Experience Seamless Full Ebook Downloads for Every Genre at textbookfull.

com

Probability on graphs random processes on graphs

and lattices Second Edition Grimmett

https://textbookfull.com/product/probability-on-graphs-
random-processes-on-graphs-and-lattices-second-edition-
grimmett/

OR CLICK BUTTON

DOWNLOAD NOW

Explore and download more ebook at https://textbookfull.com

 Recommended digital products (PDF, EPUB, MOBI) that
you can download immediately if you are interested.

Probability on Graphs Random Processes on Graphs and

Lattices Geoffrey Grimmett

https://textbookfull.com/product/probability-on-graphs-random-
processes-on-graphs-and-lattices-geoffrey-grimmett/

textboxfull.com

Random Walks and Heat Kernels on Graphs 1st Edition Martin

T. Barlow

https://textbookfull.com/product/random-walks-and-heat-kernels-on-
graphs-1st-edition-martin-t-barlow/

textboxfull.com

Generating Random Networks and Graphs 1st Edition Annibale

https://textbookfull.com/product/generating-random-networks-and-
graphs-1st-edition-annibale/

textboxfull.com

Random Graphs Geometry and Asymptotic Structure 1st

Edition Michael Krivelevich

https://textbookfull.com/product/random-graphs-geometry-and-
asymptotic-structure-1st-edition-michael-krivelevich/

textboxfull.com
Graphs, Algorithms, and Optimization, Second Edition Kocay

https://textbookfull.com/product/graphs-algorithms-and-optimization-
second-edition-kocay/

textboxfull.com

Graphs, Algorithms, and Optimization, Second Edition

William Kocay

https://textbookfull.com/product/graphs-algorithms-and-optimization-
second-edition-william-kocay/

textboxfull.com

Schaum's Outline of Probability, Random Variables, and

Random Processes, Fourth Edition Hwei P. Hsu

https://textbookfull.com/product/schaums-outline-of-probability-
random-variables-and-random-processes-fourth-edition-hwei-p-hsu/

textboxfull.com

Optimization Algorithms for Networks and Graphs, Second

Edition, James Evans

https://textbookfull.com/product/optimization-algorithms-for-networks-
and-graphs-second-edition-james-evans/

textboxfull.com

Applied Probability From Random Sequences to Stochastic

Processes Valérie Girardin

https://textbookfull.com/product/applied-probability-from-random-
sequences-to-stochastic-processes-valerie-girardin/

textboxfull.com
Cambridge University Press
978-1-108-43817-9 — Probability on Graphs
2nd Edition
Frontmatter
More Information

Probability on Graphs

Second Edition
This introduction to some of the principal models in the theory of disordered systems
leads the reader through the basics to the very edge of contemporary research, with
minimal technical fuss. Topics covered include random walks, percolation,
self-avoiding walks, interacting particle systems, uniform spanning trees, and random
graphs, as well as the Ising, Potts, and random-cluster models for ferromagnetism, and
the Lorentz model for motion in a random medium. Schramm-Löwner evolutions
(SLE) arise in various contexts. This new edition features topics in which there has
been major recent progress, including the exact value of the connective constant of the
hexagonal lattice and the critical point of the random-cluster model on the square
lattice.
The choice of topics is strongly motivated by modern applications, and focuses on
areas that merit further research. Special features include a simple account of
Smirnov’s proof of Cardy’s formula for critical percolation, and an account of the
theory of influence and sharp-thresholds. Accessible to a wide audience of
mathematicians and physicists, this book can be used as a graduate course text. Each
chapter ends with a range of exercises.

G E O F F R E Y G R I M M E T T is Professor of Mathematical Statistics in the Statistical

Laboratory at the University of Cambridge. He has written numerous research articles
on probability theory, as well as popular research books on percolation and the
random-cluster model. In addition, he is a co-author, along with David Stirzaker and
Dominic Welsh, of two successful textbooks on probability and random processes at
the undergraduate and postgraduate levels. He has served as Master of Downing
College since 2013 and was elected to the Royal Society in 2014.

© in this web service Cambridge University Press www.cambridge.org

Cambridge University Press
978-1-108-43817-9 — Probability on Graphs
2nd Edition
Frontmatter
More Information

I N S T I T U T E O F M AT H E M AT I C A L S TAT I S T I C S
TEXTBOOKS

Editorial Board
D. R. Cox (University of Oxford)
B. Hambly (University of Oxford)
S. Holmes (Stanford University)
J. Wellner (University of Washington)

IMS Textbooks give introductory accounts of topics of current concern suitable for
advanced courses at master’s level, for doctoral students and for individual study. They
are typically shorter than a fully developed textbook, often arising from material
created for a topical course. Lengths of 100–290 pages are envisaged. The books
typically contain exercises.

Other Books in the Series

1. Probability on Graphs, by Geoffrey Grimmett

2. Stochastic Networks, by Frank Kelly and Elena Yudovina
3. Bayesian Filtering and Smoothing, by Simo Särkkä
4. The Surprising Mathematics of Longest Increasing Subsequences, by Dan Romik
5. Noise Sensitivity of Boolean Functions and Percolation, by Christophe Garban and
Jeffrey E. Steif
6. Core Statistics, by Simon N. Wood
7. Lectures on the Poisson Process, by Günter Last and Mathew Penrose
8. Probability on Graphs (Second Edition), by Geoffrey Grimmett

© in this web service Cambridge University Press www.cambridge.org

Cambridge University Press
978-1-108-43817-9 — Probability on Graphs
2nd Edition
Frontmatter
More Information

Probability on Graphs

Random Processes on Graphs and Lattices

Second Edition

GEOFFREY GRIMMETT
Statistical Laboratory
University of Cambridge

© in this web service Cambridge University Press www.cambridge.org

Cambridge University Press
978-1-108-43817-9 — Probability on Graphs
2nd Edition
Frontmatter
More Information

University Printing House, Cambridge CB2 8BS, United Kingdom

One Liberty Plaza, 20th Floor, New York, NY 10006, USA
477 Williamstown Road, Port Melbourne, VIC 3207, Australia
314–321, 3rd Floor, Plot 3, Splendor Forum, Jasola District Centre,
New Delhi – 110025, India
79 Anson Road, #06–04/06, Singapore 079906

Cambridge University Press is part of the University of Cambridge.

It furthers the University’s mission by disseminating knowledge in the pursuit of
education, learning, and research at the highest international levels of excellence.

www.cambridge.org
Information on this title: www.cambridge.org/9781108438179
DOI: 10.1017/9781108528986
c Geoffrey Grimmett 2018

This publication is in copyright. Subject to statutory exception
and to the provisions of relevant collective licensing agreements,
no reproduction of any part may take place without the written
permission of Cambridge University Press.
First published 2018
Printed in the United States of America by Sheridan Books, Inc.
A catalogue record for this publication is available from the British Library.
ISBN 978-1-108-43817-9 Paperback
Cambridge University Press has no responsibility for the persistence or accuracy of
URLs for external or third-party internet websites referred to in this publication
and does not guarantee that any content on such websites is, or will remain,
accurate or appropriate.

© in this web service Cambridge University Press www.cambridge.org

Cambridge University Press
978-1-108-43817-9 — Probability on Graphs
2nd Edition
Frontmatter
More Information

Contents

Preface ix

1 Random Walks on Graphs 1

1.1 Random Walks and Reversible Markov Chains 1
1.2 Electrical Networks 3
1.3 Flows and Energy 8
1.4 Recurrence and Resistance 11
1.5 Pólya’s Theorem 14
1.6 Graph Theory 16
1.7 Exercises 18

2 Uniform Spanning Tree 21

2.1 Definition 21
2.2 Wilson’s Algorithm 23
2.3 Weak Limits on Lattices 28
2.4 Uniform Forest 31
2.5 Schramm–Löwner Evolutions 32
2.6 Exercises 36

3 Percolation and Self-Avoiding Walks 39

3.1 Percolation and Phase Transition 39
3.2 Self-Avoiding Walks 42
3.3 Connective Constant of the Hexagonal Lattice 45
3.4 Coupled Percolation 53
3.5 Oriented Percolation 53
3.6 Exercises 56

4 Association and Influence 59

4.1 Holley Inequality 59
4.2 FKG Inequality 62
4.3 BK Inequality 63

© in this web service Cambridge University Press www.cambridge.org

Cambridge University Press
978-1-108-43817-9 — Probability on Graphs
2nd Edition
Frontmatter
More Information

vi Contents

4.4 Hoeffding Inequality 65

4.5 Influence for Product Measures 67
4.6 Proofs of Influence Theorems 72
4.7 Russo’s Formula and Sharp Thresholds 80
4.8 Exercises 83

5 Further Percolation 86
5.1 Subcritical Phase 86
5.2 Supercritical Phase 90
5.3 Uniqueness of the Infinite Cluster 96
5.4 Phase Transition 99
5.5 Open Paths in Annuli 103
5.6 The Critical Probability in Two Dimensions 107
5.7 Cardy’s Formula 115
5.8 The Critical Probability via the Sharp-Threshold Theorem 126
5.9 Exercises 130

6 Contact Process 131

6.1 Stochastic Epidemics 131
6.2 Coupling and Duality 133
6.3 Invariant Measures and Percolation 135
6.4 The Critical Value 137
6.5 The Contact Model on a Tree 139
6.6 Space–Time Percolation 142
6.7 Exercises 145

7 Gibbs States 146

7.1 Dependency Graphs 146
7.2 Markov and Gibbs Random Fields 148
7.3 Ising and Potts Models 152
7.4 Exercises 154

8 Random-Cluster Model 156

8.1 The Random-Cluster, Ising, and Potts Models 156
8.2 Basic Properties 159
8.3 Infinite-Volume Limits and Phase Transition 160
8.4 Open Problems 165
8.5 In Two Dimensions 167
8.6 Proof of the Critical Point in Two Dimensions 173
8.7 Random Even Graphs 183
8.8 Exercises 186

© in this web service Cambridge University Press www.cambridge.org

Cambridge University Press
978-1-108-43817-9 — Probability on Graphs
2nd Edition
Frontmatter
More Information

Contents vii

9 Quantum Ising Model 189

9.1 The Model 189
9.2 Continuum Random-Cluster Model 191
9.3 Quantum Ising via Random-Cluster Model 193
9.4 Long-Range Order 198
9.5 Entanglement in One Dimension 199
9.6 Exercises 203

10 Interacting Particle Systems 204

10.1 Introductory Remarks 204
10.2 Contact Model 206
10.3 Voter Model 207
10.4 Exclusion Model 211
10.5 Stochastic Ising Model 214
10.6 Exercises 218

11 Random Graphs 219

11.1 Erdős–Rényi Graphs 219
11.2 Giant Component 221
11.3 Independence and Colouring 226
11.4 Exercises 231

12 Lorentz Gas 233

12.1 Lorentz Model 233
12.2 The Square Lorentz Gas 234
12.3 In the Plane 237
12.4 Exercises 238

References 240
Index 261

© in this web service Cambridge University Press www.cambridge.org

Cambridge University Press
978-1-108-43817-9 — Probability on Graphs
2nd Edition
Frontmatter
More Information

Preface

Within the menagerie of objects studied in contemporary probability theory,

a number of related animals have attracted great interest amongst proba-
bilists and physicists in recent years. The inspiration for many of these
objects comes from physics, but the mathematical subject has taken on a
life of its own and many beautiful constructions have emerged. The overall
target of these notes is to identify some of these topics, and to develop their
basic theory at a level suitable for mathematics graduates.
If the two principal characters in these notes are random walk and per-
colation, they are only part of the rich theory of uniform spanning trees,
self-avoiding walks, random networks, models for ferromagnetism and the
spread of disease, and motion in random environments. This is an area that
has attracted many fine scientists, by virtue, perhaps, of its special mixture
of modelling and problem-solving. There remain many open problems. It
is the experience of the author that these may be explained successfully to
a graduate audience open to inspiration and provocation.
The material described here may be used for personal study and also as
the bases of lecture courses of between 16 and 48 hours duration. Little is
assumed about the mathematical background of the audience beyond some
basic probability theory, but students should be willing to get their hands
dirty if they are to profit. Care should be taken in the setting of examinations,
since problems can be unexpectedly difficult. Successful examinations may
be designed, and some help is offered through the inclusion of exercises
at the ends of chapters. As an alternative to a conventional examination,
students could be asked to deliver presentations on aspects and extensions
of the topics studied.
Chapter 1 is devoted to the relationship between random walks (on graphs)
and electrical networks. This leads to the Thomson and Rayleigh principles,
and thence to a proof of Pólya’s theorem. In Chapter 2, we describe Wilson’s
algorithm for constructing a uniform spanning tree (UST), and we discuss
boundary conditions and weak limits for UST on a lattice. This chapter
includes a brief introduction to Schramm–Löwner evolutions (SLEs).

© in this web service Cambridge University Press www.cambridge.org

Cambridge University Press
978-1-108-43817-9 — Probability on Graphs
2nd Edition
Frontmatter
More Information

x Preface

Percolation theory appears first in Chapter 3, together with a short intro-

duction to self-avoiding walks. Correlation inequalities and other general
techniques are described in Chapter 4. A special feature of this part of the
book is a fairly full treatment of influence and sharp-threshold theorems for
product measures, and more generally for monotone measures.
We return to the basic theory of percolation in Chapter 5, including a full
account of Smirnov’s proof of Cardy’s formula. This is followed in Chapter
6 by a study of the contact model on lattices and trees.
Chapter 7 begins with a proof of the equivalence of Gibbs states and
Markov fields, and continues with an introduction to the Ising and Potts
models. Chapter 8 contains an account of the random-cluster model. The
quantum Ising model features in the next chapter, particularly through its re-
lationship to a continuum random-cluster model and the consequent analysis
using stochastic geometry.
Interacting particle systems form the basis of Chapter 10. This is a large
field in its own right, and little is done here beyond introductions to the
contact, voter, and exclusion models, and to the stochastic Ising model.
Chapter 11 is devoted to random graphs of Erdős–Rényi type. There are
accounts of the giant cluster, and of the chromatic number via an application
of Hoeffding’s inequality for the tail of a martingale.
The final Chapter 12 contains one of the most notorious open problems
in stochastic geometry, namely the Lorentz model (or Ehrenfest wind–tree
model) on the square lattice.
This text is based in part on courses given by the author within Part 3 of
the Mathematical Tripos at Cambridge University over a period of several
years. They have been prepared in the present form as background mate-
rial for lecture courses presented to outstanding audiences of students and
professors at the 2008 PIMS–UBC Summer School in Probability and dur-
ing the programme on Statistical Mechanics at the Institut Henri Poincaré,
Paris, during the last quarter of 2008. The book was written in part during
a visit to the Mathematics Department at UCLA (with partial support from
NSF grant DMS-0301795), to which the author expresses his gratitude for
the warm welcome received there, and in part during programmes at the
Isaac Newton Institute and the Institut Henri Poincaré–Centre Emile Borel.
Throughout this work, pointers are included to more extensive accounts
of the topics covered. The selection of references is intended to be useful
rather than comprehensive.
The author thanks four artists for permission to include their work: Tom
Kennedy (Figure 2.1), Oded Schramm (Figures 2.2–2.4), Raphaël Cerf (Fig-
ure 5.2), and Julien Dubédat (Figure 5.17). The section on influence has ben-

© in this web service Cambridge University Press www.cambridge.org

Cambridge University Press
978-1-108-43817-9 — Probability on Graphs
2nd Edition
Frontmatter
More Information

Preface xi

efited from conversations with Rob van den Berg and Tom Liggett. Stanislav
Smirnov and Wendelin Werner have consented to the inclusion of some of
their neat arguments, hitherto unpublished. Several readers have proposed
suggestions and corrections. Thank you, everyone!

G. R. G.
Cambridge
April 2010

Preface to the Second Edition

The major additions in this new edition include: a proof of the connective
constant of the hexagonal lattice (Theorem 3.14), an improved influence
theorem for general product spaces (Theorem 4.38), a streamlined proof of
exponential decay for subcritical percolation (Theorem 5.1), and a proof of
the critical point of the random-cluster model on the square lattice (Theorem
8.25).
The author is grateful to students and colleagues for their suggestions for
improvements. Special thanks are due to Naser Talebizadeh Sardari, Claude
Bélisle, Svante Janson, and Russell Lyons. Some of the writing was done
during a visit to the Statistics Department of the University of California at
Berkeley, with partial support from UC Berkeley and from the Engineering
and Physical Science Research Council under grant EP/I03372X/1.

July 2017

© in this web service Cambridge University Press www.cambridge.org

 1

Random Walks on Graphs

The theory of electrical networks is a fundamental tool for studying

the recurrence of reversible Markov chains. The Kirchhoff laws and
Thomson principle permit a neat proof of Pólya’s theorem for random
walk on a d-dimensional grid.

1.1 Random Walks and Reversible Markov Chains

A basic knowledge of probability theory is assumed in this volume. Readers
keen to acquire this are referred to [150] for an elementary introduction, and
to [148] for a somewhat more advanced account. We shall generally use the
letter P to denote a generic probability measure, with more specific notation
when helpful. The expectation of a random variable f will be written as
either P( f ) or E( f ).
Only a little knowledge is assumed about graphs, and many readers will
have sufficient acquaintance already. Others are advised to consult Section
1.6. Of the many books on graph theory, we mention [50].
Let G = (V , E) be a finite or countably infinite graph,which we generally
assume, for simplicity, to have neither loops nor multiple edges. If G is
infinite, we shall usually assume in addition that every vertex-degree is
finite. A particle moves around the vertex-set V . Having arrived at the
vertex Sn at time n, its next position Sn+1 is chosen uniformly at random
from the set of neighbours of Sn . The trajectory of the particle is called a
symmetric random walk (SRW) on G.
Two of the basic questions concerning symmetric random walk are:
1. Under what conditions is the walk recurrent, in that it returns (almost
surely) to its starting point?
2. How does the distance between S0 and Sn behave as n → ∞?
The above SRW is symmetric in that the jumps are chosen uniformly
from the set of available neighbours. In a more general process, we take a
function w : E → (0, ∞), and we jump along the edge e with probability
proportional to we .

17:53:41
.002
2 Random Walks on Graphs

Any reversible Markov chain1 on the set V gives rise to such a walk as
follows. Let Z = (Z n : n ≥ 0) be a Markov chain on V with transition
matrix P, and assume that Z is reversible with respect to some positive
function π : V → (0, ∞), which is to say that
(1.1) πu pu,v = πv pv,u , u, v ∈ V .
With each distinct pair u, v ∈ V , we associate the weight
(1.2) wu,v = πu pu,v ,
noting by (1.1) that wu,v = wv,u . Then
wu,v
(1.3) pu,v = , u, v ∈ V ,
Wu
where 
Wu = wu,v , u ∈ V.
v∈V

That is, given that Z n = u, the chain jumps to a new vertex v with probability
proportional to wu,v . This may be set in the context of a random walk on
the graph with vertex-set V and edge-set E containing all e = u, v such
that pu,v > 0. With edge e ∈ E we associate the weight we = wu,v .
In this chapter, we develop the relationship between random walks on G
and electrical networks on G. There are some excellent accounts of this
subject area, and the reader is referred to the books of Doyle and Snell
[83], Lyons and Peres [221], and Aldous and Fill [19], amongst others. The
connection between these two topics is made via the so-called ‘harmonic
functions’ of the random walk.

1.4 Definition Let U ⊆ V , and let Z be a Markov chain on V with transi-

tion matrix P, that is reversible with respect to the positive function π. The
function f : V → R is harmonic on U (with respect to P) if

f (u) = pu,v f (v), u ∈ U,
v∈V

or, equivalently, if f (u) = E( f (Z 1 ) | Z 0 = u) for u ∈ U .

From the pair (P, π), we can construct the graph G as above, and the
weight function w as in (1.2). We refer to the pair (G, w) as the weighted
graph associated with (P, π). We shall speak of f as being harmonic (for
(G, w)) if it is harmonic with respect to P.

1 Accounts of Markov chain theory are found in [148, Chap. 6] and [150, Chap. 12].

17:53:41
.002
 1.2 Electrical Networks 3

The so-called hitting probabilities are basic examples of harmonic func-

tions for the chain Z. Let U ⊆ V , W = V \ U , and s ∈ U . For u ∈ V , let
g(u) be the probability that the chain, started at u, hits s before W . That is,
g(u) = Pu (Z n = s for some n < TW ),
where
TW = inf{n ≥ 0 : Z n ∈ W }
is the first-passage time to W , and Pu (·) = P(· | Z 0 = u) denotes the
conditional probability measure given that the chain starts at u.
1.5 Theorem The function g is harmonic on U \ {s}.
Evidently, g(s) = 1, and g(v) = 0 for v ∈ W . We speak of these values
of g as being the ‘boundary conditions’ of the harmonic function g. See
Exercise 1.13 for the uniqueness of harmonic functions with given boundary
conditions.
Proof. This is an elementary exercise using the Markov property. For
u∈/ W ∪ {s},
   
g(u) = pu,v Pu Z n = s for some n < TW  Z 1 = v
v∈V

= pu,v g(v),
v∈V
as required. 

1.2 Electrical Networks

Throughout this section, G = (V , E) is a finite graph with neither loops
nor multiple edges, and w : E → (0, ∞) is a weight function on the edges.
We shall assume further that G is connected.
We may build an electrical network with diagram G, in which the edge
e has conductance we (or, equivalently, resistance 1/we ). Let s, t ∈ V
be distinct vertices termed sources, and write S = {s, t} for the source-set.
Suppose we connect a battery across the pair s, t. It is a physical observation
that electrons flow along the wires in the network. The flow is described by
the so-called Kirchhoff laws, as follows.
To each edge e = u, v, there are associated (directed) quantities φu,v
and i u,v , called the potential difference from u to v, and the current from u
to v, respectively. These are antisymmetric,
φu,v = −φv,u , i u,v = −i v,u .

17:53:41
.002
4 Random Walks on Graphs

1.6 Kirchhoff’s potential law The cumulative potential difference around

any cycle v1 , v2 , . . . , vn , vn+1 = v1 of G is zero, that is,

n
(1.7) φv j ,v j+1 = 0.
j=1

1.8 Kirchhoff’s current law The total current flowing out of any vertex
u ∈ V other than the source-set is zero, that is,

(1.9) i u,v = 0, u = s, t.
v∈V

The relationship between resistance/conductance, potential difference,

and current is given by Ohm’s law.
1.10 Ohm’s law For any edge e = u, v,
i u,v = we φu,v .

Kirchhoff’s potential law is equivalent to the statement that there exists

a function φ : V → R, called a potential function, such that
φu,v = φ(v) − φ(u), u, v ∈ E.
Since φ is determined up to an additive constant, we are free to pick the
potential of any single vertex. Note our convention that current flows uphill:
i u,v has the same sign as φu,v = φ(v) − φ(u).
1.11 Theorem A potential function is harmonic on the set of all vertices
other than the source-set.
Proof. Let U = V \ {s, t}. By Kirchhoff’s current law and Ohm’s law,

wu,v [φ(v) − φ(u)] = 0, u ∈ U,
v∈V

which is to say that

 wu,v
φ(u) = φ(v), u ∈ U,
Wu
v∈V

where 
Wu = wu,v .
v∈V

That is, φ is harmonic on U . 

17:53:41
.002
 1.2 Electrical Networks 5

We can use Ohm’s law to express potential differences in terms of cur-

rents, and thus the two Kirchhoff laws may be viewed as concerning currents
only. Equation (1.7) becomes

n
i v j ,v j+1
(1.12) = 0,
wv j ,v j+1 
j=1

valid for any cycle v1 , v2 , . . . , vn , vn+1 = v1 . With (1.7) written thus, each
law is linear in the currents, and the superposition principle follows.
1.13 Theorem (Superposition principle) If i 1 and i 2 are solutions of the
two Kirchhoff laws with the same source-set then so is the sum i 1 + i 2 .
Next we introduce the concept of a ‘flow’ on a graph.
1.14 Definition Let s, t ∈ V , s = t. An s/t-flow j is a vector j =
( ju,v : u, v ∈ V , u = v), such that:
(a) ju,v = − jv,u ,
(b) ju,v = 0 whenever u  v,

(c) for any u = s, t, we have that v∈V ju,v = 0.
The vertices s and t are called the ‘source’ and ‘sink’ of an s/t flow, and
we usually abbreviate ‘s/t flow’ to ‘flow’. For any flow j , we write

Ju = ju,v , u ∈ V,
v∈V

noting by (c) above that Ju = 0 for u = s, t. Thus,

  
Js + Jt = Ju = ju,v = 21 ( ju,v + jv,u ) = 0.
u∈V u,v∈V u,v∈V

Therefore, Js = −Jt , and we call |Js | the size of the flow j , denoted | j |. If
|Js | = 1, we call j a unit flow. We shall normally take Js > 0, in which
case s is the source and t is the sink of the flow, and we say that j is a flow
from s to t.
Note that any solution i to the Kirchhoff laws with source-set {s, t} is an
s/t flow.
1.15 Theorem Let i 1 and i 2 be two solutions of the Kirchhoff laws with
the same source and sink and equal size. Then i 1 = i 2 .
Proof. By the superposition principle, j = i 1 − i 2 satisfies the two Kirch-
hoff laws. Furthermore, under the flow j , no current enters or leaves
the system. Therefore, Jv = 0 for all v ∈ V . Suppose ju 1 ,u 2 > 0 for
some edge u 1 , u 2 . By the Kirchhoff current law, there exists u 3 such that

17:53:41
.002
6 Random Walks on Graphs

ju 2 ,u 3 > 0. Since |V | < ∞, there exists by iteration a cycle u l , u l+1 , . . . ,

u m , u m+1 = u l such that ju k ,u k+1 > 0 for k = l, l + 1, . . . , m. By Ohm’s
law, the corresponding potential function satisfies
φ(u l ) < φ(u l+1 ) < · · · < φ(u m+1 ) = φ(u l ),
a contradiction. Therefore, ju,v = 0 for all u, v. 
For a given size of input current, and given source s and sink t, there can
be no more than one solution to the two Kirchhoff laws, but is there a solution
at all? The answer is of course affirmative, and the unique solution can be
expressed explicitly in terms of counts of spanning trees.2 Consider first the
special case when we = 1 for all e ∈ E. Let N be the number of spanning
trees of G. For any edge a, b, let (s, a, b, t) be the property of spanning
trees that: the unique s/t path in the tree passes along the edge a, b in the
direction from a to b. Let N (s, a, b, t) be the set of spanning trees of G
with the property (s, a, b, t), and let N (s, a, b, t) = |N (s, a, b, t)|.
1.16 Theorem The function
1 
(1.17) i a,b = N (s, a, b, t) − N (s, b, a, t) , a, b ∈ E,
N
defines a unit flow from s to t satisfying the Kirchhoff laws.
Let T be a spanning tree of G chosen uniformly at random from the set
T of all such spanning trees. By Theorem 1.16 and the previous discussion,
the unique solution to the Kirchhoff laws with source s, sink t, and size 1 is
given by
   
i a,b = P T has (s, a, b, t) − P T has (s, b, a, t) .
We shall return to uniform spanning trees in Chapter 2.
We prove Theorem 1.16 next. Exactly the same proof is valid in the case
of general conductances we . In that case, we define the weight of a spanning
tree T as
w(T ) = we ,
e∈T
and we set
 
(1.18) N∗ = w(T ), N ∗ (s, a, b, t) = w(T ).
T ∈T T with (s,a,b,t)
The conclusion of Theorem 1.16 holds in this setting with
1  
i a,b = ∗ N ∗ (s, a, b, t) − N ∗ (s, b, a, t) , a, b ∈ E.
N
2 This was discovered in an equivalent form by Kirchhoff in 1847, [188].

17:53:41
.002
 1.2 Electrical Networks 7

Proof of Theorem 1.16. We first check the Kirchhoff current law. In every
spanning tree T , there exists a unique vertex b such that the s/t path of
T contains the edge s, b, and the path traverses this edge from s to b.
Therefore,

N (s, s, b, t) = N, N (s, b, s, t) = 0 for b ∈ V .
b∈V

By (1.17),

i s,b = 1,
b∈V


and, by a similar argument, b∈V i b,t = 1.
Let T be a spanning tree of G. The contribution towards the quantity
i a,b , made by T , depends on the s/t path π of T and equals

N −1 if π passes along a, b from a to b,

−1
(1.19) −N if π passes along a, b from b to a,
0 if π does not contain the edge a, b.

Let v ∈ V , v = s, t, and write Iv = w∈V i v,w . If v ∈ π, the contribution
of T towards Iv is N −1 − N −1 = 0 since π arrives at v along some edge of
the form a, v and departs from v along some edge of the form v, b. If
v∈/ π, then T contributes 0 to Iv . Summing over T , we obtain that Iv = 0
for all v = s, t, as required for the Kirchhoff current law.
We next check the Kirchhoff potential law. Let v1 , v2 , . . . , vn , vn+1 = v1
be a cycle C of G. We shall show that


n
(1.20) i v j ,v j+1 = 0,
j=1

and this will confirm (1.12), on recalling that we = 1 for all e ∈ E. It is

more convenient in this context to work with ‘bushes’ than spanning trees.
A bush (or, more precisely, an s/t bush) is defined to be a forest on V
containing exactly two trees, one denoted Ts and containing s, and the other
denoted Tt and containing t. We write (Ts , Tt ) for this bush. Let e = a, b,
and let B(s, a, b, t) be the set of bushes with a ∈ Ts and b ∈ Tt . The
sets B(s, a, b, t) and N (s, a, b, t) are in one–one correspondence, since
the addition of e to B ∈ B(s, a, b, t) creates a unique member T = T (B)
of N (s, a, b, t), and vice versa.

17:53:41
.002
8 Random Walks on Graphs

By (1.19) and the above, a bush B = (Ts , Tt ) makes a contribution to

i a,b of

N −1 if B ∈ B(s, a, b, t),
−N −1 if B ∈ B(s, b, a, t),
0 otherwise.
Therefore, B makes a contribution towards the sum in (1.20) that is equal to
N −1 (F+ − F− ), where F+ (respectively, F− ) is the number of pairs v j , v j+1
of C, 1 ≤ j ≤ n, with v j ∈ Ts , v j+1 ∈ Tt (respectively, v j+1 ∈ Ts , v j ∈ Tt ).
Since C is a cycle, we have F+ = F− , whence each bush contributes 0 to
the sum and (1.20) is proved. 

1.3 Flows and Energy

Let G = (V , E) be a connected graph as before. Let s, t ∈ V be distinct
vertices, and let j be an s/t flow. With we the conductance of the edge e,
the (dissipated) energy of j is defined as
 
E( j ) = 2
ju,v /we = 21 2
ju,v /wu,v .
e=u,v∈E u,v∈V
u∼v

The following piece of linear algebra will be useful.

1.21 Proposition Let ψ : V → R, and let j be an s/t flow. Then


[ψ(t) − ψ(s)]Js = 21 [ψ(v) − ψ(u)] ju,v .
u,v∈V

Proof. By the properties of a flow,

  
[ψ(v) − ψ(u)] ju,v = ψ(v)(−Jv ) − ψ(u)Ju
u,v∈V v∈V u∈V
= −2[ψ(s)Js + ψ(t)Jt ]
= 2[ψ(t) − ψ(s)]Js ,
as required. 

Let φ and i satisfy the two Kirchhoff laws. We apply Proposition 1.21
with ψ = φ and j = i to find by Ohm’s law that
(1.22) E(i ) = [φ(t) − φ(s)]Is .

17:53:41
.002
 1.3 Flows and Energy 9

That is, the energy of the true current-flow i from s to t equals the energy
dissipated in a (notional) single s, t edge carrying the same potential dif-
ference and total current. The conductance Weff of such an edge would
satisfy Ohm’s law, that is,
(1.23) Is = Weff [φ(t) − φ(s)],
and we define the effective conductance Weff by this equation. The effective
resistance is
1
(1.24) Reff = ,
Weff
which, by (1.22) and (1.23), equals E(i )/Is2 . We state this as a lemma.

1.25 Lemma The effective resistance Reff of the network between vertices
s and t equals the dissipated energy when a unit flow passes from s to t.

It is useful to be able to do calculations. Electrical engineers have devised

a variety of formulaic methods for calculating the effective resistance of a
network, of which the simplest are the series and parallel laws, illustrated
in Figure 1.1.

e f
f
Figure 1.1 Two edges e and f in parallel and in series.

1.26 Series law Two resistors of size r1 and r2 in series may be replaced
by a single resistor of size r1 + r2 .

1.27 Parallel law Two resistors of size r1 and r2 in parallel may be replaced
by a single resistor of size R, where R −1 = r1−1 + r2−1 .
A third such rule, the so-called ‘star–triangle transformation’, may be
found at Exercise 1.5. The following ‘variational principle’ has many uses.

1.28 Theorem (Thomson principle) Let G = (V , E) be a connected

graph and (we : e ∈ E) strictly positive conductances. Let s, t ∈ V ,
s = t. Amongst all unit flows through G from s to t, the flow that satisfies
the Kirchhoff laws is the unique s/t flow i that minimizes the dissipated
energy. That is,
E(i ) = inf E( j ) : j a unit flow from s to t .

17:53:41
.002
10 Random Walks on Graphs

Proof. Let j be a unit flow from source s to sink t, and set k = j − i , where
i is the (unique) unit-flow solution to the Kirchhoff laws. Thus, k is a flow
with zero size. Now, with e = u, v and re = 1/we ,
 
2E( j ) = 2
ju,v re = (ku,v + i u,v )2 re
u,v∈V u,v∈V
  
= 2
ku,v re + 2
i u,v re + 2 i u,v ku,v re .
u,v∈V u,v∈V u,v∈V
Let φ be the potential function corresponding to i . By Ohm’s law and
Proposition 1.21,
 
i u,v ku,v re = [φ(v) − φ(u)]ku,v
u,v∈V u,v∈V
= 2[φ(t) − φ(s)]K s ,
which equals zero. Therefore, E( j ) ≥ E(i ), with equality if and only if
j = i. 
The Thomson ‘variational principle’ leads to a proof of the ‘obvious’ fact
that the effective resistance of a network is a non-decreasing function of the
resistances of individual edges.
1.29 Theorem (Rayleigh principle) The effective resistance Reff of the
network is a non-decreasing function of the edge-resistances (re : e ∈ E).
It is left as an exercise to show that Reff is a concave function of the vector
(re ). See Exercise 1.6.
Proof. Consider two vectors (re : e ∈ E) and (re : e ∈ E) of edge-
resistances with re ≤ re for all e. Let i and i  denote the corresponding unit
flows satisfying the Kirchhoff laws. By Lemma 1.25, with re = ru,v ,

Reff = 21 2
i u,v re
u,v∈V
u∼v


≤ 1
2 (i u,v )2 r e by the Thomson principle
u,v∈V
u∼v


≤ 1
2 (i u,v )2re since re ≤ re
u,v∈V
u∼v

= Reff ,
as required. 

17:53:41
.002
 1.4 Recurrence and Resistance 11

1.4 Recurrence and Resistance

Let G = (V , E) be an infinite connected graph with finite vertex-degrees,
and let (we : e ∈ E) be (strictly positive) conductances. We shall consider
a reversible Markov chain Z = (Z n : n ≥ 0) on the state space V with
transition probabilities given by (1.3). Our purpose is to establish a condition
on the pair (G, w) that is equivalent to the recurrence of Z.
Let 0 be a distinguished vertex of G, called the ‘origin’, and suppose
that Z 0 = 0. The graph-theoretic distance between two vertices u, v is the
number of edges in a shortest path between u and v, denoted δ(u, v). Let
n = {u ∈ V : δ(0, v) ≤ n},
∂n = n \ n−1 = {u ∈ V : δ(0, v) = n}.
We think of ∂n as the ‘boundary’ of n . Let G n be the subgraph of G
induced by the vertex-set n . We let G n be the graph obtained from G n by
identifying the vertices in ∂n as a single composite vertex denoted In . The
resulting finite graph G n may be considered as an electrical network with
sources 0 and In . Let Reff (n) be the effective resistance of this network. The
graph G n may be obtained from G n+1 by identifying all vertices lying in
∂n ∪{In+1 }, and thus, by the Rayleigh principle, Reff (n) is non-decreasing
in n. Therefore, the limit
Reff = lim Reff (n)
n→∞

exists.
1.30 Theorem The probability of ultimate return by Z to the origin 0 is
given by
1
P0 (Z n = 0 for some n ≥ 1) = 1 − ,
W0 Reff

where W0 = v: v∼0 w0,v .
The return probability is non-decreasing as W0 Reff increases. By the
Rayleigh principle, this can be achieved, for example, by removing an edge
of E that is not incident to 0. The removal of an edge incident to 0 can have
the opposite effect, since W0 decreases while Reff increases (see Figure 1.2).
A 0/∞ flow is a vector j = ( ju,v : u, v ∈ V , u = v) satisfying (1.14)(a),
(b) and also (c) for all u = 0. That is, it has source 0 but no sink.
1.31 Corollary
(a) The chain Z is recurrent if and only if Reff = ∞.
(b) The chain Z is transient if and only
 if2 there exists a non-zero 0/∞ flow
j on G whose energy E( j ) = e je /we satisfies E( j ) < ∞.

17:53:41
.002
12 Random Walks on Graphs

e
0

Figure 1.2 This is an infinite binary tree with two parallel edges joining
the origin to the root. When each edge has unit resistance, it is an easy
calculation that Reff = 32 , so the probability of return to 0 is 23 . If the
edge e is removed, this probability becomes 12 .

It is left as an exercise to extend this to countable graphs G without the

assumption of finite vertex-degrees.
Proof of Theorem 1.30. Let
gn (v) = Pv (Z hits ∂n before 0), v ∈ n .
By Theorem 1.5 and Exercise 1.13, gn is the unique harmonic function on
G n with boundary conditions
gn (0) = 0, gn (v) = 1 for v ∈ ∂n .
Therefore, gn is a potential function on G n viewed as an electrical network
with source 0 and sink In .
By conditioning on the first step of the walk, and using Ohm’s law,
P0 (Z returns to 0 before reaching ∂n )

=1− p0,v gn (v)
v: v∼0
 w0,v
=1− [gn (v) − gn (0)]
W0
v: v∼0
|i (n)|
=1− ,
W0
where i (n) is the flow of currents in G n , and |i (n)| is its size. By (1.23) and
(1.24), |i (n)| = 1/Reff (n). The theorem is proved on noting that
P0 (Z returns to 0 before reaching ∂n ) → P0 (Z n = 0 for some n ≥ 1)

17:53:41
.002
 1.5 Recurrence and Resistance 13

as n → ∞, by the continuity of probability measures. 

Proof of Corollary 1.31. Part (a) is an immediate consequence of Theorem

1.30, and we turn to part (b). By Lemma 1.25, there exists a unit flow i (n)
in G n with source 0 and sink In , and with energy E(i (n)) = Reff (n). Let i
be a non-zero 0/∞ flow; by dividing by its size, we may take i to be a unit
flow. When restricted to the edge-set E n of G n , i forms a unit flow from 0
to In . By the Thomson principle, Theorem 1.28,

E(i (n)) ≤ i e2 /we ≤ E(i ),
e∈E n

whence
E(i ) ≥ lim E(i (n)) = Reff .
n→∞

Therefore, by part (a), E(i ) = ∞ if the chain is recurrent.

Suppose, conversely, that the chain is transient. By diagonal selection,3
there exists a subsequence (n k ) along which i (n k ) converges to some limit
j (that is, i (n k )e → je for every e ∈ E). Since each i (n k ) is a unit flow
from the origin, j is a unit 0/∞ flow. Now,

E(i (n k )) = i (n k )2e /we
e∈E

≥ i (n k )2e /we
e∈E m

→ j (e)2 /we as k → ∞
e∈E m
→ E( j ) as m → ∞.

Therefore,
E( j ) ≤ lim Reff (n k ) = Reff < ∞,
k→∞

and j is a flow with the required properties. 

3 Diagonal selection: Let (x (n) : m, n ≥ 1) be a bounded collection of reals. There

m
exists an increasing sequence n 1 , n 2 , . . . of positive integers such that, for every m, the
limit lim k→∞ x m (n k ) exists.

17:53:41
.002
14 Random Walks on Graphs

1.5 Pólya’s Theorem

The d-dimensional cubic lattice Ld has vertex-set Zd and edges between any
two vertices that are Euclidean distance one apart. The following celebrated
theorem can be proved by estimating effective resistances.4
1.32 Pólya’s theorem [242] Symmetric random walk on the lattice Ld in
d dimensions is recurrent if d = 1, 2 and transient if d ≥ 3.
The advantage of the following proof of Pólya’s theorem over more stan-
dard arguments is its robustness with respect to the underlying graph. Sim-
ilar arguments are valid for graphs that are, in broad terms, comparable to
Ld when viewed as electrical networks.

Proof. For simplicity, and with only little loss of generality (see Exercise
1.10), we shall concentrate on the cases d = 2, 3. Let d = 2, for which case
we aim to show that Reff = ∞. This is achieved by finding an infinite lower
bound for Reff , and lower bounds can be obtained by decreasing individual
edge-resistances. The identification of two vertices of a network amounts
to the addition of a resistor with 0 resistance, and, by the Rayleigh principle,
the effective resistance of the network can only decrease.

0 1 2 3

Figure 1.3 The vertex labelled i is a composite vertex obtained by

identifying all vertices with distance i from 0. There are 8i − 4 edges of
L2 joining vertices i − 1 and i .

From L2 , we construct a new graph in which, for each k = 1, 2, . . . ,

the set ∂k = {v ∈ Z2 : δ(0, v) = k} is identified as a singleton. This
transforms L2 into the graph shown in Figure 1.3. By the series/parallel
laws and the Rayleigh principle,

n−1
1
Reff (n) ≥ ,
8i − 4
i=1

whence Reff (n) ≥ c log n → ∞ as n → ∞.

Suppose now that d = 3. There are at least two ways of proceeding.
We shall present one such route, taken from [222], and we shall then sketch
4 An amusing story is told in [243] about Pólya’s inspiration for this theorem.

17:53:41
.002
 1.5 Pólya’s Theorem 15

Cu Fu,v

(Fu,v )
0

Figure 1.4 The flow along the edge u, v is equal to the area of the
projection (Fu,v ) on the unit sphere centred at the origin, with a suitable
convention for its sign.

the second, which has its inspiration in [83]. By Corollary 1.31, it suffices
to construct a non-zero 0/∞ flow with finite energy. Let S be the surface
of the unit sphere of R3 with centre at the origin 0. Take u ∈ Z3 , u = 0,
and position a unit cube Cu in R3 with centre at u and edges parallel to the
axes (see Figure 1.4). For each neighbour v of u, the directed edge [u, v
intersects a unique face, denoted Fu,v , of Cu .
For x ∈ R3 , x = 0, let (x) be the point of intersection with S of the
straight line segment from 0 to x. Let ju,v be equal in absolute value to the
surface measure of (Fu,v ). The sign of ju,v is taken to be positive if and
only if the scalar product of 12 (u + v) and v − u, viewed as vectors in R3 , is
positive. Let jv,u = − ju,v . We claim that j is a 0/∞ flow on L3 . Parts (a)
and (b) of Definition 1.14 follow by construction, and it remains to check
(c).

The surface of Cu has projection (Cu ) on S. The sum Ju = v∼u ju,v
is the integral over x ∈ (Cu ), with respect to surface measure, of the
number of neighbours v of u (counted with sign) for which x ∈ (Fu,v ).
Almost every x ∈ (Cu ) is counted twice, with signs + and −. Thus the
integral equals 0, whence Ju = 0 for all u = 0.
It is easily seen that J0 = 0, so j is a non-zero flow. Next, we estimate
its energy. By an elementary geometric consideration, there exist ci < ∞

17:53:41
.002
16 Random Walks on Graphs

such that:
(i) | ju,v | ≤ c1 /|u|2 for u = 0, where |u| = δ(0, u) is the length of a
shortest path from 0 to u,
(ii) the number of u ∈ Z3 with |u| = n is smaller than c2 n 2 .
It follows that
 ∞
 c1 2
E( j ) ≤ ju,v ≤
2
6c2 n 2 2 < ∞,
v∼u
n
u =0 n=1

as required. 
Another way of showing Reff < ∞ when d = 3 is to find a finite upper
bound for Reff . Upper bounds can be obtained either by increasing individual
edge-resistances or by removing edges. The idea is to embed a tree with
finite resistance in L3 . Consider a binary tree Tρ in which each connection
between generation n−1 and generation n has resistance ρ n, where ρ > 0. It
is an easy exercise using the series/parallel laws that the effective resistance
between the root and infinity is
∞
 ρ n
Reff (Tρ ) = ,
2
n=1

which we make finite by choosing ρ < 2. We proceed to embed Tρ in Z3

in such a way that a connection between generation n − 1 and generation
n is a lattice-path with length of order ρ n . There are 2n vertices of Tρ in

generation n, and their lattice-distance from 0 is of order nk=1 ρ k , that
is, order ρ n . The surface of the k-ball in R3 is of order k 2 , and thus it is
necessary that
c(ρ n )2 ≥ 2n ,
√
which is to say that ρ > 2.
√
Let 2 < ρ < 2. It is now fairly simple to check that Reff < c Reff (Tρ ).
This method was used in [138] to prove the transience of the infinite open
cluster of percolation on L3 . It is related to, but different from, the tree
embeddings of [83].

1.6 Graph Theory

A graph G = (V , E) comprises a finite or countably infinite vertex-set V
and an associated edge-set E. Each element of E is an unordered pair u, v
of vertices, written u, v. Two edges with the same vertex-pairs are said
to be in parallel, and edges of the form u, u are called loops. The graphs
discussed in this text will generally contain neither parallel edges nor loops,

17:53:41
.002
 1.6 Graph Theory 17

and this is assumed henceforth. Two vertices u, v are said to be joined (or
connected) by an edge if u, v ∈ E. In this case, u and v are the endvertices
of e, and we write u ∼ v and say that u is adjacent to v. An edge e is said
to be incident to its endvertices. The number of edges incident to vertex u
is called the degree of u, denoted deg(u). The negation of the relation ∼ is
written .
Since the edges are unordered pairs, we call such a graph undirected (or
unoriented). If some or all of its edges are ordered pairs, written [u, v, the
graph is called directed (or oriented).
A path of G is defined as an alternating sequence v0 , e0 , v1 , e1 , . . . , en−1 ,
vn of distinct vertices vi and edges ei = vi , vi+1 . Such a path has length
n; it is said to connect v0 to vn , and is called a v0 /vn path. A cycle or circuit
of G is an alternating sequence v0 , e0 , v1 , . . . , en−1 , vn , en , v0 of vertices
and edges such that v0 , e0 , . . . , en−1 , vn is a path and en = vn , v0 . Such
a cycle has length n + 1. The (graph-theoretic) distance δ(u, v) from u to
v is defined to be the number of edges in a shortest path of G from u to v.
We write u  v if there exists a path connecting u and v. The relation 
is an equivalence relation, and its equivalence classes are called components
(or clusters) of G. The components of G may be considered either as sets of
vertices or as graphs. The graph G is connected if it has a unique component.
It is a forest if it contains no cycle, and a tree if in addition it is connected.
A subgraph of the graph G = (V , E) is a graph H = (W, F) with
W ⊆ V and F ⊆ E. The subgraph H is a spanning tree of G if V = W
and H is a tree. A subset U ⊆ V of the vertex-set of G has boundary
∂U = {u ∈ U : u ∼ v for some v ∈ V \ U }.
Lattice-graphs are the most important type of graph for applications in
areas such as statistical mechanics. Lattices are sometimes termed ‘crys-
talline’ since they are periodic structures of crystal-like units. A general
definition of a lattice may confuse readers more than help them, and instead
we describe some principal examples.
Let d be a positive integer. We write Z = {. . . , −1, 0, 1, . . . } for the
set of all integers, and Zd for the set of all d-vectors v = (v1 , v2 , . . . , vd )
with integral coordinates. For v ∈ Zd , we generally write vi for the i th
coordinate of v, and we define

d
δ(u, v) = |u i − vi |.
i=1

The origin of Zdis denoted by 0. We turn Zd into a graph, called the d-

dimensional (hyper)cubic lattice, by adding edges between all pairs u, v of
points of Zd with δ(u, v) = 1. This graph is denoted as Ld , and its edge-set

17:53:41
.002
18 Random Walks on Graphs

Figure 1.5 The square, triangular, and hexagonal (or ‘honeycomb’)

lattices. The solid and dashed lines illustrate the concept of ‘planar
duality’ discussed after (3.7).

as Ed : thus, Ld = (Zd , Ed ). We often think of Ld as a graph embedded

in Rd , the edges being straight line-segments between their endvertices.
The edge-set EV of V ⊆ Zd is the set of all edges of Ld both of whose
endvertices lie in V .
The two-dimensional cubic lattice L2 is called the square lattice and is
illustrated in Figure 1.5. Two other lattices in two dimensions that feature
in this text are drawn there also.

1.7 Exercises
1.1 Let G = (V, E) be a finite connected graph with unit edge-weights. Show
that the effective resistance between two distinct vertices s, t of the associated
electrical network may be expressed as B/N , where B is the number of s/t bushes
of G, and N is the number of its spanning trees. (See the proof of Theorem 1.16
for an explanation of the term ‘bush’.)
Extend this result to general edge-weights we > 0.
1.2 Let G = (V, E) be a finite connected graph with strictly positive edge-
weights (we : e ∈ E), and let N ∗ be given by (1.18). Show that
1  ∗ 
i a,b = N (s, a, b, t) − N ∗ (s, b, a, t)
N∗
constitutes a unit flow through G from s to t satisfying Kirchhoff’s laws.
1.3 (continuation) Let G = (V, E) be finite and connected with given conduc-
tances (we : e ∈ E), and let (xv : v ∈ V ) be reals satisfying v xv = 0. To G
we append a notional vertex labelled ∞, and we join ∞ to each v ∈ V . Show that
there exists a solution i to Kirchhoff’s laws on the expanded graph, viewed as two
laws concerning current flow, such that the current along the edge v, ∞ is xv .

17:53:41
.002
Exploring the Variety of Random
Documents with Different Content
back
back
back
back
back
back
back
back
back
back
back
back
back
back
back
back
Welcome to our website – the ideal destination for book lovers and
knowledge seekers. With a mission to inspire endlessly, we offer a
vast collection of books, ranging from classic literary works to
specialized publications, self-development books, and children's
literature. Each book is a new journey of discovery, expanding
knowledge and enriching the soul of the reade

Our website is not just a platform for buying books, but a bridge
connecting readers to the timeless values of culture and wisdom. With
an elegant, user-friendly interface and an intelligent search system,
we are committed to providing a quick and convenient shopping
experience. Additionally, our special promotions and home delivery
services ensure that you save time and fully enjoy the joy of reading.

Let us accompany you on the journey of exploring knowledge and

personal growth!

textbookfull.com