrations,
the angle nAb. Correspondingly, the non-bonded charge concentration in
NH, occupiesa larger fraction of the surfaceof maximum chargeconcentra-
tion than does a bonded one. Thus a non-bonded chargeconcentration,while
radially more contracted than a bonded one, is laterally more diffuse and
7 .2 A P H Y S I C A LB A S I SF O R T H E \ ' S E P R\ l O D E L 267

o
bo
! H X
z Q 5

. . . 13 r O 3 3 t r , r < t \O
y N
L
-nr: .l r::o9c9 n
(.) i'g - c..] C r-.1 C c..l - -
()
() Eos
O F - \ O c o O \ o o
l( t r { ^ + ..; o. o. O c.i
. i c - r _ _ S S
bo
X L
^
\ () a.l
o\ r.) oo <f, o c-l oo * t.) t--
O\(..l .atart/aO\t$ @ O\
l - a . l c a - V ) $ t r ) - t r ) l - - \ O
C)
c . ;
\ c . i O V i + O J J 6 >

- OO \O
I
OOtf)O.\OoOOt-$
o o l r - . a o n v ^ ' - o l - - $
ca F* .A \O ca C- \O \O \O ca

R
Z *o"3 ,:,i,d,;,-i;A.-i' ;. 4

:\
c t o q n n 9 q q q
A \
oo -o O 6 q ' r , 6 - . \ l &
c < ^ - . r 3 C - C 3
(^t5
iJ
* * - - -
f- i)
c,l tr)v)+$at .7)o\
E i , * = \o o,oj J+i .r; r- +
ta)
v.) C.l C\] Cr.l

c
A
w " ^ c
e ) d
a t- r- bO
O . € - - o O f - - v - ) Q o O r O
l / N
O. >,r-- \O r-r - $ = W .O 4 &

I L
{'g : E : - - ' . - ' E = C
^ -
- : : ;
C) C) " ? € n - Q n q e " ? < ? e q
t
Eo -o O 9 $ r C - r r l F - = $ C , = $
t r { ^ L - A - A A L - A N

R <-o:_ i:- i:- i: **

0.) 9 P
bo d
co P^.r, oo r-.o O 1r"r- - R o
U
:- #= E q€eafiQ
l-
t

- t r .
- = . . = c = o o ;! ' f c\
I

0)
Q u u
a.l O(r)--ca .a€ \O
- <t <t <t f- |f) ta) c..l bo
\O t-O\$OOtf, \O\O t
,*o'S -i
9 g v
-i^iA,iA
v v v
/rA .: a

,:
d
tr)cic\
*"x 3 r v)t--$oo.a( t'- O

e< E ^ o.o.d,ri .'lVl^ri

A A A A A - -
F-.oi i

< x lr:-

R:
Q

F- r-- F- co r-. o F-. co c.] \o ca

4 y rC r-- O O $ .i .a -t O r- c..l O

i g"s q n q n q " ? " 1 v ? q v ?

- O - - # C \ -
q a

a
6)
H
-t '?
- - + - F _ . . _ . _ ' \ J I
X
-=2aca2aoo a 6 c t
268 CHEMICAL MODELSAND THE LAPLACIAN 7.2

possessesa larger surface area. In HrO where there are two non-bonded
charge concentrations,the angle between the bonded concentrations is
further reducedto 103".This angle is to be compared with an angle of 138"
between the non-bonded concentrations. The effects on the local charge
concentrationsin the VSCC of A of replacing A by its neighbouring element
in the samefamily of the periodic table and of increasingthe electronegativity
of the ligands are also illustrated by the data in Table 7.2. Subsiduary
postulate (3) can be illustrated using the sizesof the C-C bonded maxima in
ethane,ethylene,and acetylenewhich, expressedin terms of surfaceareas,are
0.62, 1.24,and 1.97 42, respectively.By including the size of a non-bonded
charge concentration in the VSCC of carbon in singlet methylene,which is
O.g4 L2,one establishesthat, for carbon, a doubly bonded charge concentra-
tion occupiesa larger fraction of the VSCC than does a non-bonded one. The
same is true for sulphur in SO, where the size of the non-bonded charge
concentrationis 1.56 42, while the sizeof the doubly bonded concentrationis
t.82 L2.
All of these observations regarding the properties of the local charge
concentrations in the VSCC of the A atom are in complete accord with the
tenets of the VSEPR model. The primary observation in the lO-electron
hydrides is that of the presenceof four local concentrationsof charge in the
valence shell of A. A non-bonded charge concentration is larger than a
bonded one and occupiesa larger fraction of the surfaceof chargeconcentra-
tion than does a bonded one. The angle subtended by two non-bonded
concentrationsis greater than that subtendedby two bonded concentrations
in an equilibrium geometry.

7.2.3 valence-shellchargeconcentrationswith five and six maxima

The molecule ClF. possesses five pairs of electronsin the valenceshell of the
Cl atom. The most probable arrangementof five electron pairs in the VSEPR
model is a trigonal bipyramid. The five sitesare not equivalentand, hence,the
model predicts that the two non-bonded pairs of electronsin ClF3 should
occupy two of the three spatially less restricted equatorial sites, thereby
predictingthe'T-shaped'geometryobservedfor this molecule.In addition to
studying the properties of the VSCC of the Cl atom in the equilibrium
geometry, two other constrained higher-energygeometries were studied
(Bader et al. 1984).They were obtained by assigningall three fluorine atoms
to equivalent equatorial sites to yield a planar Drn geometry l92kJlmol
higher in energyor to one axial and two equatorialsitesto yield a tripod-like
geometry,263 kJ/mol less stable than the 'T-shaped'geometry.The calcu-
lated equilibrium geometry is in good agreement with experiment. The
calculatedvalues,followed by the experimentalones in parentheses, of the
7.2 APHYSICALBASISFORTHE\'SEPR\IODEL 269

e q u a t or ial and ax i a l b o n d l e n g th s a n d F C F angl e are 1.68(1.70) A,

1 .6 1 (1. 60) A and 85. 4 " (8 7 .5 ' re . e h e fm a ps of the negati veof the
), s p e c ti re l l R
Laplacian of p for two planesof the equilibrium geometryare displayedin
Fig. 7.6.The propertiesof the local chargeconcentrationsin the VSCC of the
A atom are given in Table 7.3.

clFs

Frc. 7.6. Reliefmaps of - Y2p for the equatorial (upper) and axial (lower) planesof ClF.. The
chlorine atom exhibits three shells of charge concentration. The equatorial plane shows the
presenceof two non-bonded and one smaller bonded charge concentration. In the axial plane
there are three bonded chargeconcentrationsand a fourth apparent maximum which is actuall,l"
a n o t h e rv i e w o f t h e ( 3 , - 1 ) c r i t i c a l p o i n t b e t w e e nt h e t w o n o n - b o n d e dm a x i m a .T h u s t h e V S C C
of Cl possesses two non-bonded and three bonded concentrationsof charge,
 C ) a
bo\.,
L
d l -
. c \ J c . t l
o r ' aa ra) R
c c
z 5 t l I

o oo O c.l
Eos c- A J
tr{:-' v oo co

C- ln tr)
.-j o<i o\ c'i
\ o r \ o

T}

U -a
Pot
...l
,r?
-t "\ <-
L
-
N
0)
€
v?
t t -
t\
v) L)
(.)
bo rn
d li

q
c)
.:
c)

(-)
c)
ri c a t / . )
q Eo -o V.)
tr{l. .-i -; O\ \O

o.S z v c-) [--

r'- E
a r h
E :
I ,.t
F S P= ^ c.; ..i
qJ
< -o-'- oo
N
F-

t\
\o - O\ t-
R
^.= ..; + a..: A'
CJ N c..t a'l -

y N \o
\ O $
{"9 A - i
I
a
A
N
r o \ S o o \ o
O \ - F- c.l
\ t r ) . a c a $ ( r )
N
L A A
(.) O r - * c.l O.l
r- oO \o \o f-
\O \O \O
C) A A A A - i
C)
bo
l<
()
d d
O
Li L l -

() v 9
A

9
A

fr
R )

O X 5 x - x
F.l F]

l-

(.) E t-L LL LL I,L

=.:
13n , , ,
7 -
= C
Z . a
 ; A
tl. \-/
- A l

r- F. t-- (\l |'- .O $

- -
r t t t t t l

t-- oo
a t
A .i
aa v c n
-
-
c \ o (a)

J 'Ev
a\.1
n
R
L
I
-( ) r )
O t---
a

8 - : J
(.)
P o "
a

: $
(.) O ^
i

eJ

- o a
c
0 ) 3
r- T A P N
q c t l
d iA : * ' O c-.
c.l - oo o\

t r ) o o o
+ .d c'i
o o o o c o

v ? 9 q : n o q
* t n O \ O \ a a
N - Ctl c.l C.l

$ O \ O \ * c o S
O\ $ 6 t--' f-- \O
. { ^ ^ . ^ . - i - i

c \ l C O - * \ O O \
$ \ O o o c . \ O
aa C.n C-r .a @ \O
-i

c o l r 4

tr--CD-u
I 11 =-O-[
-l
O r!-(/)-t!..
J\
!r.o \ tt" = \
a L L I : g 4 L L U - U L L U -
 CHEMICAL MODELS AND THE LAPLACIAN 7.2

There are five local chargeconcentrationspresentin the valenceshell of the

chlorineatom in all threegeometries.The upper diagram in Fig. 7.6 is for the
equatorialplane of the equilibrium geometryand showstwo relativelylarge
non-bonded concentrationsand one bonded concentration opposite the
single F present in this plane. The second diagram contains all four nuclei,
and three bonded chargeconcentrationsare present.The fourth maximum in
this two-dimensionaldisplay is not a (3, - 3) critical point, i.e.not a charge
concentration,but another view of the (3, - 1) critical point located between
the two non-bonded chargeconcentrations.This and other examplesdemon-
strate that the local charge concentrations defined by the Laplacian of the
charge density are able to recover violations of the octet rule that the Lewis
model must assume to occur for atoms in the third and later rows of the
periodic table.
In the equilibrium geometry of ClF3 the size and magnitude of the local
charge concentrations in the VSCC of Cl decrease in the order: non-
bonded > equatorial bonded > axial bonded (Table 7.3).The angle between
the two equatorialnon-bondedmaxima is relativelylarge and equal to 148",
while the axial bonded-non-bonded angle is opened to 97'. All of these
observationsare in accord with the VSEPR model of molecular geometry.In
the two geometriesof higher energy,the nClb angle is decreasedto 90" in the
D.n geometry while in the least stable tripod-like geometry the correspond-
ing angle is further reduced to 77' . The angle betweenthe axial bonded and
equatorial non-bonded charge concentrationsin this latter geometry is also
less than 90o, and equals 86". Except for planar ClF3, the smallest inter-
maxima angle is a bClb angle.
The moleculeSFn also exhibits an expandedvalencyshell,with five pairs of
valence electrons on sulphur. The observed bisphenoid geometry of this
molecule is correctly predicted by assuming that the non-bonded pair
occupies an equatorial site in the most probable trigonal bipyramidal
geometry.Five local charge concentrationsare indeed found in the VSCC of
the sulphur atom. The non-bonded maximum is the largest of the charge
concentrations with respectto magnitude and size (Table 7.3).The relative
propertiesof the equatorial and axial bonded maxima in SFo are the sameas
those discussedabove for the equilibrium geometry of ClF3, including the
observation that the equatorial maxima are of greater magnitude and more
tightly bound than are the axial maxima.
The square-basedpyramid geometry observedfor ClFr is rationalized on
the basis of the most probable octahedral arrangement of six electron pairs.
The unique CIF bond lying along the fourfold axis and opposed to the non-
bonded pair has a length of 1.62A compared to the remaining CIF bonds
which have a length of 1.72A. the FCIF angle formed with the unique F is
87.5'. Figure 7.7 shows the unique bonded and the opposed non-bonded
charge concentrationsand two of the four equivalent bonded concentrations
1 .2 A P H Y S I C A LB A S I SF O R T H E V S E P RM O D E L 273

clF5
Ftc. 1.7. A relief map of -Y'p in a symmetry plane of ClF, containing the chlorine nucleus,
the axial fluorine nucleus, and two of the four fluorine nuclei in the base of the square-base
pyramid. The figure clearly illustrates the relative sizes of the charge concentrations, non-
bonded > axial bonded > basal bonded.

in this molecule.The diagram illustrates that, in accordancewith the model,

there are a total of six local concentrationsof charge presentin the VSCC of
Cl in this molecule, five bonded and one non-bonded.
The single non-bonded maximum in ClF, is of greater magnitude than
those of the two such maxima in ClF. (Table 7.3).In ClF3 and SFn the axial
fluorines are nearestto the non-bonded maxima and they possessthe longest
bond lengths. Correspondingly, their bonded maxima are of smaller magni-
tude and are lesstightly bound than are the equatorial maxima. In ClFr the
four equivalent fluorines are nearest to the non-bonded maximum. They
possessthe longest bond lengths and their associatedbonded maxima on
chlorine are of smaller magnitude and are lesstightly bound than that for the
single fluorine with the shortest bond length. The bond angle formed by a
long and a short bond is calculatedto be very similar for all three molecules,
85.5 + 1.1o,as are the corresponding angles between their bonded maxima,
83.6 + 0.9'. In all three molecules the angle between the non-bonded max-
imum on chlorine or sulphur and the bonded maximum of the longest ligand
bond is greaterthan 90", ranging from 96.5' in ClF, to 98.6'in SFn.It is dlso
found that the fluorine iigands with the longest bond lengths possessthe
largest net negative charges (Table 7.3). Thus the behaviour of the local
charge concentrations, as defined by the Laplacian of the charge density,
mimic all of the consequencesof the repulsionsassumedto exist betweenthe
localized pairs of electrons in the VSEPR model.
1 1 I
CHEMICAL MODELSAND THE LAPLACIAN 7.2

7.2.4 Effectsof charge transfer on the properties of the VSCC

The degree of charge transfer accompanying bond formation affects the

topology of an atom's VSCC. The interatomic surfaceseparatingthe basins
of the two atoms linked by a bond path is defined by the topology of the
chargedensity as made evident in its gradient vector field. While the zero-flux
property of such surfacesensuresthat the integral of the Laplacian over the
basin of an atom equalszero for either a free or a bound atom (eqn (6.88))'
there is no direct constraint on the topology of the Laplacian within the
atomic basin of a bound atom or even a requirementthat the VSCC must
persist on bonding, or remain within the atomic surface of the given atom.
Thus, an atom which losesmost of its valencecharge density on bonding, as
Li does in many of its compounds,does not possessa VSCC in its bonded
form. In the other examplesthe VSCC initially associatedwith the free atom
is sharedwith more electronegativebonded neighbours.For a bond between
identical atoms and for polar bonds between dissimilar atoms, there are, in
general,two bonded maxima, one associatedwith each atom. Thus, a single
bonded pair of the Lewis model is representedby two bonded maxima-one
in each VSCC. The sameis true for multiple bonds. The bonded pair or pairs
in the Lewis model of the C-C bonds in ethane,ethylene,and acetylene,for
example, are all respresentedby a bonded maximum of increasing size on
each carbon atom through the series.
From the many examples of Laplacian distributions that have now been
studied it appearsthat fluorine is unique in its behaviour. The charge density
of a fluorine atom is tightly bound and very localizedin all of its compounds.
In other than ionic compounds,the VSCC of fluorine exhibitsonly a torus of
non-bonded charge encircling the bond axis and no bonded charge concen-
tration, not even in F, (Bader and Essen 1984).Rather than three negative
curvatures in - Y'p (which denote a local maximum in - Ytp), the two
curvatures perpendicular to the bond axis are positive for the bonded and
non-bonded axial critical points in the VSCC of F. Such a (3, + 1) critical
point denotesthe presenceof a local minimum in the surfaceof the sphereof
maximum charge concentration. The radial curvature at these points is still
negative-the sphere of charge concentration persists-and, becauseof the
magnitude of this curvature, its sign dominates the value of the Laplacian.
Only in HF has the VSCC of fluorine been found to exhibit a bonded
maximum.
The uniform torus of charge concentration of F can be perturbed by the
presenceof low-lying unoccupied orbitals centred on a neighbouring atom
(MacDougall and Bader 1986).In thesecasesthere is a polarization of the n
density on F towards its bonded neighbour, counter to the direction of the
much greater transfer of o density.
Finally, in the caseof a Lewis acid with an incomplete valenceshell,such as
BHr, the surface of the sphere of maximum charge concentration is pun-
n a
t.) THE LAPLACIANOF P AND CHE\1IC{L REACTIVITY 275

ctured.Thus, the VSCC of B in BH, or of C in CH i is reducedto a belt of

chargeconcentrationin the plane containing the protons. There is, in these
molecules,direct accessto the region of chargedepletionassociatedwith the
core from either side of this plane, and they are very electrophilicin nature.

7.2.5 Summary

The examples and accompanying discussion that have been given above
demonstrate a remarkable mapping of the number and properties of the
localizedelectron pairs assumedin the VSEPR model of molecular geometry
on to the number and properties of the local maxima found in the VSCC of
an atom, as determined by the Laplacian of the electronic charge density. So
faithful is the mapping ih recovering all of the nuancesof the VSEPR model
that one can conclude that the properties of the Laplacian of the charge
density as displayed in the VSCC of the central atom provide the physical
basis for this model and for the Lewis model of electron pairs. It has also
been demonstratedhere and in other examples(Bader et al. 1988)that the
total energy of a molecule is minimized for the geometry which maximizes
the separations between the charge concentrations, but at this point no
theoreticalreason has been advancedas to why this should be so.
A discussion of why the Laplacian exhibits local concentrations of elec-
tronic chargewhere the VSEPR model assumeslocalizedelectronpairs to be
is given in Section87.2,which dealswith the localizationof the Fermi hole. It
is shown there that the number and relativepositionsof the local maxima in
the VSCC of the A atom in a moleculeAXn area consequenceof the partial
localization of electron pairs in its valence region which results from the
ligand field operating in concert with the Pauli principle. With this final
demonstration, all of the postulatesof the VSEPR model, including the effect
of the Pauli principle on the most likely arrangement of electron pairs, are
recoveredin the properties of the VSCC of the central atom. A knowledge of
these properties may be used to predict molecular geometriesin a modified
VSEPR model that requires only a single postulate, namely, that the most
stable geometry of a molecule is the one which maximizes the separations
between local maxima in the VSCC of the central atom.

7.3 The Laplacian of p and chemical reactivity

7.3.1 The Laplacian and the local energy of p

The Laplacian of the charge density plays a central role in the theory of
atoms in molecules where it appears as an energy density, that is,
L(r) : - (h2l4m)Y'p(r). (7.r)
 n 1
276 CHEMICAL MODELSAND THE LAPLACIAN t.J

The integral of l,(r) over an atom Q to yield I(O),

L(Q) : n ' t + mY)2p(r)dr

[*
Jrtrlor:(-
( - h2t4mr asto,r)Vp(r)'n(r): g, (7.2\
f
vanishesbecauseof the zero-flux boundary condition, egtr (6.7),which defines
an atom in a molecule.The demonstrationthat an atom is an open quantum
subsystem is obtained by a variation of Schrodinger's energy functional
Ell/,O], eqn (5.58),for a stationary state and, as shown in Chapter 8, by a
variation of the action integral,the time integralof the LagrangianI lY,Ql,
for a time-dependentsystem.In each casethe zero-flux boundary condition is
introduced by imposing the variational constraint that

, : o
Y2p(r)dr}
{J:
at every stageof the variation. The possibility of introducing the constraint in
this manner is a consequenceof the property of the functionals E l(,, Q] and
glY,Ol that, at the point of variation where the appropriate Schrodinger
equation is satisfied,they both reduce to an integral of the density Z(r). The
property given in eqn (7.2\ is common for an atom and for the total system
and it is this property which endowsthem with similar variationalproperties,
thereby making possible the generalization of the principle of stationary
action to an atom in a molecule.
The two kinetic energydensitiesK(r)and G(r)differ by Z(r)(eqn (5.48))and
it is becauset(O)vanishes for an atom that f(O), the electronickinetic energy
of an atom, enjoys the satisfactionof the sameidentity as does f for the total
system,that is,
r(Q) :,K(O) :G(O). (7.3)

It was demonstratedin Chapter 6 that the densityL(r) appearsin the local

expressionfor the virial theorem,eqn (6.31).This is an important result,since
it relatesa property of the charge density to the local contributions to the
energy; it is repeated here
(h2l4m)Vtp (r) : 2G(r) + /" (r). (7.4)
The electronic potential energy density t^ (:r),the virial of the forces exerted
on the electrons(eqn (6.30)),and the electronickinetic energy density, G(r)
(eqn (5.49)),define the electronicenergydensity,
: G(r)+ t'' (r).
8.111 (7.5)
 LEACTI\.lT\' 277
7 . 3 T H E L A P L A C I A N O F p A N D C H E I \ 1 1 C . \R

Because7,(O)vanishesfor an atom, integration of eqn (7.'ll over the basin of

an atom yields the atomic virial theorem
2T(Q): _ y'(o). (7.6)

the electronicenergyof an atom in a moleculesatisfies

and, as a consequence,
the identities,

E"(O):
I"E"(r)dr: -f(O) :*f(A). (7.7)

The potential energydensity f (r) is everywherenegative,while the kinetic

energy density is everywherepositive. Thus, the sign of the Laplacian of the
charge density determines,via eqn (7.4),which of thesetwo contributions to
the total energyis in excessover their averagevirial ratio of 2:1. In regionsof
spacewherethe Laplacian is negatiueand electronicchargeis concentrated,the
potential energydominatesboth the local total electronicenergy E"(r) and the
local uirial relationship.Wherethe Laplacian is positiueand electronicchargeis
Iocally depleted,the kinetic energy is in local excess.
An energy density is dimensionally equivalent to a force per unit area or a
pressure.Thus, the Laplacian may alternativelybe viewed as a measureof the
pressureexertedon the electroniccharge density relative to the value of zero
required to satisfy a local statement of the virial theorem, i.e. !'(r) + 2G(r)
: 0. In regions where the Laplacian is negative,the charge density is tightly
bound and compressedabove its averagedistribution. In regionswhere the
Laplacian is positive, the charge density is expanded relative to its average
distribution,the pressureis positive,and the kinetic energyof the electronsis
dominant.

7.3.2 Lewis acid-base reactions

The discussion so far has focused on the properties of the local charge
concentrations of the Laplacian distribution and on how they recover the
Lewis and VSEPR models of electron pairs. The Lewis model, however,
encompasseschemical reactivity as well through the concept of a generalized
acid-base reaction. Complementary to the local maxima in the VSCC of an
atom for the discussion of reactivity are its local minima. A local charge
concentration is a Lewis baseor a nucleophile,while a local chargedepletion
is a Lewis acid or an electrophile.A chemical reaction correspondsto the
'lump' in the VSCC of the basecombining with the 'hole' in
combination of a
the VSCC of the acid. The recovery of this fundamental chemical property is
graphically illustrated in Fig. 7.8 which illustrates the facility for the non-
bonded charge concentration on the carbon of carbon monoxide to react
with the hole present on the boron atom in BH.. The zero envelopeof Y2p
278 CHEMICAL MODELS AND THE LAPLACIAN 7.3

shown in this figure, since it displays the sites of the local charge concentra-
tions and chargedepletions,definesa system'sreactivesurface.In terms of the
local virial theorem, eqn (7.4),the reaction of a nucleophile with an electro-
phile is a reaction of a region with excesspotential energy on the base atom
with a region of excesskinetic energy on the acid atom. The accompanying
rearrangementof the charge is such that, at every stageof the reaction, Z(O)
remains equal to zero for each atom. Thus, reductions in the magnitudes of
the local concentrationsor depletionsof charge require opposing changesin
other parts of the atom to satisfy the constraint on its charge distribution
given in eqn (7.2).
The effectof the replacementof H by F in methane on the properties of the
VSCC of the carbon atom (seeFig. 7.9)illustratesthe internal compensations
which maintain the satisfactionof eqn (7.2).While, as pointed out previously
(Fig. 7.4),the structure of the atomic graph for carbon is left unchangedby the
substantial charge transfer from carbon to fluorine which accompaniesthe
substitutionof H by F (q(C) increasesfrom + 0.24 in methaneto * 0.87 in
fluoromethane,q(H) changesfrom - 0.06 to - 0.04,and q(F) : - 0.74),it
does causemajor changesin the extent of charge concentration and deple-
tion. As a result of the charge tlansfer, the bond critical point for the C*F
bond path is found only 0.43 A from the C nucleus and, as discussedin
Section7.2.4,the associatedbonded chargeconcentration now lies within the
boundary of the F atom. If the loss of this charge concentration and its
associatedexcessof potential energy were not compensatedfor, the energy of
the carbon atom would increasesignificantly and, in addition, the zero-flux
surface condition would no longer be satisfied. Consequently, the three
remaining bonded maxima, those linking the protons, increasein magnitude,
the values of - Yzp increasingfrom 0.72 au in methane to I .42 au in methyl
fluoride, as do its valuesat the intervening saddlepoints (Fig. 7.9).As a result
of this increasein the concentration of charge about the perimeter of the face
of the VSCC opposite the F, the extent of charge depletion in this face is
increased.Thus the value of Y2p at the (3, * 1) critical point in the centre of
each face increasesfrom + 0.12 au in methane to + 0.30 au in the face
opposite F and to + 0.20 au in the faces opposed to the H atoms. These
(3, + 1) critical points define the minima in the sphereof charge concentra-
tion in the VSCC of carbon, the sitesof maximum chargedepletion.They are
also the sitesof nucleophilic attack and the carbon of CH.F is clearly more
susceptibleto such attack than is the carbon of methane. The attack is
predicted to occur at the ring critical point opposite the F atom, as indicated
by the arrow in Fig. 7.9. The carbon of methanol exhibits similar behaviour,
the value of Y2p at the face critical point opposite the oxygen being only
slightly less positive than the corresponding value in methyl fluoride.
The positions of the local charge concentration and depletion, together
with their magnitudes,are determined by the positions of the corresponding
Frc.7.8. Isovalue surfacesol -Y2p for carbon monoxide (blue) and borane (red).The surface
'lump'
drawn is the V2p : 0 surface,the reactive surface.The moleculesare oriented so that the in
the VSCC of carbon is alignedwith the hole in the VSCC of boron. The VSCC of boron is reduced
to a belt-likedistribution lying in the plane of the protons,giving the baseatom direct accessto the
core of boron. The zero-valuedsurfacesencompassingthe inner shell charge concentrationon
carbon and boron are also evident. Note the torus of charge depletion encircling the carbon
nucleus.This featurecorrespondsto the localization of the z* orbitals on carbon.
7.3 THE LAPLACIAN OF P AND CHEMICAL REACTIVITY 279

Ftc. 7.9. Contour and relief maps of the Laplacian distribution for CHo and CH.F, in a plane
containing HCH nuclei in the former and HCF nuclei in the latter. Solid contours denote
positive values, dashed contours negative values of Y2p. The contour values in these and
succeedingdiagrams of Yzp are given in the Appendix (Table A2); the outermost contour in each
plot is + 0.002au. The inner contour encircling each nucleus in the contour plots of V2p
encompassesthe negative spike-like charge concentration present on each nucleus. The relief
maps display the negative of the Laplacian distributions. The relief displays are mirror images of
the nuclear arrangements shown in the contour plots. There is a spike-like charge concentration
at each nucleuswhich is somewhat broader for a proton. The arrow points to the (3, + 1) critical
point which is most susceptible to nucleophilic attack.

critical points in the VSCCs of the respectivebase and acid atoms. This
information enablesone to predict positions of attack within a moleculeand,
hence, the geometries of approach of the reactants. For example, a keto
oxygen in the formamide moleculehas two large non-bonded chargeconcen-
trations in the plane of the nuclei (Y'p : - 6.25 and - 6.30 au), while the
280 CHEMICAL MODELS AND THE LAPLACIAN 7.3

nitrogen atom exhibits two such maxima of lesser magnitude (Y'p :

-2.l4au) above and below this plane (Fig. 7.10).On the basis of this
information one correctly predicts that the formamide moleculewill preferen-
tially protonate at the keto oxygen (Birchall and Gillespie 1963),specifically
at the position of the largest of the two charge concentrations and in the

-6.25

F*-"{

Frc. 7.10. Displays of -Y2p for the formamide molecule in the plane containing the nuclei
and in a perpendicular plane containing the C-N axis. The values refer to the non-bonded
maxima in the VSCCs of the oxygen and nitrogen atoms. The VSCC of carbon exhibits three
bonded maxima in the upper figure. The Laplacian is positive over much of the VSCC of carbon
in the plane illustrated in the lower diagram. The two points in this plane where V2p is most
positive are the 'holes' in the VSCC of carbon, the points of nucleophilic attack (as indicated by
an arrow,.
t.5 THE LAPLACIANOF P AND CHE\lICAL RE.{CTIVITY 281

plane of the nuclei.There are holes in the VSCC of a carbonyl carbon and
ihey determinethe position of nucleophilicattack at this atom. Theseholes
are above and below the plane of the nuclei of the keto grouping (seeFig.
7.10) and the correspondingcritical points for a number of ketones are
positionedto form anglesof 110 * 1' with respectto the C:O bond axis (see
Fig.7.20,as well).This is the angle of attack predictedfor the approach of a
nucleophileto a carbonyl carbon (Burgi and Dunitz 1983).
Simiiar predictions have been made for the Michael addition reaction,
specificallyfor the nucleophilic attack of an unsaturated carbon in acrylic
u.id, cHz:cH-corH, and methyl acrylic acid (carroll et a/. 1989).The
properties of the Laplacian distribution correctly predict that the attack
o..u6 at the terminal carbon of the methylene group, the carbon of the
unsubstituted acid being most reactive, and that the approach of the nu-
cleophile will be from above or below the plane of the nuclei along a line
forming an angle of 115' with the C:C bond axis,the latter prediction being
in agreement with calculations of the potential energy surface for this
reaction.The reader is referredto Bader and Chang (1989)for a discussionof
the use of the Laplacian distribution in the prediction of the sites of
electrophilic attack in a seriesof substituted benzenes.
Elecirostaticpotential maps have been used to make predictions similar to
these (Scrocco and Tomasi 1978).Such maps, however, do not in general
reveal the location of the sitesof nucleophilic attack (Politzer et al. 1982),as
the maps are determined by only the classicalpart of the potential. The local
virial theorem, eqn (7.4),determinesthe sign of the Laplacian of the charge
density. The potential energy density f F) (eqn (6.30))appearing in eqn (7'4)
involves the full quantum potential. It contains the virial of the Ehrenfest
force (eqn (6.29)),the force exertedon the electronicchargeat a point in space
(eqns (O.fO)and (6.17)).The classicalelectrostaticforce is one component of
this total force.
The Laplacian distribution has been used to predict the structures and
geometriei of a large number of hydrogen-bondedcomplexesby aligning the
side of the
if, * 3) critical point, a local chargedepletion on the non-bonded
proton in the acid HF, with the (3, - 3) critical point of the base, a local
ioncentration of charge, for which - V'p attains its largest value (Carroll
et a1.1988).The moleculargraphsfor the non-linearcomplexesin this study
are shown in Fig. 7.11.With only a few exceptions,the geometriesof the
complexespredicted in the SCF calculations (which agree with experiment
wheie comparison is possible),are those predicted by the properties of the
Laplacian as outlinedabove. Figure 7.12lilustrates the Laplacian distribu-
tion for three of the basesinvolved in this study. They form an interestingset,
as the Lewis model of localized pairs fails for two of these molecules.The
Laplacian, sinceit is model-independentand instead reflectsthe properties of
the charge distribution, correctly predicts the observed structures of the
 ) **--f
**J --\
l 12
1
l
I
I

{
I T-r
l \
F,
7
\ , \
\
"-"u
13
2 t\
\

8
\ F
A
,r'ffi* }/ { \ \

H
t
H
3 II
14
----/
A-<_J+_f
t a 9
t1 Ft
"---"-----*'
4
\ 15
I
I
\ l
t w
f I
I
*--t 10
*-#-f
5 16

,t
t /
I
. . J
ffi

11
'\
6 17
FIc. 7.11. Molecular graphs of non-linear hydrogen-bondedcomplexes.The dots denote the
positions of the bond critical points. This point in HF is close to the proton. The basesin the
c o m p l e x e si n n u m e r i c a lo r d e r a r e : ( l ) N r O ; ( 2 ) H C P ; ( 3 ) H r O ; ( 4 ) H r S ; ( 5 ) N H . ; ( 6 ) P H r ; ( 7 ) O . :
( 8 ) S O r ( c i s ) ' ;( 9 ) S O , ( t r a n s ) ;( 1 0 ) H r C O ; ( l l ) H F ; ( 1 2 ) C I H ; ( 1 3 ) N r S ; ( 1 4 ) H ' C S ; ( 1 5 ) C O r :
(16) SCO;(17) HCl. The latter three are secondarystructures,the primary structuresfor (15) and
(16) being linear; for the complex between HCI and HF, the primary structure is (12).Note the
cyclic structures obtained for (13) and (14).
I

7.3 THE LAPLACIAN OF p AND CHEtl1lC{L REACTIVITY 283

Ftc.7.l2. Relief plots of the negativeof the Laplacian distributions for NtO (top), SCO, and
OCO (bottom). An oxygen atom is on the right-hand side in each case.
 284 CHEMICALMODELSANDTHELAPLACIAN 7.3

complexes.The oxygen in NNO possesses a torus of charge concentration

(which appearsas two maxima in the plane shown in Fig. 7.12) for which
- Y'p: 5'63 au, while the singlenon-bonded charge
concentrationon the
terminal N has a magnitude of 2.90 au. (This arrangement of non-bonded
charges,as does that for the bonded charge concentrations,agreeswith the
Lewis model, :N: : :x: :Q.) The Laplacian predictsthe proton to bond with
the oxygen and to form a bent structure with a bond angle of 103.1.
compared to the calculatedvalue of 104.4'.Contrary to the Lewis model, the
oxygen in SCO exhibits only a single non-bonded maximum which is co-
linear with the internuclear axis. This maximum is also the largest of the
chargeconcentrationsin this moleculeand one thus predicts a linear complex
SCO-HF, as is both observed and calculated. In oCo, the non-bonded
charge concentration on each O forms a torus about the molecular axis. in
correspondencewith the Lewis model. The value of - yrp for the torus,
which forms an angle with the internuclear axis of 149",is 4.74 au. As is clear
from the display of the Laplacian for oCo, however, each oxygen atom is
capped by a non-bonded charge concentration which is of almost constant
value and the value of - Y2p in the VSCC of oxygen on the internuclear axis
is 4.72 au. The properties of the Laplacian are thus consistent with the
experimental result of a floppy complex with a bond angle within the range
180 + 30". The calculatedequilibrium structure is found to be linear, but its
energy differs from that of a secondarybent structure with a bond angle of
165' by only 0.4 kJlmol.
Further examples of using the Laplacian distribution to predict the
geometriesof acid-base reactions are provided by the crystal structures of
(Il + )(AsFi )z and (Sl *XAs Fi)z . 0.6so2 (Gillespie-et
al.19g3;passmoreet al.
1980).The If+ cation is modelled by the isoelectronicCl?* cation. The
directions of approach to the holes in the VSCC of the chlorine atoms are
indicatedby arrows in Fig.7.13. Thesepredictedorientationsof attack by a
nucleophile are precisely the positions of the nearest neighbour fluorine
atoms of the counter-ions in the Il+ crystal. The holes in the VSCC of the
sulphur atoms in Sl* are also indicated in Fig. 7.13.They form an acute angle
with the S-S bonds and are in the plane of the nuclei. In this casea single
nearest-neighbourfluorine atom of a counter-ion lying on the bisector of the
S-S bond binds to each such pair of holes. This prediction for Sl + is in
contrast to that of frontier orbital theory. According to this theory a
nucleophileshould approach the cation from above or below the plane of the
nuclei, to maximize overlap with the z* LUMO (lowest unoccupied molecu-
lar orbital) of the cation.
The topology of the VSCC of the carbon atom in the triplet state of
difluoromethyleneis homeomorphic to that for the oxygen atom in the water
molecule.This is illustratedin Fig. 7.14which displays - yrp for the triplet
moleculein planesanalogousto thoseshown for water in Fig. 7.5.In this case
7.3 THE LAPLACIAN OF P AND CHEMICAL REACTIVITY 285

F r c . 7 . 1 3 . R e l i e fm a p s o f - Y ' p i o r S l * , S r N r , a n d C l ? * . T h e a r r o w s i n d i c a t et h e d i r e c t i o no f
nucleophilic attichment to the cations by the fluorides of the anions in the crystal.
286 CHEMICAL MODELSAND THE LAPLACIAN a
t.3
a

tct, tcF2

Ftc. 7.14. Reliefplots of the negativeof the Laplacian distributions for triplet and singletstates
oi CFr. The lower diagrams are for the plane containing the nuclei, the upper ones for the
perpendicular symmetry plane containing the C nucleus,the plane containing the non-bonded
charge maxima. There are two non-bonded maxima in the triplet, one in the singlet.The point
'a'
labelled is a (3, - l) critical point in the VSCC of triplet carbon. There is no radial maximum
or lip at the point labelled 'h' and its mirror image and the VSCC of singlet carbon exhibits
'holes'
at thesetwo points. The maxima presentin the VSCCs of the F atoms are not shown as
they are larger by a factor of l0 than those on the carbons.

the two local maxima in the non-bonded plane correspond to separate

concentrationsof charge arising from the two unpaired electrons.The plane
of the nuclei exhibits two bonded maxima linking the carbon with the
fluorines and a second view of the (3, - 1) critical point between the non-
bonded maxima, a point labelled'a'in the diagram. In contrast to this, the
two corresponding planes for the singlet speciesof this molecule, while
exhibiting two bonded maxima, exhibit a single,symmetrically placed, non-
bonded charge concentration. The non-bonded charge concentration in the
singlet state corresponds to a single electron pair and, since it is a local
maximum in - Y'p, it appearsas a maximum in both planes.This example
servesto illustrate that the topology of the Laplacian extends beyond the
Lewis model of electron pairs and enablesone to determine the location and
magnitudes of local concentrations of electronic charge even when the
t.) THE LAPLACIAN OF P AND CHEl\lICALREACTIVITY 287

maxima correspond to the presence of a single unpaired electron

'h'
(MacDougall and Bader 1986).The point in the VSCC of carbon labelled
in Fig. 7.14 and its mirror point in the singlet state denote the presenceof
holes in the sphere of maximum charge concentration,a point where a
nucleophile has direct accessto the core of the carbon atom.
The above exampleshave demonstrated that the relative approach of the
reactantsin a Lewis acid-base reaction can be predicted by aligning a region
of charge concentration with a region of charge depletion, the regions being
determined by their Laplacian distributicns. In many cases,but not all, the
local concentration of charge predicted by the Laplacian to be the site of
nucleophilic activity coincides with corresponding properties of HOMO
(highestoccupied molecular orbital). Similarly, the local depletion of charge
predicted by the Laplacian to be the site of the electrophilic activity coincides
with the propertiesof LUMO. In this way the Laplacian of the chargedensity
recoversthe predictions of the frontier orbital model of chemical reactivity
(Fukui lg7l).In this model the relative orientation of reactantsis determined
by requiring that the overlap of HOMO and LUMO be a maximum. In
larger systems where the manifold of one-electron states in the HOMO-
LUMO energy region is dense,as in Sl* for example, the ordering of the
orbitals is not well-determined and this model is difficult to apply. The
Laplacian of the charge density appears to be more stable in its ability to
predict the sites of charge depletion or concentration. This function is
determinedby a property of the total chargedensity of a systemand is model-
independent,assuming a proper representationof the charge density.
Perhaps the most important information contained in the state function
that must be sacrificedin its reduction to the charge density is the informa-
tion regarding its phase. In the orbital model, the relative phases of the
orbitals determine their symmetry properties and their effectiveoverlap. The
Laplacian of the charge distribution, by recovering predictions based upon
orbital phase information, provides a bridge betweenorbital models and the
understandingof reactivity basedupon the chargedensity.This bridge can be
further demonstrated and better understood through the use of a model
based upon second-order perturbation theory which predicts the essential
featuresof the relaxation in the chargedensity for a given displacementof the
nuclei, the latter being the source of the perturbation. This is done by
approximating the vibrationally induced relaxation in p by the transition
density obtained from a mixing-in of the lowest-energy excited state of
appropriate symmetry, the symmetry being determined by the perturbing
nuclear displacement.If one further assumesthe largest relaxation in p to be
causedby mixing in the excited state of lowest energy,one is able to predict
the nuclear displacementthat results in the smallest energy increase(Bader
1960, 1962).This method has been used to predict the signs of interaction
constants in vibrational potential functions (a direct test of the assumption
288 CHEMICAL MODELSAND THE LAPLACIAN 7.4

that the vibrational mode that causesa mixing-in of the lowest-energyexcited

state has the smallest force constant) and the dissociative pathways of
unimolecular reactions and transition-state complexes. It has also been
shown that the relative approach of the reactants, as determined by the
properties of their Laplacian distributions, leads to the formation of a
complex in which the most facile relaxation of the charge density is the one
which leads to further motion along the reaction coordinate defined by this
approach of the reactants(Baderand MacDougall 1985).The model demon-
stratesthe manner in which the nature and, in particular, the symmetry of the
lowest excited state determines the course of a chemical reaction (Bader
1962). Within the orbital model, the relevant transition density can be
approximated by the product of an orbital from the ground state with one
from the excited state. In many casesthis corresponds to the product of
HOMO with LUMO and, in this w&y, the second-order perturbation
approach to reactivity justifies the assumption of the frontier molecular
orbital model. In some cases,the symmetry of LUMO is such as to yield a
transition density of irrelevant symmetry, that is, one which does not
correspond to the symmetry of any nuclear motion of the complex, and it is
for this reason that the frontier orbital model sometimesfails.

7.4 The characterizationof atomic interactions

7.4.1 Definition of atomic interactions

The gradient vector field of the charge density identifies the set of atomic
interaction lines within a molecule. These interactions, which define the
molecular structure, can be characterizedin terms of the properties of the
Laplacian of the charge density. The local expressionof the virial theorem.
eqn (7.4),relates the sign of the Laplacian of p to the relative magnitudes of
the local contributions of the potential and kinetic energy densitiesto their
virial theorem averages.By mapping those regions where Y'p < 0, the
regionswhere electronicchargeis concentrated,one is mapping those regions
where the potential energy density makes its dominant contributions to the
lowering of the total energy of the system,eqn (7.5).It is found that such a
mapping of the potential energydensity definesa spectrum of possibleatomic
interactions, with those characterizedas shared at one extreme and those
characterizedas closed-shellat the other (Bader and Ess6n 1984).
As discussedin Chapter 2, the interaction of two atoms leads to the
formation of a critical point in the chargedensity at which p has one positir e
curvature, labelled A, and two negative curvatures labelled )t and ir. The
phase portrait of such a (3, - 1) critical point is displayed in Fig. 2.6. The
position coordinate of the critical point is denoted by r". The pair of
7.4 C H A R A C T E R I Z A T I O NO F A T O \ 1 I C I \ T E R A C T I O N S 289

eigenvectors associatedwith the negativeeigenraluesof the Hessianof p at r.,

the two negativecurvaturesof p (r.), generatea setof gradientpaths of p, all of
which terminateat the critical point and definethe interatomic surface.The
unique eigenvectorassociatedwith the positiveeigenvaluegeneratesa pair of
gradient paths, each one of which originates at the critical point and
terminatesat one of the nuclei,therebydefininga line along which the charge
density is a maximum with respectto any neighbouring line, the atomic
interactionline.In a bound system,such a line is called a bond path.
Embodied in the classicalnotion of structure is the idea of relating the
properties of a molecule to a set of pairwise atomic interactions which
dominate and char acterrzethe mechanics underlying the structure. The
interactions may be attractive or repulsive-the definition of structure in
terms of the dominant interatomic interactionsis essentialto the understand-
ing of either situation. It was shown in Chapter 6 that the quantum force
acting on the electronic charge in the basin of an atom O, the quantity F(O)
defined in eqns (6.14)and (6.15),is given by the integral of the force AdS
exerted on each element dS of its atomic surfaceby its neighbouring atoms.
(The quantity F'is the quantum stresstensordefinedin eqn (6.12).)An atomic
surface S(O) is, in general, composed of a number of interatomic surfaces
S(OlO') and the force F(C))can be expressedas
F ( Q :) - $ d S ( Q ) F ' ' n-: I r , , * n $ d s ( o l c ) ' ) F ' n . (7.8)

The sum in eqn (7.8)runs over the atoms, O', neighbouringatom O-over
eachatom linked to Q by an atomic interactionline. Thus eqn (7.8)provides
the physical basis for the model in which a molecule is viewed as a set of
interacting atoms. It isolates,through the definition of structure, the set of
pairwise atomic interactions which determine the force acting on each atom
in a molecule for any configuration of the nuclei.
Since the two perpendicular curvatures of p whose eigenvectorsdefine a
surface S(O|O') are negative,the charge density is a maximum at r" in the
interatomic surfaceand charge is locally concentratedthere with respectto
points in the surface.The curvature of p along the interaction line is positive,
charge density is locally depletedat r, relative to neighbouring points along
the line, and p is a minimum at r" along this line. Thus, the formation of a
chemicalbond and its associatedinteratomicsurfaceis the result of a competi-
tion betweenthe perpendicularcontractionsof p towards the bond path which
lead to a concentrationor compressionof chargealong this line,and the parallel
expansionof p awayfrom the surfocewhich leadsto its separateconcentration
in eachof the atomic basins.Thesign of the Laplacian of p atthe bond critical
point, the quantity Y2p(r"),determineswhich of the two competing effectsis
dominant and, becauseof the appearanceof V2p(r.)in the local expressionfor
the virial theorem, eqn (7.4),its sign also servesto summarize the essential
mechanicalcharacteristicsof the interaction which createsthe critical point.
190 CHEMICAL MODELS AND THE LAPLACIAN 7.4

There is, therefore,an intimate link betweenthe topological propertiesof p(r)

and its Laplacian, the trace of the Hessianof p, and, through the propertiesof
the Laplacian, one may begin to bridge the gap between the form of the
charge distribution and the mechanicswhich govern it.
The classification of interactions in both bound and unbound states is
presentedfirst, followed in Section7.5 by a discussionrelating the parameters
of this schemeto the requirementsfor obtaining a bound state.

7.4.2 Classification of atomic interactions

Properties of the charge density at the (3, - 1) critical point for a number of
diatomic moleculesare given in Table 7.4 and, for some polyatomic mole-
cules,in Table 7.5.Contour and relief maps of the Laplacian distributions for
some of thesemoleculesare shown in Figs 7.15-7.17.
When Ytp(r") < 0 and is large in magnitude, p(r") is also large, and
electronic charge is concentratedin the internuclear region as a result of the
dominance of the perpendicularcontractions of p towards the interaction line
or, equivalently in thesebound systems,towards the bond path. The result is
a sharing of electronic charge by both nuclei, as is found for interactions
usually characterizedas covalent or polar and they shall be referred to as
sharedinteractions.In sharedinteractions,as illustrated in Figs 7.15-7.17, the
region of spaceover which the Laplacian is negativeand which contains the
interatomic critical point is contiguous over the valence regions of both
atoms and the VSCCs of the two atoms form one continuous region of charge
concentration.The interaction is dominated by the lowering of the potential
energyassociatedwith the formation of the (3, - 1) critical point. In a shared
interaction, the atoms are bound as a consequenceof the lowering of the
potential energy associated with the concentration of electronic charge
between the nuclei (eqns (7.4) and (7.5)).
This concentration of electronic charge in the interatomic surface is
reflectedin the relatively large values of p(r"), the value of p at the (3, - 1)
critical point, for the moleculesentered under shared interactions in Tables
7.4 and 7.5. In all but one of these examples,the ratio of the perpendicular
contractions of p to its parallel expansion,as measuredby the ratio lArllAr,is
greater than unity. In cases of tight binding, as evidenced by the large
negativevaluesof Y2p(r")in Nr, for example,A, as well as 2, and )", are large
in magnitude,but the ratio lArlll, is still greaterthan unity. Occupation of
the antibonding 2n orbital of AB or of the corresponding z, orbital of Ar,
causesan increasein 2r. Compare, for example,2. for NO, NO-, and O,
wit h t hat f or Nr .
The second limiting type of atomic interaction is that occurring between
closed-shellsystems,such as found in noble gas repulsive states, in ionic
 7.4 C H A R A C T E R I Z A T I O NO F A T O I V l I CI \ T E R A C T I O N S 291

0 ) H
s( 3<t r V ) c C < . _ . a o c , l
o l - S t i a
B
O \ c C
+ - 3 o ' o \ o \ t
. { . : : A - { A J
: < _ T
0 ( ) d + + * * * O A
z 5 =
*a.l
ro ro a ,
u i : \o d \o € o\ co r-- ctl ci o' O $ - \ C
g
)
c
=
:
--<<-"q n v?cl1t.1qnoq
- C - l - - t - - *
g v v

a ' 5
= o F - \ o r - o \ o v \ o
rt r.) o\ # 1^.t n o 51- q: !? q o o c'l
6 d i r c l o o O O O O O O * -
. t r r , r i t o l O O O O O O O O ._
v c g

)l
?
9 )
-,
€ O c . ) $ * O t t $ ! - ! 9 Q f - *
>| Q F ro .o ol + t r* q\ q. -t r,
( ' 1 c{
O - t O q ! % ) \ O . . n
bo ' d
= - ^t , r r r - X ( )
(.) - ' = , { , i A - i
- , J | ) t J J J u v 9 9
O
A
v
A
v
6 - C O C C -
-v i v d d d a a - A

d O

tf
.9 -o - .a a{ \O $ c-) O - 9 ci tn :. t-- O
q o i9 r o o o p Q 9 y Q n n = P =
t r g ' c t . r o c . ( . . r o R =(J = (J= =(J . t(J 5 \r i i 9
A o .a a.l a.l a.J a
E 8 : - d o o o o i i o o o o i
o =
L

-\ $ aa
5**\o
Ca Ca c}\ C-
o \ r - o o o $ t f , \ o r -a )
* * C\l * t # * - O - l c >
. t r ) =
Q - l . . iJ : d c > o i o o i o o C)>i ()
c \ - C \

a o-;
! t

r - O
S
() O\ $ r- oO $ .. O\ r- - \9 I O, \9 q,
+ : F - r o - c o s | - n O Q S \ o \ o ?l'r'-:
o s a
Li
C O - n o . c o
6 6 oo c-r O O
O s | -
W O
c . l - ? Q S { , =
r- V (^.1 .o $ 5
E : \ a o c t
C) - o c J - - C ) . . r c ) c ) o o o o - - a :
3 s
' J v =
R
a r : : -
N C)
..) It- oO f-- O ca O $ \O oO rt 1' I-- l-- - ,- =r S ;
L
; 56 ii 6 oo co c- .-) oo c.l O ot - c'l - - ' J
"n + - r- r- $ r-, t- S!' 3o r-- .',
(-) 0.)
O.
5 '
O.
5 o \ o t r l $ O r - , - O q O O -
It
t*
0 ) - , ' ' L 3 5 : - r : ; i i i i i i i c
OO ri = ? -
rr I A
_ 0.) r l l l l l l l l l l l l lq 2 l l
U '. =,;
= v L

t .a c) ca * r-- - $ (.) r* !' tr) c{ \o

d od - rri ,rr o.t O + ta, trt Q) t'| c'r O c d ' - v

t-- C\ rr) tr) OO - tn tn 9 iO i rO - rt ( ) 0 ) ,

in = cj O 6 O (^'l c. $ 6l c{ $ ca \o
lr'i ()
- l c ; i . . i . . j; o J o i i o i i
N
r r l l l l ;Y.;
; k:c
( ) - ] L

OO O O\ 1.a q1 cA t-- $ t-- c\ oO

tat
oO
tf,
$
(.l
q) c <= €
c.l .r^, * cn rn .- \O * !a 9 E) - ; O -
c-t r- .rr ci .. q\ ci t't l.) q\
i- c.l oi !'
i\ J a.- .rr rn,ri O * Q O O O O O
o o o i i o ( t o o o o i o i i

r-(-- 5
E r c''l --
(^] 9 t.).]O\q!!a 7/
o'r)oo$.f, aaQ
4 - O \ o r ' - r - - ! ! O A a o l \ o c ' l o O
5 < - - - - . 1 h q q q 1 9 n 9 - * :
E i c.; c.i c.i c-i c-t - c. .. $ .. $ .. $ .a
o t r
6l . 6E ) c0q? C
;
)

g
L

- E ;-i;^i
'- -.1 .:
C ) o . = ^ ^ - ^ N ^ A , * - * 1 , ^ r x r t l ^ d>ic

B ; E r, ^fr-l
EC'l*"d tt,t!;?t
,tilaiir,-F
; -
v
- . =
l^ -
:
\O

i kl-^xcoi 3 T , z < ) z z Y >

?I<3
a 6 i = t ' z z z O O l + - + +
 I

292 C H E M I C A L M O D E L SA N D T H E L A P L A C I A N

-
c a . a \ O r O - O o O $ \ O \O tt- (:- t-- \o oo
O O O . O \ c a $ - O \ $ . a ^ A t - -
.a c...1 * ca - a.l - al c.l ooooo @ O o o
a\ 4 , - ^ l A d - i - : , { ; ; -i

r\
= v ) . a t . ) . a O O O
ra ca \o
-tr)\t
F- ca (^.l
F - c ' l .a ntr)
c 9 q 9 c . l q = q n - € + - -
c!- - i - i - i A A / i
r\ :s$'^\oNc..l

= O t - - O O a - r O : O \
$ c-r air l- c.) $ c-r
r a \ O
CO O\
r = c a O
:tsf-
N
ca
O\
O\
oo
(l/.)
- (-.l (..l
v v v v v v v v v
rt - i A A - i A A

$ O . . r ( - . l O C \ O - l - - -.aO tr) rr) ca

t--- \o c.] co @ ca * o ra) 9
\O<'C..i.a*.aca <t.a g v 9 9

ri ; A A A A A
v v v v v v v \ g l g

-\ -
c a t \ O O
.-)
J(
.: .1 99\v?clcl
S ca C-i t-- F- O\ O\
n,.?
O\ A t - - .
c..l c\l c.n c..l
oo t-.

^ $ c . . l - - - - - . : A . : A -i zi

iJ O c - l O \ \ O * c a \ O o o O \ o$trl o\ tr)O
O\ OO C\] $ l--. (..l cO cA ra) - O r $ F - r.l.) (--
r-
\\ OO \O O\ OO Lr) c-r $ ca F- - A N
r-- s
- c ' l . . ]. l = q 9 9 q\
- i A A / i A - i

iJ t \Or..l $oO-n cjc'..l OOc-'l \o

. a $ F - c O I t - c O $ c O O $ r O - - ci co
\O I-* I-- N * O rO cO (...l ('t.a$ N
f \ a t t $ . n F - O O o O o O
oiiio..i ..i ^i r..i - i - i A ^ ri -i
c ) s
€ h
d
t t t t l t t t l l l t l

, e
- r- O c..l (^..l oO cO l--
a. tr- tr- 6t\ f-- tr: cO
a'l (^.l
CO O
f-- \O c'.1
< t O -
\O
r aa co
+ o l - - o o - \ o - c o c \ ..l :f, $ C']
R o O f - - $ < t f - O - o O @
\ oooii6i c.i c.i t^.i A A A - i A A
L
qJ t l t t l t l t t i l
U
R
r. c n - f , v ) * $ \ o \ o \ o . a cacoC) c- V *
R o r e - r - . a o o - $ c . l $ (..l O\ C\ $ @ -
o o - r o ( ^ a F - v . a iO-Or
U - Q 9 v ? q n n
$ \ o
q q - - - \O s . a

- . ' . . . . - - - - - ^ r A r F - - i A A - i A - i
N
t l t t t t l
t--CO ca OO O\(.) .a\O co a-l $ O\ O\ t--
c.t \o 6l o\ l-- o a-l !t $ O\ \O co \O O\ t-t
\O (..l V') !f, f-- O\ OO O OO * a\ c.l
.a .a (\.l C\ C-.1 .a .i $ .a
- ; ^ ; / i . - i ^ : - i . i . i A / i A - i A A

O \ O \ \ O O o O ( l . l O - N - - + - o o -
o o * o o . . ) s c o o r o \ * ta) \O tr) 6l oo rt
9 9 c 9 c q q \ \ 9 r : oq.1v? q qoq
c..l c.l c..l c\l o.l - - - - aa .a .a ca cas
o . ' A X A
F 5 i p g,c' '=
qr 4, - = a .i Er
c F ! F F E
t-
{":Sig*"ci ?
: : -
-
F ==.==
E
o
E '= E E B - ; I 1 E E E EE A
.=.=.=.===i .= j
i -=_=_==.. .i=
f E E E E E E ! -z: c"=, i g' , g9 E
, *E 7 :LEs ;
e; e==
€
(.)
C)
E::€EEzz= t
E E = E; ;r r E -E lE: c} i} : r Zi z
( n \9J 9\ J?\ J! w, vF C: C = ":
2 t Z<
7.4 OF ATO\IIC I\TERACTIONS
CHARACTERIZATION 293

bonds, in hydrogen bonds, and in van der W'aals molecules.They are

illustratedby the secondset of entriesin Tables 7.4 and 7.5.One anticipates
that such interactionswill be dominated b1 the requirementsof the Pauli
exclusionprinciple.Thus, for closed-shell p(r.) is relativelylow in
inteructiorrs.
value and the value of Y2p(r") is positire. The sign of the Laplacian is
determinedby the positive curvature of p along the interaction line, as the
exclusionprinciple leads to a relative depletion of charge in the interatomic
surface.Theseinteractionsare dominated by the contraction of chargeaway
from the interatomic surfacetowards each of the nuclei.The Laplacian of p is
positive over the entire region of interaction and the kinetic energy contribu-
tion to the virial from this region is greater than the contribution from the
potential energy. The spatial displays of the Laplacian of p given in Figs
7.15-7.17areatomic-likefor the closed-shellinteractions.The regionswhere
the Laplacian is negative are, aside from polarization effectsto be discussed
later, identical in form to those of a free atom or ion. Thus, the spatial regions
where the potential energydominates the total energyare confined separately
to each atom, reflecting the contraction of the charge towards each nucleus,
away from the region of the interatomic surface.In a closed-shellinteraction
which gives a bound state, the atoms are bound as a consequenceof the
charge which is concentrated within the basin of each atom. The ratio
lArllA. < 1 in all the examplesof closed-shellinteractions.
The almost complete interatomic transfer of one electronic charge indi-
cated in Table 7.4 for the ionic systemsis verified by the nodal structure of the
correspondingLaplacian maps.The cations,Li*, Na*, and K*, all lack the
outer nodes associatedwith the valencedensity distribution of the isolated
atom. Thus, Li in LiCl has but one negativeregion rather than two, Na in
NaCl has two rather than three,and K in KF has three rather than four. The
reader is referred to Fig. E7.2 for displays of the charge distributions and
interatomic surfacesfor some of these systems.Another characteristic of a
closed-shellinteraction exemplified by the alkali halides and discussedin
Section E7.1 is the separatelocalization of the electrons within the basin of
each atom, as determined by the spatial localization of the Fermi hole.
Data for the hydrogen bonds in the dimers of HF and HrO are given in
Table 7.5. A hydrogen bond results from the interaction of two closed-shell
systemsand the properties of p at the associatedbond critical point reflectall
of the characteristicsassociatedwith such interactions:a low value for p(r")
and Y2p(r") > 0. Each HF or HrO fragment is easily recognizable in the
display of the Laplacian for the correspondingdimer. None of the character-
istics associatedwith the sharing and accumulation of charge are evident in
the neighbourhood of a hydrogen bond critical point. The negativeregionsof
the Laplacian are separately localized in each HF or HrO fragment in a
pattern similar to that found in the monomers. The O-H and F-H inter-
actions in the monomers and in the monomer fragments present in the
 B2 Ar2

I
@b
IN2

I
\/w
r^i7il

I
O2 Licl

i
wINO
r/l
I

NaCl
7.4 CHARACTERIZATION OF ATOlIlIC I\TERACTIONS 295

hydrogen-bonded complexes are typical of polar. shared interactions with

vrp\rJ < 0. Thus, a hydrogen bond is defined to be one in which a hydrogen
atom is bound to the acid fragment by a sharedinteraction and to the baseby
a closed-shellinteraction.
The Laplacian map for the dimer of HF is topologically equivalent (i.e.
continuously deformable)into that for Ner. Why certain closed-shellinter-
is
actions are stable with respect to separatedspecieswhile others are not
the polarization of
related to the extent of interatomic charge transfer and to
the atomic distribution, a topic discussedin Section 7.5.
The Laplacian
-of distributions for Sl* and SrN, in Fig. 7.13 provide further
examples sharedinteractions,the unequal sharing of the chargeconcentra-
final
tion in the latter molecule being an example of a polar system' The
molecule in this figure, Cl?* , illustrates both types of interactions. This
moleculeconsistsoft*o identical tightly bound Clj- ions, with a bond length
The
of 1.88A, weakly linked together at an equilibrium separation of 2.77A.
value of
two short bonds are shared interactions with the relatively large
-0'227 au'
0.206au for p(r.) at the bond critical point and with Y'p(r")--
The two long bonds have p(r") : 0'032au and Y2p(r"): + 0'082au' In the
the
short bonds, electronic charge is concentratedbetween the nuclei along
path. In the long bonds, electronic charge is concentrated separately in
bond
the baiins of the atoms linked by the long bond paths. Of the two peaks that
appear within the VSCC of each Cl atom and along the-axisof a long bond,
: - 0.541au)
the peak on the bonded side is of greater magnitude (V2p(r")
-
than is that on the non-bondedside(v2p(r') : 0'510au)' Thus' the charge
density of each Cl atom, while localized within the basin of the atom with
1..rp..i to the long interactions, is polarized in the direction of its bonded
neighbour.
Also included in Tables 7.4 and 7.5 are the values of the kinetic energy
density G(r") at the interatomic critical point and its parallel and perpendicu-
(5.48))
lar components.It is clear from the definition of the density G(r) (eqn

with shared and

Frc. 7.15. Contour maps of the Laplacian of the chargedensity for molecules
interactrons. Positive values of Y2p are denoted by solid contours, negativevaluesby
closed-shell
by the innermost
dashed contours. The Laplacian is also negative within the region bounded
the remaining figures'
solid contour enclosingthe nucleus for all uto-t beyond He in this and
Appendix) are not as
The values of the contJurs in theseand succeedingfigures(seeTable A2 in
as the extent and relative positioning of the regions where V2p is either positive or
important
one anothei when v2p undergoes a change in sign' The
negative. Some contours overlap
surfacewith
outermost conrour rn each plot is + 0.002au. The intersectionof each interatomic
- l) critical points, the point
the plane of the figure is also shown. The values of Y2p at the (3,
(which in^thesemoleculesis
where the interatomic surfaceintersectsthe atomic interaction line
coincident with the internuclear axis) are recorded in Table 7.4 under Vtp(r"). In a shared
of both atoms and is a
interaction the regron of charge increaseis contiguous over the basins
interaction
result of the contractions in pierpendicular to thi interaction line. In a closed-shell
of charge concentration are separately localized within each atom and the interaction
the regions
is dominated by the contractions in p towards each of the nuclei'
296 CHEMICAL MODELS THE LAPLACIAN
7 .4 C H A R A C T E R I Z A T I O NO F A T O \ { I C I \ T E R A C T I O N S 297

that it is everywherepositiveand finite. Its r aiue is discontinuousat a nuclear

position becauseVr/ is discontinuousthere.The conditionson the first-order
density matrix for nuclear-electroncoalescencehar e been given by Bingle
(1 9 6 3)and by P ac k a n d B ro w n (1 9 6 6 ).T h e s ee rpressi onsmay be used to
show that the limiting value of G(r) at a nuclear position is IZtp(O) for a
spherically averaged density (Bader and Bedd all 1972).The density G(r) is
expressiblein terms of three contributions along orthogonal axes. The
relativevaluesof the componentparallelto the internuclearaxis (G11)and one
of the components perpendicular to this axis, (G,) faithfully reflect the
magnitudes of the corresponding curvatures of p at r". For the shared
interactions,Gr(r") t Gtt(r"),while just the reversesituation is found for the
closed-shellinteractions.In addition, as anticipatedon the basisof the local
virial theorem (eqn (7.4)),the kinetic energy per electronic charge, the ratio
G(r")1pft"),is lessthan unity for the sharedinteractionsand greaterthan unity
for the closed-shellinteractions. Thus, when the positive curvature of p is
large and dominated by the contraction of the charge towards the nuclei, the
kinetic energy per electron is absolutely large and the value of its parallel
component exceeds that of a perpendicular component. In the shared
interactions,the accumulation of charge in the internuclear region leads to a
softening of the gradients of p and of the correspondingcurvature of p along
the interaction line, and the parallel component of G(r) is correspondingly
less than its perpendicular components. The dominance of these latter
components again mirrors the corresponding dominance of the perpendicu-
lar contractions of p towards the bond axis in sharedinteractions.Becauseof
the concentration of charge and the concomitant negative values of the
Laplacian of p over the sameregion, the potential energyis dominant and the
kinetic energy per electron is absolutely small.
Extraordinarily small values of G(r"\lp(r"), of the order of 0.03au, are
exhibited by the non-nuclear maxima, the pseudoatoms, in the metallic
clustersof Li and Na atoms illustratedin Fig. 2.11.The samesmall valuesof
kinetic energy per electron are reflectedin the ratio of the averagevalues of
f(A) to l{(O) for the pseudoatomsand, in accordancewith the Heisenberg
uncertainty principle, they indicate that the charge density of the pseudo-
atoms is loosely bound and unconfined.The Laplacian distributions for these

Frc. 7.16. Relief maps of the negative of the Laplacian distribution of p to contrast the
distinguishingfeaturesof the shared and closed-shelllimits of atomic interactionsas represented
by Nr"and,{rr, respectively.The map for F, is intermediate between the two limits. While
ir r(1") is positwe for F, as found for Ar,, its value of p(r") is three times larger than that for-Art.
Elecironii charge is accumulatedrn the binding region of Fr, as is typical of a shared interaction,
but is concentiated in the atomic basins, as is typical of a closed-shell interaction' While
Vrp(r") > 0 for both Ar, and F2, the Laplaciandistribution is a minimum at r. for Ar,. but a
(2d.2p)
m a x i m u m a t t h e . u m e p b i n t i n F r . T h e c h a r g ed e n s i t i e sa r e c a l c u l a t e du s i n g a 6 - 3 1 1 G
basis in an MP2 calculation.
298 CHEMICAL MODELSAND THE LAPLACIAN 7.4

q . - , - 0 ) c !
- t \ -
c n '= ^:^=

a g ;= E = Z
= = i j
) - t
= ::.
L ! v i '

Er! eEZ
, F - . ; c l ?
' ! - -

.:ilZ|i_
t ' f \ J J L -

F ' Y ,c .- o a ? -) =-

f" cl F i i 7 :
v - - - - - l
i
B'; i o, l= |
= y r = 5 . - |
i = 3 2 . = r : =|

g-
c r f : - . i t i . . I

I
o
(o
(D
I iEi g;E
?!=,I
:t'
ile*

to\
- \
sl
o

!
].4 C H A R A C T E R | Z A , T | O No F A T o \ 1 I C I \ T E R A C T I o N S

clustersshow the metal atoms to be missing their outer charge concentra-

tions,indicatingthat they are presentprimarill. as ionic cores(Fig' 7'18)'The
chargedensityof the coresis highly localizedas indicatedby valuesin excess
of 90 per cent for the contained Fermi correlation. The charge on the
pseudoatoms,
^Fermi however, is very delo calized with values for the contained
correlation being only 30 per cent of that required for complete
localization. Aside from the inner-shell charge concentrationsof the metallic
cores, the negative values of the Laplacian distribution are confined within
the boundaries of the pseudoatoms.The value of V2p(r.) is positive at the
bond critical point linking a pseudoatom to a metallic core, but negative
at a critical point linking two pseudoatoms. The properties of p at the
critical point linking two pseudoatoms are characteristicof a weak, shared
- 2 .6 x 10-a au, and G(r" )l p$" )
i n te rac t ion:p( r " ) : 0 .0 0 5 6 a u , Y2 p (r" ):
: 0.10 au. The ionic cores are linked to one another through the pseudo-
atoms and the study of these clusters gives a model of Group I metals
consisting of positively charged metal atoms with very localized charge
distributions immersed in and bound by an intermeshed network of nega-
tively chargedpseudoatoms.Metallic binding is a result of the lowering of the
potential energy associatedwith the charge concentrations of the valence
density, and these concentrations are confined within the boundaries of the
pseudoatoms. These same charge concentrations have a very low kinetic
,n.rgy per electron and the potential energy lowering associatedwith the
formation of the pseudoatoms is obtained without a large accompanying
increasein the kinitic energy. As a consequence,the electron density of the
pseudoatomsshould be mobile under the influence of an electric field. The
quantum mechanical current for a core density is simply the diamagnetic
current circulating around the nucleus and the metal atoms in theseclusters
will not contribute appreciably to the electrical conduction. It is the highly
delocalizeddensity oi the network of pseudoatomsthat is responsiblefor the
binding in a metallic system and for its conducting properties'
Observations regarding the behaviour of the parallel and perpendicular
components of the kinetic energy and their relation to the gradients and
.uruutur.s of p in molecular systemswere first made by Bader and Preston
(1969)for the moleculesH, and Her. They studied the spatial properties of
G(r) and K(r) (eqn (5.48))and their relation to l,(r) (eqn (7.1)),the density
proportional
'C(.j to the negativeof the Laplacian of p. The differing behaviour of
in the binding region of a bound and an unbound system is made very
clear by comparing the plots of this function these authors give for H, and
Her. It wu, ,ho*n1hat the value of L(r) in a region where the Laplacian of p
is nlgative is a measure of the extent to which the charge density can be
concentratedin regions of low potential energy beyond that anticipated on
classicalgrounds.The kinetic energydensity G(r) and its componentsmay be
 1 1
300 CHEMICAL MODELSAND THE LAPLACIAN

.:..:,
t : :

F r c . 7 . 1 8 . ( a ) C o n t o u r d i s p l a y o f V 2 p ( r ) f o r t h e N a o c l u s t e ri n t h e p l a n e o f t h e n u c l e i .T h e
locations of the two non-nuclear attractors are indicated by stars and of the bond critical point:
by dots. The contour values are given in the Appendix (Table A2). The inner contour
encompassinga sodium nucleusalso encompassesthe innermost region of chargeconcentration
The third, or valence shell of charge concentration of the Na atoms has essentiallyail been
transferred to the pseudoatoms and the region where Y'p < 0 is almost completely localized
w i t h i n t h e i r b o u n d a r i e s .( b ) C o n t o u r d i s p l a y o f V 2 p ( r ) f o r A r H F . T h e i n n e r m o s t c o n t o u r
encompassingthe Ar or the F nucleus enclosesthe innermost region of charge concentration.
The Ar atom, unlike the Na atoms in (a). possesses all three quantum shellsby exhibiting three
alternating shellsof chargeconcentration and chargedepletion.Contrast the relativeplanarity oi
the Ar-H interatomic surfacewith the curved nature of the H-F surface.The same contrasting
behaviour in the shapesof the interatomic surfacesfor the closed-shellinteraction, V2p(r") > 0.
a n d t h e s h a r e di n t e r a c t i o n .V 2 p ( r . ) < 0 . l i n k i n g t h e p r o t o n a r e f o u n d f o r a l l h y d r o g e n - b o n d e d
7.4 C H A R A C T E R I Z A T I O NO F A T O M I C I \ T E R A C T I O N S 301

relateddirectly to the gradientsof the orbital densitiesp,and their occupa-

tion numbers n,,
G(r): (h' l8m)l1,Yp,'Yp,,
p,. (7.e\
For a one-electron system,G(r) is relatedto the gradientsof p(r) itself.For
Hr, wherejust oneorbital(1o*)and,for Her, wherejust two orbitals(1o,and
1o")makethemajorcontributions to p,theaboveobservations regarding the
parallelingbehaviourof G_r(r)and Gq1(r) with that of the corresponding
gradientsand curvaturesof p for the sharedand closed-shell interactionsare
madeclearby eqn (7.9).However, the data in Tables7.4 and 7.5 showthat
theseare generalobservations and that, evenin many-electron systems, the
local behaviourof the kineticenergydensitymay be relatedto the gradients
and curvaturesof the total chargedensity.This correlationis partially
accountedfor by theory through eqn (7.4).When V2p(r)> 0 and the
Laplacianis dominatedby the positivecurvatureof p (contractionof p
towardseachnucleus), the largercontributionto the Laplaciancomesfrom
the kineticenergy-and,by observation, primarilyfrom its parallelcompon-
ent.Correspondingly, when the Laplacianof p is negativeand p is concen-
trated as a resultof contractionsperpendicular to the bond path, not only
does the potential energy make the dominant contribution to the virial but
one observes that the perpendicular components of G(r) dominate its parellel
component. Thus, one concludes that the kinetic energy dominates the
contributions to the virial and to the energy in regions of space where its
parallelcomponentis dominant.This occursin regionswhereV2p(r)> 0.
Conversely,in regionswhere the perpendicularcomponentsto G(r) are
largest,the potentialenergymakesthemajorcontributionto the virial and to
the energyof the system.
As notedby Baderand Preston(1969),the differingbehaviourof G;1(r) and
G_r(r)in H, is so extreme as to be reflected in their integrated average values.
Thus,the ratio of the averagevaluesof the parallel(7p1) to the sum of the
(I_,-
perpendicular ) components of the kinetic energy for H, at its equilibrium
separationwasfound to be 0.3569 as compared to the value of one-halffor
the separatedatoms.As noted by these authors, this result is consistent
with the original observation made by Couls on (l94la,b; Coulson and
Duncanso n l94la,b)in his study of the electronic momentum distribution for
the H, molecule.He found that the mean component of the electronic

systems(see,for example,the displays for the water and hydrogen fluoride dimers given in Fig.
i.tl1. I small bonded maximum is formed within the valenceshell of chargeconcentration of the
Ar atom in forming the complex with HF. The value of this maximum in the Laplacian
distribution, equal to - l.l9 au, and its position, 1.088au from the Ar nucleus,representonly
small perturbaiions of the spherical surface of maximum charge concentration present in the
VSCC of the free Ar atom (seeTable 7.1).
302 CHEMICAL MODELSAND THE LAPLACIAN 7.4

momentum in the direction of the bond axis is decreasedwhile the mean

component perpendicular to the bond axis is increased.For the closed-shell
interaction in Her, just the opposite behaviour is obtained. Bader and
Preston noted that, in Her, Glr(r)> G_r(r)and they found T,lT, > 0.5. This
dominance of the parallel over perpendicular components of I, both locally
and overall, is anticipated on the basisof the Laplacian map for He, which is
everywhere positive except for small spherical regions encompassingeach
nucleus(Fig. 7.15).Thesepropertiesof the kinetic energyare also evident in
the electronmomentum distribution of Her. Ramirez (1983)found that this
distribution in He, is ellipsoidal with the major axis parallel to the line of
interactions,just the opposite of the situation found in H, where the major
axis of the momentum distribution is perpendicular to the bond axis, as
found by Coulson (I94la,b).
For systemscontaining atoms other than H and He, one doesnot find such
large deviationsin the ratio of ft lfTrfrom the value of one-half,sincethe
average kinetic energy components in these systemsare dominated by the
core energies.The contributions of the components to the local value of G(r)
in the valenceregion, however,do parallel the behaviour found in H, or He2,
as exemplified by the data in Tables 7.4 and 7.5.

7.4.3 Hydrogen bonds and van der Waals molecules

The van der waals molecules, NeHF and ArHF, although very weakly
bound, can also be classified as hydrogen-bonded complexes, since the
hydrogen in these molecules is bonded to the acid fragment by a shared
interaction and to the base by a closed-shellinteraction. The charge density'
at the critical point linking the hydrogen to the base atom in the van der
Waals moleculesexhibits the samecharacteristicsas it does in the hydrogen-
bonded molecules.This is illustrated by the data in Table 7.5.Theseare weak
interactions: the charge accumulated along the bond path and the stresses
induced in p causedby a van der Waals interaction are small, as evidencedby
the valuesof p(r") and by the magnitudesof the three contributing curvatures
to Y2p(r "). The two kinds of interactions also have common featuresin terms
of the changesin the properties of the atoms brought about by the formation
of the complex (Table 7.6). (The atom of a polyatomic base engaged in
hydrogen bonding is denoted by the symbol B and the remainder of the
molecule by X.) tn addition to the van der Waals molecules,NeHF and
ArHF, which have dissociation energies(D") = l kJ/mol, data are given for
hydrogen-bondedcomplexesof HF with hydrogen cyanide and the nitrogen
molecule.Thesetwo moleculesprovide examplesof normal (D" > 15kJ/molt
and weak hydrogen bonds, respectively.Similar studieshave been published
for a large number of hydrogen-bondedsystems(Carroll and Bader 1988).
 7.4 C H A R A C T E R I Z A T I O NO F A T O I \ { I C I N T E R A C T I O N S 303

Table 7.6
(a) Properties of HF nith HCN, Nr, Ne, and Ar*
of the hydrogenbond in compleres

XB-HF D rH rts lAr"l lArul I lftl po(rn) po(r" ) Lpo

HCN HF 25.65 1 . 1 6 32 .3 9 1 0 . 9 2 1 . 1 8 2.10 0.0135 0.0137 0.0272

NN HF 7.87 1 . 3 5 9 2 .5 6 4 0 .7 2 0 .9 1 1 .6 3 0.0079 0.0082 0.0161
Ne-HF 0.71 1 . 5 1 82.469 0.56 0.38 0.94 0.0050 0.0040 0.0090
Ar HF 0.76 1.6 6 9 3 . 1 7 2 0.41 0.55 0.96 0.0033 0.0037 0.0070

(b) Changesin atomic properties

XBHF O A N(O) A IM (O )l L Q,,(Q)

HCNHF F 0.044 - 0.0140 - 0.0257

H - 0.012 - 0 .0 1 8 4 0 .0 9 8 5
N 0.038 0 .0 3 3 5 0 .2 2 7 7
C - 0.046 0.0258 - 0.2089
H - 0.022 - 0.003r - 0.0182

NNHF F 0.019 - 0.0037 - 0.0091

H - 0.004 - 0.0093 0.0640
N 0.029 - 0.0041 0.1106
N - 0.044 - 0.0009 - 0.0239

NeHF F 0.006 - 0.0002 - 0.0052

H - 0.005 - 0 .0 1 0 6 0 .0 4 3 4
Ne - 0.001 0.0391 - 0.0487

ArHF F 0.005 - 0.0004 - 0.0008

H 0.004 0.0004 0.0316
Ar - 0.008 0.0439 0 . 1 281

* D . i n k J / m o l . O t h e r q u a n t i t i e sa r e i n a t o m i c u n i t s .Results
fromMP2 6-311G(2d,2p)lMP26-311G
(2d,2p) calculations.The D" values are corrected for BSSEusingthe full supermolecule basisset
(Boysand Bernardi 1970).

The data are obtained from large basis set calculations(seeTable 7.6)which
include electron correlation (second-order Moeller-Plesset perturbation
theory) and which are corrected for basis set superposition error (BSSE),
(Boys and Bernafii 1970).While the resultsfor the stronger hydrogen bonds
are qualitatively the same with or without the inclusion of correlation, the
weaker interactions require the better description of the wave function.
The D" value for NeHF in Table 7.6 compares well with the value of
0.78 kJ/mol calculatedby ONeil et al. (1989)using a large basis set on a size
 i

304 CHEMICAL MODELSAND THE LAPLACIAN 7.4

consistent calculation in the coupled electron pair approximation (CEPA)

and corrected for BSSE. Hutson and Howard (1982)have determined an
intermolecular potential for ArHF by u least squares fitting of molecular
beam spectraldata. In the caseof HF complexesthere is little information in
thesespectra on the absolute well depth. Their estimate of 2.6 kJ/mol for D"
of ArHF probably does correctly indicate however, that the calculatedvalue
is underestimatedfor this system.
A hydrogen bond resultsfrom the mutual penetration of the van der Waals
envelopes of the H atom of the acid and of the atom B of the base, the
strength of the interaction increasingwith the degreeof mutual penetration
(Carroll and Bader 1988).The bonded radii of the B and H atoms-the
distance of a bond critical point to the corresponding nucleus-for the B-H
bond in the complexes are given in Table 7.6. Each value is less than the
correspondingvan der Waals non-bonded radius-the distanceof the H or B
nucleusto its non-bonded 0.001-aucontour in the isolated reactant-and the
valueslAro I and IAr, I are the measuresof the extent of penetration of the van
der Waals envelope to yield the bonded radii found in the complex. Also
listed in Table 7.6,under the headingspo(rrr)and po(rn),are the values of p in
the isolated acid and base molecules at the point of penetration, i.e. at the
position determined by the bonded radius rH or ru. The sum of these two
values,Lpo, is only slightly smallerin value than p(r"), the value of p at the
hydrogen bond critical point (Table 7.5).Unlike a sharedinteraction, there is
no large increase in charge density at the bond critical point and in the
interatomic surfacerelative to the unperturbed densities.The final density is
instead determined primarily by the extent of penetration-the greater the
penetration the larger the value of p(r") and, in general, the stronger the
resulting hydrogen bond. To obtain hydrogen bonds with D" > 25 kJ/mol.
the van der Waals envelopeof the H atom must be penetratedby = I au and
that of the B atom by a still larger amount. At thesepenetrations,both of the
unperturbed densitieshave values in excessof 0.01 au.
Argon, the softer of the two base atoms, is penetrated the most by the
hydrogen in the van der Waals complexes,while Ne, the harder of the tir tr
bases,penetratesthe hydrogen atom's density to a greater extent than doe:
Ar. The sum of the B and H penetrationsare nearly equal in thesetwo weaklr
bound complexes.While the penetration of the argon atom is greater than
that of the neon atom, the values of po(ru) at these penetrations are nearlr
equal, as the argon valence density is more diffuse than that of neon. The
greater penetration of the hydrogen by the harder baseneon givesrise to the
largest value of p(r"). The greater penetration of a given acid atom by the
harder of two base atoms appearsto be general(Carroll and Bader 1988).In
the complexesHF-HF and HCI-HF for example,the F atom penetratesthe
hydrogen of HF to almost twice the extent than does the Cl atom, 0.81
compared to 0.42 au, respectively,while the two baseatoms are penetratedbr
7.4 C H A R A C T E R I Z A T I O N O F A T O \ 1I C L \ T E R A C T I O N S 305

similar amounts.In a like vein,the pair of complexes.HCI-HF and HF-HCI,

allows for a comparison of soft and hard basesinteracting with hard and soft
acids,respectively.The charge distribution of HF and HCI can be compared
in Fig. 6.3.The non-bondedradius of H is 2.01au in the hard acid HF and
2.28 au in the soft acid HCl. The penetration of the soft acid H by the hard
baseF is more than twice the penetration of the hard H in HF by the soft base
Cl--{.96 au compared to 0.42 au-while the base atoms are penetratedto a
nearly equal extent by the hard and soft H atoms. Unlike the weakly bound
van der Waals complexes,penetration is overriding for a hydrogen bond with
the result that HF-HCI is the energeticallyfavoured complex'
A study of the changesin the atomic moments (Table 7.6(b))indicateshow
the atomic chargedistributions changeas a result of the penetration of the H
and B non-bonded densities,as occurs in the formation of a stable complex.
In all casesof regular hydrogen bond formation, that is, with D" ) 5 kJ/mol,
there is a relatively small transfer of charge from the base to the acid,
equalling 0.03 and 0.02 e for the complexeswith HCN and Nr, respectively.
In the samecases,the hydrogen of the acid loseschargeand the A atom of the
acid HA gains charge. In a polyatomic base,the base atom B gains charge
while a slightly greater amount of charge is lost by the remainder of the base
molecule X. These changesare exemplifiedby the complexeswith HCN and
Nr. Experimentally, it is observed that the dipole moment of a hydrogen-
bonded complex is greater than the vector sum of the dipoles of the isolated
reactants, an effect termed dipole moment enhancement.Theory demon-
stratesthat this enhancementis a result of a loss of charge by the tail of the
base (the X group) and a gain by the head of the acid (the F atom in the
present cases).This effective transfer of char.ge,while small in absolute
amount, is across the length of the complex and it createsa moment which
adds to the head-to-tail alignment of the acid and base dipoles to yield a
moment which is greater than the sum of the reactant moments.
The changesin the atomic populations encounteredin the formation of the
van der Waals complexesare extremelysmall. There is a predictqd transfer of
0.008e from the baseto the acid in ArHF, while the correspondingtransfer of
0.001e in NeHF is equal to the error in an integrated atomic population.
There are significantchangesin the first and secondmoments of the atoms in
these complexes,however, and these are discussednext.
The atomic densitiesof the isolated reactants,as determined by the first
moment M(O)(eqn (6.45)),are polarizedin a direction counter to the direction
of charge transfer. Thus in HF, F is polarized towards H which is itself
polarized away from F and, in HCN, the atoms are polarized in the direction
N to C to H. In Nr, the atoms are polarizedinto their non-bondedregions.
The atoms in these linear molecules have positive values for Q,"(O), the
quadrupole polarization with respect to the internuclear axis (eqn (6.48)),
implying a transfer of electronic charge from along the axis to a torus-like
 I

306 CHEMICAL MODELSAND THE LAPLACIAN 7.4

distribution about this axis.The largestchangesin thesemomentsare found

for the H and B atoms and the results for the HCN complex are typical of
most-hydrogen-bonded complexes.The chargedistributionsof the H atom of
the acid and of the B atom of the base change so as to remove density from
along their axis of approach and thereby facilitate the penetration of their
non-bonded charge distributions. The data in Table 7.6(b\ show that the
polarrzation of H away from the F is reducedand that of the N atom towards
the C of HCN is increased.Thus, the chargedistributions of both the H and B
atoms polarize away from one another in the formation of the complex.
Similarly, the quadrupole polarization of the same two atoms is increased,
indicating that an increasedamount of electronic charge is removed from
along the internuclear axis, the axis of approach, and placed in the torus-like
distribution about the axis. This correspondsto a o to z promotion of the
density.The same polarizationsoccur in the weaker complex vrith Nr. The N
atom acting as a basepolarizesslightly away from the H atom on forming the
complex and its original atomic polarizatron is decreased.
The changesin the atomic moments of the Ar and Ne atoms dominate the
formation of the van der Waals complexes.The dipole polarizations of the
noble gas atoms exceedin magnitude the correspondingchangesfor the base
atom in the regular complexesand in eachcasethey are polarized towardsthe
H atom of the acid. The polarization of the Ar atom is greaterthan that of the
Ne atom. In forming the complex with Ne the charge density on hydrogen is
polarized away from the baseatom while, in the Ar complex,it is polarized to
a much lesserextent towards the base atom. These differing polarizations are
consistentwith the observation made above that Ne, the harder of the two
bases,penetratesthe H atom to a greater extent than does the Ar atom. The
hydrogen atom undergoesthe same quadrupolar polarization in the forma-
tion of the van der Waals molecules as in the regular hydrogen-bonded
complexes but it is of smaller magnitude, being again smallest for the Ar
complex. The Ne and Ar atoms exhibit differing behaviour in this respect,Ne
increasing the amount of electronic charge along the axis of approach, Ar
decreasingit.
The hydrogen atom is destabilizedin the formation of a hydrogen bond.
the extent of destabilization paralleling the strength of the interaction
(Carroll and Bader 1988).This is true for the present examplesas well. The
A-H bond of the acid is slightly perturbed on hydrogen bond formation and.
in general, the A-H bond is lengthened.This lengthening is reflected in a
correspondingdecreasein the value of p(r"). Thesechangesare small and the
bond of the acid retains the characteristicsof a shared interaction. This is
illustrated by the data in Table 7.5 for the hydrogen bonds in the water and
hydrogen fluoride dimers. In the present examples the H-F bond is
lengthened by 0.0160au in HCNHF and by 0.0044au in NNHF, with
associateddecreasesin p(r") of 0.0154 and 0.0045au. The H-F bond is
1 .4 C H A R A C T E R I Z A T I O NO F A T O \ I I C I \ T E R A C T I O N S 307

perturbed to a much smaller extent in the formation of van der Waals

complex, the changesin the H F bond length and its value of p(r") being
predicted to be less than one-thousandth of an au for the Ne and Ar
complexes.The H-F internuclear separation is actually calculatedto de-
creaseby 0.00042au rn the complex with Ne. the most stable of the two
complexes.
In van der Waals moleculesand regular hydrogen-bondedcomplexes,a
hydrogen atom links an acid to a basefragment with a shared and a closed-
shell interaction, respectively.Both types of complexesresult from the mutual
penetration of the non-bonded van der Waals radii of the H atom of the acid
and of the terminal atom B of the base to yield a value for the density at the
H-B critical point which is only slightly greater than the sum of the
unperturbed densities.For the regular hydrogen-bonded complexes,the B
atom is penetratedto a greaterextent than the H atom and the strength of the
interaction increaseswith the degreeof penetration.In thesemoleculesthe H
and the B atoms polarize so as to facilitate the mutual penetration of their
non-bonded densities.The noble gas atoms are polarized to a greater extent
in the formation of the van der Waals complexesthan is a base atom of a
hydrogen bond. While the dipole and quadrupole polarizations of argon are
greater than those of neon, the latter atom penetrates and perturbs the
hydrogen atom of the acid to a greater extent. Thus the effecton the binding
energy of the greater penetration of hydrogen by neon is offset by the greater
polarizability of argon.
Unlike the polarization of the base atom in a regular hydrogen bond
interaction, the dipolar polarization of a noble gas atom is towards the
hydrogen.In the relativelyweak complex of HF with Nr, the changein the
polarrzation of the base N away from H is very small. This interaction is
transitional betweenthe two patterns of atomic polarizationsthat result from
the mutual penetration of closed-shellsystemswith little or no accompanying
chargetransfer,the featurescommon to van der Waals and hydrogen-bonded
interactions.

7.4.4 Intermediate interactions

The examplesso far consideredillustrate that atomic interactionsexhibit two

limiting setsof behaviour of the charge density at the (3, - 1) critical point.
One set is the opposite of the other in terms of the values of p(r") and of the
regions of charge concentration and depletion and their associatedmechan-
ical consequences, as determined by the sign of the Laplacian of p. In these
examples,the interatomic critical point is situated relatively far from a nodal
surfacein the Laplacian of p. In some molecules,however,the critical point is
located closeto a nodal surfaceinY2p. The atomic basinsneighbouringthe
 NaH

:) \
/
Ir
l @l'-
t t t
t '

r
MgH

BH
GV \{-}

\t A r H

I ,tr In)
@ rt Y
\
CH siH

,*'- /

@
/

(@* \
\

NH PH
7.4 C H A R A C T E R I Z A T I O NO F A T O \ 1 I C I \ T E R A C T I O N S 309

OH

clH

Ftc. 7.19. Contour maps of the Laplacian of p for second- and third-row diatomic hydrides.
The intersection of the interatomic surfacewith the plane of the diagram is also shown. These
maps illustrate the transition from closed-shellto shared atomic interactions.

interatomic surfaceexhibit opposite behaviour with respectto the sign of the

Laplacian of p in thesecasesand the interactions are transitional in charac-
ter. The transition from closed-shellto shared interactions is illustrated in
Fig. 7.19 for the second-and third-row diatomic hydrides, AH. The proper-
ties of p(r) at the interatomic critical points for thesemoleculesare given in
Table 7.7. The first member of each series, LiH and NaH, exhibits the
characteristicsof a closed-shellinteraction-the value of p(r") is small and
Y'p(r") > 0. The charge density of a hydrogen atom is easily polarized,
particularly when it is negativelycharged.This polarrzation is evident in the
disposition of the negativeregion of Y2p relative to the position of the proton.
The Laplacian distributions and the propertiesof p at r" for the moleculesCH
to FH and SH to CIH are typical of sharedinteractions.In CH there is a close
to zero transfer of charge and an equal sharing of the chargeincrease.In NH,
OH, and HF there is an increasingpolarrzation of the sharedcharge towards
A. The transition to shared interaction occurs earlier in the second-rowthan
in the third-row elementsas anticipated on chemical grounds (seeFig. 6.2)
and is understandablein terms of the greatercore sizeof a third-row element
compared to that of a corresponding second-row element.
, j1 0 CHEMICAL MODELS AND THE LAPLACIAN

bo
L
d - oo \f, N co tr) o o \o r.) r) o\ rf, -
* \O U-i cl C.l OO rO - Or c.] O\ t-- O\ !f,
O\ oO tt- O .i rn F- cO C- oO F- '.l-) O C..l
: <
z 5 + + + + | | t + + + + + +|

OO\F -I- -
t-- O \O Cj t-- OO # O\ ca - car F- - C:l
- rf) o\ o 1{.) <> tr) F- a..l oo .a \o \o ca
r\
O --.-.
or\ .a - - C-l N O, O O\ O\ t--. (..l N

'T
*=
q \ocooo.a-
ri - \o l- o c..l o\ \o \o \o tr) \o
.iO\NO\$ca.Otr*O\@cA
-l O O O N - t a ) O , O O O O * . a n
:\ oioc.i . r i 6 i c . io i o i i J c ' i
\
\
\)
$ cA \O CO \O O - O\ O N ..1 O\ N O
O \O O N $ \O S \O ta) $ f- a'l N -
N ta \O O\ * t- O OO - \O r- F- I-- N
(..l .i ci ta t \O ca * N a-l .a \O O $
i i J J,..j J 6i i o o i i c.j -.i

q
oO ca t-- $ r $ O\ oO .a O ca ai - r-
A (n \O t.) !-) .a \O
O O O\ O O\ O O $
@ l.-. ca t-'- tt- $ - i C\ O\ \O .a oO (...l
F- }i ra) .a <f, *
a..l oo c.l c-n ca cr) $ ca - <t
F r E o o o o o J J i o i o i o o
o h

- O\
-
N
ci
N
cl
(^.l
F-
O\
o\
OO
\o
r
-
t
o\
tr)
o\
!f,
n
\O
\o
(^..l
...l
N
.a
ta)
..)
\O O \O r .a c,.\ O cO \O O \O N t--. O
O (^.1 $ F- .1 oO cO O O - - N ci rO
il o i o i J - ' 6 i i i i i o d i
R
r t t t t t t t t t t t t l

* oo t: o\ $ c.] c.t o t .o o c.l c-r :f

F- ca S @ .a oO t-- c..l tf, cO .i $ \O N
R tn \o co ca o \f, oo ca oo co .n - \o oo
N * - tn O \O \O C-t * * - - - tf) t-
L. o o o J J o . i + o i i i o o i
a

U t l t t t l t l
L
R

U F- tr) .a F- tr) \O $ t-- O\ ca :f, O OO O

O \O $ @ O\ t oO .i Ct.l $ ca O\ Ctl d
rf, o\ oo tr- ca a- oo c-r r t--. - \a) - ra)
O O - N ca ca .a O O O - - C.l c.l
A A /i A A
v v
^iv v
Av v
-i v -i v
v
-ig 9 -i -i -i -{ -i

tr) co \o -f, - sf ca \c) - $ $ co - or

-. .i .a c.l \o .a ca \o t*- - r-- o Lrr o
q v?.1 - R cg \ v? c.l - og \ n (\.l
9
ca (^'.1 C\l (\.l - - - .O ca c.} (\.l c\] C..l

I. l.' Ltat;,- :
f- s= n- F h ' F-n : : f F- n- s n
= E= - =
:;a-ZcGztiatio
 C H A R A C T E R I Z A T I O NO F A T O \ I I C I \ T E R A C T I O N S 311
7,4

The valuesof p(r") show a monotonic increasethrough both series,being a

minimum at the closed-shelllimit. In BeH and BH and in MgH to SiH the
valence charge density remaining on A is strongly polarized into the non-
field
bonded region of the- A atom (as a consequenceof the net negative
exerted on it by the negatively charged hydrogen atom) where it forms a
increase. For these molecules and for PH the ratio,
separateregion of charge
those characteristicof the two
l;,i1A, exhibits values intermediatebetween
fi-iti"g types of interactions. The atomic interactions in thesemoleculeshave
prop.rii., which bridge those of the ionic systems at the closed-shelllimit and
those of the covalent and polar systems of the shared interactions. In
molecules.such as BeH, BH, and SiH where the nodal surface inY2 p is nearly
coincident with the interatomic surface, the atomic interactions correspond
on the A atom which is dominated by
to a relatively hard core of density
interacting with a softer, more polariz-
contractions towards the A nucleus,
is
able region of chargeconcentration on the bonded side of the proton which
dominated by contractions of p towards the bond path.
The carbon-oxygen and carbon-sulphur bonds provide further examples
sur-
of intermediate'inieractions.The charge distributions and interatomic
C-O andlor C-S bonds are
faces for a number of molecules containing
populations and first moments for
shown in Fig. 6.4.Tab\e7.8liststhe atomic
the data for their (3, - 1) or bond
the atoms in these bonds, together with
the
critical points. The position of the C-O interatomic surface relative to
node in the VSCC bf tn. carbon atom is illustrated for the formaldehyde
molecule in Fig. 7.20. The positions of the bonded and non-bonded charge
concentrations in the VSCCs of the carbon and oxygen atoms are also
indicated,and one seesthat, while there are two bonded maxima for the C-O
of
bond, both lie within the basin of the oxygen atom becauseof the transfer
chargefrom C to O. This is a generalobservation for all the C-O bonds with
the result that the bond critical point liesjust within the inner shell of charge
of the
depletion. The same observation holds for the C-S bonds because
charge transfer from S to C.
Transfer of the order of one electronic charge or more does not imply the
presenceof an ionic interaction. The charge distributions and the properties
of tn. atoms in thesemoleculesdo not begin to approach those of closed-shell
point. The
systems,a fact reflected in the properties of p at the bond critical
flrst observation for this set of moleculesis that, while Y'p(r") > 0 for these
bonds, the value of p(r.) is also large in value, much larger than the vaiues
in turn, is a result of the
associatedwith the closed-shellor ionic limit. This,
which are much larger
large magnitude of the curvatures of p at r"' values
than those for the closed-shelllimit, but comparable to those for the shared
limit (Tables 7.4 and 7.5). These values reflect the large stressesthat are
present in the interatomic charge distribution, stresseswhich lead to a large
of a
accumulation of charge in the interatomic surfaceand to the formation
 a
- C)
CO t-_ qo tr_ \O O\ \O ..) bo
AO CO - a.l - rO (\.l Or
q c ' l q - Q Q Q v ? !
r\
- c . ] - c \ ] c \ - - -

c o - F - t - - . a l r ) - o \
O $ t f , c o N $ O \ c \
O . : t S O v ) N O \ O \
CO \O O\ \O a! .a ca OO
c ' i + c . i c d . j J i d

L n c A O \ F - O O C O O \ -
- c..l \o \o \o o\ o\ o x
O O . a O \ O O x t - t r )
- CO c.: tai - c.l ca r-l
J - ' J J O O O O
N

t t l l l t t l
tt
L
o\ ca o\ t-= @ oo o\ co
O\ (..l \O \O \O O\ O\ O
$ c a O \ O O r < t - c O
a..l CO .A rr - a.l ti (\.]
q - J J i O O o O
\ t t r t t t t l
, L

f .a \O cO *f, \O O c..l O'

\
t\
O , O \
q/') l--
O t n o O t . ) O \ -
V-r O. tf) a! tr.) \O
3
>' N .tO.- tr)ooCOt*-.a
x
A v v 9 9 v 9 9 9
O
c)
oO O\ \O S '- OO ca C'1 a . 2
O \ o c . t O - + $ . n
c - , O c O O \ a . l $ o O V Q 6
\ V t r ) ! f , $ N c . l c ' l N
i.)
- -i
v v
^'
v 9
^'
9 v 9 9

L
- z
q
(r- $ O. v^, a.l \O S O\
O\ CO CO a'.1 (.) ca Or Fr
oqE l)a
r a ) o \ t - - o o N $ $ $
g > i i 6 > J J J - i . t a
f - S
o R F
2 2
=
+ + + + | | | |
:i r

- ? a
F \ J a ' l o O O \ O \ O O o O c O
- _ - <€E
* A N - A A A _ A

U) c>-'i.'i ..iii-j X'o 9

9 | | l + * +
. a 3 . O . a V . O O t / - )
a N , t ?
O \O l-= $ O aO t-- t/') O i L
L \f, (\.] O\ f- - t-- cO -
al o.lcac!Nt/)\oc otr ) * . 6

J J J J i c > i i
q
Oa I i I t + + + +
v - C )

- s s
- ( ) +
. - , O - O r O r - O O O
V-) \O V-i a7j a-r tt- t-- t--
$ O.l O\ rO rO tar OC - . a ! Y

V) a\l .-, v) l-- F- -, oc \o

- - - n i - - - i O - i-c r
(, * E : -
r-
A\ s + + + + + | | | l ) J -

A - Q.)

S* f - - c O \ O \ O C C \ O S - a 9
a..l Oo o .a a..J \f, co .a co f .r
$ O . a c \ $ O r - - c c
g !
Ir- tr- t-- t-- @ co I-* co 9 ( l o
',J-
T\ = ^

l-- !..) o f- F< oo 6l crl

- x z
.a C \O ea t-- - t\ (-.l
( ^ . . l - * - O T O T O O O = = a

ni r^i ni e'i oi ni r..i ..i - - c )

d p "
v U ! z \ ) A V ) A A
\-/ \,/ \-/ ,.-l V W !, U F 2 0
o c X

Y u i
t v A o- a v { r U Z q l
( ) ' - x

J- ' : - Y U U q N tr ;q
a - 4 6 6 Q T , t + *
7.4 C H A R A C T E R I Z A T I O NO F A T O \ 1 I C I \ T E R A C T I O N S 313

H'^-r.
Hru-u

Ftc. 7.20. (top) Contour plots of Y2p in the formaldehyde molecule for the plane containing
the nuclei (left-handside)and the plane perpendicularto this and along the C-O axis (right-hand
side).The dashed (solid) lines denote regions of charge concentration (depletion) with contour
values given in the Appendix (Table A2). The Laplacian is also negative within the region
bounded by innermost solid contour encompassingthe C and O nuclei. The former diagram
shows the positions of the three bonded charge concentration (solid squares)in the VSCC of
carbon and the single bonded and two non-bonded concentrationsin the VSCC of oxygen.The
latter diagram shows the two (3, + 1) critical points in - Y2p (solid triangles)which determine
the sites of nucleophilic attack at carbon. The positions of the zero-flux surfaces are also
'reactive surface'for formaldehyde'
indicaied. (bottom) The zero envelope

concentration of chargeon the electronegativeside of the interatomic surface

and to a corresponding depletion of charge on the electropositiveside. The
large positive curvature, 1.., found for the C-O bond dominates the inter-
action and the properties of the bond. Such a bond is perturbed only slightly
by substitution, of a hydrogen in formaldehyde for example.When ,i, is low
in value, as for an ethylenic bond, atomic substitution can more readily lead
to a shift in the interatomic surface,corresponding to a transfef of charge
 I

314 CHEMICAL MODELSAND THE LAPLACIAN 7.4

across the surface and to a change in the characteristicsof the bond (Slee,
1986a\.
The atomic interactions in C-O and C-S are clearly intermediate in
character: p(r.) is large as there is a large accumulation of charge in the
interatomic surfaceresulting from the considerablecontractions of p towards
the bond path, but the interaction is dominated by the still larger positive
stressin p along the interaction line giving a tight distribution with a large
kinetic energy per electron. The ability to summarize the details of the
mechanics of an interaction in terms of the properties of p at r" and the
properties of the atoms whose nuclei are linked by the associatedinteraction
line precludes the need to assign oversimplifying labels to describe the
interaction. It is, however, important to reahze that the parameters which
summarizean interaction do exhibit a continuous spectrum of valueslinking
limiting situations that can be classified as shared and closed-shellinter-
actions.
It is also possible to observe behaviour transitional between shared and
closed-shellinteractions in the absenceof charge transfer. In the homonuc-
lear s er iesB r + F r, b o th p (r.) a n d l Y 2 p (r. )l i ncreaseto a maxi mum at N ,
and decreaseto minimum values atF r. Their behaviour parallelsthe binding
energiesof these molecules,as it does the occupation of the 12,, bonding
orbital (fully occupied at Nr) and of the 12, antibonding orbital (fully
occupied at F r). Occupation of the 12, orbital leads to an increasein the
localization of the charge on each atom and this has the expected conse-
quences on the properties of p at r", namely, a decreasein p(r") and an
increase in the positive curvature ,1. and, hence, in Y'p(rr). In F, the
contraction of p towards each nucleus dominates the interaction and the
VSCC of each F atom is localized in a near-sphericalatomic-like shell, as is
typical of the closed-shelllimit (Fig. 7.16).As demonstratedin Fig. 7.16.
however, the behaviour of the Laplacian at r. for F, is just opposite to that
found for Arr. The characteristicsof p at r" for F , are intermediate between
t hos e f or 02 a n d N e , (T a b l e 7 .4 );p (r" ):0.296 au, Y 2p(r" ):0.233 au.
At : Az - - 0.131au,A, : 1.694au, and the ratio G(r")lp$") : 0.86au. The
ratio of the perpendicular to parallel kinetic energy contributions at r" is
0.751, a value which, while less than unity as typical of closed-shellinter-
actions, is considerably greater than the value for Ner.
As discussedin detail in Section E7.1,both the charge density and the pair
density of the electrons of a F atom are very localized within its basin, as is
characteristicof a closed-shellinteraction. There is only a7 per cent exchange
of the electrons of a F atom in F, with the electrons on the neighbouring
atom. The contribution to the exchangeintegralsarising from the interatomic
electron exchange (a measure of the covalent character in the molecular
orbital model of electronic structure) is correspondingly small. In contrast.
the correspondingextentsof electronlocalization in C, and N, are 73 and 7E
7.5 BOUNDAND UNBOL'\D ST{TES 315

per cent, respectively.For these two molecules the interatomic surfaces

defining the atoms maximize the extent of tnteratomic exchange,as is
characteristicof a sharedinteraction.Thus. the strong potentialfield exerted
on the valenceelectronsin a fluorine atom. u'hile providing the major source
of binding in fluorides via charge transfer to the fluorine atom, is also the
causeof the weak binding found in the F, molecule.
There is no suggestionthat the atomic interactionin the Ft moleculeis at
the closed-shelllimit which requiresthat p(r") be small in valuein addition to
Y'p(r") beinggreaterthan zero.It is, however,demonstratedthat the binding
in this molecule is qualitatively different from that found in Nr (seeFig. 7.16)
and that the differencesare made quantitative by the properties of each
systemat the critical point in its charge density, formed as a consequenceof
the interaction of the two atoms.

7.5 Atomic interactionsin bound and unboundstates

We now consider the problem of further classifyinga given interaction as

belongingto a bound or unbound state of a system,i.e. bound or unbound
with respectto a separationof the systeminto the two fragmentslinked by
the correspondinginteractionline. This problem is first consideredfrom the
electrostaticpoint of view using the Hellmann-Feynman theorem to relate
the form of the chargedistribution to the forcesexertedon the nuclei. This is
followed by a discussionof the energeticsof an interaction in terms of the
virial theorem.This discussionof the energyis linked to the propertiesof the
charge density via the Laplacian of p using the local expressionof the virial
theorem, eqn (7.4).

7.5.1 The electrostatictheorem and chemicalbinding

The Hellmann-Feynman theorem determinesthe derivative of the energy E

with respectto a parameter s appearingin the Hamiltonian (Feynman 1939).
This theorem is derived in eqn (5.36) as a special case of the hypervirial
theorem. It was pointed out there via eqn (5.37)that, when s is taken to be a
nuclear position coordinate, one obtains the electrostatic theorem. This
theorem states that the force exerted on nucleus a, the quantity F o, is Z oe
times the electrostaticfield exertedat the nucleusby the other nuclei and by
the distribution of electronic charge.An explicit expressionfor F, is given in
eqn (6.60).Accepting the quantum mechanicalexpressionfor the distribution
of electronic charge as given by p(r), the theorem is a statement of classical
electrostaticsand therein lies both its appeal and usefulness.The theorem
also applies beyond the Born-Oppenheimer approximation, Garcia-Sucre
and Castells(1989).
316 CHEMICAL MODELSAND THE LAPLACIAN 7.5

Following Berlin (1951)the expressionfor the axial componentof the force

on either nucleus in a diatomic molecule (with referenceto the coordinate
system shown in Fig. 7.21) can be expressedas

F", : + l ( t ' = ' n " * t o = ' n u ) r , r , d-zz , Z p t R: 2 Fr. (7.10)

2J\ ri rfi /
When the elementof electroniccharge - ep\)dr is placed so that the first of
the two following conditions is fulfilled,

( 2""'!' > o binding

+ -2"991%)
P t (7.1 )
\-' rl rfi / |..0antibinding.
then its effectis to draw the two nuclei together, or to exert a binding force.
When the charge element is placed so as to satisfy the second condition, its
effectis to exert a larger force on one nucleusthan on the other and draw the
nuclei apart. Such a chargeelementexertsan antibinding force on the nuclei
which acts in concert with the nuclear force of repulsion. In this manner the
spaceof a diatomic systemcan be partitioned into a binding region,the open
region in Fig. 7.21(a),andantibindingregions,the hatchedareasin the same
figure (Berlin 1951).They are separatedby two surfacesof revolution defined
by the condition that the expressionin eqn (7.11)equalzero.Chargeelements
at a point on such a surfaceexert equal forces on both nuclei and tend to
neither increasenor decreasethe separationbetweenthe nuclei. Early studies
on molecular binding used density differencedistributions, Ap : p(molecule)
- p(atoms), in conjunction with the concepts of binding and antibinding
regions to study the relation betweenthe redistribution of electronic charge
which resultsfrom the formation of a molecule and chemical binding (Bader
et al. 1967a,b;Bader 1981).It was proposed that the density difference
distribution, or L,p map, may be taken as the pictorial representationof the
'bond
density'.
The forces acting on the nurrleiin a diatomic molecule and directed along
the internuclearaxis,expressedin terms of the internuclearseparationR, the
quantity F(R), is given by - dE(R)/dR,Fig. 7.21b.E(R) is the toral energyof
the molecule as determined by the fixed-nucleusHamiltonian given in eqn
(6.8).The changein the total energyE(R) as the internuclearseparationR is
decreased from R, to R, is given by
fRr

LE,, : l F(R)dR. (7.r2)

Jn'
Two necessaryconditions for a state to be bound are that the forces on the
nuclei vanish at some Rz = R. and that LEl- < 0. Thus, in addition to the
forcesvanishing,there must be net attractive forcesacting on the nuclei, that
is, forces drawing the nuclei together and decreasingthe separation R, over
7 .5 BOUNDAND UNBOU\D STATES 317

-3
-4
-5
0.5

(b)
Ftc.7.21. (a) The coordinate system used to define Berlin's binding and antibinding regions
and a display of these regions for the caseof a homonuclear diatomic molecule.(b) Plots oi the
total energy of a diatomic molecule E(R) versus the internuclear separation R the potential
energycurve lor motion of the nuclei.Superimposedon this curve is the curve for F(R) definedas
F(R) : - dE(R)/dR. From the E versus R curve, the forces on the nuclei are attractive when
F(R) < 0. The area lying betweenthe force curve and the R axis betweenthe limits R : r and
R : R. equals the binding energy of the molecule.The larger the attractive forceson the nuclei,
forces which attain their maximum value at the inflexion point R,, the larger is the binding
energy.

some range of internuclear separations R > R" for AEr2 < 0. The energy
increasesfor an increaseor decreasein R about R" for a bound state.These
statementsare illustrated in Fig. 7.21b for the formation of a bound state
from the separatedatoms.
 I

318 C H E M I C A L M O D E L SA N D T H E L A P L A C I A N 7.5

In order for the Hellmann-Feynman forces on the nuclei to be attractive

and so yield a bound state,the electronic charge must be distributed so that
its contribution to the force from the binding region exceedsboth the nuclear
force of repulsion and the contributions to the electronic force from charge
density in the antibinding regions.This can be accomplishedby accumulating
electronic charge in the binding region andfor by concentrating it along the
internuclearaxis within the binding region (cos0" : cos 0B: I in eqn (7.11)).
It is for this reasonthat the formation of a (3, - 1)criticalpoint in the charge
densityis a necessarycondition for the formation of a bound state,sincesuch
a critical point implies the existenceof an atomic interaction line-a line
linking the nuclei along which the chargedensityis a maximum with respect
to any neighbouring line. As emphasizedin Sections2.4.2 and 3.2.4,the
presence of such a line indicates that electronic charge is accumulated
between the nuclei, with p(r) reaching its maximum values along the bond
path. If theforces resultingfro* the accumulationof electroniccharge in the
bindingyregion are sfficient to exceedthe qntibindingforces ouer a range oJ
separationsto yield an equilibriumconfiguration,then the state is boundand the
atomic interactionline is calleda bond path.
The Laplacian distributions for the shared interactions illustrated in Figs
1.15 to 7.17 clearly indicate the predominance of the concentration of
electroniccharge over the binding region for such interactions.In all of these
moleculesY'p < 0 over the binding region for each interacting pair of nuclei.
Thus, in a sharedinteraction, electronicchargeis concentratedin the binding
region as a result of the contraction of p towards the bond path and this
concentration extends over the basins of both atoms. It is the shared
concentration of chargewhich exertsthe net attractive forceson the nuclei for
R ) R. and balancesthe forcesof repulsionat R : R". While Y'p < 0 in the
antibinding regions of nuclei as well, the amount of chargeaccumulatedthere
is lessthan that accumulatedin the binding region, and the net force exerted
on the nuclei is one of attraction.The atomic graph of the N atom in Nr, for
example, consistsof two maxima in - Y'p, one bonded, the other non-
bonded with magnitudes3.15 au and 2.52 au, respectively(valuesfrom an
M P 2 c alc ulat i o n ,s e eF i g .7 .1 6 ).T h e m a x i m a are l i nked by a torus of (2,0)
critical points in - V2p which encirclethe bond axis and have a magnitude of
0.74 au. The bonded maxima of the two nitrogen VSCCs are linked by a
(3, - 1)criticalpoint in - Y2p ofmagnitude2.6l au and the regionof charge
concentrationis contiguousover both atomic basins.
In a shared interaction it is the electronic charge concentrated in the
neighbourhood of the interatomic surfaceand extending over both atomic
basins which exerts the attractive forces on the nuclei. In the closed-shell
interactionsthe region in the neighbourhoodof the interatomiccritical point
is one of charge depletion and charge is instead concentrated separatelyin
each atomic basin in shell-likeregionssimilar to those found in an isolated
t.) BOUND AND UNBOUND STATES 319

atom or ion (Figs 7.15_ 7.17). Clearly, such an interaction can result in a
of
bound or unbound ,tut. depending upon the direction of the polarization
distribution. The directions of these atomic polarizations
each atomic charge
of an
can be evident inlhe Laplacian distributions. Consider, as an example
Laplacian distribution for the Ar, molecule' The atomic
unbound state, the
graph for an argon uio* in this molecule, like that for nitrogen in Nr'
- Y'p'Unlike nitrogen, however,
.oniirt, of two axial critical points in
where both critical points denote local charge concentrations, the critical
point in the binding iegion of Ar is a (3, + 1) critical point, denoting a local
minimum in the rptr.r. of charge concentration, while that in its antibinding
of Y2p
region is a (3, - 3i maximum, alocal chargeconcentration.The values
- 1.26 and - 1.34 au' respectively (values from an
at these two points are
critical points of rank two encircle the
MP2 calculation). Two tori of
- 2, the other of signature0' and link the
internuclear axis, one of signature
atomic
two axial critical points to complete the atomic graph' The two
central minimum for which Yt p :
graphs are themselueslinked by a-broad
of maximum
+ O.+Sau. Thus, for this unbound state, there is a region
depletionin electronic charge in the central binding region and the VSCC of
.urh uto- is polarizedinto its antibinding region (see Fig. 7'16). The major
Nt and unbound Art is the
differencein the Laplacian distributions of bound
presenceof a considerable concentration of charge in the central binding
VSCCs, and its replacement in
iegion of the former linking the two atomic
the latter molecule by a region of charge depletion'
Interactions at the closed-shelllimit can result in a bound state if accom-
panied by charge transfer, examples being ionic and hydrogen-bonded
systems.ine rnuige distributions shown in Fig. 87.2 fot diatomic molecules
fornied from Group I and Group VII elementsand their Laplacian distribu-
tions shown in Fig.7.15 demonstratethat such systemsapproximate the
given
model of two nearly sphericalchargedistributions. The atomic charges
in Tables 7.4 and E7.1 bear out the ionic model, as do the measures of the
degreeof localization of the pair density within each of the separate atomic
boundariesgiven in Table Ei .1. As a consequenceof Gauss'slaw, the binding
in a model ionic system A+B- is achieved by apolarization of both ions,
consideredto be initially spherical,in a direction counter to the direction of
charge transfer. It is shown that the charge distributions of the systems
consideredhere recover the dictates of this model'
Gauss'slaw of electrostaticsstatesthat the field exertedat a point external
to a sphericalcharge distribution is equal to that which would be obtained if
all thi charge were concentratedat its centre as an equivalent point charge.
Thus, the cationic nucleus is attracted by the net negative field exerted
on it by the anion, and the charge distribution of the cation must polarize
away fiom the anion to balance this net attractive force on its nucleus and
achieveelectrostaticequilibrium. The nucleus of the anion is repelled by the
 -11(l CHEMICAL MODELSAND THE LAPLACIAN 7.5

net positive field of the cation, and the charge distribution of the anion
must polarizetowards the cation to balancethe repulsiveforce exertedon its
nucleus.Thus. in an ionic system,both nuclei are bound by the forcesexerted
on them by the electronic charge localized on the anion and polarized
towards the cation. That is, the chargeaccumulationin the binding region is
localizedwithin the basin of the anionic nucleus(Baderand Henneker 1965).
The polarizatton of the VSCC of the Cl atom in LiCl and NaCl towards the
cation is evident in the displaysof Y2p for thesemoleculesin Fig.7.l5.
The contributions to the forces on the nuclei in such moleculesshow that
they achieveelectrostaticequilibrium by a polarization of the atomic charge
distributions in a direction counter to the transfer of approximately one
electronic charge.The electroniccontribution to the force on nucleusa in the
moleculeAB is given by the expression

F;:F;(A) +F;(B) (7.13)

where F;(A) is the force exerted on nucleus a by the electronic charge in its
own basin,the basin of atom A and F;(B)is the forceexertedon nucleusa by
the electronic charge in the basin of atom B. For the cation, Fj(A) is
antibinding and F;(B) is binding, while for the anion the corresponding
contributions are both binding as predictedby the model. Not only are the
directions of the forcesas predicted by the model, so are their magnitudes.In
-,
the idealizedlimit of A * B the electronic contributions to the forces on the
two nuclei may be expressedin terms of equivalent charges Q"(o) as

Fi : Q,lR')(Q"(A)+ 0,(B)),
(7.14)
F " B : Q p l R ' ) ( Q p @ )+ Q f ( A ) ) .

The charge Q,6) or Qo(B)is the charge which, when placed at the second
nuCleus,reproducesthe force on the cationic or anionic nucleus,respectively,
to account for the polarization of their atomic densities,antibinding for the
cation and binding for the anion. Ideally, Q,(A) : * 1 and ep(B) : - l.
The charges Q"(B) and Qf(A) should equal - (Zo * 1) and - (2,- 1),
respectively.For sphericalions, theselatter effectivechargesshould equal the
negative of the correspondingelectronicpopulations l{(B) and l{(A). The
charge equivalentsof the electroniccontributions to the forcesfor some ionic
systemsare given in Table 7.9. The chargeequivalentsof the forcesFi(B) and
F;(A) are indeed close in value to the actual populations of the B and A
atoms which are given in brackets. The contributions to the forces from
polarizations of the atomic densities also approach their limiting charge
equivalentsof * 1. The sums of the chargeequivalentsfor nucleusa and of
those for nucleusB must equal - Z tt and - Zo, respectively,for electrostatic
equilibrium. Forces on nuclei are notoriously difficult to calculatequantitat-
ively and only in the caseof LiF do the numbers in Table 7.9 sum to yield
7.5 BOUNDAND UNBOU\D STATES 321

Table 7.9
Charge o.f
equiualents* electronicc'rtntributiltlto nuclearforces

AB Q,6) Q,B) QB(B| Qp(A)

LiF 0. 7 7 - 9.76(9.941 - 0.99 - 2.04(2.06)

LiCI 0.60 - 1 1 .8 2
(1 1 .9 2 ) - 0.91 - 2.07(2.07)
KF 0. 7 3 - 9 .8 6(9 .9 3 ) - 0.71 - 11.84(18.07)

A2 A2

c2 - 0. 8 6 -s.13(6.00) F2 -0.42 -'8.52(e.00)

N2 - 1.01 - 5 .9 8(7 .0 0 ) He, + 0 .71 - t.e4(2.00)
o2 - 0. 7 8 - 7 .2 1 (8 .0 0 ) Ar, + 0 .33 - t 7. 6 r( 1 8 . 0 0 )

*0"(A) : F:(A) ' R'lZn, Q,@) : F;(B)' R'lZ, with correspondingdefinitions for
nuil.u. B. A negativesign means that force is attractive, i.e. directed towards the
other nucleus.

small net forces.The experimental bond length was employed in the calcu-
lation of the two remaining systems(Mcl-ean and Yoshimine 1967)and the
errors, while larger, are smaller than any of the individual contributions to
the forces.
Also included in Table 7.9 are chargeequivalentsof the forcesin homo-
nuclear diatomic moleculesfor bound and unbound systems.In the bound
systems,which result from shared interactions,both contributions to the
electronic force on a nucleus are binding. The descreeningof the second
nucleus and the force resulting from the atornic polarization are both a
maximum for the strongest bond with the greatest charge accumulation
betweenthe nuclei,Nr. Becauseof the contraction of the chargedensity away
from the interatomic surfaceand its separatelocalization in the basin of each
atom in the closed-shellinteractions,the descreeningof the secondnucleusis
much less for the unbound systems.The charge equivalents of Fi(B) ap-
proach - l{(B) in value for He, and Arr, as found for the ionic closed-shell
systems.The atomic contributions for the unbound systemsare, however,
strongly antibinding as a result of the polarrzation of each atomic density
away from the internuclear region. The F, molecule, which is intermediate
betweenthe two limits of interaction and possesses the smallestdissociation
energy of the bound systems,is seen to possessthe largest screeningof the
second nucleus and the smallest binding contribution from the atomic
polarizations for the bound systems.
The operator for the electronic force exerted on a nucleus is proportional
to llr!, the inverseof the squareof the distancefrom the nucleus(eqn(7.10)),
and is, therefore,dominated by the polarization of the chargedensity closeto
322 CHEMICAL MODELS AND THE LAPLACIAN 7.5

the nucleus.In the expressionfor an atomic first moment M(Q), however,the

chargedensityis weightedby the operator ro (eqn (6.45)),and this property is
dominated by the polarization of the valencedensity, particularly its diffuse
component.It is possible,therefore,for the atomic forceF;(A)and the atomic
first moment M(A) to be oppositely directed,as indeed occurs for the bound
ground statesof the homonucleardiatomics C, to Fr. As evidencedby the
data in Table 7.9, the atomic force in this seriesis binding and the inner
density is polarized towards the secondnucleus.The atomic first moment in
each of thesemoleculesis, however, directed away from the secondnucleus,
the magnitudesof this moment for Cr, Nr, Or, and F, being, respectively,
0. 172, 0. 590,0 .4 3 8 ,a n d 0 .1 9 7 a u . Bo th m o ments attai n thei r maxi mum
magnitude for the nitrogen molecule,the most tightly bound of thesesystems.
In an ionic molecule both the atomic force and atomic first moment of the
anion are directed towards the cation.
A study of the contributions of the individual molecular orbital densitiesto
the total force exerted on a nucleus in homonuclear diatomic moleculeswas
made by Bader et al. (1967a). This study shows that the atomic polarizations
binding the nuclei in thesemoleculesdo not arise primarily from the 1o, and
1o,, molecular orbitals whose densitiescorrespond to slightly polarized 1s-
like atomic distributions. Indeed, in C, these orbital densities exert small
antibinding forces on the nuclei. It is the density of the 2o, orbital in these
moleculeswhich is primarily responsiblefor the atomic forceswhich bind the
nuclei, while the densitiesof the 2o, and 3o" orbitals are responsiblefor the
oppositely directed atomic first moments.
The requirementsof binding, as viewed through the electrostatictheorem,
emphasizethe existenceof an atomic interaction line as a necessarycondition
for a state to be bound, whether it be at the shared or closed-shelllimit of
interaction. The differing properties associated with the distributions of
electronic charge at the shared and closed-shelllimits of interaction are
reflected in the differing mechanismsby which the forces on the nuclei are
balanced to achieve electrostaticequilibrium in the two cases.

7.5.2 The virial theorem and chemical bonding

The Laplacian of p determinesthe regionsof spacewhere electronicchargeis

concentrated.This function, coupled with the electrostatictheorem, enables
one to qualitatively characterizea chargedistribution as binding or antibind-
ing with respectto a given interaction in a molecule.These same regions of
space,those where Y'p < 0, are the regions in which the potential energy
makes its dominant contributions to the virial of a system. Thus, the
Laplacian of p provides a link betweenthe forcesacting on the nuclei and the
energeticsof an interaction, a link we now pursue.
7.5 BOUND AND UNBOU\D STATES 323

The molecularstatementof the virial theorem(Slater 1933)is given in eqn

(6.65).For a diatomic molecule,the virial of the external forcesacting on the
nuclei assumes the simple form - R(dE dR) and in this case the virial
theorem becomes
2T: -V- R(dEldR). (7.15)
:
?. is the average kinetic energy of the electrons, V (tS, *h"te t, the
: T * Z is
complete potenlial energy operator, is defined in eqn (6'9) and E
the eigenvalueof the fixed nucleusHamiltonian definedin eqn (6'8)'The force
is zero for the separated atoms and for the equilibrium separatioo R.'
Defining A7" and nV to be the differencesin theseenergiesfor the molecular
values minus the separated atom values, the virial theorem yields the
following well-known constraints on the energy changesaccompanying the
formation of a bound state,i.e. one for which LE < 0' Af > 0, AV < 0' and
a
2LT: - LV. Thus, the virial theorem ascribes molecular stability to
decreasein the potential energy,a change which necessarily dernands that the
kinetic energy increase.Since,next to the nuclear positions, the internuclear
region .^.rt, the lowest potential, such an interpretation is consistent with
the requirementof the electrostatictheorem of binding, that electroniccharge
be accumulatedin the binding region (Slater 1963).It is also consistentwith
the local statement of the virial theorem, eqn (7.4), which states that the
potential energy makes its dominant stabilizing contributions to the virial
and to the lowering of the total energy in those regions where electronic
charge is concentraied,that is where the Laplacian distribution is negative'
a
The preceding discussion has demonstrated that this corresponds to
binding region. The relation between the local
concentration of charge in the
p is
and integrated forms of tttt virial theorem provided by the Laplacian of
subject to further investigation below. However, it is clear at this point that
the electrostatic view that binding is obtained as a consequenceof the

Table 7.10
General behauiourof dTldR and dvldR in the formation of a bound
molecularstate

Range of value of R F(R) dr(RydR d z(R)/dR dr(RydR

I m >R>Rt <0 >0 s0* s0'

R:Ri <0 0 > 0 < 0
II Ri>R>R. <0 <0 > 0 < 0
R:R" 0 <0 > 0 < 0
III R(R" >0 <0 > 0 < 0
IV R<R" >0 <0 < 0 > 0

* <0for R ) Ri, t > 0 for R ) Ri.

324 CHEMICAL MODELSAND THE LAPLACIAN 7.5

eventual balancing of attractive forceson the nuclei and the view of the virial
theorem that bonding is a result of the lowering of the potential energyare, in
fact, ascribing a common origin to a bound state: that electronic charge be
concentrated in the internuclear or binding region (Slater 1963,pp. 39-40).
The study of the properties of the Laplacian distribution for bound systems
demonstratesthat this is where electronic charge is concentrated in bound
systemsto achieveelectrostaticequilibrium and simultaneouslybring about
the stabilizing decreasein the potential energy of the system.*The local and
integral virial theorems place restrictions on the accompanying changesin
the kinetic energy as well, and thesewill also be illustratedand discussed.
The virial and electrostatictheorems are connectedby the relation

dE : - r(R)dR (7.161
which shows,as demonstratedin Fig. 7.21,that the energy will decreaseover
a decreasingrange of R values only if the forces acting on the nuclei are
attractive-that is, electronic charge is accumulated in the binding region
The two theorems may be combined to yield constraints on the signs of
dTldR and dVldR for rangesof R valuesas definedby extremain the f'(R)
and E(R) curves(Bader 1981).The subtractionof Z from both sidesof eqn
(7.15)followed by differentiation with respectto R yields

d z ( R ) / d R: 2 F ( R )+ R d F ( R ) / d R ( 7. t 7|
with F(R)definedas in eqn (7.16).The addition of 2v to eachsideof eqn (7.15)
followed by differentiation with respect to R yields a corresponding ex-
pressionfor dVldR,

dz(R)idR: - 3F(R)- Rdr(R)/dR. ( 7 . 1)8

i It is perhaps important to mention that
one can find articles in the literature which dispute
these conclusions.Since the conclusionsare based upon physical laws, one is assuredthat the
criticismsemploy non-physicalarguments,as is indeed the case.For example,the argument thar
a build-up of electronicchargein the internuclearor'bond'region causesan increassratherthan
a decreasein the potential energy of a system is obtained from the analysisof a system which
does not obey the virial theorem.That is, the result is obtained for a very approximaie (unscaledr
wave function, one that predicts the total energy to decreaseas a resuft oi A f < 0 and LV > 0.
a result in direct violation of the virial theorem which it is used to criticize! Another
argument, which ascribes molecular stability to a decreasein kinetic energy, measures the
changes in energiesaccompanying the formation of a bound state to those-of an arbitrarilr
defined promoted state,which, sinceit is not a realizablequantum state,does not obey the virial
or electrostatictheorems.In addition, the promoted state is defined so as to yield the desired
result.The conclusion of the electrostatictheorem regarding the necessityof the accumulation oi
electronic charge in the binding region has been criticized on the basis of maps obtained b1
subtracting from a total molecular density distribution, an arbitrarily defined rei".ence densitl:.
which again requiresthat one invoke a physically unrealizablestate of the system.In briet if one
stays within the bounds of physics and avoids opinions as to what arbitrarily defined reference
states should be used to obtain a desired result, the theorems and the mutually consistenr
interpretations they provide are inviolate. If one does desire an opinion regarding these two
theorems, one can refer to Slater (1972) who regarded them u, it*o of the moit powerful
theorems applicable to moleculesand solids'.
 BOUND AND UNBOUND STATES 325
7 .5

The points at which F(R) equalszero and whereit attainsits minimum value
its
enableone to definerangesfor valuesof R in which the signsof F(R) and
in the
derivativesare uniquely determined(Table 7.10).The inflexion point
to the point at which f' (R) is a minimum, a point
curve for E(R).orr.rponds
labelled Ri. At this point the magnitude of the binding forcesexerted on the
nuclei by the charge density is a maximum'
of
In range I (Tabie 7.10),where oo > R > Ri, F(R) and dF(R)/dR are
opposite sign and the signs of the derivatives of Z and V are not uniquely
determined.However, for large valuesof R, thesesignsshould be determined
by the sign of just dF(R)/dR, since this term is multiplied by R in the
to
e*pressions for the derivatives. One thus predicts the kinetic energy
initlatty decrease,dTldR > 0, and the potential energy to increase for the
two
approich of two atoms. This is the observedresult for the approach of
neutral atoms in S states, and is, in general,true for the approach of neutral
molecules for these changesin T and V are characteristic of the so-called
London or van der Waals' forces.These forces appear as the leading term
- 6C6lR1,in the perturbation expansionfor long-rangeinteractions'
Feynman,in his original paper on the electrostatictheorem(1939),sugges-
ted that the long-rangeforce on each nucleusshould result from its attraction
to the centroid of its own chargedistribution.This view is borne out by the
diagram of the changein the atomic chargedensitiesshown in Fig. 1.22 for
thelpproach of two ground-statehydrogen atoms at a separationof 8 au'

Z;==\

rc)D
(,;i===-'''
./,/ /...-\'.\
ii'r'r','-:\;i,,
'li1,'i,;-,;,,
ttliil' ,,-)i'i
,tl','i.:V
)\\.'i::'-'l i::-t:it2
\\i------j;; >----:-?:tj

H, (Das and
Ftc. j.22. Density clifferencemap calculatedfrom a correlated wave function for
each in its
Wahl 1966)for a separationof S au. ttre density of the unperturbed hydrogen atoms,
2S ground state, is subtracted from the molecular density using the same nuclear position
and
coordinates.The overlap of the atomic densitiesat this internuclear separation is negligible
by the
the differencemap illusirates the polarization of each of the atomic densitiesas caused
(
correlative interaition of the two electrons. The solid and dashed contours, increase * )
( - r e s p e c t i v e l yf,r o m t h e z e r o c o n t o u r i n t h e o r d e r * 2 x l 0 - " * 4 x l0-"'
or decrease ),
+ 8 x i 0 - " w i t h n b e g i n n i n ga t 5 a n d d e c r e a s i n gi n s t e p so f u n i t y .
-126 CHEMICAL MODELS AND THE LAPLACIAN 7.5

This diagram is obtained from a highly correlated wave function by sub-

tracting the unperturbed atomic densities from the correlated molecular
density. The transfer of electronic charge from the antibinding regions to the
binding region, which results from the polarization of each atomic density
towards the other, is a consequenceof the correlative interactions between
the electrons on the two centres.One can do no better at explaining this
picture than to give Feynman's prediction of it by quoting his 1939 paper.

The Schrodinger perturbationtheoryfor two interactingatomsat a separationR,

large comparedto the radii of the atoms, leads to the result that the charge
distributionof eachis distortedfrom centralsymmetry,a dipole momentof order
1/Ri beinginducedin eachatom.The negativechargedistributionof eachatom has
its centerof gravitymovedslightlytowardthe other.It is not the interactionof these
dipoleswhichleadsto van der Waals'force,but ratherthe attractionof eachnucleus
for the distortedchargedistributionof its own electronsthat givethe attractivelf R1
force.
Thus the long-rangeforcesare boot-strap forces---cachnucleusis pulled by
its own distorted charge distribution. The form of the charge rearrangement
found to result from this correlative interaction between the electrons and
shown in Fig. 7.22 provides a ready explanation for the initial decreasein Z
and increasein V. Chargedensity is removed from the region of each nucleus
where G(r), the kinetic energydensity,has its maximum value and is placed in
a diffuse distribution in the binding region. This accounts for AI < 0. The
charge density is also tightly bound in the region of the nucleus and its
removal from this region to a region further removed from the nucleusraises
the potential energy of the system.Thus, an initial attractive interaction can
result from a decreasein T and an increasein V, but never for the energy
change for a bound state at electrostatic equilibrium, not even a van der
Waals' minimum, where one must have LT > 0 and LV < 0.
At R : Ri, where the binding force attains its largest magnitude, the signs
of dTldR( < 0) and d VldR( > 0) are uniquely determined.Thus, the decrease
in energyfor R ) R, can result from a decreasein ?'and in spite of an increase
rn V. However, for R ( R,, up to R : R., T and Z must begin to increase
and decrease, respectively. In Hr, for example, as R decreasesfrom
i n f i n i t y , i n i t i a l l y d v l d R < 0 a n d L V > 0 . H o w e v e r ,a t R : R i : 2 a u ,
LV: - 0.0625au and, hence,dvldR changessign beforeR: R,. Thus for
Hr, the presentanalysisshows that the chargerearrangementsfor R ( 2 au
lead to a decreasein the potential energy of the systemand to an increasein
its kinetic energy.The contributionsfrom F(R)dR to the energyin the range
R, ) R > R" must dominate over thosein the range oo > R > R,, sincethe
total changesin T and Z determiningAE(R") must be such that Ar > 0 and
LV <0.
T he s ignso f d fd R a n d d v ttd R a t R : R . w i l l persi stfor some ^. ^.,
since, initially, F(R) is small and dF/'dR is large. However, F(R) rapidly
1.5 BOUND AND U\BOL-\D STATES 327

increasesin this range (tendingto infinitl t rr hile dF dR remainsfinite' Thus,

eventually,for some R ( R", the signsof d r dR and d ,' dR will change,and
v and ?"will increaseand decrease.respectirell'.uith R.
The ideasunderlying the equating of E. the total energyof a molecule,to
the sum of electronic and nuclear energies E. and En were developed in
Section6.3. The electronicenergy E. is giren b1'
E,:T+Y (7.19)

where {, the electronic potential energy.is the virial of the Ehrenfestforces

exerted on the electrons
- [ r ' Y ' 7 ( r ) d r: I (7.20)
potential
where F is the quantum stresstensor, eqn (6.13).The electronic
in E : T + V by
energy is relatedio the total potential energy V appearing
the expression
v : 4n + V",+ V,, - t - I,x" 'YoE. (7.21)

Sincethe nuclei are stationary, En is purely potential and is equal to the virial
of the electrostaticforces acting on the nuclei
En: - .Y (7.22)
I"x" oE.

The virial theorem may be stated in terms of the electronic contribution

toEas
T- -Er: -+N (7.23)

or, in terms of the total energy, as

T - - E - I " X " ' Y , E- - + ( V + I " X ' ' V " E ) . (7'24)

For an isolated atom, or for a molecule in an equilibrium configuration one

h a s ?" : - E " : - E.l n g e n e ra l ,h o w e v e r,fo r a mol ecul e

-E,t -E or T- -E"1 -E (7'25)

T-

d e pendingon whe th e rth e re a re re p u l s i v e(Z > l E l ) or attracti ve(f < l E l )

forces acting on the nuclei.
The Laplacian of the chargedensity as it appearsin the local expressionof
the virial theorem, eqn (7.4),compares twice the kinetic energy density G(r)
not with contributioni to the potential V,but with f ,the electronicpotential
energy.This is an important distinction from the point of view of whether or
not a systemis bound. Consider an interaction whereY2p is predominantly
negativeover the binding region and net forcesof attraction act on the nuclei.
In this case the local contribution to the virial of the Hellmann-Feynman
forces exerted on the electrons,which, as explained following eqn (6.60),is
 328 CHEMICAL MODELS AND THE LAPLACIAN 7.5

given by
- .y
I,X".Fo. : h^ + I,X" oE, (7.26)
is greater than 2".. Thus, becauseof the concentration of charge in the
binding region, the total potential energy Z is more negative than t (eqn
( 7 . 2 1 ) ) , l v l > 2 T , a n d l E l > r . T h u s ,E d e c r e a s ewsi t h a d e c r e a sien R , u n t i l
\n rncreasesto a value such that equilibrium is attained and v : f . It is,
howeuer,the concentrationof negatiuecharge in the binding region resulting
from contractionsin p perpendicularto the bond path which causesthe local
contributions to Y to exceed the contributionsfro^ the kinetic energyfor
R t R", thereby leading to the creation of attractiueforces and, hence,to a
bound, shared interaction.In a system where Y2p is predominantly positive
over a binding region and net forcesof repulsion act on the nuclei, the kinetic
energy contributions to the virial dominate over the internuclear region as a
result of the contraction of the electronic charge density towards the nuclei
and its consequent depletion in this region. In this situation Y2p is most
negative in the antibinding regions, v is less negative than r, zT> lvl,
repulsive forces are exerted on the nuclei, and the system is unbound. In an
ionic bond the transfer of chargeto the anion for R ) R. and the polarization
of its charge towards the cation yield net attractive contributions to t; V
exceeds{ in magnitude (eqn (7.21)l becausethe forces are attractive and
2T< tVt.
A comparison of the Laplacian distribution for a shared interaction with
its shell-like structure in the separatedatoms shows that the major changein
this function on the formation of such an interaction is the creation of a
region over which Y'p < 0 extending over the basins of both atoms. This
concentration of charge results from a contraction of p perpendicular to the
interaction line and it leads to a local lowering of the potential energy.The
magnitude of this lowering in the potential energy is greater than the
contributions to the kinetic energy from the same region. In particular, the
component of the kinetic energy parallel to the bond path is decreasedin
value as a result of the decreasesin the corresponding gradients and
curvaturesof p.
The differing propertiesof the chargedistributions for bound and unbound
statesare made very clear by comparing the Laplacian distributions of p for
Hr and Her (Fig. 7.15).In the former, the atomic interactionis dominated by
the contractions in p perpendicular to the bond path while in the latter it is
dominated by the contractions in p towards each of the nuclei.
It has been suggestedthat Hj and H, are bound, not becauseof the
accumulation of electronic charge in the 'bond' region, but becauseof the
contraction of the charge towards the nuclei. In actual fact, the major
differencebetween the Laplacian distribution for bound H, and unbound
He, is that the former but not the latter possesses accumulated,sharedcharge
7.5 BOUND AND UNBOUNDSTATES 329

in the binding region,while the distribution of the latter but not the former is
dominated by the contractionsof the chargedensitl' towards the nuclei.It is
important to read all the information that is containedin thesemaps of the
Laplacian distribution. Becauseof the appearanceof V2p(r) in the local
expressionof the virial theorem,thesemaps also revealthe role of the local
contributionsof the kinetic energyto the total energyof the system.As noted
in Section 7.4,2,the kinetic energy components parallel and perpendicularto
the interactionline mirror the behaviour of the correspondinggradientsand
curvatures of p. For Hr, Ytp < 0 in the binding region becauseof the
contraction of charge towards the interaction line, and the parallel gradient
and curvature of p arc softened.Consequently,the kinetic energyper electron
is small in value, a result in accord with the local virial theorem. Thus for H,
at R., the ratio of the two perpendicularcomponentsto the parallel compon-
ent of the averagekinetic energy,the ratio TtlTtt, is greater than 2.0 and
equals2.80 (Baderand Preston 1969).For Her,Y'p > 0 in the region of the
interatomic surfaceand the interaction is dominated by the contraction of p
towarrjs each nucleus.The parallel gradient and curvature of p are large,as is
the parallel component of the kinetic energywhich exceedsthe perpendicular
components.The kinetic energy per electron is also absolutely large. These
observations are again in accord with the local statement of the virial
theorem which statesthat the kinetic energy is dominant in regions where
Yrp > 0. F or He, at R : 2 .0 a u , th e ra ti o T rl T l i s l essthan tw o and equal s
1.81(Bader and Preston 1969).
The binding of the nuclei in ring or cage structures is made particularll'
clearby the Laplaciandistribution of the chargedensity.In cyclopropanethe
magnitudes of the two negative curvatures of p at a carbon-carbon bond
critical point are unequal (Table 7.5).The principal axis of the curvature of
smallestmagnitude is directed at the ring or (3, + 1) critical point at the
centre of the ring surface.Thus, a carbon--carbonbond of cyclopropanehas a
noticeable ellipticity (e : 0.490) and electronic charge is preferentially con-
centrated in the ring surface (Fig. 7.I7). The Laplacian of p shows a
concentration of valencecharge density over the surfaceof the ring with the
exception of a small region in the neighbourhood of the (3, + 1) critical
point. (The principal axes of the two positive curvatures of a ring critical
point lie in and define the ring surface.)It is this concentration of electronic
charge in the interior of the ring structure which binds the ring nuclei. The
large extent of this concentration of charge is reflected in a noticeable
curvature of the bond paths away from the geometricalperimeter of the ring
(Fi g. 7. 17) .
Electron-deficient molecules, such as the boranes, form ring and cage
structuresto gain maximum stability from a minimum amount of electronic
glue by a very pronounced delocalization of charge over ring surfaces.The
2
molecular graph for BuHu shows each boron atom to be linked to one
3 3 0 C H E M I C A L M O D E L S A N D T H E L A P L A C I A N T . 5

cage
hydrogen atom and four other boron atoms to form an octahedral
densitl
structure bounded by eight ring surfaces.The Laplacian of the charge
Fig' 7'23' The
for one such three-memberedboron ring surfaceis shown in
as illustra-
boron-boron bonds exhibit extremeellipticitiesequal to 3.58and,

the two nuclei and the

Ftc. j.23. Contour maps of the Laplacian of p for the plane containing
nuclei in the octahedral
two bridging protons in BrHu and ior a face containing three boron
.of.."f.T."*2. The boni paths and bond critical points are also shown. The inner contour
a region which is
around each boron nucleusencirclesthe inner region ofcharge concentration,
inturnencompassedbytheinnershellofchargedepletion.
 BOUND AND UNBOU\D STATES 331
7.5

ted in Fig. 7.23, electronic charge is concentratedalmost evenly over the

whole of the ring surface.The contractionof p along the axis perpendicularto
the ring surfaceat the ring critical point is also large and it dominatesthe two
positivecurvaturesof p in the ring surfaceso that V2p is negativeeven at the
iing critical point. The value of p at a ring critical point is 0.120au, only
slightly lessthan its value of 0.127 au at a boron-boron bond critical point.
The same observationshold true for four-membered rings involving one or
two bridging hydrogen atoms in moleculessuch as BrHu or BrHn. Figure
7.23illustratesthe Laplacian of p in the plane of the bridging hydrogen nuclei
of BrHu. In this four-memberedring the bond paths are inwardly curved and
electronic charge is again delocalized over the valence region of the ring
surface.The ellipticities of the B-H bridging bonds equal 0.561 and their
valuesof p(r.) :0.119au exceedthe value of p at the ring critical point by
only 0.013au. Thus, in three- and four-membered rings of electron-deficient
structures, electronic charge is preferentially delocalized over and concen-
trated in the ring surface where it servesto bind all nuclei of the ring. The
description of such moleculesas globally delocalizedis particularly apt (King
and Rouvray 1977).

7 .5 .3 A s um m ar Y

The electrostaticand virial theorems, when coupled through the use of the
Laplacian of the electronic charge distribution, provide an understanding
to
and classificationof the mechanicsunderlying atomic interactions leading
as
both bound and repulsive states. The regions of charge concentration
the Lapiacian, Y'p < 0, can be used in conjunction with the
defined by
electrostatictheorem to determine whether the forces on the nuclei will be
attractive or repulsive.The electronicchargein thesesameregionsmakes the
dominant stabilizing contributions to the potential and total energiesof the
system and, in this way, the Laplacian provides a link between force and
energy. If the charge is concentrated in the binding region, the forces are
attractive, the magnitude of the total potential energy Z exceedsthat of {,
and the virial of thi nuclear forcescontributes to the binding of the system.If
chargeis concentratedin the antibinding regions,the forcesare repulsive,the
,nugiittrd e of f exceeds that of V and the virial of the nuclear forces
destabilizesthe system.As explainedin Section6.4.4,thisis what occurswhen
ethane in its minimum-energy staggered conformation undergoes a rigid
rotation into its'frozen'eclipsedconformation.Charge concentrationscon-
tribute to a lowering of E": T + {, the electronic energy of the system'
wherever they occur. Even non-bonded charge concentrationscontribute to
the lowering of the electronic energy. However, the effect of local charge
concentrations on the total energy E is determined by whether they exert
332 CHEMICAL MODELSAND THE LAPLACIAN E7.l

binding or antibinding forces on the nuclei, leading to a stabihzation of the

total energy in the former caseand to its destabilization in the latter.
The Laplacian maps, again through the local virial theorem, also provide
information regarding the local contributions to the kinetic energy.The low
value of the gradient in directions parallel to the axis leads to small
contributions to the kinetic energy density G(r) and hence to T, while the
contraction of p(r) perpendicular to the axis gives large contributions. The
charge density in the binding region of a stable molecule is, therefore,
distributed in such a way as to keep the accompanying increasein kinetic
energy to a minimum. The otherwiselarge increasein G(r) resulting from the
contraction of p(r) perpendicular to the bond, the effect responsiblefor the
reduction in the potential energy,is partially offset by the smaller contribu-
tions to G(r) and henceto ?'as a result of the softeningof the gradient of p(r)
along the bond path (Bader and Preston 1969).
In addition to these local properties of p and its Laplacian distribution
which serve to provide a broad basis for the classificationof atomic inter-
actions,the theory of atoms in moleculesmakes possiblethe determination of
the average values of all the mechanical properties of an interacting atom.
The use of an atomic population and of the fluctuation in this value (which
determinesthe degreeto which the electronsare localized within the bound-
aries of the atom) have also been usedin the classificationand understanding
of atomic interactions. One can also determine the changesin the energies
undergone by the individual atoms in each interaction, as illustrated for the
origin of the barrier to rotation in CrHu. Dependingupon the nature of the
problem, the theory enables one to obtain as detailed a view of an atomic
interaction as one desires,stressingany particular set of properties as one
will, all in terms of the mechanical properties of the atoms involved.

E7.l The pair density and the localization of electrons

This section details the mathematics and the physics which enable one to
determinethe extent to which some averagenumber of electronsare localized
to a portion of the total spaceavailableto them. The localizationof electronic
charge is determined by the quantum mechanical distribution function
{*td*rd"r...d", ( e q n( 1 . 1 ) )C
. o n s i d e rt h e p a r t i t i o n i n go f a t o t a l s y s t e m
containing N electronsinto two mutually exclusiveregions Q and O'. The
probability of the event that n electrons will be found in Q when the
remaining (l/ - n) are confined to O' is

: o#, 0,.,
P"(o) J I,d'"I dxn*t

o-, ;'r'rI (x1,x2, x,ry) (E7.1)

J
 a l l

E7.l PAIRDENSIT\ JJJ

(N),the many-particleprobability densitl'.is a diagonalelementof the

wheref
spinlessl{-particle density matrix,

f ( t , ( * r , x 2 , . . x r o ): I r p i n . / * ( x r , X 2 .. . x { ) t l t( x ' r , x ' 2., . . { v ) l * , : * l

where x, denotes a product of space and spin variables,x;: rioi. The

binomial coefficientin eqn (E7.1)accounts for the indistinguishability of the
electrons,so that it matters not which n electronsare consideredto be in O
when the remaining (l{ - n) are in O'. Since the wave function is assumedto
be normalized, one has
l"P,1o;: t' (F.7.2)

The averagenumber of particles in a given region Q is

lUtOl: l^nP^(Ql - I p(r)dr, (F.7.3)

J"
a result which, as indicated in eqn (E7.3),must be equivalent to the integra-
tion of the charge density p(r) over the same region of space.If a singleevent
has a dominant probability, it is referredto as the leading event. If P,(O) is a
leading event and if the probability distribution of the other events is
symmetrical about P,(Q), then ,V(Q) will equal n. One must in general,
however, distinguish between the number of electrons localized in a given
region to yield a leading event and the average occupation number of the
region.
Daudel and co-workers(Daudel 1953,1968;Daudel et al. 1955;Aslangul
et al.1972)werethe first to provide a definition of the localizationof electrons
basedsolely upon information contained in the statefunction. They proposed
that information theory (Shannon1948)beusedto definethe'best'divisionof
'loges'. According to this theory our knowledge
a system into regions or
about a given systemis a maximum when the'missing information function'
I(P^,Q) is minimized where
I(P ' ,Q) : -I" p " (Q)l n P" (Q ). (87.4)

If the probability of one particular event (n particles in Q) approaches a

certaintv.
P,(Q) -- 1 and -r 0. (E7.s)
L^*.P^(Q)
In this instanceour knowledge of the systemwould be maximal as we would
know with certainty that n electronsare in O and (A/ - n) in Q'. Correspond-
ingly, in this case 1(P, O) approacheszero. At the other extreme, when all
eventsare equally probable, our knowledge of the specificstate of the system
is minimal and the missing information function is maximized.
'best loges' as the ones which minimize /(P", O),
Using the definition of the
Daudel et al. (1974)appliedthis theory to a number of diatomic and triatomic
moleculescontainingup to six electrons,namely,LiH, BeH+, BH, and BeHt.
334 CHEMICAL MODELS AND THE LAPLACIAN E 7.l

It was indeed possiblein thesemoleculesto partition total spaceinto regions

so that the missing information function was simultaneouslyminimized for
each region. The quantum probabilities were in each case dominated b1
Pr(O) and the most probable partitioning was found to be the one in which
pairs of electrons were localized in well-definedregions of space.
Figure E7.1 illustrates the searchfor the best sphericalloge in a two-loge
partitioning of the six-electronBH(X t> * ) molecule.Plotted in this figure are
the valuesof P"(O)and /(P", Q), where the volume O refersto a sphericalloge
centred on the boron nucleusof variable radius r. Aside from the trivial peaks
in Po(Q)for r : 0 and in Pu(A) for r : co, the singlemost probableevent is
obtained for a loge of radius 0.7 au which simultaneouslymaximizesPr(Q)
and minimizes I(P^, Q). The satelliteeventsP, (O) and Pr(Q) have small and
almost equal probabilities and thus N1O),the averagenumber of electronsin
this most probable core loge, is two, as determined by the leading evenr

1.0

0.9

0.8

+ 0'6 A
L
--
\4 u.c e i.o
z
P^/O\
z
E
c U.6

G
S 0 .6

1.0 2.0 3.0 4.0 5.0 0.0 1.0 2.0 3.0 4.0 5.0
r--- > r ____+

Ftc. E7.1. The variation in the probabilities P"(Q), the missing information function 1(P,Or
and the fluctuation in lV1O.1,
ntlg. Ot with the radius r (in au) of a spherical loge centred on the
boron nucleusin BH (Xt t * ). The positions of the maxima in the P"(O)valuesfor each value of n
are also indicated on the plot for A(N. O).
Et.r PAIRDE\SIT\ 335

pr(O). In this casethere is an 85 per cent probabilitl of finding two electrons

in a sphereof radius 0.7 au. Note from Fig. E7.l that maximizing a single
event,P, (A) for example,doesnot necessarilllead to a minimum in 1(P', O)'
Wh a t is im por t ant to n o te i s th a t 1 (P" .Q ) c a n be extremi zedand that i t
exhibitsonly the single(non-trivial)minimum: thus there is but a single'best'
core loge. The readeris referredto the original paper (Daudel et al. l974a,b)
for details regardingthe multi-loge partitioning'
The study by Daudel et al. (1974a,b)also showedthat the fluctuation in the
averagenumber of particlesin a region paralleledcloselythe behaviour of the
missing information function (seeFig. 87.1, for example).The fluctuation in
the averagenumber of electrons for a region O is defined as

A(l{, Q): ^/2tol- tlutotl' :L^n'P"(o) - ll,nP^(o)l'. (E7'6)

Clearly, the fluctuation in the averagepopulation of a region Q (denotedhere

UV N(O)) also approacheszero when a single event dominates the distribu-
tion. A difficulty with using quantum probabilities to determine the extent of
spatial localization is that their evaluation requires the use of the fuil -n{th-
ord.t density matrix. Becauseof this, the calculation of P"(O)rapidly becomes
prohibitive with increasing lf. The fluctuation A(l{, O) however, can be
ixpressed entirely in terms of the diagonal elementsof just the second-order
density matrix (Bader and Stephens1974).
A diagonal element of the second-order density matrix ot pair density,
p(rr,rr), as it is referredto here,is definedas
p(rr.,rz): (l{(l{ - \12)J do, ! do,J d*. J d**,lt*'lt (87'7)
where, as before, x, denotes a product of spin and space coordinates and
d*, : do,dr,. The expressionfor l{2 (Q) can be expressedin terms of p(rt, rr)
AS

,v'tol-- -r l d rt
J' t d r r p ( rt , ,rr*.l"d r , p ( r t )
where the charge density p(rr), or numberdensity as it is referred to here, is
(E7.8)

normalized to N electrons.Thus, the fluctuation around N(O) for a region Ct

can be written
r | -.,) -tN(')l'.
A(l{, O):2 | dr, I dt, P(rt,rr) + l{(
(E7.e)
Jn Jn
There is a great conceptual advantagein abandoning the missing informa-
tion 1(P", O) in favour of the fluctuation A(N, O) to determine the localiza-
tion of electrons.Lennard-Jones(1952)pointed out that the extent to which a
set of indistinguishable particles is spatially localized is determined by the
system'spair density,the same distribution function which determinesthe
336 C H E M I C A L M O D E L SA N D T H E L A P L A C I A N E7.1

fluctuation in l{. For a systemof electrons,the relevant properties of the pair

density are a consequenceof the so-called Fermi correlation which results
from the antisymmetrization of the state function as required by the Pauli
principle (Bader and Stephens1975).
The relationship betweenthe Fermi correlation and the spatial localization
of electronsis developedby defining, as McWeeny (1960)does,a correlation
functionf (rr,rr)in an expressionwhich determinesthe extent to which the
pair density deviatesfrom a simple product of number densities,

p ( rt , r ) : i p F r )p ( rr ) [ 1 + f ( rr , r r ) 1 . (E7.10)
The pair density can never equal the product of number densitiesas given by
the first term on the right-hand side of eqn (E7.10) since the product
LpFJp$r) integratesto ].N2 or |.n{ pairs in excessof the total number of
distinct pairs +l{(lf - 1)that can be formed from l{ particlesand the number
obtained by integration of the pair density (eqn (E7.7)).Hence,the correlation
factor/(r r,rz) must correct for the improper counting of pairs as given by a
product of number densities. Because of this requirement one finds that
(McWeeny 1960)
- 1 (E7.11)
J d . ,p ( r ) f ( r t , t z ) :
when eqn (E7.10)isintegratedover the coordinater, (the left-handsideof eqn
(E7.10)yielding ((l{ - \12)p(rr) in this step)for any fixed value of r,. This
result correspondsto the removal of one electron and the total effect of the
correlation over all spaceis to remove the jl{ self-pairsimproperly counted
by the product of number densities,that is,

1 I d t ' J d t r p ( r) p ( r ) f ( r t , t z ) : - N1 2 . (87.r2)
Thus, p\)f $t,rr) is a distribution function that, for a fixed value of rr.
describesthe decreasein the density of other electrons by the equivalent of
one unit of charge.Its effectis, therefore,to createa hole, a correlation hole, in
the pair distribution function.
In this work we are particularly concerned with the spatial extent of the
effectsof this self-pairingcorrelation on the motions of the electronsand with
whether or not the net effect of this correlation for any one particle, the
correlation hole, may be localized to one particular region of space.
In an electronicsystemit is essentialto distinguish betweenthe correlation
of electronswith identical spin, Fermi correlation, from that of electronswith
opposite spins, Coulomb correlation. The correlation term f (rr,rr) as de-
fined in eqn (E7.10)will measureboth types of correlation (McWeeny 1960).
However, the limiting value of the correlation hole as expressedin eqn (E7.1I )
arisesonly from the correlation betweenelectronsof the samespin, the Fermi
hole. That part of f (rr,r2) that refersto the correlation betweenelectronsof
opposite spin contributes zero when integrated over all space. If the co-
87.1 PAIR DENSIT\

o\ci ri i
o rd i n at esof ar r e- s p i ne l e c tro na re h e l d fi x e d d u ring the i ntegratl trn
o n l y th e re m o v a l o f r-spi n densi tl contnbute: t"
sp a c ein eqn ( E 7. li;, th .n
- 1 as a consequence of the orthogonalitr of the spin
the limiting value of
form tu'o
functions. Conceptually.it arisesbecausethe r and B electrons
can occur only within a set of identical particles.
distinct sets,and self-pairing
density matrix, obtained from a single
The pair density or r..o.rd-order
of orthogonal spin functions @, is given in
determinantal function composed
expression for the pair density with that given
eqn (E1.4).Comparison of that
a spin at rt
ineqn (E7.10)yieldsfor the Fermi hole for a referenceelectronof

1i F) Q fF) 0; F) l L i 0 f $ ' )d '( " ) '

h o( ry ,rz ): - L i L ] 0 , F , Q
(E7.13)
spin
Only orbitalsof a spincontributeasintegration(or summation)overthe
from p-spin orbitalsto The
vanish'
coordinateswill .uur. any contributions
summationin the denominatorof eqn (E7.13)is the a-spindensity,

p'(:r): I: dl(r')d,(t'). (87'14)

Corresponding expressions obtainfor the electronsof p spin.Integrationof

-
ho(rr,rr)overr, for fixedr, yields 1 corresponding to the removalof one
electronicchargeof a spin.One alsonotesthat
h o ( : r r , r z ): -po(;rr) when 12:11, ( E 71 s )
position of
correspondingto the completeremoval of a-spindensityfrom the
of
the referenceelectron.This result is a direct consequence the antis.vmmetr\'
principie
imposed on Y and hence on p(rt,rr) by the Pauli exclusion
(McWeeny 1960). Because of this antisymmetry the exchange of space
the sign of the
coordinates of two electrons of identical spin must change
corresponding Pair densitY.Thus,
- P o ( \ z ,rr) , (E7.16)
P ' ( rr , r z ) :
space co-
a result which demands that the pair density vanish when the oo(rr,rr)
ordinatesof two electronsof identical spin are coincident,or that f
: - 1 for rt:12. This limiting extremein correlation is not obtained for
of
electrons of opposite spin and there is a finite probability of electrons
'colliding' in real space. The state function must, however'
opposite spin
just as it does
.*ttiUit u .urp for the coalescenceof two electroniccoordinates
and
for the coalescenceof an electron with a nucleus (Kato 1957;Roothaan
Weiss 1960).
given in eqn
The effect oi the Fermi correlation on the pair density as
yields a
(E7.16)is so pronounced that even a single determinantal function
of
good descripiion of the local properties of the Fermi hole' The addition
One can
Coulomb correlation does not changeits significant spatial features'
of the Pauli principle as correcting for the self-pairing of
view the net result
338 CHEMICAL MODELSAND THE LAPLACIAN E7.1

electronsof identical spin in a non-trivial way, as determinedby the spatial

propertiesof the Fermi hole (seeeqn (87.13),for example).Slater(1951)was
led to proposethat the exchangechargepotential be proportional to ptl: on
the basis of a modelling of the Hartree-Fock expressionfor the Fermi hole.
We now show that the extent to which sets of electronsmay be localized in
real space is determined solely by the spatial properties of the Fermi hole
(Bader and Stephens1974, 1975).
The averagenumber of pairs of electronsin a region O, a quantity denoted
by Dr(Q,Q), is determinedby integrating both coordinatesin p(rr,rr) over
C),a result which is equivalent to expressingthe result in terms of the event
probabilitiesP,(O),

- 1): }(N'(o)- ,vtott.

Dr({l,o): iL,p,(a)n@ @7.17)
Consider now the properties of the region (A) in the limiting situation when
P"(Q), the probability of one particular event,equalsunity and the probabilit-
ies of all other events are thus zero. In this situation lf(O), the average
number of electrons in O, equals n. The number of pairs which can be
formed is n(n - l)12: N1O;1N(O)- \12, the number of distincr pairs which
can be formsd from an isolated and distinct set of n objects.When the pair
population of a region Q is expressibleas .rU1Q; (l{(O) - 1)12it is referredto as
'pure pair' population.
being a A set of electronswith such properties would
be perfectlylocalized in the region Q, i.e.they would behaveas a separateand
distinguishable subset of electrons and, since the self-pairing is totally
accountedfor within the region, the wave function for the total systemcould
be expressedas a simple product of separatelyantisymmetrized wave func-
tions for the n electronsin Q and the (l{ - n) in Q'. This is the limiting caseof
localization in an electronicsystemand we show that it occurs when the total
Fermi hole for each of the n electronsin O is confined totally to the region O.
Using eqn (E7.10),the expressionfor the averagenumber of pairs in a
region O may be expressedso as to incorporate the correlation term to yield

D2(O,O) :l dt,
J
o"P(rt'
rz): (r{(O)'+F(O,O))12 (E7.18)

where

F(O,O) : i o , ,dir ,p ( r, )p ( rr )f ( r, , r r ) .
JO JO
(E7.1e)

F(Q, Q) is a measure of the total correlation which is contained within the

region (O). For a Hartree-Fock wave function, which includes only Fermi
correlation, the magnitude of F(O, O) is a measure of the total (integrated)
Fermi hole of the l{(A) particles that lies within the region O. From eqns
(E7.11) and (E7.19),the limiting value of the correlationis seento be - lgtO).
E7.l PAIRDE}iSIT\ 339

Thus, only if the correlation hole for each of the .\-(O) particlesin O is totally
contained within Q for all valuesof rr in Q will Dr(Q.Q) correspondto the
exact number of pairs which can be formed from a distinct set of lg(O)
particles.That is, in this limiting situation the ,VtO)electrons(wherelf lO; is
necessarilyan integer)generatea pure pair population with
- r)12.
Dr({),Q): N(O)(l\/(o) (E7.20)

Otherwise,ltr1Q;. tru' (O)l1/2 and, from eqn (E7.l7), D 2(O' O) t ,V(O)(N-(O)

- \l2.Only when the relationshipexpressedin eqn (87.20) applies is the
number of pairs in a region determined by (or in any way related to) the
averagenumber of particleswithin this sameregion. In this limit both NtO)
and D r(O, O) yield pure populations as their values are determined by a
single event.
It is now easy to demonstrate that the fluctuation in the averageelectron
population of a region Q vanisheswhen the Fermi correlation for the l{1O;
electronsin Q is totally containedin O. Equation (F.7.9)for A(lV. O)can be re-
expressedin terms of the averagenumber of pairs in O as
+ N(o)- tl{(o)l'.
A(r{,Q): 2D2(Q,C)) (E7.2r)
Substitution of eqn (E7.18)for Dr({1,O) into eqn (87.21)yields
n (N, O) : l{(O) + F(O,O). (E7.22)

The fluctuation attains its minimum value of zero under the same condition
that Dr(O, O) is reduced to a pure pair population, that is, when one
particular event has a probability of unity. From eqn (E7.22)the limiting
value of F(Q, O) is equal to - lV(O), a value which, as stated previousll'.
implies that the total Fermi correlation hole for each electronin O is totally
contained within the region Q.
The magnitude of the limiting value of F(O, O), which equals N"(O)
+ l{B(O), corresponds to the Fermi holes for both sets of electrons being
totally contained within O for any motion within the boundaries of O. Thus,
all other electronsof either spin are totally excluded from O and there is no
exchangeof electronsin O with electrons outside of the region Q. The ratio
lF(O, O/N(O)l therefore, provides a quantitative measure of the extent to
which the average number of electrons lg-(O) in O are localized to that
particular region of space.The quantity,

/(Q): lF(O,QyN(o)l x 100, (87.23)

is a measure of the per cent localization of the electrons to a region Q. A
localizationof 100 per cent correspondsto the Fermi hole oJ'eachelectronbeing
totally containedin the region {2.
The expressionfor F(O, O) is particularly simple for a Hartree-Fock or
single determinantal wave function. Multiplication of the expressionfor the
340 CHEMICAL MODELSAND THE LAPLACIAN E7.l

Fermi hole given in eqn (87.13)by p"(r r) and integrationover the coordinates
of both electronsyields, together with a correspondingtreatment for the
electrons of p spin,

F(Q, Q) : -Li,S:'j(O) -L!,Sl;tOl : Foo(f),A) -r Fpp(Q,Q) (87.24)

where S,r(O)is the overlap of the spin orbitals @, and $, over the region O.
Complete localization is, of course,possibleonly for an isolated system.
What is remarkable.however.is the extent to which the electronsof atoms in
an ionic molecule approach this limit of perfect localization, with /(O) values
in excessof 95 per cent not being uncommon. In systems,such as the fluorides
and chloridesof lithium and sodium displayedin Fig. E7.2,the atomic surface
of zero flux is found to minimize the fluctuation in the atomic populations
and, thus, the magnitude of the correlation hole per particle is an extremum
for such atoms. The properties of the number and pair densitiesfor these

---.--.-.--..--_-

=r':

LiF LiCl

NaF NaCl

F t c . E 7 . 2 . C o n t o u r m a p s of the charge distribution showing the position of the interatomic

surfacesfor ionic diatomic molecules.The outer contour is 0.002au with succeedinsvalues as
given in the Appendix (Table A).
 E7.1 PAIRDE\SIT\- 341

T a b l eE 7 . l
irt iottir'.sl'stenisx
Propertieso'l'numberantl puir tlertsiric.'

AB Pa Yt (rn 4(A) : , ) D 2 ( AB
/ ( , 4| / t B t D , r A .A ; D 2 ( BB , )
- q(Bt (no) ('ro)

LiF 2. 955 0. 0802 0.7018 + 0 . 9 3 7 9 , s . 59 9 1 l.l4 44.4 20.4

Licl 3.825 0.0462 0 .2 6 5 7 + 0 . 9 3 0 9 60 9 9 . 5 1.l5 151.7 37.1
NaF 3.629 0.0543 0 .4 6 5 5 + 0.944 99.5 98.1 45.6 44.5 99.9
NaCl 4.461 0.0358 0.2004 + 0 .9 1 1 9 8 .9 9 9 .4 4 5.9 1 51 . 5 180.6

* Calculated from functions close to the Hartree-Fock limit (Mclean and Yoshimine 1967). All
*
m o l e c u l e si n 1 E g r o u n d s t a t e .A l l d i m e n s i o n e dq u a n t i t i e si n a t o m i c u n i t s .

molecules AB are summarized in Table 87.1. The atomic populations,

together with their derived charges,and the atomic pair populations are close
+ -
to the limiting values anticipated for the ionic model A B for these
systems-net charges of * I with the electrons localized in two separate
closed-shell structures. The localization of the electrons in the separate
atomic basins,as measuredby the localization of the Fermi hole for both a
and B electrons,is in excessof 95 per cent in every case.Correspondingly,the
atomic pair populationsare closein value to those associatedwith pure pair
populations resulting from complete localization of the electrons in each
a to m ic bas in.T hes ev a l u e sa re 1 .0fo r L i + ,4 5 .0 fo r N a* and F-, and 153for
-.
Cl The actual pair populations are, in each case,slightly in excessof the
values determined by the pure pair limit l{(O)(l{(C)) - 1)12using the ob-
served values of lf (O). In the limit of complete localization, the number of
pairs formed betweenthe electronson the anion and cation is maximized and
should equal simply l{(,4)N(B). Thesevalues,followed by thosefor the ionic
model A+B- in brackets, are to be compared with those listed under
Dr(A , B ) in T ableE 7 .1 .T h e y a re 2 0 .5(2 0 )fo r L i F , 3 7.1(36)for Li C l , 100(100)
for NaF, and 180.7(180)for NaCl. To the extent that the valuesof Dr(A, B)
match the limiting values,there is no exchangeof the electronson either ion
with those of its neighbour.
The spatial properties of the Fermi hole that give rise to this physical
localization of electronic charge within the basins of the separateatoms is
illustrated in Fig. 87.3. The display of the Fermi hole, since it representsthe
remoualof spin density equiualentto one electron,is a picture of the spreading
out or delocalizationof the chargedensityof the referenceelectronwhenplaced
at f r.
While it is possibleto find regions of spacebounded by surfacessuch that
the contained Fermi correlation is maximized to yield localized groupings of
electrons, these regions correspond to atomic cores or localized atomic
populationsand not, in general,to localizedpairs of bonded and non-bonded
342 CHEMICAL MODELS AND THE LAPLACIAN E7.l

LiF

l e l

NaCl

r l
e i
l

i i
--,./ / /
/ i /

-:'

, ' '

Flc. E7.3. Contour maps o[ the density of the Fermi hole for various positions of the reference
electron (denoted by a star) in LiF and NaCl. The positions of the interatomic surfacesare also
shown. The contours used in this and the following figure are in au and decreasein value from
t h e o u t e r m o s tc o n t o u r i n w a r d s i n s t e p so f - 2 x l } n , -4x l0', and -8x 10'with n
beginning at - 3 and increasing in steps of unity. Referenceelectron is: (a) 0.20 au from Li
n u c l e u s ; ( b ) 0 . 5 8a u f r o m F n u c l e u s ; ( c ) 0 . 6 a
0 u f r o m N a ; a n d ( d ) 1 . 6 6a u f r o m C l n u c l e u s I. n ( a ) t h e
referenceelectron is necessarilywithin the s-like core of Li. In the remaining plots the reference
electron lies beyond the inner s-like core. The magnitude of the contours in (a) are close to one-
half the value of corresponding spatial contours of the total density of the Li atom in Fig. E7.2.
That is, the Fermi density approachesp' : pP : pl2, the density of the e or B electronsover the
basin of the Li atom, leading to the almost total exclusionof all other a and B electronslrom this
basin for this position of the referenceelectron.

electronsas anticipatedon the basisof the Lewis model (Baderand Stephens

1975).In the hydrides past BH. for example,there are no regions of space
outside of the core for which the fluctuation can be minimized, i.e. in which
the electronsare physically localized.Methane is borderline in its behaviour,
but the motions of the valenceelectronsin the remaining hydrides,NHr,
HrO, and HF, are so strongly intercorrelatedthat no best partitioning is
found and the localizedpair conceptlosesits meaning.A bond loge in HF, for
 L O C A L C H A R G EC O N C E \ T R . { T I O N S 343
87.2

per cent of
example,which has a number population of tu'o. containsonly 53
for complete localization, while /(Q) equals
the Fermi correlation required
loge in the same molecule. Attempts to
only 49 per cent for a non-bonded
with average populations of two fail
define bonded loges in N, or F,
17 per
completelyas the containedFermi correlationamountsto only 28 and
complete localization' The atomic
cent; respectively,of that required for
per cent of the
surface of ,.ro flux in N, yields regions which contain 78
Fermi correlation required for complete localization of the seven electronsto
the basin of the atom- In this example,which one anticipates to be typical of a
atom is delocalized into the
delocalizedsystem,the charge density of either
per cent as a result of molecule
region of the other to the extent of 22
highly
foimation. This percentagedelocalizatronincreasesif one excludesthe
N atom is 28 per cent
localized core. In N, the valence density of a
1 per cent over the second
delocalized over the second valence region and
Even the
core. In F, the electronson each F atom are 93 per cent localized.
adjacent atom' The
valencedensity is delocalizedto only 9 per cent over the
and highly localized
electronicchargeof a fluorine atom is both tightly bound
pronounced
within the basin of the atom. The difficulty in disrupting the very
intraatomic Fermi correlation in a F atom and the resulting small degreeof
delocalization of its electronsprovides an explanation for the weak binding
found in F, comParedto that in Nr.
in a
In summary, electrons can be localized within an atomic boundary
within the
closed-shellinteraction. They can also be separately localized
inner or valenceshells,but not to bonded regions which cross interatomic
portion
surfaces,nor, in general,to non-bonded regionswhich occupy only a
pair'
of a valence shell, to yield the Lewis model of the localized electron
not
In terms of the orbital model, it is clear from eqn (87.24), that it is
pairs
sufficientthat orbitals be localizable to achievephysicallocalization of
to
of electrons.Instead,each orbital, for both a and B spins,must be localized
:-0' Such separate
a separateregion of space,i.e' S,,(Q): 1 and St*;(O)
localization is approached for the same systems, LiH, BeH ,, and BHr, for
example,where ihe intrapair correlation energy dominates the total correla-
tion energy, but it is not a general result. Localized orbitals are not that
localized and they are, in general, not locali zed to separate regions of space'
the requirement for physic al locahzation of electron pairs.

Fj.2 Local charge concentrations and partial localization

of the Fermi hole

It is demonstrated through a study of the Fermi hole density that the local
the
decreasein the potential resulting from the approach of each ligand to
partial condensationof the pair density to yield
central atom dois result in a
344 CHEMICAL MODELSAND THE LAPLACIAN F.7.2

patterns of localization as anticipated by the VSEPR model. It is demon-

strated that there is a correspondencebetweenthis partiallocalization of the
pair density and the appearanceof local charge concentrationsin the VSCC
of the central atom.
The properties of the Fermi hole for a physically localized pair of electrons
are illustrated by the core electronsin methane.In this molecule a sphere of
radius 0.53 au centred on the carbon nucleusdefinesa region of spacewhich
maximizes the contained Fermi correlation to yield a pair of electronswhich
are 88 per cent localized.The localized nature of the Fermi hole is illustrated
in Fig. E7.4.In spite of the unsymmetricalposition of the referenceelectron,
the hole is nearly symmetrical with respectto the nucleus,and it corresponds
closely to a plot of the a density distribution in the region of the core. With
the exception of the contours of very low value extending out to the proton,
this plot remains essentially unchanged for any position of the reference
electron within this localized core region. Thus the core region is maximally
isolated from the remainder of the system. The fluctuation in its average
population of two is minimized, as is the exchangeof its electronswith those
in the remainder of the system.
While there is localization into shells in a free atom, as reflected in the
Laplacian distribution as well, there is no possibility of defining a localized
pair within the L or M shells,as the Fermi hole for each electron is spread
over its respectiveshell. It has already been noted that the uniformity of the
outer shell of charge concentration is disrupted with the approach of ligands
to the central atom with the resultant formation of local concentrations of
charge.The formation of bonded charge maxima in the VSCC of the central
atom is understandablein view of a decreasein the electron-nuclearpotential
in each of the resulting internuclear regions of the molecule. However, the
decreasein potential is, in general,not sufficient to condensethe Fermi hole
of a single electron to a degreesufficient to localize a pair of electronsfrom
the remainder of the system as is found for the core region of an atom. This
situation is also illustrated for the methane molecule.While it is not possible
to define a valenceregion which maximizes the contained Fermi correlation
in this molecule (Bader and Stephens1975),the division of spaceexcluding
the core into four equivalent bonded domains does represent the best
partitioning of its valence space in that it minimizes the sum of the fluctu-
ations in the populations of the individual regions. The pair of electrons in
each such bonded region is 69 per cent localized. It appears from many
examplesthat it is not possibleto maximrze the contained Fermi correlation
for a region of spacecontaining on the averagea pair of electronswhen the
total Fermi correlation falls below 70 per cent. The diffuse nature of the
Fermi hole in the valenceregion of this molecule which preventsthe physical
localization of a pair of electrons is illustrated in Fig. E7.4 for a reference
electronlocatedat the bonded maximum in the VSCC of carbon.This hole is
Frc. E7.4. Contour maps of the Fermi hole density in methane for various positions of the
referenceelectron,as indicated by a star. Each map is for a plane containing the carbon and two
hydrogen nuclei as labelled in map (b) (a) Referenceelectron,0.35 au from C nucleus,is within
the localized core of radius 0.53 au. (b) Referenceelectron is at position of bonded maximum in
VSCC of carbon, 1.02au from nucleus.(c) Referenceelectron is 0.69 au from nucleus.The Fermi
density is a maximum at the C nucleusand is lesslocalizedin the valenceregion of the C-H bond
than in map (b). (d) Referenceelectron is 1.28au from C nucleus. The area within the inner
(0.08au) C-H 'bond' contour is decreasedrelative to map (b) and more density is placed within
the core and non-bonded regions.(e) Referenceelectron is moved off bonded maximum towards
second proton. It is 0.99 au from nucleus and still within neighbourhood of bonded maximum.
Map (e) is essentiallyunchanged from map (b). (f) Referenceelectron, at same distance from
nucleusas in map (e),is moved further, to a point on boundary of two bonded regions.This map
representsthe maximum possible delocalization of the Fermi hole for this distance of the
referenceelectron from the C nucleus.
346 CHEMICAL MODELSAND THE LAPLACIAN 87.2

much lesslocalized than that for the core electrons,extending into the non-
bonding region as well. Its shape remains insensitiveto the position of the
reference electron only for motion of the reference electron within the
neighbourhood of the bonded maximum. As the referenceelectron is re-
moved from this region, the hole becomesmore delocalizedand its overlap
with the core and neighbouring bonded regionsincreases.For this reasonone
cannot maximize the contained Fermi correlation nor, equivalently,minim-
ize the fluctuation in its population; any variation in the boundary of a
valencedomain causesthe averageelectronpopulation to increasefasterthan
the contained Fermi correlation.
While the valencedensity is not physically localized into isolated pairs of
electronsas is the core density,there is partial localization of the pair density
as reflectedin the value of 0.69 for the fraction of the total possible Fermi
correlation for an averagepopulation of two electrons.Rather than a single
electron pair, each bonded domain in methane contains an average of 1.3
distinct pairs of electrons as opposed to the 1.75 pairs which would be
obtained if the electronswere fully delocahzedover the entire valenceregion.
Of equal importance is the observation that the Fermi hole for a reference
electron at, or in the neighbourhood of, a bonded maxima in the VSCC of the
carbon atom attains its minimum values along the axes of the other C-H
bonds. Recalling that the Fermi hole is a display of the spread or delocaliza-
tion of the density of the referenceelectron, one seesthat this density is
concentrated along a tetrahedral or threefold axis to maximally avoid the
three other corresponding bonded regions.
The Fermi hole for the referenceelectron at a bonded maxima in the VSCC
of the carbon atom has the appearanceof the density of a directed sp3 hybrid
orbital of valencebond theory or of the density of a localizedbonding orbital
of molecular orbital theory. Luken (1982, 1984) has also discussed and
illustrated the properties of the Fermi hole and noted the similarity in
appearance of the density of a Fermi hole to that for a corresponding
localized molecular orbital. We emphasizehere again that localized orbitals
like the Fermi holes shown above for valence electrons are, in general,not
sufficiently localized to separateregions of spaceto correspond to physically
localized or distinct electron pairs. The fact that the Fermi hole resembles
localized orbitals in systemswhere physical localization of pairs is not found
further illustrates this point.
It should be borne in mind that the resemblanceof a Fermi hole density to
that of alocalized valenceorbital is obtained only when the referenceelectron
is placed in the neighbourhood of a local maximum in the VSCC. The Fermi
hole and hence the density of the referenceelectron are much more de-
localized for general positions throughout the valence region (see Fig.
E7.4(f)). Locahzed molecular orbitals thus overemphasizeelectron localiz-
ability and do not provide true representationsof the extent to which
electrons are spatially localized.
87.2 L O C A L C H A R G EC O N C E \ T R A T I O N S 347

There is partial localization of the ralence density in methane. The

condensation into four partially localized pairs of electrons arranged along
four tetrahedral axesis a result of the combined effectsof the ligand field and
the Pauli exclusion principle describedabove. Most important is that this
partial localization of the pair density is reflected in the properties of the
VSCC of the carbon atom which undergoesa correspondingcondensation
into four local concentrations of electronic charge. These properties of the
pair density are not just the result of the tetrahedral symmetry of the ligand
field in methane because,as we will now see,the Fermi hole exhibits the same
behaviour in the ammonia molecule.
The Fermi holes obtained for a referenceelectron at the positions of a
bonded and the non-bonded maxima in the VSCC of the nitrogen atom in
ammonia are illustrated in Fig. E7.5.This figure shows that there is a partial
condensationof the pair density in ammonia to yield four partially localized
pairs of electronsarranged along four axeswhich are approximately tetrahe-
drally directed. The pair density is less localized than in methane with the
bonded and non-bonded domains containing, respectively,61 and 55 per cent
of the total possible Fermi correlation for average electron populations of
two. There are, on average, 1.38 distinct pairs of electrons in a bonded
domain and 1.45in a non-bondeddomain of ammonia, and the pair density
is further removed from the limit of localizedpairs than it is in methane.As in
methane, the pattern of delocalization of the density of a bonded or of the
non-bonded referenceelectron is such as to maximally avoid the three other
axesof pair localizatron.Thus, in agreementwith the postulated behaviour of
the electron pairs in the VSEPR model, the most probable arrangementof
the partially localized pairs of electrons in ammonia is approximatell'
tetrahedral,eventhough the ligand field is only threefold.Also, in accord with
the VSEPR model, the Fermi hole for the non-bonded referenceelectron is
laterally more diffuse than is that for a bonded referenceelectron,a property
also reflectedin the properties of the correspondingmaxima in the VSCC of
the nitrogen atom.
Of equal importance to the VSEPR model is the observation that, for
geometrieswhich, on the basis of the VSEPR model, do not maximize the
distance between the electron pairs, the Fermi holes are less localized and
exhibit greater overlap with the holes for neighbouring domains. As an
example, the Fermi hole for a referenceelectron on the threefold axis in
planar ammonia, situated at one of the two symmetrically related non-
bonded charge concentrations,is strongly delocalizedover both sidesof th'e
plane of the nuclei (Fig. E7.5). The density of the non-bonded reference
electron clearly overlaps the holes for reference electrons placed at the
bonded maxima more in this geometry than in the most stable pyramidal
geometry. The Fermi holesfor referenceelectronsplaced at the coruesponding
bonded and non-bondedmaxima in the VSCC of the central atom are most
localized and least mutually ouerlappingfor those geometriesw'hich in the
348 CHEMICAL MODELSAND THE LAPLACIAN 87.2

Flc. E7.5. Contour maps of the Fermi hole density for pyramidal (a, b) and planar (c, d)
ammonia. In maps (a) and (b)the referenceelectron is positioned at the non-bonded and bonded
maxima, respectively,in the VSCC of the nitrogen atom. Note that the Fermi density is more
contracted towards the core in NH. than it is in CHn, as are the maxima in its VSCC. Maps (c)
and (d) are correspondingplots for planar ammonia. The density of the non-bonded Fermi hole.
map (c), is more delocalized than that for the pyramidal geometry, map (a). In the planar
geometry,contours of the non-bonded Fermi hole density encompassthe N-H internuclearaxis.
Clearly, maps (c) and (d) overlap one another to a greater extent than do maps (a) and (bl the
electron pairs are more localized in pyramidal than in planar ammonia.

VSEPR model maximize the interpair separationsor, equiualently,maximize

the separationsbetweenthe local maxima in the VSCC of the central atom.
It is important to emphasizethat the correlation which exists betweenthe
properties of the Fermi hole and the VSEPR model is made most evident
through the use of the properties of the Laplacian of the charge density. The
correspondence is greatest when the position of the reference electron
L

ES
REFERE\C 349

c o i n c i d e sw i t h a l o c a l m a x i m u m i n t h e V S C C o f t h e c e n t r a l a t o m . T h e F e r m i
holes are individually most localized for such a coincidencein position, and
they are least overlapping for the geometr)' u'hich maximizes the distances
between the local maxima in the VSCC of the central atom.

References

Aslangul,R., Constanciel,R., Daudel, R., and Kottis. P. (1912).Adt:.quantumChem.6,

93.
B a d e r ,R . F . W . ( 1 9 6 0 )M . o l . P h v s . 3 ,l 3 l .
B a d e r ,R . F . W . ( 1 9 6 2 ) C . a n .J . C h e m . 4 0 ,1 1 6 4 .
B a d e r ,R . F . W . ( 1 9 8 1 )l.n T h e . / b r c ec o n c e p ot f c ' h e m i s t r - r ' ( eBd. .M . D e b ) ,p p . 3 9 - 1 3 6 .
Van Nostrand Reinhold.New York.
Bader, R. F. W. and Beddall, P. M. (1972).J. Chem. Phys.56,3320.
B a d e r ,R . F . W . a n d C h a n g ,C . ( 1 9 8 9 )J. . P h v s .C h e m 9 3 . 2 9 4 6 .
B a d e r ,R . F . W . a n d E s s e n H , . ( 1 9 8 4 )J. . C h e m .P h l s . 8 0 , 1 9 4 3 .
B a d e r ,R . F . W . a n d H e n n e k e r ,W . H . ( 1 9 6 5 )J. . A m . c h e m .S o c . 8 7 , 3 0 6 3 .
B a d e r ,R . F . W . a n d M a c D o u g a l l ,P . J . ( 1 9 8 5 )J. . A m . c ' h e mS. o c .1 0 7 , 6 7 8 8 .
Bader, R. F. W. and Preston,H. J. T. (1969).Int. J. quantumChem.3, 327.
B a d e r ,R . F . W . a n d S t e p h e n sM , . E . ( 1 9 1 4 )C . h e m .P h v s .L e t t . 2 6 , 4 4 5 .
B a d e r ,R . F . W . a n d S t e p h e n sM, . E . ( 1 9 7 5 )J. . A m . c ' h e m . ! o c . 9 7 , 7 3 9 1 .
B a d e r ,R . F . W . , H e n n e k e r ,W . H . , a n d C a d e ,P . E . ( 1 9 6 1 u ) J. . C h e m .P f t r ' s . 4 6 . 3 3 4 1 .
B a d e r ,R . F . W . , K e a v e n y ,L , a n d C a d e , P . E . ( 1 9 6 7 b ) .J . C h e m .P h - l ' . s . 4 7 . 3 3 8 1 .
B a d e r ,R . F . W . , M a c D o u g a l l ,P . J . , a n d L a u , C . D . H . ( 1 9 8 4 )J. . A m . t ' h e m S . o r ' .1 0 6 .
1594.
B a d e r ,R . F . W . , G i l l e s p i eR , . J . , a n d M a c D o u g a l l ,P . J . ( 1 9 8 8 )J. . A n t .t ' h e n tS. o c ' l.l 0 .
7329.
Berlin, T. (1951).J. Chem.Phys. 19, 208.
Bingel, W. A. (1963).Z. Naturforsch. l8a, 1249.
B i r c h a l l ,T . a n d G i l l e s p i e R
, . J . ( 1 9 6 3 ) .C a n . J . C h e m . 4 1 , 2 6 4 2 .
B o y s ,S . F . a n d B e r n a r d i ,F . ( 1 9 7 0 ) M . o l . P h y 5 .1 9 , 5 5 3 .
Burgi, H, B. and Dunitz, J. D. (19831.Ac'c.Chem.Res. 16, 153.
Cade, P. E. and Huo, W. M. (1973).At. Data nut'\.Data T'ables12,415.
C a d e ,P . E . a n d H u o , W . M . ( 1 9 7 5 ) A . t . D a t a n u c l .D a t a T a h l e s1 5 , 1 .
C a d e ,P . E . a n d W a h l , A . C . ( 1 9 7 4 \ .A t . D a t a n u c l .D a t a T a b l e s1 3 , 3 3 9 .
Carroll, M. T. and Bader, R. F. W. (1988).Mol. Phys.65, 695.
Carroll, M. T., Chang, C., and Bader, R. F. W. (1988).Mol. Phys.63, 387.
Carroll, M. T., Cheeseman, J. R., Osman, R., and Weinstein,H. (1989).J. Phys.Chem.
93.5120.
C o u l s o n ,C . A . ( l 9 4 l a ) . P r o c . C a m b r i d g ep h i l . \ o c . 3 7 , 5 5 .
Coulson, C. A. (l94lb). Proc. Cambridgephil. Soc.37,74.
Coulson, C. A. and Duncanson,W. E. (l94la). Proc. Carnbridgephil. Soc'.37,67.
Coulson, C. A. and Duncanson,W. E. (l94lb). Proc. Cambridqephil.loc.37,406.
D a s , G . a n d W a h l , A . C . ( 1 9 6 6 )J. . C h e m .P h y s . 4 4 , 8 7 .
Daudel, R. (1953).C. R. Acad. Sci.,Paris 237,601.
350 REFERENCES

Daudel, R. (1968).Thefundamentalsof theoreticalchemistry.Pergamon Press,Oxford.

Daudel, R., Brion, H., and Odiot, S. (1955).J. Chem.Phys.23,2080.
Daudel, R., Bader, R. F. W., Stephens,M. E., and Borrett, D. S. (1974a).Can. J. Chem.
52,1310.
Daudel, R., Bader, R. F. w., Stephens,M. 8., and Borrett, D. s. (1974b).Can. J. chem.
52, 3077.
Daudel, R., Stephens,M. E., Kapuy, E., and Kozmutza, c. (1976).chem. phys.Lett.
40,194.
Davidson, E. R. (1972).Reu.mod. Phys.44,451.
Ebbing, D. D. and Henderson,R. C. (1965).J. Chem. Phys.42,2225.
Feynman, R. P. (1939).Phys. Reu.56,340.
F u k u i , K . ( 1 9 7 1 )A . c c . C h e m .R e s .4 , 5 7 .
Garcia-Sucre,M. and Castells,V. (1989).Int. J. QuantumChem.Symp. 23,79.
Gilbert, T. L. and Wahl, A. C. (1967).J. Chem.Phys.47,3425.
Gillespie, R. J. (1972).Molecular geomety. van Nostrand Reinhold, London.
Gillespie,R. J., Kapoor, R., Faggiani, R., Lock, c. J. L., Murchie, M., and passmore,J.
( 1 9 8 3 )J. . c h e m .S o c .C o m m u n . 8 .
H u t s o n , J . M . a n d H o w a r d , B . J . ( 1 9 8 2 )M . ol. Phys.45,79l.
Kato, W. A. (1957).Commun.pure appl. Math.l0, 151.
King, R. B. and Rouvray, D. H. (1977).J. Am. chem.9oc.99,7834.
Kutzelnigg,W. (1973).Fortschr.Chem.Forsch..41,31.
Lennard-Jones,J. E. (1952).J. Chem. Phys.20, 1024.
L e w i s ,G . N . ( 1 9 1 6 )J. . A m . c h e m .S o c . 3 8 , 7 6 2 .
L e w i s ,G . N . ( 1 9 3 3 )J. . P h y s .C h e m .l , 1 7 .
Luken, W. L. (1984).Croat. Chem.Acta 57,1283.
Luken, W. L. (1982).Theor.Chim. Acta 61,265.
MacDougall, P. J. and Bader, R. F. W. (1986).Can. J. Chem.64, 1496.
Mclean, A. D. and Yoshimine,M. (1967).IBM J. Res.Deu. 12,206.
McWeeny, R. (1960).Reu.mod. Phys.32,335.
Morse, P. and Feshbach,H. (1953).Methods of theoretical physics,Part 1. McGraw-
Hill, New York.
ONeil, S. V., Nesbitt, D. J., Rosmus, P., Werner, H.-J., and Clary, D. C. (1939).J.
C h e m .P h y s . 9 l , 7 l l .
Pack, R. T. and Brown, W. B. (1966).J. Chem.Phys.45,556.
Passmore,J., Sutherland,G., and white, p. s. (1980).J. chem. Soc.chem. commun.
330.
Politzer, P., Landry, S.J., and Warnheim, T. (1982).J. phys.Chem.86,4767.
Ramirez,B. I. (1983).Chem.Phys. Lett.94, 180.
Roothaan,C. C. J. and Weiss,A. W. (1960).Reu.mod. phys.32,194.
Sagar,R. P., Ku, A. c., Smith, v. H., and Sinas,A. M. (1988).J. chem.phys. gg,4367.
S c r o c c oE , . a n d T o m a s i ,J . ( 1 9 7 8 ) A . d u . q u a n t u mC h e m . l l , 1 1 6 .
S h a n n o n ,E . ( 1 9 4 8 ) B . e l l . S y s t .t e c h .J . 2 7 , 3 7 9 .
S h i , Z . a n d B o y d , R . J . ( 1 9 8 8 )J. . C h e m .P h y s . 8 8 , 4 3 7 5 .
Sinanoglu,O. (1962).J. Chem.Phys.36, 706.
S i n a n o g l uO , . a n d S k u t n i k ,B . ( 1 9 6 8 ) C . h e m . P h y s .L e t t . 1 , 6 9 9 .
S l a t e r ,J . C . ( 1 9 3 3 ) J. . C h e m .p h v s . l . 6 8 7 .
 REFERENCES 35r

S l a t e r J, . C . ( 1 9 5 1 ) P
. hysR
. eu.81,385.
Vol. I. McGraw-Hill, New
Slater,J. C. (1963). Quantumtheory of moleculesantl sttlid.s.
York.
Slater,J. C. (1972).J. Chem. Phys.57,2389.
Slee,T. (19864).J. Am. chem'Soc. 108, 606.
Slee,T. (19S6b).J. Am. chem.Soc. 108, 7541.
Slee.T.. Larouche,A., and Bader, R. F. W. (1988).J. Phl's.Chem.92, 6219.
 8

T H E A C T I O N P R I N C I P L EF O R A Q U A N T U M
SUBSYSTEM

Quantum mechanicsinvolves two distinct setsof hypotheses-the generalmathemat-

ical scheme of linear operators and state vectors with its associated probability
interpretation and the commutation relations and equations of motion for specific
dynamical systems.It is the latter aspectthat we wish to develop,by substitutinga
single quantum dynamical principle for the conventional array of assumptionsbased
on classicalHamiltonian dynamics and the correspondenceprinciple.
Julian Schwinger(1951)

8.1 A common basis for classicaland quantum mechanics

8. 1. 1 I nt r odu c ti o n

It has been demonstratedin Chapter 5 that Schrodinger'soriginal derivation

of the wave equation can be generahzedto provide a quantum description of
an atom in a molecule. His derivation uses the calculus of variations as
applied to an energy functional of r! and its extension to an atom in a
molecule entails two essentialsteps:the introduction of a particular type of
operator through its identification with the variation of the statefunction and
a variation of the surfacebounding the atomic subsystem.The replacementof
the variations in the state function by the action of an operator of the form
- (ielfiG$ is, as recognized by Schwinger, a fundamental step with far-
reaching consequences. Such operators are known as generatorsof infinitesi-
mal unitary transformations.Dirac (1958) pointed out that they are the
quantum counterparts of the generators of infinitesimal contact trans-
formations in classical mechanics and the correspondence between the
properties of these two sets of generatorsis responsiblefor the fundamental
similarity in the structure of classicaland quantum mechanics.Becauseof this
correspondence,Schwingerwas able to state a singleprinciple-the principle
of stationary action-from which either classicalor quantum mechanicscan
be derived in its entirety. It is through this principle that one is able to extend
the predictions of quantum mechanics to the properties of a subsystemof
some total system,that is, to an atom in a molecule,an open quantum system.
Becauseof its central role in the theory of atoms in molecules,and no less
becauseof the corresponding role it plays in providing a unifying structure
for physics,a r6sum6of what the principle of stationary action is and of how
it is obtained is given here.
8.1 C L A S S I C A LA N D Q U A \ T U \ 1 \ l E C H A N I C S 3s3

The presentationis precededby a summarv of the principal mathematical

aspectsof quantum mechanicsas requiredfor the developmentof the theory,
including, in particular, the properties of infinitesimal unitary trans-
formations.This reviewis not abstractbut is paralleledby a demonstrationof
the elegantmanner in which thesepurely'mathematicalconstructsaccount
for the propertiesthat are unique to a quantum system.Section8.1 concludes
with a review of infinitesimalcontact transformationsin classicalmechanics
as required to develop the analogy between the classical and quantal
equationsof motion, a developmentmade possibleby infinitesimalcanonical
transformations-the genericname for infinitesimal transformations in both
classicaland quantum mechanics.A readeris referredto books by Messiah
(1958) or Roman (1965) for more detailed accounts of the mathematics
underlying the framework of quantum mechanics.In particular, Roman gives
a detailed discussion of canonical transformations and their role in
Schwinger's action principle. A highly recommended treatise on classical
mechanicsis the one bv Goldstein (1980)'

8.1.2 State vectors,state functions,and transformationfunctions

The mathematical formalism of quantum mechanicsis expressedin terms of

linear operators, which represent the observablesof a system, acting on a
state vector which is a linear superposition of elements of an infinite-
dimensionallinear vector spacecalledHilbert space.We requirea knowledge
ofjust the basicpropertiesand consequences of the underlyinglinear algebra.
using mostly those postulates and results that have direct physical conse-
quences.Each state of a quantum dynamical system is exhaustivelycharac-
tertzed by a state vector denoted by the symbol lY). This vector and its
complex conjugatevector (Yl which describes the samestate of the system,
are expressedin Hilbert space. The product clY), where c is a number which
may be complex, describesthe same state.
In general,for any pair of such vectorsdenoted by la) and lb) we can define
a number (bla) which is called the scalar product of (bl with la). This
product satisfieseqn (8.1) when acting on a linear combination of state
vectors.
(bl{f c'la,)}: I c,(bla,) (8.1)

as well as the relation

(bla): ((alb))*. (8.2)

That is, the scalar product equals the complex conjugate of its transpose.
From this it follows that (a a) : (ala)* and is a real number.
A measurablephysical quantity or observableis representedby a linear
Hermitian operator ri. The action of a linear operator on a statevector la) is
354 ACTION PRINCIPLEFOR A QUANTUM SUBSYSTEM 8.1

to transform it into another vector rhla).It is linear sinceits action on a linear

combination of state vectorsis given by

ri{l,c,la,)}: I,r,cila,). (8.3)

The complexconjugateof the vectorcbla)is (olriI wherea.rtis the Hermitian
conjugateor adjointof cD.Replacementof lb) by cDlb)in eqn (8.2)yields
(blrhlla): {(clc0lb)}*, (8.4)
an equationwhichrelatesthe actionof cbto that of its Hermitianadjointcirt.
An operatoris said to be Hermitianif
rh : ritt, (8.5)
a propertywhich yieldsthe importantresult
(blrnla): { (alcDlb)*. (8.6)
}
Useof eqn(8.2)again,
with (al replacedby (aliand with lb) replaced by tlb),
yieldsthroughthe useof eqn (8.4)with c; : itt,,the additionalresult

lrtit : Ql,f . (8.7)

The only observablevaluesfor a property Q are one of the eigenvaluesar,

of the associatedlinear Hermitian operator cir as defined in the eigenvalue
equation

,hla,) : rloiktti) (8.8)

where, since cDis Hermitian, the eigenvaluesare real. An eigenvectorlar,) is

labelled by its associatedeigenvaluerr-r,.By assumption,all of the eigenvectors
which satisfy eqn (8.8) form a complete set. That is, they span the Hilbert
spaceand, as illustrated below, any arbitrary state vector can be expandedin
terms of this set of functions. Thus, the eigenvectorsof a Hermitian operator
either form a complete orthogonal set or may be combined to yield such a set.
When normalized to unity they thus satisfy the expression

(a,lat) - 6ii. (8.e)

The orthogonality extendsto the linearly independentmembersof a degener-
ate set. An eigenvector is labelled with its eigenvalue for the property O
because,when the systemis in a state describedby this vector, one is certain
that the measurementof the property O will yield the particular eigenvalue
co,.Since the ro, are the only observable values of the property O, actual
measurementof this property will carry the state vector lY) of a systemover
into one of the eigenvectorsla-r,).In this way the measurementof a physical
property placesthe systemin a specificstate.Once the measurementhas been
carried out to vield the value (D;.on€ can be certain that the state vector
8.1 C L A S S I C A LA N D Q U A N T U M M E C H A N I C S

one cannot predict ri htch trl

lY) : lar,).Without such previouspreparation.
the eigenvalueswill be observedin a given measurement.
princrple'-ri
thii latter situation is handled within the theorl''through the
set oi
superpositionwherein the state vector is expanded in the complete
with the property C)' an expansion which ultimatell
eigenvectorsassociated
property'
ull,o*, for the observationof any one of the possiblevaluesof the
Thus, the state vector is expanded as
(8.10)
lY) : Irt; lr;).
product
A coefficientcu in this expansion is evaluated by forming the scalar
the orthonormality property expressedin eqn
of lY) with (a;ul and uring
(8.9)to yield
c o: ( c o p l Y ) ' (8'11)

The physical interpretation of this expansion and the manner in which it can
O are
Ur ,rtei to predict the outcome of a measurement of the property
for the average value of the property
obtained by consideringthe expression
postulate,is
Q obtained for a systemin the state lY). This averagevalue, by
given by
( ci ,) = (Yl c D l Y ) : IrIu c l c o (a ,l cD
l oo) (8' 12)

(8.8) and the

when the state vector is normalized to unity. Using eqn
orthogonality conclition of eqn (8.9) this averagevalue reduces to

(6) : L,lttl'rt. (8.13)

Q by
The value (cD) is the one that would be obtained for the property
and it is equal to the value of
averaging the results of many measurements
that particular value' where
each auweightedby the probability of observing
sum of
the weighting factor is given by |tolt.For lY) normalizedto one, the
setting ar equal to the unit
the weilhting factors is unity, as can be seenby
obtaining the value 0)krn
operatorin eqn (8.12).Thus lcul2is the probability of
a singlemeasurement.Using eqn (8.11)this probability is given by

l r o l ': l ( r u l Y ) l t (8.14)
Equation
and the quantity (roolY) is thus the amplitude of this probability'
(8.10)is the representationof the state vector in terms of the and the
co-basis
probability amplitude (arlY) is called the representatiue.
Two linear operators possessa common set of eigenvectorsonly if the two
with the
operators commute. Suppose two observablescDand f associated
pioperties Q and T, respectively,possessa common set of eigenvectors,
r i) : a,lai, r i)
,i)lrtt,,
and
iloti,tl): ril@;,ri).
3 5 6 A C T I O N P R I N C I P L EF O R A Q U A N T U M S U B S Y S T E M 8.1

Measurement of the property O carries the state vector over into an

eigenvectorof cDwith the particular value for this property equal to co,.Since
an eigenvectorof rrris also an eigenvectorof f, the samemeasurementhas also
prepared the system to yield a particular value for the property T and it is
clear that subsequentmeasurementof this latter property will yield the value
rr. Thus, the valuesof the propertiesQ and T may be simultaneouslyknown
and measuredfor a given state of a system. It is also clear that, if two
observablesdo not commute, their values cannot be measured simultan-
eously, for measurementof the second property will force the system into a
new state which is no longer an eigenvectorof the first observableand, hence,
no longer possessiveof a sharp value for the first property.
In this manner the mathematical formalism incorporates the important
experimentalresult that not all physicalobservablescan be known simultan-
eously for a quantum system,the incompatibility of a knowledge of the
position and its conjugate momentum for a particle being an example of this
behaviour.The non-compatibilityof thesetwo observablesis enshrinedin the
Heisenbergcommutation rules

14,iil : th6;i,
1 4 ,Q
, i):0, (8.1s)
l f i ,P i ) : 0 .
These rules are a part of the formalism of quantum mechanics that is
recoveredfrom Schwinger'saction principle.
For any given system there exists a maximal number of mutually com-
muting or compatible observables.The maximum information about a given
quantum system is obtained by the simultaneous measurement of this
maximal set of compatible observables,a processcalled the maximal measure-
ment.The resulting set of eigenvaluesis used to classifythe state of the system.
For example, within the Russell-Saunderscoupling scheme for a many-
electron atom, the observablesfor the energy, for the square of both the
orbital and spin angular momenta, along with those for a single component
of both angular momenta, the set {fr, L',52, L",.f,1, for* a mutually com-
muting set of operators. Thus, each state of a many-electron atom is
characterized by giving the results of the maximal measurement as
lE, L, M y, S, Mr) where the values of the angular momenta are expressedin
terms of their associatedquantum numbers.
We are now ready to address the consequencesof a change in the
representation of the state function through a change in the set of basis
vectorsusedin the expansion,eqn (8.10).We will incorporatethe resultthat a
given set of functions will, in general,be simultaneouseigenfunctionsof a set
of commuting operators.Assumea set of mutually commuting operatorscD,,
6r , . . . , 6, an d d e n o tea p o s s i b l ec o l l e c te dset of ei genval ues
by the si ngl e
8.1 C L A S S I C A LA N D Q U A \ T U \ I \ I E C H { \ I C S l,s7

symbol o4 and label the associatedeigenrectc-rr cgrrespt)ndingl1.t'r.)' The

is
state vector expressedin the ro-representation
(8.16)
l Y) : l ,(c ,r,l Y) t' t,' ' ,.
The representative(ro,lY) may alternativell.be replacedb;-'the stateJunction
and denoted by lt(o,l to f ield
in the or-representation
(8.17)
lY) : l,lt(o,)lo,),
The state vector lY) is a vector spanning Hilbert space and is a symbol
representingall possiblevaluesof the set of eigenvalues 62r.The statefunction
particular basevector of this spaceand
{,(r,)is the projection of lY) on to a
The
signifiesa state with a particular set of values for the eigenvaluescr-r,'
the t-
same state vector can be expressedin terms of another representation,
representation,where r, denotesa set of eigenvaluesobtained from the action
of its
o f t h e m u t u a l l y c o m m u t i n gs e t o f o b s e r v a b l etsr , i r , . . . , f , o n o n e
si mu lt aneouseigenv e c to rs
l t-)'

lY) : I",/(t.)lt,), (8.18)

where f (r") is the state function in the t-representation' The connection

between the two representationsis ea'sily obtained by forming the scalar
p ro d uc t of lY ) wit h (o ,l i n e q n (8 .1 8 )'
( a r , l Y ): $ ( a , ) : l , ( r " l z , ) r / ( r , ) . (8.1e)

The quantity (a,r,lr,),the scalarproduct of the base vectorsused in the two

representations,is the probability amplitude of finding the crr,basestatein the
,,tur. state and is callid the transformation.function. The state function in the
@-representation, which represents the probability amplitude of observinga
eigenvalues
particular set of rr-r for the system described by lY), is obtained
iro- the ?-representation by summing over all of its base states. The
possibility of changing representations is fundamental to quantum mechanics
and the work of both Feynman and Schwinger is concerned with the
evaluation of the representativewhich enablesone to passfrom a representa-
tion at one time to another at a later time.
If one rewriteseqn (8.16)in the form

lY) : l,lro,)(a;,lY), (8.20)

it is evident that the effectof the summation term is to multiply lY) by the
forming a complete set
unit operator, i. Equation (8.20)is a result of the lcr-r'.)
and hence the identitY
(8.21)
l,lr,,) (ot,l: 1
condition.
is called the completeness
3 5 8 A C T I O N P R I N C I P L EF O R A Q U A N T U M S U B S Y S T E M 8.1

The concept of a representationand the associateddefinitionshave been

introduced for the caseof a discretespectrumof eigenvalues. This conceptis
readily extendedto the situation where the commuting set of observables
possesses a continuouseigenvaluespectrum.Denote by o, a setof eigenvalues
for the setof mutually commuting operators,d1, dr, . .. , d" with correspond-
ing eigenfunctionsla,), which possessa continuous set of eiggnvalues.The
analogueof eqn (8.20)for the expansionof lY) in the d-representationis

lY) : Jla")da,(a,lY) (8.22)

where the'r' summation is replacedby a correspondingintegration over the
continuous eigenvaluespectrum.The quantity (*,lY) is again the represent-
ative or state function in the d-representation.Let B" denote a second set of
mutually commuting observableswith basislp,). The two representations of
lY) using the two differentbasesare related by (compareeqn (8.19))
(a,lY) : J(n,lB,)dB.(B,lY) (8.23)
where (u,l[],) is again called the transformation function. If one introduces a
third setof mutually commuting observablesy, and their associatedbasis17,),
then we have as a specialcaseof eqn (8.23)
(a,lt) : J(a,lf")d\,(f "lt) (5.24)
The relation expressedin eqn (8.24)is known as the multiplicatiuelaw of
tr ansfor mation func t i ons.
An important example of a maximal set of commuting observableswith a
continuousspectrumof eigenvalues is provided by the operatorsf, represent-
ing the position coordinatesof a set of particles.The state function in terms of
the coordinate representationis given by

lY) : !lq,>aq,<q,|Y>
: !lq,>dq,,lt
h,) (8.25)
where the ket lq,) denotesthe simultaneouseigenvectorcorrespondingto the
set of eigenvaluesQ, (r: 1, . . . , n) for a system with n particles.In this
representation the representativeor state function (q"lY) is Schrcidinger's
waueJunction rlt(q,).The state vector denotesall possiblefunctional values of
rL(q,),while the state function itself representsthe value of this function for
one particular set of coordinates q,, the commuting set of eigenvalues.
Considertwo setsof the commuting setof coordinateoperators,one at time /,
the other at time r'. Using eqn (8.25),one obtains for the relation between
thesetwo basesin analogy with eqn (8.23)the expression
(q,,, t'lY) : rL(q,,.t') : I(q,,, t' lq,,t) dq,rlt(q,,t). (8.26)
Equation (8.26)is an important result. It shows that a knowledge of the
transformationfunction or transitionamplitude (q,,, t'\q,,t) sufficesto deter-
mine the dynamical behaviour of the systemwith time, since it relatesthe
8.1 C L A S S I C A LA N D Q U A N T U N ' IN {E C H A N I C S 359

statefunction at time r' to that at time r. Knowledge of how a systempasses

from its state at one time to that at a later time implies knowledge of the
system'sequation of motion. The path integralequationof Feynman and the
differential form of this equation as given by Schwingerare equations for the
determinationof the transformationfunction in eqn (8.26).
Von Neumann has demonstratedthat, for a systemwith a finite number of
degreesof freedom, the transformation from one representationto another
discussedabove is what is termed a unitary transformation. It is in terms of
unitary transformations that one is able to demonstrate a correspondence
betweenclassicaland quantum mechanicsand it is upon this correspondence
that Schwinger based his formulation of quantum mechanics.

8.1.3 Unitary transformations

The state of a classicalsystem is specifiedin terms of the values of a set of

coordinatesq and conjugatemomenta p at some time f, the coordinatesand
momenta satisfyingHamilton's equationsof motion. It is possibleto perform
a coordinate transformation to a new set of ps and qs which again satisfy
Hamilton's equation of motion with respectto a Hamiltonian expressedin
the new coordinates.Such a coordinatetransformationis called a canonical
transformation and, while changing the functional form of the Hamiltonian
and of the expressionsfor other properties,it leavesall of the numerical
values of the properties unchanged.Thus, a canonical transformation offers
an alternativebut equivalentdescriptionof a classicalsystem.One may ask
whether the same freedom of choosing equivalentdescriptionsof a system
exists in quantum mechanics.The answer is in the affirmative and it is a
unitary transformation which is the quantum analogue of the classical
canonical transformation.
A unitary transformation is one which, when applied to both the state
function and the observablesof a system,leavesthe description of the system
unchanged.Denote by 0 an operator with the property that its Hermitian
conjugateor adjoint is equal to its inverse,that is
-
00t:0t0:i or 0t : 1j t. (8.27)

a unitaryoperatorand0 is usedto performsimultan-

Equation(8.21)defines
eoustransformations cir
on both the statevectorlY) and on the observables
suchthat
lY)-lY'):t7lV), (8.28)
-
rb-- cit': 0rir0 t (8.2e)
It is easyto show that this transformation leavesthe description of the system
unchanged.The transformed operator rh' has the same eigenvaluespectrum
3 6 0 A C T I O N P R I N C I P L EF O R A Q U A N T U M S U B S Y S T E M 8.1

as does6. If initially one has the eigenvalueeqn (8.8),the action of ri' on lar!)
yields the same eigenvalus0ri,

,it'lto',): CAC-'01r,) : Crblot,): otijl@i) : 1x,;1.11'i).(8.30)

The expectationvalue of an observableis also left unchanged
: (Y, C- t 0aC-1 iy)
(Y'lct'lY'): ( 0Y, rb'UV>
: (YlcDlV) (8.31)
as is the normalization of iY)

( Y ' l Y ' ): ( i y , u w ) : ( y , 0 - t 0 V > : ( y l y ) . (8.32)

It is also easilydemonstratedthat operator equationsfor sums and products
of operatorsare castinto equivalentexpressions by a unitary transformation.
We shall be particularly interestedin the effectsof a unitary transformation
which is infinitely close to unity as applied separatelyto the observablesor
state vectorsof a system.Consider the operator

U:t-(ielh)G (8.33)
where e denotes a real infinitesimal quantity and d is a linear Hermitian
operator. If we retain terms only to first order in e, the inverseof 0 is

U'':l+(itih)G. (8.34)
- 1-
That ls,00 iignoring the terms in *. Sinced is Hermitian, the adjoint
o f 0 i s a l s o g i v e n b y e q n ( 8 . 3 4 )H
. enceCt:0-t a n d 0 i s u n i t a r y .T h e
infinitesimal operator rd when used in this manner is referred to as the
generator of the infinitesimal transformation.
The effect of such a transformation on an observable& is

&".-&': {i - (ielilG}e{i+ (klfiG} : &.+ (jelh)[d,

d]. (8.35)
Following Schwinger, the infinitesimal transformation induced on the
observabled by the generator ed is defined to be

6 & .: &.- d , : (i e l h )[G,d ] : (i l h)[F, d] (8.36)

where the infinitesimal Hermitian operator F : tG. The same trans-
formation, when applied to the state vector, yields

la)-- lu'): CloS: la) - (ilh)Fla), (8.37)

and the generator F induces an infinitesimal change in the state vector
given by
dla) : la') - le) : - (ilh)Flu). (8.38)
The concept of an infinitesimal unitary transformation can be used to
provide a differentialcharacterizationfor the representativeof a state,or for a
8.1 C L A S S I C A LA N D Q U A N T U \ , 1M E C H A N I C S 36r

transformationfunction.The changein the representative in eqn (8.23)when

the commuting set of operators i, is altered b1 the^unitarytransformation
generatedby the infinitesimalHermitian operator F, is given by
d(a"lY): (da.lY): (i /r)(z,lFlY) (8.3e)

d(a,lY) : (ilh)!(a,lFlp,
)dB,(B,lY). (8.40)
Similarly,the transformationfunction (a,l;,) of eqn (8.24)can be character-
ized by the effectof altering the two commuting setsi, and y, into ct^- dd and
y - 6y as induced by the two infinitesimal generating operators { and Fr.
One obtains
d ( e , l ; ' , :) ( d r , l 7 , )* ( r , l d ; ' , ): ( i l h l ( r , l F ' - F , l t , ) (8.41)

or
- (ilh)i(e"l0'r)d0',(0',lFrlv,).
d(a,l),): (ilh)l(o,lF"lB,)d/i,(f,ly,> (8'a2)

It was emphasizedabove that the transformation function determines the

temporal evolution of a system.Equation(8.41)or (8.42)relatesthe differ-
ential of a transformationfunction to the action of infinitesimalgenerators
acting on the two associatedbase states.In his action principle, Schwinger
postulates an alternative expressionfor the infinitesimal change in the
transformationfunction, proposing that this changebe given by a variation
of the quantum action integral betweenthe same two states.
The above has demonstratedthe important result that an infinitesimal
unitary transformation can be used to generatechangesin the dynamical
variablesof a quantum system.Correspondingchangescan be made in a
classicalsystem through the use of generatorsof infinitesimal canonical
transformations.Illustrating this correspondencewill make clear that all
possiblechanges,spatial or temporal, in a classicalor quantal system are
'canonical
describablein terms of generatorsof corresponding infinitesimal
transformations'.The deep understandingof physicsafforded by Schwinger's
principle derives from the common basis it provides for both classicaland
quantum mechanicsin terms of such transformations.For this reasonit is
desirable to develop, in parallel, the classical case as is done in the next
section. A reader already familiar with classicalHamiltonian mechanicsor
one eager to proceed with the quantum developmentmay omit this section,
at least on first reading for the latter readers.

8.1.4 Canonical transformationsand classicalmechanics

The concept of a canonical transformation is fundamental to the

Hamiltonian formulation of classical mechanics, the formulation which
362 ACTION PRINCIPLE FOR A QUANTUM SUBSYSTEM 8.1

provides the usual bridge from classicalto quantum mechanics.The essen-

tials of this formulation and of the role played by a canonical transformation
are introduced here.
The classicalequations of motion for any generalizedset of coordinates
can be obtained from Hamilton's principle which statesthat the motion of a
systemfrom time r, to time r, is such that the line integral
I
I

ftt

wn I (8.43)
t. L d t
.Jtl

where the Lagrangian L : T - V, is an extremum for the path of motion.

(There is a growing movement (Feynman 1964)to call the quantity wn the
action integral and to change the name of this principle to the principle
of least action, names which historically referred to the variation of the
time integralof a sum of terms p,q,for all degreesof freedom.We shall follow
these suggestedchanges in nomenclature.)In Chapter 5 it was shown
that Schrodinger's equation for a stationary state is obtained as the
Euler-Lagrange equation in the extremization of the integral E\ltl by finding
that state function ,1, which extremized this integral. According to the
principle of leastaction, the classicalequationsof motion can be obtained in
a similar manner by finding that trajectory connecting the times t, and t,
which extremizesthe action integral. The Lagrangian is a function of the
coordinatesand their time derivatives

L(q,q,t): T - V -l,i*a? - V(q) (8.44)

where q and rf without the subscriptdenote correspondingsetsof coordinates
for all particles.Supposethe proper setsof generalizedcoordinates at /, and
/, to be givenby q(tr)and q(t), respectively, and then imagineconnectingthe
initial and final setsof coordinateswith a space-time trajectory in configura-
tion spaceas pictured in Fig.8.1. There is only one true trajectory and its
form is predictedby the equationsof motion. Alternatively, we may derive the
equations of motion by using the principle of least action to find that
trajectory connecting r, and t, such that the action integral attains its
minimum value. This is clearly once again a problem for the calculus of
variations.
one denotesa point on the true path at any time r by the symbol q(t) and a
point on the varied path by q'(r). Then, as for the variation of a state function
in eqn. (5.69),one has
q ' ( t ) : q ( t l+ 6 q ( t ) (8.45)
where 6q(r)indicatesan arbitrary changein the path at time r. The variations
in the path are to vanish at the two time end-points and the end-points
themselvesare not changed.A possiblesuch varied path is shown in Fig. 8.1.
As detailed in Chapter 5 for the variation of a general integral 1 (eqn (5.68)),
8.1 C L A S S I C A LA N D Q U A N T U \ 1 \ l E C H A N I C S 363

q
q2
v a r i e dp a t h

(a)

(b)

Frc. 8.1. Schematicrepresentationsof actual and varied paths linking initial and final states.In
(a) the time end-points are fixed and the variations in the path vanish at these points. In (b) the
time end-points are varied, as is the path at these points'

the first-order change in W' causedby this variation of q is given by

ftz

+ (ALIAQ)6Q,}dI.
swrz: I I, {GLl0u)6qt (8.46)
Jtt

Using the methods developed in Chapter 5, one rids the expression of

variations in ri, using an integration by parts
d{@Llaq)lq,}ldt : @Llaq)i'6q,* {d@Llaq)ldt}6q, (8.47)

to obtain

SWrz: - d@LlaQ)ldt}6q,dt
+ LlaLlaq)dq)!/,' (8.48)
I \,{@Llaq)
3 6 4 A C T I O N P R I N C I P L EF O R A Q U A N T U M S U B S Y S T E M 8.1

Sinceeach6q, is assumedto vanish at the time end-points,the condition that

I
I
the variation in the action integral vanish,i.e.that q,(r)is such as to minimize
the action, is given by

- dGLlaq,)ldt}6q,d/
6wtz: i"I, {@Llc.q,) : 0. (8.49)

For this resultto be true for any and all variationsin q, which vanish at the
time end-points,it is necessary and sufficient that one obtain the Euler
equation
ALlAq,-d(ALlAilldt:0. (8.50)

For a system with N degreesof freedom, et, i: 1 to l/, this equation is

obtained for each of the l{ coordinatesq,. These are Lagrange'sequations of
motion, the equations of motion for a system obeying classicalmechanics.
Thus, the Lagrangian,which minimizesthe value of the action integral along
the true trajectory betweenthe times t, and f ,, is also the function which
yields the equationsof motion when insertedinto the Euler equation(8.50).
A conceptuallyricher statementof classicalmechanicsis obtained using
Hamilton's formulation. This entails a change in variablesfrom the set of
position coordinatesand their dependentvelocities,which, together with r,
determine the Lagrangian, to a set consisting of the coordinates q and their
conjugatemomenta p, the momentum p, conjugateto the coordinate 4; being
defined as

Pi: ALl04i. (8.s

1)

The ps and qs are treated as independentcoordinatesin the definition of the

Hamiltonian, H(p, q, t). The coordinate transformation from (q, q, t) to
(q,p, r) is accomplishedusing a Legendre transform of the form

H(q, p, t) : L,q,p, - L(q, 4, t). (8.52)

For a conservativesystemfor which the kinetic energy Z is a homogeneous

function of the Qjs, the summation in eqn (8.52) yields 2T and the
Hamiltonian reducesto the total energy of the system

H:27-Q-V):T*V:E. (8.s3)
The classicalequationsof motion in Hamilton's form can be obtained from
a modification of Hamilton's principle as outlined above.The Lagrangianin
the action integral is expressedin terms of the Hamiltonian using eqn (8.52)
to yield the integral
ft'
t -
I : I {L,p,q,- H(p,q, t)}dt. (8.54)
t .
.) tl
8.1 CLASSICAL AND QLIA\TL' \I \I ECHANICS 365

The extremizationof 1.
I a )

d l { L , p , q-, H l P .q . t )i d r : o . (8.ss)
J tr

is carried out betweenfixed time end-points by' making independentvari-

ations in both the p, and q,. This is easily'doneand. after usingan integration
by parts to rid the expressionof the terms involving drir, one obtains
ftz
,t: 0 H t i : p-) 6 q , ( i ; +a H l a q , ) } d r : 0 (8.56)
. J, , L , { 6 n , @ , -
which, for arbitrary valuesof the independentvariations dp, and 5q,, yields
the set of Euler equations
- AHlAq,. (8.57)
Q i : A H fA P , , ii:
Theseare Hamilton's canonicalequations.One may determinethe temporal
behaviour of a classical system with l{ degrees of freedom by solving
Lagrange's A/ second-order differential equations with the constants of
integration being fixed by the 2l{ initial values of the coordinates and
velocities which determine the initial state of the system, or by solving
Hamilton's 2l{ first-order equations for the same initial state.
It is within the Hamiltonian formulation of classicalmechanicsthat one
introducesthe conceptof a canonicaltransformation. This is a transformation
from some initial set of ps and qs, which satisfythe canonicalequationsof
motion for H(p, q, t) as given in eqn (8.57),to a new set Q and P, which
dependupon both the old coordinatesand momenta with definingequations.

Qi : Qi(p, q, t),
( 8 . 5)8
Pt: Pi(P,q,tl,

and which also satisfythe canonicalequationsof motion for a new Hamilton-

ian K(P, Q, t). The name contact transformationis also applied to such a
changeof coordinates.The original coordinatessatisfythe variation principle
statedin eqn(8.55)and, if the P andQ are to be canonicalcoordinates,they
must simultaneouslysatisfy a correspondingvariation principle stated in
terms of P, Q, and K. This doesnot imply that the integrandsfor the two sets
of coordinatesmust be equal, but that they differ at most by a total time
derivative of an arbitrary function F. The function is called the generating
function of the transformation and as demonstratedin Goldstein (1980),it
totally determinesthe transformationequations.
The generating functions are functions of one of the old coordinates and
one of the new. as well as of the time. There are, therefore, four such
generatingfunctions and we are interestedin the one of the form F(q, P, r). In
termsof this generator,the equationsfor obtaining the old momenta and new
366 A C T I O N P R I N C I P L EF O R A Q U A N T U M S U B S Y S T E M 8.1

coordinatesare

P,: iFl?q, and Q i : A Fl } P i . (8.5e)

With the particular choice of L,q,P, for F, one obtains the identity trans-
f or m at ion f or w h i c h p i : P i a n d Q ,: Q i .It i s to emphasi zedthat any of the
canonical transformations, which possibly change the functional form of a
property and are classicalanaloguesof a unitary transformation in quantum
mechanics,leavesthe values of the properties for a system unchanged.This
situation is to be contrasted with the effects of an infinitesimal canonical
transformation, introduced next, which is the analogue of an infinitesimal
unitary transformationof quantum mechanics.
Consider a coordinate transformation which differs from the identity
transformation discussedabove by an infinitesimal amount for each coordin-
ate, i.e.
Qi: Q i* 6 q i ,
(8.60)
Pt: pi * 6pi.

The dq, and 6p,do not representvirtual changeshere as did the variations in
the coordinates6q, rn the variation of the action integral to describea purely
mathematical changein the system'strajectory betweentwo times. Here they
represent real infinitesimal changes in the coordinates and momenta. The
generating function in this case will differ only by an infinitesimal amount
from the one describing the identity transformation and hence can be
written as
F : l,QiPi -t eG(q,P) (8.61)
where again e denotes an infinitesimal parameter of the transformation.
Through the use of the defining equations for this generating function,
eqn (8.59),one obtains

6p,: Pi-Pr: -tAGlAq, and 6q,:Qi-ei:e)GlApi. $.62)

Since Pr differs from p, only by an infinitesimal and the final derivative in
eqn (8.62)already involves e, P, in the derivative of G has been replacedby p,
to retain only first-order terms. It is usual practice to refer to G as well as F as
the generator,and eqns (8.62)define infinitesimal canonical or contact trans-
formations with the generator G.
To illustrate that thesegeneratorscausereal displacementsin a systemwe
take the important example where G(p, q) is set equal to the Hamiltonian
H(p, q) and e to a small time interval dr. The changes induced in the
coordinates and momenta by this transformation are found to be

6 p , : - ( 0 H l } q i ) d:r p , d t : d p , ,
(8.63)
6 q , : ( 0 H l A f i d t: 4 , d t : d q i .
8.1 C L A S S I C A L A N D Q U A N T U I I l N IE C H A N I C S 367

Thus, the transformationchangesthe coordinatesand momenta at the time r

into the values they have at the time r * dr as predicted by Hamilton's
equationsof motion, eqn (8.57).We shall demonstrateshortly that - tidr is
the temporal generatorof an infinitesimalunitary transformationin quan-
tum mechanics.After noting that the variation with time of an observablein
the Heisenbergrepresentationof quantum mechanicsmay be looked upon as
the continuous unfolding of a unitary transformation,Dirac (1930)went on
'In
to state, classicalmechanics the dynamical variables at time t -f 6t are
connected with their values at time r by an infinitesimal contact trans-
formation and the whole motion may be looked upon as the continuous
unfolding of a contact transformation. We have here the mathematical
foundation of the analogy between the classicaland quantum equations of
motion . . .'
There remains but one important concept to complete our summary of the
role of canonical transformations in classicalmechanics,that of the Poisson
bracket. Let F(p, q) and G(p, q) denote two mechanical properties of the
svstem.Their Poisson bracket is defined as
- (aF1)fi(acl0q,)1.
{F,c} : I,l@Flaq)()Glap,) (8.64)

The so-calledfundamental Poisson bracket relations are given by

- 6u.
{ p , ,p i } : o , { q , ,q i } : o , { q , ,p i } (8.6s)

In terms of theseresultsit can be shown that the value of a Poissonbracketis

invariant under a canonical transformation of the coordinates.Like the
correspondingcommutator relationshipsin quantum mechanics,to which it
is related by the expression

{ r , c }: ( - i l D l F , G l ,
the Poisson bracket notation may be used to summarize the fundamental
relations of classicalphysics.The equations of motion may be expressedas
- 0Hl0q,: ii.
{ q , ,H } : A H l A p i : 4 i and { p , ,H } : (8.66)

These are specialcasesof the general expressionfor the time derivative of a

property. The total time derivative of a property F(p, q) which does not
possessan explicit time dependenceis

dFldt : Ii l@FI ail4; + @FI Ap)i,f,

which, using the equations of motion (8.57),can be re-expressedas

dF ldt: I, l@FI 0q)(0HI afi - (aFlafi(aHI )q')l

or
d Fl d t : { F , H } . (8.67)
368 ACTION PRINCIPLEFOR A QUANTUM SUBSYSTEM 8.1

Thus, one has the important result that all functionswhosePoissonbrackets

with the Hamiltonian vanish will be constantsof the motion and, conversely,
the Poisson brackets of all constants of the motion with ^FImust be zero.
I
The properties of the angular momentum are usefully summarized by the
relations (which are more familiar in their correspondingquantum mechan-
ical form than they are in classicalmechanics):the componentsof the angular
momentum vector L are related by

{Lt, Li} : Lx (8.68)

and L2 and one of the components of L, L' n are in turn related by

{L',L'n}:0. (8.69)

Equation (8.65)shows that the Poissonbracket of two canonicalmomenta

vanishes.Thus, combining the resultsof the abovetwo equations,one has the
important result that, if one of the components of L and its magnitude are
chosen as canonical variables, the two perpendicular components cannot
simultaneouslybe canonicalvariables.
Of particular importance to the forthcoming develgpmentis the statement
of the changeinduced in a property A(p, q)by the generatorof an infinitesi-
mal canonical transformation G as expfessedin the Poissonbracket notation.
The change in the property A obtained as a result of an infinitesimal
canonical transformation of the coordinates is given by

5A : A(qt + 6q,, p, + 6p) - A(q,, pt),

A Taylor's seriesexpansion to first-order in the infinitesimals yields

6A : LI,r al aq)6qi+ @Alap,)6pif,

which, using the equationsof transformation(8.62),can be re-expressed
as

Iap)- @Alafi @G
6A: el,{pt1aq)(aG Iaq)}

6A : {A, G}. (8.70)

Thus, in analogy with the quantum result for the change induced in an
observableby an infinitesimal unitary transformation (eqn (8.36)),the change
in a property A causedby an infinitesimal contact transformation is given by
the Poisson bracket of .4 with the generator G. The Poisson bracket of a
constant of the motion with H vanishes,as does its correspondingcom-
mutator in quantum mechanics.Thus, the equation

6H : e{H, Gt,: - (ilh)elH, dl : 0

statesthe important resultthat the constantsof the motion are the generating
functions of those infinitesimal canonical transformations which leave the
Hamiltonian invariant in both quantum and classicalmechanics.
8 .2 THE QUANTUM ACTIO\ PRI\CIPLE 369

The analoguousbehaviourof the Poissonbracketand the commutator has

been used to establish a correspondencebetu'eenclassicaland quantum
mechanics.It is, however,shown in the next sectionfollowing the derivation
of Schwinger'squantum action principlethat the correspondence goesdeeper
and that the analogousbehaviourof the Poissonbracketand commutator is
a consequenceof the properties of infinitesimalcanonical transformations
which are common to both mechanics.

8.2 The quantum action principle

8.2.1 The principle of stationary action

The developmentgiven herefollows closelythat given by Schwinger(1951).A

base vector system la,, t) is specifiedby the eigenvaluesa" of some complete
set of commuting operators d, at the time f. One can consider having two
similarlyconstructedoperator setsat times f , and tz, &,r and &,,r,respectively,
which possessthe same eigenvaluespectrum a, and are, therefore,connected
by a unitary transformationas outlined in eqns(8.28)and (8.29).Thus,
&,2: 0 rr&,r0
,] ,
- 0rzld,,tr), where art : dr2 : cr.
lq,,tr)
The transformationfunction connectinga base state at time r, with that of
another representationderivedfrom the commuting set of observablesf, at
time r, may be viewedas the matrix of C;] in the original eigenvectorsystem
d,r. That is,

( a , , t z l 0 , , I r ) : ( a , , t r l C l r l \ , , t r ) : ( d , , t t l } r r t l f r , t, r ) . (8.71)

As previouslydiscussed, a descriptionof the temporalevolution of a systemis

accomplishedby stating the relationship between eigenvectorsassociated
with different times or, in other words, by exhibiting the transformation
function in eqn (8.71).One may expectthat the quantum dynamicallaws will
find their proper expressionin terms of the transformation function and we
now presentSchwinger'sdevelopment(1951)of a differentialformulation of
this type.
Schwinger'sreasoningis, with minor modification, as follows. The oper-
ator (l ,I describesthe developmentof the systemfrom time r, to time t, and
involvesnot only the detaileddynamical characteristics of the systemin this
space time region,but also the choiceof commuting operatorsd,, and &,,,at
the times r, and tr. Any infinitesimalchangein the quantitieson which the
transformationfunction dependsinducesa correspondingalterationin 01.rt,

5 ( o , , t r l,f. t , ) : ( x , . t r l 6 Ur r 'l f , , r r ) . (8.72)

3 7 0 A C T I O N P R I N C I P L EF O R , AQ U A N T U M S U B S Y S T E M 8 . 2

T he quant i ty i 0 rrd i l rt i s H e rm i ti a n a s a resul t of the uni tary property

C - t : C ' o , C C ' : 1 . B e c a u so e f t h r sp r o p e r t yo n e h a s

dii* : - CaCr
and, by us ingth e i d e n ti ty i n e q n (8 .7 )fo l l o w edby the useof the aboveresul t.
(ii,ti')* : - i,ti C' : + iCdj'
as r equir edf o r a H e rm i ti a n o p e ra to r(e q n(8 5)).A ccordi ngl y,one can defi ne
l p e ra to r6 fl tt
t he Her m it ian i n fi n i te s i ma o

bi'r',: -ihCtrb}rt',

which has the dimensionsof action as indicatedby the i ncl usi onof l ?on the
r ight - hands id e o f th e d e fi n i ti o n .M u l ti p l i c ati on of the above by (i l i l },rl

( 8 . 7 )3

and t he us e o f th i s re s u l t fo r d U ,rt i n e q n ( 8.72)gi ve

d ( r , , t r 1 0 , , 1) , : ( i i h ) ( x , .t 2 1 6 / ti r l [ J , t, r ) . (8.74)

T he Her m it i a n o p e ra to r a i i s th e g e n erati ngoperator of i nfi ni tesi mai

t r ans f or m at i o n s . T h e c o mp o s i ti o n l a w of transformati on functi ons.
eqn ( 8. 24) .imp o s e sre s tri c ti o n so n th i s g e nerator,B ecauseof thi s l aw . the
gener at ingop e ra to rsmu s t s a ti s fya n a d d i tivel aw of composi ti on,that i s.

6 i'l rl ),,,r: ) : J (",, t tl 6 li/trl 0,,t r.>dll,<IJ,,t zl),,,

(a,,t 11 t z)
+ J (2,,trllJ,,tr) d0,(|J,, trl\fr'rrli,,r.) (s.z,s
r

t)Ni'rr: dli'tz f di-r.' (8'76r

The basic assumptionmade by Schwingeris that the infinitesimalgener-

at ing oper at o rr\l l ' r, i s o b ta i n e db y a v a ri a ti onof the quanti ti escontai nedi n
a Her m it ian o p e ra to r N l ' r, w h i c h , b e c a u s eof the addi ti verequi rementgi ven
in eqn ( 8. 76),mu s t h a v e th e g e n e ra lfo rm
ftz

f ,r: v [r]dr (8.71r

I Jtl

where i' and I are referredto as action integral and Lagrange function
oper at or s .r e s p e c ti v e l yT. h e L a g ra n g eo p erator i s an i nvari ant H ermi ti an
function of the field Y and its first derivativeswith respectto spaceand time.
W e m ay s t a teth e re s u l ti n gd y n a m i c a lp ri n ci pl ei n the form of a di fferenti al
s t at em entf or th e tra n s fo rma ti o nfu n c ti o nc onnecti ngtw o coordi nateei gen-
8.2 THE QUANTUM ACTIO\ PRI\CIPLE 371

statesat times t, and tr,

6 ( q , r , t z l Q , tr,r ) : ! l h ) ( q , . . t 2 d r i ' r t l Q , t , tt )

: (ith)(r,,.',s t )0r1,r,,,,,).
(8.78)
IuJ,'
If variationsare effectedin a quantum mechanicalsystem,the corresponding
changein the transformationfunction betweenthe eigenstateslq,r,tr) and
of the action integral
lq,r,ir,) is ilh times the matrix elementof the variation
connectingthe two states.Equation (8.78)will be referredto as the quantum
action principle.
Assuming that the parametersof the systemare not altered,the variation of
the action integral in eqn (8.78)arisesonly from infinitesimal changesof the
setsof commuting observablesat the two times r, and rr. We have previously
seen(eqn (8.41))that such transformationsmay be characterizedin terms of
the generatorsof infinitesimalunitary transformationsF(tr) and F(rr), which
a ct o n t he eigenv e c to rsl q ,r,tr) a n d l q ,r.,tr)' C o mpari ng eqns (8' 41)and
(8.78)one obtains for such variations

6trr: Fft) - F(r,), (8.7e)

which is the operatorprincipleof stationaryaction It statesthat the action
integral operator is unaltered by infinitesimal variations in state functions
betweenthe times f , and f r, being affectedonly by the action of generatorsat
the two time end-points.The principle of stationaryaction is the operational
statementof the theory. It statesthat the variation of the action integraldoes
not vanish as in Hamilton's principle,but insteadequalsthe differencein the
effectsof infinitesimal operators acting at the two time end-points.This result
demands that the variation of the action integral be generalizedto include
variations of the state functions and of the time at the time end-points. The
principle of stationary action then implies the equations of motion of the
system as obtained in Hamilton's principle, and the end-point variations
define the generators of the infinitesimal canonical transformations which
induce changesin the dynamical properties of the system.In this way a single
dynamical principle, the principle of stationary action, recoversnot only the
equation of motion, but definesthe observables,their equationsof motion,
and the commutation relations.

8.2.2 Applying the principle of stationary action

The first step in applying the principle of stationary action is to generalizethe

variation of the action integral to include a variation of the time end-points
and to retain the variations at the end-points in order to define the generator
F(r). We shall expressthe Lagrangianoperator in terms of the completeset of
372 A C T I O N P R I N C I P L EF O R A Q U A N T U M S U B S Y S T E M 8.2

commuting generalizedcoordinate operators denoted by Q" introduced in

eqn (8.25),and their time derivativesri,,,

ftz
o-rr: 9 (a,,, t)dt.
Q,,, (8.80)
),,
While this is a formal approach,it maximizesthe correspondencewith the
classicalcase.In fact, if one assumesthe coordinateoperatorsand their time
derivativesto be instead their classicalanalogues,one obtains the corres-
ponding resultsfor a classicalsystem,therebyillustratingthe generalityof the
method and the similarity in the structureunderlyingthe two mechanicsthat
is made evident using this approach.The samevariation is performedwithin
the Schrodingerrepresentationin the following sectionin the extensionof the
principle to a subsystem.
The differencebetweenthe generalizedvariation of the action integral and
the restrictedone is illustrated by comparing the form of the varied paths for
the classicalspace-time trajectoriesdepicted in Fig. 8.1(a),(b).The varied
path may end at differenttimes and the variation in the coordinatesdoesnot
vanish at the time end-points. The generahzedvariation corresponds to
retaining the first-order changesbetweenthe Lagrangian evaluatedalong the
varied path, integrated betweenthe varied time limits, and its unvaried value
integratedbetweenthe unvaried time end-points.That is,

f t>* 6tt
t -
6f t z :
I 9 (A,,i 6i.u,6,,t 66,,,
t + dr)d(r+ dr)
Jt1+dt1

-l ftt
9 ( Q , ,q, , t ,t ) d t . ( 8 . 8| )
Jtr

To first-orderin the infinitesimalsthis reducesto the changein I along the

varied path betweenthe unvaried time end-points and the unvaried integrand
times the variations in the time at the two time end-points,

ftz
6{tt I 69 (e,,,q,,,t)dt * I (Au,Q,,,t)6tl',1 ( 8 . 8)2
I
Jtl

The variation of the Lagrangian operator is formally identical to the varia-

tion of the classicalaction integralas developedin eqns(8.45)-(8.48). We ma1
take over the final result in eqn (8.48) completely in its corresponding
operator form, retaining in this casethe terms involving 6q,, at the time end-
points as these variations are no longer required to vanish. The addition of
this result to the end-point variations irr eqn (8.82)yields, for the generai
8 .2 THE QUANTUM ACTIO\ PRINCIPLE 1 1 4
J I J

variation of the action integral operator.

ftz

6 f i r z : | r c v t e Q-, ,d { c Y t q , ,d1t' ,6Q,,dt

Jtr

+ 91q,,.
+ {@sle6,)d4,, Q ,,,t) at| l ' ,' ,. (8.83)

The end-point variations can be recast into a more useful form. One first
notesthat the completechangein a coordinateoperator at a time end-point,
denoted by the symbol Lq,,, can ariseas a result of a variation in the operator
or as a result of a variation in the time
L 1 ' n : 6 4 "+ 4 " 6 t . (8.84)
An end-point variation in eqn (8.83),expfessedin terms of Aq,,, is given by
- e (4u,Q,,,t)\ 6t.
@sla6,)til,,- lav 1a$,;6,, (8.8
s)
In accordancewith the classicaldefinition of the Hamiltonian, eqn (8.52),and
recalling that the derivative of the Lagrangian with respectto a velocity (eqn
(8.51))is the momentum conjugate to the correspondingcoordinate, the
Hamiltonian operator is definedas

fr(Q,,,
a91a$,,y - 9(q,,,6,,,
: 691a$,,)6,, t). (8.86)
By using this result in eqn (8.85),an end-point variation arising from the
generalizedvariation of the action integral operator at a time r, a quantity
denoted by the operator F1r;,is given by
- fr(q,,,
F(r): {@s1a6,,)LQn as laA,)l.t}. (8.87)
Incorporating this result into eqn (8.83)yields,for the generalizedvariation.

ftz
6t,,, I
II
- d@sI aA,,)ldt}
{@slAA,,) 6q,,dt+ Fft11- F(r,). (8.88)

Comparison of this result with the operator statementof stationary action

given in eqn (8.79) shows that the quantity multiplied by the arbitrary
variationsrn Au must vanish as it doesin the restrictedvariation of the action
integral,thereby yielding the equation of motion

Ag lAA,,: d@g lA6,)ldt

and the statementof stationary action
6fr, : F(t) - F(r')' (8'89)

The comparison also leads to the identification of the generatorsof the

infinitesimal unitary transformationsof the principle of stationary action
374 ACTION PRINCIPLEFOR A QUANTUM SUBSYSTEM 8.2

with the terms derived from the end-point variations rn Au and in the time as
given in eqn (8.!7) for the operator F(r).
A generatorF(r)may be decomposedinto two parts:one which determines
the purely temporal changesin a system,the generator - fi6t, the other
deErmining all possible spatial changes in a system, the generator
(Ag lAq,)5Q,,.Theuseof thesegeneratorsin the derivationof the Heisenberg
equationsof motion for the observablesand of the commutation relationsare
illustratednext. Theserelations,togetherwith the equation of motion for the
system, provide a complete description of the properties of a quantum
dynamical system,and all are derived from a single dynamical principle.
For a purely temporal change the complete variation in an observable
6), L(h, equals zero and, from eqn (8.84),the whole of its change results from
the change in the time end-point. By eqn (8.84)this change is given by
6ri : - ,iAt. According to eqn (8.36),the changein the observablecDcaused
I
I
by the operator - Hdt acting as the generator of an infinitesimal unitary I

transformation is

6rh: - (ilh)yfi,,;1at: - Aat (8.e0)

6 : (ilh)LH,(i)), (8.e1)
which is Heisenberg'sequation of motion for the observablerD.If one takes If
in this equation to be the classicalHamiltonian and multiplies the com-
mutator by (ih), one obtains the expressionfor the time rate of change of the
observablecDin the Poissonbracket notation (eqn (8.67)).As in a classical
system,the Hamiltonian is the variable conjugate to the time and the action
of the generator - fiat is to propagate the system through time.
When there is no changein the time end-point, L4.":^6Q,,and, from eqn
(8.87),the generatorof a pure displacementis given by (Ag lA4,)54u. We may
derive the commutation relations previously introduced in eqn (8.15) by
consideringthe effectof this generator on the coordinate operator f,, and the
conjugate momentum 1",. According to eqn (8.36)the change in f", induced
by this generatoris given by

6Q,:01ilrcYl\A,)f.i,,,
A,f
: (ilD@slr,i,)16r,,,
4"1+ (ilh)Kasli,d,),i,l6r,, (8.e2)
where the subscript r has been dropped, since all operators refer to the same
time, and the product in the commutator has been expanded.The variation
dri, is arbitrary and can be taken to be tQ, and the first of the commutators in
the expandedproduct yields,as a component of the simplestsolution to eqn
(8.92),the result

IQ,,Q,]: O, (8.93a)
8.2 THE QUANTUM ACTIO\ PRI\CIPLE 375

which is the first of the commutator relations.This leaves

64,:\lh)rcs -d,.a,f
d1,,. (8.e4)
Equation (8.94)requiresthe satisfactionof the secondcommutator relation

leg lAq,,4"l: lp,.4,f : - ih6,, (8.e3b)

where aS 1a$,hasbeenidentifiedwith the conjugatemomentum P,. Follow-
ing Roman (1965),we apply the samedisplacementof the coordinate to the
momentum operator which must yield zero and one has

l,l : 0
6i,: (ilh)1i,54,,
: 0lh)p,16A,,, P,f6q,: 0.
p,)+ (ilh)1p,, (8.e5)
Again taking 6q, eqval to eQ,,the first of the commutators in the expanded
product yields - tp,. The resulting operator equation must be valid for any
state including one for which p,lY):0, which, for arbitrary 6Q,,yieldsthe
last of the Heisenberscommutator relations
[ P , , P r ]: o . (8.93c)

Thus, Schwinger's principle of stationary action yields the equation of

motion for the system,Heisenberg'sequation of motion for each observable,
and the commutation relations. The Heisenbergequations and the com-
mutator relationsare obtained through the use of generatorsof infinitesimal
canonical transformations, a fundamental set of operations which are intro-
duced by identifying them with the end-point variations in a generahzed
variation of the action princip^le.The generatorsdefine the conjugate pairs of
variables,momentum ag lAq, with coordinate Q,, and the Hamiltonian H
with time. All physical observablesare constructed from the canonically
conjugate pairs of variables and the displacement and temporal operators
thus generateall possiblechangesin the dynamical variablesof a systemand
predict the consequencesof these changes. All of this is obtained for a
classical as well as for a quantum system with the fundamental Poisson
bracket relations (eqn (8.65))replacing the commutation relations. It is clear
that Schwinger'svariational principle,by providing a common mathematical
foundation through its use of infinitesimal canonical transformations,
provides a single dynamical principle from which classical and quantum
mechanicscan be derived.
A form of the principle of stationary action that will prove of great use in
later applicationsis obtainedby expressingthe principle of stationaryaction
(eqn (8.79))for an infinitesimal time interval, that is, as a variation of the
Lagrangefunction operator. Equation (8.79)is re-expressed as

6tr, : @Frcoat. (8.e6)

l,',,'
376 ACTION PRINCIPLEFOR A QUANTUM SUBSYSTEM 8.3

By dividing both sidesof eqn (8.96)by tz - r, and subjectingthe result to the

limit of At --+0, one obtains an expressionfor the principle of stationary
action in terms of a variation of the Lagrangefunction

6gltl : dFldt. (8.e7)

By using Heisenberg'seqn (8.91)for the time derivativeof an observable,the
result can be equivalentlyexpressedas

a v s t l : $ l h ) l nF, l + 0 F 1 a t (8.e8)
where the term aFlat is included for those casesin which F possesses an
explicit time dependence. The ffict of an infinitesimal canonical nans-
formation on the Lagrangefunction is equal to the time rate of changeof the
generator of the transformation.
The stageis now set for the demonstrationthat the principle of stationary
action can be extended to a particular class of subsystemsof some total
system,that is, to a particular classof open systems.It will be shown that the
variation of the action integral for such a subsystemrecoversthe primary
resultof Schwinger'sprinciple-that the changein its action integralis totally
determined by the generatorsof infinitesimal unitary transformations acting
at the two time end-points,with no contributionsfrom the interveningtimes.

8.3 Atomic action principle

8.3.1 Propertiesof the quantum mechanicalLagrangian

The Schrodinger representationwill be used in the development of the

subsystemaction principle(Baderet al.1978;Srebreniket al.1978;Bader and
Nguyen-Dang 1981).The action integral for a total system, l/rrrlyl, is
defined as the time integral of the Lagrangian glY,rl, which is in turn
obtained by integration of the Lagrangian density Llv, vy, \p, rl over the
coordinates of all of the particles in the system

7rt2lY): f" O[Y,r]dr: f" O,lOrUY,VY,v, rl. (8.ee)

J,, J,, J
The quantum mechanical Lagrangian density for a systemof many particles
interacting via a many-particlepotential energy operator 2 is
- W*V)
LlY, VY,W,tf : (ihl2)(Y*.P
- (h2l2m)I,v,Y*.v,Y
- tv*v. (9.100)
Thatthisis thecorrectLagrangian
density
is demonstrated
by showing
that
the variation of the resultingaction integral with respectto Y and Y* yields
8 .3 ATOMICACTIONPRI\CIPLE 377

Schrodinger'sequations,

ihY : frY and - ihY* : .dY* r f: - ( i 2 t u l , Y f + V

(8.101)
There is no lossin generalityby consideringa Lagrangiandensityfor a single
particle in this demonstration.The variation of il'rrlYf, using the methods
developedin eqns(5.66)-(5.69), after integrationby parts to rid the expression
of terms in dVY and d'P, is
rt' r (
1eY
6{rr[Y] : I dt I a'{aUaY - v . G L r *-Y! d) + l r *
-v'@Llav'
J I
J,, draYj
Ptt P
t

+ | dr Qas?uavY).ni)Y
Jt, J

+ | orturlu*;avlii + cc. (8.102)

J
The Lagrangran, as are the functionals considered in the derivation of
Schrodinger's stationary-state equation, is a functional of VY, and the
associatedsurfaceterm in eqn (8.102)(compareeqn (5.59))vanishesbecause
of the natural boundary condition (eqn (5.62))that VY'n : 0 on the bound-
ary at infinity or by demandingthat dY vanish on the sameboundaries.The
variationsat the time end-pointsalso vanishin this restrictedvariation which
is the analogueof Hamilton's principleor the principle of leastaction and the
resulting Euler-Lagrange equation, obtained by demanding that the varia-
tion in VrrlYf vanishfor arbitrary dY, yieldsSchrodinger'seqn (8.101).This
is easily verified by evaluatingthe partial derivativesof I in eqn (8.102)in
those terms multiplied by dY.
We may retain a formal analogy with the variation of the classical
Lagrangian (eqns (8.48)and (8.49))and with the variation of the quantum
Lagrangianintegraloperator (eqn(8.83))bydefiningthe functionalderivative
of I with respectto Y (and correspondinglyV*) to be

- v' @LlavY)}.
av 1av: {aLlaY (8.103)

The variation of I with respectto \i/ is already of the sameform for both the
classicaland quantum cases.
The Lagrangian density,eqn (8.100),exhibits a most important property
when the statefunction Y satisfiesSchrodinger'seqns (8.101).By using eqns
(8 .1 0 1)t o r id eqn ( 8. 1 0 0 )o f th e te rm s i n tP a n d .P * , the Lagrangi andensi ty
reduces to the difference between the two kinetic energy expressionsas
introduced in eqn (5.46) in terms of the identity involving the quantity
V21y*Y). Denoting by Z0 the particular form assumedby the Lagrangian
density when Schrodinger'sequation is assumedto hold, as opposedto its
 378 ACTION PRINCIPLEFOR A QUANTUM SUBSYSTEM 8.3

general functional dependenceon Y, VY, and V, one has

Lo : - (h2l4m)
I, {y* vl y + yvi y* + 2v iv*. V,V}, (g.104) II
which, using eqn (5.46),can be re-expressed
as I
Lo : _ (h 2 l 4 m)I,v,r(Y * Y ). (8.10s)
The sameidentity involving the Laplacianof Y*Y was usedto transform the
integrand in the variation of the functional ElY,Ol. Equation (5.83)details
the result of integrating one of the terms appearingin the sum found in eqn
(8.105);the use of Gauss's theorem transforms each term into a surface
integral of the quantity Y*v,Y.n and its complex conjugate,terms which
vanish at the boundary of the system at infinity. Thus, at the point of
variation where Schrodinger'sequation applies,the Lagrangian integral go
vanishes,

soyy,tl: - f t 2 1 4 m ) J d.r.,. J d r " L , v ? ( y * y ): g . (8.106)

Becausethe many-particle Lagrangian density Z,0reducesto a sum of single-
particle operators, one may define an effective single-particle Lagrangian
density 9o (r, r) by the usual recipe of summing over all spins and integrating
over the coordinatesof all the electronsbut one, a step which expressesthe
result in terms of the chargedensity p(r, /) as
go (r, t) : - th2 l4mV)y2p(r, t).
[dr' Lo : (8.107)
The vanishing of the Lagrangian integral goly,r] is seen to be a conse-
quencd of the vanishing of the flux in the gradient vector field of the charge
density at the infinite boundary of the system
golY, tf :
ldrgo(r, r)
: - (h2l4mN)$dS1r,
t)Yp(r, r).n(r,/) : 0. (8.108)
Thus, the action integral for a closed quantum system described by
Schrodinger'sequation vanishesfor any time interval Ar becauseof the zero-
flux surfacecondition that

V p (r).n (r) : g a t e v e ry p o i nt on the boundary. (g.109)

It is this property that is common to the correspondingLagrangian and
action integralsfor a quantum subsystem,the chemicalatom.

8.3.2 Atomic action and Lagrangian integrals

The subsystem or atomic Lagrangian integral is defined by the standard

mode of integration used in the definition of the subsystem functional
E lY, Q] for a stationary state and for the definition of subsystemproperties.
8.3 ATOMICACTIO\ PRI\CIPLE 379

That is,

glY,c), t ) : forJo,t(v.rY.Y, r) (8.110)

Jo

Correspondingly,the atomic action i n te g ra li s

{rrlY ' Q] : j,' orv lY,{r,r l . (8.11l)

When Y and Y* satisfy Schrodinger'sequations.the atomic Lagrangian

integral reducesto

golY. o. rl : ft' |+mN d rV 2p (r, t)

[,

: - (h2lumxt$
astr,r)Vp(r,r)'n(r,r). ( 8 . I1 2 )

Since the zero-flux boundary condition (eqn (8.109))is also satisfied by an

atom, that is, by a quantum subsystem,the atomic action and Lagrangian
integralsvanish as they do for a total closedsystem.Indeed,one may view the
vanishing of the action integral over some total systemas being the result of
the action integral vanishing separatelyover the space of each atom in the
system.
As a systemin a given quantum state changesand evolveswith time under
the action of the generator - HAt,the atomic surfacesevolvein a continuous
manner and the property of exhibiting a zero flux in Vp(r) is continuously
maintained. Thus, the atomic action integral will always vanish
f,, f
f or r lY , O l : - (h ' l a ml ,{ l d rQd S Vp (., r)' n(r, r) : g (8.113)
I
Jt, J

for any time interval, finite or infinitesimal.One may imagine an open system
with an arbitrarily definedsurfacewhose Lagrangian integral might vanish at
some time r becauseof the vanishing in the net flux of Vp(r, r) across its
surface(eqn (8.112)).However, in general,as the systemchangedwith time, a
net flux in Vp(r, r) through the surfacewould arise and the action integral for
sometime interval would not vanish as it must for the total system.Thus, the
satisfactionof eqn (8.113)for all time intervals,as found for the total system,
demandsthat the open systembe defined by a surfacethat itself evolveswith
time in a manner determined by a property of the open system,this property
being one of azero flux in Yp at every point on its surface(eqn (8.109)).
The condition,startedin eqn (8.113),that the atomic action integralvanish
for all time intervals may be taken as the quantum definition of an atom. It is
a direct consequenceof the topological definition of an atom as the union of
-T8OACTION PRINCIPLEFOR A QUANTUM SUBSYSTEM 8.3

an attractor and its basin and is the basis for the particular variational
propertiespossessed by an atom consideredas an open quantum system.

8.3.3 Variation of the atomic action integral

We are now in a position to perform a generalizedvariation of the action

integral for an open system to demonstrate that Schwinger's principle of
stationaryaction can be extendedin such a manner as to provide a quantum
definition of an atom in a molecule.We shall be consideringthe changein the
atomic action integral w , r.lY , ol of eqn (8.I 11) ensuingfrom variationsdY
in the state function Y such that the following conditions are fulfilled.

1. The trial function (D,where @ : Y + dY, and its first and secondderivat-
ives vanish whenever an electronic position vector is of infinite length.

2. rn terms of o, a region o(<D,r)can be definedthat is bounded by a zero-flux

surfacein Vp, where

Vpr(r, r)'n(r, r) : 0 for all points on the surfaceof Q (8.114)

where the trial density is defined as

Po(r,r) : NIdr'@*(., {, t)@(r,l, t). (8.1 5)

Moreover, it is required that, as O tends to Y at any time /, O(O, r) is
continuouslydeformableinto the region O(r) : Q(Y, r)associatedwith the
atom. The region Q(<D,r) thus representsthe atom in the varied total
system,which is describedby the trial function O(r, r', /) just as O(r)
representsthe atom when the total system is in the state described by
Y(r,r', r).
Requiring the fulfilment of condition (2) amounts to imposing the varia-
tional constraint that the divergenceof Vp. integrate to zero at all stagesof
the variation. i.e. that

Io : : o
Y2P'dr (8.1 6)
Jt,',,,
for all admissible<Dand for all r, which implies

dlY : : o.
,{1,,,o'p(r,t)dr} (8.117)
To impose the variational constraint given in eqn (8.117)and thereby
define a particular class of open system, one must vary the surface of the
subsystem.This requirementnecessarilyleads to the relaxation of the usual
variational constraintthat the dY vanish at all boundariesof the systemand
at the time end-points as is done in the restrictedvariation of the action
8.3 ATOMIC ACTION PRINCIPLE 381

integral in Hamilton's principle to obtain the equationsof motion. Such a

variation was performedabove in Section8.3.1in the restrictedvariation of
the quantum mechanicalaction integral to demonstratethat the quantum
mechanicalLagrangian density defined in eqn (8.100)does indeed lead to
Schrodinger'sequation of motion. The reader is aware that it is the gen-
eralization of the variation of the action integral obtained by the removal of
the constraintsthat the variationsin Y vanish at the time end-pointsand by
the additional variation of the time end-points themselvesthat leads to
Schwinger'sprinciple of stationary action. Thus, the variation of the atomic
action integral with the necessaryretention of the 6Y on the boundaries and,
hence,at the time end-points perforce leadsto the same generalizationof the
variation principle. What is now demonstratedis that, if the subsystemaction
integral is restricted to regions satisfying the above variational constraint,
this generalized variation also leads to the same physical principle of
stationary action.
The techniquesof the calculusof variations, now familiar to the teader, are
applied to the expressionfor {rrlY,Ol (eqn (8.111))with the Lagrangian
densityas definedin eqn (8.100).Includedis a variation of the time end-points
as detailed in eqns(8.81)and (8.82),and a variation of the surfaceof the
subsystemin the manner previously detailed in eqn (5.74).Variations in Y
and Y* at the time end-points are also allowed in this generahzedvariation of
the action integral.Following thesesteps,one obtains the following explicit
results for variations in Y and the complex conjugates of these (cc) for
variations in Y*

F r - F F
I ' z | |
- -
6l//t2lY,el : I ar l or lot'{( ihY* Hv*)dY+ cc}
Jr' Jo J
f t r f l
Y*dY' n(r)* ccl
+ | ar Qos1o,ryI dr't - h212m1{v
Jr' J J
1,, f
r) 9(r)615(A,r) + cc
+ | dr QOS(O,
t . I
.J .l .)

( " f f
r) | dr'l( - ihl2)(as(o,
+ | arQosto. *6v + cc)
r)l6t)Y
J,, J J
( | r __ -^ll"
+ ]$hlz)| o. I dr'Y*dY+ cc+ glY, o' rJdrf l
t Jn J )'",r.,
18)

As in previous chapters,the position vector of the electron whosecoordinates

are integrated over Q is denoted by r. The first surfaceterm in eqn(8.118)
arisesfrom the integration by parts used to rid the expressionof variations in
-rEl A C T I O N P R I N C IP L E F O R A Q U A N T U M S U B S Y S T E M 8.3

f ,Y. terms of the form

@ L l i ] v J ) d v:, Y- # v , Y * d v , Y . ( 8 .1 e )

This stepwas previouslydetailedin the variation of I llt, ol in eqn (5.76)and

the identity precedingit. As describedthere,all the resultingsurfaceterms of
V,Y* 'n vanish save the one arising from the gradient operator for the
electronwhosecoordinatesare integratedover the (finite)surfacecoordinates
of the region C),as only for this electronis the position vector in the surface
not of infinite length..Theintegration by parts used to rid the expressionof
the term involvine dY

QLIAY)6Y: (ihl\Y*d\i' (8.120)

requires specialcomment.This step is accomplishedusingthe identity

*6Y\
r'(iht2)Y
l,',,*#
{J,..I.
: (ih o,
t41,,' + Y*d\i'
{ I"o,J0,,1.i,*dY )
+ asto, r)Jdr'(as(o,
r)/dr)Y-u*l ( 8 . 1r2
)
f
The time derivativeof the surfaceappearingin this expressiondenotesthe
time rate of changein the surfacebounding the unvariedregion O(Y, r). That
is, the time derivative of an integral with a time-dependentboundary must
include a contribution arisingfrom the changein the boundary with time. It
is t he us eof t he i d e n ti tyi n e q n (8 .1 2 1 )th a lte a dsto the i ncl usi onof the terms
involving Y*6Y at the two time end-points.Since the time end-points are
themselvesvaried, one also obtains the term glY,Q,tfdt at the time end-
points, as noted previouslyin the generalizedvariation of the action integral
operator.

8.3.4 Principle of stationary action in the Schrodingerrepresentation

Beforepursuing the variation of the atomic action integral, it is helpful to first

recoverthe statementof the principle of stationary action in the Schrcidinger
representationfor the total system.If one setsthe boundary of the region O at
infinity in eqn (8.118) to obtain the variation of the total system action
integral ftrlYf, and restrictsthe variation so that dY vanishesat the time
end-pointsand the end-pointsthemselvesare not varied,then only the terms
multiplied by the variations dY in the first integral on the right-hand side
remain.The Euler equation obtained by the requirementthat this restricted
8.3 ATOMIC ACTIO\ PRI\CIPLE 383

s (8.101).
va ri a tionof t he ac t io n i n te g ra lv a n i s hi s Sc h ro d i n ger'eqn If the end-
point variations are restored for the variatron of l/ t2lV), one obtains, in
addition to the equation of motion, Schwinger'sprincipleof stationaryaction
stated in the Schrodingerrepresentation
6frr[Y] : {(ihl2)!dtY*dY+ cc + YlY, tl6t}li1 (8.t22)
Equation (8.122)states that the action integral respondsonly to changes
induced in the systemat the time end-points.In analogy with the classical
result given in eqn (8.52)which enablesa changefrom the coordinatesQ,4, t
to q, p, t as employed in eqn (8.87) in the formulation of the end-point
operatorsin the variation of the action integraloperator,one may expressthe
Schrodinger-basedaverageHamiltonran.Tf [Y] as the Legendretransform of
I lY, tl,
- 'p*Y)
.f lYl: |{Jdr(v*tiw + (HY*)Y)} : (ihl2)[dr(Y*\i'
- g [Y, r] - (h2l4m)
JarV2p1rl. (8.123)
This transformationenablesone to rewrite eqn (8.122)in the form
6frr[Y] : {(ihl2)[drY*AY* cc- tr lY],lr] ll;
: F(t) - r(rr) ( 8r. 2 4 )
where the symbol AY denotesthe generalvariation
AY:av+ti'ar.
The principle of stationary action identifiesthe end-point variations u ith
the generators of temporal and spatial unitary transformations. In the
Schrodingerrepresentationone obtains the averagevalue of the generator
F(r),as denotedby the symbol F(r). The expressiongiven here for F(r) can be
compared with that obtained by the variation of the action integral operator
in eqn (8.87).For a temporal change,AY : 0, dr is arbitrary and the change
in 71.r,F|f] reducesto the differencein the averagevalues of - H6t at the
two time end-points, thereby identifying the generator of temporal develop-
ment with the operator - H6t. Similarly, for a purely spatial change,dr : 0
and
6frr[Y] : ]{Jdr[Y*Fv + (rY)*Y] ] li? (8.125)
where in this situation AY : dY : ( - ilh\FY and F is an infinitesimal
Hermitian operator. It may be a function of the coordinates (inducing a
gauge transformation) andfor of the momenta (inducing a translation of
coordinates).If both sidesof eqn (8.125)are divided by t t - t , and the result
is subjectedto the limit Ll -- 0, one obtains, to first-order, an expressionfor
69lY, tf,
691Y, t) : i{d(F)ldt * cc}, (8.126)
384 A C T I O N P R I N C I P L EF O R A Q U A N T U M S U B S Y S T E M 8.3

which, through Heisenberg'srelation for the time rate of change of the

averagevalue of an observable,may be re-expressedas
6 s l Y , r l : ( +{)( i l h ) < l H , r l + . g F l a*t Sc c } . (8.127)
)
Equations (8.126)and (8.127) are identical in form to the corresponding
results obtained for the variation of the Lagrange function operator in
eqns(8.97)and (8.98).They are variational statementsof the Heisenberg
equation of motion for the observableF in the Schrodinger representation.
When Y describesa stationary state

ih AY li t: .fiV : EY(t, r) : exp(iEtlh)\fu(r), (8.128)

the Lagrangianintegralreducesto the form
9lY , tf : - (Y, Y>(8 bll - E) (8.12e)
where 8blll is the energyfunctional as defined in eqn (5.63),but used here for
the many-electroncase.It differs from the functional I [/] in eqn (5.58)only
-
in that the normalizationconstraintis handled by the factor <rlr,,lt> 1 rather
than by an undeterminedmultiplier. The variations in this casewill be of the
form dY(2, r) : exp(iEtlh)6rlt(r)and it follows that
- <r1,,{)-'ag
681,,1/l: [Y,r]. (8.130)
Thus, for a stationary state with a normalized state function /, eqn (8.127)
assumesthe form

6sbll: - +{(ilh)<lH,
Fl) + cc}: s, (8.1
31 )
which may be equivalently stated in terms of the functional glltl as
6El,l'l: - +{(ilh)(tti,
Fl) * cc}: 6 (8.132)
as previouslygiven in eqn (5.101).The derivationof theselatter two equations
from the principle of stationary action substantiatesthe claim made at the
conclusion of Chapter 5 that eqn (5.101),and hence eqn (8.131),are state-
ments of this principle for a stationary state. As in the time-dependentcase,
the principle yields the equation of motion for the state function in addition
to the variational statementof the equation of motion for an observableF,
which, for a stationary state,is the hypervirial theorem.

8.3.5 Atomic statementof the principle of stationary action

We return to eqn (8.118)for the case where Q refers to a region of space

bounded by a surface of zero flux in the gradient vector field of the charge
density.In addition to retainingthe terms at the time end-pointswhich arise
from the generalization of the variation and which are present for a total
systemas well, we must considerthe terms arisingfrom the non-vanishingof
8.3 A T O M I C A C T I O N P RI \ C I P L E 385

dY on the surfaceof the subsystem,the changeof this surfacewith time, and,

finally, the term arising from the variation of the surfaceof the subsystem.
The symbol dS(O,r) in this latter term denotesthe infinitesimalnormal shift
of a surfaceelementcentredon r, as conditionedby the continuousvariation
in the admissibletrial function (D: Y + dY, in a small neighbourhoodof the
statefunction Y. This continueddevelopmentof eqn (8.118)istaken when the
variation is complete.That is, we are at the point of variation, Schrodinger's
equationapplies,and the expressionfor 6Wrr[Y, O] is consequentlyreduced
to the surfaceand time end-point terms.It was shown in eqn (8.108)that the
Lagrangian integral vanishes when Schrodinger'sequation is satisfied.We
have already demonstrated in a formal sensethat the presenceof this term
does indeed define - frat as the generator of a temporal transformation.
Operationally, this role is taken over by eqn (8.127)in the Schrodinger
representationand there is no loss in generality to simply set the Lagrangian
integral at the two time end-points equal to zero for the remainder of the
discussion.
Another consequence of the applicability of Schrodinger'sequation to the
expressionfor 6f ,r[V, O] at the point of variation is that the Lagrangian
density appearing in the term arising from the variation of the surfaceis, as
shown in eqn (8.107),proportional to Ytp(r), the Laplacian of the charge
density.Thus, at the point of variation,the expressionfor the variation of the
atomic action intesral is

- h 2l 2 m ) ( v v * 6 Y .n ( r )+ c c )
6{rr[Y, O] :
f orrn,,)ld,'rr
l,',,*{
- (ihl2)(As(o,
r)/dr)Y*dY* ccl

+
f
f
asto,r)( - h2l4mN)v2p(r)drs(Q,
r) +..1
( f f Jl"
+ 10h12) | o' I dr'Y*dY + cc| | . (8.133)
I Jo J Jl',
It is through this property of the Lagrangian density at the point of variation,
as it appearsin the surfacevariation, that the consequencesof the constraint
g i ve n in eqns ( 8. 116 )a n d (8 .1 1 7 )a re i n tro d u c e d i nto the vari ati onal ex-
pression.The same procedure was followed in the derivation of the principle
for a stationary state in which casethe integrand of the functionalEbl/,O]
was also found to be proportional to the Laplacian of the charge density at
the point of variation. Here, as there, the imposition of the variational
constraintis accomplishedusing the identity given in eqn (8.134).This result
is obtained when one carriesout the variation of the Laplacian integral for
the region O, including a variation of its surface,in the manner prescribedin
386 A C T I O N P R I N C I P L E F O R A Q U A N T U M S UB S Y S T E M 8.3

e q n( 5 . 7 4 )

ar{tn'rc*rulJ : {n'A*x)!n
a.v'e(,)} drd* {V2p(r)}

+ ft2l4mrDasto,r)Yzp(r)6*S(Q,r).
f (8.134)
The variation of the integrand indicated in eqn (8.134) was considered
previously and the result is given in eqn (5.91).As emphasizedthere, the
Laplacian is a divergenceexpression-its variation yields only a surfaceterm;
thus the imposition of the constraint does not affect the Euler-Lagrange
equationsthat have beenobtained at this point. According to the constraint,
eqn(8.117),the sum of variations of the integrand and surfacemust equal
zero and, thus, the result of imposing the constraint is to replacethe surface
variation with the term arising from the variation of Y2p(r),i.e.
- th2l+mx) r) : (h2l4m)$Os1O,
r)Y2p(r')6rS(Q, r)Jdr'{(vv*)dy
$ds1o,
+ Y * d V Y ) ' n ( r*) c c . ( 8 . 1 3 5 )
Substitution of this re s u l ti n to e q n (8 .1 3 3 )y i el ds

6il'rr[Y, Q] : - (h2 .s(o,r)fdz,r(vy*)dy

t4m){J,:.'f
- Y*V(6Y)I'n(r)
+ ,.1
(
+{ - (thlz)l,"ar
osto, ;lsl;,)y*6*
.)Jdr'( **
(. f )
(
+ { (ihl4loarJar'Y*dY
. ..1 (8.136)
t [:
One recognizesthe first term in eqn (8.136)as the variation in the quantum
mechanicalcurrent density (eqns (5.94)and (5.95)).It is obtained by com-
bining the surfaceterm arising from the variation with respectto VY with the
surface term arising from the imposition of the variational constraint,
eqn (8.135).Thus the uariation of the surfaceof the subsystemtogetherwith the
restriction that the subsystembe an atom boundedby a zero-flux surfaceceuses
the quantummechanicalcuruentdensityto appearin the uariation of the action
integral, a term whose presenceis a necesscryrequirementfor the proper
descriptionof the propertiesof an open system.It is now demonstrated that
eqn (8.136)is the atomic equivalentof the principle of stationary action.
The variation of Y is identified with the action of a generator of an
infinitesimal unitary transformations on the corresponding state function, a
8.3 ATOMIC ACTION PRINCIPLE 387

fundamental identification in the development of the principle of stationary

action.Following the sameprocedurethat was usedto obtain the principleof
stationary action for the total sy^stemin the Schrodinger representation
(eqn (8.125)),we set dY : ( - |IUFY where F is an infinitesimal Hermitian
operator. The operator F refers to the electron whose coordinates are
integrated over the region O. This substitution yields

L7'[t2[Y, O] : F(O, t) - F(Q' rt)

- { @ S l A l p r ( rr ), - } ( J ' ' n + c c ) } .
I,,.,$orrn,,r (8.137)

All of the quantitiesappearingin eqn (8.137)have been previously defined:

the term F(Q, r)arising from the end-point variations is the atomic averageof
the generator^f (definedin eqn(8.138);compare eqn(6.34)for a stationary
state),
r r
F(Q, r) :
J:
drpr(r,t) : (N t2)| Ar I dr' {Y*FY + ( F Y ) * Y ) .
Jo J

prF,t) is the correspondingproperty density (compareeqn (6.35))'

(8.138)

pei,t): (Nlz)ldr'{Y*FY+ (.f9*Y}, (8.139)

and Jo(r, r) the associatedvector current density,

- (vY*)FY},
Jo(r,r) : (I'{hl2mi)Jdr'{Y*v(FU (8140)

which is lr/ times the current density previously defined in eqn (5.98).The
variation L{rlY, O] : N 67/rrr,[Y, O] because the atomic averages are
defined for a density which integratesto l{.
It was emphasizedin Chapter 6 that the definition of an atomic stationary
state property is determined by the form of the atomic stationary state
functional EbL, Q]. In preciselythe samemanner, the definition of an atomic
property in the generaltime-dependentcaseis determined by the form of the
atomic Lagrangran integra| glY,,O, r]. In both the stationary-state and
time-dependentcases,the modes of integration used in the definition of the
atomic functionals that are required to give atomic statementsof the principle
of stationary action are those which yield atomic averagesof an associated
effectivesingle-particledensity. The property density peF) is obtained in the
same manner as is the charge density itself, by summing over all spins and
averagingover the coordinates of all the electronsbut one as denoted by the
symbol Jdr'. The quantity p.(r)dr, the contribution to the averagevalue of F
from the volume element dr, is thus l{ times the value of the property F for a
single electron as determined by its interactions with the averagedmotion of
all remaining particles in the system.
388 A C T I O N P R I N C IP L E F O R A Q U A N T U M S U B S Y S T E M 8.3

The variation of the atomic action integral differs from the result obtained
for the total system(eqn(8.125))in that, in addition to the differencein the
averagevaluesof the generatorsat the two time end-points,a time integral of
surfacecontributions arising from the flux in the current density and from the
change in the surface with time appears in the atomic expression.We now
demonstratethat both equationsare equivalentexpressionsof the principle
of stationary action-that the changein action is determinedby the contribu-
tions of the generatorsacting at the two time end-points.To seethis resultin
the equation for the variation of the atomic action integral, consider the
expressionfor the time rate of changeof an atomic expectationvalue which is
easily obtained by using Schrodinger'stime-dependenteqns(8.101)in the
expressionfor the time derivativeof an atomic average(seeeqn (5.aa)).This
yields

dF(Q,t)ldt: (t12){(ilh)<lH,Fl)n
+ cc}
r(r,t)
r)l7t\p
+ $ds(o,r){(ds(o,
- iQo'n(r)+ cc)). (8.141)

The first term on the right-hand side of eqn (8.141) represents,by

Heisenberg'sequation of motion, the total changein the expectationvalue of
an observablefor a systemwith infinite boundaries.It describesthe contribu-
tion to the time rate of changeof an observablearising from the interior of the
system.An open system,a system with finite boundaries,has further contri-
butions correspondingto a changein its surfacewith time and to a net flux in
the property current density acrossthe surface.Thesetwo terms appear with
different signs in eqn(8.141)because,when ASIAt^>0 correspondingto an
expansion of the volume Q, the averagevalue of F is increased,while, when
Jr.n > 0 correspondingto a net outflow of densityp, from O, the result is a
decreasein the averagevalue of F.
Now it is the essenceof the principle of stationary action that the total
change in action is equal to the differencein the values of the generator ^f(r)
evaluated at the two time end-points (eqn (8.79)).In the Schrodinger re-
presentation this corresponds to 6frr[Y] equalling the differencein the
averagevalues of the generator at the time end-points (eqn (8.125)).Accord-
ing to eqn (8.141)the differencein the averagevaluesof the generatorat two
times for an open system is
Ptt

F ( { l, t r ) - F (Q, tr): dr(dF(O,t)ldt)

I
J tr

ftt

dttL{$lh)([ri, F])n * cc]

I
J tr

+ ot,es tot)p
o- } ( J o ' n+ c c ) ) 1 . (8.t42)
f
8.3 A T O M I CA C T I O \ P R I \ C I P L E 389

From this equation it is clear that the differencein the a\erage valuesof a
property for an open system between t\4o times implicitly includes, in
addition to the differencein valuesdetermineduithin the boundariesof the
syste mas des c r ibedby th e c o mmu ta to ra \e ra s e .a n i ntegratedcontri buti on
over the time-like surface,arising from the flur in the property acrossthe
surfaceat each time r and from the change in the surfacewith time. This
situationis illustratedin Fig. 8.2for a tw o-dimensionalopen system.Thus, to
obtain a statementof the principle of stationarv action for an open system
equivalentto that for the total system.one must subtractfrom the difference
in the open-systemaveragevaluesthe rntegralover the time-like surfaceof
the boundary contributionsto this difference.This is preciselythe expression
given for the changein the atomic action integralin eqn (8.137).Comparison
o f th i s ex pr es s ionf o r L l ,/^ t2 l Yf w i th e q n (8 .1 4 2 )show s that the terms
subtracted from the differencein the end-point averagevalues of the gener-
ator are just those which account for the surface contributions to this
differenceintegrated over the time-like surfaceconnecting the two time end-
points. Thus what remains is the differencein the values of the generator at

0,(tr)

Frc. 8.2. Pictorial representationof the space-time development of a two-dimensional sub-

systemQ(r) of a total system.The space-likesurfacesare denoted by O(r),each being bounded b1'
a surfaceS(r, r). The collection of the latter constitute a time-like surface,which, together with
O(rr), and O(rr), define the space-time integration volume in'llrrrfY,O]. The principle of
stationary action statesthat the total change in action is equal to the differencein the values oi
the generatorsF1r;evaluatedin the two space-likesurfacesat r, and rr. There is no contribution
to 67/.r, from the interior of the space-time volume swept out as a consequenceof the s)'stem
evolvins in time or from the time-like surface.
390 A C T I O N P R I N C I P L EF O R A Q U A N T U M S U B S Y S T E M 8.3

the two time end-points averagedover the interior of the subsystem,the

essenceof the principle of stationary action. The changein the atomic action
integral can also be expressedas
ftt

L%r, [Y, ol: I dt{(ilh)(U{, Fl)o * cc}12, (8.143)

Jt,

an expressionwhich is equivalent in form and content to the corresponding

expressionfor the total system

ftz
Lfr, [Y] I dt{(ilh)(Ifr,Fl) + cc}lz. (8.144)
I
Jtr

Equation (8.144)is an alternativeform of the expressiongiven in eqn (8.125)

for the total system.The principle of stationary action for a subsystemcan be
expressedfor an infinitesimal time interval in terms of a variation of the
Lagrangtan integral, similar to that giv^enin eqn (8.127)for the total system.
For the atomic Lagrangian, assumingF to have no explicit time dependence,
this statementis

6slY,o, rl : (t12){(ilh)<tri,
Fl)o + cc}. (8.145)
This result may be restatedin the form appropriate for a stationary state by
using the relationship
glY, C),rl : - (Y, Y)a@ [/, O] - E), (8.146)
correspondingto the total systemexpressiongiven in eqn (8.129).This yields
the statements

68bL,ol: - +{(ilh)
<lH,Fl)" + cc}l(!,{)n (8.147)
and
: - +{(ilh)(tri,
6EbL,Ql Fl)n + cc} (8.148)
as the analoguesof the statementsfor the total systemgiven in eqns (8.131)
and (8.132).Equation (8.148)is, of course,identical to that obtained directly
from the variation of Schrodinger's functional for a stationary state as
describedin Chapter 5 and stated in eqn (5.100).
The derivation of the principle of stationary action for an atom in a
molecule(eqn(8.143))yields Schrodinger'sequation of motion for the total
system,identifies the observablesof quantum mechanicswith the variations
of the state function, definestheir averagevalues,and givestheir equations of
motion. We have demonstratedin Chapter 6 how one can use the atomic
statementof the principle of stationary action given in eqn (8.148)to derive
the theorems of subsystemquantum mechanics and thereby obtain the
mechanicsof an atom in a molecule.The statementof the atomic action
8.3 ATOMIC ACTION PRINCIPLE 391

principle given in eqn(8.145) plays the same role in the general time-
dependent case, as will be illustrated in the following section. Equations
(8.143)and(8.145)represent an extensionof quantum mechanics.They enable
one to obtain a quantum mechanical description of the properties of any
region of spacebounded by a surfaceof zero flux in the gradient vector field
of the chargedensity.In this sense,the mechanicsof a total systemis obtained
as a special caseof these more general equations.

8.3.6 Examplesof unitary transformations

The variation of the functional?ftlr, Q] for a stationary state can be equated

to an infinitesimal times the flux in the current density of the generator
through the surface of the atom (eqn (5.99)).The corresponding statement
for the variation of the atomic Lagrangian integral, obtained by using
e q n s( 8. 141)and ( 8. 1 4 5 ),i s
- dsl(05l0t)p,
6s lY , o, rl : (r l2I\0{dF(O)/dr $
- IQo'n + cc)l). (8.14e)
One may qse this expression to evaluate 69 [Y, O, r] for any particular
generator F, correspondingto the use of eqn (5.99)for the evaluation of
6Ebll,Q] as was done in Chapter 6. However,we shall take the opportunih'
of evaluating69 for various generatorsto illustratethe propertiesof unitarl'
transformations.
The action of the generator of a unitary transformation, as representedb1
the Hermitian oper ator F, can be determinedthrough the use of the operator
d defined as
0:e-itDF and 0':e+1lh)F (8.1
50)
The sameexpressionappliesif F is an infinitesimal Hermitian operator equal
to ed, for, as r r 0, the seriesexpansion for an exponential yields, to first-
order in e,
0: i -(itilF: i -(ilhfi. (8.151)

As demonstratedin Chapter 6 for a stationary state,the atomic force law is

obtained by setting the generator equal to the momentum conjugate to the
coordinater, the electroniccoordinatethat is integratedover the basin of the
atom O. It follows immediatelyfrom eqn (8.35),which showsthe effectof an
infinitesimal unitary transformation on an operator, that the generator
F : t.p induces an infinitesimal uniform translation of the electronic co-
ordinate r by - a. Each component of the vector e is an arbitrary, infinitesi-
mal real number, fixed for all r and, from eqn (8.35),
r' : jr0r : r - ( i t , l h ) ' [ p , r ]: r - r . (8.152)
392 ACTION PRINCIPLEFOR A QUANTUM SUBSYSTEM 8.3

It is easilyshown that the sameresult is obtained for finite e through the use
of the generator p in a unitary transformation. This is most easilv done bv
parametrizing t7 as
-(i/fr)(e'n'
0G): t (8.153)
For ( :1,00):0 a n d , f o r ( : 0 , t 7 ( O:) l . D i f f e r e n t i a t i o no f t h e t r a n s -
formed coordinate denoted by r(O by the parameter ( yields

d E r ( O :A { 0 G ) r u f ( O:) - ( i l U C ( O t e . p , . l 1 7 ' ( 0 _
: t. (8.154)
The integration
of this resultover ( between
the limits0 and 1 yields
r(1) : r(0) - e. It is clear that the generatorp has no effecton a momentum
coordinate.In parallel with the resultsfor F: r'p, a uniform translation in
momentum space is instead obtained through the use of the generator
F : e'r in a unitary transformation.
A perhaps more interesting example of a unitary transformation is pro-
vided by the 'virial op-erator' r'p. The Hermitian form of this operator
denoted by f , where t is given by the equivalent expressions

r : i t . p + p . r ): r . p - ( 3 1 2 ) i h (8.1ss)
when multiplied by the real number e and used as the generator of a unitary
transformation, causesthe electronic position coordinate to be multiplied by
the factor ( : e'. Hence, the virial operator is called the scaling operator,
sinceit generatesa scalingof the electroniccoordinates.This accountsfor the
well-known result that the variational property of the energy and the
satisfactionof the virial theorem are related by the operation of scaling the
electroniccoordinatesof the wave function.The useof the singleproduct r.p
rather than the Hermitian averageshown in eqn (8.155),as was done in the
derivation of the atomic virial theorem in Chapter 6 by determining the
atomic averageof its commutator with Ii, is permissible,since the difference
between these two forms is simply the constant term 3i12.
To obtain a scaling in a positive sensewe shall take F : - t.t to define
the unitary operator

t71r;: s(ilh)ei (8.1s6)

Proceedingas in eqn (8.154)by taking the derivative of the transformed
coordinate r with respectto the parameter,one has
: d.{ C14r0r(e)}: (ilryC@)l/, r)0, @): r(e).
d.r(e) (8.157)
The integrated form of this result between the limits 0 and e is

r(e):re':r(, (8.158)
that is, the coordinate r is scaledby the factor ( : e'.If e is taken to be an
infinitesimal and F, therefore, becomes the generator of an infinitesimal
8 .4 V A R I A T I O N A L B A S I SF O R A T O M I C P R O P E R T I E S 393

unitary transformation,the coordinater is transformedinto r(1 * e),keeping

only the first-orderterm in the expansionof the exponential.The sameresult
is of course obtained using eqn (8.35).
We next demonstratethat the action of i1e; on the statefunction yieldsa
properly normalized function with the coordinate r scaled by ( (Epstein
1974).That is,
t 7 ( r ) Y ( r ,r ' ) : 6 t t z t Y ( ( r , r ' ) (8.159)

where,as before,r' denotesall the electroniccoordinatessaver. This result is

proved by noting that with X : 0 G)Y one has the identity,

hdx
X : Y when r : 0, (8.160)
i d c : l ' X '

and then showing that this relation is satisfiedwhen X is given by the right-
hand side of eqn (8.159).The differentiationgives

hdx '.
h i w ( r ( e) ) o ' r ( c ) l
. , I : ltlizli',2Y(ir. r,) + ;. '' ; ^ r, ( 8 . 611 )
I Or I a r(r) at )
which, together with the result that

l Y ( r ( e) ) f r (' e) l
1l : r'VY(r, r'), ( 8r 6 2 )
at14 ar 1,:o
yields the desired result when e : 0.

8.4 Variational basisfor atomic properties

8.4.1 Variational derivation of the atomic force law

The statement of the atomic action principle given in eqn (8.1a5)is a

variationalprinciple which enablesone to derivethe propertiesof an atom in
a molecule-it is an atomic variation principle. We shall use it first to derive
the atomic statementof the Ehrenfestforce law, the equation of motion of an
atom in a molecule.This is accomplishedthrough a variation of g [Y, Q]
with generator F : t'p. As detailed above, such a generator induces an
infinitesimal uniform translation of the electronic coordinate r by - E.
Denoting by Y, the image of the state function Y in the infinitesimal unitary
transformation generatedby F, one has

Y , ( r , r ' , t ) : 1 f - 6 tn :te ' 0 )Y (r,r' , t) : Y(r - q r' , t\. (8.163)

The variation in glY, Q] is given by the first-orderdifferencebetween9(t\

and 9k: 0) or
6 g l Y , Q l : s ' [ V ,I ( e ) f , : o . (8.164)
_ r 9 . 1 A C T I O N P R T N C I P L EF O R A Q U A N T U M S U B S Y S T E M 8.4

Thus the atomic variation principle statesthat

e - Y , 9[ V , , Q ]l , = o: I e ' { ( i l h ) < l f i , l l ) " * c c } . (8.165)

Y, for Y in the expression
Substituting for glY,Ol (eqns(8.100) and
(8.110)),differentiatingwith respectto a, and noting that

V r Y r l e = o: - VY,

one obtains for eqn (8.165)(including a variation of the surfacedS: e'n)

v u y l y , , Q l l , : o : i o .i o r , { + ' r , , o v *+) .vi *, v \ i ,

J o J t
- (v.P*)Y
-.P*vv;
h2_
),((vv,Y*).viY)
2mu'"

+ ( v v j Y ) . v , Y -* ) t l v v * Y + Y * v Y ) i
I
h 2 f
- dsv' on' (8.166)
^^- f
The dyadic notation has been used to express the terms arising from the
kinetic energy part of the Lagrangian. The explicit form for one such term is

( v v i y * ) . v i y : I o /._- ( I?, 2r F
Y* eY\
tr (8.t67)
"f,)
wherero(r(t))isthe componentof r(r,)along the directionof the unit vectorep,
k : l, 2, 3. It follows from eqn (8.167)that
( v v , y * ) . v , y: v i . ( v , y v y * -) v y * v ? y . (9.168)
Using eqn (8.168)and its complex conjugate together with Schrodinger's
equation,eqn (8.166)becomes

yu?[v,,o]1,-o: I o,Jio.'{+ih(y*vv
+ v*v\p
Jo
- (V'P*)Y- (VY*)\P))

vY + vyvy* } .n
+ (h'tz*l$asfdr'{vY*
"
, "
- f t 2, 4 m N ) v2Rn,
tds
I
L

8.4 V A R I A T I O N A L B A S I SF O R , { T O \ I I C P R O P E R T I E S 395

a result which may be equivalentlyexpressedin terms of J(r) as given in eqn

( 8 . 1 4 0f)o r F : i a s

:' -
ff.|:or'^J, #'' fasv'nn
t,.Y,e[Y,.cl]1,.-,,

v \ P V Y * ] '(n8.. 1 6 e )
+ , + 6 o t f o r ' [ v Y * v+Y
ZMJ J
Since the subsystemis bounded by a zero-flux surface at all stagesof the
variations generatedby F, the zero-flux variational constraint (eqn (8.117))
applies.In the presentcaseit assumesthe form
'7i z^ o l' (f \l
: o' ( 8l '7 o )
' Y 2o v d r
J" ) 1 ,n
:
Using the same procedure as above, we readily find the left-hand side of eqn
(8 .1 7 0)t o be giv en b y

rrrr P P
, . ' ;. Y" ,\ J1I" v ' p * , d).l - - E+4 m d o t l d r ' [ ( v v )YY*+ v v *v Y
7z / r \l

4m /tt_o J J
+ VYVY* + Y*(VVY)I'n
v 2 f
* ' ' ; , t d s v 2 n n: o . ( 81 7 1 )

for the .onrtruint to the right-handsideof

The additionof this expression
eqn(8.169)yields

e ' y , e [ v , , o ] 1 , :,oN^: [, L { *oi 'd* r - 6 o t a" } g.n2)

JQ J )
where

€ . . : N h 'l ,
F'1r; + Y*V(VY)- VY*VY - VYVY*} (8.173)
; ldr'{v(vY*)Y

tensorpreviously
is theOuun,u-stress in eqn(6.12).As
defined emphasized
there, the stress tensor plays a dominant role in the description of the
mechanicalpropertiesof an atom in a molecule.It has the dimensionsof an
energy density or, equivalently,of force per unit area.Its dot product with
n(r), the outwardly directed normal to S(Q,r), gives the force exerted on the
associatedelementof the atomic surface.As shown in eqn (6.13),F(r) is totally
determinedby the one-electrondensity matrix J'tt)(r, r').
Evaluation of the commutator appearingin eqn (8.165)demonstratesthat
this variation of glY,Q] yields the atomic averageof the total force acting
396 A C T I O N P R I N C I P L EF O R A Q U A N T U M S U B S Y S T E M 8.4

on the electrorrdescribedby the position vector r,

e . \ { ( i l h ) ( [ f , 0 ] ) "+ c c ]: ' . ( - v t ) n : e.(l/N)F(o
r ),. (8.174)
By combining the resultsgiven in eqns (8.172)and (8.174)one obtains from
the atomic variational principle the integrated atomic force law

F(Q, r) :* f JO
I at)-
dr(aJ$)
fotlrrlr)
n(r). (8.17s)

Equation (8.175)is a generalizationof Ehrenfest'stheorem (Ehrenfest 1927).

This theorem relatesthe forces acting on a subsystemor atom in a molecule
to the forces exerted on its surface and to the time derivative of the
momentum density mJ(r).It constitutesthe quantum analogueof Newton's
equation of motion in classical mechanics expressedin terms of a vector
current density and a stresstensor, both defined in real space.

8.4.2 Differential force law

Within aLagrangian formulation of the mechanicsof a field as usedhere one

may define an energy momentum tensor whose components summanze the
principal properties of the field (Morse and Feshbach 1953; Landau and
Lifshitz 1975).We show that the divergenceequations satisfiedby the spatial
components of this tensor for the Schrodingerfield yield the differentialform
of the atomic force law, eqn (8.175)(Bader 1980).
Define a component of the energy-momentum tensor as

\n: Yu@LlAY,) + cc - 6r,L, (8.176)

where p and v run over the 3l{ spatial degreesof freedom and the time, and
Yn = AY lAx". The spatialcomponentsWj,(j referringonly to the 3l{ spatial
coordinates)satisfy the set of divergenceequations

u' r y ' :- Y
L .0*n A*i
$n7)
where0LlAxt vanishesunless,L possesses an explicitdependence
on x,. In the
case of the Lagrangian density defined here (eqn (8.100)),

ALlAx,: - Y* PtlAx,)Y
and eqn (8.177)yields

- y , 'v, , t v : t : # ( v ; y * ) y - y * v ; v l -
fiZrvo.[(vuvjy*)y
+ Y*(VftVjY)- VuY*VjV - VkYV;Y*]. (8.I 78)
8.4 V A R I A T I O N A L B A S I SF O R A T O M I C P R O P E R T I E S 397

Equation (8.178)has beenpreviouslyobtainedby Pauli (1958)and by Epstein

(1975),who termed it the differential force law. The integration of this
expressionover the coordinatesof all electronsbut one by the usual recipe
yields (Bader 1980)an equation governinga force density

F (r, r) : m i .J F )I 0 t - V' F ' (r). (8'179)

Equation (8.179)is the differentialform of the integrated atomic force law

g i ve n in eqn ( 8. 175),

F(O,r) : I drF1r,
r;. (8.180)
Jo

As previouslydiscussed in Chapter 6 in eqn (6.17) and following for a

stationarystate,the force density F(r, r) is the total instantaneous force
exertedon the electronat r , by the nuclei and the remaining electrons

(rl- ril
- v l t : F ' : - ' eL2z-'"2' l r, ,! : - 2 +' I e L' 7j r .r - , -
- X"l3 lr, r;13

: I o F " o* I , ' , F " r " ; ( 8r.8 1 )

averagedover the motions of the remainingelectronsin the systemto yield

F ( r ,r ) : I " F " " p ( t ) - 2 ! & r i " r , z f ( 2 ) ( r , ,r r ) ( 81 8 2 )

wh e re ; t z ) ( r 1, r r ) is th e d i a g o n a l e l e m e n to f th e tw o-densi tymatri x. eqn

(E1.3).While F(r, r) clearlyinvolvestwo-body forces,accordingto eqn (8.179)
it is totally determined by J(r) and F'(r), which are functionals of the one-
densitymatrix, ptt)(r, r') evaluatedat the point r : r'in real space.The force
density F(r, r) may be interpreted as the total force exertedon the element of
chargedensitylocatedat r and obeying an equation of motion in real space,
eqn (8.179).This equation is identical in form and physical content to
Cauchy'sfirst equationof motion of classicalcontinuum mechanics(Malvern
te6e).
The variational derivation of the integral atomic force law, eqn (8.175),is
applicable only to a region of spacebounded by a zero-flux surfacein Vp(r),
i.e. to an open systemwhose Lagrangian integral vanishesat the point of
variation. Thus the variational derivation of the atomic force,

t , ' Y , 9 [ Y , Q ] 1 , : o : e ' F ( Q ,t ) l l ' { , (8.183)

and its law of motion (eqn(8.175))may be regardedas a quantum mechanical

definition of an atom in a molecule.
 398 A C T I O N P R I N C I P L EF O R A Q U A N T U M S U B S Y S T E M 8.4

For a stationary state the correspondingexpressionsare

e'Y,8[Y, Q]1,: o : (e F(O)/l{)lQt, {)" (8.184)

where

F(Q): -
f ds(r)F(r)'n(r). (8.185)

In generalphysical terms,the averageforce exertedon some enclosedvolume

is equal to the negativeof the integral of the pressureacting on each element
of its surface-a statementof the physicalcontent of eqn (8.185).This same
equation demonstratesthat the effect of the environment on an open
quantum system Q is determined by just the flux in the forces across the
boundary of o (Bader 1980).(Seeeqn (6.18)and accompanyingdiscussion.)

8.4.3 variational derivation of the atomic virial theorem

The general time-dependent virial theorem for an atom in a molecule is

derived from the atomic variational principle. We shall find a close connec-
tion betweenthe expressionsso obtained for the virial and those derived in
the previous sectionfor the force. In particular, the differentialforce law leads
directly to a corresponding local expressionfor the virial theorem.
The generator of the infinitesimal unitary transformation, whose atomic
projection determinesthe variation in glY,o], is in this caseF: - e{r
where e is an infinitesimal real number and where {r is the virial operator,
eqn (8.155).As discussedin Section 8.3.6,the action of this generator is
equivalent to a scaling of the electronic coordinate r by the factor ( : | + e.
We shall use the latter of the two forms given for t in eqn (8.155).Sincethe
constant term - (312)idoes not contribute to the variation in g or to the
commutator averagein eqn (8.145),the generator^Fis setequal to - er.p, the
term responsible for generating the scaling of the coordinates. Thus the
transformed state function Y,(r, z', r) is given as

Y , ( . ,r ' , t ) : ( l + ( i l h ) s r . 0 ) Y ( r ',, ) : 6 t r z Y ( ( r , t ' , t \

and the variationalprinciplenow statesthat

*t[Y,,
o ] r , = o: - ;t(;)<tn'ior)n+..1 (8.186)

Writing the expression for g [Y, Q] explicitly in terms of e, differentiating

with respectto e, and noting that

( c d e ) 1 5 -: s r ' V ,
8.4 V A R I A T I O N A L B A S I SF O R A T O M I C P R O P E R T I E S 399

o n e o b t ains( wit h 65 : - er' n )

: o'Fr'.
iro[y,,o]1,=o - .ir*VV;
+ Y*v\i'-(v'i^t)Y
J Il,o*-'i'
62
- ^ L ( ( v v , Y * ) . v j Y+ ( v v i Y ) ' v , Y * )
zm-'

- rzlvy*Y
+ Y*vY)-l+ +nt|\
L,$asv'p.'n
I J

+ f o,io,'( - 2!)o*.
^/
vY (8.1
87)
Jo J \
Equation (8.187)can be rearrangedusing eqn (8.168),Schrodinger's equa-
tion, and the identities,
v,.(v,Y*vY)'r : Vi'[(V,Y*VY)'r] - (V'Y*VY)'V,r,

V,'(V,YVY*)'r : V'' [(V,YVY*)'r] - (VTVVY*)'Vir,

to yield

e m l . / ';
: -;.l()o'(, aJ\ fi f
e]t,=0
[v,, osv' rrn
^v )+ 4mrv
f
- h2f r + vYVY*]'n' (8'188t
2a9otJdt'r'[vY*vY
As the atom is to be bounded by a zero-flux surface at all stages of the
variation, the variational constraint of eqn (8.117)must be imposed.This
constraint in the present casedemands that the variation

y !( | ,'r,',0.)l : 4 m t , J fo,'.tv(vY*)Y
4 n r r r d e \ J , r / 1 , = o *{6ot
'rr
J
+ Y * V ( V Y )+ V Y * V Y + V Y V Y * I ' n

-$osv'p.n) :o
+ f arv'p1.1 (8.18e)
Jo J )

vanishes in the limit e -- 0. The addition of the right-hand side of this

constraint to eqn (8.188)yields,for the constrainedvariation,
d ml AJ (
: - drr'"u+
iol*,.oll,=o ;J" fdstr)r'F(r)'n(r)
l z r
* (8.190)
+nrl,,drV2r(r)
-trxl ACTION PRINCIPLEFOR A QUANTUM SUBSYSTEM 8.4

n hereF(r) is the stresstensorpreviouslydefinedin eqn (8.173).Evaluatingthe

c om m ut at or o f th e ri g h t-h a n ds i d e o f e q n ( 8.186),
( t i n ; 1 f r , i ' 0 1 ):. , 2 ( Y * , - ( h ' l 2 m ) V ' Y ) o+ ( Y * , - r ' V t Y > n ,
thereby yielding
o
N;9 [Y,,Q] l,:o : - 2 T u ( O )- { o ( A ) . ( 81. e1 )
ot
The quantity Tk(A) is the average kinetic energy of the subsystemdefined
specificallyin terms of the density K(r) (eqns(5.48)and (5.49))and {o(Q) is
the virial of the forces exerted on an electron in the subsystem resulting
from its instantaneousinteractionswith the nuclei and other electronsin the
system(seeeqn (6.19))

/"o(a) : i orJdr'{- r.v21w*v.

JO
(8.te2)

Equating the results given in eqns (8.190)and (8.191), one obtains an

expressionfor the virial theorem of an opensystemwhich satisfiesthe zero-
flux boundary condition (8.109)

- 2Tk(O): {,,(!-) + {!(O) + r'i(Ol + (h,l4ml drV2p(r).

I
JO

The final term on the right-hand side of this equation vanishesbecauseof the
zero-flux surfacecondition thereby yielding
- 2To(O): {o(A) + {s(O) + {i(O). (8.193)
The term {t(A) is the virial of the forcesexertedon the surfaceof the
subsystem,a term expressible
in termsof the stresstensorpreviouslydefined
in eqns(8.173)and(6.12),
/.r(A) : f dS(r)r'F(r)'n(r). (8.1e4)
(F''n is the outwardly directed force exerted per unit area of surface and
r'7 ' n is the virial of this force.)The term {b(Q), specificto a time-dependent
system,is the virial of the forces arising from the time rate of change of the
momentum density rnJ throughout the subsystem,

/ ;(l)): - ^ r'6J16t. (8.1e5)

Irdr
It was pointed out in the discussion of the atomic virial theorem for a
stationary state (eqn (6.23))that, while the values of the individual contribu-
tions to the virial of the subsystemare dependent upon the choice of origin
for the vector r, their sum, which determines the total subsystemvirial, is
independentof the choiceof origin. It is clear from eqn (8.193)that the same
8 .4 V A R I A T I O N A L B A S I SF O R A T O \ 1 I C P R O P E R T I E S 401

origin independenceof the total virial

) rp(O)
l ' ( Q ) : / ; ( O )+ Y ; ( O+ (8.1e6)
is obtainedfor the generaltime-dependentcase.The basin and surfacevirials
are identical with their stationary-statecounterparts,eqns (6.19)and (6.21),
respectively;the generalstatementof the atomic virial theorem given in eqn
(8.193)reducesto the stationary-stateexpression,eqn (6.23),when 1/'n(Q): 0.

8.4.4 Local virial relationship

The averageof the atomic virial {o(Q), eqn (8.192),as defined by the atomic
variational principle, is the virial of the quantum mechanicalforce density as
defined in the differential force law, averaged over the atomic volume. By
taking the virial of F(r, r) in eqn (8.179),one obtains
r ' F ( r ,t ) : m r ' A JI 0 t - r ' V ' F (8.1e7)
One proceedsas in the caseof a stationarystate(eqn(6.25)and following).
Usingthe identity
V'(r'F): TrF * r'V'F
whereTrF denotesthe traceor spur of the stresstensor,eqn (8.197) gives
r ' F ( r ,t l : m r 'A JI A t * T r A - V ' 1 r ' F ' ) . (8.198)
From eqn (8.173)for the definition of F(r), Tr F may be variously expressed
as (seeeqns (6.26)and (6.27))
TrF(r;: -K(r) -G(r) : -2K(r)-(h'l4m)Y2p(r)
: - 2G(r)+ (h2l4m)Y'p(r). ( 8 1. 9 9 )
Its integralover an atomicvolumeyieldsthe uniquevalue

J: dr Tr F(r): - 2r(Oy

Using the final identity given in eqn (8.199),eqn (8.198)may be expressedas

(8.200)

- 2G(r): r' F(r,r) + v'(r'F'(r))- mr'AJ$)I at

- (hzl4m)Ytp(r). (8.201)
Integration of eqn (8.201)over an atomic volume for which the integral of
Ytp(r) vanishesyields, term for term, the atomic virial theorem for a time-
dependentsystem(eqn (8.193))orfor a stationarystate(eqn(6.23)).Thus, eqn
(8.201)is, in terms of its derivation and its integratedform, a local expression
of the virial theorem. The atomic virial theorem provides the basis for the
definition of the averageenergy of an atom, as discussedin Chapter 6.
 402 ACTION PRINCIPLEFOR A QUANTUM SUBSYSTEM 8.4

8.4.5 Summary of the atomic variational principle

The principle of stationary action provides a complete quantum mechanical

description of a system in terms of the changesinduced in the system by
infinitesimal unitary transformations. The generators of such trans-
formations describe all possible system changes,caused both by displace-
ments or by a temporal development(Roman 1965).It has been shown that
this principle, which is a variational principle, appliesuniquely to a connected
region of real space bounded by a surface of zero flux in Vp(r). Thus, the
principle of stationary action definesan atom as being a bounded region of
space,the variational properties of which, as determined by the action of
infinitesimal unitary transformations, are identical to those obtained for the
total system.The generatorsof such transformations describereal physical
changesin the system.Thus,a singleuariationalprincipleseruesto desuibe the
propertiesof a total systemand of an atom within the system.
The operational statement of this principle is the atomic variational
principle

6 s l Y , Q l : ( + ) { ( i l h ) < l f r , r l*) "c c }

or, for a stationary state,

6 E l Y , Q l : - ( + {) ( i l h ) ( [ f , F ] ) " * c c ] l < r l t , , l t ) n .

The generator F causesan infinitesimal transformation in the state function

describing the total system. The physical nature of this transformation is
identified through the commutator of /i with F. The atomic value of the
change caused by the transformation is given by the projection of the total
system value over the atomic volume. Since a generator can be found to
describeany changein a system,all atomic propertiesand the laws governing
these properties can be derived from this principle. These derivations of
atomic properties parallel the derivations of the correspondingpropertiesfor
the total system,since both are describedby the same variational principle.
For example,F : ef : ,iN, where l{ is the total number of electronsin
the system,generatesa simple infinitesimal transformation, which leavesthe
Lagrangian 9lY,O] invariant. In addition, since l{ is a constant of the
motion for the total system,tli, lf ] : 0. However, the time rate of change of
the averageelectronicpopulation of an atom, l{(Q), is not zero in generaland
the equation of continuity governing the time evolution of l{(O) is obtained
directly from the equivalent statement of the atomic variational principle,
eqn (8.149),as

asl
"q.l" drp(r) : lV(Q):
fo'{r'" *cc)-
a r pI '
(8.202)
8.5 A T O M S I N A N E L E C T R O M A G N E T I CF I E L D 403

As illustrated earlier, setting the generatorequal to e'p definesthe atomic

force and the variational principle leads to the integrai atomic force law, or
the equation of motion for an atom in a molecule.Finally, it was shown that,
when F : - €r'f, the commutator definesthe electronickinetic energyand
virial for an atom, and the variational principle yields the relationship
between these quantities, the atomic virial theorem. These three relation-
ships-the equation of continuity, the equation of motion, and the virial
theorem-form the basisfor the understanding of the mechanicsof an atom
in a molecule.
Within the Lagrangian formalism one obtains a definition of the
energy-momentum tensor for the Schrodingerfield. The many-particle diver-
gence equation satisfied by the spatial components of this tensor, when
integrated in the standard manner, yields a differential expression of the
atomic force law. The virial of the force density defined in this manner givesa
local expressionfor the atomic virial theorem. Thus, from a singleformalism
one obtains a description of the local mechanical properties of an atom as
well as of their averagevalues.
This chapter concludeswith the demonstration that the atomic statement
of the principle of stationary action obtains in the presenceof an electromag-
netic field. Thus, atoms continue to exist in the presenceof applied electric
andlor magnetic fields and the responseof each atom to such fields and its
contribution to the magnetic and electricpropertiesof the total systemcan be
determined.

8.5 Atoms in an electromagnetic field

8.5.1 The Lagrangian and Hamiltonian

We shall first find that Lagrangian for a system of charged particles in an

electromagneticfield which, through the principle of least action, gives the
correct equation of motion. The electric and magnetic fields of the electro-
magnetic field, E and B, respectively,are related to the scalar and vector
potentials, S and A, by the equations

1 A A- V 6
6- ^ (8.203)
cAt
and
B:VxA. (8.204)
It is clear that @ and A are not uniquelydetermined
by theseequations,as
one can modify both potentialsthroughthe additionof the scalary$, t), i.e.

A'(r.r) : A(r, r) + VZ(r,r) (8.20s)

404 A C T I O N P R I N C I P L EF O R A Q U A N T T ] MS U B S Y S T E M 8.5

and
Q ' F , t ) : Q ( r ,r ) - ( l l c ) ( A X g , t ) l A r ) (8.206)

without affecting the fields E and B in any way. Since the state function
dependsupon the potentials,this so-calledchangein gaugeof the potentials
representedin eqns (8.205)and (8.206)changesthe state function as well. The
wave function in the new gaugeis related to the function in original gaugeby
a gauge transformation, a unitary transformation representedby

Y'(r, t) : si'xt't)lh'Y(r,t). (8.207)

If the physical results are to remain unchanged under a unitary trans-
formation, it is necessaryto transform the operators as well as the state
functions (eqns (8.28)and (8.29)).A gauge transformation has no effect on a
coordinate operator but the momentum operator p is changed into
0 + @lc)VX.Thus, to maintain gaugeinvariancein propertiesdeterminedby
the momentum operator, p must be replaced by a new operator. In the
presenceof an electromagneticfield or just a magnetic field, the momentum
operator p is replacedby the expression

rt:fi-(elc)A(r,t) (8.208)
for a particle of charge e. The use of i will insure that all averagesover a
momentum operator will be gauge-invariant as the extra term (elc)VX is
exactly cancelledby the changein A on going to the new gauge.With the
momentum expressedas in eqn (8.208),one is free to chooseany gaugeand
the one most convenientfor our purposesis called the transversegauge.In
this gaugeone setsV'A -- 0, a choicewhich has the consequence of causing
t he quant it iesp a n d A to c o m m u te ,p ' A : A ' p, si ncethei r commutator i s
-ihv.A.
Through the use of eqn (8.208)and the inclusion of the scalarpotential @,
L(V,VY, \p), the many-particle Lagrangian given in eqn (8.100),can be
modified for the presence of external electric and magnetic fields. The
modified Lagrangian L(Y,VY, W, A,O) can be expressedas

L(Y,vY,\P,A,O): L(Y,vY,\n +, lLf lri , {V,Y*

\).t ^
.A(r,)Y
2mC

- A ( r , ) Y *. V' t ,^Y) ) - : - r . A 2 ( r , ) Y * Y
2*r2Li'

- (8.20e)
L,ti6{ri)Y*Y
where the sum _/ runs over all the chargedparticlesin the system.
The
action intesralis definedas
associated
P t -)
t l
P

f ,r.lY,A, Qf : I l a rL ( Y ,v Y ,Y, A, 6)dl (8.210)

Jt' J
 8.5 A T O M S I N A N E L E C T R O M A G N E T I CF I E L D 405
I

I Va ri at ion of the action integral with respectto Y*, subject to the natural
I Uo u ndar y condition that V,Y'n : 0 on the surface at infinity and the
condition that the variationsin Y* vanishat the time end-points,yieldsas the
equation of motion the Euler-Lagrange equation of the variation. This
statementof Schrodinger'sequation, the one appropriate for use when the
systemis in the presenceof electric and magnetic fields, is

ihY :
fit,(i''- 1o,,,,)'
Y + (t +l,e,Q(r;))Y. (8.211)

This result may be re-expressedin terms of I?o, the usual field-freeHamilton-

ian. and fit. an interaction Hamiltonian. as

rltY:tiow+IitY (8.2r2)
where li' ir given by

e -
H': - + A(r')'p')
2^r),(Pr'A(r')

. #I,A'(',) + lyi|g'). ( 8 . 231)

It will be recalled (eqn (5.22))that the first of Ehrenfest'srelationships.

obtained by taking the commutator of Ii with the position vector i. states
that the momentum operator p is given by mdildr. If one usesthe Hamiiton-
ian fl?0+ frt in the commutator with i, then one correctlydeterminesthat thc
mo m ent um is now g i v e n b y f i n e q n (8 .2 0 8 ),th a t i s, ndi /dr : p - (e c )A .
Before proceedingwith the definition and variation of the corresponding
atomic Lagrangian and action integrals,it is important to verify that, w'hen
Schrodinger'seqn (8.211)is satisfied,the many-particle Lagrangian in the
presenceof the external fields(eqn (8.209))still reducesto a sum of terms,each
of which is proportional to the Laplacianoperatorfor a singleelectronacting
on the product Y*Y. It is through this property of the Lagrangian that the
consequencesof the variational constraint which demands that the sub-
systembe bounded by a surfaceof zero flux in the gradient vector field of p is
incorporatedinto the variation of the atomic action integral.Substitutingfor
tP and tP* in eqn (8.209)through the use of eqn (8.211)and its complex
conjugate,one finds as before that

vy.\i,.A.dr : I,
L0(y. v + yviy*)- o,*}
{#(y*vi
\ !},o,*. /
-h2_
:;L,v1(Y*Y) (8.2t4)

wherethe final equality is obtainedusing the identity developedin eqn (5.46).

406 ACTION PRINCIPLEFOR A QUANTUM SUBSYSTEM 8.5

In analogy with eqn (8.110),the atomic Lagrangranintegral in the presence

of external fields is defined as

g lY,O, A, 4, tf : I Ot I dr'L(Y, vY, .P,A, d)

J o J

: I a ,s ( r , A ,d ) (g.215)
Jo
and the associatedaction integral as
t,, glY, O. A, Q,t).
{rr.lY, Q, A, Qf : (8.216)
I dr
At the point of variation,whereV unl W* satisfySchrodinger's
equation,the
atomic Lagrangianintegral,becauseof eqn (8.214),reducesto a surface
integralof the flux in Vp throughthe surfaceof the subsystem
P r t

golY,,o, A, 0, tl : - (h,ftmNl dry2p(r,

I drsro(r, A, Q): I t)
Jo Ja
P
: - th214mx)0dS(O,
' I t)yp(y,
r).n(r,r). (8.2t7\
al

Thus, the atomic Lagrangian and action integrals in the presenceof an

electromagneticfield, like their field-free counterparts, vanish as a conse-
quence of the zero-flux surfacecondition (eqn (8.109)).These properties are
common to the corresponding integrals for the total system and it is a
consequenceof this equivalencein properties that the action integralsfor the
total systemand each of the atoms which compriseit have similar variational
properties.

8.5.2 Definition of an atom in an external field

The definition of an atom in a moleculein the presenceof an electromagnetic

field is accomplishedby defining its properties using the principle of station-
ary action. This principle for a total systemand for an atom within the system
is obtained as in the field-freecase,through a variation of the corresponding
action integrals.The variation is generalizedto include a variation of the time
end-points and by the retention of the variations in the state function at the
time end-points. It is the end-point variations that are identified with the
generators of infinitesimal unitary transformations. In this manner, the
changein the action integral is equated to the action of infinitesimal unitary
transformations acting on the state functions at the two time end-points, a
relationship known as the principle of stationary action. At every stageof the
variation of the atomic action integral, which includes a variation of its
8 .5 A T O M S I N A N E L E C T R O M A G N E T I CF I E L D 407

surface,the variation is subject to the constraint given in eqns (8.116)and

(8.117).This constraintis equivalentto demandingthat the atom be bounded
by a surfaceof zero flux in the gradient vector field of the charge density at
every stage of the variation (eqn (8.11a)).The result of such a generalized
variation of the atomic action integral (given in eqn (8.216))is (compare the
field-freeresult,eqn (8.118))

6 f , r [ Y , O ,A , 0 f

: orJar'{( - r?oy*- t'y*)dv + cc}

- ift\i,*
l,',,'drJ
* - h2l2m) *
,)Jo,'{
a, asro, n(.)+ c.l
l,',,'f {(o*. #rtrlv-)av
* arf osto ,l{s't, A,Q)6,rs(e,
r)+ cc
l,',,'
+ - ihlz)(as(o,
r)l,,t)Y*dv
+ cc)
Jar't
+ o. . ..} (8.218)
{rinta{ Jo,'*.r- [:
The first set of terms multiplying the variations in Y and the corresponding
complex conjugate set yield Schrodinger's equation for Y and Y* (eqn
(8.211))for the restrictedvariation wherethe natural boundary conditionsare
imposed, the end-point variations are set equal to zero, and O equals all
space.The equations of motion are assumedfrom this point on (they are in
any caseimplied by the principle of stationary action, eet (8.79)).Since the
Lagrangian integral vanishesat the point of variation, its contribution to the
change in action at the time end-points has been omitted. The Lagrangian
density at the point of variation 90(r, A,0),, appearingin the secondof the
surfaceterms multiplied by the variation in the surfaceof the subsystem,is
defined in eqn (8.217)and is seento be proportional to the Laplacian of the
chargedensity.It is through this term that the variational constraint, as given
in eqn (8.117),is incorporated into the expressionfor the variation of the
atomic action integral. As previously detailed in eqn (8.134),the satisfaction
of the constraint given in eqn (8.117)enablesone to replacethe surfaceterm,
involving the variation of the surfaceof the subsystem,with an integral of the
variation of Y2p. Sincethe Laplacian is a divergenceexpression,its variation
as detailed in eqn (8.135)yields only a surfaceterm and, thus, the Euler
equation obtained in the variation is left unchanged.Substitution of the result
givenfor the variation of the Laplacianin eqn (8.135)for the surfacevariation
term in eqn (8.218)together with the use of Schrodinger'seqn (8.211)yields
 ACTION PRINCIPLEFOR A QUANTUM SUBSYSTEM 8.5

the following expressionfor the variation of the atomic action inteeral

6 w " r r [ Y ,O ,A . d ] : o'{ottn,
r)(ihl2){u*;"1ry
nlry+ c c
l,',,'
- + ..)
lot)Y*av
Jor'tu,
+ (ih\{[o o, . ..1[: (82le)
Ja,'Y*dY
The quantity jB is the vector current density for a particle in the presenceof an
electromagneticfield and is defined as
jr(r) : j(r) - (elmc)p'(r)A(r)

: ( h l 2 mi )J d r' [y * V v - Vy*yl - (el mc)Jdz,y* yA (r). (8.220)

The first term j(r) is the normal vector current density previously defined(eqn
(5.11))and present in the expressionsfor the variation of the energy and
action functionals for the field-freecase(eqns(5.96)and (8.136). The second
term owes its existenceto the presenceof a magnetic field and is simply the
charge density p(r) multiplied by the vector potential A. The two terms are
called the paramagneticand diamagneticcurrents,respectively.The presence
of the two contributions to the total current in the presenceof an electromag-
netic field is a consequenceof the velocity being determinedby film - (elcm)A
rather than just by fitlm.Thus,as in the variation of the field-freefunctionals,
the variation of the surfaceof the subsystem,together with the restriction that
the subsystem be bounded by a zero-flux surface, causes the quantum
mechanicalcurrent density to appear in the variation, thereby transforming a
generalmathematical relation into a physical result. The presenceof the full
magnetic current j" is essentialto the description of the propertiesof an open
system in a magnetic field.
Proceeding as before in the field-free case, the variations in the state
function are replaced by operators which act as generators of infinitesimal
unitary transformations.That is, dY : ( - ilh\FY where F is an infinitesimal
Hermitian operator 1F : ed). Introducing the notion of generatorsinto the
result for the variation of the atomic action integral yields

L f , . r [ Y , Q ,A , 0 ) : F ( O ,t ) - F ( Q ,r r ) - ot(o, r) {(0SlAt)po
1,",'orf
- tJi.(.)+ j(J.(r)+ cc)l'n(r)). (8.221)
The result is expressed
in terms of property averagesfor l/ electrons, so
L,7f : N6lf. The atomicaveragesof the generatorat the time end-points
F(O,0 and the corresponding propertydensityp, have been previously
8.5 ATOMSIN AN ELECTROM.{GNETIC
FIELD 409

d e fi n ed,eqns ( 8. 138 )a n d (8 .1 3 9 ),re s p e c ti \e l )' a

. s has the paramagneti c
contribution to the magneticcurrent densityfor property F, eqn (8.140).The
diamagneticcontribution to the total magneticcurrent densityfor property F
is given by

Ji(t) : - (elmc)p,(r)A(r). (8.222)

To re-expressthis result in the form given in eqn (8.143)for the field-free
case,we need the Heisenbergequation of motion for F(Q, r) for a system in
the presenceof a rnagneticfield. This is obtained in the same manner as eqn
(8.141)in the field-freecase,using eqn (8.211)for the Schrodingerequation of
motion to give

t)ldt : (+){(ilh)<lH,
d.F(Q, Fl)o + cc}
+ f ds(Q,r\{@slrl.)po
+ cc)l'n(r)).
tJF(.)+ +(JF(r) (8.223)

Comparison of this expressionwith that for the change in action in eqn

(8.221)shows that the terms subtractedfrom the end-point averagesof the
generatorare just those which account for the surfacecontribution to this
differenceintegrated over the time-like surfaceconnecting the two time end-
points.Thus, as in the field-freecase,the changein the atomic action integral
can be expressedentirelyin terms of the interior averagesof the generatoras

l,'
Lf ,,2[Y, O, A, 0f :
I
Jtr
dt{(ilh)(tF,Fl)o + cd112, (8.224)

a result equivalent to the statementsof stationary action obtained for

the total system(eqn (8.144))and for an atom (eqn (8.143))in the absence
of an electromagneticfield. Statementscorresponding to eqn (8.145)for
6 g l Y,O , A , @l and to e q n s (8 .1 4 7 )a n d (8 .1 4 8 )for the stati onary-state
functionals 9llt,o, A, @] and Ebll, Q,A, d] are also obtained, with the
understandingthat the Hamiltonian appearingin theseexpressionsincludes
the interaction Hamiltonian definedin eqn (8.213)'
Thus, the definition of an atom as a region of spacebounded by a zero-flux
surface in the gradient vector field of the charge density with properties
predicted by quantum mechanicspersistswhen the systemis in the presence
of an electromagneticfield and the chargedensityis correspondinglypertur-
bed. We shall demonstratethis by determining the Ehrenfestforce acting on
an atom in a molecule in the presenceof an electromagneticfield and the
correspondingstatementof the atomic virial theorem.This is followed by an
illustration of how a field-dependentproperty, namely the polarizability,can
be defined and determined for an atom in a molecule.
-110 A C T I O N P R I N C I P L EF O R A Q U A N T U M S U B S Y S T E M 8.5

8.5.3 Atomic force and virial theoremsin the presenceof external fields

The statementof the atomic action principle which enablesone to calculate

the propertiesof an atom in a moleculein the presenceof an electromagnetic
field is (compareeqn (8.145)for the field-freecase)
6slY,O, A, Of : i{\lh)<lH, Fl)o + cc}. (8.225)
The Hamiltonian appearingin eqn (8.225)for a systemof electrons,charge
: - e, and nuclei with general charge * Z,e is

f i : 1 , , { r t 1l 2 m- L o Z , r r l x " - r , l ) - t - e O F ) }

+ II e 2 ( l r- i t ; l ) - '
i<j

- xpl)-t *Loz,eo(x")
+ II zoZoez|x"
a<p

:L,in'+ t + O (8.226)
with the momentumoperatorfr, givenby
rti: Fi + @lc)A(r,); (8.227)
the potential energy contributions have been separatedinto the normal field-
free term V and another, (D,containing the effectsof the external field @.One
may determine the variation in the atomic Lagrangian by subjectingthe state
function to the infinitesimal unitary transformation generatedby the oper-
ator F. This procedurewas illustratedin Section8.4 in the derivation of the
atomic force law and the atomic virial theorem, the samederivations that are
to be consideredhere.However,as pointed out in the caseof the stationary-
state variations, the variation of the functional can be evaluatedin a simpler
manner by using the expressionobtained prior to the substitution of the
Heisenbergequation of motion. Thus, the expressionfor the variation of the
atomic action integral given in eqn (8.221), when divided by the time
differenceL,t: tz - tt and taken to the limit of A,t -- 0, yields an alternative
expressionfor the variation of the atomic Lagrangian

6glY,Q, A, Ql: d F ( O ) / d r- $ d S ( Q ,r ) { ( 0 s l ? t ) p o
- tJF(')'n(r)+ 1(J"(.)'n(r)
+ cc)l). (8.228)
Note that the result in eqn (8.228)is given for the generalcaseof a particle of
charge e, as is done throughout Sections 8.5.1 and 8.5.2.When used in
conjunction with the electronic Hamiltonian in eqn (8.226)to obtain ex-
pressionsfor the electronic force and virial, the charge must be replaced by
- e. Equating the results for 69 in eqns (8.225)and (8.228)is equivalent to
using the Heisenbergequation of motion (eqn (8.223))for the generator F.
However, the equating of the two separateexpressionsemphasizesthat the
8.5 A T O M S I N A N E L E C T R O M A G N E T I CF I E L D 4tl

resultsare obtained from a variational expression,the atomic statementof

the principle of stationary action, one that is restrictedto subsystemswhich
satisfy the zero-flux boundary condition. Equation (8.225)is, in effect,a
variational statementof the Heisenbergequationof motion for the generator
F. Rs has been stressedin previousapplicationsof the atomic statementsof
the action principle, the mode of integration of the variational functional
determines the definition of the average value of an atomic property. The
coordinatesappearing in the generator .F of the action principle and denoted
by the position vector r are those of the electron whose coordinates are
integrated over the basin of the atom, the coordinates of the remaining
electrons being integrated over all space and the spins of all electrons
summed.
Ehrenfest'sfirst relation, the definition of the velocity and its associated
momentum, is obtained using the commutator of Ii and the position vector r.
This commutator, using the above Hamiltonian and multiplied by m, does
indeed yield f as the momentum of an electron in an electromagneticfield
: ;.
m(ilh)[ri,l] : +(ilh)1rt2,i1 (8.22e)
The evaluationof the right-hand sideof eqn (8.228)for F : m? and equating
the result to the right-hand side of eqn (8.225)then yields an expressionfor
the atomic average of the momentum fr in the presence of a varying
electromaeneticfield

z(o) :*1, drr0p(r)l0t+ r) + cc

;^{ds(o, {J,(r)'n(r)
p,A(r)' n(r)1.
- (elmc) (8.230)

The atomic statementof the Ehrenfestforce law is obtained by setting the

generatorF equal to the momentum f. The commutator of the many-particle
Hamiltonian eqn (8.226) and fr can be expressedas
(ilh)lH,frl : \lh)lv - eQ$), rtf
0l + 6l2mh)1rt}, (8.231)

where 2 representsthe potential energy contributions from just the internal

electric fields, as defined in eqn (8.226).All of the potential energy terms
commute with the vector potential A(r). The first of the commutators on the
right-hand side of eqn (8.231) gives the force exerted on the electron at
r1 : r by the nuclei,the other electrons(seeeqn (8.181)),and the externalfield
QF),
: - vv -r eYS$\,
\lh)lt - e6@,01 (8.232)
while the second,which does not vanish as one might guessat first sight,gives
the force exerted on the electron by the magnetic field
fr) :
(il2mh)1rt,2, - (el2mc){fx B - B x f}. (8.233)
412 A C T I O N P R I N C IP L E F O R A Q U A N T U M S U B S Y S T E M 8.5

This latter commutator doesnot vanish becausethe componentsof f do not

commute amongstthemselves. Considerthe commutators of :i2 formed with
the component i?". One has [f] , rt,] :0 and
rt) : fi.n1rt,,,,i,1
\rttr, + lftr..n,lfr,y
: - t/,f:)
{t(*-'1)
('x/
. (+- ?)r"}
tr/')
(8.234)
\c/|.'\dy \,'y
Each term in the curly brackets in eqn (8.234)is the negative of the z
component of the magnetic field B : V x A; hence,

rt,f: *(:) {rt,n"

rrttr, + B"rty} (8.235)

Similarly,

lrt],rt; : -.(:){rt,', + Bnrt') (8.236)

and, thus, as required to obtain eqn(8.233),

: - "(:) {n,rt"- 8"rt,+ rt,"ny

ffr2,rt,,1 - rt,n"}

: --(:){(n * rt).-(r x B),} (8.237)

where (B x f), denotes the x-component of B x f. The averagingof the

commutators in eqns (8.232) and (8.233),as requiredby eqn (8.225),and
multiplication by N gives the force exerted on the electronsin the atom C),a
quantity denoted by F(O, B, 0, t)

N
F ( OB
. ,d , t ) : . { f i n < f f i , n ) ) r , * c c }: ( - V t ) n + ( e V @ ( r ) ) n
L

- (el4mc)t((,tt B - B x ft))o * cc) (8.238)

where ( - vt;n lsee eqns (6.10)and (6.17))is the force in the absenceof
external fields and the remaining terms give the atomic average of the
contributions of the external electric and magnetic fields to the Ehrenfest
force.
This resultis to be equatedto the time rate of changeof f and the flux in its
paramagneticand diamagnetic currents through the surfaceof the atom
using eqn (8.228).Thus the forces exerted on the electrons in an atom O
arising from the electric fields within and external to the molecule and from
8.5 A T O M S I N A N E L E C T R O M A G N E T I CF I E L D 413

an external magnetic field are equivalently expressedas

F ( Q ,B , 6 , t ) : m ( @ J , l A r ) ) o- $ d S ( C lr,) F ( r ) ' n ( r )

* f ds(o.';[111';r6;
+ J(r)A(r)|
:
'
+ p(r)Atr)Atr)]'n(r) (8.23e)

where F'(r) is the quantum mechanical stress tensor, eqn (8.173).In the
absenceof external fields the surface integral of the stress tensor and the
atomic averageof the time rate of changeof the current densityJ(r)determine
the force, eqn (8.175).The stresstensor has the dimensions of an energy
density, that is, of a force per unit area or a pressure,and the force acting on
the interior of the atom is given by the integral of the pressureacting on each
of its surface elements. A current density and (elmc)Ap both have the
dimensionsof a velocity densityand their multiplication by (elc)A,as occurs
in eqn (8.239),converts them into energy densitiesor pressures.Thus, the
secondsurfaceterm appearingin the atomic force law measuresthe contribu-
tion to the force acting on the interior of the atom that results from the
paramagnetic and diamagnetic pressuresacting on its surface.
For a stationarystateof a systemin the presenceof uniform externalfields.
the expressionfor the force becomes

F ( OB
, ,i l : < - VV)n+ (eV@(r)),,
-(el4mc){((,i t B- B x f))o* cc}. (8.240)

These interior forces are generatedin their entirety by the mechanical and
magnetic pressuresacting on the atom's surface,i.e.

F ( Q ,B , d ) : -
$ d S ( Or,) F ( r ) ' n ( r )

!$ortn,'y[1,o1.;.r1,;
+ J(r)A(r))

+ (8.24r)
fto@N')A(')]'n(r).
As for a stationary state in the absenceof fields (eqns (6.14)and (6.18)),eqn
(8.241) states that the force acting on an atom in a molecule is totally
determined by the flux in the forces acting on its surface.The same result is
true in the generalcaseof time-varying fields (eqn (8.239))if one includes the
basin term arising from the changein the current with time with other basin
contributions to the force actins on the interior of the atom.
4r4 ACTION PRINCIPLEFOR A QUANTUM SUBSYSTEM 8.5

The atomic virial theorem for a systemin the presenceof an electromag-

netic field is obtained by setting the generator F equal to i.f. The com-
mutators requiredfor the evaluationof 69, eqn (8.225),are readilyevaluated
giving
(ilh)lH,?'rt): (l2mh){\rt',il .,4+ i.1rt,2,
rtf}
| (ilh)lv- e$(r),?'rt1. (8.242)

The results obtained for these commutators parallel those for the field-free
case,eqn (8.191).Thus the first commutator yieldstwice the operator for the
kinetic energy,expressedhere in terms of f (seeeqn (8.226))

( i l 2 m h ) l r t . t , f ) .:f r r t 2l m : 2 i n , (8.243)

while the last yields the virial of the gradient of the potential energy operator
V and external field @(r)with respectto the electron of position vector i, the
virial of the force operator given in eqn (8.232)

(ilh)lv - eg(r),i'rtf : - r. V t + tr. Vd(r). (8.244)

The commutator of rt,2with f gives the magnetic force acting on an electron

(eqn (8.237))and the remainingcommutator in eqn (8.242)givesthe virial of
this magnetic force

( i l 2 m h ) r ' j r t ' :, r t-f {f x B - B ",4}. (8.24s)

#r'
The averagingof thesecommutators in the manner determinedby the atomic
action principle (eqn (8.225))serves to define the atomic averagesof the
kinetic energy and of the virials of the internal and external forces acting on
the basin of the atom. i.e.

^
-N{ t i l t l ( [ H , i . f ] ) " + c c ] : (/ (/ 0 + : "n ( r ')\)2 \ i ^ + ( - r . v t S n
z' c | )
\\' /n'
e(r'V@(r))"

^*t(r.(f x B - B x f ) ) , ,* c c )

: ,;:;, + rb(o)+ "/ro(0)+ rB(0). (s.246)

Thesecontributionsareequalto thetermsgivenby the alternativeexpression

for the variationin the atomicLagrangianfor the generatori.f,, eqn(8.228).
8.5 A T O M S I N A N E L E C T R O M A G N E T IFCI E L D 415

which are

6 glY , Q . B . Q f : (m(r '6J, "(r)/dr))n

+ $ d s ( Qr, ) { ( - h 2l 4 m ) Y p ( r ) ' n (-r ) r ' E ( r ) ' n ( r ) }

* :f ds(o,
,), + J(r)A(r))
[{A(r)J(r)

* ! , p ( r ) A ( r ) A ("r,)rl,
: - t;(o) - {r(a) - 7'3(o). (8.247)

The first term (1](A)) is the virial of the force over the basin of the atom
arising from a varying magnetic field. The first of the terms in the first surface
integral vanishesbecauseof the zero-flux surfacecondition defining the atom.
The second is the virial of the mechanical pressuresacting on the atomic
surfaceas determinedby the stresstensor F. The secondsurfaceintegral is the
virial of the magnetic pressuresacting on the atomic surface.Equating the
two expressionsfor the variation in the atomic Lagrangran given in eqns
(8.246)and (8.247)followed by some rearrangementyields the virial theorem
for an atom in a molecule in the presenceof an electromagneticfield,
- 2T*(Q\: {(Q, B, d) : {v(Q) + {o(a) + {B(o)

+ rr(o) + {s(o) + {8t1c21. ( 82 4 8 )

In addition to the basin virial {o(A) arising from the internal forces,which is
identicalto that obtained in the field-freecase(eqn (8.192)),there are furthcr
contributions to the virials of the forcesacting on the electronsin the basin of
the atoms resulting from their interaction with the external fields.The surface
virial %(A) reduces to the flux in the virial of the mechanical stresses
-r-? . n A s i n the caseof a
in t he ab s e n c eo f a n e x te rn a lfi e l d ,e q n (8.194).
stationary state,only the value of the total atomic virial is independentof the
choice of origin for the vector r.
As discussedin Chapter 6, the atomic statementof the virial theorem, since
it definesthe atomic averagesof the kinetic and potential energies,servesto
define the energy of an atom in a molecule.The averageelectronic potential
energy is equal to the virial of the forces acting over the basin of the atom.
Thus, the energy of an atom in a molecule in the presenceof an electromag-
netic field is defined to be
+ r(Q, B, d) : - r"(O).
E(o) : T*(lo-) (8'249)
This energy possessesthe characteristic,as do all atomic properties, that it
yields the averagevalue for the total systemwhen summed over all the atoms
416 ACTION PRINCIPLEFOR A QUANTUM SUBSYSTEM 8.5

in the molecule
E":LLE(Q) : (l1) (8.250)

where the Hamiltonian operator is definedin eqn (8.226).The secondequality

assumesthat no external mechanical forces act on the nuclei. The atomic
contributions to the molecular polarizabllity and magnetic susceptibilityare
defined and discussedin the followins section.

8.5.4 Atomic contributions to molecular polarizability and susceptibility

An atomic property, as discussedin Chapter 6, is determined by averaginga

correspondingproperty density over the basin of the atom. The property
density,however,has the essentialcharacteristicthat it includesthe effectsof
all the particlesin the system,as determinedby the property in question.This
is a result of the operator being defined for the interactions experiencedby a
singleelectronand of the mode of averagingthis operator over the motions of
the remaining particles in the system. The density which determines the
Ehrenfest force acting on the electrons provides an example of an atomic
property which illustrates all of these points. The operator is - V1 t where
the gradient is taken with respectto the coordinates of a singleelectron,and
Z is the many-particle potential energy operator. As illustrated and discussed
in Chapter 6, the operator - V1V determinesthe total force exerted on the
electronat r, by all of the other particlesin the system.(It is worth noting that
if one sums this operator for all of the electrons, the electron-electron
contribution to the force vanishes.)The averagingof this operator over all the
coordinates but 11 = r by the usual recipe, as determined by the atomic
action principle, yields the force density
F ( r ,r ) : l f j d r ' Y * ( - V r V ) V . (8.2s1)
This density is l{ times the force exertedon the electron at r by the remaining
particlesin the system,the motions of the latter particles being averagedover
the total system.Sincethe electronsare indistinguishable,multiplication by
l{ gives the force exerted on the total density of electronsat the point r and
integration of the density over the basin of the atom yields the total Ehrenfest
force acting on the averageelectron population of atom Q. Becauseof the
single-particlenature of the definition of a property density, the sum of the
atomic contributions over the atoms in the molecule yields the molecular
value for the property. The atomic properties are additive, an essential
characteristicfor the atoms of chemistry. It should be emphasizedthat the
definition of an atomic property and its consequencesare a result of the
existenceof the atomic principle of stationaryaction,as opposedto being the
result of necessary,but otherwisearbitrarily imposed definitions.
8.5 A T O M S I N A N E L E C T R O M A G N E T I CF I E L D 417

The determinationof a property densityat somepoint in a moleculeby the

total distribution of particles in the systemis essentialto the definition of
atomic contributionsto the electricand magneticpropertiesof a system.The
densities for properties resulting from the molecule being placed in an
external field must describehow the perturbed motion of the electron at r
dependsupon the field strength everywhereinside the molecule,a point that
has been emphasizedby others (Maaskant and Oosterhotr 1964). This
requirementis met by the definition of an atomic property as determined by
the theory of atoms in molecules.Property densitiesfor a molecule in the
presenceof external electric and magnetic fields have been defined and
discussedby Jamesonand Buckingham (1980)and the presentintroduction
follows their presentation.
In the presenceof an external field, additional terms contribute to p, the
electronic dipole moment of a molecule.These contributions may be ex-
panded in powers of the applied field
p: p(o\ * pttt * p{zt * (8.252)

This expressionmay then be used to definecorrespondingquantitieswhich

describethe responseof a system to the external field

I t : l t ' o *' d - 6 + i f : 6 2 + t 8 . l - r lt

Equation (8.253)definesthe polarizabilitytensora as the quantity determrn-

ing the flrst-orderresponseof a molecularchargedistribution to the applied
field. The state function can also be expandedin terms of the field e to girt'

Y: y t o l + Y ( 1 t 8+ t V ( 2 ) 8 2 (8.15-lt

and, correspondingly,the chargedensity correct to first-order in the field is

given by

p(r): 'fo' , Y *Y

r,'Id,,{Vtot*\y(0)
+ [Ytot*ty{t) 1 ty(t)*ty(o)]d +

: p( o) ( r )* p{rt$)8 + (8.255)

where,as before,Jdr'implies a summationover all spinsand integrationover

all of the electroniccoordinateswith the exceptionof those denoted by the
position vector r. The first-ordercorrectionto the electronicdipole moment
is related to the first-order correction to the density through the equation

tt: - e J r p ( r ) d r : - r [ r p t o ' ( r ) d-r , l r p ( t ) ( r ) d r ' 8 , (8.256)

418 ACTION PRINCIPLEFOR A QUANTUM SUBSYSTEM 8.5

which, upon comparison with eqn (8.253),gives

F ( r ): _ e l r p t r t ( r ) d rg. : d . g . (8.2s7\
One may define the electric polarizability density a(r) (Jameson and
Buckingham 1980)as

a( r ) : - e r p (r)(r1: - e l ,,l tytt)* .y(o)), (8.253)

J d r' 1 ytot* rY (1)1
which has the property of yielding the polarizability of the molecule when
integrated over all space.The integration of a(r) over the basin of an atom
does not, however, yield the total atomic contribution to the polarizability
tensor of a molecule.This is becausethe surfaceof the atom is also perturbed
by the applied field I and the total atomic contribution to the atomic
polarrzability consists of both a basin and a surface term. The atomic
contribution to the electronic dipole moment in the presenceof the field is
obtainedby restrictingthe integrationof eqn (8.256)to the basin of an atom.
We denote the basin of the atom in the field by Or : Oo * dQ where eo is the
unperturbed basin. Retaining terms to first-order only yields

,l rp(r)dr

_,T,,
JOr

,l
JOo
rp(o)(r)dr -rl
JQo
rP{rt6r'n rp(o)(r)dr

- (C))+ otttlo).
P(o)
Thus,the first-ordercorrectionto the atomicelectronicdipolemomentis

t(') (O): -,1," -,T,, r p ( r )( r ) d r ' 8 rp(o)(r;dr

where the final term is dependentupon the field d through the changein the
basin resulting from the change in the atomic surface.The atomic contribu-
tion to the polarization, a(O), is determined by a basin and a surface
contribution,
a(O):dr(Q) +ar(Q). (8.25e)
The basin contribution can be expressedin terms of the polarizability density
a( r ) as

a u (O ) : lr a(r)dr (8.260)
JO

and the surfacecontribution

ar(Q) : - r l r p ( o ) ( rd) r . B - |
Jdo
8.5 A T O M S I N A N E L E C T R O M A G NE T I C F I E L D 4t9

The expressiongivenin eqn (8.259)fora(O)isequivalentto the resultgivenby

the calculus of variations for the first-order changein an atomic property. In
the present case,the variation in p(O) is

dp(o) : -,[, rdp(r)dr'E- ' r)dS(O,r)rp(r),

$ottn'
that is, as a variation of the property over the basin, plus a contribution
from the variation of the atomic surface.The surface contributions cancel
when summed over all the atoms in a molecule becausethe shifts in the
interatomic surface between atoms Q and Q' are of opposite sign, that
is dS(AlO') : - 65(C)'lQ),and the polarizabllity of the molecule is obtained
by summingjust the basin contributions
a : Io d o (O ). (8.261)

To put the definition of this property into direct correspondencewith the

definition of other atomic properties,as one for which the property density at
r is determined by the effect of the field over the entire molecule,we express
the perturbed density in terms of the first-order corrections to the state
function. This is done in a succinct manner by using the concept of a
transition density (Longuet-Higgins 1956).The operator whose expectation
value yields the total electronic charge density at the position r ma1' be
expressedin terms of the Dirac delta function as

D ( r ): l { d ( r - r ' ) . ( 82 6 1 )

From the definitiveproperty of the d function,the zero-ordercontribution to

the dipole moment is, for example,given by (the unperturbed state functions
a re denot edby l0) , l k ))

' : -
to)dr
lt(r) : - el{(0lrl0)
e(0lJrf(r)drl0)
l,o
showingthat the electronicdensityof the unperturbedstateis
p o o ( r:) ( 0 l D ( r ) 1 0:) p ( o ) ( r ) . (8.263)

The transitiondensitypor(r)is similarlydefinedas

por(r): (010(r)l/c) (8.264)

The transition density describesa charge distribution whose dipole moment

is equal to the transition moment connectingstatesl0) and lk), the matrix
elementwhich determinesthe admixture of the state lk) to the state function
for the molecule when it is perturbed by an external electric field'
The first-order correction to the charge density for a molecule in the
presenceof an external electricfield, as definedin eqn (8.255)and expressedin
 420 ACTION PRINCIPLE FOR A QUANTUM SUBSYSTEM 8.5

terms of the first-order correctionsto the state function l0), is

Dou
p , , , ( r ) : 2' LFkl , ^ , (r r ) (8.265)
Eo - kPot
where Doo is the transition moment connectingthe ground state with the
excited state lk)

Dor : - e[rpoodr. (8.266)

The first-order correction to the chargedensity is independentof the choice of
origin for the dipole moment operator, as the extra contribution to Doo
arising from a shift dR in origin is - edR(Olk), a term which vanishes
becauseof the orthogonality of the zero-order states.The expressionfor the
electric polarizability density a(r) using the above expressionfor the pertur-
bed density is

Dno
c ( r ): - 2 ( tLrg;-TrrPor
-' (r). (8.267)

One may obtain from this equation an expression for the polarization
susceptibility

( 0 l r d ( r- r " ) l k ) ( / . l r d ( r- r ' ) l 0 )
d(r'rr"):2r'L,o . (8.268)
Eo-Eu

This expression,without the specificperturbations givenby - €r, is what

Stone(1985)calledthe chargesusceptibilityof a system.It containsall of the
informationneededto evaluateany of the molecularpolarrzabilities. The
polarizabilitydensity,asdefinedin eqn(8.258),
is obtainedby integratingthe
polarizatronsusceptibility
over the coordinater"
a(r'): [dr" u,1r',r"1, (8.26e)
an operation in complete analogy with the procedure used to obtain the
definition of the Ehrenfest force density given in eqn (8.251). Thus, the
contribution of the basin term to the atomic polarizability is determined by a
double intesration.
1

:
du(O) d,(r',r"), (g.270)
Jrdr'Jdr"
the first describing how the density at r' dependsupon the effect of the field
over the entire molecule,the secondyielding the contribution of theseeffects
to a particular region of space.
A parallel developmentof the atomic contribution to the magnetic suscep-
tibility of a molecule is readily obtained, the chargedensity being replacedby
the vector current density.The magneticdipole moment m is definedin terms
8.5 A T O M S I N A N E L E C T R O M A G N E T I CF I E L D 421

of the currentdensityJ(r) as
1
^:iJr x J(r)dr, (8.27l)

and the magneticdipole momentdensitym(r),in analogywith the electric

dipoledensitygivenin eqn (8.256),
is

m(r):fr x "l(r). (8.272)

The expansion of the magnetic moment in powers of the applied uniform

magnetic field B yields

rr:p(0) * m ( 1 )+ m ( 2 )+ . . . : m ( 0 )+ L . B + (5.213)

where m(0)is the intrinsic magneticmoment and the field-inducedmoment

m(1)- I'B, is proportional to the magneticsusceptibility;. Thus, a magne-
trzation density 1(r) can be defined as (Jamesonand Buckingham 1980)

r(r).8: +r x J$)1r) (8.274)

where, in analogy with eqn (8.258)for the definition of the polarizability

density in terms of the first-order density, J(t)$) is the first-order correction
to the current density in the presenceof a magnetic field. The expressionfor
this field-inducedcurrent (compareeqn (8.220))including the field strength B
and the sign of the electronic charge is
-ehBf e2
Jb ', ( r ) : - # 1 )*y y rto r- y rt0 } v* Y ( 1 , } - :- A p,o,(r). (8.275)
lA t' { y rt
r m J m c

By- using the above two equations, the magnetization density can be ex-
pressedin terms of the zero-order and first-order corrections to the state
function

XF): #(Y(1)*r x VY(o)- Y(orrx VY(l)*)

e2
- ;---p(o'(r)(r'rI - rr). (8.276)
+mc-

The atomic contribution to the magnetic susceptibilityis obtained in analogy

with the definition of a(O),by a basin contribution obtained by integration of
/r) over the basin of the atom,

/u(A) - JI "7(r)dr. (8.277)

422 ACTION PRINCIPLEFOR A QUANTUM SUBSYSTEM 8.6

but the surfacecontribution vanishesin this caseand the molecular value

is s iv en bv t he s u m o f th e b a s i n c o n tri b u ti o ns
X : T,stxu(A)' (8.278)
While an electricfield resultsin a first-order changein the electron density
and causesmeasurableshifts in the interatomic surfaces,the first-order
correction to p(r) vanishesfor an applied magneticfield, sinceY(1) is purely
imaginary. The reader is referred to recent work (Laidig and Bader 1990;
Keith and Bader 1993;Bader and Keith 1993)for the application of these
ideas .
Other magneticproperties,such as the nuclearmagneticshieldingdN, can
be expressedin terms of a density by using the perturbed vector current in
eqn (8.275)(Jamesonand Buckingham 1980).Thus, the magneticshieldingof
a given nucleusin a moleculecan be expressedas a sum of atomic contribu-
tions.Becauseof the rrr r dependenceof this property,one anticipatesthat the
major contribution to oN will be from the basin of the atom containing the
nucleusin questionoN1Q1. The effectof a changingenvironmenton aN(Q)and
on its transferability between molecules can be studied in a direct and
quant it at iv em a n n e r.

8.6 The action principle-in the past and in the future

The idea that natural behaviour can be consideredto be the result of the
operation of certain minimum principleshas a long history, beginningwith
Fermat's principle of least time. This principle, which arose in the study of
optics, was enunciated in the seventeenthcentury. In 1744 Maupertuis
proclaimed that nature acts in such a way as to render a certain quantity,
which he termed the'action', a minimum, but it was Lagrangewho gave the
first correct formulation of the principle of least action for the generalcase.
Lagrange'sformulation of the principle sufferedfrom the necessityof impos-
ing the law of conservation of energy from the outset and the principle
languished in disuse until Hamilton enunciated the principle in what is
generallyrecognizedto be its most effectualform, Hamilton's principle (eqn
(8.43)).Hamilton later establisheda formal analogy betweenmechanicsand
optics using the original statementof the principle of leastaction and it was
this relationshipthat de Broglie and Schrodingerbuilt upon to establishthe
new wave mechanics.(A reader,interestedin the developmentof the action
principle and its underlyingphilosophy,is referredto the book by Yourgrau
and M andels t a m(1 9 6 8 ).)
I n 1933 Dir a c p u b l i s h e da p a p e r e n ti tl e d' The Lagrangi ani n quantum
mechanics'.After presentinga discussionas to why the Lagrangianformula-
tion of classicalmechanicscould be consideredto be more fundamentalthan
t he appr oac hb a s e do n th e H a mi l to n i a n th e ory,D i rac w ent on to say' For
8.6 THE ACTION PRINCIPLE 423

these reasons it would seem desirable to take up the question of what

correspondsin the quantum theory to the Lagrangian method of classical
mechanics.'In the openingsectionof this chapter,attention was drawn to the
transformation function or transition amplitude. This function relates the
state function at one time to that at another (eqn (8.26)) and it thus
determinesthe dynamical behaviour of a system.Dirac was the inventor of
transformation theory. Through repeateduse of the multiplicative law of
transformation functions (eqn (8.24))he was able to expressthe function
connectingstatesat times f , and trby a sequenceof transformationfunctions
for times intermediatebetweenthe initial and final times.Taken to the limit
of the successive intermediatetimes differingonly infinitesimallyone from the
next, the multiplicative law yields a product of all the transformation
functions associated with the successiveinfinitesimal increments of time.
Dirac then stated that the transformation function associatedwith the time
displacement from r to t + dt corresponds to expf(ilh)Ldt), where the
LagrangianL was to be considered as a function of the coordinatesat time r
and at time r + dr, rather than of the coordinates and velocities.The
transformation function then becomes exp l(ilh)Wf where W, the action
integral equal to Jl, dr betweenthe limits r, and fr, is interpreted as the sum
over all the individual coordinate-dependent terms in the succession of values
of r. With this construct Dirac was able to answer the question of what in
quantum mechanics corresponds to the classical principle of stationary
action.
Feynman built on this work and in 1948it culminatedin his path integral
formulation of quantum mechanics.In the classicallimit consideredby Dirac,
only one trajectory connectsthe systemat time f r to that at time r, and he
limited his discussionto this case.What Feynman did was to considerall the
trajectoriesor paths that connectthe statesat the initial and final times,since
he wishedto obtain the correspondingquantum limit. Each path has its own
value for the action W and all the values of exp[(i/h)W) must be added
togetherto obtain the total transition amplitude.Thus, the expressionfor the
transition amplitude between the states lq,t) and lq,r) is the sum of the
elementarycontributions, one from each trajectory passingbetweenQ,t 3t
time r, and q,, at time tr.Each of thesecontributionshas the samemodulus,
b u t i t s phas eis t he c l a s s i c aal c ti o n i n te g ra l(l l h l l Ldr for the path. Thi s i s
expressedas,

(q,z.tzlq,r.tr):
; I'-t{r't^tl,',,''a'
}aq,t'r (8.27e)

The differential 6q,(t) indicates that one must integrate over all paths
connectinEQ,r at r, and Q,zet t, and 1/N is a normalizing factor.
Schwinger'squantum action principle (1951)is developedin Section 8.2,
and a statementof this principle,in which the action integral and Lagrange
4 2 4 A C T I O N P R I N C I P L EF O R A Q U A N T U M S U B S Y S T E M 8.6

function appear as operators,is repeatedas

6 ( q , t , t z l q , t , t t :) ( iI h )( q ,z , r 2 1b6i r' r l q ,r , t r )

: (ith) e ,,,)
luj,' rta,la,,
(r,,,r, (8.280)

It is a differential statementof Feynman's path integral formulation and,

while Schwingerdevelopedit independently,it can be obtained as a conse-
quenceof Feynman'sprinciple (seefor example,Yourgrau and Mandelstam
1968).The conceptual advantagesafforded by these action principles in
formulating the laws of quantum mechanicsare contained in the quotation
from Schwingerwhich opens this chapter.These principles representmore
than alternativeformulationsof the laws of quantum mechanics.In fact, they
may provide the real foundation of quantum mechanicsand thus of physical
t heor y( G ell- Ma n n 1 9 8 9 ).On e fi n a l q u o ta ti onfrom S chw i nger(1951)regard-
ing the possibleusesof the quantum action principleis relevantto the present
work, ' . to present a general theory of quantum field dynamics which
unifiesseveralindependentlydevelopedproceduresand which may provide a
framework capableof admitting fundamentallynew physicalideas'.The idea
t hat m at t er is c o m p o s e do f a to m s i s n o t n e w . It i s i ndeedamongstthe ol dest
of scientific hypotheses.However, the idea of finding atoms within the
framework of quantum mechanicsis perhapsa new physicalidea and this is
what Schwinger'sprinciple makes possible.
The samemultiplicativecompositionlaw of transformationfunctionsused
by Dirac is usedby Schwingerto show that the 6i'; the generatingoperators
of infinitesimaltransformations.must satisfyan additive law of composition,

6fr'rr:6i't2+6i,-23. (8.28
r)
The possibilitydevelopedin this treatiseof a well-definedphysicalpartition-
i n g s c h e m e[ Q , ( r ) i; : 1 , 2 , . . . , n l a t a n y t i m e r , a s a f f o r d e db y t h e z e r o - f l u x
b o u n d a r vc o n d i t i o n

Vp (., f)' n (r, f) : 0 fo r a l l p o i n t s i n the boundary. (8.282)

ens ur est he s p a ti a l l ya d d i ti v ep ro p e rty o f th e acti on

f l ' r r : T . , w[ 'e , ] , , (8.283)

and of it s v ar i a ti o n

6 ' i, . : I , 6 i ^ 7 A ; r , (8.284)
wher e

1,.
w' lQ,f, , : J r I, l Q , ( t ) f d t . (8.28s)
 REFERENCES 425

01A,0) is the projection of the Lagrangian function operator on to the

subsystem,that is, on to the atom Qr(r). Thus, the action principle can be
expressedas a sum over the changesin the action operatorsfor all the atoms
in the total system.This decompositionof the action is illustrated in eqn
(8.286)for the generalcaseof a systemwhosestatevector is characterizedby
the eigenvaluesr of some completeset of commuting observables
6 ( u r , t z l a r r, r ) : ( l h ) L , ( a r , t r l \ O - l l J r r l a r , t r ) . (8.286)

This result, or a corresponding expression in terms of Feynman's formula-

tion, enables one to follow a particular atom or, more generally, a particular
open system, through any change in state. The open systems satisfying the
boundary condition in eqn (8.282) can assume many forms, from individual
atoms to linked grouping of atoms varying in size up to the total system.
interacting weakly or strongly with their neighbours. The questions that can
be posed in terms of the paths these fragments follow are manifold.

References

B a d e r ,R . F . W . ( 1 9 8 0 )J. . C h e m .P h v s . 7 3 , 2 8 ' 7 1 .
B a d e r ,R . F . W . a n d K e i t h , T . A . ( 1 9 9 3 ) .J . C h e m .P h y s . 9 9 , 3 6 8 3 .
B a d e r ,R . F . W . a n d N g u y e n - D a n g ,T . T . ( 1 9 8 1 ) .A d u , q u a n t u mC h e m .1 4 , 6 3 .
Bader,R. F. W., Srebrenik,S.,and Nguyen-Dang,T. T. ( 1978).J . C hem.Pft1's. 68, 3680.
D i r a c , P . A . M . ( 1 9 3 3 ) .P h v s i k Z . e i t s . S o v , . j e t u n i3
o .n6 4 .
Dirac, P. A. M. (1958).The principlesof quantummechanlcs. Oxford University Press,
Oxford.
Ehrenfest,P. (1921).Z. Pht's.45,455.
Epstein, S. T. (1974).The t,ariationmethod in quantumchemistry.Academic Press.
New York.
E p s t e i n ,S . T . ( 1 9 7 5 ) J. . C h e m .P f t y s .6 3 , 3 5 7 3 .
F e y n m a n ,R . P . ( 1 9 4 8 ) R . er. mod. Phvs.20,36l.
Feynman, R. P., Leighton, R. B., and Sands, M. (1964).The Feynman lec'tureson
physics,Vol. II. Addison-Wesley,Reading,Massachusetts. p. l9-8.
G e l l - M a n n ,M . ( 1 9 8 9 )P . h , r ' s i cTs o - d a v 4 2 ( 2 ) , 5 0 .
Goldstein, H. (1980). Classi<'almechanics(2nd edn). Addison-Wesley,Reading.
Massachusetts.
J a m e s o nC , . J . a n d B u c k i n g h a m .A . D . ( 1 9 8 0 )J. . C h e m . P h 1 ' s . 7 3 , 5 6 8 4 .
K e i t h , T . A . a n d B a d e r ,R , F . W . ( 1 9 9 3 ) .J . C h e m . P h y s . 9 9 , 3 6 6 9 .
L a i d i g , K . E . a n d B a d e r ,R . F . W . ( 1 9 9 0 ) .J . C h e m .P h y s . 9 3 , 7 2 1 3 '
L a n d a u . L . D . a n d L i f s h i t z , E . M . ( 1 9 7 5 ) .T h e t ' l a s s i t ' ut lh e o r t ' o . l - . h e l d sV.o l ' I I '
Pergamon Press,Oxford.
L o n g u e t - H i g g i n sH. . C . ( 1 9 5 6 ) P . r o r ' .r - S t t c .L. o n d o n A 2 3 5 . 5 3 7 '
M a s s k a n tw . . J . A . a n d o o s t e r h o f fL , J . ( 1 9 6 4 )M . ol. Phvs.8.3l9.
s e d i u n tP. r c - n t i c c -
o t h e m e c h a n i cos. f ' uc o n t i n u o um
M a l v e r n ,L . E . ( 1 9 6 9 ) I. t ' r t r o d u t ' t i ot n
Hall. Enelewood Cliffs. New' Jersey.
 426 REFERENCES

M e s s i a h .A . ( 1 9 5 8 ) Q. u a n t u mm e c l ' t a n i t 'vso, l . I . N o r t h - H o l l a n d ,A m s t e r d a m .
M o r s e . P . M . a n d F e s h b a c h .H . ( 1 9 5 3 ) .M e t h o t l s o / ' t h e o r e t i c a lp h 1 , s i t sV . ol. I.
McGraw-Hill. New York.
P a u l i ,w . ( 1 9 5 8 )r. n H a n d b u c h d e rP h v s i k , v o l . 5 ,P a r t I ( e d .S . F i u g g e ) p, . 1 . S p r i n g e r -
Verlag, Berlin.
Roman, P. (1965). Aduancedquantum theorl'. Addison-wesley, Reading, Massa-
chusetts.
S c h w i n g e rJ, . ( 1 9 5 1 ) .P h v s .R e t : . 8 2 , 9 1 4 .
S r e b r e n i k , S . , B a d e r , R . F . w . , a n d N g u y e n - D a n g , T . T . ( 1 9J7. C 8 )h.e m . p h y s . 6 8 , 3 6 6 7 .
S t o n e ,A . J . ( 1 9 8 5 ) M
. o l . P h y s .5 6 , 1 0 6 5 .
Yourgrau, W. and Mandelstam, S.(1968). Variational principlesin dynamicsand
quantumtheory.Dover, New York.
 A PPEN DIX

Table Al
U nits

Length I au : a o : 0 . 5 2 9 1 1 7A
: 5 . 2 9 1 7 7x l 0 1 1 m

Elementarycharge 1 a u : e : 1 . 6 0 2 1 9x l 0 - 1 eC
:4.8029 x l0- to esu
3
Charge density 1 a u : e l a 3 o : 6 ' 7 4 8 3e A
: 1 . 0 8 1 x2 1 0 1 2
Cm

Laplacian density I au : elal:24.099 eA'-t

: 3 . 8 6 1 1x 1 0 3 2 C m - s

Dipole moment I au : €ao : 2.542DebYes

: 8 . 4 1 9x 1 0 - r o C m
t
Energy 1 a u : e 2 1 a o: 6 2 1 . 5 1 k c a l / m o l -
: 2 . 6 2 5 5x 1 0 3k J m o l - '
: 2 J . 2 1 2e Y
: 2 . 1 9 4 7 x 1 0 sc m I

Table A2
Laplacian
for mapsof the electronicchargedensityand of the
Contour ualues
tli str i huti on funt't i on

Unlessotherwisestated,the outermost charge density contour equals0.002au.

T h e c o n t o u r si n c r e a s ei n o r d e r 2 x l 0 ' , 4 x 1 0 ' , 8 x 1 0 nw i t h n i n c r e a s i n gi n s t e p so f
unity from n : - 3. In certain maps,as is statedin the figure caption, the outermost
contour shown equals0.001au, the remaining contours increasingas above.
The contours of the Laplacian of the electroniccharge density (dashedcontours
ytp < 0, solid contours Y'p > 0) increaseand decreasefrom a zero contour in steps
- f 2 x 1 0 , . + 4 x 1 0 " . + 8 x 1 0 ' , b e g i n n i n g w i t h n :- 3 a n d i n c r e a s i n g i n s t e p s o f
u n i t y . T h e i n n e r m o s ts o l i d c o n t o u r e n c o m p a s s i nag n u c l e u sw i t h Z > | e n c l o s e tsh e
innermost region of chargeconcentration.
ltE APPENDIX

-
O, € f) t- S'- $ (}\ oO .a O ca 6.1 - f- O\ .f,W O - c.]
$ O O O\ \t'r \O t.) ta O .a C|\ \O O <) -f, .a O N i OO cO
C cO t-- .-r t-- t--- $ - - a\ O\ \O .a cO C..l $ $ S O\ \O oO
\O a-l tr) .a $ OO - N .! .A .a $ cf) - \f, O - c.l tr) F- O
ar
.\ Q ( J ( J ( J ( J ( J - - O O O C O O O O O O O O -
-
4
C- O\ C\.1 N N O. CO (.) $ V-) $ \O C-l C! V) \O tn <- F- OO N
--. a6 lC-O\tO-OrO\tr)\Oa. l (.O.a OOF= \O (.,l 6l .a
N
q \? O \O t .a O\ O qa \O O \O c.l r= C O c..l \O J $ O,
O\ O N :f, tt- c'.1 oO co O O - * N.a \O - O O - - -
tl o o o i o - i - : c . i o i o i o i o i o i i o i
,i l l l l l l , l r l r lr lr rr t l t l t l l l
lt lt l
l
l
oo

S - cO r- O, $ .-l N O \f, ca O 6l .a $ \O $ t-- \O'- oo

OO l-- .a $ OO .a a.] l-- c\l $ OO ca <t \O a.l OO 'r) O\ \n (\l -
t tn \o oo q! o s co ca oo oo (.i - ro oo ...l oo o \o oo o
ca - - t.) O \O rO ca - - - - - t-r F- O O ..l ca.f, F
\ J i c o J J c . i + i o o o i o o o o o o i c
\ N
t r t t i t l

*
oO > r ca t-- $ \O $ t-- O\ ci $ O oO C O O\ |rr .a .a Ct.l
al Q \9 !f, oo O. tr) co ci c\ $ .. O, ol * c. \O $ c> r- O
F- sl. q. co l-- ca Ir- oo ca tr) I-- r n - r) - - \i 6 ro oo
I o . l O O - 6 1 . i . a . . O O O - - o ] ( \ l
- c o o o o o i c i c o o i c i
O O O
o i i i o -
- - -

N c0<< <t t.i

Cl € t O\ $ OO O\ CO an N .-, <f O\ O\ N O\ oO <t - rO
O' \O tn r) t- ..) .tl - f- Or - cA Ol - ct\ C-l O,r) \O rr) tr) -
* :\ 6i c.; .d .i c.; .d ..; .d .i .d + + + $ ..i + + ,i.i ..i .d c.; -j
(..) sL
4 C q J
q
l ^

)
\ \

: .
- v
sf, t-- OO rr) (..l l-- CO = ca g\ .a .-.l $ \O OO - Or Or t-- n $ a-l
A \ 1 , v? c.l cq \ cl O "n O O co \O $ c.l ro rrr S O \O q. -r
n -
E'FX \ c.t .. c\ c-.1 ..t .l c.i .-i .. r.; ..; 6i c.i 6i c.i c.i + + + ..i ..i ."i
-
V cO c9 *f, rO - .a l-- \q oO tr- O\'\O \O oO - t-- W oO g * -
i ,
! oq 9 .1 cq v? cl q I - Q =t - oq 9 9 \ c\ oq o: q q
r- c.l * rf, s ca c.).a (.1 c.l r \f, sf, $.-1 aa o.i ca * -. j - -

Q \O - tl.) \O f--.a O O O.f, OO..l oo - \O OO OO O \O

O
- Q
o
O
r--
L S
\.) .a
tQ
\o
cO "a
.i co
€
\o
f-
€
>
Q o.
tO .r, .f,
co rc
Cl r-
o co
\O
co
ca (.) |.^)
cn oo r--
O.
-\
+
c\
a- \o $ ca l-- v) .i a-l \o \o n $ eo o r- d o !t * o\ oo
t\
\ i J J i O i O O J J J J J J i " - i c . i c . i. - iJ - '
iJ

\
(J
Q
Q
$
$
O\
...l
t/''
=
$
oO
I-*
o.]
O\
C
cO O
O'
O \O
.f, f-
N
cO
cO
cn
a..l
cO
\O
i
l--
O,
c-.t
rO
.a - O
v-r
$
t- V W W O
9.t A Q v', c..l \O {-- rO - cO N $ I-- f.- \O - t.t OO cn
f: .! O O .a $ $ $ oO \O r/-r .i .a lr) \O rO S c-r c.i - -
\ i J J J , - j J J J - j i J J J _ ' J , . . j J j _ ' J J
I
I

.n A A + ^ t y ^ ^ - N - A t -

= \ a3 x e s : € = = s = - s - s = = = = r
n q n n : q cc F- r.).t - co F r, * O - .h 6 ii - o,
- rv) c..l ol c..t - J J .-, cn "d oi e.i N c-i ,ri ,rj + .d cd rd .,i
-
-' 0)

,.
A \ a
h"h ^ h r n = r..ru.ru.rx = n = i.r h-h ;. = u.rr h
N : d - N - - ^ t _ ^ t - N N _ a ] - N :

\
-
L
ct)
I = = = = = = - = *= = - * = - J€ e 9 1 Q L
a E r d E - z ? ! z > ? i E A a : : f f I : :
 APPENDIX 429

d ^,.e 2 ='2
5 < n ! = 9 2
*,>. ? - ? hIJ*
- - \O O. t- c'i - r oO'icaO - -
\O $c tlS $ $ -...- t ' a= 6*
( . , ) r ' 1
a - O.O + ; lC c) r- - \ qp O tt- \O.a.-,.4 S I + F- . . { d
,7, r, O. - a{ O )O t.) rO "a \O C .O - v-' lt- r', - O =
- l r f -6 Pr,:
O c.i +,n ca - O F- O\ $ rO oo * oo cn O O\ O\ \O _ : . e = _

- C i c.i c.i O J c.i .d c"; J ..; .-i i J J -: J J s.i- S= i En,

. " : f e ( J
.-*-t < . 3 -9
cO tai tr) \O (-'l f- ca O l-- ,;= | tr
ca
-
t--
:
C\
O,
\O V) OO
- c..l O'
*
O\ \O O
$
.a \O O)
Q
c'l .a \O tr) ca
S
- c'l €- s .i *e . e =t
g (r) cr\ O\ oO t/') -t C..l t.) \O OO .a \t OO l- .t)
6
- -
O,
- <+ <+ A <t - Or Or f\ F- f-- O\ \O t-- q'
:f
6r\ r-- i .: 3O! - -. 2 = H
. o r o i
- - - c > c > o i J c > o i J i J J o J c > i A A
! { : v
9 \ l l - - ! =

. -3 i- - ^
r r r l l l l l l l l l l l l l l {
. .l 3 r l- -l ,;'{
6' - + Y Ed l *O - .= &q -
=.(

8
I
S = S 3 ? 5 = € ooD ' iS
i.t O cc & .r. - =,
aQSEiRe!
tr) .O O,..r T, O = T, .l 2z;a -: : ,i:j
- l l 9 " q a q- ?
F== -- c o
." i -d -- n -- ,, - -a . l-l T ?
-
-:
ru - OO O..,..1c>
1
C - ^ 35'
-
=
=
J
4
(J
Y *- 7 /- .
r r - l l l l - : E d 9 " <
?^-., :- = =-0r ? E o
t! oO

'.?
-
5=9i. = . :
!l
O\ OO O(t.i t.lO\O(.l<t- $l. <f O\a'l Oct,qf s : l d
i J . , - L

- rrl s.. trr - $ 6 or c- O c! * ci cl - q S : . e ?

-
rn c't en .- - cn -.t * o\ c'.1 o\ $ v) \o o\ s ;-
* r- cil rt
- V) t-- tn .a V-) .a C\ 10\o_ :,.9
c> O O - # a\.1 ca .a C.l $ a.l . ! ^ " ; 2
q 6 : j J -
= : -{.{,i -i,: -t,i -i A zi A A A A A d d Y L --
: J i J t U V V 9 v v v , ' . - I
J f r

- c . ; 5 ' n = F . 2

E = tI E:l=
a < < c a <\O < < c a @< c aa.l< < (...l :=:i
:' J ;- . ^F - U-
tJ *'e'I
-L r IJ - . , . -
,.i t/^) \O - t- CO r', cn
.l qoq\u?q-iq9\nnnn"1
OO \O t $ OO S
" j ' ' lO' ' ?9
l? .; -3 I ir u\ 'I -: ; rEg=?t = . =
.+ .+ a.) ca a7) aa a7) aa aa aa ca aa .a aa aa ca a', ca aa a') ==
^ , -

-.l:
, - 4 ^
"
d
L
-o
-
- - -r Y
i - L U
I- c
- l r q 4 - . - . E j =
: ' : J a ! ' - - ? -
a.i .a
OO Or $ O\ - \O f-- r) f-' rO a't \t CO q) q) $
cr\ -
"q e =q n Q - q v? cl q n n n q n - € ? 1 I ir-r:.c;?-!-n,3
.,^'*ZL=.i o r
c.i cn c.i ca ca a-r ai ca a-) a.l aa ca a', (.'] aa aa a"l a'r c'l a'l ;= - c=j c c .Y. ? co .oa: €E -
9 € . 7 : i , E2 E
(r)
co
F-. ta) O
- ro,-
*
€ O\
t/.) -
@
oo
-
ci
t.)
o\
ci
c.l
tar
$
-
(.)
..i c.i cd ..; .i
t--
F-
O\
\o
(.'l
S
ca f= q)
c-r c'..1 O
"d c.i .d c'.i c.i
Sf
or
$
@
E *-
3 :: sE zie;*
g d: er,rtr
+ + ; + "d .d + c.; t o i (.t -O = iq -,

=-:€r-i= s-.
.a * - |r) cA r/.) O\ $ tr) F- t-- (^.l t-- O $ O\ O Q Q :s:!:€ =Ez
,.n r
- -
CC)
t- N.+
q1.l F
c) -
si
c)
rc
A
-
oO
O q..c.
t.) O.'l.a
$ 3. r--- -.C
tr) \O S.tl
n c?
Q
$ 5r - si l E F-Ft=i
I n - q 9 n c9 q n n n q 9 q c\ q - -9- F3_a=
= l J : q o . r X * : E
a - l ( \ . l N * - -
J !
*. r ' -3 E t -: - o . l
-q - o
Q a 9
E : - - F a n l - - - o 1 0
Q x . : - c - = a - -
-
f-r O\ O\ $ OO r) 611 \O r/-, a-r aa oO ca O c'l O CO Q E - I 9 ' 7 , : 9 i ;
+ : o, :i'o ot rr lo io qr t^! c'r 90 t n - - O Q - - ^ ^ a -

e6 J c- O c! - \o $
! i i z E ZE ; z
d oo 5, ci\ - c> \o c''.1 l-- Cr $
.. - - - o\ n co co t-- & F- t-- t-- O o. O - - c.
; : : c > o - i c > i i o i c > J o * ' - j J J
: i s : ==r € e
E - o E ; . = 1 . : ' j
A - - - - /- - - - - - - N n -l'r, C d i = ' E I S F =
5
-
-
-,-
- +
-.
-
a-
.rr
O
"n
O
=
a.l
tn
F-
=
Or
cc $
xt -
c\
..
qJ
O
:c
I
S
r--
O
0O
.l
oo q\
.f,C
09 6_loEgFsz;^
- 6 c.t rr,r, 6 F .f, .-l c. c.) c.l - $ O - $ o] $ \Q 3 ! a 6 ? : E z -
.+' ..; .n c.i c.i ..; c'i c.i c.i ..i ..i c.i c.i c.i c.i c..i c.i .-i c.i c.i = a : i a = = ; €
- .: .c'E =.- _ = !.)
at S-:n 9d.ts.- i
a ?.2'5f:=.!1",
= x = t t , : i l I c

5Fa a F5
F n n n n flr tn n h,-h'h'F iaiEzz=i,*
t s r :
" ' . ; ; ; -

: b r : 2 ! 4 i = ^ =
l5 = " x ; i =- i=22c
> i- d i
4 a 4 Q + - T , Z C L e Z O s Z C I Q I i = = 9 a . - l - - -
8 3 # 3 8 H E o o . o( , ( - u/ - 2 2 2 6 oi l}luX =rEi 3 ./
430 APPENDIX

+ c9 O O\ .f, (.) O oo v-) l-- rO tn

= € O $ oO or\ O n o!\ O
$ O r- O\ @ oO F- - - c.l \O l-- - oi cb co \O - - O, O,
o+i=-i r,r;Jcdoo,c.i ++ii c > r . . - J c .\ io
o.l N O\ \O al cO \O $ r-- - O\ oo .a O \O - S oO O, co 6
3 - a] C.l C\.l t\ - - - c..l S .a ca .a .a (\] N (r) .a C..l (^..t C.l

tt $ cO r-- Or - O c\] tr) a] oO t-- L/) $ \O l--

q O\ oO \O oO .a $
C-l V \O - rO l- .a $ \O * $ C! :f, f- Ot - O\ .a V) O\ ca
e f O\
\O
Sy?S+J.'i
-
, r ;J i . . i
o \ ra +€i o<jrossi
ca O\ V) ..) c.] OO .a - oO r, cO i tt.- - r 6 -
3 ..).aa\lc\_

v (-a q!) o\ oo s .a tr) co o co a1 a{ o\ r-- o\ - t-. \o \o

$
ct,".) O f- cr.t - $ \c) $ - - r.) C- $ oo - r- iO v-r 6 trr
i a = t? o) o + i cd,r; o<; c.i J o. J O J i ..; ; €
\o .a \o \o \o r.) $ c..1 oo r \o lr- ri s co i ro 6 co o] o.t
3 - * - a \ c . . l N ( \ c . . t 6 t -
\
q $,- N V O\ (-.l co O\ \O S trr ca $ - r- .l Y c-l - r..t
!a (\l oo tr) Q O\ = n. c.l .! .a - O a-l - rO - rO c.t cn
v) oc c-, - rn O
q t-r rC C C rO 6 - oo \O .a .a r, ^l F r- "n
I
fl
i - ' i i 6 > i i i . . i i o J i o i i . o = + i o
t t t t l i

! ; .
I - oO $ Qp C- t/', c..l Q Or c..t - Q r/) \O n O a.t O ca \O
< b !.,
r-:
!! r- c. Q \g \g
O oo tr) C O O
qt Q
O \o
rr g\ = o
- r-- O -
$ \o \o
+ \O -
cn N <F w n
-; ; i O C)
o ? i o 6 i J c i c i o . d J . . + + c . ; i J d - - -
Ol r t t t t l

t..) o\\ - .a - \O oC .l V) C! C}\ f-- rO Or ca - oO \O t..) @ |.a)

q oo r-- O..l t--- S O.. C (.).t - O O C. r -
- S 5 O
R
C c-, e,S - - i - .-, .-, 5' r-, 5 = "n 6 O .O S c,1
i i i i i i o i i c i o o i o i

6 l
q

x r t t t l l t t
: : - d d
l

v) = o - \o c.l n .a oo = $ ci - f- f- * co a..l v oo c
q, Q.r trl O oo N trt = O F- I--.o co e c, < + =
L * (\ + -
o o Or co $ o F c-] o, + tr) - - 16 d.r i 5 c
q i i J J o c, i i i J o.i : r i - d - C - - -
!r l t t t l
L 2

Q c\.l oO $ (\l.a,- - \O tr) $ oo $ c-t O $ - <t rO rr

tr) -
Q : iO c! .-l co \O - O' c-.t Or r.- O\ + - 6 rO ob - "-,
O O\ oo r- O co t.) r-. oo r- oo F vl 6 ol 6 { F 56 O, o.
-; -i -i -i A A A A A --i
e
A
e
.:
9
-i
,J ^i
__t
^i
,j .:
:__, -l
:j -l
-J -:
-J (:^' ^'
-_

\ l l

f ** + + + + + + * * o + +

6 x 5 h rn 5 F n F n n s n F F F n b r n | ] F n F
C)

>,- I = - = : E r : r - = * = ; g - - E J 6 6 U Z a
(A< rJd nCz-GZ ><taaOi =-.5-:
 APPENDIX 431

6)
.-
a

r o. $ oo o c-l c.l t-- c.t o' !a o t-- oo \o oo c'i o o \o c'l

I-- * C- \O rn o o o \o o' o $ t/) o n o\ co t.) 09 €,r,
+ .d .o ; .i 5 -;,-i O,r; o a i + "d o .J J :r ?g .i
ri o. - .r;,rj - o s - € - - .rl $ .', .. c1 - O o' o'
: + + cn cl cn ca c.r .a c{ c! .a c\ a-] r-'l c.l c\ ct''l c'l - -

.2
O * c..l c.) oo t-- - rO oo - fQ O O O ta O\ $ O, q 9 \O
6 t xj .O 6 O. n ca O ta O .'l O\ O\ \o $ oo oo S n c'J
oi ; J : - c.i S c.i..o o,r; i,.; "d > d + ql cj 9 .o
+ vi t co C] t o' ;* \.) \o oi \.) $ c. O # - o\ O o\ o\
* C.l C..l (\ - -
C.i - - - * * - - - * C.{ (.)

tr) OO C.l r) (..l \O - *'t OO O\ O t-- OO * or OO -. O (.'l \O

n \ C e i A'? \ v?n q ! 9 - q -q q 9 n ! . . l.c '5o
.ri w s $ .\ r-- F- o Q !O $ O 9 a> .r, t, r= .! Cl il
..i - oo r.l "n rr: O A .O O t ) t.l O - c'.1 C O O' !

a t J - + * - * - t - * - -
C)

\C \C c\ C C - $ - oj $ $ .a d) O r-- .'l .-l !a/ vt -. Q

- \O oO c-.1 Or v', - \O oO - Q9 oO tr) \9 .O O\ v') O .'l !? v) .2
i-i = \i ii L v', O. r-- + c\ O tri n O c\ $ v') c.'l \ n \O

rt r t l
I
t t l

c o

- C\ ca - O - rt - (\l L.) O\ ca c'l ..) F- (\l W O Ln q \O 9E

rO i.i c"i J Or - - c") c.] CO
f-- O\ - \O oo O\ ca C''l \9 t/', i o =
- \ "1 {O - Q e n -- q oq v? "1 1 q n v? q \ q I - =
d x "
h
- c.t ctl O - O ca - O - - - * - - -
d N O O .E.5 >\
l l l l l * a ?
d * i
* C.l C\ C\ * - O .- f- f-t
H.o ?
- rO O CO .a F- O\ $ O\ O Q
co 6 di I t t 96 F- S \o ..1 F-. o' I c'l o, 9 \? .?'o o.
C,l < O\ 6...ta1 r.r1 N f-,n * f-- # tr; O'-i n O O A - -
' i i 5
i ; - J - i i i i J c > c > c > J o o i o o i i
l l l l l l l 3
cd
o?
X.q
- Y i > ,
o x.o
t-- c'.l .-) o\ r.) o\ o. Ln <- o\ .a c\ - - O O O \O l-- @ F-
- O' Or cn O' O] oO \O (tl t oO f- qA O F 6 Qr t- c? - O\ trs H
tn r.? *
O - - v', O s c'.1 o.l O c\ co - o\ t= $ o\ oo V o.l " X'1
= - - o ; - - - c.i J J o o J J i o o i id .q o.= <
: a L -
Y 6 (.)-O

fo t: .i 9- 5 . o
9 v _
Z e . ^ t

.a \o v) oo - cn $ .i \O O tn co Q d.E [i.=
oo oo I O Q Q Q =
-.
- .n .r; ri dr * 6 co c-, rr
Ln
cn O c..l $
c',
oo
tt-
Q
O
o.
.f, $
Q
O
O
(.'l
Q
O SaEe
i =€ g
d + co N \o o\ 6 O co o, O -
c ) i C , J J i - J o i o c , J J i i i o o o i
62> E
f" ! ^€
- - H - ^

a
5 e ; H
1 )

*
' -
- ? v

+ + + + - i , + + + o * + - * * - * - * * .---^
n t dr n NH -H NH -n +n Fj - r t N H N NN iN Nt r in EN -
n -H Nt r N .! or{ L
- - C > - < - a
J ' 5. - il
9
-
)
(
F
F
)
v
A
+
a

A
.
l t
-
U
,
-
= - H
. A ( ' - 7 ^ "
* T V t , t , i a t J zcocc0 L
J C C co co co co co co
'I 7 I + 4 ? = Q i r -
t'r z z z O Q l! U w
i 1g"q
* N + *
432 APPENDIX

Table A5
Radii oJ' isolated ground-stute spherical
atoms and ions*

Atom r" (0.001) r" (0.002)

H 2 .8 8 2.53
He 2 .5 0 2.27
Li 4.l8 3.29
Be 4.17 3.68
B 3.91 3.51
C 3 .6 2 3.27
N 3.35 3.06
o 3 . 18 2.92
F 3.03 2.78
Ne 2 .8 9 2.66
Na 4 .2 5 3.26
Mg 4.60 4.01
AI 4.61 4.08
Si 4.45 4.01
P 4 .2 3 3.85
S 4.07 3.72
CI 3.91 3.58
Ar ). t.L 3.42
Li* r.82 r.69
Na- 2.47 2.30
F- 3.49 3.20
cl- 4 .3 6 3.92

x I n a t o m i c u n i t s .C a l c u l a t e df r o m a t o m i c f u n c t i o n so f
C f e m e n t i ,E . a n d R o e t t i , C . ( 1 9 7 4 1A , tomic Data and
N u c l e a r D a t a T a b l e s1 4 . 1 7 7 .
 INDEX

action integral and electronegativity 196-9

atomic 379 i n i o n i c s y s t e m s3 2 1 , 3 4 1
constraints on variation of atomic in hydrogen bonded systems 303-7
action 380 relation to dipole moments 199
classical,with forces of constraint 187 sigma and pi contributions to 203-7, 306
classical.variation of 362 atomic properties
operator 370 additivity of 181
for system in electromagneticfield 404 definition of 180-l
vanishing of in quantum mechanics 379 property density for 180
variation of atomic action 381-7 relation to charge distribution 136
variation of operator 372 transferability of 2-3, 135-7, 210-19, 235
variation of in Schrtidinger atomic quadrupole moment
representation 376-7, 40'7-9 in barriers to rotation 231 3
atom definition of 183
action integral lor 379 in hydrogen bonded and van der Waals
action principle for 390 complexes 305-7
bonded radius of 202 relation to pi population 207
Dalton's theory of I atomic surface
definition of in terms of principle of definition of 29
s t a t i o n a r ya c t i o n l 6 l , 3 9 0 lorces acting on 174,413
definition of in terms of topology of magnetic pressureon 413
charge density 28 variation of I 58-9, 385-6
electron repulsion energy for 193 atomic variation principle 161, 1'/0,402
e n e r g yo f 1 9 0 atomic virial
equivalenceof topological and quantum definition of 177
definitions 29 and electronic potential energy 191
non-bonded radius 199 atomic virial theorem
nuclear repulsion energy for 194 and definition of electronic energy of
observational basis for 133-6 atom 191
surface of 29 for stationary state 177
atomic basin for system in electromagneticfield 415
coordinate system for 103 lor time-dependentsystem 401
definition of 28 atomic volume 182
atomic charge 182;see a/so atomic attractor
population basin of 28
atomic first moment 183, 305, 199, 231,233 definition of 28
atomic force law, integrated atomic force law
and model of pair-wise atomic
interactions 175,289 basin virial 176, 192
for stationary state 174 behaviour space 55
for system in electromagneticfield 413 Berlin's binding and antibinding regions 316
for time-dependentsystem 396 bifurcation structure 6l
atomic graph 262 bond critical point 33
atomic interaction line 32,289 bonded atoms
atomic interactions definition of 32-3
c o n d i t i o n sf o r a b o u n d s t a t e 3 3 , 3 1 8 , 3 2 8 and electrostatictheorem 316-22
classificationof 290-7. 307-14 necessaryand sufficientconditions for
atomic population 32-3,73-5
in compounds of carbon, sulphur, and in propellanes 71-5
o x y g e n 3 11 - 1 3 and virial theorem 322-28
definition of 182 bond ellipticity
in diatomic molecules 197 in boranes 330-l
 434 TNDEX

bond ellipticity (contd.) formation of 289

and changesin structure 84 6 necessaryconditions for existenceof 316,
definitionoi 78 318
in three-memberedrings 84, 329-31 chemistry
bond energy 242 5 cornerstone of 137
bond order 75 definition of I
bond path 33, 73-5, 318 c l o s e d - s h e liln t e r a c t i o n
bond path angle 76 definition of 293
bond path length 78 in hydrogen bonds 302
in van der Waals complexes 304
co-dimension I l3
cage structure 37 completenesscondition 357
calculrrsof variations 150-4 conflict structure 63,69-70
c a n o n i c a lt r a n s f o r m a t i o n c o n j u g a t em o m e n t u m 3 6 4
definition of 365 conservation law 139
infinitesimal 366,368, 369 contact transformation 365
Dirac and 367 control space 55
catastrophe,catastrophe theory co-rank 1 13
of the elementary type 1 12 critical point
the ellipticumbilic I 15 c o - r a n k 11 2
the fold 113 definitionof 16
formation of a cage 121 and definition of molecular
and Morse theorem I I I structure 21-2, 37-9
openingof a ring I 14 d e g e n e r a t e1 8 , 6 1 , I 1 2
and splittingtheorem 112 Morse 111
catastrophe point 65,90 rank and signature of 18
catastrophe set 90 singularity 112
characteristicset 39 stable and unstable manifolds of 92
charge density and structural stability 94
in compounds of carbon, oxygen, and Taylor seriesexpansion of charge
sulphur 2M density 49, ll2
definition of 6,132-3 current density, see vector current density
determination of 7-8 curvature 16
and determination of atomic
properties 2-3, 136, 209-16
in diatomic molecules 134,200-1,340 density functional theory 214,235-6
in diborane 20 density matrices
dominant form in 14 expansion of first-order matrix 237
in ethane 227 first-order l l
in ethylene 15,30 Nth-order 333
gradient vector field of 22 second-order I 1, 335
Hessian matrix of 17. 47 differential force law
in hydrocarbons2ll f o r s t a t i o n a r ys t a t e 1 7 4
in lithium clusters 44 for time-dependentsystem 396
focal maxima in 14,40,43 Dirac's quantum correspondenceof classical
long-range behaviour of 45 action principle 422-3
and nuclear-electroncusp 45 dyadic 165
in orbital approximation 9
in propellanes 74, 122
r o l e i n d e f i n i n gs t r u c t u r e 1 3 2 - 3 E h r e n f e s tf o r c e l 4 l , 1 7 4 , 1 7 8 , l 9 l , 4 1 6
in sodium chloride 23.340 Ehrenfestrelations 141
in tetrahedrane 21. 38 electron delocalization 341
topological properties of 14 electronegativity 196-9
value at nuclear position 40, 45 electron localization
in water molecule 56,221,263 in atomic cores 251
chemical bond and Daudel's loge model 333 5
classificationof 290 and the Fermi hole 250-1. 338-42
 IN DEX 435

and fluctuation in electron and principle of stationary action 371

population 335-9 temporal generator 366, 3'74
i n i o n i c s y s t e m s3 4 1 gradient vector field
and the Lewis model 248 for bifurcation point 56, 100, 116,124
and the pair density 250, 336-8 for cage structure 38
physical localization 251, 334 of charge density 22-4
electronic energy for coalescenceof bond critical point and
of atom in molecule 191 nuclear maximum 45-6
in bound and unbound states 327 for conflict structure 57, 62,69, 98
definition of 187-9 equivalenceof 57-8, 89
variation principle for 187 for non-nuclear attractor 44
electronic Hamiltonian 189 of nuclear potential 97 101
electrostatictheorem phase portraits of 25-7
a n d B e r l i n ' sb i n d i n g r e g i o n 3 1 6 for ring structure 36
charge equivalents of forces on nuclei 320 and zero flux surfacecondition 29
a n d F e y n m a n o n l o n g - r a n g ea t t r a c t i v e Green's theorem 138
forces 325
and necessarycondition for chemical
binding 318 Hamiltonian
and viral theorem 323-6 classical 364
energy-momentum tensor 396 quantum mechanical 373
equivalencerelation Hamiltonian operator
applied to molecular graphs 55, 58, 89 in fixed-nucleusapproximation 172
applied to vector fields 57, 89 in presenceof electromagneticfield 405,
and structural stability 89 410
e s s e n t i avl a r i a b l e s 1 1 2 Hamilton's equations 364-5
Euler-Lagrange equation 152 Hamilton's principle 362, 364
Heisenbergcommutation relations
derivation of 374-5
Fermi hole and fundamental Poisson bracket
definition of 336 relations 367
d e n s i t yo f 3 3 7 statement of 356
and electron delocalization 341 Hellmann-Feynman theorem 145, 315
and electron localization 250-1, 337-40 Hermitian conjugate 355
and local charge concentrations 344-8 Hermitian operator 138
Feynman's path integral formulation of homeomorphism
quantum mechanics 423 between structure diagrams 97-101
force density 174, 178, 397, 401, 416 definition of 89
hybridization model
and atomic population 220 1,233
gauge transformation 4M and ring strain 219
Gauss'stheorem 139, 166 hypervirial theorem
generating function 365 a t o m i c s t a t e m e n to f 1 6 1 , 1 7 0
generator of infinitesimal canonical definition of bond properties 238-40
transformation derivation of 144
in classicalmechanics 366 and Hellmann-Feynman theorem 145
and Poisson bracket formulation of 368 for a subsystem 146
in quantum mechanics 359-60 v a r i a t i o n a l s t a t e m e n to f 1 6 1
generator of infinitesimal unitary
transformation
as defined by principle of stationary infinitesimal generator,see generator of
a c t i o n 3 7 3 .3 8 3 infinitesimal unitary transformation
a n d d e r i v a t i o no f c o m m u t a t i o n interatomic surface
relations 375 definition of 29
a n d d e r i v a t i o n o f H e i s e n b e r g ' se q u a t i o n o f formation of 31,289
motion 374 and interatomic forces 175,289
i n i f i n i t e s i m a lo o e r a t o r 3 6 0 inversion barrier 232-4
 436 INDEX

kinetic energy densities for cage structure 37

atomic averages 148, 276 definition of 33
chemical bonding and 323-9 d i s p l a y so f 3 4 , 7 2 - 3 , 2 8 2
in classificationof atomic for ring structure 35
interactions 295-302 molecular structure
components of 297, 301-2, 329 c h a n g e si n 6 0 - 4 , 9 4
definition of l4'7 definition of 58, 89
and local virial theorem 277 as distinct from molecular geometry 53_.4
relation to curvatures of charee generic nature of 53-4
density 301 in hydrocarbon molecules 7l-3
in hydrogen-bonded complexes 281-2
in propellanes 7l-5
Lagrange function operator stable 89
definition of 370 unstable 90
effect of infinitesimal canonical molecular structure hypothesis l-3, 53
transformation on 376 multiplicative law of transformation
Lagrange'sequations 364 functions 358,423
Lagrangian
classical 362
at point of variation 378, 405 natural boundary condition 152
quantum mechanical 376 non-nuclear attractor 43-4, 297-9
single-particledensity 378 non-transversalintersection 93
Lagrangian integral nuclear energy 187
atomic 379,406 nuclear potential
definition of 376 topology of 97
effect of infinitesimal generator on 384, 410 and mechanicsof charge density 101
for system in electromagneticfield 404
v a n i s h i n go f 3 7 8 , 3 7 9
Laplacian of the charge density observables
and binding of ring nuclei 329-31 average value of 140, 355
at a bond critical point 86 commuting 355-56
bonded and nonbonded charge compatible 356
concentrations 262-4 definition of 137
and characterizationof atomic equation of motion for 140
interactions 289 Hermitian property of 138
for C-C and C-H bonds 87 open system,see subsystem
definition of 18
and frontier orbital model 287
and Lewis acid-base reactions 2'78-81 pair density
and local charge concentration 253,258 definition of 335
and local charge depletion 277-81 and electron localization 250, 336-9
and localization of Fermi hole 344-8 path parameter 24,lO4
and local virial theoren 276 perturbation theory
prediction of molecular approach 281-6 and first-order correction to charge
reactive surface 2'18, 313 density 419-20
and shell structure 256-9 and symmetry rule for reaction
and symmetry of lowest energy transition mechanisms 287
density 288 phase portrait 25-7
v a l e n c es h e l l c h a r g ec o n c e n t r a t i o n Poincare-Hopf relationship 362
(vscc)2s8,262 Poisson bracket 367
and VSEPRmodel 265-73,347-9 polarizability
Lewiselectronpair model 248,260,265 atomic contribution 418
density 418
tensor 4l 7
magneticsusceptibility
421 potential energy surface 95
maximalmeasurement 356 principle of least action 362,422
moleculargraph principle of stationary action
 INDEX 437

for an atom, general statement of 387-90 structure diagram 65-6, 90-l

for an atom in electromagnetic field 409 subsystem
for an atom in stationary state 161 action integral for 379
and derivationof commutation Ehrenfestforce law for 174.396, 413
relations 374 electronic energy of 190-1
and expressionsfor infinitesimal Hamiltonian-based functional 162
generating operators 373 hypervirial theorem for 146
operator principle of 371 Lagrangian integral for 379
in Schrddinger representation 383 loss of Hermiticity for 139, 171
f o r a s t a t i o n a r ys t a t e 1 5 4 , 1 6 1 , 3 8 4 principle of stationary action for 161
from variation of action integral Schrodinger energy functional 154
operator 372-3 variation of action integral for 381-7
principle of superposition 355 variation of energy functional for 155-9,
probability amplitude 355 162-4
variation of surface 155
v i r i a l t h e o r e m f o r 1 ' l ' 7 , 4 0 14
, 15
quantum action principle 3'll, 423-4 surfacevirial 176, 241,415
quantum boundary condition 29, 380 surprisal 244
symmetry of transition density and reaction
coordinate 287
reactive surface 2'18,313
regular point 65
tensor 165
representation
trace theorem of Davidenko 107
basis for 355-6
trajectory of gradient vector field
change in basis 356
z-limit set 105
unitary transformation of 359
analytical expressionfor 49
representative
definition ol 22-4
definitionof 355-8
rr-r-limitset 104
change under infinitesimal unitary
transformation function 357
transformation 361
transition density
ring structure 35
definition of 419
rotation barrier 222-32
and polarizability density 420
symmetry of and reaction coordinate 287
transversalintersection 93
Schrodinger'senergy functional 150, 154
Schrodinger'swave function 358
Schrodinger'swave mechanics 150 unitary transformation
Schwinger'squantum action principle 371, infinitesimal 360
423-4 on an observable 359
shared interaction 290 and operator principle of stationary
singularity in charge density, see critical action 369-71
point and scaling of electronic
state function 6. 357 coordinates 392-3
state vector 353 on a state vector 359
strain energy and translation of electronic
in hydrocarbons 80-3 coordinates 391
origin of 219-21
relation to bond path angle 78
stresstensor van der Waals envelope 182 3, 304
definition of 142,173 vector current density
and Ehrenfest force 174 a n d a t o m i c h y p e r v i r i a lt h e o r e m 1 6 1 , 1 7 0
and local virial theorem 178 definitionof 139
trace ol 177 for a property density 160, 387
structural region 90 and time-dependenceof atomic
structural stability property 388
criteria for and theorem of 94 and variation o[ atomic functionals 159.
definition of 89, 94 t 6 4 . 1 7 0 . 3 8 6 .4 0 8
438 INDEX

vector potential of electromagneticfield 403 and chemical bonding 323-'7

virial and Heisenberg'sequation 143
definition of 144 for an isolated atom 144
a n d e l e c t r o n i cH a m i l t o n i a n 1 8 9 local statement of 177-8,401
and electronic potential energy 187 in presenceof electromagneticfield 415
of nuclear-nuclear repulsive lorces 188, for a time dependent system 400
230-1 VSEPR model of molecular geometry 265,
and partitioning of a potential J+l
energy l8G8
and potential energy contributions for an
atom 192-4
virial field zero-flux surface
definition of 189 definition of 29
and transferability of atomic and quantum boundary condition 158,
properties 136 380
virial theorem and transferability of atom 135
a t o m i c s t a t e m e n to f 1 7 7 . 4 0 0 .4 1 5 universal nature of 3l-2