Discrete Mathematics In Statistical Physics

Introductory Lectures Martin Loebl download

https://ebookbell.com/product/discrete-mathematics-in-
statistical-physics-introductory-lectures-martin-loebl-4110230

Explore and download more ebooks at ebookbell.com

Here are some recommended products that we believe you will be
interested in. You can click the link to download.

Discrete Mathematics In Statistical Physics Martin Loebl

https://ebookbell.com/product/discrete-mathematics-in-statistical-
physics-martin-loebl-1436522

Difference Equations Discrete Dynamical Systems And Applications Idcea

2022 Gifsuryvette France June 1822 Springer Proceedings In Mathematics
Statistics 444 1st Edition Llus Alsed I Soler

https://ebookbell.com/product/difference-equations-discrete-dynamical-
systems-and-applications-idcea-2022-gifsuryvette-france-
june-1822-springer-proceedings-in-mathematics-statistics-444-1st-
edition-llus-alsed-i-soler-55950574

Zerosum Discretetime Markov Games With Unknown Disturbance

Distribution Discounted And Average Criteria Springerbriefs In
Probability And Mathematical Statistics 1st Edition J Adolfo
Minjrezsosa
https://ebookbell.com/product/zerosum-discretetime-markov-games-with-
unknown-disturbance-distribution-discounted-and-average-criteria-
springerbriefs-in-probability-and-mathematical-statistics-1st-edition-
j-adolfo-minjrezsosa-11045630

Probabilistic Methods In Discrete Mathematics Reprint 2020 V F Kolchin

Editor V Ya Kozlov Editor V V Mazalov Editor Yu L Pavlov Editor Yu V
Prokhorov Editor

https://ebookbell.com/product/probabilistic-methods-in-discrete-
mathematics-reprint-2020-v-f-kolchin-editor-v-ya-kozlov-editor-v-v-
mazalov-editor-yu-l-pavlov-editor-yu-v-prokhorov-editor-51040670
Inevitable Randomness In Discrete Mathematics Jozsef Beck

https://ebookbell.com/product/inevitable-randomness-in-discrete-
mathematics-jozsef-beck-5252212

Prolog Experiments In Discrete Mathematics Logic And Computability

James L Hein

https://ebookbell.com/product/prolog-experiments-in-discrete-
mathematics-logic-and-computability-james-l-hein-5733276

A First Course In Discrete Mathematics Ian Anderson

https://ebookbell.com/product/a-first-course-in-discrete-mathematics-
ian-anderson-4199374

Topics In Discrete Mathematics Dedicated To Jarik Neetil On The

Occasion Of His 60th Birthday 1st Edition Michael E Adams

https://ebookbell.com/product/topics-in-discrete-mathematics-
dedicated-to-jarik-neetil-on-the-occasion-of-his-60th-birthday-1st-
edition-michael-e-adams-879850

Connections In Discrete Mathematics A Celebration Of The Work Of Ron

Graham 1st Edition Steve Butler

https://ebookbell.com/product/connections-in-discrete-mathematics-a-
celebration-of-the-work-of-ron-graham-1st-edition-steve-butler-9970994
Martin Loebl

Discrete Mathematics in Statistical Physics

Martin Loebl

Discrete Mathematics
in Statistical Physics
Introductory Lectures
Bibliographic information published by the Deutsche Nationalbibliothek
The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie;
detailed bibliographic data are available in the Internet at http://dnb.d-nb.de.

Dr. Martin Loebl

Charles University
Department of Applied Mathematics
Institut of Theoretical Computer Science
Malostranske 25
CZ-118 00 Praha
Tschechische Republik
E-Mail: [email protected]

1st Edition 2010

All rights reserved

© Vieweg +Teubner | GWV Fachverlage GmbH, Wiesbaden 2010
Editorial Office: Ulrike Schmickler-Hirzebruch | Nastassja Vanselow
Vieweg+Teubner is part of the specialist publishing group Springer Science+Business Media.
www.viewegteubner.de

No part of this publication may be reproduced, stored in a retrieval system or

transmitted, in any form or by any means, electronic, mechanical, photocopying,
recording, or otherwise, without the prior written permission of the copyright holder.

Registered and/or industrial names, trade names, trade descriptions etc. cited in this publication
are part of the law for trade-mark protection and may not be used free in any form or by any means
even if this is not specifically marked.

Cover design: KünkelLopka Medienentwicklung, Heidelberg

Printing company: STRAUSS GMBH, Mörlenbach
Printed on acid-free paper
Printed in Germany

ISBN 978-3-528-03219-7
dedicated to Zbyněk and Jaroslava, my parents
Preface
The purpose of these lecture notes is to briefly describe some of the basic con-
cepts interlacing discrete mathematics, statistical physics and knot theory. I
tried to emphasize a ’combinatorial common sense’ as the main method. No
attempt of completeness was made. The book should be accessible to the stu-
dents of both mathematics and physics. I profited from previous books and
expositions on discrete mathematics, statistical physics, knot theory and others,
namely [B1], [BRJ], [BB], [J1], [KG], [LL], [MN], [MJ], [MT], [S0], [S3], [SM],
[WFY], [WD], [KSV]. Most of the material contained in the book is introduc-
tory and appears without a reference to the original source. This book has been
an idea of my editor Martin Aigner. I would like to thank to him for his support
and help. Many other colleagues helped me with the book. Mihyun Kang, Jirka
Matoušek, Iain Moffatt, Jarik Nešetřil, Dominic Welsh and Christian Kratten-
thaler read earlier versions, and without their extensive comments the book
would probably not exist. I had enlightening discussions on several topics dis-
cussed in the book, in particular with Martin Klazar, Roman Kotecký, Ondřej
Pangrác, Gregor Masbaum, Xavier Viennot and Uli Wagner. Marcos Kiwi saved
the whole project by gently teaching me how to draw pictures amd Winfried
Hochstaettler drew one; I am sure you will be able to detect it. Large part of
the book was written during my visit, in the whole year 2006, at the School of
Mathematics and the Centro Modelamiento Matematico, Universidad de Chile.
I want to thank my colleagues there for wonderful hospitality, and gratefully
acknowledge the support of CONICYT via project Anillo en Redes, ACT-08.
But of course, the seminal ingredient in the process of making the book was the
creative environment of my home department of applied mathematics and the
institute of theoretical computer science at the Charles University, Prague.
Some theorems and observations in the book appear without a proof. Usually
a pointer is given to a book (preferentially) or to a paper where a proof can be
found. If no pointer is given, then I believe (possibly mistakenly) that it should
be possible to prove the statement in an elementary and not very complicated
way. The reader is encouraged to write down such proofs as exercises. The first
five chapters concentrate on the introductory discrete mathematics. Chapters
six and seven are devoted to the partition functions, and chapter eight is an in-
troduction to the theory of knots. The last chapter describes two combinatorial
technics which solve the 2D Ising and dimer problems.

Prague, September 2009

Martin Loebl
Contents
Preface VII

1 Basic concepts 1
1.1 Sets, functions, structures . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Algorithms and Complexity . . . . . . . . . . . . . . . . . . . . . 4
1.3 Generating functions . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.4 Principle of inclusion and exclusion . . . . . . . . . . . . . . . . . 7

2 Introduction to Graph Theory 13

2.1 Basic notions of graph theory . . . . . . . . . . . . . . . . . . . . 13
2.2 Cycles and Euler’s theorem . . . . . . . . . . . . . . . . . . . . . 18
2.3 Cycle space and cut space . . . . . . . . . . . . . . . . . . . . . . 20
2.4 Flows in directed graphs . . . . . . . . . . . . . . . . . . . . . . . 25
2.5 Connectivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
2.6 Factors, matchings, and dimers . . . . . . . . . . . . . . . . . . . 29
2.7 Graph colorings . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
2.8 Random graphs and Ramsey theory . . . . . . . . . . . . . . . . 38
2.9 Regularity lemma . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
2.10 Planar graphs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
2.11 Tree-width and excluded minors . . . . . . . . . . . . . . . . . . 47

3 Trees and electrical networks 51

3.1 Minimum spanning tree and greedy algorithm . . . . . . . . . . . 51
3.2 Tree isomorphism . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
3.3 Tree enumeration . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
3.4 Electrical networks . . . . . . . . . . . . . . . . . . . . . . . . . . 57
3.5 Random walks . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62

4 Matroids 65
4.1 Examples of matroids . . . . . . . . . . . . . . . . . . . . . . . . 67
4.2 Greedy algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
4.3 Circuits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
4.4 Basic operations . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
4.5 Duality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
4.6 Representable matroids . . . . . . . . . . . . . . . . . . . . . . . 73
4.7 Matroid intersection . . . . . . . . . . . . . . . . . . . . . . . . . 74
4.8 Matroid union and min-max theorems . . . . . . . . . . . . . . . 74
X CONTENTS

5 Geometric representations of graphs 77

5.1 Topological spaces . . . . . . . . . . . . . . . . . . . . . . . . . . 77
5.2 Planar curves: Gauß codes . . . . . . . . . . . . . . . . . . . . . . 82
5.3 Planar curves: rotation . . . . . . . . . . . . . . . . . . . . . . . . 87
5.4 Convex embeddings . . . . . . . . . . . . . . . . . . . . . . . . . 88
5.5 Coin representations . . . . . . . . . . . . . . . . . . . . . . . . . 91
5.6 Counting fatgraphs: matrix integrals . . . . . . . . . . . . . . . . 93

6 Game of dualities 101

6.1 Edwards-Anderson Ising model . . . . . . . . . . . . . . . . . . . 101
6.2 Max-Cut for planar graphs . . . . . . . . . . . . . . . . . . . . . 103
6.3 Van der Waerden’s theorem . . . . . . . . . . . . . . . . . . . . . 105
6.4 MacWilliams’ theorem . . . . . . . . . . . . . . . . . . . . . . . . 106
6.5 Phase transition of 2D Ising . . . . . . . . . . . . . . . . . . . . . 108
6.6 Critical temperature of the honeycomb lattice . . . . . . . . . . . 110
6.7 Transfer matrix method . . . . . . . . . . . . . . . . . . . . . . . 113
6.8 The Yang-Baxter equation . . . . . . . . . . . . . . . . . . . . . . 116

7 The zeta function and graph polynomials 119

7.1 The Zeta function of a graph . . . . . . . . . . . . . . . . . . . . 119
7.2 Chromatic, Tutte and flow polynomials . . . . . . . . . . . . . . 124
7.3 Potts, dichromate and ice . . . . . . . . . . . . . . . . . . . . . . 128
7.4 Graph polynomials for embedded graphs . . . . . . . . . . . . . . 131
7.5 Some generalizations . . . . . . . . . . . . . . . . . . . . . . . . . 135
7.6 Tutte polynomial of a matroid . . . . . . . . . . . . . . . . . . . 138

8 Knots 141
8.1 Reidemeister moves . . . . . . . . . . . . . . . . . . . . . . . . . . 142
8.2 Skein relation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
8.3 The knot complement . . . . . . . . . . . . . . . . . . . . . . . . 144
8.4 The Alexander-Conway polynomial . . . . . . . . . . . . . . . . . 146
8.5 Braids and the braid group . . . . . . . . . . . . . . . . . . . . . 148
8.6 Knot invariants and vertex models . . . . . . . . . . . . . . . . . 149
8.7 Alexander-Conway as a vertex model . . . . . . . . . . . . . . . . 150
8.8 The Kauffman derivation of the Jones polynomial . . . . . . . . . 150
8.9 Jones polynomial as vertex model . . . . . . . . . . . . . . . . . . 153
8.10 Vassiliev invariants and weight systems . . . . . . . . . . . . . . . 153

9 2D Ising and dimer models 157

9.1 Pfaffians, dimers, permanents . . . . . . . . . . . . . . . . . . . . 157
9.2 Products over aperiodic closed walks . . . . . . . . . . . . . . . . 162

Bibliography 173

List of Figures 181

Index 183
Chapter 1

Basic concepts

In this introductory chapter we first present some very basic mathematical for-
malism. Then we introduce algorithms and complexity. The chapter ends with
basic tools of discrete calculations.

1.1 Sets, functions, structures

We will use symbols N,Z,P,Q,R, and C to denote the sets of the natural,
integer, positive integer, rational, real, and complex numbers. If not specified
otherwise then i,j,k,m ,n are non-negative integers. We sometimes denote by
[n] the set {1,...,n}. We denote by |X | the cardinality of a set X . A function
f from a set X to a set Y is called one-to-one or injective if for x,y ∈ X x = y
implies f(x) = f(y), it is surjective or onto if for each y ∈ Y there is x ∈ X
such that f(x) = y, and it is a bijection if it is both one-to-one and onto. If
X ,Y are finite then a bijection is also called a permutation. Let |X | = n and
|Y | = m . There are n! = n(n − 1)(n − 2) · · · 1 permutations of X ; n! is the
factorial function. The number of all functions from X to Y is m n and the
number of one-to-one functions from X to Y is m (m − 1) · · · (m − n + 1). The
number of surjective functions does not have such a nice formula, it may be
written with the help of the principle of inclusion and exclusion as

m  

i m
(−1) (m − i)n . (1.1)
i=0
i

The Kronecker delta function is defined by δ (x,y) = 1 if x = y, and zero

X
otherwise. If X is a set, then we denote
X  by 2 the set of all the subsets of X ;
X n
|2 | = 2 . We further denote by k the set of all subsets of X of cardinality
k. We have
   
 X  n n!
 k  = k = k!(n − k)! .
 
2 CHAPTER 1. BASIC CONCEPTS

The binomial theorem says that

n  
 n k n−k
(x + y)n = x y .
k
k=0
n
The symbol k is called the binomial coefficient. The multinomial coefficient
is defined by  
n n!
=
k1 ,· · · ,km k1 ! · · · km !
and the multinomial theorem says that
 
 n
(x1 + · · · + xm )n = xk1 · · · xkmm .
k1 ,· · · ,km 1
k1 +···+km =n

A very good estimate of the factorial function n! is given by Stirling’s formula

which approximates n! by (2π n)1/2 ( ne )n .
If Y ⊂ X , then the incidence vector of Y will be denoted by i(Y ); i(Y ) is the
0,1 vector indexed by the elements of X , where [i(Y )]z = 1 if and only if z ∈ Y .
We will sometimes not distinguish between a set and its incidence vector.
An ordered pair is usually denoted by (x,y) where x is the first element of the
pair. A binary relation on X is any subset of X × X = {(x,x′ ); x ∈ X ,x′ ∈ X }.
Any function X → X is a binary relation on X . A partially ordered set, or
poset for short, is a pair (X ,), where X is a set and  is a binary relation
on X that is reflexive (x  x), transitive (x  y and y  z imply x  z), and
(weakly) antisymmetric (x  y and y  x imply x = y). The binary relation 
is itself called a partial ordering. A partial ordering where any pair of elements
is comparable is called a linear ordering. An important example of a linear
ordering is the lexicographic ordering. Let a = (a1 ,...,an ) and b = (b1 ,...,bm )
be two strings of integers. We say that a is lexicographically smaller than b if a
is an initial segment of b or aj < bj for the smallest index j such that aj = bj .
Let (X ,) be a poset and Y ⊂ X . We say that Y is a chain if the induced
ordering (Y,) is linear.
The symbol F will denote a field; F will usually be equal to R or C, or to the
finite 2-element field G F (2) = ({0,1},+,×) with addition and multiplication
modulo 2. The symbol Fd denotes the vector space of dimension d over F. The
elements of Fd are called vectors; for x ∈ Fd we write x = (x1 ,· · · ,xd ). We
will understand vectors as both row and column vectors. The scalar product
of two vectors x,y is xy = x1 y1 + · · · + xd yd . A set {x1 ,· · · ,xk } of vectors of
k
Fd is linearly independent if, whenever i=1 ai xi = 0 and each ai ∈ F, then
a1 = a2 = · · · = ak = 0. The dimension dim({x1 ,· · · ,xk }) of a set of vec-
tors {x1 ,· · · ,xk } is the maximum number of linearly independent elements in
{x1 ,· · · ,xk }. A subspace of a vector space Fd is any subset of Fd which is closed
under addition, and multiplication by an element of F. Two subspaces X ,Y are
isomorphic if there is a bijection f : X → Y such that for each a,b ∈ X and
c ∈ F, f(a + b) = f(a) + f(b) and f(ca) = cf(a). The orthogonal complement
1.1. SETS, FUNCTIONS, STRUCTURES 3

of a subspace X ⊂ Fd is the subspace {y ∈ Fd ; xy = 0 for each x ∈ X }.

Let A = (aij ) be a matrix of n rows and m columns, with entries from field
F. We say that A is an n × m matrix. If n = m then A is a square ma-
trix. The determinant of a square n × n matrix A is defined by det(A ) =
 sign(π) n
π (−1) i=1 aπ(i)i , where the sum is over all permutations π of 1,· · · ,n
and sign(π ) = |{i < j; π (i) > π (j)}|. The determinant characterizes linearly
independent vectors. A set of n vectors of length n is linearly independent if
and only if det(A ) = 0, where A is the matrix whose set of columns (or rows)
is formed by the vectors. The determinant of a matrix may be calculated ef-
ficiently by the Gaussian elimination. The permanent of matrix A is defined
analogously as the determinant,  but the (−1)sign(π) term is omited from each
n
summand. Hence, Per(A ) = π i=1 aπ(i)i . There is no efficient algorithm to
calculate the permanent. The identity matrix is the square matrix A such that
aii = 1 and aij = 0 for i= j. The trace of a square matrix A , denoted by tr(A ),
is defined by tr(A ) = i aii .
The symbol Rd also denotes the Euclidean space of dimension d. A curve in Rd is
the image of a continuous function f : [0,1] → Rd . A curve is simple if it is one-
to-one, and it connects its endpoints f(0),f(1). A curve is closed if f(0) = f(1).
The Euclidean norm of x ∈ Rd is ||x|| = (xx)1/2 . A set {x0 ,x1 ,· · · ,xk } of
k k
vectors of Rd is affinely independent if, whenever i=0 ai xi = 0, i=0 ai = 0
and each ai ∈ R, then a0 = a1 = · · · = ak = 0. For two points x0 ,x1 affine inde-
pendence means x0 = x1 ; for three points it means that x0 ,x1 ,x2 do not lie on
a common line; for four points it means that they do not lie on a common plane;
and so on. The rank of a set of points of Rd , denoted by rank({x0 ,· · · ,xk }), is
the maximum number of affinely independent elements in {x0 ,· · · ,xk }.
There is a simple relation between linear and affine independence: x0 ,· · · ,xk are
affinely independent if and only if x1 − x0 ,· · · ,xk − x0 are linearly independent.
This happens if and only if the (d+1)-dimensional vectors (1,x0 ),...,(1,xk ) are
linearly independent. An affine subspace is any subset A ⊂ Rd which contains,
for each pair of its elements x,y, the line through x,y. A hyperplane in Rd is
a (d − 1)-dimensional affine subspace, i.e., a set of the form {x ∈ Rd : ax = b}
for some nonzero a ∈ Rd and b ∈ R. A (closed) half-space has the form
{x ∈ Rd : ax ≤ b} for some nonzero a ∈ Rd and b ∈ R.
A set C ⊂ Rd is convex if for every x,y ∈ C , the segment {ax+(1−a)y : 0 ≤ a ≤
1} between x and y is contained in C . The convex hull of a set X ⊂ Rd is the
intersection of all convex sets containing X , and it is denoted by conv(X ). Each
x ∈ conv(X ) may be written as a convex combination of elements of X : there
k
are x1 ,· · · ,xk ∈ X and real numbers a1 ,· · · ,ak ≥ 0 such that i=1 ai = 1 and
k
x = i=1 ai xi .
A convex polytope is the convex hull of a finite subset of Rd . Each convex
polytope can be expressed as the intersection of finitely many half-spaces. Con-
versely, by the Minkowski-Weyl theorem, if an intersection of finitely many
half-spaces is bounded, then it is a convex polytope. A face of a convex poly-
tope P is P itself or a non-empty intersection of P with a hyperplane that does
not dissect P (i.e., not both of the open half-spaces defined by the hyperplane
4 CHAPTER 1. BASIC CONCEPTS

intersect P in a non-empty set).

1.2 Algorithms and Complexity

Algorithmic considerations are important for many concepts of both discrete
mathematics and statistical physics. We make only basic algorithmic remarks
in this book, and therefore the following exposition on algorithms and complex-
ity is very brief.
Informally, an algorithm is a set of instructions to be carried out mechanically.
Applying an algorithm to its input we get some output, provided that the se-
quence of the instructions prescribed by applying the algorithm terminates. The
application of an algorithm is often called a computation. Usually inputs and
outputs are strings (words, finite sequences) from a finite alphabet; a basic ex-
ample are binary words, i.e., finite sequences of 0,1. The notion of an algorithm
is usually formalized by the definition of a Turing machine.
A Turing machine consists of the following components:

• a finite set S called the alphabet,

• an element b ∈ S called the blank symbol,

• a subset A ⊂ S called the external alphabet; we assume b ∈

/ A,

• a finite set Q whose elements are called states of the Turing machine,

• an initial state s ∈ Q ,

• a transition function, i.e., a function

t: Q × S → Q × S × {−1,0,1}.

A Turing machine has a tape that is divided into cells. Each cell carries one
symbol from S. We assume that the tape is infinite, thus the content of the
tape is an infinite sequence s = s0 ,s1 ,· · · of elements of S.
A Turing machine also has a read-write head that moves along the tape and
changes symbols. If the head is in position p along the tape, it can read symbol
sp and write another symbol in its place.
The behaviour of a Turing machine is determined by a control device. At each
step of the computation, this device is in some state q ∈ Q . The state q and
the symbol sp under the head determine the action performed by the Turing
machine: the value of the transition function, t(q,sp ) = (q′ ,s′ ,p′ ), contains the
new state q′ , the new symbol s′ to be written in the place of sp , and the shift
p′ ∈ {−1,0,1} of the position of the head. If the head bumps into the left
boundary of the tape (that happens when p + p′ < 0), then the computation
stops.
Next we describe the input given to the Turing machine, and how the output is
obtained. Let A ∗ denote the set of all the strings (finite sequences) of elements
1.2. ALGORITHMS AND COMPLEXITY 5

of A . Inputs and outputs to the Turing machine with the external alphabet A
are strings from A ∗ . An input string I is written on the tape and followed by
the blank symbol b. Initially, the head is at the beginning (left end) of I. If
the Turing machine stops (by bumping into the left boundary of the tape), we
read the tape from left to right starting from the left end until we reach some
symbol that does not belong to A . The initial segment of the tape until that
symbol will be the output of the Turing machine.
Every Turing machine computes a function from a subset of A ∗ to A ∗ . There are
functions that are not computable. A Turing machine is obviously an algorithm
in the informal sense. The converse assertion is called the
Church-Turing thesis: Any algorithm can be realised by a Turing machine.
Note that the Church-Turing thesis is not a mathematical theorem, but rather
a statement about our understanding of the informal notion of algorithm.
Complexity classes. The computability of a function does not guarantee
that we can compute it in practice since an algorithm may require too much
time. The idea of an effective algorithm is usually formalized by the notion
of polynomial algorithms. We say that a function T on the positive integers
is of polynomial growth if T (n) ≤ cnd for all n and some constants c,d. We
say that a function f defined on the binary strings of {0,1}∗ is computable in
polynomial time if there exists a Turing machine that computes f in time T (n)
of polynomial growth, where n is the length of the input. Such a Turing machine
is called a polynomial algorithm. Polynomial time encoding plays a crucial role.
For instance, if the input is an integer N in the unary representation then the
input size is |N | but if the representation is binary, the input size is only log(|N |).
The class of all functions computable in polynomial time is denoted by P . We
should remark here that computability in polynomial time does not guarantee
practical computability either, but it is a good indication for it.
A special class of algorithmic problems are the decision problems. In a decision
problem, we want the answer to be yes or no. This clearly may be modeled as a
function from a subset of A ∗ to {0,1} where 0 encodes no and 1 encodes yes. It
is customary to call such functions predicates. One can think about predicates
as about properties: the predicate indicates for each string whether it has the
property (yes) or does not have the property (not). Hence the algorithmic
problem to compute a predicate may be formulated as the algorithmic problem
to test the corresponding property.
Another basic complexity class, the class N P , is usually defined only for the
predicates. We say that a predicate R (x,y), where x and y are binary strings,
is polynomially decidable if there is a Turing machine that computes it in time
of polynomial growth (the size of the input is |x| + |y|).
The class N P is the class of all predicates f for which there is a polynomial
growth function T (n) and a polynomially decidable predicate R of two variables
so that f(x) = 1 if and only if there is y such that |y| < T (|x|) and R (x,y) = 1.
Informally, N P is the class of the predicates (i.e., properties), for which there
is a certificate (coded by y) that can be checked in polynomial time. Most of
the properties discussed in this book belong to N P .
Clearly P ⊂ N P . Over the past 30 years intensive research has been directed
6 CHAPTER 1. BASIC CONCEPTS

towards proving that the inclusion is strict. The question whether P = N P

is today one of the fundamental problems of both mathematics and computer
science.
Reducibility. When can we say that one problem is algorithmically at least
as hard as another problem? We model the efficiency by the polynomial time
complexity, and so the answer is naturally given by the following notion of poly-
nomial reducibility: we say that a predicate f1 is polynomially reducible to a
predicate f2 if there exists a function g ∈ P so that f1 (x) = f2 (g(x)), for each
input string x.
A predicate f ∈ N P is called N P -complete if any predicate in N P is polynomi-
ally reducible to it. The predicates that are N P -complete are the most difficult
predicates of N P : if some N P -complete predicate is in P then P = N P . It
is customary to speak about N P -complete problems rather than N P -complete
predicates. The existence of an efficient algorithm to solve an N P -complete
problem is considered to be very unlikely.
A seminal result in algorithmic complexity is that N P -complete predicates
(problems) exist. This was proved independently by Cook and Levin. Many
natural N P -complete problems are known, see [GJ].

1.3 Generating functions

A useful way of counting is provided by generating functions. If f is a func-
tion fromthe non-negative integers, we can consider its (ordinary) generating
function n≥0 f(n)xn and its exponential generating function n≥0 f(n)xn /n!.

The generating functions are formal power series, since we are not concerned
with letting x take particular values, and we ignore questions of convergence.
This formalism is convenient since we can perform various operations on the
formal power series, for instance
   
  
 an xn  +  bn xn  = (an + bn )xn ,
n≥0 n≥0 n≥0
   
  
 an xn /n! +  bn xn /n! = (an + bn )xn /n!
n≥0 n≥0 n≥0

and   
  
 an xn   bn xn  = cn xn
n≥0 n≥0 n≥0
  
  
 an xn /n!  bn xn /n! = dn xn /n!
n≥0 n≥0 n≥0
n n n
where cn = i=0 ai bn−i and dn = i=0 i ai bn−i .
These operations coincide with the addition and multiplication of functions
1.4. PRINCIPLE OF INCLUSION AND EXCLUSION 7

when the power series converge for some values of x. Let us denote by C[[x]] the
set of all formal power series n≥0 an xn with complex coefficients. Addition and
multiplication in C[[x]] are clearly commutative, associative and distributive,
thus C[[x]] forms a commutative ring where 1 is the unity. Formal power series
with the coefficients in a non-commutative ring (like the square matrices of the
same size) are also extensively considered; they form a non-commutative ring
with unity.
If F (x) and G (x) are elements of C[[x]] satisfying F (x)G (x) = 1 then we
write G (x) = F (x)−1 . It is easy to see that F (x)−1 exists if and only if
a0 = F (0) = 0. If F (x)−1 exists then it is uniquely determined. We have
((F (x)−1 )−1 = F (x).
Example 1.3.1. Let a = 0 and ( n≥0 an xn )(1 − ax) = n≥0 cn xn , where a is
 
a non-zero complex number. Then from the definition of multiplication we get
c0 = 1 and cn = 0 for n > 0. Hence we may write

an xn = (1 − ax)−1 .
n≥0

The identity may be derived in the same way in every ring of formal power series
over a (not necessarily commutative) ring with unity. Hence, for instance, for
square complex matrices it can be written as

A n xn = (I − A x)−1 .
n≥0

This is of course just the formula for summing a geometric series. Informally
speaking, if we have an identity involving power series that is valid when the
power series are regarded as functions (when the variables are sufficiently small
complex numbers), then the identity remains valid when regarded as an identity
among formal power series. Formal power series may naturally have more than
one variable.

1.4 Principle of inclusion and exclusion

Let us start with the introduction of a paper of Whitney, which appeared in
Annals of Mathematics in August 1932:
”Suppose we have a finite set of objects (for instance books on a table), each
of which either has or has not a certain given property A (say of being red). Let n
be the total number of objects, n(A ) the number with the property A , and n(A¯)
the number without the property A . Then obviously n(A¯) = n−n(A ). Similarly,
if n(A B ) denotes the number with both properties A and B , and n(A¯B̄ ) the
number with neither property, then n(A¯B¯ ) = n − n(A ) − n(B ) + n(A B ), which
is easily seen to be true. The extension of these formulas to the general case
where any number of properties is considered is quite simple, and is well known
to logicians. It should be better known to mathematicians also; we give in this
paper several applications which show its usefulness.”
It is known today, under the name principle of inclusion and exclusion (PIE).
8 CHAPTER 1. BASIC CONCEPTS


Theorem 1.4.1. Suppose A 1 ,...,A n are finite sets, and A J = i∈J A i . Then
 
 n   n 
 A i = (−1)k−1 |A J |.
 
 
i=1 k=1 [n]
J∈( k )

Proof. We proceed by induction on n. The case n = 2 is clear. In the induction

step,    
 n  n−1 
 
 A i =  A i ∪ A n =
 
   
i=1 i=1
 
n−1
 n−1 
k−1 
|A n | + (−1) ΣJ∈([n− 1]) |A J | −  (A i ∩ A n ) =

k  
k=1 i=1
n

(−1)k−1 ΣJ∈([n]) |A J |.
k
k=1

Let (X ,≤) be a finite partially ordered set (poset). For example, the set of
all subsets of a finite set S equipped with the relation ’⊂’ forms a poset called
the Boolean lattice. Let F be a field and let us denote by F(X ) the collection of
all functions f : X × X → F, with the property that f(x,y) = 0 only if x ≤ y.
We equip the set F (X ) with the convolution product

(f ∗ g)(x,y) = f(x,z)g(z,y).
x≤z≤y

It is straightforward to verify that the convolution product is associative.

We recall from the introduction that the Kronecker delta function is defined by
δ (x,y) = 1 if and only if x = y, and it is zero otherwise. The Kronecker delta
function acts as the identity function with respect to the convolution product,
since δ ∗ f = f ∗ δ = f. Another basic function is the zeta function defined by
ζ (x,y) = 1 if x ≤ y, and ζ (x,y) = 0 otherwise.
Theorem 1.4.2. Let f ∈ F(X ) be a function such that f(x,x) = 0 for each
x ∈ X . Then there is unique function g so that f ∗ g = g ∗ f = δ .
Proof. We can define g inductively by first letting
g(x,x) = 1/f(x,x),
and then letting
 f(z,y)
g(x,y) = − g(x,z) .
f(y,y)
x≤z<y

Hence g ∗ f = δ and so g is the left inverse of f. Similarly we can define

function h which is the right inverse of f. By associativity, g = g∗δ = g∗(f∗h) =
(g ∗ f) ∗ h = δ ∗ h = h.
1.4. PRINCIPLE OF INCLUSION AND EXCLUSION 9

Another basic function is the inverse of the zeta function; it is called the
Möbius function and denoted by µ(x,y). We immediately have

Exercise 1.4.3. µ(x,x) = 1 and µ(x,y) = − x≤z<y µ(x,z).
Exercise 1.4.4. Prove by induction on |B − A | that the Möbius function of the
Boolean poset (2n ,⊂) is given by µ(A ,B ) = (−1)|B−A| .
Exercise 1.4.5. Let (X ,≤) be an interval in the set of integers. Then µ(x,x) = 1,
µ(x,x + 1) = −1 and µ(x,y) = 0 otherwise.
Let us state the Möbius inversion formula (MIF).

Theorem 1.4.6. Let (X ,≤)be a finite poset and let f,g be two functions
from
 X to F. Then g(x) = y≤x f(y) for all x ∈ X if and only if f(x) =
y≤x g(y)µ(y,x) for all x ∈ X .

Proof.  
g(y)µ(y,x) = f(z)µ(y,x) =
y≤x y≤x z≤y
 
µ(y,x) ζ (z,y)f(z) =
y≤x z∈X
 
 
 ζ (z,y)µ(y,x) f(z) =
z∈X z≤y≤x

δ (z,x)f(z) = f(x).
z∈X

If (X ,≤1 ) and (Y,≤2 ) are two posets then their direct product is the poset
(Z ,) where Z = X × Y = {(x,y); x ∈ X ,y ∈ Y } and (x,y)  (x′ ,y′ ) if x ≤1 x′
and y ≤2 y′ .

Theorem 1.4.7. The Möbius function of the direct product of two posets is the
product of their Möbius functions: µ((x1 ,y1 )(x2 ,y2 )) = µ1 (x1 ,x2 )µ2 (y1 ,y2 ).

Proof. This can be proven in a straightforward way by induction on the number

of pairs (u,v) that lie between (x,y) and (x′ ,y′ ), and using Exercise 1.4.3.

The following theorem is an easy consequence of Exercise 1.4.3.

Theorem 1.4.8. Let P be a finite poset and let P ′ be P with adjoined smallest
and largest elements (denoted by 0 and 1). Let ci be the number of chains in P ′
between 0,1 of length i. Then µ(0,1) = c0 − c1 + · · · .

This means that µ(0,1) may be interpreted as the Euler characteristic of the
abstract simplicial complex associated with P ′ (see Section 5.1).
Applying Theorem 1.4.6 to the Boolean lattice, and using Exercise 1.4.4, we get
10 CHAPTER 1. BASIC CONCEPTS

Corollary 1.4.9. Let F be a function on the set2S of all subsets of an n-

element set S to a field F. For K ⊂ S let G (K ) = L⊂K F (L). Then

F (K ) = (−1)|K−L|G (L).
L⊂K

Corollary 1.4.9 implies a formula which gives a simple exponential algorithm

to calculate a permanent.
Corollary 1.4.10. Let A be an n × n matrix. Then
 
 n
 
PerA = (−1)n−|S|  aij  .
S⊂{1,··· ,n} i=1 j∈S

Next we consider the poset (X ,) where X = {1,2,· · · ,n} and a  b if a

divides b.
Theorem 1.4.11. Let µ be the Möbius function of (X ,). Then µ(1,1) = 1,
µ(k,n) = µ(1,nk ) if k divides n, µ(1,n) = (−1)m if n is a product of m distinct
primes, and µ(1,n) = 0 otherwise.
αk
Proof. The first claim is simple. For the second one let n = pα 1 · · · pk
1
be
the unique factorization of n into primes. Clearly, we only need to consider
the Möbius function of (X ′ ,), where X ′ consists of all positive integers that
divide n. Clearly, (X ′ ,) is the direct product of the linearly ordered sets
(X i ,), i = 1,· · · ,k, where X i = {1,pi ,· · · ,pα
i }. The theorem thus follows
i

from Theorem 1.4.7 and Exercise 1.4.5.

The Euler function φ (n) is equal to the number of integers 1 ≤ k ≤ n such

that k and n have the greates common divisor equal to 1. For example φ (9) = 6.
A classical formula that precedes all general Möbius inversion formulas reads as
follows.
Theorem 1.4.12. 
φ (n) = n (1 − 1/p),
p|n

where the product is over all distinct primes p dividing n.

Proof. Let d divide n. Then φ ( nd ) equals the number of integers 1 ≤ k ≤ n such
that the greatest common divisor of k,n is d, since any such integer k is of the
form k = dk′ where 1 ≤ k′ ≤ nd and the greatest common divisor of k′ and nd is
1.
We take the function f  in the Möbius inversion formula 1.4.6 to be the Euler
function φ . Let g(x) = yx f(y). We observe that g(n) = n, since for each
1 ≤ k ≤ n there is unique d such that the greatest common divisor of k and n
is d. The inversion formula and Theorem 1.4.11 thus give
  n
φ (n) = µ(d,n)d = µ(1,d) .
d
dn dn
1.4. PRINCIPLE OF INCLUSION AND EXCLUSION 11

Let p1 ,· · · ,pk be the distinct primes that divide n. Theorem 1.4.11 gives
   
n n n n
φ (n) = n − + + ··· + + + ··· − ··· =
p1 p2 p1 p2 p1 p3
k
 1
n (1 − ).
i=1
pi
Another Random Scribd Document
with Unrelated Content
 Fig. 21.—Rothersand Lighthouse.

The object of these several forms of optical apparatus is not only to produce characteristics or
distinctions in lights to enable them to be readily recognized by mariners, but to utilize the light
rays in directions above and below the horizontal plane, and also, in the case of revolving or
flashing lights, in azimuths not requiring to be illuminated for strengthening the beam in the
direction of the mariner. It will be seen that the effective condensation in flashing lights is very
much greater than in fixed belts, thus enabling higher intensities to be obtained by the use of
flashing lights than with fixed apparatus.
 Catoptric System.—Parabolic reflectors, consisting of small facets of silvered glass set in
plaster of Paris, were first used about the year 1763 in some of the Mersey lights by Mr
Hutchinson, then dock master at Liverpool (fig. 29). Spherical metallic reflectors were
introduced in France in 1781, followed by parabolic reflectors on silvered copper in 1790 in
England and France, and in Scotland in 1803. The earlier lights were of fixed type, a number
of reflectors being arranged on a frame or stand in such a manner that the pencils of
emergent rays overlapped and thus illuminated the whole horizon continuously. In 1783 the
first revolving light was erected at Marstrand in Sweden. Similar apparatus were installed at
Cordouan (1790), Flamborough Head (1806) and at the Bell Rock (1811). To produce a
revolving or flashing light the reflectors were fixed on a revolving carriage having several
faces. Three or more reflectors in a face were set with their axes parallel.

A type of parabolic reflector now in use is shown in fig. 30. The sizes in general use vary
from 21 in. to 24 in. diameter. These instruments are still largely used for light-vessel
illumination, and a few important land lights are at the present time of catoptric type,
including those at St Agnes (Scilly Islands), Cromer and St Anthony (Falmouth).
 Fig. 22.—Courses of various Lighthouse Towers.

Fig. 23.—Perspective drawing of Dovetailed Fig. 24.—Section

Stone (Wolf Rock). of Dovetail.
 Dioptric System.—The first adaptation of dioptric lenses to lighthouses is probably due to T.
Rogers, who used lenses at one of the Portland lighthouses between 1786 and 1790.
Subsequently lenses by the same maker were used at Howth, Waterford and the North
Foreland. Count Buffon had in 1748 proposed to grind out of a solid piece of glass a lens in
steps or concentric zones in order to reduce the thickness to a minimum (fig. 31). Condorcet
in 1773 and Sir D. Brewster in 1811 designed built-up lenses consisting of stepped annular
rings. Neither of these proposals, however, was intended to apply to lighthouse purposes. In
1822 Augustin Fresnel constructed a built-up annular lens in which the centres of curvature
of the different rings receded from the axis according to their distances from the centre, so
as practically to eliminate spherical aberration; the only spherical surface being the small
central part or “bull’s eye” (fig. 32). These lenses were intended for revolving lights only.
Fresnel next produced his cylindric refractor or lens belt, consisting of a zone of glass
generated by the revolution round a vertical axis of a medial section of the annular lens (fig.
33). The lens belt condensed and parallelized the light rays in the vertical plane only, while
the annular lens does so in every plane. The first revolving light constructed from Fresnel’s
designs was erected at the Cordouan lighthouse in 1823. It consisted of 8 panels of annular
lenses placed round the lamp at a focal distance of 920 mm. To utilize the light, which would
otherwise escape above the lenses, Fresnel introduced a series of 8 plain silvered mirrors, on
which the light was thrown by a system of lenses. At a subsequent period mirrors were also
placed in the lower part of the optic. The apparatus was revolved by clockwork. This optic
embodied the first combination of dioptric and catoptric elements in one design (fig. 34). In
the following year Fresnel designed a dioptric lens with catoptric mirrors for fixed light, which
was the first of its kind installed in a lighthouse. It was erected at the Chassiron lighthouse in
1827 (fig. 35). This combination is geometrically perfect, but not so practically on account of
the great loss of light entailed by metallic reflection which is at least 25% greater than the
system described under. Before his death in 1827 Fresnel devised his totally reflecting or
catadioptric prisms to take the place of the silvered reflectors previously used above and
below the lens elements (fig. 28). The ray Fi falling on the prismoidal ring ABC is refracted in
the direction i r and meeting the face AB at an angle of incidence greater than the critical, is
totally reflected in the direction r e emerging after second refraction in a horizontal direction.
Fresnel devised these prisms for use in fixed light apparatus, but the principle was, at a later
date, also applied to flashing lights, in the first instance by T. Stevenson. Both the dioptric
lens and catadioptric prism invented by Fresnel are still in general use, the mathematical
calculations of the great French designer still forming the basis upon which lighthouse
opticians work.
 Fig. 25.—Dassen Island Fig. 26.—Cape San Thomé
Lighthouse (cast iron). Lighthouse.

Fig. 28.—Catadioptric or
Fig. 27.—Dioptric Prism.
Reflecting Prism.

Fresnel also designed a form of fixed and flashing light in which the distinction of a fixed
light, varied by flashes, was produced by placing panels of straight refracting prisms in a
vertical position on a revolving carriage outside the fixed light apparatus. The revolution of
the upright prisms periodically increased the power of the beam, by condensation of the rays
emergent from the fixed apparatus, in the horizontal plane.
 The lens segments in Fresnel’s early apparatus were of polygonal form instead of
cylindrical, but subsequently manufacturers succeeded in grinding glass in cylindrical rings of
the form now used. The first apparatus of this description was made by Messrs Cookson of
Newcastle in 1836 at the suggestion of Alan Stevenson and erected at Inchkeith.

In 1825 the French Commission des Phares decided upon the exclusive use of lenticular
apparatus in its service. The Scottish Lighthouse Board followed with the Inchkeith revolving
apparatus in 1835 and the Isle of May fixed optic in 1836. In the latter instrument Alan
Stevenson introduced helical frames for holding the glass prisms in place, thus avoiding
complete obstruction of the light rays in any azimuth. The first dioptric light erected by the
Trinity House was that formerly at Start Point in Devonshire, constructed in 1836.
Catadioptric or reflecting prisms for revolving lights were not used until 1850, when Alan
Stevenson designed them for the North Ronaldshay lighthouse.

Dioptric Mirror.—The next important improvement in lighthouse optical work was the
invention of the dioptric spherical mirror by Mr (afterwards Sir) J. T. Chance in 1862. The
zones or prisms are generated round a vertical axis and divided into segments. This form of
mirror is still in general use (figs. 36 and 37).

Fig. 30.—Modern
Fig. 29.—Early Reflector and Lamp (1763).
Parabolic Reflector.

Azimuthal Condensing Prisms.—Previous to 1850 all apparatus were designed to emit light
of equal power in every azimuth either constantly or periodically. The only exception was
where a light was situated on a stretch of coast where a mirror could be placed behind the
flame to utilize the rays, which would otherwise pass landward, and reflect them back,
passing through the flame and lens in a seaward direction. In order to increase the intensity
of lights in certain azimuths T. Stevenson devised his azimuthal condensing prisms which, in
various forms and methods of application, have been largely used for the purpose of
strengthening the light rays in required directions as, for instance, where coloured sectors are
provided. Applications of this system will be referred to subsequently.

Optical Glass for Lighthouses.—In the early days of lens lights the only glass used for the
prisms was made in France at the St Gobain and Premontré works, which have long been
celebrated for the high quality of optical glass produced. The early dioptric lights erected in
the United Kingdom, some 13 in all, were made by Messrs Cookson of South Shields, who
were instructed by Léonor Fresnel, the brother of Augustin. At first they tried to mould the
lens and then to grind it out of one thick sheet of glass. The successors of the Cookson firm
abandoned the manufacture of lenses in 1845, and the firm of Letourneau & Lepaute of Paris
again became the monopolists. In 1850 Messrs Chance Bros. & Co. of Birmingham began the
manufacture of optical glass, assisted by M. Tabouret, a French expert who had been a
colleague of Augustin Fresnel himself. The first light made by the firm was shown at the
Great Exhibition of 1851, since when numerous dioptric apparatus have been constructed by
Messrs Chance, who are, at this time, the only manufacturers of lighthouse glass in the
United Kingdom. Most of the glass used for apparatus constructed in France is manufactured
at St Gobain. Some of the glass used by German constructors is made at Rathenow in Prussia
and Goslar in the Harz.

The glass generally employed for lighthouse optics has for its refractive index a mean value
of µ = 1.51, the corresponding critical angle being 41° 30′. Messrs Chance have used dense
flint glass for the upper and lower refracting rings of high angle lenses and for dioptric
mirrors in certain cases. This glass has a value of µ = l.62 with critical angle 38° 5′.

Fig. 31. Fig. 32. Fig. 33.

Buffon’s Lens. Fresnel’s Annular Lens. Fresnel’s Lens Belt.

Occulting Lights.—During the last 25 years of the 19th century the disadvantages of fixed
lights became more and more apparent. At the present day the practice of installing such,
except occasionally in the case of the smaller and less important of harbour or river lights,
has practically ceased. The necessity for providing a distinctive characteristic for every light
when possible has led to the conversion of many of the fixed-light apparatus of earlier years
into occulting lights, and often to their supersession by more modern and powerful flashing
apparatus. An occulting apparatus in general use consists of a cylindrical screen, fitting over
the burner, rapidly lowered and raised by means of a cam-wheel at stated intervals. The cam-
wheel is actuated by means of a weight or spring clock. Varying characteristics may be
procured by means of such a contrivance—single, double, triple or other systems of
occultation. The eclipses or periods of darkness bear
much the same relation to the times of illumination as
do the flashes to the eclipses in a revolving or flashing
light. In the case of a first-order fixed light the cost of
conversion to an occulting characteristic does not
exceed £250 to £300. With apparatus illuminated by
gas the occultations may be produced by successively
raising and lowering the gas at stated intervals.
Another form of occulting mechanism employed
consists of a series of vertical screens mounted on a
carriage and revolving round the burner. The carriage
is rotated on rollers or ball bearings or carried upon a
small mercury float. The usual driving mechanism
employed is a spring clock. “Otter” screens are used
in cases when it is desired to produce different
periods of occultations in two or more positions in
azimuth in order to differentiate sectors marking
shoals, &c. The screens are of sheet metal blacked
and arranged vertically, some what in the manner of
the laths of a venetian blind, and operated by
mechanical means.

Leading Lights.—In the case of lights designed to

act as a lead through a narrow channel or as direction
lights, it is undesirable to employ a flashing
apparatus. Fixed-light optics are employed to meet
such cases, and are generally fitted with occulting
mechanism. A typical apparatus of this description is
that at Gage Roads, Fremantle, West Australia (fig.
38). The occulting bright light covers the fairway, and
is flanked by sectors of occulting red and green light
marking dangers and intensified by vertical
condensing prisms. A good example of a holophotal
direction light was exhibited at the 1900 Paris
Exhibition, and afterwards erected at Suzac lighthouse
(France). The light consists of an annular lens 500 Fig. 34.—Fresnel’s Revolving
Apparatus at Cordouan Lighthouse.
mm. focal distance, of 180° horizontal angle and 157°
vertical, with a mirror of 180° at the back. The lens
throws a red beam of about 4½° amplitude in azimuth, and 50,000 candle-power over a
narrow channel. The illuminant is an incandescent petroleum vapour burner. Holophotal
direction lenses of this type can only be applied where the sector to be marked is of
comparatively small angle. Silvered metallic mirrors of parabolic form are also used for the
purpose. The use of single direction lights frequently renders the construction of separate
towers for leading lights unnecessary.
 If two distinct lights are employed to indicate the line of navigation through a channel or
between dangers they must be sufficiently far apart to afford a good lead, the front or
seaward light being situated at a lower elevation than the rear or landward one.

Coloured Lights.—Colour is used as seldom as possible as a distinction, entailing as it does

a considerable reduction in the power of the light. It is necessary in some instances for
differentiating sectors over dangers and for harbour lighting purposes. The use of coloured
lights as alternating flashes for lighthouse lights is not to be commended, on account of the
unequal absorption of the coloured and bright rays by the atmosphere. When such distinction
has been employed, as in the Wolf Rock apparatus, the red and white beams can be
approximately equalized in initial intensity by constructing the lens and prism panels for the
red light of larger angle than those for the white beams. Owing to the absorption by the red
colouring, the power of a red beam is only 40% of the intensity of the corresponding white
light. The corresponding intensity of green light is 25%. When red or green sectors are
employed they should invariably be reinforced by mirrors, azimuthal condensing prisms, or
other means to raise the coloured beam to approximately the same intensity as the white
light. With the introduction of group-flashing characteristics the necessity for using colour as
a means of distinction disappeared.

Fig. 35.—Fixed Apparatus at Chassiron Lighthouse (1827).

Fig. 36.—Vertical Section. Prism of Dioptric

Spherical Mirror.
 High-Angle Vertical Lenses.—Messrs Chance of Birmingham have manufactured lenses
having 97° of vertical amplitude, but this result was only attained by using dense flint glass
of high refractive index for the upper and lower elements. It is doubtful, however, whether
the use of refracting elements for a greater angle than 80° vertically is attended by any
material corresponding advantage.

Fig. 37.—Chance’s Dioptric Spherical Mirror.

Group Flashing Lights.—One of the most useful distinctions consists in the grouping of two
or more flashes separated by short intervals of darkness, the group being succeeded by a
longer eclipse. Thus two, three or more flashes of, say, half second duration or less follow
each other at intervals of about 2 seconds and are succeeded by an eclipse of, say, 10
seconds, the sequence being completed in a period of, say, 15 seconds. In 1874 Dr John
Hopkinson introduced the very valuable improvement of dividing the lenses of a dioptric
revolving light with the panels of reflecting prisms above and below them, setting them at an
angle to produce the group-flashing characteristic. The first apparatus of this type
constructed were those now in use at Tampico, Mexico and the Little Basses lighthouse,
Ceylon (double flashing). The Casquets apparatus (triple flashing) was installed in 1877. A
group-flashing catoptric light had, however, been exhibited from the “Royal Sovereign” light-
vessel in 1875. A sectional plan of the quadruple-flashing first order apparatus at Pendeen in
Cornwall is shown in fig. 39; and fig. 55 (Plate 1.) illustrates a double flashing first order light
at Pachena Point in British Columbia. Hopkinson’s system has been very extensively used,
most of the group-flashing lights shown in the accompanying tables, being designed upon
the general lines he introduced. A modification of the system consists in grouping two or
more lenses together separated by equal angles, and filling the remaining angle in azimuth
by a reinforcing mirror or screen. A group-flashing distinction was proposed for gas lights by
J. R. Wigham of Dublin, who obtained it in the case of a revolving apparatus by alternately
raising and lowering the flame. The first apparatus in which this method was employed was
erected at Galley Head, Co. Cork (1878). At this lighthouse 4 of Wigham’s large gas burners
with four tiers of first-order revolving lenses, eight in each tier, were adopted. By successive
lowering and raising of the gas flame at the focus of each tier of lenses he produced the
group-flashing distinction. The light showed, instead of one prolonged flash at intervals of
one minute, as would be produced by the apparatus in the absence of a gas occulter, a group
of short flashes varying in number between six and seven. The uncertainty, however, in the
number of flashes contained in each group is found to be an objection to the arrangement.
This device was adopted at other gas-illuminated stations in Ireland at subsequent dates.
The quadriform apparatus and gas installation at Galley Head were superseded in 1907 by a
first order bi-form apparatus with incandescent oil vapour burner showing five flashes every
20 seconds.

Fig. 38.—Gage Roads Direction Light.

Flashing Lights indicating Numbers.—

Captain F. A. Mahan, late engineer secretary
to the United States Lighthouse Board,
devised for that service a system of flashing
lights to indicate certain numbers. The
apparatus installed at Minot’s Ledge
lighthouse near Boston Harbour,
Massachusetts, has a flash indicating the
number 143, thus: - ---- ---, the dashes
indicating short flashes. Each group is
separated by a longer period of darkness than
that between successive members of a group.
The flashes in a group indicating a figure are Fig. 39.—Pendeen Apparatus. Plan at Focal
Plane.
about 1½ seconds apart, the groups being 3
seconds apart, an interval of 16 seconds’
darkness occurring between each repetition. Thus the number is repeated every half minute.
Two examples of this system were exhibited by the United States Lighthouse Board at the
Chicago Exhibition in 1893, viz. the second-order apparatus just mentioned and a similar light
of the first order for Cape Charles on the Virginian coast. The lenses are arranged in a
somewhat similar manner to an ordinary group-flashing light, the groups of lenses being
placed on one side of the optic, while the other is provided with a catadioptric mirror. This
system of numerical flashing for lighthouses has been frequently proposed in various forms,
notably by Lord Kelvin. The installation of the lights described is, however, the first practical
application of the system to large and important coast lights. The great cost involved in the
alteration of the lights of any country to comply with the requirements of a numerical system
is one of the objections to its general adoption.

Plate I.
Fig. 54.—FASTNET LIGHTHOUSE—FIRST ORDER Fig. 55.—PACHENA POINT LIGHTHOUSE, b.c.—FIRST
SINGLE-FLASHING BIFORM APPARATUS. ORDER DOUBLE-FLASHING APPARATUS.
Plate II.

Fig. 56.—OLD EDDYSTONE LIGHTHOUSE. Fig. 57.—EDDYSTONE LIGHTHOUSE.

 Fig. 58.—ILE VIERGE LIGHTHOUSE. Fig. 59.—MINOT’S LEDGE LIGHTHOUSE.

Hyper-radial Apparatus.—In 1885 Messrs

Barbier of Paris constructed the first hyper-radial
apparatus (1330 mm. focal distance) to the
design of Messrs D. and C. Stevenson. This had a
height of 1812 mm. It was tested during the
South Foreland experiments in comparison with
other lenses, and found to give excellent results
with burners of large focal diameter. Apparatus of
similar focal distance (1330 mm.) were
subsequently established at Round Island, Bishop
Rock, and Spurn Point in England, Fair Isle and
Sule Skerry (fig. 40) in Scotland, Bull Rock and
Tory Island in Ireland, Cape d’Antifer in France,
Pei Yu-shan in China and a lighthouse in Brazil.

The light erected in 1907 at Cape Race,

Newfoundland, is a fine example of a four-sided
hyper-radial apparatus mounted on a mercury
float. The total weight of the revolving part of
the light amounts to 7 tons, while the motive Fig. 40.—Sule Skerry Apparatus.
clock weight required to rotate this large mass at
a speed of two complete revolutions a minute is only 8 cwt. and the weight of mercury
required for flotation 950 ℔. A similar apparatus was placed at Manora Point, Karachi, India,
in 1908 (fig. 41).

The introduction of incandescent and other burners of focal compactness and high
intensity has rendered the use of optics of such large dimensions as the above, intended for
burners of great focal diameter, unnecessary. It is now possible to obtain with a second-order
optic (or one of 700 mm. focal distance), having a powerful incandescent petroleum burner
in focus, a beam of equal intensity to that which would be obtained from the apparatus
having a 10-wick oil burner or 108-jet gas burner at its focus.

Stephenson’s Spherical Lenses and Equiangular Prisms.—Mr C. A. Stephenson in 1888

designed a form of lens spherical in the horizontal and vertical sections. This admitted of the
construction of lenses of long focal distance without the otherwise corresponding necessity of
increased diameter of lantern. A lens of this type and of 1330 mm. focal distance was
constructed in 1890 for Fair Isle lighthouse. The spherical form loses in efficiency if carried
beyond an angle subtending 20° at the focus, and to obviate this loss Mr Stephenson
designed his equiangular prisms, which have an inclination outwards. It is claimed by the
designer that the use of equiangular prisms results in less loss of light and less divergence
than is the case when either the spherical or Fresnel form is adopted. An example of this
design is seen (fig. 40) in the Sule Skerry apparatus (1895).

Fixed and Flashing Lights.—The use of these lights, which show a fixed beam varied at
intervals by more powerful flashes, is not to be recommended, though a large number were
constructed in the earlier years of dioptric illumination and many are still in existence. The
distinction can be produced in one or other of three ways: (a) by the revolution of detached
panels of straight condensing lens prisms placed vertically around a fixed light optic, (b) by
utilizing revolving lens panels in the middle portion of the optic to produce the flashing light,
the upper and lower sections of the apparatus being fixed zones of catadioptric or reflecting
elements emitting a fixed belt of light, and (c) by interposing panels of fixed light section
between the flashing light panels of a revolving apparatus. In certain conditions of the
atmosphere it is possible for the fixed light of low power to be entirely obscured while the
flashes are visible, thus vitiating the true characteristic of the light. Cases have frequently
occurred of such lights being mistaken for, and even described in lists of light as, revolving or
flashing lights.

”Cute” and Screens.—Screens of coloured glass, intended to distinguish the light in

particular azimuths, and of sheet iron, when it is desired to “cut off” the light sharply on any
angle, should be fixed as far from the centre of the light as possible in order to reduce the
escape of light rays due to divergence. These screens are usually attached to the lantern
framing.

Divergence.—A dioptric apparatus designed to bend all incident rays of light from the light
source in a horizontal direction would, if the flame could be a point, have the effect of
projecting a horizontal band or zone of light, in the case of a fixed apparatus, and a cylinder
of light rays, in the case of a flashing light, towards the horizon. Thus the mariner in the near
distance would receive no light, the rays, visible only at or near the horizon, passing above
the level of his eye. In practice this does not occur, sufficient natural divergence being
produced ordinarily owing to the magnitude of the flame. Where the electric arc is employed
it is often necessary to design the prisms so as to produce artificial divergence. The measure
of the natural divergence for any point of the lens is the angle whose sine is the ratio of the
diameter of the flame to the distance of the point from centre of flame.

In the case of vertical divergence the mean height of the flame must be substituted for the
diameter. The angle thus obtained is the total divergence, that is, the sum of the angles
above and below the horizontal plane or to right and left of the medial section. In fixed
dioptric lights there is, of course, no divergence in the horizontal plane. In flashing lights the
horizontal divergence is a matter of considerable importance, determining as it does the
duration or length of time the flash is visible to the mariner.

Feux-Éclairs or Quick Flashing Lights.—One of the most important developments in the

character of lighthouse illuminating apparatus that has occurred in recent years has been in
the direction of reducing the length of flash. The initiative in this matter was taken by the
French lighthouse authorities, and in France alone forty lights of this type were established
between 1892 and 1901. The use of short flash lights rapidly spread to other parts of the
world. In England the lighthouse at Pendeen (1900) exhibits a quadruple flash every 15
seconds, the flashes being about ¼ second duration (fig. 39), while the bivalve apparatus
erected on Lundy Island (1897) shows 2 flashes of 1⁄3 second duration in quick succession
every 20 seconds. Since 1900 many quick flashing lights have been erected on the coasts of
the United Kingdom and in other countries. The early feux-éclairs, designed by the French
engineers and others, had usually a flash of 1⁄10th to 1⁄3rd of a second duration. As a result of
experiments carried out in France in 1903-1904, 3⁄10 second has been adopted by the French
authorities as the minimum duration for white flashing lights. If shorter flashes are used it is
found that the reduction in duration is attended by a corresponding, but not proportionate,
diminution in effective intensity. In the case of many electric flashing lights the duration is of
necessity reduced, but the greater initial intensity of the flash permits this loss without
serious detriment to efficiency. Red or green requires a considerably greater duration than do
white flashes. The intervals between the flashes in lights of this character are also small, 2½
seconds to 7 seconds. In group-flashing lights the intervals between the flashes are about 2
seconds or even less, with periods of 7 to 10 or 15 seconds between the groups. The flashes
are arranged in single, double, triple or even quadruple groups, as in the older forms of
apparatus. The feu-éclair type of apparatus enables a far higher intensity of flash to be
obtained than was previously possible without any corresponding increase in the luminous
power of the burner or other source of light. This result depends entirely upon the greater
ratio of condensation of light employed, panels of greater angular breadth than was
customary in the older forms of apparatus being used with a higher rotatory velocity. It has
been urged that short flashes are insufficient for taking bearings, but the utility of a light in
this respect does not seem to depend so much upon the actual length of the flash as upon its
frequent recurrence at short intervals. At the Paris Exhibition of 1900 was exhibited a fifth-
order flashing light giving short flashes at 1 second intervals; this represents the extreme to
which the movement towards the reduction of the period of flashing lights has yet been
carried.

Mercury Floats.—It has naturally been found impracticable to revolve the optical apparatus
of a light with its mountings, sometimes weighing over 7 tons, at the high rate of speed
required for feux-éclairs by means of the old system of roller carriages, though for some
small quick-revolving lights ball bearings have been successfully adopted. It has therefore
become almost the universal practice to carry the rotating portions of the apparatus upon a
mercury float. This beautiful application of mercury rotation was the invention of Bourdelles,
and is now utilized not only for the high-speed apparatus, but also generally for the few
examples of the older type still being constructed. The arrangement consists of an annular
cast iron bath or trough of such dimensions that a similar but slightly smaller annular float
immersed in the bath and surrounded by mercury displaces a volume of the liquid metal
whose weight is equal to that of the apparatus supported. Thus a comparatively insignificant
quantity of mercury, say 2 cwt., serves to ensure the flotation of a mass of over 3 tons.
Certain differences exist between the type of float usually constructed in France and those
generally designed by English engineers. In all cases provision is made for lowering the
mercury bath or raising the float and apparatus for examination. Examples of mercury floats
are shown in figs. 41, 42, 43 and Plate I., figs. 54 and 55.
 Fig. 41.—Manora Point Apparatus and Lantern.

Multiform Apparatus.—In order to double the power to be obtained from a single apparatus
at stations where lights of exceptionally high intensity are desired, the expedient of placing
one complete lens apparatus above another has sometimes been adopted, as at the Bishop
Rock (fig. 13), and at the Fastnet lighthouse in Ireland (Plate I., fig. 54). Triform and
quadriform apparatus have also been erected in Ireland; particulars of the Tory Island triform
apparatus will be found in table VII. The adoption of the multiform system involves the use
of lanterns of increased height.

Twin Apparatus.—Another method of doubling the power of a light is by mounting two

complete and distinct optics side by side on the same revolving table, as I shown in fig. 43 of
the Île Vierge apparatus. Several such lights have been installed by the French Lighthouse
Service.

Port Lights.—Small self-contained lanterns and lights are in common use for marking the
entrances to harbours and in other similar positions where neither high power nor long range
is requisite. Many such lights are unattended in the sense that they do not require the
attention of a keeper for days and even weeks together. These are described in more detail
in section 6 of this article. A typical port light consists of a copper or brass lantern containing
a lens of the fourth order (250 mm. focal distance) or smaller, and a single wick or 2-wick
Argand capillary burner. Duplex burners are also used. The apparatus may exhibit a fixed
light or, more usually, an occulting characteristic is produced by the revolution of screens
actuated by spring clockwork around the burner. The lantern may be placed at the top of a
column, or suspended from the head of a mast. Coal gas and electricity are also used as
illuminants for port lights when local supplies are available. The optical apparatus used in
connexion with electric light is described below.

”Orders” of Apparatus.—Augustin Fresnel divided the dioptric lenses, designed by him, into
“orders” or sizes depending on their local distance. This division is still used, although two
additional “orders,” known as “small third order” and “hyper-radial” respectively are in
ordinary use. The following table gives the principal dimensions of the several sizes in use:—

Table II.

Vertical Angles of Optics.

Focal (Ordinary Dimensions.)
Order. Distance, Holophotal Optics.
Dioptric
mm. Lower Upper
Belt only. Lens.
Prisms. Prisms.
Hyper-Radial 1330 80° 21° 57° 48°
1st order 920 92°, 80°, 58° 21° 57° 48°
2nd order 700 80° 21° 57° 48°
 3rd order 500 80° 21° 57° 48°
Small 3rd order 375 80° 21° 57° 48°
4th order 250 80° 21° 57° 48°
5th order 187.5 80° 21° 57° 48°
6th order 150 80° 21° 57° 48°

Lenses of small focal distance are also made for buoy and beacon lights.
Fig. 42.—Cape Naturaliste Apparatus.
 Fig. 43.—Île Vierge Apparatus.

Light Intensities.—The powers of lighthouse lights in the British Empire are expressed in
terms of standard candles or in “lighthouse units” (one lighthouse unit = 1000 standard
candles). In France the unit is the “Carcel” = .952 standard candle. The powers of burners
and optical apparatus, then in use in the United Kingdom, were carefully determined by
actual photometric measurement in 1892 by a committee consisting of the engineers of the
three general lighthouse boards, and the values so obtained are used as the basis for
calculating the intensities of all British lights. It was found that the intensities determined by
photometric measurement were considerably less than the values given by the theoretical
Welcome to our website – the perfect destination for book lovers and
knowledge seekers. We believe that every book holds a new world,
offering opportunities for learning, discovery, and personal growth.
That’s why we are dedicated to bringing you a diverse collection of
books, ranging from classic literature and specialized publications to
self-development guides and children's books.

More than just a book-buying platform, we strive to be a bridge

connecting you with timeless cultural and intellectual values. With an
elegant, user-friendly interface and a smart search system, you can
quickly find the books that best suit your interests. Additionally,
our special promotions and home delivery services help you save time
and fully enjoy the joy of reading.

Join us on a journey of knowledge exploration, passion nurturing, and

personal growth every day!

ebookbell.com