An Introduction To Grids Graphs and Networks

AN INTRODUCTION TO GRIDS,
GRAPHS, AND NETWORKS
AN INTRODUCTION
TO GRIDS, GRAPHS,
AND NETWORKS
C. Pozrikidis
3
Oxford University Press is a department of the University of Oxford. It furthers the Universitys
objective of excellence in research, scholarship, and education by publishing worldwide.
Oxford New York
Auckland Cape Town Dar es Salaam Hong Kong Karachi
Kuala Lumpur Madrid Melbourne Mexico City Nairobi
New Delhi Shanghai Taipei Toronto
With offices in
Argentina Austria Brazil Chile Czech Republic France Greece
Guatemala Hungary Italy Japan Poland Portugal Singapore
South Korea Switzerland Thailand Turkey Ukraine Vietnam
Oxford is a registered trade mark of Oxford University Press in the UK and certain other
countries.
Published in the United States of America by
Oxford University Press
198 Madison Avenue, New York, NY 10016
Oxford University Press 2014

All rights reserved. No part of this publication may be reproduced, stored in a retrieval system,
or transmitted, in any form or by any means, without the prior permission in writing of
Oxford University Press, or as expressly permitted by law, by license, or under terms agreed with
the appropriate reproduction rights organization. Inquiries concerning reproduction outside
the scope of the above should be sent to the Rights Department, Oxford University Press,
at the address above.
You must not circulate this work in any other form
and you must impose this same condition on any acquirer.
Library of Congress Cataloging-in-Publication Data
Pozrikidis, C. (Constantine), 1958 author.
An introduction to grids, graphs, and networks / C. Pozrikidis.
p. cm.
Includes bibliographical references and index.
ISBN 9780199996728 (alk. paper)
1. Graph theory. 2. Differential equations, PartialNumerical solutions. 3. Finite differences. I. Title.
QA166.P69 2014
511.5dc23
2013048508
1 3 5 7 9 8 6 4 2
Printed in the United States of America
on acid-free paper
CONTENTS
Preface
xi
1. One-Dimensional Grids
1.1. Poisson Equation in One Dimension
1.2. Dirichlet Boundary Condition at Both Ends
1.3. NeumannDirichlet Boundary Conditions
1.4. DirichletNeumann Boundary Conditions
1.5. Neumann Boundary Conditions
10
1.6. Periodic Boundary Conditions
13
1.7. One-Dimensional Graphs

1.7.1. Graph Laplacian
1.7.2. Adjacency Matrix
1.7.3. Connectivity Lists and Oriented Incidence Matrix
16
17
18
19
1.8. Periodic One-Dimensional Graphs

1.8.1. Periodic Adjacency Matrix
1.8.2. Periodic Oriented Incidence Matrix
1.8.3. Fourier Expansions
1.8.4. Cosine Fourier Expansion
1.8.5. Sine Fourier Expansion
20
21
22
22
24
24
2. Graphs and Networks

2.1. Elements of Graph Theory
2.1.1. Adjacency Matrix
2.1.2. Node Degrees
2.1.3. The Complete Graph
2.1.4. Complement of a Graph
2.1.5. Connectivity Lists and the Oriented Incidence Matrix
26
26
26
28
29
29
30
vi / / C O N T E N T S
2.1.6.
2.1.7.
2.1.8.
2.1.9.
Connected and Unconnected Graphs

Pairwise Distance and Diameter
Trees
Random and Real-Life Networks
30
30
31
31
2.2. Laplacian Matrix

2.2.1. Properties of the Laplacian Matrix
2.2.2. Complete Graph
2.2.3. Estimates of Eigenvalues
2.2.4. Spanning Trees
2.2.5. Spectral Expansion
2.2.6. Spectral Partitioning
2.2.7. Complement of a Graph
2.2.8. Normalized Laplacian
2.2.9. Graph Breakup
32
33
34
35
36
36
36
38
38
39
2.3. Cubic Network
39
2.4. Fabricated Networks

2.4.1. Finite-Element Network on a Disk
2.4.2. Finite-Element Network on a Square
2.4.3. Delaunay Triangulation of an Arbitrary Set of Nodes
2.4.4. Delaunay Triangulation of a Perturbed Cartesian Grid
2.4.5. Finite Element Network Descending from an Octahedron
2.4.6. Finite Element Network Descending from an Icosahedron
41
42
43
43
43
44
45
2.5. Link Removal and Addition

2.5.1. Single and Multiple Link
2.5.2. Link Addition
46
47
49
2.6. Infinite Lattices

2.6.1. Bravais Lattices
2.6.2. Archimedean Lattices
2.6.3. Laves Lattices
2.6.4. Other Two-Dimensional Lattices
2.6.5. Cubic Lattices
50
50
53
56
57
58
2.7. Percolation Thresholds

2.7.1. Link (Bond) Percolation Threshold
2.7.2. Node Percolation Threshold
2.7.3. Computation of Percolation Thresholds
59
59
61
62
3. Spectra of Lattices
3.1. Square Lattice
3.1.1. Isolated Network
3.1.2. Periodic Strip
67
67
68
69
C O N T E N T S / / vii
3.1.3. Doubly Periodic Network

3.1.4. Doubly Periodic Sheared Network
73
77
3.2. Mbius Strips

3.2.1. Horizontal Strip
3.2.2. Vertical Strip
3.2.3. Klein Bottle
79
80
83
84
3.3. Hexagonal Lattice

3.3.3. Alternative Node Indexing
86
87
89
92
3.4. Modified Union Jack Lattice

93
94
95
3.5. Honeycomb Lattice

3.5.2. Brick Representation
3.5.4. Alternative Node Indexing
98
99
101
102
110
3.6. Kagom Lattice

111
112
115
3.7. Simple Cubic Lattice
122
3.8. Body-Centered Cubic (bcc) Lattice
124
3.9. Face-Centered Cubic (fcc) Lattice
126
4. Network Transport
130
4.1. Transport Laws and Conventions

4.1.1. Isolated and Embedded Networks
4.1.2. Nodal Sources
4.1.3. Linear Transport
4.1.4. Nonlinear Transport
130
130
131
132
133
4.2. Uniform Conductances

4.2.1. Isolated Networks
4.2.2. Embedded Networks
133
134
134
4.3. Arbitrary Conductances

4.3.1. Scaled Conductance Matrix
4.3.2. Weighed Adjacency Matrix
135
136
136
viii / / C O N T E N T S
4.3.3.
4.3.4.
4.3.5.
4.3.6.
4.3.7.
4.3.8.
Weighed Node Degrees

Kirchhoff Matrix
Weighed Oriented Incidence Matrix
Properties of the Kirchhoff Matrix
Normalized Kirchhoff Matrix
Summary of Notation
137
138
139
139
140
141
4.4. Nodal Balances in Arbitrary Networks

4.4.2. Embedded Networks and the Modified Kirchhoff Matrix
4.4.3. Properties of the Modified Kirchhoff Matrix
142
142
142
143
4.5. Lattices
4.5.1. Square Lattice
4.5.2. Mbius Strip
4.5.3. Hexagonal Lattice
4.5.4. Modified Union Jack Lattice
4.5.5. Simple Cubic Lattice
145
145
149
150
150
151
4.6. Finite Difference Grids
153
4.7. Finite Element Grids

4.7.1. One-Dimensional Grid
4.7.2. Two-Dimensional Grid
156
156
157
5. Greens Functions
161
5.1. Embedded Networks

5.1.1. Greens Function Matrix
5.1.2. Normalized Greens Function
161
162
163
5.2. Isolated Networks

5.2.1. MoorePenrose Greens Function
5.2.2. Spectral Expansion
5.2.3. Normalized MoorePenrose Greens Function
5.2.4. One-Dimensional Network
5.2.5. Periodic One-Dimensional Network
5.2.6. Free-Space Greens Function in One Dimension
5.2.7. Complete Network
5.2.8. Discontiguous Networks
164
164
166
167
168
169
171
171
172
5.3. Lattice Greens Functions

5.3.1. Periodic Greens Functions
5.3.2. Free-Space Greens Functions
173
173
175
5.4. Square Lattice

5.4.1. Periodic Greens Function
5.4.2. Free-Space Greens Function
177
177
179
C O N T E N T S / / ix
5.4.3. Helmholtz Equation Greens Function

5.4.4. Kirchhoff Greens Function
190
190
5.5. Hexagonal Lattice

191
191
192
5.6. Modified Union Jack Lattice

196
197
198
5.7. Honeycomb Lattice

200
201
203
5.8. Simple Cubic Lattice

206
206
207
5.9. Body-Centered Cubic (bcc) Lattice
209
5.10. Face-Centered Cubic (fcc) Lattice
211
5.11. Free-Space Lattice Greens Functions

5.11.1. Probability Lattice Greens Function
212
213
5.12. Finite Difference Solution in Terms of Greens Functions
216
6. Network Performance
220
6.1. Pairwise Resistance

6.1.3. One-Dimensional Network
6.1.4. One-Dimensional Periodic Network
6.1.5. Infinite Lattices
6.1.6. Triangle Inequality
6.1.7. Random Walks
220
221
223
225
226
226
227
227
6.2. Mean Pairwise Resistance

6.2.1. Spectral Representation
6.2.2. Complete Network
6.2.3. One-Dimensional Isolated Network
6.2.4. One-Dimensional Periodic Network
6.2.5. Periodic Lattice Patches
228
228
229
229
230
231
6.3. Damaged Networks

6.3.1. Damaged Kirchhoff Matrix
234
235
236
x // CONTENTS
6.3.3. One Damaged Link

6.3.4. Clipped Links
238
240
240
6.4. Reinforced Networks
240
6.5. Damaged Lattices

6.5.1. One Damaged Link
6.5.2. Effective-Medium Theory
6.5.3. Percolation Threshold
242
242
245
246
6.6. Damaged Square Lattice
247
6.7. Damaged Honeycomb Lattice
251
6.8. Damaged Hexagonal Lattice

6.8.1. Longitudinal Transport
6.8.2. Lateral Transport
255
255
257
Appendices
A. Eigenvalues of Matrices
259
A.1. Eigenvalues and Eigenvectors
259
A.2. The Characteristic Polynomial

A.2.1. Eigenvalues, Trace, and the Determinant
A.2.2. Powers, Inverse, and Functions of a Matrix
A.2.3. Hermitian Matrices
A.2.4. Diagonal Matrix of Eigenvalues
260
261
262
262
263
A.3. Eigenvectors and Principal Vectors

A.3.1. Properties of Eigenvectors
A.3.2. Left Eigenvectors
A.3.3. Matrix of Eigenvectors
A.3.4. Eigenvalues and Eigenvectors of the Adjoint
A.3.5. Eigenvalues of Positive Definite Hermitian Matrices
263
264
264
265
266
266
A.4. Circulant Matrices
267
A.5. Block Circulant Matrices
268
B. The ShermanMorrison and Woodbury Formulas
269
B.1. The Woodbury Formula
269
B.2. The ShermanMorrison Formula
273
References
278
Index
281
PREFACE
Cartesian, curvilinear, and other unstructured grids are used for the numerical solution of ordinary and partial differential equations using finite difference, finite
element, finite volume, and related methods. Graphs are broadly defined as finite
or infinite sets of vertices connected by edges in structured or unstructured configurations. Infinite lattices and tiled surfaces are described by highly ordered graphs
parametrized by an appropriate number of indices. Networks consist of nodes connected by physical or abstract links with an assigned conductance in spontaneous or
engineered configurations. In physical and engineering applications, networks are
venues for conducting or convecting a transported entity, such as heat, mass, or
digitized information according to a prevailing transport law. The performance of
networks is an important topic in the study of complex systems with applications in
energy, material, and information transport.
The analysis of grids, graphs, and networks involves overlapping and complementary topics that benefit from a unified discussion. For example, finite difference
and finite element grids can be regarded as networks whose link conductance is
determined by the differential equation whose solution is sought as well as by the
chosen finite difference or finite element approximation. Particular topics of interest include the properties of the node adjacency, Laplacian, and Kirchhoff matrices;
the evaluation of percolation thresholds for infinite, periodic, and finite systems; the
computation of the regular and generalized lattice Greens function describing the
response to a nodal source; the pairwise resistance of any two nodes; the overall characterization of the network robustness; and the performance of damaged networks
with reference to operational and percolation thresholds.
My goal in this text is to provide a concise and unified introduction to grids,
graphs, and networks to a broad audience in the engineering, physical, biological,
and social sciences. The approach is practical, in that only the necessary theoretical
and mathematical concepts are introduced. Theory and computation are discussed
alongside, and formulas amenable to computer programming are provided. The prerequisite is familiarity with college-level linear algebra, calculus, and elementary
numerical methods.
One important new concept is the distinction between isolated and embedded
networks. The former stand in isolation as though they were suspended in vacuum,
xi
xii / / P R E FA C E
whereas the latter are connected to exterior nodes where a nodal potential, such as
temperature, pressure, or electrical voltage, is specified. Regular Greens functions
describing the discrete field due to a nodal impulse are available in the case of embedded or infinite networks, whereas generalized Greens functions describing the
discrete field due to a nodal impulse in the presence of distributed sinks are available
in the case of isolated networks. Discrete Greens functions can be used as building
blocks for computing general solutions subject to given constraints.
This book is suitable for self-study and as a text in an upper-level undergraduate
or entry-level graduate course in sciences, engineering, and applied mathematics.
The material serves as a reference of terms and concepts and as a resource of topics
for further study.
C. Pozrikidis
September, 2013
AN INTRODUCTION TO GRIDS,
GRAPHS, AND NETWORKS
ONE-DIMENSIONAL GRIDS
/// 1 ///
A finite difference grid for solving ordinary or partial differential

equations consists of rectilinear or curvilinear grid lines that can be regarded as conveying links intersecting at nodes. This interpretation provides us with a point of
departure for making an analogy between numerical grids, mathematical graphs,
and physical or abstract networks. We begin in this chapter by developing finite
difference equations for an elementary ordinary equation with the objective of identifying similarities between grids and graphs, and then we generalize the framework
to higher dimensions.
1.1 POISSON EQUATION IN ONE DIMENSION
Consider the Poisson equation in one dimension for an unknown function of one
variable, f (x),
(1.1.1)
d2f
+ g(x) = 0,
dx2
to be solved in a finite domain, [a, b], where g(x) is a given source function. When
g(x) = 0, the Poisson equation reduces to Laplaces equation. When g(x) = f (x),
the Poisson equation reduces to Helmholtzs equation, where is a real or complex
constant.
A numerical solution can be found on a uniform finite difference grid with K
divisions defined by K + 1 nodes, as shown in Figure 1.1.1. Nodes numbered 0 and
i1
a
x
i+1
K+1 K+2
FIGURE 1.1.1 A finite difference with K uniform divisions along the x axis.
Dirichlet or Neumann boundary conditions are specified at the two ends
of the solution domain.
2 // AN INTRODUCTION TO GRIDS, GRAPHS, AND NETWORKS
K +2 are phantom nodes, lying outside the solution domain, introduced to implement
the Neumann boundary condition, when specified, as discussed later in this chapter.
Applying the Poisson equation at the ith node, approximating the second derivative with a central difference by setting
fi1 2fi + fi+1
(1.1.2) f (xi )
+ O(x2 )
2
x
with an error of order x2 , and rearranging, we obtain the difference equation
(1.1.3) fi1 + 2fi fi+1 = x2 gi
to be applied at an appropriate number of nodes. To simplify the notation, we have

denoted
(1.1.4) fi f (xi ),
gi g(xi ).
The signs on the left- and right-hand sides of (1.1.3) were chosen intentionally to
conform with standard notation in graph theory regarding the Laplacian, as discussed
in Section 1.7.
Collecting all available difference equations and implementing the boundary conditions provides us with a system of linear algebraic equations for a suitable number
of unknown nodal values contained in a solution vector, ,
(1.1.5) L = b,
where the centered dot denotes the matrixvector product. The size and specific form
of the coefficient matrix, L, solution vector, , and vector on the right-hand side, b,
depend on the choice of boundary conditions. Several possibilities are discussed in
this chapter.
Factorization
We will see that, for any type of boundary conditionsNeumann, Dirichlet, or

periodicthe coefficient matrix of the linear system admits the factorization
(1.1.6) L = R RT ,
where R is a square or rectangular matrix, the superscript T denotes the matrix

transpose, and the centered dot denotes the usual matrix product (e.g., [35]). This factorization can be regarded as the discrete counterpart of the definition of the second
derivative as the sequential application of the first derivative,
(1.1.7)
d2
d d
=
.
2
dx dx
dx
One-Dimensional Grids // 3
It is important to note that the commutative property R RT = RT R is not always

satisfied.
The counterpart of the factorization (1.1.6) in n dimensions is
(1.1.8) 2 = ,
where
(1.1.9) 2 =
2
2
2
+
+
+
x2n
x12 x12
is the scalar Laplacian operator,

(1.1.10) =
+
+ +
x1 x2
xn
is the vectorial gradient operator, and the centered dot denotes the inner vector
product. In two dimensions n = 2, and in three dimensions n = 3.
Exercise
1.1.1 Helmholtz equation
Write the counterpart of the difference equation (1.1.3) for the Helmholtz equation
in one dimension,
(1.1.11)
d2f
+ f = 0,
dx2
where is a real or complex constant.

1.2 DIRICHLET BOUNDARY CONDITION AT BOTH ENDS
When the Dirichlet boundary condition is specified at both ends of the solution
domain, the first and last values, f1 and fK+1 , are known. Collecting the difference
equations (1.1.3) for the interior nodes, i = 2, . . . , K, we obtain a system of linear
equations,
(1.2.1) LDD DD = bDD ,
where
(1.2.2)
DD
f2
f3 ,
..
.
fK1
fK
DD
x2 g2 + f1
x2 g3
..
.
x2 gK1
x2 gK + fK+1
are (K 1)-dimensional vectors and
2
1
1
2
1
0
..
..
(1.2.3) LDD =
.
.
0
0
0
0
0
0
0
1
2
..
.
..
.
0
0
0
..
.
0
0
0
..
.
0
0
0
2
1
0
1
2
1
0
0
0
..
.
1
2
is (K 1) (K 1) symmetric tridiagonal Toeplitz matrix. By definition, a Toeplitz

matrix consists of constant diagonal lines. The superscript DD emphasizes that the
Dirichlet condition is specified at both ends.
Decomposition
We can decompose
(1.2.4) LDD = 2 I DD ,
where I is the (K 1) (K 1) identity matrix and
0
1
0
1
0
1
1
0
0
.
.
.
DD
..
..
..
(1.2.5)
=
.
..
0
0
0
0
0
0
0
0
0
0
0
..
.
0
0
0
..
.
0
1
0
1
0
1
0
0
0
..
.
1
0
is a (K 1) (K 1) symmetric bidiagonal Toeplitz matrix with zeros along the

diagonal.
Eigenvalues and Eigenvectors
The eigenvalues of the matrices DD and LDD are

(1.2.6)
m = 2 cos m
and
(1.2.7) Lm = 2 2 cos m = 4 sin2 12 m ,
where
(1.2.8) m =
for m = 1, . . . , K 1.
The corresponding shared eigenvectors, u(m) , normalized so that their length is

equal to unity, u(m) u(m) = 1, are
1/2
2
(m)
(1.2.9) uj =
sin(jm )
K
for m, j = 1, . . . , K 1. It is interesting that all eigenvectors are pure harmonic waves,
with higher-order eigenvalues corresponding to shorter wavelengths.
Factorization
We can factorize
T
(1.2.10) LDD = RDD RDD ,
where
(1.2.11) RDD
1
0
0
..
.
1
1
0
..
.
0
1
1
..
.
0
1
..
.
..
.
0
0
0
..
.
0
0
0
..
.
0
0
0
0
0
0
0
0
0
1
0
0
1
1
0
0
1
1
0
0
0
..
.
0
1
is a rectangular (K 1) K matrix implementing forward difference approximations

to the first derivative. The transpose of RDD ,
(1.2.12) RDD
1
1
0
..
.
0
1
1
..
.
0
0
1
..
.
..
.
0
0
0
..
.
0
0
0
..
.
0
0
0
0
0
0
0
0
0
0
0
0
1
1
0
0
0
1
1
0
0
0
0
..
.
1
1
is a rectangular K (K1) matrix implementing backward difference approximations

to the first derivative.
Exercises
1.2.1 Sinusoidal Source
Solve the linear system (1.2.1) for a = 0 and g(x) = sin2 (2 x/b), where is a
constant. The boundary conditions specify that f (0) = 0 and f (b) = fb , where fb is a
given constant. Carry out computations for K = 2, 4, 8, 16, and 32, and discuss the
accuracy of the numerical results with reference to the exact solution.
1.2.2 Factorization
Confirm the factorization (1.2.10).
1.3 NEUMANNDIRICHLET BOUNDARY CONDITIONS
Now assume that the Neumann boundary condition is prescribed at the right end of
the solution domain, x = a, specifying that
(1.3.1) f (x1 ) = q1 ,
where q1 is a given constant, while the Dirichlet boundary condition is prescribed

at the left end of the solution domain, x = b. specifying the value of fK+1 . Following standard practice, we introduce a phantom node labeled zero, as shown in
Figure 1.1.1, approximate the first derivative with second-order accuracy using a
central difference as
f2 f0
(1.3.2) f (x1 )
+ O(x2 ),
2x
and write
(1.3.3) f0 = f2 + 2x q1 .
The difference equations for i = 1, . . . , K provide us with a system of linear

equations,
(1.3.4) LND ND = bND ,
where
(1.3.5)
ND
f1
f2 ,
..
.
fK1
fK
are K-dimensional vectors and
1
1
..
ND
(1.3.6) L
=
.
0
0
ND
x2 g1 + x q1
x2 g2
..
,
.
x2 gK1
2
x gK + fK+1
1
2
1
2
1
..
.
0
1
2
..
.
..
.
0
0
0
..
.
0
0
0
..
.
0
0
0
0
0
0
2
1
0
1
2
1
0
0
0
..
.
1
2
is a KK symmetric, tridiagonal, nearly Toeplitz matrix. If the first diagonal element

were equal to 2, this matrix would have been a perfect Toeplitz matrix.
Decomposition
We can decompose
(1.3.7) LND = 2 I ND ,
where I is the K K identity matrix and
(1.3.8) ND
1
1
0
..
.
0
0
1
0
1
..
.
0
1
0
..
.
..
.
0
0
0
..
.
0
0
0
..
.
0
0
0
0
0
0
0
1
0
1
0
1
0
0
0
..
.
1
0
is a K K square, symmetric, tridiagonal, nearly Toeplitz matrix. If the first diagonal

element were equal to 0, this would have been a perfect Toeplitz matrix.
The eigenvalues of ND and LND are

(1.3.9)
m = 2 cos m
and
(1.3.10) Lm = 2 2 cos m = 4 sin2 12 m ,
where
(1.3.11) m =
m 1/2
K + 1/2
for m = 1, . . . , K.
(m)
(1.3.12) uj

4 1/2
cos j 12 m
2K + 1
for m, j = 1, . . . , K. All eigenvectors are pure harmonic waves.
Factorization
We can factorize
T
(1.3.13) LND = RND RND ,
where
(1.3.14) RND
1
1
0
..
.
0
1
1
..
.
0
0
1
..
.
..
.
0
0
0
..
.
0
0
0
..
.
0
0
0
0
0
0
0
0
0
1
1
0
0
1
1
0
0
0
..
.
0
1
is a K K square lower bidiagonal K K Toeplitz matrix implementing backward

difference approximations to the first derivative. Its transpose,
1
1
0
0
0
0
0
1
0
0
0
0
0
1
0
0
0
0
T
..
..
..
..
..
..
..
(1.3.15) RND =
.
.
.
.
.
.
. ,
0
0
1
1
0
0
0
0
0
0
1
1
0
0
0
0
0
1
is a K K square upper bidiagonal K K Toeplitz matrix implementing forward
difference approximations to the first derivative.
Exercise
1.3.1 Factorization
Confirm by direct multiplication the factorization (1.3.13).
1.4 DIRICHLETNEUMANN BOUNDARY CONDITIONS
In the third case study, we assume that a Dirichlet boundary condition specifying
the value of f1 is prescribed at the left end of the solution domain, and a Neumann
boundary condition specifying that
(1.4.1) f (xK+1 ) = qK+1
is prescribed at the right end of the solution domain, where qK+1 is a given constant. We proceed by introducing a phantom node numbered K + 2, as shown in
Figure 1.1.1, approximate the first derivative with second-order accuracy as
fK+2 fK
(1.4.2) f (xK+1 )
+ O(x2 ),
2x
and obtain
(1.4.3) fK+2 = fK + 2x qK+1 .
The difference equations for i = 2, K + 1 provide us with a linear system,

(1.4.4) LDN DN = bDN ,
where
(1.4.5)
DN
f2
f3 ,
..
.
fK
fK+1
2
1
..
DN
(1.4.6) L
=
.
0
0
DN
x2 g2 + f1
x2 g3
..
.
1
2
x2 gK
x2 gK+1 + x qK+1
1
2
1
..
.
0
1
2
..
.
..
.
0
0
0
..
.
0
0
0
..
.
0
0
0
0
0
0
2
1
0
1
2
1
0
0
0
..
.
Decomposition
We can decompose
(1.4.7) LDN = 2 I DN ,
0
1
0
..
.
..
.
0
0
0
..
.
0
0
0
..
.
0
0
0
0
1
0
1
0
1
is a K K symmetric, tridiagonal, nearly Toeplitz matrix.
1
1
is a K K symmetric, tridiagonal, nearly Toeplitz matrix.
where I is the N N identity matrix and
0
1
1
0
1
0
.
..
DN
(1.4.8)
= ..
.
0
0
0
0
0
0
0
0
0
..
.
1
1
10 / / A N I N T R O D U C T I O N T O G R I D S , G R A P H S , A N D N E T W O R K S
The eigenvalues of DN and LDN are

(1.4.9)
m = 2 cos m
and
(1.4.10) Lm = 2 2 cos m = 4 sin2 12 m ,
where
(1.4.11) m =
m
K+
1
2
1
2
for m = 1, . . . , K.

(m)
(1.4.12) uj
4
2K + 1
1/2
cos
Kj+
1
2
for m, j = 1, . . . , K. All eigenvectors are pure harmonic waves.

Factorization
We can factorize
T
(1.4.13) LDN = RDN RDN = RND RND ,

T
where RDN = RND and the matrix RND is given in (1.3.15).
Exercise
1.4.1 Eigenvalues and eigenvectors
Confirm by direct substitution the eigenvalues and eigenvectors given in (1.4.10) and
(1.4.12).
1.5 NEUMANN BOUNDARY CONDITIONS
In the fourth and most important case, the Neumann boundary condition is prescribed
at both ends of the solution domain,
(1.5.1) f (x1 ) = q1 ,
f (xK+1 ) = qK+1 .
O n e - D i m e n s i o n a l G r i d s / / 11
where q1 and qK+1 are two given constants. Working in the familiar way, we collect
the difference equations for i = 1, . . . , K + 1 into a linear system,
(1.5.2) LNN NN = bNN ,
where
(1.5.3) NN
f1
f2
..
.
fK
fK+1
bNN
1
2
x2 g1 + x q1
x2 g2
= ...
x2 gK
1
2
2 x gK+1 + x qK+1
are (K + 1)-dimensional vectors and
(1.5.4) LNN
1
1
0
..
.
1
2
1
..
.
0
1
2
..
.
..
.
0
0
0
..
.
0
0
0
..
.
0
0
0
0
0
0
0
0
0
2
1
0
1
2
1
0
0
0
..
.
1
1
is a (K + 1) (K + 1) symmetric tridiagonal matrix. If the first and last diagonal

elements were equal to 2, this would have been a perfect Toeplitz matrix.
Decomposition
We can decompose
(1.5.5) L = 2 I NN ,
where I is the (K + 1) (K + 1) identity matrix and
(1.5.6) NN
1
1
0
..
.
0
0
1
0
1
..
.
0
1
0
..
.
..
.
0
0
0
..
.
0
0
0
..
.
0
0
0
0
0
0
0
1
0
1
0
1
0
0
0
..
.
1
1
is a nearly upper and lower bidiagonal matrix. Note the presence of two nonzero top
and bottom diagonal elements.
The eigenvalues of NN and LNN are

(1.5.7)
m = 2 cos m
and
(1.5.8) Lm = 2 2 cos m = 4 sin2
1
2 m
where
(1.5.9) m =
m1
K+1
for m = 1, . . . , K + 1.
2 1/2

(m)
(1.5.10) uj = Am
cos j 12 m
K+1
for m, j = 1, . . . , K + 1, where Am = 1, except that A1 = 1/ 2. The presence of a

zero eigenvalue of the Laplacian, L1 = 0, corresponding to a constant eigenvector,
confirms that the Laplacian matrix is singular. The rest of the eigenvectors are pure
harmonic waves.
Cursory inspection reveals the interesting identity
(1.5.11) f LNN f =
K

( fi fi+1 )2 0
i=1
for any arbitrary nodal field, f, which demonstrates that the matrix LNN is positive
semidefinite. If u is an eigenvector of LNN with corresponding eigenvector , then
(1.5.12) u LNN u = u u 0.
This inequality confirms that the eigenvalues of LNN are zero or positive.
It is worth remarking that the eigenvalues of the Laplacian matrix are approximations of those of the Laplace equation, , in the interval [a, b], satisfying the
equation
(1.5.13)
d2 u

+
u=0
2
dx
x2
with homogeneous Neumann boundary conditions at both ends, u (a) = 0 and

u (b) = 0, where u(x) is an eigenfunction, L = Kx, and L = b a. We find that
m 1 2
m 1 x
(1.5.14) m =
2 , um (x) = cos
,
K
K x
for m 1. The eigenvalues of the Laplacian matrix, Lm , agree with the eigenvalues
m for small m and large K.
Factorization
We can factorize
T
(1.5.15) LNN = RNN RNN ,
where
(1.5.16) RNN
1
1
0
..
.
0
1
1
..
.
0
0
1
..
.
..
.
0
0
0
..
.
0
0
0
..
.
0
0
0
0
0
0
0
0
0
0
0
0
1
1
0
0
0
1
1
0
0
0
0
..
.
1
1
is a (K+1)K rectangular matrix implementing backward difference approximations

to the first derivative. Its transpose,
1
1
0
0
0
0
0
0
1
1
0
0
0
0
0
0
1
0
0
0
0
.
.
.
.
.
.
.
NNT
.
.
.
.
.
.
.
.
.
(1.5.17) R
=
.
.
.
.
.
.
.
. ,
0
0
1
1
0
0
0
0
0
0
0
1
1
0
0
0
0
0
0
1
1
is a K (K + 1) matrix implementing forward difference approximations to the first
derivative.
Exercise
1.5.1 Eigenvalues of the Laplacian
(a) Derive the eigenvalues and eigenvectors shown in (1.5.14). (b) Prepare and discuss a plot of the eigenvalues given in (1.5.14) and those of the Laplacian matrix for
K = 2, 4, 8, 16, and 32.
1.6 PERIODIC BOUNDARY CONDITIONS
When the solution of the differential equation (1.1.1) is required to be periodic, we

specify that f1 = fK+1 and compile the difference equations for i = 1, . . . , K to obtain
a linear system,
(1.6.1) LP P = bP ,
where
f1
f2
..
.
(1.6.2)
fK1
fK
P
g1
g2
..
.
b = x
gK1
gK
P
P
(1.6.3) L =
2
1
0
..
.
1
2
1
..
.
0
1
1
..
.
..
.
0
0
0
..
.
0
0
0
..
.
0
0
1
0
0
0
0
0
0
2
1
0
1
2
1
1
0
0
..
.
1
2
is a K K symmetric and nearly tridiagonal matrix. Note the presence of a northeastern and a southwestern element, both equal to 1, implementing the periodicity
condition.
Decomposition
We can decompose
(1.6.4) LP = 2 I P ,
where I is the K K identity matrix and
0
1
0
..
.
P
(1.6.5) =
0
1
1
0
1
..
.
0
1
0
..
.
..
.
0
0
0
..
.
0
0
0
..
.
0
0
0
0
0
0
0
1
0
1
0
1
1
0
0
..
.
1
0
is a K K symmetric, nearly bidiagonal Toeplitz matrix. Note the presence of two

unit corner elements, equal to 1, implementing the periodicity condition.
The eigenvalues of P and LP are

(1.6.6)
m = 2 cos m
and
(1.6.7) Lm = 2 2 cos m = 4 sin2 12 m
for m = 1, . . . , K, where
(1.6.8) m =
m1
2 .
K
The corresponding shared eigenvectors, u(m) , normalized so that their norm is

(m)
(1.6.9) uj
1
= exp (i j m )
K
for m, j = 1, . . . , K, where i is the imaginary unit and an asterisk denotes the complex
conjugate. The presence of a zero eigenvalue of the Laplacian, L1 = 0, corresponding
to a constant eigenvector, confirms that the matrix LP is singular. The rest of the
eigenvectors are pure harmonic waves.
Complex eigenvectors appear because two eigenvalues, m1 and m2 , are identical
when
(1.6.10) m1 + m2 = K + 2.
The real part of the complex exponential in (1.6.9) can be retained for one eigenvalue,
yielding a cosine, and the imaginary part can be retained for the other eigenvalue,
yielding a sine.
Cursory inspection reveals the interesting identity
(1.6.11) f LP f =
K

(fi fi+1 )2 0,
i=1
where f is an arbitrary nodal field satisfying the mandatory periodicity condition fK+1 = f1 , which demonstrates that the matrix LNN is positive semidefinite.
Consequently, the eigenvalues of LP are zero or positive.
Factorization
We can factorize
T
(1.6.12) LP = RP RP = RP RP ,
where
P
(1.6.13) R =
1
1
0
..
.
0
1
1
..
.
0
0
1
..
.
..
.
0
0
0
..
.
0
0
0
..
.
0
0
0
0
0
0
0
0
0
1
1
0
0
1
1
1
0
0
..
.
0
1
is a K K square nearly lower bidiagonal matrix implementing backward difference

approximations. Its transpose,
(1.6.14) RP
1
0
0
..
.
1
1
0
..
.
0
1
1
..
.
..
.
0
0
0
..
.
0
0
0
..
.
0
0
1
0
0
0
0
0
0
1
0
0
1
1
0
0
0
0
..
.
1
1
is a K K square nearly upper bidiagonal matrix implementing forward difference

approximations. Note the presence of one nonzero corner element implementing the
periodicity condition.
Exercise
1.6.1 Eigenvalues and eigenvectors
Confirm by direct substitution the eigenvalues and eigenvectors given in (1.6.7) and
(1.6.9).
1.7 ONE-DIMENSIONAL GRAPHS
The finite difference grid discussed previously in this chapter is now regarded as
a graph consisting of N nodes, also called vertices, connected by L = N 1 links
(edges), as illustrated in Figure 1.7.1. In an alternative interpretation, the finite difference grid is a network consisting of conducting or conveying links. For example,
the links can be regarded as segments of a fluid-carrying pipe.
Links:
Nodes:
i
2
i1
L
i+1
FIGURE 1.7.1 Illustration of a one-dimensional graph consisting of N

nodes connected by L = N 1 links.
1.7.1 Graph Laplacian
The N N matrix L LNN , corresponding to two Neumann boundary conditions

discussed in Section 1.5, is the Laplacian of the one-dimensional network, given by
(1.7.1) L =
1
1
0
..
.
1
2
1
..
.
0
1
2
..
.
..
.
0
0
0
..
.
0
0
0
..
.
0
0
0
0
0
0
0
0
0
2
1
0
1
2
1
0
0
0
..
.
1
1
Note that the sum of the elements in each row or column is zero. Sometimes, the
graph Laplacian is also called the combinatorial Laplacian.
The eigenvalues of L are given by
(1.7.2) m = 2 2 cos n = 4 sin2 12 n
for n = 1, . . . , N, where
(1.7.3) n =
n1
.
N
The corresponding eigenvectors, u(n) , normalized so that u(n) u(n) = 1, are given by
(n)
(1.7.4) ui
= An
2 1/2
N

cos (2j 1) n
for i, n = 1, . . . , N, where An = 1, except that A1 = 1/ 2. The presence of a zero

eigenvalue, 1 = 0, corresponding to a uniform eigenvector with equal elements,
confirms that the Laplacian matrix is singular. The rest of the eigenvectors are pure
harmonic waves.
1.7.2 Adjacency Matrix
In graph theory, an N N adjacency matrix is introduced, A, defined such that Aij = 1

if nodes i and j are connected by a grid line or link, and Aij = 0 otherwise, with the
convention that Aii = 0. Thus, by convention, the diagonal line of the adjacency
matrix is zero.
In the case of the one-dimensional grid presently considered, the adjacency
matrix is
0
1
0
..
.
(1.7.5) A =
0
0
1
0
1
..
.
0
1
0
..
.
..
.
0
0
0
..
.
0
0
0
..
.
0
0
0
0
0
0
0
1
0
1
0
1
0
0
0
..
.
1
0
The eigenvalues of this matrix are
N+1
(1.7.6) n = 2 cos
for n = 1, . . . , N.
The eigenvalues of the adjacency matrix provide us with measure of the network
properties, independent of node and link labeling. In particular, the number of paths
that return to an arbitrary node after s steps have been made, summed over all starting
nodes, is
(1.7.7) ns =
N

sn ,
n=1
where s is an integer. We observe that n0 = N, in agreement with physical intuition.

In our one-dimensional network, ns = 0 if s is an odd integer and ns = 0 if s is an
even integer. Physically, an even number of steps are necessary for an equal number
of forward and backward steps. For a one-dimensional network with N = 13 nodes,
we find that
(1.7.8) n2 = 26,
n4 = 74,
n6 = 236,
n8 = 794,
n10 = 2756.
Node Degrees
The degree of the ith node, denoted by di , is defined as the number of links attached
to the node, which is equal to the sum of the elements in the corresponding row or
column of the adjacency matrix, A. In the case of the one-dimensional grid presently
considered, we have
(1.7.9) d1 = 1,
di = 2,
dN = 1,
for i = 2, . . . N 1.
Laplacian in Terms of the Adjacency Matrix
The graph Laplacian of the one-dimensional grid is given by

(1.7.10) L = D A,
where D is a diagonal matrix whose ith diagonal element is equal to the corresponding node degree, di .
1.7.3 Connectivity Lists and Oriented Incidence Matrix
The number of links in the one-dimensional network is L = N 1. It is useful to

introduce two L-dimensional connectivity lists, k and l, defined such that the label of
the first node of the mth link is km and the label of the second node of the mth link is
lm . In the case of the one-dimensional grid presently considered, we have
(1.7.11) km = m,
lm = m + 1
for m = 1, . . . , L. An N L oriented incidence matrix can be introduced, R, defined

such that Ri,m = 0, except that
(1.7.12) Rkm ,m = 1,
Rlm ,m = 1.
If nodes and links are labeled sequentially, as shown in Figure 1.7.1, we obtain the
rectangular N (N 1) matrix
(1.7.13) R =
1
1
0
..
.
0
1
1
..
.
0
0
1
..
.
..
.
0
0
0
..
.
0
0
0
..
.
0
0
0
0
0
0
0
0
0
0
0
0
1
1
0
0
0
1
1
0
encountered previously in Section 1.5.
0
0
0
..
.
1
1
Laplacian in Terms of the Oriented Incidence Matrix
The graph Laplacian is given by

(1.7.14) L = R RT .
In fact, this factorization is valid for arbitrary node and link labeling and for general
higher-dimensional graphs.
Exercise
1.7.1 Node and link labeling
Derive the connectivity lists and the oriented incidence matrix for an arbitrary node
and link labeling scheme of your choice.
1.8 PERIODIC ONE-DIMENSIONAL GRAPHS
Shown in Figure 1.8.1 is a closed or periodic one-dimensional graph consisting of N

unique nodes connected by L = N links. The N N periodic graph Laplacian is
2
1
0
0
0
1
1
2
1
0
0
0
1
1
0
0
0
0
..
..
..
..
..
..
..
(1.8.1) L =
.
.
.
.
.
.
. ,
0
0
2
1
0
0
0
0
0
1
2
1
1
0
0
0
1
2
where the two nonzero northeastern and southwestern corner elements implement
the periodicity condition, as discussed in Section 1.6.
i
i1
i
i+1
2
Links:
L
N
Nodes:
FIGURE 1.8.1 Illustration of a periodic

one-dimensional graph consisting of N
unique nodes connected by L = N links.
The first and last nodes numbered 1
and N + 1 coincide.
The periodic Laplacian is a circulant matrix. By definition, each row of an

arbitrary circulant matrix derives from the previous row by shifting each element
to the right by one place and then returning the last element to the first place, as
discussed in Section A.4, Appendix A.
The eigenvalues of the periodic Laplacian are
(1.8.2) n = 2 2 cos n = 4 sin2 12 n
for n = 1, . . . , N, where
(1.8.3) n =
n1
2 .
N
The corresponding eigenvectors, u(n) , normalized so that u(n) u(n) = 1, are

(n)
(1.8.4) ui
1
= exp(i in )
N
for n, j = 1, . . . , N, where i is the imaginary unit and an asterisk denotes the complex conjugate. The presence of a zero eigenvalue, 1 = 0, corresponding to a
uniform eigenvector, confirms that the periodic Laplacian is singular. The rest of
the eigenvectors are pure harmonic waves.
A discrete Fourier orthogonality property states that
(1.8.5)
N

j=1

2
N
exp i jp
=
0
N
if p = sN,
otherwise,
where p and s are zero or arbitrary integers. This property ensures that
(1.8.6) u(s) u(r) = sr ,
that is, the eigenvectors comprise an orthonormal set.

1.8.1 Periodic Adjacency Matrix
The N N periodic adjacency matrix is a circulant matrix,
0
1
0
0
0
1
0
1
0
0
1
0
0
0
0
.
.
.
.
..
.
..
..
..
..
(1.8.7) A = ..
.
0
0
0
0
1
0
0
0
1
0
1
0
0
0
1
1
0
0
..
.
1
0
Two nonzero corner elements appear due to the periodicity condition. The degrees
of all nodes are the same, di = 2 for i = 1, . . . , N.
The eigenvalues of the periodic adjacency matrix are
n 1
(1.8.8) n = 2 cos
for n = 1, . . . , N.
The number of steps defined in (1.7.7) are zero when s is zero or an odd integer
and nonzero when s is an even integer. When N = 13, we find that
(1.8.9) n2 = 26,
n4 = 78,
n6 = 260,
n8 = 910,
n10 = 3276.
1.8.2 Periodic Oriented Incidence Matrix
If we label nodes and links sequentially, as shown in Figure 1.8.1, we will obtain a
square N N oriented incidence matrix,
(1.8.10) R =
1
1
0
..
.
0
1
1
..
.
0
0
1
..
.
..
.
0
0
0
..
.
0
0
0
..
.
0
0
0
0
0
0
0
0
0
1
1
0
0
1
1
1
0
0
..
.
0
1
The periodic Laplacian is given by

(1.8.11) L = D A = R RT ,
where D = 2 I is a diagonal matrix hosting the degree of the N nodes.

1.8.3 Fourier Expansions
A real periodic nodal field, , can be expanded in a Fourier series so that

(1.8.12) i =
M

cp exp i pk(i 1)
p = M
for i = 1, . . . , N, where k 2 /N is the wave number of the longest wave, i is the

imaginary unit, and cp are complex Fourier coefficients. To ensure that the expanded
nodal field is real, we require that cp = cp , where an asterisk denotes the complex
conjugate. The truncation level, M, is discussed later in this section.
An equivalent representation in terms of sines and cosines, arising by resolving the Fourier coefficients and complex exponentials into their real and imaginary
parts, is
(1.8.13) i =

1
a0 +
ap cos (i 1)pk + bp sin (i 1)pk ,
2
p=1
where
(1.8.14) ap = 2
(cp ),
bp = 2 (cp )
for p = 0, . . . , M, with the understanding that b0 = 0, where

and denote the real
and imaginary parts. Accordingly,
(1.8.15) cp = 12 (ap + i bp ).
Using Fourier orthogonality properties (e.g., [35]), we find that
(1.8.16) cp =
N
1 (i 1)p
i
N
i=1
or
(1.8.17) cp =
1 + 2 p + 3 2p + + N (N 1)p ,
N
where
(1.8.18) = exp(ik).
These formulas indicate that

N
1
(1.8.19) c0 = a0 =
i ,
N
i=1
that is, the Fourier constant a0 is the mean of all nodal values.
When N is odd, we truncate the Fourier sum at M = (N 1)/2 and compute cp for
p = 0, . . . , M. When N is even, we truncate the Fourier sum at M = N/2, compute cp
and bp for p = 0, . . . , M 1, using formula (1.8.17), and set
(1.8.20) cM =
1 2 + 3 N .
N
The alternating signs arise because N/2 = exp(i ) = 1.
1.8.4 Cosine Fourier Expansion
If a real periodic nodal field, , is symmetric with respect to the midpoint of the
network, that is,
(1.8.21) i = N+2i ,
we may use the cosine Fourier expansion

(1.8.22) i =
1
2
a0 +
N
1

ap cos[(i 1)pk ],
p=1
where ap are cosine Fourier coefficients.

Using Fourier orthogonality properties, we find that
(1.8.23) ap =
1 + 2 cos(kp) + 3 cos(2kp) + + N cos[(N 1)kp]

N
for p = 1, . . . , N 1, and
N
2
(1.8.24) a0 =
i
N
i=1
that is, the Fourier constant a0 is twice the arithmetic mean of the nodal values.
The associated complex Fourier series is
(1.8.25) i =
N1

cp exp i (i 1)pk ,
p = (N 1)
where cp = 12 ap .
1.8.5 Sine Fourier Expansion
If a real periodic nodal field, , is antisymmetric with respect to the midpoint of the
network, that is,
(1.8.26) i = N+2i ,
we may use the sine Fourier expansion

(1.8.27) i =
N
1

p=1
where ap are cosine Fourier coefficients.
bp sin[(i 1) pk ],
Using Fourier orthogonality properties, we find that

(1.8.28) bp =
2 sin(kp) + 3 sin(2kp) + + N sin[(N 1)kp]

N
for p = 1, . . . , N 1.
The associated complex Fourier series is
(1.8.29) i =
N
1

cp exp i (i 1) pk ,
p = (N 1)
where cp = 12 i ap for p = 1, . . . , N 1 and cp = 0.
Exercise
1.8.1 Link labeling
Confirm that the factorization L = R RT is independent of link labeling.
/// 2 ///
GRAPHS AND NETWORKS
A graph is broadly defined as a collection of N nodes, also called

vertices, connected by L links, also called edges (e.g., [54]). The number of nodes,
N, is the order of a graph and the number of links, L, is the size of a graph. In graph
theory, a graph is typically denoted as G (V , E ), where the set V contains the vertices
and the set E contains the edges. In science, engineering, and other applications,
a graph represents a network consisting of conductive or convective pathways, as
discussed in Chapter 4. The terms graph and network will be used interchangeably
in our discourse.
2.1 ELEMENTS OF GRAPH THEORY
One of the most attractive features of graph theory is that nodes and links can
be labeled arbitrarily, independently, and in an uncorrelated fashion, as shown in
Figure 2.1.1(a), where the nodes are marked as filled circles. Eight nodes and twelve
links define this network, N = 8 and L = 12. Note that links numbered 2 and 7 do not
cross at a node. The network shown in Figure 2.1.1(a) is reminiscent of a structural
truss.
2.1.1 Adjacency Matrix
In graph theory, an N N adjacency matrix is introduced, A, defined such that Aij = 1

if nodes i and j are connected by a link, and Aij = 0 otherwise, where i, j = 1, . . . , N.
By convention, the diagonal elements of the adjacency matrix are zero. By construction, the adjacency matrix is symmetric. For example, the adjacency matrix of the
network shown in Figure 2.1.1(a) is the 8 8 matrix shown in Figure 2.1.1(b).
The total number of links in a network, L, is equal to the number of ones in the
upper or lower triangular part of the adjacency matrix,
(2.1.1) L =
N1
N

i=1 j=i+1
Aij =
N
i1

i=2 j=1
1
Aij .
2
N
Aij =
i=1 j=1
The fraction 1/2 in front of the last double sum accounts for the inherent symmetry
of A.
26
G r a p h s a n d N e t w o r k s / / 27
(a)
7
8
11
9
6
10
12
8
5
5
6
2
1
3
3
1
(b)
A=
0
1
1
1
0
0
0
0
1
0
1
1
0
0
0
0
1
1
0
0
1
0
0
0
1
1
0
0
1
1
0
0
0
0
1
1
0
0
1
1
0
0
0
1
0
0
1
0
0
0
0
0
1
1
0
1
0
0
0
0
1
0
1
0
(c)
R=
1
0
1
0
0
0 1
0
0
0
0
0
1 1
0 1
0
0
0
0
0
0
0
0
0
1 1
0
0
1
0
0
0
0
0
0
0
0
0
1 1
0
1 1
0
0
0
0
0
0
0
0
1 1
0
0
0 1
0
1
0
0
0
0
0
0
0
1 1
0
0
0
0
0
0
0
0
0
0
0
1
1 1
0
0
0
0
0
0
0
0
0
0
0
1 1
FIGURE 2.1.1 (a) Illustration of a typical graph consisting of N = 8

nodes, also called vertices, connected by L = 12 links, also called
edges. (b) The corresponding 8 8 adjacency matrix and (c) the
corresponding 8 12 oriented incidence matrix. In this conceptual
depiction, edges are allowed to cross over without intersecting at a
node.
Spectrum of the Adjacency Matrix
The eigenvalues and eigenvectors of the node adjacency matrix, A, denoted by i

for i = 1, . . . , N contain useful information on the structure of the graph. Although
the layout of the adjacency matrix depends on the node labeling, the eigenvalues are
independent of node labeling.
Since A is symmetric, it has real eigenvalues and a complete set of orthogonal
eigenvectors. The sum of the eigenvalues of A is equal to the trace of A, which is
zero. A necessary but not sufficient condition for two graphs to be isomorphic is that
the spectra of the corresponding adjacency matrices are identical.
Suppose that we begin traveling on a continuous path departing from the ith node
along a chain of s links, so that we end up at the jth node. The number of possible
pathways is equal to the ij component of the matrix power As . The sum of the number
of pathways that return to a starting node after s steps have been made is
N

(2.1.2) ns =
si .
i=1
We find that
(2.1.3) n0 = N,
n1 = 0,
n2 = 2L,
n3 = 6T,
where T is the number of triangles formed by the links.

2.1.2 Node Degrees
The degree of the ith node, denoted by di , is defined as the number of links attached
to the node, connecting the node to its nearest neighbors. A node and its nearest
neighbors define a neighborhood.
By construction, di is equal to the number of ones in the ith row or column of the
adjacency matrix. The degree of an isolated node is zero. For example, the degrees
of the eight nodes comprising the network shown in Figure 2.1.1(a) are
(2.1.4)
d1 = 3,
d2 = 2,
d3 = 3,
d4 = 4,
d5 = 4,
d6 = 2,
d7 = 3,
d8 = 2.
In the case of an infinite network consisting of a regular lattice, the vertex degrees
are also called the lattice coordination number, as discussed in Section 2.6.
The sum of the degrees of all nodes in a finite network is equal to twice the
number of all links,
(2.1.5)
N

di = 2L.
i=1
Consequently,
(2.1.6)
N
2
=
,
L dav
where
(2.1.7) dav
N
1
di
N
i=1
is the average or mean node degree. Equation (2.1.6) is also valid for an infinite
network where N and L are infinite but their ratio is well defined.
2.1.3 The Complete Graph
By definition, each node of a complete graph is connected to every other node, as

shown in Figure 2.1.2. Consequently, all off-diagonal elements of the adjacency matrix are equal to unity, that is, the adjacency matrix is the complement of the identity
matrix. The degree of each node is N 1, and the number of links is equal to the number of elements in the strictly upper or lower triangular part of the adjacency matrix,
1
(2.1.8) L = N(N 1).
2
For N = 3, we find that L = 3, describing a triangle. For N = 5, we find that L = 10,
as shown in Figure 2.1.2. Sometimes a complete graph is also called a clique. In
qualitative terms, the complete graph describes the best connected network.
2.1.4 Complement of a Graph
The union of a graph and its complement forms a complete graph. Consequently,
the complement of an arbitrary graph with adjacency matrix A is another graph with
adjacency matrix
(2.1.9) A = Ac A,
where Ac is the adjacency matrix of the complete graph. For example, the adjacency
matrix of the complement of the graph shown in Figure 2.1.1(a) is
0
0
0
0
1
1
1
1
0
0
0
0
1
1
1
1
0
0
0
1
0
1
1
1
0
0
1
0
0
0
1
1
(2.1.10) A =
.
1
1
0
0
1
1
0
0
1
1
1
0
0
1
0
1
1
1
1
1
0
0
0
0
1
1
1
1
0
1
0
0
6
3
1
10
3
FIGURE 2.1.2 Illustration of a complete graph consisting of N = 5 nodes (vertices) connected by L = 10

links (edges).
In this case, the complement contains a higher number of links than the original
graph.
2.1.5 Connectivity Lists and the Oriented Incidence Matrix
It is helpful to introduce two connectivity lists represented by the L-dimensional

vectors k and l, defined such that the label of the first node of the mth link is km and
the label of the second node of the mth link is lm , where m = 1, . . . , L. These two
connectivity lists can be arranged into L 2 edge list. The adjacency matrix can be
extracted from the edge list, and vice versa.
For example, the 12-dimensional connectivity lists of the 12 links comprising the
network shown in Figure 2.1.1(a) are
(2.1.11)
k = [ 1, 2, 3, 2, 4, 5, 1, 4, 6, 5, 7, 8 ],
l = [ 2, 3, 1, 4, 5, 3, 4, 6, 7, 7, 8, 5 ].
If nodes, links, or both are relabeled, the connectivity lists undergo corresponding
permutations.
In an undirected graph, discussed exclusively in this book, because the order of
the end points is immaterial, km and lm can be switched freely for each m. This is not
true in the case of a directed graph, also called a digraph, where an ordered part of
end points defines an arrow. In-degrees and out-degrees are defined in a digraph.
It is useful to introduce an N L oriented incidence matrix, R, defined such that
Ri, j = 0, except that
(2.1.12) Rkm ,m = 1,
Rlm ,m = 1
for m = 1, . . . , L. For example, the oriented incidence matrix of the network shown
in Figure 2.1.1(a) is the 8 12 matrix shown in Figure 2.1.1(c). Typically, but not
always, the number of links is much greater than the number of nodes, L N, and
the matrix R resembles a horizontal strip.
2.1.6 Connected and Unconnected Graphs
A graph is connected if at least one continuous path of links can be found leading
us from an arbitrary node to any other arbitrary node. If a continuous path cannot be
found, the graph is unconnected. Fragments and islands consisting of isolated nodes
or groups of nodes are found in an unconnected graph. The number of islands in
an unconnected graph can be diagnosed from the number of zero eigenvalues of the
Laplacian matrix, as discussed in Section 2.2.
2.1.7 Pairwise Distance and Diameter
A physical or abstract length or weight can be assigned to each link of a graph. The
length of each link of an unweighed graph is set to unity by convention, whereas the
length of a link in an weighed graph is set to a specified link weight, as discussed in

Chapter 4. In both cases, length is measured in predetermined units appropriate for
the physical, engineering, or information system under consideration.
The pairwise distance between two selected nodes is the minimum length of the
shortest path between these nodes. In the case of an unweighed graph, the pairwise
distance is an integer expressing the number of links along the shortest path between
the two nodes. The maximum pairwise distance over all pairs of nodes is the graph
diameter.
2.1.8 Trees
We saw that a complete graph describes the best connected network for a given number of nodes, in that any pair of nodes is connected by a link. The number of links in
a complete network scales with N 2 .
On the opposite part of the spectrum lies a tree network distinguished by the
absence of cyclical paths, as shown in Figure 2.1.3. The number of links in a tree
network is less by one than the number of nodes, L = N 1. If an arbitrary link is
clipped, a connected tree network breaks up into two disconnected tree networks.
Metaphorically speaking, a tree network is on the verge of disintegration.
2.1.9 Random and Real-Life Networks
A random graph with N vertices is characterized by the probability, p, that any pair
nodes is connected by a link, independent of any other connections. The expected
node degree is
(2.1.13) < d >= p (N 1),
5
6
5
4
4
1
3
2
2
3
FIGURE 2.1.3 Illustration of a tree network consisting

of N = 7 nodes (vertices) connected by L = 6
links (edges). The absence of triangles and cycles is a
distinguishing feature of a general tree network.
and the expected number of links is

(2.1.14) < L > = p 12 N(N 1).
When p = 1, we obtain a complete graph where the expected values are equal to the
corresponding actual values.
The degree distribution in a random graph is described by a binomial function,

N1
(2.1.15) PN (d) =
pd (1 p)N1k ,
d
where d N 1. The first large parentheses on the right-hand side denote the
combinatorial,

(2.1.16)
m
k
m +1
m!
=
,
k ! (m k) !

l
=1
where l is the minimum of k and m k. As N , the binomial distribution tends

to the Poisson distribution.
Real-Life Networks
Deterministic and random networks encountered in real life are described by node
degree distributions that differ significantly from the binomial or Poisson distribution (e.g., [32]). Node degree distributions are often skewed to the right or exhibit
a power-law behavior. Theoretical models of real-life networks have been proposed
according to their indented physical, engineering, biological, sociological, or other
application in different specializations.
Exercises
2.1.1 Complement of a graph
Draw the complement of the graph shown in Figure 2.1.1(a).
2.1.2 Node clustering
The clustering index of the ith node is defined as i = mi /(mi )max , where mi is the
number of links connecting its neighbors. Show that (mi )max = 12 di (di 1).
2.2 LAPLACIAN MATRIX
Let D be a diagonal matrix whose ith diagonal element is equal to the degree of
the ith node, di . The N N graph Laplacian matrix, L, is defined in terms of the
adjacency matrix, A, the degrees of the nodes encapsulated in D, and the oriented
incidence matrix, R, as
(2.2.1) L = D A
or
(2.2.2) L = R RT .
By construction, the sum of the elements in each row or column of L is zero.

The factorization (2.2.2) shows that the Laplacian is given by the sum of the
tensor product of L vectors,
(2.2.3) L =
L

(m) (m) ,
m=1
where (m) is the mth column of R and denotes the tensor product of two vectors.
Specifically, (m) (m) is an N N matrix with components
(m) (m)
(2.2.4) [(m) (m) ]ij = i j .
We recall that the vector (m) is filled with zeros, except that the entry corresponding
to the first end node is 1 and the entry corresponding to the second end node is 1,
and find that two diagonal components of the tensor product are equal to 1 and two
off-diagonal components are equal to 1. Thus, the matrix (m) (m) has only four
nonzero components.
For example, the 8 8 Laplacian matrix of the network shown in Figure 2.1.1(a)
is given by
(2.2.5) L =
3
1
1
1
0
0
0
0
1
3
1
1
0
0
0
0
1
1
3
0
1
0
0
0
1
1
0
4
1
1
0
0
0
0
1
1
4
0
1
1
0
0
0
1
0
2
1
0
0
0
0
0
1
1
3
1
0
0
0
0
1
0
1
2
Note that the sum of the elements in each row or column is zero.
2.2.1 Properties of the Laplacian Matrix
Being a real and symmetric matrix, the Laplacian matrix, L, has real eigenvalues and
a complete set of mutually orthogonal eigenvectors. The eigenvalues and eigenvectors of L provide us with a wealth of information on the structure of the underlying
network. Since the sum of the elements in each row of L is zero, a vector with equal
components is an eigenvector of L corresponding to the null eigenvalue.
Let an N-dimensional vector, , contain the nodal values of a discrete field at the
N nodes of a network. We find that
(2.2.6) L =
L

(km lm )2 0,
m=1
which demonstrates that the Laplacian is positive semidefinite. Consequently, the

eigenvalues of L, denoted by i , are either zero or positive. The sum of the eigenvalues of L is equal to the trace of L, which is equal to the trace of D, which is equal to
the sum of the degrees of all nodes.
We may assume that the eigenvalues have been ordered so that
(2.2.7) 0 = 1 2 N ,
where the first eigenvalue, 1 , is always zero. Further or all other eigenvalues may
also be zero.
The second smallest eigenvalue, 2 , is of particular interest in spectral graph theory. The value of 2 is sometimes called the algebraic connectivity of the network.
We know that 2 > 0 only when the graph is not connected, that is, when the graph
consists of two or more unconnected subgraphs. This observation suggests that 2 is
a sensible measure of the contiguity of a network represented by a graph. The maximum value 2 = N is attained for a complete graph. More generally, the number of
zero eigenvalues of L is equal to the number of isolated nodes or clusters of nodes.
The set of eigenvalues of a graph consisting of a number of disconnected subgraphs is the union of the eigenvalues of the constituent subgraphs, where each
subgraph contributes a zero eigenvalue. Other properties of the Laplacian eigenvalues
are reviewed by Mohar [30].
Let u(i) be the eigenvector of the Laplacian corresponding to the ith eigenvalue.
We know that the eigenvector u(1) corresponding to the zero eigenvalue, 1 = 0, is
filled with ones. Orthogonality of the set of eigenvectors requires that u(i) u(1) = 0
for i > 1, yielding
(2.2.8)
N

(i)
uj = 0
j=1
for i > 1, which shows that the mean value of the components of any but the first
eigenvector is zero.
2.2.2 Complete Graph
All elements of the Laplacian matrix of a complete graph are equal to 1, except for
the diagonal elements that are equal to N 1,
(2.2.9) L =
N1
1
1
..
.
1
N1
1
..
.
1
1
N1
..
.
..
.
1
1
1
..
.
1
1
1
..
.
1
1
1
..
.
1
1
1
1
1
1
1
1
1
N1
1
1
1
N1
1
1
1
N1
One may confirm that L2 = NL and, more generally,

(2.2.10) Lk = N k1 L,
for any positive integer, k.

The first eigenvalue of the Laplacian matrix is zero, and the rest of the eigenvalues
are equal to the number of nodes, N,
(2.2.11) 1 = 0,
n = N
for n = 2, . . . , N.
One useful set of eigenvectors, u(n), normalized so that their lengths are equal to
unity, u(n) u(n) = 1, is

(n)
(2.2.12) ui
1
= exp(i i n ),
N
for n = 1, . . . , N, where
(2.2.13) n =
n1
2
N
and i is the imaginary unit, i2 = 1. Because of the pronounced multiplicity of the

eigenvalues, other sets of eigenvectors can be chosen.
2.2.3 Estimates of Eigenvalues
Estimates for the magnitudes of the second smallest and largest eigenvalues of the
arbitrary graph, 2 and N , are available (e.g., [27, 30]). For example, it can be shown
that the second eigenvalue satisfies the inequality
(2.2.14) 2
N
min(di ).
N1 i
The last eigenvalue satisfies the inequality

(2.2.15)
N
max(di ) N max(di + dj ),
i,j
N1 i
for any pair of nodes, i and j, are connected by a link. We conclude that, if all node
degrees are zero, N = 0 and all eigenvalues are also zero.
2.2.4 Spanning Trees
A spanning tree is a continuous chain of links that visit all N nodes of a network in
the absence of local loops. Kirchhoffs spanning-tree theorem states that the number
of spanning trees in a network is
(2.2.16) nt =
1
2 N .
N
In fact, nt is the absolute value of any minor of the graph Laplacian.

2.2.5 Spectral Expansion
An arbitrary nodal field encapsulated in a vector, , can be expressed as a weighed

sum of eigenvectors of the Laplacian matrix, u( j ) , so that
(2.2.17) =
N

cj u( j ) ,
j=1
where cj are appropriate coefficients. In index notation,

(2.2.18) i =
N

(j)
cj ui
j=1
for i = 1, . . . , N.
Assume that the eigenvectors have been normalized such that their norm is unity,
(
j
)
u u( j ) = 1, where an asterisk denotes the complex conjugate. Exploiting the
orthogonality of the eigenvectors, we obtain
(2.2.19) cj = u( j ) .
The spectral expansion in terms of the eigenvectors shown in (2.2.17) is the discrete counterpart of the Fourier expansion of a continuous function in terms of
trigonometric functions or orthogonal polynomials.
2.2.6 Spectral Partitioning
We have remarked that the eigenvector corresponding to the zero eigenvalue of the
Laplacian matrix, 1 , is uniform over the nodes of a network. Higher eigenvectors
partition the network into two or a higher number of pieces (spectral partitioning).
To partition a network, we may group together nodes whose eigenvector components
corresponding to a specified eigenvalue have the same sign. The eigenvalue with
the second smallest magnitude, 2 , is chosen for division into two fragments, while
higher eigenvalues are chosen for division into a higher number of fragments.
The success of spectral partitioning relies on the zero-mean property expressed

by (2.2.8), roughly stating that an equal number of eigenvector components with positive and negative sign appear. More sophisticated partitioning methods are available
(e.g., [11]).
Square Network
As an example, the spectral partitioning of a square network is shown in Figure 2.2.1.

Positive components of an eigenvector are marked as filled circles, negative components are marked as dots, and zero components are unmarked. The network shown
consists of N = 172 = 289 nodes connected by L = 544 links. The degrees of the 4
corner nodes is 2, the degrees of the 60 edge nodes is 3, and the degrees of the 225
interior nodes is 4.
Exact expressions for the eigenvalues and eigenvectors of the Laplacian of the
square network are available, as discussed in Chapter 3. The first nine eigenvalues
corresponding to the eigenvectors shown in Figure 2.2.1 are = 0, 0.0341 (double),
0.0681, 0.1351 (double), 0.1691 (double), and 0.2701, accurate to the fourth decimal
place.
FIGURE 2.2.1 Spectral partitioning of a Cartesian network consisting of a complete set

of horizontal and vertical links.
2.2.7 Complement of a Graph
The Laplacian of the complement of a graph, indicated by a prime, is given by

(2.2.20) L = D A = (Dc D) (Ac A) = Lc L,
where the superscript c denotes the complete graph.

Let P () be the characteristic polynomial of the Laplacian of a graph, L. The
characteristic polynomial of the Laplacian of the complement of the graph, L , is
(2.2.21) P () = (1)N1
P (N ).
N
Corresponding eigenvalues are related by

(2.2.22) 1 = 0,
i+1 = N Ni+1
for i = 1, . . . , N 1. In the case of a complete graph, Ni+1 = N and i+1 = 0.

2.2.8 Normalized Laplacian
Suppose that none of the degrees of the vertices is zero, that is, isolated nodes do
not appear. A normalized incidence matrix,
R, and the corresponding normalized

Laplacian, L, can be defined as
(2.2.23)
R D1/2 R,
T

L
R
R ,
where R is the oriented incidence matrix. Subject to these definitions, we have

(2.2.24) L = R RT = D1/2
L D1/2
and
(2.2.25)
L = D1/2 L D1/2 = I D1/2 A D1/2 ,
where I is the N N identity matrix. By construction, all diagonal components of

the normalized Laplacian are equal to unity,
Lii = 1. The off-diagonal components

are Lij = 1/ di dj if nodes i and j are connected by a link, and zero otherwise.
The normalized Laplacian is a positive semidefinite matrix, having one zero ei
genvalue corresponding to an eigenvector whose ith component is di . However, the
rest of the eigenvalues are not necessarily equal to those of the Laplacian matrix. In
fact, the eigenvalues of the normalized Laplacian lie in the range [0, 2], whereas those
of the Laplacian lie in the range [0, ). The normalized Laplacian finds applications
in the theory of random walks.
2.2.9 Graph Breakup
A graph, G , can be broken into two pieces, G1 and G2 , by removing a set of links, E .
By construction, one end point of each removed link belongs to G1 , and the second
end point belongs to G2 . We are interested in finding the smallest possible cut set, E ,
that separates G into the two largest possible pieces. A measure of the quality of a
cut and fragility of G is the scalar
(2.2.26) h
|E |
,
|G1 ||G2 |
where the vertical bars denote an appropriate magnitude (volume). Cheegers

constant is defined as
(2.2.27) hG max h.
G1
Cheegers theorem relates Cheegers constant to the second eigenvalue of the

normalized Laplacian, 2 ,

(2.2.28) hG 22 ,
hG 12 2 .
As 2 tends to zero, indicating graph fragmentation, Cheegers constant also tends to
zero.
Exercise
2.2.1 Normalized Laplacian
Derive the normalized Laplacian of the graph shown in Figure 2.1.1(a).
2.3 CUBIC NETWORK
A three-dimensional network in physical space can be projected onto a plane for

better visualization. For example, a cubic network consisting of N = 8 nodes and
L = 12 links can be projected onto a planar network, as shown in Figure 2.3.1(a).
Nodes and links are labeled arbitrarily in this illustration. The corresponding 8 8
node adjacency matrix is shown in Figure 2.3.1(b).
The 12-dimensional connectivity lists of the 12 links defining the link end points
are
(2.3.1)
k = [ 1, 2, 3, 4, 5, 6, 7, 8, 1, 2, 3, 4 ],
l = [ 2, 3, 4, 1, 6, 7, 8, 5, 6, 7, 8, 1 ].
The 8 12 oriented incidence matrix is shown in Figure 2.3.1(c). The degree of each
node is 3, and the graph Laplacian is
(2.3.2) L = 3 I A = R RT ,
(a)
6
6
7
6
10
7
2
8
11
6
10
7
4
12
11
3
12
(b)
A=
0
1
0
1
0
1
0
0
1
0
1
0
0
0
1
0
0
1
0
1
0
0
0
1
1
0
1
0
1
0
0
0
0
0
0
1
0
1
0
1
1
0
0
0
1
0
1
0
0
1
0
0
0
1
0
1
0
0
1
0
1
0
1
0
(c)
R=
1
0
0
1
0
0
0
0 1
0
0
0
1 1
0
0
0
0
0
0
0 1
0
0
0
1 1
0
0
0
0
0
0
0 1
0
0
0
1 1
0
0
0
0
0
0
0 1
0
0
0
0 1
0
0
1
0
0
0
1
0
0
0
0
1 1
0 0
1
0
0
0
0
0
0
0
0
1 1
0
0
1
0
0
0
0
0
0
0
0
1 1
0
0
1
0
FIGURE 2.3.1 (a) Illustration of a cubic network and its projection on the
plane. (b) The adjacency matrix and (c) the oriented incidence matrix.
The cubic network consists of N = 8 nodes (vertices) connected by
L = 12 links (edges). Nodes and links are labeled arbitrarily in this
example.
where I is the 8 8 identity matrix. Making substitutions, we find that
3
1
0
1
0
1
0
0
1
3
1
0
0
0
1
0
0
1
3
1
0
0
0
1
0
1
3
1
0
0
0
1
(2.3.3) L =
0
0
0
1
3
1
0
1
1
0
0
0
1
3
1
0
0
1
0
0
0
1
3
1
0
0
1
0
1
0
1
3
The eight eigenvalues of L are = 0, 2 (triple), 4 (triple), and 6. The corresponding eigenvectors are illustrated in Figure 2.3.2, where positive components are
marked with filled (green) circles, negative components are marked with hollow (red)
circles, and zero components are unmarked. The sets of filled or hollow circles provide us with a spectral partitioning of the cubic network. Note that in the case of the
FIGURE 2.3.2 Spectral partitioning based on the eigenvectors of the

Laplacian matrix of a cubic network corresponding to eigenvalues
= 0, 2, 2, 2 (first row) and = 4, 4, 4, 6 (second row). Positive components are marked with filed circles, negative components are marked with hollow circles, and zero components are
unmarked.
highest eigenvalue, = 6, each filled circle has three nearest hollow circles, and each
hollow circle has three nearest filled circles.
Exercises
2.3.1 Node and link labeling
Derive the Laplacian matrix for a node and link labeling scheme of your choice
that is different than that shown in Figure 2.3.1. Confirm that the eigenvalues of the
Laplacian remain unchanged.
2.3.2 Diagonal link
Derive the Laplacian matrix when a diagonal link of your choice is added to a cubic
network.
2.4 FABRICATED NETWORKS
Finite or closed, planar or three-dimensional networks can be fabricated by the finite

element subdivision of a parental structure or by the Delaunay triangulation based on
a specified set of nodes, as discussed in this section. In the case of the finite element
tessellation, the network links are the element edges and the network nodes are the
element vertices. Each network can be mapped onto another isomorphic network that
can be partitioned in similar ways in terms of the eigenvectors of the Laplacian. In
the following discussion, the number of nodes with degree d is denoted as nd , where
the sum of all nd is the number of nodes, N.
2.4.1 Finite-Element Network on a Disk
A network associated with a finite element assembly of three-node triangles on a

disk, generated by the successive subdivision of a hard-coded, four-element parental
structure, is shown in Figure 2.4.1(a). The number of nodes is N = 145, the number
of lines is L = 400, and the node degree distribution is n3 = 4, n4 = 29, and n6 = 112,
indicating a dominant hexagonal structure.
The first few eigenvalues of the Laplacian are = 0, 0.1436 (double), 0.3388,
0.3523, 0.4890, 0.7154 (double), and 0.9322 (double). The multiple eigenvalues
are due to the inherent fourfold symmetry of the network. The corresponding
(a)
(b)
FIGURE 2.4.1 Spectral partitioning of a network arising from a finite

element grid on (a) a disk and (b) a square.
eigenvectors implementing spectral partitioning are shown with circular symbols in

Figure 2.1.1(a). In all illustrations presented in this section, positive components of
an eigenvector are marked with a filled circle, negative components are marked with
a dot, and zero components are unmarked.
2.4.2 Finite-Element Network on a Square
A network arising from a finite element assembly of triangles on a square, generated

by the successive subdivision of a hard-coded, eight-element parental structure, is
shown in Figure 2.4.1(b). The number of nodes is N = 172 = 289, the number of
links is L = 800, and the node degree distribution is n3 = 8, n4 = 56, and n6 = 224,
indicating a nearly hexagonal structure.
The first few eigenvalues of the Laplacian are = 0, 0.0564 (double), 0.1436
(double), 0.2540 (double), 0.3388, 0.3597, and 0.4890. The corresponding eigenvectors implementing spectral partitioning are shown in Figure 2.4.1(b). Comparing
this partitioning with that shown in Figure 2.2.1 for the Cartesian network reveals
significant differences in the spatial structure of high-order eigenvectors.
2.4.3 Delaunay Triangulation of an Arbitrary Set of Nodes
A network arising from the Delaunay triangulation based on an arbitrary set of nodes
in the xy plane is shown in Figure 2.4.2(a). The underlying Voronoi tessellation consisting of polygons, performed by a Matlab function, is indicated by the dashed (red)
lines. Each point inside a Voronoi cell is nearest to the corresponding central node
than to any other node.
The network shown in Figure 2.4.2(a) consists of N = 19 nodes connected by
L = 45 links arising from the Delaunay triangulation. The node degree distribution is
broad: n3 = 2, n4 = 7, n5 = 4, and n6 = 6. The first few eigenvalues of the Laplacian
are = 0, 0.0564, 0.7950, 1.3897, 2.1478, 2.8924, 3.1845, 3.7750, 4.4834, 4.7087,
and 5.1862. The corresponding eigenvectors implementing spectral partitioning are
shown in Figure 2.1.2(c). We observe that the second and third eigenvectors divide
the network into two different contiguous pieces.
2.4.4 Delaunay Triangulation of a Perturbed Cartesian Grid
A network arising from Delaunay triangulation of a perturbed Cartesian grid is

shown in Figure 2.4.2(b). To generate this network, nodes are distributed on a Cartesian grid with spacings x and y and are then displaced randomly along the x and
y axes by distances x and y, where is a uniform deviate. The links are determined by Delaunay triangulation performed by a Matlab function. The underlying
Voronoi tessellation is indicated by the dashed (red) lines in Figure 2.4.2(b).
The particular network shown in Figure 2.4.2(b) consists of N = 92 = 81 nodes
connected by L = 208 links. The node degree distribution is broader than that of
(a)
(b)
1
1
0.8
0.6
0.6
0.4
0.4
0.2
0.2
0.8
0.2
0.2
0.4
0.4
0.6
0.6
0.8
0.8
1
1
0.5
0
x
0.5
0.5
0
x
0.5
(c)
FIGURE 2.4.2 A network arising from the Delaunay triangulation of (a) a set of
arbitrary points or (b) a perturbed square lattice. (c) Spectral partitioning of the
network shown in (a). The spectral partitioning of the network shown in (b) is
illustrated in Figure 2.4.3.
the corresponding Cartesian network, n2 = 2, n3 = 5, n4 = 22, n5 = 20, n6 =

18, n7 = 11, and n8 = 3. The first few eigenvalues of the Laplacian matrix are
= 0, 0.1965, 0.2050, 0.3547, 0.7003, 0.8131, 0.8908, 0.9472, 1.4048, and 1.5030.
Multiple eigenvalues do not appear due to the lack of symmetry. The corresponding
eigenvectors implementing spectral partitioning are shown in Figure 2.4.3.
2.4.5 Finite Element Network Descending from an Octahedron
A closed network associated with a finite element grid of triangles on a sphere,

generated by the successive subdivision of an octahedral assembly, is shown in
FIGURE 2.4.3 Spectral partitioning of a network produced by the Delaunay

triangulation of a set of nodes deployed on a perturbed square lattice.
Figure 2.4.4(a). The number of nodes is N = 258, the number of links is L = 768, and
the node degree distribution is bimodal (n4 = 6 and n6 = 252), indicating a nearly
hexagonal structure. As seen previously, the number of links is significantly higher
than the number of nodes.
The first several eigenvalues of the Laplacian matrix are = 0, 0.1648 (triple),
0.3946 (double), 0.5691 (triple), and 0.8253 (triple). The corresponding eigenvectors
implementing spectral partitioning are shown in Figure 2.4.4(a).
2.4.6 Finite Element Network Descending from an Icosahedron
A closed network associated with a finite element grid of triangles on a sphere,

generated by the successive subdivision of an icosahedral assembly, is shown in Figure 2.4.4(b). The number of nodes is N = 162, the number of links is L = 480, and
the node degree distribution is bimodal, n5 = 12 and n6 = 150, indicating a nearly
hexagonal network.
The first few eigenvalues of the Laplacian matrix are = 0, 0.2643 (triple),
0.7715 (quintic), and 1.3707 (triple). The corresponding eigenvectors implementing
spectral partitioning are shown in Figure 2.4.4(b).
(a)
(b)
FIGURE 2.4.4 Spectral partitioning of a network associated with a finite element grid arising from the subdivision of (a) an octahedron or (b) an icosahedron on a
sphere.
Exercise
2.4.1 Delaunay triangulation
Generate a graph based on the Delaunay triangulation of a set of nodes of your
choice.
2.5 LINK REMOVAL AND ADDITION
In science, engineering, and other applications, a graph typically describes a physical

or abstract network. Links can be attenuated or removed due to damage, or added to
enhance the performance of the network, as discussed in Chapter 6. Link clipping
or addition alters the Laplacian matrix and may have a profound influence on the
overall performance of the network.
2.5.1 Single and Multiple Link
Suppose that one link numbered m is removed (clipped) from a network, where m =
1, . . . , L and L is the total number of links in the pristine state. If L0 is the pristine
Laplacian before clipping, then
(2.5.1) L = L0 (m) (m)
will be the altered Laplacian after clipping, where (m) is the mth column of the
pristine oriented incidence matrix before link removal, R0 , and denotes the tensor
product of an ordered pair of vectors. The ij component of the symmetric matrix
(m) (m) is
(m) (m)
(2.5.2) [(m) (m) ]ij = i j .
We recall that the vector (m) is filled with zeros, except that the entry corresponding
to the first end node is 1, and the entry corresponding to the second end node is 1.
The number of links, L, and the degrees of the two nodes defining the broken link
are reduced by one unit after link clipping.
Using Cauchys interlacing theorem, we find that the eigenvalues of the altered
matrix, L, interlace those of the pristine matrix, L0 , that is,
(2.5.3) 0 = 1 = 01 2 02 N 0N ,
which means that all eigenvalues move toward zero, in agreement with physical
intuition (e.g., [15]).
One-Dimensional Network
In the case of a one-dimensional network discussed in Section 1.7, the pristine

Laplacian is
0
(2.5.4) L =
1
1
0
..
.
1
2
1
..
.
0
1
2
..
.
..
.
0
0
0
..
.
0
0
0
..
.
0
0
0
0
0
0
0
0
0
2
1
0
1
2
1
0
0
0
..
.
1
1
After the mth link has been clipped, the altered Laplacian is

(2.5.5) L =
L0m
0
0
L0Nm

,
where L0m is the m m pristine Laplacian and L0Nm is the (N m) (N m) pristine

Laplacian.
For example, when m = 3, the altered Laplacian is
(2.5.6) L =
1
1
0
0
0
0
..
.
1
2
1
0
0
0
..
.
0
1
1
0
0
0
..
.
0
0
0
1
1
0
..
.
0
0
0
1
2
1
..
.
0
1
2
..
.
0
0
0
..
.
0
0
0
0
0
0
..
.
0
0
0
0
0
0
..
.
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
2
1
0
1
2
1
0
0
0
0
0
0
..
.
1
1
Removing one link in a one-dimensional graph results in a disconnected graph.

The eigenvalues of the damaged Laplacian are
r1
,
m 2
s1
s = 4 sin2
N m 2
(2.5.7) r = 4 sin2
for r = 1, . . . , m and s = 1, . . . , N m. Note that two eigenvalues are zero due to

network disruption. The union of these eigenvalues interlaces those of L0 , as shown
in Figure 2.5.1.
Multiple Link Removal
If several links are clipped from a network, corresponding terms are subtracted from
the right-hand side of (2.5.1). Suppose that M links are clipped, where M L. The
Laplacian after clipping is
0.2
0
0.2
0.5
1.5
2.5
3.5
FIGURE 2.5.1 Eigenvalue spectrum of the Laplacian of a one-dimensional

network with N = 16 nodes after the fourth link has been clipped, m = 4.
The vertical lines mark the eigenvalues before clipping, and the square
and symbols mark the eigenvalues after clipping.
(2.5.8) L = L0
M

(mi ) (mi ) ,
i=1
where mi is the label of the ith clipped link and 1 mi L. If all links are clipped
(M = L), the Laplacian reduces to the null matrix. A general theorem on the interlacing of the eigenvalues after multiple clippings is not available, except when M = 1 or
L. However, the second eigenvalue after multiple clippings, 2 , is guaranteed to be
less than that before clipping [10].
The number of zero eigenvalues of the Laplacian after clipping, N0 , is equal to the
number of isolated nodes or clusters of nodes. If no links are clipped in a connected
network (M = 0), the number of zero eigenvalues is precisely equal to one, N0 = 1. If
all links are clipped (M = L), the number of zero eigenvalues is equal to the number
of nodes, N0 = N. These observations suggest that the number of zero eigenvalues is
a useful diagnostic of the operational state of a network.
When a small number of pL links remain intact in a randomly clipped, almost
devastated network, we obtain

L
(2.5.9) N0 N pL = N 1 p
,
N
irrespective of the network structure, where p 0 is the percentage of active links.
Higher-order terms in p depend on the network structure [37].
2.5.2 Link Addition
Suppose that one link labeled L + 1 is added to an existing graph with L links. If L0
is the Laplacian before addition, then
(2.5.10) L = L0 + (L+1) (L+1)
will be the Laplacian after addition, where (L+1) is the L + 1 column of the new
oriented incidence matrix, R. The number of links, L, and the degrees of the two
nodes defining the new link increase by one unit after addition. However, unless
new nodes are introduced, the number of nodes, N, remains unchanged. Cauchys
interlacing theorem can be used to relate the eigenvalues of the Laplacian before and
after link addition.
Suppose that new links are added to an existing graph. If L0 is the Laplacian
before addition, then
(2.5.11) L = L0 +
(L+p) (L+p)
p=1
will be the Laplacian after addition, where (L+p) is the L + p column of the new
oriented incidence matrix, R.
Exercise
2.5.1 Periodic one-dimensional graph
Prepare the counterpart of Figure 2.5.1 for a periodic one-dimensional graph
discussed in Section 1.8.
2.6 INFINITE LATTICES
Structured networks forming infinite lattices are convenient theoretical models for
studying the structural and transport properties of idealized states. Infinite lattices
are typically visualized as crystals in physical two- or three-dimensional space. Other
isomorphic representations can be obtained by compressing, stretching, or deforming
these physical states.
The node degree of a lattice, d, is also called the lattice coordination number.
When not all node degrees are equal, a mean coordination number can be defined as
the arithmetic average of all node degrees, dav . We recall from (2.1.6) that the ratio
of the number of nodes to the number of links is
(2.6.1)
N
2
=
.
L dav
Although both N and L are infinite, the ratio N/L is well-defined, determined by the
mean coordination number.
2.6.1 Bravais Lattices
The node position of a two-dimensional Bravais lattice in physical state can be

identified by two indices, i1 and i2 ,
(2.6.2) xi1 ,i2 = x0,0 + i1 a1 + i2 a2 ,
where a1 and a2 are two corresponding base vectors. In three dimensions, three
indices are employed, i1 , i2 , and i3 , and the nodal positions are
(2.6.3) xi1 ,i2 ,i3 = x0,0,0 + i1 a1 + i2 a2 + i3 a3 ,
where a1 , a2 , and a2 are three base vectors.

Physically, a Bravais lattice appears the same, independent of the choice of the
anchor node, x0,0 in two dimensions or x0,0,0 in three dimensions.
Reciprocal Lattice
A two-dimensional Bravais lattice has a reciprocal lattice whose base vectors, b1 and
b2 , satisfy the relation
(2.6.4) ai bj = 2 ij ,
where ij is Kroneckers delta defined such that ij = 0 if i = j and ij = 1 if i = j.

A three-dimensional Bravais lattice has a reciprocal lattice whose base vectors,
b1 , b2 , and b3 , also satisfy equation (2.6.4). Using the properties of the triple mixed
product, we find that
(2.6.5) b1 =
2
a2 a3 ,
b2 =
2
a3 a1 ,
b3 =
2
a1 a2 ,
where
(2.6.6) = a1 (a2 a3 )
is the volume of the unit cell in the physical space. The nodes of the reciprocal lattice
are located at
(2.6.7) lp1 ,p2 ,p3 = p1 b1 + p2 b2 + p3 b3 ,
where p1 , p2 , and p3 are three integers.

Periodic Functions
Consider a function, f (x), that is repeated periodically in the direction of each base
vector so that
(2.6.8) f (x) = f (x + i1 A1 + i2 A2 + i3 A3 )
for any triplet of integers, i1 , i2 , and i3 , where

(2.6.9) A1 = N1 a1 ,
A 2 = N2 a2 ,
A 3 = N3 a3 ,
and N1 , N2 , and N3 are specified integers determining the size of the periodic box.
The periodic function can be expanded in a Fourier series,

(2.6.10) f (x) =
cp1 ,p2 ,p3 exp i kp1 ,p2 ,p3 x ,
p1 ,p2 ,p3
where cp1 ,p2 ,p3 are Fourier coefficients,

(2.6.11) kp1 ,p2 ,p3 = p1 B1 + p2 B2 + p3 B3
are wave numbers, p1 , p2 , and p3 are three integers, and

(2.6.12) B1 =
1
b1 ,
N1
B2 =
1
b2 ,
N2
B3 =
1
b3
N3
are fractions of the reciprocal base vectors. The sum in (2.6.10) is computed over a
finite portion of the reciprocal lattice, called the discrete Brillouin zone or Wigner
Seitz cell.
One-Dimensional Lattice
In the case of a uniform one-dimensional lattice along the x axis, we omit the
subscript 1 indicating the x direction and set
(2.6.13) a = a e,
b=
2
e,
a
k=
2
e,
Na
where a is the node separation and e is the unit vector along the x axis. The Fourier
expansion of a periodic function is
(2.6.14) f (x) =

p

2 x
cp exp i p
.
N a
Evaluating this expansion on the lattice nodes, xi = (i 1) a, yields the Fourier series
(1.8.12). The discrete Brillouin zone is discussed at the conclusion of Section 1.8.3.
Two-Dimensional Cartesian Lattice
In the case of a uniform two-dimensional Cartesian (square) lattice in the xy plane,

we set
(2.6.15) a1 = a e1 ,
a2 = a e2
and
2
e1 ,
a
b2 =
2
e2 ,
a
2
e1 ,
N1 a
k2 =
2
e2 ,
N2 a
(2.6.16) b1 =
and also define

(2.6.17) k1 =
where a is the common node separation in each direction and e1 , e2 are unit vectors
along the first and second directions, which can be identified with the x and y axes.
Consequently,
(2.6.18) f (x, y) =

p1 ,p2
cp1 ,p2

2 x
2 y
exp i p1
+ p2
.
N1 a
N2 a
Evaluating this expansion at nodes defined by the grid lines

(2.6.19) xi1 = (i1 1)a,
yi2 = (i2 1)a
for i1 = 1, . . . , N1 and i2 = 1, . . . , N2 yields the Fourier series discussed in Section 3.1.4. The discrete Brillouin zone of the two-dimensional lattice is discussed
near the end of Section 3.1.4.
2.6.2 Archimedean Lattices
An Archimedean lattice consists of an infinite doubly periodic array regular polygons. In particular, each node is surrounded by the same sequence of polygons.
Precisely 11 Archimedean lattices can be found, as shown in Figure 2.6.1. The
notation (nm , kl , . . . ) signifies that each node is surrounded sequentially by m nsided polygons, followed by l k-sided polygons and possibly other similar polygons
indicated be the three dots [13].
Square, A(44 ) Lattice
The Archimedean 44 lattice, also known as the square lattice, is a Bravais lattice
consisting of a doubly periodic array of empty squares, as shown in Figure 2.6.1(a),
(a)
(b)
(c)
(d)
(e)
(f)
FIGURE 2.6.1 Illustration of the first six Archimedean lattices, including (a)
the square, (b) honeycomb, (c) hexagonal, (d) kagom (e) A(3, 122 ), and
(f ) bathroom tile lattice. The dashed lines in (a, b, c, f ) describe the dual
lattice. The dual of the bathroom tile lattice (f ) is the Union Jack lattice
shown in Figure 2.6.2(a). (Continued on next page)
(g)
(h)
(i)
(j)
(k)
FIGURE 2.6.1 (Continued) Illustration of the last five Archimedean lattices,

including the (g) cross (h) ruby, (i) maple leaf, (j) the A(33 , 42 ), and (k) the
puzzle lattice. The dashed lines in (j) describe the dual lattice shown in
Figure 2.6.2(b).
where two base vectors are drawn with arrows. The notation 44 signifies that each
node is surrounded by four squares. The lattice coordination number is d = 4 and the
ratio of the number of nodes to the number of links is N/L = 2/d = 1/2.
Hexagonal or Triangular, A(36 ) Lattice
The Archimedean 36 lattice, also known as the hexagonal or triangular lattice, is a

Bravais lattice consisting of two doubly periodic arrays of vacant triangles, as shown
in Figure 2.6.1(b), where two base vectors are drawn with arrows. The notation 36
signifies that each node is surrounded by six triangles. The hexagonal lattice arises
from a sheared square lattice by introducing one family of slanted parallel links. The
lattice coordination number is d = 6 and the ratio of nodes to links is N/L = 1/3.
Honeycomb, A(63 ) Lattice
The Archimedean 63 lattice, also known as the honeycomb lattice, consists of a doubly periodic array of vacant hexagonal tiles, as shown in Figure 2.6.1(c). The notation
63 signifies that each node is surrounded by three hexagons. The hexagonal lattice
is a composite lattice consisting of two displaced triangular lattices, as discussed in
Section 3.5. The lattice coordination number is d = 3 and the ratio of the number of
nodes to the number of links is N/L = 2/d = 2/3.
Kagom, A(3, 6, 3, 6) Lattice
The Archimedean (3, 6, 3, 6) lattice, also known as the kagom lattice, tiles the
plane with triangles and hexagons, as shown in Figure 2.6.1(d). The Japanese word
kagom means woven bamboo lattice. The notation (3, 6, 3, 6) signifies that each
node is surrounded sequentially by one triangle, one hexagon, another triangle, and
another hexagon. The kagom lattice is a composite Bravais lattice consisting of
three displaced hexagonal lattices, as discussed in Section 3.6. The lattice coordination number is d = 4 and the ratio of the number of nodes to the number of links is
N/L = 1/2.
Star, A(3, 122 ) Lattice
The Archimedean (3, 122 ) lattice, also known as the star lattice, tiles the plane with
triangles and dodecagons (12-sided polygons), as shown in Figure 2.6.1(e). The notation (3, 122 ) signifies that each node is surrounded sequentially by one triangle and
two dodecagons. The lattice coordination number is d = 3 and the ratio of the number
of nodes to the number of links is N/L = 2/3.
Square Octagon, Bathroom Tile, A(4, 82 ) Lattice
The Archimedean (4, 82 ) lattice, also known as the square octagon or bathroom tile
lattice, covers the plane with squares and octagons, as shown in Figure 2.6.1(f ). The
notation (4, 82 ) signifies that each node is surrounded sequentially by one square and
two octagons. The lattice coordination number is d = 3 and the ratio of the number
of nodes to the number of links is N/L = 2/3. The bathroom tile lattice arises from
the square lattice shown in Figure 2.6.1(a) by replacing alternating nodes with small
tilted squares.
Cross, A(4, 6, 12) Lattice
The Archimedean (4, 6, 12) lattice, also known as the cross lattice, tiles the plane
with squares, hexagons, and dodecagons, as shown in Figure 2.6.1(g). The notation
(4, 6, 12) signifies that each node is surrounded sequentially by one square, one hexagon, and one dodecagon. The lattice coordination number is d = 3 and the ratio of
the number of nodes to the number of links is N/L = 2/3.
Ruby or Bounce, A(3, 4, 6, 4) Lattice
The Archimedean (3, 4, 6, 4) lattice, also known as the ruby or bounce lattice, covers the plane with triangles, squares, and hexagons, as shown in Figure 2.6.1(h).
The notation (3, 4, 6, 4) signifies that each node is surrounded sequentially by one
triangle, one square, one hexagon, and another square. The lattice coordination number is d = 4 and the ratio of the number of nodes to the number of links is
N/L = 1/2.
Maple Leaf, Snub Hexagonal, A(34 , 6) Lattice
The Archimedean (34 , 6) lattice, also known as the snub hexagonal or maple leaf
lattice, covers the plane with triangles and hexagons, as shown in Figure 2.1.1(i). The
notation (34 6) signifies that each node is surrounded sequentially by four triangles
and one hexagon. The lattice coordination number is d = 5 and the ratio of the
number of nodes to the number of links is N/L = 2/d = 2/5.
Bridge, A(33 , 42 ) Lattice
The Archimedean (33 , 42 ) lattice, also known as the bridge lattice, tiles the plane with
triangles and squares, as shown in Figure 2.6.1( j ). The notation (33 42 ) signifies that
each node is surrounded sequentially by three triangles and two squares. The lattice
coordination number is d = 5 and the ratio of the number of nodes to the number of
links is N/L = 2/5.
Puzzle, Snub Square, A(32 , 4, 3, 4) Lattice
The Archimedean (32 , 4, 3, 4) lattice, also known as the snub square or puzzle lattice,
tiles the plane with triangles and squares, as shown in Figure 2.6.1(k). The notation
(32 , 4, 3, 4) signifies that each node is surrounded sequentially by two triangles, one
square, another triangle, and another square. The lattice coordination number is d = 5
and the ratio of the number of nodes to the number of links is N/L = 2/d = 2/5.
2.6.3 Laves Lattices
Laves lattices, denoted by the prefix D, are the duals of the Archimedean lattices.
A Laves lattice arises by introducing vertices in the middle of the tiles (faces) of
an Archimedean lattice and then connecting the vertices to cross the edges of the
Archimedean lattice.
The dual of the square lattice is the same square lattice, the dual of the hexagonal
lattice is the honeycomb lattice, and the dual of the honeycomb lattice is the hexagonal lattice, as shown in Figure 2.6.1(ac). The dual lattices of the remaining eight
Archimedean lattices are non-Archimedean lattices. Because all vertices do not have
the same degree, only a mean coordination number can be defined. Two examples
illustrated in Figure 2.6.2 are discussed in the remainder of this section.
(a)
(b)
FIGURE 2.6.2 Illustration of (a) the Union Jack, D(4, 82 ), lattice and
(b) the D(33 , 42 ) lattice.
Union Jack, Tetrakis, Kisquadrille, D(4, 82 ) Lattice
The D(4, 82 ) Laves lattice, also called the Union Jack, tetrakis, or kisquadrille lattice,
is shown in Figure 2.6.2(a). The node degrees are d = 4, 8, the mean node degree is
d = 6, and the ratio of the number of nodes to the number of links is N/L = 1/3.
Pentagonal, D(33 , 42 ) Lattice
The D(33 , 42 ) Laves lattice, also called the pentagonal lattice, is shown in Figure 2.6.2(b). The node degrees are d = 3, 4, the mean node degree is d = 10/3,
and the ratio of the number of nodes to the number of links is N/L = 3/5.
2.6.4 Other Two-Dimensional Lattices
A variety of other lattices have been proposed. The martini lattice tiles the plane with
triangles and enneagons (nine-sided polygons), as shown in Figure 2.6.3(a) [13, 39].
The lattice coordination number is d = 3 and the ratio of the number of nodes to the
number of links is N/L = 2/d = 2/3. The martini lattice arises from the honeycomb
lattice by replacing every other junction around each hexagon with a triangle, thereby
introducing three additional edges.
(a)
(b)
FIGURE 2.6.3 Illustration of (a) the martini lattice and (b) the bow-tie
lattice.
The bow-tie lattice, shown in Figure 2.6.3(b), tiles the plane with triangles, and
rectangles. The node degrees are d = 4 and 6, the mean node degree is d = 5, and the
ratio of the number of nodes to the number of links is N/L = 2/d = 2/5.
2.6.5 Cubic Lattices
Three Bravais cubic lattice are known, including the simple cubic lattice, the bodycentered cubic (bcc) lattice, and the face-centered cubic (fcc) lattice, as shown in
Figure 2.6.4.
(a)
(b)
(c)
FIGURE 2.6.4 Illustration of (a) the simple cubic lattice, (b) the body-centered cubic (bcc)
lattice, and (c) the face-centered cubic (fcc)
lattice. Links are shown as solid lines and
lattice reference lines are shown as broken
lines.
Simple Cubic Lattice
The simple cubic lattice is a Bravais lattice consisting of empty cubes, as shown
in Figure 2.6.4(a), where the three base vectors are drawn with arrows. The lattice
coordination number is d = 6 and the ratio of the number of nodes to the number of
links is N/L = 2/d = 1/3.
Body-Centered Cubic Lattice
The body-centered cubic (bcc) lattice is a Bravais lattice consisting of two displaced
simple cubic lattices, as shown in Figure 2.6.1(b), where the three base vectors are
drawn with arrows. The lattice coordination number is d = 8 and the ratio of the
number of nodes to the number of links is N/L = 2/d = 1/4.
Face-Centered Cubic lattice
The face-centered cubic (fcc) lattice is a Bravais lattice arising from the simple cubic lattice by introducing one node at the center of each square face, as shown in
Figure 2.6.1(c), where the three base vectors are drawn with arrows. The lattice
coordination number is d = 12 and the ratio of the number of nodes to the number of links is N/L = 2/d = 1/6. The fcc lattice accommodates the densest possible
array of spheres.
Exercise
2.6.1 Cartesian networks
Compute the reciprocal base vectors of a three-dimensional Cartesian network with
base vectors a1 = a e1 , a2 = b e2 , and a3 = c e3 , where a, b, and c are three constants
and e1 , e2 , and e3 , are Cartesian unit vectors.
2.7 PERCOLATION THRESHOLDS
With reference to the infinite lattices discussed in Section 2.6, now we address the
important concept of percolation threshold determining the functional and operational state of a damaged network. A distinction between the link and the node
percolation threshold must be made at the outset according to the cause of the damage
inflicted on a given pristine state.
2.7.1 Link (Bond) Percolation Threshold
Suppose that a fraction of links, qlink , are clipped randomly from a large test section
of a pristine lattice containing L links, where 0 qlink 1. The fraction of intact
links is plink = 1 qlink . This means that the probability that an arbitrary link is intact
is plink .
(a)
(b)
(c)
(d)
FIGURE 2.7.1 Structure of a doubly periodic 16 16 square lattice after

M links have been clipped for (a) qlink = M/L = 0.1719, (b) 0.3809, (c)
0.5840, and (d) 0.7754, where L is the number of links in the pristine
state. Unremoved links are shown as solid segments and removed
links are shown as broken segments.
As an example, the damaged states of a square or honeycomb doubly periodic

network are shown in Figures 2.7.1 and 2.7.2 for four values of qlink . Because of the
inflicted damage, isolated clusters of nodes appear at sufficiently high values of the
damaged fraction, qlink .
As the sizes of the periodic boxes shown in Figures 2.7.1 and 2.7.2 increase in
both directions, the probability that a node belongs to a cluster spanning the periodic
box vanishes above a critical threshold, qlink
c . The corresponding probability,
(2.7.1) plink
= 1 qlink
c
c ,
defines the link or bond percolation threshold. Physically, as plink tends to plink
c from
link
lower values, the mean cluster size becomes infinite. Conversely, as p
tends to
plink
from
higher
values,
the
mean
cluster
size
becomes
finite.
c
In the case of a one-dimensional network consisting of an infinite or closed chain
of links, the link percolation threshold is precisely zero, plink
= 0. The reason is that
c
all links must be intact for a cluster spanning the entire network to appear.
(a)
(b)
(c)
(d)
FIGURE 2.7.2 Structure of a doubly periodic 16 16 honeycomb lattice

after M links have been clipped for (a) qlink = M/L = 0.2044, (b) 0.3893,
(c) 0.5612, and (d) 0.8047, where L is the number of intact links. Intact
links are shown as solid segments and removed links are shown as
broken segments.
2.7.2 Node Percolation Threshold
Now suppose that a fraction of nodes, qnode , are removed randomly from a large test
section of pristine lattice containing N nodes, along with the links originating from
each node, where 0 qnode 1. The fraction of remaining nodes is pnode = 1qnode .
This means that the probability that an arbitrary node is intact is pnode . Since a link
is intact only if both end nodes are present, the corresponding probability that a link
is present is
2
(2.7.2) p = pnode .
As an example, the damaged states of a square or honeycomb doubly periodic network are shown in Figures 2.7.3 and 2.7.4. As in the link removal problem, because
of the damage inflicted, isolated clusters of nodes are observed at sufficiently high
values of qnode .
As the sizes of the periodic boxes shown in Figures 2.7.3 and 2.7.4 increase in
both directions, the probability that an unremoved node belongs to a cluster spanning
the periodic box vanishes above a critical threshold, qnode
. The corresponding
c
probability,
(2.7.3) pnode
= 1 qnode
,
c
c
is the node or site percolation threshold. It is important to note that pnode

is not related
c
to plink
by
(2.7.2).
c
(a)
(b)
(c)
(d)
FIGURE 2.7.3 Structure of a doubly periodic 16 16 square lattice

after K nodes have been removed for (a) qnode = K /N = 0.1797,
(b) 0.3945, (c) 0.5977, and (d) 0.7773, where N is the number of
nodes. Intact links are shown as solid segments, removed links are
shown as broken segments, and unremoved nodes are marked as
circles inside a period.
In the case of a one-dimensional network consisting of an infinite or closed chain

of links, the node percolation threshold is zero, pnode
= 0. The reason is that all links
c
must be intact for a cluster spanning the entire network to appear.
2.7.3 Computation of Percolation Thresholds
To compute the link percolation threshold, we may consider the ghost of a lattice
consisting of invisible links playing the role of nameplates. In the bond percolation
problem, functional links are gradually introduced to replace the ghost links until
long-range connectivity is established at the bond percolation threshold. In the site
percolation problem, nodes and their associated links are introduced until long-range
connectivity is established at the site percolation threshold.
Exact link and bond percolation thresholds are known only for a few lattice geometries [40, 43, 44, 57]. Remarkably accurate percolation thresholds have been
calculated by numerical methods for other lattices (e.g., [28, 42, 58]). A comprehensive compilation accompanied by an extensive list of references is available at
(a)
(b)
(c)
(d)
FIGURE 2.7.4 Structure of a doubly periodic 16 16 honeycomb lattice

after K nodes have been removed for (a) qnode = K /N = 0.1953, (b)
0.4141, (c) 0.5938, and (d) 0.7969, where N is the number of nodes. Unremoved links are shown as solid segments, removed links are shown
as broken segments, and unremoved nodes are marked as circles
inside a period.
the Internet site: http://en.wikipedia.org/wiki/Percolation_threshold. Results for a

several lattices are shown in Table 2.7.1.
The link (bond) percolation threshold of an Archimedean lattice and its corresponding Laves lattice add up to unity. The link percolation threshold of a lattice
is precisely equal to the node percolation threshold of its covering lattice. For
example, the link percolation threshold of the honeycomb lattice is equal to the node
percolation threshold of the kagom lattice.
Correlations
Graphs of the percolation thresholds against the lattice coordination number or mean
node degrees, d, are shown in Figure 2.7.5. Partially successful efforts have been
made to derive universal formulas for these thresholds in terms of the lattice coordination number and possibly other parameters (e.g., [53]). Of particular interest
are simple formulas that provide us with easily computable estimates for use in
engineering risk analysis and design.
For the link percolation problem, Vyssotsky et al. [48] proposed the approximation

(2.7.4) plink
,
c
2D
d
link
31
pc 3D
,
2d
TABLE 2.7.1 (a) Link (Bond) and (b) Node (Site) Percolation Thresholds for Several Lattices
pnode
c
1.0
1.0
Square
0.5a
0.59275
2.6.1(a)
Hexagonal
0.34730b
0.5c
2.6.1(b)
0.65270d
0.69704
2.6.1(c)
2.6.1(d)
Lattice
Open chain
2
2
2
plink
c
Space
Honeycomb
Figure
Kagom
0.52441
0.65270d
A(3, 122 )
0.74042
0.80790e
2.6.1(e)
Bathroom tile
0.67680
0.72972
2.6.1(f )
Cross
0.69373
0.74781
2.6.1(g)
Ruby
0.52483
0.62182
2.6.1(h)
Maple leaf
0.43431
0.57950
2.6.1(i)
A(33 , 42 )
0.41964
0.55021
2.6.1(j)
Puzzle
0.41414
0.55081
2.6.1(k)
D(33 , 42 )
3, 4
0.58035
0.64708
2.6.2(a)
0.23220
0.5c
2.6.2(b)
3 13
D-Bathroom tile
4, 8
Martini
0.70711f
0.76482g
2.6.3(a)
0.404518h
0.547
2.6.3(b)
0.31160
2.6.4(a)
Bowtie
4, 6
Simple cubic
0.24881
bcc
0.18029
0.246
2.6.4(b)
fcc
12
0.12016
0.19923
2.6.4(c)
Diamond
0.43
a This is an exact value [43, 44].

b The exact value is a root of the cubic equation x3 3x + 1 = 0, given by plink = 2 sin [43, 44].
c
18
c This is an exact value.
d The exact value is 1 2 sin [40]. The Kagom lattice is the covering lattice of the honeycomb
18
lattice.
e The exact value is (1 2 sin )1/2 [40].
18
f The exact value is 1/ 2 [57].

g The threshold is a root of the quartic equation x4 3x3 + 1 = 0 [57].
h The threshold is a root of the quintic equation x5 6x3 + 6x2 + x 1 = 0.
in two or three dimensions, where d is the lattice coordination number or mean node
degree. These estimates, represented by the solid and broken lines in Figure 2.7.5(a),
are in good agreement with known exact results.
Formula (2.7.4) for a two-dimensional lattice is consistent with the more general
formula
(a)
1
0.9
0.8
link
pc
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
6
12
10
(b)
1
0.9
0.8
node
p
c
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
7
d
10
11
12
FIGURE 2.7.5 Percolation thresholds plotted against

the lattice coordination number or mean node
degress, d, for (a) link and (b) node percolation. Twodimensional lattices are represented by circles or
crosses and three-dimensional lattices are represented by squares or symbols. The solid and broken lines represent, respectively, the predictions of
(2.7.4) for two- or three-dimensional lattices.
(2.7.5) plink

c
2
d
N
,
L
where d is the mean node degree, N is the number of nodes, and L is the number of
links. This approximation is motivated by the functional dependence of the number
of zero eigenvalues of the Laplacian on q [37]. Although more involved formulas for
predicting link percolation thresholds in regular lattices have been proposed (e.g.,
[46]), their practical utility is called into question and their generalization to finite
and inhomogeneous networks is unclear.
Scher and Zallen [41] introduced the notion of critical node percolation density,
c , defined with respect to the distance of a node from its nearest neighbor. The
distance is identified with the diameter of a disk in two dimensions or a sphere in
thee dimensions. If is the filling factor, defined as the fraction of the plane or space
occupied by all circles or spheres, then
(2.7.6) pnode

c
1
c .
It remarkable that c is nearly constant, equal to 0.44 for a two-dimensional lattice

or 0.154 for a three-dimensional lattice.
Exercise
2.7.1 Link percolation thresholds
Verify from the results shown in Table 2.7.1 that the link (bond) percolation threshold
of an Archimedean lattice and its corresponding Laves lattice add up to unity.
/// 3 ///
SPECTRA OF LATTICES
The nodes of a two- or three-dimensional regular lattice, regarded

as a structured network, can be identified by two or three indices assigned to the individual lattice directions. The adjacency and Laplacian matrices can be compiled by
inspection and their properties can be assessed by standard analytical methods. Several lattice networks with different boundary conditions are discussed in this chapter
and the spectrum of their Laplacian is delineated. The results will find applications in
Chapter 5 for computing of lattice Greens functions and in Chapter 6 for analyzing
the performance of conducting networks.
3.1 SQUARE LATTICE
A network whose structure is isomorphic to that of a square lattice consists of two

intersecting one-dimensional arrays of links. A rectangular patch of a square lattice
containing N1 links in the first direction, parametrized by the index i1 , and N2 links in
the second direction, parametrized by the index i2 , is shown in Figure 3.1.1. Isolated,
periodic, doubly periodic, and other configurations of a distributed nodal field can
be envisioned.
N2
i2
2
1
1
i1
N1
FIGURE 3.1.1 Illustration of a rectangular patch of a square

network containing N1 links in the first direction and N2
links in the second direction. All links are assumed to
have the same conductance.
67
A theorem due to Fiedler states that the eigenvectors of the Laplacian matrix for
certain types of boundary conditions are tensor products of those of the constituent
one-dimensional graphs, and the eigenvalues are the sums of the eigenvalues of the
Laplacian of the constituent one-dimensional graphs. [10]. This property reflects the
separability of the discrete Laplace operator in Cartesian coordinates.
3.1.1 Isolated Network
The total number of nodes in the isolated network shown in Figure 3.1.1 is
(3.1.1) N = (N1 + 1)(N2 + 1)
and the total number of links is

(3.1.2) L = N1 (N2 + 1) + (N1 + 1)N2 .
Note that the number of links is significantly higher than the number of nodes.
The nodal values of a nodal scalar field, , can be compiled in a sequence of
horizontal layers from the bottom where i2 = 1 to the top where i2 = N2 + 1, into an
N-dimensional vector
(1)
(2)
..
(3.1.3) = .
,
(N2 )
(N2 +1)
where the subvectors
(3.1.4) (1)
1,1
2,1
..
.
N1 +1,1
...,
(N2 +1)
1, N2 +1
2, N2 +1
..
.
N1 +1, N2 +1
encapsulate horizontal profiles. The Laplacian matrix consists of N2 + 1 rows of

(N1 + 1) (N1 + 1) square tridiagonal blocks, F, E, and I, in the following
configuration:
F
I
0
0
0
0
I
E
I
0
0
0
I
E
0
0
0
0
.
..
..
..
..
..
..
(3.1.5) L =
.
.
.
.
.
. ,
..
0
0
0
E
I
0
0
0
0
I
E
I
0
0
0
0
I
F
S p e c t r a o f L a t t i c e s / / 69
where I is the (N1 + 1) (N1 + 1) identity matrix. When N1 = 3, we have
2
1
(3.1.6) F =
0
0
1
3
1
0
0
0
,
1
2
0
1
3
1
3
1
E=
0
0
1
4
1
0
0
1
4
1
0
0
.
1
4
The two entries of F correspond to corner nodes, the three entries of F and E
correspond to boundary nodes, and the four entries of E correspond to interior
nodes.
The eigenvalues of the Laplacian matrix are
(3.1.7) n1 , n2 = 4 sin2 12 n1 + 4 sin2 12 n2
or
(3.1.8) n1 , n2 = 4 2 cos n1 2 cos n2 ,
where
(3.1.9) n1 =
n1 1
,
N1 + 1
n2 =
n2 1
N2 + 1
for n1 = 1, . . . , N1 + 1 and n2 = 1, . . . , N2 + 1.
The corresponding eigenvectors, u(n1 , n2 ) , normalized so that their lengths are
equal to unity, u(n1 , n2 ) u(n1 , n2 ) = 1, are
n ,n

(3.1.10) ui11, i2 2 = An1 Bn2

cos
(N1 + 1)(N2 + 1)
i1
1
2

1
n1 cos i2
n2
2
for n1 , i1 =
1, . . . , N1 + 1 and
n2 , i2 = 1, . . . , N2 + 1, where An1 = 1, Bn2 = 1, except
that A1 = 1/ 2 and B1 = 1/ 2.
The spectral partitioning of a 17 17 network is shown in Figure 2.2.1. Positive eigenvector components are marked with filled circles, negative components are
marked with dots, and zero components are unmarked.
3.1.2 Periodic Strip
With continued reference to the rectangular patch of the square lattice shown in Figure 3.1.1, now we assume that a nodal scalar field, , is periodic in the first direction
so that
(3.1.11) 1,i2 = N1 +1,i2
for i2 = 1, . . . , N2 + 1. The vector of unknown nodal values is
(1)
(2)
..
(3.1.12) = .
,
(N2 )
(N2 +1)
(3.1.13) (1)
1,1
2,1
..
.
N 1,1
1
N1 ,1
...,
(N2 +1)
1,N2 +1
2,N2 +1
..
.
N 1,N +1
1
2
N1 ,N2 +1
encapsulate horizontal profiles. The total number of entries in the vector is

N = N1 (N2 + 1). Note that nodal values along the left side are not included in the
vector of unknowns, as they are periodic images of those along the right side of the
rectangular strip.
The Laplacian matrix consists of N2 +1 rows of N1 N1 square, nearly tridiagonal
blocks, F and E,
F
I
0
0
0
0
I
E
I
0
0
0
0
I
E
0
0
0
.
..
..
..
..
..
..
.
(3.1.14) L =
.
.
.
.
.
. ,
.
0
0
0
E
I
0
0
0
0
I
E
I
0
0
0
0
I
F
where I is the N1 N1 identity matrix. For example, when N1 = 5, we have
3
1
0
0
1
1
3
1
0
0
(3.1.15) F =
1
3
1
0
0
0
0
1
3
1
1
4
1
0
0
1
1
4
1
0
0
0
1
4
1
0
0
0
1
4
1
1
0
0
1
4
and
(3.1.16) E =
The three entries of F correspond to the bottom and top edge nodes. The northeastern
and southwestern one entries of F and E implement the periodicity condition.
(3.1.17) n1 , n2 = 4 sin2 12 n1 + 4 sin2 12 n2
or
(3.1.18) n1 , n2 = 4 2 cos n1 2 cos n2 ,
where
(3.1.19) n1 =
n1 1
2 ,
N1
n2 =
n2 1
N2 + 1
for n1 = 1, . . . , N1 and n2 = 1, . . . , N2 + 1.
The corresponding eigenvectors, un1 ,n2 , normalized so that their lengths are equal
to unity, un1 , n2 un1 , n2 = 1, are

(3.1.20)
uni11,,in2 2
= An1

2
1
exp i i1 n1 cos
i2
n2
2
N1 (N2 + 1)
for n1 , i1 = 1, . . . , N1 and n2 , i2 = 1, . . . , N2 + 1, where i is the imaginary

unit, an
asterisk denotes the complex conjugate, and An1 = 1, except that A1 = 1/ 2.
Shown in Figure 3.1.2 is a network with N1 = 16 and N2 = 8 divisions. The first
few eigenvalues of the Laplacian are = 0, 0.1206, 0.1522 (double), 0.2729 (double), 0.4679, 0.5858 (double), 0.6203 (double), 0.7064 (double), and 1.0000. The
corresponding eigenvectors implementing spectral partitioning are also displayed.
Vertical Strip
With continued reference to the square network shown in Figure 3.1.1, now we
assume that the nodal scalar field, , is periodic in the second direction,
(3.1.21) i1 ,1 = i1 , N2 +1
for i1 = 1, . . . , N1 . The vector of unknown nodal values is
(1)
(2)
..
.
(3.1.22) =
(N2 1)
(N2 )
FIGURE 3.1.2 Spectral partitioning of a periodic strip of a square lattice with N1 = 16

and N2 = 8 divisions. Positive eigenvector components are marked with filled circles,
negative components are marked with dots, and zero components are unmarked.
(3.1.23) (1) =
1,1
2,1
..
.
N1 +1,1
...,
(N2 ) =
1, N2
2, N2
..
.
N1 +1, N2

N = (N1 + 1)N2 . Note that the nodal values along the top side of the rectangular
strip are not included in the vector of unknowns.
The Laplacian consists of N2 rows of (N1 + 1) (N1 + 1) square tridiagonal blocks
and two negative unit corner blocks implementing the periodicity condition in the
second direction,
E
I
0
0
0
I
I
E
I
0
0
0
I
E
0
0
0
0
.
..
..
..
..
..
..
(3.1.24) L =
.
.
.
.
.
. ,
..
0
0
E
I
0
0
0
0
0
I
E
I
I
0
0
0
I
E
where E is an (N1 + 1) (N1 + 1) tridiagonal matrix and I is the (N1 + 1) (N1 + 1)

identity matrix. The northeastern and southwestern identity blocks implement the
periodicity condition in the second direction. When N1 = 4, we have
3
1
0
0
0
1
4
1
0
0
(3.1.25) E = 0
1
4
1
0
.
0
0
1
4
1
0
The three corner entries correspond to the left and right boundary nodes.
(3.1.26) n1 , n2 = 4 sin2 12 n1 + 4 sin2 12 n2
or
(3.1.27) n1 , n2 = 4 2 cos n1 2 cos n2 ,
where
(3.1.28) n1 =
n1 1
,
N1 + 1
n2 =
n2 1
2
N2
for n1 = 1, . . . , N1 + 1 and n2 = 1, . . . , N2 .
The corresponding eigenvectors, un1 , n2 , normalized so that their lengths are equal
to unity, un1 ,n2 un1 , n2 = 1, are

2
n1 ,n2
(3.1.29) ui1 , i2 = Bn2
cos i1 12 n1 exp(i i2 n2 )
(N1 + 1)N2
for n1 , i1 = 1, . . . , N1 + 1 and
n2 , i2 = 1, . . . , N2 , where i is the imaginary unit and
Bn2 = 1, except that B1 = 1/ 2.
3.1.3 Doubly Periodic Network
With reference to the rectangular network shown in Figure 3.1.1, now we assume
that the nodal scalar field is periodic in two directions so that
(3.1.30) 1,i2 = N1 +1,i2 ,
i1 ,1 = i1 ,N2 +1 .
The vector of unknown nodal values inside each period is
(1)
(2)
..
(3.1.31) = .
,
(N2 1)
(N2 )
(3.1.32)
(1)
1,1
2,1
..
.
...,
(N2 )
N1 ,1
1, N2
2, N2
..
.
N1 , N2

N = N1 N2 . Note that the nodal values along the right and top boundaries of the
rectangular strip are not included in the vector of unknowns.
The Laplacian consists of N2 rows of N1 N1 nearly tridiagonal blocks, E, upper
and lower diagonal negative unit blocks, I, and two negative unit corner blocks
implementing the periodicity condition in the second direction,
(3.1.33) L =
E
I
0
..
.
I
E
I
..
.
0
I
E
..
.
..
.
0
0
0
..
.
0
0
0
..
.
0
0
I
0
0
0
0
0
0
E
I
0
I
E
I
I
0
0
..
.
I
E
where I is the N1 N1 identity matrix. The northeastern and southwestern corner blocks of L implement the periodicity condition in the second direction. When
N1 = 5, we have
(3.1.34) E =
4
1
0
0
1
1
4
1
0
0
0
1
4
1
0
0
0
1
4
1
1
0
0
1
4
The northeastern and southwestern corner elements of E implement the periodicity

condition in the first direction.
An eigenvalue, , and the corresponding eigenvector, u, of the doubly periodic
Laplacian, L, satisfy the equation
(3.1.35) 4 ui1 , i2 ui1 1, i2 ui1 +1, i2 ui1 , i2 1 ui1 , i2 +1 = ui1 , i2
at any node, (i1 , i2 ).

We find that the eigenvalues are given by
(3.1.36) n1 , n2 = 4 sin2 12 n1 + 4 sin2 12 n2
or
(3.1.37) n1 , n2 = 4 2 cos n1 2 cos n2 ,
where
(3.1.38) n1 =
n1 1
2 ,
N1
n2 =
n2 1
2
N2
for n1 = 1, . . . , N1 and n2 = 1, . . . , N2 . We can write

(3.1.39) n1 = (n1 1) k1 ,
n2 = (n2 1) k2 ,
where the parameters

(3.1.40) k1 =
2
,
N1
k2 =
2
N2
are directional wave numbers.


n ,n
(3.1.41) ui11, i2 2 =
exp i (i1 n1 + i2 n2 )
N1 N2
for n1 , i1 = 1, . . . , N1 and n2 , i2 = 1, . . . , N2 , where i is the imaginary unit and

an asterisk denotes the complex conjugate. The eigenvalues given in (3.1.37) can
be computed by substituting the eigenvectors given in (3.1.41) into (3.1.35) and
simplifying the resulting expression.
Block Circulant Matrices
An alternative method of deriving the eigenvalues hinges on the observation that the
doubly periodic Laplacian (3.1.33) is a block circulant matrix, that is, a circulant
matrix whose scalar elements are replaced by constituent matrices. A theorem due to
Friedman [12] states that the spectrum of eigenvalues of this matrix is the union of
the spectra of the following N1 N1 matrices:
(3.1.42) L(n2 ) = exp(in2 ) I + E exp(in2 ) I
or
(3.1.43) L(n2 ) = 2 cos n2 I + E
for n2 = 1, . . . , N2 . For example, when N1 = 4,
(3.1.44) L(n2 )
4 2 cos n2
1
=
0
1
1
4 2 cos n2
1
0
0
1
4 2 cos n2
1
0
.
1
4 2 cos n2
Using expression (1.8.2) for the eigenvalues of the one-dimensional periodic Laplacian, we recover the eigenvalues displayed in (3.1.37).
A doubly periodic network with N1 = 12 and N2 = 8 divisions is shown
in Figure 3.1.3. The first few eigenvalues of the Laplacian matrix are = 0,
0.2679 (double), 0.5858 (double), 0.8537 (quadruple), 0.1351 (double), 1.0000 (double), and 1.5858 (quadruple). The corresponding eigenvectors implementing spectral
partitioning of a doubly periodic field are also shown.
Fourier Expansions on a Cartesian Grid
A real, doubly periodic nodal field, , defined over an N1 N2 square lattice can be
expanded into a doubly Fourier series so that
M1

(3.1.45) i1 , i2 =
M2

cp1 , p2 exp i (i1 1) p1 k1 + (i2 1) p2 k2 ,
p1 =M1 p2 =M2
where M1 and M2 are two appropriate truncation levels and cp1 , p2 are complex Fourier coefficients. If the number of intervals, N1 , is odd, we truncate the double Fourier
sum at the value M1 = (N1 1)/2. If N1 is even, we truncate the double Fourier sum
at the value M1 = N1 /2. Similar truncation levels apply to M2 (e.g., [35]).
To ensure that the right-hand side of (3.1.45) is real, we require that
(3.1.46) cp1 , p2 = cp1 , p2 ,
where an asterisk denotes the complex conjugate. The complex Fourier coefficients
are given by
(3.1.47) cp1 , p2 =

1
p
2p
(N 1)p
q1 + 11 q2 + 2 1 q3 + + 1 1 1 qN1 ,
N1 N2
where
p
2p2
(3.1.48) qm = m,1 + 22 m, 2 + 2
m, 3 + + y(N2 1)p2 m, N2
and we have defined

(3.1.49) 1 = exp(i k1 ),
2 = exp(i k2 ).
FIGURE 3.1.3 Spectral partitioning of a periodic square lattice with N1 =

12 and N2 = 8 divisions inside each period. Positive eigenvector
components are marked with filled circles, negative components are
3.1.4 Doubly Periodic Sheared Network
The nodal field of a doubly periodic Cartesian network that is sheared along first axis
axis satisfies the periodicity conditions
(3.1.50) 1, j = N1 +1, j ,
i,1 = i+r, N2 +1 ,
where r is a specified integer. The Laplacian matrix is given in (3.1.33), except that
the northeastern corner block, I, is replaced by

(3.1.51) J =
0
Ir
IN1 r
0

,
FIGURE 3.1.4 Spectral partitioning of a sheared periodic square lattice with

N1 = 12 and N2 = 12 divisions inside each period, for shearing level r = 6.
Positive eigenvector components are marked with filled circles, negative
components are marked with dots, and zero components are unmarked.
and the southwestern corner block of L is replaced by JT , where

identity matrix. For example, when N1 = 8 and r = 5, we have
0
0
0
0
0
1
0
0
0
0
0
0
0
0
1
0
0
0
0
0
0
0
0
1
0
0
0
0
0
0
0
1
(3.1.52) J =
0
1
0
0
0
0
0
0
0
0
1
0
0
0
0
0
0
0
0
1
0
0
0
0
0
0
0
0
1
0
0
0
Ip is the p p
When r = 0 or N1 , the matrix J reduces to the N1 N1 identity matrix, I.

The spectral partitioning of a sheared network with N1 = 12 and N2 = 12
divisions is shown in Figure 3.1.4.
Exercises
3.1.1 Particle vibrations
The particles of a two-dimensional crystal are arranged on a square lattice
parametrized by two indices, i1 and i2 , in the xy plane. Small departures from the
equilibrium position generate restoring forces. The motion of the (i1 , i2 ) particle is
governed by Newtons law,
(3.1.53) m

d 2 xi1 , i2
= k xi1 +1, i2 + xi1 1, i2 + xi1 , i2 +1 + xi1 , i2 1 4 xi1 , i2 ,
2
dt
where m is the particle mass, k is a spring constant, and t stands for time. In the case
of harmonic oscillations,
(3.1.54) xi1 , i2 = wi1 , i2 exp(i t),
where i is the imaginary unit, is the angular frequency, and wi1 , i2 is an eigendisplacement. Derive and solve an algebraic eigenvalue problem for the eigenfrequencies and eigendisplacements.
3.1.2 Periodic Laplacian
Confirm by direct substitution that the eigenvalues given in (3.1.36) and corresponding eigenvectors given in (3.1.37) satisfy equation (3.1.35).
3.2 MBIUS STRIPS
A section of a Cartesian strip can be twisted by 180 around its length into the configuration shown in Figure 3.2.1(a). A sequence of twisted Cartesian strips can be
glued together to form a helical strip. A finite twisted strip can be bent, and the
narrow edges can be attached to yield the Mbius strip shown in Figure 3.2.1(b).
(a)
(b)
FIGURE 3.2.1 (a) A Cartesian strip is twisted by 180 along its

length to give a helical strip. (b) The helical strip can be bent and
the narrow edges can be attached to yield the Mbius strip.
3.2.1 Horizontal Strip
The nodal profile of a Mbius strip in the direction of the first index, i1 , satisfies the
reverse periodicity condition
(3.2.1) 1,i2 = N1 +1, N2 +2i2
for i2 = 1, . . . , N2 . For example, the southwestern nodal value is equal to the

northeastern nodal value,
(3.2.2) 1,1 = N1 +1, N2 +1 .
The vector of nodal values encapsulating N = N1 (N2 + 1) unknowns is
(1)
(2)
..
.
(3.2.3) =
(N2 )
(N2 +1)
where
(3.2.4)
(1)
1,1
2,1
..
.
N 1,1
1
N1 ,1
...,
(N2 +1)
1, N2 +1
2, N2 +1
..
.
N 1, N +1
1
2
N1 , N2 +1
The Laplacian matrix consists of N2 +1 rows of N1 N1 square tridiagonal blocks,

F and E, and a chain of sparse backdiagonal blocks,
(3.2.5) L =
F
I
0
..
.
I
E
I
..
.
0
I
E
..
.
..
.
0
0
J
..
.
0
J
0
..
.
0
0
J
0
J
0
J
0
0
E
I
0
I
E
I
J
0
0
..
.
I
F
where I is the N1 N1 unit matrix. The N1 N1 matrix J is null, expect that the
northeastern and southwestern corner elements are equal to unity. When N2 is even,
the central element of L is E J. When N1 = 4,
3
1
(3.2.6) F =
0
0
1
3
1
0
0
0
,
1
3
0
1
3
1
4
1
E=
0
0
1
4
1
0
0
1
4
1
0
0
1
4
and
1
0
(3.2.7) I =
0
0
0
1
0
0
0
0
1
0
0
0
,
0
1
0
0
J=
0
1
0
0
0
0
0
0
0
0
1
0
.
0
0
The three entries of F correspond to the bottom and top edge nodes.
(3.2.8) n1 , n2 = 4 sin2 12 n1 , n2 + 4 sin2 12 n2
or
(3.2.9) n1 , n2 = 4 2 cos n1 , n2 2 cos n2 ,
where
(3.2.10) n1 , n2 =
n 1 1 + n2
2 ,
N1
n2 =
n2 1
N2 + 1
for n1 = 1, . . . , N1 and n2 = 1, . . . , N2 + 1, where n2 = 0 if n2 is odd and n2 = 1/2 if

n2 is even [56]. Formally, we write
(3.2.11) n2 =
1 + (1)n2
.
4
Note that these expressions differ from those given in (3.1.19) for the periodic
rectangular network only by the presence of n2 in the first fraction.
to unity, un1 , n2 un1 , n2 = 1, are given by

2
n1 , n2
(3.2.12) ui1 , i2 = An1
exp i i1 n1 , n2 ) cos i2 12 n2
N1 (N2 + 1)
for n1 , i1 = 1, . . . , N1 and n2 , i2 = 1, . . . , N2 + 1, where i is the imaginary
unit, an
asterisk denotes the complex conjugate, and An1 = 1, except that A1 = 1/ 2.
A Mbius network with N1 = 16 and N2 = 8 divisions is shown in Figure 3.2.2.
The first few eigenvalues of the Laplacian matrix are = 0, 0.1522 (double), 0.1590
(double), 0.4577 (double), 0.4679, 0.5858 (double), 0.6202 (double), and 1.0095
(double). The corresponding eigenvectors implementing spectral partitioning are
shown in Figure 3.2.2.
FIGURE 3.2.2 Eigenvectors on a Mbius strip of a square network

with N1 = 16 and N2 = 8 divisions. Positive eigenvector components are marked with filled circles, negative components
are marked with dots, and zero components are unmarked.
3.2.2 Vertical Strip
The nodal profile of the vertical Mbius strip satisfies a reverse periodicity condition
in the second direction,
(3.2.13) i,1 = N1 +2i,N2 +1 .
For example, the southwestern nodal value is equal to the northeastern nodal value,
1,1 = N1 +1,N2 +1 . The vector of unknown nodal values encapsulating N = (N1 +1)N2
unknowns is
(1)
(2)
..
.
(3.2.14) =
(N2 1)
(N2 )
where
(3.2.15) (1)
1,1
2,1
..
.
N ,1
1
N1 +1,1
...,
(N2 )
1,N2
2,N2
..
.
N ,N
1 2
N1 +1,N2
The Laplacian matrix consists of N2 rows of (N1 + 1) (N1 + 1) tridiagonal blocks

in addition to two (N1 + 1) (N1 + 1) corner blocks,
E
I
0
0
0
J
I
E
I
0
0
0
I
E
0
0
0
0
..
..
..
..
..
..
..
(3.2.16) L =
.
.
.
.
.
.
. ,
0
0
E
I
0
0
0
0
0
I
E
I
J
0
0
0
I
E
where I is the (N1 + 1) (N1 + 1) unit matrix and J is the (N1 + 1) (N1 + 1) unit
back-diagonal matrix. When N1 = 4, we obtain the 5 5 matrices
(3.2.17) E =
3
1
0
0
0
1
4
1
0
0
0
1
4
1
0
0
0
1
4
1
0
0
0
1
3
J=
0
0
0
0
1
0
0
0
1
0
0
0
1
0
0
0
1
0
0
0
1
0
0
0
0
(3.2.18) n1 ,n2 = 4 sin2 12 n1 ,n2 + 4 sin2 12 n2
or
(3.2.19) n1 , n2 = 4 2 cos n1 , n2 2 cos n2 ,
where
(3.2.20) n1 =
n1 1
,
N1 + 1
n1 , ln2 =
n2 1 + n1
2
N2
for n1 = 1, . . . , N1 + 1 and n2 = 1, . . . , N2 , where n1 = 0 if n1 is odd and = 1/2 if

n1 is even.
to unity, un1 , n2 un1 , n2 = 1, are given by

(3.2.21)
uni11,,in2 2
= Bn2

2
cos i1 12 n1 exp i i2 n1 ,n2
(N1 + 1)N2
for n1 , i1 = 1, . . . , N1 + 1 and n2 , i2 = 1, . . . , N2 , where i is the imaginary

unit, an
asterisk denotes the complex conjugate, and Bn2 = 1, except that B1 = 1/ 2.
3.2.3 Klein Bottle
The Klein bottle consists of two attached Mbius strips that are glued together along
one side and then folded to produce a bottle. The nodal field of the Klein bottle
satisfies the reverse periodicity condition of the Mbius strip in the first direction
and the usual periodic condition in the second direction,
(3.2.22) 1,i2 = N1 +1, N2 +2i2 ,
i1 ,1 = i1 , N2 +1
for i1 = 1, . . . , N1 + 1 and i2 = 1, . . . , N2 + 1. The vector of unknown nodal values is
(1)
(2)
..
.
(3.2.23) =
(N2 1)
(N2 )
where
(3.2.24) (1)
1,1
2,1
..
.
N 1,1
1
N1 ,1
...,
(N2 +1)
1, N2 +1
2, N2 +1
..
.
N 1, N +1
1
2
N1 , N2 +1
The Laplacian matrix consists of N2 rows of N1 N1 tridiagonal blocks, E, a chain

of sparse backdiagonal blocks, and two unit corner blocks,
E
I
0
0
0
J I
I
E
I
0
J
0
0
I
E
J
0
0
.
.
.
.
.
.
.
,
..
..
..
..
..
..
..
(3.2.25) L =
0
0
0
E
I
0
0
J
0
I
E
I
J I
0
0
0
I
E
where I is the N1 N1 identity matrix. The N1 N1 matrix J is null, except that the
northeastern and southwestern corner elements are equal to unity. When N1 = 4, we
have
4
1
0
0
0
0
0
1
1
0
4
1
0
0
0
0
,
.
(3.2.26) E =
J=
0
1
4
1
0
0
0
0
0
(3.2.27) n1 , n2 = 4 sin2 12 n1 ,n2 + 4 sin2 12 n2
or
(3.2.28) n1 , n2 = 4 2 cos n1 , n2 2 cos n2 ,
where
(3.2.29) n1 ,n2 =
n1 1 + n2
2 ,
N1
n2 =
n2 1
2
N2
for n1 = 1, . . . , N1 and n2 = 1, . . . , N2 + 1, and

0
for n2 = 1, 2, . . . , k,
(3.2.30) n2 = 1
for
n2 = k + 1, . . . , N2 .
2
The crossover threshold, k, is given by k = (N2 + 1)/2 if N2 is odd, or k = N2 /2 if N2

is even [45, 56].

(3.2.31)
uni11,,in2 2
= An1
2
N1 N2
1/2
exp i i1 n1 , n2

1
cos i2 2 n2
for n1 , i1 = 1, . . . , N1 , i2 = 1, . . . , N2 , and n2 = 1, . . . , k, where i is the imaginary
unit, an asterisk denotes the complex conjugate, and An1 = 1, except that A1 = 1/ 2.
When n2 = k + 1, . . . , N2 , the cosine is replaced by a sine on the right-hand side of
(3.2.31). When N2 is even and n2 = N2 /2, the cosine yields a sawtooth wave.
Exercise
3.2.1 Mbius strips and Klein bottle
(a) Confirm the eigenvalues and eigenvectors of the horizontal Mbius strip. (b)
Repeat for the vertical Mbius strip. (c) Repeat for the Klein bottle.
3.3 HEXAGONAL LATTICE
The hexagonal lattice arises from the square lattice by adding one right- or leftleaning slanted link inside each square cell, dividing it into two triangular cells. A
rectangular patch of a hexagonal network consisting of N1 links in the first direction,
N2 links in the second direction, and one left-leaning slanted link inside each square
cell is shown in Figure 3.3.1(a). As in the case of the square lattice, the nodes are
parametrized by two indices, i1 and i2 .
Natural State
The natural state of the hexagonal lattice patch shown in Figure 3.3.1(a) consists of
arrays of equilateral triangles in the xy plane, as shown in Figure 3.3.1(b). The nodes
fall on a Bravais lattice with base vectors
(3.3.1) a1 = a (1, 0),
a2 = a
1
(1, 3),
2
where a is the triangle edge length. The nodal positions are

(3.3.2) xi1 , i2 = x1,1 + (i1 1)a1 + (i2 1)a2 ,
for i1 = 1, . . . , N1 + 1 and i2 = 1, . . . , N2 + 1, where x1,1 is the arbitrary position of the

first node. The reciprocal base vectors of the hexagonal lattice, b1 and b2 , satisfying
by definition ai bj = 2ij are given by

2
1
2
2
(3.3.3) b1 =
1, ,
b2 =
0, ,
a
a
3
3
where ij is Kroneckers delta. The nodes of the reciprocal lattice are located at
(3.3.4) kn1 , n2 = (n1 1) b1 + (n2 1) b2 ,
where n1 and n2 are two integers.
The total number of nodes in the isolated network depicted in Figure 3.3.1 is
(3.3.5) N = (N1 + 1)(N2 + 1)

(3.3.6) L = N1 (N2 + 1) + (N1 + 1)N2 + N1 N2 .
The nodes shown in Figure 3.3.1(a) can be compiled in a sequence of horizontal layers from the bottom where i2 = 1 to the top where i2 = N2 + 1. With this convention,
(a)
N2
i2
2
1
1
N1
i1
(b)
a2
a1
N2
i2
y
2
x
1
1
i1
N1
FIGURE 3.3.1 (a) Illustration of a hexagonal network containing N1 links

in the first direction, N2 links in the second direction, and an appropriate
number of cross-links links. A hexagonal cell can be identified around
each interior node. (b) The network has been sheared to its physical
configuration consisting of stacked equilateral triangles.
a nodal field, , is hosted by the N-dimensional vector
(1)
(2)
..
(3.3.7) = .
,
(N2 )
(N2 +1)
where
(3.3.8) (1)
1,1
2,1
..
.
(N2 +1)
...,
N1 +1,1
1, N2 +1
2, N2 +1
..
.
N1 +1, N2 +1
The Laplacian matrix consists of N2 + 1 rows of (N1 + 1) (N1 + 1) tridiagonal

blocks, F, E, and G, in the following configuration:
F
J
0
0
0
0
JT
E
J
0
0
0
T
0
J
E
0
0
0
.
.
.
.
.
.
.
.
.
.
.
.
.
(3.3.9) L =
.
.
.
.
.
.
. ,
0
0
0
E
J
0
0
0
0
JT
E
J
0
0
0
0
JT
G
where J is the (N1 + 1) (N1 + 1) lower bidiagonal unit matrix. For example, when
N1 = 4, we have
2 1 0 0 0
4 1
0
0
0
1 4 1 0 0
1
6 1
0
0
(3.3.10) F =
E=
6 1
0
0 1 4 1 0 ,
0 1
.
0 0 1 4 1
0
0 1
6 1
0 0 0 1 3
0
0
0 1
4
and
(3.3.11) G =
3
1
0
0
0
1
4
1
0
0
0
1
4
1
0
0
0
1
4
1
0
0
0
1
2
J=
1
1
0
0
0
0
1
1
0
0
0
0
1
1
0
0
0
0
1
1
0
0
0
0
1
The two entries of F and G correspond to the southwestern and northeastern corner
nodes, the four entries of F and E correspond to the edge nodes, and the six entries
of E correspond to interior nodes. Analytical expressions for the eigenvalues and
eigenvectors of the Laplacian are not available.
With continued reference to Figure 3.3.1(a), now we assume that a nodal scalar field,
, is periodic in the directions of both indices, i1 and i2 , so that
(3.3.12) 1, i2 = N1 +1, i2 ,
i1 , 1 = i1 , N2 +1 .
The vector of unknown nodal values encapsulating N = N1 N2 unknowns inside each

periodic is
(1)
(2)
..
.
(3.3.13) =
(N2 1)
(N2 )
where
(3.3.14) (1) =
1,1
2,1
..
.
...,
(N2 ) =
N1 ,1
1,N2
2,N2
..
.
N1 ,N2
The Laplacian matrix consists of N2 rows of N1 N1 nearly tridiagonal blocks,

E, upper and lower diagonal blocks, K, and two corner blocks implementing the
periodicity condition in the second direction,
(3.3.15) L =
E
KT
0
..
.
K
E
KT
..
.
0
K
E
..
.
..
.
0
0
0
..
.
0
0
0
..
.
KT
0
0
..
.
0
0
K
0
0
0
0
0
0
E
KT
0
K
E
KT
0
K
E
where K is the N1 N1 lower bidiagonal unit matrix with a unit northeastern corner
element, K(1, N1 ) = 1. When N1 = 5, we have
(3.3.16) E =
6
1
0
0
1
1
6
1
0
0
0
1
6
1
0
0
0
1
6
1
1
0
0
1
6
K=
1
1
0
0
0
0
1
1
0
0
0
0
1
1
0
0
0
0
1
1
1
0
0
0
1
The northeastern and southwestern elements of E implement the periodicity condition in the first direction.
An eigenvalue, , of the doubly periodic Laplacian, and the corresponding
eigenvector, u, satisfy the equation
(3.3.17)
6 ui1 , i2 ui1 1, i2 ui1 +1, i2 ui1 , i2 1 ui1 , i2 +1

ui1 1, i2 +1 ui1 +1, i2 1 = ui1 , i2
at any node.
We find that the eigenvalues are given by
(3.3.18) n1 , n2 = 4 sin2 12 n1 + 4 sin2 12 n2 + 4 sin2
1
2

n1 n2
or
(3.3.19) n1 ,n2 = 6 2 cos n1 2 cos n2 2 cos(n1 n2 )
for n1 = 1, . . . , N1 and n2 = 1, . . . , N2 , where

(3.3.20) n1 =
n1 1
2 ,
N1
n2 =
n2 1
2 .
N2
The corresponding eigenvectors, un1 ,n2 , normalized so that their lengths are equal
to unity, un1 ,n2 un1 ,n2 = 1, are

n ,n
(3.3.21) ui11, i2 2 =
exp i (i1 n1 + i2 n2 )
N1 N2
for n1 , i1 = 1, . . . , N1 and n2 , i2 = 1, . . . , N2 , where i is the imaginary unit

and an asterisk denotes the complex conjugate. The eigenvalues given in (3.3.18)
can be derived by substituting (3.3.21) into (3.3.17) and simplifying the resulting
expression.
The doubly periodic Laplacian (3.3.15) is a block circulant matrix. A theorem due
to Friedman [12] states that the spectrum of this matrix is the union of the spectra of
the following N1 N1 circulant matrices,
(3.3.22) L(n2 ) = exp(in2 ) KT + E exp(in2 ) K
or
(3.3.23) L(n2 ) = cos n2 (K + KT ) + E + sin n2 (K KT ),
where n2 = 1, . . . , N2 . When N1 = 4,
6 2a
1 a b
1 a + b
6 2a
(3.3.24) L(n2 ) =
0
1 a + b
1 a b
0
0
1 a b
6 2a
1 a + b
1 a + b
0
,
1 a b
6 2a
where a = cos n2 and b = sin n2 . Using established formulas for the eigenvalues
of circulant matrices (Section A.5, Appendix A), we find that the eigenvalues of the
matrix L(n2 ) are given by
(3.3.25) n1 , n2 = 6 2a (1 + a + b) exp(in1 ) (1 + a b) exp(in1 )
for n1 = 1, . . . , N1 . Simplifying the right-hand side of this expression, we recover the

eigenvalues shown in (3.3.19).
Natural State
With reference to the physical lattice shown in Figure 3.3.1(b), we introduce base
vectors pertaining to the periodic patch,

(3.3.26) A1 = N1 a1 = N1 a 1, 0 ,
A2 = N2 a2 = N2 a 12 1, 3 .
The associated reciprocal base vectors are

2
1
(3.3.27) B1 =
1, ,
N1 a
3
2
B2 =
N2 a

2
0, .
3
The nodes of the reciprocal lattice are

(3.3.28) kn1 , n2 = (n1 1) B1 + (n2 1) B2 ,
where n1 and n2 are two arbitrary integers.

Every link in the natural state is parallel to one of the following three consecutive
link vectors attached to an arbitrary node:
(3.3.29) 1 = a1 ,
2 = a2 ,
3 = a1 a2 .
The eigenvalues of the Laplacian matrix can be expressed in the form

(3.3.30) n1 ,n2 = 6 2
3

cos(kn1 ,n2 r ),
r=1
where
(3.3.31) kn1 , n2 1 = n1 ,
kn1 , n2 2 = n2 ,
kn1 , n2 3 = n1 n2 .
Expression (3.3.30) provides us with a useful mnemonic rule.
3.3.3 Alternative Node Indexing
An alternative node indexing scheme of the hexagonal lattice is shown in Figure 3.3.2(a). The base vectors of the associated network in the natural state, shown
in Figure 3.3.2(b),

(3.3.32) a1 = a 1, 0 ,
a2 = a

1
1, 3 ,
2
form an angle of 120 . The corresponding reciprocal base vectors are

2
1
(3.3.33) b1 =
1, ,
3
a
2
b2 =
a

2
0, .
3
We may readily confirm that a1 b1 = 2 , a2 b2 = 2 , and a1 b2 = 0, as required.

(3.3.34)
ui1 +1, i2 +1 ui1 1, i2 1 = ui1 , i2
at any node, (i1 , i2 ). The Laplacian matrix is given in (3.3.9), except that the matrix
K is the transpose of that described after equation (3.3.9).
(a)
N2
i2
2
1
1
i1
N1
(b)
a2
N2
a1
i2
y
2
x
1
1
2
a
i1
N1
FIGURE 3.3.2 (a) Alternative node indexing of the hexagonal network containing N1 links in the first direction, N2 links in the second direction, and
an appropriate number of cross links. (b) The network has been deformed
to demonstrate the natural state consisting of stacked equilateral triangles.
The angle between the two base vectors, a1 and a2 , is 120 .
FIGURE 3.3.3 Periodic spectral partitioning of the hexagonal

network based on an eigenvector.
The eigenvectors of the Laplacian are given in (3.3.21). The corresponding

eigenvalues are
(3.3.35) n1 , n2 = 4 sin2 12 n1 + 4 sin2 12 n2 + 4 sin2
1
2

n1 + n2
or
(3.3.36) n1 , n2 = 6 2 cos n1 2 cos n2 2 cos(n1 + n2 ).
Note that these expressions differ from those shown in (3.3.19) only by the plus sign
in the argument of the last cosine. In spite of this change in sign, the spectrum of
the Laplacian remains unchanged. A typical spectral partitioning of the hexagonal
network is shown in Figure 3.3.3.
Exercises
3.3.1 Alternative node indexing for an isolated network
(a) Deduce the structure of the Laplacian matrix of the isolated network for the node
indexing scheme described in Figure 3.3.2. (b) Confirm expression (3.3.30).
The particles of a two-dimensional crystal are arranged on a hexagonal lattice.
Small departures from the equilibrium position generate restoring forces. Derive
and solve an algebraic eigenvalue problem for the eigenfrequencies and eigen
displacements [7].
3.4 MODIFIED UNION JACK LATTICE
A rectangular patch of a modified Union Jack lattice containing N1 links in the first
direction, N2 links in the second direction, and two slanted links inside each cell is
N2
i2
2
1
1
N1
i1
FIGURE 3.4.1 Illustration of the modified Union Jack lattice containing N1 links in the first direction, N2 links in the second
direction, and two noncrossing transverse links inside each cell.
shown in Figure 3.4.1. The network nodes are marked as filled circles. Note that the
slanted links do not intersect at a node inside each cell but rather bypass one another.
If they intersected, the modified Union Jack lattice shown in Figure 3.4.1 would
reduce to the regular Union Jack lattice, which is a Laves lattice, as discussed in Section 2.6.3. The coordination number of the modified Union Jack lattice is uniform,
d = 8.
The total number of nodes in an isolated patch is

(3.4.1) N = (N1 + 1)(N2 + 1)

(3.4.2)
L = N1 (N2 + 1) + (N1 + 1)N2 + 2N1 N2 .
As in the case of the rectangular and hexagonal networks discussed previously in

this chapter, the nodes can be labeled in a sequence of layers from the bottom where
i2 = 1 to the top where i2 = N2 + 1.
A nodal scalar field, , is encapsulated in the N-dimensional vector
(1)
(2)
..
.
(3.4.3) =
(N2 )
(N2 +1)
where
(3.4.4) (1)
1,1
2,1
..
.
...,
(N2 +1)
N1 +1,1
1, N2 +1
2, N2 +1
..
.
N1 +1, N2 +1
The Laplacian consists of N2 + 1 rows of (N1 + 1) (N1 + 1) tridiagonal blocks,

F and E, in the following configuration:
(3.4.5) L =
F
JT
0
..
.
J
E
JT
..
.
0
J
E
..
.
..
.
0
0
0
..
.
0
0
0
..
.
0
0
0
0
0
0
0
0
0
E
JT
0
J
E
JT
0
0
0
..
.
J
F
where J is the (N1 + 1) (N1 + 1) tridiagonal unit matrix. For example, when N1 = 4,
F=
(3.4.6)
3
1
0
0
0
1
5
1
0
0
0
1
5
1
0
0
0
1
5
1
0
0
0
1
3
E=
1
1
J=
0
0
0
1
1
1
0
0
0
1
1
1
0
5
1
0
0
0
0
0
1
1
1
0
0
0
1
1
1
8
1
0
0
0
1
8
1
0
0
0
1
8
1
0
0
0
1
5
The three entries of F correspond to corner nodes, the five entries of F and E
correspond to boundary nodes, and the eight entries of E correspond to interior
nodes.
Assume that the nodal scalar field of an infinite modified Union Jack lattice, , is
periodic in two directions, so that
(3.4.7) 1, i2 = N1 +1, i2 ,
i1 ,1 = i1 , N2 +1 .
The vector of unknown nodal values inside each period is
(1)
(2)
..
.
(3.4.8) =
(N2 1)
(N2 )
where
(3.4.9) (1) =
1,1
2,1
..
.
...,
(N2 ) =
N1 ,1
1,N2
2,N2
..
.
N1 ,N2
The vector encapsulates N = N1 N2 unique unknowns.

The Laplacian consists of N2 rows of N1 N1 nearly tridiagonal blocks and two
corner blocks implementing the periodicity condition in the second direction, in the
following configuration:
E
K
0
0
0
KT
KT
E
K
0
0
0
T
0
K
E
0
0
0
..
..
..
..
..
..
..
,
(3.4.10) L =
.
.
.
.
.
.
.
0
0
0
E
K
0
0
0
0
KT
E
K
K
0
0
0
KT
E
where K is the N1 N1 nearly tridiagonal unit matrix implementing the periodicity
condition in the second direction. For example, when N1 = 5,
(3.4.11) E =
8
1
0
0
1
1
8
1
0
0
0
1
8
1
0
0
0
1
8
1
1
0
0
1
8
K=
1
1
0
0
1
1
1
1
0
0
0
1
1
1
0
0
0
1
1
1
1
0
0
1
1
The northeastern and southwestern elements implement the periodicity condition in

the first direction.
(3.4.12)

ui1 1, i2 +1 ui1 +1, i2 1 ui1 +1, i2 +1 ui1 1, i2 1 = ui1 , i2
at any node, (i1 , i2 ). The eigenvalues of the Laplacian are given by

n1 , n2 = 4 sin2 12 n1 + 4 sin2 12 n2 + 4 sin2 12 n1 n2

(3.4.13)

+ 4 sin2 12 n1 + n2 ,
which can be restated as
(3.4.14) n1 , n2 = 8 2 cos n1 2 cos n2 2 cos(n1 + n2 ) 2 cos(n1 n2 )
or
(3.4.15) n1 ,n2 = 8 2 cos n1 2 cos n2 4 cos n1 cos n2 ,
where
(3.4.16) n1 =
n1 1
2 ,
N1
n2 =
n2 1
2
N2
for n1 = 1, . . . , N1 and n2 = 1, . . . , N2 .
to unity, un1 ,n2 un1 ,n2 = 1, are

n ,n
(3.4.17) ui11, i2 2 =
exp i(i1 n1 + i2 n2 )
N1 N2
for n1 , i1 = 1, . . . , N1 and n2 , i2 = 1, . . . , N2 , where i is the imaginary unit and an

asterisk denotes the complex conjugate. Note that the eigenvectors are identical to
those of the square and hexagonal networks.
To derive the eigenvalues, we observe that the doubly periodic Laplacian (3.4.10) is
a block circulant matrix. A theorem due to Friedman [12] states that the spectrum of
this matrix is the union of the spectra of the following N1 N1 circulant matrices:
(3.4.18) L(n2 ) = exp(in2 )KT + E exp(in2 )K
for n2 = 1, . . . , N2 . Rearranging, we obtain

(3.4.19) L(n2 ) = cos n2 (K + KT ) + E.
When N1 = 4, we find that
(3.4.20) L(n2 )
8 2a
1 2a
=
0
1 2a
1 2a
8 2a
1 2a
0
0
1 2a
8 2a
1 2a
1 2a
0
,
1 a
8 2a
where a = cos n2 . Using established expressions for the eigenvalues of circulant matrices (Section A.5, Appendix A), we find that eigenvalues of L(n2 ) are
given by
(3.4.21) n1 ,n2 = 8 2a (1 + 2a) exp(in1 ) (1 + 2a) exp(in1 )
for n1 = 1, . . . , N1 . Simplifying the right-hand side, we recover (3.4.15).
Exercise
3.4.1 Circulant matrices
Derive the eigenvalues of the matrix L(n2 ) shown in (3.4.21).
3.5 HONEYCOMB LATTICE
A patch of a honeycomb network consisting of hexagonal cells with side length a

inscribed in a circle of radius b is shown in Figure 3.5.1(a). The nodes are arranged
on two different Bravais lattices with common base vectors,

a2 = a 12 1, 3 ,
(3.5.1) a1 = a 1, 0 ,
where a =
3 b is the distance of a node from its second nearest neighbor.
Nodes in the first lattice, designated as lattice A, are shown as open circles connected by dashed lines, and nodes in the second lattice, designated as lattice B, are
shown as filled circles connected by dotted lines in Figure 3.5.1(a). Nodes on lattice A are parametrized by a pair of indices, (iA1 , iA2 ), and nodes on lattice B are
parametrized by another pair of indices, (iB2 , iB2 ), where iA1 , iB1 = 1, . . . , N1 + 1 and
iA2 , iB2 = 1, . . . , N2 + 1.
The position of nodes on lattice A is described by
(3.5.2) xiA , iA = xA1,1 + iA1 1 a1 + iA2 1 a2 ,

1 2
and the position of nodes on lattice B is described by
(3.5.3) xiB , iB = xA1,1 + + iB1 1 a1 + iB2 1 a2 ,

1 2
where

(3.5.4)
xB1,1
xA1,1
=a
1
2
1
1,
3
is the inner displacement of the two constituent lattices.
(a)
N2 + 1
N2 + 1
b
N2
N2
iB2
iA
2
a2
1
1
1
1
iA
1
2
a1
iB1
N1
N1
N1 + 1
N1 + 1
(b)
N2+1
2
1
1
2N1
2N1 + 2
FIGURE 3.5.1 (a) Illustration of a honeycomb network containing N1 cells in

the first direction and N2 cells in the second direction. (b) Equivalent isomorphic representation where the network collapses vertically into a brick
wall. In the example shown, N1 = 4 and N2 = 3. Links are drawn with
solid lines.
The numbers of nodes and links in an isolated network are twice those of the
corresponding square network,
(3.5.5) N = 2 (N1 + 1)(N2 + 1)
and
(3.5.6) L = 2 N1 (N2 + 1) + 2 (N1 + 1)N2 .
The nodes of each constituent Bravais lattice can be counted in a sequence of horizontal layers from the bottom where i2 = 1 to the top where i2 = N2 + 1, as in the case
of the square and hexagonal networks. A scalar nodal field, , can be accommodated
into an N-dimensional vector
(1)
(1)
A
B
(2)
(2)
A

.
,
(3.5.7) =
,
A = ..
,
B = ..
(N2 )
2)
(N
A
B
(N2 +1)
(N2 +1)
A
B
where
(1)
(3.5.8) A,B
1,1
2,1
..
.
Nx +1,1
(N +1)
...,
A,B2
1,N2 +1
2,N2 +1
..
.
Nx +1,N2 +1
A,B
A,B
are horizontal profiles of the constituent lattices A or B.

The Laplacian matrix consists of four (N1 + 1) (N2 + 1) square blocks, in the
following configuration:

(3.5.9) L =
B
C
A
BT
where A is a diagonal matrix consisting of N2 + 1 blocks of N1 + 1 elements in the

following order:
(3.5.10) 1, 2, . . . , 2, 2,
2, 3, . . . , 3, 3,
...,
2, 3, . . . , 3, 3,
2, 3, . . . , 3, 3,
and C is another diagonal matrix consisting of N2 + 1 blocks of Nx + 1 elements in

the following order:
(3.5.11) 3, 3, . . . , 3, 2,
3, 3, . . . , 3, 2,
...,
3, 3, . . . , 3, 2,
2, 2, . . . , 2, 1.
Note that the sequence (3.5.11) is the reverse of the sequence (3.5.10). The matrix B
has the block lower bidiagonal form
J
I
0
..
.
(3.5.12) B =
0
0
0
J
I
..
.
0
0
J
..
.
..
.
0
0
0
..
.
0
0
0
..
.
0
0
0
0
0
0
J
I
0
0
J
I
0
0
0
..
.
0
J
where I is the (N1 + 1) (N1 + 1) identity matrix and J is the (N1 + 1) (N1 + 1)
lower bidiagonal unit matrix containing ones along the diagonal and lower diagonal.
When N1 = 4, we obtain the 5 5 matrix
(3.5.13) J =
1
1
0
0
0
0
1
1
0
0
0
0
1
1
0
0
0
0
1
1
0
0
0
0
1
3.5.2 Brick Representation
In the illustration shown in Figure 3.5.1(b), the network displayed in Figure 3.5.1(a)
has been compressed vertically into a brick wall. Dashed lines in Figure 3.5.1(b) pass
through type A nodes marked as hollow circles, and dotted lines pass through type
B nodes marked as filled circles. The nodes are identified by an index i1 that ranges
from 1 to 2N1 + 2 in the first direction and an index i2 that ranges from 1 to N2 + 1 in
the second direction.
The nodes of the brick network can be compiled in a sequence of horizontal
layers from bottom where i1 = 1 to top where i2 = N2 + 1, as in the case of the
rectangular and hexagonal networks. A scalar nodal field, , can be arranged into an
N-dimensional vector
(1)
(2)
..
(3.5.14) = .
,
(N2 )
(N2 +1)
where
(3.5.15) (1)
1,1
2,1
..
.
...,
(N2 +1)
2N1 +2,1
1,N2 +1
2,N2 +1
..
.
2N1 +2,N2 +1
are (2N1 + 2)-dimensional blocks.

The Laplacian matrix of the isolated network consists of N2 + 1 rows of (2N1 +
2) (2N1 + 2) blocks, in the following configuration:
F
K
0
0
0
0
KT
E
K
0
0
0
KT
E
0
0
0
0
..
..
..
..
..
..
..
(3.5.16) L =
.
.
.
.
.
.
. .
0
0
E
K
0
0
0
0
0
KT
E
K
0
0
0
0
KT
G
The tridiagonal blocks, F, E, and G, display the node degrees along the diagonal and
implement horizontal links. For example, when N1 = 1, we obtain the 4 4 blocks
1
1
0
0
2
1
0
0
1
1
3
1
0
3
1
0
,
,
(3.5.17) F =
E=
0
1
2
1
0
1
3
1
0
and
2
1
(3.5.18) G =
0
0
1
3
1
0
0
1
3
1
0
0
.
1
3
The upper diagonal blocks implement upward links from type B nodes, while the
lower diagonal blocks implement downward links from type A nodes. For example,
when N1 = 2, we obtain the 6 6 blocks
(3.5.19) K =
0
1
0
0
0
0
0
0
0
0
0
0
0
0
0
1
0
0
0
0
0
0
0
0
0
0
0
0
0
1
0
0
0
0
0
0
Every other element along the lower diagonal is zero.

The eigenvalues of the Laplacian matrix (3.5.16) are identical to those of the
Laplacian matrix (3.5.9).
Assume that a scalar nodal field, , deployed over an infinite honeycomb lattice is
periodic in the direction of each base vector, so that
A
A
(3.5.20) 1,
i2 = N1 +1, j ,
iA1 ,1 = iA1 , N2 +1
for the constituent lattice A and

B
B
(3.5.21) 1,
i2 = N1 +1, j ,
iB1 ,1 = iB1 , N2 +1
for the constituent lattice B. The vector of unique unknown nodal values inside each
period, encapsulating N = 2N1 N2 unknowns, is

(3.5.22) =
A
B

,
(1)
A
(2)
A
..
.
A =
2 1)
(N
A
(N )
A 2
(1)
B
(2)
B
..
.
B =
2 1)
(N
B
(N )
B 2
where
(1)
(3.5.23) A, B
1,1
2,1
..
.
N 1,1
1
N1 ,1
...,
(N )
A,B2
A, B
1,N2
2,N2
..
.
N 1,N
1
2
Nk ,N2
A, B
for N1 1 and N2 2.
The Laplacian matrix consists of two diagonal blocks hosting the lattice coordination number, along with two off-diagonal square blocks,

(3.5.24) L =
3 IM
BT
B
3 IM

,
where IM is the M M identity matrix and M = N1 N2 is half the number of

unique nodes. The M M matrix B has the following nearly lower bidiagonal block
structure:
K
I
0
..
.
(3.5.25) B =
0
0
0
K
I
..
.
0
0
K
..
.
..
.
0
0
0
..
.
0
0
0
..
.
I
0
0
..
.
0
0
0
0
0
0
K
I
0
0
K
I
0
0
K
where I is the N1 N1 identity matrix and K is the N1 N1 lower bidiagonal

unit matrix, except that the northeastern element is set to unity. For example, when
N1 = 5,
(3.5.26) K =
1
1
0
0
0
0
1
1
0
0
0
0
1
1
0
0
0
0
1
1
1
0
0
0
1
The northeastern block of L implements the periodicity condition in the second direction, and the northeastern element of K implements the periodicity in the first
direction. The sum of the elements of B in each row or column is equal to the lattice
coordination number, 3.

eigenvector, u, satisfy the equations
(3.5.27) 3 uAi1 , i2 uBi1 , i2 uBi1 1, i2 uBi1 , i2 1 = uAi1 , i2
and
(3.5.28) 3 uBi1 , i2 uAi1 , i2 uAi1 +1, i2 uAi1 , i2 +1 = uBi1 , i2 .
From the second equation, we find that

1
(3.5.29) uBi1 , i2 =
uA + uAi1 +1, i2 + uAi1 , i2 +1 .
3 i1 , i2
Substituting this expression into (3.5.27) to eliminate nodal values on lattice B in

favor of those on lattice A, we obtain
(3.5.30)
(3 )2 uAi1 , i2 = 3uAi1 , i2 + uAi1 +1, i2 + uAi1 1, i2 + uAi1 , i2 +1 + uAi1 , i2 1

+ uAi1 1, i2 +1 + uAi1 +1, i2 1 .
Now substituting

(3.5.31) u = exp i(i1 n1 + i2 n2 ) ,
we obtain a quadratic equation,

(3.5.32) ( 3)2 = 3 + 2 cos n1 + 2 cos n2 + 2 cos(n1 n2 ),
where i is the imaginary unit and

(3.5.33) n1 =
n1 1
2 ,
N1
n2 =
n2 1
2 .
N2
The roots of the quadratic equation are given by

(3.5.34)
n1 ,n2 = 3 3 + 2 cos n1 + 2 cos n2 + 2 cos(n1 n2 )
1/2
The sum of two conjugate eigenvalues is

(3.5.35) +n1 , n2 + n1 , n2 = 6.
The product of two conjugate eigenvalues is

(3.5.36) +n1 , n2 n1 , n2 = 6 2 cos n1 2 cos n2 2 cos(n1 n2 ).
Accordingly,
(3.5.37)
1
1
3
+
=
.
+n1 , n2 n1 , n2 3 cos n1 cos n2 cos(n1 n2 )
Eigenvectors
The eigenvectors of the Laplacian matrix L, normalized so that their Euclidean norms
are equal to unity, consist of appropriate arrangements of the following nodal field
on the constituent Bravais lattice A:

n ,n
(3.5.38) (ui11, i2 2 )A =
exp i (i1 n1 + i2 n2 )
2N1 N2
for n1 , i1 = 1, . . . , N1 , and the following nodal field on the constituent Bravais

lattice B:

1
n1 , n2 B
n1 , n2 A
n1 , n2 A
n1 , n2 A
(3.5.39) ui1 , i2
=
ui1 , i2
+ ui1 +1, i2 + ui1 , i2 +1
,
3
n1 , n2
or
(3.5.40)
uni11,,in2 2
B
=

1
n1 , n2 A
i n1
i n2
u
1
+
e
+
e
i1 , i2
3
n1 , n2
for n2 , i2 = 1, . . . , N2 , where i is the imaginary unit. The spectral partitioning of a

periodic honeycomb lattice with N1 = 9 and N2 = 4 is shown in Figure 3.5.2.
Brick Representation
In the illustration shown in Figure 3.5.1(b), the periodic patch of the honeycomb
network displayed in Figure 3.5.1(a) has been compressed vertically into a brick
wall. The periodicity condition requires that
(3.5.41) 1, i2 = N1 +1, i2 ,
2, i2 = N1 +2, i2 ,
i1 , 1 = i1 , N2 +1
for i1 = 1, . . . , 2N1 + 2 and i2 = 1, . . . , N2 + 1. The nodes of the brick network can be

compiled in a sequence of horizontal layers from bottom where i2 = 1, to top where
i2 = N2 , as in the case of the periodic rectangular and hexagonal lattices. A scalar
nodal field, , is encapsulated in an N-dimensional vector
(1)
(2)
..
(3.5.42) = .
,
(N2 1)
(N2 )
where
(3.5.43) (1)
1,1
2,1
..
.
2N1 ,1
are horizontal profiles.
...,
(N2 )
1, N2
2, N2
..
.
2N1 , N2
FIGURE 3.5.2 Spectral partitioning of a periodic honeycomb lattice with N1 = 9 and N2 = 4 divisions inside
each period in the natural state. Positive eigenvector
components are marked with filled circles, negative
components are marked with dots, and zero components are unmarked.
The Laplacian matrix consists of N2 rows of (2N1 ) (2N1 ) blocks,
(3.5.44) L =
E
KT
0
..
.
K
E
KT
..
.
0
K
E
..
.
..
.
0
0
0
..
.
0
0
0
..
.
KT
0
0
..
.
0
0
K
0
0
0
0
0
0
E
KT
0
K
E
KT
0
K
E
The nearly tridiagonal blocks, E, display the lattice coordination number 3 along
the diagonal and implement horizontal links. The upper diagonal blocks implement
upward links originating from type B nodes. The lower diagonal blocks implement
downward links originating from type A nodes. The corner blocks implement the
periodicity condition in the second direction.
Detailed inspection reveals that
A
J
0
0
0
JT
JT
A
J
0
0
0
JT
A
0
0
0
0
.
..
..
..
..
..
..
(3.5.45) E =
.
.
.
.
.
.
..
0
0
0
A
J
0
0
0
0
JT
A
J
J
0
0
0
JT
A
and
J
0
0
..
.
(3.5.46) K =
0
0
0
J
0
..
.
0
0
J
..
.
..
.
0
0
0
..
.
0
0
0
..
.
0
0
0
0
0
0
J
0
0
0
J
0
0
0
0
..
.
0
J
where

(3.5.47) A =
3
1
1
3

,
J=
0
1
0
0

.
For example, when N1 = 2, we obtain the 4 4 blocks
3
1
(3.5.48) E =
0
1
1
3
1
0
0
1
3
1
1
0
,
1
3
0
1
K=
0
0
0
0
0
0
0
0
0
1
0
0
.
0
0
An eigenvalue, , of the doubly periodic Laplacian (3.5.44), and the corresponding eigenvector, u, satisfy the equation
(3.5.49) 3ui1 , i2 ui1 +1, i2 ui1 1, i2 ui1 , i2 +1 = ui1 , i2
at any node. The eigenvalues and eigenvectors are identical to those of the Laplacian
matrix (3.5.24) discussed in the preceding section. An alternative derivation relies on
the block circulant structure of the Laplacian.
The doubly periodic Laplacian (3.5.44) is a block circulant matrix. A theorem due
to Friedman [12] states that the spectrum of this matrix is the union of the spectra of
the following 2N1 2N1 circulant matrices:
(3.5.50) L(n2 ) = exp(in2 ) KT + E exp(in2 ) K
or
(3.5.51) L(n2 ) = cos n2 (K + KT ) + E i sin n2 (K KT ),
where
(3.5.52) n2 =
n2 1
2
N2
for n2 = 1, . . . , N2 . For example, when N1 = 2,
(3.5.53) L(n2 )
3
1 c
=
0
1
1 c
3
1
0
0
1
3
1 c
0
,
1 c
3
where c = exp(in2 ).
More generally,
(3.5.54) L(n2 )
S
JT
0
..
.
J
S
JT
..
.
0
J
S
..
.
..
.
..
.
0
0
0
..
.
JT
0
0
..
.
0
0
J
0
0
0
0
0
0
S
JT
0
J
S
JT
0
J
S
is a block circulant matrix, where

T
(3.5.55) S = cJ + A c J =
3
1 c
1 c
3

.
The spectrum of L(n2 ) is the union of the spectra of the following 2 2 Hermitian
matrices:
(3.5.56) (n1 , n2 ) = exp(in1 ) JT + S exp(in1 ) J
where
(3.5.57) n1 =
n1 1
2
N1
for n1 = 1, . . . , N1 . Explicitly,

(3.5.58)
(n1 , n2 )
3
1 (c + d)
1 (c + d)
3

,
where d = exp(in1 ). The eigenvalues are the roots of the characteristic polynomial
of (n1 ,n2 ) satisfying the quadratic equation (3.5.32), given in (3.5.38). The sum of
FIGURE 3.5.3 Spectral partitioning of a periodic brick (honeycomb) lattice with N1 = 9 and N2 = 4 divisions
inside each period. Positive eigenvector components
are marked with filled circles, negative components are
two conjugate eigenvalues is the trace of (n1 ,n2 ) , that is, the sum of the diagonal
elements. The product of two conjugate eigenvalues is the determinant of (n1 ,n2 ) .
The spectral partitioning of a periodic brick lattice with N1 = 9 and N2 = 4 is shown
in Figure 3.5.3.
3.5.4 Alternative Node Indexing
In an alternative representation, the nodes of the honeycomb lattice are identified by

two indices, i1 and i2 , corresponding to two base vectors, a1 and a2 , that form an
angle of 120 , as shown in Figure 3.5.4(a), given by
(3.5.59) a1 = a (1, 0),
a2 = a 12 (1,
3).
Nodes on lattice A are drawn as open circles connected by dashed lines, and nodes
on lattice B are drawn as filled circles connected by dotted lines. The position of
nodes at lattice A is
(3.5.60) xiA , iA = xA1,1 + (iA1 1) a1 + (iA2 1) a2 ,
1 2
and the position of nodes at lattice B is

(3.5.61) xiB , iB = xA1,1 + + (iB1 1) a1 + (iB2 1) a2 ,
1 2
where
(3.5.62)
xB1,1
xA1,1

1
1
=a
1,
2
3
is the inner displacement of the two constituent lattices.

The Laplacian matrix is given by (3.5.24)(3.5.26), provided that the matrix K
defined in (3.5.26) is replaced by its transpose. The eigenvalues of the new Laplacian
are

(3.5.63)
n1 ,n2 = 3 3 + 2 cos n1 + 2 cos n2 + 2 cos(n1 + n2 )
1/2
The eigenvectors on lattice A are given in (3.5.38) and the eigenvectors on lattice B
are given by

1
n1 , n2 B
n1 , n2 A
n1 , n2 A
n1 , n2 A
(3.5.64) ui1 , i2
=
ui1 , i2
+ ui1 1, i2 + ui1 , i2 +1
3
n1 , n2
or
(3.5.65)
uni11,,in2 2
B
=

1
n1 , n2 A
u
1 + ei n1 + ei n2 .
i1 , i2
3 n1 , n2
(a)
N2 + 1
N2 + 1
b
N2
N2
iB2
y
a2
iA
2
1
1
1
N1
iB1
2
1
iA
1
a1
N1 + 1
N1
N1 + 1
(b)
N2 + 1
2
1
1
2N1
2N1 + 2
FIGURE 3.5.4 (a) Alternative indexing of a honeycomb network containing

N1 cells in the first direction and N2 cells in the second direction. (b) Alternative representation where the network collapses vertically into a brick
wall. For the configuration shown, N1 = 4 and N2 = 3. Nodes on lattice
A are shown as open circles connected by dashed lines, and nodes on
lattice B are shown as filled circles connected by dotted lines.
The equivalent brick representation is shown in Figure 3.5.4(b). The Laplacian

matrix is given in (3.5.44), provided that the matrix K is replaced by the transpose of
than shown in (3.5.46).
Exercise
Assume that the particles of a two-dimensional crystal are arranged on a honeycomb lattice. Small departures from the equilibrium position generate restoring
forces. Derive and solve an algebraic eigenvalue problem for the eigenfrequencies
and eigendisplacements [6, 7].
3.6 KAGOM LATTICE
A rectangular patch of a kagom lattice is shown in Figure 3.6.1(a). Although the

lattice coordination number is identical to that of the square lattice, d = 4, the
two lattices are distinct. Three families of nodes falling on different Bravais lattices,
(a)
N2+1
N2+1
N2
N2
i2
i2
B
A
2
2
1
1
1 1
2 2
i1 i1
N +1
N1 N1 N +1 1
1
(b)
y
C
x
a2
a1
FIGURE 3.6.1 (a) Illustration of a kagom network involving three families

of nodes. (b) The network has been sheared into its natural state where all
links have the same length. Links are drawn as bold lines. Type A nodes
are shown as hollow circles, type B nodes are shown as solid circles,
and type C nodes are shown as solid squares.
identified as families A, B, and C, are shown with hollow circles, filled circles, or
filled squares in Figure 3.6.1(a). Each family is parametrized by a pair of indices,
i1 and i2 , where i1 = 1, . . . , N1 + 1, i2 = 1, . . . , N2 + 1, and N1 , N2 are the patch
dimensions. Type A nodes lie at the intersection of vertical and horizontal solid lines,
type B nodes lie at the intersection of solid and dotted lines, and type C nodes lie at
the intersection of solid and dashed lines.
The corresponding natural state of the network is illustrated in Figure 3.1.1(b).
Each family of nodes falls on a Bravais lattice with two base vectors a1 and a2 that
are identical to those of the hexagonal lattice.
The total number of nodes in the isolated network shown in Figure 3.6.1 is
(3.6.1) N = 3(N1 + 1)(N2 + 1)

(3.6.2) L = 6N1 N2 + 2N1 + 2N2 + 1.
The nodes are compiled in a sequence of horizontal layers from the bottom where
i2 = 1 to the top where i2 = N2 + 1. A nodal field, , can be arranged in an Ndimensional vector
(1)
(2)
(3.6.3) = ..
,
(N
)
2
(N2 +1)
where the constituent vectors
A
1,1
B
1,1
C
. 1,1
(1)
(3.6.4)
..
NA1 +1,1
NB +1,1
1
NC1 +1,1
...,
(N2 +1)
A
1,N
2 +1
B
1,N
2 +1
C
1,N
2 +1
..
.
NA1 +1,N2 +1
NB +1,N +1
1
2
NC1 +1,N2 +1
consist of ordered triplets of A, B, C nodes.

The Laplacian is a block tridiagonal matrix consisting of N2 + 1 rows of 3(N1 +
1) 3(N1 + 1) symmetric tridiagonal blocks, F, E, and G, along with sparse lower
and upper diagonal blocks, in the following configuration:
F
J
0
0
0
0
JT
E
J
0
JT
E
0
0
0
..
..
..
..
..
..
..
(3.6.5) L =
.
.
.
.
.
.
. ,
0
0
0
E
J
0
0
0
0
...
JT
E
J
0
0
0
0
JT
G
where J is a 3(N1 + 1) 3(N1 + 1) sparse matrix.
For example, when N1 = 2, we obtain the 9 9 matrices
2
1
1
0
0
0
0
1
3
1
1
0
0
0
1
1
3
0
0
0
0
0
1
0
3
1
1
0
(3.6.6) F = 0
0
0
1
3
1
1
0
0
0
1
1
4
0
0
0
0
0
1
0
3
0
0
0
0
0
0
1
0
0
0
0
0
0
1
0
0
0
0
0
0
1
2
1
0
0
0
0
0
0
1
1
4
(3.6.7) E =
(3.6.8) G =
3
1
1
0
0
0
0
0
0
1
4
1
1
0
0
0
0
0
1
1
3
0
0
0
0
0
0
0
1
0
4
1
1
0
0
0
0
0
0
1
4
1
1
0
0
0
0
0
1
1
4
0
0
0
0
0
0
0
1
0
4
1
1
0
0
0
0
0
0
1
2
1
0
0
0
0
0
0
1
1
4
3
1
1
0
0
0
0
0
0
1
4
1
1
0
0
0
0
0
1
1
2
0
0
0
0
0
0
0
1
0
4
1
1
0
0
0
0
0
0
1
4
1
1
0
0
0
0
0
1
1
2
0
0
0
0
0
0
0
1
0
4
1
1
0
0
0
0
0
0
1
2
1
0
0
0
0
0
0
1
1
2
(3.6.9) J =
0
0
1
0
0
0
0
0
0
0
0
0
0
0
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
0
0
0
0
0
0
0
0
0
0
0
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
and
T
(3.6.10) J =
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
0
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
0
0
0
0
0
0
0
0
0
0
1
0
1
0
0
The two and three entries along the diagonal lines of F, E, and G, correspond to edge
and corner nodes. The unit elements of J correspond to upward links from type C
to type A and B nodes. The nonzero elements of JT correspond to downward links
from type A or B to type C nodes.
When N1 = 2 and N2 = 1, the first component of the product L
corresponding to the (1, 1)A node reads
A
B
C
1 = 2 1,1
1 1,1
1 1,1
(3.6.11)
A
B
C
A
B
C
+ 0 2,1
+ 0 2,1
+ 0 2,1
+ 0 3,1
+ 0 3,1
+ 0 3,1
A
B
C
A
B
C
+ 0 1,2
+ 0 1,2
+ 0 1,2
+ 0 2,2
+ 0 2,2
+ 0 2,2
A
B
C
+ 0 3,2
+ 0 3,2
+ 0 3,2
,
the second component corresponding to the (1, 1)B node reads

A
B
C
2 = 1 1,1
3 1,1
1 1,1
(3.6.12)
A
B
C
A
B
C
1 2,1
+ 0 2,1
+ 0 2,1
+ 0 3,1
+ 0 3,1
+ 0 3,1
A
B
C
A
B
C
+ 0 1,2
+ 0 1,2
+ 0 1,2
+ 0 2,2
+ 0 2,2
+ 0 2,2
A
B
C
+ 0 3,2
+ 0 3,2
+ 0 3,2
,
and the third component corresponding to the (1, 1)C node reads
A
B
C
3 = 1 1,1
1 1,1
3 1,1
(3.6.13)
A
B
C
A
B
C
+ 0 2,1
+ 0 2,1
+ 0 2,1
+ 0 3,1
+ 0 3,1
+ 0 3,1
A
B
C
A
B
C
1 1,2
+ 0 1,2
+ 0 1,2
+ 0 2,2
+ 0 2,2
+ 0 2,2
A
B
C
+ 0 3,2
+ 0 3,2
+ 0 3,2
.
The coefficients are consistent with the entries of the matrices (3.6.6)(3.6.10).
Assume that the nodal scalar field of an infinite kagom network, , is periodic in
the direction of each base vector, so that
A
A
(3.6.14) 1,
i2 = N1 +1, j ,
iA1 , 1 = iA1 , N2 +1
for the constituent lattice A

B
B
(3.6.15) 1,
i2 = N1 +1, j ,
iA1 , 1 = iB1 , N2 +1
for the constituent lattice B, and

C
(3.6.16) 1,i
= NC1 +1, j ,
2
iC1 ,1 = iC1 , N2 +1
for the constituent lattice C. The nodal field, , can be accommodated in a vector
incorporating 3N1 N2 unknowns,
(1)
(2)
..
.
(3.6.17) =
(N2 1)
(N2 )
where
(3.6.18) (1)
A
1,1
B
1,1
C
1,1
..
.
NA1 ,1
NB ,1
1
NC1 ,1
...,
(N2 )
A
1,N
2
B
1,N
2
C
1,N
2
..
.
NA1 ,N2
NB ,N
1 2
NC1 ,N2
Subject to these definitions, the Laplacian takes the form of a nearly tridiagonal
block circulant matrix consisting of N2 rows of 3N1 3N1 blocks, in the following
configuration:
E
J
0
0
0
JT
JT
E
J
0
0
0
JT
E
0
0
0
0
.
..
..
..
..
..
..
.
(3.6.19) L =
.
.
.
.
.
.
. .
0
0
0
E
J
0
0
0
0
JT
E
J
J
0
0
0
JT
E
The nearly tridiagonal blocks, E, display the lattice coordination number 4 along
the diagonal. The northeastern and southwestern corner blocks implement the
periodicity condition in the second direction.
The matrix E takes the block circulant form
A
B
0
0
0
BT
BT
A
B
0
0
0
BT
A
0
0
0
0
..
..
..
..
..
..
..
,
(3.6.20) E =
.
.
.
.
.
.
.
0
0
A
B
0
0
0
0
0
BT
A
B
B
0
0
0
BT
A
where
(3.6.21) A =
1
1
1
1 ,
4
1
4
1
B=
1
0
0
0 .
0
0
0
0
The matrix J takes the lower bidiagonal block circulant form
C
0
0
0
0
DT
C
0
0
0
DT
C
0
0
0
.
.
.
.
..
.
..
..
..
..
(3.6.22) J =
.
..
0
0
0
C
0
0
0
0
DT
C
0
0
0
0
DT
DT
0
0
..
.
0
0
C
where
(3.6.23) C =
0
1
0
0 ,
0
0
0
0
D=
0
0
0
0
0
0
1 ,
0
For example, when N1 = 3, we obtain the 9 9 matrices
(3.6.24) E =
4
1
1
0
0
0
0
1
0
1
4
1
1
0
0
0
0
0
1
1
4
0
0
0
0
0
0
0
1
0
4
1
1
0
0
0
0
0
0
1
4
1
1
0
0
0
0
0
0
0
1
0
0
0
0
0
0
0
0
0
0
0
1
0
0
0
1
1
4
0
0
0
0
0
0
0
1
0
4
1
1
1
0
0
0
0
0
1
4
1
0
0
0
0
0
0
1
1
4
and
(3.6.25) J =
0
0
1
0
0
0
0
0
0
0
0
0
0
0
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
0
0
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
An eigenvalue of the doubly periodic Laplacian, , and the corresponding

eigenvector, u, satisfy the equations
(3.6.26)
C
(4 )uAi1 ,i2 uBi1 1,i2 uBi1 ,i2 uC
i1 ,i2 ui1 ,i2 1
= 0,
(4 )uBi1 ,i2
(4 )uC
i1 ,i2
= 0,
C
A
uC
i1 +1,i2 1 ui1 ,i2 ui1 +1,i2
uAi1 ,i2 uAi1 ,i2 +1 uBi1 1,i2 +1
uAi1 ,i2
uBi1 ,i2
= 0.
The eigenvalues can be calculated by eliminating the lattice B and C nodes in favor
of the lattice A nodes, and then setting

(3.6.27) u = exp i(i1 n1 + i2 n2 ) ,
where i is the imaginary unit and

(3.6.28) n1 =
n1 1
2 ,
N1
n2 =
n2 1
2 .
N2
Simplifying, we derive a cubic equation.

In an essentially equivalent approach, we note that the doubly periodic Laplacian (3.6.19) is a block circulant matrix. A theorem due to Friedman [12] states that
the spectrum of this matrix is the union of the spectra of the following 3N1 3N1
circulant matrices:
(3.6.29) L(n2 ) = exp(in2 )JT + E exp(in2 ) J
or
(3.6.30) L(n2 ) = cos n2 (J + JT ) + E i sin n2 (J JT ),
where
(3.6.31) n2 =
n2 1
2
N2
for n2 = 1, . . . , N2 . Making substitutions, we find that
(3.6.32) L(n2 )
P
QA
0
..
.
Q
P
QA
..
.
0
Q
P
..
.
..
.
0
0
0
..
.
0
0
0
..
.
QA
0
0
..
.
0
0
Q
0
0
0
0
0
0
P
QA
0
Q
P
QA
0
Q
P
where the superscript A denoted the matrix adjoint defined as the complex conjugate
of the transpose,
4
(3.6.33) P = c CT + A c C =
1
1 c
0
(3.6.34) Q = c D + B = 1
0
1
4
1
0
0
0
1 c
,
1
4
0
c ,
0
and c exp(in2 ). The spectrum of L(n2 ) is the union of the spectra of the following
3 3 Hermitian matrices:
(3.6.35) (n1 , n2 ) = exp(in1 ) QA + P exp(in1 )Q,
where
(3.6.36) n1 =
n1 1
2
N1
for n1 = 1, . . . , N1 , and the superscript A denotes the matrix adjoint. Explicitly,
(3.6.37) (n1 ,n2 )
=
1 d
1 c
1 d
4
1 c d
1 c
1 cd ,
4
where d = exp(in1 ).
The trace of (n1 ,n2 ) is

(3.6.38) T trace (n1 ,n2 ) = n1 ,n2 + +n1 ,n2 + n1 ,n2 = 12,
where n1 , n2 , +n1 , n2 , and n1 , n2 are the three eigenvalues.

The determinant of (n1 , n2 ) is

(3.6.39) D det (n1 , n2 ) = n1 , n2 +n1 , n2 n1 , n2 .
Performing the calculations, we find that

(3.6.40) D = 36 12 cos n1 12 cos n2 12 cos(n1 n2 ).
The negative of the characteristic polynomial of (n1 , n2 ) is

(3.6.41) Pn1 , n2 () det I (n1 , n2 ) = ( n1 ,n2 )( +n1 , n2 )( n1 , n2 ).
Carrying out the multiplications, we obtain

(3.6.42) Pn1 , n2 () = 3 T 2 + E D ,
where I is the 3 3 identity matrix and

(3.6.43) E = n1 , n2 +n1 , n2 + +n1 , n2 n1 , n2 + n1 , n2 n1 , n2 .
We find that

E = det

(3.6.44)
4
1 d
+ det
1 d
4

+ det
1 cd
4
4
1 c d

4
1 c
1 c
4

Computing the three 2 2 determinants and consolidating the sum, we find that
(3.6.45) E = 42 2 cos n1 2 cos n2 2 cos(n1 n2 ).
Accordingly,
(3.6.46)
1
n1 , n2
1
+n1 , n2
1
n1 , n2
E
21 cos n1 cos n2 cos(n1 n2 )
=
.
D 18 6 cos n1 6 cos n2 6 cos(n1 n2 )
For convenience, we denote

a = T = 12,
(3.6.47) b = E = 42 2 cos n1 2 cos n2 2 cos(n1 n2 ),
c = D = 36 + 12 cos n1 + 12 cos n2 + 12 cos(n1 n2 ).
The roots of the characteristic polynomial can be found using Cardanos formula,
yielding
(3.6.48) n1 , n2 = + d cos ,
3
3
,
n1 , n2 = d cos
3
3
where
(3.6.49) d = 2
1
3
|p|
1/2
= arccos

q
,
2 (|p|/3)3/2
and

1 2
a = 2 23 + cos n1 + cos n2 + cos(n1 n2 ) ,
3
(3.6.50)

2 3 1
q=c+
a ab = 4 1 + cos n1 + cos n2 + cos(n1 n2 ) .
27
3
p=b
As an example, the spectral partitioning of a kagom network with N1 = 6 and

N2 = 5 divisions is shown in Figure 3.6.2.
Exercise
Assume that the particles of a two-dimensional crystal are arranged on a kagom
lattice. Small departures from the equilibrium position generate restoring forces.
Derive and solve an algebraic eigenvalue problem for the eigen-frequencies and
eigendisplacements [7].
FIGURE 3.6.2 Spectral partitioning of a periodic kagom lattice with N1 =

6 and N2 = 5 divisions inside each period in the natural state. Positive
eigenvector components are marked with filled circles, negative components are marked with dots, and zero components are unmarked.
3.7 SIMPLE CUBIC LATTICE
Our analysis for the square lattice in Section 3.1 can be extended directly to the
simple cubic lattice associated with a Cartesian grid with N1 , N2 , and N3 divisions,
as shown in Figure 3.7.1. The coordination number of the simple cubic lattice is
d = 6.
Isolated Network
The number of nodes in an isolated network is

(3.7.1) N = (N1 + 1)(N2 + 1)(N3 + 1)
and the number of links is

(3.7.2) L = N1 N2 (N3 + 1) + N1 (N2 + 1)N3 + (N1 + 1)N2 N3 .
(3.7.3) n1 , n1 , n3 = 4 sin2 12 n1 + 4 sin2 12 n2 + 4 sin2 12 n3
or
(3.7.4) n1 , n1 , n3 = 6 2 cos n1 2 cos n2 2 cos n3 ,
N2+1
i2
1 i1
1
i3
N1+1
N3+1
FIGURE 3.7.1 Illustration of a rectangular slab of

a simple cubic network containing N1 links in
the first direction, N2 links in the second direction, and N3 links in the third direction. All links
are assumed to have the same conductance. In
the configuration shown, N1 = 2, N2 = 2, and
N3 = 1.
where
(3.7.5) n1 =
n1 1
,
N1 + 1
n2 =
n2 1
,
N2 + 1
n3 =
n3 1
N3 + 1
for n1 = 1, . . . , N1 + 1, n2 = 1, . . . , N2 + 1, and n3 = 1, . . . , N3 + 1.
The corresponding eigenvectors, un1 , n2 , n3 , normalized so that their lengths are
equal to unity, un1 , n2 , n3 un1 , n2 , n3 = 1, are
23/2
n ,n ,n
ui11, i2 2, i3 3 = An1 Bn2 Cn3
(N1 + 1)(N2 + 1)(N3 + 1)
(3.7.6)

cos i1 12 n1 cos i2 12 n2 cos i3 12 n3
for n1 , i1 = 1, . . . , N1 + 1, n2 , i2 = 1, . . . , N2 + 1,
and n3 , i3 =1, . . . , N3 + 1,
where
An1 = 1, Bn2 = 1, and Cn3 = 1, except that A1 = 1/ 2, B1 = 1/ 2, and C1 = 1/ 2.
Triply Periodic Network
The eigenvalues of the triply periodic Laplacian matrix are
(3.7.7) n1 , n2 , n3 = 4 sin2 12 n1 + 4 sin2 12 n2 + 4 sin2 12 n3
or
(3.7.8) n1 , n2 , n3 = 6 2 cos n1 2 cos n2 6 cos n3 ,
where
(3.7.9) n1 =
n1 1
2 ,
N1
n2 =
n2 1
2 ,
N2
n3 =
n3 1
2
N3
for n1 = 1, . . . , N1 , n2 = 1, . . . , N2 , and n3 = 1, . . . , N3 . We can write

(3.7.10) n1 = (n1 1)k1 ,
n2 = (n2 1)k2 ,
n3 = (n3 1)k3 ,
where the parameters

(3.7.11) k1 =
2
,
N1
k2 =
2
,
N2
k3 =
2
N3

The corresponding eigenvectors, normalized so that their lengths are equal to
unity, un1 ,n2 ,n3 un1 ,n2 ,n3 = 1, are

n ,n ,n
(3.7.12) ui11, i2 2, i3 3 =
exp i(i1 n1 + i2 n2 + i3 n3 )
N1 N2 N3
for n1 , i1 = 1, . . . , N1 , n2 , i2 = 1, . . . , N2 , and n3 , i3 = 1, . . . , N3 , where i is the

imaginary unit and an asterisk denotes the complex conjugate.
Exercise
3.7.1 Periodic cubic lattice
Derive the eigenvalues and eigenvectors of the simple cubic lattice subject to (a) the
periodicity condition in the first direction and (b) the periodicity condition in the first
and second directions.
3.8 BODY-CENTERED CUBIC (BCC) LATTICE
The nodes of the body-centered cubic (bcc) lattice can be parametrized by three
indices, i1 , i2 , and i3 , as shown in Figure 3.8.1. The lattice coordination number is
d = 8. In the Cartesian coordinates defined in Figure 3.8.1, the base vectors of the
associated Bravais lattice are
(3.8.1)
a1 = a 12 (ex + ey + ez ),
a2 = a 12 (ex ey + ez ),
a3 = a 12 (ex + ey ez ),
where ex , ey , and ex are unit vectors along the x, y, and z axes, respectively. The
reciprocal lattice base vectors are
(3.8.2)
b1 =
2
a
(ey + ez ),
b3 =
2
a
b2 =
2
a
(ez + ex ),
(ex + ey ).
The reciprocal lattice of the bcc lattice defines the face-centered cubic (fcc) lattice
discussed in Section 3.9.
i1
i3
y
x
i2
FIGURE 3.8.1 Node indexing of the body-centered cubic (bcc)

network in terms of three indices, i1 , i2 , and i3 .
The eigenvectors of the triply periodic Laplacian matrix, normalized so that their
lengths are equal to unity, un1 , n2 , n3 un1 , n2 , n3 = 1, are

n ,n ,n
(3.8.3) ui11, i2 2, i3 3 =
exp i (i1 n1 + i2 n2 + i3 n3 )
N1 N2 N3

The equation defining the eigenvalues, , and associated eigenvectors, u, specifies that
(3.8.4)
8 ui1 , i2 , i3 ui1 1, i2 , i3 ui1 +1, i2 , i3 ui1 , i2 1, i3 ui1 , i2 +1, i3

ui1 , i2 , i3 1 ui1 , i2 , i3 +1 ui1 +1, i2 +1, i3 +1 ui1 1, i2 1, i3 1 = ui1 , i2 , i3 .
Substituting the expression given in (3.8.3) and simplifying, we obtain

n1 , n2 , n3 = 8 2 cos n1 2 cos n2 2 cos n3
(3.8.5)
2 cos(n1 + n2 + n3 ),
where
(3.8.6) n1 =
n1 1
2 ,
N1
n2 =
n2 1
2 ,
N2
n3 =
n3 1
2 .
N3
It is useful to introduce three new variables, 1 , 2 , and 3 , defined such that

(3.8.7)
n1 = 1 + 2 + 3 ,
n2 = 1 2 + 3 ,
n3 = 1 + 2 3 .
Conversely,
(3.8.8) 1 = 12 (n2 + n3 ),
2 =
1
2
(n3 + n1 ),
3 =
1
2
(n1 + n2 ).
Substituting expressions (3.8.7) into (3.8.5) and simplifying, we obtain

(3.8.9) n1 , n2 , n3 = 8 1 cos 1 cos 2 cos 3 .
The eigenvectors given in (3.8.3) can be expressed in the form

, , 3
(3.8.10) ui 1, i ,2i

1 2 3

1
exp i (i1 1 + i2 2 + i3 3 ) ,
N1 N2 N3
where
(3.8.11) i1 = i1 + i2 + i3 ,
i2 = i1 i2 + i3 ,
i3 = i1 + i2 i3 .
Conversely,
(3.8.12) i1 = 12 (i2 + i3 ),
i2 =
1
2
(i3 + i1 ),
i3 =
1
2
(i1 + i2 ).
Note that the indices i1 , i2 , and i3 , are not independent. For example, if i1 is odd or
even, then i2 is also odd or even.
The relative position of the nodes in physical space is
(3.8.13) xi1 , i2 , i3 x0,0,0 = i1 a1 + i2 a2 + i3 a3 = i1 a1 + i2 a2 + i3 a3 ,
where
a1 =
(3.8.14)
1
2
1
2
(a2 + a3 ) =
a3 =
a2 =
1
2
(a3 + a1 ) =
(a1 + a2 ) =
1
2
a ez
a ex ,
1
2
1
2
a ey ,
are Cartesian base vectors associated with the primed indices.

When N1 = N2 = N3 = N , we obtain
m1 1
2 ,
N
2 =
m2 1
2 ,
N
3 =
(3.8.16) m1 = 12 (n2 + n3 ),
m2 =
1
2
m3 =
(3.8.15) 1 =
m3 1
2 ,
N
where
(n3 + n1 ),
1
2
(n1 + n2 ).
Exercise
3.8.1 Base vectors
Confirm that the base vectors shown in (3.8.2) are the reciprocal of those shown in
(3.8.1).
3.9 FACE-CENTERED CUBIC (FCC) LATTICE
The nodes of the face-centered cubic (fcc) lattice can be parametrized by three indices, i1 , i2 , and i3 , as shown in Figure 3.9.1. The lattice coordination number is
d = 12. In the Cartesian coordinates defined in Figure 3.9.1, the base vectors of the
associated Bravais lattice are
(3.9.1)
a1 = a 12 (ey + ez ),
a2 = a 12 (ez + ex ),
a3 = a 12 (ex + ey ),
i3
y
i1
x
z
i2
a
FIGURE 3.9.1 Node indexing of the face-centered cubic (fcc) network in

terms of three indices, i1 , i2 , and i3 . Links are drawn as solid lines.
Nodes are located at the intersection of two dashed lines or two dotted
lines. The indices, i1 , i2 , and i3 vary in the directions of the three base
vectors.
where ex , ey , and ez , are unit vectors along the x, y, and z axes, respectively. The
reciprocal lattice base vectors are
b1 =
(3.9.2)
2
2
(ex + ey + ez ),
b2 =
(ex ey + ez ),
a
a
2
b3 =
(ex + ey ez ).
a
The reciprocal lattice defines the body-centered cubic (bcc) lattice discussed in
Section 3.8.
The eigenvectors of the triply periodic Laplacian matrix, normalized so that their
lengths are equal to unity, un1 ,n2 ,n3 un1 ,n2 ,n3 = 1, are
n ,n ,n
(3.9.3) ui11,i2 2,i3 3 =

exp i(i1 n1 + i2 n2 + i3 n3 )
N1 N2 N3

The equation defining the eigenvalues, , and associated eigenvectors, u, specifies that
12 ui1 , i2 , i3 ui1 1, i2 , i3 ui1 +1, i2 , i3 ui1 , i2 1, i3 ui1 , i2 +1, i3
(3.9.4)
ui1 , i2 +1, i3 1 ui1 , i2 1, i3 +1 ui1 1, i2 , i3 +1 ui1 +1, i2 , i3 1
ui1 +1, i2 1, i3 +1 ui1 1, i2 +1, i3 1 = ui1 , i2 , i3 .
Substituting the eigenvectors given in (3.9.3) and simplifying, we obtain

n1 , n2 , n3 = 12 2 cos n1 2 cos n2 2 cos n3
(3.9.5)
2 cos(n1 n2 ) 2 cos(n2 n3 ) 2 cos(n3 n1 ),
where
(3.9.6) n1 =
n1 1
2 ,
N1
n2 =
n2 1
2 ,
N2
n3 =
n3 1
2 .
N3
It is useful to introduce three new variables, 1 , 2 , and 3 , such that

(3.9.7) n1 = 2 + 3 ,
n2 = 3 + 1 ,
n3 = 1 + 2 .
Conversely,
(3.9.8)
1 =
1
2
(n1 + n2 + n3 ),
3 =
1
2
2 =
1
2
(n1 n2 + n3 ),
(n1 n2 n3 ).
Substituting expressions (3.9.7) into (3.9.5) and simplifying, we obtain

(3.9.9) n1 , n2 , n3 = 4 3 cos 1 cos 2 cos 2 cos 3 cos 3 cos 1 .
The eigenvectors given in (3.9.3) can be expressed in the form

, , 3
(3.9.10) ui 1, i ,2i

1 2 3

1
exp i (i1 1 + i2 2 + i3 3 ) ,
N1 N2 N3
where
(3.9.11) i1 = i2 + i3 ,
i2 = i3 + i1 ,
i3 = i1 + i2 .
Conversely,
(3.9.12)
i1 =
1
2
(i1 + i2 + i3 ),

i3 =
1
2
i2 =
1
2
(i1 i2 + i3 ),
(i1 + i2 i3 ).
Note that the indices i1 , i2 , and i3 , are not independent. For example, if i2 = 0 and
i3 = 0, the index i1 is even.
The distance of a node from a designated zero node in physical space is
(3.9.13) xi1 , i2 , i3 x0,0,0 = i1 a1 + i2 a2 + i3 a3 = i1 a1 + i2 a2 + i3 a3 ,
where
(3.9.14)
a1 =
1
2
a3 =
a2 =
1
2
(a1 a2 + a3 ) =
(a1 + a2 a3 ) =
1
2
a ez
1
2
(a1 + a2 + a3 ) =
1
2
a ex ,
1
2
a ey ,
are Cartesian base vectors associated with the primed indices.

When N1 = N2 = N3 = N , we obtain
(3.9.15) 1 =
m1 1
2 ,
N
2 =
m2 1
2 ,
N
3 =
m3 1
2 ,
N
where
(3.9.16)
m1 = 12 (n1 + n1 + n2 ),
m3 =
1
2
m2 =
1
2
(n1 n2 + n3 ),
(n1 + n2 n3 ).
Exercise
3.9.1 Base vectors
Confirm that the base vectors shown in (3.9.2) are the reciprocal of those shown in
(3.9.1).
/// 4 ///
NETWORK TRANSPORT
In science, engineering, biological, and other applications, a

graph describes a physical or abstract, conductive, convective, or mechanical network (e.g, [32]). Heat, electricity, mass, or any other suitable transported entity can
be supplied, generated, or consumed at the nodes. The rate of a transported entity
through a link is typically determined by a driving potential according to a convective
or conductive law involving the link conductance. Introducing an appropriate transport law provides us with a complete set of governing equations that determines the
operational state of the network. The basic concepts involved and the pertinent mathematical framework are discussed in this chapter with emphasis on linear networks
operating at steady state.
4.1 TRANSPORT LAWS AND CONVENTIONS
Consider a transported entity, such as heat, associated with a scalar field, , such a
temperature, over an arbitrary network, as shown in Figure 4.1.1. If is electrical
voltage, the transported quantity is electricity through an electrical grid. If is pressure, the transported quantity is volume or mass of a transported gas or liquid along a
pipeline. Other abstract scalar fields pertinent, for example, to information exchange
are possible.
4.1.1 Isolated and Embedded Networks
Selected nodes of a network can be connected to external nodes where the potential,
, is held at a specified value in lieu of a Dirichlet boundary condition, as shown in
Figure 4.1.1. For convenience, these external nodes will be called Dirichlet nodes.
It is important to note that Dirichlet nodes are included neither in the network
configuration nor in the graph describing the network, but are regarded as exterior
anchor points. In the absence of Dirichlet nodes, we obtain an isolated network. If at
least one Dirichlet node is present, we obtain an embedded network.
130
N e t w o r k Tr a n s p o r t / / 131
7
11
8
9
6
Dirichlet node
10
12
8
5
4
4
3
2
2
Dirichlet node
Dirichlet node
3
FIGURE 4.1.1 Illustration of a conducting network consisting of N = 8

nodes connected by L = 12 links. Three selected nodes of this network, labeled 2, 3, and 6, are connected to external Dirichlet nodes
where the potential associated with a transported entity is held at
a constant value.
Embedding Matrix
It will be convenient to introduce an N N diagonal matrix, J, called the embedding

matrix, that is filled with zeros, except that Jii = 1 if the ith node of the network is
connected to a peripheral Dirichlet node for i = 1, . . . , N. For the network shown in
Figure 4.1.1 where nodes 2, 3, and 6 are connected to Dirichlet nodes,
(4.1.1) J =
0
0
0
0
0
0
0
0
0
1
0
0
0
0
0
0
0
0
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
In the case of an isolated network, the matrix J is filled with zeros. The diagonal
vector of the matrix J, denoted by j, will be employed in the analysis of the network.
4.1.2 Nodal Sources
A transported entity associated with a scalar nodal field, , can be supplied, consumed, removed, or dissipated at all or selected nodes of a network at a rate that is
denoted by si , where i = 1, . . . , N. By convention, si , is positive in the case of supply
or generation and negative in the case of removal or dissipation. In the case of an
isolated network, steady state is possible only if the sum of all nodal sources and
sinks is zero. If this condition is not met, accumulation will take place.
4.1.3 Linear Transport
The rate of linear transport along the mth link of a network defined by two ordered
end-nodes labeled k and l is
(4.1.2) qm = cm (k l ),
where cm is the real or complex link conductance.

In the case of electricity, i is the node voltage, qm is the electrical current, and
cm is the electrical conductance, which is the inverse of the electrical resistance.
In the case of heat transport through a network of rods or wires, is the
temperature and
(4.1.3) cm =
kA
,
L
where k is the thermal conductivity of the rod material, A is the rod cross-sectional
area, and L is the rod length.
In the case of fluid flow through a network of pipes or tubes, qm is the volumetric flow rate, is the pressure, p, and cm is the hydraulic conductance. Using
Poiseuilles law, we find that, in the case of transport through a circular tube of radius
a and length L,
(4.1.4) cm =
a4
,
8L
where is the fluid viscosity (e.g., [36]). The higher the fluid viscosity, the longer the
tube length, and the smaller the tube diameter, the lower the conductance. Poiseuilles
law applies under a restricted set of conditions ensuring laminar flow. A nonlinear
law must be employed to describe unsteady turbulent flow.
The difference in the driving potential between the second and first node of the
mth link
(4.1.5) m l k ,
can be expressed in terms on the oriented incidence matrix, R, as

(4.1.6) m = Rl,m l + Rk,m k =
N

Rj,m j .
j=1
Stacking all these differences in an L-dimensional vector, we obtain

(4.1.7) = RT ,
where the vector encapsulates the nodal values of the potential and the superscript
T denotes the matrix transpose.
4.1.4 Nonlinear Transport
In the case of nonlinear transport, the link conductance itself depends on the driving
potential. A nonlinear transport law may prescribe that
(4.1.8) qm = cm (k l )q ,
where q is a positive exponent that is different than unity. If linear transport is

possible only in one direction but cannot occur in the opposite direction, we may
write

(4.1.9) qm =
cm (k l )
0
if k l > 0,
otherwise.
Concisely,
(4.1.10) qm = cm (k l ) H(k l ),
where H(w) is the Heaviside function defined such that H(w) = 1 if w > 0 and
H(w) = 0 if w < 0. In the remainder of this book, we discuss exclusively linear
networks.
Exercises
4.1.1 Electrical and optical conductances
Discuss (a) the electrical conductance of a copper cable and (b) the optical
conductance of a fiber-optic cable.
4.1.2 Nonlinear transport
Discuss a natural or engineering system where a nonlinear transport law should be
employed.
4.1.3 Embedding matrix
What is the structure of the embedding matrix, J, when each node of a network is
connected to an external Dirichlet node?
4.2 UNIFORM CONDUCTANCES
It is instructive to consider the idealized case of a network with uniform link

conductances, c. The simplified setting serves as a convenient point of departure
for introducing basic concepts and deriving governing equations for more general
networks.
4.2.1 Isolated Networks
Balancing the rates of transport at the ith node of an isolated network in the absence
of link dissipation, attrition, supply, or removal yields the balance equation
(4.2.1)
Qm = si ,
where the index m ranges over all links sharing the ith node,
(4.2.2) Qm = cm (i j ),
and j is the label of the second node of the mth link. Substituting into (4.2.1) this
liner transport law and compiling all N equations, we obtain a linear system for the
nodal values of encapsulated in a vector, ,
(4.2.3) L =
1
s.
c
Because the Laplacian matrix, L, is singular, its inverse does not exist and a solution
of the linear system either is not possible or can be found up to an arbitrary constant.
Multiple solutions differing by a constant exist only when the right-hand side of
the linear system (4.2.3) is orthogonal to the eigenvector corresponding to the null
eigenvalue,
(4.2.4) s = 0,
where the N-dimensional vector is filled with ones. This condition requires that
the sum of all nodal sources and sinks is precisely zero. Physically, when the sinks
are balanced by sources, an isolated network does not have a point of reference for
anchoring the nodal field of a transported quantity at steady state.
4.2.2 Embedded Networks
In the case of an embedded network, we balance the rates of transport at each node
in the possible presence of a nodal source or sink and obtain the linear system
(4.2.5) L = +
1
s,
c
where
(4.2.6) L L + J
is the modified Laplacian matrix and J is the embedding matrix defined in (4.1.1).
The vector on the right-hand side is null, except that i is the value of at the
Dirichlet node connected to the ith network node. In the absence of Dirichlet nodes,
J = 0 and = 0.
For example, the modified Laplacian matrix of the embedded network shown in
Figure 4.1.1 is
(4.2.7) L =
3
1
1
1
0
0
0
0
1
4
1
1
0
0
0
0
1
1
4
0
1
0
0
0
1
1
0
4
1
1
0
0
0
0
1
1
4
0
1
1
0
0
0
1
0
3
1
0
0
0
0
0
1
1
3
1
0
0
0
0
1
0
1
2
The sum of the elements in each row or column is not necessarily zero.
It is important to note that, unless the embedding matrix J is null, the modified
Laplacian matrix, L, is nonsingular and the solution of the linear system (4.2.6) is
unique.
Let be an N-dimensional vector filled with ones and j be the diagonal vector of
J. For the network shown in Figure 4.1.1, we have

(4.2.8) j = 0, 1, 1, 0, 0, 1, 0 .
Since L = 0, we have
(4.2.9) L = J = j,
which confirms that, unless j is null, is not an eigenvector of L corresponding to

the null eigenvalue.
Exercise
4.2.1 Modified Laplacian
Confirm that the modified Laplacian matrix displayed in (4.2.7) is nonsingular.
4.3 ARBITRARY CONDUCTANCES
A generalization is necessary in the case of arbitrary link conductances. In the case

of capillary blood flow, hydraulic conductances may differ because of the different
lengths and diameters of the individual capillary segments. In the case of information network transport, the link conductances may be adjusted to reflect preferred or
undesirable transmission venues.
4.3.1 Scaled Conductance Matrix
It is convenient to introduce a reference conductance, c, and express the conductance

of the mth link as
(4.3.1) cm = c m
for m = 1, . . . , L, where m are dimensionless zero or positive coefficients, called the

scaled link conductance, link weight, or edge weight, and L is the number of links.
For future reference, we formulate an L L diagonal matrix,
, called the scaled
conductance matrix, whose mth diagonal element is equal to m ,
1
0
0
0
0
0
0
2
0
0
0
0
0
3
0
0
0
0
.
..
..
..
..
..
..
.
(4.3.2)
=
.
.
.
.
.
.
. .
0
0
L2
0
0
0
0
0
0
0
L1
0
0
0
0
0
0
L
The average scaled link conductance is
1
1
(4.3.3)
m = trace(
).
L
L
L
m=1
In the case of uniform conductances,

is the identity matrix and = 1.
The adjacency matrix, node degrees, and Laplacian matrix must be generalized
to incorporate the edge weights, m .
4.3.2 Weighed Adjacency Matrix
The N N weighed adjacency matrix, , is defined such that ij = m if nodes i and
j are connected by a link labeled m, and ij = 0 otherwise. If all conductances are
equal to c, the weighed adjacency matrix reduces to the adjacency matrix containing
ones and zeros.
For the network shown in Figure 4.1.1 consisting of N = 8 nodes and L = 12
links, the 8 8 weighed adjacency matrix is
0
1
3
7
0
0
0
0
1
0
2
4
0
0
0
0

2
0
0
6
0
0
0
3
0
0
0
0
7
4
5
8
(4.3.4) =
.
0
0
6
5
0
0
10
12
0
0
0
8
0
0
9
0
0
0
0
0
10
9
0
11
0
0
0
0
12
0
11
0
We emphasize that Dirichlet nodes, if present, are excluded from the network.
4.3.3 Weighed Node Degrees
The weighed degree of the ith node, also called the strength of the node, is defined
as
(4.3.5) i =
m ,
where the sum is over all links sharing the ith node. Consequently, i is equal to
the sum of all nonzero elements in the ith row or column of the weighed adjacency
matrix, .
For the network shown in Figure 4.1.1, we have
1 = 1 + 3 + 7 ,
2 = 1 + 2 + 4 ,
4 = 4 + 5 + 7 + 8 ,
(4.3.6) 6 = 8 + 9 ,
3 = 3 + 2 + 6 ,
5 = 5 + 6 + 10 + 12 ,
7 = 9 + 10 + 11 ,
8 = 11 + 12 .
The individual weighed degrees, i , can be arranged along the diagonal line of an
otherwise null N N matrix
(4.3.7) =
1
0
0
..
.
0
2
0
..
.
0
0
3
..
.
..
.
0
0
0
..
.
0
0
0
..
.
0
0
0
..
.
0
0
0
0
0
0
0
0
0
N2
0
0
0
0
N
N1
0
The average scaled link conductance is

1
1
m =
trace(),
L
2L
L
(4.3.8)
m=1
where the factor of two in the denominator arises because each link belongs to two
nodes.
4.3.4 Kirchhoff Matrix
The N N weighed graph Laplacian matrix, also called the Kirchhoff matrix or the
admittance matrix, is given by
(4.3.9) K = ,
which reveals that K is a symmetric matrix. For the network shown in Figure 4.1.1,
we have
(4.3.10) K =
1
1
3
7
0
0
0
0
1
2
2
4
0
0
0
0
3
2
3
0
6
0
0
0
7
4
0
4
5
8
0
0
0
0
6
5
5
0
10
12
0
0
0
8
0
6
9
0
0
0
0
0
10
9
7
11
0
0
0
0
12
0
11
8
By construction, the sum of all elements in each row or column of the Kirchhoff
matrix is zero.
The Kirchhoff matrix for a one-dimensional network takes a tridiagonal form, as
shown in Figure 4.3.1. The Kirchhoff matrix for a periodic one-dimensional network
takes a nearly tridiagonal circulant form, as shown in Figure 4.3.2.
FIGURE 4.3.1 Illustration of a one-dimensional isolated network consisting of N nodes connected by L = N 1 links
and the associated Kirchhoff matrix.
FIGURE 4.3.2 Illustration a periodic one-dimensional network consisting of N unique nodes connected by L = N links and the
associated Kirchhoff matrix.
4.3.5 Weighed Oriented Incidence Matrix
An alternative representation of the Kirchhoff matrix is

(4.3.11) K = R
RT T ,
where R is the N L oriented incidence matrix and

(4.3.12) R
1/2
is a modified oriented incidence matrix defined with respect to the edge weights. The
diagonal elements of the square root,
1/2 , are the square roots of
, while the rest
of the elements are zero.
4.3.6 Properties of the Kirchhoff Matrix
The Kirchhoff matrix, K, shares many of the properties of the Laplacian matrix, L,
discussed in Section 2.2. Let an N-dimensional vector, , contain the nodal values of
a function at the N nodes of an arbitrary network. For any nodal field encapsulated
in a vector, , we find that
(4.3.13) K =
L

m (km lm )2 0,
m=1
where km and lm are the end-nodes of the mth link. Since m 0, K is positive
semidefinite. Consequently, the eigenvalues of K, denoted by i , are either zero or
positive. The sum of the eigenvalues is equal to the trace of K, which is equal to the
trace of , which is equal to sum of the degrees of all nodes.
We may assume that the eigenvalues of K are ordered so that
(4.3.14) 0 = 1 2 N .
Note that the first eigenvalue, 1 , is always zero. Further eigenvalues may also be
zero.
A vector filled with ones, denoted by , is an eigenvector of K corresponding to
the null eigenvalue,
(4.3.15) K = 0,
independent of the link weights. The reason is that the sum of the elements in any
row of K is zero.
A network can be partitioned into two or more pieces based on the eigenvectors
of the Kirchhoff matrix, as discussed in Section 2.2.5. The link conductances have
an important effect on the resulting subgraphs.
Weyls Theorem
Weyls theorem states that increasing the conductance of any one link does not decrease the magnitude of the eigenvalues of the Kirchhoff matrix. The double negative
in this statement means that the magnitude of each eigenvalue either increases or
stays constant when the conductance of any one link is increased. Conversely, decreasing the scaled conductance of any one link does not increase the magnitude of
the eigenvalues. This behavior is in agreement with physical intuition concerning the
effect of the individual links on the overall performance of a network.
4.3.7 Normalized Kirchhoff Matrix
In the absence of unconnected nodes with zero degrees, a normalized weighted in can be
and the corresponding normalized Kirchhoff matrix, K,
cidence matrix, ,
introduced:
1/2 ,
(4.3.16)
=

T .
K
Subject to these definitions, we have

1/2
(4.3.17) K = T = 1/2 K
and
= 1/2 K 1/2 = I 1/2 1/2 ,
(4.3.18) K
where I is the N N identity matrix. All diagonal components of the normalized
Kirchhoff matrix are equal to unity,
Kii = 1. The off-diagonal components are
(4.3.19)
Kij =
1
i j
if nodes i and j are connected by a link, and zero otherwise.

4.3.8 Summary of Notation
We have discussed networks with uniform and varying conductances and introduced
parallel concepts and corresponding notation. Terms and definitions are summarized
in Table 4.3.1. For a network where all links have the same conductance, c, the Kirchhoff matrix, K, reduces to the Laplacian matrix, L. Correspondingly, the modified
Kirchhoff matrix, K, reduces to the modified Laplacian matrix, L.
TABLE 4.3.1 Notation and Definitions for Networks with Nonuniform and Uniform Conductances
Consisting of N Nodes and L Linksa
Size
Nonuniform
Uniform
Scaled link conductance
Scaled Dirichlet-link conductance
Weighed node degree
di
Reference link conductance
Weighed adjacency matrix
NN
Weighed degree matrix (diagonal)
NN
Kirchhoff matrix
NN
K=
L =DA
Weighed oriented incidence matrix
NN
Link conductance matrix (diagonal)
LL
Weighed embedding matrix (diagonal)
NN
Modied Kirchhoff matrix
NN
K= K+T
L=L+J
a In the last column, L is the Laplacian matrix and I is the identity matrix. When all links
have the same conductance, c, the Kirchhoff matrix, K, reduces to the Laplacian matrix, L.
Correspondingly, the modified Kirchhoff matrix, K, reduces to the modified Laplacian matrix, L.
Exercise
4.3.1 Normalized Kirchhoff matrix
Derive the normalized Kirchhoff matrix of the network shown in Figure 4.1.1.
4.4 NODAL BALANCES IN ARBITRARY NETWORKS
Systems of linear equations for the nodal values of a potential, , in a linear network
with arbitrary link conductances can be derived by compiling the balance equations
at the individual nodes. The procedure is analogous to that discussed in Section 4.2
for networks with uniform link conductances.
In the case of an isolated networks with arbitrary link conductances, we obtain the
linear system
(4.4.1) K =
1
s,
c
where the vector s incorporates the N nodal sources, si . Because the Kirchhoff matrix,
K, is singular, a solution exists only when the right-hand side is orthogonal to the
eigenvector corresponding to the null eigenvalue,
(4.4.2) s = 0,
where the N-dimensional vector is filled with ones. When this condition is met, the
solution is defined up to arbitrary constant, independent of the link conductances.
4.4.2 Embedded Networks and the Modified Kirchhoff Matrix
In the presence of Dirichlet nodes, it is convenient to introduce an N N diagonal

matrix, T, that is filled with zeros, except that
(4.4.3) Tii = i
if the ith network node is connected to a Dirichlet node with an external link with
conductance ci , where summation is not implied over the repeated index, i. In the
absence of Dirichlet nodes, the matrix T is null. We refer to the matrix T as the
weighed embedding matrix.
For the network shown in Figure 4.1.1 where nodes 2, 3, and 6 are connected to
Dirichlet nodes, we obtain
(4.4.4) T =
0
0
0
0
0
0
0
0
0
2
0
0
0
0
0
0
0
0
3
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
6
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
If all diagonal elements are zero, we obtain an isolated network.

The nodal values of a transported field in the presence of Dirichlet nodes satisfy
the linear system
(4.4.5) K = T +
1
s,
c
where
(4.4.6) K K + T
is the modified Kirchhoff matrix. The vector is null, except that i is the value
of at the Dirichlet node connected to the ith network node. For illustration, the
modified Kirchhoff matrix of a one-dimensional network involving three Dirichlet
nodes, labeled 1, 3, and N, is shown in Figure 4.4.1.
It is important to remember that, unless the matrix T is null, the modified
Kirchhoff matrix, K, is nonsingular.
4.4.3 Properties of the Modified Kirchhoff Matrix
Let the N-dimensional vector contain the nodal values of a potential at the N nodes
of an embedded network. We find that
(4.4.7) K =
L

m=1
m (km lm ) +
2
N

i i2 0,
i=1
which reveals that K is positive semidefinite. Consequently, the eigenvalues of K,

denoted by i , are zero or positive.
We may assume that the eigenvalues are ordered so that
(4.4.8) 0 1 2 N .
FIGURE 4.4.1 Illustration of a one-dimensional embedded network consisting of N nodes connected by L = N 1 links and
the associated modified Kirchhoff matrix, K.
In the absence of Dirichlet nodes, K is singular and 1 = 0.

Let be an N-dimensional vector filled with ones. Since K = 0, we find that
(4.4.9) K = T = ,
where is the diagonal vector of T. In the case of an isolated network, is filled

with zeros.
Spectral Expansion
It is useful to introduce the diagonal matrix of eigenvalues of the modified Kirchhoff

matrix, K,
1
0
0
0
0
2
0
0
..
.
.
.. ,
.
..
..
..
(4.4.10) = .
.
0
0
N1
0
0
0
0
N
and formulate the matrix of the corresponding eigenvectors, u(i) ,
.
..
(1)
(4.4.11) U =
u(2)
u(N1)
u(N)
u
where each eigenvector is normalized so that its norm is equal to unity, u(m) u(m) = 1
for m = 1, . . . , N, and an asterisk denotes the complex conjugate. By definition, we
have
(4.4.12) K u(m) = m u(m) ,
and thus
(4.4.13) K U = U .
A set of orthonormal eigenvectors can be chosen so that

(4.4.14) U1 = UA ,
where the superscript A denotes the matrix adjoint, defined as the complex conjugate
of the transpose. Accordingly, we obtain
(4.4.15) K = U UA ,
representing the spectral expansion of the modified Kirchhoff matrix.
Exercise
4.4.1 Modified Kirchhoff matrix of a periodic network
Derive the modified Kirchhoff matrix of a one-dimensional periodic network.
4.5 LATTICES
In Chapter 3, we studied the properties of infinite structured networks with uniform conductances associated with regular lattices. The results can be extended in
a straightforward fashion to lattices with nonuniform conductances.
4.5.1 Square Lattice
Consider a rectangular patch of a square network, as shown in Figure 4.5.1, and

assume that the conductance of all horizontal links is 1 c and the conductance of all
vertical links is 2 c, where c is a reference conductance and 1 , 2 are two arbitrary
dimensionless constants. The eigenvalues and eigenvectors of the Kirchhoff matrix
can be computed explicitly in terms of 1 and 2 for isolated, singly periodic, and
doubly periodic configurations. When 1 = 2 = , the Kirchhoff matrix is K = L,
where L is the Laplacian matrix, and the spectrum of the Kirchhoff matrix is the
same as that of the Laplacian matrix.
N2
i2
2
1
1
i1
N1
FIGURE 4.5.1 Illustration of a rectangular section of a

square network containing N1 links in the first direction
and N2 links in the second direction. The conductance
of all horizontal links is 1 c and the conductance
of all vertical links is 2 c, where c is a reference
conductance.
Isolated Network
The Kirchhoff matrix corresponding to the Laplacian matrix of an isolated network

shown in (3.5.16) is
K=
(4.5.1)
1 + 2 I
2 I
0
..
.
2 I
1 + 22 I
2 I
..
.
0
2 I
1 + 22 I
..
.
..
.
0
0
0
0
0
0
0
0
0
..
.
0
0
0
..
.
0
0
0
..
.
0
0
0
..
.
1 + 22 I
2 I
0
2 I
1 + 22 I
2 I
0
2 I
1 + 2 I
where I is the (N1 + 1) (N1 + 1) identity matrix and is the Laplacian matrix of a
one-dimensional isolated network with N1 + 1 nodes. When N1 = 3, we have
1
1
(4.5.2) =
0
0
1
2
1
0
0
1
2
1
0
0
.
1
2
The eigenvalues of the Kirchhoff matrix are

(4.5.3) n1 , n2 = 4 1 sin2
1
2

n1 + 4 2 sin2 12 n2
or
(4.5.4) n1 , n2 = 2(1 + 2 ) 2 1 cos n1 2 2 cos n2 ,
where
(4.5.5) n1 =
n1 1
,
N1 + 1
n2 =
n2 1
N2 + 1
for n1 = 1, . . . , N1 + 1 and n2 = 1, . . . , N2 + 1. The corresponding eigenvectors are

given in (3.1.10).
Periodic Strip
The Kirchhoff matrix corresponding to the Laplacian matrix of the periodic network
given in (3.1.14) is shown in (4.5.1), where I is the N1 N1 identity matrix and is
the Laplacian of a one-dimensional periodic network with N1 unique nodes. When
N1 = 4, we have
2
1
(4.5.6) =
0
1
1
2
1
0
0
1
2
1
1
0
.
1
2
The northeastern and southwestern corner elements implement the periodicity

condition.
The eigenvalues of the Kirchhoff matrix are
(4.5.7) n1 , n2 = 4 1 sin2
1
2

n1 + 4 2 sin2 12 n2
or
(4.5.8) n1 , n2 = 2 (1 + 2 ) 2 1 cos n1 2 2 cos n2 ,
where
(4.5.9) n1 =
n1 1
2 ,
N1
n2 =
n2 1
N2 + 1
for n1 = 1, . . . , N1 + 1 and n2 = 1, . . . , N2 + 1. The corresponding eigenvectors are

given in (3.1.20).
Doubly Periodic Network
The Kirchhoff matrix corresponding to the Laplacian matrix of a doubly periodic

network shown in (3.1.33) is given by
1 + 2 I
2 I
0
2 I
1 + 22 I
2 I
0
2 I
1 + 22 I
.
.
.
..
..
..
..
K=
.
0
0
0
0
0
0
2 I
0
0
0
0
2 I
0
0
0
0
0
0
.
.
.
.
,
..
..
..
..
(4.5.10)
1 + 22 I
2 I
0
2 I
1 + 22 I
2 I
0
2 I
1 + 2 I
where I is the N1 N1 identity matrix and is the Laplacian of a one-dimensional periodic network with N1 unique nodes inside each period. For example, when N1 = 4,
we have
1
1
0
1
1
2
1
0
.
(4.5.11) =
0
1
2
1
1
The northeastern and southwestern corner elements implement the periodicity

condition.
The eigenvalues of the Kirchhoff matrix are given by
(4.5.12) n1 , n2 = 4 1 sin2
1
2

n1 + 4 2 sin2 12 n2
or
(4.5.13) n1 , n2 = 2(1 + 2 ) 2 1 cos n1 2 2 cos n2 ,
where
(4.5.14) n1 =
n1 1
2 ,
N1
n2 =
n2 1
2
N2
for n1 = 1, . . . , N1 and n2 = 1, . . . , N2 . The corresponding eigenvectors are given in

(3.1.41).
4.5.2 Mbius Strip
The Kirchhoff matrix corresponding to the Laplacian matrix

shown in (3.2.5) is
1 + 2 I
2 I
0

+
2
I
2
1
2
2I
0
2 I
1 + 22 I
.
.
..
..
..
K=
.
0
0
0
0
0
0
1 J
0
0
(4.5.15)
of the Mbius strip
..
.
..
.
0
0
0
..
.
0
0
0
..
.
1 J
0
0
..
.
1 + 22 I
2 I
0
2 I
1 + 22 I
2 I
0
2 I
1 + 2 I
where I is the N1 N1 identity matrix and is the Laplacian of an isolated onedimensional network with N1 nodes. The N1 N1 matrix J is null, except that the
northeastern and southwestern corner elements are equal to unity. For example, when
N1 = 3, we have
2
1
0
0
0
0
0
1
1
0
2
1
0
0
0
0
,
.
(4.5.16) =
J=
0
1
2
1
0
0
0
0
0
0
1
2
1
0
0
0
The eigenvalues of the Kirchhoff matrix are given by
(4.5.17) n1 , n2 = 4 1 sin2
1
2

n1 ,n2 + 4 2 sin2 12 n2
or
(4.5.18) n1 , n2 = 2 (1 + 2 ) 2 1 cos n1 , n2 2 2 cos n2 ,
where
(4.5.19) n1 =
n1 1 +
,
N1
n2 =
n2 1
N2 + 1
for n1 = 1, . . . , N1 and n2 = 1, . . . , N2 + 1, where = 0 if n2 is odd and = 1/2 if n2

is even [56]. The corresponding eigenvectors are given in (3.2.12).
a2
N2
a1
i2
2
1
1
i1
N1
FIGURE 4.5.2 Illustration of a rectangular section of a hexagonal network containing N1

links in the first direction, N2 links in the second direction, and one inclined link inside each
cell. The conductance of all links in the first direction is 1 c, the conductance of all
links in the second direction is 2 c, and the conductance of all other links is 3 c, where
c is a reference conductance.
4.5.3 Hexagonal Lattice
Consider a periodic patch of a hexagonal network, as shown in Figure 4.5.2, and

assume that the conductance of all links in the first direction is 1 c, the conductance
of all links in the second direction is 2 c, and the conductance of all other links is
3 c, where c is a reference conductance.
The eigenvalues of the Kirchhoff matrix in the doubly periodic configuration are
(4.5.20) n1 , n2 = 4 1 sin2
1
2

n1 + 4 2 sin2 12 n2 + 4 3 sin2 12 (n1 n2 )
or
(4.5.21) n1 , n2 = 2 (1 + 2 + 3 ) 2 1 cos n1 2 2 cos n2 23 cos(n1 n2 )
for n1 = 1, . . . , N1 and n2 = 1, . . . , N2 , where

(4.5.22) n1 =
n1 1
2 ,
N1
n2 =
n2 1
2 .
N2
The corresponding eigenvectors are same as those of the doubly periodic Laplacian,
given in (3.3.21).
4.5.4 Modified Union Jack Lattice
Consider a periodic patch of a modified Union Jack lattice, as shown in Figure 4.5.3,
and assume that the conductances of all links in the first direction is 1 c, the conductances of all links in the second direction is 2 c, the conductances of all links
inclined toward the first direction is 3 c, and the conductances of all links inclined
N2
i2
2
1
1
N1
i1
FIGURE 4.5.3 Illustration of a modified Union Jack lattice containing

N1 links in the first direction, N2 links in the second direction, and
two noncrossing transverse links inside each cell. The conductance of all links in the first direction is 1 c, the conductance of
all links in the second direction is 2 c, the conductance of all links
inclined toward the first direction is 3 c, and the conductance of
all links inclined toward the second direction is 4 c.
toward the second direction is 4 c, where c is a reference conductance and 1 4

are dimensionless coefficients.
The eigenvalues of the doubly periodic Kirchhoff matrix are given by

n1 , n2 = 4 1 sin2 12 n1 + 4 2 sin2 12 n2 + 4 3 sin2 12 n1 n2
(4.5.23)

+ 4 4 sin2 12 n1 + n2
or
(4.5.24)
n1 ,n2 = 2 (1 + 2 + 3 + 4 ) 1 cos n1 2 cos n2 3 cos(n1 n2 )

4 cos(n1 + n2 ),
where
(4.5.25) n1 =
n1 1
2 ,
N1
n2 =
n2 1
2
N2
for n1 = 1, . . . , N1 and n2 = 1, . . . , N2 . The corresponding eigenvectors are same as

those of the doubly periodic Laplacian, given in (3.4.17).
4.5.5 Simple Cubic Lattice
Consider a simple cubic network whose nodes are arranged on a Cartesian grid, as
shown in Figure 4.5.4. The conductance of all links in the first direction is 1 c, the
conductance of all links in the second direction is 2 c, and the conductance of all
links in the third direction is 3 c, where c is a reference conductance and 1 , 2 , and
3 are three arbitrary dimensionless constants.
N2 + 1
i2
i1
N1 + 1
i3
N3 + 1
FIGURE 4.5.4 Illustration of a rectangular slab of a

simple cubic network containing N1 links in the first
direction, N2 links in the second direction, and N3
links in the third direction. For the configuration
shown, we have N1 = 2, N2 = 2, and N3 = 1. The
conductance of all links in the first direction is 1 c,
the conductance of all links in the second direction
is 2 c, and the conductance of all links in the third
direction is 3 c, where c is a reference conductance.
Isolated Network
The eigenvalues of the Kirchhoff matrix for an isolated network are

(4.5.26) n1 , n2 , n3 = 4 1 sin2
1
2

n1 + 4 2 sin2 12 n2 + 4 3 sin2 12 n3
or
(4.5.27) n1 , n2 , n3 = 2 (1 + 2 + 3 ) 2 1 cos n1 2 2 cos n2 2 3 cos n3 ,
where
(4.5.28) n1 =
n1 1
,
N1 + 1
n2 =
n2 1
,
N2 + 1
n3 =
n3 1
N3 + 1
for n1 = 1, . . . , N1 + 1, n2 = 1, . . . , N2 + 1, and n3 = 1, . . . , N3 + 1, The corresponding

eigenvectors are given in (3.7.6).
The eigenvalues of the Kirchhoff matrix for a triply periodic simple cubic network
are given by
(4.5.29) n1 , n2 , n3 = 4 1 sin2
1
2

n1 + 4 2 sin2 12 n2 + 4 3 sin2 12 n3
or
(4.5.30) n1 , n2 , n3 = 2 (1 + 2 + 3 ) 2 1 cos n1 2 2 cos n2 2 3 cos n3 ,
where
(4.5.31) n1 =
n1 1
2 ,
N1
n2 =
n2 1
2 ,
N2
n3 =
n3 1
2
N3
for n1 = 1, . . . , N1 , n2 = 1, . . . , N2 , and n3 = 1, . . . , N3 . The corresponding eigenvectors

are given in (3.1.41).
Exercise
4.5.1 Cubic lattices
(a) Derive the eigenvalues and eigenvectors of the triply periodic Kirchhoff matrix
associated with the body-centered cubic (bcc) lattice. (b) Repeat (a) for the facecentered cubic (fcc) lattice.
4.6 FINITE DIFFERENCE GRIDS
In Section 1.1, we saw that one-dimensional graphs and their Laplacian arise from
uniform finite difference grids for solving the Laplace or Poisson equation in one
dimension. Two- and higher-dimensional graphs and their Laplacian arise from
corresponding Cartesian or curvilinear grids.
As an example, we consider the Poisson equation in the xy plane for an unknown
function f (x, y),
(4.6.1) 2 f =
2f 2f
+
+ g(x, y) = 0,
x2 y2
where g(x, y) is a specified source term and

(4.6.2) 2 =
2
2
+
x2 y2
is the Laplacian operator expressed in Cartesian coordinates.

Cartesian Grid
To implement the finite difference method, we introduce a Cartesian grid with uniform grid spacings, x and y, is shown in Figure 4.6.1. Applying the Poisson
equation at the (i, j) node and approximating the second partial derivatives with
central differences,
(4.6.3)
2f
x2 i,j
fi1, j 2fi, j + fi+1, j

+ O(x2 ),
x2
j
y
i
x
FIGURE 4.6.1 A Cartesian finite difference grid used to

solve the Poisson equation in two dimensions.
and
(4.6.4)
2f
y2 i, j
fi, j1 2fi, j + fi, j+1

+ O(y2 ),
y2
we obtain the difference equation

(4.6.5)
fi1, j 2fi, j + fi+1, j fi, j1 2fi, j + fi, j+1

+
+ gi,j = 0.
x2
y2
Rearranging, we obtain
(4.6.6) 2(1 + )fi, j fi+1, j fi1, j fi, j1 fi, j+1 = x2 gi, j ,
where = (y/x)2 . Compiling all difference equations and implementing specified boundary or periodicity conditions provides us with a linear system involving
the Kirchhoff or modified Kirchhoff matrix for the square lattice, as discussed in
Chapter 1 for the corresponding problem in one dimension.
Isolated networks arise when the Neumann boundary condition is specified
around the solution domain, and embedded networks arise when the Dirichlet
boundary condition is entirely or partially employed.
Interpolated Field
When x = y = a, corresponding to = 1, we obtain the interior difference

equation
(4.6.7) 4fi, j fi+1, j fi1, j fi, j 1 fi, j +1 = a2 gi, j .
Compiling all difference equations and implementing the boundary or periodicity conditions, we obtain system (4.2.3) or (4.2.5) with the Laplacian or modified
Laplacian of the square lattice.
(a)
(b)
(c)
a
3
3
0
0
4
FIGURE 4.6.2 Computational stencils of the Laplacian on (a) a square,

(b) a honeycomb, (c) and a hexagonal lattice.
In fact, a detailed error analysis of the square computational stencil illustrated in

Figure 4.6.2(a) with x = y = a reveals that

1
1
4f 4f 2
+
a
x4 y4 0
(4.6.8) 2 f 0 2 4f0 f1 f2 f3 f4 +
12
a
(e.g., [35], p. 508). This means that the discrete (network) solution describes exactly
linear, quadratic, and cubic continuous fields constructed by interpolation.
Honeycomb Grid
The computational stencil of the Laplacian on a honeycomb lattice is shown in

Figure 4.6.2(b). When the x axis is aligned with the first link, as shown in the
illustration, we find that
(4.6.9) ( 2 f )0
1
3f
4
3f
3f
3
a
0
1
2
3
3a2
6 x3
xy2 0
(e.g., [3], p. 507; [35], p. 511). Similar approximations can be written when the first
link is aligned with the y axis. Compiling all difference equations and implementing boundary or periodicity conditions, we obtain system (4.2.3) or (4.2.5) with the
Laplacian or modified Laplacian of the honeycomb lattice.
Hexagonal Grid
In the case of the hexagonal lattice illustrated in Figure 4.6.2(c) where each node is
shared by six links, we obtain
1
2
6
f
fi
( 4 f )0 a2 + ,
0
16
3a2
6
(4.6.10) ( 2 f )0
i=1
where 4 = 2 2 is the biharmonic operator (e.g., [35], p. 511). Compiling all

difference equations and implementing boundary or periodicity conditions, we obtain
system (4.2.3) or (4.2.5) with the Laplacian or modified Laplacian of the hexagonal
lattice.
Helmholtz Equation
Other differential equations can be solved by finite difference methods. Consider the
Helmholtz equation in two dimensions,
(4.6.11) 2 f =
2f 2f
+
+ kf = 0,
x2 y2
where k is a real or complex constant. The counterpart of the difference equation

(4.6.7) on a square grid is
(4.6.12) (4 + k) fi, j fi+1, j fi1, j fi, j 1 fi, j +1 = 0.
A generalized equation is

(4.6.13) t fi, j 12 fi+1, j + fi1, j + fi, j 1 + fi, j +1 = 0,
where t and are arbitrary unrelated coefficients. Compiling all difference equations
and implementing boundary or periodicity conditions, we obtain a linear system that
is similar to (4.2.3) or (4.2.5).
Exercise
4.6.1 Finite difference discretization
Formulate a linear system for solving the Poisson equation on a uniform Cartesian
lattice when the Neumann boundary condition is specified around the four edges of
a rectangular solution domain.
4.7 FINITE ELEMENT GRIDS
The finite element method provides us with a venue for deriving systems of algebraic
equations for the nodal values of an unknown function that satisfies a given ordinary
or partial differential equation (e.g., [34]). The nodes define segments in one dimension or geometrical elements with various shapes in two and three dimensions.
The algebraic equations can be derived by various methods, including the method of
Galerkin projection and the method of least squares minimization.
4.7.1 One-Dimensional Grid
Consider a one-dimensional finite element grid consisting of straight segments,

called finite elements, as shown in Figure 4.7.1. Our objective is to compute a
numerical solution of the one-dimensional Laplace equation,
(4.7.1)
d2 f
= 0,
dx2
1
Elements
Nodes
i
2
i1
i +1
x
N
FIGURE 4.7.1 A one-dimensional finite element grid consisting of a chain of

straight segments connected at nodes.
subject to suitable boundary conditions. Applying the Galerkin finite element method
under the assumption that the finite element solution varies linearly across the length
each element, we obtain an algebraic equation associated with the ith interior node,
(4.7.2)
1
hi1

fi1 +
1
hi1

1
1
+
fi fi+1 = 0,
hi
hi
where hi = xi+1 xi is the element length (e.g., [34]).

The finite element grid can be regarded as a one-dimensional network, and equation (4.7.2) can be regarded as a nodal balance involving links with conductances
(4.7.3) ci =
1
a 1
=
= i c,
hi hi a
i =
a
,
hi
c=
1
,
a
and a is a reference length. The conductance is inversely proportional to the element

size, in agreement with physical intuition. The Laplacian matrix can be assembled
by collecting the finite element equations at each node.
4.7.2 Two-Dimensional Grid
Next, we consider a two-dimensional finite element grid consisting of three-node

triangles (e.g., [34]). An example of a grid generated by Delaunay triangulation based
(a)
(b)
D
E
H
C
F
A
B
G
FIGURE 4.7.2 (a) A two-dimensional finite element grid consisting of

three-node triangles with straight edges generated by Delaunay triangulation based on a specified set of nodes. (b) A neighborhood
of the finite element grid where element edges are interpreted as
network links. The dotted lines describe the underlying Voronoi
tessellation.
on a specified set of nodes is shown in Figure 4.7.2(a). Our objective is to compute

a numerical solution of the two-dimensional Laplace equation,
(4.7.4)
2f 2f
+
= 0,
x2 y2
subject to suitable boundary conditions.

A typical neighborhood of a finite element grid is shown in Figure 4.7.2(b). Node
A is connected with element edges to six adjacent nodes, BG. We denote the area
of a triangular element formed by three vertices, X, Y, and Z, by AXYZ , the square
of the length of a straight segment connecting nodes X and Y by 2XY , and the vector
By definition,
connecting an oriented pair of nodes X and Y by XY.
XY.

(4.7.5) 2XY = XY
Applying the Galerkin finite element method under the assumption that the finite
element solution varies linearly over each triangular element with respect to x and y,
we obtain an algebraic equation associated with the interior node labeled A,

(4.7.6)
X fX = 0,
X=A, B, . . . , G
where the scalar coefficients X are given by

A =
2BC
2
2
2
2
2
+ CD + DE + EF + FG + GB ,
AABC AACD AADE AAEF AAFG AAGB
B =
GB

CB

AG
AC
+
,
AAGB
AABC
C =
BC

DC

AB
AD
+
,
AABC
AACD
D =
CD

ED

AC
AE
+
,
AACD
AADE
E =
DE
FE

AD
AF
+
,
AADE
AAEF
F =
EF
GF

AE
AG
+
,
AAEF
AAFG
G =
FG

BG

AF
AB
+
.
AAFG
AAGB
(4.7.7)
Using elementary geometry, we confirm that

(4.7.8) A =
X ,
X=B, . . . , H
and thus
(4.7.9)

X=A, . . . , H
X = 0.
The finite element grid may thus be regarded as a two-dimensional network, and
equation (4.7.5) may be regarded as a nodal balance involving links originating from
point A with conductances
(4.7.10) cAX = X .
for X = B, . . . , H.
As an exercise, we consider the hexagonal finite element assembly shown in
Figure 4.6.2(c). The area of each triangular element is A = 43 a2 . The preceding
formulas give
(4.7.11) X =
a2
4
=
A
3
for X = B, . . . , H, consistent with the finite difference derivation.

The interpretation of the finite element edges as network links hinges on the independence of the link conductance on the node where the finite element equation is
applied. To demonstrate this subtlety, we apply the finite element equation at point
C in Figure 4.7.1(b), which is connected by element edges to four nodes labeled A,
B, H, and D, and obtain the equation
(4.7.12)
X fX = 0,
X=C,A,B,H,D
where
C =
2AB
2
2
2
+ BH + HD + DA ,
ACAB ACBH ACHD ACDA
(4.7.13) A =
DA

BA

CD
CB
+
,
ACDA
ACAB
B =
AB

HB

CA
CH
+
,
ACAB
ACBH
H =
BH
DH

CB
CD
+
,
ACBH
ACHD
D =
HD
AD

CH
CA
+
.
ACHD
ACDA
Using elementary geometry, we confirm that

(4.7.14) C =
X .
X=A,B,H,D
Equation (4.7.12) can be regarded as a nodal balance involving links originating from
point C with conductances
(4.7.15) cCX = X
C
C
A
B
D
D
FIGURE 4.7.3 The conductance of a link

connecting nodes A and B in a twodimensional finite element grid for solving Laplaces equation is defined in terms
of the angles C and D according to
(4.7.17).
for X = A, B, H, D. The aforementioned interpretation of element edges as network

links with unique conductance hinges on the observation that
(4.7.16) cAC = cCA = C = A .
These results indicate that the conductance of an edge connecting nodes A and B
in a triangular finite element grid, as shown in Figure 4.7.2, is given by
(4.7.17) cAB =
CB

DB

AC
AD
= cot C | + cot| D ,
AABC
AADB
where the angles C and D are defined in Figure 4.7.3. Conversely, when the link
conductivities are computed from (4.7.17), the nodal field of the underlying network that is consistent with a finite element grid represents a solution of Laplaces
equation.
Similar conclusions are reached in the analysis of three-dimensional tetrahedral
finite element grids (e.g., [34]).
Exercise
4.7.1 Three-dimensional grid
Derive an expression for the link conductance of a three-dimensional finite element
grid consisting of tetrahedral elements (e.g., [34]).
/// 5 ///
GREENS FUNCTIONS
In the context of networks, a Greens function represents the

nodal field established when a source is applied at a specified node. In discussing
network Greens functions, it is imperative to make a distinction between embedded
networks connected to their environment through Dirichlet nodes and isolated networks distinguished by the absence of Dirichlet nodes, as discussed in Section 4.1.
Infinite lattices are special realizations of embedded networks. Because of the singular nature of the Laplace or Kirchhoff matrix, the MoorePenrose Greens function,
also called a generalized Greens function associated with a matrix pseudo-inverse,
must be employed in the case of isolated networks.
Regular and generalized Greens functions are elementary mathematical devices
useful in theoretical analysis and practical applications. The networks discussed in
this chapter are assumed to be connected, that is, to be devoid of fragments, islands,
and isolated nodes, unless stated otherwise.
5.1 EMBEDDED NETWORKS
When selected nodes of a network are attached to Dirichlet nodes, the modified
Kirchhoff matrix introduced in Section 4.4, given by
(5.1.1) K = K + T,
is invertible, where K is the Kirchhoff matrix. We recall that the N N matrix T is

filled with zeros, except that Tii = i if the ith node is connected to a Dirichlet node
by an external link with conductance ci , where c is a reference conductance.
The Greens function vector associated with the ith node, denoted by g(i), is the
N-dimensional nodal field satisfying the equation
(5.1.2) K g(j) = e(j) ,
where the unit vector e(j) is filled with zeros, except that the jth element is equal to
(j)
unity, ej = 1. Since K is invertible, a unique solution can be found, given by
(5.1.3) g(j) = K1 e(j)
161
for j = 1, . . . , N. Physically, the nodal field associated with the Greens function is
established when a unit source is applied at the jth node, while the potential associated with the transported field is held at the reference value of zero at the Dirichlet
nodes supporting the network. For example, in the case of fluid flow through a network or capillary tubes, fluid is injected at one node, while the pressure is held at the
reference value of zero at peripheral Dirichlet nodes.
5.1.1 Greens Function Matrix
The N N Greens function matrix contains in its columns all nodal Greens function
vectors,
(j)
(5.1.4) Gij gi .
By definition,
(5.1.5) K G = I,
G = K1 ,
where I is the N N identity matrix. Thus, the Greens function matrix is simply the
inverse of the modified Kirchhoff matrix.
Because the modified Kirchhoff matrix is symmetric, the Greens function matrix
is also symmetric,
(5.1.6) Gij = Gji .
Thus, the N N Greens function matrix encompasses in its columns or rows all
nodal Greens function vectors.
In terms of the Greens function matrix, the solution of the linear system (4.4.5)
governing network transport,
(5.1.7) K = T +
1
s,
c
is given by
(5.1.8) = G ,
where
(5.1.9) T +
1
s,
c
and the vector s encompasses the nodal sources. These expressions are consistent
with the definition g(j) = G e(j) .
G r e e n s Fu n c t i o n s / / 163
Spectral Expansion
Using (5.1.5) and the spectral expansion of the modified Kirchhoff matrix stated in
(4.4.15), we find that the Greens function admits the spectral expansion
(5.1.10) G = U 1 UA ,
where the superscript A denotes the matrix adjoint, that is, the complex conjugate of
the transpose of the underlying matrix. The corresponding sum representation is
N

1 (s)
(5.1.11) G =
u u(s) ,
s
s =1
where u(s) are the eigenvectors of the augmented Kirchhoff matrix, normalized so
that
(5.1.12) u(s) u(s) = 1,
and an asterisk denotes the complex conjugate. In index notation,

(5.1.13) Gij =
N

1 (s) (s)
u u .
s i j
s=1
Since all eigenvalues are nonzero, the sum is well-defined. This representation is also
valid in the case of multiple eigenvalues supporting an orthonormal set of distinct
eigenvectors.
5.1.2 Normalized Greens Function
The nodal field due to a point source responsible for the Greens function can be
normalized so that it takes the reference value of zero at the application point. The
corresponding normalized Greens function, indicated by a tilde, is defined as
ij Gij Gjj ,
(5.1.14) G
where summation is not implied over the repeated index, j. By definition, the
diagonal elements of G are zero:
jj = 0.
(5.1.15) G
Using (5.1.13), we find that the spectral expansion of the normalized Greens
function is
ij =
(5.1.16) G
N

1
(s) (s) (s)
ui uj uj .
s
s=1
It is important to note that the normalized Greens function is not necessarily

symmetric, that is,
ij = G
ji ,
(5.1.17) G
in general. The reason is that Gii is not necessarily equal to Gjj . An exception occurs
in the case of an infinite regular lattice.
Exercise
5.1.1 One-dimensional network
Compute the Greens function of the one-dimensional network shown in Figure 4.4.1
for uniform link conductances, i = 1, 1 = 1, 3 = 1, and N = 1.
5.2 ISOLATED NETWORKS
In the absence of Dirichlet nodes, T = 0, the modified Kirchhoff matrix, K, reduces

to the Kirchhoff matrix, K, which is singular due to the presence of a zero eigenvalue with a corresponding uniform eigenvector, , even in the absence of network
fragments and islands, as presently assumed.
Let the N-dimensional vector be filled with ones. Because the right-hand side
of the linear system defining a Greens function vector,
(5.2.1) K g(j) = e(j) ,
does not satisfy the compatibility condition e(j) = 0, a solution cannot be found
and the Greens function of an isolated network is poorly defined.
5.2.1 MoorePenrose Greens Function
To circumvent this difficulty, we introduce a nodal field, h(j) , established when a unit
source is applied at the jth node, while a uniform distribution of sinks is simultaneously applied at all nodes, so that the total strength of the point source and sinks is
zero. By definition, the nodal field h(j) satisfies the linear system
1
(5.2.2) K h(j) = e(j) .
N
Since = N,

(5.2.3)
1
e
N
(j)

= 0,
the compatibility condition (4.4.2) is fulfilled, and the solution of the linear system
(5.2.2) can be found up to an arbitrary uniform nodal field.
To render the solution unique, we may specify that

(5.2.4) h(j) = 0,
that is, we may stipulate that the N elements of h(j) add up to zero.
Next, we put the individual nodal fields h(j) for j = 1, . . . , N, at the columns of a
matrix H satisfying the equation
(5.2.5) K H = H K = I ,
where
(5.2.6) I I
1
,
N
I is the N N identity matrix, and all components of the N-dimensional vector and
N N matrix are equal to unity (e.g., [26]). Explicitly,
1
(5.2.7) I =
N
N1
1
1
..
.
1
N1
1
..
.
1
1
N1
..
.
..
.
1
1
1
..
.
1
1
1
..
.
1
1
1
..
.
1
1
1
1
1
1
1
1
1
N1
1
1
1
N1
1
1
1
N1
It will be noted that the matrix on the right-hand side is the Laplacian of a complete
graph.
The imposed condition (5.2.4) requires that the sum of the elements in each row
or column of H is zero:
(5.2.8) H = H = 0.
Consequently and because = N, we have

(5.2.9) K = 0,
H = 0,
I = 0, K I = K,
H I = H.
In fact, the matrix H is the MoorePenrose inverse of the Kirchhoff matrix

satisfying the equations
(5.2.10) K H K = K,
H K H = H.
For any vector c that is orthogonal to the uniform eigenvector , satisfying c = 0,

we have
(5.2.11) K d = c,
where
(5.2.12) d H c
and the vector d is also orthogonal to .

To compute the generalized Greens function H, we note that

1
(5.2.13) (K + ) H + 2 = I
N
and set
(5.2.14) H = (K + )1
1
.
N2
The inverse of the deflated Kirchhoff matrix on the right-hand side of (5.2.13),
(K + )1 , is well-defined [26].
Network transport is governed by the linear system (4.4.1):
(5.2.15) K =
1
s.
c
Multiplying this system by H and using (5.2.5), we obtain the solution

(5.2.16) =
1
1
H s + ( ) .
c
N
The second term on the right-hand side contributes an inconsequential uniform nodal
field.
5.2.2 Spectral Expansion
Using (5.2.12), we find that if 1 = 0 and s are the eigenvalues of the Kirchhoff
matrix, K, for s = 2, . . . , N, then 1 = 0 and s = 1/s for s = 2, . . . , N are eigenvalues of the generalized Greens function H, and the corresponding eigenvectors are
identical. By definition, we have
(5.2.17) H U = U 0 ,
where
0
0
..
.
(5.2.18) 0 =
0
0
0
1/2
..
.
..
.
0
0
..
.
0
0
..
.
0
0
1/N1
0
0
1/N
is a regularized matrix of inverse eigenvalues, excluding the troublesome infinite

inverse eigenvalue.
A set of eigenvectors can be chosen so that
(5.2.19) U1 = UA ,
where the subscript A denotes the matrix adjoint, that is, the complex conjugate of
the transpose. The spectral expansion of H is
(5.2.20) H = U 0 UA .
Explicitly,
N

1 (s)
(5.2.21) H =
u u(s)
s
s=2
or
(5.2.22) Hij =
N

1 (s) (s)
u u ,
s i j
s=2
where u(s) are the eigenvectors of the Kirchhoff matrix normalized so that
u(s) u(s) = 1, and an asterisk denotes the complex conjugate. Note that summation
begins at s = 2.
5.2.3 Normalized MoorePenrose Greens Function
The nodal field due to a point source responsible for the MoorePenrose Greens
function can be normalized to take the reference value of zero at the application point,
yielding the corresponding normalized MoorePenrose Greens function, indicated
by a tilde
ij Hij Hjj .
(5.2.23) H
By definition, we have
jj = 0,
(5.2.24) H
where summation is not implied over the repeated index, j. The spectral expansion of the normalized generalized Greens function follows from the representation
(5.2.22):
ij =
(5.2.25) H
N

1
(s) (s) (s)
ui uj uj .
s
s=2
It is important to note that the normalized MoorePenrose Greens function is not

necessarily symmetric.
5.2.4 One-Dimensional Network
As an example, we consider an isolated one-dimensional network consisting of N

nodes and L = N 1 links with same conductance, c, as shown in Figure 5.2.1(a).
Using the eigenvalues and eigenvectors of the Laplacian matrix given in (1.7.2) and
(1.7.4), we obtain the MoorePenrose Greens function
N

1
1
cos i 12 s cos j 12 s ,
(5.2.26) Hij =
2 1
2N
s = 2 sin 2 s
where s = (s 1) /N. Note that summation begins at s = 2. The elements of H are

plotted in Figure 5.2.1(b).
(a)
Links:
Nodes: 1
i
2
i1
(b)
i+1
(c)
Hij
Hij
6
4
2
0
2
4
6
15
10
j
14 16
8 10 12
2 4 6
i
1
0
1
2
3
4
5
6
7
8
15
10
j
12 14 16
8 10
2 4 6
i
FIGURE 5.2.1 (a) Illustration of a one-dimensional isolated network consisting of N nodes connected by L = N 1 links. Graph of (b) the MoorePenrose Greens function and (c) the normalized
MoorePenrose Greens function illustrating the loss of symmetry for N = 16.
An alternative representation is

N 1

1
1
1
cos i
(5.2.27) Hij =
pk cos j 12 pk ,
2 1
2N
2
p =1 sin 2 pk
where k = /N is the fundamental wave number and p s 1. Note that summation
begins at p = 1.
The normalized MoorePenrose Greens function is given by the corresponding
sum representation

1
ij = 1

H
2 1
2N
s = 2 sin 2 s
(5.2.28)

cos i 12 s cos j 12 s
cos j 12 s ,
N
which can be rearranged into

1
i,j = 1

H
2 1
2N
pk
(5.2.29)
p = 1 sin
2

cos i 12 pk cos j 12 pk
cos (j 12 )pk .
N 1
are plotted in Figure 5.2.1(c), demonstrating the absence of

The elements of H
symmetry.
5.2.5 Periodic One-Dimensional Network
As another example, we consider a one-dimensional network with N periodically

repeated nodes and L = N links of the same conductance, as shown in Figure 5.2.2(a).
Using the eigenvalues and eigenvectors of the periodic Laplacian given in (1.8.2) and
(1.8.4), we obtain the generalized Greens function

N
1 exp i (i j) s
,
(5.2.30) Hij =
4N
sin2 12 s
s=2
where s = (s 1)2 /N and i is the imaginary unit.
Alternative representations are

N 1
N 1
1 cos (i j) pk
1 cos (l 1)pk
,
(5.2.31) Hij =
2 1
2 1
4N
4N
pk
sin
pk
p =1 sin
p
=1
2
2
where k = 2 /N is the fundamental wave number, p s 1, and l = i j + 1.
(a)
i
i1
i
i+1
2
Links:
L
N
Nodes:
(b)
(c)
4
3
2
1
0
1
2
15
0
1
Hij
Hij
2
3
4
5
15
10
j
14 16
8 10 12
2 4 6
i
10
j
14 16
8 10 12
2 4 6
i
FIGURE 5.2.2 (a) Illustration of a one-dimensional periodic network consisting of nodes connected by links. Graph of (b) the MoorePenrose Greens function and (c) the normalized
MoorePenrose Greens function for N = 16.
Comparing the last expression in (5.2.31) with the cosine Fourier expansion
(1.8.27), we obtain the complex Fourier coefficients c0 = 0 and
(5.2.32) cp =
1
1

8N sin2 1 pk
2
for p = 1, . . . , N 1. A graph of the periodic MoorePenrose Greens function is shown in Figure 5.2.2(b). Because of translational invariance, all diagonal
components are equal.
The normalized MoorePenrose Greens function is given by

N
1 1 exp i (i j) s

(5.2.33) Hij =
,
4N
sin2 1
s=2
2 s
where s = 2 (s 1)/N, which can be restated as

N 1
k 1 cos (i j)pk

.
(5.2.34) Hi,j
8
sin2 12 pk
p=1
A graph of the normalized Greens function is shown in Figure 5.2.2(c). The apparent
symmetry is due to translational invariance along the periodic array for equal link
conductances in the absence of end effects.
5.2.6 Free-Space Greens Function in One Dimension
In the limit N , the sum in (5.2.34) reduces into an integral and the right-hand
side provides us with the Greens function of the one-dimensional infinite lattice,
m H
j+m,j = 1
(5.2.35) L
8
1 cos(m)
d
sin2 12
or
m =
(5.2.36) L
1
4
2
0
1 cos(m)
d,
1 cos
where = pk. Performing the integration, we obtain

m = 1 |m|.
(5.2.37) L
2
This expression is the exact counterpart of the Greens function of Laplaces equation
in one dimension,
(5.2.38)
d2f
+ 1 (x) = 0,
dx2
given by
(5.2.39) G = 12 |x|,
where 1 (x) is the one-dimensional Dirac delta function.

5.2.7 Complete Network
In the case of a complete network described by a complete graph, as illustrated

in Figure 2.1.2, we use the eigenvalues and eigenvectors of the Laplacian given in
(2.2.11) and (2.2.12) and obtain

N 1
1
2
(5.2.40) Hij = 2
exp i q p
N
N
q=1
where p = i j. A discrete Fourier orthogonality property states that
(5.2.41)
N

q=1

2
N
exp i qp
=
0
N
if p = sN,
otherwise,
where p and s are zero or arbitrary integers. Consequently,

1
(5.2.42) Hij = 2
N
N1
1
if
if
i = j,
i = j
and
(5.2.43) H =
1
L.
N2
We have found that the MoorePenrose Greens function of a complete network is

proportional to the corresponding Laplacian.
5.2.8 Discontiguous Networks
Suppose that an isolated network is unconnected, consisting of p isolated fragments.

For example, when p = 2, an island consisting of N2 nodes may float in an ambient
network consisting of N1 nodes, where N = N1 + N2 . The inverse of the Kirchhoff
matrix, K, does not exist due to the presence of p zero eigenvalues. The elements of
the first eigenvector, (1) , are equal to unity over the first nodal set and zero over the
second nodal set, whereas the elements of the second eigenvector, (2) , are equal to
unity over the second nodal set and zero over the first nodal set.
A MoorePenrose Greens function, H, can be introduced, satisfying the equation
p

1 (s)
(5.2.44) K H = I
(s) ,
Ns
s=1
where I is the N N identity matrix. The elements of the N-dimensional vector (s)
are equal to unity over the sth nodal set and zero over the complement of the sth
nodal set, so that
(5.2.45) K (s) = 0.
Noting that (s) (s) = Ns , we find that

(5.2.46) K H (s) = 0.
In addition, we require that

(5.2.47) H (s) = (s) H = 0,
for s = 1, . . . , p. For any vector, c, orthogonal to the span of e(s) , satisfying c e(s) = 0,
we find that
(5.2.48) K d = c,
where
(5.2.49) d = H c,
and the vector d is also orthogonal to the span of e(s). In fact, H is a MoorePenrose
pseudoinverse satisfying the equation
(5.2.50) K H K = K.
The eigenvalues of H are s = 1/s for s = 1, . . . , N, except that s = 0 if s = 0.

The corresponding eigenvectors are the same as those of K.
Exercise
5.2.1 One-dimensional network
Confirm that (5.2.26) satisfies equation (5.2.5).
5.3 LATTICE GREENS FUNCTIONS
In Chapters 2 and 3, we discussed infinite lattices and studied the spectra of their
doubly or triply periodic Laplacian. The results are useful in deriving specific
expressions for periodic and free-space Greens functions.
5.3.1 Periodic Greens Functions
An infinite two- or three-dimensional lattice admits periodic Greens functions representing the nodal field due to a doubly or triply periodic array of point sources. In
two dimensions, each node is parametrized by two indices, i1 and i2 , and the periodic
MoorePenrose Greens function is denoted by
j ,j
(5.3.1) Hi11 , i22 .
j ,j
Physically, Hi11 , i22 represents the nodal value at the (i1 , i2 ) vertex due to a source
applied at the (j1 , j2 ) vertex and its periodic images. Three indices are employed in
three dimensions. The triply periodic MoorePenrose Greens function is denoted by
j ,j ,j
(5.3.2) Hi11 ,i22,i33 .
It is convenient to introduce two shifting indices in two dimensions, defined as

(5.3.3) m1 = i1 j1 ,
m 2 = i2 j2 ,
and write
j ,j
j ,j
(5.3.4) Hi11 ,i22 = Hj11 +m2 1 , j2 +m2 Mm1 ,m2 ,
where Mm1 , m2 is the MoorePenrose periodic lattice Greens function. The nodal
value at the point source corresponds to m1 = 0 and m2 = 0. Corresponding
definitions are made in three dimensions.
Normalized Greens Functions
The periodic MoorePenrose Greens function can be normalized so that it takes the
reference value of zero at the source point and its images. In two dimensions, the
normalized Greens function, indicated by a tilde, is defined as
j1 , j2 Hj1 , j2 Hj1 , j2 .
(5.3.5) H
i1 , i2
i1 ,i2
j1 , j2
The corresponding normalized periodic generalized lattice Greens function is
m ,m Mm ,m M0,0 .
(5.3.6) M
1 2
1 2
By construction,
1 2 = 0,
(5.3.7) H
j1 , j2
j ,j
0,0 = 0.
M
Three indices are employed in three dimensions.
5.3.2 Free-Space Greens Functions
The free-space lattice Greens functions represents the nodal field due to a solitary
point source. The free-space Greens functions can be derived from the periodic
Greens function by letting the size of the periodic patch tend to infinity, obtaining
j ,j
(5.3.8) Gi11,i22
j ,j
lim
N1 , N2
Hi11 , i22
j ,j
in two dimensions. Physically, Gi11, i22 represents the nodal value at the (i1 , i2 ) node due
to a source applied at the (j1 , j2 ) node. The corresponding lattice Greens function is
(5.3.9) Lm1 ,m2
lim
N1 ,N2
Mm1 ,m2 .
j ,j
(5.3.10) Gj11+m2 1 , j2 +m2 Lm1 ,m2 ,
irrespective of the location of the nodal source determined by j1 and j2 . The

normalized lattice Greens function, indicated by a tilde, is defined as
m ,m Lm ,m L0,0 .
(5.3.11) L
1 2
1 2
Analogous definitions are made in three dimensions.
The nodal distribution induced by a point source diverges at a logarithmic rate
with respect to distance from the point source in two dimensions, and it decays like
the inverse of the distance in three dimensions. This behavior is consistent with those
of the Greens function of Laplaces equation in an entire plane in two dimensions or
space in three dimensions, G , satisfying the forced Laplace equation
(5.3.12) 2 G + (x x0 ) = 0,
given by
(5.3.13) G =
1
r
ln ,
2
a
G=
1
4 r
in two or three dimensions, where r = |x x0 |, x is the position of a field point, x0

is the position of the point source, a is an arbitrary length, and is the Dirac delta
function in two or three dimensions.
C
C
A
B
C
FIGURE 5.3.1 Nearest neighbors (B) and second nearest neighbors (C) of a node (A) where
a source is applied on a honeycomb lattice.
Nearest Neighbors
Consider an infinite honeycomb lattice with identical link conductances, and assume that a point source with strength s is applied at a node labeled A, as shown
in Figure 5.3.1. Balancing the rates of transport of the entity associated with the corresponding nodal potential, , at that node and exploiting the inherent geometrical
symmetry of the honeycomb arrangement, we obtain
(5.3.14) c(B A ) + c(B A ) + c(B A ) + s = 0.
m ,m , we find that
In terms of the normalized lattice Greens function, L
1 2
nn ,
(5.3.15) B A = s L
where the subscript nn indicates the nearest neighbor. Making substitutions, we
obtain
1
d
nn = ,
(5.3.16) L
where d = 3 is the lattice coordination number.
In fact, expression (5.3.16) applies for any one-, two-, or three-dimensional simple lattice, provided that d is set equal to the lattice coordination number. We recall
that d = 2 for the one-dimensional lattice, d = 4 for the square lattice, and d = 6 for
the hexagonal (triangular) or simple cubic lattices.
Second Nearest Neighbor in the Honeycomb Lattice
In the particular case of the honeycomb lattice, but not more generally, we write a
balance at a nearest neighbor of the node where the point source is applied and obtain
nn 2L
snn = 0,
(5.3.17) 3 L
where the subscript snn indicates a second nearest neighbor, marked as node C in
Figure 5.3.1. Accordingly,
snn = 1 .
(5.3.18) L
2
This value will be confirmed by alternative methods in Section 5.7.
Exercise
5.3.1 Honeycomb lattice
Count the number of third and fourth nearest neighbors of a node on the honeycomb
lattice shown in Figure 5.3.1(b).
5.4 SQUARE LATTICE
Consider an infinite square lattice supporting a doubly periodic nodal field, as shown
in Figure 5.4.1. Each periodic test section contains N1 square cells in the first direction and N2 square cells in the second direction. The eigenvalues and eigenvectors of
the doubly periodic Laplacian are given in (3.1.36) and (3.1.41).
5.4.1 Periodic Greens Function
Substituting the eigenvalues and eigenvectors of the doubly periodic Laplacian

into the general expression (5.2.22), we derive the doubly periodic MoorePenrose
Greens function
(5.4.1)
j ,j
Hi11 ,i22

N1
N2

1
1 j1 ) n1 + (i2 j2 ) n2
exp i (i

=
,
2 1
2 1
4N1 N2
sin
+
sin
n1 = 1 n2 = 1
2 n1
2 n2
N2
i2
2
1
1
i1
N1
FIGURE 5.4.1 Illustration of a periodic patch of a square

lattice consisting of N1 links in the first direction and N2
links in the second direction.
where the prime indicates that the singular term, n1 = 1, n2 = 1, is excluded from the
sum, and
(5.4.2) n1 =
n1 1
2 ,
N1
n2 =
n2 1
2 .
N2
j ,j
We recall that Hi11 ,i22 is the field value at the (i1 , i2 ) node due to a source applied at
the (j1 , j2 ) node and its doubly periodic images. The first index parameterizes the first
direction, and the second index parametrizes the second direction.
In terms of the shift indices, m1 i1 j1 and m2 i2 j2 , we obtain the more
compact expression
(5.4.3) Mm1 ,m2

N1
N2

1
2 n2
exp
i m1n1 + m
,
=
2 1
2 1
4N1 N2
sin
+
sin
n1 = 1 n2 = 1
2 n1
2 n2
where Mm1 ,m2 is the MoorePenrose periodic lattice Greens function defined in
(5.3.4).
Fourier Coefficients
Defining p1 = n1 1 and p2 = n2 1, we obtain the equivalent representation
(5.4.4) Mm1 ,m2

N
1 1 N
2 1

1
1 k1 + m2 p2 k2
exp
i m1 p

,
=
4N1 N2
sin2 1 p k + sin2 1 p k
p1 = 0 p2 = 0
2 1 1
2 2 2
where the prime indicates that the singular term, p1 = 0, p2 = 0, is excluded from
the sum, and
(5.4.5) k1 =
2
,
N1
k2 =
2
N2
are directional wave numbers. This expression reveals that the Fourier coefficients of
the double Fourier series representing the Greens function are
(5.4.6) cp1 ,p2 =
except that c0,0 = 0.
1
1
,
8N1 N2 sin2 1 p k + sin2 1 p k
1
1
2
2
2
2
Normalized Greens Function
Next, we normalize the Greens function so that the point source generates a nodal
field that takes the reference value of zero at the application point. The normalized
Greens function, indicated by a tilde, is given by
N
1 1 N
2 1
1
2 p2 k2 )
1
cos (m1p1 k1 + m

(5.4.7) Mm1 ,m2 =
k1 k2
2
2 1
2 1
16
p1 = 0 p2 = 0 sin 2 p1 k1 + sin 2 p2 k2
or
m ,m =
(5.4.8) M
1 2
N
1 1 N
2 1
1
1 cos (m1 p1 k1 + m2 p2 k2 )
k
k
,
1
2
2
8
2 cos (p1 k1 ) cos (p2 k2 )
p1 = 0 p2 = 0
where k1 and k2 are the directional wave numbers defined in (5.4.5).

5.4.2 Free-Space Greens Function
In the limit N1 and N2 , the double sum in (5.4.7) or (5.4.8) reduces into
a double integral, yielding the normalized free-space Greens function of the infinite
square lattice,
m ,m = 1
(5.4.9) L
1 2
16 2
2
0
2
0
1 cos(m1 1 + m2 2 )
d1 d2
sin2 12 1 + sin2 12 2
or
m ,m = 1
(5.4.10) L
1 2
8 2
2
0
1 cos(m1 1 + m2 2 )
d1 d2 ,
2 cos 1 cos 2
where 1 and 2 are auxiliary integration variables. By construction,

0,0 = 0.
(5.4.11) L
By the fourfold symmetry of the square lattice, we have
m,0 = L
m,0 = L
0,m = L
0,m
(5.4.12) L
for any positive or negative integer, m. By symmetry across a diagonal line, we have
p,q = L
q,p
(5.4.13) L
for any pair of integers, p and q.
Alternative integral representations where the two terms in the argument of the
cosine in the numerator of the integrand are separated are
m ,m = 1
(5.4.14) L
1 2
4 2

0
1 cos(m1 1 ) cos(m2 2 )
d1 d2
sin2 12 1 + sin2 12 2
1 cos(m1 1 ) cos(m2 2 )
d1 d2 .
2 cos 1 cos 2
and
m ,m =
(5.4.15) L
1 2
1
2 2

0
Note that the limits of integration have been changed.

Using (5.4.10), we find that
1,0 + L
0,1 = 1
(5.4.16) L
8 2
2
d1 d2 ,
0
yielding
1,0 = L
1,0 = L
0,1 = L
0,1 = 1 ,
(5.4.17) L
4
in agreement with the general expression (5.3.16).
Horizontal and Vertical Profiles
Setting in (5.4.9) m2 = 0 or m1 = 0, we obtain a simplified expression for the nodal

profile in the first or second direction intercepting the nodal source:
m,0 = L
0,m = 1
(5.4.18) L
16 2
2
0
1 cos(mt)
dv dt,
sin 12 t + sin2 12 v
2
where v and t are two integration variables representing 1 and 2 , or vice versa.
Performing the integration with respect to v, we obtain
m,0 = L
0,m =
(5.4.19) L
1
8
1 cos(mt)
1/2 dt,
1
2 1
sin 2 t 1 + sin 2 t
which can be restated as

m,0 = L
0,m = 1
(5.4.20) L
4
where w = 12 t.
1 cos(2mw)
sin w(1 + sin2 w)1/2
dw,
For m = 1, we use a trigonometric identity and perform a straightforward

integration to recover (5.4.17):

1
sin w
1

(5.4.21) L1,0 = L0,1 =
dw = .
2
1/2
2 0 (1 + sin w)
4
For m = 2, we perform the integration to obtain
2,0 = G0,2 =
(5.4.22) L
sin w cos2 w
2
(1 + sin
w)1/2
dw = 1 +
2
.
Using (5.4.20), we find that

m,0 L
m1,0 = 1
(5.4.23) L
4
cos(2 mw) cos[2 (m 1)w]

sin w (1 + sin2 w)1/2 w
dw.
Simplifying the integrand, we obtain

m,0 L
m1,0 = 1
(5.4.24) L
2

0
sin[(2m 1)w]
(1 + sin2 w)1/2
dw.
Unfortunately, the definite integral can be found by analytical methods in terms of

elementary functions, and this prevents us from developing a recursion relation.
Diagonal Profile
A diagonal node corresponds to m1 = m2 = m, where m is arbitrary. Applying the balance equation defining the Greens function at the node labeled m1 = 1 and m2 = 0,
1,1 = L
1,1 and rearranging, we extract the first diagonal
noting that, by symmetry, L
value,
1,1 =
(5.4.25) L

1
2,0 = 1 .
4 L1,0 L
2
1,0 =
Note that this is higher in absolute value than the nearest-neighbor value, L

L0,1 = 1/4.
To obtain the diagonal profile of the Greens function, we set in (5.4.9) m1 =
m2 = m and derive the expression
m,m = 1
(5.4.26) L
16 2
2
0
1 cos[m (1 + 2 )]
d1 d2 ,
sin2 12 1 + sin2 12 2
which is equivalent to
m,m = 1
(5.4.27) L
8 2
2
0
1 cos[m (1 + 2 )]
d1 d2
2 cos 1 cos 2
or
m,m
(5.4.28) L
1
=
16 2
1 cos(m1 ) cos(m2 )
d1 d2 .
sin2 12 1 + sin2 12 2
It is helpful to introduce two new variables, w and v, such that

(5.4.29) 1 = w + v,
2 = w v,
where
(5.4.30) w = 12 (1 + 2 ),
v=
1
2
(1 2 ).
Substituting this transformation into (5.4.27), we obtain

2 2
1
1 cos(2mw)

(5.4.31) Lm,m =
dv dw.
2
16 0 0 1 cos w cos v
Performing the inner integration with respect to v, we obtain a simple integral
representation:
m,m = 1
(5.4.32) L
4
1 cos 2mw
dw.
sin w
When m = 1, we use a trigonometric identity and compute the integral by elementary

methods to recover (5.4.25).
Using (5.4.32), we find that the diagonal profile satisfies a simple one-term
recursion relation,

1
cos 2mw cos[2(m 1)w]

(5.4.33) Lm,m Lm 1, m 1 =
dw.
4 0
sin w
Using a trigonometric identity to simplify the integrand, we obtain

1

(5.4.34) Lm, m Lm 1, m 1 =
sin[(2m 1)w] dw
2 0
and then
m,m = L
m 1, m 1
(5.4.35) L
1
1
.
2m 1
Accordingly, we have
m,m =
(5.4.36) L
m
1 1
2q 1
q=1
0,0 = 0.
for m 1, where L
Subdiagonal Profile
For convenience, we define the elements of the subdiagonal line,

m + 1,m .
(5.4.37) Km L
Writing a balance at a diagonal node and exploiting the symmetry across the
diagonal, we obtain
m,m Km 1 ,
(5.4.38) Km = 2 L
where K0 = 1/4 is the value at the second nearest neighbor of the point source. For
example,
1,1 K0 =
(5.4.39) K1 = 2 L
2 1
+ .
4
Using (5.4.38), we obtain a recursion relation for the first subdiagonal array
m,m L
m 1,m 1 ) =
(5.4.40) Km Km 2 = 2 (L
or
(5.4.41) Km = Km 2
2
1
.
2m 1
Thus,
(5.4.42) Km =
m/2
1 2 1
4
4q 1
q=1
when m is even, and
(5.4.43) Km =
(m1)/2
1 2
1
4
6q 1
q=0
when m is odd.
2
1
2m 1
Recursive Relations
The Greens function can be built using relations originating from nodal balances
based on (a) the recursion relation (5.4.35) for the diagonal profile and (b) from the
recursion relation (5.4.38) for the subdiagonal profile.
For points at the first axis, m2 = 0, we have
m + 1,0 = 4 L
m,0 L
m 1,0 2L
m,1 .
(5.4.44) L
For any other point, we have
m + 1,m = 4 L
m ,m L
m 1, m L
m ,m + 1 L
m ,m 1 .
(5.4.45) L
1
2
1 2
1
2
1 2
1 2
An algorithm for building the nodal field based on these recursive relations is implemented in the Fortran code shown in Table 5.4.1(a). Note that indices of arrays are
allowed to take zero (or negative) values, assuming that these are declared at the beginning of the code. The output generated by the code is shown in Table 5.4.1(b), and
a graph of the Greens function is shown in Figure 5.4.2(a). Unfortunately, numerical
instability arises sufficiently far from the point source.
One-Dimensional Integral Representation
A useful expression for the Greens function arises by introducing a complex

variable, z, defined such that
(5.4.46) z exp(i 1 ),
cos 1 =
1
2

z+
1
z

,
d1 = i
dz
,
z
and recasting the integral representation (5.4.10) into the form

m ,m = i
(5.4.47) L
1 2
4 2
2
0
1 z|m1 | exp(i m2 w)
dz dw,
z2 2 (2 cos w)z + 1
where the closed integration path is the unit circle in the z plane and we have set
w = 2 [1, 5, 47].
The roots of the denominator of the fraction inside the integral provide us with
the pole of the integrand with respect to z. Setting z = e , we find that the real
number satisfies the equation
(5.4.48) cosh = 2 cos w.
For a pole to reside inside the unit circle, must be positive. A graph of as a
function of w is shown in Figure 5.4.3.
TABLE 5.4.1 (a) Fortran Code for Computing the Normalized Greens Function on an Infinite Square
Lattice by Recursion and (b) Output of the Code
(a)
pi = 3.14159265358979D0
mmax = 8
L(0,0) = 0.0D0
L(1,0) = -0.25D0
L(2,0) = -1.00D0+2.0D0/pi
Do m=1,mmax
! diagonal elements
L(m,m) = L(m-1,m-1) -1.0D0/pi/(2.0D0*m-1)

End Do
Do m=1,mmax-1
! first subdiagonal elements
L(m+1,m) = 2.0D0*L(m,m) - L(m,m-1)

End Do
Do m=1,mmax-1
! rest of the elements
L(m+1,0) = 4.0D0*L(m,0) - L(m-1,0)- 2.0D0*L(m,1)

Do l=1,mmax-m-1
L(m+l+1,l) = 4.0D0*L(m+l,l)-L(m+l-1,l)-L(m+l,l+1)-L(m+l,l-1)
End Do
End do
Do j=1,mmax
Do i=0,j-1
L(i,j)=L(j,i)
End Do
End Do
(b)
0.000 -0.250 -0.363 -0.430 -0.477 -0.513 -0.542 -0.567 -0.588
-0.250 -0.318 -0.387 -0.440 -0.482 -0.516 -0.544 -0.568 -0.589
-0.363 -0.387 -0.424 -0.462 -0.496 -0.525 -0.551 -0.573 -0.593
-0.430 -0.440 -0.462 -0.488 -0.514 -0.538 -0.560 -0.580 -0.599
-0.477 -0.482 -0.496 -0.514 -0.534 -0.553 -0.572 -0.590 -0.606
-0.513 -0.516 -0.525 -0.538 -0.553 -0.569 -0.585 -0.600 -0.615
-0.542 -0.544 -0.551 -0.560 -0.572 -0.585 -0.598 -0.611 -0.624
-0.567 -0.568 -0.573 -0.580 -0.590 -0.600 -0.611 -0.622 -0.634
-0.588 -0.589 -0.593 -0.599 -0.606 -0.615 -0.624 -0.634 -0.644
(a)
0
0.2
0.4
0.6
0.8
20
10
0
10
m2
20
10
20
10
20
m1
(b)
0
0.2
0.4
0.6
0.8
1
50
50
0
m2
0
50
50
m1
FIGURE 5.4.2 (a) Nodal distribution of the normalized

Greens function over a square lattice computed by recursive relations. (b) The entire distribution can be computed
by combining the recursive relations with the far-field
asymptotics. Numerical instability arises far from the
point source.
The residue of the integrand at a pole, R, arises by evaluating the ratio of the
numerator and the derivative of the denominator with respect to z at a pole, finding
(5.4.49) R =
1 exp(|m1 | + i m2 w)
1 exp(|m1 | + i m2 w)
=
.
2 (e 2 + cos )
2 sinh
Using the residue theorem, we find that

m ,m = i
(5.4.50) L
1 2
4 2
(2 i) R dw,
yielding
m ,m = 1
(5.4.51) L
1 2
2
1 exp(|m1 | + i m2 w)
dw,
sinh
2
1.8
1.6
1.4
1.2
1
0.8
0.6
0.4
0.2
0
0
0.2
0.4
0.6
0.8
w/(2)
FIGURE 5.4.3 Graph of the pole location, , against the integration variable, w, for computing the Greens function on a
square lattice.
provided that > 0. By symmetry, we also have

m ,m = 1
(5.4.52) L
1 2
2
1 exp(|m2 | + i m1 w)
dw.
sinh
Expressing the denominator of the integrand in (5.4.51) with respect to w using

(5.4.48), we obtain
m ,m = 1
(5.4.53) L
1 2
4
1 exp (|m1 | + i m2 w)
1/2 dw.
sin 12 w 1 + sin2 12 w
When m1 = 0, we recover precisely (5.4.20) for the horizontal or vertical profile.

A Mathematica script that computes the Greens function based on the integral
representation (5.4.51) adapted from Atkinson and van Steenwijk [1] is listed below:
sigma[omega_] := ArcCosh[2- Cos[omega]];
lgfs[m1_,m2_]:=
Simplify[
-1/(2*Pi)
*Integrate[
(1-Exp[-Abs[m1]*sigma[omega]]*Cos[m2*omega])/
Sinh[sigma[omega]]
,{omega,0,Pi}]
];
m ,m , Based
TABLE 5.4.2 Exact Values of the Normalized Greens Function on a Square Lattice, L
1
2
on the Computation of a One-Dimensional Integral for m1 , m2 = 0, 2, . . . .
14
1 + 2
17
+ 12
20 + 368
6
401
+ 1880
4
6
1 2
4
23
2 3
49 40
70 3223
15
4
3
2
14 3
118
3 + 15
1118
97
4 + 15
23
15
1 12
2 5
499
4 35
176
105
20
12 + 21
563
315
Note: The lower triangular part of this symmetric matrix arises by reflection.
Results generated by this script are shown in Table 5.4.2. These numerical predictions are consistent with those shown in Table 5.4.1 obtained by recursive
relations.
Far-Field Asymptotics
To study the behavior far from the source point, we note that, for small w, the solution
of the algebraic equation (5.4.48) is
(5.4.54) w + ,
where the three dots indicate higher-order terms. Accordingly, for large |m1 |, the
representation (5.4.51) yields

1
1 exp[(|m1 | + i m2 ) w
m ,m
L
dw
1 2
2
w
0

(5.4.55)
1
1
+
dw
sinh w
0
[5, 47]. Performing the integrations, we obtain the approximation
m ,m
(5.4.56) L
1 2
1
real [E ()] + ln 8 ln ,
2
2
where
(5.4.57) = (|m1 | i m2 )
and

(5.4.58) E (z)
1 et
dt
t
is the exponential integral. For large ||, we have
(5.4.59) real [E ()] ln || + E = 12 ln m21 + m22 + E + ln ,
where
(5.4.60) E = 0.577215665
is the Euler constant. Substituting this expression into (5.4.56), we obtain

m ,m
(5.4.61) L
1 2

1
1
ln m21 + m22 + E + ln 8 .
2
2
This asymptotic formula carries an error on the order of 104 around the edges of the
square |m1 | 16 and |m2 | 16.
Like the Greens function of Laplaces equation in two dimensions, the corresponding normalized lattice Greens function diverges at a logarithmic rate.
However, the lattice Greens function is zero at the forced node, whereas the Greens
function of Laplaces equation in two dimensions takes an infinite value at the
pole.
Comparing the asymptotic expression (5.4.61) with (5.4.36), we derive the
identity
(5.4.62)
lim
m

q =1
2
ln m = E + 2 ln 2.
2q 1
The sum on the left-hand side is an approximation to an integral computed by the

trapezoidal rule,
(5.4.63)
m

q=1
2
1
1

2q 1
2m 1

1
2
dx = ln(2m 1).
2x 1
The second and third terms on the left-hand side are included to render the weights
of the summed terms equal to 1/2 at the first and last points.
5.4.3 Helmholtz Equation Greens Function
Corresponding results can be derived for the generalized balance equation (4.6.13)
originating from the Helmholtz equation, repeated below for convenience:
(5.4.64) t fi, j

1
fi+1, j + fi1, j + fi, j1 + fi, j+1 = 0,
2
where t and are arbitrary coefficients (e.g., [31]).

The counterparts of the integral representations (5.4.14) and (5.4.15) for the freespace Greens function are
m ,m =
(5.4.65) L
1 2
1
8 2

0
t
2
1 cos(m1 1 ) cos(m2 2 )
d1 d2
1 + sin2 12 1 + sin2 12 2
and
m ,m = 1
(5.4.66) L
1 2
2

0
1 cos(m1 1 ) cos(m2 2 )
d1 d2 .
t
cos 1 cos 2
The Laplace Greens functions (5.4.14) and (5.4.15) arise for t = 4 and = 2.
Morita [31] developed a three-term recursive relation for the diagonal elements,
m,m ,
Lm L
(5.4.67) Lm+1
4m
=
2m + 1
t2
1
2 2
Lm
2m 1
Lm1 ,
2m + 1
where L0 = 0 and L1 is available in terms of complete elliptic integrals. When t = 4

and = 2, the term inside the parentheses is equal to unity. The nodal field can be
produced by recursion, as discussed in Section 5.4.2.
The asymptotic behavior of the Helmholtz lattice Greens function far from the
point source has been studied with reference to wave scattering (e.g., [29]).
5.4.4 Kirchhoff Greens Function
The periodic of free-space Greens function of the Laplace matrix corresponds to

networks with uniform link conductances, c. In the case of a square network with arbitrary conductances in the directions of the indices i1 and i2 , we use the eigenvalues
of the Kirchhoff matrix given in (4.5.3) or (4.5.4) and obtain the free-space Greens
function
m ,m = 1
(5.4.68) L
1 2
8 2
2
0
1 cos(m1 1 + m2 2 )
d1 d2 ,
1 + 2 1 cos 1 2 cos 2
where 1 and 2 are dimensionless conductance coefficients. The double integral can
be evaluated by numerical methods.
Exercise
5.4.1 Integral representation of the free-space Greens function
Derive the integral representation (5.4.53) from (5.4.51).
5.5 HEXAGONAL LATTICE
Consider a patch of a hexagonal lattice in its natural state supporting a doubly periodic nodal field, as shown in Figure 5.5.1. The base vectors of the underlying Bravais
lattice are
(5.5.1) a1 = a (1, 0) ,

a2 = a 12 1, 3 ,
where a is the distance between two nearest neighbors. Each periodic test section
contains N1 triangular cells in the first direction, and N2 triangular cells in the second
direction. The eigenvalues and eigenvectors of the doubly periodic Laplacian are
given in (3.3.36) and (3.3.21).
Working as in Section 5.4.1 for the square lattice, we derive the normalized periodic
Greens function
N1 1 N
2 1
k1 k2
1 cos (m1 p1 k1 + m2 p2 k2 )

(5.5.2) Mm1 ,m2 =
2 1
2 1
2 1
16 2
sin
p
k
+
sin
p
k
+
sin
(p
k
+
p
k
)
2 2
p1 = 0 p2 = 0
2 1 1
2 2 2
2 1 1
a2
N2
a1
i2
y
2
x
1
1
i1
N1
FIGURE 5.5.1 Illustration of a Periodic Patch of a Hexagonal Lattice Consisting of N1 triangles in the first direction and N2 triangles in the second
direction. The base vectors, a1 and a2 , determined the node numbering
scheme.
or
m ,m =
(5.5.3) M
1 2
N1 1 N
2 1
k1 k2
1 cos (m1 p1 k1 + m2 p2 k2 )

,
2
3 cos(p1 k1 ) cos(p2 k2 ) cos(p1 k1 + p2 k2 )
8
p1 = 0 p2 = 0
where
(5.5.4) k1 =
2
,
N1
k2 =
2
N2
The normalized free-space Greens function is given by the integral representation

m , m = 1
(5.5.5) L
1
2
16 2
2
0
2
2
sin
2
1
2 1
1 cos(m1 1 + m2 2 )

d1 d2
+ sin2 12 2 + sin2 12 (1 + 2 )
or
m ,m =
(5.5.6) L
1 2
1
8 2
1 cos(m1 1 + m2 2 )
d1 d2 .
3 cos 1 cos 2 cos(1 + 2 )
By symmetry, we have
m,0 = L
0,m = L
m,m = L
m,m
(5.5.7) L
for any positive or negative integer, m. Using (5.5.6), we find that
1,0 + L
0,1 + L
1,1 = 1
(5.5.8) L
8 2
2
d1 d2 ,
0
yielding the nearest-neighbor value

1
6
1,0 = L
0,1 = L
1,1 = L
1,1 = ,
(5.5.9) L
in agreement with the more general expression (5.3.16) for lattice coordination
number d = 6.
It is convenient to introduce two new variables, w and v, such that

(5.5.10) 1 = w + v,
2 = w v,
where
(5.5.11) w = 12 (1 + 2 ),
v=
1
2
(1 2 ).
Substituting these transformations into (5.5.6), we obtain

m ,m =
(5.5.12) L
1 2
1
8 2
2
0
1 exp[ i(m1 m2 )v] exp[ i(m1 + m2 )w]

dv dw,
3 cos(w + v) cos(w v) cos 2w
1 exp[ i(m1 m2 )v] exp[ i |m1 + m2 |w]

dv dw.
3 2 cos w cos v cos 2w
which can be rearranged into

m ,m = 1
(5.5.13) L
1 2
8 2
2
0
Next, we introduce a complex variable z, defined such that

(5.5.14) z exp( i v),
cos v =
1
2

z+

1
,
z
dv = i
dz
,
z
and find that

m ,m = i
(5.5.15) L
1 2
8 2
2
0
1 z|m1 m2 | exp[ i (m1 + m2 ) w]

dz dw,
z2 cos w z (3 cos 2w) + cos w
where the closed integration path is the unit circle centered at the origin of the z plane
[5].
The roots of the denominator provide us with the poles of the integrand with
respect to z, satisfying the equation

1
(5.5.16) z +
cos w = 3 cos 2w.
z
Setting z = e , we find that the real number satisfies a nonlinear algebraic
equation,
(5.5.17) cosh =
3 cos 2w 2 cos2 w
=
.
2 cos w
cos w
For a pole to reside inside the unit circle, must be positive. A graph of as a
function w is shown in Figure 5.5.2 in the interval [0, 12 ]. A solution cannot be
5
4.5
4
3.5
3
2.5
2
1.5
1
0.5
0
0
0.2
0.4
0.6
0.8
w/(2)
FIGURE 5.5.2 Graph of the pole location, , against the integration variable, w , for the hexagonal lattice. The dashed
line represents the linear dependence for small w.
found in the interval [ 12 , 32 ], and this means that the contour integral in (5.5.15)
is zero.
The residue of the integrand, R, arises by evaluating the ratio of the numerator
and the derivative of the denominator with respect to z at a pole, finding
1 z|m1 m2 | exp[ i (m1 + m2 )w]
.
2 e cos w 3 + cos 2w
(5.5.18) R =
Substituting 3 cos 2w = 2 cosh cos w into the denominator and simplifying, we

obtain
(5.5.19) R =
1 1 z|m1 m2 | exp[ i (m1 + m2 )w]

.
2
cos w sinh
Using the residue theorem, we find that

m ,m = i
(5.5.20) L
1 2
8 2
(2 i) R dw
and then
m ,m =
(5.5.21) L
1 2
1
2
/2
1 e|m1 m2 | cos[(m1 + m2 )w]

dw,
cos w sinh
provided that > 0 [1]. Eliminating the dependent variable from the denominator
of the integrand, we obtain
m ,m = 1
(5.5.22) L
1 2
/2
0
1 e|m1 m2 | cos[(m1 + m2 )w]

dw
[ (3 cos 2w)2 2 cos 2w 2 ]1/2
or
m ,m = 1
(5.5.23) L
1 2
/2
0
1 e|m1 m2 | cos[(m1 + m2 )w]

dw.
(cos2 2w 8 cos 2w + 7)1/2
Note that the denominator becomes zero at w = 0.

A radial profile arises by setting m1 = m2 = m, yielding
m,0 = L
0,m = L
m,m = L
m,m
L
/2
1
1 cos 2mw
=
dw
(5.5.24)
2
0
( cos 2w 8 cos 2w + 7 )1/2

1
1 cos(mv)
=
dv,
2
2 0 ( cos v 8 cos v + 7 )1/2
where v = 2w. A Mathematica script that computes the radial profile based on
(5.5.21) is presented next:
sigma[w_] := ArcCosh[2/Cos[w]- Cos[w]];
lgft[m1_,m2_] := Simplify[ -1/(2*Pi) *Integrate[
(1-Exp[-Abs[m1-m2]*sigma[w]]*Cos[(m1+m2)*tau])
/(Cos[w]*Sinh[sigma[w]]),{w,0,Pi/2}
]
];
lgft[1,1]
Results for the five nearest neighbors of the forced node, marked as nodes AE, are
shown in Figure 5.5.3(a) [1].
To study the behavior of the free-space Greens function far from the point source,
we note that, for small w, the solution of equation (5.5.17) is
(5.5.25)
3 w + ,
represented by the dashed line in Figure 5.5.2, where the three dots indicate higherorder terms. For large |m1 m2 |, the integral representation (5.5.21) yields

/2 1 exp |m m | 3 w + i (m + m ) w
1
2
1
2
m ,m 3
L
dw
1 2
6
w
0
(5.5.26)
#
/2
3
1
+
dw .
cos w sinh w
0
Performing the integrations, we obtain
m ,m 3 ( real [E ()] + 0.097723 ),

(5.5.27) L
1 2
6
where E it the exponential integral with complex argument defined in (5.4.58), and

(5.5.28) = |m1 m2 | 3 i (m1 + m2 )

.
2
For large ||, we have

(5.5.29) real [E ()] ln |m1 m2 |
3 i(m1 + m2 )

+ E,
2
yielding
(5.5.30) real [E ()] 12 ln m21 + m22 m1 m2 + E + ln ,
where E = 0.577215665 is the Euler constant. Substituting this expression into

(5.5.27), we obtain
m ,m
(5.5.31) L
1 2

3

ln m21 + m22 m1 m2 + 1.81967 .
6
Although this expression was derived under the assumption that |m1 m2 | is large, it
does apply for arbitrarily large |m1 | or |m2 |. The asymptotic formula provides us with
remarkably accurate results, as shown in the last column of Figure 5.5.3(a).
The entire nodal distribution of the Greens function is shown in Figure 5.5.3(b).
Like the Greens function of the square lattice, the Greens function for the hexagonal
lattice diverges at a logarithmic rate.
Exercises
5.5.1 Free-space Greens function
Confirm by numerical integration that the predictions of formula (5.5.24) are
consistent with those shown in figure 5.5.3(a) for m = 1, 2, 3.
5.5.2 Far field
Derive the counterpart of (5.5.31) for the point indexing scheme shown in
Figure 3.3.1.
5.6 MODIFIED UNION JACK LATTICE
Consider the modified Union Jack lattice shown in Figure 5.6.1, supporting a doubly
periodic nodal field. Each periodic test section contains N1 square cells in the first
direction and N2 square cells in the second direction inside each period. The eigenvalues and eigenvectors of the doubly periodic Laplacian were given in (3.4.13) and
(3.3.21).
D
(a)
D
D
C
D
B
B
C
C
A
B
D
D
C
C
D
Node
A
B
C
D
E
L1,0 = L0,1 = L1,1 = L1,1

L1,2 = L2,1
L2,0 = L0,2 = L2,2 = L2,2
L1,3 = L3,1 = L2,3 = L3,2
L3,0 = L0,3 = L3,3 = L3,3
Exact
1
1
3
43 + 2
5
5
2
27
24
2 +
61 = 0.1667
3 = 0.2180
3 = 0.2307
3 = 0.2566
3 = 0.2681
Asymptotic
0.1672
0.2177
0.2309
0.2566
0.2682
(b)
FIGURE 5.5.3 (a) Normalized free-space Greens function of the hexagonal

lattice at the five nearest neighbors of the point source, AE. (b) Nodal
distribution of the normalized Greens function over the hexagonal lattice.
Working as in Section 5.4.1, we derive the normalized MoorePenrose doubly

periodic Greens function
m1 ,m2 =
M
N1 1 N
2 1
k1 k2

16 2
p1 =0 p2 =0
(5.6.1)
sin
1
2 p1 k1

+ sin
1 cos (m1 p1 k1 + m2 p2 k2 )

+ sin2 12 (p1 k1 + p2 k2 ) + sin2 12 (p1 k1 p2 k2 )
1
2 p2 k2
N2
i2
2
1
1
i1
N1
FIGURE 5.6.1 Illustration of a periodic patch of a modified Union Jack lattice consisting of N1 square cells
in the first direction and N2 square cells in the second
direction.
or
(5.6.2)
m ,m = k1 k2
M
1 2
8 2
N
1 1 N
2 1
p1 = 0 p2 = 0
1 cos (m1 p1 k1 + m2 p2 k2 )
,
4 cos(p1 k1 ) cos(p2 k2 ) cos(p1 k1 + p2 k2 ) cos(p1 k1 p2 k2 )
where k1 = 2 /N1 and k2 = 2 /N2 are directional wave numbers, and the prime
indicates that the singular term, p1 = 0 and p2 = 0, is excluded from the double sum.
The normalized free-space Greens function is given by the integral representation

m1 ,m2 =
L
(5.6.3)

sin2
1
16 2
2
0
1
1 + sin2
2
2
0
1 cos(m1 1 + m2 2 )

d1 d2
1
2 1
2 1
2 + sin
(1 + 2 ) + sin
(1 2 )
2
2
2
or
m ,m = 1
L
1 2
8 2
(5.6.4)
2
0
2
0
1 cos(m1 1 + m2 2 )
d1 d2 .
4 cos 1 cos 2 cos(1 + 2 ) cos(1 2 )
m,0 = L
0,m ,
(5.6.5) L
m,m = L
m,m
L
for any positive or negative integer, m. Using (5.6.4), we find that

1,0 + L
0,1 + L
1,1 + L
1,1 = 1
L
8 2
2
d1 d2 ,
0
yielding
1,0 + L
1,1 = 1 .
(5.6.6) L
4
1,0 = L
1,1 .
However, it should be noted that L
It is convenient to write
(5.6.7) 1 = w + v,
2 = w v,
where
(5.6.8) w = 12 (1 + 2 ),
v=
1
2
(1 2 )
are two new variables. Substituting these transformations into (5.6.4), we obtain
m ,m =
(5.6.9) L
1 2
1
2 2

0
2 1
0
exp[ i (m1 m2 )v] exp[ i (m1 + m2 )w]

dv dw.
4 2 cos w cos v cos 2w cos 2v
Next, we introduce a complex variable z such that

(5.6.10) z exp( iv),
cos v =
1
2

1
z+
,
z
cos 2v =
1
2

1 2
z+
1,
z
dv = i
dz
z
and find that

m ,m = i
(5.6.11) L
1 2
2 2

0
1 z|m1 m2 | exp[ i (m1 + m2 ) w]

z(z2 + 1) cos w (5 cos 2w) z2 + 12 (z2 + 1)2
z dz dw,
where the closed integration path is the unit circle in the z plane.
The roots of the denominator provide us with the pole of the integrand with respect to z. Setting z = e , we find that the real number cosh = (z2 + 1)/(2z)
satisfies the equation
(5.6.12) 2 2 + 2 cos w 5 + cos 2w = 0.
Solving this quadratic equation and retaining the positive root, we obtain
1/2
(5.6.13) cosh = 12 cos w + 3 34 cos2 w
.
For a pole to reside inside the unit circle, must be positive. The residue of the
integrand, R, arises by evaluating the ratio of the numerator and the derivative of the
denominator divided by z at a pole, finding
(5.6.14) R =
1 z|m1 m2 | exp[ i (m1 + m2 )w]

,
2 e cos w 5 + cos 2w + 2( 2)
where sinh . Eliminating the expression 5 cos w in favor of the rest of the
terms in (5.6.12) and simplifying, we obtain
(5.6.15) R =
1 1 z|m1 m2 | exp[ i (m1 + m2 ) w]

.
2
cos w sinh + sinh 2
It is instructive to compare this residue with that shown in (5.5.19) for the hexagonal
lattice. Now using the residue theorem, we find that
m ,m = i
(5.6.16) L
1 2
2 2
(2 i) R dw
and thus
m ,m = 1
(5.6.17) L
1 2
2

0
1 e|m1 m2 | cos[(m1 + m2 )w]

dw,
sinh cos w + sinh 2
provided that > 0. Numerical computations show that

1,0 = 0.12101 . . . ,
(5.6.18) L
1,1 = 0.12899 . . . ,
L
in agreement with (5.6.6). The entire nodal distribution can be computed by

conventional numerical integration.
Exercise
5.6.1 Far field
Derive an expression for the far-field behavior of the free-space Greens function.
5.7 HONEYCOMB LATTICE
Consider a honeycomb lattice supporting a doubly periodic nodal field, as shown

in Figure 5.7.1. Each periodic test section contains N1 hexagonal cells in the first
direction and N2 hexagonal cells in the second direction. The nodes are distributed
on two triangular Bravais lattices, denoted as A and B, as discussed in Section 3.5.
Nodes on lattice A are shown as hollow circles connected by dashed lines, and nodes
on lattice B are shown as filled circles connected by dotted lines in Figure 5.7.1.
N2 + 1
N2 + 1
b
N2
N2
B
i2
y
iA
2
a2
2
1
a1
N1
i1B
2
iA
1
N1 + 1
N1
N1 + 1
FIGURE 5.7.1 Illustration of a periodic patch of a honeycomb lattice consisting of two

hexagonal lattices, A and B, containing N1 cells in the first direction and N2 cells
in the second direction. For the configuration shown, N1 = 4 and N2 = 3. Nodes
on lattice A are shown as open circles connected by dashed lines, and nodes on
lattice B are shown as filled circles connected by dotted lines.
Without loss of generality, we assume that the point source associated with the
Greens function is applied at a node on lattice A. The eigenvalues of the doubly
periodic Laplacian were given in (3.5.63) as

(5.7.1)
n1 ,n2 = 3 3 + 2 cos n1 + 2 cos n2 + 2 cos(n1 + n2 )
1/2
where
(5.7.2) n1 =
n1 1
2 ,
N1
n2 =
n2 1
2
N2
for n1 = 1, . . . , N1 and n2 = 1, . . . , N2 . The corresponding eigenvectors were given

in (3.5.38) and (3.5.65).
The normalized MoorePenrose doubly periodic Greens function on the constituent

lattice A is given by
m ,m )A =
(5.7.3) (M
1 2
N1
N2

1
1
1

+
1 exp[i(m1 n1 + m2 n2 ) ] ,
+
2N1 N2
n1 ,n2 n1 ,n2
n1 =1 n2 =1
where the prime indicates that the singular term, n1 = 1 and n2 = 1, is excluded from
the sum. Consolidating the two terms inside the sum, we obtain
m ,m )A =
(5.7.4) (M
1 2
N1
N2
+ + n1 ,n2

1
n1 ,n2
1 exp[i(m1 n1 + m2 n2 ) ] .
+
2N1 N2
n1 ,n2 n1 ,n2
n1 =1 n2 =1
Substituting expression (5.7.1) for the eigenvalues and simplifying, we obtain

N1
N2

1 exp[i m1 n1 + m2 n2 ]
3

m ,m ) =
(5.7.5) (M
1 2
2N1 N2
3 cos n1 cos n2 cos(n1 + n2 )
A
n1 =1 n2 =1
or
m ,m )A =
(M
1 2
(5.7.6)
N1
N2
3

4N1 N2
n1 =1 n2 =1
1 exp[i( m1 n1 + m2 n2 ) ]

.
1
1
1
sin2 n1 + sin2 n2 + sin2 (n1 + n2 )
2
2
2
Physically, this expression provides us with the nodal field generated at the point
( j1 + m1 , i2 + m2 ) of the constituent Bravais lattice A when a source is applied at the
0,0 )A = 0. Expression
point ( j1 , j2 ) of the same lattice. By construction, we obtain (M
(5.7.6) shows that the nodal values of the Greens function on lattice A are three times
those on a hexagonal lattice.
Lattice B
Working in a similar fashion for the nodes of the second constituent lattice B, we
obtain the periodic lattice Greens function
N1
N2
1
(Mm1 ,m2 ) =
2N1 N2
(5.7.7)
n1 =1 n2 =1
B
$
1
+n1 ,n2
1
1
1
+
3 +n1 ,n2 n1 ,n2 3 n1 ,n2
exp[i (m1 n1 + m2 n2 ) ] ( 1 + ei n1 + ei n2 ).
Physically, this expression provides us with the nodal field generated at the point
( j1 + m1 , i2 + m2 ) of the constituent Bravais lattice B when a source is applied at the
point ( j1 , j2 ) of lattice A. The second and fourth fractions inside the tall parentheses
originate from the eigenvectors of the doubly periodic Laplacian on lattice B.
We find that

1
1
1
1
1
1
=
+ 3 +n1 ,n2 n1 ,n2 3 n1 ,n2

D
3+ D 3 D
(5.7.8) n1 ,n2
2
1
=
=
,
9 D 3 cos n1 cos n2 cos(n1 + n2 )
1
where
(5.7.9) D = 3 + 2 cos n1 + 2 cos n2 + 2 cos(n1 + n2 ).
Substituting this expression into (5.7.7), we obtain

N1
N2
1
2N1 N2
(Mm1 ,m2 )B =
(5.7.10)
n1 =1 n2 =1

exp[i (m1 n1 + m2 n2 ) ]
1 + ei n1 + ei n2 .
The normalized Green function, indicated by a tilde, is given by

m ,m )B =
(5.7.11) (M
1 2
N1
N2
1
Am1 ,m2

,
2N1 N2
n1 =1 n2 =1
where
(5.7.12)
Am1 ,m2 = 3 cos(m1 n1 + m2 n2 ) cos[(m1 1) n1 + m2 n2 )]

cos[m1 n1 + (m2 + 1) n2 )].
We may confirm that

1
m 1,m )A + (M
m ,m +1 )A ,
(Mm1 ,m2 )A + (M
1
2
1 2
3
m ,m )B =
(5.7.13) (M
1 2
which is consistent with the linear equation defining the Greens function at the
(m1 , m2 ) node of lattice B.
The normalized free-space Greens function on lattice A is given by the integral

representation
m1 ,m2 )A =
(5.7.14) (L
3
16 2
2

0
sin2
1
2 1
1 cos(m1 1 + m2 2 )

d1 d2
+ sin2 12 2 + sin2 12 (1 + 2 )
or
m ,m )A =
(5.7.15) (L
1 2
3
8 2
2
0
2
0
1 cos(m1 1 + m2 2 )
d1 d2 .
3 cos 1 cos 2 cos(1 + 2 )
0,0 )A = 0. Comparing the representations (5.7.14) and

By definition, we have (L
(5.7.15) with those derived in Section 5.5, we find that the nodal values of the Greens
function on lattice A are three times those on a triangular lattice.
m,0 )A = (L
0,m )A = (L
m,m )A = (L
m,m )A
(5.7.16) (L
for any positive or negative integer, m. Using (5.7.15), we obtain
2 2
3
A

(5.7.17) (L1,0 + L0,1 + L1,1 ) = 2
d1 d2 ,
8 0 0
yielding
1,0 )A = (L
0,1 )A = (L
1,1 )A = (L
1,1 )A = 1 ,
(5.7.18) (L
2
which is the value at the second nearest neighbor derived earlier in (5.3.18).
Lattice B
Writing a balance equation at an arbitrary node of the constituent lattice B, we obtain

m ,m )B = 1 (L
m ,m )A + (L
m 1,m )A + (L
m ,m +1 )A .
(5.7.19) (L
1 2
1 2
1
2
1 2
3
This equation allows us to generate the nodal field on lattice B in terms of the nodal
field on lattice A. For m1 = 0 and m2 = 0, we obtain

0,0 )B = 1 (L
1,0 )A + (L
0,+1 )A = 1 .
(5.7.20) (L
3
3
1
3
0,0 )B = (L
1,0 )B = (L
0,1 )B = ,
(5.7.21) (L
which is the value at the nearest neighbor.
The Greens function on lattice B is given by the integral representation
2 2

B
1

Lm1 ,m2 =
16 2 0 0
Am1 ,m2 (1 , 2 )
(5.7.22)

d1 d2
2 1
2 1
2 1
sin
1 + sin
2 + sin
(1 + 2 )
2
2
2
or

m ,m B = 1
(5.7.23) L
1 2
8 2
2
0
Am1 ,m2 (1 , 2 )
d1 d2 ,
3 cos 1 cos 2 cos(1 + 2 )
where
(5.7.24)
Am1 ,m2 (1 , 2 ) = 3 cos(m1 1 + m2 2 )

cos[(m1 1) 1 + m2 2 )] cos[m1 1 + (m2 + 1) 2 )].
Summary
The first several nearest neighbors of a node on a honeycomb lattice are identified
in Figure 5.2.2(a). Nodes A, B, and C fall on lattice A involving the source point,
and nodes A , B , and C fall on lattice B. The corresponding values of the Greens
function are given in a table under the illustration in Figure 5.7.2(a). The entire
distribution of the Greens function is plotted in Figure 5.7.2(b).
(a)
C
B
B
B
C
C
C
A
A
A
A
C
No de
A
A
B
C
B
C
Value
(L0,0 )B
(L1,0 )A
(L1,1 )B =
(L2,0 )B =
(L2,1 )A
(L2,0 )A
1
3
1
3
2 LA + LB
LA + LB + LC
0.3
13 = 0.3333
12 = 0.5
1 3 = 0.5513
76 + 1 3 = 0.6153
1 3 3 = 0.6540
4 + 6 3 = 0.6920
(b)
FIGURE 5.7.2 (a) Normalized free-space Greens function of

the honeycomb lattice at the six nearest neighbors of a point
source applied at the central node. (b) Nodal distribution of
the normalized Greens function over the hexagonal lattice.
Exercises
5.7.1 Lattice B
Confirm by numerical integration that the integral representation (5.7.22) reproduces
(5.7.21).
5.7.2 Alternative node indexing
Derive expressions for the free-space Greens function for the point indexing scheme
shown in Figure 3.5.1.
5.8 SIMPLE CUBIC LATTICE
Consider a slab of a simple cubic network containing N1 links in the first direction,
N2 links in the second direction, and N3 links in the third direction, as shown in
Figure 5.8.1. The lattice is parametrized by three indices, i1 , i2 , and i3 running in
three perpendicular directions.
The triply periodic normalized MoorePenrose periodic Greens is given by

m ,m ,m = k1 k2 k3
M
1 2 3
32 3
(5.8.1)
N
3 1
1 1 N
2 1 N

p1 = 0 p2 = 0 p3 = 0
1 cos(m1 p1 k1 + m2 p2 k2 + m3 p3 k3 )

2 1
2 1
2 1
sin
p1 k1 + sin
p2 k2 + sin
p3 k3
2
2
2
i2
i1
i3
FIGURE 5.8.1 Illustration of a rectangular

slab of a simple cubic network containing
N1 links in the first direction, N2 links in the
second direction, and N3 links in the third direction. The conductances of all links are
assumed to be the same. In the configuration shown, N1 = 2, N2 = 2, and N3 = 1.
or
m ,m ,m = k1 k2 k3
(5.8.2) M
1 2 3
16 3
N
1 1 N
2 1 N
3 1

p1 = 0 p2 = 0 p3 = 0
1 cos(m1 p1 k1 + m2 p2 k2 + m3 p3 k3 )
,
3 cos(p1 k1 ) cos(p2 k2 ) cos(p3 k3 )
where
2
,
N1
(5.8.3) k1 =
k2 =
2
,
N2
k3 =
2
N3
are directional wave numbers, and the prime after the summation symbol indicates
that the troublesome term ( p1 = 0, p2 = 0, and p3 = 0) is excluded from the sum.
The free-space Greens function is given by the triple integral representation

2 2

m ,m ,m = 1
L
1 2 3
32 3
(5.8.4)
1 cos(m1 1 + m2 2 + m3 3 )

d1 d2 d3
2 1
2 1
2 1
sin
1 + sin
2 + sin
3
2
2
2
or
m ,m ,m = 1
(5.8.5) L
1 2 3
16 3
2 2
0
2
0
1 cos(m1 1 + m2 2 + m3 3 )
d1 d2 d3 .
3 cos 1 cos 2 cos 3
Alternative integral representations where the three terms in the argument of the
cosine in the numerator of the integrand are separated are
m ,m ,m = 1
L
1 2 3
4 3
(5.8.6)

0
1 cos(m1 1 ) cos(m2 2 ) cos(m3 3 )
d1 d2 d3
sin2 12 1 + sin2 12 2 + sin2 12 3
and
m ,m ,m = 1
L
1 2 3
2 3
(5.8.7)

0
1 cos(m1 1 ) cos(m2 2 ) cos(m2 3 )

d1 d2 d3 .
3 cos 1 cos 2 cos 3
Note the new limits of integration.

m,0,0 = L
0,m,0 = L
0,0,m
(5.8.8) L
for any positive or negative integer, m.
Using the integral representation (5.8.5), we find that the sum of the three nearestneighbor values is
1,0,0 + L
0,1,0 + L
0,0,1 =
(5.8.9) L
2 2
1
16 3
1
d1 d2 d3 = ,
2
yielding
1
6
1,0,0 = L
0,1,0 = L
0,0,1 = ,
(5.8.10) L
It is interesting to consider the nodal distribution in the (i1 , i2 ) plane corresponding to a fixed value of i3 . Setting in (5.8.5) m3 = 0, we obtain
1
2
m ,m ,0 =
(5.8.11) L
1 2

0
Fm1 ,m2 () d3 ,
where
1
(5.8.12) Fm1 ,m2 () 2
8
2
0
1 cos(m1 1 + m2 2 )
d1 d2
cos 1 cos 2
and 3 cos 3 . The integral representation (5.8.12) arises from the Greens
function for the square lattice given in (5.4.10) by replacing the 2 in the denominator
of the fraction of the integrand with .
Alternative representations of the Greens function have been developed in terms
of complete elliptic integrals and products [2123]. Unfortunately, an efficient
method for computing the Greens function is not available.
Far from the point source, in the limit as m1 or m2 or m3 tends to infinity, the
normalized Greens function tends to the asymptotic value

1
d1 d2 d3

(5.8.13) L =
.
16 3 0 0 0 3 cos 1 cos 2 cos 3
A detailed analysis shows that
2
=
(5.8.14) L
2

18 + 12 2 10 3 7 6 K 2 ()
or

31 2 1
2 11

= 0.25273 . . . ,
192
24
24
= 1
(5.8.15) L
3
where = (2

3)( 3 2), K is the complete elliptic integral of the first kind,

/2
(5.8.16) K()
dw
1 2 sin2 w
and is the Gamma function [14, 20, 21, 50]. In contrast with the logarithmic growth
of the Greens function of the square lattice, the normalized Greens function of the
cubic lattice tends to a constant value far from the point source.
Exercise
5.8.1 Numerical evaluation of a triple integral
Write a code that computes the triple integral in (5.8.6) using the trapezoidal rule and
confirm the value given in (5.8.10) (e.g., [35]).
5.9 BODY-CENTERED CUBIC (BCC) LATTICE
Using the results of Section 3.8, we find that the normalized free-space Greens function of the body-centered cubic network shown in Figure 5.9.1 is given by the integral
representation
m ,m ,m = 1
L
1 2 3
32 3
(5.9.1)
2 2
0
1 cos(m1 1 + m2 2 + m3 3 )

d1 d2 d3
1
1
1
1
sin2
1 + sin2
2 + sin2
3 + sin2
(1 + 2 + 3 )
2
2
2
2
or
(5.9.2)
m ,m ,m = 1
L
1 2 3
16 3
2 2
0
2
0
1 cos(m1 1 + m2 2 + m3 3 )
d1 d2 d3 ,
4 cos 1 cos 2 cos 3 cos(1 + 2 + 3 )
where the relative node indices, m1 , m2 , and m3 , correspond to the base vectors, a1 ,
a2 , and a3 , as shown in (3.8.1).
By symmetry,
m,0,0 = L
0,m,0 = L
0,0,m
(5.9.3) L
i1
i3
y
x
i2
FIGURE 5.9.1 Illustration of the body-centered cubic

(bcc) lattice. The nodes are parametrized by three indices, i1 ,i2 , and i3 , corresponding to the base vectors,
a1 , a2 , and a3 , shown in (3.8.1).
for any positive or negative integer, m. Following a procedure similar to that

described in previous sections, we find that we have
1
8
1,0,0 = L
0,1,0 = L
0,0,1 = ,
(5.9.4) L
The nodes can be identified by an alternative trio of primed indices,
(5.9.5) i1 = i1 + i2 + i3 ,
i2 = i1 i2 + i3 ,
i3 = i1 + i2 i3 ,
corresponding to a Cartesian base, as discussed in Section 3.8, The corresponding

lattice Greens function is
2 2 2
1 cos(m1 1 + m2 2 + m3 3 )
1

(5.9.6) Lm ,m ,m =
d1 d2 d3 ,
1 2 3
1 cos 1 cos 2 cos 3
64 3 0 0 0
where
m1 = m1 + m2 + m3 ,
(5.9.7)
m2 = m1 m2 + m3 ,
m3 = m1 + m2 m3 .

1 cos(m1 1 ) cos(m2 2 ) cos(m3 3 )
1

(5.9.8) Lm ,m ,m = 3
d1 d2 d3 .
1 2 3
1 cos 1 cos 2 cos 3
8 0 0 0
Exercise
Write a code that computes the triple integral in (5.9.1) using the trapezoidal rule and
confirm the value given in (5.9.4) (e.g., [35]).
5.10 FACE-CENTERED CUBIC (FCC) LATTICE
Using the results of Section 3.9, we find that the normalized free-space Greens
function of the body-centered cubic network shown in Figure 5.10.1 is given by the
integral representation
m ,m ,m = 1
(5.10.1) L
1 2 3
32 3
2 2
0
2
0
1 cos(m1 1 + m2 2 + m3 3 )
d1 d2 d3 ,
D
where
D = sin2
(5.10.2)
+ sin2
1
2

1 + sin2 12 2 + sin2 12 3 + sin2 12 (1 2 )
1
2

(2 3 ) + sin2 12 (3 1 ) ,
or
m ,m ,m =
(5.10.3) L
1 2 3
1
16 3
2 2
0
2
0
1 cos(m1 1 + m2 2 + m3 3 )
d1 d2 d3 ,
E
where
(5.10.4)
E = 6 cos 1 cos 2 cos 3

cos(1 2 ) cos(2 3 ) cos(3 1 ).
The node indices, m1 , m2 , and m3 (corresponding to i1 , i2 , and i3 ), are associated

with the base vectors a1 , a2 , and a3 , as shown in (3.9.1).
m,0,0 = L
0,m,0 = L
0,0,m
(5.10.5) L
i3
y
i1
x
z
i2
a
FIGURE 5.10.1 Illustration of the face-centered cubic (fcc) lattice.

The nodes are parametrized by three indices, i1 , i2 , and i3 ,
corresponding to the base vectors, a1 , a2 , and a3 shown in (3.9.1).
for any positive or negative integer, m. Following a procedure similar to that

described in previous sections, we find that
1,0,0 = L
0,1,0 = L
0,0,1 =
(5.10.6) L
1
,
12

The nodes can be identified by an alternative trio of indices, i1 , i2 , and i3
(corresponding to a different base), defined as
(5.10.7) i1 = i2 + i3 ,
i2 = i3 + i1 ,
i3 = i1 + i2 ,
as discussed in Section 3.9. The corresponding normalized lattice Greens function

is
2 2 2
m ,m ,m = 3
L
1 2 3
128 3 0 0 0
(5.10.8)
1 cos(m1 1 + m2 2 + m3 3 )

d1 d2 d3 ,
3 cos 1 cos 2 cos 2 cos 3 cos 3 cos 1
where
(5.10.9) m1 = m2 + m3 ,
m2 = m3 + m1 ,
m3 = m1 + m2 .
(5.10.10)
m ,m ,m = 3
L
1 2 3
16 3

0 0 0
1 cos(m1 1 ) cos(m2 2 ) cos(m3 3 )

d1 d2 d3 .
3 cos 1 cos 2 cos 2 cos 3 cos 3 cos 1
Exercise
Write a code that computes the triple integral in (5.10.2) using the trapezoidal rule
and confirm the value given in (5.10.6) (e.g., [35]).
5.11 FREE-SPACE LATTICE GREENS FUNCTIONS
In Section 5.2.6, we found that the normalized free-space Greens function associated
with the Laplacian matrix of an infinite lattice in one dimension is given by the
integral representation
m = 1 1
(5.11.1) L
d 2
2
0
1 cos(m)
d,
1 ()
where d = 2 is the lattice coordination number and () = cos is a structure

function arising from the eigenvalues of the periodic Laplacian.
The normalized free-space Greens function of a Bravais lattice in two dimensions admits the unified integral representation
m ,m =
(5.11.2) L
1 2
1 1
d 4 2
2
0
1 cos(m1 1 + m2 2 )
d1 d2 ,
1 (1 , 2 )
where d is the lattice coordination number and (1 , 2 ) is a structure function

arising from the eigenvalues of the doubly periodic Laplacian.
The normalized free-space Greens function of a Bravais lattice in three dimensions admits the unified integral representation
m ,m ,m = 1 1
(5.11.3) L
1 2 3
d 8 3
2 2
0
+ m2 2 + m3 3 )
d1 d2 d3 ,
1 (1 , 2 , 3 )
2 1 cos(m
0
1 1
where d is the lattice coordination number and (1 , 2 , 3 ) is a structure function

arising from the eigenvalues of the triply periodic Laplacian.
The structure function, , is tabulated in Table 5.11.1 for several Bravais lattices
along with the lattice coordination number, d. It should be noted that the form of
the structure function is not unique for each lattice type, but depends on the node
indexing scheme associated with the choice of base vectors.
The normalized free-space Greens function of a composite lattice consisting of
two or a higher number of Bravais lattices admits a more involved representation. For
the honeycomb lattice consisting of two interwoven hexagonal lattices, we found that
m ,m = 1
(5.11.4) L
1 2
8 2
2
0
2
0
1 cos(m1 1 + m2 2 )
d1 d2
1 (1 , 2 )
on the lattice hosting the nodal source, where

(5.11.5) (1 , 2 ) = 13 [cos 1 + cos 2 + cos(1 2 ) ] .
5.11.1 Probability Lattice Greens Function
Suppose that a random walker wanders over the nodes of a Bravais lattice, starting
at a node labeled m = 0 in one dimension, m1 = 0 and m2 = 0 in two dimensions, or
m1 = 0, m2 = 0, and m3 = 0 in three dimensions. The walker stays in its current position with probability 1 z, and jumps indiscriminantly to one of its nearest neighbors
with probability z.
In one dimension, we introduce the the probability lattice Greens function
1
(5.11.6) Pm (z) =
2
2
0
cos(m)
1
d =
.
1 z()
1 z2
In two dimensions, we introduce the probability lattice Greens function

(5.11.7) Pm1 ,m2 (z) =
1
4 2
2
0
2
0
cos(m1 1 + m2 2 )
d1 d2 .
1 z(1 , 2 )
In three dimensions, we introduce the probability lattice Greens function

(5.11.8) Pm1 ,m2 ,m3 (z) =
1
8 3
2 2
0
2
0
cos(m1 1 + m2 2 + m3 3 )
d1 d2 d3
1 z(1 , 2 , 3 )
(e.g., [16]). These representations are strikingly similar to those of the free-space
Greens function discussed earlier in this section.
The probability that the walker returns to the origin after any number of steps is
(5.11.9) 0 (z) = 1
1
,
P0 (z)
where
(5.11.10) P0 (z) =
1
4 2
2
0
2
0
d1 d2
1 z(1 , 2 )
TABLE 5.11.1 Tabulation of the Structure Function, , of Several Lattices with Coordination Number d.
Lattice
One-dimensional
cos
Square
Hexagonal
Modied Union Jack
Simple cubic
Body-centered cubic (bcc)
1 (cos + cos )
1
2
2
1 cos + cos + cos( )
1
2
1
2
3
1 cos + cos + cos( + )
1
2
1
2
4

+ cos(1 2 )

1 cos + cos + cos
1
2
3
3
1 cos + cos + cos
1
2
3
4
+ cos(1 + 2 + 3 )
Body-centered cubic (bcc)
Face-centered cubic (fcc)
12
Face-centered cubic
12
cos 1 cos 2 cos 3

1 cos + cos + cos
1
2
3
6
+ cos(1 2 ) + cos(2 3 )

+ cos(3 1 )
1 cos cos + cos cos
1
2
2
3
3

+ cos 3 cos 1
Note: The minus or plus sign in the argument of the cosine for the hexagonal and honeycomb
lattices apply when the base vectors in the natural state form a 60 or 120 angle.
in two dimensions. Similar equations can be written in two, three, and higher
dimensions. For example, in the case of the square lattice, we have
(5.11.11) P0 (z) =
2
K(z),
where K is the complete elliptic integral of the first kind. Since P0 (1) is infinite in
two dimensions, 0 (1) = 1, which shows that the walker is certain to return to the
origin after an unspecified number of steps.
Of particular interest is the Taylor series expansion of the lattice Greens function
with respect to z. In two dimensions, we have
(1)
(2)
2
(5.11.12) Pm1 ,m2 (z) = a(0)
m1 ,m2 + am1 ,m2 z + am1 ,m2 z + .
(n)
The coefficients, am1 ,m2 , are the probabilities that a walker starting at (0, 0) is located
at (m1 , m2 ) after n steps. The structure function is given in terms of the single-step
probabilities by the Fourier expansion
(5.11.13) (1 , 2 ) =

a(1)
m1 ,m2 exp [ i(m1 1 + m2 2 ) ],
m1 m2
where i is the imaginary unit. Similar interpretations apply in one and three
dimensions.
(1)
For example, in the case of the square lattice, all am1 ,m2 are zero, except that
(1)
(5.11.14) a1,0 = 14 ,
(1)
a0,1 = 14 ,
expressing equal probabilities in four directions. Applying (5.11.13) reproduces the

structure function shown in the second entry of Table 5.11.1.
In one dimension, we find that
(5.11.15)
(n)
a0
1
= n
2
n
n/2

=
1
n!
n
2 [(n/2)!]2
if n is even, and an = 0 if n is odd, where the large parentheses denote the

combinatorial and the exclamation mark denotes the factorial.
Exercise
5.11.1 Eigenvalues of the Laplacian
Confirm by numerical computation the coefficients given in (5.11.15).
5.12 FINITE DIFFERENCE SOLUTION IN TERMS OF GREENS FUNCTIONS
In Section 4.5, we saw that the Laplacian matrix arises from the finite difference
discretization of the Laplacian of an unknown function of two variables in two
dimensions, f (x, y), or three variables in three dimensions, f (x, y, z), on a uniform
Cartesian finite difference grid. The Kirchhoff matrix arises from corresponding
discretizations on a nonuniform or non-Cartesian grid.
As an example, we consider a uniform Nx Ny Cartesian grid described by two
indices, i and j, covering a rectangular area in the xy plane, as shown in Figure 5.12.1.
Our objective is to compute a numerical solution of the Poisson equation,
(5.12.1) 2 f + g(x, y) = 0,
where g(x, y) is a given distributed source. Using the five-point formula to approximate the Laplacian at the (i, j) node, we obtain the finite difference equation
(5.12.2) 2 (1 + )fi, j fi+1, j fi1, j (fi, j1 + fi, j+1 ) = x2 gi, j ,
where = (y/x)2 (e.g., [35]). In the case of a square grid, x = y, we set = 1.
The finite difference solution at the (i, j) node can be expressed as a linear superposition of the fields due to (a) boundary nodal sources with a priori unknown
strength sij and (b) interior nodal sources with strength x2 gi,j . In the case of Laplaces equation, g = 0, only boundary nodal sources are employed. Introducing the
normalized free-space Greens function of the infinite square lattice corresponding
m ,m , we write
to the prevailing value of , L
1 2
(1)
(2)
(3)
(4)
(5)
(5.12.3) fij = fij + fij + fij + fij + fij ,
where
(1)
(5.12.4) fij
Nx

i p,j 1
sp,1 L
p =1
Ny
y
j
y
1
1
x
Nx
i
x
FIGURE 5.12.1 A Cartesian finite difference grid used to

solve the Poisson equation in two dimensions.
is a nodal distribution at the bottom,
(5.12.5)
(2)
fij
Ny

i Nx 1,j q
sNx +1,q L
q =1
is a nodal distribution at the right,
(3)
(5.12.6) fij
Nx

i p,j N 1
sp,Ny + 1 L
y
p =1
is a nodal distribution at the top,
(5.12.7)
(4)
fij
Ny

i 1,j q
s1,q L
q =1
is a nodal distribution at the left, and
(5.12.8)
(5)
fij
= x
Ny
Nx

i p,j q
gp,q L
p =2 q =2
is a nodal distribution in the interior of the solution domain. Thanks to the linearity of
the governing equations, expression (5.12.3) satisfies the difference equation (5.12.3)
for any boundary and interior source terms.
The representation in terms of the lattice Greens function is inspired by the
integral representation of the solution of Laplaces equation in terms of the corresponding Greens function. The boundary source terms, sij , must be computed to
satisfy the boundary conditions around the four edges of the rectangle.
In most applications, we impose the Dirichlet boundary condition, specifying the
boundary values of f , or the Neumann boundary condition, specifying the normal
derivative. After the solution has been found, the nodal field can be reconstructed
from the point source distribution. The efficiency of this approach hinges on the
availability of the lattice Greens function.
In the presence of a source term, the computational cost for assembling the linear
system is proportional to the product Nx Ny , and the computational cost for solving
the linear system is proportional to the sum Nx + Ny . In the absence of a source term,
both costs are proportional to Nx + Ny .
(a)
0.6
0.4
1
0.2
0.5
sn
0.2
0.5
0.4
1
30
0.6
20
40
(b)
2
30
20 25
10 15
i1
30
20 25
10 15 i
1
30
20 25
15
10
i1
20
0.8
60
n
80
100
i2
120
10
x 103
0
0.08
2
0.06
sn
0.04
0.02
0
30
10
20
i2
12
20
40
80
100
120
0.5
0.4
0.3
0.2
0.1
0
0.1
0.2
0.3
0.4
0.5
1
0.5
sn
(c)
60
n
10
0.5
1
30
i2
20
40
60
n
80
100
120
20
10
FIGURE 5.12.2 Finite-difference solution of the Laplace or Poisson equation computed in terms
of the square lattice Greens function. The strength of nodal sources around the four edges
of a square domain starting from the southwestern point and moving along the bottom,
right, top, and left is shown in the left column, where n is a node count. The finite difference
solution of (a) Laplaces or (b) Poissons equation is shown in the right column. (c) Same as
(a) but with the zero-flux condition along the bottom.
Dirichlet Boundary Condition
Assume the the boundary values of f are prescribed as a Dirichlet boundary condition. Applying (5.12.3) at the boundary nodes and rearranging the difference
equations, we formulated a system of linear equations with a unique solution for the
2 (Nx + Ny ) unknown strengths of the boundary point sources sij . The linear system
can be solved by a direct or iterative method.
A solution of Laplaces equation, gp,q = 0, on a 32 32 grid is shown in
Figure 5.12.2(a). In this case, the boundary conditions specify half a sinusoidal wave
along the bottom and top sides and a full sinusoidal wave along the left and right
sides with equal amplitude.
A solution of Poissons equation with uniform source term gp,q = 1/Nx2 and
the homogeneous Dirichlet boundary condition on a 32 32 grid is shown in
Figure 5.12.2(b). Physically, the nodal distribution describes the velocity profile of
Poiseuille flow inside a square duct or the deformed shape of an elastic membrane
attached to a square frame.
Neumann Boundary Condition
To implement the Neumann boundary condition specifying the normal derivative,

we approximate the normal derivative with a one-side finite difference with a desired
degree of accuracy (e.g., [35]). The emerging algebraic equation is then used to compute the boundary nodal sources, as in the case of the Dirichlet boundary condition
discussed in the previous section. A pertinent finite difference solution is shown in
Figure 5.2.2(c). Other types of boundary conditions can be handled in similar ways.
Exercise
5.12.1 Constant field
Compute a finite difference solution with the Dirichlet boundary condition specifying the same boundary values around the four edges of a square. Discuss the
computed boundary source distribution.
/// 6 ///
NETWORK PERFORMANCE
The efficiency and performance of a conductive or convective

network depends on the node connectivity and conductance of the individual links.
Of particular interest is the pairwise resistance determining the rate of transport of
a suitable entity associated with a nodal potential across an arbitrary pair of nodes.
The pairwise resistance also admits a probabilistic interpretation in the context of
random walks. The sum of all pairwise resistances over all possible sets of nodes
provides us with an overall measure of the network performance. Link disruption or
clipping weakens a network, whereas link addition improves the performance of a
network. Pertinent concepts and quantitative measures are discussed in this chapter
for isolated and embedded networks.
6.1 PAIRWISE RESISTANCE
Suppose that a transported entity associated with a scalar nodal potential, , is supplied at a rate s at the ith node and withdrawn at the same rate from the jth node of a
network, as illustrated in Figure 6.1.1. The induced difference in the potential across
s
+s
j
i
FIGURE 6.1.1 A transported entity, such as

heat, is supplied at a rate s at the ith node
and withdrawn at the same rate from the jth
node of an embedded or isolated network.
The dashed lines connect selected nodes to
external Dirichlet nodes.
220
N e t w o r k P e r f o r m a n c e / / 221
this pair of nodes can be used to define a corresponding pairwise resistance, in that,
the more pathways connecting the two nodes, the lower the associated pairwise resistance. Nodes belonging to disconnected parts of an unconnected network register
an infinite pairwise resistance.
In the case of heat or mass transfer through a conductive network of rods or
conduits, is the temperature or species concentration and s is the rate of heat or
mass transport. In the case of fluid flow through a capillary tube network, is the
pressure and s is the volumetric or mass fluid rate. In the case of electricity transport,
is the electrical voltage and s is an electrical current. Physically, the pairwise
resistance arises when the ith node is connected to a positive battery pole or Ohm
meter, while the jth node is connected to the negative battery pole or Ohm meter,
or vice versa. The Ohm meter will register a resistance that depends on the overall
structure of the network.
Consider an embedded network where the Dirichlet nodes are grounded to zero
slpotential. Using (5.1.8), we express the nodal field induced by a source applied
the ith node and a sink applied at the j node in terms of the Greens function matrix,
G , as

(6.1.1) = rs G e(i) e(j) ,
where e(i) and e(j) are unit vectors, and
(6.1.2) r
1
c
is a reference resistance associated with a reference conductance, c. The difference

in the induced nodal field at the source and sink is

(6.1.3) i j = e(i) e(j) ,
yielding
(6.1.4) i j = rs e(i) e(j) G e(i) e(j) .
This equation motivates defining the dimensionless pairwise resistance

(6.1.5) Rij

i j
(i) (j)
= e e
G e(i) e(j) .
rs
Carrying out the multiplications, we obtain

(6.1.6) Rij = e(i) G e(i) + e(j) G e(j) e(i) G e(j) e(j) G e(i) ,
yielding
(6.1.7) Rij = Gii + Gjj Gij Gji .
Taking into consideration the symmetry of the Greens function, Gij = Gji , we obtain
(6.1.8) Rij = Gii + Gjj 2 Gij
for i = j. In vector notation, the N N pairwise resistance matrix is given by

(6.1.9) R = ( ) G + G ( ) 2 G ,
where the N-dimensional vector and N N matrix are filled with ones,
denotes the tensor product, and G is the diagonal part of G .
Identities
Premultiplying (6.1.9) by the augmented Kirchhoff matrix, K, and recalling that, by

definition, K G = I and also K = , as discussed in Section 4.4, we obtain
(6.1.10) K R = ( ) G + K G ( ) 2 I,
where I is the identity matrix. Postmultiplying this equation by K, we obtain

(6.1.11) K R K = ( ) G K + K G ( ) 2 K.
We recall that the vector contains the scaled conductances of links connecting
network nodes to external Dirichlet nodes.
Spectral Expansion
Substituting into (6.1.8) the spectral expansion of the Greens function given in
(5.1.13), we obtain
N

1
(n) (n)
(n) (n)
(6.1.12) Rij =
ui uj
ui uj
,
n
n=1
where an asterisk denotes the complex conjugate [56].
Representation in Terms of the Normalized Greens Function
Expression (6.1.8) can be rearranged as

(6.1.13) Rij = (Gij Gii ) (Gij Gjj ),
yielding
ij G
ji ,
(6.1.14) Rij = G
ij is the normalized Greens function, defined such that G
ii = 0, where
where G
summation is not implied over the repeated index, i. We recall that the normalized
ij is not necessarily equal to
Greens function is not necessarily symmetric, that is, G
ji . In contrast, Gij is always equal to Gji .
G
A network is isolated in the absence of Dirichlet nodes, T = 0 and K = K, where K is

the Kirchhoff matrix. Introducing the MoorePenrose Greens function, H, working
as in Section 6.1.1 for an embedded network, and using (5.2.16), we obtain
(6.1.15) Rij = (e(i) e(j) ) H (e(i) e(j) ),
yielding
(6.1.16) Rij = Hii + Hjj 2 Hij .
In vector notation, we have

(6.1.17) R = ( ) H + H ( ) 2 H,
where the N-dimensional vector and the N N matrix are filled with ones,
denotes the tensor product, and H is the diagonal part of H. Simplifications occur
in the case of an infinite regular lattice where the diagonal components of H are all
equal.
Identities
Useful identities can be derived from (6.1.17). Premultiplying (6.1.17) by the

Kirchhoff matrix, K, and recalling that K = 0 and K H = I , we obtain
(6.1.18) K R = K H ( ) 2 I ,
where the matrix

(6.1.19) I I
1

N
was introduced in (5.2.6), and I is the N N identity matrix. Equation (6.1.18) is the
counterpart of (6.1.10).
Postmultiplying (6.1.18) by I , recalling that I = 0, and noting that I 2 = I ,
we obtain
(6.1.20) K R I = 2 I .
Because all matrices involved in this equation are symmetric, we can write

(6.1.21) trace K R I = trace K I R = trace K R ,
yielding

(6.1.22) trace K R) = 2 trace(I ) = 2 (N 1).
We conclude that [55]

(6.1.23) trace K R = K : R = 2 (N 1)
where the colon denotes the double dot product, that is, the sum of the products of
corresponding elements of the two matrices on either side.
Postmultiplying equation (6.1.18) by K, we obtain
(6.1.24) K R K = 2 K.
Postmultiplying this equation by an arbitrary symmetric matrix, S, we obtain

(6.1.25) K R K S = 2 K S,
which is the counterpart of (6.1.11). Because all matrices involved in this equation
are symmetric, we can write

trace K R K S = trace K S K R
(6.1.26)
= (K S K) : R = 2 trace(K S).
A chain of identities can be derived by setting S = Kn , where n is a positive integer.

Spectral Expansion
Substituting into (6.1.16) the spectral expansion of the Greens function given in
(5.2.22), we obtain
N

1
(s) (s)
(s) (s)
(6.1.27) Rij =
ui uj
ui uj
,
s
s=2
where an asterisk denotes the complex conjugate. Note that summation begins at
s = 2 to skip the zero eigenvalue, 1 = 0.
Representation in Terms of the Normalized Greens Function
Expression (6.1.16) can be rearranged into

(6.1.28) Rij = (Hij Hii ) (Hij Hjj ),
yielding
ij H
ji ,
(6.1.29) Rij = H
ij is the normalized Greens function defined such that H
ii = 0, where
where H
summation is not implied over the repeated index, i. We recall that the normalized
ij is not necessarily equal
Greens function is not necessarily symmetric, that is, H
ji .
to H
Complete Network
In the case of a complete network with identical link conductances, c, we use the
MoorePenrose Greens function given in (5.2.42) and find that
(6.1.30) Rij =
2
N
for any nodal pair, i and j. This is the minimum possible pairwise resistance for any
uniform network.
6.1.3 One-Dimensional Network
Substituting into the general expression (6.1.16) the MoorePenrose Greens function for a one-dimensional isolated network with uniform conductances, given in
(5.2.26), we obtain the pairwise resistance
(6.1.31) Rij =
1
2N
N

s=2
cos
1
2
2
s cos j 12 s
,

1
sin 2 s
where s = (s 1) /N. It can be shown by algebraic manipulation, or else confirmed

by numerical computation, that
(6.1.32) Rij = |i j|,
in agreement with physical intuition [56]. The same result can be obtained by substituting into (6.1.29) the corresponding normalized MoorePenrose Greens function
given in (5.2.28).
6.1.4 One-Dimensional Periodic Network
Substituting into the general expression (6.1.16) the MoorePenrose Greens function for a one-dimensional periodic network with uniform conductances, given in
(5.2.30), we obtain the corresponding pairwise resistance
(6.1.33) Rij =
N
1 exp(i i s ) exp(i j s ) 2

,
4N
sin 1
s=2
where s = 2(s1) /N. It can be shown by algebraic manipulation, or else confirmed

by numerical computation, that
(6.1.34) Rij =

1
|i j| N |i j| ,
N
in agreement with physical intuition [56]. The same results is obtained by substituting into (6.1.29) the corresponding normalized periodic MoorePenrose Greens
function given in (5.2.34).
6.1.5 Infinite Lattices
In the case of an infinite regular lattice with a uniform coordination number d in one,
two, or three dimensions, the diagonal components of the Greens function matrix
are equal. Expression (6.1.8) for the pairwise resistance simplifies to

(6.1.35) Rij = 2 Gii Gij .
This relation applies for the square, hexagonal, modified Union Jack, honeycomb,
cubic, or any other appropriate lattice.
In terms of the normalized Greens function, denoted by a tilde, we obtain the
simplified expression
ij .
(6.1.36) Rij = 2 G
Since all pairwise resistances are positive, every component of the normalized lattice
Greens function must be negative.
Using expression (5.3.16) for the nearest-neighbor Greens function, we obtain
the nearest-neighbor pairwise resistance
(6.1.37) Rnn =
2
,
d
where d is the lattice coordination number. For example, d = 2 for the onedimensional lattice, d = 4 for the square lattice, and d = 6 for the hexagonal
(triangular) or simple cubic lattice.
6.1.6 Triangle Inequality
The pairwise resistance obeys a triangle inequality stating that

(6.1.38) Rij Rik + Rkj ,
where i, j, and k is an arbitrary triplet of nodes [8, 26]. In the case of an embedded
network, the inequality implies that
ij + G
ji G
ik + G
ki + G
kj + G
jk .
(6.1.39) G
In the case of an isolated network, the inequality implies that
ij + H
ji H
ik + H
ki + H
kj + H
jk .
(6.1.40) H
For a uniform infinite lattice, we have
ij G
ik + G
kj .
(6.1.41) G
6.1.7 Random Walks
The pairwise resistance of an isolated network admits a physical interpretation in the

context of random walks. With reference to the Kirchhoff matrix, K, we define the
probability that a compulsory random walker jumps from the ith to the jth node,
(6.1.42) pi, j
Ki,j
,
Ki,i
provided that pi, i = 0. By construction,

(6.1.43)
N

pi, j = 1,
j=1
as required. Note that pi, j is not necessarily equal to pj, i .

The first passage probability, P, is defined as the probability that the random
walker starts at node and reaches node before returning to . It can shown that
(6.1.44) P, =
1
K, R,
(e.g., [38]). In the case of uniform conductances, K, is the degree of the node.
Exercise
6.1.1 One-dimensional networks
(a) Confirm (6.1.32) by numerical computation. (b) Repeat for (6.1.34).
6.2 MEAN PAIRWISE RESISTANCE
To assess the overall transport efficiency of a network, we require a global measure

of the node pairwise resistance. One such measure is the mean resistance, defined as
the scaled sum of all elements of the pairwise resistance matrix, R,
(6.2.1) Rmean
N
N
N1
N
1 1
Rij =
Rij .
2N
N
i=1
j=1
i=1
j=i+1
The product N Rmean is sometimes called the effective network resistance or the
resistance distance (e.g., [9, 26]).
The mean pairwise resistance is a mathematically sound and physically intuitive
measure of the overall network robustness (e.g., [8, 9]). Complete networks are the
most robust and tree networks are the least robust connected networks, in agreement
with physical intuition.
Substituting into (6.2.1) expression (6.1.8) for an embedded network or expression (6.1.16) for an isolated network, and noting that the sum of elements in each
row or column of G or H is zero, we obtain
(6.2.2) Rmean = trace(G )
for an embedded network or

(6.2.3) Rmean = trace(H)
for an isolated network.

If a network is unconnected, containing fragments or secluded islands of nodes,
the pairwise resistance of nodes residing in two different fragments or inside and
outside an island is infinite, and the mean resistance is not defined.
6.2.1 Spectral Representation
Since the trace of a matrix is equal to the sum of its eigenvalues, we have
(6.2.4) Rmean =
N

1
s
s=1
for an embedded network, where s are the eigenvalues of the modified Kirchhoff
matrix defined in (4.4.6). For an isolated network,
(6.2.5) Rmean
N

1
=
,
s
s=2
where s are the eigenvalues of the Kirchhoff matrix. Note that the zero eigenvalue
is excluded from the sum in (6.2.5).
Based in (6.2.5), we derive the inequality
(6.2.6)
1
N1
< Rmean
.
2
2
As expected, when 2 = 0, the mean pairwise resistance is infinite. Other tighter

bounds of the mean pairwise resistance are available (e.g., [49]).
Substituting into (6.2.1) formula (6.1.12) for an embedded network or formula
(6.1.27) for an isolated network, along with comparing the resulting expression with
(6.2.4) or (6.2.5), we find that
N
N

(s) (s) 2
u u = N,
(6.2.7)
i
j
i=1 j=i+1
where s = 1, . . . , N for an embedded network or s = 2, . . . , N for an isolated network.

6.2.2 Complete Network
In the case of a complete network with uniform link conductances, c, we substitute

into (6.2.1) the pairwise resistances given in (6.1.30) and obtain
(6.2.8) Rmean =
N1
< 1.
N
Precisely the same result is obtained by substituting into (6.2.5) the eigenvalues of
the graph Laplacian matrix given in (2.2.11). The mean resistance of a complete
network is lower than that of any other network with the same number of nodes.
6.2.3 One-Dimensional Isolated Network
In the case of a one-dimensional isolated network with uniform conductances, we

substitute into (6.2.1) the pairwise resistances given in (6.1.32) and obtain
(6.2.9) Rmean
N1
N
N1
1
1
=
(j i) =
(N i + 1)(N i)
N
2N
i=1
j=i+1
i=1
or
(6.2.10) Rmean =
N1
1
1
p (p + 1) = (N 2 1).
2N
6
p=1
We observe that the mean resistance increases as the square of the number of nodes,
N. In the case of two nodes, N = 2, connected by one link, Rmean = 1/2.
Substituting into (6.2.5) the eigenvalues of the Laplace matrix given in (1.7.2),
we obtain
(6.2.11) Rmean =
N
1
1
.
2 s1
4
sin
s=2
2N
Comparing (6.2.10) with (6.2.11), we derive the identity

N1

(6.2.12)
m=1
sin
1
m
=
2 2
(N 1),
3
=
1
(2N 2 + 1).
3
2N
which can be recast into the form
(6.2.13)
N

2
m=1
sin
1
m
2N
6.2.4 One-Dimensional Periodic Network
In the case of a one-dimensional periodic network with uniform link conductances,

we substitute into (6.2.1) the pairwise resistances given in (6.1.34) and obtain
(6.2.14) Rmean

N1
N
1
ji
=
(j i) 1
.
N
N
i=1
j=i+1
Computing the inner sum, we find that
(6.2.15) Rmean =
N1
1
(N i + 1)(N i)(N + 2i 1),
6N 2
i=1
which can be summed to

(6.2.16) Rmean =
1
(N 2 1).
12
The mean resistance of a one-dimensional periodic network is half that of a onedimensional isolated network.
Substituting into (6.2.5) the eigenvalues of the Laplacian matrix given in (1.8.2),
we obtain
(6.2.17) Rmean =
N
1
1
.
2 s1
4
sin
s=2
N
Comparing (6.2.16) with (6.2.17) we derive the identity
(6.2.18)
N1

1

sin2 m
N
m=1
=
1 2
(N 1).
3
6.2.5 Periodic Lattice Patches
In Chapter 3, we derived expressions for the eigenvalues of the Laplacian of several

lattice patches in isolated or periodic configurations. For any two-dimensional periodic lattice whose eigenvalues, n1 , n2 , are parametrized by two indices, n1 and n2 ,
the mean resistance is
(6.2.19)
N1
N2

Rmean 2D =
n1 , n2
n1 =1 n2 =1
where the integers N1 and N2 determine the size of the periodic patch and the prime
after the summation symbol indicates that the zero eigenvalue, n1 = 1 and n2 = 1, is
excluded from the sum. For a three-dimensional lattice, we obtain the corresponding
expression
(6.2.20)
N3
N1
N2

Rmean 3D =
n1 =1 n1 =1 n3 =1
1
,
n1 , n2 , n3
where the prime has a similar meaning. Expressions for the eigenvalues are shown
in Table 6.2.1 for several lattices, where
(6.2.21) n1 =
n1 1
2 ,
N1
n2 =
n2 1
2 ,
N2
n3 =
n3 1
2 .
N3
Composite lattices consisting of dual or multiple Bravais lattices are treated in

special ways. In the case of the honeycomb lattice discussed in Section 3.5, we have
N1
N2

(6.2.22) Rmean 2D =
n1 =1 n2 =1
1
n1 ,n2
1
+n1 ,n2
%
,
TABLE 6.2.1 Eigenvalues of Periodic Patches of Several Lattices with Coordination Number d
Lattice
n1 ,n2 or n1 ,n2 ,n3
Square
4 2 cos n1 2 cos n2
Hexagonal
6 2 cos n1 2 cos n2 2 cos(n1 n2 )
Modied Union Jack
8 2 cos n1 2 cos n2 4 cos n1 cos n2
Honeycomb

1 13 cos n1 + cos n2 + cos(n1 n2 )
Kagom
18 6 cos n1 6 cos n2 6 cos(n1 n2 )

Simple cubic
6 2 cos n1 2 cos n2 2 cos n3
bcc

2 cos(n1 + n2 + n3 )
fcc
12
2 cos(n1 n2 ) 2 cos(n2 n3 )
2 cos(n3 n1 )
Note: The plus or minus sign applies for different node indexing schemes associated with a
different set of base vectors.
which can be reduced into (6.2.19) with

(6.2.23) n1 ,n2 = 1 13 cos n1 + cos n2 + cos(n1 n2 ) .
The plus or minus sign correspond to different node numbering schemes associated
with different base vectors.
In the case of the kagom lattice discussed in Section 3.6, we have
N1
N2

(6.2.24) Rmean 2D =
n1 =1 n2 =1
1
n1 ,n2
1
n1 ,n2
1
+n1 ,n2
%
,
which can be reduced into (6.2.19) with

(6.2.25) n1 ,n2 =
18 6 cos n1 6 cos n2 6 cos(n1 n2 )

.
The plus or minus sign correspond to different node numbering schemes associated
with different base vectors.
Graphs of the mean resistance scaled by the number of nodes inside each period,
N, are shown in Figure 6.2.1(a) for two-dimensional lattices with N1 = N2 on a
linear-logarithmic scale. As the size of the periodic unit increases, N , the
scaled mean resistance tends to a well-defined limit,
(6.2.26) Rmean N,
where the coefficient depends on the lattice type. For simple Bravais lattices, the
coefficient decreases as the lattice coordination number becomes higher due to the
(a)
0.22
square (d=4)
hexagonal (d=6)
mod Union Jack (d=8)
honeycomb (3)
kagome(4)
0.2
0.18
Rmean/N
0.16
0.14
0.12
0.1
0.08
0.06
0.04
0.02
0.5
1.5
2.5
log N1
(b)
0.4
simple cubic (d=6)

bcc (d=8)
fcc (d=12)
0.35
Rmean/N
0.3
0.25
0.2
0.15
0.1
0.05
0.5
1.5
2.5
log N1
FIGURE 6.2.1 (a) Dependence of the scaled mean resistance

of periodic two-dimensional lattices with dimensions N1 =
N2 and (b) periodic three-dimensional lattices with dimensions
N1 = N2 = N3 .
availability of a higher number of conductive pathways. The dual honeycomb and

kajom lattices exhibit a higher scaled effective resistance.
Graphs of the scaled mean resistance for three-dimensional lattices with N1 =
N2 = N3 , shown in Figure 6.2.1(b), exhibit a similar behavior. We may conclude that
the coefficient is a sensible index of the efficiency of lattice transport.
Exercises
6.2.1 Trigonometric identity
Confirm identity (6.2.13) by direct numerical evaluation.
6.2.2 Tree network
Compute the mean pairwise resistance of a tree network with N = 4 nodes.
6.3 DAMAGED NETWORKS
Consider an arbitrary network involving L links with arbitrary conductances, and

assume that the conductances of M L links, numbered ms for s = 1, . . . , M,
are perturbed from the unperturbed value, cms , to a perturbed value indicated by a
prime, cms .
For example, the network shown in Figure 6.3.1 has L = 19 total links, M = 10
damaged links drawn with thin lines labeled
(6.3.1) m1 = 3,
m2 = 18,
...,
m3 = 5,
m10 = 8,
and L M = 9 intact lines drawn with heavy lines.

Our goal is to assess the effect of these perturbations on the overall performance
of the network. In Section 6.4, we will consider the complementary problem of link
addition.
17
19
18
16
13
15
14
12
3
1
10
11
6
4
FIGURE 6.3.1 Illustration of a network

with L = 19 total links, M = 10 damaged
links (thin lines), and L M = 9 intact
lines (heavy lines.)
6.3.1 Damaged Kirchhoff Matrix
The Kirchhoff matrix of an isolated network after the M links have been altered is
given by
(6.3.2) K = K0 +
M

ms (ms ) (ms ) ,
s=1
where a superscript 0 indicates the unperturbed state

(6.3.3) j
1
(c cj ) = j (j 1)
c j
are dimensionless coefficients, c is a reference conductance, j cj /cj = 1 is the

ratio of the perturbed to the unperturbed conductance of the perturbed link labeled
j, j cj /c, and (j) is the jth column of the pristine oriented incidence matrix, R0 ,
before link removal. Specifically, the N-dimensional vector (j) is null, except that
(j)
(j)
(6.3.4) kj = 1,
lj = 1,
where kj is the label of the first end node and lj is the label of the second end node of
the jth link.
Unperturbed links make trivial contributions to the right-hand side of (6.3.2). If
three links labeled 7, 9, and 14 are removed, then we have M = 3, m1 = 3, m2 = 9,
and m3 = 14.
It is useful to introduce a rectangular N M matrix holding in its columns the
vectors corresponding to the perturbed links,
(6.3.5) V = (m1 )
..
.
(ms )
..
.
(mM ) .
The N L matrix V encompassing all links, M = L, is the oriented incidence matrix

of the network, R. In the case of selected damaged links, we obtain a reduced N M
incidence matrix referring to the set of defective links.
Moreover, it is useful to introduce an M M diagonal matrix,
(6.3.6) Z
m1
0
..
.
0
0
m2
..
.
..
.
0
0
..
.
0
0
..
.
0
0
mM1
0
0
mM
under the stipulation that i = 1 so that the matrix Z is nonsingular. This means that
undamaged links are not allowed into the matrices V and Z.
The matrices V and Z are defined such that the sum on the right-hand side of
(6.3.2) is given by the matrix product V Z VT , so that
(6.3.7) K = K0 + V Z VT .
Neither the unperturbed nor the perturbed Kirchhoff matrix is invertible.

An expression analogous to (6.3.7) can be written for the modified Kirchhoff matrix
of an embedded network,
(6.3.8) K = K0 + V Z VT ,
provided that links connecting network nodes to Dirichlet nodes are not disrupted.
The inverse of the unperturbed modified Kirchhoff matrix, K, is the corresponding
Greens function, G ,
(6.3.9) K G = I,
where I is the N N identity matrix.

Using the generalized Woodbury formula discussed in Appendix B, we find that
the Greens function after perturbation is given by

(6.3.10) G = G 0 I V (Z1 + )1 VT G 0 ,
where I is the N N identity matrix, is an M M matrix with elements

(6.3.11) pq = (mp ) w(mq )
for p, q = 1, . . . , M, and the vector w(mq ) satisfies the linear system

(6.3.12) K0 w(mq ) = (mq ) .
We can write
(6.3.13) = VT W = VT K0 V,
where
(6.3.14) W = w(m1 )
..
.
w(ms )
..
.
0
w(mM ) = G V.
By construction, the matrix W is symmetric for any network topology due to the symmetry of the lattice Greens function. Physically, the pq component of expresses
the difference in the nodal values across the mp damaged link due to a point-source
dipole applied across the mq damaged link. In terms of the unperturbed Greens
function, we have
(6.3.15) pq = Gl0p ,lq + Gk0p ,kq Gl0p ,kq Gk0p ,lq ,
where kp and lp are the end nodes of the pth link and kq and lq are the end nodes of
the qth link. The diagonal components,
(6.3.16) pp = Gl0p ,lp + Gk0p ,kp 2 Gl0p ,kp ,
express the difference in the nodal values across a damaged link due to a point-source
dipole applied across the same link. In terms of the normalized Greens function,
0 G
0 .
(6.3.17) pp = G
lp ,kp
kp ,lp
We recall that in the case of an infinite regular lattice, but not more generally, we
have pp = 2/d, where d is the lattice coordination number.
Subject to the preceding definitions, we have
(6.3.18) G = G 0 W (Z1 + )1 WT .
It is interesting that nodal field differences corresponding to damaged links, but not
intact links, appear in the final expressions for the Greens function in the perturbed
state, G .
Perturbation Nodal Field
It is useful to introduce the matrix

(6.3.19) P W (1 + )1 VT
and obtain
(6.3.20) G = (I + P) G 0 .
The perturbation nodal field, denoted by a prime, is given by

(6.3.21) = P 0 ,
where the superscript 0 denotes the unperturbed field corresponding to the pristine
network.
6.3.3 One Damaged Link
In the case of one damaged link, M = 1, connecting nodes k and l in the pristine state,
the matrices Z and are scalars, yielding
(6.3.22) G = G 0
( 1)
ww
1 + ( 1)
and
(6.3.23) P =
( 1)
w v,
1 + ( 1)
where the coefficients and and the vectors and w are associated with the
perturbed link,
(6.3.24) w = G 0 ,
and
(6.3.25) w = wl wk = G 0 = R0kl .
Physically, the scalar represents the nodal difference across the link when electrical
current is supplied at the first node of the link and withdrawn from the second node
of the link in the pristine state. In terms of the unperturbed Greens function,
0
0
(6.3.26) wi = Gi,l
Gi,k
and
0
0
0
0 G
0 ,
(6.3.27) = Gl,l
+ Gk,k
2 Gl,k
= G
l,k
k,l
where k and l are the end points of the perturbed link.

Using expression (6.1.8), we find that the pairwise resistance matrix in the
perturbed state is given by
(6.3.28) Rij = R0ij
( 1)
(w2 + w2j 2wi wj )
1 + ( 1) i
or
(6.3.29) Rij = R0ij
( 1)
(wi wj )2 .
1 + ( 1)
The second term on the right-hand side expresses the effect of the perturbation. For
the resistance to increase when = 0, the denominator must be positive, and this
requires that 1. The equality applies in the case of a one-dimensional network.

Applying (6.3.29) for i = k and j = l, setting = R0kl , and rearranging, we obtain
(6.3.30) Rkl =
R0kl
1 + ( 1) R0kl
As the conductance of the altered ring increases, , the effective resistance

tends to zero.
An Infinite Regular Lattice with One Damaged Link
In the case of an infinite homogeneous regular lattice, = 1, we obtain =

2/d, where d is the lattice coordination number. Consequently, the altered Greens
function is
(6.3.31) G = G 0
1
2
1 + ( 1)
d
w w,
the altered projection matrix is

1
(6.3.32) P =
1 + ( 1)
2
d
w v,
and the altered pairwise resistance matrix is

(6.3.33) Rij = R0ij
1
2
1 + ( 1)
d
(wi wj )2 .
Using the altered projection matrix, we obtain

(6.3.34) v =
1
(v w) (v 0 ).
2
1 + ( 1)
d
Substituting once again v w = 2/d and simplifying, we obtain

(6.3.35) v =
1
1
d+ 1
2
(v 0 ).
Physically, v is the difference in the perturbation field and v 0 is the difference

in the unperturbed field across the defective link.
6.3.4 Clipped Links
In the case of a network whose links have the same conductance in the pristine state,
i = 1 for i = 1, . . . , N, and M clipped links with zero conductance in the perturbed
state, ms = 0 for s = 1, . . . , M, we find that Z = IM , where IM is the M M identity
matrix. Accordingly, we have
(6.3.36) Q 1 + = IM .
When the matrix Q is singular, the projection matrix P does not exist and the disturbance nodal field is not defined. Physically, isolated nodes or clusters of nodes
unconnected to their neighbors are encountered inside the network. The number of
these isolated groups is equal to the number of zero eigenvalues of the matrix Q.
Eigenvalues equal to 1 correspond to isolated nodes or clusters of nodes attached to
the Dirichlet nodes.
In the case of isolated networks, we use (5.2.14) and compute the MoorePenrose
Greens function of the perturbed network
(6.3.37) H = H0 W (1 + )1 WT ,
where
(6.3.38) W = H0 V,
= VT H0 Y,
subject to the preceding definitions for embedded networks.
Exercise
6.3.1 Perturbed network
Derive the matrix V corresponding to the damaged network shown in Figure 6.3.1.
6.4 REINFORCED NETWORKS
The analysis of Section 6.3 can be adapted to address the effect of link addition,
intended to strengthen or reinforce a network.
Consider the addition of one link labeled L + 1 with conductance cL+1 = c
anchored at nodes labeled k and l of an embedded network, as shown in Figure 6.4.1.
The Greens function matrix after link addition is given by
(6.4.1) G = G 0
w w,
1 +
k
l
M+1
FIGURE 6.4.1 Illustration of a reinforced embedded network

with one added link anchored at the kth and lth nodes, drawn
as a heavy line.
the nodal projection matrix providing us with the perturbation field due to link
addition is given by
(6.4.2) P =
w v,
1+
and the pairwise resistance matrix is given by

(6.4.3) Rij = R0ij
(wi wj )2 ,
1+
where i, j = 1, . . . , N. The vector is null, except that the lth entry is equal to 1 and
the kth entry is equal to 1. The vector w = G 0 and scalar are given in (6.3.26)
and (6.3.27) in terms of the unperturbed Greens function. Physically, the scalar
(6.4.4) = w = wl wk R0kl
represents the difference in the induced potential across the added link when current
is supplied at the first node of the link and withdrawn from the second node of the
link in the pristine state. Applying (6.4.3) for i = k and j = l, we obtain [8]
(6.4.5) Rkl =
R0kl
1 + R0kl
To address the general case of L added links, we introduce a rectangular N L

matrix holding in its columns the vectors corresponding to the added links,
(6.4.6) V = (L+1)
..
.
(i)
..
.

(L+L ) ,
and the L L diagonal matrix
(6.4.7) Z
L+1
0
..
.
0
0
L+2
..
.
..
.
0
0
..
.
0
0
..
.
0
0
L+L 1
0
L+L
under the stipulation that i = 1 so that the matrix Z is nonsingular.

The matrices V and Z are defined such that the Kirchhoff matrix after reinforcement is
(6.4.8) K = K0 + V Z VT .
Neither the original nor the reinforced Kirchhoff matrix is invertible. However, the
corresponding modified Kirchhoff matrices, K and K0 , are invertible. The concepts
and formulas discussed in Section 6.3 for link damage also apply to link addition
with sensible modifications.
Exercise
6.4.1 Reinforced lattices
Explain how a square network (d = 4) can be transformed into a hexagonal network
(d = 6) with systematic link addition.
6.5 DAMAGED LATTICES
In Section 6.3, we discussed the performance of arbitrary damaged networks and

derived general expressions for the Greens function and pairwise resistance. In this
section, we consider the particular case of networks configured as infinite regular
lattices.
6.5.1 One Damaged Link
Consider an infinite square lattice where all links have the same conductance, c, except that one defective link extending between nodes labeled A and B has a different
conductance, c , as shown in Figure 6.5.1. We are interested in assessing the effect
of the defect on the nodal distribution of a potential, , associated with a transported
entity.
A balance of the transported entity at node labeled A requires that
(6.5.1) c (B A ) + c (C A ) + c (D A ) + c (E A ) = 0.
G
B
c
C
E
A
D
FIGURE 6.5.1 Illustration of scalar transport

through an infinite square network of resistors arranged on a square lattice. The conductance of one link is different than that of
all other links.
(6.5.2) c(B A ) + c(C A ) + c(D A ) + c(E A ) + s = 0,
where the term

(6.5.3) s (c c) (B A )
is regarded as an a priori unknown nodal source applied in a pristine network with

uniform conductance, c, at node numbered A. A similar balance at node labeled B
requires that
(6.5.4) c (A B ) + c (F B ) + c (G B ) + c (H B ) = 0.
(6.5.5) c(A B ) + c(F B ) + c(G B ) + c(H B ) s = 0.
The solution of the linear system that arises by writing balance equations at all
nodes can be decomposed into a homogeneous solution, 0 , a particular solution,
(1) , due to the source (sink) in equation (6.5.2), and another particular solution,
(2) , due to the sink (source) in equation (6.5.5). The nodal value at an arbitrary
node X is
(1)
(2)
(6.5.6) X = X0 + X + X .
Considering the nodal field (1) , we note that, by symmetry, the flow rate of the
transported field is divided into d = 4 equal flow rates upon entering node A, where
d is the lattice coordination number. Consequently,
(1)
(1)
(6.5.7) s = dc A B
A similar conclusion can be reached regarding the field (2) , yielding
(2)
(1)
(6.5.8) s = dc A B
Now using equation (6.5.6), we obtain
(1)
(1)
(6.5.9) A B = A0 B0 + A B

(2)
(2)
+ A + B .
Substituting the preceding expressions for the particular solutions and rearranging,
we find that
2 s

2 c c
A B .
c

(6.5.10) AB
(A B ) A0 B0 =
=
d c d
Solving for the difference across the defective link, we obtain

(6.5.11) AB A B =
0

1
A B0 .

2 c c
1+
d c
Consequently,
(6.5.12)

AB
0
AB
1
,
+
where c /c and
(6.5.13) = 12 d 1,
which is positive since d = 4. Expression (6.5.12) is consistent with the more general
result stated in (6.3.35).
In fact, expressions (6.5.12) and (6.5.13) apply for any one-, two-, or threedimensional regular network consisting of links with equal conductances, provided
that the coefficient d is set equal to the lattice coordination number [24, 25]. In the
case of a one-dimensional lattice, d = 2, in the case of a honeycomb lattice, d = 3,
in the case of a square lattice, d = 4, and in the case of a hexagonal (triangular) or
simple cubic lattice, d = 6.
6.5.2 Effective-Medium Theory
Assume that a defective link with conductance c occurs with probability density
function (c ). The expected value of the coefficient defined in (6.5.12) is

(6.5.14) <
>=
(c ) (c ) dc.
To be consistent with the imposed boundary conditions far from the defective link,
we require that <
>= 0 and invoke the definition of to obtain an algebraic
equation for c,

(6.5.15)
0
c c
(c ) dc = 0.
c + c
In the case of a binary distribution with two possible link conductances c = c0

and c0 , we set
(6.5.16) (c ) = (1 q) (c c0 ) + q (c c0 ),
where is a specified positive coefficient, q the number density of links with

conductance c0 , and is the Dirac delta function. Also setting c = c0 , we obtain
(6.5.17)
1

(1 q) +
q = 0,
+1
+
which can be rearranged into a quadratic equation for the dimensionless coefficient ,

(6.5.18) 2 ( + 1)(1 q) 1 + ( + 1) q 1
= 0.
Substituting the value of from (6.5.13), we find that

(6.5.19)
2 1
1

d
d
(1
q)
1
+
d
q
1
= 0.
2
2
2
1
The positive root of this quadratic equation provides us with a rational estimate for
the effective conductivity of the network.
6.5.3 Percolation Threshold
When = 0, corresponding to disrupted links, and lattice coordination number

d > 2, equation (6.5.19) has the uninteresting root = 0 and the interesting root
(6.5.20) = 1
+1
q
or
1
d (1 q) 1
(6.5.21) = 2 1
,
2
d1
which is plotted in Figure 6.5.2 for several lattices. We find that = 0 at the
approximate percolation threshold
(6.5.22) pc = 1 qc
2
.
d
The fraction on the left-hand side is the ratio of number of nodes to the number
of links, N/L, according to (2.1.6). Considering the heuristic nature of the effective
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0.2
0.4
0.6
0.8
FIGURE 6.5.2 Coefficient determining the effective conductivity of a network with a binary distribution of conductances.
The solid line is for the honeycomb lattice (d = 3), the dashed
line is for the square lattice (d = 4), and the dotted dashed
line is for the hexagonal lattice (d = 6). The symbols on the
q axis represent percolation thresholds, qc = 2 sin( /18)
0.3473 for the honeycomb lattice (circle), qc = 0.5 for the
square lattice (square), qc = 1sin(/18) 0.6527 for the hexagonal (diamond), and qc = 0.7512 for the simple cubic
lattice () [28, 44, 52].
medium theory, the predictions of the critical threshold for complete disruption are
remarkably accurate.
Exercise
6.5.1 Effective medium theory
Derive the counterpart of (6.5.18) for three types of links with conductances c0 , 1 c0 ,
and 2 c0 , occurring with probabilities 1 q1 q2 , q1 , and q2 .
6.6 DAMAGED SQUARE LATTICE
Consider transport through an infinite square lattice whose nodes are parametrized by
two indices, i1 and i2 , as shown in Figure 6.6.1. All links have the same conductance,
c, except for two unrelated defective links that have different conductances, c and
c . Our objective is to assess the effect of the defects on the nodal distribution of a
transported field, . For simplicity, we assign the labels AD to the end points of the
defective links, as shown in Figure 6.6.1.
Without loss of generality, we may assume that the first defective link with
conductance c is horizontal, extending between two nodes labeled (n1 , n2 ) and
(n1 + 1, n2 ). When the defective links are parallel, the second defective link with
conductance c extends between nodes (m1 , m2 ) and (m1 + 1, m2 ), as shown in
Figure 6.6.1(a). When the defective links are perpendicular, the second defective link
with conductance c is subtended between nodes (m1 , m2 ) and (m1 , m2 + 1), as shown
in Figure 6.1.1(b).
For any relative defective link orientations, a balance of the transported entity
associated with the potential at each end node of the first defective link requires
that
c (B A ) + c (A1 A ) + c (A2 A ) + c (A3 A ) = 0,
(6.6.1)
c (A B ) + c (B1 B ) + c (B2 B ) + c (B3 B ) = 0.

(a)
(b)
i2
i2
C1
C2
m2
B3
A1
n2
C
C3
c
A2
A
A3
D3
D2
A2
A3
m1
C1
c
A
D1
C3
B3
A1
n2
B2
D3
m2
D1
B1
n1
D2
B2
C2
B1
n1
m1
FIGURE 6.6.1 Illustration of transport through an infinite square network with

two parallel or perpendicular defective links.
(6.6.2)
c (B A ) + c (A1 A ) + c (A2 A ) + c (A3 A ) + s = 0,

c (A B ) + c (B1 B ) + c (B2 B ) + c (B3 B ) s = 0,
where the term

(6.6.3) s (c c) (B A )
is regarded as an a priori unknown nodal source applied to a pristine network with

uniform conductance, c.
Working similarly with the second defective link, we derive corresponding
equations involving a nodal source with strength
(6.6.4) s (c c) (D C )
at the point C, and a nodal sink with opposite strength at the point D.
nodes can be decomposed into a homogeneous solution, 0 , a particular solution due
to a source (sink) at node A accompanied by a sink (source) at node B, denoted as
AB , and another particular solution due to a source (sink) at node C accompanied
by a sink (source) at node D, denoted as CD . The value at an arbitrary node X is
(6.6.5) X = X0 + XAB + XCD .
In terms of the lattice Greens function, GXY ,

(6.6.6) X = X0 +
s
s
GXA GXB ) +
GXC GXD ).
c
c
Physically, GXY is the potential induced at node X by a point source of unit strength
applied at point Y.
To compute the strengths of the fictitious sources, s and s , we apply equation
(6.6.6) at the end points of the defective links, obtaining
s
s
GAA GAB ) +
GAC GAD ),
c
c
s
s
B = B0 +
GBA GBB ) +
GBC GBD ),
c
c
(6.6.7)
s
s
C = C0 +
GCA GCB ) +
GCC GCD ),
c
c
s
s
D = D0 +
GDA GDB ) +
GDC GDD ).
c
c
A = A0 +
Next, we subtract the second from the first equation and the third from the second
equation, and obtain

BA + G
AB ) c c (B A )
1 (G
c
(6.6.8)

c c
AD G
AC G
BD (D C ) = 0 0
GBC + G
B
A
c
and
(6.6.9)

c c
CB G
CA G
DB (B A )
GDA + G
c

c c

1 (GDC + GCA )
(D C ) = D0 C0 ,
c
where
XY GXY GYY
(6.6.10) G
XX = 0. Solving this linear
is the normalized Greens function defined such that G
system provides for the nodal differences A B and D C and thereby allows
us to compute the strengths of the sources, s and s .
Parallel and Adjacent Defective Links
When the two defective links are parallel and adjacent, m1 = n1 + 1 and m2 = n2 ,
nodes B and C coincide. Referring to Section 5.4, we find that
1
4
AB = G
BA = G
BD = G
DB = ,
(6.6.11) G
AD = G
DA = 1 + 2 .
G
Substituting these values into equations(6.6.8) and (6.6.9), we obtain

(6.6.12)

1 c c
2 1 c c
1+
(B A )
(D B ) = B0 A0 ,
2 c
2
c

2 1 c c
1 c c
(B A ) + 1 +
(D B ) = D0 B0 .
2
c
2 c
When c = c , these equations can be added and rearranged to yield

(6.6.13) D A =

1

D0 A0 .
2 c c
1+ 1
c
The corresponding perturbation difference is

(6.6.14) DA
(D A ) (D0 A0 ).
Making substitutions, we obtain

(6.6.15)

DA
0
DA
1
+
0 0 0 , c /c, and
where DA
D
A
(6.6.16) =
2
.
2
Node Damage and Effective Medium Theory
If a node of a square lattice is damaged, the conductances of the four links sharing
the node are altered, as shown in Figure 6.6.2. In one special configuration, the conductance of all damaged links is the same, c . In the case of unidirectional transport
in the first direction, corresponding to the index i1 , the nodal values of the unperturbed potential, 0 , are independent of the second index, i2 . An effective medium
theory may then be developed following the analysis of Section 6.4.2. The analysis
culminates in equation (6.5.18) for the effective conductance coefficient, , where q
is the fraction of damaged links and the coefficient is given in (6.6.16).
In the case of clipped links, = 0, we obtain (6.5.20) and substitute the value of
from (6.6.16) to obtain
(6.6.17) = 1
1
q
2
i2
n2
n1
i1
FIGURE 6.6.2 Illustration of scalar transport

through an infinite square network of resistors
with one damaged node disrupting the operation
of four links.
[51]. The percolation threshold corresponding to = 0 is predicted to be

(6.6.18) qc
2
= 0.637,
pc = 1 qc = 1
2
= 0.363.
Using (2.7.2), we set pc = pnode

and obtain
c
2 1/2
= 0.60281,
(6.6.19) pnode
1
c
which compares favorably with the known value for the square lattice, pnode
=
c
0.59275, as discussed in Section 2.7.
Exercises
6.6.1 Perpendicular adjacent links
Derive the counterpart of system (6.6.12) for two adjacent perpendicular links, as
shown in Figure 6.6.1(b).
6.6.2 Simple cubic lattice
Derive an estimate for the node percolation threshold of the simple cubic lattice based
on the effective medium theory.
6.7 DAMAGED HONEYCOMB LATTICE
The analysis of Section 6.6 for the square lattice can be extended to the honeycomb
lattice. Consider transport through a honeycomb lattice, as shown in Figure 6.7.1.
All links have the same conductance, c, except for three adjoining defective links
that have different conductances, c , c , and c . Our objective is to assess the effect
H
I
c
c
D
A
c
B
F
J
E
FIGURE 6.7.1 Illustration of transport

through an infinite honeycomb lattice
with three adjoining defective links.
of the defects on the nodal distribution of a potential, , associated with a transported

entity. We will see that this calculation will allow us to obtain an accurate estimate
for the node percolation threshold.
Ten nodes of interest, labeled AJ, are shown in Figure 6.7.1. Balancing the rates
of transport at nodes AD provides us with four equations:
c (B A ) + c (C A ) + c (D A ) = 0,
c (A B ) + c (E B ) + c (F B ) = 0,
(6.7.1)
c (A C ) + c (G C ) + c (H C ) = 0,
c (A D ) + c(I D ) + c (J D ) = 0.
Rearranging, we obtain an identical set of equations:
c (B A ) + c (C A ) + c (D A ) + s + s + s
c (A B ) + c (E B ) + c (F B ) s
(6.7.2)
c (A C ) + c (G C ) + c (H C ) s
c (A D ) + c (I D ) + c (j D ) s
= 0,
= 0,
= 0,
= 0,
where
(6.7.3)
s = (c c) (B A ),
s = (c c) (C A ),
s = (c c) (D A )
are fictitious sources applied at the four nodes.

nodes can be decomposed into a homogeneous solution, 0 , a particular solution due
to the source (sink) at node A accompanied by a sink (source) at point B, denoted by
AB , another particular solution due to the source (sink) at node A accompanied by
a sink (source) at point C, denoted by AC , and a third particular solution due to the
source (sink) at node A accompanied by a sink (source) at point D, denoted by AD .
The nodal value at an arbitrary node, X, is
(6.7.4) X = X0 + XAB + XAC + XAD .
In terms of the lattice Greens function, GXY , we obtain the representation

(6.7.5) X = X0 +
s
s
s
(GXA GXB ) + (GXA GXC ) +
(GXA GXD ).
c
c
c
Physically, GXY is the potential induced at node X by a point source of unit strength
applied at point Y.
To compute the strengths of the sources, we apply the representation (6.7.5) at

nodes AD, obtaining
s
s
s
(GAA GAB ) + (GAA GAC ), + (GAA GAD ),
c
c
c
s
s
s
0
B = B + (GBA GBB ) + (GBA GBC ) +
(GBA GBD ),
c
c
c
(6.7.6)
s
s
s
C = C0 + (GCA GCB ) + (GCA GCC ) +
(GCA GCD ),
c
c
c
s
s
s
D = D0 + (GDA GDB ) + (GDA GDC ) +
(GDA GDD ).
c
c
c
A = A0 +
Subtracting the first from the second, third, and fourth equations, substituting expressions (6.7.3) for the fictitious sources, and rearranging, we obtain a system of
three linear equations for the differences B A , C A , and D A . The first
equation reads

c c
AB ) (B A ) c c (G
BA G
BC + G
AC ) (C A )
1
(GBA + G
c
c
(6.7.7)
c c
BD + G
AD ) (D A ) = 0 0 ,
(GBA G
B
A
c
the second equation reads
(6.7.8)

c c
c c

(GCA GCB + GAB ) (B A ) + 1

(GCA + GAC ) (C A )
c
c
c c
CD + G
AD ) (D A ) = 0 0 ,
(GCA G
C
A
c
and the third equation reads
(6.7.9)

c c
DB + G
AB ) (B A ) c c (G
DA G
DC + G
AC ) (C A )
(GDA G
c
c

c c

+ 1
(GDA + GAD ) (D A ) = D0 A0 ,
c
where
XY GXY GYY
(6.7.10) G
XX = 0.
is the normalized Greens function defined such that G
Using the results of Section 5.6, we find that
(6.7.11)
AB = G
BA = G
AC = G
CA = G
AD = G
DA = 1 ,
G
3
BC = G
CB = G
BD = G
DB = G
CD = G
DC = 1 .
G
2
Substituting these values into equations (6.7.7)(6.7.9), we obtain

2 c c
1 c c
1 c c
1+
(B A ) +
(C A ) +
(D A )
3 c
6 c
6 c
(6.7.12)
= B0 A0 ,

1 c c
2 c c
1 c c
(B A ) + 1 +
(C A ) +
(D A )
3 c
6 c
(6.7.13) 6 c
= C0 A0 ,
and

1 c c
1 c c
2 c c
(B A ) +
(C A ) + 1 +
(D A )
6 c
3 c
(6.7.14) 6 c
= D0 A0 .
Node Damage and Effective Medium Theory
If node A is damaged, the conductances of the three links sharing this node are
modified. Assume that the conductances of the three affected links is the same, given
as
(6.7.15) c = c = c = c,
where is an arbitrary positive or zero coefficient. In the case of vertical unperturbed

transport, the nodal values of the unperturbed potential are independent of horizontal
position,
0
(6.7.16) BA
B0 A0 = A0 C0 ,
D0 A0 = 0.
By symmetry, the perturbed nodal field satisfies the same equations. Equation
(6.7.14) is trivially satisfied and equation (6.7.12) or (6.7.13) yields
(6.7.17) BA B A = A C =
2
( 0 A0 ).
1+ B
The corresponding perturbation difference is

(6.7.18) BA
(B A ) (B0 A0 ).
Making substitutions, we obtain

(6.7.19)

BA
0
BA
1
.
1+
An effective medium theory can be developed following the analysis of Section 6.4.2 for one defective link, culminating in equation (6.5.18) for the effective
conductance coefficient, , where q is the fraction of damaged links and = 1. In
the case of clipped links, = 0, we substitute = 1 into equation (6.5.20) and obtain
(6.7.20) = 1 2 q
[19]. The percolation threshold corresponding to = 0 is predicted to be

(6.7.21) qc 0.5,
pc = 1 qc 0.5.

and obtain
c
1
(6.7.22) pnode
= 0.707,
c
which is in surprisingly good agreement with the exact value for the honeycomb
lattice, pnode
= 0.69704, as discussed in Section 2.7.
c
Exercise
6.7.1 Effective conductance and node percolation threshold
Derive the effective conductance and estimate the node percolation threshold for the
case of horizontal unperturbed transport.
6.8 DAMAGED HEXAGONAL LATTICE
Consider transport through a hexagonal lattice, as shown in Figure 6.8.1. To study the
performance of the network, we consider separately the case of longitudinal transport
where the unperturbed potential varies along horizontal links, and the case of lateral
transport where the unperturbed potential is constant along horizontal links.
6.8.1 Longitudinal Transport
To study the case of longitudinal unperturbed transport, we refer to Figure 6.8.1(a)

and assume links have the same conductance, c, except for two adjoining defective links that have different conductances, c and c . Repeating the analysis of
Sections 6.5 and 6.6, we derive the balance equations

(6.8.1)
c c
AB )
1
(GBA + G
c

(B A )
c c
BC + G
AC ) (C A ) = 0 0
(GBA G
B
A
c
(a)
(b)
E
D
c
A
G
B
FIGURE 6.8.1 Illustration of (a) longitudinal and (b) lateral transport through an infinite
hexagonal network.
and
(6.8.2)
c c
CB + G
AB ) (B A )
(GCA G
c

c c

+ 1
(GCA + GAC ) (C A ) = C0 A0 .
c

AB = G
BA = G
AC = G
CA = 1 ,
G
6
(6.8.3)
4 2

GBC = GCB = +
3.
3
Equations (6.8.1) and (6.8.2) then become

1 c c
2
c c
(6.8.4) 1 +
(B A ) +
31
(C A ) = B0 A0
3 c
c
and

(6.8.5)
2
31

c c
1 c c
(B A ) + 1 +
(C A ) = C0 A0 .
c
3 c
Assume that the conductances of the links is the same, c = c = c, where
is an arbitrary positive or zero coefficient. In the case of longitudinal unperturbed
transport, the nodal values of the unperturbed potential are independent of lateral
position and
0
(6.8.6) BA
B0 A0 = A0 C0 .
By symmetry, the perturbed nodal field satisfies the same equations. Equation (6.8.4)
or (6.8.5) yields
1
( 0 A0 )
1 + ( 1) B
(6.8.7) B A = A C =
and
(6.8.8)

BA
0
BA
1
,
+
where
(6.8.9) =
4 2
3,
3
1
.
The effective medium theory culminates in equation (6.5.20), yielding

(6.8.10) = 1
1
q.
1
The percolation threshold corresponding to = 0 is predicted to be

(6.8.11) qc 1 ,
pc = 1 qc .

and obtain
c
(6.8.12) pnode

c
= 0.480,
which is in surprisingly good agreement with the exact value for the hexagonal
lattice, pnode
= 0.5, as discussed in Section 2.7.
c
6.8.2 Lateral Transport
To study the case of lateral transport, we consider a more general configuration where
six links originating from a node labeled A and ending at nodes BG are damaged,
as shown in Figure 6.8.1(b). All links have the same conductance, c, except for the
six defective links that have a different conductance, c = c. The nodal value at an
arbitrary node, X, can be expressed in terms of the lattice Greens function, GXY , as

X = X0 + ( 1) (B A )(GXA GXB ) + (C A )(GXA GXC )
(6.8.13)
+(D A )(GXA GXD ) + (E A )(GXA GXE )

+(F A )(GXA GXF ) + (G A )(GXA GXG ) .
Subtracting the equation at node X = A from that at node X = B, and rearranging,

we obtain

BA + G
AB )
(B A ) 1 ( 1)(G

BA G
BC + G
AC )
( 1) (C A )(G
(6.8.14)
BA G
BD + G
AD ) (E A )(G
BA G
BE + G
AE )
(D A )(G
BA G
BF + G
AF ) (G A )(G
BA G
BG + G
AG )
(F A )(G
= B0 A0 ,
XY is the normalized Greens function defined so that G
XX = 0. Similar
where G
equations can be written for the other nodes.
Now we consider the field induced by a vertical potential gradient. By symmetry,
we have
(6.8.15) B0 A0 = A0 E0 = G0 A0 = A0 D0 ,
and
(6.8.16) F0 = A0 ,
A0 = C0 .
A set of identical equations can be written for the perturbed potential, . Equation
(6.8.14) simplifies into

BD + G
BE G
BA ) = 0 0 .
(6.8.17) (B A ) 1 ( 1)(G
B
A
1
6
BA = ,
(6.8.18) G
BD = 1 1 3,
G
3
BE = 4 + 2 3.
G
3
Substituting these values into (6.8.17), we obtain expressions (6.8.7) and (6.8.8),
where
5 1
1
(6.8.19) =
3, =
.
6
The effective medium theory yielding the node percolation threshold
(6.8.20) pnode
= 0.531,
c
which is in surprisingly good agreement with the exact value for the hexagonal
lattice, pnode
= 0.5 as discussed in Section 2.7 [18].
c
Exercise
6.8.1 Effective medium
Derive the values of and stated in (6.8.19).
APPENDIX A
EIGENVALUES OF MATRICES
A brief account of eigenvalues and eigenvectors of matrices is given in this appendix.

Further information is available in texts on linear algebra and numerical methods
(e.g., [35]).
A.1 EIGENVALUES AND EIGENVECTORS
An eigenvector, u, of an N N square matrix, A, and the corresponding eigenvalues,

, satisfy the equation
(A.1.1) A u = u,
with the understanding that the eigenvector, u, is not null, where a centered dot
indicates the regular matrix product. An equivalent statement is
(A.1.2) (A I) u = 0,
where I is the N N identity matrix. Requiring that this homogeneous equation has
a nontrivial solution for u, we find that the matrix
(A.1.3) A I =
A1,1
A2,1
AN1,1
AN,1
A1,2
A2,2
AN1,2
AN,2
A1, N1
A2, N1
AN1, N1
AN, N1
A1, N
A2, N
AN1, N
AN, N
must be singular, that is, its determinant must be zero. Conversely, the eigenvalues
of a matrix, A, render the diagonally shifted matrix A I singular. By definition, an
eigenvector belongs to the null space of the matrix A I.
If u is an eigenvector corresponding to a certain eigenvalue, then au is also an eigenvector corresponding to the same eigenvalue, for any real or complex constant a.
259
However, eigenvectors that arise from one another by multiplication with a nonzero
scalar constant are not distinct.
A real or complex matrix may have real or complex eigenvalues and associated
eigenvectors. A real matrix has either real eigenvalues or pairs of complex conjugate
eigenvalues. A real and symmetric matrix, and more generally a Hermitian complex
matrix, has only real eigenvalues. If a matrix is real, an eigenvector corresponding
to a real eigenvalue must be real if the eigenvalue is not repeated or complex if the
eigenvalue is repeated, whereas an eigenvector corresponding to complex eigenvalues is necessarily complex. If a matrix is complex, an eigenvector corresponding to
a real eigenvalue is necessarily complex.
A.2 THE CHARACTERISTIC POLYNOMIAL
Expressing the determinant of the shifted N N matrix, A I, in terms of the cofactors, we obtain an Nth-degree polynomial with respect to , called the characteristic
polynomial of the matrix A:
(A.2.1) PN () = det(A I),
where det denotes the determinant.

Monitoring the first three highest powers of in the Laplace expansion of the
determinant of the matrix A I, and noting that the constant term is the determinant
of A, we find that the characteristic polynomial takes the form
(A.2.2) PN () = ()N + c1 ()N1 + + cm ()Nm + + cN ,
where
c1 = trace(A) A1,1 + A2,2 + + AN,N ,
(A.2.3) c2 =
N
i1

(Ai,i Aj,j Ai,j Aj,i ),
i=1 j=1
cN = det(A).
When N = 2, we have c3 = det(A) and the characteristic polynomial is
(A.2.4) P2 () = 2 (A1,1 + A2,2 ) + (A1,1 A2,2 A1,2 A2,1 ).
When N = 3, we have c4 = det(A) and the characteristic polynomial is

P3 () = 3 + (A1,1 + A2,2 + A3,3 ) 2

(A1,1 A2,2 A1,2 A2,1 ) + (A2,2 A3,3 A2,3 A3,2 ) + (A3,3 A1,1 A3,1 A1,3 )
+ A1,1 (A2,2 A3,3 A2,3 A3,2 ) A2,1 (A1,2 A3,3 A1,3 A3,2 )
(A.2.5)
+ A3,1 (A1,2 A2,3 A1,3 A2,2 ).
E i g e n v a l u e s o f M a t r i c e s / / 261
Algorithms for the programmable computation of the coefficients, ci , for arbitrary

polynomials are available.
We have demonstrated that computing the eigenvalues of a matrix is equivalent
to finding the roots of its characteristic polynomial satisfying
(A.2.6) PN () = 0.
Since an Nth-degree polynomial has precisely N roots, an N N matrix is guaranteed

to have exactly N real or complex eigenvalues, 1 , 2 , . . . , N . If an eigenvalue, i ,
is repeated m times, its algebraic multiplicity is m, meaning that
PN (i ) = 0,
(A.2.7)
(m1)
PN (i )
PN (i ) = 0,
(m)
PN (i )
= 0,
...,
= 0,
(k)
where PN denotes the kth derivative. Since the coefficients of the characteristic polynomial associated with a real matrix are real, the eigenvalues must be real or appear
in pairs of complex conjugates.
Spectrum and Spectral Radius
The set of all eigenvalues of a matrix is the spectrum of eigenvalues of the matrix.
The maximum of the norm of all real and complex eigenvalues is the spectral radius
of the matrix,
(A.2.8) max |i |.
i
The spectral radius of a matrix is an important diagnostic of certain important

properties of the matrix regarded as a engine that drives a linear map.
Diagonal and Triangular Matrices
The characteristic polynomial of a diagonal or triangular matrix, A, takes the form

(A.2.9) PN () = (A1,1 )(A2,2 ) (AN,N ),
which shows that the eigenvalues are equal to the diagonal elements. A repeated
diagonal element reveals a multiple eigenvalue. For example, the N N identity
matrix has a single eigenvalue equal to unity with algebraic multiplicity m = N.
A.2.1 Eigenvalues, Trace, and the Determinant
The characteristic polynomial can be expressed in an alternative form in terms of its

roots,
(A.2.10) PN () = (1 ) (2 ) (N1 ) (N ).
In the case of a diagonal or triangular matrix, this expression is equivalent to that

shown in (A.2.9). Expanding the product on the right-hand side of (A.2.10), we find
that
(A.2.11)
PN () = ()N + ()N1 (1 + 2 + + N ) + + cm ()Nm

+ + 1 2 N1 N .
Comparing the right-hand side of this equation with the right-hand side of (A.2.11),
we derive expressions for the trace and determinant in terms of the eigenvalues:
(A.2.12) trace(A) = 1 + 2 + + N ,
det(A) = 1 2 N .
Thus, if one eigenvalue is zero, the determinant is also zero and the matrix is singular.
A.2.2 Powers, Inverse, and Functions of a Matrix
Multiplying both sides of the definition A u = u by A, we find that

(A.2.13) A2 u = A (A u) = (A u) = 2 u,
which shows that 2 is an eigenvalue of the matrix A2 with corresponding eigenvector u. Working in a similar fashion, we find that k is an eigenvalue of the matrix Ak
with corresponding eigenvector u, for any positive integer exponent, k.
Multiplying both sides of the definition A u = u by the inverse matrix A1 , we
find that u = (A1 u), and then
(A.2.14) A1 u =
1
u.
Thus, 1/ is an eigenvalue of the inverse matrix, A1 , with corresponding eigenvector u.

Working in a similar fashion, we find that, if Q(x) is an arbitrary polynomial,
then Q() is an eigenvalue of the matrix Q(A) with corresponding eigenvector u. If
Q(x) and R(x) are two arbitrary polynomials, then Q()/R() is an eigenvalue of the
matrix R1 (A) Q(A) with corresponding eigenvector u. To show this, we observe
that R() Q(A) u = Q() R(A) u.
A.2.3 Hermitian Matrices
By definition, a Hermitian matrix is equal to the complex conjugate of its transpose.

Hermitian matrices, and their inclusive real and symmetric matrices, have real eigenvalues. To show this, we take the complex conjugate of the definition A u = u,
finding that Aij uj = ui , where summation is implied over the repeated index j, and
an asterisk denotes the complex conjugate. Because the matrix A is assumed to be

Hermitian, Aij = Aji and thus Aji uj = ui . Taking the inner product of both sides
with u, we find that
(A.2.15) ui Aji uj = ui ui ,
or
uj uj = ui ui ,
where summation is implied over the repeated index i. The last equation requires that
= , which guarantees that is real.
Consider an NN Hermitian matrix, A. If the scalar xi Aij xj is real and positive for
any N-dimensional vector x, then the matrix A is called positive definite. Identifying
x with an eigenvector, we find that ui Aij uj = ui ui > 0. Since u u is real and
positive, the eigenvalue, , must also be real and positive. We conclude that a positive
definite Hermitian matrix has real and positive eigenvalues.
A.2.4 Diagonal Matrix of Eigenvalues
It is useful to introduce a diagonal

eigenvalues of a matrix, A:
1
0
.
(A.2.16) = ..
0
0
matrix, , whose diagonal entries are the N

0
2
..
.
..
.
0
0
..
.
0
0
..
.
0
0
N1
0
0
N
Note that some or all of the eigenvalues may be the same.

Next, we consider the characteristic polynomial of the matrix A, given in (A.2.1),
replace with and unity with the N N identity matrix, I, and obtain the matrix
polynomial
(A.2.17) PN () = (1 I ) (2 I ) (N I ).
We note that the ith column of the matrix enclosed by the ith set of parentheses on
the right-hand side is zero for i = 1, . . . , N, and we carry out the multiplications to
obtain
(A.2.18) PN () = 0,
which shows that the diagonal matrix of eigenvalues is a root of the characteristic
polynomial.
A.3 EIGENVECTORS AND PRINCIPAL VECTORS
If the eigenvalues of a matrix are available, the eigenvectors can be found by solving
the homogeneous linear system (A.1.2). For each eigenvalue, the linear system has
multiple solutions reflecting the arbitrary length of the corresponding eigenvector.

However, this degree of freedom can be removed by imposing a suitable constraint.
For example, we may specify the value of one eigenvector component, solve for the
rest of the components, and then scale the eigenvector so that the magnitude of one
chosen component or the length of the eigenvector is equal to unity.
A.3.1 Properties of Eigenvectors
Eigenvectors corresponding to distinct eigenvalues are linearly independent. To show

this, we express one eigenvector as a linear combination of all other eigenvectors,
multiply the linear expansion by the matrix A, use the definition of the eigenvectors,
compare the resulting equation with the original expansion of the eigenvector, and
find that the eigenvector must be the null vector, which is a contradiction.
If a matrix has N distinct eigenvalues, it is guaranteed to have N linearly independent eigenvectors that form a base of the N-dimensional space. Any vector can
be expressed as a linear combination of the eigenvectors.
If one or more eigenvalues appear multiple times, we may not be able to find N
linearly independent eigenvectors. The number of eigenvectors, k, corresponding to a
particular eigenvalue of algebraic multiplicity m, is the geometric multiplicity of the
eigenvalue. Since equation (A.1.2) has at least one family of solutions, the geometric
multiplicity satisfies the inequality 1 k m.
Hermitian matrices are guaranteed to have N linearly independent and mutually
orthogonal eigenvectors, even in the case of multiple eigenvalues. The proof relies
on the existence of the Schur normal (e.g. [35]).
Two different matrices may have the same set of linearly independent eigenvectors. For example, two Hermitian matrices that commute with respect to
multiplication share eigenvectors but not necessarily eigenvalues.
A.3.2 Left Eigenvectors
The determinant, and therefore the characteristic polynomial and eigenvalues of a

matrix, A, are the same as those of its transpose, AT . However, unless the matrix is
symmetric, the eigenvectors are different. The eigenvectors of the transpose, AT, are
also called the left eigenvectors of A.
An eigenvector of AT corresponding to an eigenvalue 1 is orthogonal to an
eigenvector of A corresponding to a different eigenvalue, 2 . To show this, we formulate the inner product of both sides of the definition, A u = 2 u, with the left
eigenvector, v, and find that vi Aij uj = 2 vi ui , where summation is implied over the
repeated indices i and j. Substituting the definition vi Aij = 1 vj and rearranging, we
obtain (1 2 ) u v = 0, which shows that u v = 0.
The number of linearly independent eigenvectors of a matrix and its transpose
corresponding to a particular multiple eigenvalue is the same. A matrix and its
transpose have identical eigenvalues and the same number of linearly independent
eigenvectors.
A.3.3 Matrix of Eigenvectors
If an N N matrix, A, has N eigenvectors, u(i) , its transpose also has N eigenvectors,

v(i) . It is useful to arrange the first set of eigenvectors at the columns of the matrix
(A.3.1) U = u(1)
u(2)
..
.
u(N1)
u(N) ,
and the second set of eigenvectors at the columns of the matrix
(A.3.2) V = v(1)
v(2)
..
.
v(N1)
v(N) .
Next, we normalize the eigenvectors so that corresponding pairs satisfy the condition
(A.3.3) v(i) u(i) = 1.
Subject to these definitions,

(A.3.4) VT U = I,
UT V = I,
(A.3.5) U1 = VT ,
V1 = UT .
which shows that
The collection, u(i) , and the collection, v(i) , provide us with two mutually orthogonal
(biorthonormal) sets.
Symmetric Matrices
Since the eigenvalues and eigenvectors of a symmetric matrix and its transpose
are identical, two eigenvectors corresponding to two different eigenvalues are
orthogonal. Consequently,
(A.3.6) U = V,
U1 = UT ,
which demonstrates that the matrix of eigenvectors, U, is orthogonal. An N N real

symmetric matrix has N real eigenvalues and N real and orthogonal eigenvectors.
A.3.4 Eigenvalues and Eigenvectors of the Adjoint
The determinant of a matrix, A, is equal to the complex conjugate of the determinant

of its adjoint, denoted by the superscript A:
(A.3.7) AA A .
T
Thus, the characteristic polynomial, and therefore the eigenvalues, are complex conjugates of those of the adjoint. However, the associated eigenvectors are not generally
associated by a simple relationship.
An eigenvector of AA corresponding to an eigenvalue, 1 , call it w, is orthogonal
to the complex conjugate of an eigenvector of A corresponding to an eigenvalue
2 , call it u , where 1 = 2 , that is, w u = 0. This property follows from the
biorthogonality of the eigenvectors of a matrix and its transpose discussed earlier in
this section.
Let us assume that AA has N eigenvectors, w(i) , arranged at the columns of a
matrix, W. Moreover, let us assume that the two sets of eigenvectors w(i) and u(i) are
normalized so that
(A.3.8) w(i) u(i) = 1.
By construction, we have
(A.3.9) U1 = WA ,
W1 = UA ,
in agreement with (A.3.5). If the matrix A is Hermitian, then W = U and U1 = UA .

A.3.5 Eigenvalues of Positive Definite Hermitian Matrices
We have seen that a positive definite Hermitian matrix, A, has real and positive
eigenvalues. Conversely, if all eigenvalues of a Hermitian matrix are positive, the
matrix is positive definite. To show this, we express an arbitrary vector, x, as a linear
combination of the eigenvectors:
(A.3.10) x = c1 u(1) + + cN u(N) .
Multiplying both sides by A, we find that

(A.3.11) A x = c1 1 u(1) + + cN N u(N) .
Next, we compute the scalar
(A.3.12) x A x = c1 u(1) + + cN u(N)

c1 1 u(1) + + cN N u(N) .
Using the orthogonality property, u(i) u(j) = 0 for i = j, we obtain
(A.3.13) x A x = c1 c1 1 u(1) u(1) + + cN cN N u(N) u(N)
or
(A.3.14) x A x = |c1 |2 1 |u(1) |2 + + |cN |2 N |u(N) |2 ,
which shows that, if all eigenvalues are positive, x A x is guaranteed to also be

positive.
A.4 CIRCULANT MATRICES
Each row of a circulant matrix derives from the previous row by shifting each element
to the right by one place, and then returning the last element to the first place. By
construction, all elements along any super- or subdiagonal line of a circulant matrix
are the same.
A 2 2 circulant matrix, a 3 3 circulant matrix, and a 4 4 circulant matrix
are shown below,

(A.4.1) A =
a
b
b
a

,
a
A= c
b
b
a
c
c
b ,
a
a
d
A=
c
b
b
a
d
c
c
b
a
d
d
c
,
b
a
where a, b, c, and d are arbitrary elements.

Circulant matrices arise in the mathematical modeling of problems involving
temporal or spatial periodicity. It is remarkable that the eigenvalues and eigenvectors
of an arbitrary circulant matrix can be found explicitly in closed form.
Let A be an N N circulant matrix and qm be an Nth complex root of unity,
satisfying qN
m = 1, given by
(A.4.2) qm = exp[ (m 1)k i ],
for i = 1, . . . , N, where k = 2 /N and i is the imaginary unit, i2 = 1. Direct

substitution shows that the eigenvalues of A are given by
(A.4.3) m = A1,1 + A1,2 qm + A1,3 q2m + + A1, N qN1
m ,
and the corresponding eigenvectors are

T
(A.4.4) u(m) = 1, qm , q2m , . . . , qN1
,
m
where m = 1, . . . , N. The first eigenvalue, 1 , is equal to the sum of the elements in

each row. The corresponding eigenvector is filled with ones:
(A.4.5) u(1) = [ 1, 1 , . . . , 1 ]T .
When N = 2, we obtain q1 = 1 and q2 = 1, yielding

(A.4.6) 1 = a + b,
2 = a b,
which are the eigenvalues of the first matrix in (A.4.1).

A.5 BLOCK CIRCULANT MATRICES
A block circulant matrix consists of N repeated matrix blocks in the arrangement of

circulant matrices:
A(1)
A(N)
..
.
(A.5.1) A =
A(3)
A(2)
A(2)
A(1)
..
.
..
.
A(N1)
A(N2)
..
.
A(N)
A(N1)
..
.
A(4)
A(3)
A(N1)
A(N1)
A(1)
A(1)
where A(i) are square matrices with the same dimensions, M M. Each row of this
matrix derives from the previous row by shifting each block to the right by one place
and then bringing the last block to the first place.
Consider the following square M M matrices defined in terms of the block
matrices, A(i) :
(N)
(A.5.2) B(m) = A(1) + qm A(2) + q2m A(2) + + qN1
m A
for m = 1, . . . , N. It can be shown that the determinant of the matrix A is the product
of the determinants of the matrices B(m) , the characteristic polynomial of A is the
product of the characteristic polynomials of B(m) , and the spectrum of eigenvalues of
A is the union of the spectra of eigenvalues of B(m) [12].
APPENDIX B
THE SHERMANMORRISON
AND WOODBURY FORMULAS
The Woodbury and ShermanMorrison formulas allow us to compute the inverse of a

matrix that is perturbed with respect to a reference matrix whose inverse is available
in an explicit or readily computable form (e.g., [17], p. 123).
B.1 THE WOODBURY FORMULA
Woodburys formula relates the inverse of a perturbed N N matrix, B, to the inverse

of an unperturbed N N matrix, A. The two matrices are related by
(B.1.1) A = B + U VT ,
where U and V are two N K matrices, K 1 is an arbitrary dimension, the superscript T denotes the matrix transpose, and a centered dot denotes the usual matrix
product. The inverse of the perturbed matrix is

(B.1.2) A1 = B1 I U (IK + G)1 VT B1 ,
where I is the N N identity matrix, IK is the K K identity matrix, and

(B.1.3) G VT B1 U
is a K K matrix.
Direct Proof
Woodburys formula can be proved by direct substitution, invoking the definition of

the matrix inverse. Using (B.1.1) and (B.1.2), we compute

(B.1.4) A A1 = (B + U VT ) B1 I U (IK + G)1 VT B1 .

269
Carrying out the multiplications and invoking the definition of the matrix G, we
obtain
(B.1.5) A A1 = I + U VT B1 U (IK + G) (IK + G)1 VT B1 ,
where the first two terms on the right-hand side correspond to the matrix I inside the
square brackets on the right-hand side of (B.1.4). Carrying out the multiplications,
we obtain
(B.1.6) A A1 = I,
as required. Other proofs based on block Gauss elimination of LU decomposition are

available.
Proof by Gauss Elimination
(B.1.7) (B + U VT ) A1 = I
and thus
(B.1.8) B A1 + U D = I,
where
(B.1.9) D VT A1
is an intermediate K N matrix. The last two equations can be collected into the
block linear system

B
U
I
A1
(B.1.10)
=
.
0
VT
IK
D
Solving the first equation for A1 , we obtain
(B.1.11) A1 = B1 U D + B1 .
Substituting this expression into the second equation of (B.1.10), we obtain the
reduced system

(B.1.12)
B
0

I
A1
=
.
D
VT B1
U
IK + G
From the second equation, we find that

1
(B.1.13) D = IK + G
VT B1 .
S h e r m a n M o r r i s o n a n d Wo o d b u r y Fo r m u l a s / / 271
Substituting this expression into the first equation yields the Woodbury formula.
The procedure described is the counterpart of the method of Gauss elimination
for solving systems of linear equations. The counterpart of the LU decomposition is
the block decomposition

(B.1.14)
B
VT
U
IK
I
VT B1
0
IK

B
U
IK + G

.
The determinant of the matrix on the left-hand side is equal to the determinant of the
matrix A = B + UVT . Since the first matrix on the right-hand side is lower triangular
with ones along the diagonal, its determinant is equal to unity. The determinant of
the second matrix on the right-hand side is equal to the product of the determinants
of the two square matrices along the diagonal, B and IK + G. Taking the determinant
of both sides of (B.1.13), recalling that the determinant of the product of two square
matrix is the product of the determinants, and rearranging, we obtain

det(A)
= det IK + G .
det(B)
(B.1.15)
When B is the N N identity matrix, I, we find that

(B.1.16) det I + U VT = det IK + VT U ,
expressing Sylvesters determinant theorem.

Alternative Proof
A third way of proving the Woodbury formula proceeds by applying the general
identity
1
(B.1.17) (B C)
=B

I+
(C B1 )n
n=1
for the square matrix C = U VT , obtaining

(B.1.18) (B + U V )
T 1
=B

I+
(U VT B1 )n .
n=1
Rearranging the sum, we find that

(B.1.19) A1 = B1 I U IK +

m=1

(G)m VT B1 ,
where m = n 1. Also applying the general identity

(B.1.20) (IK C)1 = IK +
Cn
n=1
for the K K matrix C = G, we obtain

(B.1.21) (IK + G)

= IK +
(G)n ,
n=1
which completes the proof.

Generalized Woodbury Formulas
A generalization of the Woodbury formula (B.1.2) provides us with the inverse of

the N N perturbed matrix
(B.1.22) A = B + U VT ,
where U and V are two N K matrices and is an arbitrary nonsingular K K

matrix with K 1. The inverse of the perturbed matrix is

(B.1.23) A1 = B1 I U (1 + G)1 VT B1 .
Formula (B.1.23) arises from (B.1.2) by replacing U with U. When is the K K
identity matrix we recover the standard Woodbury formula.
Further Generalization
A further generalization incorporates M deviations of a matrix of interest, A, from

an unperturbed matrix, A,
T
(B.1.24)
A = B + U(1) (1) V(1) + + U(q) (q) V(q)
+ + U(M) (M) V(M) ,
where U(q) and V(q) are collections of N Kq matrices and (q) are Kq Kq square
matrices for q = 1, . . . , M. Let
(B.1.25) K
M

q=1
Kq .
The inverse of the perturbed matrix is given in (B.1.23), where I is the N N identity
matrix,
..
.
(B.1.26) U U(1)
..
.
U(q)
U(M)
is a N K matrix,
(B.1.27) VT
V(1)
T
V(q)
T
V(M)
is a K N matrix,
(B.1.28)
(1)
0
..
.
0
(2)
..
.
..
.
0
0
..
.
0
0
..
.
0
0
0
0
(M1)
0
0
(M)
is a square block-diagonal K K matrix, and
V(1) B1 U(1)
(2)T
V
B1 U(1)
(B.1.29) G =
..
V(p) B1 U(1)
V(1) B1 U(2)
T
V(2) B1 U(2)
..
.
T
V(p) B1 U(2)
..
.
V(1) B1 U(p)
T
V(2) B1 U(p)
..
.
V(p) B1 U(p)
is a K K matrix [2]. Formula (B.1.23) corresponds to M = 1. To prove the

generalized formula, we simply observe that
T
(B.1.30) U(1) (1) V(1) + + U(M) (p) V(M) = U VT .
B.2 THE SHERMANMORRISON FORMULA
In the particular case where K = 1, the otherwise arbitrary matrices U and V reduce
into N-dimensional column vectors, u and v, and
(B.2.1) A = B + u vT .
The matrix G IK + vT B1 u is scalar and the Woodbury formula reduces to the

ShermanMorrison formula,

(B.2.2) A1 = B1 I

1
u vT B1 ,
1+s
where
(B.2.3) s vT B1 u
is a scalar (e.g., [4], p. 39). Equation (B.2.4) yields

(B.2.4)
det(A)
= 1 + s.
det(B)
It will be noted that the ShermanMorrison formula fails when s = 1, in which case
B1 u is an eigenvector of the perturbed matrix A corresponding to a zero eigenvalue.
The following Matlab script uses the internal Matlab function inv to verify the
ShermanMorrison formula:
B = [1 2 3; 2 3 4; 1 4 5];
u = [3 4 9];
v = [2 3 7];
A = B+u*v;
invA = inv(A)
invB = inv(B);
s = v*invB*u;
invA1 = invB-invB*u*v*invB/(1+s)
A prime denotes the vector or matrix transpose. The output of the code is
invA =
7.0000
-18.0000
6.2500
invA1 =
7.0000
-18.0000
6.2500
-1.0000
5.0000
-2.0000
-2.0000
4.0000
-1.2500
-1.0000
5.0000
-2.0000
-2.0000
4.0000
-1.2500
We observe that the matrix inverses computed directly or by using the Sherman
Morrison formula are identical.
If all matrices involved are scalars, N = 1 and K = 1, the ShermanMorrison
formula provides us with the identity
(B.2.5)
1
1
uv
=
1
,
b + uv b
b + uv
whose veracity can be readily confirmed.

An alternative form of (B.2.2) is

(B.2.6) A
1
I

1+s

B1 ,
where
(B.2.7) = w vT ,
w = B1 u,
s wT v.
The N N matrix satisfies the property

(B.2.8) n = sn
for any integer, n.

If A is the identity matrix, I, we obtain
(B.2.9) (I + u vT )1 = I
1
u vT ,
1+s
where s uT v.
Generalized ShermanMorrison Formula
Consider two collections of N-dimensional column vectors, u(q) and v(q) for q =
1, . . . , M, and formulate the perturbed N N matrix
T
(B.2.10) A = B + 1 u(1) v(1) + + q u(q) v(q) + + M u(M) v(M) ,
where q are arbitrary constants for q = 1, . . . , M. In compact notation, we have

(B.2.11) A = B + U Z VT ,
where
..
.
(B.2.12) U u(1)
..
.
u(p)
u(M)
is an N M matrix,
(B.2.13) V v(1)
..
.
v(p)
..
.
v(M)
is another N M matrix, and
(B.2.14) Z
0
..
.
2
..
.
..
.
0
..
.
0
..
.
M1
is an M M diagonal matrix.
Applying (B.1.23) with = Z and K = M, we find that the inverse of the
perturbed matrix A is given by

1
(B.2.15) A1 = B1 I U Z1 + G

VT B1 ,
where
(B.2.16) G VT B1 U
is an M M matrix with components
(i)
(j)
1
(i)
(j)
(B.2.17) Gij = vl B1
lm um = Blm : v u
= v(i) w(j) ,
summation is implied over the repeated indices l and m, and the vector w(j) satisfies
the linear system
(B.2.18) B w(j) = u(j)
for j = 1, . . . , M. We may set B1 U = W and obtain G VT W, where
(B.2.19) W =
w(1)
..
.
w(p)
..
.
w(M)
is an N M matrix.
The following Matlab script confirms the generalized ShermanMorrison formula for N = 3 and M = 2:
B = [1 2 3;
2 3 4;
1 4 5];
z1 = 1.4;
u1 = [3 4 9];
v1 = [2 3 7];
z2 = 3.4;
u2 = [1 2 3];
v2 = [6 5 4];
U(:,1) = u1;
U(:,2) = u2;
V(:,1) = v1;
V(:,2) = v2;
w1 = u1/B;
w2 = u2/B;
G(1,1) = v1*w1; G(1,2) = v1*w2;
G(2,1) = v2*w1; G(2,2) = v2*w2;
Z(1,1) = c1; Z(1,2) = 0.0;
Z(2,1) = 0.0; Z(2,2) = c2;
A = B + z1*u1*v1 + z2*u2*v2;
invA = inv(A)
invB = inv(B);
invA1 = invB - invB*U*inv(inv(Z)+G)*V*invB
The output of the code is
invA =
4.8017
-7.6920
2.1952
0.2443
-0.1932
-0.0256
-1.7767
2.7420
-0.7327
invA1 =
4.8017
-7.6920
2.1952
0.2443
-0.1932
-0.0256
-1.7767
2.7420
-0.7327
The matrix inverse computed directly is the same as that computed by the Sherman
Morrison formula.
REFERENCES
[1] Atkinson D., and Van Steenwijk, F. J. (1999) Infinite resistive lattices. Am. J. Phys. 67, 486492.
[2] Batista, M. (2009) A note on a generalization of ShermanMorrisonWoodbury formula.
arXiv:0807.3860.
[3] Beyer, W. H. (1987) Standard Mathematical Tables, 28th edition. CRC Press, Boca Raton, FL.
[4] Bodewig, E. (1959) Matrix Calculus. North-Holland, Amsterdam.
[5] Cserti, J. (2000) Application of the lattice Greens function for calculating the resistance of an
infinite network of resistors. Am. J. Phys. 68(10), 896906.
[6] Cserti, J., and Tichy, G. (2004) A simple model for the vibrational modes in honeycomb lattices.
Eur. J. Phys. 25, 723736.
[7] Dean, P. (1963) The vibrations of three two-dimensional lattices. Proc. Camb. Philos. Soc. 59,
383396.
[8] Ellens, W. (2011) Effective resistance and other graph measures for network robustness.
Masters thesis, Mathematical Institute, University of Leiden.
[9] Ellens W., Spieksma, F. M., Van Mieghem, P., Jamakovic A., and Kooij R. E (2011) Effective
graph resistance. Lin. Alg. Appl. 435, 24912506.
[10] Fiedler, M. (1973) Algebraic connectivity of graphs. Czech. Math. J. 23(98), 298305.
[11] Fortunato, S. (2010) Community detection in graphs. Phys. Rep. 486, 75174.
[12] Friedman, B. (1961) Eigenvalues of composite matrices. Proc. Camb. Philos. Soc. 57, 3748.
[13] Grnbaum, B., and Shephard, G. C. (1987) Tilings and Patterns. W. H. Freeman, New York.
[14] Glasser M. L., and Zucker, I. J. (1977) Extended Watson integrals for cubic lattices. Proc. Natl.
Acad. Sci. USA 74, 18001801.
[15] Godsil C. D., and Royle, G. F. (2001) Algebraic Graph Theory. Graduate Texts in Mathematics,
No 207, SpringerVerlag, New York.
[16] Guttmann, A. J. (2010) Lattice Greens functions in all dimensions. J. Phys. A: Math. Theor. 43,
305205.
[17] Householder, A. S. (1964) The Theory of Matrices in Numerical Analysis. Reprinted by Dover
Publications, Mineola, NY (1975).
[18] Joy, T., and Strieder, W. (1978) Effective medium theory of site percolation in a random simple
triangular conductance network. J. Phys. C: Solid State Phys. 11, L867L870.
[19] Joy, T., and Strieder, W. (1979) Effective-medium theory of the conductivity for a random-site
honeycomb lattice. J. Phys. C: Solid State Phys. 12, L279L281.
[20] Joyce, G. S. (1972) Lattice Green function for the simple cubic lattice. Q. J. Math. 10, 266276.
[21] Joyce, G. S. (2002) Exact evaluation of the simple cubic lattice Green function for a general
lattice point. J. Phys. A: Math. Gen. 35, 98119828.
[22] Joyce, G. S., and Delves, R. T. (2004) Exact product forms for the simple cubic lattice Green
function I. J. Phys. A: Math. Gen. 37, 36453671.
[23] Joyce, G. S., and Delves, R. T. (2004) Exact product forms for the simple cubic lattice Green
function II. J. Phys. A: Math. Gen. 37, 54175447.
278
R E F E R E N C E S / / 279
[24] Kirkpatrick, S. (1972) Classical transport in disordered media: Scaling and effective-medium
theories. Phys. Rev. Lett. 27, 17221725.
[25] Kirkpatrick, S. (1973) Percolation and conduction. Rev. Mod. Phys. 45, 574588.
[26] Klein, D. J., and Randic, M. (1993) Resistance distance. J. Math. Chem. 12, 8195.
[27] Li, J.-S., and Zhang, X.-D. (1998) On the Laplacian eigenvalues of a graph. Lin. Alg. Appl. 285,
305307.
[28] Lorenz, C. D., and Ziff, R. M. (1998) Precise determination of the bond percolation thresholds
and finite-size scaling corrections for the SC, FCC, and BCC lattices. Phys. Rev. E 57, 230236.
[29] Martin, P. A. (2006) Discrete scattering theory: Greens function for a square lattice. Wave
Motion 43, 619629.
[30] Mohar, B. (1991) The Laplacian spectrum of graphs. In: Alavi, Y., Chartrand, G., Oellermann,
O. R., Schwenk, A.J. (eds.), Theory, Combinatorics, and Applications, Vol. 2. Wiley, New York,
pp. 871898.
[31] Morita, T. (1971) Useful procedure for computing the lattice Greens functionsquare, tetragonal, and bcc lattices. J. Math. Phys. 12, 17441747.
[32] Newman, M. E. J (2003) The structure and function of complex networks. SIAM Rev. 45, 167
256.
[33] Newman, M. E. J., and Ziff, R. M. (2000) Efficient Monte-Carlo algorithm and high-precision
results for percolation. Phys. Rev. Lett. 85, 41044107.
[34] Pozrikidis, C. (2014) Introduction to Finite and Spectral Element Methods Using Matlab, 2nd
edition. Taylor & Francis/CRC, New York.
[35] Pozrikidis, C. (2008) Numerical Computation in Science and Engineering, 2nd edition. Oxford
University Press, New York.
[36] Pozrikidis, C. (2011) Introduction to Theoretical and Computational Fluid Dynamics, 2nd
edition. Oxford University Press, New York.
[37] Pozrikidis, C., and Hill, A. I. (2013) Operational thresholds of disrupted networks. Physica
Scripta 87, 015604.
[38] Redner, S. (2001) A Guide to First-Passage Processes. Cambridge University Press, New York.
[39] Scullard, C. R. (2006) Exact site percolation thresholds using a site-to-bond transformation and
the star-triangle transformation. Phys. Rev. E 73, 016107.
[40] Scullard, C. R., and Ziff, R. M. (2006) Prediction of bond percolation thresholds for the kagom
and Archimedean (3, 122 ) lattices. Phys. Rev. E 73, 045102(R).
[41] Scher, H., and Zallen, R. (1970) Critical density in percolation processes. J. Chem. Phys. 53,
37593761.
[42] Suding, P. N., and Ziff, R. M. (1999) Site percolation thresholds for Archimedean lattices. Phys.
Rev. E 60, 275283.
[43] Sykes, M. F., and Essam, J. W. (1963) Some exact critical percolation probabilities for bond and
site problems in two dimensions. Phys. Rev. Lett. 10, 34.
[44] Sykes, M. F., and Essam, J. W. (1964) Exact critical percolation probabilities for site and bond
problems in two dimensions. J. Math. Phys. 5, 11171127.
[45] Tzeng, W.-J., and Wu, F. Y. (2000) Spanning trees on hypercubic lattices and non-orientable
surfaces. Appl. Math. Lett. 13, 1925.
[46] Van Der Marck, S. C. (1997) Percolation thresholds and universal formulas. Phys. Rev. E 55,
15141517.
[47] Venezian, G. (1994) On the resistance between two points on a grid. Am. J. Phys. 62, 10001004.
[48] Vyssotsky, V. A., Gordon, S. B., Frisch, H. L., and Hammersley, J. M. (1961) Critical percolation
probabilies (bond problem). Phys. Rev. 123, 15661567.
[49] Wang, H., Kooij, R. E., and Van Mieghem, P. (2010) Graphs with given diameter maximizing
the algebraic connectivity. Lin. Alg. Appl. 433, 18891908.
[50] Watson, N. G. (1939) Three triple integrals. Q. J. Math. 10, 266276.
280 / / R E F E R E N C E S
[51] Watson, B. P., and Leath, P. L. (1974) Conductivity in the two-dimensional-site percolation
problem. Phys. Rev. B 9, 48934896.
[52] Wierman, J. C. (1981) Bond percolation on honeycomb and triangular lattices. Adv. Appl. Prob.
13, 298313.
[53] Wierman, J. C., and Naor, D. P. (2005) Criteria for evaluation of universal formulas for
percolation thresholds. Phys. Rev. E 71, 036143.
[54] Wilson, R. J. (2010) Introduction to Graph Theory. Prentice Hall, Upper Saddle River, NJ.
[55] Weinberg, L. (1958) Kirchhoffs third and fourth laws. IRE Trans. Circ. Theory 5, 830.
[56] Wu, F. Y. (2004) Theory of resistor networks: The two point resistance. J. Phys. A: Math. Gen.
37, 66536673.
[57] Ziff, R. M. (2006) Generalized celldual-cell transformation and exact thresholds for percolation. Phys. Rev. E 73, 016134.
[58] Ziff, R. M., and Gu, H. (2009) Universal condition for critical percolation thresholds of Kagomlike lattices. Phys. Rev. E 79, 020102(R).
INDEX
addition of a link, 49
adjacency matrix, 18, 26
periodic, 21
weighed, 136
admittance matrix, 138
Archimedean lattice, 53
bathroom tile lattice, 55
bcc lattice, 59, 124
Greens function, 209
biharmonic operator, 155
biorthonormal sets, 265
bounce lattice, 56
boundary condition
Dirichlet, 3
Neumann, 6
periodic, 13
bow-tie lattice, 57
Bravais lattice, 50, 86
bridge lattice, 56
Brillouin zone, 51
Cartesian grid, 153
characteristic polynomial, 260
Cheegers constant, 39
circulant matrix, 21, 267
block, 268
clique, 29
complement of a graph, 29
Laplacian of, 38
complete graph, 29, 34, 171
conductance, 130
arbitrary, 135
matrix, 136
scaled, 136
connected graph, 30
connectivity
algebraic, 34
list, 19, 30
coordination number, 28
cross lattice, 55
cubic lattice, 58
bcc, 59, 124
fcc, 59, 126
simple, 59, 122
degree of a node, 18, 28
delta function, 171
determinant, 259261
Sylvester theorem, 271
diagonal matrix, 261
of eigenvalues, 263
diameter of a graph, 30
differential equation, 1
partial, 153
digraph, 30
directed graph, 30
Dirichlet
boundary condition, 3
node, 130
discontiguous network, 172
dual lattice, 56
edge
list, 30
weight, 136
effective medium theory, 250, 254
eigenvalue, 259
algebraic multiplicity of, 261
eigenvector, 259, 263
left, 264
elliptic integral, 209
embedded network, 130, 134, 142, 161
281
282 / / I N D E X
embedding matrix, 131
weighed, 142
Euler constant, 189, 196
exponential integral, 189
fcc lattice, 59, 126
finite
difference method, 2, 153, 216
element method, 156
Fourier expansion
in one dimension, 22
in two dimensions, 76
Gamma function, 209
gradient, 3
graph, 26
complement, 29, 38
complete, 29
connected, 30
diameter, 30
directed, 30
Laplacian, 17
one-dimensional, 16
order, 26
periodic, 20
random, 31
size, 26
unconnected, 30
undirected, 30
Greens function
bcc lattice, 209
fcc lattice, 211
free-space, 175, 212
hexagonal lattice, 191
honeycomb lattice, 200
in probability theory, 213
lattice, 173
MoorePenrose, 164
normalized, 163, 167
periodic, 173
simple cubic lattice, 206
square lattice, 177
Union Jack modified lattice, 196
Greens functions, 161
grid, 1
finite difference, 153
finite element, 156
Heaviside function, 133

Helmholtz equation, 1, 156
Hermitian matrix eigenvalues, 262, 266
hexagonal
grid, 155
lattice, 86, 150
damaged, 255
honeycomb
grid, 155
lattice, 98, 176
damaged, 251
honeycomb lattice, 55
incidence matrix
normalized, 140
weighed, 139
inner
displacement, 98, 110
product, 3
isolated network, 130, 134, 142, 164, 223
kagom lattice, 55, 111
Kirchhoff
matrix, 138
damaged, 235
modified, 142
normalized, 140
properties, 139
spanning-tree theorem, 36
kisquadrille lattice, 57
Klein bottle, 84
Kroneckers delta, 51
Laplace equation, 1, 156
Laplacian
factorization, 2
matrix, 32
modified, 134
normalized, 38
operator, 3
lattice, 56
Archimedean, 53
bathroom tile, 55
bcc, 59, 124
I N D E X / / 283
bounce, 56
bow-tie, 57
bridge, 56
coordination number, 28, 50
cross, 55
cubic, 58
damaged, 242
fcc, 59, 126
in probability theory, 213
hexagonal, 54, 86
honeycomb, 55, 98, 176
kagom, 55, 111
kisquadrille, 57
maple leaf, 56
martini, 57
modified Union Jack, 93
puzzle, 56
ruby, 56
simple cubic, 59, 122
snub hexagonal, 56
snub square, 56
square, 53, 67
square octagon, 55
star, 55
tetrakis, 57
triangular, 54
Union Jack, 57
Union Jack modified Greens function, 196
Laves lattice, 56
left eigenvectors, 264
linear
system, 134
transport, 132
link
addition, 46, 49
removal, 46
weight, 136
Mbius strip, 79, 149
maple leaf lattice, 56
martini lattice, 57
matrix
block circulant, 268
circulant, 267
positive definite, 263
positive semidefinite, 12, 15

power, 262
MoorePenrose
inverse, 165
multiple eigenvalue, 261
multiplicity
algebraic, 261
geometric, 264
nearest neighbor, 28
neighborhood, 28
network
damaged, 234
discontiguous, 172
embedded, 130, 134, 142, 161
fabricated, 41
isolated, 130, 134, 142, 164, 223
reinforced, 240
transport, 130
Neumann boundary condition, 6
Newtons law, 79
node
clustering, 32
degree, 18, 28
weighed, 137
strength, 137
nonlinear transport, 133
oriented incidence matrix, 30
normalized, 38
weighed, 139
pairwise resistance, 220
mean, 228
percolation threshold, 246
bond or link, 59
site or node, 61
periodic
boundary conditions, 13
graph, 20, 69
Poiseuille law, 132
Poisson equation, 1, 216
polynomial, characteristic, 260
positive
definite matrix eigenvalues, 263
semidefinite matrix, 12, 15, 140, 143
puzzle lattice, 56
random
graph, 31
walk, 213, 227
284 / / I N D E X
resistance
distance, 228
effective, 228
pairwise, 220
ruby lattice, 56
ShermanMorrison formula, 269, 273
simple cubic lattice, 59, 122
snub
square lattice, 56
spanning tree, 36
spectral
expansion, 36
partitioning, 36
radius, 261
spectrum of a matrix, 261
square lattice, 53, 67, 145
damaged, 247
square octagon lattice, 55
star lattice, 55
structure function, 213
Sylvesters determinant theorem, 271
symmetric matrix eigenvalues, 262
tetrakis lattice, 57
Toeplitz matrix, 4
trace of a matrix, 261
transport
linear, 132
nonlinear, 133
tree, 31
spanning, 36
triangular
lattice, 54
matrix, 261
truss, 26
unconnected graph, 30
undirected graph, 30
Union Jack lattice, 57, 150
modified, 93
weight of an edge, 136

Weyls theorem, 140
WignerSeitz cell, 51
Woodbury formula, 269

An Introduction To Grids Graphs and Networks

Cargado por

Información del documento

Título original

Derechos de autor

Formatos disponibles

Compartir este documento

Compartir o incrustar documentos

Opciones para compartir

¿Le pareció útil este documento?

¿Este contenido es inapropiado?

Copyright:

Formatos disponibles

An Introduction To Grids Graphs and Networks

Cargado por

Copyright:

Formatos disponibles

AN INTRODUCTION TO GRIDS,

GRAPHS, AND NETWORKS

Oxford University Press 2014

1.1. Poisson Equation in One Dimension

1.2. Dirichlet Boundary Condition at Both Ends

1.3. NeumannDirichlet Boundary Conditions

1.4. DirichletNeumann Boundary Conditions

1.5. Neumann Boundary Conditions

1.6. Periodic Boundary Conditions

1.7. One-Dimensional Graphs

1.8. Periodic One-Dimensional Graphs

2. Graphs and Networks

Connected and Unconnected Graphs

2.2. Laplacian Matrix

2.3. Cubic Network

2.4. Fabricated Networks

2.5. Link Removal and Addition

2.6. Infinite Lattices

2.7. Percolation Thresholds

3.1.3. Doubly Periodic Network

3.2. Mbius Strips

3.3. Hexagonal Lattice

3.4. Modified Union Jack Lattice

3.5. Honeycomb Lattice

3.6. Kagom Lattice

3.7. Simple Cubic Lattice

3.8. Body-Centered Cubic (bcc) Lattice

3.9. Face-Centered Cubic (fcc) Lattice

4.1. Transport Laws and Conventions

4.2. Uniform Conductances

4.3. Arbitrary Conductances

Weighed Node Degrees

4.4. Nodal Balances in Arbitrary Networks

4.6. Finite Difference Grids

4.7. Finite Element Grids

5.1. Embedded Networks

5.2. Isolated Networks

5.3. Lattice Greens Functions

5.4. Square Lattice

5.4.3. Helmholtz Equation Greens Function

5.5. Hexagonal Lattice

5.6. Modified Union Jack Lattice

5.7. Honeycomb Lattice

5.8. Simple Cubic Lattice

5.9. Body-Centered Cubic (bcc) Lattice

5.10. Face-Centered Cubic (fcc) Lattice

5.11. Free-Space Lattice Greens Functions

5.12. Finite Difference Solution in Terms of Greens Functions

6.1. Pairwise Resistance

6.2. Mean Pairwise Resistance

6.3. Damaged Networks

6.3.3. One Damaged Link

6.4. Reinforced Networks

6.5. Damaged Lattices

6.6. Damaged Square Lattice

6.7. Damaged Honeycomb Lattice

6.8. Damaged Hexagonal Lattice

A.1. Eigenvalues and Eigenvectors

A.2. The Characteristic Polynomial

A.3. Eigenvectors and Principal Vectors

A.4. Circulant Matrices

A.5. Block Circulant Matrices

B. The ShermanMorrison and Woodbury Formulas