Around the Unit Circle: Mahler Measure,

Integer Matrices and Roots of Unity 1st

Edition James Mckee
Visit to download the full and correct content document:
https://ebookmeta.com/product/around-the-unit-circle-mahler-measure-integer-matric
es-and-roots-of-unity-1st-edition-james-mckee/
More products digital (pdf, epub, mobi) instant
download maybe you interests ...

Precalculus: Concepts Through Functions, A Unit Circle

Approach to Trigonometry 4th Edition Michael Sullivan

https://ebookmeta.com/product/precalculus-concepts-through-
functions-a-unit-circle-approach-to-trigonometry-4th-edition-
michael-sullivan/

The Star Around The Sun ■■■■ Pt 1 1st Edition Diamond

Circle Work

https://ebookmeta.com/product/the-star-around-the-
sun-%e6%98%8e%e6%97%a5%e6%98%9f%e7%a8%8b-pt-1-1st-edition-
diamond-circle-work/

The Future of Think Tanks and Policy Advice Around the

World 1st Edition James Mcgann

https://ebookmeta.com/product/the-future-of-think-tanks-and-
policy-advice-around-the-world-1st-edition-james-mcgann/

Basic Analysis IV Measure Theory and Integration 1st

Edition James K. Peterson

https://ebookmeta.com/product/basic-analysis-iv-measure-theory-
and-integration-1st-edition-james-k-peterson/
Primary Mathematics 3A Hoerst

https://ebookmeta.com/product/primary-mathematics-3a-hoerst/

Beauty and Sadness Mahler s 11 Symphonies 1st Edition

David Vernon Marina Mahler

https://ebookmeta.com/product/beauty-and-sadness-
mahler-s-11-symphonies-1st-edition-david-vernon-marina-mahler/

Linear Algebra and Matrices 1st Edition Shmuel

Friedland

https://ebookmeta.com/product/linear-algebra-and-matrices-1st-
edition-shmuel-friedland/

The Eighth Mahler and the World in 1910 1st Edition

Stephen Johnson

https://ebookmeta.com/product/the-eighth-mahler-and-the-world-
in-1910-1st-edition-stephen-johnson/

The Unity of Science 1st Edition David Bensimon

https://ebookmeta.com/product/the-unity-of-science-1st-edition-
david-bensimon/
Universitext

James McKee
Chris Smyth

Around
the Unit
Circle
Mahler Measure, Integer Matrices and
Roots of Unity
Universitext

Series Editors
Carles Casacuberta, Universitat de Barcelona, Barcelona, Spain
John Greenlees, University of Warwick, Coventry, UK
Angus MacIntyre, Queen Mary University of London, London, UK
Claude Sabbah, École Polytechnique, CNRS, Université Paris-Saclay, Palaiseau,
France
Endre Süli, University of Oxford, Oxford, UK
Wojbor A. Woyczyński, Case Western Reserve University, Cleveland, OH, USA
Universitext is a series of textbooks that presents material from a wide variety of
mathematical disciplines at master’s level and beyond. The books, often well
class-tested by their author, may have an informal, personal even experimental
approach to their subject matter. Some of the most successful and established books
in the series have evolved through several editions, always following the evolution
of teaching curricula, into very polished texts.
Thus as research topics trickle down into graduate-level teaching, first textbooks
written for new, cutting-edge courses may make their way into Universitext.

More information about this series at https://link.springer.com/bookseries/223

James McKee Chris Smyth
•

Around the Unit Circle

Mahler Measure, Integer Matrices and Roots
of Unity

123
James McKee Chris Smyth
Egham, UK Edinburgh, UK

ISSN 0172-5939 ISSN 2191-6675 (electronic)

Universitext
ISBN 978-3-030-80030-7 ISBN 978-3-030-80031-4 (eBook)
https://doi.org/10.1007/978-3-030-80031-4

Mathematics Subject Classification: 11C08, 11C20, 11R06, 05A05, 05C20, 05C22, 11R18, 11S05,
15B36

© Springer Nature Switzerland AG 2021

This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part
of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations,
recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission
or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar
methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this
publication does not imply, even in the absence of a specific statement, that such names are exempt from
the relevant protective laws and regulations and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and information in this
book are believed to be true and accurate at the date of publication. Neither the publisher nor the
authors or the editors give a warranty, expressed or implied, with respect to the material contained
herein or for any errors or omissions that may have been made. The publisher remains neutral with regard
to jurisdictional claims in published maps and institutional affiliations.

This Springer imprint is published by the registered company Springer Nature Switzerland AG
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
You shall see them on a beautiful quarto
page, where a neat rivulet of text shall
meander through a meadow of margin.
Sheridan, School for Scandal.
Preface

Consider an algebraic number a. By definition, it is the zero in some field of an

irreducible polynomial Pa with integer coefficients. The complete set of zeros of
such a polynomial constitutes a conjugate set of algebraic numbers. The study of
such sets is the subject of this book. We embed them in C and study them there. We
study them embedded in Qp , the algebraic closure of the field Qp of p-adic
numbers.
Very often, we restrict to algebraic integers, where Pa , the minimal polynomial
of a, is monic (has leading coefficient 1). We study conjugate sets of algebraic
integers as eigenvalues of matrices with integer entries. More generally, we study
algebraic integers by attaching combinatorial objects, such as graphs or directed
graphs (digraphs), to them. We study conjugate sets when they are all real. We
study them when they are all sums of roots of unity.
We are also interested in conjugate sets that are small, in some sense. For this
purpose, the Mahler measure (a height function) is the main tool. For conjugate sets
of totally real algebraic integers, their span (diameter) is also important. For totally
positive algebraic numbers, the trace is studied, too, both as the sum of the elements
of a conjugate set and as the sum of the eigenvalues of an integer symmetric matrix.
The house (maximum modulus of the elements of a conjugate set) is another
important height function. It has attracted renewed attention recently with
Dimitrov’s surprising proof of the Schinzel–Zassenhaus Conjecture. We present his
proof.
The range of values assumed by a height function is an interesting topic in its
own right. We study the set of these values for the Mahler measure of polynomials
with integer coefficients, in any number of variables. This includes ‘Lehmer’s
Conjecture’, concerning the question of whether for noncyclotomic polynomials
such Mahler measures are bounded away from 1. For algebraic integers in a
cyclotomic field, we do the same for their Cassels height (the mean square modulus
of their conjugates).

vii
viii Preface

Certain sets of square integer matrices are a particularly interesting source of

algebraic integers, through their eigenvalues. While every algebraic integer is an
eigenvalue of the companion matrix of its minimal polynomial, when we restrict to
symmetric or other sets of such matrices, we get some challenging questions. Some
of these are discussed in the book.
We now outline the contents of the chapters. Chapter 1 contains basic facts about
the Mahler measure of polynomials in one variable and some applications. We also
discuss small values of the Mahler measure of integer polynomials. Tables of these,
and also their presumed small limit points, are given in Appendix D.
Chapter 2 contains much material about the Mahler measure of integer poly-
nomials in several variables. This includes a discussion of Boyd’s Conjecture that
the set of such Mahler measures is closed. It also includes specific closed form
evaluations of particular Mahler measures.
Chapter 3 contains a version of Dobrowolski’s proof of the strongest uncondi-
tional result in the direction of the proof of Lehmer’s Conjecture. We essentially
follow the proof of Cantor and Straus, which uses the confluent Vandermonde
determinant.
Chapter 4 gives an account of Dimitrov’s recent proof of the Schinzel–
Zassenhaus Conjecture.
Chapter 5 concerns roots of unity. One topic is finding cyclotomic points (points
whose coordinates are roots of unity) on plane curves. Another topic is a discussion
of Robinson’s two problems and five conjectures concerning cyclotomic integers.
Many of our chapters concern integer square matrices. They have a visual
interpretation as signed multidigraphs. So for A ¼ ðaij Þ, the entry aij 6¼ 0 for i 6¼ j
can be regarded as a directed edge from vertex i to vertex j of weight aij . For i ¼ j
the entry aii 6¼ 0 represents a charge aii on the ith vertex. When A is symmetric with
all entries either 0 or 1 and all diagonal entries 0, we simply have a graph (A being
its adjacency matrix). If A is symmetric with all entries 1, 0 or 1 and all diagonal
entries 0, we have a signed graph. If we allow nonzero diagonal entries we have a
charged graph or charged signed graph. Throughout the book, we move freely
between the matrix and graphical interpretations of these objects.
Chapters 6 and 7 concern cyclotomic matrices. These are integer symmetric
matrices whose eigenvalues are all of the form x þ x1 for some root of unity x. It
turns out that essentially all of them are (adjacency matrices of) graphs, signed
graphs or charged signed graphs. After describing how cyclotomic matrices can be
‘grown’, the important tool of Gram vectors is introduced. The Classification
Theorem for cyclotomic matrices is then stated. The classification includes signed
graphs that tessellate a torus and charged signed graphs that tessellate a cylinder. Its
proof occupies Chap. 7.
Chapter 8 is devoted to a description of the structure of the set of all Cassels
heights of cyclotomic integers as what we call a Thue set. Further, a complete
description of the values of the Cassels heights of integers in a cyclotomic field
Qðxp Þ, where xp is a pth root of unity for a prime p, is given.
Preface ix

In Chap. 9, some structure theory is developed to show which (charged, signed)

graphs can be embedded in the toroidal or cylindrical tessellations of Chap. 6. This
will be useful in Chap. 13 for describing minimal noncyclotomic matrices.
Chapter 10 gives basic facts about the transfinite diameter of subsets of the
complex plane, as well as some applications to describing which such subsets can
contain infinitely many conjugate sets of algebraic integers. It also summarises
some results concerning the integer transfinite diameter and the monic integer
transfinite diameter. Some of these results are used in Dimitrov’s proof in Chap. 4
of the Schinzel–Zassenhaus Conjecture.
In Chap. 11, we prove or survey various lower bounds for Mahler measures that
apply when the integer polynomials have their zeros restricted in some way. One
such restriction is to bound the number of monomials of the polynomial. Another,
the totally p-adic case, is to restrict to polynomials that split completely over the
field of p-adic numbers. A third is to assume that the polynomials define abelian
extensions of the rationals. A fourth is to restrict the zeros to a compact set not
containing all of the unit circle (Langevin’s Theorem). In these situations, stronger
results can be obtained than in the unrestricted general case. In Chap. 12, another
restricted case, that of nonreciprocal polynomials, is studied.
In Chap. 13, all minimal noncyclotomic matrices are found. These are noncy-
clotomic integer square matrices for which the removal of any row and corre-
sponding column gives a cyclotomic matrix.
The explicit auxiliary function method, described in Chap. 14, is used to find the
smallest mean values of a given function f on the values of various restricted sets of
conjugate algebraic integers or numbers. Knowing that certain resultants are non-
zero expresses the problem as a linear programming problem that, when dualised,
gives an auxiliary function to be minimised. This then often gives a discrete
spectrum of the smallest mean values of f on these restricted sets. Two important
examples of this method are f ðxÞ ¼ x and f ðxÞ ¼ maxð1; log xÞ for x [ 0. The first
gives a spectrum of small values of the mean trace of totally positive algebraic
integers (the Schur–Siegel–Smyth trace problem), while the second does the same
for the absolute Mahler measure of such algebraic integers.
In Chap. 15, the analogous trace problem for positive definite integer symmetric
matrices is studied. In contrast to the trace problem of the previous chapter, this
problem can be completely solved: the trace of such an n  n matrix is at least
2n  1, and this bound is attained.
The small-span problem for conjugate sets of totally real algebraic integers is a
classical one: which such sets lie in an interval of length less than 4? In Chap. 16,
the analogous problem is solved for integer symmetric matrices.
The next three chapters concern symmetrizable matrices. Chapter 17 describes
the structure of symmetrizable matrices, as well as presenting other relevant
material needed for the following chapters. In Chap. 18, all (nonsymmetric) max-
imal symmetrizable cyclotomic matrices are found, and Chap. 19 concerns the trace
problem for nonsymmetric symmetrizable integer matrices.
x Preface

Salem graphs are a special kind of graph that can be associated with a Salem
number or quadratic Pisot number in a specific way. In Chap. 20, families of these
are constructed using certain rational functions called rational interlacing quotients
and circular interlacing quotients. Special cases of these constructions have an
application to the following chapter on minimal polynomials of integer symmetric
matrices.
Exercises in the book are distributed throughout the text, including the appen-
dices. They are of very variable difficulty, ranging from the almost trivial to the
challenging. An example of an easy one is the following.
Exercise 0.1 (Serre [Ser19, (A.7.5)]) Show that when the complex numbers w
and z are of modulus 1 then

jðw  zÞðw  zÞj ¼ jðw þ wÞ  ðz þ zÞj:

If we feel that an exercise is even trickier, and perhaps that we do not know the
answer, it is marked as a Problem. Some exercises require light use of a computer,
and some require more substantial programming: these are flagged as computational
exercises. Unsolved problems are presented either as Research Problems or Open
Problems, the former generally being more open-ended and less notorious.
Most chapters end with a section of Notes, followed by a Glossary of terms used
in that chapter. The Notes include references to the sources of some of the results
and proofs that have been presented, as well as references to related work. For
results that have been stated in the text, but not proved, references are given beside
the result.
We end this preface with a list of some books and survey articles containing
material related to that in our book, which could be consulted. First of all, we highly
recommend the recent text on Mahler measure by Brunault and Zudilin [BZ20]. It
contains some advanced material that we have referred to only briefly: for instance,
a detailed account of Deninger’s work, and treatment of Mahler measure of elliptic
curves in much greater depth than we have described. Other texts containing a
significant to amount of Mahler measure material are those of Schinzel [Sch00] and
Bombieri and Gubler [BG06]. For results on real and complex polynomials in
general, the scholarly book of Rahman and Schmeisser [RS02] is a very valuable
source of results and references.
While we treat some aspects of the spectra of graphs, the topic is well covered in
the books by Godsil and Royle [GR01] and Brouwer and Haemers [BH12]. We
refer to Pisot and Salem numbers only briefly—they are equal to their Mahler
measures! For a more extensive treatment, see the book [BDGGH+92] by Bertin
et al. The classic text [Sal63] by Salem is also well worth reading.
In our treatment of roots of unity and cyclotomic integers in Chap. 5, the
considerable literature on vanishing sums of roots of unity is not covered. See for
instance Zannier’s survey paper [Zan95] for an account of this area.
The idea of producing a book of this sort has been with us for several years.
Right from the start, Rémi Lodh at Springer gave us encouragement to pursue the
dream, and was patiently supportive of our slow progress. We owe him a huge debt
Preface xi

for nudging us at the right moments and continuing to believe in the project. We are
also appreciative of the substantial help and encouragement provided by the
reviewers of the manuscript. We thank Igor Pritsker for detailed feedback on the
chapter on the transfinite diameter. To all these and to all our friends and colleagues
who have helped us through this work, thank you: we are glad it is now finished!

Royal Holloway, UK James McKee

Edinburgh, UK Chris Smyth
May 2021
Contents

1 Mahler Measures of Polynomials in One Variable . . . . . . ....... 1

1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ....... 1
1.1.1 Polynomials over the Field C of Complex
Numbers . . . . . . . . . . . . . . . . . . . . . . . . . ....... 1
1.1.2 Polynomials over the Field Q of Rational
Numbers . . . . . . . . . . . . . . . . . . . . . . . . . ....... 2
1.2 Kronecker’s Two Theorems . . . . . . . . . . . . . . . . . . . ....... 3
1.3 Mahler Measure Inequalities . . . . . . . . . . . . . . . . . . ....... 7
1.4 A Lower Bound for an Integer Polynomial Evaluated
at an Algebraic Number . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.5 Polynomials with Small Coefficients . . . . . . . . . . . . . . . . . . . . 10
1.6 Separation of Conjugates . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.7 The Shortness of a Polynomial . . . . . . . . . . . . . . . . . . . . . . . 14
1.7.1 Finding Short Polynomials . . . . . . . . . . . . . . . . . . . 16
1.8 Variants of Mahler Measure . . . . . . . . . . . . . . . . . . . . . . . . . . 18
1.8.1 The Weil Height . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
1.9 Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
1.10 Glossary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2 Mahler Measures of Polynomials in Several Variables . . . . . . . . . . 25
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.2 Preliminaries for the Proofs of Theorems 2.5 and 2.6 . . . . . . . 28
2.3 Proof of Theorem 2.5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
2.4 Proof of Theorem 2.6 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
2.5 Computation of Two-Dimensional Mahler Measures . . . . . . . . 34
2.6 Small Limit Points of L? . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
2.6.1 Shortness Conjectures Implying Lehmer’s
Conjecture and Structural Results for L . . . ....... 35
2.6.2 Small Elements of the Set of Two-Variable
Mahler Measures . . . . . . . . . . . . . . . . . . . ....... 37

xiii
xiv Contents

2.7 Closed Forms for Mahler Measures of Polynomials

of Dimension at Least 2 . . . . . . . . . . . . . . . . . . . . . . . . . . .. 38
2.7.1 Dirichlet L-Functions . . . . . . . . . . . . . . . . . . . . . .. 42
2.7.2 Some Explicit Formulae for Two-Dimensional
Mahler Measures . . . . . . . . . . . . . . . . . . . . . . . . .. 43
2.7.3 Mahler Measures of Elliptic Curves . . . . . . . . . . . .. 46
2.7.4 Mahler Measure of Three-Dimensional
Polynomials . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 48
2.7.5 Mahler Measure Formulae for Some Polynomials
of Dimension at Least 4 . . . . . . . . . . . . . . . . . . . . . 52
2.7.6 An Asymptotic Mahler Measure Result . . . . . . . . . . 53
2.8 Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
2.9 Glossary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
3 Dobrowolski’s Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . ....... 57
3.1 The Theorem and Preliminary Lemmas . . . . . . . . . . ....... 57
3.2 Proof of Theorem 3.1: Dobrowolski’s Lower Bound
for MðaÞ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ....... 63
3.3 Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ....... 65
3.4 Glossary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ....... 66
4 The Schinzel–Zassenhaus Conjecture . . . . . . . . . . . . . . . . . . . . . . . 69
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
4.1.1 A Simple Proof of a Weaker Result . . . . . . . . . . . . . 70
4.2 Proof of Dimitrov’s Theorem . . . . . . . . . . . . . . . . . . . . . . . . . 70
4.3 Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
4.4 Glossary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
5 Roots of Unity and Cyclotomic Polynomials . . . . . . . . . . . . . . . . . . 77
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
5.2 Solving Polynomial Equations in Roots of Unity . . . . . . . . . . 80
5.3 Cyclotomic Points on Curves . . . . . . . . . . . . . . . . . . . . . . . . . 83
5.3.1 Definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
5.3.2 Lðf Þ of Rank 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
5.3.3 Lðf Þ Full of Rank 2 . . . . . . . . . . . . . . . . . . . . . . . . 84
5.3.4 Lðf Þ of Rank 2, but Not Full . . . . . . . . . . . . . . . . . 85
5.3.5 The Case of f Reducible . . . . . . . . . . . . . . . . . . . . . 85
5.3.6 An Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
5.4 Cyclotomic Integers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
5.4.1 Introduction to Cyclotomic Integers . . . . . . . . . . . . . 87
5.4.2 The Function NðbÞ . . . . . . . . . . . . . . . . . . . . . . . . 90
pffiffiffi
5.4.3 Evaluating or Estimating Nð d Þ . . . . . . . . . ..... 92
5.4.4 Evaluation of the Gauss Sum . . . . . . . . . . . . ..... 93
5.4.5 The Absolute Mahler Measure of Cyclotomic
Integers . . . . . . . . . . . . . . . . . . . . . . . . . . . . ..... 95
Contents xv

5.5 Robinson’s Problems and Conjectures . . . . . . . . . . . . . . . . . . 95

5.6 Cassels’ Lemmas for MðbÞ . . . . . . . . . . . . . . . . . . . . . . . . . . 97
5.7 Discussion of Robinson’s Problems . . . . . . . . . . . . . . . . . . . . 101
5.7.1 Robinson’s First Problem . . . . . . . . . . . . . . . . . . . . 101
5.7.2 Robinson’s Second Problem . . . . . . . . . . . . . . . . . . 105
5.8 Discussion of Robinson’s Conjectures . . . . . . . . . . . . . . . . . . 105
5.8.1 The First Conjecture . . . . . . . . . . . . . . . . . . . . . . . . 105
5.8.2 The Second Conjecture . . . . . . . . . . . . . . . . . . . . . . 106
5.8.3 The Third Conjecture . . . . . . . . . . . . . . . . . . . . . . . 106
5.8.4 The Fourth Conjecture . . . . . . . . . . . . . . . . . . . . . . 107
5.8.5 The Fifth Conjecture . . . . . . . . . . . . . . . . . . . . . . . . 107
5.9 Multiplicative Relations Between Conjugate Roots of Unity . . 108
5.10 Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ...... . 109
5.11 Glossary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ...... . 110
6 Cyclotomic Integer Symmetric Matrices I: Tools and Statement
of the Classification Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
6.2 The Mahler Measure of a Matrix and Cyclotomic Matrices . . . 114
6.3 Flavours of Equivalence: Isomorphism, Equivalence and
Strong Equivalence of Matrices . . . . . . . . . . . . . . . . . . . . . . . 117
6.4 Growing Cyclotomic Matrices . . . . . . . . . . . . . . . . . . . . . . . . 122
6.5 Gram Vectors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
6.6 Statement of the Classification Theorem for Cyclotomic
Integer Symmetric Matrices . . . . . . . . . . . . . . . . . . . . . . . . . . 131
6.7 Glossary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
7 Cyclotomic Integer Symmetric Matrices II: Proof
of the Classification Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
7.1 Cyclotomic Signed Graphs . . . . . . . . . . . . . . . . . . . . . . . . . . 137
7.2 Cyclotomic Charged Signed Graphs . . . . . . . . . . . . . . . . . . . . 140
7.3 Cyclotomic Integer Symmetric Matrices: Completion
of the Classification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
7.4 Further Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
7.5 Notes on Chaps. 6 and 7 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
7.6 Glossary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
8 The Set of Cassels Heights . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
8.1 Cassels Height and the Set C . . . . . . . . . . . . . . . . . . . . . . . . . 149
8.2 The Derived Sets and the Sumsets of C . . . . . . . . . . . . . . . . . 154
8.3 Proof of Theorem 8.4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
8.3.1 Structure and Labelling of Thue Sets . . . . . . . . . . . . 159
8.4 Cassels Heights of Cyclotomic Integers in Qðxp Þ . . . . . . . . . . 161
8.5 Proof of Theorem 8.14 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162
8.6 Proof of Theorem 8.13 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163
xvi Contents

8.7 Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168

8.8 Glossary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168
9 Cyclotomic Integer Symmetric Matrices Embedded
in Toroidal and Cylindrical Tessellations . . . . . . . . . . . . . . . . . . . . 171
9.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171
9.2 Preliminaries: Notation and Tools . . . . . . . . . . . . . . . . . . . . . 171
9.3 Cyclotomic Graphs Embedded in T2k . . . . . . . . . . . . . . . . . . . 175
9.4 Changes for Charges . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180
9.5 Glossary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 184
10 The Transfinite Diameter and Conjugate Sets of Algebraic
Integers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187
10.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187
10.2 Analytic Properties of the Transfinite Diameter . . . . . . . . . . . . 187
10.3 Application to Conjugate Sets of Algebraic Integers . . . . . . . . 193
10.4 Integer Transfinite Diameters . . . . . . . . . . . . . . . . . . . . . . . . . 197
10.4.1 The Integer Transfinite Diameter . . . . . . . . . . . . . . . 197
10.4.2 The Monic Integer Transfinite Diameter . . . . . . . . . . 200
10.5 Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 202
10.6 Glossary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 202
11 Restricted Mahler Measure Results . . . . . . . . . . . . . . . . . . . . . . . . . 205
11.1 Monic Integer Irreducible Noncyclotomic Polynomials . . . . . . 205
11.2 Complex Polynomials That are Sums of a Bounded
Number of Monomials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 206
11.3 Some Sets of Algebraic Numbers with the Bogomolov
Property . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207
11.3.1 Totally p-Adic Fields . . . . . . . . . . . . . . . . . . . . . . . 207
11.3.2 Abelian Extensions of Q . . . . . . . . . . . . . . . . . . . . . 209
11.3.3 Langevin’s Theorem . . . . . . . . . . . . . . . . . . . . . . . . 212
11.4 The Height of Zhang and Zagier and Generalisations . . . . . . . 214
11.5 The Weil Height of a When QðaÞ=Q is Galois . . . . . . . . . . . . 215
11.6 Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 216
11.7 Glossary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 218
12 The Mahler Measure of Nonreciprocal Polynomials . . . . . . . . . . . . 219
12.1 Mahler Measure of Nonreciprocal Polynomials . . . . . . . . . . . . 219
12.1.1 The Set H of Rational Hardy Functions . . . . . . . . . . 220
12.2 Proof of Theorem 12.1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 224
12.2.1 Start of the Proof . . . . . . . . . . . . . . . . . . . . . . . . . . 224
12.2.2 The Case ‘ \2k . . . . . . . . . . . . . . . . . . . . . . . . . . . 225
12.2.3 The Case ‘  2k: Proof that MðPÞ  h0 . . . . . . . . . . 227
12.2.4 The Case ‘  2k: Existence of a d, Part 1 . . . . . . . . . 229
Contents xvii

12.2.5 The Case ‘  2k: Existence of a d, Part 2 . . . . . . . . . 231

12.2.6 The Case ‘  2k: Completion of the Proof . . . . . . . . 235
12.3 Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 236
12.4 Glossary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 236
13 Minimal Noncyclotomic Integer Symmetric Matrices . . . . . . . . . . . 237
13.1 Supersporadic Matrices and Other Sporadic Examples . . . . . . . 237
13.2 Minimal Noncyclotomic Charged Signed Graphs:
Any that Are Not Supersporadic . . . . . . . . . . . . . . . . . . . . . . 241
13.2.1 The Uncharged Case . . . . . . . . . . . . . . . . . . . . . . . . 241
13.2.2 The Charged Case . . . . . . . . . . . . . . . . . . . . . . . . . 243
13.3 Completing the Classification . . . . . . . . . . . . . . . . . . . . . . . . . 243
13.4 Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 245
13.5 Glossary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 248
14 The Method of Explicit Auxiliary Functions . . . . . . . . . . . . . . . . . . 249
14.1 Conjugate Sets of Algebraic Numbers . . . . . . . . . . . . . . . . . . 249
14.2 The Optimisation Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . 250
14.2.1 Dualising the Problem . . . . . . . . . . . . . . . . . . . . . . . 251
14.2.2 Method Outline . . . . . . . . . . . . . . . . . . . . . . . . . . . 252
14.3 The Schur–Siegel–Smyth Trace Problem . . . . . . . . . . . . . . . . 254
14.3.1 Totally Positive Algebraic Integers with Small
Mean Trace . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 255
14.4 The Mean Trace of a Less Its Least Conjugate . . . . . . . . . . . . 257
14.5 An Upper Bound Trace Problem . . . . . . . . . . . . . . . . . . . . . . 258
14.6 Mahler Measure of Totally Real Algebraic Integers . . . . . . . . . 259
14.7 Mahler Measure of Totally Real Algebraic Numbers . . . . . . . . 262
14.8 Langevin’s Theorem for Sectors . . . . . . . . . . . . . . . . . . . . . . . 263
14.8.1 Further Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . 267
14.9 Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 267
14.10 Glossary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 268
15 The Trace Problem for Integer Symmetric Matrices . . . . . . . . . . . 269
15.1 The Mean Trace of a Positive Definite Matrix . . . . . . . . . . . . 269
15.2 The Trace Problem for Integer Symmetric Matrices . . . . . . . . 270
15.3 Constructing Examples that Have Minimal Trace . . . . . . . . . . 273
15.4 Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 276
15.5 Glossary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 276
16 Small-Span Integer Symmetric Matrices . . . . . . . . . . . . . . . . . . . . . 277
16.1 Small-Span Polynomials . . . . . . . . . . . . . . . . . . . . . . . . . . . . 277
16.2 Small-Span Integer Symmetric Matrices . . . . . . . . . . . . . . . . . 280
16.3 Bounds on Entries and Degrees . . . . . . . . . . . . . . . . . . . . . . . 282
16.4 Growing Small Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . 283
16.4.1 Two Rows . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 284
xviii Contents

16.4.2 Three Rows . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 284

16.4.3 Four Rows . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 285
16.4.4 Five Rows . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 285
16.4.5 Six Rows . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 286
16.4.6 Seven Rows . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 287
16.4.7 Eight Rows . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 288
16.4.8 Nine Rows . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 290
16.4.9 Ten Rows . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 291
16.4.10 Eleven Rows . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 292
16.4.11 Twelve Rows . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 292
16.4.12 Thirteen Rows . . . . . . . . . . . . . . . . . . . . . . . . . . . . 292
16.5 Cyclotomic Small-Span Matrices . . . . . . . . . . . . . . . . . . . . . . 293
16.5.1 Examples with an Entry of Modulus Greater
Than 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 293
16.5.2 Subgraphs of the Sporadic Examples . . . . . . . . . . . . 293
16.5.3 Subgraphs of Cylindrical Tessellations . . . . . . . . . . . 295
16.5.4 Subgraphs of Toroidal Tessellations . . . . . . . . . . . . . 297
16.6 The Classification Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . 298
16.7 Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 299
16.8 Glossary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 300
17 Symmetrizable Matrices I: Introduction . . . . . . . . . . . . . . . . . . . . . 301
17.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 301
17.2 Definitions and Immediate Consequences . . . . . . . . . . . . . . . . 302
17.3 The Structure of Symmetrizable Matrices . . . . . . . . . . . . . . . . 304
17.4 The Balancing Condition and Its Consequences . . . . . . . . . . . 306
17.5 The Symmetrization Map . . . . . . . . . . . . . . . . . . . . . . . . . . . 307
17.6 Interlacing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 310
17.7 Equitable Partitions of Signed Graphs . . . . . . . . . . . . . . . . . . 310
17.8 Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 314
17.9 Glossary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 314
18 Symmetrizable Matrices II: Cyclotomic Symmetrizable
Integer Matrices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 317
18.1 Cyclotomic Symmetrizable Integer Matrices . . . . . . . . . . . . . . 317
18.2 Quotients of Signed Graphs . . . . . . . . . . . . . . . . . . . . . . . . . . 324
18.3 Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 331
18.4 Glossary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 331
19 Symmetrizable Matrices III: The Trace Problem . . . . . . . . . . . . . . 333
19.1 The Trace Problem for Symmetrizable Matrices . . . . . . . . . . . 333
19.1.1 Definitions, Notation and Statement
of the Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 333
19.1.2 Proof of Proposition 19.3 . . . . . . . . . . . . . . . . . . . . 335
19.1.3 Corollaries, Including Theorem 19.4 . . . . . . . . . . . . 336
Contents xix

19.1.4 The Structure of Minimal-Trace Examples . . . . . . . . 338

19.2 Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 341
19.3 Glossary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 341
20 Salem Numbers from Graphs and Interlacing Quotients . . . . . . . . 343
20.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 343
20.2 Salem Graphs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 343
20.3 Examples of Salem Graphs . . . . . . . . . . . . . . . . . . . . . . . . . . 346
20.3.1 Nonbipartite Examples . . . . . . . . . . . . . . . . . . . . . . 346
20.3.2 Bipartite Examples . . . . . . . . . . . . . . . . . . . . . . . . . 346
20.3.3 Finding Cyclotomic Factors . . . . . . . . . . . . . . . . . . . 348
20.4 Attaching Pendant Paths . . . . . . . . . . . . . . . . . . . . . . . . . . . . 348
20.4.1 A General Construction . . . . . . . . . . . . . . . . . . . . . . 348
20.4.2 An Application to Salem Graphs . . . . . . . . . . . . . . . 350
20.5 Interlacing Quotients . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 350
20.5.1 Rational Interlacing Quotients . . . . . . . . . . . . . . . . . 350
20.5.2 Circular Interlacing Quotients . . . . . . . . . . . . . . . . . 353
20.5.3 From CIQs to Cyclotomic RIQs . . . . . . . . . . . . . . . 353
20.5.4 Salem Numbers from Interlacing Quotients . . . . . . . 357
20.6 Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 358
20.7 Glossary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 359
21 Minimal Polynomials of Integer Symmetric Matrices . . . . . . . . . . . 361
21.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 361
21.2 Small Discriminant . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 363
21.3 Small Span . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 366
21.3.1 The Cyclotomic Case . . . . . . . . . . . . . . . . . . . . . . . 366
21.3.2 The Noncyclotomic Case . . . . . . . . . . . . . . . . . . . . 367
21.3.3 Some Counterexamples to Conjecture 21.2 . . . . . . . . 367
21.4 Small Trace . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 369
21.5 Polynomials that are Not Interlaced . . . . . . . . . . . . . . . . . . . . 369
21.6 Counterexamples for all Degrees Greater than 5 . . . . . . . . . . . 370
21.6.1 Degrees 8 to 16 . . . . . . . . . . . . . . . . . . . . . . . . . . . 370
21.6.2 Degree 20 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 371
21.6.3 Degree 19 and All Degrees Greater than 20 . . . . . . . 372
21.6.4 All Together Now . . . . . . . . . . . . . . . . . . . . . . . . . . 376
21.7 Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 376
21.8 Glossary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 376
22 Breaking Symmetry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 379

Appendix A: Algebraic Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 381

Appendix B: Combinatorial Background . . . . . . . . . . . . . . . . . . . . . . . . . 399
xx Contents

Appendix C: Tools from the Theory of Functions . . . . . . . . . . . . . . . . . . 407

Appendix D: Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 409
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 419
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 431
Chapter 1
Mahler Measures of Polynomials in One
Variable

1.1 Introduction

In this chapter, we introduce some basic results about the Mahler measure of poly-
nomials in one variable and some of its applications. In later chapters, there are
further results for one-variable Mahler measures: Dobrowolski’s Theorem (Chap. 3)
and Mahler measures for restricted sets of polynomials (Chaps. 11 and 14) and for
nonreciprocal polynomials (Chap. 12).

1.1.1 Polynomials over the Field C of Complex Numbers

Polynomials form a class of fundamental mathematical objects. By introducing an

unknown variable, say z, they represent the first step in extending the basic numerical
operations of addition and multiplication to a larger domain than fields of numbers
like the complex numbers C.
Taking a typical monic polynomial of degree d,

P(z) := z d + a1 z d−1 + · · · + ad , (1.1)

say, with coefficients a1 , , . . . , ad ∈ C, the Fundamental Theorem of Algebra (see

[DF04, 14.6]) tells us that P(z) factorises over C as

(z − α1 )(z − α2 ) · · · (z − αd ) , (1.2)

say, where α1 , α2 , . . . , αd are the zeros of P(z). This factorisation is unique. By

comparing coefficients of powers of z in (1.1) and (1.2), we can express the ai (i =
1, . . . , d) in terms of the α j ( j = 1, . . . , d). Thus, we have a bijection between Cd
and the d-element multisets of C, defined by the mapping

© Springer Nature Switzerland AG 2021 1

J. McKee and C. Smyth, Around the Unit Circle, Universitext,
https://doi.org/10.1007/978-3-030-80031-4_1
2 1 Mahler Measures of Polynomials in One Variable

(a1 , . . . , ad ) → {α1 , . . . , αd } . (1.3)

1.1.2 Polynomials over the Field Q of Rational Numbers

When we restrict this mapping (1.3) to Qd ⊂ Cd , its image is a countable set of

d-element multisets of C whose elements are called algebraic numbers. When P(z)
as in (1.1) is irreducible over the rational field Q, then {α1 , . . . , αd } is a conjugate set
of algebraic numbers. A fundamental problem is to describe these sets. For instance,
how are they distributed in the complex plane? Because the ai are all real, we have
from (1.1) and (1.2) that

d 
d
(z − α j ) = P(z) = P(z) = (z − α j ) ,
j=1 j=1

so that the multiset {α1 , . . . , αd } is stable under complex conjugation.

By dropping the requirement that P(z) be monic, we can multiply it by the lcm of
the denominators of the ai , and we can instead assume that P(z) = a0 z d + · · · + ad
has integer coefficients and content (the gcd of its coefficients ai ) equal to 1. When
a0 = 1, the zeros α j are algebraic integers. Then we know that there is a monic
polynomial with integer coefficients, call it Pα (z), which has α as a zero, and is
irreducible over Q, but, over C, splits into linear factors

d
Pα (z) = (z − α j ),
j=1

say, where α = α1 . This polynomial is the minimal polynomial of α. Denote by

Cα the set {α1 , . . . , αd }. The α j are conjugates of α, and sets of the type Cα are
conjugate sets, as defined above. By Exercise 1.2, these are indeed sets—they have
no repeated elements.
What distinguishes sets Cα from arbitrary sets of d complex numbers? Firstly,
because Pα (α j ) = Pα (α j ) = 0, we know that Cα is stable under complex conjuga-
tion. In particular, it has an even number of nonreal elements. Geometrically, it is
symmetric about the real axis. Also, we know that the integers (−1)i ai (i = 1, . . . , d)
are the elementary symmetric functions of the α j . Specifically, (−1)i ai is the sum of
 
all di possible products of i distinct conjugates α j . But the sequence (a1 , . . . , ad )
is not arbitrary, since it must provide the coefficients of the irreducible polynomial
Pα (z). For example, ad cannot be 0 for d > 1. More generally, Pα (z) can have no
irreducible factor of degree d  ≤ d/2 .

Exercise 1.1 Prove that the minimal polynomial of an algebraic number is unique
and has no multiple zeros.

Exercise 1.2 Prove that two conjugate sets either coincide or are disjoint and that
the field of algebraic numbers Q is a countable disjoint union of conjugate sets.
1.1 Introduction 3

Next, note that either Pα (z) = z or ad is a nonzero integer, implying that

|α1 · · · αd | = |ad | ≥ 1 .

This shows that z is the only such polynomial having all its zeros in the open disc
|z| < 1. Furthermore, if |ad | = 1 and all α j are in the closed unit disc |z| ≤ 1, then
they are actually all on the unit circle |z| = 1. For the closed disc |z| ≤ 1, moreover,
we have the first of Kronecker’s two theorems, to which we now turn.

1.2 Kronecker’s Two Theorems

Theorem 1.3 (Kronecker’s First Theorem [Kro57]) Suppose that α is a nonzero

algebraic integer that lies, with its conjugates, in the unit disc |z| ≤ 1. Then α is a
root of unity.

Proof Suppose that α has degree d, with minimal polynomial

Pα (z) = z d + a1 z d−1 + · · · + ad ,

say. Then, because (−1)k ak is the sum of all possible distinct  k-tuples of the zeros
of P, all of which have modulus at most 1, we have |ak | ≤ dk (k = 0, . . . , d). The
same bounds apply to the coefficients of the polynomial Pr whose zeros are the r th
powers of those of P. Because by Proposition A.22 αr is also an algebraic integer,
the coefficients of Pr are integers, so there are only finitely many possibilities for the
coefficients of these polynomials. Hence, there are finitely many possibilities for all
the zeros of all the Pr , and so in particular, there are only finitely many possibilities
for the αr . Therefore, two of them must be equal, say αr = α s with r < s. Then
α s−r = 1, and we see that α is a root of unity. 

Theorem 1.4 (Kronecker’s Second Theorem [Kro57]) Suppose that β is a real alge-
braic integer that lies, with its conjugates, in the interval [−2, 2]. Then β is conjugate
to 2 cos(2π/) for some positive integer .

Proof Let Q(x) be the minimal polynomial of β, of degree d  say. Then the poly-

nomial z d Q(z + 1/z) has all its zeros on |z| = 1, so, by the First Theorem, these
are all roots of unity. Hence, β = ω + ω−1 for some primitive th root of unity
ω . By Proposition 5.2, ω is conjugate to e2πi/ , showing that β is conjugate to
2 cos(2π/). 

Kronecker’s theorems suggest that we should try to describe all ‘small’ (suitably
defined) algebraic integers that are not covered by his theorems.
One way to do this is using the Mahler measure.
The Mahler measure M(P) of a polynomial P(z) = a0 z d + · · · + a0 ∈ Z[z] with
a0 = 0, and zeros α1 , …, αd ∈ C, is defined as
4 1 Mahler Measures of Polynomials in One Variable


d
M(P) := |a0 | max(1, |αi |) .
i=1

So M(P) measures, in a single value, how far outside the unit circle the zeros of
P are. Thus, when a0 = 1, the Mahler measure M(P) is the product of the moduli
of those zeros of P that lie in the region |z| > 1 (empty products being 1). Clearly
M(P) ≥ |a0 | ≥ 1. The definition extends immediately to any Laurent polynomial
(negative powers of z allowed). Such polynomials can be written as z k P(z), where
P is a regular polynomial as in (1.1), and k is any (possibly negative) integer. Then
M(z k P(z)) is defined simply as M(P).
The nth cyclotomic polynomial n (z) is defined as the minimal polynomial of
all primitive nth roots of unity. (From Proposition 5.2(b), we know that they all
have the same minimal polynomial.) These polynomials are discussed in Chap. 5.
Throughout the book, we use the term cyclotomic polynomial to refer to any monic
integer polynomial whose zeros are all roots of unity. Thus, such polynomials are
products of polynomials n (z) for a multiset of values of n.
The following is almost immediate.
Exercise 1.5 Show that, for a nonzero algebraic integer α, the mean modulus of its
conjugates is at least 1, with equality if and only if α is a root of unity.
The nth Chebyshev polynomial (of the first kind) Tn (x), though usually defined
so that all its zeros are in [−1, 1], is more naturally defined so that they lie in [−2, 2].
So for n = 1, 2, 3, …we define Tn (x) by Tn (z + z −1 ) := z n + z −n , and then Tn (x)
is monic with integer coefficients. Its zeros are clearly at 2 cos((2 j − 1)π/n) ( j =
1, . . . , n). The following formula gives a convenient way of computing Tn (x)—for
a discussion of the resultant, see Appendix A.

Exercise 1.6 Show that the resultant resz (z 2 − x z + 1, z 2n − yz n + 1) is equal to

(y − Tn (x))2 . Thus, Tn (x) can be obtained from the coefficient of y in this expression.

Exercise 1.7 The nth Chebyshev polynomial of the second kind Un (x) is defined
by Un (z + z −1 ) := (z n − z −n )/(z − z −1 ). Find the zeros of Un (x) and show that

resz (z 2 − x z + 1, z 2n − y(z n+1 − z n−1 ) − 1) = (y − Un (x))2 (4 − x 2 ) .

In number theory, Mahler measure arose first in a paper of Lehmer [Leh33] in

1933, as a way of estimating the growth rate of integer sequences defined by a linear
recurrence. In this paper, he noted that the smallest Mahler measure M(P) > 1 for
an integer polynomial P that he could find was for P(z) = L(z), where

L(z) := z 10 + z 9 − z 7 − z 6 − z 5 − z 4 − z 3 + z + 1 ,

for which M(L) = 1.17628082 · · · , known as Lehmer’s number. He asked whether

such M(P) were always at least c, for some absolute constant c > 1. To this day,
M(L) is the smallest M(P) > 1 that has been found, but his question has not been
1.2 Kronecker’s Two Theorems 5

answered. That the answer to this question is ‘yes’ is now often called ‘Lehmer’s
Conjecture’. That the answer is ‘yes’ and with c = M(L) is the Strong Lehmer Con-
jecture. If true, the Strong Lehmer Conjecture would be an extension of Kronecker’s
First Theorem, as it would show that any irreducible monic integer polynomial P(z)
with M(P) < M(L) must be either z or a cyclotomic polynomial.
The best unconditional result in the direction of a proof of Lehmer’s Conjecture
is due to Dobrowolski [Dob79], who proved that if M(P) > 1, then for any ε > 0
there is a d0 (ε) so that
 3
log log d
M(P) > 1 + (1 − ε) (1.4)
log d

for all P of degree d ≥ d0 (ε). A little later, Cantor and Straus [CS83] proved a
version of this inequality with (1 − ε) replaced by 1/1200, but valid for all d ≥ 2.
Later, Voutier [Vou96], improved the constant 1/1200 to 1/4, again valid for d ≥ 2.
A version of Dobrowolski’s result is proved in Chap. 3, based on the method of Cantor
and Straus.
In any study of Lehmer’s question, we of course need to make use of the fact
that P is not cyclotomic, so that the zeros αi of Pare not roots of unity. Thus, we
d
know that, for all positive integers n, the product i=1 (αin − 1), being nonzero and
a symmetric function of the αi , is a nonzero integer. Using only this arithmetical
information, Blanksby and Montgomery [BM71] and later Stewart [Ste78b] used
this fact (and very different methods!) to prove for some positive constant c and
d ≥ 2 that
c
M(P) > 1 + .
d log d

In their proof, Blanksby and Montgomery used the Fourier analysis, while Stewart,
borrowing an idea from transcendence proofs, constructed an auxiliary function.
Dobrowolski, in the proof of (1.4), made use of the slightly deeper fact that, for all
d d p
j=1 (αi − α j ) is a nonzero integer multiple of p .
d
primes p, we have that i=1
In Voutier’s proof, he uses some additional arithmetic information concerning the
modulus of integer polynomial discriminants.
If P is not self-reciprocal (see Sect. A.1), then it is known from Theorem 12.1 that
M(P) ≥ M(z 3 − z − 1) = 1.3247 . . . , so, when addressing Lehmer’s question, we
can assume that P(z) is self-reciprocal.
There are several ways of coming up with the polynomial L(z). One way (see
Exercise 1.8 below for another way) is to note that it comes from the graph E 10 :

We take its adjacency matrix A = (ai j ) which has 1 in the (i, j)th place if vertices
labelled i and j are joined by an edge, and 0 otherwise. Then
6 1 Mahler Measures of Polynomials in One Variable
√ √ 
L(z) = z 5 det ( z + 1/ z)I − A .

Exercise 1.8 Show that P j (z) := (z 3 − z − 1)z n + (z 3 + z 2 − 1) has at most one

zero, αn say, in |z| > 1. Show that M(Pn (z)) M(z 3 − z − 1) = 1.3247 . . . as
n → ∞. Show too that Pn (z) is a cyclotomic polynomial for n = 1, . . . , 7, but that
for j = 8 it equals (z − 1)L(z).
It is easy to see that the Mahler measure of a polynomial with integer coefficients
is an algebraic number. In fact, this algebraic number must be an algebraic integer.
Proposition 1.9 The Mahler measure M(P) of a polynomial with integer coeffi-
cients in one variable is an algebraic integer.
Proof For each prime p, consider an algebraic closure Q p of Q p , armed with a
valuation that extends the usual p-adic valuation on Q p —see A.3. Write


d
P(z) = a0 z d + · · · + ad = a0 (z − αi ) ,
j=1


so that M(P) = ±a0 m 1 say, where m 1 = |α j |≥1 α j = α1 . . . αk say. Now let k p
be the number of α j that are in |z| p > 1 in Q p . Then,  by considering ±ak p /a0
as a sum of all products of k p of the α j , we see that |α j | p >1 |α j | p = |ak p /a0 | p ,

giving |a0 | p |α j | p >1 |α j | p = |ak p | p . Hence, |a0 | p times the product of any number
of the |α j | p (for distinct α j ) is at most |ak p | p ≤ 1. In particular, |M(P)| p ≤ 1 and
|M(P)∗ | p ≤ 1 for all conjugates M(P)∗ of M(P). Hence, by Exercise A.19, M(P)
is an algebraic integer. 
An algebraic integer β is called a Perron number if it is real and positive, and all
its conjugates except itself (if there are any) have modulus less than β.
Exercise 1.10 Show that the Mahler measure M(P) of an integer polynomial P is
a Perron number.
The Mahler measure M(P) has an integral representation via Jensen’s Theorem
(see Exercise 1.11 below) as
 1
M(P) = exp log |P(e2πit )| dt . (1.5)
0

This is the geometric mean of |P(z)| for z on the unit circle.

Exercise 1.11 Prove Jensen’s Theorem: that for any complex number a,

1
log |a| if |a| > 1;
log |e2πit − a| dt =
0 0 otherwise.

Deduce (1.5) from Jensen’s Theorem.

1.3 Mahler Measure Inequalities 7

1.3 Mahler Measure Inequalities

The inequalities in this section concern real polynomials in general—their coeffi-

cients need not be integers.
Proposition 1.12 (Gonçalves’ Inequality) Let P(z) = a0 z d + · · · + ad be a poly-
nomial with real coefficients, and with a0 ad = 0. Then

d
a02 ad2
M(P)2 + ≤ a 2j .
M(P)2 j=0

d
Proof Factorise P over C as P(z) = a0 i=1 (z + αi ). Then
 
P(z) = a0 (z + αi ) · (1 + αi z) · f (z) ,
|αi |≥1 |αi |<1

where
 z + αi
f (z) := .
|α |<1
1 + αi z
i

 
Also, put b := |αi |<1 αi and B := |αi |≥1 αi , so that a0 b = ad /B. Then P(z) =
a0 b f (z)Q(z), where
 
Q(z) := (z + αi ) · (z + 1/αi )
|αi |≥1 |αi |<1

= z d + · · · + B/b
a0 B 2
= zd + · · · + .
ad

Hence,

d 1
2
a 2j = P(e2πit ) dt
j=0 0

1
ad 2
= f (e2πit )Q(e2πit ) dt
0 B
1
ad2 2
= Q(e2πit ) dt (as | f (z)| = 1 on |z| = 1)
B2
0
 2 
a2 a0 B 2
≥ d2 1 + 2
B ad
a02 ad2
= + M(P)2 ,
M(P)2
8 1 Mahler Measures of Polynomials in One Variable

since M(P) = |a0 |B. 

The following weaker inequality suffices for some applications.

Corollary 1.13 (Specht’s Inequality) For P(z) = a0 z d + · · · + ad ∈ R[z],



 d
M(P) ≤  a 2j .
j=0

Exercise 1.14 For which polynomials P does equality in Gonçalves’ Inequality

occur?

Exercise 1.15 Modify Gonçalves’ Inequality and its proof to cover polynomials
with complex coefficients.

Lemma 1.16 (Gauss–Lucas) For any polynomial P(z) ∈ C[z], the zeros of its
derivative P  (z) lie in the convex hull of the zeros of P.
d
Proof Writing P(z) = i=1 (z − αi ), we have the well-known partial fraction expan-
sion
d
P  (z) 1
= . (1.6)
P(z) i=1
z − αi

Thus, for any zero β of P  , not a zero of P, it follows that

d d
1 1
=0= ,
i=1
β − αi i=1
β − αi

giving
d
β − αi
= 0,
i=1
|β − αi |2

d
so that β = i=1 wi αi for weights

1 1
wi = ,
W |β − αi |2

where
d
1
W = .
j=1
|β − α j |2

Of course, the result is trivially true if β is also a zero of P. 

1.3 Mahler Measure Inequalities 9

Proposition 1.17 Let P(z) be a real monic polynomial of degree d, with Mahler
measure M. Let its ‘monic derivative’ d1 P  (z) have Mahler measure M  . Then M  ≤
M.
Proof From (1.6), we have

1 d
1 1
log M  − log M = log dt .
0 d i=1
e2πit − αi

Now the right-hand side, considered as a function of one αi , with the others fixed,
is a subharmonic function of αi in the open set C \ {z : |z| = 1} (see Appendix C).
Hence, its maximum is on its boundary |z| = 1. Doing this for each αi shows that
log M  − log M has its maximum when all the αi are on |z| = 1. But then M = 1
and, by Lemma 1.16, M  = 1 too. So M  /M ≤ 1 for all polynomials P. 
A related result is the following.
Proposition 1.18 If P(z) is a real polynomial of degree d and self-reciprocal with
Mahler measure M(P) > 1, then also M( d1 P  ) is greater than 1.
This is an immediate consequence of Theorem A.8.
Exercise 1.19 For a polynomial P(z) = a0 z d + a1 z d−1 + · · · + ad ∈ C[z], show
that for j = 1, . . . , d we have
 
d
|a j | ≤ M(P) .
j

Exercise 1.20 Show that if the polynomial Q, with leading coefficient q0 , is a factor
of the polynomial P, with leading coefficient p0 , then M(Q) ≤ |q0 / p0 |M(P).

1.4 A Lower Bound for an Integer Polynomial Evaluated at

an Algebraic Number

Theorem 1.21 Given a polynomial Q(z) ∈ Z[z] of degree e and an algebraic num-
ber α of degree d, then Q(α) = 0 or

max(1, |α|)e
|Q(α)| ≥ . (1.7)
L d−1 M(α)e

Here, L is the length of Q (the sum of the moduli of its coefficients).

Proof Let α = α1 have minimal polynomial P with leading coefficient p0 and con-
jugates α j ( j = 1, . . . , d). Then from (A.2), we have
10 1 Mahler Measures of Polynomials in One Variable

1 ≤ | resz (P, Q)| = p0e |Q(α)| Q(α j ) .
j>1

Also, we clearly have

|Q(α j )| ≤ L max(1, |α j |)e ,

so that
 |Q(α)|L d−1 M(α)e
1 ≤ p0e |Q(α)|L d−1 max(1, |α j |)e = ,
j>1
max(1, |α|)e

using the definition of M(α). Hence, (1.7) holds. 

When Q is linear, we have the following.

Corollary 1.22 For any rational number p/q with q > 0 and algebraic number
α = p/q of degree d, we have

p max(1, |α|)
α− ≥ .
q q(| p| + q)d−1 M(α)

In particular,
max(1, |α|)
|α − 1| ≥ .
2d−1 M(α)

1.5 Polynomials with Small Coefficients

In this section, we consider integer polynomials as factors of integer polynomials

with small coefficients. The key connection with the Mahler measure is the following
theorem.
Theorem 1.23 Let P(z) ∈ Z[z] be monic. Then P(z) has an integer polynomial
multiple P(z)S(z) say such that all coefficients of P(z)S(z) are at most M(P) in
modulus.
Very often S(z) can be chosen so that M(S) = 1. However, this is not guaranteed
by the theorem, so that maybe M(P S) > M(P). Also, it is not claimed that in general
P S is monic (although, of course, if M(P) < 2, then it must be).

Proof Let α1 , α2 , . . . , αd be the zeros of P(z), where d is the degree of P. Note

that its zeros need not be conjugate or distinct. However, we first do the case
when the αi are assumed to be distinct. We label the α j so that α1 , . . . , α2s are
nonreal with α2 j  −1 = α2 j  for j  = 1, . . . , s, and α2s+1 , . . . , αd are all real. Put
Y := 1 + M(P) . We need two integer parameters N , L and further integers
yi ∈ {0, 1, . . . , Y − 1} (i = 1, . . . , N ), all to be specified later.
We consider the vector v = (v1 , . . . , vd ) ∈ Rd defined as
1.5 Polynomials with Small Coefficients 11
  N
  N
  N
  N

v := Re α2i yi , Im α2i yi , Re α4i yi , Im α4i yi ,...,
i=1 i=1 i=1 i=1
 N
  N
 N N N

Re α2s
i
yi , Im α2s
i
yi , α2s+1
i
yi , α2s+2
i
yi , . . . , αdi yi .
i=1 i=1 i=1 i=1 i=1

Now v j ∈ [−B j , B j ], where

B j ≤ N (Y − 1)M jN ( j = 1, . . . , d) , (1.8)

with M j := max(1, |α j |). Next, for j = 1, . . . , d divide the interval [−B j , B j ] into

L M jN intervals of equal length, and hence, divide the box dj=1 [−B j , B j ] ⊆ Rd
d
into smaller boxes. Since j=1 M j = M(P), this gives us at most L d M(P) N boxes,
together containing all the possible vectors v, with edge lengths at most

M jN 2N (Y − 1)
2N (Y − 1) < ( j = 1, . . . , d) . (1.9)
L M jN L −1
√
These lengths are less than 1/ 2 provided that
√
L ≥ 1 + 2 2N (Y − 1) . (1.10)

Now there are Y N possible vectors (y1 , . . . , y N ), so if Y N > L d M(P) N then, by the
Pigeonhole Principle, two vectors v1 , v2 , say, defined by distinct vectors (y1i ), (y2i )
say, must lie in the same box. Thus, it suffices to take
  N /d
√ Y
1 + 2 2N (Y − 1) ≤ L < . (1.11)
M(P)

This can clearly be achieved by taking N sufficiently large—see Exercise 1.25

below for an estimate for such an N√. Hence, the difference v := v1 − v2 has all
components of modulus less than 1/ 2. It is defined by the nonzero vector (y1i ) −
(y2i ) =: (yi ) say, which all lie in the set {−(Y − 1), . . . , −1, 0, 1, . . . , Y − 1}. Then
for
v = (v1 , v2 , . . . , v2s−1
 
, v2s 
, v2s+1 , . . . , vd ),

we certainly have

v22j  −1 + v22j < 1 ( j  = 1, . . . , s) and |vj | < 1, ( j = 2s + 1, . . . , d) ,

N i  N i 
so that | i=1 α j yi | < 1 ( j = 1, . . . , d). Thus, for any α j , we have | i=1 α yi | < 1
N i 
for every conjugate α of α j . Hence, every i=1 α j yi has norm less than 1, and so
12 1 Mahler Measures of Polynomials in One Variable

N  i
must be 0. Thus, all α j are zeros of the nonzero polynomial i=1 yi z , which is an
integer polynomial multiple of P, given the α j are distinct.
Finally, we outline the modifications necessary to the proof when some of the α j
are equal. If α1 = α2 = · · · = αk = α say are zeros of P(z), then P(z) and its k  th

 Nk =i 1, 2, . . . , k − 1 are all zero at z = α. So we need to replace the
derivatives for
k copies of i=1 α yi appearing in v by

N N   N   N  
i i−1 i i−2 i
α yi ,
i
α yi , α yi , . . . , α i−k+1 yi .
i=1 i=1
1 i=1
2 i=1
k − 1

Since k ≤ d, we can bound the modulus of each of these sums simply if rather crudely
by N d (Y − 1) max(1, |α|) N . The effect of this is that we need only replace the ‘N ’
appearing in (1.8), (1.9) (two places), (1.10) and the left-hand side of (1.11) by N d .
The argument is otherwise unaffected and adapts similarly if there are further sets of
equal zeros. 
Exercise 1.24 The theorem tells us that for every n the polynomial (1 + z)n is a
factor of a polynomial with coefficients in {−1, 0, 1}. Find such a polynomial.
Exercise 1.25 Find an estimate for N in (1.11).

1.6 Separation of Conjugates

In this section, we give a lower bound for the minimum distance between conjugates
of an algebraic integer. We also give a lower bound for the modulus of the derivative
of the minimal polynomial of an algebraic integer, evaluated at its conjugates. Both
of these bounds involve the Mahler measure.
Let α be an algebraic integer with conjugates α = α1 , α2 , . . . , αd . Define the
separation of α, Sep(α), by

Sep(α) = min |αi − α j | .

i=j

Also, its discriminant (α) is the square of the Vandermonde determinant

j )i, j=1,...,d . Note that
det(α i−1 (α) is not always the same as the field discrimi-
( j)
nant (Q(α)) of Q(α), which is the square of the determinant of the matrix (αi ),
where α (1) , α (2) , . . . , α (d) is a Z-basis for the ring of integers of Q(α), and where,
( j)
with σ1 , . . . , σd being the embeddings of Q(α) into C, we have αi = σi (α ( j) ).
Exercise 1.26 Prove that (α) and (Q(α)) are both integers and that (Q(α))
divides (α).
The next lemma will immediately be applied to bound the separation of an alge-
braic integer from below, but it holds more generally for any d-tuple of complex
1.6 Separation of Conjugates 13

numbers. The proof of this lemma appeals to Hadamard’s Inequality, which we

prove later as Lemma 3.2. For z = (z 1 , . . . , z d ) ∈ Cd , we define

M(z) := M((z − z 1 ) · · · (z − z d )) .

Lemma 1.27 For any z = (z 1 , . . . , z d ) ∈ Cd , we have

√
|z 1 − z 2 | ≥ 3d −(d+2)/2 M(z)1−d | det((z i−1
j )i, j=1,...,d )| .

Proof We can assume that |z 2 | ≤ |z 1 |. Let V = (z i−1

j )i, j=1,...,d . Then

2
0 1 1 ··· 1
1 z2 z3 ··· zd
z1 + z2 z 22 z 32 ··· z d2
| det(V )|2 = |z 1 − z 2 |2 z 12 + z 1 z 2 + z 22 z 23 z 33 ··· z d3
.. .. .
. . · · · ..
z 1d−2 + z 1d−3 z 2 + · · · + z 2d−2 z 2d−1 z 3d−1 · · ·
z dd−1
 d−1  d  d−1 

≤ |z 1 − z 2 |2 |i z 1i−1 |2 |z j |2i by Lemma 3.2
i=1 j=2 i=0
 d−1 

d
≤ |z 1 − z 2 |2 i 2 max(1, |z 1 |)2(d−2) max(1, |z j |)2(d−1)
i=1 j=2
3
d d−1
≤ |z 1 − z 2 |2 d M(z)2(d−1) ,
3
d−1
using i=1 i 2 = 13 (d − 1)(d − 21 )d. This gives the required lower bound for |z 1 −
z 2 |. 
We now apply this to the minimal polynomial of an algebraic integer α. For an
algebraic number α, we write M(α) for the Mahler measure M(Pα ), where Pα (z) ∈
Z[z] is the minimal polynomial of α.
Corollary 1.28 (Mahler [Mah64]) Let α be an algebraic integer of degree d. We
have √
Sep(α) > 3d −(d+2)/2 | (α)|1/2 M(α)−(d−1) .

Lemma 1.29 (Mahler [Mah64]) For any z = (z 1 , . . . , z d ) ∈ Cd , define the polyno-


mial P(z) := (z − z 1 ) · · · (z − z d ), so that P (z i ) = j =i (z i − z j ) (i = 1, . . . , d).
Then
⎧
⎨ M(z)−(d−1) if |z i | ≤ 1;
|P  (z i )| ≥ 3(d−1)/2 d −(d +d−1)/2 | det(V )| ×
2

⎩ z −(d−1)2 /2 if |z i | > 1.
14 1 Mahler Measures of Polynomials in One Variable

Here z := maxi=1,...,d |z i |.

Proof By standard column operations, we have

| det(V )|2 = |P  (z i )|2 |det (c1 · · · cd )|2 ,

where c1 = (1, z 1 , . . . , z 1d−1 )T while for j = 2, . . . , d,

c j = (0, 1, z 1 + z j , z 12 + z 1 z j + z 2j , . . . , z 1d−2 + z 1d−3 z j + · · · + z 1 z d−3

j + z d−2
j )T .

Thus, by Hadamard’s Inequality (Lemma 3.2),

 d−1 
d d+2
d
 d−2
|det (c1 · · · cd )| ≤ d max(1, |z 1 |)
2 d−1
max |z 1 |, |z j |
3 j=1
⎧
⎨ M(z) d−1
if |z i | ≤ 1;
≤ 3−(d−1) d (d +d−1)
2
×
⎩ z (d−1)2 if |z i | > 1,

which gives the required inequality. 

Corollary 1.30 For an algebraic integer α with conjugates α = α1 , α2 , . . . , αd and

minimal polynomial Pα (z), we have

d   −(d−1)
min |Pα (αi )| ≥ 3(d−1)/2 d −(d +d−1)/2
| (α)|1/2 max M(α), α (d−1)/2
2
.
i=1

1.7 The Shortness of a Polynomial

The length of a polynomial P(z) ∈ Z[z] is defined as usual to be the sum of the
absolute values of its coefficients. We define a short polynomial for P to be the
minimum-length product of P with a cyclotomic polynomial. We call this length the
shortness of P, denoted by sh(P). Note that a polynomial and its short polynomial
have the same Mahler measure. For an algebraic number α, we define the shortness
of α to be the shortness of its minimal polynomial. A nonreciprocal polynomial
clearly has shortness of at least 3, while a reciprocal noncyclotomic polynomial has
shortness of at least 5, by Exercise 1.34 below.

Exercise 1.31 Show that 0 has shortness 1, and all roots of unity have shortness 2.

Exercise 1.32 Show that the shortness of (3 + 4i)/5 is 12.

Exercise 1.33 For an algebraic number α, show that its shortness is at least M(α)
(the smallest integer not less than M(α)).
1.7 The Shortness of a Polynomial 15

Exercise 1.34 Show that all nonconstant reciprocal noncyclotomic integer polyno-
mials have shortness at least 5.

Exercise 1.35 Is a short polynomial for an algebraic number α always unique?

Exercise 1.36 Show that the shortness of (z − 1)k is even, at most 2k , and is 6 for
k = 3.

Exercise 1.37 Let P be a cyclotomic polynomial and let k be the largest multiplicity
of any zero of P. Show that sh(P) ≤ sh((z − 1)k ).

Problem 1.38 (open problem)

Find the shortness of (z − 1)k for all k.

While algebraic numbers are usually specified by their minimal polynomials,

there are some advantages in using short polynomials to specify them. Allowing
these extra cyclotomic factors, whose Mahler measures are of course all 1, means
that the length of the polynomial can be kept small, without changing the Mahler
measure. For instance, the irreducible polynomial

Q(z)
= z 112 + z 111 + z 110 + z 109 − z 106 − 2z 105 − z 104 − z 103 + z 101 + z 100 + 2z 99 + 2z 98
+ z 97 − z 95 − 2z 94 − 2z 93 − 2z 92 − 2z 91 + z 89 + 2z 88 + 3z 87 + 2z 86 + 2z 85 + z 84 − z 83
− 2z 82 − 3z 81 − 3z 80 − 2z 79 − z 78 + 2z 76 + 3z 75 + 3z 74 + 3z 73 + z 72 − z 70 − 3z 69
− 3z 68 − 3z 67 − 2z 66 + z 64 + 2z 63 + 3z 62 + 3z 61 + 2z 60 − 2z 58 − 3z 57 − 3z 56 − 3z 55
− 2z 54 + 2z 52 + 3z 51 + 3z 50 + 2z 49 + z 48 − 2z 46 − 3z 45 − 3z 44 − 3z 43 − z 42 + z 40
+ 3z 39 + 3z 38 + 3z 37 + 2z 36 − z 34 − 2z 33 − 3z 32 − 3z 31 − 2z 30 − z 29 + z 28 + 2z 27
+ 2z 26 + 3z 25 + 2z 24 + z 23 − 2z 21 − 2z 20 − 2z 19 − 2z 18 − z 17 + z 15 + 2z 14 + 2z 13
+ z 12 + z 11 − z 9 − z 8 − 2z 7 − z 6 + z 3 + z 2 + z + 1

has length 179 and Mahler measure 1.24846635. However, its shortness is only 6,
with short polynomial

P(z) = z 130 − z 114 − z 77 + z 53 + z 16 − 1 .

This is #170 in Table D.1 of Appendix D. The polynomial Q(z) is the noncy-
clotomic irreducible factor of P(z). In Chap. 5, we shall discover how to iden-
tify cyclotomic factors of polynomials, and in particular from Exercise 5.7, we
can see that, because P(z) has no factors that are polynomials in z 4 , we have
Q(z) := P(z)/ gcd(P(z), P(−z)P(z 2 )P(−z 2 )). For a discussion of a connection
between these short polynomials and related polynomials in 2 or 3 variables, see
Sect. 2.6.1.
There are 236 known integer polynomials with Mahler measures less than 1.25,
as shown in Table D.1. Computation shows that 15 of these are Mahler measures
16 1 Mahler Measures of Polynomials in One Variable

of short reciprocal pentanomials, while 214 are Mahler measures of polynomials

of reciprocal hexanomials. Of the remaining seven, three are Mahler measures of
heptanomials, three are Mahler measures of octanomials, while one is the Mahler
measure of a decanomial. These polynomials, when not irreducible, are the product
of a cyclotomic polynomial and a polynomial whose Mahler measure is the given
value. All these values are in (1, 1.25) and so are less than M(Q 1 ) = 1.255 · · · , the
smallest Mahler measure in Table D.2 in Appendix D. It is expected that there are
infinitely many Mahler measures less than M(Q 1 ). For more on the measures in that
table, see Sect. 2.6.2.

1.7.1 Finding Short Polynomials

Currently there is no known algorithm to compute the shortness of a polynomial and

its associated short polynomial. However, we can, at least in principle, compute a
related polynomial that very often, indeed maybe always, turns out to be a short one.
Given a polynomial P(z) ∈ Z[z], we define a minimum-length multiple of P to
be a nonzero polynomial of the form P(z)T (z), where T (z) ∈ Z[z], and the length
of P T is as small as possible among all polynomials of that form. We call the length
of such a P T the minimum length of P. Note that if T turns out to be cyclotomic,
then P T is a short polynomial for P, and the minimum length of P is actually
the shortness of P. But possibly T could be noncyclotomic, so that we would have
M(P T ) > M(P). Indeed, we cannot even rule out the possibility that T could be
nonmonic.
We now describe an algorithm to find a minimum-length multiple of P in the
case where P is monic, noncyclotomic and irreducible. We do not claim that the
algorithm is an efficient one, merely that it is effective.
Let  P be the length of P, and min its minimum length. We know that 3 ≤ min ≤
 P . We assume that  ≥ 3 and that we have shown that min ≥ . Our aim is to
find out whether or not there is an integer polynomial T with P T of length  and
given leading coefficient v. If there is such a T , then min =  and we are finished.
Otherwise we increase v by 1, if possible, or else increase  by 1 and reset v to 1.
Because P is monic and noncyclotomic, we know from Theorem 1.3 that P has a
zero in |z| > 1: call it α, say. We can choose α so that |α| = α , the largest modulus
of any conjugate of α (the house of α). Using α , we can get tighter bounds for 
and v.

Lemma 1.39 If the polynomial P T has length  and leading coefficient v, then


≥1+ α and v≤ . (1.12)
1+ α

Proof For some positive integer v, we can write P T as

1.7 The Shortness of a Polynomial 17

vz n ± z n−d1 ± z n−d1 −d2 · · · ± z n−d1 −d2 −···−d−v−1 ± 1 , (1.13)

where d1 > 0 and d2 , d3 , . . . , d−v−1 ≥ 0. Also, define d−v ≥ 0 by

n = d1 + d2 + · · · + d−v−1 + d−v . (1.14)

To give a coefficient of modulus greater than 1 (other than the leading coefficient),
some di must be 0, and the signs of ±z n−d1 −d2 −···−di−1 and ±z n−d1 −d2 −···−di must be
the same, to disallow cancellation. Then

| ± α n−d1 ± α n−d1 −d2 · · · ± α n−d1 −d2 −···−d−v−1 ± 1| < ( − v)|α|n−d1 ,

which is less than v|α|n if d1 > log(/v−1)

log |α|
. If log(/v−1)
log |α|
< 1 then, since d1 ≥ 1, we have
a contradiction. Hence, /v ≥ 1 + |α| = 1 + α , which gives the claimed inequali-
ties. 

Now assume that P T satisfies (1.12). Then from the proof of Lemma 1.39, we
have   
log(/v − 1)
d1 ∈ 1, 2, . . . .
log |α|

For such a d1 , we have

| ± α n−d1 −d2 ± α n−d1 −d2 −d3 · · · ± α n−d1 −d2 −···−d−v−1 ± 1| < ( − v − 1)|α|n−d1 −d2
< |vα n ± α n−d1 |

−v−1
if |α|d2 > α d1 ±1
. Hence,
  
log( − v − 1) − log |vα d1 ± 1|
d2 ∈ 0, 1, 2, . . . , ,
log |α|

where the range of values for d2 depends on the choice of the ± sign. Continuing in
this way, we have, for i = 2, , . . . , ,  − v, that

| ± α n−d1 −d2 −···−di+1 ± α n−d1 −d2 −···−di+2 · · · ± α n−d1 −d2 −···−d−v−1 ± 1|

< ( − v − i)|α|n−d1 −d2 −···−di+1 ,

which is less than

|vα n ± α n−d1 ± α n−d1 −d2 − · · · − α n−d1 −d2 −···−di |

if
−v−i
|α|di+1 > .
|vα d1 +···+di ± α d2 +···+di ± · · · ± α di ± 1|
18 1 Mahler Measures of Polynomials in One Variable

Hence, we obtain, for i = 1, 2, 3, . . . ,  − v − 1, that

  
log( − v − i) − log |vα d1 +···+di ± α d2 +···+di ± · · · ± α di ± 1|
di+1 ∈ 0, 1, 2, . . . .
log |α|

Note that the term vα d1 +···+di ± α d2 +···+di ± · · · ± α di ± 1 has length i + v <  so,
by assumption, cannot be zero. Thus, for i = 1, . . . ,  − v, each di is bounded in
terms of d1 , . . . , di−1 . If a polynomial of the form (1.13) is found that is 0 at z = α
then, because P is irreducible, the polynomial will be divisible by P, and so be of the
form P T . Then n is given by (1.14). Thus, the search is a finite one. It will eventually
succeed in finding P T of minimum length, as it will find P itself as a last resort.
This completes the algorithm description.
If T turns out to be cyclotomic, then we have found a short polynomial P T for
P.

Problem 1.40 Find a monic irreducible integer polynomial P with minimum-length

polynomial P T where T is not cyclotomic. Alternatively, prove that no such P exists.

It seems to be a challenge to extend the algorithm above to nonmonic and reducible

polynomials, including those having repeated irreducible factors. Perhaps the most
tractable such case could be when P is nonmonic and irreducible. By Corollary A.17,
if P has no zero in |z| > 1 then, for some prime p, it has a zero α in Q p with |α| p > 1.

Problem 1.41 Modify the above algorithm so that it works when P is nonmonic
and irreducible and has all its complex zeros in the disc |z| ≤ 1.

1.8 Variants of Mahler Measure

As before, for an algebraic number α, we write M(α) for the Mahler measure M(Pα ),
where Pα (z) ∈ Z[z] is the minimal polynomial of α. Three variants of Mahler mea-
sure are used in the literature:
• The Weil height
log M(α)
h(α) := log M(α) = . (1.15)
deg α

• The absolute Mahler measure M(α) := M(α)1/ deg α . Also called the exponential
Weil height.
• The logarithmic Mahler measure m(P) := log M(P) is often used for polyno-
mials P in several variables.
1.8 Variants of Mahler Measure 19

1.8.1 The Weil Height

The Mahler measure results for algebraic numbers α have an alternative formulation
using the (absolute logarithmic) Weil height h(α) defined in (1.15). In particular, the
following are easily proved.

Exercise 1.42 Show that for an algebraic number α, we have

• h(α) = 0 if and only if α is 0 or a root of unity;
• h(α −1 ) = h(α);
• h(α + 1) ≤ h(α) + log 2, with equality if and only if α = 1;
• h(α k ) = kh(α) for all positive integers k;
• h(ωα) = h(α) for any root of unity ω.

There is also a nice inequality for the Weil height, as follows.

Lemma 1.43 For any two algebraic numbers α,β, we have

h(αβ) ≤ h(α) + h(β).

Before giving its proof, we need a preliminary lemma, with corollaries. Let us
extend the p-adic valuation on Q p to an algebraic closure Q p of Q p .

Lemma 1.44 For a given prime p, suppose that


d
a0 z d + a1 z d−1 + · · · + ad−1 z + ad = a0 (z − α j ) , (1.16)
j=1

where a0 = 0, all ak are in Q p and all α j are in Q p . Then

log+ |α j | p = maxdk=0 log |ak /a0 | p .

j

Here log+ x := max(0, log x) for real positive x.

The proof follows straight from the Newton polygon of (1.16)—see Proposi-
tion A.18 of Appendix A.

Corollary 1.45 Assume further that the coefficients ai in (1.16) are in Z, with a0 > 0,
and gcd(a0 , . . . , ad ) = 1. Then

log+ |α j | p = log a0 ,
p j

where the sum is taken over all primes p dividing a0 .

20 1 Mahler Measures of Polynomials in One Variable

Proof Suppose p   a0 . Then p  ak for some k, so that

maxdk=0 log |ak /a0 | p =  log p .

Now do this for all such p. 

Corollary 1.46 For α of degree dα over Q, we have

⎛ ⎞
1 ⎝
h(α) = log+ |α j | + log+ |α j | p ⎠ .
dα j p j

Proof This comes straight from (1.15) and Proposition A.15. 

Proof (of Lemma 1.43) The proof is based on the obvious inequality

log+ (x1 x2 ) ≤ log+ x1 + log+ x2 (x1 , x2 > 0) .

Let αi (i = 1, . . . , dα ), β j ( j = 1, . . . , dβ ) and (αβ)k (k = 1, . . . , dαβ ) be the con-

jugates in C of α, β and αβ, respectively. Let G be the Galois group of the normal
closure, N say, of Q(α, β), say of degree d over Q. Applying the d automorphisms
σ of G, the numbers σ α consist of d/dα copies of the αi , the numbers σβ consist of
d/dβ copies of the β j and the numbers σ (αβ) consist of d/dαβ copies of the αi β j .
Hence, averaging the inequalities

log+ |σ (αβ)| ≤ log+ |σ α| + log+ |σβ| ,

for all σ ∈ G, we obtain

1 1 1
log+ |(αβ)k | ≤ log+ |αi | + log+ |β j | .
dαβ k
dα i
dβ j

Doing the same for each prime p, except embedding N into Q p instead of C, we
obtain
1 1 1
log+ |(αβ)k | p ≤ log+ |αi | p + log+ |β j | p .
dαβ k
dα i
dβ j

Then adding all these inequalities and applying Corollary 1.46 gives the result. 
1.9 Notes 21

1.9 Notes

The first result in the direction of Lehmer’s Conjecture where, for a nonzero non-
cyclotomic integer polynomial P of degree d, M(P) − 1 was bounded below by a
function not exponentially small in d, was due to Blanksby and Montgomery [BM71].
They proved that M(P) ≥ 1 + 1/(52d log(6d)) for such P. Fourier analysis, in par-
ticular the nonnegativity of the Fejér kernel, played a large part in the proof. A little
later Stewart [Ste78b] proved a result of similar strength, but by a transcendence-type
argument, using an auxiliary function. The following result, which we state as an
exercise, formed part of Blanksby and Montgomery’s proof.

Exercise 1.47 ([BM71, Lemma 4]) For given real ρ satisfying 0 < ρ ≤ 1 and z ∈ C
with ρ ≤ |z| ≤ ρ −1 , show that

z
|z − 1| ≤ ρ −1 ρ −1 .
|z|

Proposition 1.12 (Gonçalves’ Inequality) was slightly strengthened by Schinzel

[Sch00, Sect. 3.2, Lemma 13]. Also, Borwein, Mossinghoff and Vaaler [BMV07]
generalised the inequality from using the 2-norm to one involving the p-norm of P,
where 1 ≤ p ≤ 2.
For an extensive history of the Gauss–Lucas Lemma (Lemma 1.16), see Rahman
and Schmeisser [RS02, pp. 91–92].
Theorem 1.21 comes from Waldschmidt [Wal00, p. 46]. For more results on the
distance of α from 1, see the survey [Smy08]. Lemma 1.27 is due to Mahler [Mah64].
The ‘variants’ of the Mahler measure referred to in Sect. 1.8 are not true variants:
they are all comparable in the sense that any given value of one variant can be used
to calculate the values of the other variants. On the other hand, there are true Mahler
measure variants. One such true variant is the metric Mahler measure m Mm(α)
of a nonzero algebraic number α defined by Dubickas and Smyth [DS01a]: it is
defined to be the infimum over all k ≥ 1 and all products M(β1 )M(β2 ) · · · M(βk )
where β1 β2 · · · βk = α. They showed that it defines a metric on the (multiplicative)
×
group Q quotiented by the group of all roots of unity. This metric defines the
discrete topology on this quotient group if and only if Lehmer’s Conjecture is true.
Samuels [Sam11] proved that the infimum in the definition is always attained. Thus,
by Proposition 1.9, m Mm(α) is always an algebraic integer.

1.10 Glossary

 The integer part, or floor function: for real x, x denotes the greatest integer
not exceeding x.
α . The house of α.
22 1 Mahler Measures of Polynomials in One Variable

(α), (Q(α)). Respectively the discriminant of an algebraic integer α and the

field discriminant of the field generated by α. The former is an integer multiple
of the latter.
h(α). The (absolute logarithmic) Weil height of an algebraic number α, namely
log M(α).
L(z). Lehmer’s polynomial.
log+ x. For a positive real number x, this is the maximum of 0 and log x. Thus,
log+ x = log x if x ≥ 1, and log+ x = 0 if 0 < x < 1.
M(α). For α an algebraic number, M(α) is the Mahler measure of the minimal
polynomial (in Z[z], content 1) of α.
M(α). The absolute Mahler measure, M(α)1/ deg α .
M(P). The Mahler measure of the polynomial P.
P(z), P(z). Let P(z) be a polynomial with complex coefficients. Then P(z) is the
polynomial whose coefficients are the complex conjugates of those of P. Note
that for a specific complex number z, the value of P(z) does not generally equal
the complex conjugate of P(z), which is written P(z).
sh(P). The shortness of a polynomial P.
absolute Mahler measure. If α is an algebraic number of degree d, then the abso-
lute Mahler measure is the dth root of the Mahler measure.
absolute logarithmic Weil height. The Weil height.
algebraic integer. An algebraic number α is called an algebraic integer if it is
a zero of a monic polynomial P(z) ∈ Z[z]. For example, the algebraic integers
inside Q are precisely the integers Z, and indeed in a number field, the algebraic
integers can be viewed as a generalisation of Z in Q. The set of algebraic integers
in fact forms a ring (Proposition A.22).
algebraic number. A complex number α is called algebraic if it is a zero of some
polynomial P(z) ∈ Z[z]. There is no requirement in the definition that P be
irreducible, although of course if α is algebraic, then there is an irreducible P(z) ∈
Z[z] such that P(α) = 0.
Chebyshev polynomials. There are two families of the Chebyshev polynomials,
usually defined so as to express cos nθ and sin((n + 1)θ )/ sin θ as polynomials
in cos θ . But we prefer the less common monic versions, and define Tn (z) by
Tn (z + z −1 ) := z n + z −n and define Un (z) by Un (z + z −1 ) := (z n − z −n )/(z −
z −1 ). The zeros of Tn and Un are all real and lie in the interval [−2, 2].
conjugate. If α is an algebraic number with minimal polynomial P, then the zeros
of P are the conjugates of α. The set of all zeros of P is a conjugate set.
conjugate set. If P(z) ∈ Z[z] is an irreducible polynomial, then its zeros form a
conjugate set of algebraic numbers. The elements of a conjugate set are the Galois
conjugates of each other.
content. The content of an integer polynomial is the greatest common divisor of
its coefficients.
cyclotomic polynomial. Let ωn be a primitive nth root of unity in C. The minimal
polynomial of ωn over Q is called the nth cyclotomic polynomial. This polynomial
is in fact the minimal polynomial of any primitive nth root of unity. When we
speak simply of a cyclotomic polynomial, we mean one whose zeros are roots
1.10 Glossary 23

of unity, but we do not require irreducibility. Thus, a cyclotomic polynomial is a

product of certain nth cyclotomic polynomials, allowing the possibility of more
than one irreducible factor and the possibility of repeated factors.
discriminant. The discriminant of an algebraic integer α (compare with field
discriminant), written (α), is the square of the determinant of the Vandermonde
j )i, j=1,...,d , where α1 , . . . , αd are the conjugates of α.
matrix (α i−1
field discriminant. The field discriminant of Q(α), written (Q(α)), is the square
( j)
of the determinant of the matrix (αi )i, j=1,...,d , where α (1) , . . . , α (d) is a Z-basis
( j)
for the ring of integers of Q(α) and αi is the image of α ( j) under the ith of the
d embeddings of Q(α) into C.
house. The house of an algebraic number α is the largest modulus of any of the
conjugates of α.
Laurent polynomial. A  Laurent polynomial in z is a polynomial in z and z −1 .
n
Thus, it has the shape i=m ai z i for some m ≤ n, where m is allowed to be
negative.
Lehmer’s Conjecture. The conjecture (originally posed as a question rather than
a conjecture) that there is some c > 1 such that any integer polynomial that has
Mahler measure strictly below c has Mahler measure equal to 1.
Lehmer’s number. The Mahler measure of Lehmer’s polynomial.
Lehmer’s polynomial. The polynomial L(z) = z 10 + z 9 − z 7 − z 6 − z 5 − z 4 −
z 3 + z + 1. Among integer polynomials, this has the smallest known Mahler mea-
sure greater than 1, namely 1.17628 · · · .
length. The length of a polynomial is the sum of the absolute values of its coeffi-
cients.
logarithmic Mahler measure. The logarithm of the Mahler measure. This is par-
ticularly convenient to use for Mahler measures of polynomials in several vari-
ables, where the usual Mahler measure is defined to be the exponential of an
integral.
Mahler measure. Let P(z) = a0 z d + a1 z d−1 + · · · + a0 ∈ Z[z] with a0 = 0, and
let its zeros in C be α1 , . . . , αd . If P(z) has repeated zeros, then they are listed
with multiplicity. The Mahler measure, M(P), is defined to be the product of |a0 |
and all |αi | for which |αi | > 1:

M(P) = |a0 | |αi | .
|αi |>1

One can show that

1
log M(P) = log P(e2πit ) dt ,
0

and this integral version of the definition will be used later to define the Mahler
measure of polynomials in more than one variable.
minimum length, minimum-length multiple. Let P ∈ Z[z]. Among all integer
polynomial multiples of P, say P T where T ∈ Z[z], the smallest possible length
24 1 Mahler Measures of Polynomials in One Variable

is called the minimum length of P, and an example P T that achieves that length
is a minimum-length multiple of P.
Perron number. A Perron number is a real and positive algebraic integer β such
that |β  | < β for any conjugate β  = β.
Pisot number. A Pisot number is a real algebraic integer greater than 1, all of
whose conjugates except itself lie in the open disc |z| < 1.
separation. The separation of an algebraic integer α is the smallest modulus of
the difference between two of its conjugates.
short polynomial. Let P(z) ∈ Z[z]. A short polynomial for P is a polynomial of
minimum length of the shape P(z)Q(z), where Q(z) is a cyclotomic polynomial.
shortness. The shortness of a polynomial P(z) ∈ Z [z] is the length of a short
polynomial for P. The shortness of an algebraic integer α is the shortness of its
minimal polynomial.
Strong Lehmer Conjecture. A more precise version of Lehmer’s Conjecture,
with c = M(L(z)).
subharmonic. A real-valued continuous function of a complex variable, f (z), is
subharmonic on an open subset of the complex plane if for every closed disc
in that region the function value at the centre is bounded above by the average
1
value on the boundary of the disc: f (z) ≤ 0 f (z + r e2πit ) dt, where z is at the
centre of the disc and r is the radius. Extending f (z) to its boundary by continuity
(allowing −∞ as a value), the maximum of f (z) is attained on the boundary.
Weil height. The Weil height of an algebraic number is the logarithm of its absolute
Mahler measure.
Chapter 2
Mahler Measures of Polynomials
in Several Variables

2.1 Introduction

While the set of Mahler measures of polynomials with real coefficients clearly con-
sists of the whole of the positive real line, the situation for polynomials with integer
coefficients is far more interesting. Consider the sequence {M(z n − z − 1)}n≥2, where
 1
log M(z n − z − 1) = log |e2πint − e2πit − 1| dt .
0

As n → ∞, the term e2πint becomes increasingly uncorrelated with e2πit , so that (as
we shall see)
 1  1
lim log M(z n − z − 1) = log |e2πit1 − e2πit2 − 1| dt1 dt2 .
n→∞ 0 0

Thus, if we take this integral to be the definition of log M(z 1 − z 2 − 1), then we have
a convergent sequence of one-variable Mahler measures converging to a two-variable
Mahler measure. This suggests a more general definition of the Mahler measure. It
applies also to Laurent polynomials, so that negative exponents of the variables are
allowed.
Let k ≥ 1, zk = (z 1 , . . . , z k ) and F(zk ) be a nonzero Laurent polynomial with
integer coefficients. Then its Mahler measure M(F) is defined as
 1  1 
M(F) = exp ··· log |F(e 2πit1
,...,e 2πitk
)| dt1 · · · dtk . (2.1)
0 0

© Springer Nature Switzerland AG 2021 25

J. McKee and C. Smyth, Around the Unit Circle, Universitext,
https://doi.org/10.1007/978-3-030-80031-4_2
26 2 Mahler Measures of Polynomials in Several Variables

Our example also suggests considering the set

∞

L := {M(F) | F is a polynomial in k variables with integer coefficients} .
k=1
(2.2)
Then, conjecturally, the example above can be considerably extended, as follows.

Conjecture 2.1 (Boyd’s Conjecture) The set L is a closed subset of the real line.

Note that for M(F1 ) and M(F2 ) in L,

M(F1 )M(F2 ) = M(F1 F2 ) ∈ L , (2.3)

i.e., L is a multiplicative semigroup (indeed a monoid, since 1 = M(1) ∈ L).

Exercise 2.2 Show that for any ρ1 , . . . , ρk of modulus 1, we have

M(F(ρ1 z 1 , . . . , ρk z k )) = M(F(z 1 , . . . , z k )) .

(Thus, in particular, M(F(±z 1 , . . . , ±z k )) = M(F(z 1 , . . . , z k ))).

We shall see that L has many naturally defined closed subsets.

But what, conjecturally, does L look like? We believe that Boyd’s Conjecture 2.1
is true. This implies the truth of Lehmer’s Conjecture—see Exercise 2.7.
It may be that the set L has far more structure. Perhaps it is even a Thue set, as
defined in Sect. 8.2. The properties of L described in this chapter are consistent with
this possibility, but the evidence so far obtained is not substantial, and we are a long
way from formulating a credible conjecture. One result that would be needed is a
good estimate for the difference M (Fr(n) (z)) − M(F) in (2.7) of Proposition 2.17.
This estimate needs to good enough to be able to show that r(n) can be chosen so that
M (Fr(n) (z)) tends to M(F) either from above, from below or both. This was done
by Boyd [Boy81b, Appendix 2] for the particular polynomial F = 1 + z 1 + z 2 , but
it seems to be the only such example.
Let L1 ⊂ L denote the set of Mahler measures M(P), for P a one-variable integer
polynomial. Every F(zk ) ∈ L in k variables zk = (z 1 , . . . , z k ) specifies a particular
subset of L1 by substituting monomials in the single variable z for each variable z j .
Thus, we define

M1 (F) := {M(F(z r1 , z r2 , . . . , z rk )) | (r1 , r2 , . . . , rk ) ∈ Zk } .

The closure M1 (F) in R of this set can be described explicitly, as follows. Given
 ≥ 0 and an  × k integer matrix A = (ai j ), define the k-tuple zA by

zA := (z 1 , . . . , z  ) A := (z 1a11 · · · z a1 , . . . , z 1a1k · · · z ak ) (2.4)

Another random document with
no related content on Scribd:
Fig. 285. Clathropteris meniscoides. From Rhaetic rocks near Erlangen.
[M.S.]
 Fig. 286. Clathropteris egyptiaca. (Nat. size.) a, b, pieces of main ribs in
grooves.
What is probably the rhizome of this species has been described
by Nathorst (Rhizomopteris cruciata); it is similar to that of
Dictyophyllum, but the leaf-scars are more widely separated. This
species occurs in Upper Triassic, Rhaetic or Lower Jurassic rocks of
Scania, France, Germany, Switzerland, Bornholm, North America,
China, Tonkin, and Persia and is represented by fragments in the
Rhaetic beds of Bristol[961].

Clathropteris egyptiaca Sew.[962] Fig. 286.

The specimen on which this species was founded was discovered
in the Nubian Sandstone east of Edfu; the age of the beds is
uncertain, but the presence of Clathropteris suggests a Lower
Jurassic or Rhaetic horizon[963]. Seven strong ribs radiate through the
lamina from the summit of the petiole; at a and b small pieces of the
projecting ribs are shown in the grooves. From the main veins
slender branches are given off at right angles and, as seen in the
enlarged drawing, these again subdivide into a delicate reticulum
with free-ending veinlets.

Fig. 287. Camptopteris spiralis. (After Nathorst. Much reduced.)

Camptopteris.

Camptopteris spiralis, Nath. Figs. 282, C; 287.

Nathorst proposed this generic name for Rhaetic fronds[964]
resembling those of Clathropteris and Dictyophyllum, but differing in
the form of the pinnae and in habit. The habit of the type-species, C.
spiralis, is shown in fig. 287. An examination of the specimens in the
Stockholm Museum convinced me of the correctness of Nathorst’s
restoration[965]. Each of the forked arms of the rachis bore as many
as 150–160 long and narrow pinnae characterised by an
anastomosing venation (fig. 282, C) and by a spiral disposition due
to the torsion of the axes. The sporangia agree in essentials with
those of Dictyophyllum.

Hausmannia.
A critical and exhaustive account of this genus has been given by
Prof. Von Richter[966] based on an examination of specimens found in
the Lower Cretaceous rocks of Quedlinburg in Germany. The name
was proposed by Dunker[967] for leaves from the Wealden of
Germany characterised by a deeply dissected dichotomously
branched lamina. Andrae subsequently instituted the genus
Protorhipis[968] for suborbicular leaves with dichotomously branched
ribs from the Lias of Steierdorf. A similar but smaller type of leaf was
afterwards described by Zigno[969] from Jurassic beds of Italy as P.
asarifolius, and Nathorst[970] figured a closely allied form from Rhaetic
rocks of Sweden. While some authors regarded Hausmannia and
Protorhipis as ferns, others compared them with the leaves of Baiera
(Ginkgoales); Saporta suggested a dicotyledonous affinity for leaves
of the Protorhipis type. The true nature of the fossils was recognised
by Zeiller[971], who called attention to the very close resemblance in
habit and in soral characters to the recent genus Dipteris. A
comparison of the different species of Dipteris, including young
leaves (fig. 231, p. 297), with those of the fossil species reveals a
very striking agreement[972]. There can be no doubt, as Richter points
out, that the names Hausmannia and Protorhipis stand for one
generic type.
Hausmannia may be defined as follows:
Rhizome creeping, slender, dichotomously branched; leaf-stalks slender (2–
25 cm. long), bearing a leathery lamina (1–12 cm. long and broad), wedge-
shaped below, occasionally cordate or reniform, entire or more or less deeply
lobed into broad linear segments. The leaf is characterised by dichotomously
branched main ribs which arise from the summit of the rachis as two divergent
arms and radiate in a palmate manner, with repeated forking, through the
 lamina. Lateral veins are given off at a wide angle, and, by subdivision, form a
fairly regular network similar to that in Dictyophyllum, Clathropteris, and
Dipteris.

Fig. 288. Hausmannia dichotoma. (Specimens from the late Dr Marcus

Gunn’s Collection of Upper Jurassic plants, Sutherlandshire;
very slightly reduced.)

Hausmannia dichotoma, Dunker[973]. Fig. 288, A, B.

This Wealden species, represented in the North German flora and
in beds of approximately the same age at Quedlinburg, has been
discovered by Dr Marcus Gunn in Upper Jurassic rocks on the north-
east coast of Scotland. The lamina (12 cm. or more in length) is
divided into five to seven linear segments and bears a close
superficial resemblance to leaves of Baiera and to recent species of
Schizaea (fig. 222, p. 287). Each segment contains one or two main
ribs (fig. 288, A). A similar form is described by Bartholin[974] and by
Moeller[975] as H. Forchammeri from Jurassic rocks of Bornholm.

Hausmannia Kohlmanni, Richt. Fig. 278, F.

In this species, instituted by Richter from material obtained from
the Lower Cretaceous beds of Strohberg[976], the comparatively
slender rhizome bears fronds with petioles reaching a length in
extreme cases of 25 cm. but usually of about 10 cm. The lamina (1–
7 cm. long and 1–10 cm. broad) is described as leathery, obcordate,
and divided into two symmetrical halves by a median sinus which,
though occasionally extending more than half-way through the
lamina, is usually shallow. The venation consists of two main
branches which diverge from the summit of the petiole (fig. 278, F)
and subdivide into dichotomously branched ribs; finer veins (not
shown in the drawing) are given off from these at right angles and
form more or less rectangular meshes as in other members of the
Dipteridinae and in such recent ferns as Polypodium quercifolium
(fig. 231, D, p. 297).
The imperfect lamina represented in fig. 289 may belong to
Hausmannia Richteri or may be a distinct species; it shows some of
the finer veins connecting the shorter forked ribs, which formed part
of the reticulate ramifying system in the mesophyll. This specimen
was obtained from the plant-beds of Culgower on the
Sutherlandshire coast, which have been placed by some geologists
in the Kimmeridgian series.
The smaller type represented in fig. 278, E, is referred by Richter
to a distinct species, Hausmannia Sewardi[977], founded on a few
specimens from the Lower Cretaceous strata of Strohberg. This
species is characterised by a stouter rhizome bearing smaller leaves
consisting of a short petiole (3–4 cm. long) and an obovate lamina
(1–2 cm. long and broad). There are usually two opposite leaflets on
each leaf-stalk, and these may be equivalent to the two halves of a
single deeply dissected lamina.

Fig. 289. Hausmannia sp. Upper Jurassic, near Helmsdale, Scotland.

From a specimen in the British Museum. (Nat. size.)
It is interesting to compare these different forms of Hausmannia
with the fronds of recent species of Dipteris represented in fig. 231.
The more deeply dissected type, such as H. dichotoma, closely
resembles D. Lobbiana or D. quinquefurcata, while the more or less
entire fossil leaves (fig. 278, E, F and fig. 289) are very like the
somewhat unusual form of Dipteris conjugata shown in fig. 231, B, p.
297.
Other species of the genus are recorded from Liassic rocks of
Steierdorf[978] (Hungary) and of Bornholm[979]. Nathorst[980] has
described a small Rhaetic species from Scania: a French Permian
plant described by Zeiller[981] and compared by him with H.
dichotoma, may be a Palaeozoic example of this Dipteris-like genus.
Some segments of leaves from the Eocene beds (Middle Bagshot)
of Bournemouth, and now in the British Museum, described by
Gardner and Ettingshausen[982] as Podoloma polypodioides, bear a
close resemblance in the venation to the lamina of Dipteris
conjugata.
 CHAPTER XXII.
Marattiales (Fossil).
The discovery of Pteridosperms has necessarily led to a
considerable modification of the views formerly held that existing
genera of Marattiaceae represent survivors of a group which
occupied a dominant position in the forests of the Coal age. Mr Arber
writes:—“The evidence, formerly regarded as beyond suspicion, that
the eusporangiate ferns formed a dominant feature of the vegetation
of the Palaeozoic period, has been undermined, more especially by
the remarkable discovery of the male organs of Lyginodendron by Mr
Kidston. At best we can only now regard them as a subsidiary group
in that epoch in the past history of the vegetable kingdom[983].” Dr
Scott expresses himself in terms slightly more favourable to the view
that the Marattiaceae represent the aristocracy among the Filicales.
He says:—“We now have to seek laboriously for evidence, which
formerly seemed to lie open to us on all hands. I believe, however,
that such careful investigation will result in the resuscitation of the
Palaeozoic ferns as a considerable, though not as a dominant
group[984].” Zeiller’s faith[985] in the prospect of Marattiaceous ferns
retaining their position as prominent members of Palaeozoic floras,
though shaken, is not extinguished: he recognises that they played a
subordinate part.
Reference has already been made to the impossibility of
determining whether Palaeozoic fern-like fronds may be legitimately
retained in the Filicales, or whether they must be removed into the
ever widening territory of the Pteridosperms. The difficulty is that the
evidence of reproductive organs is very far from decisive. In the
absence of the female reproductive organs, the seeds, we cannot in
most cases be certain whether the small sporangium-like bodies on
fertile pinnules are true fern sporangia or the microsporangia of a
heterosporous pteridosperm. What is usually called an exannulate
fern sporangium, such as we have in Angiopteris and in many
Palaeozoic plants, has no distinguishing features which can be used
as a decisive test. The microsporophylls of the Mesozoic
Bennettitales produced their spores in sporangial compartments
grouped in synangia like those of recent Marattiaceae; and in the
case of Crossotheca, a type of frond always regarded as
Marattiaceous until Kidston[986] proved it to be the microsporophyll of
Lyginodendron, we have a striking instance of the futility of making
dogmatic assertions as to the filicinean nature of what look like true
fern sporangia. In all probability Dr Kidston’s surmise that the
supposed fern sporangia known as Dactylotheca, Renaultia,
Urnatopteris are the microsporangia of Pteridosperms will be proved
correct[987]. The question is how many of the supposed
Marattiaceous sporangia must be assigned to Pteridosperms? There
is, however, no reasonable doubt that true Marattiaceae formed a
part of the Upper Carboniferous flora. All that can be attempted in
the following pages is to describe briefly some of the numerous
types of sporangia recognised on Palaeozoic fern-like foliage,
leaving to the future the task of deciding how many of them can be
accepted as those of ferns. It is impossible to avoid overlapping and
some repetition in the sections dealing with true Ferns and with
Pteridosperms. The filicinean nature of the stem known as Psaronius
(see page 415) has not as yet been questioned.
The nomenclature of supposed Marattiaceous species from
Carboniferous and Permian rocks is in a state of some confusion
owing to a lack of satisfactory distinguishing features between
certain types to which different generic names have been assigned.
As we have already seen in the case of supposed leptosporangiate
sporangia, the interpretation of structural features in petrified or
carbonised sporangia does not afford an example of unanimity
among palaeobotanical experts.

Ptychocarpus.
This generic name, proposed by the late Professor Weiss[988], is
applied to a type of fructification illustrated by the plant which
Brongniart named Pecopteris unita, a species common in the Upper
Coal-Measures of England[989]. It is adopted by Kidston for fertile
specimens from Radstock which he describes as Ptychocarpus
oblongus[990], but the precise nature of the fertile pinnules of this
species cannot be determined.

Ptychocarpus unita (Brongn.[991]). Fig. 291, A, B. (= Goniopteris

unita, Grand’Eury.)
This species has tripinnate fronds with linear pinnae bearing
contiguous pinnules of the Pecopteris type (fig. 291, B), 4–5 mm.
long, confluent at the base or for the greater part of their length. On
the under surface of the fertile segments, which are identical with the
sterile, occur circular synangia (fig. 291, A) consisting of seven
sporangia embedded in a common parenchymatous tissue and
radially disposed round a receptacle supplied with vascular tissue.
The synangium is described as shortly stalked like those of Marattia
Kaulfussii (fig. 245, B′, p. 320). In shape, in the complete union of the
sporangia, and presumably in the apical dehiscence, Ptychocarpus
agrees very closely with Kaulfussia (fig. 245); but we cannot be
certain that we have not a collection of microsporangia simulating a
fern synangium.
A synangium closely resembling Ptychocarpus has been
described by Mr Watson[992] from the Lower Coal-Measures of
Lancashire as Cyathotrachus altus, but there is no convincing
evidence as to the nature of the plant on which it was borne.

Danaeites.
This generic name, instituted by Goeppert[993], has been used by
authors without due regard to the nature of the evidence of affinity to
Danaea. The type named by Stur Danaeites sarepontanus[994] (fig.
291, E) bears small pecopteroid pinnules with ovoid sporangia in
groups of 8–16 in two contiguous series on the lower face of the
lamina. The sporangia dehisce by an apical pore and are more or
less embedded in the mesophyll of the segments. No figures have
been published showing any detailed sporangial structure, and such
evidence as we have is insufficient to warrant the conclusion that the
resemblance to Danaea is more than an analogy.
Parapecopteris.

Parapecopteris neuropteroides, Grand’Eury. Fig. 290, D.

The plant described by Grand’Eury[995] from the Coal-fields of Gard
and St Étienne, and made the type of a new genus, is characterised
by pinnules intermediate between those of Pecopteris and
Neuropteris[996] and by the presence of two rows of united sporangia
along the lateral veins, as in Danaea and Danaeites.

Asterotheca.
Certain species of Pecopteris fronds from Carboniferous strata are
characterised by circular sori or synangia consisting of a small
number (3–8) of exannulate sporangia attached to a central
receptacle and free only at their apices. Strasburger[997] suggested a
Marattiaceous affinity for Asterotheca and Stur[998] describes the
species Asterotheca Sternbergii Goepp. (fig. 291, C, D) as an
example of a Marattiaceous fern. The latter author retains Corda’s
genus Hawlea[999] for the fertile fronds of the common Coal-
Measures species Pecopteris Miltoni, while on the other hand
Kidston[1000] includes this type in Asterotheca.

Pecopteris (Asterotheca) Miltoni (Artis).

1825. Filicites Miltoni, Artis, Antedil. Phyt. Pl. xiv.
1828. Pecopteris Miltoni, Brongniart, Prodrome, p. 58.
1828. Pecopteris abbreviata, Brongniart, Hist. vég. foss. p. 337, Pl.
cxv. figs. 1–4; Lindley and Hutton, Foss. Flor. Vol. iii. Pl.
184.
1845. Hawlea pulcherrima, Corda, Flor. Vorwelt, p. 90, Pl. lvii. figs.
7, 8.
1877– Hawlea Miltoni, Stur, Culm Flora, p. 293; Farne Carbon.
1888. Flora, p. 108, Pls. lix. lx.
1888. Pecopteris (Asterotheca) abbreviata, Zeiller, Flor. Valenc. p.
186, Pl. xxiv. figs. 1–4.
 Fig. 290.
A. Alethopteris lonchitica. × 2½. | For description
B. Lonchopteris rugosa. × 2. | see; Chap. xxvii.
C. Sphenopteris Hoeninghausi. × 4. |
D. Parapecopteris neuropteroides.
E. Pecopteris (Dactylotheca) plumosa [= P. (Dactylotheca) dentata
Zeiller (88)]. × 4.
(A–C, E, after Zeiller; D, after Grand’Eury.)
The fronds of this species reached a length of more than 3 metres
and a breadth of 2 metres. They are characterised by the presence
of aphlebiae[1001] appressed to the rachis and by circular sori
composed of a small number (3–6) of sporangia. In habit and in the
form of the pinnules this type is similar to Dactylotheca plumosa.
 Fig. 291.
A, B. Ptychocarpus unita.
C, D. Asterotheca Sternbergii.
E. Danaeites sarepontanus.
F. Hawlea Miltoni.
G. Hawlea pulcherrima.
H–K. Scolecopteris elegans.
(A, B, after Renault; C–G, after Stur; H, I, after Strasburger; K, after
Sterzel.)

Hawlea.
Stur[1002] retains this generic name for sori in which the sporangia
are free and united only by the proximal end to a central receptacle
(fig. 291, F, G). He describes the individual sporangia as possessing
a rudimentary annulus, a comparatively strong wall, and terminating
in a pointed distal end. He emphasises the greater degree of
cohesion between the sporangia of Asterotheca as the distinguishing
feature of that genus; but this is a character difficult to recognise in
some cases, and from the analogy of recent ferns one is disposed to
attach little importance to the greater or less extent to which
sporangia are united, at least in such cases as Asterotheca and
Hawlea when the cohesion is never complete.

Scolecopteris.
Zenker[1003] gave this name to detached fertile pinnules from the
Lower Permian of Saxony, which he described as Scolecopteris
elegans. He recognised the fern nature of the sori and suggested
that the pinnules might belong to the fronds of one of the
“Staarsteinen” (Psaronius), a view which subsequent investigations
render far from improbable. The sori, which occur in two rows on the
lower surface of the small pecopteroid segments with strongly
revolute margins (fig. 291, H–K), contain 4–5 sporangia attached to
a stalked receptacle comparable with that of Marattia Kaulfussii.
These pedicellate synangia were fully described by Strasburger[1004],
who decided in favour of a Marattiaceous alliance. The lower
portions of the distally tapered sporangia are concrescent, the distal
ends being free (fig. 291, H). Stur includes in Scolecopteris the
common species Pecopteris arborescens (fig. 376), but Kidston[1005]
states that the British example of Scolecopteris is S. polymorpha,
Brongn. from the Upper Coal-Measures.
Scolecopteris elegans Zenk. furnishes an example of a plant, or
plant fragment, which has been assigned to the animal kingdom.
Geinitz[1006] described silicified pinnules as Palaeojulus dyadicus, the
generic name being chosen because of the resemblance to
Millipedes such as the genus Julus. The mistake is not surprising to
anyone who has seen a block of siliceous rock from Chemnitz
crowded with the small pinnules with their concave surfaces formed
by the infolding of the edges. Sterzel[1007], who pointed out the
confusion between Myriapods and Filices, has published figures
which illustrate the deceptive resemblance of the pinnules, with their
curved lamina divided by lateral veins into segments, to the body of
a Millipede (fig. 291, K). He points out that Geinitz searched in vain
for the head and legs of Palaeojulus and expressed the hope that
further examination would lead to fresh discoveries: the examination
of sections revealed the presence of sporangia and demonstrated
the identity of Palaeojulus and Scolecopteris.

Discopteris.
Stur[1008] instituted this genus for fertile fronds from the Upper
Carboniferous Schatzlarer beds, including two species Discopteris
karwinensis and D. Schumanni. He described the small
Sphenopteroid pinnules as characterised by disc-shaped sori made
up of 70–100 sporangia attached to a hemispherical receptacle: the
absence of a true annulus led him to refer the genus to the
Marattiaceae. In his memoir on the coal-basin of Heraclea (Asia
Minor), Zeiller[1009] instituted the species Sphenopteris (Discopteris)
Rallii and figured sporangia resembling those described by Stur in
the possession of a rudimentary “apical annulus.” He compared the
sporangia with those of recent Osmundaceae and Marattiaceae. In
the later memoir on the Upper Carboniferous and Permian plants of
Blanzy and Creusot, Zeiller[1010] gives a very full and careful
description of fertile specimens of Sphenopteris (Discopteris)
cristata, a fern originally described by Brongniart as Pecopteris
cristata[1011]. Many of the Sphenopteroid pinnules of this
quadripinnate fern frond show the form and structure of the sori with
remarkable clearness in the admirable photographs reproduced in
Plates i.–iii. of Zeiller’s Blanzy memoir. The lobed pinnules of this
species are of oval-triangular form, 5–15 mm. long and 2·5–6 mm.
broad[1012]. An examination of the type-specimens of Discopteris from
Vienna enabled Zeiller to correct Stur’s original description of the
sori: he found that the Austrian and French specimens, though
specifically distinct, undoubtedly belong to one genus. The sori in
Discopteris cristata are globular, as in the recent genera Cyathea
and Alsophila, and frequently cover the whole face of the lamina.
The individual sporangia are 0·4–0·5 mm. long and 0·15–0·2 mm. in
diameter; they are exannulate, but for the annulus is substituted a
group of thicker-walled and larger cells in the apical and dorsal
region. The description by Stur of a hemispherical receptacle
seemed to indicate an important difference between the Austrian and
French species; but Zeiller found that this feature does not actually
exist and that it was so described as the result of misinterpretation.
Zeiller succeeded in isolating spores, 40–50 μ in diameter, from
some of the sporangia of D. cristata and found that they exhibited
the three-rayed pattern characteristic of fern-spores and which is
indicative of their formation in tetrads. The conclusion arrived at is
that the genus Discopteris, as represented by D. karwinensis, D.
cristata etc., may be regarded as a true fern and included in the
Marattiaceae. As Zeiller points out, the sori of Discopteris differ from
those of recent Marattiaceae in their pluriseriate construction and
agree in this respect with those of the Cyatheaceae. The comparison
already made[1013] between the sporangia of D. Rallii and those of
recent Osmundaceae holds good: the genus affords another
example of a generalised type, in this case probably a fern,
combining features which are now distributed among the
Marattiaceae, Osmundaceae and Cyatheaceae.
• • • • •
In addition to genera founded on true synangia or groups of free or
partially united sporangia, the literature of Palaeozoic ferns contains
several generic names applied to sporangia which occur singly on
Sphenopteroid or Pecopteroid pinnules. The following may serve as
examples; but it should be stated that these will probably be
transferred eventually to the Pteridosperms. It is, however,
immaterial whether they are dealt with here or in the chapter devoted
to the seed-bearing “ferns.”

Dactylotheca.
Zeiller[1014] created this genus for fertile fronds of Pecopteris
dentata Brongn. (= P. plumosa Artis[1015]), a common British species
in the Upper and Middle Coal-Measures. Stur[1016] included P. dentata
in his list of species of Senftenbergia, the genus to which reference
was made under the Schizaeaceae.
Pecopteris (Dactylotheca) plumosa (Artis). Figs. 290, E, 292, 293.
1825. Filicites plumosus, Artis, Antedil. Phyt. p. 17, Pl. xvii.
1828. Pecopteris plumosa, Brongniart, Hist. vég. foss. p. 348, Pls.
cxxi. cxxii.
— P. dentata, Brongniart, ibid. Pls. cxxiii. cxxiv.
— P. delicatulus, Brongniart, ibid. Pl. cxvi. fig. 6.
1832. Sphenopteris caudata, Lindley and Hutton, Foss. Flor. Vol. i.
Pl. xlviii.; Vol. ii. Pl. cxxxviii.
1834. Pecopteris serra, Lindley and Hutton, ibid. Vol. ii. Pl. cvii.
1834. Schizopteris adnascens, Lindley and Hutton, ibid. Vol. i. Pls.
c. ci.
1836. Aspidites caudatus, Goeppert, Syst. fil. foss. p. 363.
1838. Steffensia silesiaca, Presl, in Sternberg, Flor. Vorwelt, Vers.
ii. p. 122.
1869. Pecopteris silesiacus, Schimper, Trait. pal. vég. Vol. i. p. 517.
— Cyathocarpus dentatus, Weiss, Flora der jüngst. Stk. und
Roth. p. 86.
1877. Senftenbergia plumosa, Stur, Culm Flora, ii. p. 187 (293).
— S. dentata, ibid.
1886. Dactylotheca plumosa, Kidston, Cat. Palaeozoic Plants, p.
128.
1888. Dactylotheca dentata, Zeiller, Flor. Valenc. Pls. xxvi.–xxviii.
For a fuller synonymy reference should be made to Kidston’s
account of this species[1017], from which the above list is compiled.
The large fronds of this species are tri- or quadripinnate. The
pinnules vary much in shape and size and in degree of lobing,
according to their position on the frond (fig. 293). The primary pinnae
are subtended by two Aphlebiae (fig. 293, A) appressed to the
rachis, like the delicate leaves of the recent fern Teratophyllum
aculeatum (see page 301). The sporangia (0·5–0·65) are oval and
exannulate and are attached parallel to the lateral veins; they may
occupy the whole of the space between the midrib and the edge of
the pinnules. This species occurs in the Upper, Middle, and Lower
Coal-Measures of Britain, reaching its maximum in the Upper Coal-
Measures. The aphlebiae undoubtedly served to protect the young
fronds, as shown by a specimen figured by Kidston (fig. 293, B); they
may also have served other purposes, as suggested by the above
comparison with Teratophyllum, in the mature frond. Lindley and
Hutton regarded the aphlebiae as leaves of a fern climbing up the
rachis; which they named Schizopteris adnascens, a confusion
similar to that already mentioned in the description of Hemitelia
capensis (see p. 304).

Fig. 292. Dactylotheca plumosa. (After Kidston. Slightly reduced.)

 Fig. 293. Dactylotheca plumosa: A. Rachis with Aphlebiae. B, a, young
pinnae circinately folded. (After Kidston. A, B, ⅘ nat. size.)

Renaultia.
This name was proposed by Zeiller[1018] for Upper Carboniferous
fertile pinnae of the Sphenopteroid type, bearing ovoid sporangia
either singly or in marginal groups of 2 to 5 at the ends of the veins.
The appearance of the apical cells occasionally suggests the
presence of a rudimentary annulus. Kidston has recorded this type of
fructification in Britain[1019]. Stur describes fertile pinnules of the same
type under the generic name Hapalopteris[1020].

Zeilleria.