Graphs, Rings and Polyhedra

Isidoro Gitler and Rafael H. Villarreal

Departamento de Matemáticas
Centro de Investigación y de Estudios Avanzados
del Instituto Politécnico Nacional (IPN)
Ciudad de México, México
e-mail: {igitler,vila}@math.cinvestav.mx

i
Dedication
The first author dedicates this book to his mother Raquel and his son and daughters,
Emilio, Tania and Sofia; finally I would like to thank Rafael, the second author, for
kindly inviting me to such a quest.
Both authors dedicate this book to Samuel Gitler, for being an example to follow
in our lives.

iii
Contents

Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
Chapter 1. Monomial Ideals and Face Rings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.1. Primary decompositions of modules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.2. Monomial ideals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.3. Simplicial complexes and homology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
1.4. Graded modules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
1.5. Stanley-Reisner rings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
1.6. Hilbert functions and Hilbert series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
Chapter 2. Graphs, Ideals and Invariants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
2.1. Graph theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
2.2. Edge ideals and special types of graphs . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
2.3. B-graphs and bounds for Krull dimension . . . . . . . . . . . . . . . . . . . . . . . . . 52
2.4. Cohen-Macaulay and chordal graphs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
2.5. Unmixed and C-M bipartite graphs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
2.6. Shellable graphs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
2.7. Bounds for the depth and multiplicity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
2.8. Suspensions of graphs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
Chapter 3. Polyhedral Sets, Optimization and Semigroups . . . . . . . . . . . . . . . . . 93
3.1. Polyhedral sets and cones . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
3.2. Volumes and unimodular coverings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
3.3. Hilbert bases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
3.4. The integral closure of a semigroup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
3.5. Unimodularity of matrices and normality . . . . . . . . . . . . . . . . . . . . . . . . . . 127
3.6. Cut-incidence matrices and integrality . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
3.7. The elementary integral vectors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
Chapter 4. Monomial Rings and Optimization Problems . . . . . . . . . . . . . . . . . . 139
4.1. Gröbner bases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
v
vi CONTENTS

4.2. Normal rings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142

4.3. Monomial subrings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
4.4. Homogeneous subrings. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
4.5. Ehrhart and homogeneous subrings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153
4.6. Rees algebras and optimization problems . . . . . . . . . . . . . . . . . . . . . . . . . 165
4.7. Blowup algebras of ideals of covers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 188
4.8. The basis monomial ring of a matroid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 191
4.9. Computation of Hilbert functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 194
Chapter 5. Edge Subrings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201
5.1. Planar graphs and embeddings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201
5.2. The subring associated to a graph . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205
5.3. Incidence matrices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 218
5.4. Circuits of a graph and Gröbner bases . . . . . . . . . . . . . . . . . . . . . . . . . . . . 221
5.5. Edge subrings of bipartite graphs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 224
5.6. The integral closure of an edge subring . . . . . . . . . . . . . . . . . . . . . . . . . . . 231
5.7. Ehrhart rings of edge polytopes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 241
5.8. Normalizations of Rees algebras . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 244
5.9. Monomial subrings of complete graphs . . . . . . . . . . . . . . . . . . . . . . . . . . . 250
Chapter 6. Edge Cones of Graphs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 255
6.1. A combinatorial representation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 255
6.2. Studying the bipartite case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 262
6.3. Minimal covers and Rees cones . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 267
6.4. Rees cones of non bipartite graphs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 270
6.5. Irreducible covers of graphs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 272
Chapter 7. Edge Subrings and Combinatorial Optimization . . . . . . . . . . . . . . . . 275
7.1. Integrality of the shift polyhedron . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 276
7.2. The canonical module . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 279
7.3. Computing the a-invariant . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 281
7.4. A family of normal monomial subrings . . . . . . . . . . . . . . . . . . . . . . . . . . . 288
7.5. Some applications to edge subrings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 292
7.6. The a-invariant for non bipartite graphs . . . . . . . . . . . . . . . . . . . . . . . . . . . 296
Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 303
Index of Notation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 312
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 314
Preface
The aim of this book is to provide a theoretical framework to study monomial
rings and ideals associated to graphs, using techniques from graph theory, polyhedral
geometry, combinatorial optimization, and ring theory. We study certain algebraic
properties of rings (such as the Cohen-Macaulay and normality properties), but many
theoretical aspects are presented in the context of combinatorial and geometric
structures. Although we stress graph theory, we shall see that the relation goes
deeper and wider.
In the past three decades part of the trend has been to examine combinatorial
and geometric structures (graphs and hypergraphs, clutters, matroids, simplicial
complexes, polytopes, toric ideals, posets) from the view point of commutative
algebra. One of the purposes has been to understand the algebraic properties and
invariants of the corresponding graded monomial rings and to use this knowledge to
gain insight on these structures and algebras [99, 113, 162, 169]. The resulting blend
of algebra, combinatorics and polyhedra is often called Combinatorial Commutative
Algebra. The main references for the subject are [111, 133, 166, 170], some other
related references are [23, 41, 187]. In comparison with these books our approach is
different in some respects: we focus on graphs as the main combinatorial structure
and we set up applications to combinatorial optimization and blowup algebras.
The results presented here are expected to play a unifying role and to stress
connections between certain parts of mathematics (e.g. between monomial rings
and graph theory or between commutative algebra and combinatorial optimization).
We do not expect the reader to have knowledge of these fields; therefore throughout
the text we introduce the most fundamental results and concepts needed.
Some of the goals of this book are:
∙ To use graph theory and combinatorial optimization techniques to study monomial
rings and ideals associated to graphs and clutters.
∙ To build a “dictionary” between algebraic invariants of edge rings and invariants
of graphs and clutters.
∙ To study algebraic properties and invariants of edge subrings using the combina-
torial structure of graphs.
vii
viii PREFACE

∙ To study basic properties of polyhedral cones and polytopes associated to graphs

and in particular determine their irreducible representations.
∙ To study the relations between the algebraic properties of blowup algebras and
the combinatorial properties of cluters.
∙ To study the structure of certain sub families of bipartite graphs (e.g. unmixed,
Cohen-Macaulay, shellable and sequentially Cohen-Macaulay graphs).
∙ To study the structure of complete intersection toric ideals of edge subrings of
graphs and oriented graphs.
∙ To show the advantages of an algebro-combinatorial study of edge rings of graphs
and clutters by presenting some applications.
∙ To make this book as self contained as possible, accessible to graduate students
in mathematics and to people from other subjects.
The book contains a considerable number of exercises. In many cases the
exercises are used as part of the proofs.
We thank the Consejo Nacional de Ciencia y Tecnologı́a (CONACyT grant
49251-F) and the Sistema Nacional de Investigadores (SNI) for their generous
support.

Mexico City ISIDORO GITLER

RAFAEL H. VILLARREAL
Introduction

This book is about algebraic properties of monomial rings and polyhedra defined
on graphs. The algebraic and combinatorial invariants of these rings are studied in
detail.
Here the theory of monomial rings and the theory of polyhedra are centered on
the structure of graphs as its underlying subject. The deepening and broadening of
the study of monomial rings and ideals associated to graphs by all the developments
in the past decades are evidence of the interest in using graph theory methods
in commutative algebra. The reader will encounter links with other branches of
mathematics, specially with combinatorial optimization and affine geometry.
Teaching a course. Topics in commutative ring theory that are essential for
the whole book are included in Chapter 1. These topics include monomial ideals,
primary decompositions, Hilbert functions and basic facts on dimension theory. The
material of Chapter 1 can be complemented reading [3, Chapter 2], where the notion
of a tensor product is studied in detail. The essential material on graph theory and
polyhedral geometry is introduced in Sections 2.1 and 3.1 respectively.
Chapters 3 and 4 can be used to teach a course on algebraic properties of blowup
algebras of edge ideals, Ehrhart rings and homogeneous subrings. A follow-up
course may include Chapters 5–7. For people interested in ideals and graphs we
suggest to read Chapters 2 and 5. These two chapters study the structure of certain
sub families of bipartite graphs, invariants of graphs and rings, and some graph
theoretical problems arising from complete intersection toric ideals.
Contents. Let R = K[x1 , . . . , xn ] be a polynomial ring over a field K and let
G be a simple graph with vertex set V = { x1 , . . . , xn }. In order to relate graphs
with rings and polyhedral geometry consider the set { f1 , . . . , fq } of all the square-
free monomials of R of the form xi x j such that xi is adjacent to x j , and the set
𝒜 = {v1 , . . . , vq } of all the vectors of the form ei + e j such that xi is adjacent to x j ,
where ei is the ith unit vector in ℝn .
In this book we study the following monomial rings and ideals associated to the
graph G:
∙ edge ideal: I = I(G) := ( f1 , . . . , fq ),

1
2 INTRODUCTION

∙ edge ring: R/I(G),

∙ edge subring: K[G] := K[ f1 , . . . , fq ] ⊂ R,
∙ toric ideal: ideal of relations of K[G],
∙ canonical module of the edge subring of a bipartite graph G:
(0.1) ωK[G] := ({ xa ∣ a ∈ ℤn ∩ ri(ℝ+ 𝒜)}) ⊂ K[G],
where ℝ+ 𝒜 is the cone spanned by 𝒜 and ri(ℝ+ 𝒜) is its relative interior,
∙ Ehrhart ring: A(𝒫 ) := K[xa ti ∣a ∈ ℤn ∩ i𝒫 ] ⊂ R[t], where t is a new
variable and 𝒫 is the convex hull of 𝒜,
∙ Rees algebra: R[It] := K[x1 , . . . , xn , f1 t, . . . , fq t] ⊂ R[t],
∙ symbolic Rees algebra: R s (I) := R ⊕ It ⊕ I (2) t2 ⊕ ⋅ ⋅ ⋅ ⊕ I (i) ti ⊕ ⋅ ⋅ ⋅ , where
I (i) is the ith symbolic power of I,
∙ associated graded ring: grI (R) := R[It]/IR[It].
We also study the following related objects coming from polyhedral geometry and
combinatorics:
∙ edge cone: ℝ+ 𝒜, the rational polyhedral cone spanned by 𝒜,
∙ Rees cone: ℝ+ 𝒜′ , the rational polyhedral cone generated by
𝒜′ := {ei + e j + en+1 ∣ xi is adjacent to x j } ∪ {e1 , . . . , en } ⊂ ℝn+1 ,
∙ set covering polyhedron: Q(A) := { x ∈ ℝn ∣ xA ≥ 1; x ≥ 0}, where A is
the incidence matrix of G with column vectors v1 , . . . , vq ,
∙ complementary simplicial complex or independence complex:
∆G := {D ⊂ V ∣ D is an independent set of G}.
We are interested in studying algebraic properties (e.g. normality, Cohen-
Macaulay and Gorenstein properties) and invariants of these rings (such as the
a-invariant, Krull dimension, depth, multiplicity, and type). For this purpose we
make use of commutative ring theory, graph theory and optimization techniques.
In Chapter 1 we review primary decompositions of modules over commuta-
tive Noetherian rings. Then we study monomial ideals and Stanley-Reisner rings.
We discuss the connection between simplicial complexes and square-free mono-
mial ideals and present Reisner’s characterization of Cohen-Macaulay face rings in
terms of simplicial homology. This chapter also introduces Hilbert series, a topic
that has proven to be fundamental in several related areas and in applications to
combinatorics [133, 166, 170].
The main results on edge rings are presented in Chapter 2. We give bounds
for the depth and the Krull dimension of R/I(G). Then we bound the generalized
multiplicity e′ (G) of R/I(G), where e′ (G) is the number of maximal independent
sets of G. In general, e′ (G) is larger than the usual multiplicity of R/I(G). To study
 INTRODUCTION 3

the edge ring of G we examine graphs with a maximal independent cover, these
graphs are called B-graphs in the literature. This notion is at the center of several
interesting families of graphs. One has the following implications for any graph G
without isolated vertices:
edge-critical ⇒
B-graph ⇒ vertex-critical.
Cohen-Macaulay ⇒ unmixed ⇒
The independence complex ∆G introduced above is precisely the Stanley-
Reisner simplicial complex of I(G). The Cohen-Macaulay property of a given
graph can be expressed in terms of the reduced homology of the links of ∆G [145].
Unfortunately the simpler the graph G the more complicated the complex ∆G . Fol-
lowing [156, 183] we will study Cohen-Macaulay graphs by looking at both the
independence complex and the graph itself. This property is related to other im-
portant properties of simplicial complexes through the following implications about
∆G :
pure shellable ⇒ constructible ⇒ Cohen-Macaulay ⇐ Gorenstein.
vertex decomposable ⇒ shellable ⇒ sequentially Cohen-Macaulay.
We present classifications of the following families:
∙ unmixed bipartite graphs,
∙ Cohen-Macaulay bipartite graphs,
∙ Cohen-Macaulay trees,
∙ shellable bipartite graphs,
∙ sequentially Cohen-Macaulay bipartite graphs.
In Chapter 3 we introduce with certain detail several notions and results from
polyhedral geometry and linear programming. Excellent references for these topics
are [120, 151, 200]. Then we study relative volumes of lattice polytopes, uni-
modular coverings, affine semigroups and their integral closures, Hilbert bases and
optimization problems, circuits of vector matroids and matrices. We will study the
Rees cone of a clutter matrix, and characterize the max-flow min-cut property in
terms of Hilbert bases of Rees cones and the integrality of set covering polyhedra.
In Chapter 4 we deal with point configurations and their lattice polytopes, we
consider various subrings and ideals associated with them, e.g., Ehrhart rings, toric
ideals, monomial subrings and Rees algebras. Ehrhart rings are used here to compute
Hilbert functions of normalizations of zero-dimensional monomial ideals. The class
of homogeneous monomial subrings is studied in detail. We present a description of
the integral closure of a monomial subring using polyhedral geometry and integer
programming techniques. We study the relations between the algebraic properties
of blowup algebras and the combinatorial properties of clutters. Applications to
4 INTRODUCTION

commutative algebra and combinatorial optimization are presented. At the end we

study the basis monomial ideal of a matroid.
Edge subrings and their toric ideals are studied in Chapter 5. If P is a toric ideal
of a monomial subring, then
circuits of P ⊂ universal Gröbner basis of P ⊂ Graver basis of P.
For toric ideals of edge subrings we will make more precise these binomial sets and
inclusions. We give a graph theoretical description of K[G], the integral closure of
K[G] in its field of fractions. Then we compare K[G] with the Ehrhart ring A(P) and
study their multiplicities and the complete intersection property of the toric ideal of
K[G]. Here we study R[It] and its integral closure by looking closely at the matrix
whose columns are the vectors in 𝒜′ . We introduce and study a family of simple
graphs that we call ring graphs, some characterizations of this family are shown. A
graph G has the primitive cycle property (PCP) if any two primitive cycles intersect
in at most one edge. The following implications hold for any graph G:
⇒ planar
outerplanar ⇒ ring graph ⇔ PCP + contains no
subdivision of 𝒦4
⇒ PCP
If G is bipartite, then
ring graph ⇔ complete intersection ⇔ PCP + planar.
The toric ideal of the edge subring of a digraph will also be examined. This ideal is
related to the acyclic orientations of digraphs.
In order to deepen the study of the edge subring of G, we examine the edge
cone of G from the view point of polyhedral geometry, this is the aim of Chapter 6.
Our first objective in this chapter is to give an explicit combinatorial description of
the edge cone. The second goal is to study in detail the facets of the edge cone
of a connected bipartite graph and show a canonical irreducible representation of
the edge cone. As an application to commutative algebra, in Chapter 7 we use the
previous results to compute the a-invariant and the canonical module of an edge
subring. As an application to graph theory a generalization of the classical marriage
theorem will follow. Then we study the Rees algebra R[It], when G is a bipartite
graph. We determine the irreducible representation of the Rees cone ℝ+ 𝒜′ . This
turns out to be related to the minimal vertex covers of G. Some applications to Rees
algebras are presented.
Let G be a connected bipartite graph. The main purpose of Chapter 7 is to study
the canonical module of K[G] and its a-invariant using combinatorial optimization
techniques. The canonical module is a fundamental object in commutative algebra
that contains homological information of K[G], see [23]. From Eq. (0.1) it is
 INTRODUCTION 5

apparent that the convex set

conv(ℤn ∩ ri(ℝ+ 𝒜))
should give some information about ωK[G] , we will examine this convex set and show
that it has some structure. Various applications to edge subrings are presented. Some
of the techniques and results in Chapter 7 can be extended to more general situations
(other than edge subrings) by adding some extra hypothesis.
Throughout this book we will frequently use NORMALIZ [24, 25, 26], a program
created by Winfried Bruns, Bogdan Ichim and Robert Koch for computing normal-
izations of affine semigroups. It provides an invaluable computational tool to study
monomial subrings and their algebraic invariants. This program computes:
∙ the normalization (or integral closure) of an affine semigroup or, in other
terms, the Hilbert basis of a rational cone;
∙ the lattice points and the support hyperplanes of an integral polytope;
∙ the generators of the integral closure of the Rees algebra of a monomial
ideal I = (xv1 , . . . , xvq ) ⊂ R = K[x1 , . . . , xn ];
∙ the Hilbert series and Hilbert polynomial of the semigroup in the homo-
geneous case;
∙ the Ehrhart ring, the relative volume of a lattice polytope, and the multi-
plicity of a zero dimensional monomial ideal;
∙ the minimal vertex covers of a clutter.
If I is square-free, it can also be used to determine:
∙ the minimal primes of I;
∙ the reducedness of the associated graded ring grI (R);
∙ the integrality of the set covering polyhedron
Q(A) = { x∣ xA ≥ 1; x ≥ 0},
where A = (v1 , . . . , vq );
∙ the total dual integrality of the system x ≥ 0; xA ≥ 1.
There are some other computer programs that we use throughout the book, e.g.,
CoCoA [32], PORTA [36], Macaulay2 [90], Maple, and Mathematica [198].
Chapter 1
Monomial Ideals and Face Rings

In this chapter we study monomial ideals and Stanley-Reisner rings. We discuss

the connection between simplicial complexes and square-free monomial ideals and
present Reisner’s characterization of Cohen-Macaulay face rings in terms of simpli-
cial homology. Later in Chapter 2 we will study edge ideals, i.e., ideals generated
by square-free monomials of degree 2 which can be associated to graphs.
We begin this chapter by introducing some notions from commutative algebra.
In particular we recall the notion of primary decomposition of an ideal or module
and review some of its properties. A certain familiarity with rings and modules will
be assumed, which can be quickly acquired by reading [3, Chapters 1 and 2]. For
convenience we will introduce various fundamental algebraic structures. All rings
considered in this book are commutative and Noetherian, and modules are finitely
generated.

1.1. Primary decompositions of modules

A ring R is a nonempty set with two binary operations + and ⋅ (addition and
multiplication) such that the following conditions are satisfied for all elements
x, y, z ∈ R:
(a) (R, +, 0) is an abelian group,
(b) x(y + z) = xy + xz, (y + z)x = yx + zx, and
(c) x(yz) = (xy)z,
where, as usual, we write xy for x ⋅ y. Condition (a) means that for all x, y, z in R
one has:
(a1 ) x + y ∈ R,
(a2 ) x + y = y + x,
(a3 ) x + (y + z) = (x + y) + z,
(a4 ) there is an element 0 ∈ R such that x + 0 = x,
(a5 ) there exists an element − x in R such that x + (− x) = 0.
7
8 1. MONOMIAL IDEALS AND FACE RINGS

We shall consider only rings which are commutative, i.e., xy = yx for all
x, y ∈ R, and have an identity element, i.e., there is 1 ∈ R such that x1 = 1x = x for
all x ∈ R.
Let R be a ring. A subset I ⊂ R is an ideal if it satisfies:
(i) 0 ∈ I,
(ii) a + b ∈ I for all a, b ∈ I,
(iii) a ∈ R and b ∈ I imply ab ∈ I.
An ideal 𝔭 of R is prime if 𝔭 ∕= R and if a, b ∈ R and ab ∈ 𝔭 implies a ∈ 𝔭 or b ∈ 𝔭.
An ideal I of R is finitely generated if there exists f1 , . . . , fq such that
I = {a1 f1 + ⋅ ⋅ ⋅ + aq fq ∣ ai ∈ R, ∀ i} .
In this case we write I = ( f1 , . . . , fq ) or I = R f1 + ⋅ ⋅ ⋅ + R fq . A Noetherian ring R
is a commutative ring with unit with the property that every ideal of R is finitely
generated.
Modules are a natural generalization of both vector spaces and ideals, they play
a unifying role in algebra and geometry.
DEFINITION 1.1.1. Let R be a ring. An R-module M is an additive abelian group
together with a function R × M → M, written (r, m) → rm, and subject to the
following axioms, for all r, s ∈ R and m, n ∈ M:
r(m + n) = rm + rn,
(r + s)m = rm + sm,
(rs)m = r(sm),
1m = m.
EXAMPLE 1.1.2. Let R be a ring and let I be an ideal of R. Then
(a) I is an R-module.
(b) The quotient ring R/I is an R-module with addition and multiplication
(r + I) + (s + I) := (r + s) + I and r(s + I) := rs + I,
where r, s ∈ R. We set r = r + I = {r + i∣ i ∈ I }.
(c) A vector space V over a field K is a K-module.
(d) The cartesian product Rn is an R-module with the operations
(ai ) + (bi ) = (ai + bi ) and r(ai ) = (rai ) (ai , bi , r ∈ R),
where (ai ) = (a1 , . . . , an ).
(e) Let ℕ = {0, 1, 2, . . .}. An abelian group (G, +) is a ℤ-module with the
operation + and the multiplication by scalars given by:
( n times) if n ∈ ℕ, a ∈ G,
{
a + ⋅⋅⋅ + a
na =
(−a) + ⋅ ⋅ ⋅ + (−a) (−n times) if − n ∈ ℕ, a ∈ G.
 1. MONOMIAL IDEALS AND FACE RINGS 9

DEFINITION 1.1.3. Let R be a ring and let M be an R-module. A submodule N of

M is a subset of M which is an R-module under restriction of the module operations
in M.
If N is a submodule of M, then the abelian group M/N inherits an R-module
structure from M defined by r(x + N) = rx + N, where r ∈ R and x ∈ M.
Let M, N be R-modules. A mapping f : M → N is an R-module homomorphism
(or is R-linear) if
(i) f (x + y) = f (x) + f (y),
(b) f (rx) = r f (x),
for all r ∈ R and x, y ∈ M. The image and kernel of f are defined as:
im( f ) = { f (m)∣ m ∈ M } and ker( f ) = { x ∈ M ∣ f (m) = 0}
respectively. The image and the kernel of an R-module homomorphism f are
submodules of N and M respectively.
Let f : M → N be a homomorphism of R-modules. If f is injective (resp. onto),
we call f a monomorphism (resp. epimorphism). If f is a monomorphism (resp.
epimorphism) we write
f f f
M ,→ N or 0 → M → N (resp. M → N → 0).
The map f is called and isomorphism if ker( f ) = (0) and im( f ) = N, in this case we
write M ≃ N. There is an isomorphism of R-modules
M/ker( f ) ≃ im( f ).
f g
A sequence 0 → M ′ → M → M ′′ → 0 of R-modules is called a short exact
sequence if f is a monomorphism, g is an epimorphism and im( f ) = ker(g). A
sequence of R modules and homomorphisms:
fi−1 fi fi+1
⋅ ⋅ ⋅ −→ Mi−1 −→ Mi −→ Mi+1 −→ ⋅ ⋅ ⋅
is said to be exact at Mi if im( fi−1 ) = ker( fi ). If the sequence is exact at each Mi it
is called an exact sequence.
DEFINITION 1.1.4. An R-module M is Noetherian if every submodule N of M
is finitely generated, i.e., N = R f1 + ⋅ ⋅ ⋅ + R fq , for some f1 , . . . , fq .
PROPOSITION 1.1.5. [3, Proposition 6.5] If M is a finitely generated R-module
over a Noetherian ring R, then M is a Noetherian module.
COROLLARY 1.1.6. If R is a Noetherian ring and I is an ideal of R, then R/I
and Rn are Noetherian R-modules. In particular any submodule of Rn is finitely
generated.
10 1. MONOMIAL IDEALS AND FACE RINGS

THEOREM 1.1.7. (Hilbert basis theorem, [3, Theorem 7.5]) A polynomial ring
R[x] over a Noetherian ring R is Noetherian.
In particular, since any field K is a Noetherian ring, by induction on the number
of variables we obtain:
COROLLARY 1.1.8. A polynomial ring K[x1 , . . . , xn ] in a finite number of vari-
ables over a field K is a Noetherian ring.
THEOREM 1.1.9. [3, Proposition 6.2] An R-module M is Noetherian if and only
if M satisfies the ascending chain condition for submodules, i.e., for every ascending
chain of submodules of M
N0 ⊂ N1 ⊂ ⋅ ⋅ ⋅ ⊂ Nn ⊂ Nn+1 ⊂ ⋅ ⋅ ⋅ ⊂ M
there exists an integer k such that Ni = Nk for every i ≥ k.
In this book by a ring (resp. module) we shall always mean a Noetherian ring
(resp. finitely generated module).
The spectrum of a ring R will be denoted by Spec(R), it is the set of all prime
ideals of R. The minimal primes of R are the minimal elements of Spec(R) with
respect to inclusion. The maximal ideals of R are the maximal elements of the set
of proper ideals of R with respect to inclusion.
Modules of fractions and localizations. Let R be a ring, M an R-module, and S
a multiplicatively closed subset of R, i.e., x, y ∈ S imply xy ∈ S . If 1 ∈ S , then the
module of fractions of M with respect to S , or the localization of M with respect to
S , is defined by
S −1 (M) = {m/s∣ m ∈ M, s ∈ S } ,
where m/s = m1 /s1 if and only if t(s1 m − sm1 ) = 0 for some t ∈ S . In particular
S −1 R has a ring structure given by the usual rules of addition and multiplication,
and S −1 M is a module over the ring S −1 R with the operations:
m1 /s1 + m2 /s2 = (s2 m1 + s1 m2 )/s1 s2 (si ∈ S , mi ∈ M)

(a/s) ⋅ (m1 /s1 ) = am1 /ss1 (a ∈ R, s ∈ S ).

There is a canonical map ϕ : M → S −1 M, where ϕ(m) = m/1.
EXAMPLE 1.1.10. If f is an element of a ring R and S = { f n ∣ n ≥ 0}, then S −1 R
is usually written R f . For instance if R = ℂ [x] is a polynomial ring in one variable,
then R x is the ring of Laurent polynomials:
R x = ℂ [x, x−1 ].
DEFINITION 1.1.11. Let M be an R-module and let 𝔭 be a prime ideal of R. If
S = R ∖ 𝔭, in this case S −1 M is written M𝔭 and is called the localization of M at 𝔭.
 1. MONOMIAL IDEALS AND FACE RINGS 11

Primary decomposition of modules.

Let I be an ideal of a ring R. The radical of I is the ideal
rad (I) = { x ∈ R∣ xn ∈ I for some n > 0}.
In particular rad (0), denoted by nil(R), is the set of nilpotent elements of R and is
called the nilradical of R. A ring is reduced if its nilradical is zero.
PROPOSITION 1.1.12. [3, Proposition 1.1.4] If I is a proper ideal of a ring R,
then rad (I) is the intersection of all prime ideals containing I.
Let M be an R-module. The annihilator of M is given by
ann(M) = { x ∈ R∣ xM = 0}.
If m ∈ M, the annihilator of m is ann (m) = ann (Rm). It is convenient to generalize
the notion of annihilator to ideals and submodules. Let N1 and N2 be two submodules
of M, their ideal quotient or colon is defined as
(N1 : N2 ) = { x ∈ R∣ xN2 ⊂ N1 }.
In particular if I and J are two ideals of a ring R, their colon is
(I : J) = {r ∈ R ∣ rJ ⊂ I }.
If J = (x) is a principal ideal we denote (I : J) simply by (I : x).
DEFINITION 1.1.13. The set of associated primes of M, denoted by Ass(M) or
AssR (M), is the set of all prime ideals 𝔭 of R such that there is a monomorphism φ
of R-modules:
φ
R/𝔭 ,→ M.
Note that 𝔭 = ann (φ(1)) = { x ∈ R∣ xφ(1) = 0}. If I is an ideal of R, we denote the
set of associated primes of R/I simply by Ass(I). The elements of Ass(I) are called
the associated primes of I.
DEFINITION 1.1.14. The support of M, denoted by Supp(M), is the set of all
prime ideals 𝔭 of R such that M𝔭 ∕= 0.
PROPOSITION 1.1.15. [132, Theorem 6.5] The set AssR (M) is finite,
AssR (M) ⊂ Supp(M) = {𝔭 ∈ Spec(R) ∣ ann (M) ⊂ 𝔭}
and any minimal element of Supp(M) is in AssR (M).
The minimal primes of M are defined to be the minimal elements of Supp(M)
with respect to inclusion. A minimal prime of M is called an isolated associated
prime of M. An associated prime of M which is not isolated is called an embedded
prime.
12 1. MONOMIAL IDEALS AND FACE RINGS

An element x ∈ R is a zero divisor of M if there is 0 ∕= m ∈ M such that xm = 0.

The set of zero divisors of M is denoted by 𝒵 (M). If x is not a zero divisor on M
we say that x is a regular element of M.
LEMMA 1.1.16. [132, Theorem 6.1] The zero divisors of M are given by:
∪
𝒵 (M) = 𝔭.
𝔭∈AssR (M)

DEFINITION 1.1.17. A submodule N of M is said to be a primary submodule if

AssR (M/N) = {𝔭}. An ideal 𝔮 of a ring R is called a primary ideal if AssR (R/𝔮) =
{𝔭}.

PROPOSITION 1.1.18. An ideal 𝔮 of a ring R is a primary ideal if and only if

/ 𝔮 implies yn ∈ 𝔮 for some n ≥ 1.
xy ∈ 𝔮 and x ∈
PROOF. Assume 𝔮 is a primary ideal. Let x, y ∈ R such that xy ∈ 𝔮 and x ∈ / 𝔮.
Hence y is a zero divisor of R/𝔮 because yx = 0 and x ∕= 0. Since 𝒵 (R/𝔮) = {𝔭} and
rad(𝔮) = 𝔭, we get yn ∈ 𝔮 for some positive integer n. □

THEOREM 1.1.19. [132, Theorem 6.8] If N is a submodule of M, then N has

an irredundant primary decomposition, i.e., there are submodules N1 , . . . , Nr of M
with N = N1 ∩ ⋅ ⋅ ⋅ ∩ Nr such that:
(a) AssR (M/Ni ) = {𝔭i } for all i.
(b) N ∕= N1 ∩ ⋅ ⋅ ⋅ ∩ Ni−1 ∩ Ni+1 ∩ ⋅ ⋅ ⋅ ∩ Nr for all i.
(c) 𝔭i ∕= 𝔭 j if Ni ∕= N j .
REMARK 1.1.20. If N ∕= M, then
AssR (M/N) = {𝔭1 , . . . , 𝔭r },
ann (M/Ni ) is a 𝔭i primary ideal, and rad (ann (M/Ni )) = 𝔭i for all i.
COROLLARY 1.1.21. If R is a ring and I a proper ideal of R, then I has an
irredundant primary decomposition
I = 𝔮1 ∩ ⋅ ⋅ ⋅ ∩ 𝔮r
such that 𝔮i is a 𝔭i -primary ideal and AssR (R/I) = {𝔭1 , . . . , 𝔭r }.
PROOF. Let (0) = I/I = (𝔮1 /I) ∩ ⋅ ⋅ ⋅ ∩ (𝔮r /I) be an irredundant decomposition
of the zero ideal of R/I. Then I = 𝔮1 ∩ ⋅ ⋅ ⋅ ∩ 𝔮r and 𝔮i /I is 𝔭i -primary, that is,
AssR ((R/I)/(𝔮i /I)) = Ass(R/𝔮i ) = {𝔭i }. Let us show that 𝔮i is a primary ideal.
If xy ∈ 𝔮i and x ∈ / 𝔮i , then y is a zero-divisor of R/𝔮i , but 𝒵 (R/𝔮i ) = 𝔭i , hence
y ∈ 𝔭i = rad (ann (R/𝔮i )) = rad (𝔮i ) and yn is in 𝔮i for some n > 0. □
 1. MONOMIAL IDEALS AND FACE RINGS 13

Exercises
1.1.22. Let I be an ideal of a ring R. Prove that rad(I) = I if and only if for any
x ∈ R such that x2 ∈ I one has x ∈ I.
1.1.23. Let I1 , . . . , In , 𝔭 be ideals of a ring R. If 𝔭 is prime and ∩ni=1 Ii ⊂ 𝔭, prove
that Ii ⊂ 𝔭 for some i. In particular if ∩ni=1 Ii = 𝔭, then 𝔭 = Ii for some i.
1.1.24. Let I, 𝔭1 , . . . , 𝔭r be ideals of a ring R. If I is contained in ∪ri=1 𝔭i and 𝔭i
is a prime ideal for i ≥ 3, prove that I ⊂ 𝔭i for some i.
1.1.25. Let I be an ideal of a ring R. Then a prime ideal 𝔭 of R is an associated
prime of R/I if and only if 𝔭 = (I : x) for some x ∈ R.
1.1.26. Let R′ = k[x2 , . . . , xn ] and R = R′ [x1 ] be polynomial rings over a field k.
If is an ideal of R′ and 𝔭 ∈ AssR (R/(I ′ , x1 )), then
I′
(a) 𝔭 = x1 R + 𝔭′ R, where 𝔭′ is a prime ideal of R′ , and
(b) 𝔭′ is an associated prime of R′ /I ′ .
1.1.27. Let R be a ring and let 𝔮 be a proper ideal of R. If every element in R/𝔮
is either nilpotent or invertible, prove that 𝔮 is a primary ideal.
1.1.28. Let R be a ring and let 𝔮 be an ideal of R such that rad (𝔮) is a maximal
ideal. Show that 𝔮 is a primary ideal.
1.1.29. Let R be a ring and let I be an ideal. If x ∈ R ∖ I, then there is an exact
sequence of R-modules:
ψ φ
0 −→ R/(I : x) −→ R/I −→ R/(I, x) −→ 0,
where ψ(r) = xr is multiplication by x and φ(r) = r.
1.1.30. If f : M → N is a homomorphism of R-modules, then there is an induced
homomorphism of S −1 R-modules:

S −1 f : S −1 M −→ S −1 N, m/s 7→ f (m)/s.
f g
1.1.31. If 0 → M ′ → M → M ′′ → 0 is an exact sequence of R-modules, prove
that the following sequence is also exact
S −1 f S −1 g
0 −→ S −1 M ′ −→ S −1 M −→ S −1 M ′′ −→ 0.
1.1.32. If M, N are R-modules, prove Ass(M ⊕ N) = Ass(M) ∪ Ass(N).
1.1.33. Let R = ℚ[x, y, z] and let I = (xn + yn − zn ∣ n ≥ 2). Prove that
I = 𝔮1 ∩ 𝔮2 ∩ 𝔮3 is a primary decomposition of I, where
𝔮1 = (y2 , x − z), 𝔮2 = (x2 , y − z), 𝔮3 = (y3 , z3 , x2 y2 , x2 + y2 − z2 ).
14 1. MONOMIAL IDEALS AND FACE RINGS

1.1.34. Let 𝔭 be a prime ideal of a ring R. Prove that 𝔭R𝔭 is the only maximal
ideal of R𝔭 .
1.1.35. Let M = ℤr × ℤ pα1 × ⋅ ⋅ ⋅ × ℤ pαmm be a finitely generated ℤ-module, where
1
p1 , . . . , pm are prime numbers. Prove the following:
(a) M has finite length iff r = 0 (see Section 1.6).
(b) `(M) = α1 + ⋅ ⋅ ⋅ + αm , if `(M) < ∞.
(c) Ass(M) ∕= Supp(M) if r ∕= 0.
(d) 𝒵 (M) = ∪m i=1 (pi ).
(e) Ass(M) = {(p1 ), . . . , (pm ), (0)} if r ∕= 0.
(f) Ass(M) = {(p1 ), . . . , (pm )} if r = 0.

1.2. Monomial ideals

In this section we study general monomial ideals and introduce square-free mono-
mial ideals. Subsequently in Section 1.5 we will focus on square-free monomial
ideals and present some of their relevant properties.
Let R = k[x] = k[x1 , . . . , xn ] be a polynomial ring over a field k. A monomial of
R is an element of the form
λx1a1 ⋅ ⋅ ⋅ xnan (0 ∕= λ ∈ k; ai ∈ ℕ).
Throughout this book, with very few exceptions, by a monomial we really mean a
monomial with λ = 1. To make notation simpler we set
xa = x1a1 ⋅ ⋅ ⋅ xnan for a = (a1 , . . . , an ) ∈ ℕn .
DEFINITION 1.2.1. An ideal I of R is called a monomial ideal if there is 𝒜 ⊂ ℕn
such that I is generated by { xa ∣a ∈ 𝒜}. If I is a monomial ideal the quotient ring
R/I is called a monomial ring.
LEMMA 1.2.2 (Dickson). If {mi }∞ i=1 is a sequence of monomials of R, then there
is an integer ` such that mi is a multiple of some term in the set {m1 , . . . , m` } for
every i > `.
PROOF. We consider the ideal I ⊂ R generated by {mi }∞ i=1 . By the Hilbert
basis theorem I is finitely generated, see Corollary 1.1.7. It is readily seen that I
can be generated by a finite set of terms m1 , . . . , m` . Hence for each i > `, there is
1 ≤ j ≤ ` such that mi is a multiple of m j , as required. □
REMARK 1.2.3. A monomial ideal is always generated by a finite set of mono-
mials by Dickson’s lemma.
DEFINITION 1.2.4. A face ideal is an ideal 𝔭 of R generated by a subset of the
set of variables, that is, 𝔭 = (xi1 , . . . , xik ) for some variables xi j .
 1. MONOMIAL IDEALS AND FACE RINGS 15

LEMMA 1.2.5. [132, Theorem 6.3] If 0 → M ′ → M → M ′′ → 0 is a short

exact sequence of modules over a ring R, then
AssR (M) ⊂ AssR (M ′ ) ∪ AssR (M ′′ ).
PROPOSITION 1.2.6. If I ⊂ R is a monomial ideal, then every associated prime
of I is a face ideal.
PROOF. By induction on the number of variables that occur in a minimal
generating set of I consisting of monomials. Set 𝔪 = (x1 , . . . , xn ). Let 𝔭 be an
associated prime of I. If rad(I) = 𝔪, then 𝔭 = 𝔪. Hence we may assume rad(I) ∕= 𝔪.
Pick a variable x1 not in rad(I) and consider the ascending chain of ideals
I0 = I and In+1 = (In : x1 ) (n ≥ 0).
Since R is Noetherian one has I` = (I` : x1 ) for some `. There are two cases to
consider. If 𝔭 is an associated prime of (In , x1 ) for some n, then by induction
𝔭 is a face ideal because one can write (In , x1 ) = (In′ , x1 ), where In′ is an ideal
minimally generated by a finite set of monomials in the variables x2 , . . . , xn (cf.
Exercise 1.1.26). Now we assume that 𝔭 is not an associated prime of (In , x1 ) for
any n. By Exercise 1.1.29 for each n there is an exact sequence
x
1
0 −→ R/(In : x1 ) −→ R/In −→ R/(In , x1 ) −→ 0,
hence making a recursive application of Lemma 1.2.5 one obtains that 𝔭 is an
associated prime of In for all n. Since x1 is regular on R/I` one concludes that I`
is an ideal minimally generated by monomials in the variables x2 , . . . , xn , thus by
induction 𝔭 is a face ideal. □
DEFINITION 1.2.7. A monomial xa in R is called square-free if
x a = xi1 . . . xir
for some 1 ≤ i1 < ⋅ ⋅ ⋅ < ir ≤ n.
COROLLARY 1.2.8. If I is an ideal of R generated by square-free monomials and
𝔭1 , . . . , 𝔭 s are the associated primes of I, then
I = 𝔭1 ∩ ⋅ ⋅ ⋅ ∩ 𝔭 s .
PROOF. Set J = ∩i=1 s 𝔭 and note J = rad (I). We only have to check J ⊂ I
i
because I is contained in any of its associated primes. Take a monomial f in
J = rad(I) and write f = xia11 ⋅ ⋅ ⋅ xiarr , where i1 < ⋅ ⋅ ⋅ < ir and ai > 0 for all i. Since
f ` ∈ I for some ` ≥ 1, and using that I is generated by square-free monomials, we
obtain xi1 ⋅ ⋅ ⋅ xir ∈ I. Hence f ∈ I. To finish the proof observe that J is a monomial
ideal because the intersection of monomial ideals is again a monomial ideal (see
Exercise 1.2.19). □
16 1. MONOMIAL IDEALS AND FACE RINGS

DEFINITION 1.2.9. The support of a monomial xa = x1a1 ⋅ ⋅ ⋅ xnan in R is given by

supp(xa ) = { xi ∣ ai > 0}.
PROPOSITION 1.2.10. Let 𝔮 be a monomial ideal of R. Then 𝔮 is a primary ideal
if and only if, after permutation of the variables, 𝔮 has the form:
𝔮 = (x1a1 , . . . , xrar , xb1 , . . . , xbs ),
s supp(xbi ) ⊂ { x , . . . , x }.
where ai ≥ 1 and ∪i=1 1 r

PROOF. If Ass(R/𝔮) = {𝔭}, then by Proposition 1.2.6 one has that 𝔭 is equal
to (x1 , . . . , xr ). Since rad (𝔮) = 𝔭 the ideal 𝔮 is minimally generated by a set of the
form:
{ x1a1 , . . . , xrar , xb1 , . . . , xbs }.
Let x j ∈ supp(xbi ), then xbi = x j m, where m is a monomial not in 𝔮. Since 𝔮 is
primary, a power of x j is in 𝔮. Thus x j ∈ (x1 , . . . , xr ) and consequently 1 ≤ j ≤ r,
as required.
For the converse note that any associated prime 𝔭 of R/𝔮 can be written as
𝔭 = (𝔮 : f ), for some monomial f . It follows rapidly that (x1 , . . . , xr ) is the only
associated prime of 𝔮. □
COROLLARY 1.2.11. If 𝔭 is a face ideal, then 𝔭i is a primary ideal for all i.
PROPOSITION 1.2.12. If I is a monomial ideal of R, then I has an irredundant
primary decomposition I = 𝔮1 ∩ ⋅ ⋅ ⋅ ∩ 𝔮r , where 𝔮i is a primary monomial ideal for
all i and rad (𝔮i ) ∕= rad (𝔮 j ) if i ∕= j.
PROOF. Let f1 , . . . , fq be a set of monomials that minimally generate I. We
proceed by induction on the number of variables that occur in the union of the
supports of f1 , . . . , fq .
One may assume that one of the variables in ∪qi=1 supp( fi ), say xn , satisfies
i
xn ∈/ I for all i, otherwise I is a primary ideal and there is nothing to prove. Next
we permute the fi in order to find integers 0 ≤ a1 ≤ ⋅ ⋅ ⋅ ≤ aq , with aq ≥ 1, such
that fi is divisible by xnai but by not higher power of xn . If we apply this procedure
a
to (I, xnq ), instead of I, then we must choose a variable different from xn .
a
Since (I : xnq ) is generated by monomials with less than n variables, from the
equality
a a
I = (I, xnq ) ∩ (I : xnq )
it follows that one can apply recursively the argument given above to both monomial
ideals occurring in the intersection and use induction to obtain a decomposition of
I into primary monomial ideals 𝔮′1 , . . . , 𝔮′s . Finally we remove redundant primary
ideals from 𝔮′1 , . . . , 𝔮′s and group together all primary ideals with the same radical.
□
 1. MONOMIAL IDEALS AND FACE RINGS 17

The computation of a primary decomposition of a monomial ideal can be carried

out by successive elimination of powers of variables, as described in the proof of
Proposition 1.2.12. We now illustrate this procedure with a specific ideal.
EXAMPLE 1.2.13. If R = k[x, y, z] and I = (yz2 , x2 z, x3 y2 ).
3 2
* (I : x ) = (z, y )

 (J : z ) = (x , y)
I  *
2 2
H
Hj J = (I, x3 ) = (x3 , zx2 , z2 y)
H 
H
j (J, z2 ) = (z2 , x3 , x2 z).
H
H

Thus I = (z, y2 ) ∩ (x2 , y) ∩ (z2 , x3 , x2 z).

Even for monomial ideals a minimal irredundant primary decomposition is not
unique, what is unique is the number of terms in such a decomposition and also the
primary components that correspond to minimal primes [3].
EXAMPLE 1.2.14. If I = (x2 , xy) ⊂ k[x, y], then
I = (x) ∩ (x2 , xy, y2 ) = (x) ∩ (x2 , y),
are two minimal irredundant primary decompositions of I.
COROLLARY 1.2.15. Let R = k[x1 , . . . , xn ] be a polynomial ring over a field k.
If I is a monomial ideal, then I has a primary decomposition
I = 𝔮1 ∩ ⋅ ⋅ ⋅ ∩ 𝔮m ,
such that 𝔮i is generated by powers of variables for all i.
PROOF. Let 𝔮 be a primary ideal minimally generated by the set of monomials
x1a1 , . . . , xrar , f1 , . . . , f s and such that
s
∪
supp( fi ) ⊂ { x1 , . . . , xr },
i=1

where ai > 0 for all i. Observe that if f1 = x1b1 ⋅ ⋅ ⋅ xrbr and b1 > 0, then a1 > b1 and
one has a decomposition:
𝔮 = (x1b1 , x2a2 , . . . , xrar , f2 , . . . , f s ) ∩ (x1a1 , x2a2 , . . . , xrar , x2b2 ⋅ ⋅ ⋅ xrbr , f2 , . . . , f s ),
where in the first ideal of the intersection we have lowered the degree of x1a1 and
have eliminated f1 , while in the second ideal we have eliminated the variable x1
from f1 . Applying the same argument repeatedly it follows that one can write 𝔮
as an intersection of primary monomial ideals such that for each of these ideals
the only minimal generators that contain x1 are pure powers of x1 . Therefore, by
induction, 𝔮 is the intersection of ideals generated by powers of variables. Hence
the result follows from Proposition 1.2.12 and Proposition 1.2.10. □
18 1. MONOMIAL IDEALS AND FACE RINGS

Monomial ideals form a lattice. A partially ordered set or poset is a pair

P = (V, ⪯), where V is a finite set of vertices and ⪯ is a binary relation on V
satisfying:
(a) u ⪯ u, ∀u ∈ V (reflexivity).
(b) u ⪯ v and v ⪯ u, imply u = v (antisymmetry).
(c) u ⪯ v and v ⪯ w, imply u ⪯ w (transitivity).
A poset P = (V, ⪯) with vertex set V = { x1 , . . . , xn } can be displayed by an
inclusion diagram, as in the diagram
r12
6 rH @@r 4
3r Hr 2
H
@r 1
@

of the divisors of 12. In such a diagram xi is joined to x j by raising a line if xi ≺ x j

and there is no other vertex in between.
Let L be a lattice, that is, L is a poset in which any two elements x, y have a
greatest lower bound or meet x ∧ y and a lowest upper bound or join x ∨ y. The
lattice L is called distributive if
(a) x ∧ (y ∨ z) = (x ∧ y) ∨ (x ∧ z), and
(b) x ∨ (y ∧ z) = (x ∨ y) ∧ (x ∨ z), ∀ x, y, z in L.
PROPOSITION 1.2.16. If L is the family of monomial ideals of R, order by
inclusion, then L is a distributive lattice under the operations I ∧ J = I ∩ J and
I ∨ J = I + J.
PROOF. It follows from Exercises 1.2.19 and 1.2.20. □

Exercises
1.2.17. If I is a monomial ideal of R, then any associated prime 𝔭 of R/I can be
written as 𝔭 = (I : f ), for some monomial f ; cf. Exercise 1.1.25.
1.2.18. If L is a lattice, then the following conditions are equivalent:
(a) x ∧ (y ∨ z) = (x ∧ y) ∨ (x ∧ z), ∀ x, y, z in L.
(b) x ∨ (y ∧ z) = (x ∨ y) ∧ (x ∨ z), ∀ x, y, z in L.
1.2.19. Let I and J be two ideals generated by finite sets of monomials F and
G respectively, prove that the intersection I ∩ J is generated by the set
{lcm( f, g)∣ f ∈ F and g ∈ G}.
1.2.20. If I, J, K are monomial ideals, prove the equalities
I ∩ (J + K) = (I ∩ J) + (I ∩ K),
I + (J ∩ K) = (I + J) ∩ (I + K).
 1. MONOMIAL IDEALS AND FACE RINGS 19

1.2.21. If I and J are two monomial ideals, then (I : J) is a monomial ideal.

1.2.22. Find a primary decomposition of the ideal:
I = (x1 x22 , x12 x2 , x1 x32 , x12 x3 , x2 x32 , x22 x3 , x1 x2 x3 ).
1.2.23. Show that I = (x12 , x1 x2 ) is a non primary ideal with a prime radical.
1.2.24. Prove that the radical of a monomial ideal is a square-free monomial
ideal.

1.3. Simplicial complexes and homology

A finite simplicial complex consists of a finite set V of vertices and a collection ∆ of
subsets of V called faces or simplices such that
(i) If v ∈ V, then {v} ∈ ∆.
(ii) If F ∈ ∆ and G ⊂ F, then G ∈ ∆.
Let ∆ be a simplicial complex and let F be a face of ∆. Define the dimensions
of F and ∆ by
dim(F) = ∣F ∣ − 1 and dim(∆) = sup{dim(F)∣ F ∈ ∆}
respectively. A face of dimension q is called a q-face or a q-simplex.
EXAMPLE 1.3.1. Take the following triangulation of the disk with a whisker
attached.
xs1 xs3
A
A 0-simplices: x1 , x2 , x3 , x4
1-simplices: x1 x2 , x2 x3 , x3 x4 , x2 x4
A
A
A
A
A
A 2-simplices: x2 x3 x4
A
s As
x2 x4
Let F be a q-simplex of ∆ with vertices v0 , v1 , . . . , vq . We say that two total
orderings of the vertices
vi0 < ⋅ ⋅ ⋅ < viq and v j0 < ⋅ ⋅ ⋅ < v jq
are equivalent if (i0 , . . . , iq ) is an even permutation of ( j0 , . . . , jq ). This is an
equivalence relation because the set of even permutations form a group, and for
q > 1, it partitions the total orderings of v0 , . . . , vq into two equivalence classes. An
oriented q-simplex of ∆ is a q-simplex F with a choice of one of these equivalence
classes. If v0 , . . . , vq are the vertices of F, the oriented simplex determined by the
ordering v0 < ⋅ ⋅ ⋅ < vq will be denoted by [v0 , . . . , vq ].
20 1. MONOMIAL IDEALS AND FACE RINGS

Assume that k is a field. Let Cq (∆) be the k-vector space with basis consisting
of the oriented q-simplices in ∆, modulo the relations
[v0 , v1 , v2 , . . . , vq ] + [v1 , v0 , v2 , . . . , vq ].
In particular notice that dimk Cq (∆) equals the number of q-simplices of ∆. For
q ≥ 1 we define the homomorphism
∂q : Cq (∆) −→ Cq−1 (∆)
induced by
q
∑
∂q ([v0 , v1 , . . . , vq ]) = (−1)i [v0 , v1 , . . . , vi−1 , v̂i , vi+1 , . . . , vq ],
i=0

where v̂i means that the vertex vi is to be deleted. Since ∂q ∂q+1 = 0 we obtain the
chain complex C(∆) = {Cq (∆), ∂q }, which is the oriented chain complex of ∆. The
augmented oriented chain complex of ∆ is
d ∂ 
0 −→ Cd (∆) −→ Cd−1 (∆) −→ ⋅ ⋅ ⋅ −→ C0 (∆) −→ C−1 (∆) = k −→ 0,
where d = dim(∆) and (v) = 1 for every vertex v of ∆. This augmented chain
complex is denoted by (C? (∆), ). Set ∂0 =  and C−1 (∆) = k.
Let Zq (∆, k) = ker(∂q ), Bq = im(∂q+1 ) and
H̃q (∆; k) = Zq (∆, k)/Bq (∆, k), for q ≥ 0.
The elements of Zq (∆, k) and Bq (∆, k) are called cycles and boundaries respectively
and H̃q (∆; k) is the qth reduced simplicial homology group of ∆ with coefficients in
k.
REMARK 1.3.2. Note that if ∆ ∕= ∅, then H̃i (∆; k) = 0 for i < 0.
PROPOSITION 1.3.3. If ∆ ∕= ∅, then dimk H̃0 (∆; k) = r − 1, where r is the number
of connected components of ∆.
PROOF. Let V = { x1 , . . . , xn } be the vertex set of ∆ and ∆1 , . . . , ∆r its connected
components. Denote the vertex set of ∆i by Vi and pick a vertex in each one of the
Vi , one may assume xi ∈ Vi for i = 1, . . . , r.
Set βi = xi − xr and βi = βi + im(∂1 ) for i = 1, . . . , r − 1. We claim that the set
ℬ = {β1 , . . . , βr−1 } is an k-basis for H̃0 (∆; k) = ker(∂0 )/im(∂1 ).
First we prove that ℬ is linearly independent. Assume r−1 i=1 ai βi is in im(∂1 )
∑
for some ai in k, by making an appropriate grouping of terms one can write
r−1
∑ ∑ ∑
ai βi = bi j (xi − x j ) + ⋅ ⋅ ⋅ + bi j (xi − x j ),
i=1 xi ,x j ∈V1 xi ,x j ∈Vr
 1. MONOMIAL IDEALS AND FACE RINGS 21

for some bi j in k. Since V is a k-basis for C0 (∆) and the Vi are mutually disjoint one
has
∑
a` x` = bi j (xi − x j ) for 1 ≤ ` ≤ r − 1,
xi ,x j ∈V`

applying ∂0 to both sides of this equality we obtain a` = 0 for all `.

Next we show that ℬ is a set of generators. Let xi1 and x j1 be two vertices
of ∆, it suffices to prove that xi1 − x j1 + im(∂1 ) is in the subspace generated by ℬ
because ker(∂0 ) is generated by the elements of the form xi − x j . Assume xi1 ∈ Vm
and x j1 ∈ V s . There is a path xi1 , . . . , xit , xm in ∆m and a path x j1 , . . . , x j p , x s in ∆ s
such that every two consecutive vertices of these paths are a 1-face of ∆. From the
equalities
(xi1 − xi2 ) + (xi2 − xi3 ) + ⋅ ⋅ ⋅ + (xit − xm ) = xi1 − xm
(x j1 − x j2 ) + (x j2 − x j3 ) + ⋅ ⋅ ⋅ + (x j p − x s ) = x j1 − x s ,
one concludes that xi1 − x j1 + im(∂1 ) = (xm − xr ) − (x s − xr ) + im(∂1 ). □

Exercises
Let ∆ be a simplicial complex and let k be a field.

1.3.4. If dim ∆ = d and fi is the number of i-faces in ∆. Prove:

d
∑ d
∑
(−1)i rank H̃i (∆; k) = −1 + (−1)i fi .
i=−1 i=0

This number is called the reduced Euler characteristic of ∆ and is denoted by χ̃(∆).
The number χ(∆) = χ̃(∆) + 1 is the Euler characteristic of ∆.

1.3.5. Prove that there is an exact sequence

0 −→ H̃0 (∆; k) −→ H0 (∆; k) −→ k −→ 0

and conclude that H0 (∆; k) ≃ H̃0 (∆, k) ⊕ k.

1.3.6. Let V = { x1 , . . . , xn } be the vertex set of ∆ and let

1 ∂0 ∂
⋅ ⋅ ⋅ −→ C1 (∆) −→ C0 (∆) −→ C−1 (∆) = k −→ 0,

be the augmented chain complex. Prove that ker(∂0 ) is generated by the cycles of
the form xi − x j . Prove that dimk ker(∂1 ) is equal to f1 − f0 + r, where r is the
number of connected components of ∆.
22 1. MONOMIAL IDEALS AND FACE RINGS

1.4. Graded modules

Let (H, +) be an abelian semigroup. An H-graded ring is a ring R together with a
decomposition ⊕
R= Ra (as a ℤ-module),
a∈H
such that Ra Rb ⊂ Ra+b for all a, b ∈ H. Accordingly if R is an H-graded ring and
M is an R-module with a decomposition
⊕
M= Ma ,
a∈H
such that Ra Mb ⊂ Ma+b for all a, b ∈ H, we say that M is an H-graded module.
An element f ∈ M is said to be homogeneous of degree a if f ∈ Ma , in this
case we set deg( f ) = a. The elements in Ra are also called forms of degree a.
A map ϕ : M → N between H-graded modules is graded if ϕ(Ma ) ⊂ Na for
all a ∈ H. If M = ⊕a∈H Ma is an H-graded module and N = ⊕a∈H Na is a graded
submodule, that is, Na ⊂ Ma for all a, then M/N is an H-graded R-module with
(M/N)a = Ma /Na .
EXAMPLE 1.4.1. Let R = k[x1 , . . . , xn ] be a polynomial ring over a field k. We
assign degree di to xi , where d1 , . . . , dn are positive integers. For a = (ai ) in ℕn we
set ∣a∣ = a1 d1 + ⋅ ⋅ ⋅ + an dn . The induced ℕ-grading is:
Ri = k{ xa ∣ i = ∣a∣}; R = R0 ⊕ R1 ⊕ ⋅ ⋅ ⋅ ⊕ Ri ⊕ ⋅ ⋅ ⋅
This gives a ℤ-grading by setting Ri = 0 for i < 0. An ideal generated by homoge-
neous polynomials is called homogeneous or graded. If I is a homogeneous ideal,
then I is graded by Ii = I ∩ Ri . Hence R/I inherits a structure of ℕ-graded R-module
whose ith component is (R/I)i = Ri /Ii .
DEFINITION 1.4.2. The standard grading of a polynomial ring k[x1 , . . . , xn ] is
the ℕ-grading induced by setting deg(xi ) = 1 for all i.
LEMMA 1.4.3 (Graded Nakayama lemma). Let M be an ℕ-graded R-module
and let R+ = ⊕∞i=1 Ri . If N is a graded submodule of M and I ⊂ R+ is a graded ideal
of R such that M = N + I M, then N = M.
PROOF. Since M/N = I(M/N), one may assume N = (0). If x ∈ M is a
homogeneous element of degree r, then a recursive use of the equality M = I M
yields that x ∈ I r+1 M and x must be zero. □
A local ring (R, 𝔪) is a ring R with exactly one maximal ideal 𝔪.
LEMMA 1.4.4. (Nakayama; see [3, Corollary 2.7]) Let (R, 𝔪) be a local ring
and let N be a submodule of an R-module M. If I is an ideal of R contained in 𝔪
such that M = I M + N, then M = N.
 1. MONOMIAL IDEALS AND FACE RINGS 23

Let M be an R-module. The minimum number of generators of M will be

denoted by ν(M). A consequence of Nakayama’s lemma is an expression for ν(M),
when the ring R is local.
COROLLARY 1.4.5. If M is a module over a local ring (R, 𝔪), then
ν(M) = dimk (M/𝔪 M), where k = R/𝔪.
PROOF. Let α1 , . . . , αq be a minimal generating set for M and αi = αi + 𝔪 M.
After a permutation of the αi one may assume that α1 , . . . , αr is a basis for M/𝔪 M
as a k-vector space, for some r ≤ q. Set N = Rα1 + ⋅ ⋅ ⋅ + Rαr . Note the equality
M = N + 𝔪 M, then by Nakayama’s Lemma N = M. Therefore r = q, as required.
□

Exercises
1.4.6. Let I be a graded ideal of a polynomial ring R. Prove that the radical of
I is also graded.
1.4.7. Let R = ⊕i∈ℤ Ri be a ℤ-graded ring. Prove that R0 is a subring of R with
1 ∈ R0 . If Ri = 0 for i < 0, R is called a graded R0 -algebra.

1.5. Stanley-Reisner rings

Let R = k[x1 , . . . , xn ] be a polynomial ring over a field k. If I is an ideal of R
generated by square-free monomials, the Stanley-Reisner simplicial complex ∆I
/ I } and its faces are defined by
associated to I has vertex set V = { xi ∣ xi ∈
∆I = {{ xi1 , . . . , xir }∣ i1 < ⋅ ⋅ ⋅ < ir , xi1 ⋅ ⋅ ⋅ xir ∕∈ I }.
Conversely if ∆ is a simplicial complex with vertex set V contained in { x1 , . . . , xn },
the Stanley-Reisner ideal I∆ is defined as
I∆ = { xi1 ⋅ ⋅ ⋅ xir ∣ i1 < ⋅ ⋅ ⋅ < ir , { xi1 , . . . , xir } ∈
/ ∆} ,
( )

and its Stanley-Reisner ring k[∆] is defined as the quotient ring R/I∆ .
DEFINITION 1.5.1. Let I be an ideal of R generated by square-free monomials.
The quotient ring R/I is called a face ring or a Stanley-Reisner ring.
EXAMPLE 1.5.2. A simplicial complex ∆ of dimension 1 and its Stanley-Reisner
ideal: s x2
@
x1 s @sx3
@ I∆ = (x1 x3 , x1 x2 x4 , x2 x3 x4 , x1 x2 x3 x4 )
@
@ s x4
@
24 1. MONOMIAL IDEALS AND FACE RINGS

DEFINITION 1.5.3. A face F of a simplicial complex ∆ is said to be a facet if F

is not properly contained in any other face of ∆.
PROPOSITION 1.5.4. If ∆ is a simplicial complex with vertices x1 , . . . , xn , then
the primary decomposition of the Stanley-Reisner ideal of ∆ is:
∩
I∆ = 𝔭F ,
where the intersection is taken over all facets F of ∆, and 𝔭F denotes the face ideal
/ F.
generated by all xi such that xi ∈
PROOF. Let F be a face of ∆ and 𝔭F the face ideal generated by the xi such that
/ F. Note that 𝔭F is a minimal prime of I∆ if and only if F is a facet. Therefore
xi ∈
the result follow from Corollary 1.2.8. □
Krull dimension and height. By a chain of prime ideals of a ring R we mean a
finite strictly increasing sequence of prime ideals
𝔭0 ⊂ 𝔭1 ⊂ ⋅ ⋅ ⋅ ⊂ 𝔭n ,
the integer n is called the length of the chain. The Krull dimension of R, denoted by
dim(R), is the supremum of the lengths of all chains of prime ideals in R. Let 𝔭 be
a prime ideal of R, the height of 𝔭, denoted by ht (𝔭) is the supremum of the lengths
of all chains of prime ideals
𝔭0 ⊂ 𝔭1 ⊂ ⋅ ⋅ ⋅ ⊂ 𝔭n = 𝔭
which end at 𝔭. Note dim(R𝔭 ) = ht (𝔭). If I is an ideal of R, then ht (I), the height of
I, is defined as
ht(I) = min{ht(𝔭)∣ I ⊂ 𝔭 and 𝔭 ∈ Spec(R)}.
In general dim(R/I) + ht (I) ≤ dim(R). The difference dim(R) − dim(R/I) is called
the codimension of I.
Let M be an R-module. Recall that the dimension of M is given by
dim(M) := dim(R/ann (M))
and the codimension of M is codim(M) := dim(R) − dim(M).
THEOREM 1.5.5. [132, Theorem 15.4] If R[x] is a polynomial ring over a
Noetherian ring R, then dim R[x] = dim(R) + 1.
PROPOSITION 1.5.6. [187, Corollary 2.1.7] If R is a polynomial ring over a field
k and I ∕= R is an ideal of R, then
dim(R/I) = dim(R) − ht (I).
COROLLARY 1.5.7. Let ∆ be a simplicial complex of dimension d on the vertex
set V = { x1 , . . . , xn } and let k be a field. Then
/ I∆ and i1 < ⋅ ⋅ ⋅ < i s }.
dim k[∆] = d + 1 = max{ s ∣ xi1 ⋅ ⋅ ⋅ xis ∈
 1. MONOMIAL IDEALS AND FACE RINGS 25

PROOF. Let F be a facet of ∆ with d + 1 vertices. Note that by Proposition 1.5.4

the face ideal 𝔭F generated by the variables not in F has height equal to the height
of I∆ . Hence ht(I∆ ) = n − d − 1, taking into account Proposition 1.5.6 one has
dim k[∆] = d + 1. □
DEFINITION 1.5.8. Let M be an R-module. A sequence h = h1 , . . . , hn in R is
called a regular sequence of M, or an M-regular sequence, or simply an M-sequence
if (h)M ∕= M and hi ∈/ 𝒵 (M/(h1 , . . . , hi−1 )M) for all i.
Notice that positively graded modules over polynomial rings behave as modules
over local rings. Thus most of the results that hold for modules over local rings will
hold for ℕ-graded modules over polynomial rings. Sometimes we will only give
the local versions of the results. The same applies to definitions, i.e., sometimes we
will only give the local versions of the definitions. For instance the reader should
compare the local and graded versions of Nakayama’s lemma (see Lemmas 1.4.3
and 1.4.4).
LEMMA 1.5.9. [187, Lemma 1.3.6] Let M be a module over a local ring (R, 𝔪).
If h1 , . . . , hr is an M-regular sequence in 𝔪, then r ≤ dim(M).
Let M ∕= (0) be a module over a local ring (R, 𝔪). The depth of M, denoted by
depth(M), is the length of any maximal regular sequence on M which is contained
in 𝔪. In general by Lemma 1.5.9 we have the inequality depth(M) ≤ dim(M)
DEFINITION 1.5.10. M is called Cohen-Macaulay (C-M for short) if
depth(M) = dim(M),
or if M = (0). The ring R is called Cohen-Macaulay if R is Cohen-Macaulay as
an R-module. An ideal I of R is Cohen-Macaulay if R/I is a Cohen-Macaulay
R-module.
If R = k[x1 , . . . , xn ] is a polynomial ring over a field k and I is a graded ideal,
then the maximal ideal 𝔪 = (x1 , . . . , xn ) plays the role of the maximal ideal of
a local ring, i.e., in this case the depth of R/I is the largest integer r such that
there is a homogeneous sequence h1 , . . . , hr in 𝔪 with hi not a zero-divisor of
R/(I, h1 , . . . , hi−1 ) for all 1 ≤ i ≤ r, where h0 = 0.
DEFINITION 1.5.11. Let k be a field. A simplicial complex ∆ is said to be
Cohen-Macaulay over k if the Stanley-Reisner ring k[∆] is a Cohen-Macaulay ring.
PROPOSITION 1.5.12. [23, Theorem 2.1.2] If M ∕= (0) is a Cohen-Macaulay
R-module over a local ring (R, 𝔪) and 𝔭 ∈ Ass(M), then
dim(R/𝔭) = depth(M).
DEFINITION 1.5.13. An ideal I of a ring R is height unmixed or unmixed if
ht (I) = ht (𝔭) for all 𝔭 in AssR (R/I).
26 1. MONOMIAL IDEALS AND FACE RINGS

COROLLARY 1.5.14. Let R be a positively graded polynomial ring over a field k

and I a graded ideal of R. If R/I is Cohen-Macaulay, then I is unmixed.
PROOF. It follows from Propositions 1.5.12 and 1.5.6. □
DEFINITION 1.5.15. A simplicial complex ∆ is called pure if all its maximal
faces have the same dimension (maximal with respect to inclusion).
COROLLARY 1.5.16. A Cohen-Macaulay simplicial complex ∆ is pure.
PROOF. It follows from Proposition 1.5.4 and Corollary 1.5.14. □
DEFINITION 1.5.17. Let ∆ be a simplicial complex and F ∈ ∆, define lk(F), the
link of F, as
lk(F) = {H ∈ ∆∣ H ∩ F = ∅ and H ∪ F ∈ ∆}.
THEOREM 1.5.18 (Reisner [145]). Let ∆ be a simplicial complex of dimension
d. If k is a field, then the following conditions are equivalent:
(a) ∆ is Cohen-Macaulay over k.
(b) H̃i (lk F; k) = 0 for F ∈ ∆ and i < dim lk F.

We clarify that in particular a Cohen-Macaulay complex ∆, being the link of its

empty face, must satisfy H̃i (∆; k) = 0 for i < dim ∆.
DEFINITION 1.5.19. Let A, B be two rings. A homomorphism of rings is a map
ϕ : A → B such that:
(i) ϕ(a + b) = ϕ(a) + ϕ(b), ∀ a, b ∈ A,
(ii) ϕ(ab) = ϕ(a)ϕ(b), ∀ a, b ∈ A, and
(iii) ϕ(1) = 1.
Let R be a ring and let ϕ : ℤ → R be the homomorphism ϕ(a) = a ⋅ 1R , then
ker(ϕ) = nℤ, for some n ≥ 0. The integer n is called the characteristic of R and is
denoted by char(R).
PROPOSITION 1.5.20. Let (R, 𝔪) be a local ring, then either char(R) = 0 or
char(R) = pk , for some prime number p and some integer k ≥ 1.
PROOF. Let n = char(R) ≥ 2. If n = pqm, where p, q are prime numbers with
q ∕= p, then p ⋅ 1R and q ⋅ 1R are both in 𝔪 because they are non invertible, but this
is impossible because 1 = ap + bq and from the equality
1R = (a ⋅ 1R )(p ⋅ 1R ) + (b ⋅ 1R )(q ⋅ 1R ),
one derives 1R ∈ 𝔪, a contradiction. □
In general the Cohen-Macaulay condition of a simplicial complex ∆ may depend
on the characteristic of the field k, examples of this dependence are triangulations of
the projective plane [145], see the exercises at the end of this section. This condition
 1. MONOMIAL IDEALS AND FACE RINGS 27

is shown to be a topological property in [139], i.e., depends only on the underlying

topological space ∣∆∣, the geometric realization of ∆.
EXAMPLE 1.5.21. If ∆ is a discrete set with n vertices
s x5
x6 s s x4

x7 s sx3

xn s s x2
s x1

then ∆ is Cohen-Macaulay and I∆ = (xi x j ∣1 ≤ i < j ≤ n).

COROLLARY 1.5.22. If ∆ is a 1-dimensional simplicial complex, then ∆ is
connected if and only if ∆ is Cohen-Macaulay.
PROOF. By Reisner theorem ∆ is Cohen-Macaulay iff H̃0 (∆; k) = 0, because the
link of a vertex of ∆ consists of a discrete set of vertices. To complete the argument
apply Proposition 1.3.3. □
PROPOSITION 1.5.23 ([114]). If k is a field, then a simplicial complex ∆ is
Cohen-Macaulay over k if and only if lk(F) is Cohen-Macaulay over k for all
F ∈ ∆.
PROOF. ⇒) Let F be a fixed face in ∆ and ∆′ = lk(F). If G ∈ ∆′ , using the
equality lk∆′ (G) = lk∆ (G ∪ F) and Reisner criterion one infers H̃i (lk∆′ (G), k) = 0
for i < dim lk∆′ (G), thus ∆′ is Cohen-Macaulay.
⇐) Since all the links are Cohen-Macaulay, it suffices to take F = ∅ and observe
∆ = lk(∅) to conclude that ∆ is Cohen-Macaulay. □
DEFINITION 1.5.24. The q-skeleton of a simplicial complex ∆ is the simplicial
complex ∆q consisting of all p-simplices of ∆ with p ≤ q.
PROPOSITION 1.5.25. Let ∆ be a simplicial complex of dimension d and ∆q its
q-skeleton. If ∆ is Cohen-Macaulay over a field k, then ∆q is Cohen-Macaulay for
q ≤ d.
PROOF. Set ∆′ = ∆q and take F ∈ ∆′ with dim(F) < q ≤ d. Note
lk∆′ (F) = (lk∆ (F))q−∣F∣ .
Since ∆ is pure there is a facet F ′ of ∆ of dimension d containing F, hence F ′ ∖ F
is a face of lk∆ (F) of dimension d − ∣F ∣. It follows that the dimension of lk∆′ (F) is
q − ∣F ∣. Using Proposition 1.5.23 we get
H̃i (lk∆ (F); k) = 0 for i < dim(lk∆ (F)) = d − ∣F ∣,
and consequently
H̃i ((lk∆ (F))q−∣F∣ ; k) = H̃i (lk∆ (F); k) = 0 for i < q − ∣F ∣ ≤ d − ∣F ∣.
28 1. MONOMIAL IDEALS AND FACE RINGS

Observe that the equality between the homology modules follows simply because
the augmented oriented chain complexes of (lk∆ (F))q−∣F∣ and lk∆ (F) are equal up
to degree q − ∣F ∣ and thus their homologies have to agree up to degree one less.
Hence by Reisner criterion ∆q is Cohen-Macaulay. □
EXAMPLE 1.5.26. Let R = k[x1 , . . . , xn ] be a polynomial ring over a field k and
let ∆ be the (p − 2)-skeleton of a simplex of dimension n − 1, where p ≥ 2. Then
I∆ is a Cohen-Macaulay ideal and the face ring of ∆ is:
k[∆] = R/({ xi1 ⋅ ⋅ ⋅ xi p ∣ 1 ≤ i1 < ⋅ ⋅ ⋅ < i p ≤ n}).
The ideal I∆ is the basis monomial ideal of a matroid M and ∆ is a matroid complex
[101, 166].
Given a simplicial complex ∆ and ∆1 , . . . , ∆ s sub complexes of ∆ with vertex
sets V1 , . . . , V s , we define their union
∆′ = ∆1 ∪ ⋅ ⋅ ⋅ ∪ ∆ s
as the simplicial complex with vertex set ∪i=1s V and such that F is a face of ∆′ if
i
and only if F is a face of ∆i for some i. The intersection can be defined similarly.
DEFINITION 1.5.27. A pure simplicial complex ∆ of dimension d is shellable if
the facets of ∆ can be order F1 , . . . , F s such that
⎛ ⎞
∩ i−1 ∪
Fi ⎝ F j⎠
j=1

is pure of dimension d − 1 for all i ≥ 2. Here F i = {σ ∈ ∆∣ σ ⊂ Fi }. If ∆ is

shellable, F1 , . . . , F s is called a shelling.
DEFINITION 1.5.28. A simplicial complex ∆ is constructible if ∆ can be obtained
by the following recursive procedure:
(i) any simplex is constructible,
(ii) if ∆1 , ∆2 are constructible complexes of dimension d, and if ∆1 ∩ ∆2 is
constructible of dimension d − 1, then ∆1 ∪ ∆2 is constructible.
PROPOSITION 1.5.29. If ∆ is a shellable complex of dimension d, then ∆ is
constructible.
PROOF. We proceed by induction on d. If d = 0, then ∆ is a discrete set of
vertices, which is shellable. Assume d > 0 and that ∆ has at least two facets. Let
F1 , . . . , Fr be a shelling of the simplicial complex ∆. Thus the complex
∆m = (F 1 ∪ ⋅ ⋅ ⋅ ∪ F m ) ∩ F m+1
 1. MONOMIAL IDEALS AND FACE RINGS 29

is pure of dimension d − 1 for 1 ≤ m ≤ r − 1. Next we prove that ∆m is shellable.

Consider the set
ℱ = {Fi ∩ Fm+1 ∣ dim(Fi ∩ Fm+1 ) = d − 1; i ≤ m}
= {F`1 ∩ Fm+1 , . . . , F`s ∩ Fm+1 }
where `1 < ⋅ ⋅ ⋅ < ` s < m + 1 and ∣ℱ∣ = s. We claim that a face of the form
F = F`i ∩ F` j ∩ Fm+1
has dimension d − 2 for ` j < `i < m + 1. Clearly dim(F) ≤ d − 2, otherwise one
has F`i ∩ Fm+1 = F` j ∩ Fm+1 , a contradiction because ∣ℱ∣ = s. By Exercise 1.5.50,
there are vi ∈ Fm+1 ∖ F`i and v j ∈ Fm+1 ∖ F` j such that
Fm+1 ∩ F`i ⊂ Fm+1 ∩ Fk1 = Fm+1 ∖ {vi }; (k1 < m + 1),
Fm+1 ∩ F` j ⊂ Fm+1 ∩ Fk2 = Fm+1 ∖ {v j }; (k2 < m + 1).
By dimension considerations we get
Fm+1 ∩ F`i = Fm+1 ∖ {vi }, Fm+1 ∩ F` j = Fm+1 ∖ {v j }.
Hence F = Fm+1 ∖ {vi , v j }, consequently dim(F) is equal to d − 2, as claimed. It
follows readily that ℱ gives a shelling of ∆m because the complex
( )
(F `1 ∩ F m+1 ) ∪ ⋅ ⋅ ⋅ ∪ (F `k ∩ F m+1 ) ∩ (F `k+1 ∩ F m+1 ).
is pure of dimension d − 2 for 1 ≤ k ≤ s − 1. Therefore by induction hypothesis
∆m is constructible for 1 ≤ m ≤ r − 1. Hence, using induction on m, it follows that
the complex
F 1 ∪ ⋅ ⋅ ⋅ ∪ F m ∪ F m+1
is constructible for 1 ≤ m ≤ r − 1, as required. □
LEMMA 1.5.30. (Depth lemma, see [180, Corollary A.6.3]) If 0 → N → M →
L → 0 is a short exact sequence of modules over a local ring R, then
(a) If depth(M) < depth(L), then depth(N) = depth(M).
(b) If depth(M) = depth(L), then depth(N) ≥ depth(M).
(c) If depth(M) > depth(L), then depth(N) = depth(L) + 1.
PROPOSITION 1.5.31. The next implications hold for simplicial complexes:
shellable =⇒ constructible =⇒ Cohen-Macaulay.
PROOF. The first implication is Proposition 1.5.29. The second implication
follows using Exercise 1.5.53 together with the depth lemma. □
Let B and D be two rings. We say that D is a B-algebra if there is a homomor-
phism of rings ϕ : B −→ D. In this case D has a B-module structure (compatible
with its ring structure) given by b ⋅ d = ϕ(b)d for all b ∈ B and d ∈ D.
30 1. MONOMIAL IDEALS AND FACE RINGS

A homomorphism φ : D → C of B-algebras is a map φ which is both a homo-

morphism of rings and a homomorphism of B-modules. Note that φ(a ⋅ 1) = a ⋅ 1
for all a ∈ B.
Let B ⊂ D be a ring extension, i.e., B is a subring of D. Then D has in a natural
way a B-algebra structure induced by the inclusion map. If B = K is a field and D
is a K-algebra, then ϕ is injective, in this case one may always assume that K ⊂ D
is a ring extension.
Given a subset F of D, we denote the subring of D generated by F and B by
B[F]. Recall that B[F] is the intersection of all subrings of D containing B ∪ F. We
call B[F] the B-subalgebra of D spanned by F. If D = B[F] for some finite set F,
we say that D is a finitely generated B-algebra. A finitely generated K-algebra over
a field K is called an affine K-algebra.
DEFINITION 1.5.32. Let k be a field. A standard algebra is a finitely generated
ℕ-graded k-algebra (see Exercise 1.4.7)
S = S 0 ⊕ S 1 ⊕ ⋅ ⋅ ⋅ ⊕ S i ⊕ ⋅ ⋅ ⋅ = k[y1 , . . . , yr ]
such that yi ∈ S 1 for all i and S 0 = k. If we only require yi homogeneous and
deg(yi ) > 0 for all i, we say that S is a positively
⊕ graded k-algebra. The irrelevant
maximal ideal 𝔪 of S is defined by 𝔪 = S + = ∞ i=1 S i .

DEFINITION 1.5.33. Let k be a field and let S = k[y1 , . . . , yr ] be a positively

graded k-algebra with yi homogeneous of degree di . There is a graded epimorphism
ϕ : R = k[x1 , . . . , xr ] −→ S
ϕ
given by f (x1 , . . . , xr ) 7−→ f (y1 , . . . , yr ), where R is a polynomial ring graded by
deg(xi ) = di , the presentation of S is the k-algebra R/ ker(ϕ). The graded ideal
ker(ϕ) is called the ideal of relations or the presentation ideal of S .
PROPOSITION 1.5.34. [187, Proposition 2.2.20] If B, D are affine algebras over
a field k with presentations B ≃ k[x]/I1 and D ≃ k[y]/I2 , then
B ⊗k D ≃ k[x, y]/(I1 + I2 ).
THEOREM 1.5.35 ([75]). If B, D are standard algebras over a field k, then
depth(B ⊗k D) = depth(B) + depth(D).
COROLLARY 1.5.36. If B and D are two standard algebras over a field k, then
B ⊗k D is Cohen-Macaulay if and only if B and D are Cohen-Macaulay.
PROOF. As the dimension is always greater than or equal to the depth by
Lemma 1.5.9, the result follows at once using Theorem 1.5.35 together with Exer-
cise 1.5.56. □
 1. MONOMIAL IDEALS AND FACE RINGS 31

DEFINITION 1.5.37. Let ∆1 , ∆2 be two simplicial complexes with vertex sets V1 ,

V2 . The join ∆1 ∗ ∆2 is the simplicial complex on the vertex set V1 ∪ V2 with faces
F1 ∪ F2 , where Fi ∈ ∆i .
PROPOSITION 1.5.38. If ∆1 and ∆2 are simplicial complexes, then their join
∆1 ∗ ∆2 is Cohen-Macaulay if and only if ∆i is Cohen-Macaulay for i = 1, 2.
PROOF. Since one has the equality I∆1 ∗∆2 = I∆1 + I∆2 , there is a graded isomor-
phism of k-algebras
k[∆1 ∗ ∆2 ] ≃ k[∆1 ] ⊗k k[∆2 ],
see Proposition 1.5.34. To complete the proof use Corollary 1.5.36. □
Gorenstein rings. Let M ∕= (0) be a module over a local ring (R, 𝔪) and let
k = R/𝔪 be the residue field of R. The socle of M is defined as
Soc(M) = (0 : M 𝔪) = {z ∈ M ∣ 𝔪z = (0)},
and the type of M is defined as
type(M) = dimk Soc(M/xM),
where x is a maximal M-sequence in 𝔪. Observe that the type of M is well defined,
see [23]. The ring R is said to be Gorenstein if R is a Cohen-Macaulay ring of type
1. An ideal I ⊂ R is called Gorenstein if R/I is a Gorenstein ring.
REMARK 1.5.39. The following implications hold for a graded ideal of a poly-
nomial ring with coefficients in a field [23]:
complete intersection ⇒ Gorenstein ⇒ Cohen-Macaulay.

Exercises
1.5.40. Prove that a sequence f = f1 , . . . , fr of monomials of R is a regular
sequence if and only if supp( fi ) ∩ supp( f j ) = ∅ for i ∕= j.
1.5.41. Let R be a polynomial ring over a field k. Prove that the family of ideals
of R generated by square-free monomials is a sublattice of the lattice of monomial
ideals.
1.5.42. Prove that a monomial ideal I is generated by square-free monomials if
and only if any of the following conditions hold:
(a) I is an intersection of prime ideals.
(b) rad (I) = I.
1.5.43. Let ∆ be a Cohen-Macaulay simplicial complex and let F ∈ ∆. If lk(F)
is not a discrete set of vertices, prove that lk(F) is connected.
1.5.44. Let ∆ be a simplicial complex on the vertex set V and let I be its
Stanley-Reisner ideal. Take x ∈ V and set J = (I : x).
32 1. MONOMIAL IDEALS AND FACE RINGS

/ i 𝔭i , where Ass(I) = {𝔭1 , . . . , 𝔭r }.

(a) Prove that J = ∩ x∈𝔭
(b) Find a relation between ∆ and ∆ J .
1.5.45. Let R be a polynomial ring over a field k and let I be an ideal minimally
generated by square-free monomials f1 , . . . , fq . If R/I is Cohen-Macaulay and x is
a variable not in I such that x ∈ ∪qi=1 supp( fi ), then R/(I : x) is Cohen-Macaulay.
1.5.46. Let ∆ be a pure simplicial complex on the vertex set V and I = I∆ . Assume
that x is a variable in ∪qi=1 supp( fi ), where the fi are monomials that minimally
generate I. If (I : x) and (I, x) are Cohen-Macaulay, then I is Cohen-Macaulay.
1.5.47. Let ∆ be a simplicial complex and let F ∈ ∆, define the star of the face
F as
star(F) = {G ∈ ∆∣ F ∪ G ∈ ∆}.
Prove that the facets of star(F) are the facets of ∆ that contain F.
1.5.48. Let R = k[x] be a polynomial ring
∏ over a field k and ∆ a simplicial
complex with vertex set x. If F ∈ ∆ and f = xi ∈F xi , prove the equality
k[star(F)] = R/(I∆ : f ).
1.5.49. If ∆ is a Cohen-Macaulay complex and F a face of ∆, then star(F) is a
Cohen-Macaulay complex.
1.5.50. A simplicial complex ∆ is shellable in the nonpure sense of Björner and
Wachs [12] if the facets of ∆ can be listed F1 , . . . , F s such that for all 1 ≤ j < i ≤ s,
there exist some v ∈ Fi ∖ F j and some k ∈ {1, . . . , i − 1} with Fi ∖ Fk = {v}. Prove
that ∆ is shellable in the sense of Definition 1.5.27 if and only if ∆ is pure and
shellable in the sense of Björner and Wachs [12].
1.5.51. Let ∆ be a pure simplicial complex. Prove that ∆ is shellable if and only
if the facets of ∆ can be ordered F1 , . . . , F s such that for all 1 ≤ j < i ≤ s, there is
v ∈ Fi ∖ F j and k < i with Fi ∩ F j ⊂ Fi ∩ Fk = Fi ∖ {v}.
1.5.52. Let R be a ring and I1 , I2 ideals of R. Prove that there is an exact
sequence of R-modules:
ϕ φ
0 −→ R/(I1 ∩ I2 ) −→ R/I1 ⊕ R/I2 −→ R/(I1 + I2 ) −→ 0,
where ϕ(r) = (r, −r) and φ(r1 , r2 ) = r1 + r2 .
1.5.53. Let R = k[x] be a polynomial ring over a field k and let ∆1 , ∆2 be
simplicial complexes whose vertex sets are contained in x = { x1 , . . . , xn }. Prove
that there is an exact sequence of R-modules:
0 −→ k[∆1 ∪ ∆2 ] −→ k[∆1 ] ⊕ k[∆2 ] −→ k[∆1 ∩ ∆2 ] −→ 0.
 1. MONOMIAL IDEALS AND FACE RINGS 33

1.5.54. [145, Remark 3] Let R = k[a, . . . , f ] be a polynomial ring over a field k

and ∆ the following triangulation of the real projective plane ℙ2 :

b s sa

L@ A
A

L@
f A

L @ @s``
L
A ```A
c s
XX As c


L
A ` 
A XXXLs
As

A
A
d JJe




A J

a As J
s b


Show that I∆ = (abc, abd, ace, ad f, ae f, bc f, bde, be f, cde, cd f ). The ring k[∆] is
Cohen-Macaulay if k has characteristic other than 2, and is not Cohen-Macaulay
otherwise.
1.5.55. Let I be an ideal of a ring R and let x ∈ R ∖ I. If I is unmixed and
I = rad (I), then (I : x) is unmixed and ht(I) = ht(I : x).
1.5.56. If B and D are affine algebras over a field k, then
dim(B ⊗k D) = dim(B) + dim(D).
1.5.57. Let I, I1 , . . . , Im be ideals of a ring R and let k be an infinite field. If
I ⊂ ∪m
i=1 Ii and k is a subring of R, then I ⊂ Ii for some i.

1.5.58. Let R = ⊕∞ i=0 Ri be an ℕ-graded ring. Prove that R is a Noetherian ring if

and only if R0 is a Noetherian ring and R = R0 [x1 , . . . , xn ] for some x1 , . . . , xn in R.

1.6. Hilbert functions and Hilbert series

In this section we introduce Hilbert functions and Hilbert series of graded modules.
Then we define two interesting invariants of a standard algebra or module: the degree
or multiplicity and the a-invariant. The second invariant measures the difference
between the Hilbert polynomial and the Hilbert function.
An R0 -module L has finite length if there is a composition series
(0) = L0 ⊂ L1 ⊂ ⋅ ⋅ ⋅ ⊂ Ln = L,
where Li /Li−1 is a nonzero simple module (that is, Li /Li−1 has no proper submodules
other than (0)) for all i. Note that Li /Li−1 must be cyclic and thus isomorphic to
R0 /𝔪, for some maximal ideal 𝔪. The number n is independent of the composition
series and it is called the length of L. This number is usually denoted by `R0 (L) or
simply `(L).
34 1. MONOMIAL IDEALS AND FACE RINGS

DEFINITION 1.6.1. A ring R0 is Artinian if `R0 (R0 ) < ∞.

Let k be an Artinian ring and let R = k[x1 , . . . , xn ] = ⊕∞
i=0 Ri be a finitely
generated ℕ-graded algebra with deg(xi ) = di ∈ ℕ+ and R0 = k. If di =⊕ 1 for all i,
R is called a homogeneous ring. In what follows we assume that M = ∞ i=0 Mi is
a finitely generated ℕ-graded R-module of dimension d. As usual we set Mi = (0)
for i < 0 so that M becomes ℤ-graded. The Hilbert function and Hilbert series of
M are defined by
∞
∑
H(M, i) = `(Mi ) and F(M, t) = H(M, i)ti
i=0
respectively, where `(Mi ) denotes the length of Mi as a k-module. The most
interesting case, for our purposes, occurs when k is field. In this case `(Mi ) equals
the dimension of Mi as a k-vector space.
THEOREM 1.6.2. (Hilbert-Serre; see [3, Theorem 11.1]) The Hilbert series of
M is a rational function that can be written as
h(t)
F(M, t) = n for some h(t) ∈ ℤ[t].
∏
(1 − tdi )
i=1
If di = 1 for all i, then there is a unique polynomial h(t) ∈ ℤ[t] such that
h(t)
F(M, t) = and h(1) ∕= 0.
(1 − t)d
DEFINITION 1.6.3. The degree of F(M, t) as a rational function is denoted by
a(M). This number is called the a-invariant of M.
THEOREM 1.6.4. (Hilbert; see [23, Theorem 4.1.3]) If R is a homogeneous ring,
then there are integers a−d , . . . , a−1 such that
d−1 ( )
∑ i+d− j−1
H(M, i) = a−(d− j) , ∀ i ≥ a(M) + 1.
d− j−1
j=0

COROLLARY 1.6.5. If R is a homogeneous ring and d = dim(M), then there

is a unique polynomial ϕ M (t) ∈ ℚ[t] of degree d − 1 so that H(M, i) = ϕ M (i) for
i ≥ a(M) + 1.
PROOF. It is an immediate consequence of Theorem 1.6.4. □
DEFINITION 1.6.6. ϕ M (t) = cd−1 td−1 + ⋅ ⋅ ⋅ + c1 t + c0 is called the Hilbert poly-
nomial of M. The integer cd−1 (d − 1)! is called the degree or multiplicity of M and
is denoted by e(M). If d = 0, we set e(M) = `(M).
 1. MONOMIAL IDEALS AND FACE RINGS 35

REMARK 1.6.7. The leading coefficient of ϕ M (t) is equal to h(1)/(d − 1)!, where
h(t) is the polynomial F(M, t)(1 − t)d . If d = 0, then
h(1) = `(M) = dimk (M).
COROLLARY 1.6.8. If r0 = min{r ∈ ℕ ∣H(M, i) = ϕ M (i), ∀ i ≥ r}, then r0 = 0 if
a(M) < 0 and r0 = a(M) + 1 otherwise.
If a ∈ ℕ, then M(−a) denotes the regrading of M obtained by a shift of the
graduation of M, more precisely
∞
⊕
M(−a) = M(−a)i ,
i=0
where M(−a)i = M−a+i . Note that we have Mi = (0) for i < 0. In this way M(−a)
becomes an ℕ-graded R-module.
LEMMA 1.6.9. F(M(−a), t) = ta F(M, t).
PROOF. Since M(−a)i = Mi−a one has:
∞
∑
F(M(−a), t) = t a
`(Mi−a )ti−a = ta F(M, t),
i=a
where the first equality follows using that Mi−a = 0 for i = 0, . . . , a − 1. □
LEMMA 1.6.10. Let M be a graded R-module. If z ∈ Ra , then there is a degree
preserving exact sequence
z ϕ
0 −→ (M/(0 : z))(−a) −→ M −→ M/zM −→ 0 (ϕ(m) = m + zM).
PROOF. As the map ψ : M(−a) → M, given by ψ(m) = zm, is a degree zero
homomorphism one has that (0 : z)(−a) is a graded submodule of M(−a). The
exactness of the sequence follows because ψ induces an exact sequence
ı ψ ϕ
0 −→ (0 : z)(−a) −→ M(−a) −→ M −→ M/zM −→ 0. □
EXAMPLE 1.6.11. Let R be a polynomial ring over a field k with the standard
grading. If R has n variables, then
( )
1 m+n−1
F(R, t) = and H(R, m) = .
(1 − t)n n−1
EXAMPLE 1.6.12. Let R = k[x1 , x2 , x3 ] be a polynomial ring and let I be the
ideal
I = (x12 , x22 x3 , x23 ).
Let us compute the Hilbert series of R/I using elimination. Pick any monomial
involving more than one variable, say x22 x3 . The idea is to eliminate x22 from the
36 1. MONOMIAL IDEALS AND FACE RINGS

monomials containing more than one variable. From the exact sequence of graded
modules:
x2
0 −→ R/(x12 , x3 , x2 )(−2) −→
2
R/I −→R/(x12 , x22 ) −→ 0,
and applying Exercise 1.6.26, we get
(1 − t)2 (1 − t2 ) (1 − t2 )2 −t4 + 2t2 + 2t + 1
[ ]
F(R/I, t) = t2 + = .
(1 − t)3 (1 − t)3 (1 − t)
DEFINITION 1.6.13. Let R be a ring and let I be an ideal of R of height g. If
I is generated by a regular sequence with g elements we say that I is a complete
intersection. See Proposition 1.6.14
The following is a consequence of Krull principal ideal theorem (see Theo-
rem 2.2.7).
PROPOSITION 1.6.14 ([3]). Let R be a Noetherian ring and let f = f1 , . . . , fg be
a regular sequence on R, then ht ( f ) = g.
EXAMPLE 1.6.15. Let R = k[x1 , . . . , x5 ] and let I be the ideal
I = (x1 x2 , x3 x4 x5 ).
Note that I is a complete intersection, hence by Exercise 1.6.26 one has
(1 − t2 )(1 − t3 ) 1 + 2t + 2t2 + t3 5t2 − t + 2
F(R/I, t) = = = − 1 + .
(1 − t)5 (1 − t)3 (1 − t)3
Let us illustrate how the integers a−d , . . . , a−1 of Theorem 1.6.4 can be computed.
The Laurent expansion of F(R/I, t) + 1, in negative powers of (1 − t), is given by
5t2 − t + 2 a−3 a−2 a−1
3
= 3
+ 2
+ ,
(1 − t) (1 − t) (1 − t) (1 − t)
where a−1 = 5, a−2 = −9, a−3 = 6. Hence
( ) ( ) ()
i+2 i+1 i
ϕR/I (i) = 6 −9 +5 = 3i2 + 2.
i i i
DEFINITION 1.6.16. Given a simplicial complex ∆ of dimension d its f -vector
is: f (∆) = ( f0 , . . . , fd ), where fi is the number of i-faces of ∆. Note f−1 = 1.
PROPOSITION 1.6.17. If ∆ is a simplicial complex of dimension d and ( fi ) is its
f -vector, then:
d ( )
∑ j−1
H(k[∆], j) = fi , for j ≥ 1 and H(k[∆], 0) = 1,
i
i=0
d
∑ fi zi+1
F(k[∆], z) = .
(1 − z)i+1
i=−1
 1. MONOMIAL IDEALS AND FACE RINGS 37

PROOF. It is left as an exercise. □

h-vectors of standard algebras. Let S be a standard algebra of dimension d
over a field k and let
h(t) = h0 + h1 t + ⋅ ⋅ ⋅ + hr tr
be the (unique) polynomial in ℤ[t] such that h(1) ∕= 0 and satisfying
F(S , t) = h(t)/(1 − t)d ,
see Theorem 1.6.2. The h-vector of S is defined by h(S ) = (h0 , . . . , hr ).
DEFINITION 1.6.18. A set of homogeneous elements y = {y1 , . . . , yd } of S is
called a homogeneous system of parameters (h.s.o.p) if rad (y) = S + .
LEMMA 1.6.19. [187, Lemma 2.2.15] Let k be an infinite field. If S is Cohen–
Macaulay, then there exists a h.s.o.p y = {y1 , . . . , yd } such that y is a regular
sequence and yi is homogeneous of degree 1 for all i.
THEOREM 1.6.20 ([163]). Let y1 , . . . , yd a h.s.o.p for S with ai = deg(yi ) and let
A be the quotient ring S /(y1 , . . . , yd ). Then S is Cohen-Macaulay if and only if
F(S , t) = F(A, t)/(1 − ta1 ) ⋅ ⋅ ⋅ (1 − tad ).
THEOREM 1.6.21. If S is Cohen-Macaulay, then hi ≥ 0 for all i.
PROOF. One may assume k infinite, otherwise one may change the coefficient
field k using the functor (⋅) ⊗k K, where K is an infinite field extension of k. By
Lemma 1.6.19, there is a h.s.o.p y = {y1 , . . . , yd } for S , where each yi is a form in
S of degree one. Set S = S /(y)S . From Theorem 1.6.20 we derive hi = H(S , i) for
all i and consequently hi ≥ 0. □

Exercises
1.6.22. Let k be a field and let D be an affine k-algebra. Prove that D is Artinian
if and only if dimk (D) < ∞.
1.6.23. Let d, m ∈ ℕ. Prove the equality
( ) ∑ m ( )
d+m j+d−1
= .
d d−1
j=0

1.6.24. Let k be a field and let R = k[x1 , . . . , xn ] be a polynomial ring graded by

deg(xi ) = di ∈ ℕ+ . Then the Hilbert series of R is
F(R, t) = 1/(1 − td1 ) ⋅ ⋅ ⋅ (1 − tdn ).
1.6.25. Let R = k[x, y, z] be a polynomial ring over a field k and let I be the ideal
I = (xa , yb , zc , xa1 yb1 zc1 ); (a ≥ a1 , b ≥ b1 , c ≥ c1 ).
Then e(R/I) = dimk (R/I) = abc1 + (c − c1 )(a1 b + (a − a1 )b1 ).
38 1. MONOMIAL IDEALS AND FACE RINGS

1.6.26. Let R = k[x1 , . . . , xn ] be a polynomial ring over a field k graded by

deg(xi ) = di ∈ ℕ+ . If f1 , . . . , fr a homogeneous regular sequence in R and ai =
deg( fi ), then
n
∏ r
) ∏
F(R/( f1 , . . . , fr ), t) 1 − tdi = 1 − tai .
( ( )

i=1 i=1

1.6.27. Let R = k[x1 , x2 , x3 ] and let I = (x12 , x22 x3 , x23 ). Prove that the a-invariant
of R/I is 3 and H(R/I, i) = 4 for i ≥ 4.
1.6.28. Let ∆ be a simplicial complex of dimension d and let ( f0 , . . . , fd ) be its
f -vector. Prove that fd is equal to e(k[∆]), the multiplicity of k[∆].
1.6.29. Let S = k[∆] be the Stanley-Reisner ring of a simplicial complex ∆ with
vertices x1 , . . . , xn and A = S /(x12 , . . . , xn2 ). Prove that ∆ is pure if and only if all the
nonzero elements of the socle of A have the same degree.
1.6.30. If ∆ is a connected simplicial complex having all its vertices of degree
two, then ∆ is a cycle.
1.6.31. Give an example of a monomial Gorenstein ideal which is not a complete
intersection.
1.6.32. Let R = k[x1 , . . . , xn ] be a polynomial ring over a field k and let ∆ be a
simplicial complex with vertices x1 , . . . , xn . Associated to the Stanley-Reisner ideal
I∆ define another ideal
I ′ (∆) = (I∆ , x1 y1 , . . . , xn yn ) ⊂ k[x1 , . . . , xn , y1 , . . . , yn ],
and denote by S (∆) the Stanley–Reisner complex corresponding to this ideal I ′ (∆).
Prove the simplicial complex S (∆) is shellable.
1.6.33. Let R be a polynomial ring over a field k with the standard grading
and let M be a finitely generated ℕ-graded module. If M is Cohen-Macaulay and
h(t) = h0 + h1 t + ⋅ ⋅ ⋅ + hr tr is the unique polynomial in ℤ[t] such that h(1) ∕= 0 and
h(t)
F(M, t) = , where d = dim(M),
(1 − t)d
prove that hi ≥ 0 for all i.
Hint Adapt the proof of Theorem 1.6.21.
1.6.34. The projective space of dimension n over a field k, denoted by ℙnk , is the
quotient space (kn+1 ∖ {0})/ ∼ where two points a, b in kn+1 ∖ {0} are equivalent
if a = λb for some λ in k. Let S = k[x0 , x1 , . . . , xn ] be a polynomial ring over
the field k with the standard grading and let X be a finite subset of ℙnk . If I(X) is
the ideal of homogeneous polynomials of S that vanish on X, prove that S /I(X) is
Cohen-Macaulay, dim S /I(X) = 1, and dim(S /I(X))i = ∣X ∣ for i ≥ ∣X ∣ − 1. Then
conclude that the degree of S /I(X) is equal to ∣X ∣.
 1. MONOMIAL IDEALS AND FACE RINGS 39

1.6.35. Let R be a polynomial ring over an infinite field k with the standard
grading and let I be a graded ideal of R such that R/I is Cohen-Macaulay of
dimension 1. Prove that the Hilbert function of R/I is non-decreasing.
Chapter 2
Graphs, Ideals and Invariants

In this chapter we introduce graphs and take a glance through some basic
notions needed to undertake the study of edge rings. We define some families of
graphs that will play a significant role in later chapters. We begin the study of
ideals generated by square-free monomials of degree two that can be associated to
graphs. There are many common objects of study of graphs and edge rings, for
instance the independence number of a graph is equal to the Krull dimension of the
corresponding edge ring.
In this chapter we start to build a “dictionary” between algebraic invariants of
edge rings and invariants of graphs. We study algebraic properties and invariants
of edge rings using the combinatorial structure of graphs. Although we focus on
graphs, some of the methods work for clutters and hypergraphs. We study Cohen-
Macaulay graphs and present classifications of the following families: unmixed
bipartite graphs, Cohen-Macaulay bipartite graphs, Cohen-Macaulay trees, shellable
bipartite graphs, sequentially Cohen-Macaulay bipartite graphs.
For those readers interested in a comprehensive view of graph theory itself, we
recommend the books [14] and [45], which we take as our main references for the
subject. We refer to [69, 95, 138] for the related theory of facet ideals and edge
ideals associated to simplicial complexes and clutters.

2.1. Graph theory

A graph G is an ordered pair of disjoint finite sets (V, E) such that E is a subset
of the set of unordered pairs of V. The set V is the set of vertices and the set E
is called the set of edges. An edge z = {vi , v j } is said to join the vertices vi and
v j (we will at times also denote an edge by z = vi v j ). The vertices vi and v j are
referred to as the endvertices (or ends) of the edge. When working with several
graphs it is convenient to write V(G) and E(G) for the vertex set and edge set of G
respectively. Sometimes we will also write VG and EG for the vertex set and edge
set of G respectively.
41
42 2. GRAPHS, IDEALS AND INVARIANTS

Let G be a graph on the vertex set V. If z = {vi , v j } is an edge of G one says

that the vertices vi and v j are adjacent or neighbouring vertices of G. In this case it
is also usual to say that the edge z is incident with its endvertices. The degree of a
vertex v in V denoted by deg(v), is the number of edges incident with v. A vertex
with degree zero is called an isolated vertex.
Note that by definition a graph does not contain loops, a pair {v, v} (“an edge
joining a vertex to itself"); neither does it contain a pair {u, v} that occurs several
times (“that is, several edges joining the same two vertices"). If we allow any of
these type of relations as edges we will call G a multigraph. Most results on graphs
carry over to multigraphs in a natural way. There are areas and notions in graph
theory (such as plane duality, and minors) where multigraphs arise more naturally
than graphs. Terminology introduced earlier for graphs can be used correspondingly
for multigraphs.
Given two graphs G and H, a mapping ϕ, from V(G) to V(H) is called a
homomorphism if {ϕ(vi ), ϕ(v j )} ∈ E(H) whenever {vi , v j } ∈ E(G) (so if {vi , v j } is
an edge ϕ(vi ) ∕= ϕ(v j )). Two graphs G and H are isomorphic if there is a bijective map
ψ from V(G) to V(H) such that {vi , v j } ∈ E(G) if and only if {ψ(vi ), ψ(v j )} ∈ E(H).
An isomorphism from G to itself is called an automorphism. A map taking graphs
as arguments is called a graph invariant if it assigns equal values to isomorphic
graphs. The number of vertices and the number of edges are two simple examples
of graph invariants.
Let H and G be two graphs, then H is called a subgraph of G if V(H) ⊂ V(G)
and E(H) ⊂ E(G). A subgraph H is called an induced subgraph if H contains
all the edges {vi , v j } ∈ E(G) with vi , v j ∈ V(H). In this case H is said to be the
subgraph induced by V(H). An induced subgraph is denoted by H = G[V(H)] or
H = ⟨V(H)⟩. It is also denoted by H = GV(H) . A spanning subgraph is a subgraph
H of G containing all the vertices of G.
A walk of length n in G is an alternating sequence of vertices and edges
w = {v0 , z1 , v1 , . . . , vn−1 , zn , vn },
where zi = {vi−1 , vi } is the edge joining vi−1 and vi . A walk may also be written
{v0 , . . . , vn } with the edges understood, or {z1 , z2 , . . . , zn } with the vertices under-
stood. If v0 = vn , the walk w is called a closed walk. A path is a walk with all its
vertices distinct.
We say that G is connected if for every pair of vertices vi and v j there is a path
from vi to v j . Notice that G has a vertex disjoint decomposition
G = G1 ∪ G2 ∪ ⋅ ⋅ ⋅ ∪ Gr
where G1 , . . . , Gr are the maximal (with respect to inclusion) connected subgraphs
of G, the Gi are called the connected components of G. A component is called even
(resp. odd) if its order is even (resp. odd).
 2. GRAPHS, IDEALS AND INVARIANTS 43

Let G be a graph. A vertex v (resp. and edge e) of G is called a cutvertex or

cutpoint (resp. a bridge) if the number of connected components of G ∖ {v} (resp.
G ∖ {e}) is larger than that of G. A maximal connected subgraph of G without cut
vertices is called a block. A graph G is 2-connected if ∣V(G)∣ > 2 and G has no cut
vertices. Thus a block of G is either a maximal 2-connected subgraph or a bridge.
By their maximality, different blocks of G intersect in at most one vertex, which
is then a cutvertex of G. Therefore every edge of G lies in a unique block, and G
is the union of their blocks. The technique of decomposing a graph into blocks is
particularly useful when studying edge subrings of bipartite graphs and complete
intersection toric ideals, see Sections 5.2 and 5.5.
A cycle of length n is a closed path {v0 , . . . , vn } in which n ≥ 3. A cycle is even
(resp. odd) if its length is even (resp. odd). We denote by Cn the graph consisting
of a cycle with n vertices, C3 will be called a triangle, C4 a square and so on. If all
the vertices of G are isolated, G is called a discrete graph. A forest is an acyclic
graph and a tree is a connected forest. The complete graph 𝒦n has every pair of its
n vertices adjacent. For instance:
s x3 xs4 s x3
@ @ @
@sx4 xs5
@
x2 s @ s x4
@
s @s x2 s
@
@ x3 A x5
@ A


@s x1 x2 As
s x1 s x1
𝒦4 C5 Tree

The chromatic number of a graph G is the least integer r such that there exists
a homomorphism from G to 𝒦r . A colouring of the vertices of a graph G is an
assignment of colours to the vertices of G in such a way that adjacent vertices have
distinct colours. Thus the chromatic number of G, denoted by χ(G), is the minimal
number of colours in a colouring of G. A clique of a graph G is a set of vertices
that induces a complete subgraph. The clique number of G, denoted by ω(G), is the
size of the largest complete subgraph of G. The clique number and the chromatic
number are related by the inequality
ω(G) ≤ χ(G).
A graph is perfect if for every induced subgraph H, the chromatic number of H
equals the size of the largest complete subgraph of H, i.e., ω(H) = χ(H) for every
induced subgraph of G. An excellent reference for the theory of perfect graphs is
the book of Golumbic [86].
A graph G is bipartite if its vertex set V(G) can be partitioned into two disjoint
subsets V1 and V2 such that every edge of G has one end in V1 and one end in V2 .
The pair (V1 , V2 ) is called a bipartition of G. If G is connected such a bipartition is
uniquely determined.
44 2. GRAPHS, IDEALS AND INVARIANTS

A graph G is a complete bipartite graph if G is bipartite and we have that V1

and V2 are completely joined. If V1 and V2 have m and n vertices respectively, we
denote such a complete bipartite graph by 𝒦m,n . A star is a complete bipartite graph
of the form 𝒦1,n . For instance:
s s s
V1 H@H @ 

@ HH @

@H@
V2 s @s H@
Hs
𝒦3,3
DEFINITION 2.1.1. The distance d(u, v) between two vertices u and v of a graph
G is defined to be the minimum of the lengths of all possible paths from u to v. If
there is no path joining u and v, then d(u, v) = ∞.
PROPOSITION 2.1.2. A graph G is bipartite if and only if all the cycles of G are
even.
PROOF. ⇒) Let V1 and V2 be a bipartition of V(G) such that every edge of G
joins a vertex of V1 with a vertex of V2 . If {v0 , . . . , vn } is a cycle of G, one may
assume v0 ∈ V1 . Since v1 ∈ V2 , it follows at once that vi is in V1 if and only if i is
even, thus n must be even.
⇐) It suffices to prove that each connected component of G is bipartite, thus
one may assume G connected. Pick a vertex v0 ∈ V. Set
V1 = {v ∈ V(G)∣d(v, v0 ) is even} and V2 = V(G) ∖ V1 .
It follows that no two vertices of Vi are adjacent for i = 1, 2, otherwise G would
contain an odd cycle. Therefore G is bipartite. □
If z is a edge, we denote by G ∖ {z} the spanning subgraph of G obtained by
deleting z and keeping all the vertices of G. If u, v ∈ V(G) are nonadjacent vertices
of G then we denote by G + {u, v} the graph obtained from G by adding the edge
joining u and v. The removal of a vertex v from a graph G results in a subgraph
G ∖ {v} of G consisting of all the vertices in G except v and all the edges not incident
with v. For instance:
s x3 s x3
r x3 @ @
@ @ @
x2 r G z @@r x4 x2 s @s x4 x2 s @s x4
G ∖ { x1 } @ G ∖ {z}
@@ @
@ r x1 @s x1
If A is a set of vertices of G one can define G ∖ A recursively by removing one
vertex at a time.
DEFINITION 2.1.3. A set of edges in a graph G is called independent or a
matching if no two of them have a vertex in common.
 2. GRAPHS, IDEALS AND INVARIANTS 45

DEFINITION 2.1.4. Let A be a set of vertices of a graph G. The neighbor set of

A, denoted by NG (A) or simply by N(A) if G is understood, is the set of vertices of
G that are adjacent with at least one vertex of A.
PROPOSITION 2.1.5. Let G be a bipartite graph with bipartition (V1 , V2 ) and let
m and n be the number of vertices in V1 , V2 respectively. If ∣A∣ ≤ ∣N(A)∣ for all
A ⊂ V1 , then there are m independent edges in G.
PROOF. By induction on m. If m = 1, then there is at least one vertex in V2
connected to V1 , thus there is one independent edge. Assume m ≥ 2.
Case (I): First assume ∣A∣ < ∣N(A)∣ for all A ⊂ V1 with ∣A∣ < m. Let x1 ∈ V1
and y1 ∈ V2 be two adjacent vertices. Consider the subgraph H = G ∖ { x1 , y1 }
obtained from G by removing x1 and y1 . Note that for every A ⊂ V1 ∖ { x1 } one has
NH (A) = N(A) if y1 ∈ / N(A) and NH (A) = N(A)∖{y1 } otherwise. Hence by induction
there are z2 , . . . , zm independent edges in H, which together with z1 = { x1 , y1 } yield
the required number of independent edges.
Case (II): Next assume ∣A∣ = ∣N(A)∣ for some A ⊂ V1 with ∣A∣ < m. Set r = ∣A∣.
Consider the subgraph H = G ∖ (V1 ∖ A). Since NH (S ) = N(S ) for all S ⊂ A, by
induction there are independent edges z1 , . . . , zr in H.
On the other hand consider F = G ∖ (A ∪ N(A)). If S ⊂ V1 ∖ A, then
∣N(A ∪ S )∣ ≥ ∣A ∪ S ∣ = ∣S ∣ + ∣A∣,
since N(S ∪ A) = N(A) ∪ NF (S ) and r = ∣N(A)∣ we get ∣NF (S )∣ ≥ ∣S ∣. Applying
induction there are w1 , . . . , wm−r independent edges in F. Putting altogether it is
seen that there are m independent edges in G, as required. This proof is due to
Halmos and Vaughn. □
COROLLARY 2.1.6. Let G be a bipartite graph with bipartition (V1 , V2 ). If there
is an integer p ≥ 0 such that ∣A∣ − p ≤ ∣N(A)∣ for all A ⊂ V1 , then there are m − p
independent edges in G, where m = ∣V1 ∣.
PROOF. By Proposition 2.1.5 the assertion holds for p = 0. Consider the graph
F obtained from G by adding a new vertex y to V2 and joining every vertex of V1 to
y. Let A ⊂ V1 , since NF (A) = N(A) ∪ {y} one concludes ∣NF (A)∣ ≥ ∣A∣ − (p − 1).
Hence by induction there are m − p + 1 independent edges in F, it follows that there
are m − p independent edges in G. □
Let G be a graph with vertex set V. A subset C ⊂ V is a minimal vertex cover
for G if: (i) every edge of G is incident with at least one vertex in C, and (ii) there
is no proper subset of C with the first property. If C satisfies condition (i) only,
then C is called a vertex cover of G and C is said to cover all the edges of G. The
notion of a minimal vertex cover of a clutter or hypergraph is defined similarly (see
Definition 4.6.2).
The following example illustrates the notion of a minimal vertex cover:
46 2. GRAPHS, IDEALS AND INVARIANTS

qP
x5 P  q x3
BB PPxd q
4 
 C = { x1 , x2 , x4 , x6 }
B @@  is a minimal vertex cover of G.
Bd
q @d q
x6 A G
x2
A

x7Aq d
q x1

It is convenient to regard the empty set as a minimal vertex cover for a graph with
all its vertices isolated. A set of vertices of G is called independent or stable if no
two of them are adjacent. Notice the following duality: a set of vertices in G is a
maximal independent set (with respect to inclusion) if and only if its complement
is a minimal vertex cover for G.
DEFINITION 2.1.7. The vertex covering number of G, denoted by α0 (G), is the
number of vertices in a minimum vertex cover in G (the cardinality of any smallest
vertex cover in G).
The edge independence number of a graph G, denoted by β1 (G), is the number
of edges in a maximum independent set of edges in G (the size of any largest
independent set of edges in G).
The vertex independence number of a graph G, denoted by β0 (G), is the number
of vertices in a maximum independent set (the cardinality of any largest independent
set in G).
THEOREM 2.1.8 (König). If G is a bipartite graph, then β1 (G) = α0 (G).
PROOF. Let V(G) = V1 ∪ V2 be a partition of the vertex set of G such that every
edge of G joins a vertex of V1 with a vertex of V2 . Set r = β1 (G) and m = ∣V1 ∣.
Since there are r independent edges in G one obtains α0 (G) ≥ r.
Now there must exist a subset A ⊂ V1 such that ∣N(A)∣ ≤ ∣A∣ − (m − r),
otherwise if ∣N(A)∣ > ∣A∣ − (m − r) for all A ⊂ V1 , then by Corollary 2.1.6 one
would have r + 1 independent edges which is impossible. Therefore N(A) ∪ (V1 ∖ A)
is a vertex cover of G with at most r elements, thus we get α0 (G) ≤ r. □
A pairing by an independent set of edges of all the vertices of a graph G is
called a perfect matching. Thus G has a perfect matching if and only if G has an
even number of vertices and there is a set of independent edges containing all the
vertices.
Next we present a criterion for the existence of perfect matchings; see Proposi-
tion 7.6.8 for a generalization of the marriage theorem.
THEOREM 2.1.9 (Marriage theorem). If G = (V, E) is a bipartite graph with
bipartition (V1 , V2 ), then the following are equivalent
(a) G has a perfect matching.
(b) ∣A∣ ≤ ∣N(A)∣ for all A ⊂ V independent set of vertices.
 2. GRAPHS, IDEALS AND INVARIANTS 47

PROOF. Set r = β1 (G). It is clear that (a) implies (b).

(b) ⇒ (a): Since Vi is an independent set for i = 1, 2, one readily derives
∣V1 ∣ = ∣V2 ∣. By König theorem there are z1 , . . . , zr independent edges and a minimal
vertex cover A of G with r elements, hence zi ∩ A has exactly one vertex for any
i and A is an independent set. Using the fact that V ∖ A is an independent set of
vertices and from the equality N(V ∖ A) = A we conclude that
∣V ∖ A∣ ≤ ∣N(V ∖ A)∣ = ∣A∣.
It follows that ∣V1 ∣ = r, and thus z1 , . . . , zr yields a perfect matching. □
What happens if G is not bipartite? The answer is given by the following
theorem of Tutte.
THEOREM 2.1.10. (Tutte’s Theorem; see [45, Theorem 2.2.1]) A graph G has
a perfect matching if and only if c0 (G ∖ S ) ≤ ∣S ∣ for all S ⊂ V(G), where c0 (G)
denotes the number of odd components (components with an odd number of vertices)
of a graph G.
Clutters and hypergraphs. Many notions and results in graph theory admit a
natural generalization to hypergraphs and in particular to clutters. See for instance
[9], [38] and [151].
DEFINITION 2.1.11. Let V = { x1 , . . . , xn }. A hypergraph ℋ is a pair (V, E) such
that E is a subset of the set of all subsets of V. The elements of E = {E1 , . . . , Eq }
are called edges and the elements x1 , . . . , xn are called vertices.
DEFINITION 2.1.12. A clutter 𝒞 is a hypergraph, such that Ei ⊂ E j ⇒ i = j. In
particular if ∣Ei ∣ = 2 for all i, then 𝒞 is a graph.
DEFINITION 2.1.13. Let 𝒞 be a clutter with vertex set V = { x1 , . . . , xn } and let
E1 , . . . , Eq be the edges of 𝒞 . The incidence matrix of 𝒞 is the n × q matrix A = (ai j )
given by ai j = 1 if xi ∈ E j and ai j = 0 otherwise.
Directed graphs. A directed graph or digraph 𝒟 consists of a finite set V(𝒟)
of vertices together with a prescribed collection E(𝒟) of ordered pairs of distinct
points called edges or arrows. An oriented graph is a digraph having no symmetric
pair of directed edges. In other words an oriented graph is a graph together with
an orientation of its edges. A tournament is a complete oriented graph. Any
tournament has a spanning directed path according to [96].

Exercises
2.1.14. If G is a connected bipartite graph, prove that G has a unique bipartition.
2.1.15. A cutpoint of a graph is a vertex whose removal increases the number of
connected components. If u, v are two vertices at maximum distance in a connected
graph G, then u, v are not cutpoints.
48 2. GRAPHS, IDEALS AND INVARIANTS

2.1.16. If G is a regular bipartite graph (all vertices have the same degree), then
G has a perfect matching.
2.1.17. An edge cover in a graph G is a set of edges collectively incident with
each vertex of G. The edge covering number of G denoted by α1 (G), is the number
of edges in a minimum edge cover in G (the cardinality of any smallest edge cover
in G). Prove:
(a) For any graph G, β0 (G) + α0 (G) = ∣V(G)∣. If G has no isolated vertices,
β1 (G) + α1 (G) = ∣V(G)∣.
(b) A minimal edge cover is minimum if and only if it contains a maximum
matching.
(c) A maximal matching is maximum if and only if it is contained in a mini-
mum edge cover.
(d) For any graph G, β1 (G) ≤ α0 (G). If G is bipartite, α1 (G) = β0 (G).
2.1.18. Let G be a bipartite graph with minimum degree r. Then G is the union
of r edge disjoint edge covers.
2.1.19. A graph G contains a set of independent edges covering all but at most
d of the vertices if and only if c0 (G ∖ S ) ≤ ∣S ∣ + d for all S ⊂ V(G), where
c0 (G) denotes the number of odd components (components with an odd number of
vertices) of a graph G.
2.1.20. Define a 2-coloration of a set A as a surjective mapping from A onto a
set of two elements. A 2-coloration of a graph is a 2-coloration of its vertex set.
Prove that a graph is connected if for any 2-coloration of the graph there exists an
heterochromatic edge.
2.1.21. An Euler trail in a graph G is a walk that contains every edge exactly
once. An Euler circuit is a closed Euler trail. Let G be a graph with q edges. Prove
that: (i) an Euler trail is an edge bijective homomorphism ϕ : Pq −→ G where Pq
denotes the path with q + 1 vertices and q edges, and (ii) an Euler circuit is an edge
bijective homomorphism ϕ : Cq −→ G, where Cq denotes a q-cycle.
2.1.22. If G = (V1 ∪ V2 , E) is bipartite graph, prove that the mapping from V(G)
to V(𝒦2 ) that sends all the vertices from Vi to vertex i is a homomorphism from G
to 𝒦2 .
2.1.23. Let G be a bipartite graph with bipartition (A, B), ∣A∣ = a, ∣ B∣ = b and
let c ≤ a and d ≤ b be positive integers. Assume G has at most (a − c)(b − d)/d
edges. Then there exists C ⊂ A and D ⊂ B such that ∣C ∣ = c, ∣D∣ = d and C ∪ D is
an independent set in G.
2.1.24. Prove that the clique number and the chromatic number of a graph G
satisfy ω(G) ≤ χ(G).
 2. GRAPHS, IDEALS AND INVARIANTS 49

2.1.25. Let G be a perfect connected graph with vertex set V(G). Then there
are cliques K1 , . . . , K s of G such that V(G) is the disjoint union of K1 , . . . , K s and
α0 (G) = α0 (⟨K1 ⟩) + ⋅ ⋅ ⋅ + α0 (⟨K s ⟩).
2.1.26. Prove that any tournament has a spanning directed path.

2.2. Edge ideals and special types of graphs

Let G be a graph with vertices v1 , . . . , vn and let R = k[x1 , . . . , xn ] be a polynomial
ring over a field k, with one variable xi for each vertex vi , we will often identify the
vertex vi with the variable xi .
DEFINITION 2.2.1. The edge ideal I(G) associated to the graph G is the ideal of
R generated by the set of square-free monomials xi x j such that vi is adjacent to v j ,
that is,
I(G) = ({ xi x j ∣ {vi , v j } ∈ E(G)}) ⊂ R .
If all the vertices of G are isolated we set I(G) = (0). The ring R/I(G) is called the
edge ring of G.
Edge ideals are algebraic representations of graphs. These ideals occur in
computational chemistry [126], Physics [16], and combinatorics [166]. The non
zero edge ideals are precisely those ideals generated by square-free monomials of
degree two.
Given a poset P = (V, ⪯) with vertex set V = { x1 . . . . , xn } its order complex,
denoted by ∆(P), is the simplicial complex on V whose faces are the chains (linearly
ordered sets) in P, thus
k[∆(P)] = k[V]/(xi x j ∣ xi ∕∼ x j )
is the Stanley-Reisner ring of ∆(P). Here xi ∕∼ x j means that xi is not comparable to
x j . The Cohen-Macaulay property of order complexes of posets has been extensively
studied, see [23, 166]. Those simplicial complexes that occur as order complexes
have been nicely characterized by Stanley [164].
The next result establishes a one to one correspondence between the minimal
vertex covers of a graph and the minimal primes of the corresponding edge ideal.
PROPOSITION 2.2.2. Let R = k[x1 , . . . , xn ] be a polynomial ring over a field k
and let G be a graph with vertices x1 , . . . , xn . If 𝔭 is an ideal of R generated by
C = { xi1 , . . . , xir }, then 𝔭 is a minimal prime of I(G) if and only if C is a minimal
vertex cover of G.
PROOF. Note that I(G) ⊂ 𝔭 if and only if C is a vertex cover of G. Assume
that C is a minimal vertex cover of G, hence I(G) ⊂ 𝔭. By Proposition 1.2.6 any
50 2. GRAPHS, IDEALS AND INVARIANTS

minimal prime of I(G) is a face ideal, thus 𝔭 is a minimal prime of I(G). The
converse is clear. □
An analogous result holds for edge ideals of hypergraphs and clutters (see
Exercise 2.2.15).
COROLLARY 2.2.3. If G is a graph and I(G) is its edge ideal, then the vertex
covering number α0 (G) is equal to the height of the ideal I(G).
PROOF. It follows at once from Proposition 2.2.2. □
DEFINITION 2.2.4. An edge e of a graph G is critical if
α0 (G ∖ e) < α0 (G).
If all the edges in G are critical G is called edge-critical . A vertex v of G is critical
if α0 (G ∖ v) < α0 (G), and G is called vertex-critical if all its vertices are critical.
PROPOSITION 2.2.5. Let G be a graph. If α0 (G ∖ v) < α0 (G) for some vertex v
of G, then α0 (G ∖ v) = α0 (G) − 1.
PROOF. Set r = α0 (G ∖ v) and note that r + 1 ≤ α0 (G). Pick a minimal vertex
cover C of G ∖ v with r vertices. Since α0 (G ∖ v) < α0 (G), v is not an isolated vertex
and N(v) ∕= C. Thus C ∪ {v} is a minimal vertex cover of G. Hence α0 (G) ≤ r + 1
and one has the asserted equality. □
PROPOSITION 2.2.6. Let G be a graph without isolated vertices. Then G is vertex
critical if and only if any of the following equivalent conditions hold.
(a) Any vertex of G is in some minimal vertex cover of G with α0 (G) vertices.
(b) α0 (G) = α0 (G ∖ v) + 1 for any vertex v of G.
(c) β0 (G) = β0 (G ∖ v) for any vertex v of G.
PROOF. It follows rapidly from Proposition 2.2.5. □
THEOREM 2.2.7. (Krull principal ideal theorem; see [3, Corollary 11.17]) Let
I be an ideal of a ring R generated by a sequence h1 , . . . , hr . Then
(a) ht (𝔭) ≤ r for any minimal prime 𝔭 of I.
(b) If h1 , . . . , hr is a regular sequence, then ht (𝔭) = r for any minimal prime 𝔭
of I.
A ring R is a catenary ring if for every pair 𝔭 ⊂ 𝔮 of prime ideals ht (𝔮/𝔭) is
equal to the length of any maximal chain of prime ideals between 𝔭 and 𝔮. If R is a
domain, then R is catenary if and only if ht (𝔮/𝔭) = ht (𝔮) − ht (𝔭) for every pair of
prime ideals 𝔭 ⊂ 𝔮.
THEOREM 2.2.8. [23, Theorem 2.1.12] If R is a Cohen-Macaulay ring, then R
is catenary. In particular a polynomial ring k[x1 , . . . , xn ] over a field k is catenary.
 2. GRAPHS, IDEALS AND INVARIANTS 51

PROPOSITION 2.2.9. Let I be an ideal of a catenary domain R and let x ∈ R. If

ht (I) < ht (I, x), then ht (I, x) = ht (I) + 1.
PROOF. Let 𝔭 be a minimal prime of I such that ht (I) = ht (𝔭). Note that the
image of (x) + 𝔭 in R/𝔭 is a principal ideal, hence by Krull’s principal ideal theorem
(see Theorem 2.2.7) there is a minimal prime ideal 𝔮 of (x)+ 𝔭 such that ht (𝔮/𝔭) ≤ 1.
As R is a catenary domain one has
ht (I, x) − ht (I) ≤ ht (𝔮) − ht (𝔭) = ht (𝔮/𝔭) ≤ 1.
Therefore ht (I, x) ≤ ht (I) + 1, and the required equality follows. □
COROLLARY 2.2.10 ([96]). Let G be a graph. If e is an edge of G such that
α0 (G ∖ e) < α0 (G), then α0 (G) = α0 (G ∖ e) + 1.
PROOF. It follows from Corollary 2.2.3 and Proposition 2.2.9. □
PROPOSITION 2.2.11. Let G be a graph with n vertices and let I(G) be its edge
ideal. Then
n = α0 (G) + β0 (G) = ht(I) + dim R/I(G).
In particular β0 = dim R/I(G).
PROOF. It follows from Exercise 2.2.15. □
DEFINITION 2.2.12. A graph G is said to be an unmixed graph if any two minimal
vertex covers of G have the same cardinality.
DEFINITION 2.2.13. A graph G is said to be Cohen-Macaulay over the field k
(C-M graph for short) if R/I(G) is a Cohen-Macaulay ring.
PROPOSITION 2.2.14. If G is a Cohen-Macaulay graph, then G is unmixed.
PROOF. By Corollary 1.2.8 I(G) is the intersection of its minimal primes.
Therefore I(G) is an unmixed ideal by Corollary 1.5.14, thus G is unmixed by the
correspondence between minimal covers and minimal primes. □

Exercises
2.2.15. Let G be a hypergraph with vertex set X = { x1 , . . . , xn } and let R = k[X]
be a polynomial ring over a field k. The edge ideal of G is the square-free monomial
ideal: ( )
∏
I = I(G) = xi e is an edge of G ⊂ R.
xi ∈e
A set of vertices is called independent if it does not contains any edge of G. The
notion of a minimal vertex cover of a hypergraph is defined as for the case of graphs.
Let C ⊂ X. Prove that the following are equivalent:
52 2. GRAPHS, IDEALS AND INVARIANTS

(a) C is a minimal vertex cover of G.

(b) X ∖ C is a maximal independent set.
(c) The face ideal 𝔭 = (C) ⊂ R generated by C is a minimal prime of I(G).
(d) X ∖ C is a maximal face of the Stanley-Reisner complex of I(G).
2.2.16. Let G be a graph. Construct a graph G′ by appending an isolated vertex
x to G. Prove that R/I(G)[x] = R[x]/I(G′ ), α0 (G) = α0 (G′ ) and β0 (G′ ) = β0 (G) + 1.
2.2.17. Let P = (V, ⪯) be a poset with vertex set V = { x1 , . . . , x8 } ordered by the
next inclusion diagram. Prove that the Stanley-Reisner ring of the order complex
∆(P) is k[V]/(x6 x7 , x2 x3 ).
r x8
x7 r @ r x6
@
@ @ r x5
r x4
x3 r @ @ r x2
@ @ r x1

2.2.18. (Dilworth’s theorem) Using König’s theorem, prove that in any finite
poset the cardinality of any largest antichain equals the cardinality of any smallest
chain partition.
2.2.19. Prove that in a finite poset the cardinality of any longest chain equals
the cardinality of any smallest antichain partition.

2.3. B-graphs and bounds for Krull dimension

Here we will compare several families of graphs and examine some of their numer-
ical invariants. At the center lies the following notion.
DEFINITION 2.3.1. A family 𝒞 consisting of independent
∪ sets of a graph G is
said to be a maximal independent cover of G if V(G) = C∈𝒞 C and ∣C ∣ = β0 (G)
for all C ∈ 𝒞 . A graph G without isolated vertices having a maximal independent
cover is called a B-graph [8].
PROPOSITION 2.3.2 ([8]). If G is a B-graph, then G is vertex critical.
PROOF. Let v be a vertex of G and let e = {v, w} be an edge of G. Since w is
contained in an independent set A of G with β0 (G) vertices, then v is in V(G) ∖ A.
Hence α0 (G ∖ {v}) < α0 (G). □
PROPOSITION 2.3.3 ([8]). If G is an edge-critical graph or an unmixed graph,
then G has a maximal independent cover.
 2. GRAPHS, IDEALS AND INVARIANTS 53

PROOF. First assume G is edge-critical. Let v be a vertex of G. If v is isolated,

then v belongs to any maximal independent set. On the other hand if v is not
isolated, then there is an edge x = {u, v} ∈ E(G). Since G is edge-critical, then
β0 (G∖ x) = β0 (G) + 1. Hence there is a maximal independent set M with β0 (G) + 1
elements and such that u, v ∈ M. It follows that M ∖{u} is a maximal independent
set of G with β0 (G) elements. If G is unmixed, then the result follows readily
because any two maximal independent sets have the same cardinality. □

Let us summarize some of the results obtained thus far. The following implica-
tions hold for any graph without isolated vertices:
edge-critical =⇒
(α-edge-critical)
B-graph =⇒ vertex-critical
(τ-vertex-critical).
unmixed =⇒
C-M =⇒ (well-covered)
The words in parenthesis indicate the corresponding terminology in [8]. In loc. cit.
the integer α0 (G) (resp. β0 (G)) is denoted by τ(G) (resp. α(G)) and is called the
transversal number (resp. stability number) of G.
PROPOSITION 2.3.4. If G is a vertex critical graph and G ∖ {v} is not vertex
critical for all v ∈ V(G), then G has a maximal independent cover.
PROOF. We use induction on the number of vertices of G. If G has two vertices
or α0 (G) = 1, then for both cases the assumptions on G imply that G has a single
edge and the result is clear. Thus one may assume G has at least three vertices and
α0 (G) ≥ 2.
Assume G has no maximal independent cover. Let C1 , . . . , Cr be the set of all
minimal vertex covers of G with α0 (G) vertices. Pick a vertex v ∈ V(G) such that v
is in Ci for all i. By hypothesis
V(G) = C1 ∪ ⋅ ⋅ ⋅ ∪ Cr
because G is vertex-critical (Proposition 2.2.6). Hence degG (x) ≥ 2 for all x ∈ V(G)
adjacent to v, indeed if degG (x) = 1, then x ∈ C j for some j and C j ∖ { x} is a vertex
cover of G which is impossible. Thus G ∖ {v} has no isolated vertices. Since
α0 (G ∖ {v}) is equal to α0 (G) − 1, the set Ci ∖ {v} is a minimal vertex cover of
G ∖ {v} with α0 (G ∖ {v}) vertices for all i. Thus, from the equality
V(G) ∖ {v} = (C1 ∖ {v}) ∪ ⋅ ⋅ ⋅ ∪ (Cr ∖ {v})
we conclude that G ∖ {v} is vertex critical, a contradiction. □
54 2. GRAPHS, IDEALS AND INVARIANTS

REMARK 2.3.5. Let G be a graph and let A be the intersection of all minimal
vertex covers of G with α0 (G) vertices. Note that G has a maximal independent
cover if and only if A = ∅.
PROPOSITION 2.3.6. Let G be a graph and let A be the intersection of all the
minimal vertex covers of G with α0 (G) vertices. If G is vertex critical, then the graph
G ∖ A is vertex-critical, has a maximal independent cover, and β0 (G) = β0 (G ∖ A).
PROOF. Let A = {v1 , . . . , v s }. If A = ∅, then G has a maximal independent
cover and there is nothing to prove. Assume A ∕= ∅. By the arguments in the proof
of Proposition 2.3.4 the graph G ∖ {v1 } is vertex-critical and β0 (G) = β0 (G ∖ {v1 }).
Since the intersection of all the minimal vertex covers of G ∖ {v1 } with α0 (G ∖ {v1 })
vertices is equal to A ∖ {v1 }, the result follows by induction. □
DEFINITION 2.3.7. Let G be a graph with vertex set V(G) and let H and K be
subsets of V(G). We say that H is K-isolated if v is not adjacent to w for all v ∈ H
and w ∈ K.
The following examples illustrate this notion: (a) every vertex v ∈ V(G) satisfies
that {v} is {v}-isolated, (b) a vertex v ∈ V(G) is isolated if and only if {v} is V(G)-
isolated, (c) if v is a vertex of G and K is an independent set of G containing v, then
v is K-isolated, and (d) a subset H of V(G) is an independent set if and only if H is
H-isolated; moreover an independent set H of G is maximal if and only if there is
no H ⊊ K such that H is K-isolated.
THEOREM 2.3.8. If G is a B-graph, then β0 (G) ≤ α0 (G).
PROOF. Let n be the number of vertices of G. Assume β0 (G) > α0 (G). To
derive a contradiction the idea is to give a recursive procedure to construct a set
H of isolated vertices of G satisfying ∣H ∣ ≥ n − 2k, where k = α0 (G). Let M1
be a maximal independent set with n − k elements. We set H1 = K1 = M1 and
∣K1 ∣ = n − s1 , where s1 = k. Clearly H1 is K1 -isolated and ∣H1 ∣ ≥ n − 2k + s1 . For
r ≥ 1 assume we have constructed M1 , . . . , Mr independent sets of G with n − k
elements such that the sets
Hi = M1 ∩ ⋅ ⋅ ⋅ ∩ Mi , and
Ki = M1 ∪ ⋅ ⋅ ⋅ ∪ Mi (i = 1, . . . , r)
satisfy the following conditions:
(ar ) ∣K1 ∣ = n − s1 < ∣K2 ∣ = n − s2 < ⋅ ⋅ ⋅ < ∣Kr ∣ = n − sr ,
(br ) ∣Hi ∣ ≥ n − 2k + si for 1 ≤ i ≤ r, and
(cr ) Hi is Ki -isolated for 1 ≤ i ≤ r.
If Kr = V(G), then by (cr ) the set Hr is an isolated set of vertices and the procedure
stops. To continue with the procedure assume Kr ⊊ V(G). There is a vertex
 2. GRAPHS, IDEALS AND INVARIANTS 55

w ∈ V(G) ∖ Kr and there is an independent set Mr+1 with n − k vertices containing

w. Set Hr+1 = Hr ∩ Mr+1 and Kr+1 = Kr ∪ Mr+1 . One can write
∣Kr+1 ∣ = n − sr+1 ,
for some 0 < sr+1 < sr . Hence since Kr+1 = Kr ∪ (Mr+1 ∖ Kr ) one has
(2.2) ∣ Mr+1 ∖ Kr ∣ = ∣Kr+1 ∣ − ∣Kr ∣ = sr − sr+1 .
Next we claim that the following inequality holds:
(2.3) ∣ Mr+1 ∩ (Kr ∖ Hr )∣ ≤ k − sr ,
otherwise there would be an independent set T ⊂ Kr ∖ Hr with ∣T ∣ = k − sr + 1,
hence T ∪ Hr is an independent set with at least n − k + 1 elements, a contradiction
because β0 (G) = n − k. From the equality
Mr+1 = (Mr+1 ∩ Hr ) ∪ (Mr+1 ∩ (Kr ∖ Hr )) ∪ (Mr+1 ∖ Kr )
and using Eq. (2.2) and Eq. (2.3) we get
n−k = ∣Hr+1 ∣ + ∣ Mr+1 ∩ (Kr ∖ Hr )∣ + ∣ Mr+1 ∖ Kr ∣
≤ ∣Hr+1 ∣ + (k − sr ) + (sr − sr+1 ).
Hence ∣Hr+1 ∣ ≥ n − 2k + sr+1 . Observe that Hr+1 is Kr+1 -isolated. To complete the
proof observe that since Ki is strictly contained in Ki+1 for each i, then Km = V(G)
for some m and consequently H = Hm is the required set of isolated vertices. □
Let G be a graph. In what follows let ∆G be the Stanley-Reisner complex of
I(G) and let k[∆G ] be the Stanley-Reisner ring of ∆G over a field k. Notice that the
faces of ∆G are precisely the independent sets of G.
THEOREM 2.3.9 ([61]). If G is a vertex-critical graph with n vertices, then
⌊n⌋
β0 (G) = dim R/I(G) = dim k[∆G ] ≤ .
2
PROOF. By induction on n. The case n = 2 is clear. Assume n ≥ 3. One
may assume that G has no maximal independent cover, otherwise the result follows
directly from Theorem 2.3.8 and α0 (G) + β0 (G) = n. By Proposition 2.3.4 there
is v ∈ V(G) such that G ∖ {v} is vertex-critical. Thus by the induction hypothesis
β0 (G ∖ {v}) ≤ (n − 1)/2. Since β0 (G) = β0 (G ∖ {v}), see Proposition 2.2.6, the
assertion follows. □
THEOREM 2.3.10. Let G be a graph and let A be the intersection of all the
minimal vertex covers of G with α0 (G) vertices. If G is vertex critical, then
dim k[∆G ] ≤ n − ∣A∣/2.
PROOF. It follows from Proposition 2.3.6 and Theorem 2.3.9. □
56 2. GRAPHS, IDEALS AND INVARIANTS

CONJECTURE 2.3.11. Let G be a hypergraph with n vertices and without isolated

vertices such that all the hyperedges of G have k vertices. If G has a maximal
independent cover, then dim k[∆G ] ≤ n − ⌊n/k⌋, where ⌊ x⌋ denotes the integer part
of x.
EXAMPLE 2.3.12. Given n, k positive integers such that n ≥ 2(k − 1) consider
the following graph G with n vertices.

xn x2k−1 x1 x2 x3 xk−2 xk−1

s s s s s s s
H J
J H G
J HH J
s
 J HHJ
Js x2k p p p
xn−1 H
 J H
JJHH 
J 
J HH  J
s
J  H
p p p H
Js s s s s s
xn−2 x2k+1 xk xk+1 xk+2 x2k−3 x2k−2
𝒦n−2k+2
Using Proposition 2.4.3 it follows that G is Cohen-Macaulay. Note β0 (G) = k. In
particular if n = 2k or n = 2k + 1 one has β0 (G) = ⌊ n2 ⌋. Thus the bound given in
Theorem 2.3.9 is best possible.
LEMMA 2.3.13. If G is a bipartite graph with n vertices, then α0 (G) ≤ ⌊ n2 ⌋.
PROOF. Let (V1 , V2 ) be a bipartition of G. Since n = ∣V1 ∣ + ∣V2 ∣, then ∣Vi ∣ ≤ ⌊ 2n ⌋
for some i. Thus α0 (G)) ≤ n2 , □
PROPOSITION 2.3.14. Let G be a bipartite graph with n vertices. If G has no
isolated vertices and G has a maximal independent cover, then n is an even integer.
PROOF. From Lemma 2.3.13 and Theorem 2.3.9 we get n ≤ 2⌊ 2n ⌋, and conse-
quently n must be even. □

Exercises
2.3.15. If G is a graph, then there is a spanning subgraph G′ of G such that G′
is edge-critical and β0 (G′ ) = β0 (G).
2.3.16. Prove that every edge-critical bipartite graph is unmixed.
2.3.17. Prove that a cycle with nine vertices is an edge-critical graph which is
not unmixed.
2.3.18. Let G be a graph with n vertices. If G is an unmixed graph without
isolated vertices such that 2α0 (G) = n, then G has a perfect matching.
 2. GRAPHS, IDEALS AND INVARIANTS 57

2.3.19. Prove that the following graph is neither edge-critical nor vertex-critical,
is not unmixed, and { x1 , x4 , x6 , x7 }, { x2 , x3 , x5 , x6 } is a maximal independent cover.

x1
s

G  @
 @
 @

 x 4
@ x6
x2 
s
HH x 3 s s @ s x5 s

HH
@ isolated vertex
HH@@
HH@
@s
H
H
x7
2.3.20. Let G be a graph. The α0 (G)-core is the intersection of all minimal
vertex covers of G with α0 (G) vertices and the β0 (G)-core is the intersection of all
maximal independent sets of G with β0 (G) vertices. Prove the following statements:
(a) Let v ∈ V(G). If deg(v) = 0, then v ∈ β0 (G)-core. If deg(v) is greater than
or equal to α0 (G) + 1, then v ∈ α0 (G)-core.
(b) G has a cover by maximum independent sets ⇔ α0 (G)-core = ∅.
(c) isol(G) ⊂ β0 (G)-core, where isol(G) is the set of isolated vertices of G.
(d) β0 (G)-core ∖ isol(G) is an independent set and there is an inclusion
N(β0 (G)-core ∖ isol(G)) ⊂ α0 (G)-core
provided that β0 (G)-core ∖ isol(G) ∕= ∅.
(e) If α0 (G)-core = ∅, then β0 (G)-core ∖ isol(G) = ∅. The converse does not
hold as the following exercise shows.
2.3.21. Consider the following graph G:
xs1
Q
 S Q
  S Q
 S Q
 x4 
Ssx5 Qs x6
Q
s
x3  s

Q  
Q S
Q S  
Q S 
Q
QSs

x2
Verify that β0 (G)-core = ∅ and α0 (G)-core = { x1 , x2 }. Notice that G is clearly
vertex critical, β0 (G) = 2, and α0 (G) = 4. The role played by the α0 (G)-core will
become clear in Theorem 2.3.10.
58 2. GRAPHS, IDEALS AND INVARIANTS

2.3.22. Let G be a graph with q edges. If G has a maximal independent cover,

then q ≤ α0 (G)2 .
2.3.23. Prove that the complete bipartite graph G = 𝒦g,g is an unmixed graph
with α0 (G)2 edges and has a maximal independent cover.
2.3.24. If G is an edge-critical graph, then q ≤ (α0 (G) + α0 (G)2 )/2 where q is
the number of edges of G.
2.3.25. Let G be a graph with n vertices and without
⌊ ⌋isolated vertices. If G has
a maximal independent cover prove that dim k[∆G ] ≤ 2n .
2.3.26. (N. Terai) Let R = k[a, . . . , j] be a polynomial ring on 10 variables over
a field k and let
I = (abc, abd, acd, bcd, abe, ac f, adg, bch, bdi, cd j),
which is the edge ideal of the hypergraph which consists of: the boundary of a
tetrahedron (abc, abd, acd, bcd) and a new triangle on each edge of the tetrahedoron
using a new vertex (e.g., ab → abe). Prove that this ideal is unmixed of height 3 (in
fact Cohen-Macaulay).

2.4. Cohen-Macaulay and chordal graphs

Let G be a graph on the vertex set V = { x1 , . . . , xn }. We define the independence
complex ∆G of a graph G by
∆G = {A ⊂ V ∣ A is an independent set in G},
notice that ∆G is precisely the Stanley-Reisner simplicial complex of I(G). The
complex ∆G is also called the complementary simplicial complex of G.
In this section we study Cohen-Macaulay graphs, chordal graphs and the Cohen-
Macaulay property of the ideal of vertex covers of a graph. A classification of
Cohen-Macaulay trees is presented.
According to Reisner’s Theorem 1.5.18 the Cohen-Macaulay property of a given
graph can be decided once the reduced homology of the links of ∆G is computed.
Unfortunately the simpler the graph G the more complicated the complex ∆G .
Following [156, 183] we will study Cohen-Macaulay graphs by looking at both the
independence complex and the graph itself.

Constructions of Cohen-Macaulay graphs.

We begin by showing how large classes of Cohen-Macaulay graphs can be produced
and give some obstructions for a graph to be Cohen-Macaulay.
 2. GRAPHS, IDEALS AND INVARIANTS 59

PROPOSITION 2.4.1. [187, Proposition 2.2.10] Let R = k[x1 , . . . , xn ] be a polyno-

mial ring over a field k and let I be a graded ideal of R. Then there are homogeneous
polynomials h1 , . . . , hd in R+ such that
dim R/(I, h1 , . . . , hi ) = d − i, for i = 1, . . . , d,
where dim(R/I) = d. If k is infinite, then h1 , . . . , hd can be chosen in R1 .
r
w
First construction. Let H be a graph with vertex set
r z
@ V(H) = { x1 , . . . , xn , z, w} and J its edge ideal. Assume that
r r @r
@ z is adjacent to w with deg(z) ≥ 2 and deg(w) = 1. We label
x1 x2 xk the vertices of H such that x1 , . . . , xk , w are the vertices of
H adjacent to z, as shown in the figure. The next two results describe how the
Cohen-Macaulay property of H relates to that of the two subgraphs G = H ∖ {z, w}
and F = G ∖ { x1 , . . . , xk }. One has the equalities
J = (I, x1 z, . . . , xk z, zw) and (I, x1 , . . . , xk ) = (L, x1 . . . , xk ),
where I and L are the edge ideals of G and F respectively.
Assume H is unmixed with height of J equal to g+1. Since z is not isolated, there
is a minimal prime 𝔭 over I containing { x1 , . . . , xk } and such that ht(I) = ht(𝔭) = g.
It is not difficult to prove that k < n and deg(xi ) ≥ 2 for i = 1, . . . , k.
PROPOSITION 2.4.2 ([183]). If H is a Cohen-Macaulay graph, then F and G are
Cohen-Macaulay graphs.
PROOF. Set A = k[x1 , . . . , xn ] and R = A[z, w]. By Proposition 2.4.1 there exists
a homogeneous system of parameters { f1 , . . . , fd } for A/I, where fi ∈ A+ for all i.
Because of the hypothesis and the equalities
z(z − w) + zw = z2 and w(w − z) + zw = w2 ,
the set { f1 , . . . , fd , z−w} is a regular system of parameters for R/J. Hence f1 , . . . fd is
a regular sequence on R/I, that is, G is Cohen-Macaulay. To show the corresponding
property for F we consider the sequence
z ψ
0 −→ R/(I, x1 , . . . , xk , w)[−1] −→ R/J −→ R/(I, z) −→ 0
where the first map is multiplication by z and ψ is induced by a projection. The
exactness of this sequence follows from Corollary 1.2.8. Taking depths with respect
to the maximal ideal of R, by the depth lemma one has
n − g + 1 ≤ depth R/(I, x1 , . . . , xk , w), where g = ht(I).
Since (I, x1 , . . . , xk , w) = (L, x1 , . . . , xk , w), F is Cohen-Macaulay. □
PROPOSITION 2.4.3. If F and G are Cohen-Macaulay graphs and { x1 , . . . , xk }
form a part of a minimal vertex cover for G, then H is a Cohen-Macaulay graph.
60 2. GRAPHS, IDEALS AND INVARIANTS

PROOF. Consider the exact sequence of Proposition 2.4.2. Since the ends of
this sequence have R-depth equal to dim(R/J), by the depth lemma H is a Cohen-
Macaulay graph. □
COROLLARY 2.4.4. If G is Cohen-Macaulay and { x1 , . . . , xk } is a minimal vertex
cover for G, then H is Cohen-Macaulay.
PROOF. Note that in this case F is Cohen-Macaulay because I(F) = (0). □
r z Second construction. For the discussion of the second
@ construction we change our notation. Let H be a graph on
xk the vertex set V(H) = { x1 , . . . , xn , z}. Let { x1 , . . . , xk } be the
r r @ @r
x1 x2
vertices of H adjacent to z, as shown in the figure. Taking
into account the first construction one may assume deg(xi ) ≥ 2 for i = 1, . . . , k
and deg(z) ≥ 2. Setting G = H ∖ {z} and F = G ∖ { x1 , . . . , xk }, notice that the
ideals J, I and L associated to H, G and F respectively are related by the equalities
J = (I, x1 z, . . . xk z) and (I, x1 , . . . , xk ) = (L, x1 , . . . , xk ).
PROPOSITION 2.4.5 ([183]). If H is a Cohen-Macaulay graph, then F is a
Cohen-Macaulay graph.
PROOF. We set R = k[x1 , . . . , xn , z], A = k[x1 , . . . , xn ] and ht(J) = g + 1. The
polynomial f = z − x1 − ⋅ ⋅ ⋅ − xk is regular on R/J because it is clearly not contained
in any associated prime of J. Therefore there is a sequence { f, f1 , . . . , fm } regular
on R/J so that { f1 , . . . , fm } ⊂ A+ , where m = n − g − 1. Observe that { f1 , . . . , fm }
is in fact a regular sequence on A/I, which gives depth(A/I) ≥ n − g − 1. Next, we
use the exact sequence
z
0 −→ R/(I, x1 , . . . , xk )[−1] −→ R/J −→ R/(I, z) −→ 0
and ht(I, x1 , . . . , xk ) = g + 1 to conclude that F is Cohen-Macaulay. □
PROPOSITION 2.4.6. Assume that x1 , . . . , xk do not form a part of a minimal
vertex cover for G and ht(I, x1 , . . . , xk ) = ht(I) + 1. If F and G are Cohen-Macaulay,
then H is Cohen-Macaulay.
PROOF. The assumption on { x1 , . . . , xk } forces ht(J) = ht(I) + 1. From the exact
sequence
z
0 −→ R/(I, x1 , . . . , xk )[−1] −→ R/J −→ R/(I, z) −→ 0
we obtain that H is Cohen-Macaulay. □
COROLLARY 2.4.7. If G is Cohen-Macaulay and { x1 , . . . , xk−1 } is a minimal
vertex cover for G, then H is Cohen-Macaulay.
 2. GRAPHS, IDEALS AND INVARIANTS 61

Connected components of Cohen-Macaulay graphs. A nice property of

Cohen-Macaulay graphs is their additivity with respect to connected components.
LEMMA 2.4.8. Let R1 = k[x] and R2 = k[y] be two polynomial rings over a field
k and R = k[x, y]. If I1 and I2 are graded ideals in R1 and R2 respectively, then
depth(R1 /I1 ) + depth(R2 /I2 ) = depth R/(I1 + I2 ).
PROOF. The formula is a direct consequence of Proposition 1.5.34 and Theo-
rem 1.5.35. □
PROPOSITION 2.4.9. If G is a graph and G1 , . . . , Gn its connected components,
then G is Cohen-Macaulay if and only if Gi is Cohen-Macaulay for all i.
PROOF. The result follows from Lemma 2.4.8. □
Cohen-Macaulay trees. Here we give an effective description of Cohen-
Macaulay trees and show an interesting family of graphs containing all Cohen-
Macaulay trees.
PROPOSITION 2.4.10 ([74]). Let I be a monomial ideal in a polynomial ring R
over a field k and S = R/I. Then there is a polynomial ring R′ and a square-free
monomial ideal I ′ of R′ such that S = S ′ /(h), where S ′ = R′ /I ′ and h is a regular
sequence on S ′ of forms of degree one.
PROOF. Let R = k[x] and let F = { f1 , . . . , fr } be a set of monomials that
minimally generate I. Assume that one of the variables, say x1 , occurs in at least
one of the monomials in F with multiplicity greater than 1. Thus one may write
f1 = x1a1 g1 , . . . , f s = x1as g s ,
/ supp(gi ) for all i ≤ s and x1 ∈
where ai ≥ 1 for all i, a1 ≥ 2, x1 ∈ / supp( fi ) for
i > s. Set
I ′ = (x0 x1a1 −1 g1 , . . . , x0 x1as −1 g s , f s+1 , . . . , fr ) ⊂ R′ = R[x0 ],
where x0 is a new variable. We claim that x0 − x1 is a nonzero divisor of S ′ = R′ /I ′ .
On the contrary assume that x0 − x1 belongs to an associated prime 𝔭 of I ′ and write
𝔭 = (I ′ : h), for some monomial h. Since 𝔭 is a face ideal, one has xi h ∈ I ′ for
i = 0, 1. Hence using h ∈ / I ′ gives x1 h = x0 x1ai −1 gi h1 for some i and consequently
ai −2
h = x0 x1 gi h1 . Hence
{ a −1
x0 h = x02 x1ai −2 gi h1 = x0 x1 j g j M or,
fjM for some j > s.
In both cases one obtains x0 ∈ supp(M) and h ∈ I ′ , a contradiction. Since
S ′ /(x0 − x1 ) = S one can repeat the construction to obtain the asserted monomial
ideal I ′ . This proof is due to R. Fröberg. □
62 2. GRAPHS, IDEALS AND INVARIANTS

The ideal I ′ constructed above is called the polarization of I. Thus any monomial
ring is a deformation by linear forms of a monomial ring with square-free relations.
Note that I is Cohen-Macaulay (resp. Gorenstein) if and only if I ′ is Cohen-
Macaulay (resp. Gorenstein).
PROPOSITION 2.4.11. Let R = k[x1 , . . . , xn , y1 , . . . , yn ] be a polynomial ring over
a field k and M ⊂ {(i, j)∣ 1 ≤ i < j ≤ n}. Then the ideal
I = ({ xi yi , y j y` ∣ i = 1, . . . , n and ( j, `) ∈ M }),
is a Cohen-Macaulay ideal.
PROOF. Note that the polarization of the ideal
J = ({ xi2 , x j x` ∣ 1 ≤ i ≤ n and ( j, `) ∈ M })
is equal to I and apply Proposition 2.4.10. This proof was pointed out to us by
Jürgen Herzog. □
LEMMA 2.4.12. Let T be a tree. If v and w are two adjacent vertices of degree
at least two, then there is a minimal vertex cover for T containing both v and w.
PROOF. Let x be the edge joining v and w. Since T ∖ { x} has exactly two
components, say T 1 and T 2 , there are minimal vertex covers A and B for T 1 and T 2
respectively so that v ∈ A and w ∈ B. Therefore A ∪ B is the required cover for T .
□
THEOREM 2.4.13 ([183]). Let T be a tree with vertex set V and edge set E.
Then T is Cohen-Macaulay if and only if ∣V ∣ ≤ 2 or 2 < ∣V ∣ = 2r and there are
vertices x1 , . . . , xr , y1 , . . . , yr so that deg(xi ) = 1, deg(yi ) ≥ 2, and { xi , yi } ∈ E for
i = 1, . . . , r.
PROOF. ⇒) As T is a bipartite graph there are disjoint sets V1 and V2 such that
V = V1 ∪ V2 and every edge of T joins V1 with V2 . Since V1 and V2 are minimal
vertex covers for T , and T is unmixed, we have
r = ∣V1 ∣ = ∣V2 ∣ = α0 (T ).
By Theorem 2.1.8 there are r independent edges. Therefore we may assume V1 =
{z1 , . . . , zr }, V2 = {w1 , . . . , wr }, and {zi , wi } ∈ E for i = 1, . . . , r. To complete the
proof is enough to show that for each i either zi or wi has degree one. Assume
deg(zi ) ≥ 2 and deg(wi ) ≥ 2 for some i. By Lemma 2.4.12∩there is a minimal
vertex cover for T , say A, containing zi and wi . Since {z j , w j } A ∕= ∅ for j ∕= i we
conclude ∣A∣ ≥ r + 1, which is a contradiction.
⇐) The sufficiency follows from Proposition 2.4.11. □
COROLLARY 2.4.14. If G is a tree, then G is Cohen-Macaulay if and only if G
is unmixed.
 2. GRAPHS, IDEALS AND INVARIANTS 63

PROOF. If G is unmixed, using the proof above it follows that I(G) is a Cohen-
Macaulay ideal of the kind described in Proposition 2.4.11. □
For a generalization of Corollary 2.4.14 see [69]. A far reaching extension of
Corollary 2.4.14 to a certain type of clutters (that include simplicial trees) has been
shown in [138, Corollary 2.19].
COROLLARY 2.4.15. The only Cohen-Macaulay cycles are the triangle and the
pentagon.
PROOF. Let Cn = { x0 , x1 , . . . , xn = x0 } be a Cohen-Macaulay cycle of length
n. The cases n = 4, 6, and 7 can be treated separately, so we assume n ≥ 8. By
Proposition 2.4.5 Cn ∖ { x0 , x1 , xn−1 } is a Cohen-Macaulay path of length n − 3
which is impossible by Theorem 2.4.13 □
Complementary complex versus the complement. The complement G of a
graph G has vertex set V(G) and two vertices are adjacent in G if and only if they
are not adjacent in G. Along the chapter we will encounter the complement G, and
witness its usefulness to encapsulate information of the edge ideal I(G).
PROPOSITION 2.4.16. Let G be a graph and ∆G its complementary simplicial
complex. Then G ⊂ ∆G , with equality if and only if G has no triangles.
PROOF. If x = {v1 , v2 } is an edge of G, then {v1 , v2 } is an independent set of G
and x ∈ ∆G . Hence G ⊂ ∆G . Assume G has no triangles. If A = {v1 , . . . , vn } is an
independent set of G, then n = 1 or n = 2; otherwise if n > 2, then G would contain
the triangle {v1 , v2 , v3 }. Therefore A is either a point or an edge of G, thus G = ∆G .
The converse also follows rapidly. □
Simplicial complexes of edge ideals. Given a graph G on the vertex set V one
defines a simplicial complex Simp(G) on the same set of vertices:
Simp(G) = {F ∣ ∃ H a subgraph of G such that F = V(H) and H ≃ 𝒦r },
Conversely a simplicial complex ∆ defines a graph Skel(∆), the 1-skeleton of ∆, on
the same set of vertices:
Skel(∆) = {F ∈ ∆∣ dim(F) ≤ 1}.
In general one has ∆ ⊂ Simp(Skel(∆)).
PROPOSITION 2.4.17. Let ∆ be a simplicial complex and I∆ its Stanley-Reisner
ideal. Then ∆ = Simp(Skel(∆)) if and only if I∆ is generated by square-free
monomials of degree two.
PROOF. Let V = { x1 , . . . , xn } be the vertex set of ∆. Assume I∆ is generated by
square-free monomials of degree two. Take F ∈ Simp(Skel(∆)), for simplicity set
F = { x1 , . . . , xr }, that is, { xi , x j } for all 1 ≤ i < j ≤ r. If F ∕∈ ∆, then x1 ⋅ ⋅ ⋅ xr is
in I∆ , and by assumption xi x j is in I∆ for some 1 ≤ i < j ≤ r. Hence { xi , x j } ∕∈ ∆,
64 2. GRAPHS, IDEALS AND INVARIANTS

which is a contradiction. This shows ∆ = Simp(Skel(∆)). The converse also follows

rapidly. □
Simplicial complexes associated to edge ideals have been studied in the litera-
ture, they are called flag complexes, see [166, Chapter III] and the references there
for a series of amusing related problems.
Triangulated (chordal) graphs. The use of graph theoretical methods in com-
mutative algebra is implicit in G. Lyubeznik thesis [128] and explicit in the work
of R. Fröberg [76]. In both cases the central notion is that of a triangulated graph.
It will turn out that this is the right notion to express the Cohen-Macaulay prop-
erty of unmixed ideals of height two generated by square-free monomials (see
Theorem 2.4.20).
Let G be a graph and let G be its complement. The graph G is said to be
triangulated or chordal if every cycle Cn in G of length n ≥ 4 has a chord in G.
A chord of Cn is an edge joining two non adjacent vertices of Cn . Triangulated
graphs have been extensively studied, they can be constructed according to a result
of G. A. Dirac, see [46, 107, 172].
LEMMA 2.4.18. Let G be a connected graph and let d be the minimum degree
among the vertices of G. If G is not a complete graph, then there is S ⊂ V(G) such
that G ∖ S is disconnected and d = ∣S ∣.
PROOF. Let v be a vertex of G of degree d and S = N(v) the neighbor set of
v. Since G is not a complete graph and v has minimum degree there is a vertex
w ∕∈ S ∪ {v}. Thus G ∖ S has at least two components. □
PROPOSITION 2.4.19 ([46]). If H is a triangulated non complete graph, then it
can be constructed out of two smaller disjoint triangulated graphs H1 and H2 by
identifying two (possibly empty) complete subgraphs of the same size in H1 and H2 .
PROOF. One may assume that H is a connected graph with n vertices. Since H
has at least one vertex of degree less than n − 1, by Lemma 2.4.18 there is a minimal
set of vertices S such that H ∖ S is disconnected and ∣S ∣ ≤ n − 2. Hence we can
write H ∖ S = F1 ∪ F2 , where F1 and F2 are disjoint subgraphs of H. Set
H1 =< V(F1 ) ∪ S > and H2 =< V(F2 ) ∪ S >,
where < A > is the induced subgraph with vertex set A. Note H1 ∪ H2 = H and
H1 ∩ H2 =< S >.
Next we show that < S > is a complete graph. Assume there are vertices x1 and
x2 in S not connected by an edge of H. Since H ∖ (S ∖ { xi }) is connected for i = 1, 2,
there are paths p1 and p2 in H1 and H2 respectively joining x1 with x2 , whose only
vertices in S are x1 and x2 . If p1 and p2 are the shortest paths with this property,
then p1 ∪ p2 is a cycle of length greater than or equal to four without a chord, a
contradiction. □
 2. GRAPHS, IDEALS AND INVARIANTS 65

Ideals of vertex covers. Let R = k[x1 , . . . xn ] be a polynomial ring over a field

k and let I be a monomial ideal of I. The ideal of minimal covers or simply the
ideal of covers of I, denoted Ic , is the ideal of R generated by all the monomials
xi1 ⋅ ⋅ ⋅ xik such that the ideal (xi1 , . . . , xik ) is an associated prime ideal of I. This
terminology is intended to emphasize that a minimal prime “covers" or contains all
the monomials in I. If I is a Stanley-Reisner ring one has a duality:
I = (xi1 ⋅ ⋅ ⋅ xir , . . . , ) = (x j1 , . . . , x js ) ∩ ⋅ ⋅ ⋅
↕ ↕
Ic = (xi1 , . . . , xir ) ∩ ⋅ ⋅ ⋅ = (x j1 ⋅ ⋅ ⋅ x js , . . . , )
In many cases I reflects properties of Ic and viceversa, see [53, 100, 133] and
Chapter 4. The ideal Ic is also called the Alexander dual of I and is often denoted
by I ∨ .
Let G be a graph and let E = E(G) be its edge set, in this case we denote the
ideal of minimal covers of I(G) by Ic (G). One has
∩
Ic (G) = (x, y).
{x,y}∈E

Thus Ic (G) is an unmixed ideal of height two.

THEOREM 2.4.20 ([129]). Let G be a graph and G its complement. Then Ic (G)
is Cohen-Macaulay if and only if G is a triangulated graph.
PROPOSITION 2.4.21. If Ic (G) is Cohen-Macaulay, then ν(Ic (G)) ≤ g + 1. where
ν(Ic (G)) is the minimum number of generators of Ic (G) and g is the height of I(G).
PROOF. By the Hilbert-Burch Theorem (see [187]) Ic (G) is generated by the
m − 1 minors of an m × (m − 1) matrix A with homogeneous entries, where
m = ν(Ic (G)). Take f in Ic (G) of degree g, since any m − 1 minor of A has degree
at least m − 1, one obtains deg( f ) = g ≥ m − 1. □

Exercises
2.4.22. Show that a cycle Cn of length n ≥ 3 is unmixed if and only if n =
3, 4, 5, 7.
2.4.23. Let G1 , G2 be two graphs with vertex set x. If their complements are
triangle free and they have the same number of edges, then the Hilbert series of
k[x]/I(G1 ) and k[x]/I(G2 ) are identical.
2.4.24. Let G be a graph on the vertex set V = { x1 , . . . , x p } such that ht(I(G)) =
p − 2. Prove the following:
(a) G is Cohen-Macaulay if and only if ∆G is connected, and
66 2. GRAPHS, IDEALS AND INVARIANTS

(b) if I(G) is not Cohen-Macaulay, then I(G) is unmixed if and only if ∆G has
no isolated vertices.
2.4.25. If G is a Cohen-Macaulay graph with q edges. Then q ≤ (g2 + g)/2,
where g is the height of I(G).
2.4.26. If G is a Cohen-Macaulay graph and x1 , x2 are two vertices of degree 1,
prove that they cannot have an adjacent vertex in common.
2.4.27. Let G be a connected graph and let {z1 , . . . , zn } be a path with n ≥ 4. If
deg(z1 ) = 1, deg(xi ) = 2 for i = 2, . . . n − 1 and deg(zn ) ≥ 3, prove that G cannot be
unmixed.
2.4.28. Prove that G = Skel(Simp(G)), for any graph G.
2.4.29. Let T be a tree with 2r vertices. If r ≥ 2, prove that T is Cohen-
Macaulay iff ht I(T ) = r and T has exactly r vertices of degree 1. State and prove a
similar characterization valid for forests without isolated vertices.

2.5. Unmixed and C-M bipartite graphs

In this section we present some structure theorems for unmixed and Cohen-Macaulay
bipartite graphs.
LEMMA 2.5.1. The complete bipartite graph 𝒦m,n is Cohen-Macaulay if and
only if m = n = 1.
PROOF. By Exercise 1.5.43 it is enough to observe that the complementary
complex of 𝒦m,n is disconnected for m + n ≥ 3. □
LEMMA 2.5.2. Let G be an unmixed bipartite graph and let I(G) be its edge
ideal. If I(G) has height g, then there are disjoint sets of vertices V1 = { x1 , . . . , xg }
and V2 = {y1 , . . . , yg } such that: (i) { xi , yi } is an edge of G for all i, and (ii) every
edge of G joins V1 with V2 .
PROOF. The statement follows by using that g is equal to the maximum number
of independent edges of G, see Theorem 2.1.8. □
DEFINITION 2.5.3. Let G be a graph and let v ∈ V(G). The neighbor set of v,
denoted by N(v) or NG (v), is the set of vertices of G adjacent to v.
THEOREM 2.5.4 ([66]). Let G be a Cohen-Macaulay bipartite graph. If G is not
a discrete graph, then there is a vertex v ∈ V(G) of degree 1.
PROOF. We may harmlessly assume that G has no isolated vertices. Set g =
ht I(G). Let V1 and V2 be as in Lemma 2.5.2. We proceed by contradiction.
Assume deg(v) ≥ 2 for all v ∈ V1 . Let v ∈ V1 be a vertex of minimal degree.
 2. GRAPHS, IDEALS AND INVARIANTS 67

We may assume v = x1 . If deg(v) = g, then G = 𝒦g,g is a complete bipartite

graph, but this is impossible by Lemma 2.5.1. Therefore we may also assume
2 ≤ deg(x1 ) ≤ g − 1. Set r = deg(x1 ). Let N(x1 ) be the set of vertices in V adjacent
to x1 , say N(x1 ) = {y1 , . . . , yr }.
We claim that deg(xi ) = r, for all i = 1, . . . , r. For simplicity we assume
i = 2. If deg(x2 ) > r, then x2 is adjacent to y j for some r + 1 ≤ j ≤ g. Let
G′ = G ∖{ x1 , y1 , . . . , yr }. By Proposition 2.4.5 we derive that G′ is Cohen-Macaulay,
and hence G′ is unmixed. Notice that {yr+1 , . . . , yg } is a minimal vertex cover for G′
and there is also a minimal vertex cover for G′ containing { x2 , xr+1 . . . , xg }, which
is a contradiction since G′ is unmixed. Thus deg(x2 ) = r. The previous argument
also shows that N(xi ) = {y1 , . . . , yr }, for all 1 ≤ i ≤ r. Hence N(yi ) is a subset of
{ xr+1 , . . . , xg }, for all r + 1 ≤ i ≤ g. Consider the graph
H = G ∖ ({yr+1 , . . . , yg } ∪ N(yr+1 ) ∪ ⋅ ⋅ ⋅ ∪ N(yg )).
To complete the proof observe that a repeated use of Proposition 2.4.5 shows that
H is a Cohen-Macaulay bipartite graph. This is impossible since H = 𝒦r,r is a
complete bipartite graph and r ≥ 2. □
COROLLARY 2.5.5. If G is a Cohen-Macaulay bipartite graph, then G ∖ {v} is
Cohen-Macaulay for some vertex v in G.
PROOF. Let {v, z} be an edge of the graph G such that deg(v) = 1. Then by
Proposition 2.4.2 we obtain that G ∖ {z} is Cohen-Macaulay. □
LEMMA 2.5.6. Let G be an unmixed bipartite graph. Assume that G has a
bipartition V1 = { x1 , . . . , xg }, V2 = {y1 , . . . , yg } such that { xi , yi } ∈ E(G) for i =
1, . . . , g. If { xi , y j }, { x j , yk } are in E(G) and i, j, k are distinct, then { xi , yk } ∈ E(G).
PROOF. Assume that xi is not adjacent to yk . Then there is a maximal independent
set of vertices A containing xi and yk . Notice that ∣A∣ = g because G is unmixed.
Hence C = V(G) ∖ A is a minimal vertex cover of G with g vertices. Since xi and yk
are not on C, we get that y j and x j are both in C, a contradiction because C intersects
{ xi , yi } in exactly one vertex for i = 1, . . . , g. □
The following is a combinatorial characterization of all unmixed bipartite
graphs. Unmixed graphs are also called well covered [142]. Unmixed bipartite
graphs were characterized by Ravindra [144]. The advantage of the next characteri-
zation is that it admits a natural extension to hypergraphs and clutters with a perfect
matching of König type [138].
THEOREM 2.5.7 ([190]). Let G be a bipartite graph without isolated vertices.
Then G is unmixed if and only if G has a bipartition V1 = { x1 , . . . , xg }, V2 =
{y1 , . . . , yg } such that: (a) { xi , yi } ∈ E(G) for all i, and (b) if { xi , y j } and { x j , yk }
are in E(G) and i, j, k are distinct, then { xi , yk } ∈ E(G).
68 2. GRAPHS, IDEALS AND INVARIANTS

PROOF. ⇒) By Lemma 2.5.2 there is a bipartition (V1 , V2 ) satisfying (a). Hence

G must satisfy (b) by Lemma 2.5.6.
⇐) Let C be a minimal vertex cover of G. It suffices to prove that C intersects
{ x j , y j } in exactly one vertex for j = 1, . . . , g. Assume that x j and y j belong to C
for some j. Thus there are xi ∈ NG (y j ) ∖ { x j } and yk ∈ NG (x j ) ∖ {y j } such that
xi ∈/ C and yk ∈ / C. Notice that i, j, k are distinct. Indeed if i = k, then { xi , yi } is an
edge of G not covered by C, which is impossible. Therefore using (b) we get that
{ xi , yk } is an edge of C, a contradiction. □
The following nice result of Herzog and Hibi classifies all Cohen-Macaulay
bipartite graphs. A similar classification exist for other families of clutters (see
Theorem 2.6.34 and [94, 138]). The family of Cohen-Macaulay bipartite graphs
belongs to the family of uniform admissible clutters defined and studied in [71, 94,
138].
THEOREM 2.5.8 ([102]). Let G be a bipartite graph without isolated vertices.
Then G is a Cohen-Macaulay graph if and only if there is a bipartition V1 =
{ x1 , . . . , xg }, V2 = {y1 , . . . , yg } of G such that: (i) { xi , yi } ∈ E(G) for all i, (ii)
if { xi , y j } ∈ E(G), then i ≤ j, and (iii) if { xi , y j } and { x j , yk } are in E(G) and
i < j < k, then { xi , yk } ∈ E(G).
PROOF. By Proposition 2.4.9 we may assume that G is connected.
⇐) The proof is by induction on g. The case g = 1 is clear. Assume g ≥ 2.
Let S = { xr1 , . . . , xrs } be the set of all vertices of G adjacent to yg . Consider the
subgraph G′ = G ∖ S obtained from G be removing the vertices in S . We claim that
yr1 , . . . , yrs are isolated vertices of G′ . Indeed if yr j is not isolated, there is an edge
{ xi , yr j } in G′ with i < r j . Hence by condition (iii) we get that { xi , yg } is an edge
of G and xi must be a vertex in S , a contradiction. Thus by induction the graphs
G′ ∖{yr1 , . . . , yrs } and G ∖{ xg , yg } are Cohen-Macaulay. Applying Proposition 2.4.3
we get that G is Cohen-Macaulay.
⇒) By Lemma 2.5.2 there is a bipartition (V1 , V2 ) satisfying (i). By Theo-
rem 2.5.4 we may assume that deg(xg ) = 1 and deg(yg ) ≥ 2. Since G is Cohen-
Macaulay we get that G ∖ {yg } is Cohen-Macaulay, see Proposition 2.4.2. Using
induction on G it is seen that there is a bipartition of G satisfying (i) and (ii). To
complete the proof notice that G must also satisfy (iii) according to Lemma 2.5.6.
□
Theorem 2.5.7 shows that condition (iii) is in essence an expression for the
unmixed property of G, i.e., in Theorem 2.5.8 we may assume that G is unmixed
instead of assuming condition (iii).
Computing the type. Let G be a Cohen-Macaulay bipartite graph on the vertex
set V. Assume G has no isolated vertices. By Lemma 2.5.2 there are disjoint sets of
vertices V1 = { x1 , . . . , xg } and V2 = {y1 , . . . , yg } such that: (i) { xi , yi } is an edge of
 2. GRAPHS, IDEALS AND INVARIANTS 69

G for all i, and (ii) every edge of G joins V1 with V2 . Since V1 and V2 are minimal
vertex covers for G, and G is unmixed, we have g = ∣V1 ∣ = ∣V2 ∣ = height I(G).
Notice that y = { xi − yi }gi=1 is a regular system of parameters for R/I(G), where
R = k[x, y], and R/I(G) modulo (y) reduces to:
A = k[x]/(x12 , . . . , xg2 , I(H)),
where H is a graph on the vertex set V1 . The algebra A is generated, as a k-vector
space, by the image of 1 and the images of all the monomials xi1 ⋅ ⋅ ⋅ xir such that
{ xi1 , . . . , xir } is an independent set of H. If ϒ(H) denotes the set of minimal vertex
covers of H, then the type of R/I(G) is equal to ∣ϒ(H)∣, because Soc(A) is generated
by the images of all the monomials xi1 ⋅ ⋅ ⋅ xir such that { xi1 , . . . , xir } is a maximal
independent set of H.
EXAMPLE 2.5.9. Let G be the Cohen-Macaulay bipartite graph whose edge ideal
is:
I(G) = (x1 y1 , x1 y2 , x1 y3 , x1 y4 , x2 y2 , x2 y3 , x2 y4 , x3 y3 , x4 y4 ).
The graph H has edges x1 x2 , x1 x3 , x1 x4 , x2 x3 , x2 x4 . Since this graph has only three
minimal vertex covers, the type of R/I(G) is equal to three.

Exercises
2.5.10. Prove that the following is the full list of Cohen-Macaulay connected
bipartite graphs with eight vertices.
r r r rr r r rr r r rr r r r
Q
@Q @ Q
A QA
@ Q
A Q
@ Q
@Q @
A
@Q @ A@QA A@Q @QA@
@QQ @ A @Q AQ A @QQ @QAQ@
@ Q @ A @A Q A @ Q @A Q@
r r @r QQrr
@ Ar @Ar Q
Qrr Ar @r Q
Qrr r @Ar Q
@r
Q
r r r rr r r rr r r r
Q
@
A Q@ A Q
@
A Q@ A Q
A Q@
@ A A
A@QA@ A@Q @ A A@QA@ A
A @Q AQ
@ A @QQ @A A @Q AQ
@A
A @A Q @ A @ Q @A A @A Q @A
r Ar @Ar Q
Qrr
@ Ar @r Q
@
QArr Ar @Ar Q@
QAr
2.5.11. Let G be a Cohen-Macaulay bipartite graph. Show that the h-vector
of R/I(G) has length at most g = ht I(G). Can we replace Cohen-Macaulay by
unmixed?
2.5.12. Let I = (xi y j ∣ 1 ≤ i ≤ j ≤ n), draw a picture of the bipartite graph G
such that I = I(G) and prove that I is a Cohen-Macaulay ideal.
2.5.13. Let G be a Cohen-Macaulay bipartite graph. If G is not a discrete graph,
prove that G has at least two vertices of degree one.
70 2. GRAPHS, IDEALS AND INVARIANTS

2.5.14. Let G be a connected bipartite graph satisfying the conditions of Theo-

rem 2.5.8. Prove that deg(y1 ) = 1 and deg(xg ) = 1.

2.5.15. Let G be a connected Cohen-Macaulay bipartite graph with q edges. If

q is equal to g(g + 1)/2, where g = ht I(G), prove that the edge ideal of G is given
by I(G) = (xi y j ∣ 1 ≤ i ≤ j ≤ g).

2.5.16 ([190]). Let G be a bipartite graph. Then G is unmixed if and only there
is a perfect matching e1 , . . . , eg such that for any two edges e ∕= e′ and for any two
distinct vertices x ∈ e, y ∈ e′ contained in some ei , one has that (e ∖ { x}) ∪ (e′ ∖ {y})
is an edge.

2.5.17 ([144]). A connected bipartite graph G is unmixed if and only if there

is a perfect matching such that for every edge { x, y} in the perfect matching, the
induced subgraph ⟨NG (x) ∪ NG (y)⟩ is a complete bipartite graph.

2.5.18. Let G be a bipartite graph with bipartition V1 = { x1 , . . . , xg } and V2 =

{y1 , . . . , yg } such that { xi , yi } ∈ E(G) for all i and g ≥ 2. Assume that G is
connected. Following [33] we define a directed graph 𝒟 with vertex set V1 as
follows: (xi , x j ) is an edge of 𝒟 if i ∕= j and { xi , y j } is an edge of G. Prove the
following
(a) If G is a square, then 𝒟 has two vertices x1 , x2 and two arrows (x1 , x2 ),
(x2 , x1 ) forming a cycle of length two.
(b) 𝒟 is acyclic, i.e., 𝒟 has no directed cycles, if and only if there is a linear
ordering of the vertex set V1 such that all the edges of 𝒟 are of the form
(xi , x j ) with i < j.
(c) 𝒟 is called transitive if for any two (xi , x j ), (x j , xk ) in E(𝒟) with i, j, k
distinct, we have that (xi , xk ) ∈ E(𝒟). The digraph 𝒟 is transitive if and
only if G is unmixed.
(d) (Carra’Ferro-Ferrarello [33]) G is Cohen-Macaulay if and only if 𝒟 is
acyclic and transitive.
(e) ([178]) G is a Cohen-Macaulay tree if and only if 𝒟 is a tree such that
every vertex xi of 𝒟 is either a source (i.e., has only arrows leaving xi ) or
a sink (i.e., has only arrows entering xi ).
(f) If G is a Cohen-Macaulay graph, then 𝒟 has at least one source xi and at
least one sink x j .
(g) If every vertex of 𝒟 is either a source or a sink, then G is Cohen-Macaulay.
 2. GRAPHS, IDEALS AND INVARIANTS 71

2.6. Shellable graphs

We shall call a graph G shellable if the independence complex ∆G is a shellable
simplicial complex. Here, we mean the non-pure definition of shellability as intro-
duced by Björner and Wachs [12]. Because a shellable simplicial complex has the
property that its associated Stanley-Reisner ring is sequentially Cohen-Macaulay,
by identifying shellable graphs, we are in fact identifying some of the sequentially
Cohen-Macaulay graphs. We will focus on the shellability of bipartite graphs,
chordal graphs, and clutters with the free vertex property. We will classify all
sequentially Cohen-Macaulay bipartite graphs and show that chordal graphs and
clutters with the free vertex property are shellable.
DEFINITION 2.6.1. A simplicial complex ∆ is shellable if the facets (maximal
faces) of ∆ can be ordered F1 , . . . , F s such that for all 1 ≤ i < j ≤ s, there exists
some v ∈ F j ∖ Fi and some ` ∈ {1, . . . , j − 1} with F j ∖ F` = {v}.
We call F1 , . . . , F s a shelling of ∆ when the facets have been ordered with
respect to the shellable definition. For a fixed shelling of ∆, if F, F ′ ∈ ∆ then we
write F < F ′ to mean that F appears before F ′ in the ordering.
The above definition of shellable is due to Björner and Wachs [12] and is usually
referred to as nonpure shellable, although in this section we will drop the adjective
“nonpure”. Originally, in Section 1.5 the definition of shellable also required that
the simplicial complex be pure, that is, all the facets have same dimension. We will
say ∆ is pure shellable if it also satisfies this hypothesis.
DEFINITION 2.6.2. Let G be a graph whose independence complex is ∆G . We
say G is a shellable graph if ∆G is a shellable simplicial complex.
To prove that a graph G is shellable, it suffices to prove each connected compo-
nent of G is shellable, as demonstrated below.
LEMMA 2.6.3 ([178]). Let G1 and G2 be two graphs with disjoint sets of vertices
and let G = G1 ∪ G2 . Then G1 and G2 are shellable if and only if G is shellable.
PROOF. (⇒) Let F1 , . . . , Fr and H1 , . . . , H s be the shellings of ∆G1 and ∆G2
respectively. Then if we order the facets of ∆G as
F1 ∪ H1 , . . . , F1 ∪ H s ; F2 ∪ H1 , . . . , F2 ∪ H s ; . . . ; Fr ∪ H1 , . . . , Fr ∪ H s
we get a shelling of ∆G . Indeed if F ′ < F are two facets of ∆G we have two cases
to consider. Case (i): F ′ = Fi ∪ Hk and F = F j ∪ Ht , where i < j. Because ∆G1
is shellable there is v ∈ F j ∖ Fi and ` < j with F j ∖ F` = {v}. Hence v ∈ F ∖ F ′ ,
F` ∪ Ht < F, and F ∖ (F` ∪ Ht ) = {v}. Case (ii): F ′ = Fk ∪ Hi and F = Fk ∪ H j ,
where i < j. This case follows from the shellability of ∆G2 .
(⇐) Note that if F is a facet of ∆G , then F ′ = F ∩ VG1 , respectively, F ′′ = F ∩ VG2 ,
is a facet of ∆G1 , respectively ∆G2 . We now show that G1 is shellable and omit the
72 2. GRAPHS, IDEALS AND INVARIANTS

similar proof for the shellability of G2 . Let F1 , . . . , Ft be a shelling of ∆G , and

consider the subsequence
Fi1 , . . . , Fis with 1 = i1 < i2 < ⋅ ⋅ ⋅ < i s
where F1 ∩ VG2 = Fi j ∩ VG2 for i j ∈ {i1 , . . . , i s }, but F1 ∩ VG2 ∕= Fk ∩ VG2 for any
k ∈ {1, . . . , t} ∖ {i1 , . . . , i s }. We then claim that
F1′ = Fi1 ∖ VG2 , F2′ = Fi2 ∖ VG2 , . . . , F ′s = Fis ∖ VG2
is a shelling of ∆G1 . We first show that this is a complete list of facets; indeed, each
F ′j = Fi j ∩VG1 is a facet of ∆G1 , and furthermore, for any facet F ∈ ∆G1 , F ∪(F1 ∩VG2 )
is a facet of ∆G , and hence F ∪ (F1 ∩ VG2 ) = Fi j for some i j ∈ {i1 , . . . , i s }.
Because the Fi ’s form a shelling, if 1 ≤ k < j ≤ s, there exists v ∈ Fi j ∖ Fik =
(Fi j ∖ VG2 ) ∖ (Fik ∖ VG2 ) = F ′j ∖ Fk′ such that {v} = Fi j ∖ F` for some 1 ≤ ` < i j . It
suffices to show that F` is among Fi1 , . . . , Fis . Now because Fi j ∩ VG2 ⊂ Fi j and
v ∕∈ Fi j ∩ VG2 , we must have Fi j ∩ VG2 ⊂ F` . So, F` ∩ VG2 ⊃ Fi j ∩ VG2 . But
F` ∩ VG2 is a facet of ∆G2 , so we must have F` ∩ VG2 = Fi j ∩ VG2 . So F` = Fir for
some r < j, and hence, {v} = F ′j ∖ Fr′ , as desired. □
LEMMA 2.6.4. Let x be a vertex of G and let G′ = G ∖ ({ x} ∪ NG (x)). Then
∆G′ = lk∆G (x).
/ F and F ∪ { x} is a facet of ∆G .
In particular, F is a facet of ∆G′ if and only if x ∈
PROOF. If F ∈ lk∆G (x), then x ∕∈ F, and F ∪ { x} ∈ ∆G implies that F ∪ { x} is
an independent set of G. So (F ∪ { x}) ∩ NG (x) = ∅. But this means that F ⊂ VG′
because VG′ = VG ∖ ({ x} ∪ NG (x)). Thus F ∈ ∆G′ since F is also an independent
set of the smaller graph G′ .
Conversely, if F ∈ ∆G′ , then F is an independent set of G′ that does not contain
any of the vertices of { x} ∪ NG (x). But then F ∪ { x} is an independent set of G,
i.e., F ∪ { x} ∈ ∆G . So F ∈ lk∆G (x).
The last statement follows readily from the fact that F is a facet of lk∆G (x) if
and only if x ∈ / F and F ∪ { x} is a facet of ∆G . □
The property of shellability is preserved when removing the vertices { x}∪ NG (x)
and all incident edges from G for any vertex x.
THEOREM 2.6.5 ([178]). Let x be a vertex of a graph G and let G′ be the graph
G′ = G ∖ ({ x} ∪ NG (x)). If G is shellable, then G′ is shellable.
PROOF. Let F1 , . . . , F s be a shelling of ∆G . Suppose the subsequence
Fi1 , Fi2 , . . . , Fit with i1 < i2 < ⋅ ⋅ ⋅ < it
is the list of all the facets with x ∈ Fi j . Setting H j = Fi j ∖ { x} for each j = 1, . . . , t,
Lemma 2.6.4 implies that the H j ’s are the facets of ∆G′ .
 2. GRAPHS, IDEALS AND INVARIANTS 73

We claim that H1 , . . . , Ht is a shelling of ∆G′ . Because the Fi ’s form a shelling, if

1 ≤ k < j ≤ t, there exists a vertex v ∈ Fi j ∖ Fik = (Fi j ∖{ x}) ∖ (Fik ∖{ x}) = (H j ∖ Hk )
such that {v} = Fi j ∖ F` for some 1 ≤ ` < i j . It suffices to show that F` is among
the list Fi1 , . . . , Fit . But because x ∈ Fi j and x ∕= v, we must have x ∈ F` . Thus
F` = Fik for some k ≤ j. But then {v} = Fi j ∖ F` = H j ∖ Hk . So, the Hi ’s form a
shelling of ∆G′ . □
We now turn our attention to the shellability of bipartite graphs.
LEMMA 2.6.6. Let G be a bipartite graph with bipartition { x1 , . . . , xm },
{y1 , . . . , yn }. If G is shellable and G has no isolated vertices, then there is v ∈ VG
with deg(v) = 1.
PROOF. Let F1 , . . . , F s be a shelling of ∆G . We may assume that Fi =
{y1 , . . . , yn }, F j = { x1 , . . . , xm } and i < j. Then there is xk ∈ F j ∖ Fi and F`
with ` ≤ j − 1 such that F j ∖ F` = { xk }. For simplicity assume that xk = x1 . Then
{ x2 , . . . , xm } ⊂ F` and there is yt in F` for some 1 ≤ t ≤ n. Since {yt , x2 , . . . , xm } is
an independent set of G, we get that yt can only be adjacent to x1 . Thus deg(yt ) = 1
because G has no isolated vertices. □
THEOREM 2.6.7 ([178]). Let G be a graph and let x1 , y1 be two adjacent vertices
of G with deg(x1 ) = 1. If G1 = G ∖ ({ x1 } ∪ NG (x1 )) and G2 = G ∖ ({y1 } ∪ NG (y1 )),
then G is shellable if and only if G1 and G2 are shellable.
PROOF. If G is shellable, then G1 and G2 are shellable by Theorem 2.6.5. So it
suffices to prove the reverse direction. Let F1′ , . . . , Fr′ be a shelling of ∆G1 and let
H1′ , . . . , H s′ be a shelling of ∆G2 . It suffices to prove that
F1′ ∪ { x1 }, . . . , Fr′ ∪ { x1 }, H1′ ∪ {y1 }, . . . , H s′ ∪ {y1 }
is a shelling of ∆G . One first shows that this is the complete list of facets of ∆G
using Lemma 2.6.4. Indeed, take any facet F of ∆G . If y1 ∈ F, then x1 ∕∈ F because
{ x1 , y1 } is an edge of G, and by Lemma 2.6.4, F ∖ {y1 } = Hi′ for some i. On the
other hand, if y1 ∕∈ F, we must have x1 ∈ F, because if not, then { x1 } ∪ F is larger
independent set of G because x1 is only adjacent to y1 . Again, by Lemma 2.6.4, we
have F ∖ { x} = Fi′ for some i. Let F ′ < F be two facets of ∆G . There are three cases
to consider. Case (i): F ′ = Fi′ ∪ { x1 } and F = H ′j ∪ {y1 }. Since H ′j ∪ { x1 } is an
independent set of G, it is contained in a facet of ∆G , i.e., H ′j ∪ { x1 } ⊂ F`′ ∪ { x1 } for
some `. Hence (H ′j ∪ {y1 }) ∖ (F`′ ∪ { x1 }) = {y1 }, y1 ∈ F ∖ F ′ , and F`′ ∪ { x1 } < F.
The remaining two cases follow readily from the shellability of ∆G1 and ∆G2 . □
Putting together the last two results yields a recursive procedure to verify if a
bipartite graph is shellable. The next result is the first combinatorial characterization
of all shellable bipartite graphs.
74 2. GRAPHS, IDEALS AND INVARIANTS

COROLLARY 2.6.8 ([178]). Let G be a bipartite graph. Then G is shellable if

and only if there are adjacent vertices x and y with deg(x) = 1 such that the bipartite
graphs G ∖ ({ x} ∪ NG (x)) and G ∖ ({y} ∪ NG (y)) are shellable.
PROOF. By Lemma 2.6.3 it suffices to verify the statement when G is connected.
By Lemma 2.6.6 there exists a vertex of x1 with deg(x1 ) = 1. Now apply the previous
theorem. □
Next we show that all chordal graphs are shellable. Let S be a set of vertices of
a graph G. The induced subgraph GS is the maximal subgraph of G with vertex set
S . For use below we call a complete subgraph of G a clique. As usual, a complete
graph with r vertices is denoted by 𝒦r .
LEMMA 2.6.9. [172, Theorem 8.3] Let G be a chordal graph and let 𝒦 be a
complete subgraph of G. If 𝒦 ∕= G, then there is x ∕∈ V(𝒦) such that G NG (x) is a
complete subgraph.
THEOREM 2.6.10 ([178]). Let G be a chordal graph. Then G is shellable.
PROOF. We proceed by induction on n = ∣VG ∣. Let VG = { x1 , . . . , xn } be the
vertex set of G. If G is a complete graph, then ∆G consists of n isolated vertices
and they clearly form a shelling. Thus by Lemma 2.6.3 we may assume that G
is a connected non-complete graph. According to Lemma 2.6.9 there is x1 ∈ VG
such that G NG (x1 ) = 𝒦r−1 is a complete subgraph for some r ≥ 1. (To apply
Lemma 2.6.9, take 𝒦 to be any edge of G; this is clearly a complete graph.) Notice
that G{x1 }∪NG (x1 ) = 𝒦r and that 𝒦r is the only maximal complete subgraph of G that
contains x1 . We may assume that V(𝒦r ) = { x1 , . . . , xr }. Consider the subgraphs
Gi = G ∖ ({ xi } ∪ NG (xi )), which are also chordal. By induction there is a shelling
Fi1 , . . . , Fisi of ∆Gi for i = 1, . . . , r. Observe that any facet of ∆G intersects V(𝒦r )
in exactly one vertex. Thus by Lemma 2.6.4 the following is the complete list of
facets of ∆G :
F11 ∪ { x1 }, . . . , F1s1 ∪ { x1 };
F21 ∪ { x2 }, . . . , F2s2 ∪ { x2 }; . . . ;
Fr1 ∪ { xr }, . . . , Frsr ∪ { xr }.
We claim that this linear ordering is a shelling of ∆G . Let F ′ < F be two facets of ∆G .
There are two cases to consider. Case (i): F ′ = Fik ∪ { xi } and F = F jt ∪ { x j }, where
i < j. Notice that F jt ∪ { x1 } is an independent set of G because F jt ∩ V(𝒦r ) = ∅.
Thus F jt ∪ { x1 } can be extended to a facet of G, i.e., F jt ∪ { x1 } ⊂ F1` ∪ { x1 } for
some 1 ≤ ` ≤ s1 . Set F ′′ = F1` ∪ { x1 }. Hence x j ∈ F ∖ F ′ , F ∖ F ′′ = { x j }, and
F ′′ < F. Case (ii): F ′ = Fik ∪ { xi } and F = Fit ∪ { xi }, with k < t. This case follows
from the shellability of ∆Gi . □
As shown below, if a graph G is shellable, then it is also sequentially Cohen-
Macaulay. The above theorem, therefore, gives a proof to the fact that all chordal
 2. GRAPHS, IDEALS AND INVARIANTS 75

graphs are sequentially Cohen-Macaulay as first proved in [72]. To show that all
chordal graphs are sequentially Cohen-Macaulay, the authors of [72] show that
for each degree d ≥ 0, the square-free part of the Alexander dual I(G)∨ (also
defined below) in degree d has linear quotients, that is, there is an ordering of
the generators {u1 , . . . , u s } of the square-free part of I(G)∨ of degree d such that
(u1 , . . . , ui−1 ) : (ui ) = (xi1 , . . . , xit ) for i = 1, . . . , s. However, when G is shellable,
the generators of the Alexander dual I(G)∨ must have linear quotients (see [107,
Theorem 1.4(c)] and [160]); so, when G is chordal, the ideal I(G)∨ also has linear
quotients.
The following nice recent result of R. Woodroofe gives a sufficient condition
for shellability of graphs and is an extension of the fact that chordal graphs are
shellable because any vertex decomposable complex is shellable and sequentially
Cohen-Macaulay, see [11, 199].
THEOREM 2.6.11. [199, Theorem 1.1] If G is a graph with no chordless cycles
of length other than 3 or 5, then G is vertex decomposable (hence shellable and
sequentially Cohen-Macaulay).
Shellable bipartite graphs with a perfect matching. Let ∆G be the Stanley-
Reisner complex of I(G). Its facets are the maximal independent (stable) sets of
vertices of G. Following [49] we define the kth pure skeleton of ∆G as:
[k]
∆G = ⟨{F ∈ ∆G ∣ k = ∣F ∣}⟩; 0 ≤ k ≤ dim(∆G ) + 1,
where ⟨ℱ⟩ denotes the subcomplex generated by ℱ . Note that this simplicial
complex is always pure. Here we shall be interested only in the pure skeleton of ∆G
of maximum dimension. Below we study the sequentially Cohen-Macaulay property
of a module. We say that a simplicial complex ∆ is sequentially Cohen-Macaulay
if its Stanley-Reisner ring has this property. The following is an interesting result
of Duval.
THEOREM 2.6.12. [49, Theorem 3.3] A simplicial complex ∆ is sequentially
Cohen-Macaulay if and only if ∆[k] is Cohen-Macaulay for 0 ≤ k ≤ dim(∆) + 1.
COROLLARY 2.6.13. R/I(G) is Cohen-Macaulay if and only if R/I(G) is sequen-
tially Cohen-Macaulay and G is unmixed.
The following result characterizes all bipartite graphs with a perfect matching
that satisfy condition (ii) of Theorem 2.5.8.
THEOREM 2.6.14. If G is a bipartite graph with a perfect matching e1 , . . . , eg
[g]
such that ei = { xi , yi } for all i, then Γ = ∆G is pure shellable if and only if we can
order e1 , . . . , eg such that { xi , y j } ∈ E(G) implies i ≤ j.
[g]
PROOF. ⇐) It suffices to show that Γ = ∆G is shellable because this simplicial
complex is always pure. We proceed by induction on g. Each facet of Γ contains
76 2. GRAPHS, IDEALS AND INVARIANTS

exactly one vertex of each edge of the perfect matching. We set

∪
A = {yi ∣ xi ∈ N(yg )}; B = A ∪ N(yg ) = { xi , yi },
xi ∈N(yg )

where N(yg ) is the set of vertices of G that are adjacent to yg . Consider the graph
G′ = G ∖ B, obtained from G by removing all vertices of B and all edges incident
with some vertex of B.
Let F1′ = ∅ if ∣A∣ = g, in which case G′ = ∅. Else let F1′ , . . . , Fr′ be the facets
′ that do not intersect N(A), where ` = g − ∣ A∣. Here N(A) denotes
[`]
of Γ′ = ∆G
the neighbor set of A, i.e., the set of vertices of G that are adjacent to some vertex
of A. We claim that F1 = F1′ ∪ A, . . . , Fr = Fr′ ∪ A is the set of facets of Γ that
contain yg . First we show that Fk is a facet of Γ for all k. If Fk contains an edge
e = { xi , y j }, then y j ∈ A and xi ∈ Fk′ because A and Fk′ are independent. Then
xi ∈ N(A), a contradiction because N(A) ∩ Fk′ = ∅. Hence Fk is independent and
it is a facet of Γ because ∣Fk ∣ = g. Conversely, let F be a facet of Γ containing yg .
Then F ∩ N(yg ) = ∅, A ⊂ F, and F ∩ N(A) = ∅. Thus we can write F = F ′ ∪ A,
where F ′ = F ∖ A is a facet of Γ′ with F ′ ∩ N(A) = ∅, as required. By the induction
hypothesis Γ′ is shellable. Next we prove that F1′ , . . . , Fr′ is a shelling with the
linear order induced by the shelling of Γ′ . Assume Fi′ < F ′j . Since Γ′ is shellable,
there are v ∈ F ′j ∖ Fi′ and a facet F ′ of Γ′ such that F ′ < F ′j and F ′j ∖ F ′ = {v}.
It suffices to prove that F ′ does not intersect N(A). If F ′ ∩ N(A) ∕= ∅, pick x p
in the intersection. Then x p ∈ / Fi′ ∪ F ′j because Fi′ and F ′j do not intersect N(A),
consequently y p ∈ Fi′ ∩ F ′j and y p ∈ / F ′ because any facet of Γ′ contains exactly
one vertex of the edge { x p , y p }. Thus y p = v and v ∈ Fi′ , a contradiction. This
proves F ′ ∩ N(A) = ∅, as required. Thus by reordering, we have that F1′ , . . . , Fr′ is a
shelling for the simplicial complex they generate. It is rapidly seen that F1 , . . . , Fr
is also a shelling for the simplicial complex they generate.
[g−1]
Next we consider the graph G′′ = G ∖ { xg , yg } and the complex Γ′′ = ∆G ′′ . Let
F1 , . . . , Fm be the facets of Γ . By the induction hypothesis Γ is shellable. Thus
′′ ′′ ′′ ′′

we may assume that F1′′ , . . . , Fm′′ is a shelling of Γ′′ . It is not hard to see that
H1 = F1′′ ∪ { xg }, . . . , Hm = Fm′′ ∪ { xg }
is the set of facets of Γ containing xg , and that H1 , . . . , Hm is a shelling of the
simplicial complex generated by them. To finish the proof notice that
H1 , H2 , . . . , Hm , F1 , F2 , . . . , Fr ,
is clearly the complete list of facets of Γ and they form a shelling of Γ. Indeed for
any F j one has that Hk = (F j ∖ {yg }) ∪ { xg } is a facet of Γ with Fi ∖ Hk = {yg } and
Hk < F j .
⇒) The proof is by induction on g. We claim that G has a vertex of degree 1.
Let F1 , . . . , F s be a shelling of Γ. As {y1 , . . . , yg } and { x1 , . . . , xg } are facets of Γ,
 2. GRAPHS, IDEALS AND INVARIANTS 77

we may assume that Fi = {y1 , . . . , yg }, F j = { x1 , . . . , xg } and i < j. Then there is

xk ∈ F j ∖ Fi and F` with ` ≤ j − 1 such that F j ∖ F` = { xk }. Then
{ x1 , . . . , xk−1 , xk+1 . . . , xg } ⊂ F`
and there is yt in F` for some 1 ≤ t ≤ g. Since
F` = { x1 , . . . , xk−1 , yt , xk+1 , . . . , xg }
is an independent set of G, we get that yt can only be adjacent to xt . Thus deg(yt ) = 1
because G has no isolated vertices. Thus we may order e1 , . . . , eg so that deg(xg ) = 1.
[g−1]
Consider the graph G′ = G ∖{ xg , yg }. Using [178, Theorem 2.9] we obtain that ∆G ′
is a shellable complex. Hence by induction hypothesis we can order e1 , . . . , eg−1 so
that if { xi , y j } ∈ E(G′ ), then 1 ≤ i ≤ j ≤ g − 1. To finish the proof note that any
edge of G is either an edge of G′ or an edge of G containing yg . □
EXAMPLE 2.6.15 ([178]). Let G be the following bipartite graph. The ring
[5]
R/I(G) is not sequentially Cohen-Macaulay but the complex ∆G is shellable.

xt1 xt2 xt3 xt4 xt5

H
@ @HH @ @
@ @ H @ @
@ @ HHH@ @
@ @ HH
@ @
t @t @t @t
H @t
y1 y2 y3 y4 y5

[5]
A shelling of the facets of ∆G is:
{ x1 , x2 , x3 , x4 , x5 } < { x2 , x3 , x4 , x5 , y1 } < { x3 , x4 , x5 , y1 , y2 } <
{ x4 , x5 , y1 , y2 , y3 } < { x5 , y1 , y2 , y3 , y4 } < {y1 , y2 , y3 , y4 , y5 }.
COROLLARY 2.6.16. G is a Cohen-Macaulay bipartite graph if and only if: (h′1 )
[g]
∆G is shellable, g = ht I(G), and (h′2 ) G is unmixed.
PROOF. It follows using Theorem 2.5.7 together with Theorems 2.5.8 and 2.6.14.
□
Sequentially Cohen-Macaulay bipartite graphs. Next we classify all sequen-
tially Cohen-Macaulay bipartite graphs. We call a graph G sequentially Cohen-
Macaulay if R/I(G) is sequentially Cohen-Macaulay. A number of authors (for
example, see [47, 66, 70, 72, 102, 199]) have been interested in classifying or
identifying sequentially Cohen-Macaulay graphs G in terms of the combinatorial
properties of G.
78 2. GRAPHS, IDEALS AND INVARIANTS

DEFINITION 2.6.17. Let R = k[x1 , . . . , xn ]. A graded R-module M is called

sequentially Cohen-Macaulay (over k) if there exists a finite filtration of graded
R-modules
0 = M0 ⊂ M1 ⊂ ⋅ ⋅ ⋅ ⊂ Mr = M
such that each Mi /Mi−1 is Cohen-Macaulay, and the Krull dimensions of the quo-
tients are increasing:
dim(M1 /M0 ) < dim(M2 /M1 ) < ⋅ ⋅ ⋅ < dim(Mr /Mr−1 ).
As first shown by Stanley [166], shellable implies sequentially Cohen-Macaulay.
THEOREM 2.6.18. Let ∆ be a simplicial complex, and suppose that R/I∆ is
the associated Stanley-Reisner ring. If ∆ is shellable, then R/I∆ is sequentially
Cohen-Macaulay.
We now specialize to the case of graphs by providing a sequentially Cohen-
Macaulay analog of Theorem 2.6.5.
THEOREM 2.6.19 ([178]). Let x be a vertex of a graph G and let G′ be the
graph G′ = G ∖ ({ x} ∪ NG (x)). If G is sequentially Cohen-Macaulay, then G′ is
sequentially Cohen-Macaulay.
PROOF. Let F1 , . . . , F s be the facets of ∆ = ∆G , and let F1 , . . . , Fr be the
facets of ∆ that contain x. Set Γ = ∆G′ ; by Lemma 2.6.4, the facets of Γ are
F1′ = F1 ∖ { x}, . . . , Fr′ = Fr ∖ { x}.
Consider the pure simplicial complexes
∆[k] = ⟨{F ∈ ∆∣ dim(F) = k}⟩; −1 ≤ k ≤ dim(∆),
Γ[k] = ⟨{F ∈ Γ∣ dim(F) = k}⟩; −1 ≤ k ≤ dim(Γ),
where ⟨ℱ⟩ denotes the subcomplex generated by the set of faces ℱ . Recall that H
is a face of ⟨ℱ⟩ if and only if H is contained in some F in ℱ . Take a facet Fi′ of Γ
of dimension d = dim(Γ). Then Fi′ ∪ { x} ∈ ∆[d+1] and consequently { x} ∈ ∆[k+1]
for k ≤ d. Because the facets of Γ are F1′ = F1 ∖ { x}, . . . , Fr′ = Fr ∖ { x}, we have
the equality
Γ[k] = lk∆[k+1] (x)
for k ≤ d. By [49, Theorem 3.3] the complex ∆ is sequentially Cohen-Macaulay if
and only if ∆[k] is Cohen-Macaulay for −1 ≤ k ≤ dim(∆). Because ∆[k] is Cohen-
Macaulay, by Proposition 1.5.23 lk∆[k] (F) is Cohen-Macaulay for any F ∈ ∆[k] .
Thus, Γ[k] = lk∆[k+1] (x) is Cohen-Macaulay for any −1 ≤ k ≤ dim(Γ) ≤ dim(∆) − 1.
Therefore Γ is sequentially Cohen-Macaulay by [49, Theorem 3.3], as required. □
EXAMPLE 2.6.20. The six cycle C6 is a counterexample to the converse of the
above statement. For any vertex x of C6 , the graph C6 ∖ ({ x} ∪ NG (x)) is a tree,
which is sequentially Cohen-Macaulay. (A tree is a chordal graph, so by Theorem
 2. GRAPHS, IDEALS AND INVARIANTS 79

2.6.10, a tree is shellable, and hence, sequentially Cohen-Macaulay by Theorem

2.6.18.) However, the only sequentially Cohen-Macaulay cycles are C3 and C5 [72,
Proposition 4.1].
We make use of the following result of Herzog and Hibi that links the notions
of componentwise linearity and sequentially Cohen-Macaulayness. We begin by
recalling:
DEFINITION 2.6.21. Let (Id ) denote the ideal generated by all degree d elements
of a homogeneous ideal I. Then I is called componentwise linear if (Id ) has a linear
resolution for all d.
DEFINITION 2.6.22. If I is a squarefree monomial ideal, then the Alexander dual
of I = (x1,1 ⋅ ⋅ ⋅ x1,s1 , . . . , xt,1 ⋅ ⋅ ⋅ xt,st ) is the ideal
I ∨ = (x1,1 , . . . , x1,s1 ) ∩ ⋅ ⋅ ⋅ ∩ (xt,1 , . . . , xt,st ).
If I is a square-free monomial ideal we write I[d] for the ideal generated by all
the squarefree monomial ideals of degree d in I.
THEOREM 2.6.23. ([100]) Let I be a squarefree monomial ideal of R. Then (a)
R/I is sequentially Cohen-Macaulay if and only if I ∨ is componentwise linear. (b)
I is componentwise linear if and only if I[d] has a linear resolution for all d ≥ 0.
LEMMA 2.6.24. Let G be a bipartite graph with bipartition { x1 , . . . , xm },
{y1 , . . . , yn }. If G is sequentially Cohen-Macaulay, then there is v ∈ VG with
deg(v) = 1.
PROOF. We may assume that m ≤ n and that G has no isolated vertices. Let J
be the Alexander dual of I = I(G) and let L = J[n] be the monomial ideal generated
by the square-free monomials of J of degree n. We may assume that L is generated
by g1 , . . . , gq , where g1 = y1 y2 ⋅ ⋅ ⋅ yn and g2 = x1 ⋅ ⋅ ⋅ xm y1 ⋅ ⋅ ⋅ yn−m . Consider the
linear map
ϕ
Rq −→ R (ei 7→ gi ).
The kernel of this map is generated by syzygies of the form
(g j / gcd(gi , g j ))ei − (gi / gcd(gi , g j ))e j .
Since the vector α = x1 ⋅ ⋅ ⋅ xm e1 − yn−m+1 ⋅ ⋅ ⋅ yn e2 is in ker(ϕ) and since ker(ϕ) is
generated by linear syzygies (see Theorem 2.6.23), there is a linear syzygy of L of
the form x j e1 − zek , where z is a variable, k ∕= 1. Hence x j (y1 ⋅ ⋅ ⋅ yn ) = z(gk ) and gk =
x j y1 ⋅ ⋅ ⋅ yi−1 yi+1 ⋅ ⋅ ⋅ yn for some i. Because the support of gk is a vertex cover of G,
we get that the complement of the support of gk , i.e., {yi , x1 , . . . , x j−1 , x j+1 , . . . , xm },
is an independent set of G. Thus yi can only be adjacent to x j , i.e., deg(yi ) = 1. □
The next theorem generalizes a result of [66] which shows that a bipartite graph
G is Cohen-Macaulay if and only if ∆G has a pure shelling.
80 2. GRAPHS, IDEALS AND INVARIANTS

THEOREM 2.6.25 ([178]). Let G be a bipartite graph. Then G is shellable if and

only if G is sequentially Cohen-Macaulay.
PROOF. Since G shellable implies G sequentially Cohen-Macaulay (Theo-
rem 2.6.18) we only need to show the converse. Assume that G is sequentially
Cohen-Macaulay. The proof is by induction on the number of vertices of G. By
Lemma 2.6.24 there is a vertex x1 of G of degree 1. Let y1 be the vertex of G adjacent
to x1 . Consider the subgraphs G1 = G ∖ ({ x1 }∪ NG (x1 )) and G2 = G ∖ ({y1 }∪ NG (y1 )).
By Theorem 2.6.19 G1 and G2 are sequentially Cohen-Macaulay. Hence ∆G1 and
∆G2 are shellable by the induction hypothesis. Therefore ∆G is shellable by Theo-
rem 2.6.7. □
As we saw in Corollary 2.6.8, one can verify recursively that a bipartite graph is
shellable. The above theorem, therefore, implies the same for sequentially Cohen-
Macaulay bipartite graphs. In particular, we have:
COROLLARY 2.6.26 ([178]). Let G be a bipartite graph. Then G is sequentially
Cohen-Macaulay if and only if there are adjacent vertices x and y with deg(x) = 1
such that the bipartite graphs G ∖ ({ x}∪ NG (x)) and G ∖ ({y}∪ NG (y)) are sequentially
Cohen-Macaulay.
Recently Van Tuyl [177] has shown that the independence complex ∆G must be
vertex decomposable for any bipartite graph G whose edge ring R/I(G) is sequen-
tially Cohen-Macaulay. Thus Theorem 2.6.25 remains valid if we replace shellable
by vertex decomposable.
Clutters with the free vertex property are shellable. We now extend the
scope of this section to include a special family of hypergraphs called clutters.
The results of this section give a proof to a result of Faridi on the sequentially
Cohen-Macaulayness of simplicial forests.
The study of algebraic and combinatorial properties of edge ideal of clutters and
hypergraphs (e.g., Betti numbers, normality, normally torsion freeness, shellability)
is of current, see [51, 52, 60, 68, 80, 93, 95, 122, 138] and the references there. In
Chapter 4 we will study some connections of blowup algebras of edge ideal with
combinatorial optimization problems.
A clutter 𝒞 with vertex set X = { x1 , . . . , xn } is a family of subsets of X, called
edges, none of which is included in another. The set of vertices and edges of 𝒞 are
denoted by V𝒞 and E𝒞 respectively. A basic example of a clutter is a graph. For a
thorough study of clutters—that includes 18 conjectures in the area—from the point
of view of combinatorial optimization see [38].
Let R = k[x1 , . . . , xn ] be a polynomial ring over a field k and let I be an ideal
of R minimally generated by a finite set { xv1 , . . . , xvq } of square-free monomials.
We associate to the ideal I a clutter 𝒞 by taking the set of indeterminates V𝒞 =
{ x1 , . . . , xn } as the vertex set and E𝒞 = {S 1 , . . . , S q } as the edge set, where S i =
 2. GRAPHS, IDEALS AND INVARIANTS 81

supp(xvi ) is the support of xvi , i.e., S i is the set of variables that occur in xvi . For
this reason I is called the edge ideal of 𝒞 and is denoted I = I(𝒞 ). Edge ideals of
clutters are also called facet ideals [68] because S 1 , . . . , S q are exactly the facets of
the simplicial complex ∆ = ⟨S 1 , . . . , S q ⟩ generated by S 1 , . . . , S q .
A subset C ⊂ X is a minimal vertex cover of the clutter 𝒞 if: (i) every edge of
𝒞 contains at least one vertex of C, and (ii) there is no proper subset of C with the
first property. If C only satisfies condition (i), then C is called a vertex cover of 𝒞 .
Notice that 𝔭 is a minimal prime of I = I(𝒞 ) if and only if 𝔭 = (C) for some minimal
vertex cover C of 𝒞 . In particular, if D1 , . . . , Dt is a complete list of the minimal
vertex covers of 𝒞 , then
I(𝒞 ) = (D1 ) ∩ (D2 ) ∩ ⋅ ⋅ ⋅ ∩ (Dt ).
Because I = I(𝒞 ) is a square-free monomial ideal, it also corresponds to a
simplicial complex via the Stanley-Reisner correspondence. We let ∆𝒞 represent
this simplicial complex. As for the case of graphs this complex is called the
independence complex of 𝒞 . Note that F is a facet of ∆𝒞 if and only if X ∖ F is
a minimal vertex cover of 𝒞 and that F is a face of ∆𝒞 if and only if F does not
contains any edge of 𝒞 . As for graphs, we may say that the clutter 𝒞 is shellable if
∆𝒞 is shellable.
LEMMA 2.6.27. Let 𝒞 be a clutter with minimal vertex covers D1 , . . . , Dt . If ∆𝒞
is shellable and A ⊂ V𝒞 is a set of vertices, then the Stanley-Reisner complex ∆I ′ of
the ideal ∩
I′ = (Di )
Di ∩A=∅
is shellable with respect to the linear ordering of the facets of ∆I ′ induced by the
shelling of the simplicial complex ∆𝒞 .
PROOF. Let H1 , . . . , Ht be a shelling of ∆𝒞 . We may assume that Hi = V𝒞 ∖ Di
for all i. Let Hi and H j be two facets of ∆I ′ with i < j, i.e., A ∩ Di = ∅ and
A ∩ D j = ∅. By the shellability of ∆𝒞 , there is an x ∈ H j ∖ Hi and an ` < j such
that H j ∖ H` = { x}. It suffices to prove that D` ∩ A = ∅. If D` ∩ A ∕= ∅, pick
z ∈ D` ∩ A. Then z ∈/ Di ∪ D j and z ∈ Hi ∩ H j . Since z ∈ / H` (otherwise z ∈
/ D` , a
/ Hi . □
contradiction), we get z ∈ H j ∖ H` , i.e., z = x, a contradiction because x ∈
The following notion of a minor was introduced in [80, 84]. It is inspired by the
notion of a minor in combinatorial optimization [38]. An ideal I ′ is called a minor
of I if there is a subset
X ′ = { xi1 , . . . , xir , x j1 , . . . , x js }
of the set of variables X = { x1 , . . . , xn } such that I ′ is a proper ideal of R′ = k[X ∖ X ′ ]
that can be obtained from a generating set of I by setting xik = 0 and x j` = 1 for all
k, `. The ideal I is also considered to be a minor. A minor of 𝒞 is a clutter 𝒞 ′ on
82 2. GRAPHS, IDEALS AND INVARIANTS

the vertex set V𝒞 ′ = X ∖ X ′ that corresponds to a minor (0) ⊊ I ′ ⊊ R′ . Notice that

the edges of 𝒞 ′ are obtained from I ′ by considering the unique set of square-free
monomials of R′ that minimally generate I ′ . For use below we say xi is a free
variable (resp. free vertex) of I (resp. 𝒞 ) if xi only appears in one of the monomials
xv1 , . . . , xvq (resp. in one of the edges of 𝒞 ). If all the minors of 𝒞 have free vertices,
we say that 𝒞 has the free vertex property. Note that if 𝒞 has the free vertex property,
then so do all of its minors.
LEMMA 2.6.28. Let xn be a free variable of I = I(𝒞 ) = (xv1 , . . . , xvq−1 , xvq ), and
let xvq = xn xu . (a) If 𝒞1 is the clutter associated to J = (xv1 , . . . , xvq−1 ), then C
is a minimal vertex cover of 𝒞 containing xn if and only if C ∩ supp(xu ) = ∅ and
C = { xn } ∪ C ′ for some minimal vertex cover C ′ of 𝒞1 . (b) If 𝒞2 is the clutter
associated to L = (xv1 , . . . , xvq−1 , xu ), then C is a minimal vertex cover of 𝒞 not
containing xn if and only if C is a minimal vertex cover of 𝒞2 .
PROOF. (a) Assume that C is a minimal vertex cover of 𝒞 containing xn . If
C ∩ supp(xu ) ∕= ∅, then C ∖ { xn } is a vertex cover of 𝒞 , a contradiction. Thus
C ∩ supp(xu ) = ∅. Hence it suffices to notice that C ′ = C ∖ { xn } is a minimal vertex
cover of 𝒞1 . The converse also follows readily.
(b) Assume that C is a minimal vertex cover of 𝒞 not containing xn . Let xa be
a minimal generator of I(𝒞2 ), then either xu divides xa or xa = xvi for some i < q.
Then clearly C ∩ supp(xa ) ∕= ∅ because C ∩ A ∕= ∅, where A = supp(xu ). Thus C
is a vertex cover of 𝒞2 . To prove that C is minimal take C ′ ⊊ C. We must show
that there is an edge of 𝒞2 not covered by C ′ . As C is a minimal vertex cover of 𝒞 ,
there is xvi such that supp(xvi ) ∩ C ′ = ∅. If xvi is a minimal generator of 𝒞2 there is
nothing to prove, otherwise xu divides xvi and the edge A of 𝒞2 is not covered by C ′ .
The converse also follows readily. □
THEOREM 2.6.29 ([178]). If the clutter 𝒞 has the free vertex property, then the
independence complex ∆𝒞 is shellable.
PROOF. We proceed by induction on the number of vertices of 𝒞 . Let xn be a
free variable of I = I(𝒞 ) = (xv1 , . . . , xvq−1 , xvq ). We may assume that xn occurs in xvq .
Hence we can write xvq = xn xu for some xu such that xn ∈ / supp(xu ). For use below
we set A = supp(xu ). Consider the ideals J = (xv1 , . . . , xvq−1 ) and L = (J, xu ). Then
J = I(𝒞1 ) and L = I(𝒞2 ), where 𝒞1 and 𝒞2 are the clutters defined by the ideals J and
L, respectively. Notice that J and L are minors of the ideal I obtained by setting
xn = 0 and xn = 1, respectively. The vertex set of 𝒞i is V𝒞i = X ∖ { xn } for i = 1, 2.
Thus ∆𝒞1 and ∆𝒞2 are shellable by the induction hypothesis. Let F1 , . . . , Fr be the
facets of ∆𝒞 that contain xn and let G1 , . . . , G s be the facets of ∆𝒞 that do not contain
xn . Set Ci = X ∖ Gi and Ci′ = Ci ∖ { xn } for i = 1, . . . , s. Then C1 , . . . , C s is the set of
minimal vertex covers of 𝒞 that contain xn , and by Lemma 2.6.28(a) C1′ , . . . , C ′s is
the set of minimal vertex covers of 𝒞1 that do not intersect A. One has the equality
 2. GRAPHS, IDEALS AND INVARIANTS 83

Gi = V𝒞1 ∖ Ci′ for all i. Hence, by the shellability of ∆𝒞1 and using Lemma 2.6.27,
we may assume that G1 , . . . , G s is a shelling for the simplicial complex generated
by G1 , . . . , G s . By Lemma 2.6.28(b) one has that C is a minimal vertex cover of 𝒞
not containing xn if and only if C is a minimal vertex cover of 𝒞2 . Thus, F is a facet
of ∆𝒞 that contains xn , i.e., F = F ′ ∪ { xn } if and only if F ′ is a facet of ∆𝒞2 . By
induction we may also assume that F1′ = F1 ∖ { xn }, . . . , Fr′ = Fr ∖ { xn } is a shelling
of ∆𝒞2 . We now prove that
F1 , . . . , Fr , G1 , . . . , G s with Fi = Fi′ ∪ { xn }
is a shelling of ∆𝒞 . We need only show that given G j and Fi there is a ∈ G j ∖ Fi
and F` such that G j ∖ F` = {a}. We can write
G j = X ∖ C j and Fi = X ∖ Ci ,
where C j (resp. Ci ) is a minimal vertex cover of 𝒞 containing xn (resp. not
containing xn ). Recall that A = supp(xu ) is an edge of 𝒞2 . Notice the following:
(i) C j = C ′j ∪ { xn } for some minimal vertex cover C ′j of 𝒞1 such that A ∩ C ′j = ∅,
and (ii) Ci is a minimal vertex cover of 𝒞2 . From (i) we get that A ⊂ G j . Observe
that A ∕⊂ Fi , otherwise A ∩ Ci = ∅, a contradiction because Ci must cover the edge
A = supp(u). Hence there is a ∈ A ∖ Fi and a ∈ G j ∖ Fi . Since C ′j ∪ {a} is a vertex
cover of 𝒞 , there is a minimal vertex cover C` of 𝒞 contained in C ′j ∪ {a}. Clearly
a ∈ C` because C` has to cover xu and C ′j ∩ A = ∅. Thus F` = X ∖ C` is a facet of
∆𝒞 containing xn . To finish the proof we now prove that G j ∖ F` = {a}. We know
/ C` , a contradiction. Thus a ∈ G j ∖ F` . Conversely
that a ∈ G j . If a ∈ F` , then a ∈
take z ∈ G j ∖ F` . Then z ∈ / C ′j ∪ { xn } and z ∈ C` ⊂ C ′j ∪ {a}. Hence z = a, as
required. □
Soleyman Jahan and Zheng [160, Theorem 3.4] have given a generalization of
Theorem 2.6.29 using the notion a pretty clean monomial ideal.
The n × q matrix A with column vectors v1 , . . . , vq is called the incidence matrix
of 𝒞 . This matrix has entries in {0, 1}. We say that A (resp. 𝒞 ) is a totally balanced
matrix (resp. clutter) if A has no square submatrix of order at least 3 with exactly
two 1’s in each row and column. According to [151, Corollary 83.3a] a totally
balanced clutter satisfies the free vertex property. Thus we obtain:
COROLLARY 2.6.30. If 𝒞 is a totally balanced clutter, then ∆𝒞 is shellable.
Faridi [68] introduced the notion of a leaf for a simplicial complex ∆. Precisely,
a facet F of ∆ is a leaf if F is the only facet of ∆, or there exists a facet G ∕= F in ∆
such that F ∩ F ′ ⊂ F ∩ G for all facets F ′ ∕= F in ∆. A simplicial complex ∆ is a
simplicial forest if every nonempty subcollection, i.e., a subcomplex whose facets
are also facets of ∆, of ∆ contains a leaf. We can translate Faridi’s definition into
hypergraph language; we call the translated version of Faridi’s leaf a f -leaf.
84 2. GRAPHS, IDEALS AND INVARIANTS

DEFINITION 2.6.31. An edge E of a clutter 𝒞 is an f -leaf if E is the only edge

of 𝒞 , or if there exists an edge H of 𝒞 such that E ∩ E ′ ⊂ E ∩ H for all edges E ′ ∕= E
of 𝒞 . A clutter 𝒞 is an f -forest, if every subclutter of 𝒞 , including 𝒞 itself, contains
an f -leaf.
In [105, Theorem 3.2] it is shown that 𝒞 is an f -forest if and only if 𝒞 is a totally
balanced clutter. Thus we obtain:
COROLLARY 2.6.32. If the clutter 𝒞 is an f -forest, then ∆𝒞 is shellable.
We now recover the main result of Faridi [68]:
COROLLARY 2.6.33. Let I = I(∆) be the facet ideal of a simplicial forest. Then
R/I(∆) is sequentially Cohen-Macaulay.
PROOF. If ∆ = ⟨F1 , . . . , F s ⟩, then I(∆) is also the edge ideal of the clutter 𝒞
whose edge set is E𝒞 = {F1 , . . . , F s }. Now apply Corollary 2.6.32 and Theorem
2.6.18. □
Recall that a clutter 𝒞 is called d-uniform if all its edges have size d. An ideal
I ⊂ R is called normally torsion free if Ass(R/I i ) is contained in Ass(R/I) for all
i ≥ 1 and I ∕= R.
THEOREM 2.6.34 ([94]). Let 𝒞 be a d-uniform clutter. Assume that I(𝒞 ) is
normally torsion free of height two. Then 𝒞 is Cohen-Macaulay if and only if (i) 𝒞
is unmixed, and (ii) there is a partition
X 1 = { x11 , x21 }, . . . , X d = { x1d , x2d }
of X and a perfect matching e1 = { x11 , . . . , x1d }, e2 = { x21 , . . . , x2d } of 𝒞 so that all
edges of 𝒞 have the form { xi11 , . . . , xidd } for some non decreasing sequence 1 ≤ i1 ≤
⋅ ⋅ ⋅ ≤ id ≤ 2.

Exercises
2.6.35. Let 𝒦m,n be the complete bipartite graph. Prove that if m, n ≥ 2, then
𝒦m,n is not shellable. If m = 1 and n ≥ 1, prove that 𝒦m,n is shellable.
2.6.36 ([66]). Prove that a bipartite graph G is Cohen-Macaulay if and only if
∆G is pure shellable.
2.6.37. ([105], [160]) Let 𝒞 be a clutter. Prove that the following conditions are
equivalent
(a) 𝒞 is the clutter of the facets of a simplicial forest.
(b) 𝒞 has the free vertex property.
(c) 𝒞 is totally balanced.
2.6.38. Prove that no even cycle can be sequentially Cohen-Macaulay.
 2. GRAPHS, IDEALS AND INVARIANTS 85

2.6.39. [72, Proposition 4.1] Prove that the only sequentially Cohen-Macaulay
cycles are C3 and C5 .

2.7. Bounds for the depth and multiplicity

In this section we estimate the multiplicity and the depth of an edge ring. The
cover complexity of a graph will also be examined with certain detail. The depth
and arithmetical rank of edge ideals of bipartite graphs and trees has been studied
recently in [7, 122, 137].
Let G be a graph and let K be a field. The multiplicity e(G) of G is defined as
the multiplicity of the edge-ring R/I(G). Recall that e(G) is precisely the number
of faces of maximum dimension of the Stanley-Reisner complex ∆G of I(G), i.e.,
the multiplicity of G is equal to the number of independent sets of G with β0 (G)
vertices, see Exercise 1.6.28. The number of maximal independent sets of G will
be denoted by e′ (G), we call this number the independence complexity of G. This
is another invariant of G that will be studied here.
DEFINITION 2.7.1. The family of minimal vertex covers of G will be denoted
by ϒ(G). We call ϒ(G) the blocker of G or the Alexander dual of G. The integer
∣ϒ(G)∣ is called the cover complexity of the graph G.
Recall that a set of vertices of G is a maximal independent set if and only if its
complement is a minimal vertex cover for G. Thus we get that the independence
and cover complexity are equal:
e′ (G) = ∣ϒ(G)∣.
We are interested in algebraic ways of grasping this important number. We refer to
[34] and the references there for additional information about e′ (G). The following
estimate is due to E. Sperner (cf. [73, Theorem 3.1]):
THEOREM 2.7.2 ([161]). If G is a graph with n vertices, then
( )
n
∣ϒ(G)∣ ≤ .
⌊n/2⌋
THEOREM 2.7.3 ([97]). If G is a graph, then e′ (G) ≤ 2α0 (G) .
PROOF. Note that if A is the set of isolated vertices of G, then e′ (G) = e′ (G ∖ A).
Hence one may assume that G has no isolated vertices. Since V(G) ∖ C is a maximal
independent set if and only if C is a minimal vertex cover, it suffices to count the
number of minimal vertex covers of G. Let 𝒞 = ϒ(G) be the family of all minimal
vertex covers of G and let C be a fixed vertex cover of G with α0 (G) vertices. For
0 ≤ i ≤ α0 (G) consider:
𝒞i = {C ′ ∣ C ′ ∈ 𝒞 and ∣C ∩ C ′ ∣ = i}.
86 2. GRAPHS, IDEALS AND INVARIANTS

We claim that the following inequality holds:

α0 (G)
( )
(2.4) ∣𝒞i ∣ ≤ (∀ i ≥ 0).
i
We are going to show that if C ′ and C ′′ are two vertex covers in 𝒞i such that
C ∩ C ′ = C ∩ C ′′ , then C ′ = C ′′ . Using C ∩ C ′ = C ∩ C ′′ we obtain that the neighbors
sets satisfy:
(2.5) NG (C ∖ C ′ ) = NG (C ∖ C ′′ ).
Since C ′ and C ′′ are vertex covers of G one has:
NG (C ∖ C ′ ) ∪ (C ∩ C ′ ) ⊂ C ′ and NG (C ∖ C ′′ ) ∪ (C ∩ C ′′ ) ⊂ C ′′ .
Observe that NG (C ∖ C ′ ) ∪ (C ∩ C ′ ) and NG (C ∖ C ′′ ) ∪ (C ∩ C ′′ ) are vertex covers
of G, because C is a vertex cover of G. From the minimality of C ′ and C ′′ one
concludes the equalities:
NG (C ∖ C ′ ) ∪ (C ∩ C ′ ) = C ′ and NG (C ∖ C ′′ ) ∪ (C ∩ C ′′ ) = C ′′ ,
which together with Eq. (2.5) allow ′ ′′
(α0 (G)us
) to derive the equality C = C , as required.
Therefore since there are only i subsets in C with i vertices, one obtains the
inequality (2.4), as claimed. Hence
α∑0 (G) α∑
0 (G) (
α0 (G)
)
′
e (G) = ∣𝒞∣ = ∣𝒞i ∣ ≤ = 2α0 (G) . □
i
i=0 i=0

PROPOSITION 2.7.4. Let γn = sup {∣ϒ(G)∣ ∣ G has n vertices}. If {Fn } is the

Fibonacci sequence given by F0 = F1 = 1 and Fn = Fn−1 + Fn−2 , then γn ≤ Fn for
n ≥ 1.
PROOF. The proof is by induction on n. The required inequality is clearly
satisfied if n ≤ 4. Assume γk ≤ Fk for k < n and n ≥ 5. Let G be a graph
with n vertices. We write G = G1 ∪ G2 where G1 is a connected component of
G with at least two vertices. Let x be a vertex in G1 and x1 , . . . , xr the vertices
of G adjacent to x. Let ϒ1 be the set of minimal covers of G that contain x and
ϒ2 = ϒ(G) ∖ ϒ1 . If A ∈ ϒ1 notice that A ∖ { p} is a (possible empty) minimal vertex
cover for G ∖ { x}, hence ∣ϒ1 ∣ ≤ Fn−1 . If A ∈ ϒ2 , then A ∖ { x1 , . . . , xr } is a (possible
empty) minimal vertex cover for G ∖ { x, x1 , . . . , xr }, which implies ∣ϒ2 ∣ ≤ Fn−r−1 .
Therefore γn ≤ Fn as required. □
The following result applies to edge-rings associated to hypergraphs whose
edges all have p vertices (cf. Theorem 2.7.3).
PROPOSITION 2.7.5. Let R = K[x1 , . . . , xn ] be a polynomial ring over a field
K and let I be an ideal of R of height r minimally generated by homogeneous
polynomials f1 , . . . , fq of degree p ≥ 1. Then e(R/I) ≤ pr .
 2. GRAPHS, IDEALS AND INVARIANTS 87

PROOF. One may assume that K is an infinite field by considering a field

extension K ⊂ L, with L an infinite field, and using the functor (⋅) ⊗K L. It is not
hard to see that there are forms g1 , . . . , gq of degree p in I such that g1 , . . . , gr is
a regular sequence and I is generated by g1 , . . . , gq . Hence one may also assume
f1 , . . . , fr is a regular sequence. Consider the subideal I ′ = ( f1 , . . . , fr ). There is a
graded epimorphism of R/I ′ → R/I → 0, where dim(R/I ′ ) = dim(R/I). Therefore
e(R/I) ≤ e(R/I ′ ). On the other hand by Exercise 1.6.26 the Hilbert series of R/I ′ is
given by
(1 + t + ⋅ ⋅ ⋅ + t p−1 )r
F(R/I ′ , t) = .
(1 − t)n−r
Making t = 1 in the numerator yields the required bound. □
PROPOSITION 2.7.6. Let G be a graph with n vertices. If G has a maximal
independent cover 𝒞 , then ⌈n/β0 (G)⌉ ≤ e(G) and this lower bound is sharp.
PROOF. Let A1 , . . . , Ar be the sets of the cover 𝒞 . Since V(G) = A1 ∪ ⋅ ⋅ ⋅ ∪ Ar
we have n = ∣V(G)∣ ≤ rβ0 (G), which proves the bound. That the bound is the best
possible is shown in [174]. □
For some special types of Cohen-Macaulay hypergraphs 2β0 (G) is a lower bound
for the multiplicity of G [166, Theorem 8.1].
Bounds for the depth of an edge-ring. Let us recall the following expression
for the depth of K[∆G ]; a simple proof can be found in [76].
THEOREM 2.7.7 ([159]). If G is a hypergraph and ∆ = ∆G , then
depth(K[∆]) = 1 + max{i ∣ K[∆i ] is Cohen-Macaulay},
where ∆i = {F ∈ ∆ ∣ dim(F) ≤ i} is the i-skeleton of ∆.
DEFINITION 2.7.8. Let G be a hypergraph. We will denote by β′0 (G) the smallest
number of vertices in any maximal independent set, i.e., the minimum size of a
maximal independent set. This number is called the smallest independence number
of G.
PROPOSITION 2.7.9. If G is a hypergraph and ∆ = ∆G , then
depth K[∆] ≤ β′0 (G).
PROOF. Assume ∆i is Cohen-Macaulay. According to Theorem 2.7.7 it suffices
to prove that i + 1 ≤ β′0 (G). If i + 1 > β′0 (G), then there is a maximal independent
set F of G with β′0 (G) vertices, a contradiction because the complex ∆i is pure, that
is, all maximal faces of ∆ have dimension i. □
THEOREM 2.7.10. Let G be a bipartite graph without isolated vertices. If G has
n vertices, then ⌊n⌋
depth K[∆G ] ≤ β′0 (G) ≤ .
2
88 2. GRAPHS, IDEALS AND INVARIANTS

PROOF. Let (V1 , V2 ) be a bipartition of G with ∣V1 ∣ ≤ ∣V2 ∣. Note 2∣V1 ∣ ≤ n

because ∣V1 ∣ + ∣V2 ∣ = n. Since V1 is a maximal independent set of vertices one has
β′0 (G) ≤ ∣V1 ∣. Therefore using Proposition 2.7.9 we conclude
depth K[∆G ] ≤ β′0 (G) ≤ ∣V1 ∣ ≤ n/2. □
The bound above does not extend to non bipartite graphs. Next we present a
family of connected graphs showing that the depth of the edge-ring K[∆G ] can be
arbitrarily large with respect to n/2.
PROPOSITION 2.7.11. Let 𝒦r be the complete graph with r vertices and let Gr,i
be the graph obtained by attaching i lines at each vertex of 𝒦r . Then
depth(R/I(Gr,i )) = (r − 1)i + 1.
PROOF. Let R = K[x1 , . . . , xn ] be a polynomial ring and let x1 , . . . , xn be the
vertices of Gr,i . Note that Gr,i has n = r(i + 1) vertices. We proceed by induction on
r. If r = 1, then G1,i is an star and clearly K[∆G1,i ] has depth 1. Assume r ≥ 2. Set
G = Gr,i and I = I(G). Let x1 be any vertex of 𝒦r and consider the exact sequence
x
1
0 −→ R/(I : x1 ) −→ R/I −→ R/(x1 , I) −→ 0. (∗)
It is not hard to see that (I : x1 ) is a face ideal generated by i + (r − 1) variables.
Hence depth(R/(I : x1 )) = (r − 1)i + 1. On the other hand note the equality (I, x1 ) =
(x1 ) + I(Gr−1,i ). Since x1 is adjacent in G to exactly i-vertices of degree 1 and those
vertices do not occur in (I, x1 ), by induction hypothesis we derive
depth(R/(x1 , I)) = i + [(r − 2)i + 1] = (r − 1)i + 1.
Altogether the ends of the exact sequence (∗) have depth equal to (r − 1)i + 1. Thus
by the depth Lemma 1.5.30 we get that R/I has depth (r − 1)i + 1, as required. □
EXAMPLE 2.7.12. Let K = ℚ and let G = G4,3 be the following graph
x xr10
x7r r 8 x
r 11

x r r 5
@ x x9 r rx
6 12
@
@
@
x4 r r
x1 x13@
@r r x14

r @r x
x3 r r 15
x2 x16
Then by Proposition 2.7.11 we have depth R/I(G) = 10.
COROLLARY 2.7.13. Let 𝒦r be a complete graph and let G = Gr,i1 ,...,ir be the
graph obtained by attaching i j lines at each vertex v j of 𝒦r . If i j ≤ i j+1 for all j,
then
depth(R/I(G)) = i1 + ⋅ ⋅ ⋅ + ir−1 + 1.
 2. GRAPHS, IDEALS AND INVARIANTS 89

PROOF. Note that the number of vertices of G is n = i1 + ⋅ ⋅ ⋅ + ir + r. Let x1 be a

vertex of 𝒦r with degG (x1 ) = ir + (r − 1). Consider the exact sequence
x
1
0 −→ R/(I : x1 ) −→ R/I −→ R/(x1 , I) −→ 0.
As in the proof of Proposition 2.7.11 (by induction on r) we obtain:
depth R/(I : x1 ) = i1 + ⋅ ⋅ ⋅ + ir−1 + 1,
depth R/(I, x1 ) = i1 + ⋅ ⋅ ⋅ + ir−2 + ir + 1.
By the depth Lemma 1.5.30 we get depth(R/I) = i1 + ⋅ ⋅ ⋅ + ir−1 + 1, as required. □
The method of proof of Theorem 2.3.8 will be adapted to prove:
THEOREM 2.7.14 ([81]). Let G be a graph with n vertices and let
c = (β0 (G) − β′0 (G) + 2)β0 (G) − (β0 (G) − β′0 (G) + 1)n.
If G has no isolated vertices, then c ≤ 0.
PROOF. Assume c > 0. To derive a contradiction it suffices to construct a set
H of isolated vertices of G with ∣H ∣ ≥ c. There is a family ℱ = { M1 , . . . , Mm } of
maximal independent sets of G satisfying:
(a) V(G) = M1 ∪ ⋅ ⋅ ⋅ ∪ Mm ,
(b) M1 ⊊ M1 ∪ M2 ⊊ ⋅ ⋅ ⋅ ⊊ M1 ∪ ⋅ ⋅ ⋅ ∪ Mm−1 ⊊ V(G), and
(c) β0 (G) = ∣ M1 ∣ ≥ ∣ M2 ∣ ≥ ⋅ ⋅ ⋅ ≥ ∣ Mm ∣.
Consider the sets Hi = M1 ∩ ⋅ ⋅ ⋅ ∩ Mi , Ki = M1 ∪ ⋅ ⋅ ⋅ ∪ Mi , i = 1, . . . , m, clearly
Hi is Ki -isolated for all i. There are s1 > ⋅ ⋅ ⋅ > sm = 0 such that
∣K1 ∣ = n − s1 < ∣K2 ∣ = n − s2 < ⋅ ⋅ ⋅ < ∣Km ∣ = n − sm .
We claim that ∣Hi ∣ ≥ 2β0 (G) − n + si − ij=1 (β0 (G) − ∣ M j ∣) for 1 ≤ i ≤ m. To
∑
prove it we use induction on i. For i = 1 is clear. Thus we assume that the inequality
is true for an integer i with 1 < i < m. Let us prove the inequality for the integer
i + 1. First note the equality:
(2.6) ∣ Mi+1 ∖ Ki ∣ = ∣Ki+1 ∣ − ∣Ki ∣ = si − si+1 .
because Ki+1 = Ki ∪ (Mi+1 ∖ Ki ). Second note the equality
(2.7) ∣ Mi+1 ∩ (Ki ∖ Hi )∣ ≤ β0 (G) − ∣Hi ∣,
this equality holds because otherwise there would be an independent set T ⊂ Ki ∖ Hi
with β0 (G) − ∣Hi ∣ + 1 vertices and T ∪ Hi would be an independent set of G with
more than β0 (G) vertices, a contradiction.
From the equality
Mi+1 = (Mi+1 ∩ Hi ) ∪ (Mi+1 ∩ (Ki ∖ Hi )) ∪ (Mi+1 ∖ Ki ),
90 2. GRAPHS, IDEALS AND INVARIANTS

together with Eq. (2.6) and Eq. (2.7) we obtain:

∣ Mi+1 ∣= ∣Hi+1 ∣ + ∣ Mi+1 ∩ (Ki ∖ Hi )∣ + ∣ Mi+1 ∖ Ki ∣
≤ ∣Hi+1 ∣ + (β0 (G) − ∣Hi ∣) + (si − si+1 ).
Hence ∣Hi+1 ∣ ≥ 2β0 (G) − n+ si+1 − i+1
∑
j=1 (β0 (G) −∣ M j ∣), and the proof of the claim is
complete. To finish the proof of the theorem note that H = Hm is the set of isolated
vertices with the desired property. Indeed using (b) we readily obtain
∑m
∣Hm ∣ ≥ 2β0 (G) − n − (β0 (G) − ∣ M j ∣)
j=1
≥ 2β0 (G) − n − (n − β0 (G))(β0 (G) − β′0 (G)). □
An immediate consequence of Theorem 2.7.14 is:
COROLLARY 2.7.15. Let G be a graph without isolated vertices. If G has n
vertices, then
n + β0 (G)n − 2β0 (G) − β0 (G)2
depth K[∆G ] ≤ β′0 (G) ≤ x0 = ,
(n − β0 (G))
β0 (G) ≤ α0 (G)[1 + (β0 (G) − β′0 (G))].

Exercises
2.7.16. Find e(G) and e′ (G) if G is a disjoint union of stars.
2.7.17. Let F0 = F1 = 1, F2 = 2 and Fn = Fn−2 + Fn−3 for n ≥ 3. If Pn is the
path { x1 , . . . , xn }, prove that Fn = ∣ϒ(Pn )∣.
2.7.18. If T is a forest, then ∣ϒ(T )∣ ≥ α0 (T ) + 1.
2.7.19. Consider the graph G which is the disjoint union of k complete graphs
with two vertices. Prove that G satisfies e′ (G) = e(G) = 2α0 (G) and k = α0 (G).
Thus the bound of Theorem 2.7.3 is optimal. For a complete characterization of all
graphs achieving this bound see [97].

2.8. Suspensions of graphs

In this section we give graph theoretical interpretations for the quotients of certain
ideals attached to graphs.
Let G0 be a graph on the vertex set V = {y1 , . . . , yn } and I = I(G0 ). Take a
new set of variables x1 , . . . , xn and define S (G0 ) to be the new graph obtained by
attaching to each vertex yi a new vertex xi and the edge { xi , yi }. For use below we
set
K = I(S (G0 )) = (x1 y1 , . . . , xn yn , I) ⊂ R = k[x, y],
 2. GRAPHS, IDEALS AND INVARIANTS 91

where k is a field. The S in the notation is intended for suspension or attaching

whiskers to the base graph. The significance of this construction lies partly in the
following restatement of Theorem 2.4.13:
THEOREM 2.8.1. If G is a graph, then G is a Cohen-Macaulay tree if and only
if G = S (G0 ) for some tree G0 .
PROPOSITION 2.8.2. Let G0 be∑a graph and let G = S (G0 ) be its suspension.
Then the multiplicity of R/I(G) is di=−1 fi , where fi is the number of i-faces of the
Stanley-Reisner complex of I(G0 ).
PROOF. Let y1 , . . . , yn be the vertices of G0 . By Proposition 2.4.11, the set of
linear forms {yi − xi }ni=1 is a regular system of parameters for R/I(G). The reduction
of R/I(G) modulo these forms is
k[y]/(y21 , . . . , y2n , I(G0 )),
whose k-vector space basis is formed by the standard monomials, that is, the faces
of the attached simplicial complex. □
The next proposition describes the first quotient of S (G0 ) relative to the regular
sequence z = { x1 y1 , . . . , xn yn }.
PROPOSITION 2.8.3. Let G0 and G1 be two copies of the same graph on disjoint
sets of vertices {y1 , . . . , yn } and { x1 , . . . , xn } respectively. Let ϒ be the set of minimal
vertex covers of G1 and
L = ({ xi1 ⋅ ⋅ ⋅ xis ∣ { xi1 , . . . , xis } ∈ ϒ(G1 )}).
Then ((z) : K) = (z, L), where K = I(S (G0 )).
PROOF. Take { xi1 , . . . , xir } ∈ ϒ and yk y` ∈ I. Since i s ∈ {k, `} for some s
we obtain xi1 ⋅ ⋅ ⋅ xir yk y` ∈ (z). This shows (z, L) ⊂ ((z) : K). Conversely, assume
M is a monomial and M ∈ ((z) : K)) ∖ (z). Let {yk , y` } be any edge in G0 . Since
Myk y` = xt yt M1 , either xk divides M or x` divides M; in either case, M ∈ L. □
COROLLARY 2.8.4 ([187]). The type of R/I(S (G0 )) is equal to ∣ϒ(G0 )∣.
Observe that these results can be used to compute the primary decomposition
of an edge ideal:
EXAMPLE 2.8.5. Let I(G0 ) = (y1 y3 , y1 y4 , y2 y4 , y2 y5 , y3 y5 , y3 y6 , y4 y6 ). The ideal
((z) : I(S (G0 )) is generated by z = { x1 y1 , . . . , x6 y6 } together with:
x2 x3 x4 , x3 x4 x5 , x1 x2 x5 x6 , x1 x2 x3 x6 , x1 x4 x5 x6 ,
and these monomials correspond to the minimal primes of I(G0 ). The type of
R/I(S (G0 )) is 5 and the multiplicity of the face ring of I(G0 ) is 2.
92 2. GRAPHS, IDEALS AND INVARIANTS

Exercises
2.8.6. Let I = (xi y j ∣ 1 ≤ i ≤ j ≤ n) ⊂ R = k[x, y]. Prove that R/I has type n
and its a-invariant is −(n − 1).
Chapter 3
Polyhedral Sets, Optimization and Semigroups

This chapter deals with polyhedra associated with point configurations. An aim
here is to study algebraic properties of affine semigroups and to explore some links
to combinatorial properties.
In Section 3.1 we introduce several notions and some essential results from poly-
hedral geometry and linear programming (e.g., finite basis theorem, fundamental
theorem of linear inequalities, LP duality theorem, irreducible representations and
the topology of polyhedra). Excellent references for these two areas, in alphabetical
order, are [120, 151, 200]. Section 3.1 contains fairly standard material.
In the other sections of this chapter we study relative volumes of lattice poly-
topes, unimodular coverings, affine semigroups and their integral closures, circuits
of vector matroids, Hilbert bases and optimization problems. We will introduce and
study the Rees cone of a clutter matrix. Then we characterize the max-flow min-cut
property in terms of Hilbert bases of Rees cones and the integrality of set covering
polyhedra. Later on in Chapter 4 we study some other combinatorial problems and
their links to blowup algebras.

3.1. Polyhedral sets and cones

An affine space or linear variety in ℝn is by definition a translation of a linear
subspace of ℝn . Let A ⊂ ℝn . Recall that aff(A), the affine space generated by A, is
the set of all affine combinations of points in A:
aff(A) = {λ1 v1 + ⋅ ⋅ ⋅ + λr vr ∣ vi ∈ A, λ1 + ⋅ ⋅ ⋅ + λr = 1, λi ∈ ℝ}.
There is a unique linear subspace V of ℝn such that
aff(A) = x0 + V,
for some x0 ∈ ℝn . The dimension of A is defined as dim A = dim ℝ (V).
DEFINITION 3.1.1. An affine map is a function between two affine spaces that
preserves affine combinations.
93
94 3. POLYHEDRAL SETS, OPTIMIZATION AND SEMIGROUPS

A point x ∈ ℝn is called a convex combination of v1 , . . . , vr ∈ ℝn if there are

non negative real numbers λ1 , . . . , λr such that
x = λ1 v1 + ⋅ ⋅ ⋅ + λr vr and λ1 + ⋅ ⋅ ⋅ + λr = 1.
The convex hull of A, denoted by conv(A), is the set of all convex combinations
of points in A. If A = conv(A) we say that A is a convex set.
DEFINITION 3.1.2. A convex polytope P ⊂ ℝn is the convex hull of a finite set
of points v1 , . . . , vr in ℝn , that is, P = conv(v1 , . . . , vr ). We also denote a polytope
by 𝒫 .
The inner product of two vector x = (x1 , . . . , xn ) and y = (y1 , . . . , yn ) in ℝn is
defined by
⟨ x, y⟩ = x ⋅ y = x1 y1 + ⋅ ⋅ ⋅ + xn yn .
DEFINITION 3.1.3. Given x = (x1 , . . . , xn ) ∈ ℝn its Euclidean norm is
∥ x∥ = ⟨ x, x⟩.
√

We shall always assume that a subset of ℝn has the topology induced by the
usual topology of ℝn .
DEFINITION 3.1.4. A point a of a set A in ℝn is said to be a relative interior
point of A if there exists r > 0 such that
Br (a) ∩ aff(A) ⊂ A,
where Br (a) = { x∣ ∥ x − a∥ < r}.
Let A ⊂ ℝn . We denote the closure of A by A. The set of relative interior
points of A, denoted by ri(A), is called the relative interior of A. The set A ∖ ri(A)
is called the relative boundary of A and is denoted by rb(A) or ∂(A), points in rb(A)
are called the relative boundary points of A. If dim(A) = n, the relative interior of
A is sometimes denoted by Ao .
An affine space of ℝn of dimension n − 1 is called an affine hyperplane. Given
a ∈ ℝn ∖ {0} and c ∈ ℝ, define the affine hyperplane
H(a, c) = { x ∈ ℝn ∣ ⟨ x, a⟩ = c}.
Notice that any affine hyperplane of ℝn has this form. The two closed halfspaces
bounded by H(a, c) are
H + (a, c) = { x ∈ ℝn ∣ ⟨ x, a⟩ ≥ c} and H − (a, c) = H + (−a, −c).
If a and c are rational, the affine hyperplane (resp. the halfspace H + (a, c)) is called
a rational hyperplane. If c = 0, the set Ha will denote the hyperplane of ℝn through
the origin with normal vector a, that is,
Ha := H(a, 0) = { x ∈ ℝn ∣ ⟨ x, a⟩ = 0}.
 3. POLYHEDRAL SETS, OPTIMIZATION AND SEMIGROUPS 95

The two closed halfspaces bounded by Ha are denoted by

Ha+ = { x ∈ ℝn ∣ ⟨ x, a⟩ ≥ 0} and Ha− = { x ∈ ℝn ∣ ⟨ x, a⟩ ≤ 0}.
DEFINITION 3.1.5. A polyhedral set or (convex) polyhedron is a set Q in ℝn
which is the intersection of a finite number of closed halfspaces of ℝn . The set ℝn
is considered a polyhedron.
Notation The transpose of a matrix A (resp. vector x) will be denoted by At (resp.
xt ). Often a vector x will denote a column vector or a row vector, from the context
the meaning should be clear.
Thus a polyhedral set can be represented as
Q = { x ∈ ℝn ∣ Ax ≤ b}
for some matrix A and for some vector b. Note that if a = (ai ) and c = (ci ) are
vectors in ℝn , then a ≤ c means ai ≤ ci for all i. If A and b have rational entries, Q
is called a rational polyhedron.
Since the intersection of an arbitrary family of convex sets in ℝn is convex, one
derives that any polyhedral set is convex and closed.
DEFINITION 3.1.6. Let Q be a closed convex set in ℝn . An affine hyperplane
H of ℝn is called a supporting hyperplane of Q if Q is contained in one of the two
closed halfspaces bounded by H and Q ∩ H ∕= ∅.
DEFINITION 3.1.7. A proper face of a polyhedral set Q is a set F ⊂ Q such that
there is a supporting hyperplane H(y, c) satisfying the conditions:
(a) F = Q ∩ H(y, c) ∕= ∅,
(b) Q ∕⊂ H(y, c), and Q ⊂ H + (y, c) or Q ⊂ H − (y, c).
The improper faces of a polyhedral set Q are ∅ and Q itself.
DEFINITION 3.1.8. A proper face F of a polyhedral set Q ⊂ ℝn is called a facet
of Q if
dim(F) = dim(Q) − 1.
DEFINITION 3.1.9. If C ⊂ ℝn is closed under linear combinations with non
negative real coefficients, we say that C is a convex cone. A polyhedral cone is a
convex cone which is also a polyhedral set.
The set of non-negative real numbers is denoted by:
ℝ+ = { x ∈ ℝ∣ x ≥ 0}.
96 3. POLYHEDRAL SETS, OPTIMIZATION AND SEMIGROUPS

If 𝒜 is a set of points in ℝn , the cone generated by 𝒜, denoted by ℝ+ 𝒜 or cone(𝒜),

is defined as
⎧ ⎫
⎨∑ ⎬
ℝ+ 𝒜 = ai βi ai ∈ ℝ+ and βi ∈ 𝒜, for all i .
⎩ ⎭
finite
Given a vector v ∈ ℝn we define:
ℝ+ v := {λv∣ λ ∈ ℝ+ }.
DEFINITION 3.1.10. We say that C is a finitely generated cone if C = ℝ+ 𝒜, for
some finite set 𝒜 = {v1 , . . . , vq }.
PROPOSITION 3.1.11. If C is a closed convex cone and H is a supporting hyper-
plane of C, then H is a hyperplane passing through the origin.
PROOF. Let H = H(a, c). Since 0 ∈ C ⊂ H + , one has c ≤ 0. If C ∩ H = {0},
then c = 0 as required. Assume C ∩ H ∕= {0} and pick 0 ∕= z ∈ C such that ⟨z, a⟩ = c.
Using that tz ∈ C ⊂ H + (a, c) for all t ≥ 0, one derives
tc = t⟨z, a⟩ = ⟨tz, a⟩ ≥ c ∀ t ≥ 0 ⇒ c = 0. □
PROPOSITION 3.1.12. Let 𝒜 be a finite set of points in ℤn and let F be a face of
ℝ + 𝒜.
(a) If F ∕= {0}, then F = ℝ+ 𝒜′ for some 𝒜′ ⊂ 𝒜.
(b) If dim(F) = 1 and 𝒜 ⊂ ℕn , then F = ℝ+ α for some α ∈ 𝒜.
(c) If dim(ℝ+ 𝒜) = n and F is a facet defined by the supporting hyperplane
Ha , then Ha is generated by a linearly independent subset of 𝒜.
PROOF. Let F = ℝ+ 𝒜 ∩ Ha with ℝ+ 𝒜 ⊂ Ha− . Then F is equal to the cone
generated by the set 𝒜′ = {α ∈ 𝒜∣ ⟨α, a⟩ = 0}. This proves (a). Parts (b) and (c)
follow from (a). □
Most of the notions and results considered thus far make sense if we replace
ℝn by an intermediate field ℚ ⊂ 𝕂 ⊂ ℝ, i.e., we can work in the affine space
𝕂n . However with very few exceptions, like for instance Theorem 3.1.13, we will
always work in the Euclidean space ℝn or ℚn .
For convenience we state the fundamental theorem of linear inequalities:
THEOREM 3.1.13. [150, Theorem 7.1] Let ℚ ⊂ 𝕂 ⊂ ℝ be an intermediate field
and let C ⊂ 𝕂n be a cone generated by 𝒜 = {α1 , . . . , αq }. If α ∈ 𝕂n ∖ C and
t = rank{α1 , . . . , αq , α}, then there exists a hyperplane Ha containing t − 1 linearly
independent vectors from 𝒜 such that ⟨a, α⟩ > 0 and ⟨a, αi ⟩ ≤ 0 for i = 1, . . . , q.
THEOREM 3.1.14 (Farkas Lemma). Let A be an n × q matrix with entries in a
field 𝕂 and let α ∈ 𝕂n . Assume ℚ ⊂ 𝕂 ⊂ ℝ. Then either there exists x ∈ 𝕂q with
Ax = α and x ≥ 0, or there exists v ∈ 𝕂n with vA ≥ 0 and ⟨v, α⟩ < 0, but not both.
 3. POLYHEDRAL SETS, OPTIMIZATION AND SEMIGROUPS 97

PROOF. Let 𝒜 = {α1 , . . . , αq } be the set of column vectors of A. Assume that

there is not x ∈ 𝕂q with Ax = α and x ≥ 0, i.e., α is not in the cone generated by
𝒜. By Theorem 3.1.13 there is a hyperplane Hv such that ⟨v, α⟩ < 0 and ⟨v, αi ⟩ ≥ 0
for all i. Hence vA ≥ 0, as required. If both conditions hold, then
0 > ⟨v, α⟩ = ⟨v, Ax⟩ = ⟨vA, x⟩ ≥ 0,
a contradiction. □
The reader is refer to [200] for other versions of Farkas Lemma; there are quite
a few. The next result says that Farkas lemma can be used to separate a point from
a cone.
COROLLARY 3.1.15. Let C ⊂ 𝕂n be a cone generated by 𝒜 = {α1 , . . . , αq }.
If γ ∈ 𝕂n and γ ∈/ C, then there is a hyperplane through the origin H such that
γ ∈ H − ∖H and C ⊂ H + .
PROOF. Let A be the matrix with column vectors α1 , . . . , αq . By Farkas’s
Lemma there exists µ ∈ 𝕂n such that µA ≥ 0 and ⟨γ, µ⟩ < 0. Thus ⟨µ, αi ⟩ ≥ 0 for
all i. If H is the hyperplane through the origin with normal vector µ we get C ⊂ H + ,
as required. □
COROLLARY 3.1.16. Let 𝒜 be a finite set in ℤn , then
ℤ𝒜 ∩ ℝ+ 𝒜 = ℤ𝒜 ∩ ℚ+ 𝒜 and ℤn ∩ ℝ+ 𝒜 = ℤn ∩ ℚ+ 𝒜,
where ℤ𝒜 is the subgroup of ℤn generated by 𝒜.
PROOF. It follows at once from Theorem 3.1.14. □
THEOREM 3.1.17. If C ⊂ ℝn , then C is a finitely generated cone (resp. finitely
generated cone by rational vectors) if and only if C is a polyhedral cone (resp.
rational polyhedral cone) in ℝn .
PROOF. ⇒) Let (0) ∕= C = ℝ+ 𝒜, where 𝒜 = {α1 , . . . , αm } and r = dim(ℝ+ 𝒜).
Notice that aff(C) is the real vector space generated by 𝒜, because 0 ∈ C. If
aff(C) = C, then C is a polyhedral cone, because C is the intersection of n − r
hyperplanes of ℝn through the origin. Now assume that C ⊊ aff(C). Consider the
family
ℱ = {F ∣ F = Ha ∩ C; dim(F) = r − 1; C ⊂ Ha− }.
By Theorem 3.1.13 the family ℱ is non empty. Notice that ℱ is a finite set because
each F in ℱ is a cone generated by a subset of 𝒜, see Proposition 3.1.12. Assume
that ℱ = {F1 , . . . , F s }, where Fi = Hai ∩ C. We claim that the following equality
holds
(3.8) C = Ha−1 ∩ ⋅ ⋅ ⋅ ∩ Ha−s ∩ aff(C).
To show the equality we proceed by contradiction. Assume there exists α ∕∈ C and
such that α belongs to the right hand side of Eq. (3.8). By Theorem 3.1.13 there is
98 3. POLYHEDRAL SETS, OPTIMIZATION AND SEMIGROUPS

a hyperplane Ha containing linearly independent vectors α1 , . . . , αr−1 and such that

(i) ⟨a, α⟩ > 0, and (ii) ⟨a, αi ⟩ ≤ 0 for i = 1, . . . , m. Thus F = Ha ∩ C = Hak ∩ C
for some 1 ≤ k ≤ s. We may assume that α1 , . . . , αr form a basis of aff(C). Since
α ∈ aff(C) there are scalars λ1 , . . . , λr with
α = λ1 α1 + ⋅ ⋅ ⋅ + λr αr .
Using (ii) we get ⟨a, α⟩ = λr ⟨αr , a⟩ > 0. Hence λr < 0. On the other hand
⟨ak , αr ⟩ < 0 because C ⊂ Ha−k and dim(Fk ) = r − 1. Therefore
⟨ak , α⟩ = λr ⟨ak , αr ⟩ > 0,
|{z} | {z }
<0 <0
a contradiction to α ∈ Hak . −

⇐) Assume C = Hb−1 ∩ ⋅ ⋅ ⋅ ∩ Hb−r , where b1 , . . . , br ∈ ℝn . Consider the cone C ′

generated by b1 , . . . , br . From the first part of the proof we can write
C ′ = ℝ+ {b1 , . . . , br } = Hα−1 ∩ ⋅ ⋅ ⋅ ∩ Hα−m , (∗)
for some set of vectors 𝒜 = {α1 , . . . , αm } in ℝn . Next we show the equality
C = ℝ+ 𝒜. Notice that ⟨bi , α j ⟩ ≤ 0 for all i, j, because bi ∈ C ′ for all i. Thus
ℝ+ 𝒜 ⊂ C. Assume there is α ∈ C ∖ ℝ+ 𝒜. By Corollary 3.1.15 there exists a
hyperplane Ha such that ℝ+ 𝒜 ⊂ Ha− and ⟨a, α⟩ > 0. Hence ⟨αi , a⟩ ≤ 0 for all i and
by (∗) we conclude that a ∈ ℝ+ {b1 , . . . , br }. Therefore we can write
a = λ1 b1 + ⋅ ⋅ ⋅ + λr br (λi ≥ 0).
Notice that since α ∈ C we have
⟨α, a⟩ = λ1 ⟨α, b1 ⟩ + ⋅ ⋅ ⋅ + λr ⟨α, br ⟩ ≤ 0,
contradicting ⟨a, α⟩ > 0. Thus C = ℝ+ 𝒜. The respective statement about the
rationality character of the representations is left as an exercise. □
By Theorem 3.1.17 a polyhedral cone C ⊊ ℝn has two representations:
Minkowski representation: C = ℝ+ ℬ with ℬ = {β1 , . . . , βr } a finite set, and
Implicit representation: C = Hc+1 ∩ ⋅ ⋅ ⋅ ∩ Hc+s for some c1 , . . . , c s ∈ ℝn ∖ {0}.
These two representations satisfy the duality theorem for cones:
(3.9) Hβ+1 ∩ ⋅ ⋅ ⋅ ∩ Hβ+r = ℝ+ c1 + ⋅ ⋅ ⋅ + ℝ+ c s .
This equality follows from the second part of the proof of Theorem 3.1.17. The
dual cone of C is defined as
∩ ∩
C ∗ := Hc+ = Ha+ .
c∈C a∈ℬ
By the duality theorem one has C ∗∗
= C. An implicit representation of C is called
irredundant if none of the closed halfspaces Hc+1 , . . . , Hc+s can be omitted from the
intersection. See Definition 3.1.30 for another type of representations of polyhedra.
 3. POLYHEDRAL SETS, OPTIMIZATION AND SEMIGROUPS 99

REMARK 3.1.18. The left hand side of Eq. (3.9) is an irredundant representation
of C ∗ if and only if no proper subset of ℬ generates C.
COROLLARY 3.1.19. Let C ⊂ ℝn be a finitely generated cone and let
ℱ = {F ∣ F = Ha ∩ C; dim(F) = r − 1; C ⊂ Ha− } = {F1 , . . . , F s } =
∕ ∅,
where Fi = C ∩ Hai , dim(C) = r. Then C = aff(C) ∩ Ha−1 ∩ ⋅ ⋅ ⋅ ∩ Ha−s .
PROOF. It follows from the first part of the proof of Theorem 3.1.17. □
One of the fundamental results in polyhedral geometry is the following remark-
able decomposition theorem for polyhedra. See [150, Corollary 7.1b] for historical
comments and more information about this result.
THEOREM 3.1.20 (Finite basis theorem). If Q is a set in ℝn , then Q is a polyhedral
set (resp. rational) if and only if Q can be expressed as Q = P + C, where P is a
convex polytope (resp. rational) and C is a finitely generated cone (resp. rational).
PROOF. ⇒) Let Q = { x∣ Ax ≤ b} be a polyhedron in ℝn . Consider the set
{( ) }
′ x
C = x ∈ ℝ ; λ ∈ ℝ+ ; Ax − λb ≤ 0 .
n
λ
Notice that C ′ can be written as
{( ) ( )( ) }
x A −b x
C′ = x ∈ ℝn ; λ ∈ ℝ; ≤0 ,
λ 0 −1 λ
where −b is a column vector. Thus C ′ is a polyhedral cone in ℝn+1 . Using
Corollary 3.1.17 we get that C ′ can be expressed as
{( ) ( )}
′ x1 x
C = ℝ+ ,..., m (λi ≥ 0; xi ∈ ℝn ).
λ1 λm
We may assume that λi ∈ {0, 1} for all i. Set
𝒜 = { xi ∣ λi = 1} = { x1 , . . . , xr }, ℬ = { xi ∣ λi = 0} = { xr+1 , . . . , xm },
P = conv(𝒜), and C = ℝ+ (ℬ ). Notice that x ∈ Q if and only if (x, 1) ∈ C ′ . Thus
x ∈ Q if and only if (x, 1) can be written as
() ( ) ( ) ( ) ( )
x x x x x
= µ1 1 + ⋅ ⋅ ⋅ + µr r + µr+1 r+1 + ⋅ ⋅ ⋅ + µm m
1 1 1 0 0
with µi ≥ 0 for all i. Consequently x ∈ Q is and only if x ∈ P + C. Therefore we
obtain Q = P + C.
⇐) Let Q = P + C with P = conv(x1 , . . . , xr ) and C = ℝ+ { xr+1 , . . . , xm }.
Consider the following finitely generated cone
{( ) ( ) ( ) ( )}
′ x1 xr xr+1 x
C = ℝ+ ,..., , ,..., m .
1 1 0 0
100 3. POLYHEDRAL SETS, OPTIMIZATION AND SEMIGROUPS

By Corollary 3.1.17 the cone C ′ is a polyhedron. Thus there exists a matrix A and
a vector b such that C ′ can be is written as
{( ) }
′ x
C = x ∈ ℝ ; λ ∈ ℝ; Ax + λb ≤ 0 .
n
λ
Since x ∈ Q if and only if (x, 1) ∈ C ′ we conclude that Q = { x∣ Ax ≤ −b}, that is
Q is a polyhedron. This proof is due to A. Schrijver [150]. □
Thus by the finite basis theorem a polyhedron has two representations. We have
used the computer program Polyhedron Representation Transformation Algorithms
[36] (PORTA for short) to switch between these two representations. This program
is available for Unix and Windows systems. The “state of the art” tool for the algo-
rithmic treatment of convex polyhedra and finite simplicial complexes is polymake
[78], this program can be used on Unix systems only.
COROLLARY 3.1.21. A set Q ⊂ ℝn is a convex polytope if and only if Q is a
bounded polyhedral set.
PROOF. If Q = conv(α1 , . . . , αm ) is a polytope, then by the triangle inequality
for all x ∈ Q we have ∥ x∥ ≤ ∥α1 ∥ + ⋅ ⋅ ⋅ + ∥αm ∥. Thus Q is bounded.
Conversely if Q is a bounded polyhedron, then by Theorem 3.1.20 we can
decompose Q as Q = P + C, with P a polytope and C a finitely generated cone.
Notice that C = {0}, otherwise fixing p0 ∈ P and c0 ∈ C ∖ {0} we get p0 + λc0 ∈ Q
for all λ > 0, a contradiction because Q is bounded. Thus P = Q, as required. □
DEFINITION 3.1.22. Let Q be a polyhedral set and x0 ∈ Q. The point x0 is
called a vertex or an extreme point of Q if { x0 } is a proper face of Q.
PROPOSITION 3.1.23. [18, Theorem 7.2] If P = conv(α1 , . . . , αr ) ⊂ ℝn and V is
the set of vertices of P, then V ⊂ {α1 , . . . , αr } and P = conv(V).
LEMMA 3.1.24. [18, Theorem 5.6] If Q is a polyhedral set in ℝn and v ∈ Q is
not a vertex of Q, then there is a face F of Q such that v ∈ ri(F).
LEMMA 3.1.25. Let Q be a convex polyhedron in ℝn and x0 ∈ Q. If x0 is a
vertex of Q, then Q ∖ { x0 } is a convex set.
PROOF. Let H(a, c) be a proper supporting hyperplane of Q such that { x0 } =
Q ∩ H(a, c) and Q ⊂ H + (a, c). Take x, y ∈ Q ∖ { x0 } and consider z = ty + (1 − t)x
with 0 < t < 1. Note that x, y are not in H(a, c), thus
⟨z, a⟩ = t⟨y, a⟩ + (1 − t)⟨ x, a⟩ > tc + (1 − t)c = c.
Hence ⟨z, a⟩ > c, that is, z ∕= x0 and z ∈ Q, as required. □
PROPOSITION 3.1.26. If Q = C + P ⊂ ℝn with C a polyhedral cone and P a
polytope, then every vertex of Q is a vertex of P.
 3. POLYHEDRAL SETS, OPTIMIZATION AND SEMIGROUPS 101

PROOF. Let x be a vertex of Q and write x = c + p for some c ∈ C and p ∈ P.

We claim that p is a vertex of P. If p is not a vertex of P, then by Lemma 3.1.24
there is a face F of P such that p ∈ ri(F). Pick p ∕= v ∈ F. There is y ∈ F such that
p = λy + (1 − λ)v (0 < λ < 1).
Note that x ∕= y + c ∈ Q, x ∕= v + c ∈ Q, and
x = λ(y + c) + (1 − λ)(v + c) (0 < λ < 1),
thus Q ∖ { x} is not a convex set, a contradiction to Lemma 3.1.25. This proves that
p is a vertex of P. Assume c ∕= 0, then
p + λc ∕= x = p + c ∕= p + 2c (0 < λ < 1).
Consider the convex combination
x = p + c = λ0 (p + λc) + (1 − λ0 )(p + 2c),
where 0 < λ0 = 1/(2 − λ) < 1. Since p + λc and p + 2c are in Q ∖ { x}, this shows
that Q ∖ { x} is not a convex set, a contradiction. Thus c = 0 and x is a vertex of P,
as required. □
THEOREM 3.1.27. [150, Theorem 8.4] Let Q ⊂ ℝn be a polyhedral set. Then Q
has at least one vertex if and only if Q does not contains any lines in ℝn .
PROPOSITION 3.1.28. Let Q be a polyhedron containing no lines and let f (x) =
⟨ x, a⟩. If max{ f (x)∣ x ∈ Q} < ∞, then the maximum is attained at a vertex of Q.
PROOF. Let x0 ∈ Q be an optimal solution and let λ = f (x0 ) be the optimal
value. Note that
F = Q ∩ H(a, λ)
is a face of Q because Q ⊂ H − (a, λ) and x0 ∈ F. Since F contains no lines, by
Theorem 3.1.27 the face F contains at least one vertex z0 , which is also a vertex of
Q by transitivity. Thus the optimal value λ is attained at the vertex z0 , as required.
□
THEOREM 3.1.29. If Q ⊂ ℝn is a polyhedral set which contains no lines, then
Q can be written as:
Q = conv(α1 , . . . , αr ) + (ℝ+ β1 + ⋅ ⋅ ⋅ + ℝ+ β s ),
where α1 , . . . , αr are precisely the vertices of Q and β1 , . . . , β s ∈ ℝn .
PROOF. It follows from the proof of [194, Theorem 4.1.3]. □
DEFINITION 3.1.30. If a polyhedron Q in ℝn is represented as
Q = aff(Q) ∩ H + (a1 , c1 ) ∩ ⋅ ⋅ ⋅ ∩ H + (ar , cr ) ( ∗)
102 3. POLYHEDRAL SETS, OPTIMIZATION AND SEMIGROUPS

with ai ∈ ℝn ∖{0}, ci ∈ ℝ for all i, and none of the closed halfspaces H + (a1 , c1 ), . . . ,
H + (ar , cr ) can be omitted from the intersection, we say that (∗) is an irreducible
representation of Q.
THEOREM 3.1.31. [194, Theorem 3.2.1] Let Q be a polyhedral set in ℝn which
is not an affine space and let
Q = aff(Q) ∩ H + (a1 , c1 ) ∩ ⋅ ⋅ ⋅ ∩ H + (ar , cr )
be an irreducible representation of Q as an intersection of closed halfspaces. For
each i = 1, . . . , r, let Fi = Q ∩ H(ai , ci ). Then
(a) ri(Q) = { x ∈ Q∣ ⟨ x, a1 ⟩ > c1 , . . . , ⟨ x, ar ⟩ > cr };
(b) rb(Q) = F1 ∪ ⋅ ⋅ ⋅ ∪ Fr ;
(c) the facets of Q are precisely the sets F1 , . . . , Fr ;
(d) each proper face F of Q is the intersection of those facets Fi of Q such
that F ⊂ Fi .
COROLLARY 3.1.32. Let A = (ai j ) be an r × n real matrix and let c = (ci ) be a
column vector. If Q = { x ∈ ℝn ∣ Ax ≥ c} and x0 ∈ Q, then x0 is a vertex of Q if and
only if there is I ⊂ {1, . . . , r} with ∣I ∣ = n such that
{ x0 } = {(xi ) ∈ ℝn ∣ ai1 x1 + ⋅ ⋅ ⋅ + ain xn = ci for all i ∈ I }.
PROOF. Assume x0 is a vertex of Q. Hence by Theorem 3.1.31(d), x0 is the
unique solution of a system of linear equations:
ai1 x1 + ⋅ ⋅ ⋅ + ain xn = ci (i ∈ J)
for some J ⊂ {1, . . . , r}. Let [A′ ∣c′ ] be the augmented matrix of this system, by
Gaussian elimination one obtains that this matrix reduces to
In ∣ c′′
[ ]
.
0 ∣ 0
Hence the rank of A′ is equal to n. Thus there are n linearly independent rows of A′
and x0 is the unique solution of the system
ai1 x1 + ⋅ ⋅ ⋅ + ain xn = ci (i ∈ I ⊂ J)
for some I ⊂ J with ∣I ∣ = n. The converse is clear. □
REMARK 3.1.33. The result above yields a method to find the vertices of a
polyhedron. This method is by no means the best in practice. See [4, 5] for a
thorough discussion on how to find all the vertices of a polyhedron.
COROLLARY 3.1.34. If C ∕= ℝn is a polyhedral cone of dimension n, then there
is a unique irreducible representation
(3.10) C = Ha+1 ∩ ⋅ ⋅ ⋅ ∩ Ha+r , where ai ∈ ℝn ∖ {0}.
 3. POLYHEDRAL SETS, OPTIMIZATION AND SEMIGROUPS 103

PROOF. Let F be a facet of C. By Proposition 3.1.11, there is a supporting

hyperplane Ha of C such that F = C ∩ Ha . Since aff(F) = Ha , the uniqueness
follows from part (c) of Theorem 3.1.31. □
REMARK 3.1.35. If C ⊂ ℝn is an n-dimensional rational cone, then there exists
unique (up to sign) vectors a1 , . . . , ar in ℤn with relatively prime entries satisfying
Eq. (3.10).

Notation If a = (ai ) ∈ ℝn , we set ∣a∣ = ∣a1 ∣ + ⋅ ⋅ ⋅ + ∣an ∣.

THEOREM 3.1.36. Let 𝒜 be a finite set of points in ℕn ∖ {0} and let 𝒜′ be the
set of α ∈ 𝒜 such that ℝ+ α is a face of ℝ+ 𝒜 of dimension 1. Then
ℝ + 𝒜 = ℝ + 𝒜′ .
PROOF. Set 𝒜 = {α1 , . . . , αq } and S = conv(e1 , . . . , en ). Consider the convex
polytope P = S ∩ ℝ+ 𝒜. Note that P = conv(β1 , . . . , βq ), where βi = αi /∣αi ∣. There-
fore by Proposition 3.1.23 the set of vertices V(P) of P is contained in {β1 , . . . , βq }
and P is the convex hull of V(P). Since ℝ+ 𝒜 = ℝ+ V(P) it suffices to show that ℝ+ β
is a face of the cone spanned by 𝒜 for every β in V(P).
Let β ∈ V(P). There is a hyperplane H = { x ∈ ℝn ∣ ⟨ x, a⟩ = c} such that
P ∩ H = {β} and P ⊂ H − . Set a = (a1 , . . . , an ). To complete the proof we now
show that
ℝ+ 𝒜 ⊂ Hb− and ℝ+ 𝒜 ∩ Hb = ℝ+ β,
where b = (a1 − c, . . . , an − c). Since αi /∣αi ∣ ∈ P ⊂ H − for all i one has
⟨αi /∣αi ∣, a⟩ ≤ c =⇒ ⟨αi , a⟩ ≤ c∣αi ∣ = ⟨αi , (c, . . . , c)⟩ ,
thus ⟨αi , b⟩ ≤ 0 for all i. This proves ℝ+ 𝒜 ⊂ Hb− . There is i such that β = αi /∣αi ∣.
Next we show ℝ+ β ⊂ ℝ+ 𝒜 ∩ Hb . It suffices to prove β ∈ Hb . Since β ∈ H we
obtain
⟨αi /∣αi ∣, a⟩ = c = ⟨αi /∣αi ∣, (c, . . . , c)⟩ .
Hence β ∈ Hb . To prove the reverse inclusion take 0 ∕= x ∈ ℝ+ 𝒜 ∩ Hb . Then
⟨ x, a⟩ = ⟨ x, (c, . . . , c)⟩ = c∣ x∣ =⇒ ⟨ x/∣ x∣, a⟩ = c.
Thus x/∣ x∣ ∈ H. Observe that x/∣ x∣ ∈ S and x/∣ x∣ ∈ ℝ+ 𝒜, hence
x/∣ x∣ ∈ S ∩ ℝ+ 𝒜 = P.
Altogether x/∣ x∣ ∈ P ∩ H = {β}, this proves x ∈ ℝ+ β. □
THEOREM 3.1.37 (Carathéodory’s theorem). Let v1 , . . . , vq be a set of vectors in
ℝn not all of them zero. If x ∈ ℝ+ v1 + ⋅ ⋅ ⋅ + ℝ+ vq , then there is a linearly independent
set 𝒜 ⊂ {v1 , . . . , vq } such that x ∈ ℝ+ 𝒜.
104 3. POLYHEDRAL SETS, OPTIMIZATION AND SEMIGROUPS

PROOF. By induction on q. The case q = 1 is clear. If q ≥ 2, we can write

x = a1 v1 + ⋅ ⋅ ⋅ + aq vq , ai ≥ 0 for i = 1, . . . , q.
One may assume v1 , . . . , vq linearly dependent and ai > 0 for all i, otherwise the
result follows by induction. There are real numbers b1 , . . . , bq such that at least one
bi is positive and satisfying the equality
b1 v1 + ⋅ ⋅ ⋅ + bq vq = 0.
Set 0 < c = min{ai /bi ∣ bi > 0} = a j /b j . Notice the equality
x = x − c(b1 v1 + ⋅ ⋅ ⋅ + bq vq ) = (a1 − cb1 )v1 + ⋅ ⋅ ⋅ + (aq − cbq )vq ,
where ai − cbi ≥ 0 for all i and a j − cb j = 0. Therefore using induction the result
follows. □
DEFINITION 3.1.38. A set of vectors α1 , . . . , αq ∈ ℝn is affinely independent if
for every sequence λ1 , . . . , λq of real numbers satisfying
λ1 α1 + ⋅ ⋅ ⋅ + λq αq = 0 and λ1 + ⋅ ⋅ ⋅ + λq = 0,
one has λi = 0 for all i.
COROLLARY 3.1.39. Let v1 , . . . , vq be a set of vectors in ℝn and let x be in its
convex hull. Then there exists an affinely independent set 𝒜 ⊂ {v1 , . . . , vq } such
that x ∈ conv(𝒜).
PROOF. Consider 𝒜′ = {(v1 , 1), . . . , (vq , 1)}. Since (x, 1) belongs to ℝ+ 𝒜′ , the
result follows from Carathéodory’s theorem and Exercise 3.1.47. □
Linear programming. Let A be an n × q real matrix and let b, c be two real
vectors of sizes n and q respectively. Consider the primal problem
Maximize f (x) = qi=1 ci xi
∑
(∗)
Subject to Ax ≤ b and x ≥ 0.
Its dual problem is defined as
Minimize g(y) = ni=1 bi yi
∑
(∗∗)
Subject to yA ≥ c and y ≥ 0.
A solution x0 ∈ ℝq of the primal (resp. dual) problem that maximizes the objective
function f is called and optimal solution, the corresponding value f (x0 ) of the
objective function is called an optimal value.
THEOREM 3.1.40. (Duality theorem, [150, Corollary 7.1.g]) If the primal prob-
lem (∗) has an optimal solution (x1 , . . . , xq ), then the dual problem (∗∗) has an
optimal solution (y1 , . . . , yn ) such that
c1 x1 + ⋅ ⋅ ⋅ + cq xq = b1 y1 + ⋅ ⋅ ⋅ + bn yn
 3. POLYHEDRAL SETS, OPTIMIZATION AND SEMIGROUPS 105

DEFINITION 3.1.41. A rational polyhedron Q ⊂ ℝn is called integral if

Q = conv(ℤn ∩ Q).
THEOREM 3.1.42. [120, Theorem 5.12] Let Q be a rational polyhedron. Then
Q is integral if and only if any of the following statements hold:
(a) max{⟨c, x⟩∣ x ∈ Q} is attained by an integral vector for each c for which
the maximum is finite.
(b) max{⟨c, x⟩∣ x ∈ Q} is an integer for each integral vector for each c for
which the maximum is finite.
COROLLARY 3.1.43. If Q is a pointed polyhedron, that is, Q contains no lines,
then Q is integral if and only if all vertices of Q are integral.
PROOF. Let x0 be a vertex of Q. There is a supporting hyperplane H = H(a, c)
such that { x0 } = H ∩ Q and Q ⊂ H + . Since x0 is a convex combination of points
in Q ∩ ℤn , it is seen that x0 is integral. The converse follows from Theorem 3.1.42
and Proposition 3.1.28. □
PROPOSITION 3.1.44. Let Q ⊂ ℝn be a rational polyhedron and let QI =
conv(Q ∩ ℤn ) be its integral hull. Then
(a) QI is a polyhedron.
(b) If Q is a pointed, then QI is integral.
PROOF. We may assume that QI is non-empty. By the finite basis theorem we
can write Q = P + C, where C is a cone generated by integral vectors α1 , . . . , α s and
P is a polytope. Consider the linear map T : ℝ s → ℝn induced by T (ei ) = αi . Notice
that B := T ([0, 1] s ) is a polytope whose elements have the form λ1 α1 + ⋅ ⋅ ⋅ + λ s α s ,
0 ≤ λi ≤ 1 for all i. It is not hard to see that QI = conv((P+ B) ∩ℤn )+C. Since P+ B
is bounded, we get that QI is a polyhedron by the finite basis theorem. This proves
(a). To prove part (b) observe that the vertices of QI are contained in (P + B) ∩ ℤn ,
this follows from Proposition 3.1.26. Hence by Corollary 3.1.43, QI is integral.
This proof was adapted from [150]. □

Exercises
3.1.45. Prove that the affine maps f : ℝq → ℝn are precisely those maps which
can be expressed in the form f (x) = Ax + b, for some n × q matrix A and some
vector b.
3.1.46. Let Q be a polyhedral set in ℝn . If F1 , F2 are faces of Q, prove that
their intersection F = F1 ∩ F2 is a face of Q.
3.1.47. Let 𝒜 = {α0 , . . . , αr } be a sequence of distinct vectors in ℝn . Prove:
106 3. POLYHEDRAL SETS, OPTIMIZATION AND SEMIGROUPS

(a) 𝒜 is affinely independent ⇔ α1 − α0 , . . . , αr − α0 are linearly independent

⇔ (α0 , 1), . . . , (αr , 1) are linearly independent.
(b) If 𝒜 is contained in an affine hyperplane not containing the origin, then 𝒜
is affinely independent if and only if it is linearly independent.
3.1.48. Let A ⊂ ℝn . Prove that A is a linear variety in ℝn if and only if
λ1 A + λ2 A ⊂ A for all λ1 , λ2 in ℝ such that λ1 + λ2 = 1.
3.1.49. Let C ⊂ ℝn . Prove that C is a convex cone if and only if λx + µy ∈ C
for all x, y ∈ C and λ, µ ≥ 0.
3.1.50. Let π1 : ℝ2 → ℝ be the projection π1 (x1 , x2 ) = x1 . Prove that π1 is not
a closed linear map.
3.1.51. Let P be a polyhedral set in ℝn and let f : ℝn → ℝ be a linear function.
Prove that f (P) is a polyhedral set.
3.1.52. Determine the facets of the convex polytope
P = conv(±e1 , . . . , ±en ) ⊂ ℝn .
3.1.53. If C = ℝn , then C ∗ = (0).
3.1.54. Let C = He+2 ⊂ ℝ2 . Write C = ℝ+ e1 + ℝ+ (−e1 ) + ℝ+ e2 and use the
duality theorem to show that C ∗ = ℝ+ e2 .
3.1.55. Let 𝒜 = {v1 , . . . , vk }, ℬ = {u1 , . . . , um } be two sets of points in ℝn and
let 𝒞 be the finite set of all those points of the form vi + u j , where i = 1, . . . , k and
j = 1, . . . , m. Prove that
conv(𝒜) + conv(ℬ ) = conv(𝒞 ).
3.1.56. If C is a rational cone, then C I := conv(C ∩ ℤn ) = C.
3.1.57. Let Q, P, C, B be as in the proof of Proposition 3.1.44. Prove the equality
QI = conv((P + B) ∩ ℤn ) + C.
3.1.58. Let Q = { x∣ Ax ≤ b} =∕ ∅ be a polyhedron. If Q = P + C, where P is a
polytope and C is a polyhedral cone, prove that
C = {y∣Ay ≤ 0} = {y∣ x + y ∈ Q, ∀ x ∈ Q}.
The cone C is called the characteristic cone of Q.
3.1.59. (Complementary slackness) Consider the LP primal-dual pair
min{⟨ x, c⟩∣ x ≥ 0; xA ≥ b} = max{⟨y, b⟩∣ y ≥ 0; Ay ≤ c}
Let x and y be feasible solutions, i.e., xA ≥ b, Ay ≤ c and x, y ≥ 0. Use
Theorem 3.1.40 to show that the following are equivalent
 3. POLYHEDRAL SETS, OPTIMIZATION AND SEMIGROUPS 107

(a) x and y are both optimum solutions.

(b) ⟨ x, c⟩ = ⟨y, b⟩.
(c) (b − xA)y = 0 and x(c − Ay) = 0.
3.1.60. (P. Gordan [88]) The system Ax < 0 is unsolvable if and only if the
system yA = 0, y ≥ 0, y ∕= 0 is solvable. The vector inequality (ai ) < (bi ) meaning,
as usual, ai < bi for all i.

3.2. Volumes and unimodular coverings

Let 𝒜 = {v1 , . . . , vq } be a set of distinct vectors in ℤn and let
P = conv(v1 , . . . , vq )
be the convex hull of 𝒜. We set d = dim(P). A point in ℤn is called a lattice point
and P is called a lattice polytope. Here a lattice point (resp. polytope) is used as a
synonym of integral point (resp. polytope).
In the sequel to simplify the exposition and the proofs we set m = q − 1 and
α0 = v1 , . . . , αm = vq .
If Γ = {0, α1 − α0 , . . . , αm − α0 } is the image of 𝒜 under the translation
f
f : ℝn −→ ℝn , x 7−→ x − α0 ,
then aff(𝒜) = α0 + ℝΓ, where ℝΓ is the linear subspace generated by Γ. In particular,
from this expression we get
d = dim(P) = dimℝ (ℝΓ).
Let ℤΓ be the subgroup of ℤn generated by Γ. From the inclusions
ℤΓ ⊂ ℤn ∩ ℝΓ ⊂ ℝΓ
and using that any subgroup of ℤn is free, it follows that ℤn ∩ ℝΓ is a free abelian
group of rank d. A constructive proof of this fact, which is useful for computing
relative volumes, is included below.
LEMMA 3.2.1. There are vectors γ1 , . . . , γd in ℤn such that
ℝΓ ∩ ℤn = ℤγ1 ⊕ ⋅ ⋅ ⋅ ⊕ ℤγd .
PROOF. Consider the matrix C of order n × m whose column vectors correspond
to the nonzero vectors in Γ. By [140, Theorem II.9, pp. 26-27] there are invertible
integral matrices U and Q, of orders n and m respectively, such that U −1CQ has the
“diagonal” Smith normal form
D = diag{ s1 , . . . , sd , 0, . . . , 0} (si ∕= 0; si ∣ si+1 ∀ i).
108 3. POLYHEDRAL SETS, OPTIMIZATION AND SEMIGROUPS

Let u1 , . . . , un be the columns of U. We claim that the leftmost d columns of U can

serve as γ1 , . . . , γd . For convenience we regard vectors as column vectors.
Take α ∈ ℝΓ ∩ ℤn , thus α = Cβ for some β ∈ ℝm . Set β′ = (β′i ) = Q−1 β. Notice
the equalities
α = U(Dβ′ ) and U −1 α = Dβ′ = (s1 β′1 , . . . , sd β′d , 0, . . . , 0)t .
Hence si β′i ∈ ℤ for all i, and consequently α is in ℤu1 + ⋅ ⋅ ⋅ + ℤud . For the reverse
containment it suffices to prove that ui belongs to ℝΓ for i = 1, . . . , d. Using the
equality (CQ)ei = (UD)ei (ei =ith unit vector) we derive
Cqi = U(si ei ) = si (Uei ) = si ui (1 ≤ i ≤ d),
where qi = Qei is the ith column of Q. Hence ui ∈ ℝΓ for i = 1, . . . , d. □
LEMMA 3.2.2. There is an affine isomorphism
φ : aff(𝒜) −→ ℝd
such that φ(aff(𝒜) ∩ ℤn ) = ℤd , where d = dim(P).
PROOF. By Lemma 3.2.1 there are vectors γ1 , . . . , γd in ℤn such that
ℝΓ ∩ ℤn = ℤγ1 ⊕ ⋅ ⋅ ⋅ ⊕ ℤγd .
Therefore there is a linear map
ψ
ψ : ℝΓ −→ ℝd , γi 7−→ ei .
Hence the map φ = ψ f : aff(𝒜) −→ ℝd satisfies the required condition. □
Observe that φ(P) is an integral polytope of dimension d with a positive
Lebesgue measure denoted by m(φ(P)).
DEFINITION 3.2.3. The relative volume of the integral polytope P is:
vol(P) := m(φ(P)).
If P has dimension n, then the relative volume of P agrees with its usual volume.
To see that the relative volume is independent of φ we need the following fact about
the standard volume.
THEOREM 3.2.4. [194, Theorem 6.2.14] Let T : ℝn → ℝn be an affine map
given by T (x) = Ax + x0 , where A is a real matrix of order n and x0 ∈ ℝn . If P is a
bounded convex set in ℝn , then T (P) has volume ∣det(A)∣vol(P).
LEMMA 3.2.5. Let T : ℤn → ℤn be the affine map given by T (x) = Ax + β, where
A is an integral matrix of order n and β ∈ ℤn . If T is bijective, then det(A) = ±1.
 3. POLYHEDRAL SETS, OPTIMIZATION AND SEMIGROUPS 109

PROOF. Note that the matrix A determines a bijective linear map, thus A has an
inverse with integral entries. Indeed, for each i there is a column vector βi in ℤn
such that Aβi = ei . Therefore
A (β1 ⋅ ⋅ ⋅ βn ) = I,
and consequently det(A) = ±1. □
PROPOSITION 3.2.6. The relative volume of P is independent of φ.
PROOF. If φ1 is another affine isomorphism
φ1 : aff(𝒜) −→ ℝd
such that φ1 (aff(𝒜) ∩ ℤn ) = ℤd , then the affine map φ1 φ−1 : ℤd → ℤd is bijective.
Using Theorem 3.2.4 and Lemma 3.2.5 one obtains
vol(φ(P)) = vol((φ1 φ−1 )(φ(P))) = vol(φ1 (P)). □
The programs Normaliz [25] and Polyprob [43] are specially useful to compute
volumes of lattice polytopes.
EXAMPLE 3.2.7. Here we give an illustration in ℝ3 of the procedure outlined
above to compute relative volumes. We set
𝒜 = {(1, 3, 1), (1, 5, 3), (4, 9, 4), (2, 8, 5)}.
Then Γ = {(0, 0, 0), (0, 2, 2), (3, 6, 3), (1, 5, 4)} and
f
f : ℝ3 −→ ℝ3 , x −→ x − α0
where α0 = (1, 3, 1). Consider the matrix
⎡ ⎤
0 3 1
A = ⎣ 2 6 5 ⎦.
2 3 4
Next if we use Maple to compute the Smith normal form D of A we obtain unimod-
ular (invertible) matrices U and Q such that UAQ = D, where
⎡ ⎤ ⎡ ⎤ ⎡ ⎤
1 0 0 −2 1 0 0 1 0
D = ⎣ 0 1 0 ⎦; U=⎣ 1 0 0 ⎦; U −1 = ⎣ 1 2 0 ⎦.
0 0 0 1 −1 1 1 1 1
Thus, we get

ℝΓ ∩ ℤ3 = ℤ(0, 1, 1) ⊕ ℤ(1, 2, 1).

Then we define the affine map ψ : ℝΓ −→ ℝ3 induced by ψ(0, 1, 1) = (1, 0) and
ψ(1, 2, 1) = (0, 1), and the map φ : aff(𝒜) −→ ℝ given by φ = ψ f . Then
φ(P) = conv{(0, 0), (2, 0), (0, 3), (3, 1)}
110 3. POLYHEDRAL SETS, OPTIMIZATION AND SEMIGROUPS

2.
11
and vol(P) = m(φ(P)) =
PROPOSITION 3.2.8. [165, Proposition 4.6.30] The relative volume of P is given
by:
∣ℤn ∩ iP∣
vol(P) = lim .
i→∞ id
DEFINITION 3.2.9. For an abelian group (M, +) its torsion subgroup T (M) is the
set of all x in M such that px = 0 for some 0 ∕= p ∈ ℕ. The group M is torsion-free
if T (M) = (0).
LEMMA 3.2.10 ([64]). If ℬ = {β1 , . . . , β s } ⊂ ℤn , then
(a) (ℝℬ ∩ ℤn )/ℤℬ = T (ℤn /ℤℬ ). (∗)
(b) In particular, ℝℬ ∩ ℤn = ℤℬ if and only if ℤn /ℤℬ is torsion-free.
PROOF. To prove (a) consider an arbitrary class z in the left hand side of Eq. (∗)
such that z ∈ ℝℬ ∩ ℤn . This amounts to saying that the system (β1 , . . . , β s )y = z has
a real solution y. Notice that y can be taken rational, which means
z = (⌊y1 ⌋β1 + ⋅ ⋅ ⋅ + ⌊y s ⌋β s ) + ({y1 }β1 + ⋅ ⋅ ⋅ + {y s }β s ),
where ⌊yi ⌋ is the integral part of yi and {yi } is rational with ∣{yi }∣ < 1. Now,
the second parenthesis multiplied by a common multiple of the denominators of
the rationals {yi } lies in ℤℬ ; this shows that z is in the right hand side of Eq. (∗).
Conversely if z ∈ T (ℤn /ℤℬ ) ⊂ ℤn /ℤℬ , then kz ∈ ℤℬ for some 0 ∕= k ∈ ℕ, so
z ∈ ℚℬ ⊂ ℝℬ . Since also z ∈ ℤn , z is in the left hand side of Eq. (∗). This
completes the proof of (a). Part (b) follows from (a). □
LEMMA 3.2.11. [167, pp. 32-33] If H ⊂ G are free abelian∑groups of the
same rank d with ℤ-bases δ1 , . . . , δd and γ1 , . . . , γd related by δi = j gi j γ j , where
gi j ∈ ℤ for all i, j, then ∣G/H ∣ = ∣ det(gi j )∣.
THEOREM 3.2.12 ([64]). If 𝒜 = {v1 , . . . , vq } ⊂ ℤn is a set of vectors lying on
an affine hyperplane not containing the origin and P = conv(𝒜) has dimension d,
then
∣ℤ𝒜 ∩ iP∣
vol(P) = ∣T (ℤn /(v2 − v1 , . . . , vq − v1 ))∣ lim .
i→∞ id
PROOF. As pointed out at the beginning of this section for convenience of
notation we set m = q − 1 and α0 = v1 , . . . , αm = vq . Let x0 ∈ ℚn be such that
⟨αi , x0 ⟩ = 1 for all i. As ⟨αi − α0 , x0 ⟩ = 0, there is a decomposition
ℤ𝒜 = ℤα0 ⊕ ℤΓ′
with Γ′ = {α1 − α0 , . . . , αm − α0 }. Pick δ1 , . . . , δd ∈ ℤn such that
(3.11) ℤΓ′ = ℤδ1 ⊕ ⋅ ⋅ ⋅ ⊕ ℤδd .
 3. POLYHEDRAL SETS, OPTIMIZATION AND SEMIGROUPS 111

Therefore one can write

(3.12) αi = α0 + fi1 δ1 + ⋅ ⋅ ⋅ + fid δd ( fi j ∈ ℤ).
Consider the lattice polytope
P1 = conv((1, 0, . . . , 0), (1, f11 , . . . , f1d ), . . . , (1, fm1 , . . . , fmd )) ⊂ ℝd+1 .
ϕ
Observe that there is a bijective map ℤ𝒜 ∩ iP −→ ℤd+1 ∩ iP1 given by
ϕ(η0 α0 + η1 δ1 + ⋅ ⋅ ⋅ + ηd δd ) = (η0 , . . . , ηd ), (ηi ∈ ℤ),
that is, ϕ is the restriction of the linear map from ℤ𝒜 to ℤd+1 that maps each vector
into its coordinate vector with respect to the basis {α0 , δ1 , . . . , δd }. Therefore
∣ℤ𝒜 ∩ iP∣ ∣ℤd+1 ∩ iP1 ∣
lim = lim = vol(P1 ).
i→∞ id i→∞ id
To estimate vol(P1 ) consider the lattice polytope
P2 = conv((0, . . . , 0), ( f11 , . . . , f1d ), . . . , ( fm1 , . . . , fmd )) ⊂ ℝd
and note that applying the translation α 7→ α − e1 to P1 gives:
vol(P1 ) = vol(conv((0, . . . , 0), (0, f11 , . . . , f1d ), . . . , (0, fm1 , . . . , fmd )))
= vol(P2 ).
The next step is to relate vol(P2 ) with vol(P) by obtaining a second expression
for vol(P2 ). According to Lemma 3.2.1, there are γ1 , . . . , γd ∈ ℤn such that
(3.13) ℝΓ′ ∩ ℤn = ℤγ1 ⊕ ⋅ ⋅ ⋅ ⊕ ℤγd .
Writing
(3.14) δi = gi1 γ1 + ⋅ ⋅ ⋅ + gid γd (gi j ∈ ℤ; i, j = 1, . . . , d),
from (3.12) and (3.14) one derives the matrix equality
(3.15) ( fi j )(gi j ) = (ci j ),
where αi − α0 = ci1 γ1 + ⋅ ⋅ ⋅ + cid γd , 1 ≤ i ≤ m. Recall that by definition
vol(P) = vol(conv((0, . . . , 0), (c11 , . . . , c1d ), . . . , (cm1 , . . . , cmd ))).
By (3.15), the linear transformation
σ
σ : ℤd −→ ℤd , x 7−→ x(gi j ),
satisfies σ( fi1 , . . . , fid ) = (ci1 , . . . , cid ). Hence
vol(P) = vol(σ(P2 )) = ∣ det(gi j )∣vol(P2 ).
To finish the proof it suffices to show that ∣ det(gi j )∣ is the order of the torsion
subgroup of ℤn /ℤΓ′ . Now we have
(a) (b)
T (ℤn /ℤΓ′ ) = (ℝΓ′ ∩ ℤn )/ℤΓ′ = (ℤγ1 ⊕ ⋅ ⋅ ⋅ ⊕ ℤγd )/(ℤδ1 ⊕ ⋅ ⋅ ⋅ ⊕ ℤδd ),
112 3. POLYHEDRAL SETS, OPTIMIZATION AND SEMIGROUPS

where in (a) we use Lemma 3.2.10, and in (b) the identities (3.11) and (3.13). Hence,
from the identity (3.14) and Lemma 3.2.11, we get
∣T (ℤn /ℤΓ′ )∣ = ∣ det(gi j )∣,
as claimed. □
PROPOSITION 3.2.13. Let α0 , . . . , αn be a set of affinely independent points in
ℝn and ∆ = conv(α0 , . . . , αn ). Then the volume of the simplex ∆ is
α0 1 α1 − α0
⎛ ⎞ ⎛ ⎞

det ⎝ ... ... ⎠ det ⎝ ..

⎜ ⎟ ⎜ ⎟
. ⎠
αn 1 αn − α0
vol(∆) = = .
n! n!
PROOF. The asserted formula follows rapidly from the change of variables
formula [124, Chapter XII]. □
DEFINITION 3.2.14. A set ∆ in ℝn is called a lattice d-simplex if ∆ is the convex
hull of a set of d + 1 affinely independent points in ℤn .
Let ∆ be a lattice d-simplex in ℝn . The normalized volume of ∆ is defined as
d!vol(∆). If d = n, then from Proposition 3.2.13 one has
1
vol(∆) ≥ .
n!
This suggest the following:
DEFINITION 3.2.15. A lattice d-simplex ∆ in ℝn is called unimodular if
d!vol(∆) = 1.
PROPOSITION 3.2.16. Let ∆ = conv(α0 , . . . , αd ) be a lattice d-simplex. If
ℝ(α1 − α0 , . . . , αd − α0 ) ∩ ℤn = ℤγ1 ⊕ ⋅ ⋅ ⋅ ⊕ ℤγd
for some γ1 , . . . , γd ∈ ℤn and αi − α0 = j ci j γ j , then
∑

∣ det(ci j )∣
vol(∆) = .
d!
PROOF. Notice the equality vol(∆) = vol(conv(0, c1 , . . . , cd )), where ci =
(ci1 , . . . , cin ). Hence the formula follows from Proposition 3.2.13. □
LEMMA 3.2.17. If β1 , . . . , βm ∈ ℤn , then
ℝ(β1 , . . . βm ) ∩ ℤn = ℤβ1 + ⋅ ⋅ ⋅ + ℤβm + ℤδ1 + ⋅ ⋅ ⋅ + ℤδ s ,
where δ1 , . . . , δ s generate the torsion subgroup of ℤn /(β1 , . . . , βm ).
PROOF. It is left as an exercise. □
 3. POLYHEDRAL SETS, OPTIMIZATION AND SEMIGROUPS 113

LEMMA 3.2.18 ([64]). If α0 , . . . , αm ∈ ℤn , then there is an isomorphism of

groups
ϕ : T ℤn /(α1 − α0 , . . . , αm − α0 ) −→ T ℤn+1 /((α0 , 1), . . . , (αm , 1)) ,
( ) ( )

given by ϕ(α) = (α, 0).

PROOF. The map ϕ is clearly well defined and injective. To prove that ϕ is onto
consider the equation
s(α, b) = λ0 (α0 , 1) + ⋅ ⋅ ⋅ + λm (αm , 1) (λi ∈ ℤ),
where 0 ∕= s ∈ ℕ, α ∈ ℤn and b ∈ ℤ. Then
sα = λ0 α0 + ⋅ ⋅ ⋅ + λm αm ,
sb = λ0 + ⋅ ⋅ ⋅ + λm .
Hence s(α − bα0 ) = λ1 (α1 − α0 ) + ⋅ ⋅ ⋅ + λm (αm − α0 ). It follows readily that
ϕ(α − bα0 ) = (α, b), as required. □
PROPOSITION 3.2.19. [64, Proposition 3.18] Let α = α0 , . . . , αd be a set of
affinely independent vectors in ℤn defining a simplex ∆. Then the following are
equivalent:
(a) d!vol(∆) = 1.
(b) ℝ(α1 − α0 , . . . , αd − α0 ) ∩ ℤn = ℤ(α1 − α0 ) + ⋅ ⋅ ⋅ + ℤ(αd − α0 ).
(c) ℤn /(α1 − α0 , . . . , αd − α0 ) is torsion-free.
(d) ℤn+1 /((α0 , 1), . . . , (αd , 1)) is torsion-free.
PROOF. (a) ⇔ (b) follows form Proposition 3.2.16 and (b) ⇔ (c) ⇔ (d) follows
from Lemmas 3.2.17 and 3.2.18. □
DEFINITION 3.2.20. A lattice polytope P = conv(𝒜) of dimension d is said to
have a unimodular covering with support in 𝒜 if there are simplices ∆1 , . . . , ∆m of
dimension d such that the following two conditions hold
(i) P = ∪m i=1 ∆i and the vertices of ∆i are contained in 𝒜 for all i.
(ii) d!vol(∆i ) = 1.
If condition (ii) is replaced by
(ii)’ [ℤ𝒜 : ℤ𝒜i ] = 1 for all i,
where 𝒜i denotes the vertex set of ∆i , then we say that ∆1 , . . . , ∆m is a weakly
unimodular covering of P with support in 𝒜.
For a discussion on the existence of unimodular covers of rational cones see
[19] and the references there.
114 3. POLYHEDRAL SETS, OPTIMIZATION AND SEMIGROUPS

PROPOSITION 3.2.21 ([64]). If 𝒜 = {v1 , . . . , vq } ⊂ ℤn and the vectors in 𝒜 lie

in a hyperplane of ℝn not containing the origin, then any unimodular covering of
P = conv(𝒜) with support in 𝒜 is weakly unimodular.
PROOF. Let ∆i = conv(𝒜i ) be any unimodular simplex of dimension d with
vertex set 𝒜i ⊂ 𝒜 and d = dim(P). For convenience assume
𝒜i = {v1 , . . . , vd+1 }.
To show the equality ℤ𝒜 = ℤ𝒜i it suffices to prove ℤ𝒜 ⊂ ℤ𝒜i . Take an arbitrary
vector α in ℤ𝒜 and write:
α = η1 v1 + ⋅ ⋅ ⋅ + ηq vq (ηi ∈ ℤ),
(α, η) = η1 (v1 , 1) + ⋅ ⋅ ⋅ + ηq (vq , 1) (η = η1 + ⋅ ⋅ ⋅ + ηq ).
Since 𝒜 lies in an affine hyperplane not containing the origin and 𝒜i is affinely
independent, the set 𝒜i is seen to be linearly independent. Therefore
( ) ( )
v1 ⋅ ⋅ ⋅ vd+1 v1 ⋅ ⋅ ⋅ vq
d + 1 = rank = rank .
1 ⋅⋅⋅ 1 1 ⋅⋅⋅ 1
Thus one can write
(α, η) = λ1 (v1 , 1) + ⋅ ⋅ ⋅ + λd+1 (vd+1 , 1) (λi ∈ ℚ).
By Proposition 3.2.19 the abelian group
ℤn+1 /((v1 , 1), . . . , (vd+1 , 1))
is torsion-free. Hence it follows form Lemma 3.2.10(b) that λi ∈ ℤ for all i, and
consequently α is in ℤ𝒜i . □
EXAMPLE 3.2.22. Let 𝒜 = {v1 , v2 , v3 , v4 , v5 } and let
v1 = (1, 0, 1, 0, 0, 1), v2 = (1, 0, 0, 1, 1, 0), v3 = (0, 1, 1, 0, 1, 0),
v4 = (0, 1, 0, 1, 0, 1), v5 = (1, 0, 0, 1, 0, 1).
Consider
𝒜1 = {v2 , v3 , v4 , v5 }, 𝒜2 = {v1 , v3 , v4 , v5 }, 𝒜3 = {v1 , v2 , v3 , v5 },
∆1 = conv(𝒜1 ), ∆2 = conv(𝒜2 ), ∆3 = conv(𝒜3 ).
Notice that ℤ6 /ℤ𝒜 is torsion-free and also ℤ6 /ℤ𝒜i is torsion-free for all i. This
readily implies that
ℤ𝒜 = ℤ𝒜i (i = 1, 2, 3).
Using the relation v1 + v2 + v4 = v3 + 2v5 it is not hard to see that
P = conv(𝒜) = ∆1 ∪ ∆2 ∪ ∆3 .
Thus ∆1 , ∆2 , ∆3 is a weakly unimodular covering. The relative volume of P can be
computed using the procedure described before. It is left to the reader to verify that
dim(P) = 3 and vol(P) = 1/2.
 3. POLYHEDRAL SETS, OPTIMIZATION AND SEMIGROUPS 115

Exercises
3.2.23. If P is a lattice polytope of dimension d, then the relative volume of P
is given by:
∣ℤn ∩ iP∣
vol(P) = lim .
i→∞ id
Prove this formula if n = dim(P).
3.2.24. If P is the integral polytope with vertices (0, 0) and (1, 3):
y
6
3 s


2 

1 


s -
0 1 2 x
Prove that the relative volume of P is equal to 1.
3.2.25. Let 𝒜 = {v1 , . . . , vq } be a set of vectors in ℤn . Prove that the rank of
ℝ𝒜 ∩ ℤn as a free abelian group is equal to the dimension of ℝ𝒜 as a real vector
space.
3.2.26. Consider the set 𝒜 = {e1 + e2 , e2 + e3 , e3 + e4 , e1 + e4 } and the polytope
P = conv(𝒜). Prove that dim(P) = 2 and vol(P) = 1.
3.2.27. Prove that the converse of Proposition 3.2.21 fails.
3.2.28. If ∆ = conv(α0 , α1 , . . . , αd ) is a unimodular lattice d-simplex in ℝn , then
∆ ∩ ℤn = {α0 , α1 , . . . , αd }.
3.2.29. Let ∆ ⊂ ℝn be an n-dimensional lattice simplex with vertices α0 , . . . , αn
in ℤn . Prove that ∆ is unimodular if and only if either of the following two equivalent
conditions hold
(a) ℤn+1 = ℤ(α0 , 1) + ⋅ ⋅ ⋅ + ℤ(αn , 1).
(b) ℤn = ℤ(α1 − α0 ) + ⋅ ⋅ ⋅ + ℤ(αn − α0 ).

3.3. Hilbert bases

DEFINITION 3.3.1. A semigroup (S , +, 0) of ℤn is said to be finitely generated if
there exists a finite set Γ = {γ1 , . . . , γr } ⊂ S such that:
S = ℕΓ := ℕγ1 + ⋅ ⋅ ⋅ + ℕγr .
116 3. POLYHEDRAL SETS, OPTIMIZATION AND SEMIGROUPS

A set of generators Γ of S is called minimal if γi ∕= 0 ∀i and none of its elements is

a linear combination with coefficients in ℕ of the others.
LEMMA 3.3.2 (Gordan). Let 𝒜 = {v1 , . . . , vq } ⊂ ℤn and let ℤ𝒜 be the subgroup
generated by 𝒜. Then the semigroups
ℤ𝒜 ∩ ℝ+ 𝒜 and ℤn ∩ ℝ+ 𝒜
are finitely generated
PROOF. Set M = q maxi ∣v+i ∣ and N = −q maxi ∣v− i ∣. By Corollary 3.1.16 we
have
C = ℤ𝒜 ∩ ℝ+ 𝒜 = ℤ𝒜 ∩ ℚ+ 𝒜.
Let β ∈ C. We can write β = qi=1 (xi /yi ) vi , where xi ∈ ℕ and 0 ∕= yi ∈ ℕ. By the
∑
division algorithm there are non negative integers ri , ni such that xi = ni yi + ri and
0 ≤ ri < yi . Therefore one can write
q
∑ q
∑
β= ni vi + ai vi ,
i=1 i=1
∑q
where ni ∈ ℕ and ai ∈ [0, 1] ∩ ℚ. Since in C ∩ [N, M]n , we rapidly
i=1 ai vi is
obtain that
Γ := 𝒜 ∪ (C ∩ [N, M]n )
is a generating set for C.
Consider the finite set of integral points:
{a1 v1 + ⋅ ⋅ ⋅ + aq vq ∣ 0 ≤ ai ≤ 1} ∩ ℤn = {γ1 , . . . , γr }.
It is left to the reader to infer that γ1 , . . . , γr is the required set of generators for
ℤn ∩ ℝ+ 𝒜. □
REMARK 3.3.3. There are examples of submonoids of ℕn , with n ≥ 2, which
are not finitely generated, see Exercise 3.3.21.
Let 𝒜 be a finite set in ℤn and let G = ℤn or G = ℤ𝒜. Then by Gordan’s Lemma
there exists γ1 , . . . , γr ∈ ℤn ∩ ℝ+ 𝒜 such that:

G ∩ ℝ + 𝒜 = ℕ γ1 + ⋅ ⋅ ⋅ + ℕ γr ,
the γi ’s can be computed using Normaliz, an algorithm of Winfried Bruns Bogdan
Ichim, and Robert Koch [24, 25, 26].
DEFINITION 3.3.4. Let C be a rational polyhedral cone, that is, C = ℝ+ 𝒜 for
some finite set 𝒜 of integral vectors in ℝn . A finite set ℋ is called an integral
Hilbert bases of C if C = ℝ+ ℋ and
ℤn ∩ ℝ+ ℋ = ℕℋ.
 3. POLYHEDRAL SETS, OPTIMIZATION AND SEMIGROUPS 117

A Hilbert basis of C is called minimal if it does not strictly contain any other Hilbert
basis of C.
PROPOSITION 3.3.5. Let 𝒜 be a finite set in ℤn and let C = ℝ+ 𝒜. If
ℝ+ 𝒜 ∩ ℤn = ℕγ1 + ⋅ ⋅ ⋅ + ℕγr ,
then γ1 , . . . , γr is an integral Hilbert basis of C.
PROOF. Since one has the equality ℝ+ 𝒜 = ℝ+ γ1 + ⋅ ⋅ ⋅ + ℝ+ γr , we get that
γ1 , . . . , γr is an integral Hilbert basis of C. The existence follows from Lemma 3.3.2.
□
EXAMPLE 3.3.6. Let 𝒜 = {(1, 1, 0), (0, 1, 1), (1, 0, 1)} and let C = ℝ+ 𝒜. Using
Normaliz [24, 25], with its option 0, we obtain that the minimal integral Hilbert
basis of C is ℋ = 𝒜 ∪ {(1, 1, 1)}.
In Corollary 3.1.43 we considered pointed polyhedra. For cones we have the
following equivalent definition.
DEFINITION 3.3.7. Let C = { x∣ Ax ≤ 0} be a polyhedral cone. The cone C is
called a pointed cone if { x∣ Ax = 0} = (0).
LEMMA 3.3.8. Let C = { x∣ Ax ≤ 0} be a rational polyhedral cone. If C is
pointed, then there exists an integral vector b such that ⟨b, x⟩ > 0 for all 0 ∕= x ∈ C.
PROOF. We may assume that the matrix A is integral. If u1 , . . . , ut are the rows
of the matrix A, it is easy to see that b = −u1 −⋅ ⋅ ⋅− ut satisfy the required condition.
□
THEOREM 3.3.9 ([149]). Let 𝒜 be a finite set in ℤn and let C = ℝ+ 𝒜. If C is a
pointed cone, then there exists a unique minimal integral Hilbert basis of the cone
C given by
/ ℕy1 + ℕy2 ; ∀y1 , y2 ∈ (C ∖ {0}) ∩ ℤn }.
ℋ = { x ∈ C ∩ ℤn ∣ 0 ∕= x ∈
PROOF. Let Γ = {γ1 , . . . , γr } be an arbitrary integral Hilbert basis of C. To
begin with, we claim that ℋ ⊂ Γ. Let x ∈ ℋ. Since ℤn ∩ C = ℕΓ we can write
x = a1 γ1 + ⋅ ⋅ ⋅ + ar γr (ai ∈ ℕ).
Thus by construction of ℋ we get x = γi for some i, which proves the claim. To
finish the proof it suffices to prove that ℕℋ = ℕΓ, because this equality implies
ℝ+ ℋ = ℝ+ Γ and consequently ℤn ∩ ℝ+ ℋ = ℕℋ, that is, ℋ is a Hilbert basis of C.
We will proceed by contradiction assuming that the set:
/ ℕℋ}
A = { x∣ x ∈ ℕΓ = C ∩ ℤn ; x ∈
is non-empty. By Lemma 3.3.8 there exists b ∈ ℤn such that ⟨b, x⟩ > 0 for all
0 ∕= x ∈ C. Let x0 ∈ A such that
⟨ x0 , b⟩ = min{⟨ x, b⟩∣ x ∈ A}.
118 3. POLYHEDRAL SETS, OPTIMIZATION AND SEMIGROUPS

Since x0 ∈ / ℋ we can write x0 = x1 + x2 with x1 , x2 in (C ∖ {0}) ∩ ℤn . Now,

since ⟨ xi , b⟩ < ⟨ x0 , b⟩ for i = 1, 2, we have that x1 , x2 ∈ ℕℋ and x1 + x2 ∈ ℕℋ, a
contradiction. □

Rees cones of clutter matrices and their Hilbert bases.

Let A be an n × q clutter matrix with column vectors v1 , . . . , vq , that is, A is a matrix
with entries in {0, 1} such that vi ≰ v j for i ∕= j. Notice that for any clutter matrix
A of order n × q there is a clutter 𝒞 with n vertices and q edges such that A is
the incidence matrix of 𝒞 . Thus it is convenient to think of a clutter matrix as a
numerical object associated to a clutter which is a combinatorial object.
The Rees cone of A is the rational polyhedral cone in ℝn+1 , denoted by ℝ+ 𝒜′ ,
consisting of the non-negative linear combinations of the set
𝒜′ := {e1 , . . . , en , (v1 , 1), . . . , (vq , 1)} ⊂ ℝn+1 ,
where ei is the ith unit vector. Notice that ℝ+ 𝒜′ has dimension n + 1. It is not hard
to see that the set
F = ℝ+ 𝒜′ ∩ Hei (1 ≤ i ≤ n + 1)
defines a facet of ℝ+ 𝒜′ if and only if either i = n + 1 or 1 ≤ i ≤ n and ⟨ei , v j ⟩ = 0
for some column v j of A. Consider the index set
𝒥 = {1 ≤ i ≤ n ∣ ⟨ei , v j ⟩ = 0 for some j} ∪ {n + 1},
then the Rees cone has the following unique irreducible representation
( ) ( r )
∩ ∩ ∩
′ + +
(3.16) ℝ+ 𝒜 = Hei Hai
i∈𝒥 i=1

such that 0 ∕= ai ∈ ℚn+1and ⟨ai , en+1 ⟩ = −1 for all i. The set covering polyhedron
(in Proposition 4.6.3 we clarify this terminology) is defined by:
Q(A) := { x ∈ ℝn ∣ x ≥ 0, xA ≥ 1},
where 0 and 1 are vectors whose entries are equal to 0 and 1 respectively. Often we
denote the vectors 0, 1 simply by 0, 1.
LEMMA 3.3.10. Let a = (ai1 , . . . , aiq ) be the ith row of the matrix A. Set
k = min{ai j ∣ 1 ≤ j ≤ q}. If ai j > 0 for all j, then ei /k is a vertex of Q(A).
PROOF. Set x0 = ei /k. Clearly x0 ∈ Q(A), ⟨ x0 , v j ⟩ = 1 for some j, and
⟨ x0 , e` ⟩ = 0 for ` ∕= i. It is seen that x0 is a vertex of Q(A). □
THEOREM 3.3.11 ([84]). Let V be the vertex set of Q(A). Then the irreducible
representation of the Rees cone ℝ+ 𝒜′ is given by
( ) ( )
∩ ∩ ∩
ℝ + 𝒜′ = He+i +
H(α,−1) .
i∈𝒥 α∈V
 3. POLYHEDRAL SETS, OPTIMIZATION AND SEMIGROUPS 119

PROOF. Let V = {α1 , . . . , α p } be the set of vertices of Q(A) and let

ℬ = {ei ∣ i ∈ 𝒥 } ∪ {(α, −1)∣ α ∈ V }.
First we dualize Eq. (3.16) and use the duality theorem for cones to obtain
(ℝ+ 𝒜′ )∗ = {y ∈ ℝn+1 ∣ ⟨y, x⟩ ≥ 0, ∀ x ∈ ℝ+ 𝒜′ }
= He+1 ∩ ⋅ ⋅ ⋅ ∩ He+n ∩ H(v +
1 ,1)
+
∩ ⋅ ⋅ ⋅ ∩ H(v q ,1)
∑
(3.17) = ℝ + ei + ℝ + a 1 + ⋅ ⋅ ⋅ + ℝ + a r .
i∈𝒥
Next we show the equality
(3.18) (ℝ+ 𝒜′ )∗ = ℝ+ ℬ .
The right hand side is clearly contained in the left hand side because a vector
α belongs to Q(A) if and only if (α, −1) is in (ℝ+ 𝒜′ )∗ . To prove the reverse
inclusion observe that by Eq. (3.17) it suffices to show that ak ∈ ℝ+ ℬ for all k.
Writing ak = (ck , −1) and using ak ∈ (ℝ+ 𝒜′ )∗ gives ck ∈ Q(A). The set covering
polyhedron can be written as
Q(A) = ℝ+ e1 + ⋅ ⋅ ⋅ + ℝ+ en + conv(V),
where conv(V) denotes the convex hull of V, this follows from the structure of
polyhedra (see Theorem 3.1.20) by noticing that the characteristic cone of Q(A) is
precisely ℝn+ . Thus we can write
ck = λ1 e1 + ⋅ ⋅ ⋅ + λn en + µ1 α1 + ⋅ ⋅ ⋅ + µ p α p ,
where λi ≥ 0, µ j ≥ 0 for all i, j and µ1 + ⋅ ⋅ ⋅ + µ p = 1. If 1 ≤ i ≤ n and i ∈/ 𝒥 , then
the ith row of A has all its entries positive. Thus by Lemma 3.3.10 we get that ei /ki
is a vertex of Q(A) for some ki > 0. To avoid cumbersome notation we denote ei and
(ei , 0) simply by ei , from the context the meaning of ei should be clear. Therefore
from the equalities
( ) ∑ ( ) (∑ )
∑ ∑ ei ei
λi ei = λ i ki = λ i ki , −1 + λi ki en+1
ki ki
/
i∈𝒥 /
i∈𝒥 /
i∈𝒥 /
i∈𝒥

/ λi ei is in ℝ+ ℬ . From the identities

∑
we conclude that i∈𝒥
ak = (ck , −1) = λ1 e1 + ⋅ ⋅ ⋅ + λn en + µ1 (α1 , −1) + ⋅ ⋅ ⋅ + µ p (α p , −1)
∑ ∑ ∑p
= λ i ei + λi ei + µi (αi , −1)
/
i∈𝒥 i∈𝒥 ∖{n+1} i=1

we obtain ak ∈ ℝ+ ℬ , as required. Taking duals in Eq. (3.18) yields

∩
(3.19) ℝ + 𝒜′ = Ha+ .
a∈ℬ
120 3. POLYHEDRAL SETS, OPTIMIZATION AND SEMIGROUPS

Thus, by Remark 3.1.18, the proof reduces to showing that β ∈ / ℝ+ (ℬ ∖ {β}) for all
β ∈ ℬ . For this we will assume that β ∈ ℝ+ (ℬ ∖ {β}) for some β ∈ ℬ and derive a
contradiction.
Case (I): β = (α j , −1). For simplicity assume β = (α p , −1). We can write
∑ p−1
∑
(α p , −1) = λi ei + µ j (α j , −1), (λi ≥ 0; µ j ≥ 0).
i∈𝒥 j=1

Consequently
∑ p−1
∑
(3.20) αp = λi ei + µ jα j
i∈𝒥 ∖{n+1} j=1
(3.21) −1 = λn+1 − (µ1 + ⋅ ⋅ ⋅ + µ p−1 ).
To derive a contradiction we claim that Q(A) = ℝn+ + conv(α1 , . . . , α p−1 ), which is
impossible because by Proposition 3.1.26 the vertices of Q(A) would be contained in
{α1 , . . . , α p−1 }. To prove the claim note that the right hand side is clearly contained
in the left hand side. For the other inclusion take γ ∈ Q(A) and write
n
∑ p
∑ p
∑
γ = bi ei + ci α i (bi , ci ≥ 0; ci = 1)
i=1 i=1 i=1
p−1
(3.20) ∑
= δ+ (ci + c p µi )αi (δ ∈ ℝn+ ).
i=1
Therefore using the inequality
p−1 p−1
( p−1 )
∑ ∑ ∑ (3.21)
(ci + c p µi ) = ci + c p µi = (1 − c p ) + c p (1 + λn+1 ) ≥ 1
i=1 i=1 i=1
we get γ ∈ + conv(α1 , . . . , α p−1 ). This proves the claim.
ℝn+
Case (II): β = ek for some k ∈ 𝒥 . We only consider the subcase k ≤ n. The
subcase k = n + 1 can be treated similarly. We can write
∑ p
∑
ek = λi ei + µi (αi , −1), (λi ≥ 0; µi ≥ 0).
i∈𝒥 ∖{k} i=1
∑p ∑p
From this equality we get ek = i=1 µi αi . Hence ek A ≥ ( i=1 µi )1 > 0, a contra-
diction because k ∈ 𝒥 and ⟨ek , v j ⟩ = 0 for some j. □
DEFINITION 3.3.12. The system x ≥ 0; xA ≥ 1 is called totally dual integral
(TDI) if the maximum in the LP-duality equation
min{⟨α, x⟩∣ x ≥ 0; xA ≥ 1} = max{⟨y, 1⟩∣ y ≥ 0; Ay ≤ α}
 3. POLYHEDRAL SETS, OPTIMIZATION AND SEMIGROUPS 121

has an integral optimum solution y for each integral vector α with finite maximum.
PROPOSITION 3.3.13. If the system x ≥ 0; xA ≥ 1 is totally dual integral, then
Q(A) has only integral vertices.
PROOF. It follows from Theorem 3.1.42. □
DEFINITION 3.3.14. The matrix A has the max-flow min-cut (MFMC) property
if both sides of the LP-duality equation
(3.22) min{⟨α, x⟩∣ x ≥ 0; xA ≥ 1} = max{⟨y, 1⟩∣ y ≥ 0; Ay ≤ α}
have integral optimum solutions x and y for each non-negative integral vector α.
It turns out that the max-flow min-cut property is equivalent to require that a
certain blowup ring has no nilpotent elements, see Theorem 4.6.25.
PROPOSITION 3.3.15. The system x ≥ 0; xA ≥ 1 is TDI if and only if A has the
max-flow min-cut property.
PROOF. If follows from Propositions 3.3.13 and 3.1.28. □
THEOREM 3.3.16 ([84]). If A is a clutter matrix, then A has the max-flow min-cut
property if and only if Q(A) is an integral polyhedron and 𝒜′ is an integral Hilbert
basis of ℝ+ 𝒜′ .
PROOF. ⇒) By Proposition 3.3.13 the polyhedron Q(A) is integral. Next we
show that 𝒜′ is an integral Hilbert basis. Take (α, αn+1 ) ∈ ℤn+1 ∩ ℝ+ 𝒜′ . Then
Ay ≤ α and ⟨y, 1⟩ = αn+1 for some vector y ≥ 0. Therefore one concludes that the
optimal value of the linear program
max{⟨y, 1⟩∣ y ≥ 0; Ay ≤ α}
is greater than or equal to αn+1 . Since A has the MFMC property, this linear program
has an optimal integral solution y0 . By Exercise 3.3.22 there exists an integral vector
y′0 such that
0 ≤ y′0 ≤ y0 and ∣y′0 ∣ = αn+1 .
Therefore
α α
( ) ( ) ( ) ( ) ( )
A ′ A ′ A
= y0 + (y0 − y0 ) + − y0
αn+1 1 0 0 0
and (α, αn+1 ) ∈ ℕ𝒜′ , as required.
⇐) Assume that A does not satisfy the MFMC property. Then there is an
α0 ∈ ℕn such that if y0 is an optimal solution of the linear program:
max{⟨y, 1⟩∣ y ≥ 0; Ay ≤ α0 }, (∗)
then y0 is not integral. We claim that also the optimal value ∣y0 ∣ = ⟨y0 , 1⟩ of this
linear program is not integral. If ∣y0 ∣ is integral, then (α0 , ∣y0 ∣) is in ℤn+1 ∩ ℝ+ 𝒜′ . As
122 3. POLYHEDRAL SETS, OPTIMIZATION AND SEMIGROUPS

𝒜′ is a Hilbert basis we get that (α0 , ∣y0 ∣) is in ℕ𝒜′ , but this readily yields that the
linear program (∗) has an integral optimal solution, a contradiction. This completes
the proof of the claim.
Now, consider the dual linear program:
min{⟨ x, α0 ⟩∣ x ≥ 0, xA ≥ 1}.
By Proposition 3.1.28 the optimal value of this linear program is attained at a vertex
x0 of Q(A). Then by the duality theorem (Theorem 3.1.40) we get ⟨ x0 , α0 ⟩ = ∣y0 ∣ ∈/
ℤ. Hence x0 is not integral, a contradiction to the integrality of the set covering
polyhedron Q(A). □
REMARK 3.3.17. The program Normaliz [24, 25] computes the irreducible rep-
resentation and the minimal integral Hilbert basis of a Rees cone. Thus we can
effectively use Theorems 3.3.11 and 3.3.16 to determine whether a given {0, 1}-
matrix has the max-flow min-cut property, see example below.
EXAMPLE 3.3.18. Let A be the transpose of the matrix:
⎡ ⎤
1 0 0 0 1
⎢ 0 1 0 1 0 ⎥
⎣ 0 0 1 1 1 ⎦.
⎢ ⎥

1 1 1 0 0
Consider the following input file for Normaliz that we call “example.in”:
4
5
1 0 0 0 1
0 1 0 1 0
0 0 1 1 1
1 1 1 0 0
3
Applying Normaliz [24] to this file, i.e., typing “normaliz example” in the
directory where you have the executable file normaliz.exe we get the (output) file
“example.out”:
9 generators of integral closure of Rees algebra:
1 0 0 0 0 0
0 1 0 0 0 0
0 0 1 0 0 0
0 0 0 1 0 0
0 0 0 0 1 0
1 0 0 0 1 1
0 1 0 1 0 1
0 0 1 1 1 1
 3. POLYHEDRAL SETS, OPTIMIZATION AND SEMIGROUPS 123

1 1 1 0 0 1

10 support hyperplanes:
0 0 1 1 1 -1
1 0 0 0 0 0
0 1 0 0 0 0
0 0 0 0 0 1
0 0 1 0 0 0
1 0 0 1 0 -1
0 0 0 1 0 0
0 0 0 0 1 0
0 1 0 0 1 -1
1 1 1 0 0 -1

semigroup is not homogeneous

The first block shows that 𝒜′ is a Hilbert basis for the Rees cone. The second
block shows the irreducible representation of the Rees cone of A, thus using Theo-
rem 3.3.11 we obtain that Q(A) is integral. Altogether Theorem 3.3.16 proves that
the matrix A has the max-flow min-cut property.

Exercises
3.3.19. A semigroup (0) ∕= S of ℕn is called full if ℕn ∩ ℤS = S . If S ⊂ ℕn is a
full semigroup, prove that S is a finitely generated semigroup.

3.3.20. Prove that any subsemigroup of ℕ is finitely generated.

3.3.21. Let ℕ = {0, 1, 2, . . .} and ℕ+ = {1, 2, . . .}. Prove that the subsemigroup
of ℕ2 given by
S = (ℕ+ × ℕ) ∪ {(0, 0)}
is not finitely generated.

3.3.22. Let b, η1 , . . . , ηq be a sequence of non-negative integers. If

η1 + ⋅ ⋅ ⋅ + ηq ≥ b,

then there are integers 1 , . . . , q ∈ ℕ such that 0 ≤ i ≤ ηi for all i and 1 +⋅ ⋅ ⋅+q = b.

3.3.23. Prove that for any clutter matrix A of order n × q there is a clutter 𝒞 with
n vertices and q edges such that A is the incidence matrix of 𝒞 .
124 3. POLYHEDRAL SETS, OPTIMIZATION AND SEMIGROUPS

3.4. The integral closure of a semigroup

Let ℕ = {0, 1, 2, . . . } and let 𝒜 = {v1 , . . . , vq } be a finite set of vectors in ℕn ∖ {0}.
As before, ℝ+ 𝒜 denotes the rational cone generated by 𝒜:
ℝ+ 𝒜 := ℝ+ v1 + ⋅ ⋅ ⋅ + ℝ+ vq ⊂ ℝn .
DEFINITION 3.4.1. The integral closure or normalization of the affine semigroup
ℕ𝒜 := ℕv1 + ⋅ ⋅ ⋅ + ℕvq ⊂ ℕn ,
is defined as:
ℕ𝒜 = ℤ𝒜 ∩ ℝ+ 𝒜,
where ℤ𝒜 is the subgroup of ℤn generated by 𝒜.
DEFINITION 3.4.2. The semigroup ℕ𝒜 is normal or integrally closed if
ℕ𝒜 = ℕ𝒜.
We say that ℕ𝒜 is strongly normal if ℕ𝒜 = ℝ+ 𝒜 ∩ ℤn .
The algebraic invariants of semigroup rings of strongly normal semigroups are
studied in Chapters 4 and 7. These semigroups occur in the theory of Rees algebras
and monomial subrings associated to totally unimodular matrices.
PROPOSITION 3.4.3. If ℕ𝒜 is strongly normal, then ℕ𝒜 is normal.
PROOF. From the inclusions:
ℕ𝒜 ⊂ ℝ+ 𝒜 ∩ ℤ𝒜 ⊂ ℝ+ 𝒜 ∩ ℤn = ℕ𝒜
we obtain ℕ𝒜 = ℝ+ 𝒜 ∩ ℤ𝒜. □
PROPOSITION 3.4.4. ℕ𝒜 is strongly normal if and only if 𝒜 is a Hilbert basis of
ℝ + 𝒜.
PROOF. It follows rapidly from the definitions. □
Rees semigroups. The Rees semigroup of 𝒜 is by definition:
ℕ𝒜′ ⊂ ℕn+1 ,
where 𝒜′ = {(v1 , 1), . . . , (vq , 1), (e1 , 0), . . . , (en , 0)} and ei denotes the ith unit vector
in ℝn . The cone
ℝ+ 𝒜′ ⊂ ℝn+1 ,
is called the Rees cone of 𝒜.
REMARK 3.4.5. Notice that ℤ𝒜′ = ℤn+1 because we have the equality
en+1 = (v1 , 1) − ⟨(v1 , 1), (e1 , 0)⟩e1 − ⋅ ⋅ ⋅ − ⟨(v1 , 1), (en , 0)⟩en .
Hence ℕ𝒜′ = ℝ+ 𝒜′ ∩ ℤn+1 .
From this remark we obtain:
 3. POLYHEDRAL SETS, OPTIMIZATION AND SEMIGROUPS 125

PROPOSITION 3.4.6. The Rees semigroup ℕ𝒜′ is normal if and only if 𝒜′ is an

integral Hilbert basis of ℝ+ 𝒜′ .
DEFINITION 3.4.7. A matrix A is called totally unimodular if each i × i subde-
terminant of A is 0 or ±1 for all i ≥ 1.
Examples of totally unimodular matrices include incidence matrices of hy-
pergraphs without odd cycles [9] and network matrices or equivalently two-way
cut-incidence matrices of cross-free families, see [120]. A precise characterization
of totally unimodular matrices can be found in [153], which is one of the deepest
papers in the theory of matroids. See also [104] for a different perspective about
unimodular matrices.
A useful criterion for total unimodularity is the following:
THEOREM 3.4.8. (Ghouila-Houri; see [120, Theorem 5.23]) An integral matrix
A = (ai j ) of order n × q is totally unimodular if and only if for every S ⊂ {1, . . . , n}
there is a partition S = S 1 ∪ S 2 such that
∑ ∑
ai j − ai j ∈ {0, 1, −1} ∀ j = 1, . . . , q.
i∈S 1 i∈S 2

THEOREM 3.4.9 (Hoffman, Kruskal). If A is a totally unimodular matrix, then

the polyhedron Q = { x ∣ x ≥ 0; Ax ≤ b} is integral for each integral vector b.
PROOF. Assume that A has order n × q. Notice that Q contains no lines, thus
it suffices to show that Q has only integral vetices. Let x0 be a vertex of Q. By
Corollary 3.1.32 and its proof, x0 is the solution of A′ x0 = b′ for some subsystem
A′ x ≤ b′ of
( ) ( )
A b
≤
−Iq 0
such that A′ is a square matrix of order q with linearly independent rows. By
hypothesis det(A′ ) = ±1, hence x0 is integral because the inverse of A′ is an integral
matrix by Cramer’s rule. □
Next we will use linear programming techniques to study the normality of
certain Rees semigroups. The next result also follows from Theorems 3.3.16 and
3.4.9. However the method of proof of the next result is interesting because it
provides another good illustration of the use of linear optimization techniques to
study affine semigroups and blowup algebras. Blowup algebra will be introduced
in Chapter 4.
THEOREM 3.4.10 ([15]). If A = (ai j ) is a {0, 1} totally unimodular matrix of
order n × q with column vectors v1 , . . . , vq , then the Rees semigroup ℝ+ 𝒜′ is normal.
126 3. POLYHEDRAL SETS, OPTIMIZATION AND SEMIGROUPS

PROOF. Recall that 𝒜′ = {(v1 , 1), . . . , (vq , 1), (e1 , 0), . . . , (en , 0)}, where ei is the
ith unit vector in ℝn . Let (α, b) ∈ ℤ𝒜′ ∩ ℝ+ 𝒜′ , where α ∈ ℝn and b ∈ ℝ. Note
ℤ𝒜′ = ℤn+1 . One can write
(α, b) = λ1 (v1 , 1) + ⋅ ⋅ ⋅ + λq (vq , 1) + µ1 (e1 , 0) + ⋅ ⋅ ⋅ + µn (en , 0)
where λi and µ j are non negative rational numbers for all i, j. Hence one obtains
the equalities
α = λ1 v1 + ⋅ ⋅ ⋅ + λq vq + µ1 e1 + ⋅ ⋅ ⋅ + µn en ,
b = λ1 + ⋅ ⋅ ⋅ + λq .
Let A′ = (A e1 ⋅ ⋅ ⋅ en ) be the n × (q + n) matrix obtained from A by adjoining
the column vectors e1 , . . . , en . Since total unimodularity is preserved under the
operation of adding columns of unit vectors, the matrix A′ is totally unimodular.
Consider the linear program
Maximize x1 + ⋅ ⋅ ⋅ + xq (∗)
Subject to
A′ x = α and x ≥ 0,
where x is the column vector x = (x1 , . . . xq , xq+1 , . . . , xq+n ). Since the column vector
c = (c1 , . . . , cn+q ) = (λ1 , . . . , λq , µ1 , . . . , µn )
satisfies
A′ c = α, c ≥ 0 and c1 + ⋅ ⋅ ⋅ + cq = b,
one concludes that the linear program (∗) has an optimal value greater than or equal
to b, which is attained at a vertex x0 of the rational polytope
P = { x ∈ ℝn+q ∣ A′ x = α and x ≥ 0},
see Proposition 3.1.28. Observe that all the vertices of P have integral entries
because A′ is totally unimodular, see Theorem 3.4.9. Thus x0 is a vector with non
negative integral entries:
x0 = (η1 , . . . , ηq , δ1 , . . . , δn )
such that η1 + ⋅ ⋅ ⋅ + ηq ≥ b. By Exercise 3.3.22 there are integers 1 , . . . , q satisfying
0 ≤ i ≤ ηi ∀i and 1 + ⋅ ⋅ ⋅ + q = b.
Therefore
(α, b) = η1 (v1 , 0) + ⋅ ⋅ ⋅ + ηq (vq , 0) + δ1 (e1 , 0) + ⋅ ⋅ ⋅ + δn (en , 0) + (0, b)
= 1 (v1 , 1) + ⋅ ⋅ ⋅ + q (vq , 1) +
(η1 − 1 )(v1 , 0) + ⋅ ⋅ ⋅ + (ηq − q )(vq , 0) +
δ1 (e1 , 0) + ⋅ ⋅ ⋅ + δn (en , 0),
and (α, b) ∈ ℕ𝒜′ , as required. This proof was adapted from [15]. □
 3. POLYHEDRAL SETS, OPTIMIZATION AND SEMIGROUPS 127

Exercises
3.4.11. If S ⊂ ℕn is a full semigroup, prove that ℝ+ S ∩ ℤS = ℕS .
3.4.12. Let 𝒜 = {(1, 1, 0), (0, 1, 1), (1, 0, 1)}. Prove that the affine semigroup
ℕ𝒜 is normal but it is not strongly normal.

3.5. Unimodularity of matrices and normality

A minor of order r × r of a matrix A is defined as the determinant of a submatrix of
A of order r × r.
Notation For an integral matrix C ∕= (0), the greatest common divisor of all the
nonzero r × r minors of C will be denoted by ∆r (C).
LEMMA 3.5.1. If C is an integral matrix of order n × q and rank r, then
∆r (C) = ∣T (ℤn /(w1 , . . . , wq ))∣,
where wi is the ith column of C; in particular ∆r (C) = 1 if ℤn /(w1 , . . . , wq ) is
torsion-free.
PROOF. The invariant factors of ℤn /(w1 , . . . , wq ) are given by
d1 = ∆1 (C), di = ∆i (C)/∆i−1 (C), i = 2, . . . , r,
see [117, Theorem 3.9]. Using the fundamental theorem of finitely generated abelian
groups [117, pp. 187-188], we get that
T (ℤn /(w1 , . . . , wq )) ≃ ℤd1 ⊕ ⋅ ⋅ ⋅ ⊕ ℤdr .
Hence the required formula follows by noticing that d1 ⋅ ⋅ ⋅ dr = ∆r (C). □
Let A be an n × q integral matrix with column vectors v1 , . . . , vq and let b ∈ ℤn
be a column vector such that r = rank(A) = rank([A b]). For use below notice that
there is a positive integer k such that
kb = λ1 v1 + ⋅ ⋅ ⋅ + λq vq (λi ∈ ℤ ∀i). (∗)
Therefore there is a canonical epimorphism of finite groups
ϕ
ϕ : T (ℤn /ℤ𝒜) −→ T (ℤn /ℤℬ ) (α + ℤ𝒜 7−→ α + ℤℬ ),
where ℬ = 𝒜 ∪ {b} and 𝒜 = {v1 , . . . , vq }.
THEOREM 3.5.2 ([189]). The following conditions are equivalent
(i) ϕ is injective.
(ii) b ∈ ℤ𝒜.
(iii) The groups ℤn /ℤ𝒜 and ℤn /ℤℬ have the same invariant factors.
128 3. POLYHEDRAL SETS, OPTIMIZATION AND SEMIGROUPS

(iv) The matrices [A 0] and [A b] have the same Smith normal form.
(v) ∆r (A) = ∆r ([A b]), where r = rank (A).
PROOF. The implications (i) ⇒ (ii) ⇒ (iii) are straightforward. There are
invertible integral matrices Pi and Qi such that
D1 = Q1 [A 0]P1 = diag(d1 , . . . , dr , 0, . . . , 0),
D2 = Q2 [A b]P2 = diag(e1 , . . . , er , 0, . . . , 0),
are the Smith normal forms of [A 0] and [A b] respectively, that is, di , ei are positive
integers satisfying that di divides di+1 and ei divides ei+1 for all i. By the fundamental
structure theorem of finitely generated abelian groups there are isomorphisms:
T (ℤn /ℤ𝒜) ≃ (ℤ/d1 ℤ) × ⋅ ⋅ ⋅ × (ℤ/dr ℤ),
T (ℤn /ℤℬ ) ≃ (ℤ/e1 ℤ) × ⋅ ⋅ ⋅ × (ℤ/er ℤ).
Thus (iii) ⇔ (iv). Note ∆i (A) = d1 ⋅ ⋅ ⋅ di and ∆i ([A b]) = e1 ⋅ ⋅ ⋅ ei for all i. Hence (iv)
⇒ (v). To prove (v) ⇒ (i) observe ∣T (ℤn /ℤ𝒜)∣ = ∣T (ℤn /ℤℬ )∣ and consequently ϕ
must be injective, as required. □
As an immediate consequence of Theorem 3.5.2 we get:
COROLLARY 3.5.3. (I. Heger) Let A be an n × q integral matrix and let b ∈ ℤn
be a column vector. If rank(A) = rank([A b]), then the system
Ax = b
has an integral solution if and only if ∆r (A) = ∆r ([A b]).
REMARK 3.5.4. By the finite basis theorem ℝ+ 𝒜 can be expressed as a finite
intersection of closed halfspaces through the origin, that is, there is a rational matrix
C such that
ℝ+ 𝒜 = { x ∈ ℝn ∣ Cx ≤ 0}.
Thus this representation together with Theorem 3.5.2 yield a membership test to
decide when a given α in ℤn belongs to ℕ𝒜 = ℤ𝒜 ∩ ℝ+ 𝒜.
DEFINITION 3.5.5. An integral matrix A ∕= (0) is t-unimodular if all the non-zero
r × r minors of A have absolute value equal to t, where r ∕= 0 is the rank of A. If
t = 1, the matrix A is called a unimodular matrix.
PROPOSITION 3.5.6. If A is a t-unimodular matrix and vi1 , . . . , vir is a ℚ-basis
for the column space of A, then
ℤ𝒜 = ℤvi1 ⊕ ⋅ ⋅ ⋅ ⊕ ℤvir
PROOF. For each j one has ∆r (vi1 ⋅ ⋅ ⋅ vir ) = ∆r (vi1 ⋅ ⋅ ⋅ vir v j ) = 1. Then by
Theorem 3.5.2, we get v j ∈ ℤvi1 ⊕ ⋅ ⋅ ⋅ ⊕ ℤvir , as required. □
 3. POLYHEDRAL SETS, OPTIMIZATION AND SEMIGROUPS 129

THEOREM 3.5.7. [189, Theorem 2.15] If A is a t-unimodular matrix, then the

affine semigroup ℕ𝒜 is normal.
PROOF. Let b ∈ ℤ𝒜 ∩ ℝ+ 𝒜. By Theorem 3.1.37 there are vi1 , . . . , vir linearly
independent columns of A, where r is the rank of A, such that
b ∈ ℝ + vi1 + ⋅ ⋅ ⋅ + ℝ + vir . (1)
On the other hand by Proposition 3.5.6 one has
b ∈ ℤvi1 + ⋅ ⋅ ⋅ + ℤvir . (2)
Since vi1 , . . . , vir are linearly independent by comparing the coefficients of b with
respect to the two representations given by (1) and (2) one derives b ∈ ℕ𝒜. □
( )
COROLLARY 3.5.8. If the matrix A = v1 v2 ⋅ ⋅ ⋅ vq is totally unimodular, then
ℕ𝒜 is strongly normal.
PROOF. Adapting the proof of Proposition 3.5.7, it is seen that ℕ𝒜 = ℤn ∩ ℝ+ 𝒜.
We leave the details as an exercise. □
PROPOSITION 3.5.9. If 𝒜 = {v1 , . . . , vq } lie in a hyperplane not containing the
origin and the polytope P = conv(𝒜) has a unimodular covering with support in 𝒜,
then the semigroup ℕ(v1 , 1) + ⋅ ⋅ ⋅ + ℕ(vq , 1) ⊂ ℤn+1 is normal.
PROOF. Let 𝒜′ = {(v1 , 1), . . . , (vq , 1)} and let (β, b) ∈ ℝ+ 𝒜′ ∩ ℤ𝒜′ , where
β ∈ ℤn and b ∈ ℤ. There are λ1 , . . . λq ∈ ℚ+ such that
(3.23) (β, b) = λ1 (v1 , 1) + ⋅ ⋅ ⋅ + λq (vq , 1),
β = λ1 v1 + ⋅ ⋅ ⋅ + λq vq ,
b = λ1 + ⋅ ⋅ ⋅ + λq ,
and there is a unimodular d-simplex ∆i = conv(𝒜i ) with support in 𝒜 such that
β ∈ ℝ+ 𝒜i . For simplicity assume 𝒜i = {v1 , . . . , vd+1 } and set 𝒜′i = {(v1 , 1), . . . ,
(vd+1 , 1)}, where d = dim(P). Thus one can write
β = µ1 v1 + ⋅ ⋅ ⋅ + µd+1 vd+1 (µi ∈ ℚ+ ).
Since v1 , . . . , vq lie in a hyperplane not containing the origin one has
λ1 + ⋅ ⋅ ⋅ + λq = µ1 + ⋅ ⋅ ⋅ + µd+1 .
Therefore one may assume q = d + 1 in Eq. (3.23). There is a positive integer m such
that m(β, b) ∈ ℤ𝒜′i . Observe that ℤn+1 /ℤ𝒜′i is torsion free by Proposition 3.2.19.
Hence (β, b) ∈ ℤ𝒜′i . To finish the proof note that 𝒜i is a linearly independent set,
therefore using Eq. (3.23) one concludes λi ∈ ℕ for all i and (β, b) ∈ ℕ𝒜′i ⊂ ℕ𝒜′ ,
as required. □
130 3. POLYHEDRAL SETS, OPTIMIZATION AND SEMIGROUPS

PROPOSITION 3.5.10. If 𝒜 = {v1 , . . . , vq } lie in a hyperplane not containing the

origin and P = conv(𝒜) has a weakly unimodular covering with support in 𝒜, then
ℕ𝒜 is a normal semigroup.
PROOF. Let 0 ∕= β ∈ ℝ+ 𝒜 ∩ ℤ𝒜. There are λ1 . . . , λq ∈ ℚ+ such that
β/∣λ∣ = (λ1 /∣λ∣)v1 + ⋅ ⋅ ⋅ + (λq /∣λ∣)vq ,
where ∣λ∣ = λ1 + ⋅ ⋅ ⋅ + λq . Since β/∣λ∣ is in P and using that P has a weakly
unimodular covering, there is an affinely independent set 𝒜′ ⊂ 𝒜 defining a lattice
simplex ∆ = conv(𝒜′ ) such that β/∣λ∣ ∈ ∆ and ℤ𝒜′ = ℤ𝒜. Altogether we conclude
that β belongs to ℤ𝒜′ ∩ ℝ+ 𝒜′ . It is crucial to note that 𝒜′ is a linearly independent
set because the vectors in 𝒜′ lie in a hyperplane. Therefore it follows rapidly that
β ∈ ℕ𝒜′ , as required. □

Exercises
3.5.11. Let A ∕= (0) be a totally unimodular matrix of order n × q and let
𝒜 = {v1 , . . . , vq } be the set of column vectors of A. Prove the equality:
ℤn ∩ ℝ+ 𝒜 = ℕ𝒜.
cf. Corollary 3.5.8 and Lemma 6.2.10.

3.6. Cut-incidence matrices and integrality

In Section 3.3 we have only considered TDI systems of the form xA ≥ 1; x ≥ 0.
Next we introduce general TDI systems in a way which is convenient for our
purposes. See [120, Chapter 5] and [150, p. 311].
DEFINITION 3.6.1. Let B be an integral matrix. A system Bx ≥ b; x ≥ 0 is
totally dual integral (TDI) if the maximum in
min{⟨c, x⟩∣ x ≥ 0; Bx ≥ b} = max{⟨y, b⟩∣ y ≥ 0; yB ≤ c}
has an integral optimum solution y for each integral vector c with finite maximum.
PROPOSITION 3.6.2. If the matrix B is totally unimodular, then the system Bx ≥
b; x ≥ 0 is TDI.
PROOF. It follows from Theorem 3.4.9. □
PROPOSITION 3.6.3. Let B be an integral matrix and let b be an integral vector.
If the system Bx ≥ b; x ≥ 0 is TDI, then Q = { x∣ Bx ≥ b; x ≥ 0} has only integral
vertices.
 3. POLYHEDRAL SETS, OPTIMIZATION AND SEMIGROUPS 131

PROOF. It follows from Theorem 3.1.42. □

We now exhibit some totally dual integral systems derived from incidence
matrices.
Recall that a digraph G consists of a finite vertex set V(G) and a family E(G)
of ordered pairs of elements of V(G). The pairs (v1 , v2 ) ∈ E(G) are called directed
edges or arrows. The “state of the art” reference for digraphs is the book by
Bang-Jensen and Gutin [6].
DEFINITION 3.6.4. Let G be a digraph with vertex set V(G) and edge set E(G).
Given a family ℱ of subsets of V(G), the one-way cut-incidence matrix of ℱ is the
matrix B′ = (bX,e )X∈ℱ ,e∈E(G) , where
1 if e ∈ δ+ (X)
{
bX,e =
0 otherwise.
The two-way cut-incidence matrix of ℱ is the matrix B′ = (bX,e )X∈ℱ ,e∈E(G) , where
⎨ 1 if e ∈ δ+ (X)
⎧

bX,e = −1 if e ∈ δ− (X)
0 otherwise.
⎩

Here δ+ (X) = {e = (z, w) ∈ E(G)∣ z ∈ X, w ∈ / X } is the set of edges leaving the

vertex set X and δ (X) is the set of edges entering the vertex set X. If ℱ = {{v} ∣ v ∈
−

V(G)}, the matrix B′ is simply called the incidence matrix of G.

REMARK 3.6.5. An interesting case occurs when the one-way and the two-way
cut-incidence matrix of the family ℱ coincide. In this case the rows of the matrix B
correspond to directed cuts only. Recall that δ+ (X) is a directed cut of a digraph G
∕ X ⊂ V(G) and δ− (X) = ∅.
if ∅ =
It was pointed out to us by Jens Vygen that the next result can be regarded as a
slight generalization of the Lucchesi, Younger Theorem [127]. It follows using the
technique of proof given in [120, Theorem 19.10].
PROPOSITION 3.6.6 ([175]). Let G be a digraph and ℱ a family of subsets of
V(G) such that the one-way cut incidence matrix B of ℱ is equal to the two-way cut
incidence matrix of ℱ . If ℱ satisfies the following three conditions:
(a) the rows of B are nonzero,
(b) if X, Y ∈ ℱ and X ∪ Y =∕ V(G), then X ∪ Y ∈ ℱ , and
(c) if X, Y ∈ ℱ and X ∩ Y ∕= ∅, then X ∩ Y ∈ ℱ ,
then the system Bx ≥ 1; x ≥ 0 is totally dual integral and the rational polyhedron
{ x∣ Bx ≥ 1; x ≥ 0} is integral.
132 3. POLYHEDRAL SETS, OPTIMIZATION AND SEMIGROUPS

PROOF. Let us prove that the system is TDI. Let c be an integral vector such
that the maximum in
(3.24) min{⟨c, x⟩∣ x ≥ 0; Bx ≥ 1} = max{⟨y, 1⟩∣ y ≥ 0; yB ≤ c}
is attained at a vector y0 = (yF1 , . . . , yFr ), where ℱ = {F1 , . . . , Fr }, note that c ≥ 0.
Since the set of vectors y for which the maximum in Eq. (3.24) is attained is a
face of the polytope
P = {y ∈ ℝr ∣ y ≥ 0; yB ≤ c}
and since any face of P is a compact set, one may assume
r
{ r }
∑ ∑
(3.25) yFi ∣Fi ∣2 = max yi ∣Fi ∣2 ⟨y, 1⟩ = ⟨y0 , 1⟩; y ≥ 0; yB ≤ c .
i=1 i=1
We claim that the family
𝒢 = {Fi ∣ yFi > 0}
is cross-free, that is, for any Fi , F j ∈ 𝒢 , at least one of the four sets Fi ∖ F j , F j ∖ Fi ,
Fi ∩ F j , V(G) ∖ Fi ∪ F j is empty. To prove the claim suppose F1 ∖ F2 ∕= ∅, F2 ∖ F1 ∕= ∅,
F1 ∩ F2 ∕= ∅, F1 ∪ F2 ∕= V(G) for some F1 , F2 ∈ 𝒢 (for simplicity we are assuming
i = 1, j = 2). Set  = min{yF1 , yF2 } and consider the vector y′0 in ℝr whose entries
are given by
y′F1 := yF1 − , y′F2 := yF2 − , y′F1 ∩F2 := yF1 ∩F2 + , y′F1 ∪F2 := yF1 ∪F2 + ,
and y′S := yS for all other S ∈ ℱ . Let B = (bFi ,α j ) be the one-way cut-incidence
matrix of ℱ , where α1 , . . . , αq are the edges of G. Consider any edge α j = (z, w).
For each 1 ≤ j ≤ q one has the following inequality
bF1 , j (yF1 − ) + bF2 , j (yF2 − ) + bF1 ∩F2 , j (yF1 ∩F2 + ) + bF1 ∪F2 , j (yF1 ∪F2 + )
≤ bF1 , j yF1 + bF2 , j yF2 + bF1 ∩F2 , j yF1 ∩F2 + bF1 ∪F2 , j yF1 ∪F2 .
To show the inequality consider two cases. Case (I): bF1 , j = bF2 , j = 0. If i = 1 or
i = 2 and z ∈ Fi , then w ∈ Fi and bF1 ∪F2 , j = 0. If i = 1 or i = 2 and z ∈ / Fi , then
w∈ / Fi and bF1 ∩F2 , j = 0. Hence in this case yF1 ∩F2 , j = 0 and yF1 ∪F2 , j = 0. Case
(II): bF1 , j = 0 and bF2 , j = 1. By the previous considerations one has that either
yF1 ∩F2 , j = 0 or yF1 ∪F2 , j = 0.
From the inequality above we obtain y′0 B ≤ y0 B. Since ⟨y′0 , 1⟩ = ⟨y0 , 1⟩, we
obtain a contradiction with the choice of y0 because
r
∑ r
∑
yFi ∣Fi ∣2 < y′Fi ∣Fi ∣2 .
i=1 i=1
Note that for any numbers a > b ≥ c > d > 0 with a+d = b+c one has the inequality
a2 + d2 > b2 + c2 . In our case to show the strict inequality we take a = ∣F1 ∪ F2 ∣,
b = ∣F1 ∣ ≥ c = ∣F2 ∣ and d = ∣F1 ∩ F2 ∣. Thus we have shown that the family 𝒢 is
cross-free.
 3. POLYHEDRAL SETS, OPTIMIZATION AND SEMIGROUPS 133

Let B′ be the submatrix of B whose rows correspond to the elements of the

family 𝒢 . Then
(3.26) max{⟨y′ , 1⟩∣ y′ ≥ 0; y′ B′ ≤ c} = max{⟨y, 1⟩∣ y ≥ 0; yB ≤ c}.
The inequality ≤ is clear because B′ is a submatrix of B. As the maximum in the
right hand side is attained at y0 , where y0 has zero entries in positions corresponding
to rows outside B′ , the reverse inequality follows.
Now the matrix B′ , being the two-way cut-incidence matrix of a cross-free family
𝒢 , is a network matrix, hence it is totally unimodular, see [120, Theorem 5.27].
Hence the polytope
{y′ ∣ y′ ≥ 0; y′ B′ ≤ c}
is integral, and consequently the maximum in the left hand side of Eq. (3.26) is
attained at an integral vector y′ . To complete the proof that the system Bx ≥ 1;
x ≥ 0 is TDI observe that y′ can be extended (by adding zero entries) to an integral
optimum solution y of the right hand side of Eq. (3.26).
Since the system Bx ≥ 1; x ≥ 0 is totally dual integral it follows from Theo-
rem 3.1.42 that the polyhedron { x∣ Bx ≥ 1; x ≥ 0} is integral. □
COROLLARY 3.6.7. [175] Let G be a connected digraph and let
ℱ = {X ∣ ∅ ∕= X ⊊ V(G); δ− (X) = ∅}.
If B is the one-way cut-incidence matrix of ℱ , then { x∣ Bx ≥ 1; x ≥ 0} is a
(nonempty) integral polyhedron.
PROOF. First we prove that δ+ (X) ∕= ∅ for X ∈ ℱ . Assume δ+ (X) = ∅. Pick
z ∈ V(G) ∖ X and x ∈ X. If (z, w) ∈ E(G) or (w, z) ∈ E(G), then w ∈ / X. There is
an undirected path {z0 = z, z1 , . . . , zr = x} from z to x, a contradiction since zi ∈
/ X
for all i.
To verify the hypothesis of Proposition 3.6.6 note that (a) is satisfied because
δ+ (X) ∕= ∅ for X ∈ ℱ . Take X, Y ∈ ℱ . From the inequality
∣δ− (X)∣ + ∣δ− (Y)∣ ≥ ∣δ− (X ∩ Y)∣ + ∣δ− (X ∪ Y)∣,
we get δ− (X ∩ Y) = ∅ and δ− (X ∪ Y) = ∅. Thus if X ∩ Y ∕= ∅ (resp. X ∪ Y ∕= V(G)),
then X ∩ Y ∈ ℱ (resp. X ∪ Y ∈ ℱ ). Hence conditions (b) and (c) are satisfied. On
the other hand by construction of ℱ the matrix B is also the two-way cut-incidence
matrix of ℱ . □
LEMMA 3.6.8. Let G be a connected bipartite graph with bipartition (V1 , V2 )
and let ℱ be the family
ℱ = {A ∪ A′ ∣ ∅ ∕= A ⊊ V1 ; N(A) ⊂ A′ ⊂ V2 } ∪ {A′ ∣ ∅ ∕= A′ ⊂ V2 }.
134 3. POLYHEDRAL SETS, OPTIMIZATION AND SEMIGROUPS

If G is regarded as the digraph with all its arrows leaving the vertex set V2 , then the
following equality holds
ℱ = {X ∣ ∅ ∕= X ⊊ V(G); δ− (X) = ∅}.
PROOF. It follows readily from the definitions. □

Exercises
3.6.9. If A is the incidence matrix of a digraph G, prove that A is totally
unimodular.
3.6.10. Let T : ℝq → ℝn be a linear map and let Q′ ⊂ ℝq be a rational
polyhedron. If T (ℤq ) ⊂ ℤn , then T (Q′ ) is a rational polyhedron.
3.6.11. Let T : ℝq → ℝn be a linear map and let Q′ ⊂ ℝq be an integral
polyhedron. If T (ℤq ) ⊂ ℤn , then T (Q′ ) is an integral polyhedron.

3.7. The elementary integral vectors

The notion of elementary integral vector occurs in convex analysis [148] and in the
theory of toric ideals of graphs [170, 184].
If α ∈ ℝq , its support is defined as
supp(α) = {i ∣ αi ∕= 0}.
Note that α = α+ − α− , where α+ and α− are two non negative vectors with disjoint
support.
REMARK 3.7.1. If α is a linear combination of β1 , . . . , βr ∈ ℝq , then
supp(α) ⊂ supp(β1 ) ∪ supp(β2 ) ∪ ⋅ ⋅ ⋅ ∪ supp(βr ).
DEFINITION 3.7.2. Let N be a linear subspace of ℚq . An elementary vector of
N is a nonzero vector α in N whose support is minimal with respect to inclusion,
i.e., supp(α) does not properly contain the support of any other nonzero vector in N.
The concept of elementary vector arises in graph theory when N is the kernel
of the incidence matrix of a graph G or a digraph 𝒟. The reader is referred to [148,
Section 22] for a precise interpretation of the vectors in N as flows of G which are
conservative at every vertex.
LEMMA 3.7.3. If N is a linear subspace of ℚq and α, β are two elementary
vectors of N with the same support, then α = λβ for some λ ∈ ℚ.
PROOF. If i ∈ supp(α), then one can write αi = λβi for some scalar λ. Since
supp(α − λβ) ⊊ supp(α), one concludes α − λβ = 0, as required. □
 3. POLYHEDRAL SETS, OPTIMIZATION AND SEMIGROUPS 135

DEFINITION 3.7.4. Let N be a linear subspace of ℚq . An elementary integral

vector or circuit of N is an elementary vector of N with relatively prime integral
entries.
This concept is naturally related with the circuits of vector matroids, as explained
later on in this chapter.
COROLLARY 3.7.5. If N is a linear subspace of ℚq , then number of circuits of
N is finite.
PROOF. It follows from Lemma 3.7.3. □
DEFINITION 3.7.6. Two vectors α = (αi ) and β = (βi ) in ℚq are in harmony if
αi βi ≥ 0 for every i.
LEMMA 3.7.7 ([147]). Let N be a linear subspace of ℚq . If 0 ∕= α ∈ N, then
there is an elementary vector γ ∈ N in harmony with α such that supp(γ) ⊂ supp(α).
PROOF. Let β = (β1 , . . . , βq ) be an elementary vector of N whose support is
contained in supp(α). By replacing β by −β one may assume αi βi > 0 for some i.
Consider
λ j = α j /β j = min {αi /βi ∣ αi βi > 0} .
1≤i≤q
Note that zi = (αi − λ j βi )αi ≥ 0 for all i. Indeed if αi βi ≤ 0, then clearly zi ≥ 0,
and if αi βi > 0 by the minimality of λ j one has zi ≥ 0. Thus the vector α − λ j β
is in harmony with α and its support is strictly contained in the support of α. If
α − λ j β = 0 we are done, otherwise we apply the same argument with α − λ j β
playing the role of α. Since being in harmony is an equivalence relation, the result
follows by a recursive application of this procedure. □
THEOREM 3.7.8 ([147]). If N is a vector subspace of ℚq and α ∈ N ∖ {0}, then
α can be written as α = β1 + ⋅ ⋅ ⋅ + βr for some elementary vectors β1 , . . . , βr of N
with r ≤ dim N such that
(i) β1 , . . . , βr are in harmony with α,
(ii) supp(βi ) ⊂ supp(α) for all i, and
(iii) supp(βi ) ⊂ supp(β1 ) ∪ ⋅ ⋅ ⋅ ∪ supp(βi−1 ) for all i ≥ 2.
PROOF. By induction on the number of elements in the support of α. If α is not an
elementary vector of N, then by Lemma 3.7.7 there is an elementary vector α1 ∈ N
in harmony with α such that supp(α1 ) ⊂ supp(α). Using the proof of Lemma 3.7.7
it follows that there is a positive scalar λ1 > 0 such that α − λ1 α1 is in harmony with
α and the support of α − λ1 α1 is strictly contained in supp(α). Therefore the result
follows applying induction. This proof uses the original argument given in [147].
□
136 3. POLYHEDRAL SETS, OPTIMIZATION AND SEMIGROUPS

COROLLARY 3.7.9. If N is a linear subspace of ℚq , then the circuits of N

generate N as a ℚ-vector space.
PROPOSITION 3.7.10 ([58]). Let A be an n × q integral matrix and let α be a
non zero vector in ker(A). If A has rank n, then α is an elementary vector of ker(A)
if and only if there is 0 ∕= λ ∈ ℚ such that
n+1
∑
α=λ (−1)k det[vi1 , . . . , vik−1 , vik+1 , . . . , vin+1 ]eik , (∗)
k=1
for some column vectors vi j of A. Here ek is the kth unit vector in ℝq .
PROOF. ⇐) Let A′ be the submatrix of A consisting of the column vectors
vi1 , . . . , vin+1 . Since A′ has order n × (n + 1) and has rank n, ker(A′ ) is generated by
a single nonzero vector. It follows readily that the vector α given by Eq. (∗) is an
elementary vector.
⇒) Let supp(α) = { j1 , . . . , jr } and let v1 , . . . , vq be the column vectors of A.
Since α ∈ ker(A) one has
α = λ j1 e j1 + ⋅ ⋅ ⋅ + λ jr e jr and λ j1 v j1 + ⋅ ⋅ ⋅ + λ jr v jr = 0, (1)
where 0 ∕= λ jk ∈ ℚ and v jk are columns of A for all k = 1, . . . , r. By the minimality
of the support of α one may assume that v j1 , . . . , v jr−1 are linearly independent.
Thus using that rank (A) = n, there are column vectors v jr+1 , . . . , v jn+1 such that the
set
𝒜′ = {v j1 , . . . , v jr−1 , v jr+1 , . . . , v jn+1 }
is linearly independent. One can write
𝒜 = 𝒜′ ∪ {v jr } = {vi1 , . . . , vin+1 }
such that 1 ≤ i1 < ⋅ ⋅ ⋅ < in+1 ≤ q. Consider the vector
n+1
∑
β= (−1)k det[vi1 , . . . , vik−1 , vik+1 , . . . , vin+1 ]eik . (2)
k=1
Note that in Eq. (1) and Eq. (2) the coefficient of e jr is non zero, hence there are
scalars 0 ∕= λ and cik in ℚ such that
∑
α − λβ = cik vik .
ik ∕= jr
Since the support of α − λβ is strictly contained in the support of β and since β is
an elementary vector of ker(A), one obtains α − λβ = 0, as required. □
COROLLARY 3.7.11. Let A be a t-unimodular matrix of order n × q and let α be
a circuit of ker(A), then
α = 1 e 1 + ⋅ ⋅ ⋅ + q e q ,
for some integers 1 , . . . , q in {0, ±1}.
 3. POLYHEDRAL SETS, OPTIMIZATION AND SEMIGROUPS 137

PROOF. It follows readily from the explicit description of the circuits given in
Proposition 3.7.10 and from Exercise 3.7.16. □

Matroids
Let M = (E, ℐ ) be a matroid on E, i.e., there is a collection ℐ of subsets of E
satisfying the following three conditions:
(i1 ) ∅ ∈ ℐ .
(i2 ) If I ∈ ℐ and I ′ ⊂ I, then I ′ ∈ ℐ .
(i3 ) If I1 and I2 are in ℐ and ∣I1 ∣ < ∣I2 ∣, then there is an element e of I2 ∖ I1
such that I1 ∪ {e} ∈ ℐ .
The members of ℐ are the independent sets of M. It is convenient to write ℐ (M)
for ℐ and E(M) for E, particularly when several matroids are being considered. A
subset of E that is not in ℐ is called dependent. A maximal independent set of M
with respect to inclusion is called a basis. A minimal dependent set in M is called
a circuit. The reader is referred to [141, 195] for the general theory of matroids.

PROPOSITION 3.7.12. [141, Proposition 1.1.1, p. 8] Let E be the set of column

labels of an m × n matrix A over a field K, and let ℐ be the set of subsets B of E
for which the multiset of columns labeled by B is linearly independent in K m . Then
M[A] = (E, ℐ ) is a matroid.

The matroid M[A] obtained as above from the matrix A is called the vector
matroid of A. Any matroid M arising from a matrix A is called representable over
the field K.

THEOREM 3.7.13. [141, Corollary 1.2.5, p. 18] Let ℬ be a non-empty family of

subsets of X. Then ℬ is the collection of bases of a matroid on X if and only if
the following exchange property is satisfied: If B1 and B2 are members of ℬ and
b1 ∈ B1 ∖ B2 , then there is an element b2 ∈ B2 ∖ B1 such that (B1 ∖ {b1 }) ∪ {b2 } is
in ℬ .

The family of bases of a matroid satisfies the following symmetric exchange

property (see [123]):

THEOREM 3.7.14. If B1 and B2 are bases of a matroid M and b1 ∈ B1 ∖ B2 ,

then there is an element b2 ∈ B2 ∖ B1 such that both (B1 ∖ {b1 }) ∪ {b2 } and
(B2 ∖ {b2 }) ∪ {b1 } are bases of M.

It is well known [141] that all basis of a matroid M have the same number of
elements, this common number is called the rank of the matroid.
138 3. POLYHEDRAL SETS, OPTIMIZATION AND SEMIGROUPS

Elementary integral vectors and circuits of vector matroids. Let A be a non-

negative integral matrix with column vectors v1 , . . . , vq , and let M[A] be the vector
matroid on 𝒜 = {v1 , . . . , vq } over the field ℚ.
It is interesting to observe that there is a correspondence

Circuits of ker(A) −→ Circuits of M[A]

given by

α = (α1 , . . . , αq ) −→ C(α) = {vi ∣ i ∈ supp(α)}.

Thus the set of elementary integral vectors of the kernel of A is the algebraic
realization of the set of circuits of the vector matroid M[A].

Exercises
3.7.15. Prove that the elementary integral vector of the kernel of the matrix M
are precisely the row vectors of the matrix A with a ± sign
⎛ ⎞
( ) 2 −3 1 0
4 3 1 0 ⎜ 3 −4 0 1 ⎟
M= A=⎜ ⎟.
0 1 3 4 ⎝ 0 1 −3 2 ⎠
1 0 −4 3
3.7.16. Let A be an n × q matrix with entries∑ in an integral domain R and let
F1 , . . . , Fn be the rows of A. Assume that Fi = j∕=i λ j F j , for some λ j ’s in the field
of fractions of R. Consider the R-linear maps
ϕ
ϕ : Rq −→ Rn α 7−→ Aα
ϕ′
ϕ′ : Rq −→ Rn−1 α 7−→ A′ α,
where A′ is the matrix obtained from A by removing its ith row, prove that ker(ϕ) =
ker(ϕ′ ).
3.7.17. Let X = { x1 , . . . , x4 } and let
ℬ = {{ x1 , x2 }, { x2 , x3 }, { x3 , x4 }, { x1 , x4 }}.
Prove that ℬ satisfies the bases exchange property of matroids.
Chapter 4
Monomial Rings and Optimization Problems

In this chapter we deal with monomial ideals and their underlying point con-
figurations and clutters. We study various homogeneous subrings and prime ideals
associated with them, e.g., toric ideals, Ehrhart rings and blowup rings. We will
establish some connections between algebraic properties of certain graded algebras
and combinatorial optimization properties of clutters. Some applications to Rees
algebras theory and optimization are presented. In particular for edge ideals we
gain insight on the algebraic properties of their blowup rings.
We study symbolic Rees algebras of edge ideals of clutters. The study of sym-
bolic powers of edge ideals from the point of view of graph theory and combinatorics
was initiated in [156] and further elaborated on in [171, 193]. A breakthrough in
this area is the translation of combinatorial problems (e.g., the Conforti-Cornuéjols
conjecture [38], the max-flow min-cut property, the idealness of a clutter, or the in-
teger rounding property) into algebraic problems of blowup algebras of edge ideals
[17, 51, 80, 84].

4.1. Gröbner bases

In this section we review some basic facts and definitions on Gröbner bases. The
reader may consult [40] for a detailed discussion on this topic and for the missing
proofs of this section.
Let K be a field and let R = K[x1 , . . . , xn ] be a polynomial ring. The monomials
in
ℳn = { x a ∣ a ∈ ℕ n }
are called terms, i.e., the terms of R are the monomials λxa with λ = 1.
DEFINITION 4.1.1. A total order ≻ of ℳn is called a monomial order or term
order if
(a) p ર 1 for all p ∈ ℳn , and
(b) for all p, q, r ∈ ℳn , q ≻ p implies qr ≻ pr.
139
140 4. MONOMIAL RINGS AND OPTIMIZATION PROBLEMS

Two examples of monomial orders in ℳn are the lexicographical order or lex

order defined as xb ≻ xa if and only if the last nonzero entry of b − a is positive, and
the reverse lexicographical order or revlex order given by xb ≻ xa if and only if the
last nonzero entry of b − a is negative. Another quite useful order is the elimination
order on K[x1 , . . . , xn , t1 , . . . , tq ] with respect to x1 , . . . , xn . This order is given by
xa t c ≻ xb t d
if and only if deg(xa ) ≻ deg(xb ), or both degrees are equal and the last nonzero
entry of (a, c) − (b, d) is negative.
In the sequel we assume that a monomial order for ℳn has been fixed. Let f
be a nonzero polynomial in R. One can write
f = a1 m1 + ⋅ ⋅ ⋅ + ar mr ,
with ai ∈ K ∗ = K ∖ {0}, mi ∈ ℳn and m1 ≻ ⋅ ⋅ ⋅ ≻ mr . The leading term m1 of f
is denoted by in( f ). The leading coefficient a1 of f and its leading monomial a1 m1
are denoted by lc( f ) and lm( f ) respectively.
DEFINITION 4.1.2. Let I be an ideal of R. The initial ideal of I, denoted by
in(I), is given by
in(I) = ({in( f )∣ f ∈ I }).
DEFINITION 4.1.3. Let F = { f1 , . . . , fq } ⊂ R ∖ {0} be a set of polynomials in
R. One says that f reduces to g modulo F, denoted f →F g , if
g = f − (au/ lc( fi )) fi
for some fi ∈ F, u ∈ ℳn , a ∈ K ∗ such that a ⋅ u ⋅ in( fi ) occurs in f with coefficient
a.
PROPOSITION 4.1.4. The reduction relation “−→F ” is Noetherian, i.e., any
sequence of reductions
g1 −→F g2 −→F ⋅ ⋅ ⋅ −→F gi −→F ⋅ ⋅ ⋅
is stationary.
PROOF. Notice that at the ith step of the reduction some term of gi is replaced
by terms of lower degree. Therefore if the sequence above is not stationary, then
there is a never ending decreasing sequence of terms in ℳn , but this is impossible
according to Dickson’s lemma (see Lemma 1.2.2). □
DEFINITION 4.1.5. Let I ∕= (0) be an ideal of R and let F = { f1 , . . . , fr } be a
subset of I. The set F is called a Gröbner basis of I if
in(I) = (in( f1 ), . . . , in( fr )).
DEFINITION 4.1.6. A Gröbner basis F = { f1 , . . . , fr } of an ideal I is called a
reduced Gröbner basis of I if:
 4. MONOMIAL RINGS AND OPTIMIZATION PROBLEMS 141

(i) lc( fi ) = 1 ∀i, and

(ii) none of the terms occurring in fi belongs to in(F ∖ { fi }) ∀i.
Note that by the Hilbert basis theorem (see Corollary 1.1.8) a Gröbner basis of
I always exist. A nice feature of a reduced Gröbner basis is its uniqueness [40].
DEFINITION 4.1.7. Let f, g ∈ R. The S-polynomial of f and g is given by
[in( f ), in(g)] [in( f ), in(g)]
S ( f, g) = f− g,
lm( f ) lm(g)
where [ f, g] := lcm( f, g) and lcm is the least common multiple.
Given a polynomial ideal one can determine a Gröbner basis using the next
fundamental procedure:
THEOREM 4.1.8 (Buchberger [31]). If F = { f1 , . . . , fq } is a set of generators of
an ideal I of R, then one can construct a Gröbner basis for I using the following
algorithm:
Input: F
Output: a Gröbner basis G of I
Initialization: G := F, B := {{ fi , f j }∣ fi ∕= f j ∈ G}
while B ∕= ∅ do
pick any { f, g} ∈ B
B := B ∖ {{ f, g}}
r := remainder of S ( f, g) with respect to G
if r ∕= 0 then
B := B ∪ {{r, h}∣ h ∈ G}
G := G ∪ {r}

Exercises
4.1.9. Let f and g be two polynomials in R and let F = { f, g}. If lm( f ) and
lm(g) are relatively prime, then S ( f, g) →F 0.
4.1.10. Let I be an ideal of a ring R and let 0 ∕= f ∈ R, then
( f )(I : f ) = I ∩ ( f ).
4.1.11. Let R = K[x] be a polynomial ring, where K is a field. Recall that f is
a binomial in R if f = xα − xβ for some α, β in ℕn . Use Buchberger’s algorithm to
prove that an ideal of R generated by a finite set of binomials has a Gröbner basis
consisting of binomials.
4.1.12. Let I be an ideal of a ring R and f ∈ R. Prove that the following
statements are equivalent:
142 4. MONOMIAL RINGS AND OPTIMIZATION PROBLEMS

(a) f is regular on R/I.

(b) (I : R f ) = I.
(c) I = (I, 1 − t f ) ∩ R, where t is a new variable.
4.1.13. If I and J are two ideals of a polynomial ring R over a field K, then
I ∩ J = (tI, (1 − t)J) ∩ R.
4.1.14. Let I be an ideal of a ring R and f ∈ R, the saturation of I with respect
to f is
(I : f ∞ ) := {r ∈ R ∣ r f i ∈ I, for some i ≥ 1 .
}

Prove the equality (I : f ∞ ) = (I, 1 − t f ) ∩ R, where t is a variable.

4.1.15. Let ℕn be the set of n-tuples of non-negative integers endowed with the
partial order given by
(a1 , . . . , an ) ≥ (b1 , . . . , bn )
if and only if ai ≥ bi for all i. If A ⊂ ℕn , prove that A has only a finite number of
minimal elements.
4.1.16. If R = K[x0 , . . . , xn ] is a polynomial ring over a field K and
ℬ = { xiq x j − xi xqj ∣ 0 ≤ i < j ≤ n},
then ℬ is a Gröbner basis of I = (ℬ ) with respect to the revlex ordering. Prove that
the a-invariant of R/I is n(q − 1).
4.1.17. (Macaulay) Let R = K[x1 , . . . , xn ] be a polynomial ring over a field K.
If I is a graded ideal of R, then R/I and R/ in(I) have the same Hilbert function.
4.1.18. Let R = K[x0 , . . . , xn ] be a polynomial ring over a finite field K and let
I be the binomial ideal
I = ({ xiq−1 − x0q−1 ∣ 1 ≤ i ≤ n}),
where q is the number of elements of K. Prove that I is a complete intersection and
dim R/I = 1. Compute the Hilbert series and the Hilbert polynomial of R/I. Then
show that the multiplicity of R/I is equal to (q − 1)n . Prove that I is a radical ideal
and find the primary decomposition of I as an intersection of prime ideals. Prove
that xi is not a zero divisor of R/I for any i.

4.2. Normal rings

Let B ⊂ D be a ring extension, i.e., B is a subring of D. Let F be a finite subset
of D. Recall that B[F], the subring of D generated by F, is the intersection of all
subrings of D that contain B ∪ F. If D = B[F], we say that D is a B-algebra of finite
type or a finitely generated B-algebra. We say that D is a finite B-algebra if D is a
finitely generated B-module.
 4. MONOMIAL RINGS AND OPTIMIZATION PROBLEMS 143

DEFINITION 4.2.1. An element d ∈ D is integral over B, if there is a monic

polynomial 0 ∕= f (x) ∈ B[x] such that f (d) = 0. The set B of all elements d ∈ D
which are integral over B is called the integral closure of B in D.
A ring extension B ⊂ D is called an integral extension if d is integral over B for
all d ∈ D. In this case we say that D is integral over B. A homomorphism of rings
ϕ : B → D is called integral if D is integral over its image ϕ(B).
PROPOSITION 4.2.2. Let B be a subring of D. If D is a finitely generated
B-module, then D is integral over B.
PROOF. Let δ ∈ D. There are α1 , . . . , αn ∈ D such that
D = Bα1 + ⋅ ⋅ ⋅ + Bαn .
One can write δαi = j=1 mi j α j , where mi j ∈ B. Set M = (mi j ), N = M − δI and
∑n

α = (α1 , . . . , αn ). Here I denotes the identity matrix. Since Nαt = 0, one can use the
formula Nadj(N) = det(N)I to conclude αi det(N) = 0 for all i. Hence det(N) = 0.
To complete the proof note that
f (x) = (−1)n det(M − xI)
is a monic polynomial in B[x] and f (δ) = (−1)n det(N) = 0. □
COROLLARY 4.2.3. If B is a subring of D, then the integral closure B of B in D
is a subring of D.
PROOF. Let α, β ∈ B. If α and β satisfy monic polynomials with coefficients in
B of degree m and n respectively, then B[α, β] is a finitely generated B-module with
basis
{αi β j ∣ 0 ≤ i ≤ m − 1 and 0 ≤ j ≤ n − 1}.
Hence by Proposition 4.2.2 B[α, β] is integral over B. In particular α ± β and αβ
are integral over B. □
COROLLARY 4.2.4. If D is a B-algebra of finite type, then D is integral over B
if and only if D is finite over B.
PROOF. ⇒) This implication follows using the arguments given in the proof of
Corollary 4.2.3.
⇐) It follows from Proposition 4.2.2. □
PROPOSITION 4.2.5. [131, Theorem 5] Let B ⊂ D be an integral extension of
rings. If D is integral over B, then dim(B) = dim(D).
DEFINITION 4.2.6. Let B be an integral domain and let KB be its field of fractions.
The integral closure of B in KB , denoted by B, is called the integral closure or
normalization of B.
144 4. MONOMIAL RINGS AND OPTIMIZATION PROBLEMS

By Corollary 4.2.3 the integral closure of B is a subring of KB . If B = B we say

that B is integrally closed or normal. If B is not a domain we say that B is normal
if B𝔭 is a normal domain for every prime ideal 𝔭 of B. A useful fact is that a normal
ring is a direct product of finitely many normal domains (see [56]).

Exercises
4.2.7. Let ϕ : B → D be an isomorphism between two integral domains. If KB
(resp. KD ) is the field of fractions of B (resp. D), then ϕ extends to an isomorphism
ϕ from KB to KD such that ϕ(B) = D.
4.2.8. If B is a unique factorization domain, show that B is normal.
4.2.9. Let B be a domain and let x be an indeterminate over B. Then B is normal
if and only if B[x] is normal.

4.3. Monomial subrings

Let R = K[x] = K[x1 , . . . , xn ] be a polynomial ring over a field K and let F =
{ xv1 , . . . , xvq } be a finite set of distinct monomials in R such that xvi ∕= 1 for all i.
We set fi = xvi for all i.
There is an isomorphism, denoted by log, between the multiplicative semigroup
ℳn of monomials of R and the additive semigroup ℕn :
log
ℳn = {Monomials of R} −→ ℕn

xa := x1a1 ⋅ ⋅ ⋅ xnan −→ a = (a1 , . . . , an )

xa+b = xa xb −→ a + b.
The image of F under this isomorphism is denoted by 𝒜:
log
F = { xv1 , . . . , xvq } −→ 𝒜 = {v1 , . . . , vq },
i.e., log( fi ) = vi and 𝒜 := log(F).
Let ℱ be the family of all subrings D of R such that K ∪ F ⊂ D. Recall that the
monomial subring generated or spanned by F is:
∩
K[F] := D.
D∈ℱ
Thus formally K[F] is the subring of R generated by the set F ∪ K. The elements
of K[F] are polynomial expressions with coefficients in K:
∑ ( )a ( )a
ca xv1 1 ⋅ ⋅ ⋅ xvq q
finite
 4. MONOMIAL RINGS AND OPTIMIZATION PROBLEMS 145

with ca ∈ K. Let ℕ𝒜 = ℕv1 + ⋅ ⋅ ⋅ + ℕvq be the subsemigroup of ℕn generated by 𝒜.

Hence as a K-vector space K[F] is generated by the set of monomials of the form
xα with α ∈ ℕ𝒜.
Notice that K[F] is an integral domain that coincides with K[ℕ𝒜], the semigroup
ring of the semigroup ℕ𝒜:
K[F] = K[ℕ𝒜] = K[{ xa ∣ a ∈ ℕ𝒜}].
There is a graded epimorphism of K-algebras:
ϕ
ϕ : B = K[t1 , . . . , tq ] −→ K[F] −→ 0, ti 7−→ fi ,
where B is a polynomial ring graded by deg(ti ) = deg( fi ) and K[F] is a graded
subring of R graded by K[F]i = K[F] ∩ Ri . The map ϕ is given by
ϕ(h(t1 , . . . , tq )) = h( f1 , . . . , fq ), for all h ∈ B.
The kernel of ϕ, denoted by P, is called the presentation ideal or toric ideal of K[F]
with respect to F. Thus P := ker(ϕ). Sometimes we will write PF instead of P to
emphasize the set F.
PROPOSITION 4.3.1. P = ker(ϕ) = (t1 − f1 , . . . , tq − fq ) ∩ B.
PROOF. Let h ∈ ker(ϕ). Making ti = (ti − fi ) + fi we can write h as:
h = h(t1 , . . . , tq ) = γ aγ (t1 − f1 )γ1 ⋅ ⋅ ⋅ (tq − fq )γq + β aβ f β .
∑ ∑

Since h( f1 , . . . , fq ) = 0 one derives β aβ f β = 0, and consequently h is in (t1 −

∑
f1 , . . . , tq − fq ) ∩ B. The other containment also follows readily. □
THEOREM 4.3.2. Let R[t] = R[t1 , . . . , tq ] be a polynomial ring with a term order
≺ such that terms in the xi ’s are greater than terms in the ti ’s. If I is an ideal of R[t]
with a Gröbner basis G, then G ∩ K[t1 , . . . , tq ] is a Gröbner basis of I ∩ K[t1 , . . . , tq ].
PROOF. Set B = K[t1 , . . . , tq ] and I c = I ∩ B. Assume m ∈ in(I c ). There is
f ∈ I c with in( f ) = m. Hence m = λ in(g) for some g ∈ G, because G is a Gröbner
basis. Since m ∈ B and xα ≻ tβ for all α and β we obtain g ∈ G ∩ B, that is,
m ∈ (in(G ∩ B)). Thus
in(I c ) = (in(G ∩ B)),
as required. □
DEFINITION 4.3.3. A binomial f in B = K[t1 , . . . , tq ] is a difference of two
monomials, i.e., f = tα − tβ , for some α, β in ℕq . An ideal generated by binomials
is called a binomial ideal.
PROPOSITION 4.3.4. The toric ideal P is a graded prime ideal generated by a
finite set of binomials.
146 4. MONOMIAL RINGS AND OPTIMIZATION PROBLEMS

PROOF. Observe B/P ≃ K[F] and that K[F] is a Noetherian domain. If a

polynomial h0 + ⋅ ⋅ ⋅ + h s is in P and hi ∈ Bi , then ϕ(hi ) = 0 because ϕ is graded.
Hence hi ∈ Pi for all i, and P is a graded ideal.
Consider E = {t1 − f1 , . . . , tq − fq }. Let f, g ∈ E. Since the S -polynomial
S ( f, g) is again a binomial and the remainder of S ( f, g) with respect to E is also a
binomial, it follows that the output of the Buchberger algorithm (see Theorem 4.1.8)
is a Gröbner basis of L = (E) consisting of binomials. Hence by Theorem 4.3.2 and
Proposition 4.3.1 we get that P is generated by a finite set of binomials. □
COROLLARY 4.3.5. The toric ideal P has a Gröbner basis consisting of binomials
with respect to any monomial order.
PROOF. As P is a binomial ideal, the assertion follows from the proof of
Proposition 4.3.4. □
DEFINITION 4.3.6. The associated matrix of the subring K[F] is the n × q matrix,
denoted by A, whose columns are the vectors log( f1 ), . . . , log( fq ).
c
Notation For use below we set f c = f1c1 ⋅ ⋅ ⋅ fq q if c ∈ ℕq .
COROLLARY 4.3.7. If A is the associated matrix of K[F], then
P = ({tα+ − tα− ∣ α ∈ ℤq and Aα = 0}).
PROOF. Let tα − tβ be a binomial in P. If supp(α) ∩ supp(β) ∕= ∅, then we can
write f α − f β = f γ ( f α − f β ) = 0, where γi = min{αi , βi }. Note that α′ and β′
′ ′

have disjoint support, and f α − f β = 0 if and only if A(α − β) = 0. To complete the

argument use the fact that P is generated by binomials. □
DEFINITION 4.3.8. A binomial tα − tβ ∈ P is called primitive if there is no other
binomial tγ − tδ ∈ P such that tγ divides tα and tδ divides tβ .
LEMMA 4.3.9. If f is a binomial in the reduced Gröbner basis of P with respect
to some term order, then f is a primitive binomial.
PROOF. Let G be the reduced Gröbner basis of P. Suppose that f = tα+ − tα− and
α+ ≻ α− . Assume g = tβ+ − tβ− ∕= f is a binomial in P such that tβ+ divides tα+ and
tβ− divides tα− . Note that β+ ∕= α+ and β− ∕= α− , otherwise f = g. If β+ ≻ β− , then
tβ+ ∈ in(G ∖ { f }) and consequently tα+ ∈ in(G ∖ { f }), a contradiction. If β− ≻ β+ ,
then tβ− ∈ in(G ∖ { f }) and consequently tα− ∈ in(G ∖ { f }), a contradiction. □
According to [170], the set 𝒢P of all primitive binomials in a toric ideal P is
finite and is called the Graver basis of P. See Theorem 5.4.4.
DEFINITION 4.3.10. The universal Gröbner basis of a toric ideal P is the union
of all reduced Gröbner basis of P. It is denoted by 𝒰P .
 4. MONOMIAL RINGS AND OPTIMIZATION PROBLEMS 147

DEFINITION 4.3.11. Let A be the associated matrix of K[F]. If α is a circuit

of ker(A), we call the binomial tα+ − tα− a circuit of P. The set of circuits of P is
denoted by 𝒞P .
PROPOSITION 4.3.12 ([170]). If P is a toric ideal, then 𝒞P ⊂ 𝒰P ⊂ 𝒢P .
PROOF. The second inclusion follows from Lemma 4.3.9. To show the first
inclusion take a circuit f = tα+ − tα− of P. We may assume that
f = t1α1 ⋅ ⋅ ⋅ trαr − tr+1
αr+1
⋅ ⋅ ⋅ tαs s (αi ≥ 1).
Set T = {t1 , . . . , tq }. We reorder the variables as T ∖ {t1 , . . . , t s }, t1 , . . . , t s . Consider
the lexicographical order: ta ≻ tb if the first non-zero entry of a − b is positive. Let
G be the reduced Gröbner basis of P with respect to ≻ and let g = tβ+ − tβ− be a
binomial in G so that in(g) = tβ+ divides in( f ). Clearly supp(tβ+ ) ⊂ supp(tα+ ). By the
choice of term order it follows that supp(tβ− ) ⊂ supp(tα ). Thus supp(β) ⊂ supp(α).
Since α is a circuit we get equality. Thus using Lemma 3.7.3 we obtain that β = λα
for some λ ∈ ℚ, consequently α = β because the non-zero entries of α (resp. β) are
relatively prime. □
COROLLARY 4.3.13. If the associated matrix of K[F] is t-unimodular, then
𝒞P = 𝒰P = 𝒢P .
PROOF. It follows from Exercise 4.3.30. □
DEFINITION 4.3.14. A universal Gröbner basis of a polynomial ideal I is a finite
set 𝒰 ⊂ I which is a Gröbner basis of I with respect to all term orders.
PROPOSITION 4.3.15. If P is the toric ideal of K[F], then the set 𝒢P of primitive
binomials in P is a universal Gröbner basis.
PROOF. It follows from Proposition 4.3.12. □
DEFINITION 4.3.16. Let K ⊂ L be a field extension. A subset of L which
is algebraically independent and is maximal with respect to inclusion is called a
transcendence basis of L over K.
THEOREM 4.3.17. [118, Theorem 8.35] If K ⊂ L is a field extension, then any
two transcendence bases have the same cardinality.
DEFINITION 4.3.18. Let K ⊂ L be a field extension. The transcendence degree
of L over K, denoted trdegK (L), is the cardinality of any transcendence basis of L
over K.
PROPOSITION 4.3.19. [187, Corollary 2.1.13] Let K be a field and let D be a
finitely generated K-algebra. If D is an integral domain with field of fractions L,
then
dim(D) = trdegK (L).
148 4. MONOMIAL RINGS AND OPTIMIZATION PROBLEMS

PROPOSITION 4.3.20. If A is the associated matrix of K[F], then

dim K[F] = rank(A).
PROOF. Let vi = (vi1 , . . . , vin ) for i = 1, . . . , q and let A = (v ji ) be the n × q matrix
associated to K[F]. Set r = rank(A). By Proposition 4.3.19 we have the equality
dim K[F] = trdegK K(F), where K(F) is the field of fractions of K[F]. Therefore
it suffices to prove the equality r = trdegK K(F). We may assume v1 , . . . , vr is a
basis for the column space of A. We now show that f1 , . . . , fr are algebraically
independent over K. If f1 , . . . , fr satisfy a polynomial relation, then there are
a = (a1 , . . . , ar , 0, . . . , 0) ∈ ℕq and c = (c1 , . . . , cr , 0, . . . , 0) ∈ ℕq
such that f a = f c . Hence A(a) = A(c) and γ1 v1 + ⋅ ⋅ ⋅ + γr vr = 0, where γi = ai − ci .
Therefore ai = ci , that is f a = f c is a trivial relation. It remains to show that
fr+i is integral over the quotient field K( f1 , . . . , fr ). Pick s ∈ ℕ such that svr+i =
s = f λ1 ⋅ ⋅ ⋅ f λr , as required. □
λ1 v1 + ⋅ ⋅ ⋅ + λr vr , where λ j ∈ ℤ for all j. Therefore fr+i 1 r

DEFINITION 4.3.21. Let R = ⊕∞ i=0 Ri be the standard grading of R and let r ∈ ℕ.

The rth Veronese subring of R is:
R(r) := K ⊕ Rr ⊕ R2r ⊕ ⋅ ⋅ ⋅ ⊕ Rir ⊕ ⋅ ⋅ ⋅
An element of Rir is said to have normalized degree i. Thus R(r) is a standard
K-algebra generated by all the monomials of normalized degree 1.
COROLLARY 4.3.22. Let P the toric ideal of the Veronese subring R(r) . If r ≥ 2,
then P is generated by homogeneous binomials of degree two.
PROOF. Consider a polynomial ring B = K[{tα ∣ α ∈ ℕn ; ∣α∣ = r}, with one
variable tα for each monomial xα of degree r. Let ϕ be the graded homomorphism
ϕ
ϕ : B −→ R(r) induced by tα 7−→ xα , where B has the standard grading and R(r)
has the normalized ⊕grading. By Proposition 4.3.4 the toric ideal P = ker(ϕ) is a
graded ideal P = ∞ s=2 P s generated by homogeneous binomials. Hence we need
only show that all the binomials in P s are in (P2 ) for s ≥ 2. The rest of the proof
follows by induction on s. □
Normality of monomial subrings. Here we present a description of the integral
closure of S = K[F]. For convenience recall that if S is a domain with field of
fractions KS , the normalization or integral closure of S is the subring S consisting
of all the elements of KS which are integral over S .
THEOREM 4.3.23. The integral closure of K[F] is given by
K[F] = K[{ xa ∣ a ∈ ℤ𝒜 ∩ ℝ+ 𝒜}].
PROOF. Set D = K[{ xa ∣ a ∈ ℤ𝒜 ∩ ℝ+ 𝒜}]. First we show that K[F] ⊂ D. Take
z ∈ K[F], since K[F] ⊂ R and R is normal we get K[F] ⊂ R and z ∈ R. Hence one
 4. MONOMIAL RINGS AND OPTIMIZATION PROBLEMS 149

can write
z = d1 xγ1 + ⋅ ⋅ ⋅ + dr xγr ,
where di ∈ K ∖{0} and γ1 , . . . , γr distinct nonzero points in ℕn . Write γi = ni βi , with
ni equal to the gcd of the entries of γi (observe that β1 , . . . , βr are not necessarily
distinct). Consider the cone C spanned by 𝒜 and {β1 , . . . , βr }. One must show
that βi ∈ ℝ+ 𝒜 for all i. Assume on the contrary that C is not equal to ℝ+ 𝒜. By
Theorem 3.1.36 one may assume that ℝ+ β1 is a face of C not contained in ℝ+ 𝒜.
Let Ha be a hyperplane so that Ha ∩ C = ℝ+ β1 and C ⊂ Ha− . There is 1 ≤ ` ≤ r
such that
n` = sup {n j ∣ γ j = n j β1 }.
1≤ j≤r
Since z is integral over K[F] it satisfies a monic polynomial f of degree m with
coefficients in K[F]. To derive a contradiction we claim that the term xmγ` occurs
only once in the expansion of f (z) as a sum of monomials. Assume the equality
(xγ` )m = xα (xγi1 )mi1 ⋅ ⋅ ⋅ (xγit )mit ,
where mi j > 0 for all j, m ≥ tj=1 mi j and α ∈ ℝ+ 𝒜. As γ` = n` β1 , one has
∑

⟨mγ` , a⟩ = 0 and from this equality one rapidly derives

⟨α, a⟩ = ⟨βi1 , a⟩ = ⋅ ⋅ ⋅ = ⟨βit , a⟩ = 0.
Hence α, βii , . . . , βit ∈ Ha ∩ C = ℝ+ β1 . Note β1 ∈
/ ℝ+ 𝒜. Thus α = 0 and βi j = β1
for all j. Therefore t = 1 and γ` = γi1 , as claimed. Altogether we get f (z) ∕= 0,
which is impossible. This proves that γi ∈ ℝ+ 𝒜 for all i. On the other hand using
that z belongs to the field of fractions of K[F] it follows that γi ∈ ℤ𝒜 for all i.
Let us now show the other inclusion D ⊂ K[F]. Note the equality
ℤ𝒜 ∩ ℝ+ 𝒜 = ℤ𝒜 ∩ ℚ+ 𝒜,
which follows from Theorem 3.1.14. A straightforward calculation shows that xα
is in the field of fractions of K[F] if α ∈ ℤ𝒜, and xα is an integral element over
K[F] if α ∈ ℚ+ 𝒜. Hence D ⊂ K[F]. □
COROLLARY 4.3.24. K[F] is normal if and only if
ℕ𝒜 = ℤ𝒜 ∩ ℝ+ 𝒜.
PROOF. It suffices to observe the equality K[F] = K[{ xα ∣α ∈ ℕ𝒜}] and to use
the description of K[F] given above. □
COROLLARY 4.3.25. If the matrix A is t-unimodular, then K[F] is normal.
PROOF. It follows from Theorem 3.5.7 and Corollary 4.3.24. □
⊕∞
DEFINITION 4.3.26. A decomposition K[F] = i=0 K[F]i of the K-vector space
K[F] is an admissible grading if K[F] is a positively graded K-algebra with respect
to this decomposition and each component K[F]i has a finite K-basis consisting of
monomials.
150 4. MONOMIAL RINGS AND OPTIMIZATION PROBLEMS

THEOREM 4.3.27. (Danilov, Stanley, [23, Theorem 6.3.5]) If S = K[F] is

normal, then the canonical module ωS of S , with respect to an arbitrary admissible
grading, can be expressed as
(4.27) ωS = ({ xa ∣ a ∈ ℕ𝒜 ∩ ri(ℝ+ 𝒜)}).
This formula represents the canonical module of K[F] as an ideal of K[F]
generated by monomials.
THEOREM 4.3.28. (Hochster [113]) If K[F] is a normal domain, then K[F] is
Cohen-Macaulay.
PROPOSITION 4.3.29. [187, Proposition 4.2.3] If K[F] is a normal standard
graded K-algebra, then
a(K[F]) = − min{i∣ (ωK[F] )i ∕= 0}.

Exercises
4.3.30. Let K[F] = K[xv1 , . . . , xvq ] be a monomial subring and let A be the
matrix with columns v1 , . . . , vq . If A is t-unimdular and tβ+ − tβ− is a primitive
binomial of the toric ideal of K[F], then β is a circuit of ker(A).
Hint Use Corollary 3.7.11 and Lemma 3.7.7.
4.3.31. Let P be the toric ideal of a monomial subring K[xv1 , . . . , xvq ] and let A
be the matrix with columns v1 , . . . , vq . If A is unimodular, then the reduced Gröber
basis of P with respect to any term order consists of square-free binomials.
4.3.32. Prove that the integral closure of S = K[x3 , y3 , x2 y] is S [xy2 ].
4.3.33. Let 𝒜 = {v1 , . . . , vq } ⊂ ℕn and C = ℕv1 + ⋅ ⋅ ⋅ + ℕvq . Prove that the
semigroup ring K[C] is normal if and only if any element α ∈ ℤ𝒜 satisfying rα ∈ C
for some integer r ≥ 1 belongs to C.
4.3.34. Let K ⊂ L be a field extension and let E be a finite subset of L. Prove
that the collection ℐ of subsets of E that are algebraically independent over K is
the set of independent sets of a matroid on E. This type of matroids are called
algebraic.
4.3.35. Let 𝒜 = {v1 , . . . , vq } ⊂ ℤn , let K be a field, and let
S = K[xv1 , . . . , xvq ] ⊂ K[x1±1 , . . . , xn±1 ].
Prove that S is normal if and only if ℕ𝒜 = ℤ𝒜 ∩ ℝ+ 𝒜.
4.3.36. Let r ≥ 1 be an integer and let
Vr = { xi1 ⋅ ⋅ ⋅ xir ∣ 1 ≤ i1 < ⋅ ⋅ ⋅ < ir ≤ n}
 4. MONOMIAL RINGS AND OPTIMIZATION PROBLEMS 151

be the set of square–free monomials of degree r of R. The subring K[Vr ] ⊂ R is

called the square-free rth Veronese subring . Let P be the toric ideal of K[Vr ]. If
r ≥ 2, prove that P is generated by homogeneous binomials of degree two.

4.4. Homogeneous subrings

The class of homogeneous monomial subrings will be introduced and studied here.
Examples that fall into this class are Rees algebras of ideals generated by mono-
mials of the same degree, subrings generated by monomials of the same degree,
and polytopal subrings. To avoid repetitions we continue using the notation of
Section 4.3.
DEFINITION 4.4.1. The monomial subring K[F] is said to be homogeneous if
there is x0 ∈ ℚn such that ⟨vi , x0 ⟩ = 1 for i = 1, . . . , q.
PROPOSITION 4.4.2. K[F] is a homogeneous subring if and only if K[F] is a
standard graded algebra with the grading
∑
K[F]i = K(xv1 )c1 ⋅ ⋅ ⋅ (xvq )cq , where ∣c∣ = c1 + ⋅ ⋅ ⋅ + cq .
∣c∣=i

PROOF. ⇒) Let x0 ∈ ℚn such that ⟨vi , x0 ⟩ = 1 for all i. It suffices to prove

K[F] = K ⊕ K[F]1 ⊕ K[F]2 ⊕ ⋅ ⋅ ⋅ ⊕ K[F]i ⊕ ⋅ ⋅ ⋅
because one clearly has K[F]i K[F] j ⊂ K[F]i+ j for all i, j. First let us prove that
K[F]i ∩ K[F] j = {0} for i ∕= j. If this intersection is not zero, one has an equality
(xv1 )a1 ⋅ ⋅ ⋅ (xvq )aq = (xv1 )c1 ⋅ ⋅ ⋅ (xvq )cq ,
with ∣a∣ = i and ∣c∣ = j. Hence
i = ∣a∣ = ⟨a1 v1 + ⋅ ⋅ ⋅ + aq vq , x0 ⟩ = ⟨c1 v1 + ⋅ ⋅ ⋅ + cq vq , x0 ⟩ = ∣c∣ = j,
a contradiction. To finish assume fi1 + ⋅ ⋅ ⋅ + fir = 0, with fi j ∈ K[F]i j and
i1 < ⋅ ⋅ ⋅ < ir . If fir ∕= 0, using that the monomials of R form a K-basis, one
has K[F]ir ∩ K[F]i j ∕= {0} for some j < r, a contradiction. Thus fir = 0. By
induction all the fi j must be zero, as required.
⇐) Set V = aff ℚ (v1 , . . . , vq ) and r = dim(V). To begin with we claim that 0 is
∑q V. Otherwise if 0 ∈ V, write 0 = a1 v1 + ⋅ ⋅ ⋅ + aq vq with ai ∈ ℚ for all i and
not in
i=1 ai = 1. There is 0 ∕= p ∈ ℤ such that pai ∈ ℤ for all i. One may assume, by
reordering the vi , that pai ≥ 0 for i ≤ s and pai ≤ 0 for i > s. As K[F] is graded,
from the equality
(xv1 ) pa1 ⋅ ⋅ ⋅ (xvs ) pas = (xvs+1 )−pas+1 ⋅ ⋅ ⋅ (xvq )−paq
152 4. MONOMIAL RINGS AND OPTIMIZATION PROBLEMS

one derives p(a1 +⋅ ⋅ ⋅+a s ) = − p(a s+1 +⋅ ⋅ ⋅+aq ), that is, p qi=1 ai = 0, a contradiction.
∑
Thus 0 ∈/ V, and consequently r < n. It is well known that a linear variety V in ℚn
of dimension r can be written as an intersection of hyperplanes
n−r
∩
V= H(yi , ci ),
i=1
/ V, one has c j ∕= 0 for some j. To
where 0 ∕= yi ∈ ℚn and ci ∈ ℚ for all i. Since 0 ∈
finish it suffices to set x0 = y j /c j . □
COROLLARY 4.4.3. Let P be the toric ideal of K[F]. Then P is a graded ideal
with respect to the standard grading if and only if K[F] is homogeneous.
PROOF. As P is generated by a finite set of binomials, the result follows rapidly.
□
PROPOSITION 4.4.4. Let t1 , . . . , tq , t be a new set of variables. If ϕ, ψ are the
maps of K-algebras defined by the diagram
ψ ψ
K[t1 , . . . , tq ] / K[F] ti /o o/ /o / xvi
r8 O
ψ r O ϕ
ϕ r O
r 
 r
K[Ft] x vi t
then there is a unique epimorphism ψ such ψ = ψϕ. In addition ψ is an isomorphism
if and only if K[F] and K[Ft] have the same Krull dimension.
PROOF. To show the existence of ψ it suffices to prove ker(ϕ) ⊂ ker(ψ). This
containment follows at once because any binomial in ker(ϕ) clearly belongs to
ker(ψ), and ker(ϕ) is a binomial ideal.
Note that ker(ϕ) and ker(ψ) are both prime ideals. The map ψ is an isomorphism
if and only if ker(ϕ) = ker(ψ). To finish we need only observe that the last equality
holds if and only if K[F] and K[Ft] have the same dimension. □
COROLLARY 4.4.5. The map ψ : K[Ft] → K[F] is an isomorphism if and only
if K[F] is a homogeneous subring.
PROOF. If ψ is an isomorphism, then by the proof above ker(ψ) = ker(ϕ) and
K[F] is homogeneous because ker(ϕ) is homogeneous with the standard grading of
K[t1 , . . . , tq ].
Conversely assume that ker(ψ) is homogeneous relative to the standard grading.
Note that any homogeneous binomial in ker(ψ) is also in ker(ϕ), thus one has
ker(ψ) = ker(ϕ) and ψ is an isomorphism. □
PROPOSITION 4.4.6. [187, Proposition 7.7.21] If S = K[F] and I is the ideal
I = (S : S S ) = {a ∈ S ∣ aS ⊂ S },
 4. MONOMIAL RINGS AND OPTIMIZATION PROBLEMS 153

then there is a monomial xγ ∈ I. In particular I ∕= 0.

PROPOSITION 4.4.7. If S = K[F] is homogeneous, then S is a positively graded
S -module and its multiplicity is equal to the multiplicity of S .
PROOF. Set P = conv(𝒜). Recall that S is graded as in Proposition 4.4.2. As S
is homogeneous, it is seen that
S = S0 ⊕ S1 ⊕ S2 ⊕ ⋅⋅⋅ ⊕ Si ⊕ ⋅⋅⋅
is an ℕ-graded S -module whose ith component is given by
K xα .
∑
Si =
α∈ℤ𝒜∩iP

Indeed one can readily verify that S i S j ⊂ S i+ j . Let h and E be the Hilbert functions
of S and S respectively. Since S and S have the same dimension d, one can write
h(i) = a0 id−1 + terms of lower degree,
E(i) = c0 id−1 + terms of lower degree,
for i ≫ 0. One clearly has a0 ≤ c0 because S i ⊂ S i for all i. On the other hand by
Proposition 4.4.6 there is xγ ∈ S of degree m such that xγ S ⊂ S . Thus for each i
there is a one to one map
xγ
S i −→ S i+m .
Hence E(i) ≤ h(i + m) for all i, and consequently c0 ≤ a0 . Altogether a0 = c0 .
Therefore
e(S ) = a0 (d − 1)! = c0 (d − 1)! = e(S ). □

Exercises
4.4.8. If K[F] is a homogeneous monomial subring over a field K, use the
Danilov-Stanley formula to show that a(K[F]) < 0.
4.4.9. Let F = { xv1 , . . . , xvq } be a set of monomials of R. Prove that K[F] is
homogeneous if and only if 0 ∈ / V = aff ℚ (v1 , . . . , vq ) ⊂ ℚn .
4.4.10. Is the integral closure of a homogeneous subring homogeneous?
4.4.11. Is the monomial subring K[x2 , xy, y3 ] ⊂ K[x, y] homogeneous?

4.5. Ehrhart and homogeneous subrings

Throughout we shall use the following symbology and terminology which is con-
sistent with the notation introduced thus far:
154 4. MONOMIAL RINGS AND OPTIMIZATION PROBLEMS

K; k a field
R polynomial ring K[x1 , . . . , xn ]
𝒜 𝒜 = { v1 , . . . , vq } ⊂ ℕ n
F F = { x v1 , . . . , x vq }
I ideal generated by F
A n × q matrix with column vectors v1 , . . . , vq
𝒫 := conv(𝒜) ⊂ ℝn lattice polytope
R[t] the polynomial ring K[x1 , . . . , xn , t]
A(𝒫 ) Ehrhart ring K[{ xa ti ∣ a ∈ ℤn ∩ i𝒫 , i ∈ ℕ}]
K[𝒫 ] polytopal subring K[{ xα t∣ α ∈ 𝒫 ∩ ℤn }] ⊂ R[t]
R[It] Rees algebra of I: R[{ xv1 t, . . . , xvq t}]
K[F] monomial subring K[xv1 , . . . , xvq ]
K[Ft] monomial subring K[xv1 t, . . . , xvq t]
K[Ft] integral closure of K[xv1 t, . . . , xvq t].

Following [64, 156] we shall be interested in comparing these monomial subrings

and their algebraic properties. We are also interested in comparing their algebraic
invariants. One has the following inclusions:
K[Ft] ⊂ K[𝒫 ] ⊂ A(𝒫 ) ⊂ R[It],
K[Ft] ⊂ A(𝒫 ).
It will turn out that the last inclusion is an equality if and only if the group
ℤn /ℤ{vi − v1 }qi=2 is torsion free. We will show that the multiplicity of an Ehrhart
ring A(𝒫 ) equals the normalized volume of 𝒫 and present the reciprocity law of
Ehrhart.
Basic properties of Ehrhart rings.
PROPOSITION 4.5.1. dim K[𝒫 ] = dim(𝒫 ) + 1.
PROOF. First we observe the equality
aff(𝒫 ) = aff(v1 , . . . , vq ) = v1 + V,
where V = ℝ(v2 − v1 ) + ⋅ ⋅ ⋅ + ℝ(vq − v1 ). Thus one has dim(𝒫 ) = dimℝ (V). On the
other hand using the identity
⎛ ⎞ ⎛ ⎞
v1 1 v2 − v1 0
rank ⎝ ... ... ⎠ = rank ⎝ .. .. ⎟ + 1,
⎜ ⎟ ⎜
. . ⎠
vq 1 vq − v1 0
the asserted formula follows from Proposition 4.3.20. □
PROPOSITION 4.5.2. The following assertions hold:
(a) A(𝒫 ) is a finitely generated K-algebra and a normal domain,
 4. MONOMIAL RINGS AND OPTIMIZATION PROBLEMS 155

(b) K[Ft] ⊂ A(𝒫 ) is a finitely generated integral extension,

(c) K[Ft] = A(𝒫 ) ⇔ A(𝒫 ) is contained in the field of fractions of K[Ft],
(d) dim(K[Ft]) = dim(K[Ft]) = dim(A(𝒫 )) = dim(𝒫 ) + 1, and
(e) if K[F] is homogeneous, then dim(K[F]) = dim(K[Ft]).
PROOF. First let us show that the Ehrhart ring can be expressed as
(4.28) A(𝒫 ) = K[{ xα ti ∣ (α, i) ∈ ℤn+1 ∩ ℝ+ Γ}],
where Γ is the finite set {(α, 1)∣ α ∈ ℤn ∩ 𝒫} and ℝ+ Γ is the polyhedral cone
consisting of all linear combinations of Γ with non-negative coefficients. Take a
monomial xα ti in A(𝒫 ), then α ∈ ℤn ∩ i𝒫 and i ∈ ℕ. Thus we can write
(α, i) = iλ1 (v1 , 1) + ⋅ ⋅ ⋅ + iλq (vq , 1),
∑q
with i=1 λi = 1 and λ j ≥ 0 for all j. This proves that (α, i) is in ℤn+1 ∩ ℝ+ Γ.
Hence xα ti is in the right hand side of Eq. (4.28). For the reverse containment take
a monomial xα ti with (α, i) in ℤn+1 ∩ ℝ+ Γ. Note that any vector of the form (β, 1)
with β ∈ ℤn ∩ 𝒫 can be written as
(β, 1) = µ1 (v1 , 1) + ⋅ ⋅ ⋅ + µq (vq , 1)
for some non-negative real numbers µ1 , . . . , µq . Therefore the vector (α, i) can be
written as
(α, i) = ρ1 (v1 , 1) + ⋅ ⋅ ⋅ + ρq (vq , 1) (ρi ∈ ℝ+ ).
From this equation we obtain α ∈ ℤn ∩ i𝒫 . Hence xα ti ∈ A(𝒫 ). This completes the
proof of Eq. (4.28).
By Gordan’s Lemma (see Lemma 3.3.2), the monoid ℤn+1 ∩ ℝ+ Γ is finitely
generated, that is, there is a finite set ℬ = {z1 , . . . , zr } such that
(4.29) ℤn+1 ∩ ℝ+ Γ = ℕz1 + ⋅ ⋅ ⋅ + ℕzr = ℕℬ .
To prove (a) note that from Eq. (4.28) and Eq. (4.29) we get
(4.30) A(𝒫 ) = K[{ xα ti ∣ (α, i) ∈ ℕℬ}].
This proves that A(𝒫 ) is generated as a K-algebra by the monomials of R[t] corre-
sponding to the vectors z1 , . . . , zr . To see that A(𝒫 ) is normal it suffices to show
the equality ℕℬ = ℝ+ ℬ ∩ ℤℬ . Since the left hand side is always contained in the
right hand side and since ℝ+ ℬ ∩ ℤℬ = ℚ+ ℬ ∩ ℤℬ (see Corollary 3.1.16), we need
only show the containment ℚ+ ℬ ∩ ℤℬ ⊂ ℕℬ . Take an arbitrary β in ℚ+ ℬ ∩ ℤℬ and
write it as
β = λ1 z1 + ⋅ ⋅ ⋅ + λr zr (λi ∈ ℚ+ ).
There exists 0 ∕= m ∈ ℕ satisfying mλi ∈ ℕ for all i. Thus multiplying this
equality by m gives mβ ∈ ℕℬ and, by Eq. (4.29), we get mβ ∈ ℤn+1 ∩ ℝ+ Γ. Hence
β ∈ ℤn+1 ∩ ℝ+ Γ and using again Eq. (4.29) we conclude β ∈ ℕℬ , as required.
156 4. MONOMIAL RINGS AND OPTIMIZATION PROBLEMS

To prove (b) note that given xa ti in A(𝒫 ), then (xa ti ) s ∈ K[Ft] for some
0 ∕= s ∈ ℕ. Hence A(𝒫 ) is integral over K[Ft]. As a result, using (a) together
with Corollary 4.2.4, we obtain that A(𝒫 ) is a finitely generated K[Ft]-module.
Assertion (c) follows directly from (a) and (b). The first two equalities in (d) follow
from Proposition 4.2.5 because
K[Ft] ⊂ K[Ft] ⊂ A(𝒫 )
are integral extensions. To prove the third equality in (d) apply Proposition 4.5.1.
Finally (e) follows at once from Corollary 4.4.5. □
Multiplicity of monomial subrings. Let M = ⊕∞
i=0 Mi be a finitely generated
graded module over a standard graded K-algebra
S = K[y1 , . . . , yn ] = ⊕∞
i=0 S i .

This means that the Mi ’s (resp. S i ’s) are subgroups of M (resp. S ) such that
S i M j ⊂ Mi+ j for all i, j. Note S 0 = K. The term standard means that y1 , . . . , yn
have degree one, i.e., S is generated as a K-algebra by S 1 . Recall that the Hilbert
function of M is defined as
ϕ M (i) := dimK (Mi ).
An excellent exposition of the theory of Hilbert functions of graded modules and
their relation to dimension theory and multiplicities is given in the classical book
of H. Matsumura [132, Chapter 5]. For convenience we recall that, according to
Corollary 1.6.5, ϕ M is a polynomial function with rational coefficients:
ad d
ϕ M (i) = i + (terms of lower degree) (i ≫ 0)
d!
such that dim(M) = d + 1 and 0 ∕= ad ∈ ℕ. .
DEFINITION 4.5.3. The multiplicity of M, denoted by e(M), is given by

e(M) := ad = d! lim ϕ M (i)/id .

i→∞

REMARK 4.5.4. In general the subrings K[Ft], K[Ft] and A(𝒫 ) have a well
defined multiplicity. Below we clarify this assertion.
Notice that the subring K[Ft] is always homogeneous because (vi , 1) lies in the
hyperplane xn+1 = 1 for all i, thus its multiplicity exists. The multiplicity of K[Ft]
is well defined because K[Ft] is graded by
∑
(4.31) K[Ft]i = K xa t b ,
(a,b)∈ℤ𝒜′ ∩i𝒫 ′
 4. MONOMIAL RINGS AND OPTIMIZATION PROBLEMS 157

where ℬ = {(v1 , 1), . . . , (vq , 1)} and 𝒫 ′ = conv(ℬ ). If K[F] is a homogeneous

subring, let x0 be a rational vector satisfying ⟨ x0 , vi ⟩ = 1 for all i, then by Proposi-
tion 4.4.2 K[F] = ⊕i∈ℕ K[F]i is a standard graded K-algebra graded by
∑
K[F]i = K { xa ∣ xa ∈ K[F]}.
⟨a,x0 ⟩=i

Thus e(K[F]) is well defined in this case. If K[F] is homogeneous, then there is an
isomorphism of K-algebras:
ϕ ϕ
K[F] = K[{ xv1 , . . . , xvq }] ≃ K[Ft], xvi 7−→ xvi t,
see Corollary 4.4.5. In particular e(K[F]) = e(K[Ft]), because ϕ is degree preserv-
ing, i.e., ϕ maps K[F]i into K[Ft]i for all i ∈ ℕ.
The Ehrhart ring A(𝒫 ) is a normal finitely generated K-algebra, and it is a
graded K-algebra with ith component given by

K xα t i .
∑
A(𝒫 )i =
α∈ℤn ∩i𝒫

Since A(𝒫 ) is a finitely generated module over the homogeneous ring K[Ft], see
Proposition 4.5.2(b), from Theorem 1.6.4 we obtain that its Hilbert function:
E(i) = ∣ℤn ∩ i𝒫∣ = cd id + ⋅ ⋅ ⋅ + c1 i + c0 (i ≫ 0)
is a polynomial function of degree d = dim(𝒫 ) such that ci ∈ ℚ for all i and d!cd is
an integer, which is the multiplicity of A(𝒫 ). Notice that
(4.32) e(A(𝒫 )) = d!cd = d!vol(𝒫 ).
In this case the Hilbert polynomial of A(𝒫 ):
E𝒫 (x) = cd xd + cd−1 xd−1 + ⋅ ⋅ ⋅ + c1 x + c0
is called the Ehrhart polynomial of 𝒫 . Recall that the relative volume of 𝒫 is given
by
∣ℤn ∩ i𝒫∣
vol(𝒫 ) = lim ,
i→∞ id
where d = dim(𝒫 ). Hence vol(𝒫 ) is the leading coefficient of the Ehrhart polynomial
of 𝒫 . For this reason d!cd is called the normalized volume of 𝒫 .
The Hilbert series of A(𝒫 ) is:
∞
∑
F(A(𝒫 ), x) = ∣ℤn ∩ i𝒫∣ xi ,
i=0
158 4. MONOMIAL RINGS AND OPTIMIZATION PROBLEMS

this series is called the Ehrhart series of 𝒫 . Using the Hilbert-Serre theorem (see
Theorem 1.6.2) it follows that this series is a rational function that can be uniquely
written as:
h0 + h1 x + ⋅ ⋅ ⋅ + h s x s
F(A(𝒫 ), x) = ,
(1 − x)d+1
with h0 + h1 + ⋅ ⋅ ⋅ + h s ∕= 0, hi ∈ ℤ for all i and d = dim(𝒫 ). Some well known
properties are:
∙ a(A(𝒫 )) = s − (d + 1) < 0; because of the Danilov-Stanley formula.
∙ E𝒫 (i) = E(i) for all integers i ≥ 0; because of Corollary 1.6.8.
∙ hi ≥ 0 for all i; because A(𝒫 ) is Cohen-Macaulay, see Exercise 1.6.33.
∙ h0 + h1 + ⋅ ⋅ ⋅ + h s = d!vol(𝒫 ); see Remark 1.6.7.
∙ In general some of ci ’s may be negative. It is unknown whether ci ≥ 0 for
all i if the vertices of 𝒫 have {0, 1}-entries.
Consider the function E + (i) = ∣ℤn ∩ ri(i𝒫 )∣. The following is a fundamental
result in the theory of integral polytopes.
THEOREM 4.5.5. (Reciprocity law of Ehrhart [55])
E + (i) = (−1)d E𝒫 (−i) ∀ i ≥ 1.
EXAMPLE 4.5.6. To illustrate the reciprocity law consider the polytope 𝒫 :

y
6
6
5 ppr
p pppppppppppp
J p ppppppppppppppppppppppp
ppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppp

p
4
p p p
rpppppppppppppppppppppppppppppppppppprppppppppppppppppppppppppppppppppppppprppppppppppppppppppppppppp
 pppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppp
J ∣ℤ2 ∩ 𝒫∣ = 16
pppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppp
3
ppppppppppppppppppppppppppppppppppppppppppppppp
rppppppppppppppppppppppppppppppppppppprpppppppppppppppppppppppppppppppppppppprppppppppppppppppppppppppppppppppppppprpppppppppppp
J
ppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppp
∣ℤ2 ∩ ∂𝒫∣ = 6
pppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppp
J
ppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppp
2 rpppppppppppppppppppppppppppppppppppppprppppppppppppppppppppppppppppppppppppprpppppppppppppppppppppppppppppppppppppprppppppppppppppppppppppppppppppp r
J
pppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppp
E𝒫 (x) = 12x2 + 3x + 1
ppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppp
ppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppp

E + (1) = (−1)2 E𝒫 (−1) = 10
AA  ppppppppppppppppprpppppppppppppppppppppppppppppprpppppppppppppppp
1  pppppppppppppppppppppppppppppppppp
pppppppppppppppppppppppppppp
A pppppppppppppppppp
 ppppppppp p

A r -
0 1 2 3 4 5 6 x

PROPOSITION 4.5.7. Let F1 , . . . , F s be the facets of 𝒫 . Then c0 = 1 and

s
1∑
cd−1 = vol(Fi ).
2
i=1

PROOF. The assertion that c0 = 1 follows from E𝒫 (0) = E(0) = 1. Notice the
equalities
i𝒫 = ri(i𝒫 ) ∪ ∂(i𝒫 ); ∂(𝒫 ) = F1 ∪ ⋅ ⋅ ⋅ ∪ F s ,
 4. MONOMIAL RINGS AND OPTIMIZATION PROBLEMS 159

where ∂(𝒫 ) is the relative boundary. Using the reciprocity law and the inclusion-
exclusion principle [1, p. 38, Formula 2.12] we get
∣∂(i𝒫 ) ∩ ℤn ∣ = E𝒫 (i) − E + (i)
= E𝒫 (i) − (−1)d E𝒫 (−i)
= 2cd−1 id−1 + terms of lower degree
= E F1 (i) + ⋅ ⋅ ⋅ + E F s (i) + h(i),
where h is a polynomial of degree at most d − 2. Comparing leading terms, the
required formula follows. □
THEOREM 4.5.8. If 𝒫 ⊂ ℝ2 and d = 2, then
∣ℤ2 ∩ ∂𝒫∣
E𝒫 (x) = area(𝒫 )x2 + x+1
2
PROOF. From the decomposition P = ri(𝒫 ) ∪ ∂(𝒫 ) and the reciprocity law we
obtain
∣∂(𝒫 ) ∩ ℤ2 ∣ = ∣𝒫 ∩ ℤ2 ∣ − ∣ri(𝒫 ) ∩ ℤ2 ∣ = E𝒫 (1) − E + (1)
= E𝒫 (1) − E𝒫 (−1)
= (c2 + c1 + c0 ) − (c2 − c1 + c0 ) = 2c1 . □
COROLLARY 4.5.9 (Pick’s Formula). If 𝒫 ⊂ ℝ2 and d = 2, then
∣ℤ2 ∩ ∂𝒫∣
∣ℤ2 ∩ 𝒫∣ = area(𝒫 ) + +1
2
PROOF. It follows making x = 1 in the formula of Theorem 4.5.8. □
COROLLARY 4.5.10. If v1 , . . . , vq have {0, 1}-entries, then E𝒫 (−1) = 0.
PROOF. By Exercise 4.5.27 and the reciprocity law:
0 = E + (1) = (−1)d E𝒫 (−1). □
COROLLARY 4.5.11. If xv1 , . . . , xvq are square-free, then K[Ft] = K[𝒫 ].
PROOF. It follows from Exercise 4.5.27. □
PROPOSITION 4.5.12. The multiplicities of A(𝒫 ) and K[Ft] are related by
e(A(𝒫 )) = ∣T (ℤn /(v2 − v1 , . . . , vq − v1 ))∣ e(K[Ft]).
PROOF. Let ℬ = {(v1 , 1), . . . , (vq , 1)} and consider 𝒫 ′ = conv(ℬ ). Note that
there is a bijective map
ℤn ∩ i𝒫 −→ ℤn+1 ∩ i𝒫 ′ , α 7−→ (α, i),
160 4. MONOMIAL RINGS AND OPTIMIZATION PROBLEMS

and d = dim(𝒫 ) = dim(𝒫 ′ ). Thus vol(𝒫 ) = vol(𝒫 ′ ). Since ℬ lies in the affine
hyperplane xn+1 = 1, applying Theorem 3.2.12 gives
∣ℤℬ ∩ i𝒫 ′ ∣
(4.33) vol(𝒫 ) = ∣T (ℤn+1 /((v2 − v1 , 0), . . . , (vq − v1 , 0)))∣ lim .
i→∞ id
On the other hand, one has:
∣ℤℬ ∩ i𝒫 ′ ∣
(4.34) e(K[Ft]) = e(K[Ft]) = d! lim .
i→∞ id
The first equality holds because of Proposition 4.4.7, while the second equality
follows by definition of multiplicity noticing that K[Ft] is graded by
∑
(4.35) K[Ft]i = K xa t b
(a,b)∈ℤℬ∩i𝒫 ′
and using Proposition 4.5.2 (d). Hence the asserted equality follows from Eq. (4.33),
Eq. (4.34), and observing that the map
T (ℤn /(v2 − v1 , . . . , vq − v1 )) −→ T (ℤn+1 /((v2 − v1 , 0), . . . , (vq − v1 , 0)))
given by v 7−→ (v, 0) is bijective. □
Notation If A is an integral matrix, ∆i (A) will denote the greatest common divisor
of all the nonzero i × i minors of A.
COROLLARY 4.5.13. e(A(𝒫 )) = ∆r (B)e(K[Ft]), where B is the matrix with
column vectors (v1 , 1), . . . , (vq , 1) and r = rank (B).
PROOF. It follows from Lemma 3.2.18 and Proposition 4.5.12. □
THEOREM 4.5.14 ([64]). If the matrix
( )
v1 ⋅ ⋅ ⋅ vq
B= ,
1 ⋅⋅⋅ 1
has rank r, then ∆r (B) = 1 if and only if K[Ft] = A(𝒫 ).
PROOF. ⇒) Since K[Ft] ⊂ A(𝒫 ) is an integral extension of rings and A(𝒫 )
is normal one has K[Ft] ⊂ A(𝒫 ). For the other containment it suffices to prove
that A(𝒫 ) is contained in the field of fractions of K[Ft]. Let xα ti ∈ A(𝒫 )i , that is,
α ∈ ℤn ∩ i𝒫 and i ∈ ℕ. Hence the system of equations
α
( )
(4.36) By = = α′
i
has a real solution. Hence by Gaussian elimination the system has a rational solution
y. Here we regard α as a column vector. Thus the augmented matrix [B α′ ] has rank
r. In general ∆r ([B α′ ]) divides ∆r (B), so in this case they are equal. By Hegers’s
theorem (see Theorem 3.5.2) the linear system (4.36) has an integral solution.
Therefore xα ti is in the field of fractions of K[Ft], as required.
 4. MONOMIAL RINGS AND OPTIMIZATION PROBLEMS 161

⇐) By hypothesis e(K[Ft]) = vol(𝒫 )(r − 1)!. Thus from Proposition 4.5.12 we

get that ℤn /(v2 − v1 , . . . , vq − v1 ) is torsion-free. Hence by Lemma 3.2.18 the group
ℤn+1 /((v1 , 1), . . . , (vq , 1)) is also torsion-free, as required. □
COROLLARY 4.5.15. If the subgroup L = ({vi − v j ∣ i, j = 1, . . . , q}) of ℤn has
rank n, then K[Ft] = A(𝒫 ) if and only if L = ℤn .
PROOF. It follows using Lemma 3.2.18. □
COROLLARY 4.5.16. If ∆r (B) = 1, then K[Ft] is a normal domain if and only if
a(K[Ft]) < 0 and the Hilbert polynomial of K[Ft] is equal to the Ehrhart polynomial
of 𝒫 .
PROOF. ⇒) It follows from Theorems 4.3.27 and 4.5.14.
⇐) Since the a-invariants of K[Ft] and A(𝒫 ) are both negative, then
dimK K[Ft]i = dimK A(𝒫 )i (∀ i ≥ 0).
Hence noticing K[Ft]i ⊂ A(𝒫 )i , we get K[Ft]i = A(𝒫 )i for all i ≥ 0. Therefore
K[Ft] = A(𝒫 ) and consequently K[Ft] is normal. □
PROPOSITION 4.5.17. If 𝒜 lies on a hyperplane not containing the origin and 𝒫
has a weakly unimodular covering with support in 𝒜, then K[F] is normal.
PROOF. We leave to the reader to infer this result by adapting the proof of
Theorem 3.5.7, and by noticing that a set of vectors lying on an affine hyperplane not
containing the origin is affinely independent if and only if it is linearly independent.
□
THEOREM 4.5.18. If 𝒜 = {v1 , . . . , vq } ⊂ ℕn lie on an affine hyperplane not
containing the origin and 𝒫 = conv(𝒜) has a unimodular covering with support in
𝒜, then
K[F] ≃ K[Ft] = A(𝒫 ).
PROOF. Let ∆ = conv(vi1 , . . . , vid+1 ) be any simplex in the unimodular covering
of 𝒫 , where d = dim(𝒫 ). Note that
ℤn+1 /((vi1 , 1), . . . , (vid+1 , 1))
is torsion-free by Proposition 3.2.19. Therefore by Proposition 4.5.14 we obtain the
equality K[Ft] = A(𝒫 ). To finish the proof note that K[F] is normal because 𝒫 has
a weakly unimodular covering (see Proposition 3.5.10) and K[F] ≃ K[Ft] because
K[F] is homogeneous. □
REMARK 4.5.19. It is not hard to find examples where Theorem 4.5.18 fails if
we replace unimodular covering by weakly unimodular covering.
PROPOSITION 4.5.20 ([64]). If A is a unimodular matrix and the vectors v1 , . . . , vq
lie on a hyperplane not containing the origin, then K[Ft] = A(𝒫 ).
162 4. MONOMIAL RINGS AND OPTIMIZATION PROBLEMS

PROOF. There is an epimorphism of K-algebras ψ : K[Ft] −→ K[F] induced

by xvi t7−→ xvi , see Proposition 4.4.4. By hypothesis
( )
A
r = rank(A) = rank .
1⋅⋅⋅1
Hence ψ is an isomorphism. Using Corollary 4.3.25, one obtains that K[Ft] is
normal. As ∆r (A) = 1, applying Proposition 4.5.14 yields the asserted equality. □
The next result is a generalization of [156, Theorem 7.1].
THEOREM 4.5.21 ([64]). If the Rees algebra R[It] is a normal domain and
v1 , . . . , vq lie on a hyperplane
b1 x1 + ⋅ ⋅ ⋅ + bn xn = 1 (bi > 0 ∀i),
then K[Ft] = A(𝒫 ).
PROOF. Let xα ti ∈ A(𝒫 ), where α ∈ ℤn ∩ i𝒫 and i ≥ 1. One can write
(α, i) = iλ1 (v1 , 1) + ⋅ ⋅ ⋅ + iλq (vq , 1) ( i λi = 1, λi ≥ 0 ).
∑
(4.37)
Hence (α, i) ∈ ℤ𝒜′ ∩ ℝ+ 𝒜′ = ℤn+1 ∩ ℝ+ 𝒜′ , where
𝒜′ = {(e1 , 0), . . . , (en , 0), (v1 , 1), . . . , (vq , 1)}
and ei is the ith unit vector in ℝn . By the normality of R[It] one has (α, i) ∈ ℕ𝒜′ .
Thus one can write
(4.38) (α, i) = m1 (e1 , 0) + ⋅ ⋅ ⋅ + mn (en , 0) + η1 (v1 , 1) + ⋅ ⋅ ⋅ + ηq (vq , 1),
where mi and η j are in ℕ for all i, j. From Eqs. (4.37) and (4.38):
α = m1 e1 + ⋅ ⋅ ⋅ + mn en + η1 v1 + ⋅ ⋅ ⋅ + ηq vq
= iλ1 v1 + ⋅ ⋅ ⋅ + iλq vq ,
i = η1 + ⋅ ⋅ ⋅ + ηq .
Taking the inner product of α with b = (b1 , . . . , bn ) yields
⟨α, b⟩ = m1 b1 + ⋅ ⋅ ⋅ + mn bn + i,
⟨α, b⟩ = iλ1 ⟨v1 , b⟩ + ⋅ ⋅ ⋅ + iλq ⟨vq , b⟩ = i,
From the equalities above one derives mi = 0 for all i. Since
xα ti = (xv1 t)n1 ⋅ ⋅ ⋅ (xvq t)nq ,
we get xα ti ∈ K[Ft], as required. □
COROLLARY 4.5.22. If xv1 , . . . , xvq are monomials of degree r and the Rees
algebra R[It] is normal, then the multiplicity of K[F] is given by
e(K[F]) = d!vol(𝒫 ), d = dim(𝒫 ).
 4. MONOMIAL RINGS AND OPTIMIZATION PROBLEMS 163

PROOF. Notice that ⟨vi , (1/r, . . . , 1/r)⟩ = 1 for all i. Thus by Corollary 4.4.5
K[F] ≃ K[Ft], as graded K-algebras. Since e(A(𝒫 )) = d!vol(𝒫 ), using Theo-
rem 4.5.21 the result follows. □
As an application we prove that the multiplicity of a standard graded normal
Rees algebra is always a normalized volume:
PROPOSITION 4.5.23 ([63]). If xv1 , . . . , xvq are monomials of degree r ≥ 2 and
R[It] is normal, then
e(R[It]) = n!vol(𝒫 ′ ),
where 𝒫 ′ = conv(𝒜′ ) and 𝒜′ = {(v1 , 1), . . . , (vq , 1), e1 , . . . , en }.
PROOF. Notice that the vectors in 𝒜′ lie in the hyperplane:
x1 + ⋅ ⋅ ⋅ + xn − (r − 1)xn+1 = 1.
This means that R[It] is a standard graded K-algebra and dim(𝒫 ′ ) = n. Thus
by Theorem 4.5.14 we get R[It] ≃ A(𝒫 ′ ), as graded K-algebras. According to
Eq. (4.32) the multiplicity of the right hand side is n!vol(𝒫 ′ ), as desired. □
PROPOSITION 4.5.24 ([158]). If x0 ∈ ℤn is a vector with positive entries such
that ⟨ x0 , vi ⟩ = r, i = 1, . . . , q, for some r ≥ 2 and R[It] is normal, then the following
holds:
(a) The torsion subgroup of ℤn /ℤ𝒜 is a cyclic group of order dividing r.
(b) If ℤn /ℤ𝒜 is a finite group and x0 = (1, . . . , 1), then ℤn /ℤ𝒜 ≃ ℤr .
PROOF. Set ℬ = {(v1 , 1), . . . , (vq , 1)}. There is an exact sequence of finite
groups
ϕ ψ
0 −→ T (ℤn+1 /ℤℬ ) −→ T (ℤn /ℤ𝒜) −→ ℤr , (∗)
where the maps ϕ and ψ are given by
ϕ((α, b)) = α (α ∈ ℤn , b ∈ ℤ),
ψ(α) = ⟨α, x0 ⟩.
Consider the matrix B whose columns are the vectors in the set ℬ . By Theo-
rems 4.5.21 and 4.5.14 we get ∆ s (B) = 1, where s is the rank of ℤℬ . Since ∆ s (B)
is the order of T (ℤn+1 /ℤℬ ), part (a) follows using the sequence (∗). To prove (b),
note that ℤn /ℤ𝒜 is a torsion group and the ith unit vector ei maps into the element
1 under the map ψ. Hence ψ is onto, and ψ is the required isomorphism. □
Below xr will denote the set of all monomials of degree r in R. Then K[xr ] is
the rth Veronese subring R(r) of R.
DEFINITION 4.5.25. An extension K[F] ⊂ R(r) is said to be birational if K[F]
and R(r) have the same field of fractions.
164 4. MONOMIAL RINGS AND OPTIMIZATION PROBLEMS

PROPOSITION 4.5.26 ([158]). If F ⊂ xr , then K[F] ⊂ K[xr ] is a birational

extension if and only if n = rank(A) and ∆n (A) = r.
PROOF. Assume the given extension is birational. Then ℤlog(xr ) = ℤ𝒜 by
Exercise 4.5.37. By Proposition 4.5.24 (b), or by direct observation, ℤn /ℤlog(xr ) ≃
ℤr . Therefore, ∆n (A) = r.
Conversely, if ∆n (A) = r, the surjection
ℤn /ℤ𝒜 −→ ℤr , a7−→∣a∣
is an isomorphism. Since log(xr ) maps to zero under this map, one obtains ℤ𝒜 =
ℤlog(xr ), as required. □

Exercises
4.5.27. If v1 , . . . , vq ∈ ℕn ∖ {0} are vectors with {0, 1}-entries, then
conv(v1 , . . . , vq ) ∩ ℤn = {v1 , . . . , vq }.
4.5.28. If ∆ is a unimodular simplex in ℝn with vertex set 𝒜 = {α0 , . . . , αn },
then ∆ ∩ ℤn = 𝒜.
4.5.29. Let K be a field and let S be a K-algebra. We say that S is an affine
K-algebra if S = K[y1 , . . . , yr ] for some y1 , . . . , yr ∈ S . If S = K[y1 , . . . , yr ] is an
affine K-algebra of dimension r, then S is a polynomial ring.
4.5.30. Let K be a field and let 𝒫 be a unimodular simplex in ℝn with vertex set
{α0 , . . . , αn } in ℕn . Then K[𝒫 ] is a polynomial ring and
K[𝒫 ] = K[xα0 t, . . . , xαn t].
4.5.31. If 𝒫 is a unimodular lattice simplex in ℝn with vertices α0 , . . . , αn ,
prove that the polytopal subring K[𝒫 ] is not isomorphic (as a ring) to the monomial
subring K[xα0 , . . . , xαn ].
4.5.32. Let F = { xv1 , . . . , xvq } be a set of square-free monomials of degree r of
R. If 𝒫 = conv(v1 , . . . , vq ), then K[𝒫 ] ≃ K[F] as graded K-algebras.
4.5.33. Prove that A(𝒫 ) = ∞
⊕
i=0 A(𝒫 )i is a graded K-algebra with its ith com-
ponent given by
K xα t i .
∑
A(𝒫 )i =
α∈ℤn ∩i𝒫

4.5.34. If 𝒫 = conv({(0, 0, 0), (12, 1, 0), (0, 1, 1), (1, 0, 1)}), use Normaliz [24,
25] to show that E𝒫 (x) = 13/6x3 + x2 − 1/6x + 1.
 4. MONOMIAL RINGS AND OPTIMIZATION PROBLEMS 165

4.5.35. Let 𝒫 = conv((0, 0), (0, 1), (1, 0), (1, 1)) be the unit square. In the figure
below we show the integral points of 4𝒫 . Verify the following:

6 ∣ℤ2 ∩ 4𝒫∣ = 25
4 s s s s s E𝒫 (x) = (x + 1)2
3 s s s s s vol(𝒫 ) = 1
2 s s s s s A(𝒫 ) = K[Ft]
1 s s s s s F(A(𝒫 ), x) = (1 + x)/(1 − x)3
s s s s s -
0 1 2 3 4

4.5.36. Let A be an integral matrix of order n × q with equal column sum r.

Prove the equalities
( ) ( ) ( )
A A A
In (A) = In = In = rIn ,
A1 + ⋅ ⋅ ⋅ + An r1 1
where A1 , . . . , An are the rows of A and In (A) is the ideal of ℤ generated by the
n-minors of A.
4.5.37. Let F = { xv1 , . . . , xvq } ⊂ xr . Prove that k[F] ⊂ k[xr ] is a birational
extension if and only if ℤlog(F) = ℤlog(xr ).
4.5.38. Let F be a set of monomials of R = K[x1 , . . . , xn ] of the same degree
r such that ℤn /ℤlog(F) is a finite group. Prove that K[Ft] = A(𝒫 ) if and only if
ℤn /ℤlog(F) ≃ ℤr .
4.5.39. Prove that Theorem 4.5.18 fails if we replace unimodular covering by
weakly unimodular covering.
4.5.40. Let R = K[x1 , . . . , xn ] be a polynomial ring over a field K and let
𝔪 = (x1 , . . . , xn ). Prove that e(K[𝔪d ]), the multiplicity of K[𝔪d ], is equal to dn−1 .

4.6. Rees algebras and optimization problems

Let R = K[x1 , . . . , xn ] be a polynomial ring over a field K and let I be an ideal of R
of height g ≥ 2 minimally generated by a finite set
F = { x v1 , . . . , x vq }
of square-free monomials of degree at least two. For technical reasons we shall
assume that each variable xi occurs in at least one monomial of F.
Recall that a clutter with finite vertex set X is a family of subsets of X, called
edges, none of which is included in another. An important example of clutter
166 4. MONOMIAL RINGS AND OPTIMIZATION PROBLEMS

is a graph. There is a natural one to one correspondence between the family of

square-free monomial ideals and the family of clutters:
DEFINITION 4.6.1. We associate to the ideal I a clutter 𝒞 by taking the set of
indeterminates X = { x1 , . . . , xn } as vertex set and E = {E1 , . . . , Eq } as edge set,
where Ek = supp(xvk ). The ideal I is called the edge ideal or clutter ideal of 𝒞 .
To stress the relationship between I and 𝒞 we will use the notation I = I(𝒞 ). The
{0, 1}-vector vk is called the characteristic vector of Ek .
Let A be the incidence matrix of 𝒞 whose column vectors are v1 , . . . , vq . To
link algebraic properties of monomial algebras with combinatorial optimization
problems we reintroduce the set covering polyhedron of 𝒞 :
Q(A) = { x ∈ ℝn ∣ x ≥ 0; xA ≥ 1}.
This polyhedron was studied in Chapter 3 in order to characterize the max-flow
min-cut property of the clutter matrix A (see Theorem 3.3.16).
DEFINITION 4.6.2. A set C ⊂ X is a minimal vertex cover of the clutter 𝒞 if
every edge of 𝒞 contains at least one vertex in C and C is minimal with respect to
this property.
PROPOSITION 4.6.3 ([80]). The following are equivalent:
(a) 𝔭 = (x1 , . . . , xr ) is a minimal prime of I = I(𝒞 ).
(b) C = { x1 , . . . , xr } is a minimal vertex cover of 𝒞 .
(c) α = e1 + ⋅ ⋅ ⋅ + er is a vertex of Q(A).
PROOF. It is left as an exercise. □
Blowup algebras and Rees cones. The blowup algebras studied here are: (a)
the Rees algebra
R[It] := R ⊕ It ⊕ ⋅ ⋅ ⋅ ⊕ I i ti ⊕ ⋅ ⋅ ⋅ ⊂ R[t],
where t is a new variable, (b) the associated graded ring
grI (R) = R/I ⊕ I/I 2 ⊕ ⋅ ⋅ ⋅ ⊕ I i /I i+1 ⊕ ⋅ ⋅ ⋅
≃ R[It]/IR[It],
with multiplication (a + I i+1 )(b + I j+1 ) = ab + I i+ j+1 (a ∈ I i , b ∈ I j ), and (c) the
symbolic Rees algebra
R s (I) := R + I (1) t + I (2) t2 + ⋅ ⋅ ⋅ + I (i) ti + ⋅ ⋅ ⋅ ⊂ R[t],
where I (i) = 𝔭i1 ∩⋅ ⋅ ⋅∩𝔭is is the ith symbolic power of I and 𝔭1 , . . . , 𝔭 s are the minimal
primes of I.
We have the following inclusions:
K[Ft] ⊂ R[It] ⊂ R[It] ⊂ R s (I).
 4. MONOMIAL RINGS AND OPTIMIZATION PROBLEMS 167

The Rees cone of the point configuration 𝒜 = {v1 , . . . , vq } is the rational

polyhedral cone in ℝn+1 , denoted by ℝ+ 𝒜′ , consisting of the non-negative linear
combinations of the set
𝒜′ := {e1 , . . . , en , (v1 , 1), . . . , (vq , 1)} ⊂ ℝn+1 ,
where ei is the ith unit vector.
Rees cones are important in the study of invariants and algebraic properties of
certain normal Rees algebras (see Chapters 3, 4, 5 and 7) and this is one of the
reasons for the interest in these special types of cones.
By Exercise 4.6.64 the halfspace He+i occurs in the irreducible representation
of ℝ+ 𝒜′ for i = 1, . . . , n + 1. Therefore, using Corollary 3.1.34, there is a unique
irreducible representation
(4.39) ℝ+ 𝒜′ = He+1 ∩ ⋅ ⋅ ⋅ ∩ He+n+1 ∩ Ha+1 ∩ ⋅ ⋅ ⋅ ∩ Ha+r
such that 0 ∕= ai ∈ ℚn+1 and ⟨ai , en+1 ⟩ = −1 for all i.
THEOREM 4.6.4 ([84]). The function ϕ : ℚn → ℚn+1 given by α 7→ (α, −1)
induces a bijective map ϕ : V −→ {a1 , . . . , ar } between the set V of vertices of Q(A)
and the set {a1 , . . . , ar } of normal vectors that occur in the irreducible representation
of ℝ+ 𝒜′ .
PROOF. It is a direct consequence of Theorem 3.3.11. □
Notation Let dk be the unique positive integer such that dk ak has relatively prime
integral entries. We set `k = dk ak for k = 1, . . . , r.
LEMMA 4.6.5 ([65]). If 1 ≤ j ≤ r and ⟨` j , en+1 ⟩ = −1, then 1 ≤ j ≤ s.
PROOF. Let ` j = (a1 , . . . , an , −1). Since ℝ+ 𝒜′ ⊂ H`+j we get ai ≥ 0 for
i = 1, . . . , n. Consider the ideal 𝔭 of R generated by the set of xi such that ai > 0.
Note that I ⊂ 𝔭 because ⟨` j , (vi , 1)⟩ ≥ 0 for all 1 ≤ i ≤ n. For simplicity of notation
assume that ` j = (a1 , . . . , am , 0, . . . , 0, −1), where ai > 0 for all i. Take an arbitrary
vector v ∈ 𝒜′ such that v ∈ H` j , then v satisfies the equation
a1 x1 + ⋅ ⋅ ⋅ + am xm = xn+1 .
Observe that v also satisfies the equation
x1 + ⋅ ⋅ ⋅ + xm = xn+1 .
Using that H` j contains n linearly independent vectors from 𝒜′ we conclude that
H` j = Hb , where b = e1 + ⋅ ⋅ ⋅ + em − en+1 . Hence ` j = b and consequently ai = 1
for all i. It remains to show that 𝔭 is a minimal prime of I, and this follows from the
irreducibility of Eq. (4.40). □
Let 𝔭1 , . . . , 𝔭 s be the minimal primes of the edge ideal I = I(𝒞 ) and let
Ck = { xi ∣ xi ∈ 𝔭k } (k = 1, . . . , s)
168 4. MONOMIAL RINGS AND OPTIMIZATION PROBLEMS

be the corresponding minimal vertex covers of 𝒞 .

COROLLARY 4.6.6 ([65]). The irreducible representation of the Rees cone is
(4.40) ℝ+ 𝒜′ = He+1 ∩ ⋅ ⋅ ⋅ ∩ He+n+1 ∩ H`+1 ∩ ⋅ ⋅ ⋅ ∩ H`+r
where `k = −en+1 + xi ∈Ck ei for 1 ≤ k ≤ s. Moreover all the vertices of Q(A) are
∑
integral if and only if r = s.
PROOF. By Proposition 4.6.3 we obtain that `k + en+1 is a vertex of Q(A) for
k = 1, . . . , s. The integral vertices of Q(A) have entries in {0, 1} because A is a
clutter matrix. Hence using Theorem 4.6.4 and Lemma 4.6.5 the result follows
readily. □
REMARK 4.6.7. We can use Corollary 4.6.6 and Lemma 4.6.5 to effectively
compute the minimal vertex covers of a clutter 𝒞 using linear programming, see
Exercise 4.6.73.
Notation In the sequel we shall always assume that `1 , . . . , ` s are the vectors defined
as: ∑
`k = −en+1 + ei
xi ∈Ck
and that `1 , . . . , `r are the vectors of Eq. (4.40). Notice that dk = 1 for k = 1, . . . , s
and dk = −⟨`k , en+1 ⟩ for 1 ≤ k ≤ r.
DEFINITION 4.6.8. The Simis cone of the point configuration 𝒜 is the rational
polyhedral cone defined as:
(4.41) Cn(𝒜) := He+1 ∩ ⋅ ⋅ ⋅ ∩ He+n+1 ∩ H`+1 ∩ ⋅ ⋅ ⋅ ∩ H`+s .
The term Simis cone is intended to do homage to Aron Simis.
By Lemma 3.3.2 there exists a finite set ℋ ⊂ ℕn+1 such that
Cn(𝒜) = ℝ+ ℋ and ℤn+1 ∩ ℝ+ ℋ = ℕℋ.
THEOREM 4.6.9. R s (I) = K[ℕℋ].
PROOF. Recall that K[ℕℋ] = K[{ xa tb ∣ (a, b) ∈ ℕℋ}]. Take xa tb ∈ R s (I), that
is, xa ∈ 𝔭bi for all i. Hence ⟨(a, b), `i ⟩ ≥ 0 for all i or equivalently (a, b) ∈ Cn(𝒜).
Thus (a, b) ∈ ℕℋ and xa tb ∈ K[ℕℋ]. Conversely take xa tb ∈ K[ℕℋ], then (a, b)
is in Cn(𝒜) and ⟨(a, b), `i ⟩ ≥ 0 for all i. Hence xa ∈ 𝔭bi for all i and xa ∈ I (b) , as
required. □
PROCEDURE 4.6.10. In order to determine the minimal integral Hilbert basis of
Cn(𝒜) using Normaliz [25] one may proceed as follows. First one computes the
irreducible representation of the Rees cone to obtain `1 , . . . , ` s . Then one dualizes
 4. MONOMIAL RINGS AND OPTIMIZATION PROBLEMS 169

Eq. (4.41) and use Normaliz again to obtain an irreducible representation of the dual
cone:
Cn(𝒜)∗ = {y ∈ ℝn+1 ∣ ⟨y, x⟩ ≥ 0, ∀ x ∈ Cn(𝒜)}
= ℝ+ e1 + ⋅ ⋅ ⋅ + ℝ+ en+1 + ℝ+ `1 + ⋅ ⋅ ⋅ + ℝ+ ` s
= He+1 ∩ ⋅ ⋅ ⋅ ∩ He+n ∩ Hβ+1 ∩ ⋅ ⋅ ⋅ ∩ Hβ+p .
By the duality theorem for cones, the set
ℬ = {e1 , . . . , en , β1 , . . . , β p }
generates Cn(𝒜) as a cone. Since Cn(𝒜)∗ is pointed, the subcones
ℝ+ e1 , . . . , ℝ+ en+1 , ℝ+ `1 , . . . , ℝ+ ` s
are the 1-dimensional faces of Cn(𝒜)∗ , see [150, Section 8.8, pp. 105-106]. In
general ℬ is not the minimal integral Hilbert basis of Cn(𝒜), but one can use ℬ as
input for Normaliz to obtain the desired Hilbert basis. See Example 4.6.11 for a
concrete illustration of this precedure.
There is a recent version of Normaliz [24] that conputes the Hilbert basis of a
pointed cone defined by linear inequalities, i.e., with this version we can compute
the Hilbert basis of Cn(𝒜) directly.
EXAMPLE 4.6.11. Using Theorem 4.6.9, Procedure 4.6.10 and Normaliz, below
we compute the minimal generators of the symbolic Rees algebra of the edge ideal
I = I(𝒞 ) = (x1 y, x2 y, x3 y, x4 y, x5 y, x1 x2 , x2 x3 , x3 x4 , x4 x5 , x1 x5 ).
The input file for Normaliz is:
10
6
1 0 0 0 0 1
0 1 0 0 0 1
0 0 1 0 0 1
0 0 0 1 0 1
0 0 0 0 1 1
1 1 0 0 0 0
0 1 1 0 0 0
0 0 1 1 0 0
0 0 0 1 1 0
1 0 0 0 1 0
3
Using Normaliz we obtain that the dual cone Cn(𝒜)∗ is generated by the rows of
the following matrix:
170 4. MONOMIAL RINGS AND OPTIMIZATION PROBLEMS

13
7
0 1 0 1 1 1 -1
1 1 0 1 0 1 -1
1 0 0 0 0 0 0
0 0 0 0 0 0 1
1 1 1 1 1 0 -1
1 0 1 0 1 1 -1
0 1 0 0 0 0 0
1 0 1 1 0 1 -1
0 1 1 0 1 1 -1
0 0 0 1 0 0 0
0 0 0 0 0 1 0
0 0 1 0 0 0 0
0 0 0 0 1 0 0
1
We use this input file for Normaliz to compute the supporting hyperplanes of Cn(𝒜)∗
which by duality gives the following set of generators for Cn(𝒜).
23
7
0 0 0 1 0 1 1
1 0 0 0 0 1 1
0 0 0 1 1 0 1
1 0 0 0 1 1 2
0 0 0 0 1 1 1
1 1 1 1 1 0 3
1 1 0 0 0 0 1
0 0 0 1 0 0 0
0 1 0 0 0 1 1
0 0 0 0 0 1 0
0 0 1 0 0 0 0
0 0 1 0 0 1 1
0 0 1 1 0 1 2
0 1 1 0 0 1 2
0 1 1 0 0 0 1
0 0 1 1 0 0 1
1 1 0 0 0 1 2
0 1 0 0 0 0 0
1 0 0 0 0 0 0
1 1 1 1 1 2 5
0 0 0 0 1 0 0
0 0 0 1 1 1 2
1 0 0 0 1 0 1
0
 4. MONOMIAL RINGS AND OPTIMIZATION PROBLEMS 171

These row vectors may not form a Hilbert basis. To obtain a Hilbert basis we
use Normaliz once more to obtain that the minimal Hilbert basis of Cn(𝒜) has 24
generators. The extra generator is
1 1 1 1 1 1 4
The symbolic Rees algebra of I is minimally generated as a K-algebra by x1 , . . . , x5 ,
It, x1 x5 yt2 , x3 x4 yt2 , x2 x3 yt2 , x1 x2 yt2 , x4 x5 yt2 , x1 ⋅ ⋅ ⋅ x5 t3 , x1 ⋅ ⋅ ⋅ x5 yt4 , x1 ⋅ ⋅ ⋅ x5 y2 t5 .
Let ℕ𝒜′ be the subsemigroup of ℕn+1 generated by 𝒜′ , consisting of the linear
combinations of 𝒜′ with non-negative integral coefficients. The Rees algebra of I
can be written as
R[It] = K[{ xa tb ∣ (a, b) ∈ ℕ𝒜′ }]
= R ⊕ It ⊕ I 2 t2 ⊕ ⋅ ⋅ ⋅ ⊕ I i ti ⊕ ⋅ ⋅ ⋅ .
According to Theorem 4.3.23 the integral closure of R[It] in its field of fractions
can be expressed as
R[It] = K[{ xa tb ∣ (a, b) ∈ ℤn+1 ∩ ℝ+ 𝒜′ }]
= R ⊕ It ⊕ I 2 t2 ⊕ ⋅ ⋅ ⋅ ⊕ I i ti ⊕ ⋅ ⋅ ⋅ ,
where I i = ({ xa ∈ R∣ ∃ p ≥ 1; (xa ) p ∈ I pi }) is the integral closure of I i . If I = I, the
ideal I is said to be complete or integraly closed. Hence we get:
PROPOSITION 4.6.12. R[It] is a normal domain if and only if any of the following
two equivalent conditions hold:
(a) ℕ𝒜′ = ℤn+1 ∩ ℝ+ 𝒜′ , i.e., 𝒜′ is an integral Hilbert basis.
(b) I i = I i for all i ≥ 1.
If the second condition holds we say that I is a normal ideal.
PROPOSITION 4.6.13. Let Ii = I ∩ K[X ∖ { xi }]. If Ii is a normal ideal for
i = 1, . . . , n and
(4.42) H`1 ∩ H`2 ∩ ⋅ ⋅ ⋅ ∩ H`r ∩ ℝn+1
+ ∕= (0),
then R[It] is normal.
PROOF. Let xa tb = x1a1 ⋅ ⋅ ⋅ xnan tb ∈ R[It] be a minimal generator, that is (a, b)
cannot be written as a sum of two non-zero integral vectors in ℝ+ 𝒜′ . It suffices to
prove that 0 ≤ b ≤ 1 because this readily implies that xa is either a variable or a
monomial in F. Assume b ≥ 2. Since Ii is normal we may assume that ai ≥ 1
for all i. As each variable occurs in at least one monomial of F, using Eq. (4.42) it
follows that there is (vk , 1) such that ⟨(vk , 1), `i ⟩ = 0 for i = 1, . . . , r. Therefore
⟨(a − vk , b − 1), `i ⟩ ≥ 0 (i = 1, . . . , r).
Thus (a, b) − (vk , 1) ∈ ℝ+ 𝒜′ , a contradiction to the choice of xa tb . □
172 4. MONOMIAL RINGS AND OPTIMIZATION PROBLEMS

PROPOSITION 4.6.14 ([65],[84]). The following conditions are equivalent

(a) Q(A) is integral.
(b) The irreducible representation of the Rees cone has the form
ℝ+ 𝒜′ = He+1 ∩ ⋅ ⋅ ⋅ ∩ He+n+1 ∩ H`+1 ∩ ⋅ ⋅ ⋅ ∩ H`+s .
(c) R s (I) = R[It].
PROOF. (a) ⇔ (b): It follows from Corollary 4.6.6. (a) ⇔ (c): it follows from
Theorem 4.6.9. □
EXAMPLE 4.6.15. Let I = (x1 x2 x5 , x1 x3 x4 , x2 x3 x6 , x4 x5 x6 ). The figure:

1q p p p p p pp pp p p p p p p p p pp p p p p p p p p p pp p p p p p q 2
p p pp p p pp pp p pp p p pp p p p p p pp p pp p p p p
pp p p p p p p p p pp p pp p
pp p p p p pp ppp
ppp q ppp
pp p
pp p p
pp 5
p ppp p pp
p p
p pp p p p pp
p p p pp p ppp p pp
p p pp pp pp 4 p p p p pp pp ppp pp p p p pp pp pp p
p p p p p pp p q q p p pp p p
p pp pp pp pp 6 p p pp p pp pp p p
p pp p p pp pp pp
p pp p pp p pp p p p p
p pp
p pp p
pp p p p

q
3
corresponds to the clutter associated to I. This clutter will be denoted by 𝒬6 .
Using Normaliz we obtain that R[It] = R[It][x1 ⋅ ⋅ ⋅ x6 t2 ]. Thus R[It] is not normal.
The supporting hyperplanes of the facets of the Rees cone are defined by the rows
(normal vectors) of the matrix :
⎡ ⎤
1 0 1 1 0 0 −1
⎢
⎢ 0 1 1 0 0 1 −1 ⎥
⎥
⎢
⎢ 1 1 0 0 1 0 −1 ⎥
⎥
⎢
⎢ 0 0 0 0 1 0 0 ⎥
⎥
⎢
⎢ 0 0 0 0 0 1 0 ⎥
⎥
⎢
⎢ 0 0 0 0 0 0 1 ⎥
⎥
⎢
⎢ 0 0 0 1 0 0 0 ⎥
⎥.
⎢
⎢ 0 1 0 0 0 0 0 ⎥
⎥
⎢
⎢ 0 1 0 1 0 0 −1 ⎥
⎥
⎢
⎢ 0 0 1 0 0 0 0 ⎥
⎥
⎢
⎢ 1 0 0 0 0 0 0 ⎥
⎥
⎢
⎢ 0 0 1 0 1 0 −1 ⎥
⎥
⎣ 1 0 0 0 0 1 −1 ⎦
0 0 0 1 1 1 −1
 4. MONOMIAL RINGS AND OPTIMIZATION PROBLEMS 173

Hence the symbolic Rees algebra is the integral closure of the Rees algebra.
Altogether we have:
R[It] ⊊ R[It] = R s (I).
LEMMA 4.6.16. If x1 is a variable in Ck for some 1 ≤ k ≤ s such that
′ ′
xv1 = x1 xv1 , . . . , xv p = x1 xv p
′ ′
/ supp(xvi ) for i > p, then there is xv j such that supp(xv j ) ∩ Ck = ∅.
and x1 ∈
′
PROOF. If supp(xv j ) ∩ Ck ∕= ∅ for all j, then Ck ∖ { x1 } is a vertex cover of 𝒞 , a
contradiction because Ck is a minimal vertex cover. □
Let 𝔭 be a prime ideal of R. For use below we denote the localization of the
Rees algebra R[It] at the multiplicative set R ∖ 𝔭 by R[It]𝔭
LEMMA 4.6.17. Let B = R[It]. If 𝔭 ∈ Spec(R), then 𝔭 B𝔭 ∩ R = 𝔭.
PROOF. It is left as an exercise. □
LEMMA 4.6.18. Let B = R[It]. If 𝔭 is a minimal prime of I, then 𝔭 B𝔭 ∩ B is a
minimal prime of IB.
PROOF. From the equality 𝔭 B𝔭 = 𝔭R𝔭 [𝔭R𝔭 ] we get that 𝔭 B𝔭 is a prime ideal.
Hence its contraction 𝔭 B𝔭 ∩ B is also a prime ideal. To prove the minimality take
P ∈ Spec(B) such that IB ⊂ P ⊂ 𝔭 B𝔭 ∩ B. Contracting to R and using Lemma 4.6.17
we get P ∩ R = 𝔭. From this equality it follows that PB𝔭 ∩ B = P. Therefore
𝔭 B𝔭 ∩ B ⊂ PB𝔭 ∩ B = P
and consequently P = 𝔭 B𝔭 ∩ B, as required. □

Notation In the sequel Jk(dk ) will denote the ideal of R[It] given by
Jk(dk ) = ({ xa tb ∣ ⟨(a, b), `k ⟩ ≥ dk }) ∩ R[It] (k = 1, . . . , r)
and Jk will denote the ideal of R[It] given by
Jk = ({ xa tb ∣ ⟨(a, b), `k ⟩ > 0}) ∩ R[It] (k = 1, . . . , r),
where dk = −⟨`k , en+1 ⟩. If dk = 1 we have Jk(1) = Jk . By Lemma 4.6.5 we have that
dk = 1 if and only if 1 ≤ k ≤ s. In general Jk(dk ) might not be equal to the dk th
symbolic power of Jk , see for instance [182].
PROPOSITION 4.6.19. J1 , . . . , Jr are height one prime ideals containing IR[It]
and Jk is equal to 𝔭k R[It]𝔭k ∩ R[It] for k = 1, . . . , s. If Q(A) is integral, then
rad(IR[It]) = J1 ∩ J2 ∩ ⋅ ⋅ ⋅ ∩ J s .
174 4. MONOMIAL RINGS AND OPTIMIZATION PROBLEMS

PROOF. IR[It] is clearly contained in Jk for all k by construction. To show that

Jk is a prime ideal of height one it suffices to notice that the right hand side of the
isomorphism:
R[It]/Jk ≃ K[{ xa tb ∈ R[It]∣ ⟨(a, b), `k ⟩ = 0}]
is an n-dimensional integral domain, because Fk = ℝ+ 𝒜′ ∩ H`k is a facet of the Rees
cone for all k. Set Pk = 𝔭k R[It]𝔭k ∩ R[It] for 1 ≤ k ≤ s. By Lemma 4.6.18 this
ideal is a minimal prime of IR[It] (cf. [115]) and admits the following description
Pk = 𝔭k R𝔭k [𝔭k R𝔭k t] ∩ R[It]
= 𝔭k + (𝔭2k ∩ I)t + (𝔭3k ∩ I 2 )t2 + ⋅ ⋅ ⋅ + (𝔭i+1 i i
k ∩ I )t + ⋅ ⋅ ⋅

Notice that xa ∈ 𝔭b+1

∑
k if and only if ⟨a, xi ∈Ck ei ⟩ ≥ b + 1. Hence Jk = Pk .
Assume that Q(A) is integral, i.e., r = s. Take xα tb ∈ Jk for all k. Using
Eq. (4.40) it is not hard to see that (α, b + 1) ∈ ℝ+ 𝒜′ , that is xα tb+1 is in R[It] and
xα tb+1 ∈ I b+1 tb+1 . It follows that xα tb is a monomial in the radical of IR[It]. This
proves the asserted equality. □
PROPOSITION 4.6.20. If P ∈ { J1 , . . . , J s }, then
R[It] ∩ IR[It]P = P,
that is, the P-primary component of IR[It] is P.
PROOF. In general the left hand side is contained in the right hand side. To show
the reverse containment notice that we may assume
P = (x1 , . . . , xm , xv1 t, . . . , xv p t)R[It],
where 𝔭 = (x1 , . . . , xm ) is a minimal prime of I. Set C = { x1 , . . . , xm }.
Case (I): Consider xk with 1 ≤ k ≤ m. By Lemma 4.6.16 there is j such that
x = xk xα and supp(xα ) ∩ C = ∅. Thus since xα is not in P (because of the second
v j

condition) we obtain xk ∈ R[It] ∩ IR[It]P .

Case (II): Consider xvk t with 1 ≤ k ≤ p. Since
⟨(vk , 1), e1 + ⋅ ⋅ ⋅ + em − en+1 ⟩ ≥ 1,
the monomial xvk contains at least two variables in C. Thus we may assume that
x1 , x2 are in the support of xvk . Again by Lemma 4.6.16 there are j, ` such that
xv j = x1 xα , xv` = x2 xγ , and the support of xα and xγ disjoint from C. Hence the
monomial xvk xα+γ t belongs to I 2 t and xα+γ is not in P. Writing
xvk t = (xvk xα+γ t)/xα+γ ,
we get xvk t ∈ R[It] ∩ IR[It]P . □
LEMMA 4.6.21. rad (Jk(dk ) ) = Jk for 1 ≤ k ≤ r.
 4. MONOMIAL RINGS AND OPTIMIZATION PROBLEMS 175

PROOF. By construction rad(Jk(dk ) ) ⊂ Jk . The reverse inclusion follows by

noticing that if xa tb ∈ Jk , then (xa tb )dk ∈ Jkdk ⊂ Jk(dk ) . □
The following result was shown to us by Yuji Yoshino.
PROPOSITION 4.6.22. If R[It] is normal, then:
IR[It] = J1(d1 ) ∩ J2(d2 ) ∩ ⋅ ⋅ ⋅ ∩ Jr(dr ) .
PROOF. “⊂”: Assume xα tb ∈ IR[It]. Since xα ∈ I b+1 we readily obtain
(α, b + 1) ∈ ℕ𝒜′ . In particular we get (α, b + 1) ∈ ℝ+ 𝒜′ . Therefore
0 ≤ ⟨(α, b + 1), `k ⟩ = ⟨(α, b), `k ⟩ − dk
and consequently xα tb ∈ J (dk ) for 1 ≤ k ≤ r.
“⊃”: Assume xα tb ∈ J (dk ) for all k. Since the element (α, b+1) is in ℝ+ 𝒜′ ∩ℤn+1
and using that R[It] is normal yields (α, b+1) ∈ ℕ𝒜′ . It follows that xα tb ∈ I b+1 tb ⊂
IR[It]. □
A similar formula is shown in [27]. The normality of R[It] can be described in
terms of primary decompositions of IR[It], see [109, Proposition 2.1.3].
Set 𝔭k = Jk ∩ R for k = 1, . . . , r. The following two nice formulas, pointed out
to us by Wolmer Vasconcelos, describe the difference between the symbolic Rees
algebra of I and the normalization of its Rees algebra.

s
∩ r
∩
R s (I) = R[It]𝔭k ∩ R[t], R[It] = R[It]𝔭k ∩ R[t].
k=1 k=1
These representations are linked to the so called Rees valuations of the ideal I, see
[181, Chapter 8] for further details.
LEMMA 4.6.23. If I is a monomial ideal of R, then the nilradical of the associated
graded ring of I is given by
nil(grI (R)) = ({g ∈ I i /I i+1 ∣ g s ∈ I si+1 ; i ≥ 0; s ≥ 1})
= ({ xα ∈ I i /I i+1 ∣ x sα ∈ I si+1 ; i ≥ 0; s ≥ 1}).
PROOF. The first equality holds because the radical of any graded algebra is
generated by homogeneous elements.
To prove the second equality take an equivalence class 0 ∕= g ∈ I i /I i+1 such that
g s ∈ I is+1 . There are nonzero scalars λ1 , . . . , λr in K and monomials m1 , . . . , mr in
I i such that
g = λ1 m1 + ⋅ ⋅ ⋅ + λr mr
and m1 ≺ ⋅ ⋅ ⋅ ≺ mr , where ≺ is the lexicographical ordering. We may assume that
mj ∈ / I i+1 for all j. Since mrs is the leading term of g s and since I is+1 is a monomial
176 4. MONOMIAL RINGS AND OPTIMIZATION PROBLEMS

ideal, we get mrs ∈ I is+1 . Thus mr is in the nilradical of grI (R). By induction it
follows that m j is in nil(grI (R)) for all j, as required. □
DEFINITION 4.6.24. The clutter 𝒞 satisfies the max-flow min-cut (MFMC) prop-
erty if both sides of the LP-duality equation
(4.43) min{⟨α, x⟩∣ x ≥ 0; xA ≥ 1} = max{⟨y, 1⟩∣ y ≥ 0; Ay ≤ α}
have integral optimum solutions x and y for each non-negative integral vector α.
THEOREM 4.6.25 ([65, 84, 115]). The following are equivalent
(i) grI (R) is reduced.
(ii) R[It] is normal and Q(A) is an integral polyhedron.
(iii) x ≥ 0; xA ≥ 1 is a TDI system.
(iv) 𝒞 has the max-flow min-cut property.
(v) R[It] = R s (I), i.e., I i = I (i) for i ≥ 1.
PROOF. (i) ⇒ (ii). First we show the normality of R[It]. By induction we show
that I i = I i for i ≥ 1. If i = 1, then I = I because I is a clutter ideal. Assume
I i = I i . Take xα in I i+1 ⊂ I i . Then x sα ∈ I s(i+1) ⊂ I si+1 for some s ≥ 1. Hence
xα ∈ I i /I i+1 is a nilpotent element of grI (R) and consequently xα ∈ I i+1 . This proves
I i+1 = I i+1 . Notice that IR[It] is a radical ideal because grI (R) ≃ R[It]/IR[It]. As
R[It] is normal, from Proposition 4.6.22 we get r = s. Hence Q(A) is integral by
Proposition 4.6.14
(iv) ⇒ (i): Let xα ∈ I i /I i+1 be in nil(grI (R)), that is x sα ∈ I is+1 for some
0 ∕= s ∈ ℕ. By Lemma 4.6.23 we need only show xα = 0. It follows rapidly that
the maximum in Eq. (4.43) is greater than or equal to i + 1/s. By hypothesis the
maximum in Eq. (4.43) has an integral optimum solution y = (y1 , . . . , yq ). Thus
y1 + ⋅ ⋅ ⋅ + yq ≥ i + 1. Since y satisfies y ≥ 0 and Ay ≤ α we obtain xα ∈ I i+1 . This
proves xα = 0, as required.
(iii) ⇔ (iv): This is just Proposition 3.3.15. (ii) ⇔ (iv): This follows at once
from Theorem 3.3.16 and Proposition 4.6.12. (ii) ⇔ (v): Since R s (I) is a normal
domain, this follows from Proposition 4.6.14. □
COROLLARY 4.6.26. If A is totally unimodular, then grI (R) is reduced.
PROOF. By Proposition 3.6.2 the system x ≥ 0; xA ≥ 1 is TDI. Hence grI (R) is
reduced by Theorem 4.6.25. □
DEFINITION 4.6.27. Let X = { x1 , . . . , xn } and let
X ′ = { xi1 , . . . , xir , x j1 , . . . , x js }
be a subset of X. A minor of I is a proper ideal I ′ of R′ = K[X ∖ X ′ ] obtained from
I by making xik = 0 and x j` = 1 for all k, `. The ideal I is considered itself a minor.
A minor of 𝒞 is a clutter 𝒞 ′ that corresponds to a minor (0) ⊊ I ′ ⊊ R′ of I.
 4. MONOMIAL RINGS AND OPTIMIZATION PROBLEMS 177

Notice that in the definition above 𝒞 ′ is obtained from I ′ by considering the

unique set of square-free monomials of R′ that minimally generate I ′ . The notion
of a minor of a clutter is not a generalization of the notion of a minor of a graph in
the sense of graph theory [151, p. 25], see Definition 5.1.17. For instance if G is a
cycle of length four and we contract an edge we obtain that a triangle is a minor of
G, but a triangle cannot be a minor of G in our sense.
PROPOSITION 4.6.28. If I i = I (i) for some i ≥ 2 and J = I ′ is a minor of I, then
J i = J (i) .
PROOF. Assume J is the minor obtained from I by making x1 = 0. Take
xα ∈ J (i) . Then xα ∈ I (i) = I i because J ⊂ I. Thus xα ∈ I i . Since x1 ∈ / supp(xα ) it
follows that xα ∈ J i . This proves J (i) ⊂ J i . The other containment is clear because
J (i) is integrally closed.
Assume J is the minor obtained from I by making x1 = 1. Take xα ∈ J (i) .
Notice that x1i xα ∈ I (i) = I i . Indeed if x1 ∈ 𝔭k , then x1i ∈ 𝔭ik , and if x1 ∈
/ 𝔭k , then
J ⊂ 𝔭k and xα ∈ 𝔭ik . Since x1 ∈ / supp(xα ) it follows that xα ∈ J i . □

COROLLARY 4.6.29. If R s (I) = R[It], then R s (I ′ ) = R′ [I ′ t] for any minor I ′ of I.

DEFINITION 4.6.30. Let 𝒟 = ϒ(𝒞 ) be the clutter of minimal vertex covers of 𝒞 .
The clutter 𝒞 is called diadic if ∣A ∩ B∣ ≤ 2 for A ∈ 𝒞 and B ∈ 𝒟
PROPOSITION 4.6.31. If R[It] = R s (I) and 𝒞 is diadic, then R[It] is normal.
PROOF. Let xa tb = x1a1 ⋅ ⋅ ⋅ xnan tb ∈ R[It] be a minimal generator, that is (a, b)
cannot be written as a sum of two non-zero integral vectors in the Rees cone ℝ+ 𝒜′ .
We may assume ai ≥ 1 for 1 ≤ i ≤ m, ai = 0 for i > m, and b ≥ 1.
Case (I): ⟨(a, b), `i ⟩ > 0 for all i. The vector γ = (a, b) − e1 satisfies ⟨γ, `i ⟩ ≥ 0
for all i, that is γ ∈ ℝ+ 𝒜′ . Thus since (a, b) = e1 + γ we derive a contradiction.
Case (II): ⟨(a, b), `i ⟩ = 0 for some i. We may assume
{`i ∣ ⟨(a, b), `i ⟩ = 0} = {`1 , . . . , ` p }.

Subcase (II.a): ei ∈ H`1 ∩ ⋅ ⋅ ⋅ ∩ H` p for some 1 ≤ i ≤ m. It is not hard to verify

that the vector γ = (a, b) − ei satisfies ⟨γ, `k ⟩ ≥ 0 for all 1 ≤ k ≤ s. Thus γ ∈ ℝ+ 𝒜′ ,
a contradiction because (a, b) = ei + γ.
Subcase (II.b): ei ∈/ H`1 ∩ ⋅ ⋅ ⋅ ∩ H` p for all 1 ≤ i ≤ m. Since the vector (a, b)
belongs to the Rees cone it follows that we can write
(a, b) = λ1 (v1 , 1) + ⋅ ⋅ ⋅ + λq (vq , 1) (λi ≥ 0). (∗)
By the choice of xa tb we may assume 0 < λ1 < 1. Set γ = (a, b) − (v1 , 1) and notice
that by Eq. (∗) this vector has non-negative entries. We claim that γ is in the Rees
178 4. MONOMIAL RINGS AND OPTIMIZATION PROBLEMS

cone. Since by hypothesis one has 0 ≤ ⟨(v1 , 1), ` j ⟩ ≤ 1 for all j we readily obtain
⟨(a, b), `k ⟩ − ⟨(v1 , 1), `k ⟩ = 0 if 1 ≤ k ≤ p,
{
⟨γ, `k ⟩ =
⟨(a, b), `k ⟩ − ⟨(v1 , 1), `k ⟩ ≥ 0 otherwise.
Thus γ ∈ ℝ+ 𝒜′ and (a, b) = (v1 , 1) + γ. As a result γ = 0 and (a, b) ∈ R[It], as
desired. □
COROLLARY 4.6.32. [39, Theorem 1.3] If Q(A) is integral and 𝒞 is diadic, then
x ≥ 0; xA ≥ 1 is a TDI system.
PROOF. By Proposition 4.6.31 the Rees algebra R[It] is normal. To complete
the proof apply Theorem 4.6.25. □
Notation For use below we set [n] = {1, . . . , n} and 𝒫 = conv(v1 , . . . , vq ).
PROPOSITION 4.6.33. If `k has the form `k = −dk en+1 + i∈Ak ei for some
∑

Ak ⊂ [n] for k = 1, . . . , r, then A(𝒫 )[x1 , . . . , xn ] = R[It].

PROOF. The Ehrhart ring A(𝒫 ) is contained in R[It]. Thus the equality follows
adapting the proof of Proposition 4.6.31. □
DEFINITION 4.6.34. If `1 , . . . , `r satisfy the condition of Proposition 4.6.33, we
say that the Rees cone of I is quasi-ideal.
THEOREM 4.6.35. If R[It] = R s (I) and K[Ft] = A(𝒫 ), then the Rees algebra
R[It] is normal.
PROOF. From Proposition 4.6.14 r = s and the vectors `1 , . . . , `r satisfy the
hypothesis of Proposition 4.6.33. Hence
R[It] = K[Ft][x1 , . . . , xn ] = A(𝒫 )[x1 , . . . , xn ] = R[It]. □
PROPOSITION 4.6.36. If xv1 , . . . , xvq have degree d ≥ 2 and I b = I (b) for all b,
then I b is generated by monomials of degree bd for b ≥ 1.
PROOF. The monomial ideal I b has a unique minimal set of generators consisting
of monomials. Take xa in this minimal set. Notice that (a, b) is in ℝ+ 𝒜′ . Thus we
may proceed as in the proof of Proposition 4.6.31 to obtain that (a, b) is in the cone
generated by {(v1 , 1), . . . , (vq , 1)}. This yields deg(xa ) = bd. □
PROPOSITION 4.6.37. If grI (R) is reduced (resp. R[It] is normal) and I ′ is a
minor of I, then grI ′ (R′ ) is reduced (resp. R′ [I ′ t] is normal).
PROOF. Notice that we need only show the result when I ′ is a minor obtained
from I by making x1 = 0 or x1 = 1. Using Lemma 4.6.23 both cases are quite easy
to prove. □
A set of edges of the clutter 𝒞 is independent if no two of them have a common
vertex. We denote the smallest number of vertices in any minimal vertex cover of
 4. MONOMIAL RINGS AND OPTIMIZATION PROBLEMS 179

𝒞 by α0 (𝒞 ) and the maximum number of independent edges of 𝒞 by β1 (𝒞 ). These

numbers are related to min-max problems because they satisfy:
α0 (𝒞 ) ≥ min{⟨1, x⟩∣ x ≥ 0; xA ≥ 1}
= max{⟨y, 1⟩∣ y ≥ 0; Ay ≤ 1} ≥ β1 (𝒞 ).
Notice that α0 (𝒞 ) = β1 (𝒞 ) if and only if both sides of the equality have integral
optimum solutions. These two numbers can be interpreted in terms of invariants of
I. By Proposition 4.6.3 the height of the ideal I, denoted by ht(I), is equal to the
vertex covering number α0 (𝒞 ). On the other hand the edge independence number
can be expressed as:
β1 (𝒞 ) = max{r∣ ∃ a regular sequence of monomials xα1 , . . . , xαr ∈ I }.
The number in the right hand side is denoted by mgrade(I), it is called the monomial
grade of the ideal.
DEFINITION 4.6.38. If α0 (𝒞 ) = β1 (𝒞 ) we say that the clutter 𝒞 (or the ideal I)
has the König property.
DEFINITION 4.6.39. A clutter 𝒞 satisfies the packing property (PP) if all its
minors satisfy the König property, that is, α0 (𝒞 ′ ) = β1 (𝒞 ′ ) for any minor 𝒞 ′ of 𝒞 .
PROPOSITION 4.6.40. If 𝒞 satisfies PP and I = I(𝒞 ), then I 2 = I 2 .
PROOF. By induction on n. Assume I 2 ∕= I 2 and consider M = I 2 /I 2 . If
𝔭 ∕= 𝔪 = (x1 , . . . , xn ) is a prime ideal of R, then by induction M𝔭 = (0). Thus 𝔪 is
the only associated prime of M and there is an embedding
R/𝔪 ,→ M (1 7→ xa ),
where xa ∈ I 2 ∖ I 2 and xi xa ∈ I 2 for all i. Notice that by induction all the entries
ai of a are positive. We consider two cases. Assume ai ≥ 2 for some i, say i = 1.
Given a monomial xα , the monomial obtained from xα by making x1 = 1 is denoted
by xα . Then making x1 = 1 and using x1 xa ∈ I 2 gives
′

xa = xv1 xv2 xδ ,
′ ′ ′

hence xa = x1a1 xa = xv1 xv2 xγ ∈ I 2 , a contradiction. On the other hand if ai = 1 for

′

all i, then xa = x1 ⋅ ⋅ ⋅ xn ∈ I g ⊂ I 2 , where g = ht(I), a contradiction. Therefore

I2 = I2. □
THEOREM 4.6.41. (A. Lehman; see [38, Theorem 1.8]) If 𝒞 has the packing
property, then Q(A) is integral.
COROLLARY 4.6.42. If the ring grI (R) is reduced, then α0 (𝒞 ′ ) = β1 (𝒞 ′ ) for any
minor 𝒞 ′ of 𝒞 .
180 4. MONOMIAL RINGS AND OPTIMIZATION PROBLEMS

PROOF. Let 𝒞 ′ be any minor of 𝒞 and let I ′ be its clutter ideal. By Proposi-
tion 4.6.37 the associated graded ring grI ′ (R′ ) is reduced. Hence by Theorem 4.6.25
the system x ≥ 0; xA′ ≥ 1 is TDI, where A′ is the clutter matrix associated to 𝒞 ′ .
Since clutters of TDI systems have the König property we get α0 (𝒞 ′ ) = β1 (𝒞 ′ ). □
PROPOSITION 4.6.43. If 𝒞 is a graph, A is its incidence matrix, and I is its edge
ideal, then the following are equivalent:
(a) grI (R) is reduced.
(b) 𝒞 is bipartite.
(c) Q(A) is integral.
(d) 𝒞 has the packing property.
(e) 𝒞 has the max-flow-min-cut property.
(f) The system x ≥ 0; xA ≥ 1 is TDI.
(g) R[It] = R s (I), i.e., I i = I (i) for all i.
(h) R[It] = R s (I), i.e., I i = I (i) for all i.
PROOF. By Corollary 4.6.26 (b) implies (a). To show that (a) implies (b) assume
that 𝒞 has an odd cycle { xr+1 = x1 , . . . , xr } with r odd and consider the monomial
m = x1 ⋅ ⋅ ⋅ , xr . Set s = (r − 1)/2. Notice that 0 ∕= m ∈ I s /I s+1 and m2 ∈ I 2s+1 .
Hence m2 = 0, a contradiction.
Applying Theorem 4.6.25 and Proposition 4.6.31 we obtain that (a) is equivalent
to (c). By Corollary 4.6.42 condition (a) implies (d). Using Theorem 4.6.41 we
obtain that (d) implies (c). The equivalence between (a), (e), (f) and (h) was
shown in Theorem 4.6.25 and the equivalence between (c) and (g) was shown in
Proposition 4.6.14. □
COROLLARY 4.6.44. If 𝒞 has the max-flow min-cut property, then 𝒞 has the
packing property.
PROOF. It follows at once from Theorem 4.6.25 and Corollary 4.6.42. □
CONJECTURE 4.6.45. (Conforti-Cornuéjols [37]) If the clutter 𝒞 has the packing
property, then 𝒞 has the max-flow min-cut property.
To the best of our knowledge this conjecture is still open, see [38, Conjec-
ture 1.6]. The converse of Corollary 4.6.42 gives an algebraic version of this
conjecture:
CONJECTURE 4.6.46. If α0 (𝒞 ′ ) = β1 (𝒞 ′ ) for all minors 𝒞 ′ of 𝒞 , then the ring
grI (R) is reduced.
Using Theorems 4.6.41 and 4.6.25 this conjecture reduces to:
CONJECTURE 4.6.47. If α0 (𝒞 ′ ) = β1 (𝒞 ′ ) for all minors 𝒞 ′ of 𝒞 , then R[It] is
normal.
 4. MONOMIAL RINGS AND OPTIMIZATION PROBLEMS 181

DEFINITION 4.6.48. A clutter is binary if its edges and its minimal vertex covers
intersect in an odd number of vertices.
THEOREM 4.6.49 ([152]). A binary clutter 𝒞 has the max-flow min-cut property
if and only if 𝒬6 is not a minor of 𝒞 .
COROLLARY 4.6.50. If 𝒞 is a binary clutter with the packing property, then 𝒞
has the max-flow min-cut property.
PROOF. Notice that 𝒬6 is not a minor of 𝒞 because 𝒬6 does not have the König
property. Thus 𝒞 has the max-flow min-cut property by Theorem 4.6.49. □
THEOREM 4.6.51. If xv1 , . . . , xvq are monomials of degree d ≥ 2 such that grI (R)
is reduced and the matrix ( )
v1 ⋅ ⋅ ⋅ vq
B=
1 ⋅⋅⋅ 1
has rank r, then ∆r (B) = 1.
PROOF. Since R[It] is normal, using Theorem 4.5.21 we get K[Ft] = A(𝒫 ).
Therefore ∆r (B) = 1 by Theorem 4.5.14. □
Using this result we obtain that a positive answer to Conjecture 4.6.46 implies
the following:
CONJECTURE 4.6.52 ([80]). If α0 (𝒞 ′ ) = β1 (𝒞 ′ ) for all minors 𝒞 ′ of 𝒞 and
xv1 , . . . , xvq
have degree d ≥ 2, then the group
ℤn+1 /((v1 , 1), . . . , (vq , 1))
is torsion-free, or equivalently ∆r (B) = 1 where r = rank(B).
PROPOSITION 4.6.53 ([80]). If Q(A) is an integral polyhedron and xv1 , . . . , xvq
have degree d ≥ 2, then there exists a minimal vertex cover of 𝒞 intersecting every
edge of 𝒞 in exactly one vertex.
PROOF. Let B be the matrix whose columns are the vertices of Q(A). By
Theorem 4.6.41 Q(B) is an integral polyhedron. Thus the columns of B are the
characteristic vectors of the minimal vertex covers of 𝒞 .
We proceed by contradiction. Assume that for each column αk of B there exists
vik such that
vik B ≥ 1 + ek .
Consider the vector α = vi1 + ⋅ ⋅ ⋅ + vis . From the inequality
αB ≥ (s + 1, . . . , s + 1)
we obtain that α/(s + 1) ∈ Q(B). Thus by the finite basis theorem we can write
α/(s + 1) = λ2 v1 + ⋅ ⋅ ⋅ + λq vq + µ1 e1 + ⋅ ⋅ ⋅ + µn en (λi , µ j ≥ 0).
182 4. MONOMIAL RINGS AND OPTIMIZATION PROBLEMS

Therefore taking inner product with 1 we get

∣α∣ = sd ≥ (s + 1)d,
a contradiction. This proof was adapted from [51]. □
PROPOSITION 4.6.54. If Q(A) is integral and xv1 , . . . , xvq have degree d ≥ 2,
then there are X1 , . . . , Xd mutually disjoint minimal vertex covers of 𝒞 such that
X = ∪di=1 Xi and
(4.44) ∣supp(xvi ) ∩ Xk ∣ = 1 ∀ i, k.
PROOF. By induction on d. By Lemma 4.6.53 there is a minimal vertex cover
X1 of 𝒞 such that ∣supp(xvi ) ∩ X1 ∣ = 1 for all i. Consider the ideal I ′ obtained from I
by making xi = 1 for xi ∈ X1 . Then I ′ is an ideal generated by monomials of degree
d − 1. Since the integrality of Q(A) is preserved under taking minors we can apply
the induction hypothesis to get the required assertion. □
The simplest case of Proposition 4.6.54 occurs when we have a 2-partition of
X, i.e., when ∣Xi ∣ = 2 for all i. Below we will examine this case.

Minors and the packing property

Let R = K[x1 , . . . , xn ] be a polynomial ring over a field K and let
X = X1 ∪ X2 ∪ ⋅ ⋅ ⋅ ∪ Xd
be a partition of X = { x1 , . . . , xn } into d subsets of size two. We set Xi = { x2i−1 , x2i }
for i = 1, . . . , d and d ≥ 2. Note n = 2d.
Let I be an ideal of R minimally generated by a finite set of square-free mono-
mials F = { xv1 , . . . , xvq } such that
(4.45) ∣supp(xvi ) ∩ Xk ∣ = 1 ∀ i, k.
For convenience we shall always assume that each variable xi occurs in at least one
monomial of F. In the sequel A will denote the n × q incidence matrix with column
vectors v1 , . . . , vq , and 𝒞 will denote the clutter associated to I, i.e., I = I(𝒞 ).
REMARK 4.6.55. In our situation, by the pigeon hole principle, any minimal
vertex cover C of the clutter 𝒞 satisfies 2 ≤ ∣C ∣ ≤ d. Notice that for each odd
integer k the sum of rows k and k + 1 of the matrix A is equal to 1 = (1, . . . , 1). Thus
the rank of A is bounded by d + 1.
The next result shows that A has “maximal rank” if 𝒞 has a cover of maximum
possible size.
PROPOSITION 4.6.56 ([51]). Let C be a minimal vertex cover of 𝒞 . If ∣C ∣ = d ≥ 3
and 𝒞 satisfies the König property, then rank(A) = d + 1.
 4. MONOMIAL RINGS AND OPTIMIZATION PROBLEMS 183

PROOF. First notice that C contains exactly one element of each X j because
X j ∕⊂ C. Thus we may assume

C = { x1 , x3 , . . . , x2d−1 }.

Consider the monomial xα = x2 x4 ⋅ ⋅ ⋅ x2d and notice that xk xα ∈ I for each xk ∈ C

because the monomial xk xα is clearly in every minimal prime of I. Writing xk =
x2i−1 with 1 ≤ i ≤ d we conclude that the monomial

xαi = x2 x4 ⋅ ⋅ ⋅ x2(i−1) x2i−1 x2(i+1) ⋅ ⋅ ⋅ x2d

is a minimal generator of I. Thus we may assume xαi = xvi for i = 1, . . . , d. The

vector 1 belongs to the linear space generated by v1 , . . . , vq because 𝒞 has the König
property. It follows that the matrix with rows v1 , . . . , vd , 1 has rank d + 1. □

The following result gives some support to Conjecture 4.6.52.

THEOREM 4.6.57 ([51]). If A has rank r and 𝒞 has the packing property, then
( )
A
∆r = 1.
1

PROOF. By induction on r = rank(A). Since 1 is the sum of the first two rows of
A it suffices to prove that 1 is the only invariant factor of At or equivalently that the
Smith normal form of At is the “identity”. Let w1 , . . . , w2d be the columns of At and
let Vi be the linear space generated by w1 , . . . , w2i . Notice that if k is odd and we
remove columns wk and wk+1 from At we obtain a submatrix whose rank is greater
than or equal to r − 1, see Remark 4.6.55. Thus after permuting columns we may
assume
{
i + 1 if 1 ≤ i ≤ r − 1,
(4.46) dim(Vi ) =
r if r ≤ i.

Let J be the monomial ideal defined by the rows of [w1 , . . . , w2(r−1) ]. Then a
minimal set of generators of J consists of monomials of degree r − 1 and will still
satisfy condition in Eq. (4.45) with d replaced by r − 1. Furthermore J satisfies the
packing property because it is a minor of I. If [w1 , . . . , w2(r−1) ] diagonalizes (over
the integers) to the identity matrix so does At . Therefore we may assume d = r − 1
and I = J.
Let B be the matrix [w1 , . . . , w2(d−1) ] and let I ′ be the monomial ideal defined
by the rows of B, that is I ′ is obtained from I making x2d−1 = x2d = 1. Notice that
by induction B diagonalizes to an identity matrix [Ir−1 , 0] of order r − 1. Since I has
the König property we may permute rows and columns and assume that the matrix
184 4. MONOMIAL RINGS AND OPTIMIZATION PROBLEMS

At is written as:
10 10 10 ⋅⋅⋅ 10 10 ←
01 01 01 ⋅⋅⋅ 01 01
⃝ ⃝ ⃝ ⋅⋅⋅ ⃝ 10 ←
.. .. .. .. ..
. . . . .
⃝ ⃝ ⃝ ⋅⋅⋅ ⃝ 10 ←
⃝ ⃝ ⃝ ⋅⋅⋅ ⃝ 01
.. .. .. ..
. . . .
⃝ ⃝ ⃝ ⋅⋅⋅ ⃝ 01
| {z }
B
where either a pair 1 0 or 0 1 must occur in the places marked with a circle and such
that the number of 1′ s in the last column is greater than or equal to the number of
1′ s in any other column. Consider the matrix C obtained from A by removing the
rows whose penultimate entry is equal to 1 (these are marked above with an arrow)
and removing the last column. Let K be the monomial ideal defined by the rows of
C, that is K is obtained from I by making x2d−1 = 0 and x2d = 1. By the choice of
the last column and because of Eq. (4.46) it is seen that K has height two. Since K
is a minor of I it has the König property. As a consequence using row operations At
can be brought to the form:
10 10 10 ⋅⋅⋅ 10 10
10 10 10 ⋅⋅⋅ 10 01
⃝ ⃝ ⃝ ⋅⋅⋅ ⃝ 10
.. .. .. ..
. . . .
⃝ ⃝ ⃝ ⋅⋅⋅ ⃝ 10
⃝ ⃝ ⃝ ⋅⋅⋅ ⃝ 01
.. .. .. ..
. . . .
⃝ ⃝ ⃝ ⋅⋅⋅ ⃝ 01
| {z }
B1

where B1 has rank r − 1 and diagonalizes to an identity. Therefore it is readily seen

that this matrix reduces to

Ir−1 0 0 ⋅⋅⋅ 0 0 0
0 0 0 ⋅⋅⋅ 0 1 −1
0 0 0 ⋅⋅⋅ 0 a1 b1
.. .. .. .. ..
. . . . .
0 0 0 ⋅⋅⋅ 0 ar br

To finish the proof observe that by rank considerations (see Eq. (4.46)) this matrix
reduces to [Ir , 0], as required. □
 4. MONOMIAL RINGS AND OPTIMIZATION PROBLEMS 185

COROLLARY 4.6.58. If K[Ft] is normal and 𝒞 satisfies PP, then 𝒞 has the
max-flow min-cut property.
COROLLARY 4.6.59 ([51]). If 𝒞 satisfies PP and v1 , . . . , vq are linearly indepen-
dent, then 𝒞 has the max-flow min-cut property.
PROPOSITION 4.6.60 ([51]). If I is minimally non-normal and Q(A) is integral,
then rank(A) = d + 1.
PROOF. Let xα tb be a minimal generator of R[It] not in R[It] and let m = xα .
Using Proposition 4.6.36, one has deg(m) = bd. By hypothesis 𝔪 is the only
associated prime of M = I b /I b . Hence we have
∩
rad(ann(M)) = 𝔭 = 𝔪 = (x1 , . . . , xn )
𝔭∈Ass(M)

and 𝔪r ⊂ ann(M) for some r > 0. Thus for i odd we can write
xir xα = (xv1 )a1 ⋅ ⋅ ⋅ (xvq )aq xδ ,
where a1 + ⋅ ⋅ ⋅ + aq = b and deg(xδ ) = r. If we write xδ = xis1 xi+1
s2 γ
x with xi , xi+1
γ / {i, i + 1}, it is not hard to see that
not in the support of x , making x j = 1 for j ∈
r = s1 + s2 and γ = 0. Thus we get an equation:
xis2 xα = (xv1 )a1 ⋅ ⋅ ⋅ (xvq )aq xi+1
s2

with s2 > 0. Using a similar argument we obtain an equation:

w1 α
xi+1 x = (xv1 )b1 ⋅ ⋅ ⋅ (xvq )bq xiw1
with w1 > 0. Therefore
s2 +w1 v1 a1
xi+1 (x ) ⋅ ⋅ ⋅ (xvq )aq = xis2 +w1 (xv1 )b1 ⋅ ⋅ ⋅ (xvq )bq .
Consider the group ℤn /ℤ𝒜, where 𝒜 = {v1 , . . . , vq }. Since this group is torsion-
free, we get ei − ei+1 ∈ ℤ𝒜 for i odd. Finally to conclude that rank(A) = d + 1
notice that 1 ∈ ℤ𝒜. □

Exercises
4.6.61. Prove that the vertex covering number and the edge independent number
of 𝒞 satisy:
α0 (𝒞 ) ≥ min{⟨1, x⟩∣ x ≥ 0; xA ≥ 1}
= max{⟨y, 1⟩∣ y ≥ 0; Ay ≤ 1} ≥ β1 (𝒞 ).
4.6.62. Prove that the edge independence number is given by
β1 (𝒞 ) = max{r∣ ∃ a regular sequence of monomials xα1 , . . . , xαr ∈ I }.
4.6.63. Prove that the following conditions are equivalent
186 4. MONOMIAL RINGS AND OPTIMIZATION PROBLEMS

(a) α0 (𝒞 ) = β1 (𝒞 ).
(b) The monomial x1 ⋅ ⋅ ⋅ xn tg belongs to R[It], where g is the height of I.
(c) Both sides of the equality
min{⟨1, x⟩∣ x ≥ 0; xA ≥ 1} = max{⟨y, 1⟩∣ y ≥ 0; Ay ≤ 1}
have integral optimum solutions.
4.6.64. Prove the following assertions:
(a) dim(ℝ+ 𝒜′ ) = n + 1.
(b) Hei ∩ ℝ+ 𝒜′ is a facet of ℝ+ 𝒜′ for i = 1, . . . , n + 1.
(c) He+1 , . . . , He+n+1 occur in the irreducible representation of ℝ+ 𝒜′ .
Hint Use that the height of I is at least two.
4.6.65. Let A be a matrix with entries in ℕn . Consider the matrix
( )
′ A
A = ; 1 = (1, . . . , 1).
1
Prove that the vertices of Q = Q(A) and Q′ = Q(A′ ) are related by
vert(Q′ ) = {en+1 } ∪ {(β, 0)∣ β ∈ vert(Q)}.
4.6.66. Let v1 and v2 be two vectors in ℝn with entries in {0, 1} and let I =
(xv1 , xv2 )
⊂ R. Prove that the Rees algebra R[It] is normal.
4.6.67. Prove that the ideal I = (x12 , x22 ) ⊂ k[x1 , x2 ] is not normal.
4.6.68. If grI (R) is reduced (resp. R[It] is normal) and I ′ is a minor of I, prove
that grI ′ (R′ ) is reduced (resp. R′ [I ′ t] is normal).
4.6.69. Let I and I ′ be monomial ideals of R and R′ = K[x2 , . . . , xn ] respectively.
If I is square-free and I = x1 I ′ , prove that grI (R) is reduced if and only if grI ′ (R′ ) is
reduced.
4.6.70. Let I = (xv1 , . . . , xvq ) be a square-free monomial ideal of height g in a
polynomial ring R = K[x1 , . . . , xn ] and let A = (v1 , . . . , vq ). If deg(xvi ) = d ≥ 2 for
all i and Q(A) is integral, then
(a) n ≥ gd.
(b) If x1 ⋅ ⋅ ⋅ xn ∈ K[xv1 , . . . , xvq ], then n = gd and the clutter 𝒞 associated to
I has the König property. Prove that 𝒞 is vertex critical, i.e. the height of
the ideal I(𝒞 ∖ { xi }) is strictly less than the height of I = I(𝒞 ) for all i.
Hint (a) Notice that 1A ≥ d1 and use the finite basis theorem. (b) Use that the
property “Q(A) is integral” is closed under taking minors (see Proposition 4.6.28).
 4. MONOMIAL RINGS AND OPTIMIZATION PROBLEMS 187

4.6.71. Let xw1 , . . . , xwr be a set of generators of I and let B be the matrix whose
columns are w1 , . . . , wr . Then both sides of the equation
min{⟨1, x⟩∣ x ≥ 0; xA ≥ 1} = max{⟨y, 1⟩∣ y ≥ 0; Ay ≤ 1}
have integral optimum solutions if and only if both sides of the equation
min{⟨1, x⟩∣ x ≥ 0; xB ≥ 1} = max{⟨y, 1⟩∣ y ≥ 0; By ≤ 1}
have integral optimum solutions. If any of the two systems have integral optimum
solutions x, y their optimal values are equal.
4.6.72. If I is a square-free monomial ideal, prove that its symbolic Rees algebra
R s (I) is finitely generated.
4.6.73. Consider the graph:
xt1 xt5
b "
btx3 x4"
b "
t
" b
" b
t
" bt
x2 x6
Verify that the facets of the Rees cone are:
( 1) - x1 <= 0
( 2) - x2 <= 0
( 3) - x3 <= 0
( 4) - x4 <= 0
( 5) - x5 <= 0
( 6) - x6 <= 0
( 7) - x7 <= 0
( 8) - x1- x2 - x4- x5 + x7 <= 0
( 9) - x1- x2 - x4 - x6+ x7 <= 0
( 10) - x1 - x3- x4- x5 + x7 <= 0
( 11) - x1 - x3- x4 - x6+ x7 <= 0
( 12) - x1 - x3 - x5- x6+ x7 <= 0
( 13) - x2- x3- x4- x5 + x7 <= 0
( 14) - x2- x3- x4 - x6+ x7 <= 0
( 15) - x2- x3 - x5- x6+ x7 <= 0
( 16) - x1- x2- x3- x4- x5- x6+2x7 <= 0
( 17) -2x1 -2x3- x4- x5- x6+2x7 <= 0
( 18) - x1- x2- x3-2x4-2x5 +2x7 <= 0
( 19) - x1- x2- x3-2x4 -2x6+2x7 <= 0
( 20) -2x2-2x3- x4- x5- x6+2x7 <= 0
and that the equations (8)-(15) correspond to the minimal vertex covers of the graph.
For instance equation (8) says that { x1 , x2 , x4 , x5 } is a minimal vertex cover of the
graph.
188 4. MONOMIAL RINGS AND OPTIMIZATION PROBLEMS

4.6.74. If I is an unmixed ideal and 𝒞 satisfies the König property, then x1 =

x1 x2 ⋅ ⋅ ⋅ xn belongs to the subring K[{ xv1 , . . . , xvq }].
4.6.75. Let v1 and v2 be two vectors in ℝn with entries in {0, 1} and let I =
(xv1 , xv2 )
⊂ R. Prove that the Rees algebra R[It] is normal.
4.6.76. Prove that 𝒬6 is a binary clutter.

4.7. Blowup algebras of ideals of covers

Let R = K[x1 , . . . , xn ] be a polynomial ring over a field K and let I be a monomial
ideal of R of height at least two minimally generated by a set of square-free mono-
mials { xv1 , . . . , xvq } of degree d ≥ 2. Recall that the clutter 𝒞 associated to I has
vertex set X = { x1 , . . . , xn } and edge set E = {S 1 , . . . , S q }, where
S k = supp(xvk ).
As before, we will use the notation I = I(𝒞 ). Notice that if G is a graph and I is its
edge ideal, then the clutter 𝒞 associated to I is G.
DEFINITION 4.7.1. The clutter 𝒞 (or the edge ideal I(𝒞 )) is called unmixed if all
the minimal vertex covers of 𝒞 have the same cardinality.
Let 𝔭1 , . . . , 𝔭 s be the minimal primes of I(𝒞 ) and let
Ck = { xi ∣ xi ∈ 𝔭k } (k = 1, . . . , s)
be the corresponding minimal vertex covers of 𝒞 .
DEFINITION 4.7.2. Let xuk = xi ∈Ck xi for 1 ≤ k ≤ s. The ideal of vertex covers
∏
of 𝒞 is the square-free monomial ideal
Ic (𝒞 ) = (xu1 , . . . , xus ) ⊂ R.
In the literature Ic (𝒞 ) is also called the Alexander dual of I because if ∆ is the
Stanley-Reisner complex of I, then Ic (𝒞 ) is the Stanley-Reisner ideal of ∆∗ , the
Alexander-dual of ∆. The survey article [99] explains the role of Alexander duality
to prove combinatorial and algebraic theorems.
In our case Ic (𝒞 ) is an unmixed ideal of height d because I is generated by
square-free monomials of degree d.
We set 𝒜 = {v1 , . . . , vq } and 𝒜′ = {e1 , . . . , en , (v1 , 1), . . . , (vq , 1)}.

PROPOSITION 4.7.3. If R[It] = R s (I) and J = Ic (𝒞 ), then R[Jt] = R s (J).

 4. MONOMIAL RINGS AND OPTIMIZATION PROBLEMS 189

PROOF. Using Proposition 4.6.14 one has

ℝ+ 𝒜′ = He+1 ∩ ⋅ ⋅ ⋅ ∩ He+n+1 ∩ H`+1 ∩ ⋅ ⋅ ⋅ ∩ H`+s ,
where `k = (uk , −1) and xuk = xi ∈Ck xi for all k. Using the duality theorem for
∏
cones we get:
(n ) (q ) ( n+1 ) ( s )
∩ ∩ ∩ ∑ ∑
He+i H(v+
i ,1)
= ℝ + ei + ℝ+ (ui , −1) . (∗)
i=1 i=1 i=1 i=1

By Proposition 4.6.14 we need only show the equality

( n ) ( s ) (n+1 ) ( q )
∑ ∑ ∩ ∩ ∩
ℝ + ei + ℝ+ (ui , 1) = +
Hei H(vi ,−1) .
+

i=1 i=1 i=1 i=1

Clearly the left hand side is contained in the right hand side. To show the reverse
containment take (α, b) in the right hand side, where α ∈ ℝn and b ∈ ℝ. Then
0 ≤ ⟨(α, b), (vi , −1)⟩ = ⟨(α, −b), (vi , 1)⟩
for all i. Hence using Eq. (∗) we can write
s
∑ n+1
∑
(α, −b) = λi (ui , −1) + µi ei (λi ; µ j ≥ 0). (∗∗)
i=1 i=1

Since λ1 + ⋅ ⋅ ⋅ + λ s − b = µn+1 ≥ 0, there are λ′1 , . . . , λ′s such that

0 ≤ λ′i ≤ λi ; b = λ′1 + ⋅ ⋅ ⋅ + λ′s ; µn+1 = (λ1 − λ′1 ) + ⋅ ⋅ ⋅ + (λ s − λ′s ).
Therefore using Eq. (∗∗) we conclude
s
∑ n
∑ s
∑
(α, b) = λ′i (ui , 1) + µi ei + (λi − λ′i )ui ,
i=1 i=1 i=1

as desired. □
The symbolic Rees algebra of Ic (𝒞 ) can be interpreted as the subalgebra gener-
ated by monomials associated to “k-covers” [103].
COROLLARY 4.7.4. [38, Theorem 1.17] If Q(A) is integral and A′ is the incidence
matrix of the clutter of minimal vertex covers of 𝒞 , then Q(A′ ) is integral.
PROOF. It follows at once from Propositions 4.6.14 and 4.7.3. □
THEOREM 4.7.5 ([79]). Let G be a graph and let J = Ic (G) be its ideal of vertex
covers. If R[It] is equal to R s (I), then R[It] and R[Jt] are normal.
190 4. MONOMIAL RINGS AND OPTIMIZATION PROBLEMS

PROOF. It is a consequence of Proposition 4.6.31. □

If I = I(G) is the edge ideal of a bipartite graphs G and J = Ic (G), the next result
shows that gr J (R) is reduced, this complements the fact that grI (R) is reduced (see
Proposition 4.6.43).
COROLLARY 4.7.6. If G is a bipartite graph and J = Ic (G), then gr J (R) is
reduced.
PROOF. According to Theorem 4.6.25 and Proposition 4.6.43 one has the
equality R[It] = R s (I). Thus Proposition 4.7.3 yields the equality R[Jt] = R s (J).
On the other hand R[Jt] is normal by Theorem 4.7.5. Thus by Theorem 4.6.25 the
associated graded ring of J is reduced. □
If G is a graph, then R s (Ic (G)) is generated as a K-algebra by elements of
t-degree at most two [103].
An application to Rees algebras. By assigning deg(xi ) = 1 for all i and deg(t) =
−(d − 1), the Rees algebra R[It] becomes a standard graded K-algebra generated
by elements of degree 1. Below we compute the a-invariant of R[It] with respect to
this grading.
THEOREM 4.7.7 ([79]). If grI (R) is a reduced ring and I = I(G) is an unmixed
ideal of height 2, then R[It] is Gorenstein and
a(R[It]) = −(n − d + 1).
PROOF. Since R[It] is a normal domain, then according to Danilov-Stanley
formula its canonical module is the ideal of R[It] given by
ωR[It] = ({ x1a1 ⋅ ⋅ ⋅ xnan tan+1 ∣ a = (ai ) ∈ (ℝ+ 𝒜′ )o ∩ ℤn+1 }),
By Proposition 4.3.29, the a-invariant can be expressed as
a(R[It]) = −min{ i ∣ (ωR[It] )i ∕= 0}.
Using Eq. (4.40) it is seen that the vector (1, . . . , 1, 1) is in the interior of the Rees
cone. Thus a(R[It]) is at least −(n − d + 1) = − deg(x1 ⋅ ⋅ ⋅ xn t). To show the reverse
inequality take an arbitrary monomial xa tb = x1a1 ⋅ ⋅ ⋅ xnan tb in the ideal ωR[It] , that is
(a, b) ∈ (ℝ+ 𝒜′ )o . By Proposition 4.6.14 the vector (a, b) has positive entries and
satisfies
−b + xi ∈Ck ai ≥ 1 (k = 1, . . . , s).
∑
(4.47)
If b = 1 we obtain the inequality
deg(xa tb ) = a1 + ⋅ ⋅ ⋅ + an − (d − 1) ≥ n − d + 1. (∗)
Now assume 2 ≤ b. Using the normality of R[It] and Eqs. (4.40) and (4.47) it
follows that the monomial
m = x1a1 −1 ⋅ ⋅ ⋅ xnan −1 tb−1
 4. MONOMIAL RINGS AND OPTIMIZATION PROBLEMS 191

belongs to R[It]. Since xa tb = mx1 ⋅ ⋅ ⋅ xn t, the inequality (∗) also holds in this case.
Altogether we conclude the desired equality for the a-invariant. From the argument
above we obtain that ωR[It] is generated by x1 ⋅ ⋅ ⋅ xn t and thus R[It] is a Gorenstein
ring. □
COROLLARY 4.7.8 ([79]). Let J = Ic (G) be the ideal of vertex covers of G. If G
is a bipartite graph and I = I(G) is unmixed, then R[Jt] is a Gorenstein ring and
a(R[Jt]) = −(n − α0 (G) + 1).
PROOF. It follows from Theorem 4.7.7 once we recall that gr J (R) is a reduced
ring according to Corollary 4.7.6. □
REMARK 4.7.9. The a-invariant of R[It] is equal to −(β0 + 1), where β0 is the
Krull dimension of R/I (see Theorem 6.3.5).
The family of unmixed bipartite graphs contains the class of Cohen-Macaulay
bipartite graphs (see Proposition 2.2.14). Both classes have been classified (see
Theorems 2.5.7 and 2.5.8).

Exercises
4.7.10. Consider the clutter 𝒬6 whose edges are
{{ x1 , x2 , x5 }, { x1 , x3 , x4 }, { x2 , x3 , x6 }, { x4 , x5 , x6 }}.
Prove that I = I(𝒬6 ) is minimally non-normal, i.e., I is not normal and any proper
minor I ′ of I is normal.
4.7.11. Let I = I(G) be the edge ideal of a graph G. When is I minimally
non-normal?

4.8. The basis monomial ring of a matroid

Let M be a matroid of rank d on a finite set X = { x1 , . . . , xn } and let ℬ be the
collection of bases of M. Notice that ℬ is a clutter on X. Its clutter ideal I = I(ℬ )
will be called the basis monomial ideal of M.
DEFINITION 4.8.1. The set of all square-free monomials xi1 ⋅ ⋅ ⋅ xid ∈ R such that
{ xi1 , . . . , xid } ∈ ℬ will be denoted by F M and the subsemigroup (of the multiplicative
semigroup of monomials of R) generated by F M will be denoted by 𝕄 M . The basis
monomial ring of M is the ring K[F M ], the monomial subring of R generated by
F M over K.
Notice that K[F M ] is a standard graded K-algebra because all basis of M have
d elements. For use below let us state a result that implies the normality of the basis
monomial ring of a matroid.
192 4. MONOMIAL RINGS AND OPTIMIZATION PROBLEMS

PROPOSITION 4.8.2 ([196]). If xa is a monomial of degree `d for some ` ∈ ℕ

such that (xa ) p ∈ 𝕄 M for some 0 ∕= p ∈ ℕ, then xa ∈ 𝕄 M .
COROLLARY 4.8.3 ([196]). Let 𝒫 be the convex hull of all ei1 + ⋅ ⋅ ⋅ + eid such
that { xi1 , . . . , xid } is a basis of M. Then A(𝒫 ) = K[F M t]. In particular K[F M ] is
normal.
PROOF. It suffices to show the inclusion A(𝒫 ) ⊂ K[F M t]. Take xa tb ∈ A(𝒫 ),
that is, a ∈ ℤn ∩ b𝒫 . Hence xa has degree bd and (xa ) p ∈ 𝕄 M for some positive
integer p. By Proposition 4.8.2 we get xa ∈ 𝕄 M . It is rapidly seen that xa tb is in
K[F M t], as required. □
In [197, Conjecture 12] N. White conjectures in terms of exchange properties
for bases that, for every matroid M, the toric ideal P M of K[F M ] is generated by
quadrics. This has been shown for graphic matroids [13].
PROPOSITION 4.8.4. If ℬ has the packing property, then grI (R) is reduced.
PROOF. Since Q(A) is integral, by Lehman theorem, using Theorem 4.6.35 we
get that R[It] is normal, and both conditions yield that grI (R) is reduced according
to Theorem 4.6.25. □
PROPOSITION 4.8.5. Let X1 , . . . , Xd be a family of disjoint sets of variables and
let M be the transversal matroid whose collection of basis is
𝒞 = {{y1 , . . . , yd }∣ yi ∈ Xi ∀ i}.
If I is the basis monomial ideal of M, then grI (R) is reduced.
PROOF. Let 𝒟 be the clutter of minimal vertex covers of 𝒞 . Notice that the
edges of 𝒟 are precisely X1 , . . . , Xd , consequently the incidence matrix of 𝒟 is
totally unimodular. Hence Q(A) is integral by Corollary 4.7.4. Thus the result
follows applying Theorems 4.6.25 and 4.6.35 together with Corollary 4.8.3. □
Another property which is closed under taking minors is:
PROPOSITION 4.8.6. Let I = I(ℬ ). If I ′ is a minor of I, then I ′ is again the basis
monomial ideal of a matroid M ′ .
PROOF. Let I ′ be the ideal obtained from I by making xn = 1 and let G I ′ be the
unique minimal generating set of I ′ consisting of monomials. It suffices to prove
that I ′ is generated by monomials of degree d − 1. Take f ∈ G I ′ . We proceed by
contradiction. Assume that f has the form:
f = x1 x2 ⋅ ⋅ ⋅ xd
/ supp( f ). Since xn must occur as an element of some basis of M, there is
with xn ∈
a monomial in I of the form
g = xi1 xi2 ⋅ ⋅ ⋅ xid−1 xn .
 4. MONOMIAL RINGS AND OPTIMIZATION PROBLEMS 193

If supp(g) ∖{ xn } ⊂ { x1 , . . . , xd } we obtain a contradiction because of the minimality

of G I ′ . Thus we may assume xi1 ∈ / supp( f ). By the exchange property (see
Theorem 3.7.13) we may assume that the monomial
g1 = x1 xi2 xi3 ⋅ ⋅ ⋅ xid−1 xn
belongs to I. If supp(g1 ) ∖ { xn } ⊂ { x1 , . . . , xd } we obtain a contradiction. Thus we
may assume xi2 ∈ / supp( f ). Again by the exchange property we may assume that
the monomial
g2 = x1 x2 xi3 xi4 ⋅ ⋅ ⋅ xid−1 xn
belongs to I. Continuing with this procedure we conclude that the monomial
x1 x2 ⋅ ⋅ ⋅ xd−1 xn belongs to I, a contradiction. The case xn = 0 is also not hard to
show. □
Rees cone of a matroid. Let { B1 , . . . , Bq } be the collection of bases of a matroid
of rank d on the set [[1, n]] := {1, . . . , n}. Thus ∣ Bi ∣ = d for all i. For use below we
set
∑
𝒜′ = {e1 , . . . , en , fB1 , . . . , fBq }; fBi = en+1 + e j.
j∈Bi

PROPOSITION 4.8.7 ([192]). If F is a facet of ℝ+ 𝒜′ , then F has the form ℝ+ 𝒜′ ∩

Hb , where b = ei for some 1 ≤ i ≤ n + 1 or b = (bi ) is an integral vector such that
bi ∈ {0, 1} for i = 1, . . . , n and bn+1 ≥ −d.
PROOF. The proof is by induction on n. The case n = 1 is clear because the only
possibilities for b are b = (0, 1) and b = (1, −1). Let F be a facet of ℝ+ 𝒜′ . There
is a unique integral vector 0 ∕= b = (bi ) whose non-zero entries are relatively prime
and such that
(a) F = ℝ+ 𝒜′ ∩ Hb , ℝ+ 𝒜′ ⊂ Hb+ , and
(b) there is a linearly independent set A ⊂ Hb ∩ 𝒜′ with ∣A∣ = n.
If A = {e1 , . . . , en } (resp. A ⊂ { fB1 , . . . , fBq }), then b = en+1 (resp. b is the vector
(1, . . . , 1, −d)). Notice that bi ≥ 0 for i = 1, . . . , n. Thus we may assume that A is
the set {e1 , . . . , e s , fB1 , . . . , fBt }, where 1 ≤ s ≤ n − 1, s + t = n, and {1, . . . , s} is
the set of all i ∈ [[1, n]] such that ei ∈ Hb . By induction we may assume that 1 ∈ Bk
for some 1 ≤ k ≤ q. Indeed if 1 is not in ∪qi=1 Bi , then the facets of ℝ+ 𝒜′ different
from He1 ∩ ℝ+ 𝒜′ are in one to one correspondence with the facets of
ℝ+ {e2 , . . . , en , fB1 , . . . , fBq }.
Assume that 1 ∈ / B1 . Since 1 ∈ Bk ∖ B1 for some k, by the exchange property
there is j ∈ B1 ∖ Bk such that (B1 ∖ { j}) ∪ {1} = Bi for some basis Bi . Hence
fB1 + e1 = fBi + e j ⇒ ⟨ fBi , b⟩ = −⟨e j , b⟩ ⇒ ⟨ fBi , b⟩ = ⟨e j , b⟩ = 0,
194 4. MONOMIAL RINGS AND OPTIMIZATION PROBLEMS

i.e., fBi and e j belong to Hb . Thus from the outset we may assume that 1 ∈ B1 .
Next assume that t ≥ 2 and 1 ∈ / B2 . Since 1 ∈ B1 ∖ B2 , by the exchange property
there is j ∈ B2 ∖ B1 such that (B2 ∖ { j}) ∪ {1} = Bi for some basis Bi . Hence
fB2 + e1 = fBi + e j ⇒ ⟨ fBi , b⟩ = ⟨e j , b⟩ = 0,
i.e., fBi and e j belong to Hb . Notice that i ∕= 1. Indeed if i = 1, then from the equality
above we get that { fB2 , e1 , fB1 , e j } is linearly dependant, a contradiction. Applying
the arguments above repeatedly shows that we may assume that 1 belongs to Bi for
i = 1, . . . , t. We may also assume that B1 , . . . , Br is the set of all basis of M such
that 1 ∈ Bi , where t ≤ r. For 1 ≤ i ≤ r, we set Ci = Bi ∖ {1}. Notice that there is a
matroid M ′ of rank d − 1 whose collection of bases is {C1 , . . . , Cr }. Consider the
Rees cone ℝ+ 𝒜′′ generated by
𝒜′′ = {e2 , . . . , en , fC1 , . . . , fCr }.
Notice that ℝ+ 𝒜′′ ⊂ Hb+ and that {e2 , . . . , e s , fC1 , . . . , fCt } is a linearly independent
set. Thus ℝ+ 𝒜′′ ∩ Hb is a facet of the cone ℝ+ 𝒜′′ and the results follows. □
COROLLARY 4.8.8 ([192]). If I = I(ℬ ) is the basis monomial ideal of a matroid,
then R[It] is normal.
PROOF. It follows from Propositions 4.6.33, 4.8.7, and Corollary 4.8.3. □

4.9. Computation of Hilbert functions

Let R = k[x1 , . . . , xd ] be a polynomial ring over a field k, with d ≥ 2, and let
I = (xu1 , . . . , xu p ) be a zero dimensional monomial ideal of R, i.e., dim(R/I) = 0.
Let R[It] be the normalization of I. This algebra has for components the integral
closures of the powers of I,
R[It] = R ⊕ It ⊕ ⋅ ⋅ ⋅ ⊕ I i ti ⊕ ⋅ ⋅ ⋅ ⊂ R ⊕ It ⊕ ⋅ ⋅ ⋅ ⊕ I i ti ⊕ ⋅ ⋅ ⋅ = R[It].
Here we will study the normalization of I and show how to compute its Hilbert
function using linear programming techniques.
There is a unique set of monomials F = { xv1 , . . . , xvq } that minimally generate
I. We may assume that vi = ai ei for 1 ≤ i ≤ d, where a1 , . . . , ad are positive integers
and ei is the ith unit vector of ℚd . Set α0 = (1/a1 , . . . , 1/ad ). We may also assume
that {vd+1 , . . . , v s } is the set of vi such that ⟨vi , α0 ⟩ < 1, and {v s+1 , . . . , vq } is the set
of vi such that i > d and ⟨vi , α0 ⟩ ≥ 1. Consider the convex polytopes in ℚd :
P := conv(v1 , . . . , v s ), S := conv(0, v1 , . . . , vd ),
and the rational convex polyhedron
Q := ℚd+ + conv(v1 , . . . , vq ) = ℚd+ + conv(v1 , . . . , vd , vd+1 , . . . , v s ).
 4. MONOMIAL RINGS AND OPTIMIZATION PROBLEMS 195

The second equality follows from the finite basis theorem and using the equality
ℚd+ + conv(v1 , . . . , vd ) = { x ∣ x ≥ 0; ⟨ x, α0 ⟩ ≥ 1}.
PROPOSITION 4.9.1 ([191]). I n = ({ xa ∣ a ∈ nQ ∩ ℤd }) for 0 ∕= n ∈ ℕ.
PROOF. Let xα ∈ I n , i.e., xmα ∈ I nm for some 0 ∕= m ∈ ℕ. Hence
α/n ∈ conv(u1 , . . . , u p ) + ℚd+ ⊂ conv(v1 , . . . , vq ) + ℚd+ = Q
and α ∈ nQ ∩ ℤd . Conversely let α ∈ nQ ∩ ℤd . Hence xmα ∈ (I)nm for some
0 ∕= m ∈ ℕ, this yields xα ∈ (I)n . Since (I)n ⊂ I n and the later ideal is complete we
get xα ∈ I n . □
COROLLARY 4.9.2. If ⟨ui , α0 ⟩ ≥ 1 for all i, then P = conv(a1 e1 , . . . , ad ed ) and
I n = (x1a1 , . . . , xdad )n , ∀ n ≥ 1.
PROOF. Recall the following description [181, Proposition 7.22]:
I = (x⌈α⌉ ∣ α ∈ conv(u1 , . . . , u p )),
where ⌈α⌉ is the ceiling of α = (αi ). Its entries are given by ⌈α⌉i = αi if αi ∈ ℕ, and
⌈α⌉i = ⌊αi ⌋ + 1 if αi ∈ / ℕ, where ⌊αi ⌋ is the integer part of αi . Hence ⟨vi , α0 ⟩ ≥ 1
for i = 1, . . . , q and P = conv(a1 e1 , . . . , ad ed ). Now, the required equality follows
from Proposition 4.9.1. □
The Hilbert function of the filtration ℱ = {I n }∞
n=0 is defined as
f (n) = `(R/I n ) = dimk (R/I n ); n ∈ ℕ ∖ {0}; f (0) = 0.
For simplicity we call f the Hilbert function of I. From Proposition 4.9.1 we get:
COROLLARY 4.9.3 ([191]). `(R/I n ) = ∣ℕd ∖ nQ∣ for n ≥ 1.
The function f behaves as a polynomial of degree d:
f (n) = cd nd + cd−1 nd−1 + ⋅ ⋅ ⋅ + c1 n + c0 (n ≫ 0),
where c0 , . . . , cd ∈ ℚ and cd ∕= 0. The Hilbert polynomial of I is:
cd x d + ⋅ ⋅ ⋅ + c0 .
One has the equality d!cd = e(I) = e(I), where e(I) is the multiplicity of I, see [43].
We will express f (n) as a difference of two Ehrhart polynomials and then show a
positive lower bound for cd−1 .
Set d1 = dim(P). The Ehrhart function of P is χP (n) := ∣ℤd ∩ nP∣, for n ∈ ℕ.
This is a polynomial function of degree d1 :
χP (n) = ∣ℤd ∩ nP∣ = ad1 nd1 + ⋅ ⋅ ⋅ + a1 n + a0 (n ≫ 0),
where ai ∈ ℚ for all i. The polynomial E P (x) = ad1 xd1 + ⋅ ⋅ ⋅ + a1 x + a0 is called the
Ehrhart polynomial of P.
196 4. MONOMIAL RINGS AND OPTIMIZATION PROBLEMS

PROPOSITION 4.9.4 ([191]). f (n) = ES (n) − E P (n) for n ∈ ℕ. In particular

f (n) = cd nd + cd−1 nd−1 + ⋅ ⋅ ⋅ + c1 n + c0 for n ∈ ℕ and c0 = 0.
PROOF. Since E P (0) = ES (0) = 1, we get the equality at n = 0. Assume n ≥ 1.
Notice that from the decomposition Q = (ℚd+ ∖ S ) ∪ P we get
nQ = (ℚd+ ∖ nS ) ∪ nP =⇒ ℕd ∖ nQ = [ℕd ∩ (nS )] ∖ [ℕd ∩ (nP)].
Hence by Corollary 4.9.3 we obtain f (n) = ES (n) − E P (n). □
EXAMPLE 4.9.5. Let I = (x12 , x23 ) ⊂ R = k[x1 , x2 ]. It is easy to verify that
I = I + (x1 x22 ). Notice that P = conv((2, 0), (0, 3)) because (1, 2) lies above the
hyperplane x1 /2 + x2 /3 = 1 and S = conv(0, (2, 0), (0, 3)). Using Normaliz [24, 25]
we get ES (n) = 3n2 + 3n + 1 and E P (n) = n + 1. Thus f (n) = 3n2 + 2n.
EXAMPLE 4.9.6. Let I = (x14 , x25 , x36 , x1 x2 x32 ). Using NORMALIZ:
I = I + (x1 x24 , x12 x23 , x13 x22 , x12 x22 x3 , x24 x32 , x1 x35 , x23 x33 , x12 x33 , x13 x2 x3 ,
x22 x34 , x1 x23 x3 , x13 x32 , x2 x35 ).
Note that P = conv((4, 0, 0), (0, 5, 0), (0, 0, 6), (1, 1, 2)). Using NORMALIZ:
f (n) = ES (n) − E P (n) = (1 + 6n + 19n2 + 20n3 )
−(1 + (1/6)n + (3/2)n2 + (13/3)n3 ) = (35/6)n + (35/2)n2 + (47/3)n3 .
THEOREM 4.9.7. [181, Theorem 7.58] I b = II b−1 for b ≥ d.
REMARK 4.9.8. We can use polynomial interpolation together with Theo-
rem 4.9.7 and Proposition 4.9.4 to determine c1 , . . . , cd , see Example 4.9.9.
EXAMPLE 4.9.9. Let I = (x110 , x28 , x35 ). Using CoCoA [32] we obtain that the
values of f at n = 0, 1, 2, 3 are 0, 112, 704, 2176. By polynomial interpolation we
get:
f (n) = `(R/I n ) = (200/3)n3 + 40x2 + (16/3)n, ∀ n ≥ 0.
LEMMA 4.9.10. Let α = (αi ) and β = (βi ) be two vectors in ℚd+ such that αi = βi
for i = 1, . . . , d − 1, βd > αd and ⟨β, α0 ⟩ < 1. Then
(a) β ∈ conv(v1 , . . . , vd , α).
(b) If αi > 0 for i = 1, . . . , d − 1, then β ∈ conv(v1 , . . . , vd , α)o .
(c) If αi > 0 for i = 1, . . . , d − 1 and α ∈ P, then β ∈ Po .
PROOF. (a) To see that β is a convex combination of v1 , . . . , vd , α we set:
d
βd − αd
∑ [ ]
s = αi /ai = ⟨α0 , α⟩ < 1, µ = 1 − > 0,
ad (1 − s)
i=1
λi = (1 − µ)αi /ai ≥ 0, i = 1, . . . , d − 1,
λd = (βd − µαd )/ad = ((βd − αd )/ad ) + αd (1 − µ)/ad > 0.
 4. MONOMIAL RINGS AND OPTIMIZATION PROBLEMS 197

Then β = λ1 v1 + ⋅ ⋅ ⋅ + λd vd + µα and λ1 + ⋅ ⋅ ⋅ + λd + µ = 1, as required.

(b) Set V = {v1 , . . . , vd , α} and ∆ = conv(V). Since V is affinely independent,
∆ is a d-simplex. From [18, Theorem 7.3], the facets of ∆ are precisely those sets
of the form conv(W), where W is a subset of V having d points. If β is not in the
interior of ∆, then β must lie in its boundary by (a). Therefore β lies in some facet
of ∆, which rapidly yields a contradiction.
(c) By part (b) we get β ∈ conv(v1 , . . . , vd , α)o ⊂ Po , as required. □
Notation The relative boundary of P will be denoted by ∂P.
LEMMA 4.9.11. If α ∈ ∂P ∖ conv(v1 , . . . , vd ) and αi > 0 for i = 1, . . . , d, then
the vector α′ = (α1 , . . . , αd−1 , 0) is not in P.
PROOF. Notice that ⟨α, α0 ⟩ < 1. If α′ ∈ P, then by Lemma 4.9.10(c) we obtain
α ∈ Po , a contradiction. Thus α′ ∈
/ P. □
For use below we set
Ki = {a ∈ S ∣ ai = 0} = conv({v1 , . . . , vd , 0} ∖ {vi }); i = 1, . . . , d,
H = conv(v1 , . . . , vd ); K = (∪di=1 Ki ) ∖ H; L = ∂P ∖ H, if H ⊊ P.
Consider the map ψ : L → K given by
α, if αi = 0 for some 1 ≤ i ≤ d,
{
ψ(α) =
(α1 , . . . , αd−1 , 0), if αi > 0 for i = 1, . . . , d.
Take α ∈ L. Then ⟨α, α0 ⟩ < 1. Since ∂P ⊂ P ⊂ S it is seen that ψ(α) ∈ K.
LEMMA 4.9.12. ψ is injective.
PROOF. Let α, β ∈ L. Assume ψ(α) = ψ(β). If αi = 0 for some i and β j = 0 for
some j, then clearly α = β. If βi > 0 for i = 1, . . . , d and α j = 0 for some j, then by
Lemma 4.9.11 it is seen that this case cannot occur. If αi βi > 0 for all i, then α = β
by Lemma 4.9.10(c). □
Let us introduce some more notation. We set
/ supp(xv j )};
𝒜i = {v j ∣ 1 ≤ j ≤ s; xi ∈
Pi = conv(𝒜i ); Hi = conv({v1 , . . . , vd } ∖ {vi }) ⊂ Pi ⊂ Ki .
LEMMA 4.9.13. ∂P ∩ Ki = Pi for i = 1, . . . , d.
PROOF. For simplicity of notation assume i = 1. Let α = (αi ) ∈ ∂P ∩ K1 , then
α ∈ P and α1 = 0. Since α is a convex combination of v1 , . . . , v s it follows that α
is a convex combination of 𝒜1 , i.e., α ∈ P1 . Conversely assume α ∈ P1 . Clearly
α ∈ K1 ∩ P because 𝒜1 ⊂ K1 ∩ P. If α ∈ / ∂P, then α ∈ Po . Thus if dim(P) = d,
then ∑α ∈ P ⊂ S , and if dim(P) = d − 1, we can write α = λ1 v1 + ⋅ ⋅ ⋅ + λd vd such
o o

that di=1 λi = 1 and 0 < λi < 1 for i = 1, . . . , d. In both cases we get αi > 0 for
i = 1, . . . , d, a contradiction. Hence α ∈ ∂P. □
198 4. MONOMIAL RINGS AND OPTIMIZATION PROBLEMS

PROPOSITION 4.9.14 ([191]). Let Ii be the ideal obtained from I by making xi = 0

and let e(Ii ) be its multiplicity. Then
d−1
∑ e(Ii )
2cd−1 ≥ .
(d − 1)!
i=1

PROOF. Case (I): dim(P) = d. Let ES (x) = ad xd + ⋅ ⋅ ⋅ + a1 x + 1 and let

E P (x) = bd xd + ⋅ ⋅ ⋅ + b1 x + 1. Notice ci = ai − bi for all i. From the decompositions
P = Po ∪ ∂P, ∂S = K ∪ H, ∂P = L ∪ H,
and using the reciprocity law we get:
f (n) = ES (n) − E P (n)
= ESo (n) + ∣∂(nS ) ∩ ℤd ∣ − (E Po (n) + ∣∂(nP) ∩ ℤd ∣)
= (−1)d ES (−n) − (−1)d E P (−n) + ∣nK ∩ ℤd ∣ − ∣nL ∩ ℤd ∣
for 0 ∕= n ∈ ℕ. Therefore
2(cd−1 nd−1 + cd−3 nd−3 + terms of lower degree) = ∣nK ∩ ℤd ∣ − ∣nL ∩ ℤd ∣ = g(n).
We have the inclusions:
[(d−1 ) ] (d )
∪ ∪
ψ(L) ⊂ M := (∂P ∩ Ki ) ∪ Kd ∖ H ⊂ K := Ki ∖ H.
i=1 i=1

By Lemma 4.9.13 we have:

(d−1 ) d
∪ ∪
M= (Pi ∖ Hi ) ∪ (Kd ∖ Hd ) and K = (Ki ∖ Hi ).
i=1 i=1

Set h(n) = ∣nK ∩ ℤd ∣ − ∣nM ∩ ℤd ∣. Since Pi ∖ Hi ⊂ Pi , Ki ∖ Hi ⊂ Ki for all i and

because Pi ∩ P j , Ki ∩ K j are polytopes of dimension at most d − 2 for i ∕= j, by the
inclusion-exclusion principle [1, p. 38, Formula 2.12] we obtain:
d
∑ d−1
∑
d
h(n) = ∣n(Ki ∖ Hi ) ∩ ℤ ∣ − ∣n(Pi ∖ Hi ) ∩ ℤd ∣
i=1 i=1
d−1
∑
−∣n(Kd ∖ Hd ) ∩ ℤd ∣ + p(n) = (E Ki (n) − E Pi (n)) + p(n), n ≫ 0,
i=1

where p(n) is a polynomial function of degree at most d − 2. Consider the function

g(n) = ∣nK ∩ ℤd ∣ − ∣nL ∩ ℤd ∣. As g(n) ≥ h(n) and since the leading coefficient of
E Ki (n) − E Pi (n) is equal to e(Ii )/(d − 1)!, the required equality follows.
 4. MONOMIAL RINGS AND OPTIMIZATION PROBLEMS 199

Case (II): dim(P) = d − 1. There is an injective map from nP to nKd induced

by α 7→ (α1 , . . . , αd−1 , 0). Hence
∣ℤd ∩ nP∣ ∣ℤd ∩ nKd ∣
vol(P) = lim ≤ lim = vol(Kd ).
n→∞ nd−1 n→∞ nd−1
The facets of S are K1 , . . . , Kd and P. Therefore
d d−1 d−1
1∑ 1∑ 1 ∑ e(Ii )
cd−1 = −vol(P) + vol(Ki ) ≥ vol(Ki ) = . □
2 2 2 (d − 1)!
i=1 i=1 i=1

Let e0 , e1 , . . . , ed be the Hilbert coefficients of f . Recall that we have:

( ) ( ) ( )
n+d−1 n+d−2 n
f (n) = e0 − e1 d−1
+ ⋅ ⋅ ⋅ + (−1) ed−1 + (−1)d ed ,
d d−1 1
where e0 = e(I) is the multiplicity of I and cd = e0 /d!. Notice that ed = 0 because
f (0) = 0.
PROPOSITION 4.9.15. ei ≥ 0 for all i.
PROOF. Let ℱ = {I n }∞ n=0 be the integral closure filtration. Since the Rees
algebra of ℱ is R[It] and this algebra is normal and Cohen-Macaulay, we derive
that the associated graded ring grℱ (R) = ⊕i≥0 I i /I i+1 is Cohen-Macaulay (see [89,
Corollary 2.1, p. 74]). Recall that grℱ (R) is a finitely algebra over R/I, i.e.,
grℱ (R) = (R/I)[z1 , . . . , zm ], where z1 , . . . , zm are homogeneous of positive degree
(see [23, Proposition 4.5.4]). The canonical map grI (R) → grℱ (R) is integral.
Thus grℱ (R) is a positively graded Noetherian module over the standard graded
algebra grI (R). Hence by the Hilbert-Serre theorem and because grℱ (R) is Cohen-
Macaulay, we can write the Hilbert series of grℱ (R) as h(t)/(1 − x)d , where h(x) is
a polynomial with non-negative integral coefficients, see Exercise 1.6.33. To finish
the proof notice that ei = h(i) (1)/i!. □
COROLLARY 4.9.16 ([191]). e0 (d − 1) − 2e1 ≥ e(I1 ) + ⋅ ⋅ ⋅ + e(Id−1 ) ≥ d − 1.
[ () ]
PROOF. From the equality cd−1 = d!1 e0 d2 − de1 and using Proposition 4.9.14
we obtain the desired inequality. □
The inequality e0 (d − 1) ≥ 2e1 holds for an arbitrary 𝔪-primary ideal I of a
regular local ring (R, 𝔪) [143, Theorem 3.2].
EXAMPLE 4.9.17. Let 𝔪 = (x1 , . . . , xd ) and let I = 𝔪k . Then
kd d kd−1
( )
kn + d − 1
f (n) = = n + nd−1 + terms of lower degree,
d d! (d − 2)!2
e0 = kd and e1 = (d − 1)(kd − kd−1 )/2.
200 4. MONOMIAL RINGS AND OPTIMIZATION PROBLEMS

Exercises
4.9.18. Consider the ideal I = (x110 , x28 , x35 ) ⊂ R. Use CoCoA to verify that the
values of the Hilbert function f of I at n = 0, 1, 2, 3 are 0, 112, 704, 2176. Then use
interpolation to prove:
f (n) = `(R/I n ) = (200/3)n3 + 40x2 + (16/3)n.
4.9.19. Let I = (x2 , y3 , z5 , w6 , xy, z2 w, xz2 , y2 z2 ) ⊂ R = k[x, . . . , w]. Use the
methods of this section to prove:
`(R/I n ) = (7/2)n4 + (65/6)n3 + (21/2)n2 + (19/6)n; ∀ n ∈ ℕ.
Chapter 5
Edge Subrings

Here we study monomial subrings associated to graphs and their toric ideals
using Gröbner bases, graph theory, and Ehrhart rings. For this purpose we introduce
embeddings of graphs on closed surfaces, i.e., compact and without boundary, and
the central notion of a planar graph. A full description of the integral closure of
these subrings will be presented along with some connections with graph theory and
polyhedral geometry. The normalization of the Rees algebra of an edge ideal will be
examined. We introduce the family of ring graphs and show some characterizations.
Then we study the complete intersection property of toric ideals of edge subrings.

5.1. Planar graphs and embeddings

In this section we introduce the basic notions needed about embeddings of graphs
on surfaces. There is an excellent book on the embedding of graphs on surfaces
including the plane by B. Mohar and C. Thomassen [134], which we follow closely.

Embeddings of graphs
Let Y be a topological space. An arc in Y is the image
γ([0, 1]) = {γ(x) ∣ x ∈ [0, 1]}
of a continuous function γ : [0, 1] −→ Y. The arc is simple if γ is injective. The
arc α = γ([0, 1]) is said to join its endpoints γ(0) and γ(1). A simple closed arc is
defined analogously except that now we have γ(0) = γ(1). A topological space Y is
arcwise connected if any two elements of Y are connected by a simple arc in Y.
The existence of a simple arc between two points of Y determines an equiva-
lence relation whose equivalence classes are called arcwise connected components
(regions) of Y. A set A ⊂ Y separates Y if Y ∖ A is not arcwise connected. A face
of A ⊂ Y is an arcwise connected component of Y ∖ A. If Y is a Hausdorff space,
then the union of a finite number of arcs is compact and hence closed.
201
202 5. EDGE SUBRINGS

DEFINITION 5.1.1. A graph G is embedded in a topological space Y if the vertices

of G are distinct elements of Y and every edge of G is a simple arc connecting in
Y the two vertices which it joins in G, such that its interior is disjoint from other
edges and vertices.
An embedding of a graph G in the topological space Y is an isomorphism of
G with a graph G′ embedded in Y, in this case, G′ is said to be a representation
(drawing) of G in Y. If there is an embedding of G into Y we say that G can be
embedded into Y.

Planar Graphs
We review some basic properties of graphs which can be embedded in the (Eu-
clidean) plane ℝ2 . Such graphs are called planar. A graph G that is embedded in the
plane is called a plane graph . The point set of a plane graph is compact. Therefore
exactly one face of G (a region of ℝ2 ∖ G) is unbounded. It is called the unbounded
face of G (outer face, exterior face). If we want to avoid giving the unbounded face
a special status we may consider embedding the graph in the 2-sphere and since the
2-sphere minus a point is homeomorphic (steoreographic projection) to the plane,
then use this homeomorphism and its inverse to switch back and forth. So planar
graphs are precisely the graphs that can be drawn on the sphere.
A polygonal arc is a subset of ℝ2 which is the union of a finite number of straight
line segments { p + λ(q − p) ∣ 0 ≤ λ ≤ 1} (for p, q ∈ ℝ2 ) and is homeomorphic to
the closed unit interval [0, 1].
LEMMA 5.1.2. If Ω is an open arcwise connected set in the plane, then any two
points in Ω are joined by a simple polygonal arc in Ω.
PROOF. Let x, y ∈ Ω and let α be a simple arc joining x and y in Ω. For each
point z ∈ Ω there is an  > 0 such that the open disc Dz = D(z, ) is contained in Ω.
A continuous image of a compact set is compact, so α is compact in Ω. Therefore
its open cover {Dz ∣ z ∈ α} has a finite subcover {Dz1 , Dz2 , . . . , Dzk }. Now we can
construct a simple polygonal arc in the union Dz1 ∪ Dz2 ∪ ⋅ ⋅ ⋅ ∪ Dzk connecting x
and y. □
LEMMA 5.1.3. If G is a planar graph, then G has a representation in the plane
such that all edges are simple polygonal arcs.
PROOF. Let Γ be a plane graph isomorphic to G. Let p be some vertex of Γ, and
let D p be a closed disc with p as center such that D p intersects only those edges that
are incident to p. Furthermore assume that D p ∩ Dq = ∅ for every pair of distinct
vertices p, q in Γ. For each edge { p, q} of Γ let C pq be a segment of the edge { p, q}
such that C pq joins D p with Dq and has only its ends in common with D p ∪ Dq . We
can now redraw G so that we use all arcs C pq and such that the parts of the edges
 5. EDGE SUBRINGS 203

in the discs D p are straight line segments. Using Lemma 5.1.3 it is now easy to
replace each of the arcs C pq by a simple polygonal arc. □
THEOREM 5.1.4 (Jordan Curve, special case). If C is a simple closed polygonal
arc in the plane, then ℝ2 ∖ C consists of precisely two arcwise connected components
each of which has C as its boundary.
PROOF. Let P1 , P2 , . . . , Pn be the straight line segments of C. Assume without
loss of generality that none of P1 , P2 , . . . , Pn is horizontal. For each z ∈ ℝ2 ∖ C
denote by π(z) the number of segments Pi (1 ≤ i ≤ n) such that the horizontal right
half line in ℝ2 starting at z contains a point of Pi but does not contain the endpoint
of Pi that has the largest second coordinate. We let π̄(z) be π(z) reduced modulo 2.
If A is a polygonal arc in ℝ2 ∖ C, then it is easy to see that π̄(z) is constant on A.
By Lemma 5.1.3, π̄ is constant on arcwise connected components of ℝ2 ∖ C. The
points close to C must have different value of π̄ on each side. It follows that ℝ2 ∖ C
has at least two arcwise connected components.
Select a disk D such that D ∩ C is a straight line segment. If a, b, c are points
in ℝ2 ∖ C, then we can find three polygonal arcs in ℝ2 ∖ C starting at these points
and terminating in D, by first coming close to C and then follow C close enough
until we reach D. Two of the three arcs can be combined to get an arc in ℝ2 ∖ C
joining two of the points a, b, c. Therefore ℝ2 ∖ C has at most two arcwise connected
components. This argument also shows that every point on C is in the boundary of
each face of ℝ2 ∖ C. □
For the more general case of the Jordan Curve Theorem, where C is a closed
curve in the plane see [134].
DEFINITION 5.1.5. Given a graph H, we call a path 𝒫 an H-path if 𝒫 is non-
trivial and meets H exactly in its ends.
THEOREM 5.1.6. ([45, Proposition 3.1.2] and [96, Theorem 5.10]) Let G be a
graph. Then the following three conditions are equivalent
(a) G is 2-connected.
(b) G can be constructed from a cycle by successively adding H-paths to
graphs H already constructed.
(c) Every pair of vertices of G are joined by at least 2 vertex-disjoint paths.
THEOREM 5.1.7. (Euler’s formula, [45, Theorem 4.2.7]) Let G be a plane
connected graph with n vertices, q edges, and r faces. Then n − q + r = 2. If G is
2-connected, then each face has a cycle of G as boundary.
The cycles of a 2-connected plane graph G bounding the faces of G are said to
be facial. A triangulation or a quadrangulation is a connected plane graph with
polygonal edges such that each face boundary is a 3-cycle or a 4-cycle, respectively.
204 5. EDGE SUBRINGS

PROPOSITION 5.1.8. Let G be a plane graph with polygonal edges and n ≥ 4

vertices. Let q be the number of edges in G. Then:
(a) q ≤ 3n − 6 with equality if and only if G is a triangulation.
(b) If G has no triangle, then q ≤ 2n − 4 with equality if and only if G is a
quadrangulation.
PROOF. First assume that G is 2-connected. Let r denote the number of faces of
G. By Theorem 5.1.7 we have (i) n − q + r = 2. Theorem 5.1.7 implies that every
edge is in precisely two facial cycles. Therefore we have (ii) 3r ≤ 2q, with equality
if and only if G is a triangulation. Combining (i) and (ii) we get (a). Similarly we
get (b). If G is not 2-connected, then (a) or (b) applied to each block of G shows
that there is strict inequality in (a) or in (b). □
COROLLARY 5.1.9. 𝒦5 and 𝒦3,3 are non planar.
PROOF. Lemma 5.1.3 and Proposition 5.1.8 show these graphs have too many
edges to be planar. □
DEFINITION 5.1.10. A graph H is called a subdivision of a graph G if H arises
from G by replacing edges by paths. That is H is obtained by iteratively choosing
an edge {u, v}, introducing a new vertex w, deleting edge {u, v}, and adding edges
{u, w} and {w, v}.
EXAMPLE 5.1.11. The bipartite graph 𝒦2,3 and a subdivision

s s
@ @
@ @s
s
@
@s s
@@s
H
HH  H 
@
s
 H
@ H
s
 s

@ H @ H
s
@@s @@s

DEFINITION 5.1.12. Two graphs H1 and H2 are called homeomorphic if there

exists a graph G such that both H1 and H2 are subdivisions of G.
THEOREM 5.1.13. (Kuratowski; see [134]) A graph is planar if and only if it
does not contain a subdivision of 𝒦5 or a subdivision of 𝒦3,3 as a subgraph.
DEFINITION 5.1.14. A planar graph is outerplanar if it can be embedded in the
plane so that all its vertices lie on some face; it is usual to choose this face to be the
exterior face.
THEOREM 5.1.15. [96, Theorem 11.10] A graph is outerplanar if and only if it
has no subgraph homeomorphic to 𝒦4 or 𝒦2,3 except 𝒦4 ∖ {e}, where e is an edge.
 5. EDGE SUBRINGS 205

Let z = {vi , v j } be an edge of a graph G = (V, E). We denote by G/z the graph
obtained from G by contracting the edge z into a new vertex vz which becomes
adjacent to all former neighbors of vi and v j (edges of the form {vi , vk } and {v j , vk }
become multiple edges that are also deleted). Note that if G is a multigraph we may
have cycles of length one (loops) or two (pairs of multiple edges), hence the notion
of edge contraction is simpler in multigraphs, since we only remove the contracted
edge.
PROPOSITION 5.1.16. If V1 , V2 ⊂ E(G) are disjoint edge subsets of a graph G,
then
(G/V1 ) ∖ V2 = (G ∖ V2 )/V1 .
PROOF. It is left as an exercise. □
DEFINITION 5.1.17. A graph H is a minor of a graph G if H is isomorphic to a
graph obtained from a subgraph F of G by contracting a set of edges A ⊂ E(F).
Hence every minor can be obtained from G by successively contracting edges
and deleting edges and isolated vertices. The order in which these operations are
performed is unimportant. Every subgraph of a graph is also its minor, in particular
every graph is its own minor. See [45] for further information about minors.
PROPOSITION 5.1.18 (Wagner; see [134]). A graph is planar if and only if it
does not contain neither 𝒦5 or 𝒦3,3 as a minor.

Exercises
5.1.19. Show that every graph has an embedding into ℝ3 .
5.1.20. Let G be a outerplanar graph with n vertices and q edges. If G has no
triangles, then q ≤ (3n − 4)/2.

5.2. The subring associated to a graph

Let R⊕= k[x1 , . . . , xn ] be a polynomial ring over a field k with the standard grading
R= ∞ i=0 Ri induced by deg(xi ) = 1 for all i and let G be a graph on the vertex set
V = {v1 , . . . , vn }.
DEFINITION 5.2.1. The monomial subring or edge subring of the graph G,
denoted by k[G], is the k-subalgebra of R generated by the monomials corresponding
to the edges of G:
k[G] = k[{ xi x j ∣ vi is adjacent to v j }] ⊂ R.
206 5. EDGE SUBRINGS

Recall that k[G] is a standard k-algebra with the normalized grading k[G]i =
k[G] ∩ R2i . Let F = { f1 , . . . , fq } be the set of all monomials xi x j such that vi is
adjacent to v j . Since the elements in k[G] are polynomial expressions in F with
coefficients in k, there is a graded epimorphism of k-algebras
ϕ : B = k[t1 , . . . , tq ] −→ k[G], ti 7−→ fi ,
where B is a polynomial ring graded by deg(ti ) = 1 for all i. The kernel of ϕ,
denoted by P(G), is a graded ideal of B called the toric ideal of k[G] with respect to
f1 , . . . , fq .
DEFINITION 5.2.2. Let w = {v0 , v1 , . . . , vr = v0 } be an even closed walk in G
such that fi = xi−1 xi . The binomial
tw = t1 t3 ⋅ ⋅ ⋅ tr−1 − t2 t4 ⋅ ⋅ ⋅ tr
is in P(G). We say that tw is the binomial associated to w.
DEFINITION 5.2.3. A monomial f = xi x j in R is said to be an edge generator of
G if vi is adjacent to v j .
Notation For use below consider the set ℐ s of all non-decreasing sequences α =
(i1 , . . . , i s ) of length s. If a1 , . . . , aq is a sequence and α = (i1 , . . . , i s ) is an element
of ℐ s , then we set aα = ai1 ⋅ ⋅ ⋅ ais .
PROPOSITION 5.2.4 ([184]). If P = P(G) is the toric ideal of k[G], then
(a) P = ({tw ∣ w is an even closed walk}), and
(b) P = ({tw ∣ w is an even cycle}) if G is bipartite.
PROOF. The toric ideal P is graded and P = B ⋅ ∪∞
( )
s=2 P s . First we set
ℬ = {tw ∣ w is an even closed walk}.
One clearly has (ℬ ) ⊂ P. To prove the other inclusion we use induction on s. First
recall that P is a binomial ideal by Proposition 4.3.4. If s = 2, it is enough to
note that the binomials in P2 come from squares. Assume P s−1 ⊂ (ℬ ). To show
P s ⊂ (ℬ ), take tα − tβ in P s , where α = (i1 , . . . , i s ) and β = ( j1 , . . . , j s ). Define G1
as the subgraph of G having vertex set
V1 = { xi ∈ V ∣ xi divides fα }
and edge set
E1 = {{ x, y} ∈ E(G)∣ f` = xy for some ` ∈ {i1 , . . . , i s , j1 , . . . , j s }}.
Note that if fi1 ⋅ ⋅ ⋅ fim = f j1 ⋅ ⋅ ⋅ f jm for some m < s and for some ordering of the
generators then tα − tβ ∈ (ℬ ); to prove it, notice that tα − tβ can be written as
tα − tβ = tim+1 ⋅ ⋅ ⋅ tis (ti1 ⋅ ⋅ ⋅ tim − t j1 ⋅ ⋅ ⋅ t jm )
+t j1 ⋅ ⋅ ⋅ t jm (tim+1 ⋅ ⋅ ⋅ tis − t jm+1 ⋅ ⋅ ⋅ t js )
 5. EDGE SUBRINGS 207

and use induction hypothesis. Now we can assume fi1 ⋅ ⋅ ⋅ fim ∕= f j1 ⋅ ⋅ ⋅ f jm for m < s
and for any re-ordering of the fi ’s. Observe that this forces G1 to be connected.
Take v0 ∈ V1 . Since fα = fβ , it is easy to check that after re-ordering we can write
fik = v2k−2 v2k−1 and f jk = v2k−1 v2k , where 1 ≤ k ≤ s and v0 = v2s . Therefore
the even closed walk w = {v0 , . . . , v2n } satisfies tw = tα − tβ and the induction is
complete. This proves part (a).
Assume G is bipartite. The second part of the proposition can be proved
similarly if we notice that this time the condition fi1 ⋅ ⋅ ⋅ fim ∕= f j1 ⋅ ⋅ ⋅ f jm for m < s
and for any re-ordering of the fi ’s implies that the graph G1 is an even cycle. □
COROLLARY 5.2.5. If G is a graph and f = tα − tβ is a primitive binomial in
P(G), then f = tw for some even closed walk w of G.
PROOF. It follows from the proof of Proposition 5.2.4. □
COROLLARY 5.2.6. If G is a bipartite graph and f = tα − tβ is a primitive
binomial in P(G), then f = tw for some even cycle w of G.
PROOF. It follows from the proof of Proposition 5.2.4. □
LEMMA 5.2.7. Let G be a graph and let P = P(G) be the toric ideal of k[G]. If
f is a polynomial in any reduced Gröbner bases of P, then
(a) f is a primitive binomial and f = tw for some even closed walk w of the
graph G.
(b) If G is bipartite, then f is primitive and f = tw for some even cycle w of
the graph G.
PROOF. Let 𝒜 be a reduced Gröbner basis of P and take f ∈ 𝒜. Using
Corollary 4.3.5 together with the fact that normalized reduced Gröbner bases are
uniquely determined, we obtain f = tα − tβ . By Lemma 4.3.9 f is primitive, thus
by Corollary 5.2.5 and Corollary 5.2.6 f has the required form. □
PROPOSITION 5.2.8. If G is a graph, then the set
{tw ∣ tw is primitive and w is an even closed walk}
is a universal Gröbner basis of P(G).
PROOF. It follows from Lemma 5.2.7. □
PROPOSITION 5.2.9. If G is a bipartite graph, then the set
{tw ∣ w is an even cycle}
is a universal Gröbner basis of P(G).
PROOF. It follows from Lemma 5.2.7. □
DEFINITION 5.2.10. A chord of a cycle c of a graph G is any edge of G joining
two non adjacent vertices of c. A cycle without chords is called primitive.
208 5. EDGE SUBRINGS

PROPOSITION 5.2.11. If a is a cycle and ta ∈ (tc ∣ c is a cycle, c ∕= a), then a has

a chord.
PROOF. If ta ∈ (tc ∣ c is a cycle, c ∕= a), then each term of ta must be divisible by
a term of tc for some c ∕= a. □
PROPOSITION 5.2.12. If G is a bipartite graph, then P(G) is a prime ideal
minimally generated by the set {tc ∣ c is a primitive cycle of G}.
PROOF. It follows from Proposition 5.2.11. □
The cycle space of a graph. Let us describe the cycle space of a graph G over
the two element field F = ℤ2 . We denote the edge set of G by
E(G) = { f1 , . . . , fq }
and the vertex set of G by
V = { x1 , . . . , xn } .
Let C0 and C1 denote the vector spaces over F of 0-chains and 1-chains respectively.
Recall that a 0-chain of G is a formal linear combination
∑
ai xi
of vertices and a 1-chain is a formal linear combination
∑
bi fi

of edges, where ai ∈ F and bi ∈ F. The boundary operator ∂ is the linear

transformation defined by
∂( fk ) = ∂({ xi , x j }) = xi + x j .
A cycle vector is a 1-chain of the form f1 + ⋅ ⋅ ⋅ + fr where f1 , . . . , fr are the edges
of a cycle of G. The cycle space 𝒵 (G) of G over F is equal to ker(∂). The vectors
in 𝒵 (G) can be regarded as a set of edge-disjoint cycle vectors. A cycle basis for
G is a basis for 𝒵 (G) which consist entirely of cycle vectors, such a basis can be
constructed as follows:
REMARK 5.2.13. If G is connected, then G has an spanning tree T . The subgraph
of G consisting of T and any edge in G not in T has exactly one cycle, the collection
of all cycle vectors of cycles obtained in this way form a cycle basis for G. In
particular
dimF 𝒵 (G) = q − n + 1.
DEFINITION 5.2.14. If G is a graph, the number dimF 𝒵 (G) is called the cycle
rank of G and is denoted by rank(G).
 5. EDGE SUBRINGS 209

LEMMA 5.2.15. Let G be a connected graph and let 𝒵e (G) be the subspace of
𝒵 (G) of all cycle vectors of G with an even number of terms. Then
{
q − n + 1 if G is bipartite, and
dimF 𝒵e (G) =
q−n otherwise.
PROOF. Let c1 , . . . , c` , c`+1 , . . . , cm be a cycle basis for G, where c1 , . . . , c` are
even cycles and c`+1 , . . . , cm odd cycles. It is clear that a bases for 𝒵e (G) is given
by {c1 , . . . , c` , c`+1 + cm , . . . , cm−1 + cm }. □
DEFINITION 5.2.16. Let G be a simple graph. The incidence matrix of G,
denoted AG , is the integer matrix with entries in {0, 1} whose columns are the
vectors ei + e j such that xi is adjacent to x j .
LEMMA 5.2.17. If If G is a connected graph with n vertices and AG its incidence
matrix, then rank(AG ) = n − 1 if G is bipartite and rank(AG ) = n if G is non bipartite.
PROOF. Pick a spanning tree T of G. Clearly the rank of AT is n − 1. Assume G
is bipartite. It suffices to observe that any edge e of G not in T lies in an even cycle
C such that the edges of C different from e lie in T and that vectors that correspond
to the edges of an even cycle are linearly dependant. Now we assume that G is non
bipartite. It suffices to observe that there is an edge e of G not in T that lies in an
odd cycle C (see Exercise 5.2.59) whose edges lie in E(T ) ∪ {e} and that the set of
vectors that correspond to E(T ) ∪ {e} are linearly independent. □
PROPOSITION 5.2.18 ([184]). Let G be a connected graph and let P be the toric
ideal of the edge subring k[G]. Then
ht(P) = dimF 𝒵e (G).
PROOF. It follows from Lemmas 5.2.17, 5.2.15, and Proposition 4.3.20. □
COROLLARY 5.2.19. If G is a connected graph with n vertices and k[G] is its
edge subring, then
{
n if G is not bipartite, and
dim(k[G]) =
n − 1 otherwise.
PROOF. It follows from Proposition 5.2.18 and Lemma 5.2.15. □

Notation The number of primitive cycles of a graph G will be denoted by

frank(G).
PROPOSITION 5.2.20. If G is a graph, then 𝒵 (G) is generated by primitive cycles.
In particular rank(G) ≤ frank(G).
PROOF. Let c1 , . . . , cr be a basis for the cycle space of G. Notice that if some c j
has a chord, we can write c j = c′j + c′′j , where c′j and c′′j are cycle vectors of smaller
length that c j . Applying this observation repeatedly we get that the cycle space is
generated by primitive cycles. □
210 5. EDGE SUBRINGS

COROLLARY 5.2.21. Let G be a graph. The following are equivalent:

(a) rank(G) = frank(G).
(b) the set of primitive cycles is a basis for the cycle space of G.
(c) the set of primitive cycles is linearly independent.
DEFINITION 5.2.22. Let G be a graph. The toric ideal P(G) is called a complete
intersection if it can be generated by r binomials, where r is the height of P(G).
PROPOSITION 5.2.23 ([154]). Let G be a connected bipartite graph with n vertices
and q edges. Then q − n + 1 = frank(G) if and only if the toric ideal P of k[G] is a
complete intersection.
PROOF. ⇒) By Proposition 5.2.12 the toric ideal P is minimally generated by
the binomials corresponding to primitive cycles. Since the height of P is precisely
q − n + 1, see Proposition 5.2.18, we get that P is a complete intersection.
⇐) If P is a complete intersection, then the minimum number of generators of P
is equal to the height of P, which in our case is q − n + 1. Thus by Proposition 5.2.12
we obtain q − n + 1 = frank(G), as required. □
LEMMA 5.2.24. Let G be a graph and let G1 , . . . , Gr be its blocks. Then
rank(G) = frank(G) if and only if rank(Gi ) = frank(Gi ) for all i.
PROOF. It follows rapidly using the maximality of the blocks and applying
Corollary 5.2.21. □
PROPOSITION 5.2.25. If G is an outerplanar graph, then
rank(G) = frank(G).
PROOF. By induction on q, the number of edges of G. Notice that G is
outerplanar if and only if its blocks are outerplanar. Thus by Lemma 5.2.24 we
may assume that G is 2-connected. Fix an embedding of G in the plane so that
all its vertices lie on the exterior face. Pick an edge e which is on the boundary
of the exterior face and such that e belongs to a primitive cycle c. To apply the
induction hypothesis notice that e cannot belong to a primitive cycle different from
c, because this would imply that the graph G contains a subdivision of 𝒦2,3 , which
is impossible by Theorem 5.1.15. Thus the number of edges of G ∖ {e} is q − 1
and the number of primitive cycles of G ∖ {e} is one less than that of G. Hence by
induction we get rank(G) = frank(G). □
EXAMPLE 5.2.26. The graph below is planar but not outerplanar and it satisfies
rank(G) = frank(G) = 3. Notice that not all the regions of this embedding are
bounded by primitive cycles.
s

s s s

s s s
@
@
@s
 5. EDGE SUBRINGS 211

The family of graphs satisfying the equality rank(G) = frank(G) can be con-
structed. In order to describe this family we need to introduce a new notion. The
structure of 2-connected graphs (see Theorem 5.1.6) suggest the following:
DEFINITION 5.2.27. A graph G is a ring graph if each block of G which is not
a bridge can be constructed from a cycle by successively adding H-paths of length
at least 2 that meet graphs H already constructed in two adjacent vertices.
In Example 5.2.26 a ring graph is shown. Families of ring graphs include trees
and cycles. These graphs are planar by construction.
REMARK 5.2.28. Let G be a 2-connected ring graph and let C be a fixed primitive
cycle of G, then G can be constructed from C by successively adding H-paths of
length al least 2 that meet graphs H already constructed in two adjacent vertices.
LEMMA 5.2.29. [1, Lemma 7.78, p. 387] Let G be a graph with vertex set V. If
G is 2-connected and deg(v) ≥ 3 for all v ∈ V, then G contains a subdivision of 𝒦4
as a subgraph.
LEMMA 5.2.30. Let G be a graph. If rank(G) = frank(G) and x, y are two non
adjacent vertices of G, then there are at most two vertex disjoint paths joining x and
y.
PROOF. Assume that there are tree vertex disjoint paths joining x and y:
𝒫1 = { x, x1 , . . . , xr , y}, 𝒫2 = { x, z1 , . . . , zt , y},
𝒫3 = { x, y1 , . . . , y s , y},
where r, s, t are greater than or equal to 1. We may assume that the sum of the
lengths of the 𝒫i ’s is minimal. Consider the cycles
c1 = { x, x1 , . . . , xr , y, zt , . . . , z1 , x},
c2 = { x, z1 , . . . , zt , y, y s , . . . , y1 , x},
c3 = { x, x1 , . . . , xr , y, y s , . . . , y1 , x}.
Thus we are in the following situation:

x1 c3 xr
s s s s s
c1
z1s zst y
x s s s

c2
y1 s s s s s ys
212 5. EDGE SUBRINGS

Observe that, by the choice of the 𝒫i ’s, a chord of the cycle c1 (resp. c2 , c3 ) must
join xi and z j (resp. zi and y j , xi and y j ) for some i, j. Therefore we can write
n1
∑ n2
∑ n3
∑
c1 = ai , c2 = bi , c3 = di
i=1 i=1 i=1
where a1 , . . . , an1 , b1 , . . . , bn2 , d1 , . . . , dn3 are distinct cycle vectors of primitive cy-
cles of G and ci is the cycle vector corresponding to ci . Thus from the equality
c3 = c1 + c2 we get that the set of cycle vectors of primitive cycles is linearly
dependent, a contradiction to Corollary 5.2.21. □
DEFINITION 5.2.31. A graph G has the primitive cycle property (PCP) if any
two primitive cycles intersect in at most one edge.
LEMMA 5.2.32. Let G be a graph. If rank(G) = frank(G), then G has the
primitive cycle property.
PROOF. Let c1 , c2 be two distinct primitive cycles. Assume that c1 and c2
intersect in at least two edges. Thus c1 and c2 must intersect in two non adjacent
vertices. Hence there are two non adjacent vertices x, y in c1 and a path 𝒫 =
{ x, x1 , . . . , xr , y} of length at least two that intersect c1 in exactly the vertices x, y:

x1 𝒫 xr
s s s s s

x s s s s sy
c1
s s s s s

This contradicts Lemma 5.2.30. Hence c1 and c2 have at most one edge in common.
□
The converse of Lemma 5.2.32 fails, as a counterexample consider the complete
graph 𝒦4 .
THEOREM 5.2.33. If G is a graph, then rank(G) = frank(G) if and only if G is a
ring graph.
PROOF. Let G1 , . . . , Gr be the blocks of G. Notice that rank(G) = frank(G) if
and only if rank(Gi ) = frank(Gi ) for all i, see Lemma 5.2.24. Thus we may assume
that G is 2-connected.
⇒) The proof is by induction on n, the number of edges of G. The result is clear
for any cycle. We claim that G has a least one vertex of degree 2. If deg(v) ≥ 3
for all v ∈ V(G), then by Lemma 5.2.29 there is a subgraph H ⊂ G which is a
subdivision of 𝒦4 . If 𝒦4 is a subgraph of G, then G has four distinct triangles whose
 5. EDGE SUBRINGS 213

cycle vectors are linearly dependent, a contradiction to Corollary 5.2.21. If 𝒦4 is

not a subgraph of G, then there are two vertices x, y in H which are non adjacent in
G. Since H is a strict subdivision of 𝒦4 it follows rapidly that there are at least three
non adjacent paths joining x and y, a contradiction to Lemma 5.2.30. Let v ∈ V(G)
be a vertex of degree 2. By Lemma 5.2.32 there is a unique primitive cycle c of
G containing v. Thus the graph G′ = G ∖ {v} satisfies rank(G′ ) = frank(G′ ) and
consequently G′ is a ring graph, then using Remark 5.2.28 it is seen that G is a ring
graph as well.
⇐) By induction on the number of vertices it is not hard to see that any ring
graph G satisfies the equality rank(G) = frank(G). □
An immediate consequence of Theorem 5.2.33 is:
COROLLARY 5.2.34. Let G be a graph. If rank(G) = frank(G), then G is planar.
PROPOSITION 5.2.35. Let G be a bipartite graph and let G1 , . . . , Gr be its blocks,
then
k[G] ≃ k[G1 ] ⊗k ⋅ ⋅ ⋅ ⊗k k[Gr ] and P(G) = (P(G1 ), . . . , P(Gr )).
PROOF. Since G is bipartite it follows from Exercise 5.2.61. □
DEFINITION 5.2.36. A graph G is called a complete intersection if the toric ideal
P(G) of k[G] is a complete intersection.
The complete intersection property of the edge subring of a bipartite graph
was first studied in [48], and later in [82, 154]. This property is independent of k
according to [135, Theorem 3.9].
PROPOSITION 5.2.37 ([154]). If G is a bipartite graph, then G is a complete
intersection if and only if rank(G) = frank(G).
PROOF. Let G1 , . . . , Gr be the blocks of G. By Proposition 5.2.35 G is a c.i. if
and only if Gi is a c.i. for all i. On the other hand by Proposition 5.2.23 Gi is a c.i.
if and only if rank(Gi ) = frank(Gi ). To finish the proof apply Lemma 5.2.24. □
COROLLARY 5.2.38. If G is a bipartite graph and G is a complete intersection,
then G is a planar graph.
PROOF. It follows from Proposition 5.2.37 and Corollary 5.2.34. □
COROLLARY 5.2.39. If G is a bipartite graph, then G is a complete intersection
if and only if G is a ring graph.
PROOF. By Proposition 5.2.37 G is a complete intersection if and only if
rank(G) = frank(G) and the result follows from Theorem 5.2.33. □
DEFINITION 5.2.40. Let g1 = tα1 − tβ1 , . . . , gr = tαr − tβr be a finite sequence of
homogeneous binomials of degree at least 2 in the polynomial ring B = k[t1 , . . . , tq ].
We say that ℬ = {g1 , . . . , gr } is a foliation if the following conditions are satisfied:
214 5. EDGE SUBRINGS

(a) tαi and tβi are square-free for all i,

(b) supp(tαi ) ∩ supp(tβi ) = ∅ for all i, and
j
(c) ∣(∪i=1 Ci ) ∩ C j+1 ∣ = 1 for 1 ≤ j < r, where Ci = supp(gi ) for all i.
The notion of foliations is related to complete intersections as the next results
shows.
PROPOSITION 5.2.41. If ℬ = {g1 , . . . , gr } is a foliation, then the binomial ideal
I = (ℬ ) is a complete intersection.
PROOF. By the constructive nature of ℬ we can order the variables t1 , . . . , tq
so that the leading terms of g1 , . . . , gr , with respect to the lex order, are relatively
prime. Then using Exercise 4.1.9, it is seen that ℬ is a Gröbner basis. Since
dim(B/I) = dim(B/(in(g1 ), . . . , in(gr ))),
we obtain that the height of I is equal to r, as required. □
COROLLARY 5.2.42. If G is a 2-connected bipartite graph with at least four
vertices, then the toric ideal P(G) is a complete intersection if and only if it is
generated by a foliation.
PROOF. It follows from Corollary 5.2.39, using the constructive nature of a ring
graph. □
PROPOSITION 5.2.43. If G is a graph, then rank(G) = frank(G) if and only if G
satisfies PCP and G does not contain a subdivision of 𝒦4 as a subgraph.
PROOF. It follows, along the lines of the proof Theorem 5.2.33, using induction
on the number of vertices of G. □
THEOREM 5.2.44 ([119]). The edge subring k[G] of a bipartite graph G is a
complete intersection if and only if G is planar and any two primitive cycles of G
have at most one common edge.
PROOF. ⇒) It follows from Lemma 5.2.32 and Corollary 5.2.38.
⇐) See [119, Theorem 3.5]. □
Let us summarize some of the results obtained thus far. The following implica-
tions hold for any graph G:
⇒ planar
outerplanar ⇒ ring graph ⇔ PCP + contains no
⇕ subdivision of 𝒦4
rank = frank ⇒ PCP
If G is bipartite, then
ring graph ⇔ complete intersection ⇔ PCP + planar.
 5. EDGE SUBRINGS 215

Toric ideals of oriented graphs. Let G be a connected simple graph with n

vertices and q edges and let 𝒪 be an orientation of the edges of G, i.e., an assignment
of a direction to each edge of G. Thus 𝒟 = (G, 𝒪) is an oriented graph.
To each oriented edge e = (xi , x j ) of 𝒟, we associate the vector ve defined as
follows: the ith entry is −1, the jth entry is 1, and the remaining are zero. The
incidence matrix A𝒟 of 𝒟 is the n × q matrix with entries in {0, ±1} whose columns
are the vectors of the form ve , with e an edge of 𝒟. For simplicity of notation we
set A = A𝒟 . The set of column vectors of A will be denoted by 𝒜 = {v1 , . . . , vq }.
Let M[A] be the vector matroid on 𝒜 over the field ℚ, whose independent sets are
the independent subsets of 𝒜.
Recall that there is a one to one correspondence:
Circuits of ker(A) −→ Circuits of M[A] = cycles of G
given by

α = (α1 , . . . , αq ) −→ C(α) = {vi ∣ i ∈ supp(α)}.

LEMMA 5.2.45. The circuits of M[A] are precisely the cycles of G, A is totally
unimodular, and rank(A) = n − 1
PROOF. It follows from [85, pp. 343-344] and [150, p. 274]. □
In contrast if we consider the incidence matrix AG of G, then we have the
following description of the circuits.
PROPOSITION 5.2.46. The circuit of the vector matroid M[AG ] are exactly the
even cycles, or two edge disjoint odd cycles meeting at exactly one vertex, or two
vertex disjoint odd cycles joined by an arbitrary path.
PROOF. It will follow from Proposition 5.4.1. □
Consider the subring k[𝒟] := k[xv1 , . . . , xvq ] ⊂ k[x1±1 , . . . , xn±1 ] of the oriented
graph 𝒟. There is an epimorphism of k-algebras
ϕ : B = k[t1 , . . . , tq ] −→ k[𝒟], ti 7−→ xvi ,
where B is a polynomial ring. The kernel of ϕ, denoted by P𝒟 , is called the toric
ideal of k[𝒟]. Notice that P𝒟 is no longer a graded ideal of B. However if the
vertices of 𝒟 can be linearly order such that all edges of 𝒟 are of the form (xi , x j )
with i < j, then P𝒟 is homogeneous with respect to the positive grading induced by
deg(t` ) = j − i, where t` maps to xi x j and (xi , x j ) is an edge of 𝒟 (see Exercise 2.5.18).
The toric ideal P𝒟 is a prime ideal of height q − n + 1 generated by binomials
and k[𝒟] is a normal domain. Thus any minimal generating set of P𝒟 must have at
least q − n + 1 elements.
216 5. EDGE SUBRINGS

If 0 ∕= α ∈ ker(A) ∩ ℤn we associate the binomial

tα = t α + − t α − .
Notice that tα ∈ P𝒟 .
Given a cycle c, we split c in two disjoint sets of edges c+ and c− , where c+
consists of the edges of c oriented clockwise and c− = c ∖ c+ . The binomial
∏ ∏
tc = ti − ti
vi ∈c+ vi ∈c−

belongs to P𝒟 . If c+ = ∅ or c− = ∅ we set
∏ ∏
ti = 1 or ti = 1.
vi ∈c+ vi ∈c−

DEFINITION 5.2.47. The toric ideal P𝒟 is called a complete intersection if P𝒟

can be generated by q − n + 1 binomials.
PROPOSITION 5.2.48. P𝒟 is generated by the set of all binomials tc such that c
is a cycle and this set is a universal Gröbner basis.
PROOF. Let 𝒰𝒟 be the set of all binomials of the form tα such that α is a circuit
of ker(A). Since A is totally unimodular, by Corollary 4.3.13 the set 𝒰𝒟 form a
universal Gröbner basis of P𝒟 . Notice that the circuits of ker(A) are in one to one
correspondence with the circuits of the vector matroid M[A]. To complete the proof
it suffices to observe that the circuits of M[A] are precisely the cycles of G, see
Lemma 5.2.45. □
COROLLARY 5.2.49. P𝒟 is generated by the set of binomials corresponding to
primitive cycles.
PROOF. If a cycle c is not primitive, then it has a chord dividing c into two cycles
c1 and c2 . It is not hard to verify that tc is a linear combination of tc1 and tc2 . □
In general the complete intersection property of P𝒟 depends on the orientation
of G. However we have:
COROLLARY 5.2.50. If G is a ring graph, then P𝒟 is a complete intersection for
any orientation of G.
PROOF. By Corollary 5.2.49, P is generated by q − n + 1 binomials. □
A problem here is to characterize the graphs with the property that P𝒟 is a
complete intersection for all orientations of G. Apart from ring graphs, it is easy
to verify that 𝒦4 has this property. In [146] it is shown that for a given graph G
there is always an acyclic orientation of G such that P𝒟 is a complete intersection
generated by the cycle basis of a spanning tree, where 𝒟 = (G, 𝒪). Recall that an
oriented cycle of 𝒟 is a cycle with all its arrows oriented in the same direction and
that 𝒟 is acyclic if it has not oriented cycles.
 5. EDGE SUBRINGS 217

Exercises
5.2.51. Let G be a graph with q edges and P(G) the toric ideal of the edge
subring k[G]. If f is a primitive binomial in P(G) prove that deg( f ) ≤ q if q is even
and deg( f ) ≤ q − 1 otherwise.
5.2.52. Let G be a graph. If f = tα − tβ is a primitive binomial in the toric ideal
P(G), then the entries of α satisfy αi ≤ 2 for all i.
5.2.53. Let G be a graph. Then G is planar (resp. outerplanar) if and only if
each block of G is planar (resp. outerplanar).
5.2.54. Let G be a ring graph. Prove that G is outerplanar if and only if G does
not contains a subdivision of 𝒦2,3 as a subgraph.
5.2.55. Let G be a graph. Prove that G satisfies PCP if G does not contains a
subdivision of 𝒦2,3 as a subgraph.
5.2.56. Let G be a graph. Prove that if G satisfies PCP, then G does not contains
a subdivision of 𝒦2,3 as an induced subgraph.
5.2.57. Let G be a ring graph with n ≥ 3 vertices and q edges. If G is 2-connected
and has no triangles, then q ≤ 2(n − 2).
5.2.58. Prove that any connected graph has a spanning tree.
5.2.59. Prove that a graph G is bipartite if and only if every cycle in some cycle
basis of G is even.
5.2.60. If G is a connected non bipartite graph with n vertices, prove that there
is a connected subgraph H of G with n vertices and n edges and with a unique cycle
of odd length.
5.2.61. Let G be a graph that can be written as G = G1 ∪ G2 , where G1 and
G2 are subgraphs with at most one vertex in common. If G1 is bipartite, then
P(G) = (P(G1 ), P(G2 )) and
k[G] ∼
= k[G1 ] ⊗k k[G2 ].
5.2.62. Let G be a connected bipartite ring graph. Prove that P(G) is minimally
generated by a Gröbner basis consisting of binomials.
5.2.63. Let G be a bipartite graph. If G is a subdivision of 𝒦5 or a subdivision
𝒦3,3 prove that the binomials of P(G) that correspond to primitive cycles do not
form a regular sequence.
5.2.64. Prove that any subdivision of 𝒦3,3 has cycle rank equal to six.
218 5. EDGE SUBRINGS

5.2.65. Let G be a simple graph and let A be its incidence matrix. Let M[A]
be the vector matroid of A. If G is not bipartite, prove that the bases of M[A] are
precisely the spanning subgraphs of G whose connected components are unicyclic
graphs with an odd cycle. If G is bipartite, prove that the bases of M[A] are the
spanning trees of G.
5.2.66. Let 𝒟 be an acyclic oriented graph with q edges and n vertices. If P𝒟 is
a complete intersection, then P𝒟 is generated by q − n + 1 binomials corresponding
to primitive cycles of 𝒟.

5.3. Incidence matrices

Let G be a graph simple with vertex set V = { x1 , . . . , xn } and with edge set E =
{z1 , . . . , zq }. If zi = { xi j , xik } define vi = ei j + eik , where ei is the ith canonical vector
in ℝn . The incidence matrix AG associated to G is the n × q integer matrix whose
columns are the vectors v1 , . . . , vq .
EXAMPLE 5.3.1. Consider a triangle G with vertices x1 , x2 , x3 . The incidence
matrix of G is: ⎛ ⎞
1 0 1
AG = ⎝ 1 1 0 ⎠ ,
0 1 1
with the vectors v1 = e1 + e2 , v2 = e2 + e3 , and v3 = e1 + e3 corresponding to the
edges z1 = { x1 , x2 }, z2 = { x2 , x3 }, and z3 = { x1 , x3 }.
REMARK 5.3.2. Let A be a square submatrix of AG . In [92] it is shown that
either det(A) = 0 or det(A) = ±2k , for some integer k such that 0 ≤ k ≤ τ0 , where
τ0 is the maximum number of vertex disjoint odd cycles in G. Moreover for any
such value of k there exists a minor equal to ±2k .
PROPOSITION 5.3.3. If A is the incidence matrix of a bipartite graph G, then A
is totally unimodular.
PROOF. By induction on the number of columns of A. Let (V1 , V2 ) be a
bipartition of G and let B be a square submatrix of A. If a column of B has at most
one entry equal to 1, then by induction det(B) = 0, ±1. Thus we may assume that
all the columns of B have two entries equal to 1. Hence
∑ ∑
Fi = Fi = (1, . . . , 1),
xi ∈V1 xi ∈V2

where x1 , . . . , xn are the vertices of G and F1 , . . . , Fr are the rows of B. Therefore

the rows of B are linearly dependent and det(B) = 0. □
 5. EDGE SUBRINGS 219

Recall that a graph G is bipartite if all its cycles are of even length, see Propo-
sition 2.1.2. Thus any tree and in particular any vertex is a bipartite graph. The
number of bipartite connected components of G will be denoted by c0 , and the num-
ber of non bipartite connected components will be denoted by c1 . Thus c = c0 + c1
is the total number of components of G.
LEMMA 5.3.4 ([92]). If G is a graph with n vertices and A = AG is its incidence
matrix, then rank(A) = n − c0 .
PROOF. Let G1 , . . . , Gc be the connected components of G, and ni the number of
vertices of Gi . After permuting the vertices we may assume that A is a “diagonal”
matrix A = diag(AG1 , . . . , AGc ), where AGi is the incidence matrix of Gi . By
Lemma 5.2.17 the rank of AGi is equal to ni − 1 if Gi is bipartite, and is equal to ni
otherwise. Hence the rank of A is equal to n − c0 . □
For use below we define the matrix B by:
( )
v1 ⋅ ⋅ ⋅ vq
B= .
1 ⋅⋅⋅ 1
Observe that rank(B) = rank(A) because the last row of B is a linear combination of
the first n rows.
For simplicity we keep the notation introduced above throughout the rest of this
section. We shall assume, if need be, that the graph G has no isolated vertices.
LEMMA 5.3.5. If G is a unicyclic graph with a unique odd cycle and r is the
rank of B, then ∆r (B) = 1.
PROOF. If G is an odd cycle of length n, then the matrix B′ obtained from B by
deleting any of its first n rows has determinant ±1. The matrix B′ is in fact totally
unimodular by Theorem 5.8.14.
We proceed by induction. If G is not an odd cycle, then G has a vertex x j of
degree 1. Thus the jth row of A has exactly one entry equal to 1, say a jk = 1.
Consider the matrix C obtained from B by deleting the jth row and the kth column.
Thus using induction ∆r−1 (C) = 1, which gives ∆r (B) = 1. □
LEMMA 5.3.6. Let G be a graph with connected components G0 , . . . , Gm . If G0
is non bipartite, then there exists a subgraph H of G with connected components
H0 , . . . , Hm such that H0 is a spanning unicyclic subgraph of G0 with a unique odd
cycle and Hi is a spanning tree of Gi for all i ≥ 2.
PROOF. The existence of H2 , . . . , Hm is clear because any connected graph has
a spanning tree, see Exercise 5.2.58. Take a spanning tree T 0 of G0 . For each
edge e of G0 not in T 0 , the graph T 0 ∪ {e} has exactly one cycle. According to
Remark 5.2.13 the set Z(T ) of such cycles form a basis for the cycle space of G0 .
Since a graph is bipartite if and only if every cycle in some cycle basis is even, see
220 5. EDGE SUBRINGS

Exercise 5.2.59, there is an edge e of G0 such that T 0 ∪ {e} is a unicyclic connected

graph with a unique odd cycle. □
PROPOSITION 5.3.7. If G is a graph with exactly one non bipartite connected
component and r = rank(B), then ∆r (B) = 1.
PROOF. Let G0 , . . . , Gm be the components of G, with G0 non bipartite, and let
H be as in Lemma 5.3.6. If A′ is the incidence matrix of H, then using that G and
H have the same number of non bipartite components one has rank(A) = rank(A′ ).
Therefore ∆r (B) = 1 by Lemma 5.3.5. □
PROPOSITION 5.3.8. Let G be a non bipartite graph and let A be its incidence
matrix. If A has rank r, then ∆r (A) = 2∆r (B).
PROOF. Set ℬ = {(v1 , 1), . . . , (vq , 1)}. It suffices to prove that there is an exact
sequence of groups
ϕ ψ
0 −→ T (ℤn+1 /ℤℬ ) −→ T (ℤn /ℤ𝒜) −→ ℤ2 −→ 0.
For β = (a1 , . . . , an ) ∈ ℤn and b ∈ ℤ, define
ϕ(β, b) = β and ψ(β) = a1 + ⋅ ⋅ ⋅ + an .
It is not hard to verify that ϕ is injective and that im(ϕ) = ker(ψ).
To show that ψ is onto consider a non bipartite component G1 of G. If x1 , . . . , x s
are the vertices of G1 and v′1 , v′2 , . . . , v′m are the columns of the incidence matrix of
G1 , then ℤ s /(v′1 , v′2 , . . . , v′m ) is a finite group. It follows that for any 1 ≤ j ≤ s, the
element e j is in the torsion subgroup of ℤn /ℤ𝒜, hence ψ(e j ) = 1, as required. □
COROLLARY 5.3.9 ([92]). If G is a graph and c0 (resp. c1 ) is the number of
bipartite (resp. non bipartite) components, then
ℤn /ℤ𝒜 ≃ ℤn−r × ℤc21 = ℤc0 × ℤc21 ,
where r = n − c0 is the rank of AG .
PROOF. Let G1 , . . . , Gc be the components of G and let 𝒜i be the set of vectors
in 𝒜 corresponding to the edges of the subgraph Gi . Since the vectors in 𝒜i can
be regarded as vector in ℤηi , where ηi is the number of vertices in Gi , there is a
canonical isomorphism
ℤn /ℤ𝒜 ≃ ℤη1 /ℤ𝒜1 ⊕ ⋅ ⋅ ⋅ ⊕ ℤηc /ℤ𝒜c .
To finish the proof apply Proposition 5.3.7 and Proposition 5.3.8 to derive
{
ℤ2 if Gi is non bipartite,
ℤηi /ℤ𝒜i ≃
ℤ if Gi is bipartite.
In the second isomorphism we are using the fact that incidence matrices of bipartite
graphs are totally unimodular, see Proposition 5.3.3. □
 5. EDGE SUBRINGS 221

COROLLARY 5.3.10. If c1 is the number of non bipartite components of G and

r is the rank of B, then
{ c −1
21 if c1 ≥ 1,
∆r (B) =
1 if c1 = 0.
PROOF. It follows from Proposition 5.3.8 and Corollary 5.3.9. □
DEFINITION 5.3.11. Let U be a square matrix with entries in a commutative ring
R. The matrix U is called unimodular if det(U) is a unit in R.
THEOREM 5.3.12 ([92]). If G is a graph with n vertices and AG is its incidence
matrix, then there are unimodular integral matrices U, V such that
( )
D 0
S = UAG V = ,
0 0
where D = diag(1, 1, . . . , 1, 2, 2, . . . , 2), n − c is the number of 1’s and c1 is the
number of 2’s.
PROOF. It follows using Corollary 5.3.9 together with the fundamental structure
theorem of finitely generated modules over a principal ideal domain. See [117,
Chapter 3]. □
The matrix S is called the Smith normal form of AG and the numbers in the
diagonal matrix D are the invariant factors of AG .

Exercises
5.3.13. Let G be a cycle of length n and let A be its incidence matrix. Prove that
det(A) = ±2 if n is odd and det(A) = 0 if n is even.
5.3.14. Let U be a square matrix with entries in a commutative ring R. Prove
that U is unimodular if and only if U is invertible.

5.4. Circuits of a graph and Gröbner bases

Let G be a graph. In this section we study the toric ideal P(G) of the edge subring
k[G] from the view point of Gröbner bases, and examine the elementary integral
vectors of the incidence matrix of G. For general results on Gröbner bases of toric
ideals a standard reference is [170, Chapter 4].
Let G be a graph on the vertex set V with edge set E and incidence matrix M.
We set N = ker(M), the kernel of M in ℚq . Observe that a vector α ∈ N ∩ ℤq
determine the subgraph Gα of G having vertex set
Vα = { x ∈ V ∣ x divide f α+ }
222 5. EDGE SUBRINGS

and edge set Eα = {{ x, y} ∈ E ∣ xy ∈ supp( f α+ ) ∪ supp( f α− )}, where f1 , . . . , fq

are the square-free monomials corresponding to the edges of G, that is, the edge
generators of G. Note that we are using the following notation:
Notation For α = (α1 , . . . , αq ) ∈ ℕq and f1 , . . . , fq in a commutative ring with
α
identity we set f α = f1α1 ⋅ ⋅ ⋅ fq q . The support of f α is defined as the set supp( f α ) =
{ fi ∣ αi ∕= 0}.
The next result gives a geometric description of the elementary integral vectors
or circuits of the kernel of the incidence matrix M of a graph G. This result has
been generalized to general multigraphs [59]. In loc. cit. an algorithm is given for
constructing a basis for the null space of M with respect to any fixed spanning tree
of G.
PROPOSITION 5.4.1 ([184]). Let G be a graph with incidence matrix M and let
N be the kernel of M in ℚq . Then a vector 0 ∕= α ∈ ℤq ∩ N is an elementary vector
of N if and only if
(i) Gα is an even cycle, or
(ii) Gα is a connected graph consisting of two odd cycles (with at most one
common vertex) joined by a path.
PROOF. Assume α ∈ ℤq ∩ N is an elementary vector of N so that Gα is not an
even cycle. From the equation f α+ = f α− we obtain a walk in Gα
w = { x0 , x1 , . . . , xn = x0 , xn+1 , . . . , x` },
so that x1 , . . . , x`−1 are distinct vertices, n is odd and x` is in { x1 , . . . , x`−2 }. The
walk w can be chosen so that xi xi+1 ∈ supp( f α+ ) if i is even and xi xi+1 ∈ supp( f α− )
otherwise. We claim that x` is not in { x1 , . . . , xn−1 }. Assume x` = xm for some
1 ≤ m ≤ n − 1. If m and ` − n are odd then the even closed walk
w1 = { x0 , x1 , . . . , xm , x`−1 , x`−2 , . . . , xn = x0 }
yields a relation f δ = f γ with supp(δ − γ) properly contained in supp(α), which is
a contradiction. If m is odd and ` − m is even then we can consider the walk
w1 = { xn , xn+1 , . . . , x` = xm , xm+1 , . . . , xn }
and derive a contradiction, the remaining cases are treated similarly. We may now
assume x` = xm for n ≤ m ≤ ` − 2. The cycle
C1 = { xm , . . . , x` = xm }
must be odd, otherwise C1 would give a relation f δ = f γ with supp(δ − γ) a proper
subset of supp(α). The walk w gives a relation f δ = f γ with
supp( f δ ) ⊂ supp( f α+ ) and supp( f γ ) ⊂ supp( f α− ).
 5. EDGE SUBRINGS 223

Hence supp(δ − γ) is a subset of supp(α), and since the support of α is minimal one
has the equality supp(δ − γ) = supp(α). Therefore
Vα = { x 1 , . . . , x ` }
and Gα is as required. The converse follows from Lemma 5.2.17. □
DEFINITION 5.4.2. A circuit of a graph G is either an even cycle or a subgraph
consisting of two odd cycles (with at most one common vertex) joined by a path.
COROLLARY 5.4.3. Let G be a graph with incidence matrix M. If λ = (λi ) is an
elementary integral vector of N = ker(M), then ∣λi ∣ ≤ 2 for all i.
PROOF. Let λ be an elementary integral vector of N. Since Gλ is an even cycle
or two edge disjoint cycles joined by a path we obtain an elementary integral vector
α with supp(α) = supp(λ) and so that ∣αi ∣ ≤ 2 for all i. Because any two elementary
vectors with the same support are dependent we obtain λ = α. □
THEOREM 5.4.4. Let G be a connected graph with q edges and let P be the toric
ideal of k[G]. Then the total degree of a polynomial in any reduced Gröbner basis
of P is less or equal than 2q dim Ze (G).
PROOF. Set N = {α ∈ ℚq ∣ Mα = 0} where M is the incidence matrix of
G. Assume that F is the reduced Gröbner basis of P and take f = f (t) ∈ F.
Using Corollaries 4.3.5 and 4.3.7, together with the fact that normalized reduced
Gröbner basis are uniquely determined, we obtain f (t) = tα − tβ . Notice that
supp(α) ∩ supp(β) = ∅, otherwise take i in the intersection and write f (t) = ti (ta − tb ).
Since ta − tb reduces with respect to F it follows that f (t) reduces with respect to
F ∖{ f }, which is impossible. We can now write f (t) = tα+ − tα− for some α ∈ N ∩ℤq .
By Theorem 3.7.8 we can write
α = a1 λ1 + ⋅ ⋅ ⋅ + ar λr (ai ∈ ℚ+ )
for some elementary integral vectors λi each in harmony with α such that supp(λi ) ⊂
supp(α) for all i and r ≤ dim N. Hence
α+ = a1 (λ1 )+ + ⋅ ⋅ ⋅ + ar (λr )+ and α− = a1 (λ1 )− + ⋅ ⋅ ⋅ + ar (λr )− .
If ai > 1 for some i then αi + and αi − are componentwise larger than λi+ and
λi− respectively, note that this is not possible by the previous argument. Hence
0 < ai ≤ 1. Using Corollary 5.4.3 it now follows that the entries of α satisfy
∣αi ∣ ≤ 2 dim Ze (G) for all i, which gives the asserted inequality. □

Exercises
5.4.5. Let G be a graph with incidence matrix M and N = ker(M) ⊂ ℚq . Give an
example to show that the set of elementary integral vectors of N does not determine
the presentation ideal of k[G].
224 5. EDGE SUBRINGS

5.4.6. Let G be a graph with incidence matrix M and N = ker(M) the kernel
of M in ℚq . Let λ1 , . . . , λ s be the elementary integral vectors of N. Prove that
ℰN = [0, λ1 ] + ⋅ ⋅ ⋅ + [0, λ s ] is a polytope.
5.4.7. If N is the kernel of the incidence matrix of a graph G and P is the toric
ideal of k[G], prove that U = {tα+ − tα− ∣ α ∈ ℰN ∩ ℤq } is a universal Gröbner basis
for P.
5.4.8. Prove that the complete graph 𝒦5 is not a circuit and consists of two edge
disjoint odd cycles of length 5.

5.5. Edge subrings of bipartite graphs

Let R = k[x1 , . . . , xn ] be a polynomial ring over a field k and let G be a bipartite
graph with vertex set V(G) = { x1 , . . . , xn }.
If f1 , . . . , fq are the edge generators of G, that is, the monomials that correspond
to the edges of G, then the presentation ideal P(G) of the edge subring k[G] is the
kernel of the homomorphism of k-algebras
B = k[t1 , . . . , tq ] −→ k[G]
induced by the assignment ti 7→ fi . By abuse of notation, we will sometimes refer
to ti as “the edge ti ”.
Let c be an even closed walk, c = ti1 , . . . , ti2r . For any such walk, write
tc = ti1 ti3 ⋅ ⋅ ⋅ ti2r−1 − ti2 ti4 ⋅ ⋅ ⋅ ti2r .
By Proposition 5.2.4 the generators of P(G) correspond to cycles in G:
P(G) = ({tc ∣ c is a cycle in G}).
Here we will consider graphs embedded on surfaces. The goal of this section
is to extract information about the edge subring k[G], by studying the particular
geometry of the graph G. We restrict ourselfs to compact, 2-dimensional surfaces
without boundary. These are compact connected Hausdorff topological spaces in
which every point has a neighborhood that is homeomorphic to the plane ℝ2 . Every
surface is homeomorphic to a space obtained from the 2-sphere by adding handles
(torus, double torus, etc.) and crosscaps (projective plane, klein bottle, etc.).
DEFINITION 5.5.1. An embedding of a graph G in the surface Σ is cellular if
every face of G is homeomorphic to an open disc in ℝ2 .
We define the genus g(G) and nonorientable genus g̃(G) of a graph G as the
minimum h and the minimum k, respectively, such that G has an embedding into the
orientable compact surface Σh of genus h, i.e., Σh is the sphere S 2 with h handles,
respectively into the nonorientable surface ℵk of genus k. Embeddings into Σg(G) are
 5. EDGE SUBRINGS 225

minimum genus embeddings. Similarly a nonorientable minimum genus embedding

of a graph G is an embedding into ℵg̃(G) .
PROPOSITION 5.5.2 ([134]). Every minimum genus embedding of a connected
graph G is cellular.
In what follows we will consider bipartite connected graphs embedded on
orientable surfaces. We always assume minimum genus embeddings. Furthermore
we assume that all our embeddings are simple, that is, the regions induced by
the embedding are open 2-cells bounded by simple circuits (closed walks with no
repeated vertices or edges). We refer to the regions as faces of the embedding. We
will denote the set of faces of the embedding by F(G). If c is a closed walk that
bounds a face we say that c is an atomic cycle.
DEFINITION 5.5.3. Let c be an even closed walk of G, c = ti1 , . . . , tis . A
polarization of c consists in choosing one of the sets
{ti1 , ti3 , . . . , tis−1 } or {ti2 , ti4 , . . . , tis }.
Let {c1 , . . . , cr } be the set of atomic cycles of G. A polarization 𝒫 of G is a set
𝒫 = {𝒫1∪ , . . . , 𝒫r } where 𝒫i is a polarization of ci for all i, 𝒫i ∩ 𝒫 j = ∅ for i ∕= j and
E(G) = ri=1 𝒫i .
THEOREM 5.5.4 ([82]). Let G a bipartite connected graph simply embedded on
an orientable surface 𝒮 , then G has a polarization.
PROOF. Let G be a graph with vertex set V(G) = {v1 , . . . , vn }, edge set E(G) =
{z1 , . . . , zq } and face set F(G), embedded on the surface 𝒮 . Let m(G) be the medial
graph of G with vertex set V(m(G)) = {u1 , . . . , uq } (see [85, Section 17.2]). Recall
that m(G) is a 4-regular graph. By construction, m(G) is also cellularly embedded
on 𝒮 . In particular the faces of m(G), F(m(G)), can be 2-colored, where one color
class corresponds to the set of vertices of G, denoted by F1 and the other class
corresponds to the faces of G, denoted by F2 . We color with black the faces in F1
and with white the faces in F2 .
Let us construct a new graph m′ (G) from m(G). For ui ∈ V(m(G)), let
{yi1 , yi2 , yi3 , yi4 } be the vertices adjacent to ui . Split ui into two new vertices u′i
and u′′i , so that u′i is adjacent to yi1 and yi2 , u′′i is adjacent to yi3 and yi4 and add a new
edge between u′i and u′′i , as shown in the following figure.

u u u u
yi1 f1 yi2 yi1 f1 yi2
@ @
@ @
f3 u@
i
u
f4 =⇒ ′ u
f3 u@ u u′′ f
@
@@ @
@
i
@@
@@@@@
@@i 4
@
@@ @@
@@ @@@
@@@@
@@
@ @
@ @ @
@
yi3 u@@
@
@@ f2@
@
@
@ @
@ @uyi
@ yi3 u@@
@
@
@
@ @ f2@
@
@ @@@
@@@
@@@uyi
@
@
@ @@@ @
@
@
@
@ 4 @
@
@ @@@
@ @ @@ @
@@@@
@@ 4
@@
@@@
@@
@@@
@@@
@@@
@ @@
@@@
@@
@@@
@@@@@@
@@@@
@@@@
226 5. EDGE SUBRINGS

Repeat the process for all vertices of m(G) to obtain m′ (G).

It is easy to see that the graph m′ (G) is bipartite by showing that all its cycles
are even. Since any cycle can be obtained as the symmetric difference of atomic
cycles, we only consider these cycles. Let C be an atomic cycle, if C is in F1 ,
then it corresponds to a vertex v ∈ V(G); the cycle associated to this vertex has, by
construction, a length twice the degree of v in G, so it is always even. Now if C is
in F2 then it corresponds to a cycle of F(G). Since G is bipartite, then the cycle is
also even, because in the splitting process no vertex is added to the boundary of a
white face.
Since m′ (G) is bipartite, we take a 2-coloring of its vertices. A polarization is
obtained by fixing one of these color classes and considering the set of vertices in
this class for every atomic cycle in F2 . □

We illustrate with an example the constructions given in the proof.

EXAMPLE 5.5.5. Let G = 𝒦4,4 and consider the following embedding of G in
the torus 𝒯 .
vs1 vs2 r r
c8 c3 @ c8 c8 c3 c8
s s s s
@
@ v6 @ u1 u2 u3 u4
v5 s c1 @s c6 @ sv5 r c1 r c6 r
@ @ u
s5
u6 s u
s7
u8 s
@ @
c2 v3@s c5 @sv4 c2 c2 r c5 r
c2
@ @ s s s s
@ s u9 u10 u11 u12
v7 s c4 @ c7 @ sv7
@ r c4 r c7 r
@ v8@ u
s 13
u14 s u
s 15
u16 s
c8@@s c3 @
@ s c8 c8 r
c3 r
c8
v1
v2
1. 𝒦4,4 embedded in 𝒯 2. The medial graph m(𝒦4,4 )
vs1 vs2
c8bc c3 c8 c8 @@ c3 @@ c8
ppppp pp t bcppp pt
pp
t bcppp pp t bcppp @@v6 @@
c1 c6 v5 s c1 @s c6 @ sv5
bc t
ppppp bc t
ppppp
ppppp ppppp
@ @
t bc t bc @@ @
@
c2 c5 c2 c2 v3@s
@ c5 @ @s v4 c2
bcppp t bcppp t
pp t ppp pp t ppp
bcpp bcpp
@@ @
@
@
@ @
c4 c7 v7 s c4 @s c7 @@ sv7
p bc t
ppppp p bc t
ppppp v
ppppp ppppp
@
t bc t bc @ 8@
@
c8 c3 c8 c8@@@s c3 @ @ s c8
@
v1 v2
′ 4. The polarization of 𝒦4,4
3. m (𝒦4,4 ) with a 2-coloring
of its vertices
 5. EDGE SUBRINGS 227

Figure 1 shows 𝒦4,4 embedded in the torus 𝒯 ; figure 2 shows the medial graph of
𝒦4,4 in the torus; in figure 3 the color classes are represented by the white vertices
and the black vertices and in figure 4 the polarization is obtained by taking pairs of
edges with lines in the interior of each atomic cycle.
From this embedding we obtain the following polarization
⎨ 𝒫1 = {t1 , t8 }, 𝒫2 = {t2 , t9 }, 𝒫3 = {t3 , t10 } ⎬
⎧ ⎫

𝒫= 𝒫4 = {t4 , t11 }, 𝒫5 = {t5 , t12 }, 𝒫6 = {t6 , t13 }

𝒫7 = {t7 , t14 }, 𝒫8 = {t15 , t16 }
⎩ ⎭

where
t1 = { x1 , x5 }, t2 = { x3 , x5 }, t3 = { x1 , x6 }, t4 = { x3 , x7 }
t5 = { x3 , x8 }, t6 = { x2 , x6 }, t7 = { x4 , x8 }, t8 = { x3 , x6 }
t9 = { x4 , x7 }, t10 = { x2 , x8 }, t11 = { x1 , x8 }, t12 = { x4 , x6 }
t13 = { x4 , x5 }, t14 = { x2 , x9 }, t15 = { x1 , x7 }, t16 = { x2 , x5 }
as is shown in figure 4.
LEMMA 5.5.6. Let G be a bipartite graph simply embedded on an orientable
surface and let {c1 , . . . , cr } be the atomic cycles of G. If 𝒫 = {𝒫1 , . . . , 𝒫r } is a
polarization of G, then by relabeling the edges one may assume that
ti ∈ 𝒫i for all i = 1, . . . , r − 1
/ ci if 1 ≤ i < j ≤ r − 1.
and t j ∈
PROOF. We use induction on r, the number of atomic cycles. If r − 1 = 1 the
lemma is clear. Take an edge t in the atomic cycle cr such that t ∈ / 𝒫r , label this
edge tr−1 . The edge tr−1 appears in an atomic cycle distinct from cr , that we label
cr−1 , and is contained in 𝒫r−1 . Let G′ be the graph obtained from G by deleting all
the edges and vertices that appear in the atomic cycles cr−1 and cr . Recall that G′
has r − 1 atomic cycles {c1 , . . . , cr−2 , c′r−1 }, where c′r−1 is the atomic cycle obtained
from cr−1 and cr when the edges and vertices in cr−1 ∩ cr are deleted. Apply the
induction hypothesis to G′ , with the polarization induced by G, labeling with t1
through tr−2 , to finish the proof. □
PROPOSITION 5.5.7. Let G be a bipartite graph simply embedded on an ori-
entable surface. If c1 , . . . , cr are the atomic cycles of G, then
{tc1 , . . . , tcr } ∖ {tci }
is a regular sequence for 0 ≤ i ≤ r,
PROOF. Exchange the labels of the atomic cycles ci and cr . Assume we have a
labeling of the edges as in Lemma 5.5.6. Let ≻ be the lexicographic ordering on
B = k[t1 , . . . , tq ] with
t1 ≻ t2 ≻ ⋅ ⋅ ⋅ ≻ tr−1 ≻ ⋅ ⋅ ⋅ ≻ tq ,
228 5. EDGE SUBRINGS

let in( f ) be the initial term of the polynomial f with respect to this ordering and let
in(I) be the ideal generated by the initial term of the polynomials in I with respect
to this ordering. Hence
∏
mi = in(tci ) = {tk ∈ 𝒫i }.

We have that gcd(mi , m j ) = 1 for all i ∕= j, because 𝒫i ∩ 𝒫 j = ∅. Therefore

ht(tc1 , . . . , tcr−1 ) = ht(in(tc1 , . . . , tcr−1 )) = ht(m1 , . . . , mr−1 ) = r − 1
and tc1 , . . . , tcr−1 is a regular sequence. □
EXAMPLE 5.5.8. Let 𝒦4,4 be embedded in the torus 𝒯 as in Example 5.5.5.
Then considering any of the subsets with seven binomials from the set
⎨ t1 t8 − t2 t3 , t2 t9 − t4 t13 , t3 t10 − t6 t11 ⎬
⎧ ⎫

ℛ= t4 t11 − t5 t15 , t5 t12 − t7 t8 , t6 t13 − t12 t16

t7 t14 − t9 t10 , t15 t16 − t1 t14
⎩ ⎭

we obtain a regular sequence.

Edge subrings of bipartite planar graphs. Let G be a bipartite planar graph

with a fixed embedding in the plane. From the algebraic point of view we may
limit ourselves to graphs which are 2-connected, see Proposition 5.2.35. Under this
assumption, atomic cycles will actually be cycles according to Theorem 5.1.7. In
the sequel, G is a 2-connected bipartite planar graph with n vertices, q edges and r
regions.
The ideal A(G) ⊂ P(G) generated by ({tc ∣ c is an atomic cycle of G}), is called
the atomic ideal.
PROPOSITION 5.5.9. If ri is the number of atomic cycles of length 2i of the graph
G, then the multiplicity of k[G] is given by
e(k[G]) ≤ 2r2 3r3 ⋅ ⋅ ⋅ srs
and this is a sharp bound for the multiplicity of k[G].
PROOF. First observe that r is the height of P(G) because the height of P(G) is q −
n+1 and this number is the dimension of the cycle space of G, see Proposition 5.2.18.
There is a graded epimorphism of k-algebras:
B/A(G) −→ B/P(G) ≃ k[G],
where B = k[t1 , . . . , tq ]. Since the Hilbert polynomials of B/A(G) and B/P(G) have
the same degree we obtain that e(A(G)) is an upper bound for the multiplicity of
 5. EDGE SUBRINGS 229

k[G]. As B/A(G) is a complete intersection its Hilbert series is given by

r
∏
1 − zai
( )

i=1
F(B/A(G), z) = ,
(1 − z)q
where ai is the degree of the binomial tci associated to the atomic cycle ci . Hence
r
∏
1 + z + ⋅ ⋅ ⋅ + zai −1
( )

i=1
F(B/A(G), z) = ,
(1 − z)n−1
and consequently e(A(G)) = 2r2 3r3 ⋅ ⋅ ⋅ srs .
To construct a planar graph with multiplicity e(A(G)), recall that P(G) is a
complete intersection if and only if any two cycles in G with no chord have at most
one edge in common, see Theorem 5.2.44. Thus any connected bipartite planar
graph G without cut vertices such that P(G) is a complete intersection would reach
this bound provided it has ri atomic cycles of length 2i for each i. An specific such
graph can be obtained starting with a cycle of length 2i and gluing ri cycles of this
length one by one, where at each step the new cycle added share exactly one edge
with the previous graph, having constructed this graph we continue with the same
procedure adding cycles of lengths 3i and so on. □
PROPOSITION 5.5.10 ([48]). If c1 , . . . , cr are the atomic cycles of G, then there
is a monomial order such that tc1 , . . . , tcr is a Gröbner basis for A(G), and both
tc1 , . . . , tcr and its leading terms sequence in(tc1 ), . . . , in(tcr ) are regular sequences.
PROOF. It follows from Proposition 5.5.7 and its proof. □
EXAMPLE 5.5.11. Let G be the following graph and consider the polarization
obtained by taking pairs of dashed lines in each atomic cycle.

t2
∙ ∙ _ _ _ ∙
t1
 c1  c2
 
_ _ _
∙ _ _ _ ∙ ∙
t3
c3  t c4 
 4

_ _ _
∙ ∙ ∙

Notice that with the lex ordering t1 ≻ ⋅ ⋅ ⋅ ≻ t4 ≻ t5 ≻ ⋅ ⋅ ⋅ ≻ t12 the binomials

tc1 , . . . , tc4 form a Gröbner basis for the ideal they generate.
230 5. EDGE SUBRINGS

THEOREM 5.5.12 ([82]). Let ri be the number of atomic cycles in G of length 2i.
Then
e(k[G]) = 2r2 3r3 ⋅ ⋅ ⋅ srs
if and only if A(G) = P(G).
PROOF. If A(G) = P(G), then P(G) is a complete intersection and the formula
for the multiplicity follows from the proof of Proposition 5.5.9.
Conversely assume the equality e(k[G]) = 2r2 3r3 ⋅ ⋅ ⋅ srs . Let c1 , . . . , cr be
the atomic cycles of G. By Proposition 5.5.10 there is a monomial order ≺ of
B = k[t1 , . . . , tq ] such that tc1 , . . . , tcr is a Gröbner basis for A(G), and both tc1 , . . . , tcr
and its leading terms sequence in(tc1 ), . . . , in(tcr ) are regular sequences. By Propo-
sition 5.2.9 the set of binomials {tc ∣ c is a cycle in G} form a universal Gröbner
basis of P(G). Thus in(P(G)), the initial ideal of P(G) with respect to ≺, is a
square-free monomial ideal. From the proof of [170, Proposition 13.15] and using
that a shellable simplicial complex is Cohen-Macaulay it follows that B/in(P(G)) is
Cohen-Macaulay. Note that
e(B/in(P(G))) = e(B/P(G)) = e(B/A(G)) = e(B/in(A(G))).
Since the multiplicity e(B/in(P(G))) (resp. e(B/in(A(G))) is the number of minimal
primes of in(P(G)) (resp. in(A(G))) and all those primes have the same height r,
using primary decompositions it follows readily that
in(A(G)) = (in(tc1 ), . . . , in(tcr )) = in(P(G)).
Hence P(G) = A(G) as required. □
COROLLARY 5.5.13 ([82]). Let ri be the number of atomic cycles in G of length
2i. If
e(k[G]) = 2r2 3r3 ⋅ ⋅ ⋅ srs ,
then k[G] is a complete intersection.
The converse fails. Indeed in Example 5.2.26 the edge subring k[G] is a
complete intersection, e(k[G]) = 8, and 22 3 = 12.

Exercises
5.5.14. Let G be a bipartite connected graph. Let ri be the number of primitive
cycles of G of length 2i. If G is a ring graph, prove that
e(k[G]) = 2r2 3r3 ⋅ ⋅ ⋅ srs ,
5.5.15. Let G be a bipartite graph and let G1 , . . . , Gr be their blocks, then
∑r ∏r
a(k[G]) = a(k[Gi ]) and type(k[G]) = type(k[Gi ]).
i=1 i=1
 5. EDGE SUBRINGS 231

5.5.16. Let G be a bipartite connected planar graph. Prove that the atomic ideal
A(G) depends on the embedding of G in the plane.
5.5.17. Show that the following embedding in the plane of the bipartite graph
consisting of two edge disjoint squares that meet at a point has atomic cycles which
are not cycles.
s

A@

A@

A @
s s

A s @s
@ @
@ @s
@
@s

5.6. The integral closure of an edge subring

Let G be a graph and let k[G] be the edge subring of G over an arbitrary field k.
In general, it is difficult to certify whether k[G] has a given algebraic property or
to obtain a measure of its numerical invariants. This arises because the number
q of monomials is usually large. Our goal here is to give a construction for the
integral closure or normalization k[G] of k[G], for this purpose the underlying
graph theoretic aspects are very helpful.

A combinatorial description of normalizations

Let us begin by proving the normality of edge subrings associated to graphs which
are bipartite or nearly bipartite.
COROLLARY 5.6.1 ([156]). If G is a bipartite graph and k is a field, then the
edge subring k[G] is a normal domain.
PROOF. It follows from Corollary 4.3.25 and Proposition 5.3.3. □
PROPOSITION 5.6.2. Let G be a connected graph and k a field. If G has at most
one odd cycle, then k[G] is normal.
PROOF. We may assume that G has a unique odd cycle, for otherwise G is a
bipartite graph and by Corollary 5.6.1 k[G] is normal. Pick an edge e of G which is
on the unique odd cycle and set H = G ∖ {e}. Notice that H is a connected bipartite
graph because e is not a bridge (it is on a cycle). Hence k[H] is a normal domain of
dimension dim k[G] − 1 (see Proposition 5.2.18). Consider the epimorphism
ϕ
k[H][t] −→ k[G] −→ 0, induced by ϕ(t) = fe ,
232 5. EDGE SUBRINGS

where t is a new indeterminate and fe is the monomial corresponding to e. Since

k[H][t] and k[G] are both Noetherian domains of the same dimension, ϕ is an
isomorphism. Therefore k[G] is normal. □
LEMMA 5.6.3. Let G be a graph without even cycles. Then the intersection of
any two cycles is either a point or the empty set.
PROOF. Let Z1 = {z0 , z1 , . . . , zn = z0 } and Z2 = {w0 , w1 , . . . , wm = w0 } be two
distinct cycles of G such that Z1 ∩ Z2 ∕= ∅. We may assume z0 = w0 and w1 ∕∈ Z1 .
Let j = min{i ≥ 1∣ wi ∈ Z1 }, and write w j = z s . If 0 < s < n, considering the cycles
{w0 , w1 , . . . , w j = z s , z s−1 , . . . , z0 }
and {w0 , w1 , . . . , w j = z s , z s+1 , . . . , zn }
of lengths s+ j and j+n− s, respectively, we obtain that n is even, which is a contradic-
tion. Therefore s = 0 or s = n. To conclude observe that Z2 = {w0 , w1 , . . . , w j−1 , z0 },
hence Z1 ∩ Z2 = {z0 }. □
DEFINITION 5.6.4. A graph G having just one cycle is said to be unicyclic.
DEFINITION 5.6.5. Let G be a graph without even cycles. A cycle Z of G is said
/ Z.
to be a terminal cycle if N(Z) = Z ∪ {v}, for some v ∈
LEMMA 5.6.6. Let G be a connected graph without even cycles. Assume that
Z1 ∩ N(Z2 ) = ∅, for any two cycles Z1 , Z2 of G. If G is not a unicyclic graph, and
deg(v) ≥ 2 for all v ∈ V(G), then G has at least two terminal cycles.
PROOF. Let Z1 , . . . , Zr be the cycles of the graph G. We now construct a new
graph T having vertex set
( r
)
∪ ∪
V(T ) = {Z1 , . . . , Zr } V(G) ∖ Zi .
i=1

Thus the vertices of T are the cycles of G, together with the vertices of G not lying
on any cycle of G. Two points Y1 , Y2 of T are adjacent if and only if Y1 ∩ N(Y2 ) ∕= ∅,
here we regard a vertex v of G as the one point set {v}. Notice that T is a tree and
that the terminal cycles of G correspond to the end points of T . Hence G must have
at least two terminal cycles. □
The integral closure in terms of bow ties. Let G be a graph with vertex set
V = { x1 , . . . , xn } and let R = k[x1 , . . . , xn ] be a polynomial ring over a field k.
For use below let 𝒜 = {v1 , . . . , vq } be the set of ei + e j such that xi is adjacent to
x j , where ei is the ith unit vector in ℝn , and let { f1 , . . . , fq } be the set of monomials
xi x j in R such that { xi , x j } is an edge of G. Note
k[G] = k[ f1 , . . . , fq ] ⊂ R.
 5. EDGE SUBRINGS 233

One has the following description of the integral closure of k[G]

k[G] = k[{ xa ∣ a ∈ ℝ+ 𝒜 ∩ ℤ𝒜}],
where ℝ+ 𝒜 is the cone in ℝn generated by 𝒜 and ℤ𝒜 is the subgroup of ℤn generated
by 𝒜. See Theorem 4.3.23.
DEFINITION 5.6.7. A bow tie of a graph G is a subgraph w of G consisting of
two primitive odd cycles with at most one common vertex
sz1 s z8
z2 
s 

J 
J
Js zs6 s


z0 = z5 z7 = z10Z
z3 X
s Z
XX Xsz4 ZZsz9
Z1 = {z0 , z1 , . . . , zr = z0 } and Z2 = {z s , z s+1 , . . . , zt = z s },
joined by a path {zr , . . . , z s }. In this case we set Mw = z1 ⋅ ⋅ ⋅ zr z s+1 ⋅ ⋅ ⋅ zt .
If w is a bow tie of a graph G, as above, then Mw is in the integral closure of
k[G]. Indeed if fi = zi−1 zi , then
(5.48) z21 ⋅ ⋅ ⋅ z2r = f1 ⋅ ⋅ ⋅ fr , z2s ⋅ ⋅ ⋅ z2t−1 = f s+1 ⋅ ⋅ ⋅ ft ,
which together with the identities
∏ ∏
Mw = fi fi−1 and Mw2 = f1 ⋅ ⋅ ⋅ fr f s+1 ⋅ ⋅ ⋅ ft
i odd i even
r<i≤s
gives Mw ∈ k[G]. Putting the two equations above together one obtains
∏ ∏ ∏ ∏
(5.49) fi2 fi = fi2 fi .
i odd i odd i even i even
r<i≤s i∈ / (r, s] r<i≤s i∈ / (r, s]
There are two special cases of this equality that we would like to record. If r = s,
that is Z1 and Z2 intersect at a vertex, then one derives
∏ ∏
(5.50) fi = fi ,
i odd i even
which is a square-free relation. If s = r + 1, that is Z1 and Z2 are joined by an edge
{zr , z s }, then one has
∏ ∏
(5.51) fi = f s2 fi ,
i odd i even
i ∕= s
which is a one sided square-free relation.
234 5. EDGE SUBRINGS

LEMMA 5.6.8. Let G be a connected graph without even cycles, V the vertex set
of G, and f1 , . . . , fq the monomials defining the edges of G. Assume the conditions:
(a) Z1 ∩ N(Z2 ) = ∅, for any two cycles Z1 , Z2 of G, and
(b) deg(v) ≥ 2 for all v ∈ V and G has an even number of vertices.
Then v∈V v is in k[ℬ ], where ℬ = { f1 , . . . , fq } ∪ { Mw ∣ w is a bow tie}.
∏

P∏ROOF. The proof is by induction on the number of cycles of the graph G. Set
z = v∈V v. If G has exactly two cycles, then G is a bow tie and we can write
z = f δ, where f ∈ k[G] and δ = AG is the product of the vertices in the two cycles.
We may now assume that G has at least three cycles.
By Lemma 5.6.6 there are two terminal cycles
Z1 = { x0 , x1 , . . . , xr = x0 }, Z2 = {y0 , y1 , . . . , y s = y0 }.
There is a path { xr , xr+1 , . . . , xr1 }, so that r + 1 ≤ r1 , deg(x j ) = 2 for all r < j < r1 ,
and deg(xr1 ) ≥ 3. Likewise there is a path {y s , y s+1 , . . . , y s1 }, so that s + 1 ≤ s1 ,
deg(y j ) = 2 for all s < j < s1 , and deg(y s1 ) ≥ 3. We may assume r1 and s1 even,
for otherwise, if r1 is odd, we write
∏
z = (x1 x2 )(x3 x4 ) ⋅ ⋅ ⋅ (xr xr+1 ) ⋅ ⋅ ⋅ (xr1 −2 xr1 −1 ) v,
v∈V1

where V1 = V ∖ { x1 , . . . , xr1 −1 }, and then use induction. It is not hard to show that
{ x1 , . . . , xr1 } ∩ {y1 , . . . , y s1 −1 } = { x1 , . . . , xr1 −1 } ∩ {y1 , . . . , y s1 } = ∅.
Set
σ1 = (xr+1 xr+2 ) ⋅ ⋅ ⋅ (xr1 −2 xr1 −1 ) and σ2 = (y s+1 y s+2 ) ⋅ ⋅ ⋅ (y s1 −2 y s1 −1 ).
If xr1 ∕= y s1 then we write
s
∏ r
∏ ∏
z = σ1 σ2 yi xi v,
i=1 i=1 v∈V1

where V1 = V ∖{ x1 , . . . , xr1 −1 , y1 , . . . , y s1 −1 }, hence using induction we get z ∈ k[ℬ ].

We may now assume that r1 , s1 are even and xr1 = y s1 . Consider the subgraph
H = G ∖ { x1 , . . . , xr1 −1 , y1 , . . . , y s1 −1 }.
If deg(xr1 ) ≥ 4, then applying the induction hypothesis to H we obtain z ∈ k[ℬ ].
Assume deg(xr1 ) = 3. Note that H has a terminal cycle
Z3 = {z0 , z1 , . . . , zt = z0 }
and there is a path {zt , zt+1 , . . . , zt1 }, so that t + 1 ≤ t1 , deg(z j ) = 2 for all t < j < t1 ,
and deg(zt1 ) ≥ 3, by the previous arguments we are reduced to the case r1 , s1 , t1
even and xr1 = y s1 = zt1 . Observe that this forces the equality
V = { x1 , . . . , xr1 , y1 , . . . , y s1 −1 , z1 , . . . , zt1 −1 }.
 5. EDGE SUBRINGS 235

To complete the proof we use the identity

r
∏ s
∏
z = (z1 z2 )(z3 z4 ) ⋅ ⋅ ⋅ (zt1 −1 zt1 )σ1 σ2 xi yi ,
i=1 i=1
to derive z ∈ k[ℬ ]. □
Given a graph G we set 𝒜G = {log( f1 ), . . . , log( fq )}, where f1 , . . . , fq are the
monomials corresponding to the edges of G and log( fi ) is the exponent vector of fi .
The affine subsemigroup of ℕn generated by 𝒜G will be denoted by CG . Thus
∑
CG = ℕα.
α∈𝒜G
The cone generated by CG will be denoted by ℝ+CG , it is equal to
{ r }
∑
ℝ+ C G = ai αi ai ∈ ℝ+ , αi ∈ 𝒜G , r ∈ ℕ .
i=1
Note that k[G] is equal to the affine semigroup ring k[CG ]. This cone will be studied
in detail in Chapter 6.
LEMMA 5.6.9. Let G1 and G2 be two finite sets of monomials in disjoint sets of
indeterminates. Then k[G1 ∪ G2 ] ≃ k[G1 ] ⊗k k[G2 ] and
k[G1 ∪ G2 ] ≃ k[G1 ] ⊗k k[G2 ] ≃ k[G1 ] k[G2 ],
is the subring generated by the generators of k[G1 ] and k[G2 ].
PROOF. Set G = G1 ∪ G2 . By Theorem 4.3.23 we have
k[G] = k[{ xa ∣ a ∈ ℤCG ∩ ℝ+CG }].
On the other hand there is a decomposition
ℤCG ∩ ℝ+CG = (ℤCG1 ∩ ℝ+CG1 ) ⊕ (ℤCG2 ∩ ℝ+CG2 ).
Using this equality the assertion follows readily. □
COROLLARY 5.6.10. Let G be a graph and let G1 , . . . , Gr be its connected
components. Then k[G] is normal if and only if k[Gi ] is normal for each i.
For convenience we recall the following result.
THEOREM 5.6.11 ([29]). Let F be a finite set of square-free monomials of degree
r in R. If dim k[F] = n and n ≥ 2r ≥ 4, then k[F] is ⌈generated as a k-algebra by
elements of normalized degree less or equal than n − nr .
⌉

THEOREM 5.6.12 ([157]). Let G be a graph and let k be a field. Then the integral
closure k[G] of k[G] is generated as a k-algebra by the set
ℬ = { f1 , . . . , fq } ∪ { Mw ∣ w is a bow tie}.
236 5. EDGE SUBRINGS

PROOF. Let V = { x1 , . . . , xn } and E = {e1 , . . . , eq } be the vertex set and edge

set of G, respectively. According to Theorem 4.3.23 one has:
k[G] = k[{ xa ∣ a ∈ ℤCG ∩ ℝ+CG }].
The proof is by induction on q, the number of edges of G. The case q = 1 is clear.
Assume q ≥ 2 and that the result is true for graphs with less than q edges. If
ei = { x j , xk }, we set fi = x j xk .
First we remove all the isolated vertices of G, without affecting the subring
k[G]. By Lemma 5.6.9 and the induction hypothesis we may assume that G is a
connected graph. We may also assume that deg(v) ≥ 2 for all v ∈ V; because if
deg(v) = 1 and fi = vx j , then setting G′ = G ∖ {ei } we have k[G] = k[G′ ][ fi ], the
latter being a polynomial ring over k[G′ ], hence
k[G] = k[G′ ][ fi ]
and fi is not part of a bow tie, which by induction yields the required equality.
Let z = xa ∈ k[G] be a minimal (irreducible) generator of the integral closure
of k[G]. There is a positive integer m such that
mλq b
(5.52) zm = f1mλ1 ⋅ ⋅ ⋅ fq = f1b1 ⋅ ⋅ ⋅ fq q , (λi ∈ ℤ and bi ∈ ℤ+ ∀i).
Observe that if bi = λi = 0 for some i, then each bowtie of G ∖ {ei } is a bowtie of
G and the induction hypothesis yields z ∈ k[ℬ ]. Hence we may further assume that
all the monomials f1 , . . . , fq must be present in Eq. (5.52). There are two cases to
consider.
Case (I): Assume that G contains an even cycle Z. We set vi = log( fi ). For con-
venience we use the edges of G to represent the cycle Z by the sequence {ei1 , . . . , eik },
where two consecutive edges are adjacent. If bir = 0 for some 1 ≤ r ≤ k, then using
the relation
k
∑ k
∑
(5.53) (−1) j vi j = (−1) j log( fi j ) = 0,
j=1 j=1

that comes from the even cycle, we may rewrite Eq. (5.52) as
mµq b
zm = f1mµ1 ⋅ ⋅ ⋅ fq = f1b1 ⋅ ⋅ ⋅ fq q , (µi ∈ ℤ ∀i and µir = bir = 0),
which by induction yields z ∈ k[ℬ ]. We may now assume bir > 0 for all 1 ≤ r ≤ k.
We may harmlessly assume that bi1 ≤ ⋅ ⋅ ⋅ ≤ bik . Using Eq. (5.52) and Eq. (5.53)
we obtain
q
∑ k
∑ ∑ ∑ ∑
ma = bi vi = bi1 vi j + ci vi = 2bi1 vi j + ci vi ,
i=1 j=1 i∕=i1 j even i∕=i1
 5. EDGE SUBRINGS 237

where the ci ’s are non negative integers. The net effect is that now we may rewrite
Eq. (5.52) as
mδq d
zm = f1mδ1 ⋅ ⋅ ⋅ fq = f1d1 ⋅ ⋅ ⋅ fq q , (δi ∈ ℤ, di ∈ ℤ+ ∀i),
and di1 = δi1 = 0. Hence by induction z ∈ k[ℬ ].
Case (II): Assume that all the cycles of G are odd. Let Z1 and Z2 be two distinct
cycles of G, represented by the sequences of edges {ei1 , . . . , eik } and {e`1 , . . . , e`s },
respectively. If Z1 ∩ Z2 ∕= ∅, then according to Lemma 5.6.3 Z1 ∩ Z2 = { x}, for some
x ∈ V, say x ∈ supp( fi1 ) ∩ supp( f`1 ). Using the identity
∑ ∑ ∑ ∑
vi j + v` j = vi j + v` j
j odd j even j even j odd
we may proceed as in Case (I). If Z1 ∩ Z2 = ∅ and N(Z1 ) ∩ Z2 ∕= ∅, say ft = xy, with
x ∈ supp( fik ) and y ∈ supp( f`1 ). Again using the equality
∑ ∑ ∑ ∑
(5.54) vi j + v` j = 2vt + vi j + v` j ,
j odd j odd j even j even
we may proceed as in Case (I) to derive z ∈ k[ℬ ].
Therefore we are reduced to the case N(Z1 ) ∩ Z2 = ∅, for any two cycles Z1 , Z2 .
Observe that if fi correspond to an edge of some cycle, then bi > 0, for otherwise if
bi = 0 one may use Eq. (5.49) to eliminate fi from Eq. (5.52) and use induction to
obtain z ∈ k[ℬ ]. In particular any variable that occurs in a cycle is in the support of
z. The proof now reduces to showing the following two subcases.
b
Case (II) (a): If x j is not in the support of f1b1 ⋅ ⋅ ⋅ fq q , for some j, consider the
set
D = {i ∣ x j ∈ supp( fi±λi )}.
The set D is nonempty because ∑ G has no isolated vertices and all the edges of G are
present in Eq. (5.52). Then i∈D λi = 0 and bi = 0, for all i ∈ D. Taking squares
in both sides of Eq. (5.52) gives
2λ 2b
(z2 )m = ( f12λ1 ⋅ ⋅ ⋅ fq q )m = f12b1 ⋅ ⋅ ⋅ fq q .
Note that for any i ∈ D, we can use that G is connected and Eq. (5.49) to eliminate
the monomial fi from the equation above. Hence by induction we get z2 ∈ k[ℬ ].
Therefore using that z is a minimal generator of k[G] we can write
(5.55) z2 = Mw1 ⋅ ⋅ ⋅ Mwk fi1 ⋅ ⋅ ⋅ fi p ,
where w1 , . . . , wk are distinct bow ties and fi1 , . . . , fi p are distinct monomials. If
k = 0 and w is any bowtie of G, then using Eq. (5.48) and the minimality of z yields
z = Mw .
Assume k > 0 using Eq. (5.48) one may assume that cycles occurring in the
bow ties w1 , . . . , wk are mutually disjoint.
238 5. EDGE SUBRINGS

Given an odd cycle

Z1 = {z0 , z1 , . . . , zr = z0 }

occurring in wi for some 1 ≤ i ≤ k, we claim that there are z ∈ Z1 and x ∈ V ∖ Z1

such that fi j = zx for some 1 ≤ j ≤ p. Let us proceed by contradiction. First note
that z2i divides the right hand side of Eq. (5.55) for all 1 ≤ i ≤ r. Hence either z1 z2 or
zr z1 occur in fi1 , . . . , fi p , say fi1 = z1 z2 . Observe that z2 z3 cannot occur in fi2 , . . . , fi p ,
for otherwise z42 would divide the right hand side of Eq. (5.55) and thus either z1 z2 or
z2 z3 would have to appear twice in the sequence fi1 , . . . , fi p , a contradiction. Hence
z3 z4 must occur in the sequence fi2 , . . . , fi p , say fi2 = z3 z4 . Repeating the argument
one obtains that z5 z6 , . . . , zr−2 zr−1 , zr z1 occurrs in the sequence fi1 , . . . , fi p . Thus z41
divides the right hand side of Eq. (5.55), and consequently either z1 z2 or z1 zr occur
twice, a contradiction. This completes the proof of the claim.
We set fi1 = zx, z = zr , x = zr+1 , and S = ∪ki=1 supp(Mwi ). If zr+1 ∈ / S , then since
2
zr+1 divides the right hand side of Eq. (5.55) there is a vertex zr+2 such that zr+1 zr+2
occurrs in the sequence fi2 , . . . , fi p . If zr+2 ∈ / S we repeat the argument. Continuing
with this procedure leads to the following two possibilities:
(i) there is r < s such that zi zi+1 occurrs in the sequence fi1 , . . . , fi p for r ≤ i ≤
s − 1, z s ∈ S , and zi ∈ / S for r < i < s, or

(ii) there is an odd cycle

Z2 = {z s , z s+1 , . . . , zt = z s }

disjoint from the cycles in w1 , . . . , wk such that the “edge” zi zi+1 occurs in fi1 , . . . , fi p
for r ≤ i < t
Subcase (i): Since the cycles in w1 , . . . , wk are mutually disjoint one may
assume that w1 is a bow tie with odd cycles Z1 , Z2 , where Z2 is the odd cycle
Z2 = {z s , z s+1 , . . . , zt = z s }. If s is even one has

r−3 s−2
⎡ ⎤
s−1
∏ ∏2 ∏2
Mw1 (zi zi+1 ) = ⎣(z1 zr )(zr−1 zr ) (z2i z2i+1 ) (z2i z2i+1 )2 ⎦
⎢ ⎥
i=r i=1 r+1
i= 2
t−3
⎡ ⎤
2
∏
⋅ ⎣(z s z s+1 )(z s zt−1 ) (z2i z2i+1 )⎦ ,
⎢ ⎥
s+2
i= 2
 5. EDGE SUBRINGS 239

and if s is odd one has

r
∏ s−1
∏ t−1
∏
zi (zi zi+1 ) zi
i=1 i=r i=s
r−1 s−1 t−2
∏2 2
∏ 2
∏
= (z2i−1 z2i ) (z2i−1 z2i )2 (z s z s+1 )(z s zt−1 ) (z2i−1 z2i ).
i=1 r+1 s+3
i= 2 i= 2

In both cases using Eq. (5.55) and the minimality (irreducibility) of z one derives
z = fi for some i.
Subcase (ii): Since G is connected using Eq. (5.48) and the equality
t−1
∏ s−1
∏
(zi zi+1 ) = (zi zi+1 ) ⋅ (z s z s+1 ⋅ ⋅ ⋅ zt−1 )2
i=r i=r

one can reduce to the argument given in subcase (i). Indeed if Z1 , Z2′ are the cycles
of w1 , then we form two bowties consisting of the cycles Z1 , Z2 and Z2 , Z2′ . Thus
one may apply the argument of subcase (i).
Case (II) (b): It remains to consider the case supp(zm ) = V, that is one can write
z = x1a1 ⋅ ⋅ ⋅ xnan , with ai ≥ 1 for all i. By Corollary 5.6.11 we obtain
⌈ ⌉ that k[G] is
generated as a k-algebra by monomials of degree at most 2(n − n2 ). Altogether
we get
n ≤ deg(z) ≤ 2 n − ⌈n/2⌉ ,
( )

hence n must be even and z = x1 ⋅ ⋅ ⋅ xn . Using Lemma 5.6.8 we conclude that

z ∈ k[ℬ ], as required. □
There is a version of Theorem 5.6.12, due to Hibi and Ohsugi [112], that allows
loops in the graph G.
Next we clarify that there are two special types of bowties that are irrelevant for
the description of the integral closure of k[G].
PROPOSITION 5.6.13. Let w be a bowtie consisting of two odd cycles Z1 , Z2 that
meet at exactly one vertex or are disjoint and connected by an edge, that is, w has
essentially one of the forms:
∙??? ∙ ∙??? ∙
?? Z1  ?? 
??  ?? 
??  ?? 

∙?? Z1 ∙ ∙??? Z2
 ???  ??
 ??  ??
 Z2 ??  ??
∙ ∙ 
∙ ∙
then Mw is in k[G].
240 5. EDGE SUBRINGS

PROOF. It is left as an exercise. □

As an immediate consequence of Theorem 5.6.12 and Proposition 5.6.13 one
has the following full characterization of when k[G] is integrally closed.
COROLLARY 5.6.14. Let G be a connected graph. Then k[G] is normal if and
only if for any two odd cycles Z1 , Z2 with at most one common vertex either Z1 ∩ Z2
consists of a single vertex, or Z1 and Z2 are connected by an edge.
Thus the smallest connected graph G such that k[G] is non normal is the Hochster
configuration of Example 5.8.2, consisting of a graph G with two disjoint triangles
joined by a 2-path.
DEFINITION 5.6.15. A graph G is said to satisfy the odd cycle condition if every
two vertex disjoint odd cycles of G are joined by an edge in G.
The odd cycle condition has come up in the literature in connection with the
normality of edge subrings [112, 156, 157] and the description of the circuits of a
graph [148, 184]; it also occurred before [77]. This notion has been rediscovered
during the study of monomial subrings of graphs.
COROLLARY 5.6.16 ([130]). If G is a graph such that k[G] is normal, then the
toric ideal of k[G] is generated by the binomials that correspond to the circuits of
the kernel of the incidence matrix of G.
PROOF. (Sketch) Let P = ⊕i≥2 Pi be the presentation ideal of k[G] and let I be the
circuit ideal of G, i.e., I is generated by the binomials that correspond to the circuits
of the kernel of the incidence matrix of G. Using induction and the geometric
description of the circuits of the incidence matrix of G given in Proposition 5.4.1, it
follows that Pi ⊂ I for all i ≥ 2. □

Exercises
5.6.17. Prove that k[G] is normal for every connected graph G with six vertices,
where k is a field.
5.6.18. Let G be a graph and let G1 , . . . , Gr be the connected components of
G. If k is a field, prove that k[G] is normal if and only if Gi satisfies the odd cycle
condition for all i.
5.6.19. Let G be a graph and let w be a bowtie of G with odd cycles
Z1 = { f1 , . . . , fr } and Z2 = { f s+1 , . . . , ft }
joined by the path { fr+1 , . . . , f s }. Here fi has a double meaning, it stands for an edge
of G and also it stands for its corresponding monomial. Assume f = f1b1 ⋅ ⋅ ⋅ ftbt ,
where bi ∈ ℕ for all i. Use Eq. (5.49) recursively to prove that f can be rewritten
 5. EDGE SUBRINGS 241

as f = f1c1 ⋅ ⋅ ⋅ ftct for some c1 , . . . , ct in ℕ such that one of the following three
conditions occur:
(i) c j = 0 for some f j that occur in Z1 or Z2 , or
(ii) ck = 0 for some r < k ≤ s, or
(iii) ck = 1 for some r < k ≤ s and ci ≥ 1 for r < i ≤ s.
5.6.20. Prove that if f can be written in such a way that condition (iii) above
holds and (i) does not hold, prove that f 2 can be rewritten as
f 2 = f1d1 ⋅ ⋅ ⋅ ftdt
for some d1 , . . . , dt in ℕ, where dk = 0 for some r < k ≤ s.
5.6.21. Use the last two exercises to give a proof of Theorem 5.6.12 without
using Lemma 5.6.8 or Corollary 5.6.11.

5.7. Ehrhart rings of edge polytopes

Let G be a graph on the vertex set V = { x1 , . . . , xn }. The edge polytope of G is the
lattice polytope
𝒫 = conv(𝒜) ⊂ ℝn ,
where 𝒜 = {v1 , . . . , vq } is the set of vectors in ℕn of the form ei + e j such that xi is
adjacent to x j . Let R = K[x1 , . . . , xn ] be a polynomial ring over a field K. Recall
that the Ehrhart ring of 𝒫 is
A(𝒫 ) = K[{ xα ti ∣α ∈ ℤn ∩ i𝒫}] ⊂ R[t]
and the edge subring of G is the K-subalgebra
K[G] = K[xv1 , . . . , xvq ] ⊂ R = K[x1 , . . . , xn ].
If A is the incidence matrix of G and K[𝒫 ] is the polytopal subring
K[𝒫 ] = K[xα t∣ α ∈ ℤn ∩ 𝒫 ] ⊂ R[t],
then one has the equalities
dim K[G] = rank(A) = n − c0 and dim K[𝒫 ] = dim(𝒫 ) + 1,
where c0 is the number of bipartite components of G (see Chapter 4). By Exer-
cise 4.5.27 one has
K[G] ≃ K[𝒫 ] ⊂ A(𝒫 ),
and consequently dim(𝒫 ) = n − c0 − 1.
For use below we define the matrix B by:
( )
v1 ⋅ ⋅ ⋅ vq
B= .
1 ⋅⋅⋅ 1
242 5. EDGE SUBRINGS

Observe that rank(B) = rank(A) because the last row of B is a linear combination of
the first n rows.
For simplicity we keep the notation introduced above throughout the rest of this
section. We shall assume, if need be, that the graph G has no isolated vertices.
If G is a connected graph, then K[P] = A(𝒫 ), see [112, p. 423]. Next we
generalize this result by characterizing those graphs for which equality occurs.
THEOREM 5.7.1. Let G be a graph and let 𝒫 be its edge polytope, then equality
K[𝒫 ] = A(𝒫 ) occurs if and only if G has at most one non bipartite connected
component.
PROOF. ⇒) If G has two non bipartite components G1 , G2 , take
Z1 = { x0 , x1 , . . . , xr = x0 }, Z2 = { xr+1 , xr+2 . . . , xr+s = xr+1 }
two odd cycles of lengths r, s − 1 in G1 , G2 respectively. Consider the vector
β = e1 + ⋅ ⋅ ⋅ + er + er+1 + ⋅ ⋅ ⋅ + er+s−1
and note the equality
[( r−1
) ( r+s−2
)]
m er + e1 ∑ ei + ei+1 er+s−1 + er+1 ∑ ei + ei+1
β= + + +
2 m m m m
i=1 i=r+1

where m = r + s − 1. Hence β ∈ ℤn ∩ p𝒫 , where p = m/2, that is, xβ t p belongs

to A(𝒫 ) = K[𝒫 ]. Hence (β, p) is in the field of fractions of the polytopal subring
K[𝒫 ]. There are integers λ1 , . . . , λq such that
e1 + e2 + ⋅ ⋅ ⋅ + er+s−1 + pen+1 = λ1 (v1 , 1) + ⋅ ⋅ ⋅ + λq (vq , 1).
Since Z1 and Z2 are in different connected components we obtain
e1 + ⋅ ⋅ ⋅ + er = j λ j v j ,
∑

where the sum is taken over all j such that v j correspond to an edge of G1 . Let x0 be
the vector with all its entries equal to 1. Taking inner products in the last equality
one has:
r = ⟨e1 + ⋅ ⋅ ⋅ + er , x0 ⟩ = 2 j λ j ,
∑

a contradiction because r is odd.

⇐) If G is a bipartite graph, then K[P] = A(𝒫 ) by Proposition 4.5.20. If G is not
a bipartite graph the result follows from Proposition 5.3.7 and Proposition 4.5.14.
□
THEOREM 5.7.2. If G is a graph and 𝒫 is its edge polytope, then
{ c −1
2 1 e(K[G]) if G is non bipartite,
e(A(𝒫 )) = vol(𝒫 )d! =
e(K[G]) if G is bipartite,
where d = dim(𝒫 ) and c1 is the number of non bipartite components of G.
 5. EDGE SUBRINGS 243

PROOF. By Corollary 4.5.13 one has e(A(𝒫 )) = ∆r (B)e(K[𝒫 ]), where r is the
rank of A. From Corollary 5.3.9 one derives
∆r (A) = ∣T (ℤn /ℤ𝒜)∣ = 2c1 .
Thus the result follows readily from Proposition 5.3.8 and using the fact that A is
totally unimodular if c1 = 0. □
LEMMA 5.7.3. Let G be a connected non bipartite graph with n vertices and let
ℬ = {v1 , . . . , vn } be a set of linearly independent columns of the incidence matrix
of G. If K[G] is normal and 0 ∕= β ∈ ℝ+ ℬ ⊂ ℝn , then β can be written as
β = µ1 v′1 + ⋅ ⋅ ⋅ + µn v′n (µi ≥ 0),
such that v′1 , . . . , v′n
are linearly independent columns of the incidence matrix of G
defining a spanning subgraph of G which is unicyclic connected and non bipartite.
PROOF. For simplicity we identify the edges of G with the columns of its
incidence matrix. There are non negative real numbers λ1 , . . . , λn such that
(5.56) β = λ1 v1 + ⋅ ⋅ ⋅ + λn vn (λi ≥ 0).
Consider the subgraph H of G defined by the edges v1 , . . . , vn and the subgraph
H ′ of G defined by the set of edges {vi ∣λi > 0}. Using that v1 , . . . , vn are linearly
independent one rapidly derives that H is a spanning subgraph of G whose connected
components are unicyclic and non bipartite (cf. [170, Lemma 9.5]). Let H1 , . . . , H s
be the components of H and Zi the unique odd cycle of Hi , for i = 1, . . . , s. First
assume that H ′ has at most one odd cycle, say Z1 . For each cycle Zi with i ≥ 2 there is
v ji in Zi with λ ji = 0. Construct a graph K by removing v j1 , . . . , v js from H and then
adding an edge connecting Z1 with Zi for each i ≥ 2, such construction is possible
thanks to Corollary 5.6.14. This new graph K is a spanning connected subgraph of
G, is non bipartite and has a unique cycle Z1 . Thus in this case Eq. (5.56) is itself
the required expression for β. Next we assume that H ′ has at least two odd cycles
Z1 , Z2 . Let v be an edge joining Z1 with Z2 . Using that the edges of Z1 ∪ Z2 ∪ {v}
are related as in Equation (5.54), we can rewrite
β = µ1 v′1 + ⋅ ⋅ ⋅ + µn v′n (µi ≥ 0),
where v′1 , . . . , v′n are linearly independent edges such that the graph H ′′ determined
by the set {v′i ∣ µi > 0} has one cycle less than H ′ , thus by induction the result
follows. □
Next we give a description of the normal edge subrings having a unimodular
covering with support in the set of vertices of the edge polytope. The connected
version of the next result was shown in [112].
THEOREM 5.7.4. Let G be a graph with vertex set V = { x1 , . . . , xn } and let 𝒫
be its edge polytope. If K[G] is normal, then 𝒫 has a unimodular covering with
support in the vertices of 𝒫 if and only if G has at most one non bipartite component.
244 5. EDGE SUBRINGS

PROOF. ⇒ By Theorem 4.5.18 K[Ft] = A(𝒫 ). Hence by Theorem 5.7.2 we

obtain that the number of non bipartite components of G is at most 1.
⇐ Let G1 , . . . , Gc be the components of G and let ηi be the number of vertices
of Gi . Assume that G1 is non-bipartite and G2 , . . . , Gc are bipartite. As usual we
denote the set of columns of the incidence matrix of G by 𝒜 = {v1 , . . . , vq }.
To construct a unimodular covering of 𝒫 consider any subgraph H of G with
connected components H1 , . . . , Hc such that Hi is a spanning tree of Gi for i > 1
and H1 is a unicyclic connected spanning subgraph of G1 with a unique odd cycle.
For each such graph H consider its edge polytope ∆H . It is easy to see that ∆H is a
simplex of dimension d = n − c. From Propositions 5.3.7 and 3.2.19 it follows that
∆H is a simplex of normalized volume equal to 1.
We claim that the family of all simplices ∆H is a unimodular covering of 𝒫 .
Let β ∈ 𝒫 . By Carathéodory’s Theorem (see Theorem 3.1.37) one can write (after
permutation of edges):
β = λ1 v1 + ⋅ ⋅ ⋅ + λm vm (λi ≥ 0),
where m = n − (c − 1) and v1 , . . . , vm are linearly independent. Note that v1 , . . . , vm
define a subgraph H of G. For each i there is a subgraph Hi of Gi such that
H = H1 ∪ ⋅ ⋅ ⋅ ∪ Hc . Since v1 , . . . , vm are linearly independent, Hi is bipartite for
i > 1 and hence the number of edges of Hi is at most ηi − 1 for all i > 1. On the
other hand H1 has at most η1 edges. Altogether the number of edges of H is at most
η1 + (η2 − 1) + ⋅ ⋅ ⋅ + (ηc − 1), but since the number of edges of H is exactly this
number one concludes that the number of edges of Hi (resp. H1 ) is ηi − 1 (resp.
η1 ) for i > 1 (resp. i = 1). Hence H2 , . . . , Hc are trees. Since G1 is connected and
non bipartite applying Lemma 5.7.3 we can rewrite β such that H1 is a spanning
subgraph of G1 which is connected unicyclic and non bipartite. Therefore β ∈ ∆H ,
as required. If G1 is bipartite the proof follows using similar arguments as above.
Indeed is suffices to make G1 = ∅ and proceed as above. □

5.8. Normalizations of Rees algebras

First we discuss some typical behavior of the integral closure of the powers of an
edge ideal.
PROPOSITION 5.8.1. If I is the edge ideal of a graph, then I 2 is integrally closed.
PROOF. It is left as an exercise. □
M. Hochster pointed out the next example showing that the proposition does
not extend to the next power of I.
EXAMPLE 5.8.2 (Hochster’s configuration). Let G be the graph
 x2 sH 5. EDGE SUBRINGS s x 245
 6
H HHsx1 x4
s
x5 
s

 HH
x3 s HHs x
7

Note that f = (x1 x2 x3 )(x5 x6 x7 ) satisfies f ∈ I 3 ∖ I 3 .

DEFINITION 5.8.3. Let G be a graph. A Hochster configuration of order t (H-
configuration for short) of G consists of two odd cycles C2r+1 and C2s+1 satisfying
the following conditions:
(i) C2r+1 ∩ N(C2s+1 ) = ∅ and t = r + s + 1.
(ii) No chord of either C2r+1 or C2s+1 is an edge of G.
Given such a configuration, we consider the product of the monomials corre-
sponding to the two cycles to produce new generators of the integral closure of the
appropriate power of I(G).
PROPOSITION 5.8.4. If I is the edge ideal of a graph G admitting an H-
configuration of order t, then I t ∕= I t .
PROOF. The monomial obtained by multiplying all the variables in the two
cycles of the configuration is an element of I t ∖ I t . Indeed, let
δ = z1 ⋅ ⋅ ⋅ z2r+1 and γ = w1 ⋅ ⋅ ⋅ w2s+1
be the monomials corresponding to the two cycles in the configuration. By defini-
tion, we have r + s + 1 = t. Therefore δγ ∈ I t−1 ∖ I t . On the other hand δ2 ∈ I 2r+1
and γ2 ∈ I 2s+1 , thus (δγ)2 ∈ I 2(r+s+1) = I 2t . This shows that δγ ∈ I t . □
CONJECTURE 5.8.5 ([156]). The edge ideal of a graph G is normal if and only
if G admits no H-configurations.
In Corollary 5.8.10 we show that the conjecture has a positive answer.
The cone C(G), over the graph G, is obtained by adding a new vertex t to G and
joining every vertex of G to t.
EXAMPLE 5.8.6. A pentagon and its cone:
ts
BS
s  sB S

QQ   QQ B S
s
 Qs s

 Q Ss
A
A  B

As s

As B s

G C(G)
246 5. EDGE SUBRINGS

PROPOSITION 5.8.7. Let G be a graph and let C(G) be the cone over G. Then
there is an isomorphism
ϕ
R[I(G)t] ≃ k[C(G)].
PROOF. Let V = { x1 , . . . , xn } be the vertex set of G and let
R[I(G)t] = k[{ x1 , . . . , xn , t fi ∣ 1 ≤ i ≤ q}]
be the Rees algebra of the edge ideal I(G). We assume that f1 , . . . , fq are the
monomials in the xi ’s corresponding to the edges of G. Let
k[C(G)] = k[{txi , f j ∣ 1 ≤ i ≤ n, 1 ≤ j ≤ q}]
be the monomial subring of the cone. As these two algebras are integral domains
of the same dimension by Propositions 4.3.20 and 5.2.18, it follows that there is an
isomorphism
ϕ : R[I(G)t] −→ k[C(G)], induced by ϕ(xi ) = txi ; ϕ(t fi ) = fi . □
COROLLARY 5.8.8. Let G be a connected graph and let I = I(G) be its edge
ideal. Then k[G] is normal if and only if the Rees algebra R[It] is normal.
PROOF. ⇒) Let C(G) be the cone over G. By Proposition 5.8.7 there is an
isomorphism
R[I(G)t] ≃ k[C(G)].
Let Mw be a bowtie of C(G) with edge disjoint cycles Z1 and Z2 joined by the path
/ Z1 ∪ Z2 ∪ P, then w is a bowtie
P, it is enough to verify that Mw ∈ k[C(G)]. If t ∈
of G and Mw ∈ k[G]. Assume t ∈ Z1 ∪ Z2 , say t ∈ Z1 . If Z1 ∩ Z2 ∕= ∅, then
Mw ∈ k[C(G)]. On the other hand if Z1 ∩ Z2 = ∅, then Mw ∈ k[C(G)] because in
this case Z1 and Z2 are joined by the edge {t, z}, where z is any vertex in Z2 . It
remains to consider the case t ∈/ Z1 ∪ Z2 and t ∈ P, since G is connected there is a
path in G joining Z1 with Z2 . Therefore Mw = Mw1 for some bowtie w1 of G and
Mw ∈ k[G].
⇐) By Theorem 4.5.21 we obtain that k[G] is normal. □
In the corollary above the hypothesis that G is connected is essential, e.g., if
G consists of two disjoint triangles, then k[G] is a polynomial ring and R[It] is not
normal.
PROPOSITION 5.8.9. Let G be a graph and let G1 , . . . , Gr be the connected
components of G. Then R[I(G)t] is normal if and only if either:
(i) G is bipartite, or
(ii) exactly one of the components, say G1 , is non bipartite and K[G1 ] is a
normal domain.
PROOF. One may proceed as in the proof of Corollary 5.8.8. □
 5. EDGE SUBRINGS 247

COROLLARY 5.8.10 ([156]). The edge ideal of a graph G is normal if and only
if G admits no H-configurations.
REMARK 5.8.11. (a) Let G be a graph consisting of two disjoint triangles. Note
that G itself is an H-configuration but G is not contained in a bowtie, since the two
triangles cannot be joined by a path.
(b) If two disjoint odd cycles Z1 , Z2 of a graph G are in the same connected
component, then the two cycles form an H-configuration if and only if Z1 , Z2 are
not connected by an edge of G and Z1 , Z2 have no chords in G.
Integral closure of Rees algebras. Let G be a graph with vertices x1 , . . . , xn
and let I = I(G) be its edge ideal. Consider the endomorphism φ of the field
k(x1 , . . . , xn , t) defined by xi 7→ xi t, t 7→ t−2 . It induces an isomorphism
φ
R[It] 7−→ k[C(G)], xi 7→ txi , xi x j t 7→ xi x j ,
where R = k[x1 , . . . , xn ] is a polynomial ring over a field k and C(G) is the cone over
G obtained by adding a new variable t.
We thank Wolmer Vasconcelos [179, 180, 181] for showing us how to get the
description of the integral closure of R[It] given below. First we describe the integral
closure of k[C(G)].
If Z1 = { x1 , x2 , . . . , xr = x1 } and Z2 = {z1 , z2 , . . . , z s = z1 } are two edge disjoint
odd cycles in C(G). Then one has
Mw = x1 ⋅ ⋅ ⋅ xr z1 ⋅ ⋅ ⋅ z s
for some bow tie w of C(G), this follows readily because C(G) is a connected graph.
In particular Mw is in the integral closure of k[C(G)]. Observe that if t occurs in
Z1 or Z2 , then Mw ∈ k[C(G)] because in this case either Z1 and Z2 meet at a point,
or Z1 and Z2 are joined by an edge in C(G); see Proposition 5.6.13. Thus in order
to compute the integral closure of k[C(G)] the only bow ties that matter are those
defining the set:
/ supp(Mw )}.
ℬ = { Mw ∣ w is a bowtie in C(G) such that t ∈
Therefore by Theorem 5.6.12 we have proved:
PROPOSITION 5.8.12. k[C(G)] = k[C(G)][ℬ ].
To obtain a description of the integral closure of the Rees algebra of I note that
using φ together with the equality
r+1 s−1
2
∏ 2
∏
(tz1 ) (x2i−1 x2i ) (z2i z2i+1 )
i=1 i=1
Mw = ,
tx1
248 5. EDGE SUBRINGS

where xr+1 = x1 , we see that Mw is mapped back in the Rees algebra in the element
r+s r+s
Mw t 2 = x1 x2 ⋅ ⋅ ⋅ xr z1 z2 ⋅ ⋅ ⋅ z s t 2 .
If ℬ′ is the set of all monomials of this form, then one obtains:
PROPOSITION 5.8.13. R[It] = R[It][ℬ ′ ].
Rees semigroups of bipartite graphs. Here we use another approach to prove
the normality of the Rees algebras of a bipartite graph by looking closely at the
matrix defining a Rees algebra.
THEOREM 5.8.14 ([83]). Let G be a simple bipartite graph with n vertices and
q edges and let A = (ai j ) be its incidence matrix. If e1 , . . . , en are the first n unit
vectors in ℝn+1 and C is the matrix
a11 . . . a1q e1 ⋅ ⋅ ⋅ en
⎛ ⎞
⎜ .. .. .. ⎟
C=⎜ . . .
⎜ ⎟
⎝ an1 . . . anq
⎟
⎠
1 ... 1
obtained from A by adjoining a row of 1’s and the column vectors e1 , . . . , en , then C
is totally unimodular.
PROOF. Suppose that {1, . . . , m} and {m + 1, . . . , n} is the bipartition of the
graph G. Let C ′ be the matrix obtained by deleting the last n − m columns from
C. It suffices to show that C ′ is totally unimodular. First one successively subtracts
the rows 1, 2, . . . , m from the row n + 1. Then one reverses the sign in the rows
m+1, . . . , n. These elementary row operations produce a new matrix C ′′ . The matrix
C ′′ is the incidence matrix of a directed graph, namely, consider G as a directed
graph, and add one more vertex n + 1, and add the edges (i, n + 1) for i = 1, . . . , m.
The matrix C ′′ , being the incidence matrix of a directed graph, is totally unimodular
(see Exercise 3.6.9). As the last m column vectors of C ′′ are
e1 − en+1 , . . . , em − en+1 ,
one can successively pivot on the first nonzero entry of ei − en+1 for i = 1, . . . , m
and reverse the sign in the rows m + 1, . . . , n to obtain back the matrix C ′ . Here a
pivot on the entry c′st means transforming column t of C ′′ into the sth unit vector
by elementary row operations. Since pivoting preserves total unimodularity [141,
Lemma 2.2.20] one derive that C ′ is totally unimodular, and hence so is C. In the
next specific example we display C ′ and C ′′ . This proof was pointed out to us by
Bernd Sturmfels (see [83, p. 20]). □
EXAMPLE 5.8.15. To illustrate the constructions in the second proof above
consider the complete bipartite graph G = 𝒦2,3 . In this case the two ℤ-row equivalent
 5. EDGE SUBRINGS 249

matrices C ′ and C ′′ are

⎡ ⎤ ⎡ ⎤
1 1 1 0 0 0 1 0 1 1 1 0 0 0 1 0
⎢0 0 0 1 1 1 0 1⎥ ⎢ 0 0 0 1 1 1 0 1⎥
⎢ ⎥ ⎢ ⎥
⎢1 0 0 1 0 0 0 0⎥ ⎢−1 0 0 −1 0 0 0 0⎥
⎢ ⎥, ⎢ ⎥.
⎢0
⎢ 1 0 0 1 0 0 0⎥
⎥
⎢ 0
⎢ −1 0 0 −1 0 0 0⎥
⎥
⎣0 0 1 0 0 1 0 0⎦ ⎣ 0 0 −1 0 0 −1 0 0⎦
1 1 1 1 1 1 0 0 0 0 0 0 0 0 −1 −1

respectively. Note that pivoting in the last two columns of C ′′ amounts to adding
row 1 to row 6, and then adding row 2 to row 6.

There is a family of {0, 1} matrices called complement totally unimodular that

preserve total unimodularity when adjoining a row or column consisting of 1’s, see
[173, Lemma 12.3.4].

COROLLARY 5.8.16 ([156]). If A is the incidence matrix of a bipartite graph,

then the Rees semigroup

ℕ(v1 , 1) + ⋅ ⋅ ⋅ + ℕ(vq , 1) + ℕe1 + ⋅ ⋅ ⋅ + ℕen ⊂ ℕn+1

is normal, where ei denotes the ith unit vector of ℝn+1 .

PROOF. It follows from Theorems 5.8.14 and 3.5.7. □

Exercises
5.8.17. Let F = { f1 , . . . , f6 } and I = (F), where

f1 = x1 x2 , f2 = x3 x4 x5 , f3 = x1 x3 , f4 = x2 x4 , f5 = x2 x5 , f6 = x1 x5 .

Prove that k[F] is normal and that R[It] is not normal.

5.8.18. Let G be a graph on the vertex set V = { x1 , . . . , xn } and choose a new

set of vertices y1 , . . . , yn , define S (G), the suspension of G, to be the graph obtained
by attaching to each vertex xi the new vertex yi and the edge { xi , yi }. If I(G) is a
normal ideal, then I(C(G)) and I(S (G)) are normal ideals.

5.8.19. Let G be a graph and let m = (r + s)/2 be the least positive integer such
that G has two vertex disjoint odd cycles of lengths r and s. If I = I(G) is the edge
ideal of G, prove that I i is integrally closed for i < m.

5.8.20. If G is any non discrete graph with n vertices and C(G) its cone, prove
that the edge subring k[C(G)] has dimension n + 1.
250 5. EDGE SUBRINGS

5.8.21. Let G be a graph with vertices x1 , . . . , xn and I ⊂ R = k[x1 , . . . , xn ] its

edge ideal. If ϕ, ψ are the maps of k-algebras defined by the diagram
ϕ ϕ ϕ
R[t1 , . . . , tq ] / R[It] ti /o o/ /o / T fi xi /o o/ o/ / xi
O O
ψ O ψ O ψ
O O
  
k[C(G)] fi xti

where C(G) is the cone of G and f1 , . . . , fq are the edge generators, then ker(ϕ) =
ker(ψ).
5.8.22. Let R = k[x1 , x2 , x3 , y1 , y2 , y3 , y4 ] and I = I1 J3 + I3 J1 , where Ir (resp. Jr )
denote the ideal of R generated by the square-free monomials of degree r in the xi
variables (resp. yi variables). If z = x1 x2 x3 y21 y2 y3 y4 , then z2 ∈ I 4 and z ∈ I 2 ∖ I 2 .
5.8.23. Let G be a graph. Explain the differences between a circuit of G and an
H-configuration of G.
5.8.24. Consider the matrices
⎛ ⎞
0 0 1 0
⎜
⎜ 0 0 1 0 ⎟
⎟
⎛ ⎞ ⎜ 0 1 0 0 ⎟
1 0 0 1 ⎜ ⎟
⎜ 0
⎜ 0 1 0 0 ⎟ ( )
1 0 1 ⎟
⎟, ⎟, A
.
⎜ ⎟
V=⎜
⎝ 0 A=⎜ 1 0 0 0 B=
0 1 1 ⎠ ⎜ ⎟ 1111
⎜ 1 0 0 0 ⎟
1 1 1 1 ⎜ ⎟
⎜
⎜ 1 0 0 1 ⎟
⎟
⎝ 0 1 0 1 ⎠
0 0 1 1
Note det(V) = −2. Prove that A is totally unimodular but B is not.
5.8.25. Let G be a bipartite graph and let I(G) be its edge ideal. Prove that the
toric ideal of the Rees algebra of I(G) has a universal Gröbner basis consisting of
square-free binomials.
5.8.26. Let I be the edge ideal of a graph G. Prove that I is minimally non-
normal (i.e., I is not normal and any proper minor I ′ of I is normal) if and only if G
consists of two vertex disjoint odd cycles.

5.9. Monomial subrings of complete graphs

Here we treat two special types of monomial subrings attached to complete graphs
and compute their Hilbert series, multiplicity and a-invariant.
 5. EDGE SUBRINGS 251

Complete bipartite graphs. Let 𝒦m,n be the complete bipartite graph with
bipartition V1 = { x1 , . . . , xm }, V2 = {y1 , . . . , yn } and let
B = k[{ti j ∣ 1 ≤ i ≤ m, 1 ≤ j ≤ n}]
be a polynomial ring over a field k. Consider the graded homomorphism of k-
algebras
ψ : B −→ k[𝒦m,n ] = k[xi y j ’s], ψ(ti j ) = xi y j .
One of the results that simplify the study of an edge subring associated to a
complete bipartite graph is the fact that its toric ideal is a determinantal prime ideal:
THEOREM 5.9.1. [30, Theorem 2.5] Let X = (xi j ) be an m × n matrix of
indeterminates over a field k and let It (X) be the ideal generated by the t-minors of
X. If m ≤ n and 1 ≤ t ≤ m + 1, then
height(It (X)) = (m − t + 1)(n − t + 1).
PROPOSITION 5.9.2. The toric ideal P of k[𝒦m,n ] is generated by the 2 × 2 minors
of a generic m × n matrix.
PROOF. We claim that the kernel of ψ is generated by the 2 × 2 minors of a
generic m × n matrix. Indeed let X = (ti j ). It is clear that the ideal I2 (X), generated
by the 2 × 2 minors of X, is contained in P = ker(ψ). Since the graph is bipartite,
by Corollary 5.2.19, one has that the dimension of k[𝒦m,n ] is equal to m + n − 1.
Therefore
height(P) = mn − (m + n − 1) = (m − 1)(n − 1) = height(I2 (X)),
the latter equality follows from Theorem 5.9.1. Since P and I2 (X) are both prime
ideals, we have I2 (X) = P. □
The Cohen-Macaulayness and Gorensteiness of algebras that include k[𝒦m,n ],
has been dealt with in [28].
PROPOSITION 5.9.3. [187, Proposition 9.1.3] Let P be the toric ideal of the edge
subring k[𝒦m,n ]. If n ≥ m, then the Hilbert series of B/P is given by
m−1
∑ (m − 1)(n − 1)
zi
i i
F(B/P, z) = i=0 .
(1 − z)n+m−1
COROLLARY 5.9.4. If n ≥ m, then the a-invariant and the multiplicity of the
edge subring k[𝒦m,n ] are given by:
( )
m+n−2
a(k[𝒦m,n ]) = −n and e(k[𝒦m,n ]) = .
m−1
PROOF. It follows from Remark 1.6.7. □
252 5. EDGE SUBRINGS

PROPOSITION 5.9.5. Let G be an spanning connected subgraph of 𝒦m,n . If

n ≥ m, then the a-invariant of k[G] satisfies

a(k[G]) ≤ a(k[𝒦m,n ]) = −n.

PROOF. From Corollary 5.9.4 one has a(k[𝒦m,n ]) = −n. Hence using the
Danilov-Stanley formula and Proposition 4.3.29 the inequality follows. □
Complete graphs. Let R = k[x1 , . . . , xn ] be a polynomial ring over the field k,
and let 𝒦n be the complete graph on the vertex set X = { x1 , . . . , xn }. For i < j we
set fi j = xi x j . Let
k[𝒦n ] = k[{ fi j ∣ 1 ≤ i < j ≤ n}]
be the k-subring of R spanned by the fi j . Consider the homomorphism

ψ : Bn = k[{ti j ∣ 1 ≤ i < j ≤ n}] → k[𝒦n ], induced by ti j → fi j .

The ideal Pn = ker(ψ) is the presentation ideal or toric ideal of k[𝒦n ].

PROPOSITION 5.9.6 ([185]). If the terms in Bn are ordered lexicographically by

t12 ≺ ⋅ ⋅ ⋅ ≺ t1n < t23 ≺ ⋅ ⋅ ⋅ ≺ t2n ≺ ⋅ ⋅ ⋅ ≺ t(n−1)n , then the set

ℬ = {ti j tk` − ti` t jk , tik t j` − ti` t jk ∣ 1 ≤ i < j < k < ` ≤ n}

is a minimal generating set for Pn and ℬ is a reduced Gröbner basis for Pn .

EXAMPLE 5.9.7. If k[{ti j ∣ 1 ≤ i < j ≤ 5}] has the lex ordering induced by

t45 ≺ t34 ≺ t35 ≺ t23 ≺ t24 ≺ t25 ≺ t12 ≺ t13 ≺ t14 ≺ t15 .

Then the reduced Gröbner basis for the toric ideal P of k[𝒦5 ] is equal to:
t35 t24 − t45 t23 , t34 t25 − t45 t23 , t45 t23 t13 − t34 t35 t12 ,
t25 t13 − t35 t12 , t35 t14 − t45 t13 , t23 t14 − t34 t12 ,
t34 t15 − t45 t13 , t23 t15 − t35 t12 , t24 t15 − t45 t12 .
t24 t13 − t34 t12 t25 t14 − t45 t12 .
THEOREM 5.9.8 ([185]). Let Bn /Pn be the presentation of k[𝒦n ]. If n ≥ 3, then
the Hilbert series Fn (z) of Bn /Pn satisfies:
⌊ n2 ⌋ ( )
n n(n − 3) ∑ n m
(1 − z) Fn (z) = 1 + z+ z
2 2m
m=2

COROLLARY 5.9.9. The multiplicity e(k[𝒦n ]), of the ring k[𝒦n ], is equal to
2n−1 − n and its a-invariant is equal to −⌈n/2⌉.
 5. EDGE SUBRINGS 253

Bounding the multiplicity. Let G be a connected graph with n vertices and q

edges. Then according to Corollaries 5.9.4 and 5.9.9 one has sharp bounds for the
multiplicity of the edge subring:
⎧ n−1
⎨2 − n, if G is non bipartite,
e(k[G]) ≤ (
m−1 , if G is a spanning subgraph of 𝒦m,r .
⎩ m+r−2)

It is easy to produce examples where these bounds are very lose. The extreme
cases being the unicyclic connected non bipartite graphs and the trees, because
in both cases the multiplicity of k[G] is equal to 1. In our case adding edges
to G increases the volume of the edge polytope (see Exercise 4.5.27), hence the
multiplicity is an increasing function of the edges. If G is non bipartite in the
average adding an edge to G increases the multiplicity by
2n−1 − n − e(k[G])
(n) .
2 −q
For planar bipartite graphs these bounds can be improved, see Section 5.5.

Exercises
5.9.10. Let A = k[x1 , . . . , xm ] and R = k[y1 , . . . , yn ] be two polynomial rings
over a field k, prove that the Segre product
(A0 ⊗k R0 ) ⊕ (A1 ⊗k R1 ) ⊕ ⋅ ⋅ ⋅ ⊂ A ⊗k R
is isomorphic to S = k[xi y j ’s] ⊂ k[xi ’s, y j ’s].
5.9.11. Let R = k[x1 , . . . , xn ] be a polynomial ring over a field k. Set 𝔪 =
(x1 , . . . , xn ). Prove that the toric ideal of the Rees algebra R[𝔪t] is the ideal generated
by the 2-minors of the following matrix of indeterminates:
( )
x1 x2 ⋅ ⋅ ⋅ xn
X= .
t1 t2 ⋅ ⋅ ⋅ tn
Chapter 6
Edge Cones of Graphs

Let G be an arbitrary simple graph. In this we chapter we give an explicit

combinatorial description of the edge cone of G. As an application we prove the
classical marriage theorem. We study in detail the facets of the edge cone of a
connected bipartite graph, as an application we compute the a-invariant of the Rees
algebra of the edge ideal of G. In Chapter 7 these results will be applied to compute
the a-invariant and the canonical module of an edge subring.
The main tools used to show these results are linear algebra (Farkas’s Lemma,
incidence matrices of graphs, Carathéodory’s Theorem, Heger’s Theorem), graph
theory and polyhedral geometry (finite basis theorem, facet structure of polyhedra).
The proofs require a careful graph theoretical analysis.

6.1. A combinatorial representation

Let G be a simple graph with vertex set V = V(G) = {v1 , . . . , vn } and edge set
E = E(G) = {z1 , . . . , zq }. Thus every edge zi is an unordered pair of distinct vertices
zi = {vi j , vik } and we set αi = ei j + eik , where ei is the ith unit vector in ℝn . Recall
that the incidence matrix of G, denoted by MG , is the n × q matrix whose columns
are precisely the vectors α1 , . . . , αq .
We set 𝒜G (or simply 𝒜 if G is understood) equal to the set {α1 , . . . , αq }. Since
αi represents an edge of G, sometimes αi is called an edge or an edge vector. The
edge cone of G, denoted by ℝ+ 𝒜, is by definition the cone generated by 𝒜. Notice
that ℝ+ 𝒜 = ∕ (0) if G is not a discrete graph. By Corollary 5.3.9 one has
n − c0 (G) = rank (MG ) = dim ℝ+ 𝒜,
where c0 (G) is the number of bipartite components of G.
LEMMA 6.1.1. If vi is not an isolated vertex of the graph G, then the set
F = Hei ∩ ℝ+ 𝒜 is a proper face of the edge cone.
PROOF. Notice that F ∕= ∅ and ℝ+ 𝒜 ⊂ He+i . Since vi is not an isolated vertex
ℝ+ 𝒜 ∕⊂ Hei . □
255
256 6. EDGE CONES OF GRAPHS

Let A be an independent set of vertices of G. The supporting hyperplane of the

edge cone defined by ∑ ∑
xi = xi
vi ∈A vi ∈N(A)
will be denoted by HA . The sum over an empty set is defined to be zero.
LEMMA 6.1.2. If A is an independent set of vertices of the graph G and F =
ℝ+ 𝒜 ∩ HA , then either F is a proper face of the edge cone or F = ℝ+ 𝒜.
PROOF. It suffices to prove ℝ+ 𝒜 ⊂ HA− . Take and edge {v j , v` } of the graph G.
If {v j , v` } ∩ A ∕= ∅, then e j + e` is in HA , else e j + e` is in HA− . □
DEFINITION 6.1.3. The support of a vector β = (βi ) ∈ ℝn is defined as
supp(β) = {βi ∣ βi ∕= 0}.
LEMMA 6.1.4 ([186]). Let G1 , . . . , Gr be the connected components of G. If
G1 is a tree with at least two vertices and G2 , . . . , Gr are unicyclic non bipartite
graphs, then ker(MGt ) = (β) for some β ∈ ℝn with supp(β) = {1, −1} such that
V(G1 ) = {vi ∈ V ∣ βi = ±1}.
PROOF. If G is a tree, using induction on the number of vertices it is not hard to
show that ker(MGt ) = (β), where supp(β) = {1, −1}. On the other hand if G = Gi for
some i ≥ 2, then MGt is non singular and ker(MGt ) = (0). The general case follows
readily by writing MGt as a “diagonal” matrix
MGt = diag(MGt 1 , . . . , MGt m ),
where MGi is the incidence matrix of the graph Gi . Observe the equality V(G1 ) =
{vi ∈ V ∣ βi = ±1}. □
DEFINITION 6.1.5. For any set of vertices S of a graph G, the induced subgraph
⟨S ⟩ is the maximal subgraph of G with vertex set S .
To determine whether HA defines a facet of ℝ+ 𝒜 consider the subgraph L =
L1 ∪ L2 , where L1 is the subgraph of G with vertex set and edge set
V(L1 ) = A ∪ N(A) and E(L1 ) = {z ∈ E(G)∣ z ∩ A ∕= ∅}
respectively, and L2 =<S > is the subgraph of G induced by S = V ∖ V(L1 ). The
vectors in 𝒜 ∩ HA correspond precisely to the edges of L.
REMARK 6.1.6. Let F be a facet of ℝ+ 𝒜 defined by the hyperplane Ha . If
the rank of MG is n, then it is not hard to see that Ha is generated by a linearly
independent subset of 𝒜. See Proposition 3.1.12.
THEOREM 6.1.7 ([186]). If rank(MG ) = n, then F is a facet of ℝ+ 𝒜 if and only
if either
 6. EDGE CONES OF GRAPHS 257

(a) F = Hei ∩ ℝ+ 𝒜 for some i, where all the connected components of G ∖ {vi }
are non bipartite graphs, or
(b) F = HA ∩ ℝ+ 𝒜 for some independent set A ⊂ V such that L1 is a
connected bipartite graph, and the connected components of L2 are non
bipartite graphs.
PROOF. Assume that F is a facet of ℝ+ 𝒜. By Remark 6.1.6 we may assume
that there is a ∈ ℝn such that
(i) F = ℝ+ 𝒜 ∩ Ha , ⟨a, αi ⟩ ≤ 0 ∀ i ∈ {1, . . . , q}, and
(ii) α1 , . . . , αn−1 are linearly independent vectors in Ha .
Consider the subgraph D of G whose edges correspond to α1 , . . . , αn−1 and its
vertex set is the union of the vertices in the edges of D. Let Γ be the transpose of
the incidence matrix of D, and let D1 , . . . , Dr be the connected components of D.
Without loss of generality one may assume
Γ1 0 . . . 0
⎡ ⎤
⎢ 0 Γ2 . . . 0 ⎥
Γ = ⎢ .. .. ⎥ , rank(Γ) = n − 1,
⎢ ⎥
.. . .
⎣. . . .⎦
0 0 . . . Γr
where Γi is the transpose of the incidence matrix of Di . We set ni and qi equal to
the number of vertices and edges of Di respectively. Let Θ be the matrix with rows
α1 , . . . , αn−1 . We consider two cases.
Case (I): Assume Θ has a zero column, hence it has exactly one zero column
and D has vertex set V(D) = V ∖ {vi }, for some i. This implies
r
∑ r
∑
ni = qi = n − 1.
i=1 i=1

Since ni ≤ qi for all i, one obtains ni = qi for all i. Note that c0 (D) = 0, by
Lemma 5.3.4. Therefore D1 , . . . , Dr are unicyclic non bipartite graphs. As any
component of G ∖ {vi } contains a component of D, we are in case (a).
Case (II): Assume that all the columns of Θ are non zero, that is, D has vertex
set equal to V. Note that in this case
r
∑ r
∑
(6.57) n= ni = 1 + qi .
i=1 i=1

Since rank (Γ) = n − c0 (D) = n − 1, we get that D has exactly one bipartite
component, say D1 . Hence n1 − 1 ≤ q1 and ni ≤ qi for all i ≥ 2, which together
with Eq. (6.57) yields q1 = n1 − 1 and ni = qi for all i ≥ 2. Altogether D1 is a tree
and D2 , . . . , Dr are unicyclic non bipartite graphs. By Lemma 6.1.4 there is β ∈ ℝn
258 6. EDGE CONES OF GRAPHS

such that
ker(Γ) = (β) and supp(β) ⊂ {1, −1}.
We may assume a = β, because Ha = Hβ and a ∈ (β). Set
A = {vi ∈ V ∣ ai = 1} and B = {vi ∈ V ∣ ai = −1}.
Note that A ∕= ∅, because D has no isolated vertices and its vertex set is V. We will
show that A is an independent set in G and B = N(A), where the neighbor set of A
is taken with respect to G.
On the contrary if {vi , v j } is an edge of G for some vi , v j in A, then αk = ei + e j
and by (i) we get ⟨a, αk ⟩ ≤ 0, which is impossible because ⟨a, αk ⟩ = 2. This proves
that A is an independent set.
Next we show N(A) = B. If vi ∈ N(A), then αk = ei + e j for some v j in A, using
(i) we obtain ⟨a, αk ⟩ = ai + 1 ≤ 0 and ai = −1, hence vi ∈ B. Conversely if vi ∈ B,
since D has no isolated vertices, there is 1 ≤ k ≤ n − 1 so that αk = ei + e j , for some
j, by (ii) we obtain ⟨a, αk ⟩ = −1 + a j = 0, which shows that v j ∈ A and vi ∈ N(A).
From the proof of Lemma 6.1.4 we obtain V(D1 ) = A ∪ N(A). Therefore
α1 , . . . , αn−1 are in HA and Ha = HA . Observe that L1 (see notation before Re-
mark 6.1.6) cannot contain odd cycles because A is independent and every edge of
L1 contains a vertex in A. Hence L1 is bipartite, that L1 is connected follows from
the fact that D1 is a spanning tree of L1 . Note that every connected component of
L2 is non bipartite because it contains some Di , with i ≥ 2.
Conversely assume that F is as in (a). Since the vectors in 𝒜 ∩ Hei correspond
precisely to the edges of G ∖ {vi } one obtains that F is a facet of the edge cone. If
F is as in (b), a similar argument shows that F is also a facet. □
COROLLARY 6.1.8 ([186]). Let G be a connected non bipartite graph. Then a
vector x = (x1 , . . . , xn ) ∈ ℝn is in ℝ+ 𝒜 if and only if x is a feasible solution of the
system of linear inequalities
− xi ≤ 0, i = 1, . . . , n
∑ ∑
vi ∈A xi −
vi ∈N(A) xi ≤ 0, for all independent sets A ⊂ V.
PROOF. Let ℝ+ 𝒜 = Hb−1 ∩ ⋅ ⋅ ⋅ ∩ Hb−m be the irreducible representation of the edge
cone as an intersection of closed half-spaces. By Theorem 3.1.31 the set Hbi ∩ ℝ+ 𝒜
is a facet of ℝ+ 𝒜, and we may apply Theorem 6.1.7. □
THEOREM 6.1.9. If G is a connected graph and F is a facet of the edge cone of
G, then either
(a) F = ℝ+ 𝒜 ∩ { x ∈ ℝn ∣ xi = 0} for some 1 ≤ i ≤ n, or
(b) F = ℝ+ 𝒜 ∩ HA for some independent set A of G.
PROOF. By Theorem 6.1.7 one may assume that G is a bipartite graph with
n ≥ 3 vertices. Notice that
(6.58) F = ℝ+ ℬ ,
 6. EDGE CONES OF GRAPHS 259

where ℬ = {αi ∈ 𝒜 ∣ ⟨αi , a⟩ = 0}. Since dim(F) = n − 2 one may assume that
α1 , . . . , αn−2 are linearly independent vectors in ℬ . There is 0 ∕= a ∈ ℝn such that
(i) F = ℝ+ 𝒜 ∩ Ha , ⟨a, αi ⟩ = 0 ∀ i ∈ {1, 2, . . . , n − 2}, and
(ii) ⟨a, αi ⟩ ≤ 0 for i ∈ {1, 2, . . . , q}.
Consider the subgraph D of G whose edges correspond to α1 , . . . , αn−2 and
its vertex set is the union of the vertices in the edges of D. Set k = ∣V(D)∣. By
Corollary 5.3.9 one has:
n − 2 = rank (MD ) = k − c0 (D),
where MD is the incidence matrix of D and c0 (D) is the number of bipartite com-
ponents of D. Thus 0 ≤ n − k = 2 − c0 (D). This shows that either c0 (D) = 1 and
k = n − 1 or c0 (D) = 2 and k = n.
Case (I): Assume c0 (D) = 1 and k = n − 1. Set V(D) = {v1 , . . . , vn−1 }. As D is
a tree with n − 2 edges and ⟨αi , a⟩ = 0 for i = 1, . . . , n − 2, applying Lemma 6.1.4,
one may assume that a = (a1 , . . . , an−1 , an ), where ai = ±1 for 1 ≤ i ≤ n − 1.
First assume an = 0. Set
A = {vi ∈ V ∣ ai = 1} and B = {vi ∈ V ∣ ai = −1}.
Note that ∅ = ∕ A ⊂ V(D), because D is a tree with at least two vertices. We will
show that A is an independent set of G and B = NG (A).
If A is not an independent set of G, there is an edge {vi , v j } of G for some vi , v j
in A. Thus αk = ei + e j and by (ii) we get ⟨a, αk ⟩ ≤ 0, which is impossible because
⟨a, αk ⟩ = 2. This proves that A is an independent set.
Next we will show NG (A) = B. If vi ∈ N(A), then αk = ei + e j for some v j in A,
using (ii) we obtain ⟨a, αk ⟩ = ai + 1 ≤ 0 and ai = −1, hence vi ∈ B. Conversely if
vi ∈ B, since D has no isolated vertices, there is 1 ≤ k ≤ n − 2 so that αk = ei + e j ,
for some j, by (i) we obtain ⟨a, αk ⟩ = −1 + a j = 0, which shows that v j ∈ A and
vi ∈ N(A). Therefore Ha = HA and F is as in (b).
Next we will assume an > 0. In this case consider a′ = −en . Note the following
(c) ⟨αi , a′ ⟩ = 0 for i = 1, . . . , n − 2.
(d) If α j ∈ ℝ(α1 , . . . , αn−2 ) ⇒ ⟨α j , a′ ⟩ = ⟨α j , a⟩ = 0.
(e) If α j ∈/ ℝ(α1 , . . . , αn−2 ), then ⟨α j , a′ ⟩ = −1 and ⟨α j , a⟩ < 0.
To prove (e) assume α j ∈ / ℝ(α1 , . . . , αn−2 ), then α j = ek + en , because otherwise
the “edge” α j added to the tree D form a graph with a unique even cycle, a contra-
diction. Thus ⟨α j , a′ ⟩ = −1. On the other hand ⟨α j , a⟩ < 0, because if ⟨α j , a⟩ = 0,
then Ha contains the linearly independent vectors α1 , . . . , αn−2 , α j , a contradiction
to dim(F) = n − 2.
Therefore given 1 ≤ j ≤ q, then ⟨α j , a⟩ = 0 if and only if ⟨α j , a′ ⟩ = 0. Notice
the equality
(6.59) ℝ+ 𝒜 ∩ Ha′ = ℝ+ ℬ ′ ,
260 6. EDGE CONES OF GRAPHS

where ℬ ′ = {αi ∈ 𝒜 ∣ ⟨αi , a′ ⟩ = 0}. Hence from Eqs. (6.58) and (6.59) we derive
F = ℝ+ 𝒜 ∩ Ha′ , as required. The case an < 0 can be treated similarly.
Case (II): Assume c0 (D) = 2 and k = n. Let D1 and D2 be the two components
of D and set V1 = V(D1 ) and V2 = V(D2 ).
Using Lemma 6.1.4 we can relabel the vertices of the graph D to write a = rb+sc,
where 0 ∕= r ≥ s ≥ 0 are rational numbers and
b = (b1 , . . . , bm , 0, . . . , 0), c = (0, . . . , 0, cm+1 , . . . , cn ),
where V1 = {v1 , . . . , vm }, bi = ±1 for i ≤ m, and ci = ±1 for i > m. Set a′ = b.
Note the following:
(f) ⟨αi , a′ ⟩ = 0 for i = 1, . . . , n − 2.
(g) If α j ∈ ℝ(α1 , . . . , αn−2 ) ⇒ ⟨α j , a′ ⟩ = ⟨α j , a⟩ = 0.
(h) If α j ∈ / ℝ(α1 , . . . , αn−2 ), then ⟨α j , a⟩ < 0 and ⟨α j , a′ ⟩ = −1 .
Let us prove assertion (h). Assume α j ∈ / ℝ(α1 , . . . , αn−2 ). Since
dim(F) = n − 2
by (ii) we have ⟨α j , a⟩ < 0. Observe that if an “edge” αk has vertices in V1 (resp.
V2 ), then ⟨αk , a⟩ = 0. Indeed if we add the edge αk to the tree D1 (resp. D2 ) we get
a graph with a unique even cycle and this implies that α1 , . . . , αn−2 , αk are linearly
dependent, that is, ⟨αk , a⟩ = 0. Thus α j = ei + e` for some vi ∈ V1 and v` ∈ V2 .
From the inequality
⟨α j , a⟩ = r⟨α j , b⟩ + s⟨α j , c⟩ = rbi + sc` < 0
we obtain bi = −1 = ⟨α j , a′ ⟩, as required. In particular note that Ha′ is a supporting
hyperplane of ℝ+ 𝒜 because ℝ+ 𝒜 ⊂ Ha−′ . We may now proceed as in case (I) to
obtain the equality
F = ℝ+ 𝒜 ∩ Ha = ℝ+ 𝒜 ∩ Ha′ .
Using the fact that D1 has no isolated vertices, one may also proceed as in case (I)
to prove that Ha′ = HA for the independent set of vertices given by A = {vi ∣ bi = 1}.
□
THEOREM 6.1.10. If G is a connected graph, then
( ) (n )
∩ ∩ ∩
ℝ+ 𝒜 = HA− He+i ,
A∈ℱ i=1
where ℱ is the family of all the independent sets of vertices of G and He+i is the
closed halfspace { x ∈ ℝn ∣ xi ≥ 0}.
PROOF. Set C = ℝ+ 𝒜. From Theorem 6.1.9 and Corollary 3.1.19 we get
( ) (n )
∩ ∩ ∩
C = aff(C) ∩ HA− He+i .
A∈ℱ i=1
 6. EDGE CONES OF GRAPHS 261

We may assume that G is bipartite, otherwise aff(C) = ℝn and there is nothing to

prove. If (V1 , V2 ) is the bipartition of G, then aff(ℝ+ 𝒜) is the hyperplane
∑ ∑
xi = xi ,
vi ∈V1 vi ∈V2

because dim(ℝ+ 𝒜) = n − 1. As HV−1 ∩ HV−2 = HV1 = aff(C), we obtain the required

equality. □
The next two results give an explicit representation by closed halfspaces of the
edge cone of an arbitrary graph.
COROLLARY 6.1.11. If G is a graph, then
( ) (n )
∩ ∩ ∩
ℝ+ 𝒜 = HA− He+i , (∗)
A i=1

where the first intersection is taken over all the independent sets of vertices of G
and He+i = { x ∈ ℝn ∣ xi ≥ 0}.
PROOF. Let G1 , . . . , Gr be the connected components of G. For simplicity of
notation we assume r = 2 and V(G1 ) = {v1 , . . . , vm }. There is a decomposition
ℝ+ 𝒜 = ℝ+ 𝒜G1 ⊕ ℝ+ 𝒜G2 .
Let δ be a vector in the right hand side of Eq. (∗). One can write
δ = (δi ) = β + γ = (δ1 , . . . , δm , 0, . . . , 0) + (0, . . . , 0, δm+1 , . . . , δn ).
Let A be an independent set of G1 . Note NG (A) = NG1 (A), hence
∑ ∑ ∑
δi ≤ δi = δi .
vi ∈A vi ∈NG (A) vi ∈NG1 (A)

Applying Theorem 6.1.10 yields β ∈ ℝ+ 𝒜G1 . Similarly one has γ ∈ ℝ+ 𝒜G2 . Hence
δ ∈ ℝ+ 𝒜G , as required. □
COROLLARY 6.1.12. If G1 , . . . , Gr are the connected components of G, then
( ) (n )
∩ ∩ ∩
−
ℝ+ 𝒜 = HA Hei ,
+

A i=1

where the first intersection is taken over all the independent sets of vertices of the
connected components Gi of G and He+i = { x ∈ ℝn ∣ xi ≥ 0}.
PROOF. It follows at once from Corollary 6.1.11. □
262 6. EDGE CONES OF GRAPHS

COROLLARY 6.1.13. Let G be a graph. Then a vector x = (xi ) ∈ ℝn is in ℝ+ 𝒜

if and only if x is a solution of the system of linear inequalities
− xi ≤ 0, i = 1, . . . , n
∑ ∑
vi ∈A xi − vi ∈N(A) xi ≤ 0, for all independent sets A ⊂ V.
PROOF. It follows at once from Corollary 6.1.11. □
THEOREM 6.1.14. If G is a graph and F is a facet of the edge cone of G, then
either
(a) F = ℝ+ 𝒜 ∩ { x ∈ ℝn ∣ xi = 0} for some 1 ≤ i ≤ n, or
(b) F = ℝ+ 𝒜 ∩ HA for some independent set A of G.
PROOF. By Corollary 6.1.11 we can write ℝ+ 𝒜 = aff(ℝ+ 𝒜) ∩ H1− ∩ ⋅ ⋅ ⋅ ∩ Hr− for
some hyperplanes H1 , . . . , Hr , such that none of the halfspaces H −
j can be omitted
in the intersection and each H j is either of the form H−ei or H j = HA for some
independent set A. By Theorem 3.1.31(c), the facets of ℝ+ 𝒜 are precisely the sets
F1 , . . . , Fr , where Fi = Hi ∩ ℝ+ 𝒜. □

Exercises
6.1.15. Let G = 𝒦n be the complete graph on n vertices. Then a vector x ∈ ℝn
is in ℝ+ 𝒜G if and only if x = (x1 , . . . , xn ) satisfies

∑ − xi ≤ 0, i = 1, . . . , n
xi − j∕=i x j ≤ 0, i = 1, . . . , n.
In addition if n ≥ 4, these equations define all the facets of ℝ+ 𝒜G .
6.1.16. If G is a triangle with vertices { x1 , x2 , x3 }, then the edge cone of G has
three facets defined by
x1 ≤ x2 + x3 , x2 ≤ x1 + x3 , x3 ≤ x1 + x2 .
Note that x1 ≥ 0, x2 ≥ 0 and x3 ≥ 0 define proper faces of dimension 1.

6.2. Studying the bipartite case

For connected bipartite graphs we will present sharper results about the irreducible
representations of edge cones. Let G be a connected bipartite graph with bipartition
(V1 , V2 ). To avoid repetitions we will continue using the notation introduced in
Section 6.1.
PROPOSITION 6.2.1. If A is an independent set of G such that A ∕= Vi for i = 1, 2,
then F = ℝ+ 𝒜 ∩ HA is a proper face of the edge cone.
 6. EDGE CONES OF GRAPHS 263

PROOF. Assume N(A) = V2 . Take any vi ∈ V1 ∖ A and any v j ∈ V2 adjacent to

/ HA . This means that ℝ+ 𝒜 ⊈ HA . Thus we may assume N(A) ∕= Vi
vi , then ei + e j ∈
for i = 1, 2.
Case (I): N(A) ∩ Vi ∕= ∅ for i = 1, 2. If the vertices in N(A) ∩ Vi for i = 1, 2 are
only adjacent to vertices in A, then pick vertices vi ∈ N(A) ∩ Vi and note that there
is no path between v1 and v2 , a contradiction. Thus there must be a vector in the
edge cone which is not in HA .
Case (II): A ⊊ V1 . If the vertices in N(A) are only adjacent to vertices in A.
Then a vertex in A cannot be joined by a path to a vertex in V2 ∖ N(A), a contradiction.
As before we obtain ℝ+ 𝒜 ∕⊂ HA . □
PROPOSITION 6.2.2. Let ℱ the family of independent sets A of G such that
HA ∩ ℝ+ 𝒜G is a facet. If A is in ℱ and Vi ∩ A ∕= ∅ for i = 1, 2, then the halfspace
HA− is redundant in the following expression of the edge cone
( ) (n )
∩ ∩ ∩
ℝ+ 𝒜 = aff(ℝ+ 𝒜) ∩ HA− He+i .
A∈ℱ i=1

PROOF. Set 𝒜 = {α1 , . . . , αq }. One can write A = A1 ∪ A2 with Ai ⊂ Vi for

i = 1, 2. There are α1 , . . . , αn−2 linearly independent vectors in HA ∩ ℝ+ 𝒜, where
n is the number of vertices of G. Consider the subgraph D of G whose edges
correspond to α1 , . . . , αn−2 and its vertex set is the union of the vertices in those
edges. Note that D cannot be connected. Indeed there is no edge of D connecting
a vertex in NG (A1 ) with a vertex in NG (A2 ) because all the vectors α1 , . . . , αn−2
satisfy the equation ∑ ∑
xi = xi .
vi ∈A vi ∈NG (A)
It follows that D is a spanning subgraph of G with two connected components D1
and D2 (which are trees) such that V(Di ) = Ai ∪ NG (Ai ), i = 1, 2. Therefore HAi is a
proper support hyperplane defining a facet Fi = HAi ∩ ℝ+ 𝒜, that is A1 , A2 are in ℱ .
Since HA−1 ∩ HA−2 is contained in HA− the proof is complete. □
PROPOSITION 6.2.3. If A2 ⊊ V2 and F = HA2 ∩ ℝ+ 𝒜 is a facet of the edge cone
of G, then
{ −
− ′ HA1 ∩ aff(𝒜′ ) where A1 = V1 ∖ N(A2 ) ∕= ∅, or
HA2 ∩ aff(𝒜 ) =
He+i ∩ aff(𝒜′ ) for some vi with G ∖ {vi } connected,
where 𝒜′ = 𝒜 ∪ {0}.
PROOF. Let us assume G has p vertices v1 , . . . , v p and V1 is the set of the
first m vertices of G. Set 𝒜 = {α1 , . . . , αq }. There are α1 , . . . , α p−2 linearly
independent vectors in the hyperplane HA2 . Consider the subgraph D of G whose
edges correspond to α1 , . . . , α p−2 and its vertex set is the union of the vertices in
264 6. EDGE CONES OF GRAPHS

those edges. As G is connected, either D is a tree with p − 1 vertices or D is a

spanning subgraph of G with two connected components.
If D is a tree, write V(D) = V(G) ∖ {vi } for some i. Note
⟨α j , αA2 ⟩ = −⟨α j , ei ⟩ ( j = 1, . . . , q),
where ∑ ∑
α A2 = ei − ei .
vi ∈A2 vi ∈N(A2 )
Indeed if the “edge” α j has vertices in V(D), then both sides of the equality are zero,
otherwise write α j = ei + e` . Observe vi ∈ / A2 and v` ∈ N(A2 ) because as HA2 is
a facet it, cannot contain α j , thus both sides of the equality are equal to −1. As a
consequence since aff(𝒜′ ) = ℝ(α1 , . . . , α p−2 , α j ) for some α j = ei + e` we obtain
⟨α, αA2 ⟩ = −⟨α, ei ⟩ (∀ α ∈ aff(𝒜′ )).
Therefore
HA−2 ∩ aff(𝒜′ ) = He+i ∩ aff(𝒜′ ),
as required.
We may now assume D is not a tree. We claim A1 = V1 ∖ N(A2 ) ∕= ∅. If
V1 = N(A2 ). Take vi ∈ V2 ∖ A2 and {vi , v j } and edge of D containing vi . Hence
since v j ∈ N(A2 ) we get ⟨ei + e j , αA2 ⟩ = −1, a contradiction because ei + e j is in
HA2 . Thus A1 ∕= ∅. Since all the vectors in aff(𝒜′ ) satisfy the linear equation
m
∑ p
∑
xi = xi ,
i=1 i=m+1
we obtain
⎧ ⎫
⎨ ∑ ∑ ⎬
HA−2 ∩ aff(𝒜′ ) = x ∈ aff(𝒜′ ) xi ≤ xi .
⎩ ⎭
vi ∈V1 ∖N(A2 ) vi ∈V2 ∖A2

Hence we need only show V2 ∖ A2 = N(A1 ). The inclusion N(A1 ) ⊂ V2 ∖ A2 holds

in general. For the reverse containment take vi ∈ V2 ∖ A2 . There is v j such that
{vi , v j } is an edge of D. If v j ∈ N(A2 ), then ⟨ei + e j , αA2 ⟩ = −1, a contradiction
because ei + e j ∈ HA2 . Hence v j ∈ / N(A2 ) and vi ∈ N(A1 ). □
From the proof of Proposition 6.2.3 we get:
LEMMA 6.2.4. Let F = HA ∩ ℝ+ 𝒜 be a facet of ℝ+ 𝒜 with A ⊊ V1 . Then
(a) If N(A) = V2 , then A = V1 ∖ {vi } for some vi ∈ V1 and F = Hei ∩ ℝ+ 𝒜.
(b) If N(A) ⊊ V2 , then F = HV2 ∖N(A) ∩ ℝ+ 𝒜 and N(V2 ∖ N(A)) = V1 ∖ A.
PROPOSITION 6.2.5. Let A ⊊ V1 . Then F = HA ∩ ℝ+ 𝒜 is a facet of ℝ+ 𝒜 if and
only if
 6. EDGE CONES OF GRAPHS 265

(a) ⟨A ∪ N(A)⟩ is connected with vertex set V(G) ∖ {v}, v ∈ V1 , or

(b) ⟨A ∪ N(A)⟩ and ⟨(V2 ∖ N(A)) ∪ (V1 ∖ A)⟩ are connected and their union is
a spanning subgraph of G.
Moreover any facet has the form HA ∩ ℝ+ 𝒜 for some A ⊊ Vi , i ∈ {1, 2}.
PROOF. The first statement follows readily from Lemma 6.2.4 and using part of
the proof of Theorem 6.1.10. The last statement follows combining Theorem 6.1.10
with Proposition 6.2.2. □
REMARK 6.2.6. In Proposition 6.2.5 the case (a) is included in case (b). To see
this fact, take N(A) = V2 and note that ⟨(V2 ∖ N(A)) ∪ (V1 ∖ A)⟩ must consist of a
point. The condition in case (a) is equivalent to require that G ∖ {v} connected and
in this case F = Hei ∩ ℝ+ 𝒜, where v = vi corresponds to the unit vector ei .
LEMMA 6.2.7. Let F be a facet of ℝ+ 𝒜. If F = HA ∩ ℝ+ 𝒜 = HB ∩ ℝ+ 𝒜 with
A ⊊ V1 and B ⊊ V1 , then A = B.
PROOF. Set V1 = {v1 , . . . , vm } and V2 = {vm+1 , . . . , vm+n }. Recall that aff(ℝ+ 𝒜)
is defined by the equality x1 + ⋅ ⋅ ⋅ + xm = xm+1 + ⋅ ⋅ ⋅ + xm+n .
Case (I): N(A) = V2 . Then by Lemma 6.2.4 (after permutation of vertices)
A = {v1 , . . . , vm−1 }. Hence any x ∈ F satisfies
∑ ∑
xi − xi = − xm
vi ∈A vi ∈N(A)

and thus F = Hem ∩ ℝ+ 𝒜. If vm ∈ B, then {vm , v j } ∈ E(G) for some v j in N(B),

thus em + e j ∈ HB and consequently em + e j ∈ Hem , a contradiction. Hence vm ∈ / B,
that is, B ⊂ A. If N(B) = V2 , then by Lemma 6.2.4 A = B. Assume V2 ∖ N(B) ∕= ∅,
to complete the proof for this case we will show that this assumption leads to a
contradiction. First note that vm is not adjacent to any v j ∈ V2 ∖ N(B). Indeed if
{vm , v j } ∈ E(G), then em + e j ∈ HB . Thus em + e j ∈ Hem , a contradiction. Therefore
by the connectivity of G at least one vertex vi ∈ V1 ∖ B must be adjacent to both
a vertex v j ∈ V2 ∖ N(B) and a vertex vk ∈ N(B), which is impossible because
ei + ek ∈ Hem and ei + ek ∈ / HB .
Case (II): N(A) ⊊ V2 and N(B) ⊊ V2 . We begin by considering the subcase
A ∩ B ∕= ∅. Take v0 ∈ A ∩ B and v0 ∕= v ∈ B. By Proposition 6.2.5 the subgraph
⟨ B ∪ N(B)⟩ is connected, hence there is a path of even length
𝒫 = {v0 , v1 , v2 , . . . , v2r−1 , v2r = v}
such that v2i ∈ B for all i. Note that v2 ∈ A. If v2 ∈/ A, then e1 + e2 ∈ HB and
e1 + e2 ∈ / HA , a contradiction. By induction we get v2i ∈ A ∩ B for all i. Hence
v ∈ A. This proves B ⊂ A, a similar argument proves A = B.
Now suppose A ∩ B = ∅. We claim N(A) ∩ N(B) = ∅, otherwise if {v j , vk } is
an edge with v j ∈ B and vk ∈ N(A) ∩ N(B), then e j + ek ∈/ HA because v j ∈/ A and
e j + ek ∈ HB , a contradiction.
266 6. EDGE CONES OF GRAPHS

We may now assume A ∩ B = N(A) ∩ N(B) = ∅. Observe A ∪ B ∕= V1 because

if V1 = A ∪ B, then G would be disconnected with components ⟨A ∪ N(A)⟩ and
⟨ B ∪ N(B)⟩. Take v j ∈ V1 ∖ (A ∪ B) such that v j is adjacent to some vk in N(A) ∪ N(B).
This choice is possible because G is connected. Say vk ∈ N(A). Note vk ∈ / N(B).
Then e j + ek ∈ HB and e j + ek ∈/ HA , a contradiction. □
Putting together the previous results we get the following canonical way of
representing the edge cone. The uniqueness follows from Lemmas 6.2.4 and 6.2.7.
THEOREM 6.2.8. There is a unique irreducible representation
ℝ+ 𝒜 = aff(ℝ+ 𝒜) ∩ (∩ri=1 HA−i ) ∩ (∩i∈ℐ He+i )
such that Ai ⊊ V1 for all i and ℐ = {i∣vi ∈ V2 is not cut vertex }.
COROLLARY 6.2.9. There is an irreducible representation
ℝ+ 𝒜 = aff(ℝ+ 𝒜) ∩ (∩ri=1 HA−i ) ∩ (∩i=1
s
HB−i )
such that Ai ⊊ V1 for all i and Bi ⊊ V2 for all i.
LEMMA 6.2.10. ℤn ∩ ℝ+∑ 𝒜 = ℕ𝒜. In particular if (β1 , . . . , βn ) is an integral
vector in the edge cone, then ni=1 βi is an even integer.
PROOF. Let α ∈ ℤn ∩ ℝ+ 𝒜. By Carathéodory’s Theorem (see Theorem 3.1.37)
and after an appropriate permutation of the αi ’s we can write
α = η1 α1 + ⋅ ⋅ ⋅ + ηr αr (ηi ≥ 0),
where r is the rank of MG and α1 , . . . , αr are linearly independent vectors in 𝒜. On
the other hand, since the submatrix M ′ = (α1 ⋅ ⋅ ⋅ αr ) is totally unimodular because
G is bipartite, by Theorem 3.5.2 the system of equations M ′ x = α has an integral
solution. Hence α is a linear combination of α1 , . . . , αr with coefficients in ℤ. It
follows that ηi ∈ ℕ for all i, that is, α ∈ ℕ𝒜. The other containment is clear. □
Next we recover the marriage theorem for bipartite graphs, see [14, 96]. For a
generalized version of the marriage theorem see Proposition 7.6.8.
THEOREM 6.2.11 (Marriage Theorem). If G is a bipartite graph, then G has a
perfect matching if and only if ∣A∣ ≤ ∣N(A)∣ for every independent set of vertices A
of G.
PROOF. Note that G has a perfect matching if and only if the vector β =
(1, 1, . . . , 1) is in ℕ𝒜. By Lemma 6.2.10 β is in ℕ𝒜 if and only if β ∈ ℝ+ 𝒜. Thus
the result follows from Corollary 6.1.13. □
COROLLARY 6.2.12. Let G = 𝒦m,n be the complete bipartite graph with m ≤ n.
If V1 = {v1 , . . . , vm } and V2 = V ∖ V1 is the bipartition of G, then a vector z ∈ ℝm+n
is in ℝ+ 𝒜 if and only if z = (x1 , . . . , xm , y1 , . . . , yn ) satisfies
x1 + ⋅ ⋅ ⋅ + xm = y1 + ⋅ ⋅ ⋅ + yn ,
 6. EDGE CONES OF GRAPHS 267

xi ≥ 0 for i = 1, . . . , m, and yi ≥ 0 for i = 1, . . . , n. In addition if m ≥ 2, the

inequalities define all the facets of ℝ+ 𝒜.
EXAMPLE 6.2.13. If G = 𝒦1,3 is the star with vertices {v, v1 , v2 , v3 } and center
x, then the edge cone of G has three facets defined by xi ≥ 0, i = 1, 2, 3. Notice that
x = 0, define a proper face of dimension 1.

Exercises
6.2.14. Let G be a graph with n vertices. If the incidence matrix of G has rank
n. Prove: (a) if A is an independent set of vertices of G, then the set F = ℝ+ 𝒜 ∩ HA
is a proper face of the edge cone, and (b) the set F = Hei ∩ ℝ+ 𝒜 is a proper face of
the edge cone.

6.3. Minimal covers and Rees cones

Let us fix some notation. In this section G will denote a connected bipartite graph
with vertex set V = V(G) = {v1 , . . . , vn }, edge set E(G), and height of I(G) greater
than or equal to 2.
The polyhedral cone of ℝn+1 generated by the set of vectors
𝒜′ = {ei + e j + en+1 ∣ vi is adjacent to v j } ∪ {ei ∣ 1 ≤ i ≤ n} ⊂ ℝn+1
is called the Rees cone of G and it will be denoted by ℝ+ 𝒜′ . Note that the Rees
cone has dimension n + 1.
Recall that the cone C(G) of G is the graph obtained by adding a new vertex
vn+1 to G and joining every vertex of G to vn+1 . If A is an independent set of C(G)
we define ∑ ∑
αA = ei − ei ,
vi ∈A vi ∈N(A)
where N(A) is the neighbor set of A in C(G), consisting of all vertices of C(G) that
are adjacent to some vertex of A.
LEMMA 6.3.1. Let ℝ+ ℬ be the polyhedral cone in ℝn+1 generated by
ℬ = {ei + e j ∣ {vi , v j } ∈ E(C(G))}.
Then F is a facet of ℝ+ ℬ if and only if
(i) F = ℝ+ ℬ ∩ Hei , for some 1 ≤ i ≤ n, or
(ii) F = ℝ+ ℬ ∩ HαA , where A is a maximal independent set of C(G).
PROOF. ⇒) Applying Theorem 6.1.7 to the graph C(G) it follows that we can
write F as in (i) or we can write F = ℝ+ ℬ ∩ HαA for some independent set A of C(G)
such that the induced subgraph ⟨V ∪{vn+1 }∖ (A ∪ N(A))⟩ has non bipartite connected
268 6. EDGE CONES OF GRAPHS

components. Since this induced subgraph is bipartite one has V ∪{vn+1 } = A ∪ N(A),
that is, A is a maximal independent set of C(G).
⇐) If F is as in (i), note G ∖ {vi } is connected and non bipartite. Hence F is a
facet. Assume that F is as in (ii). First note V ∪ {vn+1 } = A ∪ N(A) because A is a
maximal independent set of C(G). Consider the subgraph L1 of C(G) with vertex
set A ∪ N(A) and edge set
E(L1 ) = {z ∈ E(C(G))∣ z ∩ A ∕= ∅}.
One can rapidly verify (by considering a bipartition of G and showing that L1 has
only even cycles) that L1 is connected and bipartite. Therefore F is a facet by
Theorem 6.1.7. □
THEOREM 6.3.2 ([83]). F is a facet of the Rees cone ℝ+ 𝒜′ if and only if
(a) F = ℝ+ 𝒜′ ∩ Hei for some 1 ≤ i ≤ n + 1, or
(b) F = ℝ+ 𝒜′ ∩ { x ∈ ℝn+1 ∣ − xn+1 + vi ∈C xi = 0} for some minimal vertex
∑
cover C of G.
PROOF. ⇒) Since the Rees cone is of dimension n + 1, there is a unique a = (ai )
in ℤn+1 with gcd{ai ∣ ai ∕= 0} = 1 such that F = ℝ+ 𝒜′ ∩ Ha and ℝ+ 𝒜′ ⊂ Ha+ . Hence
the entries of a must satisfy ai ≥ 0 for 1 ≤ i ≤ n. Consider the vector
b = (bi ) = (2a1 + an+1 , . . . , 2an + an+1 , −an+1 ).
Using the equalities
2⟨ei + e j + en+1 , a⟩ = ⟨ei + e j , b⟩ if {vi , v j } ∈ E(G),
2⟨ei , a⟩ = ⟨ei + en+1 , b⟩ if 1 ≤ i ≤ n,
we obtain that F ′ = ℝ+ ℬ ∩ Hb is a facet of ℝ+ ℬ with ℝ+ ℬ ⊂ Hb+ . Thus from
Lemma 6.3.1 we can write b in one of the following three forms:
λe
⎧
1 ≤ i ≤ n,
⎨ i

λ(1, . . . , 1, −1),

b=

 −λαA for some maximal
independent set A of G,
⎩

for some integer λ > 0. In the first and second case we get a = ei with 1 ≤ i ≤ n
and a = en+1 respectively. Now consider the case b = −λαA with A ⊂ V a maximal
independent set of G. Note A ∪ N(A) = V ∪ {vn+1 } and vn+1 ∈ / A. Hence the entries
of b satisfy
⎨ −λ if vi ∈ V ∖ A,
⎧

bi = λ if vi ∈ A,
−λ if i = n + 1.
⎩

Thus ai = 0 if vi ∈ A. If vi ∈ V ∖ A, then ai = −an+1 . It follows that an+1 = −1.

Therefore setting C = V ∖ A we fall into case (b).
 6. EDGE CONES OF GRAPHS 269

⇐) It follows using the same type of arguments as above. □

As a consequence we get the irreducible representation of the Rees cone:
COROLLARY 6.3.3 ([83]). ℝ+ 𝒜′ is the intersection of the closed halfspaces given
by the linear inequalities
∑ xi ≥ 0 i = 1, . . . , n + 1,
− xn+1 + vi ∈C i ≥ 0
x C is a minimal vertex cover of G,
and none of those halfspaces can be omitted from the intersection.
REMARK 6.3.4. A consequence of Corollary 6.3.3 is that we can use this result
to compute the minimal vertex covers of G using linear programming.
Canonical modules and a-invariants. Let I = I(G) be the edge ideal of G. By
Corollary 5.6.1 the Rees algebra R[It] is a normal domain and a standard graded K-
algebra. Then by the Danilov-Stanley formula (see Theorem 4.3.27) the canonical
module of R[It] is the ideal of R[It]:
ωR[It] = ({ x1a1 ⋅ ⋅ ⋅ xnan tan+1 ∣ a = (ai ) ∈ (ℝ+ 𝒜′ )o ∩ ℤn+1 }),
where (ℝ+ 𝒜′ )o is the topological interior of the Rees cone.
For use below β0 (resp. α0 ) will denote the vertex independence (covering)
number of G. In algebraic terms β0 is the Krull dimension of the edge ring R/I(G)
and α0 is the height of I(G), see Chapter 2.
THEOREM 6.3.5. If a(R[It]) is the a-invariant of R[It] with respect to the grading
induced by deg(xi ) = 1 and deg(t) = −1, then
a(R[It]) = −(β0 + 1).
PROOF. By Proposition 4.3.29, the a-invariant of R[It] can be expressed as
a(R[It]) = −min{ i ∣ (ωR[It] )i ∕= 0}
Let a = (ai ) be a vector in (ℝ+ 𝒜′ )o ∩ ℤn+1 . By Corollary 6.3.3 a satisfies ai ≥ 1 for
1 ≤ i ≤ n + 1 and for any minimal vertex cover C of G we have
∑
−an+1 + vi ∈C ai ≥ 1.
Let C be a vertex cover of G with α0 elements and let A = V ∖ C. Note β0 = ∣A∣.
Hence if m = x1a1 ⋅ ⋅ ⋅ xnan tan+1 , then
∑ ∑
deg(m) = ai + ai − an+1 ≥ β0 + 1.
vi ∈A vi ∈C
This proves a(R[It]) ≤ −(β0 + 1). On the other hand using Corollary 6.3.3 and the
assumption α0 ≥ 2 we get that the monomial m = x1 ⋅ ⋅ ⋅ xn tα0 −1 is in ωR[It] and has
degree β0 + 1. Thus a(R[It]) ≥ −(β0 + 1). □
COROLLARY 6.3.6. type(R[It]) ≥ α0 − 1.
270 6. EDGE CONES OF GRAPHS

PROOF. By Corollary 6.3.3 the monomials x1 ⋅ ⋅ ⋅ xn ti , with i = 1, . . . , α0 − 1, are

distinct minimal generators of ωR[It] . □

Exercises
6.3.7. Let G be a connected bipartite graph and let Q be the polyhedron defined
as the set of x = (xi ) ∈ ℝn+1 satisfying

∑i = 1, . . . , n + 1, and
xi ≥ 1 for
− xn+1 + vi ∈C xi ≥ 1 for any minimal vertex cover C of G.
If G is unmixed and α0 ≥ 3, then a = (1, . . . , 1) and b = (1, . . . , 1, α0 − 1) are
vertices of Q.

6.4. Rees cones of non bipartite graphs

In this section G will denote a connected non bipartite graph with vertex set V =
V(G) = { x1 , . . . , xn }, edge set E(G), and height of I(G) greater than or equal to 2.
Recall that Rees cone of G is the rational polyhedral cone in ℝn+1 , denoted by
ℝ+ 𝒜′ , consisting of all the non-negative linear combinations of the set of vectors
𝒜′ := {e1 , . . . , en , (v1 , 1), . . . , (vq , 1)} ⊂ ℝn+1 ,
where v1 , . . . , vq is the set of all vectors ei + e j in ℝn such that { xi , x j } is an edge of
G. The Rees cone of G has a unique irreducible representation
(6.60) ℝ+ 𝒜′ = He+1 ∩ He+2 ∩ ⋅ ⋅ ⋅ ∩ He+n+1 ∩ H`+1 ∩ H`+2 ∩ ⋅ ⋅ ⋅ ∩ H`+r
such that 0 ∕= `i ∈ ℤn+1 and the non-zero entries of `i are relatively prime for
i = 1, . . . , r. It is easy to see that the first n entries of each `i are non-negative
and that the last entry of each `i is negative, the second assertion follows from the
irreducibility of Eq. (6.60).
Algebras of covers. Let R = k[x1 , . . . , xn ] be a polynomial ring over a field k
and let I = I(G) = (xv1 , . . . , xvq ) be the edge ideal of the graph G. We denote the
ideal of vertex covers of G by J and the symbolic Rees algebra of J by R s (J).
Let a = (a1 , . . . , an ) ∕= 0 be a vector in ℕn and let b ∈ ℕ. We say that a is a
b-cover of G or that a is a cover of G of degree b if ⟨vi , a⟩ ≥ b for i = 1, . . . , q. The
algebra of covers of G is the k-subalgebra of k[t] generated by all monomials xa tb
such that a is a b-cover of G. It is not hard to see that the algebra of covers of G is
equal to R s (J). We say that a b-cover a of G is reducible if there exists an i-cover c
and a j-cover d of G such that a = c + d and b = i + j. If a is not reducible, we call a
irreducible. The irreducible covers of G of degree 0 and 1 are the canonical vectors
and the characteristic vectors of the minimal vertex covers of G respectively.
 6. EDGE CONES OF GRAPHS 271

The minimal generators of R s (J), as a k-algebra, correspond to the irreducible

covers of G. Since R s (J) is finitely generated as a k-algebra (see Theorem 4.6.9) we
get that the number of irreducible covers of G is finite. In general each `i occuring
in Eq. (6.60) determines a minimal generator of R s (J). Indeed if `i = (a, −d), then
a is an irreducible d-cover of G. It would be interesting to characterize when the
irreducible representation of the Rees cone of G determine the irreducible covers.
Cover algebras associated to graphs and to weighted simplicial complexes have
been studied in [79, 103], see also Section 4.7.
LEMMA 6.4.1. If 0 ∕= a ∈ ℕn is an irreducible k-cover of G, then 0 ≤ k ≤ 2 and
0 ≤ ai ≤ 2 for i = 1, . . . , n.
PROOF. If k = 0 or k = 1, then a = ei for some i or a = ei1 + ⋅ ⋅ ⋅ + eir for some
minimal vertex cover { xi1 , . . . , xir } of G. Assume k ≥ 2.
Case (I): ai ≥ 1 for all i. Clearly 1 = (1, . . . , 1) is a 2-cover. If a − 1 ∕= 0, then
a − 1 is a k − 2 cover and a = 1 + (a − 1), a contradiction. Hence a = 1 and k = 2.
Case (II): ai = 0 for some i. We may assume ai = 0 for 1 ≤ i ≤ r and ai ≥ 1
for i > r. Notice that the set A = { x1 , . . . , xr } is linearly independent and consider
its neighbor set NG (A). We may assume that NG (A) = { xr+1 , . . . , x s }. Clearly
ai ≥ k ≥ 2 for i = r + 1, . . . , s. Write
a = (0, . . . , 0, ar+1 − 2, . . . , a s − 2, a s+1 − 1, . . . , an − 1)+
(0, . . . , 0, 2, . . . , 2, 1, . . . , 1) = b + c.
| {z } | {z } | {z }
r s−r n−s
Clearly c is a 2-cover. If b ∕= 0, it is not hard to see that b is a (k − 2)-cover and
a = b + c, a contradiction. Hence b = 0. Therefore ai = 2 for r < i ≤ s, ai = 1 for
i > s, and k = 2. □
COROLLARY 6.4.2 ([103]). R s (J) is generated as a k-algebra by elements of
t-degree at most two.
COROLLARY 6.4.3. If `i = (`i1 , . . . , `in , −`i(n+1) ), then 0 ≤ `i j ≤ 2 for j =
1, . . . , n and 1 ≤ `i(n+1) ≤ 2.
PROOF. It suffices to observe that (`i1 , . . . , `in ) is an irreducible cover of G of
degree `i(n+1) and to apply Lemma 6.4.1. □

Exercises
6.4.4. Consider the graph G whose edge ideal is
I = (x1 x2 , x2 x3 , x1 x3 , x3 x4 , x4 x5 , x3 x5 , x5 x6 , x6 x7 , x5 x7 , x7 x8 , x8 x9 , x7 x9 ).
Prove that there exists an irreducible 2-cover a of G such that the supporting
hyperplane H(a,−2) does not define a facet of the Rees cone of G.
272 6. EDGE CONES OF GRAPHS

6.4.5. Let G be a graph and let J be the ideal of vertex covers of G. ¿Can you
give an example where the irreducible representation of the Rees cone of J does not
determine R s (I(G))?

6.5. Irreducible covers of graphs

Let G be a graph with vertex set V = { x1 , . . . , xn }. Here we will give a graph
theoretical description of the irreducible b-covers of G.
LEMMA 6.5.1. a = (1, . . . , 1) is an irreducible 2-cover of G if and only if G is
non bipartite and has no isolated vertices.
PROOF. ⇒) If xi is an isolated vertex, then a − ei is a 2-cover and a = ei +(a − ei ),
a contradiction. If G is bipartite, then G has a bipartition that yields a decomposition
of a as a sum of two 1-covers, a contradiction.
⇐) Clearly a cannot be the sum of a 0-cover and a 2-cover. Assume that
a = b + c, where b, c are 1-covers. Let Cr be an odd cycle of length r of G. Notice
that any vertex cover of Cr must contain a pair of adjacent vertices, otherwise Cr
would have a perfect matching, a contradiction because r is odd. Hence the vertex
covers of G corresponding to b and c must contain a pair of adjacent vertices, a
contradiction because b and c are complementary vectors and the complement of a
vertex cover is an independent set. □
PROPOSITION 6.5.2. Let 0 ∕= a = (a1 , . . . , an ) ∈ ℕn . Assume that G has no
isolated vertices.
(i) If G is bipartite, then a is an∑ irreducible b-cover of G if and only if a = ei
for some 1 ≤ i ≤ n or a = xi ∈C ei for some minimal vertex cover C of G.
(ii) If G is non-bipartite, then a is an irreducible b-cover if and only if a has
one of the following forms:
(a) (0-covers) a = ei for some 1 ≤ i ≤ n,
∑
(b) (1-covers) a = xi ∈C ei for some minimal vertex cover C of G,
(c) (2-covers) a = (1, . . . , 1),
(d) (2-covers) up to permutation of vertices
a = (0, . . . , 0, 2, . . . , 2, 1, . . . , 1)
| {z } | {z }
∣A∣ ∣NG (A)∣

for some independent set of vertices A ∕= ∅ of G such that

(d1 ) NG (A) is not a vertex cover of G and V ∕= A ∪ NG (A),
(d2 ) the induced subgraph ⟨V ∖ (A ∪ NG (A))⟩ has no isolated vertices
and is not bipartite.
 6. EDGE CONES OF GRAPHS 273

PROOF. (i) ⇒) By Corollary 4.7.6 and Theorem 4.6.25, we have R s (J) = R[Jt],
where J is the ideal of vertex covers of G. Hence it suffices to observe that xa tb is a
minimal generator of R s (J). The proof of the converse is straightforward.
(ii) ⇒) By Lemma 6.4.1 we have 0 ≤ b ≤ 2 and 0 ≤ ai ≤ 2 for all i. If b = 0
or b = 1, then clearly a has the form indicated in (a) or (b) respectively.
Assume b = 2. If ai ≥ 1 for all i, the ai = 1 for all i, otherwise if ai = 2 for
some i, then a − ei is a 2-cover and a = ei + (a − ei ), a contradiction. Hence a = 1.
Thus we may assume that a has the form
a = (0, . . . , 0, 2, . . . , 2, 1, . . . , 1).
We set A = { xi ∣ ai = 0} ∕= ∅, B = { xi ∣ ai = 2}, and C = V ∖ (A ∪ B). Observe that A
is an independent set because a is a 2-cover and B = NG (A) because a is irreducible.
Hence it is seen that conditions (d1 ) and (d2 ) are satisfied. Using Lemma 6.5.1, the
proof of the converse is direct. □

Exercises
6.5.3. Let G be a bipartite graph and let J = Ic (G) be the ideal of covers of G.
Prove that J is normal.
6.5.4. Give an example of a graph G such that the ideal of covers Ic (G) is not
normal.
Chapter 7
Edge Subrings and Combinatorial Optimization

Let G be a connected bipartite graph on the vertex set V = {v1 , . . . , v p } and let
R = k[x1 , . . . , x p ] = ⊕∞
i=0 Ri be a polynomial ring over a field k with the standard
grading induced by deg(xi ) = 1. Recall that the edge subring of G is the k-subalgebra
k[G] := k[{ xi x j ∣ vi is adjacent to v j }] ⊂ R,
graded with the normalized grading k[G]i = k[G] ∩ R2i . By Theorem 4.3.28 and
Corollary 5.6.1, k[G] is a normal Cohen-Macaulay standard k-algebra.
The main purpose of this chapter is to study the canonical module of k[G] and its
a-invariant using combinatorial optimization techniques. The canonical module is a
fundamental object in commutative algebra that contains homological information,
see [23, pp. 136–146].
The set of vectors αk = ei + e j ∈ ℝ p such that vi is adjacent to v j will be denoted
by 𝒜G := {α1 , . . . , αq } (or simply by 𝒜 if G is understood). The incidence matrix
of G will be denoted by A, recall that the columns of A are precisely the vectors in
𝒜. By Lemma 6.2.10 we have
(7.61) ℕ𝒜 = ℤ p ∩ ℝ+ 𝒜.
Thus according to the Danilov-Stanley formula, the canonical module of the edge
subring k[G] is the ideal given by
ωk[G] = ({ xa ∣ a ∈ ℕ𝒜 ∩ ri(ℝ+ 𝒜)})
(7.61)
(7.62) = ({ xa ∣ a ∈ ℤ p ∩ ri(ℝ+ 𝒜)}) ⊂ k[G].
From this formula it is apparent that the convex set
conv(ℤ p ∩ ri(ℝ+ 𝒜))
should give some information about ωk[G] . It will turn out that this convex set
is an integral polyhedron whose vertices correspond to minimal generators of the
canonical module. We will show how to compute a generating set for ωk[G] in terms
of the incidence matrix of G and the vertices of a certain blocking polyhedron.
The cone ℝ+ 𝒜 is called the edge cone of G. Some explicit irreducible repre-
sentations of the edge cone were shown in Chapter 6, we will make use of these
275
276 7. EDGE SUBRINGS AND COMBINATORIAL OPTIMIZATION

representations to find good expressions for both the relative interior of the edge
cone and conv(ℤ p ∩ ri(ℝ+ 𝒜)). As a result we obtain combinatorial expressions for
the a-invariant of k[G] and show how it can be computed using linear programming.
At the end we explain how some of the results presented here can be adapted to
hold for a certain family of normal monomial subrings.

7.1. Integrality of the shift polyhedron

In the sequel G will denote a connected bipartite simple graph with p = n+m vertices
and bipartition (V1 , V2 ). We will assume that the vertices in V1 are v1 , . . . , vm , the
vertices in V2 are vm+1 , . . . , vn+m , and 2 ≤ m ≤ n.
Consider the family
ℱ = {A ∪ A′ ∣ ∅ ∕= A ⊊ V1 ; N(A) ⊂ A′ ⊂ V2 } ∪ {A′ ∣ ∅ ∕= A′ ⊂ V2 },
where N(A) is the neighbor set of A. For each X = A ∪ A′ ∈ ℱ we associate the
following vector ∑ ∑
βX = ei − ei ∈ ℝm+n ,
vi ∈A vi ∈A′
note that if A = ∅ the vector βX is a {0, −1}-vector. Let C ′ be the matrix whose rows
are the vectors in {βX }X∈ℱ . The matrix C ′ plays an important role here because
according to Theorem 6.2.8 the edge cone of G can be written as:
ℝ+ 𝒜 = aff(ℝ+ 𝒜) ∩ { x∣ C ′ x ≤ 0}.
DEFINITION 7.1.1. The shift polyhedron of the edge cone of G is defined as the
rational polyhedron:
Q = aff(ℝ+ 𝒜) ∩ { x∣ C ′ x ≤ −1}.
From the finite basis theorem (Theorem 3.1.20) the shift polyhedron can be
written as the sum of a unique cone and a polytope. In our case:
Q = ℝ+ 𝒜 + conv(β1 , . . . , βr ),
where β1 , . . . , βr are the vertices of Q. Recall that Q is an integral polyhedron if
Q = conv(ℤ p ∩ Q). As Q is a pointed polyhedron, it is integral if and only if
β1 , . . . , βr are integral vectors (Corollary 3.1.43).
REMARK 7.1.2. Let G be a graph without isolated vertices and take a vector
β in ri(ℝ+ 𝒜). If F is a proper face of ℝ+ 𝒜, then β ∕∈ F. This follows from
Theorem 3.1.31.
A reason for introducing the shift polyhedron is that its integral points define
the canonical module, that is:
LEMMA 7.1.3. ℤ p ∩ Q = ℤ p ∩ ri(ℝ+ 𝒜).
 7. EDGE SUBRINGS AND COMBINATORIAL OPTIMIZATION 277

PROOF. Let C ′ be the matrix defining the shift polyhedron. Note that any row of
C ′ defines a proper face of the edge cone, except the row with the first m entries equal
to 0 and the last n entries equal to −1. Thus equality follows from Remark 7.1.2
and Theorem 3.1.31. □
The shift polyhedron provides a gadget which allows to use combinatorial
optimization techniques to study the edge subring k[G].
THEOREM 7.1.4. The shift polyhedron Q = aff(ℝ+ 𝒜) ∩ { x∣ C ′ x ≤ −1} is an
integral polyhedron.
PROOF. If G is regarded as the digraph with all its arrows leaving the vertex set
V2 , then it is seen that one has the equality C ′ A = − B′ , where B′ is the one-way
cut-incidence matrix of the family ℱ . Let b be any vector in ℝm+n such that the
following maximum is finite
(7.63) max{⟨ x, b⟩∣ x ∈ Q}.
According to Theorem 3.1.42 it suffices to prove that the maximum in Eq. (7.63) is
attained by an integral vector. As Q ⊂ ℝ+ 𝒜, any vector x ∈ Q can be written as
x = A x̃ for some x̃ ≥ 0, x̃ ∈ ℝq , where q is the number of edges of G. Hence
{⟨b, x⟩∣ x ∈ Q} = {⟨bA, x̃⟩∣ x̃ ≥ 0; B′ x̃ ≥ 1} =⇒

(7.64) max{⟨ x, b⟩∣ x ∈ Q} = max{⟨bA, x̃⟩∣ x̃ ≥ 0; B′ x̃ ≥ 1}.

By Corollary 3.6.7 and Lemma 3.6.8 the polyhedron
Q′ = {x̃ ∈ ℝq ∣ x̃ ≥ 0; B′ x̃ ≥ 1}
is integral. Hence, again by Theorem 3.1.42 the maximum in the right hand side of
Eq. (7.64) is attained by an integral vector x̃0 ∈ Q′ , thus the maximum in Eq. (7.63)
is attained by the integral vector A x̃0 ∈ Q, as required. □
REMARK 7.1.5. The linear transformation x̃ 7→ A x̃ maps Q′ onto Q. Note
that dim(Q′ ) = q and dim(Q) = m + n − 1. Thus one can shorten the proof of
Theorem 7.1.4 using Exercise 7.1.8.
EXAMPLE 7.1.6. Consider the following bipartite graph G and make G a digraph
with edges α1 , . . . , α6 as shown below.
v1 v2 v3
V1 H
s
α2 s s

α4
Y
6HH6 6
α1 HH α6
V2 s α 3 s α5
HHs
v4 v5 v6
278 7. EDGE SUBRINGS AND COMBINATORIAL OPTIMIZATION

The family ℱ consists of the following subsets that occur in the directed cuts:
δ+ ({v1 , v4 , v6 }) = {α3 , α6 }, δ+ ({v5 }) = {α4 , α5 },
δ+ ({v2 , v4 , v5 }) = {α1 , α5 }, δ+ ({v6 }) = {α2 , α6 },
δ ({v3 , v5 , v6 }) = {α2 , α4 },
+ δ ({v4 , v5 }) = {α1 , α3 , α4 , α5 },
+

δ+ ({v1 , v2 , v4 , v5 , v6 }) = {α5 , α6 }, δ+ ({v4 , v6 }) = {α1 , α2 , α3 , α6 },

δ ({v1 , v3 , v4 , v5 , v6 }) = {α3 , α4 },
+ δ+ ({v5 , v6 }) = {α2 , α4 , α5 , α6 },
δ+ ({v2 , v3 , v4 , v5 , v6 }) = {α1 , α2 }, δ+ ({v4 , v5 , v6 }) = {α1 , α2 , α3 , α4 , α5 , α6 }.
δ+ ({v4 }) = {α1 , α3 },
As pointed out in the proof of Theorem 7.1.4 the one-way cut-incidence matrix B′
of ℱ is related to the incidence matrix A of G and to the matrix C ′ by the equality
C ′ A = − B′ , that is, one has the matrix equality:
⎡ ⎤ ⎡ ⎤
1 0 0 −1 0 −1 0 0 1 0 0 1
⎢
⎢ 0 1 0 −1 −1 0 ⎥
⎥
⎢
⎢ 1 0 0 0 1 0 ⎥
⎥
⎢
⎢ 0 0 1 0 −1 −1 ⎥
⎥
⎢
⎢ 0 1 0 1 0 0 ⎥
⎥
⎢ 1 1 0 −1 −1 −1 ⎥⎡ ⎤ ⎢ 0 0 0 0 1 1 ⎥
⎢ ⎥ 1 1 0 0 0 0 ⎢ ⎥
⎢ 1 0 1 −1 −1 −1 ⎥ ⎢ 0 0 1 1 0 0 ⎥
⎢ ⎥⎢ 0 0 1 1 0 0 ⎥ ⎢ ⎥
⎢ 0 1 1 −1 −1 −1 ⎥⎢ ⎥ ⎢ 1 1 0 0 0 0 ⎥
⎢ ⎥⎢ 0 0 0 0 1 1 ⎥ ⎢ ⎥
⎢ 0 0 0 −1 0 0 ⎥⎢ ⎥ = −⎢ 1 0 1 0 0 0 ⎥
⎢ ⎥⎢ 1 0 1 0 0 0 ⎥ ⎢ ⎥
⎢ 0 0 0 0 −1 0 ⎥⎢ ⎥ ⎢ 0 0 0 1 1 0 ⎥
⎢ ⎥⎣ 0 0 0 1 1 0 ⎦ ⎢ ⎥
⎢ 0 0 0 0 0 −1 ⎥ ⎢ 0 1 0 0 0 1 ⎥
⎢ ⎥ 0 1 0 0 0 1 ⎢ ⎥
⎢
⎢ 0 0 0 −1 −1 0 ⎥
⎥
⎢
⎢ 1 0 1 1 1 0 ⎥
⎥
⎢
⎢ 0 0 0 −1 0 −1 ⎥
⎥
⎢
⎢ 1 1 1 0 0 1 ⎥
⎥
⎣ 0 0 0 0 −1 −1 ⎦ ⎣ 0 1 0 1 1 1 ⎦
0 0 0 −1 −1 −1 1 1 1 1 1 1

The integral polyhedron Q′ has two vertices (0, 1, 1, 0, 1, 0), (1, 0, 0, 1, 0, 1) that
map under A onto the vector (1, . . . , 1) of Q.
COROLLARY 7.1.7. If Q is the shift polyhedron of the edge cone of G, then
Q = conv(ℤ p ∩ ri(ℝ+ 𝒜)).
PROOF. Clearly conv(ℤ p ∩ ri(ℝ+ 𝒜)) ⊂ Q because any proper face of the
edge cone lies in its relative boundary, see Remark 7.1.2. To show the reverse
inclusion observe that using Theorem 3.1.31 we rapidly obtain Q ⊂ ri(ℝ+ 𝒜). As a
consequence ℤ p ∩ Q ⊂ ℤ p ∩ ri(ℝ+ 𝒜). Taking convex hulls on both sides gives that
Q is contained in conv(ℤ p ∩ ri(ℝ+ 𝒜)) because Q is integral. □

Exercises
7.1.8. Let T : ℝq → ℝn be a linear transformation and let Q′ be an integral
polyhedron in ℝq . If T (ℤq ) ⊂ ℤn , then T (Q′ ) is an integral polyhedron.
 7. EDGE SUBRINGS AND COMBINATORIAL OPTIMIZATION 279

7.2. The canonical module

In our situation S = k[G] is a standard graded k-algebra and a normal domain. The
a-invariant of S , denoted a(S ), is the degree as a rational function of the Hilbert
series of S . If ωS is the canonical module of S , then by Proposition 4.3.29 we have:
a(S ) = −min{ i ∣ (ωS )i ∕= 0}.
THEOREM 7.2.1. If Q is the shift polyhedron of the edge cone of G, then
{ }
∣ x∣
a(k[G]) = −min x∈Q .
2
PROOF. By Lemma 7.1.3 we have ℤ p ∩ Q = ℤ p ∩ ri(ℝ+ 𝒜). Hence the inequality
“≤” follows from Eq. (7.62) and noticing that we are using the normalized grading
on k[G]. As the minimum above is attained at a vertex β of Q it suffices to observe
that β has integral entries by Theorem 7.1.4. □
It is interesting to observe that the a-invariant of k[G] can be interpreted in
combinatorial terms as the next result shows.
PROPOSITION 7.2.2. If G is the digraph with all its arrows leaving the vertex set
V2 , then the following three numbers are equal
(a) −a(k[G]), minus the a-invariant of k[G].
(b) The minimum cardinality of an edge set that contains at least one edge of
each directed cut.
(c) The maximum number of edge disjoint directed cuts.
PROOF. Let A be the incidence matrix of G and let C ′ be the matrix defining the
shift polyhedron. As C ′ A = − B′ , then by Proposition 3.6.6 and duality one has that
the optimum values in the equality
min{⟨1, x̃⟩∣ x̃ ≥ 0; B′ x̃ ≥ 1} = max{⟨y, 1⟩∣ y ≥ 0; yB′ ≤ 1}
are attained by integral vectors. By looking at B′ as a one-way cut-incidence matrix
it follows that the two numbers in (b) and (c) are equal.
On the other hand note:
⟨1, A x̃⟩/2 = ⟨1, x̃1 α1 + ⋅ ⋅ ⋅ + x̃q αq ⟩/2 = (2 x̃1 + ⋅ ⋅ ⋅ + 2 x̃q )/2 = ⟨1, x̃⟩,
where α1 , . . . , αq are the column vectors of A. Using the equality
min{⟨1, x⟩/2∣ x ∈ aff(ℝ+ 𝒜); C ′ x ≤ −1} = min{⟨1, A x̃⟩/2∣ x̃ ≥ 0; B′ x̃ ≥ 1}
= min{⟨1, x̃⟩∣ x̃ ≥ 0; B′ x̃ ≥ 1},
and Theorem 7.2.1 one derives that the numbers in (a) and (b) are equal. □
It is well known that for arbitrary digraphs the numbers in (b) and (c) are equal,
see for instance [120, Theorem 19.10].
280 7. EDGE SUBRINGS AND COMBINATORIAL OPTIMIZATION

PROPOSITION 7.2.3. If β is a vertex of the shift polyhedron Q, then xβ is a

minimal generator of ωk[G] .
PROOF. By Eq. (7.62), Remark 7.1.2, and Theorem 7.1.4 we get xβ ∈ ωk[G] .
There are c ∈ ℚm+n and b ∈ ℚ such that
{β} = { x∣⟨ x, c⟩ = b} ∩ Q and Q ⊂ { x∣⟨ x, c⟩ ≤ b}.
If α1 , . . . , αq are the columns of the incidence matrix of G, then by definition of Q
one has αi + β ∈ Q for all i. Thus
⟨αi + β, c⟩ = ⟨αi , c⟩ + b ≤ b ⇒ ⟨αi , c⟩ ≤ 0 (i = 1, . . . , q).
Assume there is α ∈ Q and η1 , . . . , ηq in ℕ such that
β = η1 α1 + ⋅ ⋅ ⋅ + ηq αq + α =⇒
b = ⟨β, c⟩ = η1 ⟨α1 , c⟩ + ⋅ ⋅ ⋅ + ηq ⟨αq , c⟩ + ⟨α, c⟩ ≤ ⟨α, c⟩ ≤ b.
Hence ⟨α, c⟩ = b and by (i) we get α = β. Thus xβ is a minimal generator of the
canonical module ωk[G] . □
DEFINITION 7.2.4. If C ′ A = − B′ , the polyhedron
Q′ = {x̃ ∈ ℝq ∣ x̃ ≥ 0; B′ x̃ ≥ 1}
is called the blocking polyhedron.
THEOREM 7.2.5. The blocking polyhedron Q′ is an integral polyhedron.
PROOF. It follows from Corollary 3.6.7 and Lemma 3.6.8. □
The converse of Proposition 7.2.3 is not true in general, see for instance Exam-
ple 7.3.7. However any minimal generator of ωk[G] is in the image under A of the
blocking polyhedron.
PROPOSITION 7.2.6. Let Q be the shift polyhedron of the edge cone of G. If xβ
is a minimal generator of ωk[G] , then there is a vertex α̃ of the blocking polyhedron
Q′ such that Aα̃ = β.
PROOF. Let α1 , . . . , αq be the column vectors of the incidence matrix A and let
C′ be the matrix defining Q. As A is totally unimodular, then by Carathéodory’s
theorem (Theorem 3.1.37) and Heger’s theorem (after permuting the αi ’s) we can
write:
β = η1 α1 + ⋅ ⋅ ⋅ + ηr αr (ηi ∈ ℕ ∖ {0}; r ≤ p − 1 ≤ q),
where p = m + n. We claim that ηi = 1 for all i. Assume ηi > 1. Take any row v of
C ′ . Observe that ⟨v, αk ⟩ is equal to 0 or −1 for any αk . The vector
β′ = η1 α1 + ⋅ ⋅ ⋅ + ηi−1 αi−1 + (ηi − 1)αi + ηi+1 αi+1 + ⋅ ⋅ ⋅ + ηr αr
satisfies ⟨β′ , v⟩ ≤ −1 because ⟨β, v⟩ ≤ −1. Thus β′ ∈ Q, a contradiction because
β′ = β − αi and xβ is minimal. Thus ηi = 1. The vector α̃ = e1 + ⋅ ⋅ ⋅ + er satisfies
Aα̃ = β, and from C ′ A = − B′ we get α̃ ∈ Q′ . Consider the linear program:
 7. EDGE SUBRINGS AND COMBINATORIAL OPTIMIZATION 281

min x̃r+1 + ⋅ ⋅ ⋅ + x̃q

(∗)
subject to B′ x̃ ≥ 1; x̃ ≥ 0.
Note that 0 is the optimum value of this linear program because α̃ is in Q′ . By
Theorem 7.2.5 there is an integral vertex γ̃ of Q′ where the minimum is attained.
Hence γ̃i = 0 for i > r. By Exercise ∑ 7.2.8 the vector γ has {0, 1}-entries. If γ̃k = 0
for some 1 ≤ k ≤ r, then Aγ̃ = i∕=k i αi ∈ Q, where i ∈ {0, 1} for all i ∕= k, a
contradiction to the minimality of xβ . Thus α̃ = γ̃, as required. Observe that the last
part of the argument works even if q = r, which is the case of a tree. □
THEOREM 7.2.7. The canonical module ωk[G] of k[G] is generated by the set
{ xAα̃ ∣ α̃ is a vertex of Q′ },
where Q′ is the blocking polyhedron.
PROOF. It follows noticing that the blocking polyhedron Q′ is integral and using
Proposition 7.2.6. □
The vertices of the shift polyhedron are not enough to determine the canonical
module, see Example 7.3.7.

Exercises
7.2.8. If B′ is a {0, 1}-matrix, prove that any integral vertex of the following
polyhedron is a {0, 1}-vector:
Q′ = {x̃ ∈ ℝq ∣ x̃ ≥ 0; B′ x̃ ≥ 1}.

7.3. Computing the a-invariant

In order to be able to use the results of Section 7.2 in an efficient way we need to
introduce better representations of the shift polyhedron.
Recall that a set of vertices of G is called independent if no two of its vertices
are adjacent. For each independent set of vertices A of G consider the vector
∑ ∑
αA = ei − ei ,
vi ∈A vi ∈N(A)

There exists an irreducible representation of the edge cone as an intersection of

closed halfspaces of the form (see Corollary 6.2.9):
− −
(7.65) ℝ+ 𝒜 = aff(ℝ+ 𝒜) ∩ Hα−A ∩ ⋅ ⋅ ⋅ ∩ Hα−Ar ∩ H−e i
∩ ⋅ ⋅ ⋅ ∩ H−e is
,
1 1

where for each i either Ai ⊊ V1 or Ai ⊊ V2 and none of the half-spaces can be omitted
from the intersection. Let us denote by C the matrix whose rows are the vectors
282 7. EDGE SUBRINGS AND COMBINATORIAL OPTIMIZATION

αA1 , . . . , αAr , −ei1 , . . . , −eis and by C ′ the matrix defining the shift polyhedron, as
defined in Section 7.1.
DEFINITION 7.3.1. The shift polyhedron of the edge cone of G with respect to
C is defined as
aff(ℝ+ 𝒜) ∩ { x∣ Cx ≤ −1}.
LEMMA 7.3.2. If Q is the shift polyhedron of the edge cone of G and x ∈ Q,
then xi ≥ 1 for all i = 1, . . . , m + n.
PROOF. Let (V1 , V2 ) be the bipartition of G. If G ∖ {vi } is connected, then the
hyperplane Hei defines a facet of the edge cone and consequently xi ≥ 1. In case
G = 𝒦1,n is a star the lemma follows readily.
Thus one may assume G ∖ {vi } disconnected, vi ∈ V1 , and G not a star. For
simplicity of notation set i = 1. Let G1 , . . . , Gr be the connected components of the
subgraph G ∖ {v1 }. One may assume that the first component G1 has at least two
vertices. Since
− x1 = x2 + ⋅ ⋅ ⋅ + xm − xm+1 − ⋅ ⋅ ⋅ − xm+n
it suffices to prove that the right hand side of this equality is less or equal than
−1. Let us denote by (V1 (Gi ), V2 (Gi )) the bipartition of Gi , where the first set of
vertices is contained in V1 and the second set in V2 . Consider A = V1 (G1 ) and note
NG (A) = V2 (G1 ). The subgraph ⟨(V1 ∖ A) ∪ (V2 ∖ NG (A))⟩ is connected because all
the components of G ∖ {v1 } are connected to x1 . Therefore the hyperplane
∑ ∑
xi − xi = 0
vi ∈A vi ∈NG (A)

defines a facet of the edge cone and consequently

∑ ∑
xi − xi ≤ −1.
vi ∈A vi ∈NG (A)

If V1 (Gk ) ∕= ∅, then NG (V1 (Gk )) = V2 (Gk ). Observe that every component Gk defines
a valid inequality for the edge cone:
∑ ∑
xi − xi ≤ 0,
vi ∈V1 (Gk ) vi ∈V2 (Gk )

where V1 (Gk ) = ∅ if Gk consists of a single vertex and the sum over an empty set is
defined as zero. Therefore
⎛ ⎞ ⎛ ⎞
∑ ∑ ∑ ∑ ∑
⎝ xi − xi ⎠ + ⎝ xi − xi ⎠ ≤ −1.
vi ∈V1 (G1 ) vi ∈V2 (G1 ) k>1 vi ∈V1 (Gk ) vi ∈V2 (Gk )

As the left hand side of this equality is x2 + ⋅ ⋅ ⋅ + xm − xm+1 − ⋅ ⋅ ⋅ − xm+n , the proof
is complete. □
 7. EDGE SUBRINGS AND COMBINATORIAL OPTIMIZATION 283

PROPOSITION 7.3.3. If Q is the shift polyhedron of the edge cone of G and x ∈ Q,

then ∑ ∑
xi − xi ≤ −1
vi ∈A vi ∈A′
for all A ⊊ V1 (resp. A ⊊ V2 ) such that N(A) ⊂ A′ ⊂ V2 (resp. N(A) ⊂ A′ ⊂ V1 ).
PROOF. It suffices to prove the case A′ = N(A). If N(A) = V2 , then
∑ ∑ ∑
xi − xi = − xi ≤ −1
vi ∈A vi ∈N(A) vi ∈V1 ∖A

because of Lemma 7.3.2.

Assume B = V2 ∖ N(A) ∕= ∅. If the subgraph ⟨A ∪ N(A)⟩ is disconnected, there
is a partition A1 , . . . , Ar of A such that ⟨Ai ∪ N(Ai )⟩ are the connected components
of ⟨A ∪ N(A)⟩ for i = 1, . . . , r. Since
⎛ ⎞
∑ ∑ ∑ r ∑ ∑
xi − xi = ⎝ xi − xi ⎠ ,
vi ∈A vi ∈N(A) j=1 vi ∈A j vi ∈N(A j )

one may assume that ⟨A ∪ N(A)⟩ is connected. The following is not hard to establish:
A ⊂ V1 ∖ N(B) and N(A) = N(V1 ∖ N(B)).
Set A1 = V1 ∖ N(B) and observe N(A) = N(A1 ). Notice that the subgraph ⟨A1 ∪ N(A1 )⟩
is connected because ⟨A ∪ N(A)⟩ is connected. Now in order to shift our attention
to A1 observe that
∑ ∑ ∑ ∑
xi − xi ≤ xi − xi (∗)
vi ∈A vi ∈N(A) vi ∈A1 vi ∈N(A1 )

and V1 ∖ A1 = N(V2 ∖ N(A)) = N(V2 ∖ N(A1 )). Hence if ⟨ B ∪ N(B)⟩ is connected,

then the hyperplane
∑ ∑
xi − xi = 0
vi ∈A1 vi ∈N(A1 )
defines a facet of the edge cone and the required inequality follows. Thus it only
remains to consider the case in which the subgraph ⟨ B ∪ N(B)⟩ is disconnected. Let
B1 , . . . , Br be a partition of B such that ⟨ Bi ∪ N(Bi )⟩ are the connected components
of ⟨ B ∪ N(B)⟩ for i = 1, . . . , r. Set
A2 = A1 ∪ N(B1 ) ∪ ⋅ ⋅ ⋅ ∪ N(Br−1 ).
We claim N(A2 ) = N(A1 ) ∪ B1 ∪ ⋅ ⋅ ⋅ ∪ Br−1 . The left hand side clearly contains
the right hand side. For the reverse containment take v ∈ N(A2 ) and w ∈ A2 such
that {v, w} is an edge of G. If v is not in the right hand side, then v ∈ Br and
w ∈ N(Br ). Thus w must be in A1 because ⟨ Bi , N(Bi )⟩ is a component of ⟨ B, N(B)⟩,
a contradiction.
284 7. EDGE SUBRINGS AND COMBINATORIAL OPTIMIZATION

Next we claim that ⟨A2 ∪ N(A2 )⟩ is connected. To see this observe that for each
i at least one vertex of ⟨ Bi , N(Bi )⟩ is adjacent to a vertex of ⟨A1 , N(A1 )⟩ because G is
connected. Note V2 ∖ N(A2 ) = Br and V1 ∖ A2 = N(Br ). Hence since ⟨A2 ∪ N(A2 )⟩
and ⟨ Br ∪ N(Br )⟩ are connected we derive that the hyperplane
∑ ∑
xi − xi = 0
vi ∈Br vi ∈N(Br )
defines a facet of the edge cone. Thus using
⎛ ⎞ ⎛ ⎞
∑ ∑ ∑ ∑ r−1
∑ ∑ ∑
xi − xi ≤ ⎝ xi − xi ⎠ + ⎝ xi − xi ⎠
vi ∈A1 vi ∈N(A1 ) vi ∈A1 vi ∈N(A1 ) j=1 vi ∈N(B j ) vi ∈B j
∑ ∑
= xi − xi ≤ −1
vi ∈Br vi ∈N(Br )
the proof is complete. □
The next result says that the shift polyhedron with respect to C is just the shift
polyhedron.
THEOREM 7.3.4. The shift polyhedron Q of the edge cone of G is given by
Q = aff(ℝ+ 𝒜) ∩ { x∣ Cx ≤ −1}.
PROOF. It follows from Lemmas 7.3.2, 6.2.7 and Proposition 7.3.3. □
REMARK 7.3.5. If CA = − B and = C′ A − B′ ,
where A is the incidence matrix
of G, then B and B′ define the same blocking polyhedron.
REMARK 7.3.6. The matrices C ′ and B′ are useful for theoretical reasons. Notice
that Theorem 7.3.4 and Remark 7.3.5 simplify the task of finding the generators of
ωk[G] by a reduction of the number of inequalities when using C and B to compute
the vertices of the shift and blocking polyhedra. In conclusion we can effectively
compute a set of generators of the ideal ωk[G] . See Example 7.3.7 for an illustration.
EXAMPLE 7.3.7. Consider the following bipartite simple graph G:
vt1 vt9 vt5 vt11

v4t A

A

A
t

X XXX tv v tH t A
v7 X2 10  H v3
H 


 XXXXXHH
A

  X H A


t


XXH XHXAAt
v8 v6

with bipartition V1 = {v1 , . . . , v6 } and V2 = {v7 , . . . , v11 }. The incidence matrix of

G, denoted by A, is the transpose of the following matrix:
 7. EDGE SUBRINGS AND COMBINATORIAL OPTIMIZATION 285

columns of A corresponding monomial

1 0 0 0 0 0 1 0 0 0 0 x1x7
1 0 0 0 0 0 0 1 0 0 0 x1x8
0 1 0 0 0 0 1 0 0 0 0 x2x7
0 1 0 0 0 0 0 0 1 0 0 x2x9
0 0 1 0 0 0 0 1 0 0 0 x3x8
0 0 1 0 0 0 0 0 0 1 0 x3x10
0 0 1 0 0 0 0 0 0 0 1 x3x11
0 0 0 1 0 0 0 0 0 1 0 x4x10
0 0 0 0 1 0 0 0 0 1 0 x5x10
0 0 0 0 0 1 0 0 0 1 0 x6x10
0 0 0 0 0 1 0 0 0 0 1 x6x11
0 0 0 0 1 0 0 0 1 0 0 x5x9
0 0 0 1 0 0 0 0 0 0 1 x4x11
1 0 0 0 0 0 0 0 1 0 0 x1x9
0 1 0 0 0 0 0 0 0 1 0 x2x10
0 0 0 0 0 1 1 0 0 0 0 x6x7
0 0 0 0 1 0 0 0 0 0 1 x5x11
In Example 7.4.5 we will show “real” input files for PORTA and explain how they
can be converted into output files, see also [43, Example 2.1] and [200, pp. 11–13].
Using PORTA we obtain an irreducible representation of ℝ+ 𝒜, which immediately
yields the following representation of Q by linear inequalities:
-xi<= -1 for i=7,...,11
x1+x2+x3+x4+x5+x6-x7-x8-x9-x10-x11 == 0
x1+x3+x4+x5+ x6-x7-x8-x9-x10-x11 <= -1
x1+x2+x4+x5+ x6-x7-x8-x9-x10-x11 <= -1
x1+x2+x3+x5+ x6-x7-x8-x9-x10-x11 <= -1
x1+x2+x3+x4+ x6-x7-x8-x9-x10-x11 <= -1
x1+x2+x3+x4+x5 -x7-x8-x9-x10-x11 <= -1
x4 -x10-x11 <= -1
x2 -x7 -x9 -x10 <= -1
x4+ x6-x7 -x10-x11 <= -1
x4+x5 -x9-x10-x11 <= -1
x3+x4 -x8 -x10-x11 <= -1
x1+x2 -x7-x8-x9-x10 <= -1
x3+x4 +x6-x7-x8 -x10-x11 <= -1
x3+x4+x5 -x8-x9-x10-x11 <= -1
x2 +x4+x5+x6-x7 -x9-x10-x11 <= -1
x1 -x7-x8-x9 <= -1
x2+x3+x4+x5+x6-x7-x8-x9-x10-x11 <= -1
Using PORTA we get that the vertices of the shift polyhedron are:
1 2 1 1 1 1 1 1 1 1 3 1 1 1 1 1 1 1 1 1 2 1
2 1 1 1 1 1 1 1 1 1 3 1 1 1 1 1 1 1 1 2 1 1
286 7. EDGE SUBRINGS AND COMBINATORIAL OPTIMIZATION

2 1 1 1 1 1 1 1 1 3 1 1 1 1 1 1 1 2 1 1 1 1
1 1 1 1 1 1 1 1 1 1 2
From the equality CA = − B (see Remark 7.3.5), we obtain that the corresponding
blocking polyhedron is defined by the linear inequalities:
xi>=0 for i=1,...,17 x10+x11+x15+x17+x6+x7+x9>=1
x15+x3+x4>=1 x1+x10+x12+x14+x16+x6+x8+x9>=1
x5+x6+x7>=1 x1+x15+x17+x3+x6+x7+x9>=1
x13+x8>=1 x10+x11+x14+x15+x4+x6+x7>=1
x12+x17+x9>=1 x10+x11+x15+x17+x2+x9>=1
x10+x11+x16>=1 x10+x12+x16+x5+x6+x8+x9>=1
x1+x16+x3>=1 x1+x15+x17+x2+x3+x9>=1
x2+x5>=1 x10+x11+x14+x15+x2+x4>=1
x12+x14+x4>=1 x1+x14+x6+x7>=1
x10+x15+x6+x8+x9>=1 x12+x16+x3+x4+x5>=1
x11+x13+x17+x7>=1 x1+x14+x2>=1
Using PORTA once more we get that the blocking polyhedron Q′ has 173 vertices.
The distinct images of these vertices under the incidence matrix A are:
1 2 1 1 1 1 1 1 1 1 3 vertex of the shift polyhedron Q
2 1 1 1 1 1 1 1 1 1 3 vertex of Q
2 1 1 1 1 1 1 1 1 3 1 vertex of Q
1 1 1 1 1 1 1 1 1 1 2 vertex of Q
1 1 1 1 1 1 1 1 1 2 1 vertex of Q
1 1 1 1 1 1 1 1 2 1 1 vertex of Q
1 1 1 1 1 1 2 1 1 1 1 vertex of Q
2 1 1 1 1 1 1 1 1 2 2 not a vertex of Q but correspond to a
minimal generator of the canonical module
1 1 1 1 1 2 1 1 2 2 1 this and the following vectors correspond to
1 1 2 1 1 1 1 1 2 2 1 redundant monomials in the canonical module
.....................
Altogether using Theorem 7.2.7 we get that ωk[G] is minimally generated by eight
monomials corresponding to the first eight vectors above. Thus the Cohen-Macaulay
type of the algebra k[G] is 8, this means that the rank of the last module of syzygies
in the homogeneous resolution of k[G] is 8.
REMARK 7.3.8. NORMALIZ [24, 25] can be used in practice to compute the
a-invariant of k[G] through the computation of the Hilbert series of k[G]. It is also
possible but less efficient to compute this invariant using algebraic systems such
as Macaulay2 [90] or CoCoA [32]. From Theorems 7.2.1 and 7.3.4 we obtain an
effective method to compute the a-invariant of k[G] that only requires a description
of the shift polyhedron by linear inequalities and to solve a linear program. See
Example 7.3.9.
EXAMPLE 7.3.9. Consider the following bipartite graph G:
 7. EDGE SUBRINGS AND COMBINATORIAL OPTIMIZATION 287

v4
s
bipartition:
v7 s @sv8
@
@ V1 = {v1 , v2 , v3 , v4 }
@sv3
V2 = {v5 , v6 , v7 , v8 }
v6
s

v1 s @sv2
@
@
@sv5
In order to estimate the a-invariant of G we set up the next linear program using the
following input file for Mathematica
vars:={x1,x2,x3,x4,x5,x6,x7,x8}

f:=(x1+x2+x3+x4+x5+x6+x7+x8)/2

ieq:={x1+x2+x3+x4-x5-x6-x7-x8 == 0, -x1<= -1, -x2<= -1,-x4<= -1,

-x5<= -1, -x7<= -1,-x8 <= -1,x4-x7-x8 <= -1,
-x3-x4+x7+x8 <= -1,x3+x4-x6-x7-x8 <= -1}

ConstrainedMin[f,ieq,vars]
where the set of inequalities were found using PORTA and they come from an
irreducible representation of the edge cone as in Eq. (7.65). The answer found is
that the optimal value of this linear program is equal to 5 and is attained at the vertex
(1, 1, 2, 1, 1, 2, 1, 1). Hence by Theorems 7.2.1 and 7.3.4 we get that the a-invariant
of k[G] is equal to −5.

Exercises
7.3.10. To compute the vertices of the shift polyhedron Q using PORTA we
need a “valid” point, i.e., a point in Q. Prove that if 𝒜 = {α1 , . . . , αq } is the set of
column vectors of the incidence matrix of a bipartite graph G, then the following
point is valid: α = α1 + ⋅ ⋅ ⋅ + αq

7.3.11. Consider the bipartite graph G:

vr5 vr2
 T
 v6r T
v1r JJr Trv8
T
T
v3 

Tr r

v7 v4

Prove that the canonical module of k[G] is minimally generated by three monomials
and that the shift polyhedron has vertices 1, (2,1,1,1,1,1,1,2) and (1,2,1,1,1,1,2,1).
288 7. EDGE SUBRINGS AND COMBINATORIAL OPTIMIZATION

7.4. A family of normal monomial subrings

Here we explain how the methods introduced so far can be used to study the
canonical module and the a-invariant of certain more general monomial subrings,
other than edge subrings associated to bipartite graphs. For this purpose we extend
the notion of a shift polyhedron.
Let 𝒜 = {α1 , . . . , αq } be a set of distinct points in ℕ p ∖ {0} and let
ℝ+ 𝒜 = aff(ℝ+ 𝒜) ∩ Ha−1 ∩ ⋅ ⋅ ⋅ ∩ Ha−r (ai ∈ ℤ p )
be an irreducible representation (of the polyhedral cone generated by 𝒜) as an
intersection of closed half-spaces, such that the non-zero entries of ai are relatively
prime for all i. The polyhedron
Q = aff(ℝ+ 𝒜) ∩ { x ∈ ℝ p ∣ Cx ≤ −1}
is called the shift polyhedron of ℝ+ 𝒜 relative to C, where C is the integral matrix
with rows a1 , . . . , ar . If ℤ p ∩ ri(ℝ+ 𝒜) ∕= ∅, then Q ∕= ∅.
EXAMPLE 7.4.1. For cones in ℝ2 , the notion of shift polyhedron is consistent
with geometry:

76
p pp
( )
p ppppppp

6 p pp
ppppppppppppppppppppp
 1 −2
 p pppppppppppp pp p C=
5 p ppppppppppppp p

pppppppppppp
−2 3
pppppp
 
p p p

ppppppppppppppp
pppppppppppppppppp p

pppppp
 
4 p p p p p p p pppppppppp p 
p ppppppppppppppppp
ppppppppppppp p 
ppppppppppppppppppppppp pppp 

p p
p pp pppp rppppp pp ppp pp p p p

3 p p pppppppppppppppppppppppp p
 ℝ+ 𝒜 = { x∣ Cx ≤ 0}
p ppppppppppppppppppppp p 
p pppppppppp p p 

2 p pp
ppppppp p p
ppppppppp p
ppp
p p
1  pp p p
p

ppp p
r
 -
0 1 2 3 4 5 6 7 8 9 10
Here the shift polyhedron of ℝ+ 𝒜 with respect to C is Q = { x∣ Cx ≤ −1}
PROPOSITION 7.4.2. If Q is the shift polyhedron of ℝ+ 𝒜 w.r.t C, then
(7.66) ℤ p ∩ Q = ℤ p ∩ ri(ℝ+ 𝒜),
and conv(ℤ p ∩ ri(ℝ+ 𝒜)) is an integral polyhedron if ℤ p ∩ ri(ℝ+ 𝒜) ∕= ∅.
PROOF. To prove the equality it suffices to note that a point α is in ri(ℝ+ 𝒜)
if and only if α ∈ ℝ+ 𝒜 and ⟨α, ai ⟩ < 0 for i = 1, . . . , r, see Theorem 3.1.31.
The second assertion follows by taking convex hulls in Eq. (7.66) and observing
that conv(ℤ p ∩ Q) is the integer hull of Q, which is an integral polyhedron by
Proposition 3.1.44. □
 7. EDGE SUBRINGS AND COMBINATORIAL OPTIMIZATION 289

REMARK 7.4.3. Let R = k[x1 , . . . , x p ] be a polynomial ring over a field k and let
S = k[{ xα1 , . . . , xαq }] ⊂ R
be the monomial subring of R generated by { xα1 , . . . , xαq } over k. Assume that the
following three conditions hold:
(a) there exists 0 ∕= x0 ∈ ℚ p such that ⟨ x0 , αi ⟩ = 1 for all i, i.e., S is a
homogeneous subring (see Chapter 4),
(b) ℕ𝒜 = ℤ p ∩ ℝ+ 𝒜, and
(c) the shift polyhedron Q, relative to C, of the cone ℝ+ 𝒜 is integral.
Conditions (a) and (b) imply that S is a standard graded algebra and a normal
domain whose canonical module is the ideal of S given by
(7.67) ωS = ({ xa ∣ a ∈ ℕ𝒜 ∩ ri(ℝ+ 𝒜)})
(b)
(7.68) = ({ xa ∣ a ∈ ℤ p ∩ ri(ℝ+ 𝒜)})
(7.66)
(7.69) = ({ xa ∣ a ∈ ℤ p ∩ Q}).
Note that S is graded as follows. A monomial xa with a ∈ ℕ𝒜 has degree i if and
only if ⟨a, x0 ⟩ = i. It is not hard to adapt the proof of Theorem 7.2.1 to show that
the a-invariant of the ring S is given by
a(S ) = −min {⟨ x0 , x⟩∣ x ∈ Q} .
From the proof of Proposition 7.2.3 it follows that xβ is a minimal generator of ωS
for any vertex β of Q. Using the proof of Corollary 7.1.7 we get
Q = conv(ℤ p ∩ ri(ℝ+ 𝒜)).
EXAMPLE 7.4.4. Let k[G] = k[x1 x2 , x2 x3 , x1 x3 ] be the edge subring of a triangle
G. If ⎡ ⎤
−1 −1 1
C = ⎣ −1 1 −1 ⎦ ,
1 −1 −1
then ℝ+ 𝒜 = { x∣ Cx ≤ 0} and Q = { x∣ Cx ≤ −1}, where 𝒜 is the set {e1 + e2 , e2 +
e3 , e1 +e3 }. Conditions (a) and (c) of Remark 7.4.3 are satisfied. However condition
(b) is not satisfied since α is in ℤ3 ∩ ℝ+ 𝒜 and is not in ℕ𝒜. Thus in this example
S = k[G] is a normal domain whose canonical module cannot be expressed using
Eq. (7.68), instead one should use the more complicated formula given in Eq. (7.67).
EXAMPLE 7.4.5. Consider the following input file for PORTA that we call
cone.poi:
DIM = 12

CONE_SECTION
290 7. EDGE SUBRINGS AND COMBINATORIAL OPTIMIZATION

1 1 1 0 0 0 0 0 0 0 0 0
0 1 1 1 0 0 0 0 0 0 0 0
0 0 0 1 1 1 0 0 0 0 0 0
0 0 0 0 1 1 1 0 0 0 0 0
0 0 0 0 0 0 1 1 1 0 0 0
0 0 0 0 0 0 0 1 1 1 0 0
0 0 0 0 0 0 0 0 0 1 1 1
1 0 0 0 0 0 0 0 0 0 1 1

END
Let A be the matrix above and let a1 , . . . , a8 be their rows. Consider the monomial
subring S = k[xa1 , . . . , xa8 ]. In your PORTA directory type: xporta -T cone.poi.
Then PORTA creates the following output file which is called cone.poi.ieq:
DIM = 12

VALID
0 0 0 0 0 0 0 0 0 0 0 0

EQUALITIES_SECTION
( 1) +x11-x12 == 0
( 2) +x8-x9 == 0
( 3) +x5-x6 == 0
( 4) +x2-x3 == 0
( 5) +x1 -x3+x4 -x6+x7 -x9+x10-x11 == 0

INEQUALITIES_SECTION
( 1) -x3 <= 0
( 2) -x4 <= 0
( 3) -x6 <= 0
( 4) -x7 <= 0
( 5) -x9 <= 0
( 6) -x10 <= 0
( 7) -x12 <= 0
( 8) -x3+x4-x6 <= 0
( 9) -x4+x6-x7 <= 0
( 10) -x6+x7-x9 <= 0
( 11) -x7+x9-x10 <= 0
( 12) -x9+x10-x12 <= 0
( 13) -x3+x4-x6+x7-x9 <= 0
( 14) -x4+x6-x7+x9-x10 <= 0
( 15) -x6+x7-x9+x10-x12 <= 0
( 16) -x3+x4-x6+x7-x9+x10-x12 <= 0

END
 7. EDGE SUBRINGS AND COMBINATORIAL OPTIMIZATION 291

The linear equalities and inequalities in this file define the cone generated by the ai ’s.
The first block defines the affine space and the second block defines the facets of the
cone. To compute the point representation of the shift polyhedron you may proceed
as follows. In your PORTA directory type: xporta -T shift.ieq, where shift.ieq is the
input file:
DIM = 12

VALID
2 2 2 2 2 2 2 2 2 2 2 2

INEQUALITIES_SECTION
( 1) +x11-x12 == 0
( 2) +x8-x9 == 0
( 3) +x5-x6 == 0
( 4) +x2-x3 == 0
( 5) +x1 -x3+x4 -x6+x7 -x9+x10-x11 == 0

( 1) -x3 <=-1
( 2) -x4 <=-1
( 3) -x6 <=-1
( 4) -x7 <=-1
( 5) -x9 <=-1
( 6) -x10 <=-1
( 7) -x12 <=-1
( 8) -x3+x4-x6 <=-1
( 9) -x4+x6-x7 <=-1
( 10) -x6+x7-x9 <=-1
( 11) -x7+x9-x10 <=-1
( 12) -x9+x10-x12 <=-1
( 13) -x3+x4-x6+x7-x9 <=-1
( 14) -x4+x6-x7+x9-x10 <=-1
( 15) -x6+x7-x9+x10-x12 <=-1
( 16) -x3+x4-x6+x7-x9+x10-x12 <=-1

END
Then PORTA will create the following output file which is called shift.ieq.poi:
DIM = 12

CONE_SECTION
( 1) 0 0 0 0 0 0 0 0 0 1 1 1
( 2) 0 0 0 0 0 0 0 1 1 1 0 0
( 3) 0 0 0 0 0 0 1 1 1 0 0 0
( 4) 0 0 0 0 1 1 1 0 0 0 0 0
( 5) 0 0 0 1 1 1 0 0 0 0 0 0
292 7. EDGE SUBRINGS AND COMBINATORIAL OPTIMIZATION

( 6) 0 1 1 1 0 0 0 0 0 0 0 0
( 7) 1 0 0 0 0 0 0 0 0 0 1 1
( 8) 1 1 1 0 0 0 0 0 0 0 0 0

CONV_SECTION
( 1) 1 1 1 1 1 1 1 1 1 1 1 1

END
These two blocks define the point representations of the cone section and the convex
section of the shift polyhedron (see the finite basis theorem). The first block gives
the generators of the cone section and the second block gives the vertices of the shift
polyhedron. Therefore we obtain a(S ) = −4 = − 12 3 = deg(1) and type(S ) = 1.

Exercises
7.4.6. Let G = 𝒦4 and let C(G) = 𝒦5 be its cone. Prove that the shift polyhedron
of the cone ℝ+ 𝒜C(G) relative to some matrix C has the property that one of its vertices
does not correspond to a monomial in the canonical module of k[C(G)].
7.4.7. Consider the complete graph G = 𝒦4 . Prove that the shift polyhedron
(relative to some matrix C) of the Rees cone is not integral.

7.5. Some applications to edge subrings

As a first direct application we will recover an optimal upper bound for the a-
invariant of the edge subring k[G].
PROPOSITION 7.5.1 ([187]). If G is a connected bipartite graph with a bipartition
(V1 , V2 ) and ∣V1 ∣ = m ≤ ∣V2 ∣ = n, then
a(k[G]) ≤ −n.
PROOF. Let V1 = {v1 , . . . , vm }. Take a monomial xβ in the canonical module
of k[G], that is, β = (βi ) is an integral vector in ri(ℝ+ 𝒜G ). One may assume
∑n+mG is a star and the result is clear. Thus βi ≥ 1 by
m ≥ 2 because if m∑= 1 the graph
Remark 7.1.2, and i=1 βi = i=m+1 βi because the edge cone lies in the hyperplane
m

m
∑ m+n
∑
xi = xi .
i=1 i=m+1

deg(xβ ),
the normalized degree of xβ , is equal to i=m+1 βi .
∑m+n
Hence Therefore
deg(xβ ) ≥ n, which proves the required inequality. □
 7. EDGE SUBRINGS AND COMBINATORIAL OPTIMIZATION 293

COROLLARY 7.5.2 ([22]). Let G = 𝒦m,n . If n ≥ m, then

a(k[G]) = −n.
PROOF. By Corollary 6.2.12 and Proposition 7.5.1 it suffices to observe that the
monomial xβ with
β = (n − m + 1, 1, . . . , 1, 1, . . . , 1 )
| {z } | {z }
m entries n entries
is in the canonical module of k[G] and has degree equal to n. □
In Example 7.5.3 we show a 2-connected G graph where the a-invariant of k[G]
does not reach the upper bound of Proposition 7.5.1.
EXAMPLE 7.5.3. Consider the following bipartite graph G

vs6 vs3
@ @
@ @
v1 s @sv2 v7 s @s v8
@ @
@ @
@s @s
v5 v4

The bipartition of G is V1 = {v1 , . . . , v4 }, V1 = {v5 , . . . , v8 }. To compute the

a-invariant of k[G] using Mathematica we set up the next linear program
vars:={x1,x2,x3,x4,x5,x6,x7,x8}
f:=(x1+x2+x3+x4+x5+x6+x7+x8)/2
ieq:={x1+x2+x3+x4-x5-x6-x7-x8== 0, -x1<= -1, -x2<= -1,
-x3<= -1, -x4<= -1, -x5<= -1, -x6<= -1,
-x7<= -1, -x8<= -1, x3-x6-x7-x8<= -1,
x4-x5-x7-x8<= -1, x1+x2-x5-x6<= -1}
ConstrainedMin[f,ieq,vars]
The answer is
{5, {x1 -> 1, x2 -> 1, x3 -> 1, x4 -> 2, x5 -> 1,
x6 -> 2, x7 -> 1, x8 -> 1}}.
Thus a(k[G]) = −5 and the optimum value is attained at (1, 1, 1, 2, 1, 2, 1, 1), which
is one of the four vertices of the shift polyhedron.
A general formula for the type of a determinantal ring is due to J. Brennan, see
[30, p. 115]. We recover a special case of this formula by describing the generators
of the canonical module of a determinantal ring for the case of the 2-minors of a
generic m × n matrix.
294 7. EDGE SUBRINGS AND COMBINATORIAL OPTIMIZATION

PROPOSITION 7.5.4. Let G = 𝒦m,n . If n ≥ m, then k[G] is a level algebra, i.e.,

the canonical module is generated in a single degree, and:
( )
n−1
type(k[G]) = .
m−1
PROOF. One can rapidly see form Corollary 6.2.12 that the canonical module of
k[G] is minimally generated by the set of monomials of the form
x1a1 ⋅ ⋅ ⋅ xm
am
y1 ⋅ ⋅ ⋅ yn ,
such that a1 + ⋅ ⋅ ⋅ + am = n and ai ≥ 0 for all i. Thus the number of such partitions is
the type. As the canonical module is generated in a single degree the algebra k[G]
is level. □
PROPOSITION 7.5.5. If G = G(m, n) is a bipartite connected graph with 1 ≤
m ≤ n, then a(k[G]) = −m if and only if m = n and 1 = (1, . . . , 1) is in ri(ℝ+ 𝒜).
PROOF. ⇒) By Remark 7.1.2 the canonical module ωk[G] of k[G] is generated
by monomials xβ such that all the entries of β are positive integers. As n = m we
derive that x1 = x1 x2 ⋅ ⋅ ⋅ xm+n is in the canonical module and it is the only monomial
of normalized degree n in ωk[G] .
⇐) It follows from Proposition 7.5.1. □
DEFINITION 7.5.6. Let G be a graph. A cycle containing all the vertices of G is
said to be a Hamilton cycle, and a graph containing a Hamilton cycle is said to be
Hamiltonian.
COROLLARY 7.5.7. If G is a Hamiltonian bipartite graph with 2n vertices, then
a(k[G]) = −n.
PROOF. Let H be a Hamiltonian cycle of G and let 2n be the number of vertices
of G. Note a(k[H]) = −n (because k[H] is the ring of a hypersurface of degree n) and
dim k[H] = dim k[G]. Hence a(k[H]) ≤ a(k[G]) and consequently a(k[G]) = −n.
□
COROLLARY 7.5.8. If G = 𝒦n,n is a complete bipartite graph, then the a-
invariant of k[G ∖ {e}] is equal to −n.
PROOF. Note that for n ≥ 3 the graph G ∖ {e} is Hamiltonian. □
Decomposing a bipartite graph into blocks. Let G be a connected bipartite
graph. Next we describe in a condense form a technique that can be used to simplify
the computation of the a-invariant and the type of k[G]. It is based in the graph
theoretical notion of block (see Section 5.5) and the following result.
PROPOSITION 7.5.9. If B and C are two standard k-algebras, then
a(B ⊗k C) = a(B) + a(C).
 7. EDGE SUBRINGS AND COMBINATORIAL OPTIMIZATION 295

If in addition B and C are Cohen-Macaulay, then

type(B ⊗k C) = type(B) ⋅ type(C).
PROOF. The first equality follows from the proof of [187, Proposition 4.1.14].
For the second equality see [75]. □
COROLLARY 7.5.10. If A and B are two standard algebras over a field k, then
A ⊗k B is Gorenstein if and only if A and B are Gorenstein.
PROOF. It follows from Corollary 1.5.36 and using that the type of the tensor
products of two Cohen-Macaulay standard algebras is the product of their types, see
Proposition 7.5.9. □
PROPOSITION 7.5.11. Let G be a connected bipartite graph and let G1 , . . . , Gr
be its blocks, then
r
∑ ∏r
a(k[G]) = a(k[Gi ]) and type(k[G]) = type(k[Gi ]).
i=1 i=1

PROOF. Since G is connected and bipartite one has

k[G] ≃ k[G1 ] ⊗k k[G2 ] ⊗k ⋅ ⋅ ⋅ ⊗k k[Gr ].
Hence the result follows from Proposition 7.5.9. □
The next theorem gives a combinatorial obstruction for an edge subring to be
Gorenstein. It suffices to consider the class of Gorenstein subrings of 2-connected
bipartite graphs, see Proposition 7.5.11
THEOREM 7.5.12. Let G be a bipartite graph with bipartition (V1 , V2 ). If k[G]
is Gorenstein and G is connected, then G has a perfect matching and furthermore
∣A∣ < ∣N(A)∣ for all A ⊊ V1 .
PROOF. Let xβ be the generator of ωk[G] . For each 1 ≤ i ≤ m + n choose an
spanning tree of G ∖ {vi } and enlarge this to an spanning tree 𝒯i of G. Note that the
monomial xγi with
γi = (deg𝒯i (v1 ), . . . , deg𝒯i (vi ), . . . , deg𝒯i (vn+m ))
| {z }
1
is the generator of the canonical module of k[𝒯i ]. Hence since aff(ℝ+ 𝒜i ) =
aff(ℝ+ 𝒜), where ℝ+ 𝒜i is the edge cone of 𝒯i , we get xγi ∈ ωk[G] . Therefore
βi = 1, and consequently β = 1. Noting that β is in the affine space generated by the
edge cone of G it follows m = n. Thus from Theorem 6.2.8 we obtain ∣A∣ < ∣N(A)∣
for all A ⊊ V1 . □
From the proof of Theorem 7.5.12 we obtain:
296 7. EDGE SUBRINGS AND COMBINATORIAL OPTIMIZATION

COROLLARY 7.5.13. Let G = G(m, n) be a connected bipartite graph. If G has

no cut vertices and k[G] is a Gorenstein domain, then G has a perfect matching and
the a-invariant of k[G] is equal to −n.
If one of the blocks of a graph G is a bridge e one can say a bit more about the
relationship between G and G ∖ {e}.
PROPOSITION 7.5.14. Let G be a graph with a bridge e. If G ∖ {e} = G1 ∪ G2
with G1 and G2 disjoint graphs such that k[Gi ] is Gorenstein for i = 1, 2, then k[G]
is Gorenstein.
PROOF. There is an isomorphism k[G] ≃ (k[G1 ] ⊗k k[G2 ])[t], where t is a
variable. Since the tensor product of Gorenstein rings is Gorenstein and a polynomial
ring over a Gorenstein ring is Gorenstein, the result follows. □
From the isomorphism above we also obtain.
PROPOSITION 7.5.15. Let G be a graph with a bridge e. If G ∖ {e} = G1 ∪ G2
with G1 and G2 disjoint graphs, then
a(G) = a(G1 ) + a(G2 ) − 1.

Exercises
Notice that in Proposition 7.5.14 we can relate the generators of the canonical
modules of k[G] and k[G ∖ {e}] using the following:
7.5.16. Let G be a graph with n vertices and without isolated vertices and let
𝒜 = {α1 , . . . , αq }. If β = ek + e` ∈
/ ℝ𝒜 and α ∈ ri(ℝ+ 𝒜), then α + β ∈ ri(ℝ+ ℬ ),
where ℬ = 𝒜 ∪ {β}.

7.6. The a-invariant for non bipartite graphs

Here we study the a-invariant for non-bipartite graphs. Let G be an arbitrary
simple graph. We grade the edge subring k[G] ⊂ R with the normalized grading
k[G]i = k[G] ∩ R2i , where R is a polynomial ring over the field k with the standard
grading. If k[G] is normal, then by Theorem 4.3.27, the canonical module ωk[G] of
k[G] is given by
(7.70) ωk[G] = ({ xa ∣ a ∈ ℕ𝒜G ∩ ri(ℝ+ 𝒜G )}).
DEFINITION 7.6.1. The join G1 ∗ G2 of two disjoint graphs G1 and G2 consists
of G1 ∪ G2 and all the lines joining G1 with G2 .
 7. EDGE SUBRINGS AND COMBINATORIAL OPTIMIZATION 297

PROPOSITION 7.6.2. [29, Corollary 3.6] If F is a finite set of square-free mono-

mials of the same degree r and dim K[F] = n, then
⎨ − ⌈n/r⌉ ,
⎧
if n ≥ 2r,
a(K[F]) ≤
− ⌈n/(n − r)⌉ , if n ≤ 2r, n ∕= r.
⎩

PROPOSITION 7.6.3. Let G1 and G2 be two connected graphs on n vertices. If

k[G1 ] and k[G2 ] are normal, then
a(k[G1 ∗ G2 ]) = −n.
PROOF. Let V1 = {v1 , . . . , vn } and V2 = {vn+1 , . . . , v2n } be the vertex set of G1
and G2 respectively. One may assume n ≥ 2. Set G = G1 ∗ G2 . Note that the vector
β = (1, . . . , 1) ∈ ℝ2n satisfies
∑ ∑
xi ≤ −1 + xi and xi ≥ 1,
vi ∈A vi ∈N(A)

for all independent sets A of G, because ∣A∣ ≤ n − 1 < ∣N(A)∣. Hence by Corol-
lary 6.1.8 and Lemma 5.3.4 we get β ∈ (ℝ+ 𝒜G )o ∩ ℕ𝒜G . Since k[G] is normal by
Corollary 5.6.14, we apply Eq. (7.70) to get xβ ∈ ωk[G] . This implies −n ≤ a(k[G]).
On the other hand by Proposition 7.6.2, we conclude a(k[G]) = −n. □
PROPOSITION 7.6.4. Let G be a connected non bipartite graph on n vertices. If
k[G] is normal, then
⌈ ⌉
n+1
a(k[G]) − 1 ≤ a(k[C(G)]) ≤ − .
2
PROOF. Let ω1 and ω2 be the canonical modules of k[G] and k[C(G)] respec-
tively. Set 𝒜 = 𝒜G and V = V(G). By Lemma 5.3.4 one has

ri(ℝ+ 𝒜) = (ℝ+ 𝒜)o ⊂ ℝn and ri(ℝ+ 𝒜C(G) ) = (ℝ+ 𝒜C(G) )o ⊂ ℝn+1 .

As k[G] is normal and according to Eq.(7.70), we may pick β ∈ (ℝ+ 𝒜)o ∩ ℕ𝒜 so
that deg(xβ ) is equal to −a(k[G]). Set γ = (β1 + 1, β2 , . . . , βn , 1). To prove the first
inequality we will show that γ is in (ℝ+ 𝒜C(G) )o ∩ ℕ𝒜C(G) , that is, one must show
that γ is not on any facet of ℝ+ 𝒜C(G) .
Let F be a facet of ℝ+ 𝒜C(G) and H the hyperplane defining F. By Theorem 6.1.7
there are three cases to consider. If H = Hei for some 1 ≤ i ≤ n, then γ ∕∈ H;
for otherwise β would be on the proper face of ℝ+ 𝒜 determined by Hei , which
contradicts the choice of β (see Remark 7.1.2). If H = Hen+1 , note ⟨γ, en+1 ⟩ = 1 and
γ ∕∈ H. Assume that H = HA , for some independent set in C(G). If vn+1 ∈ A, then
γ satisfies ∑
xn+1 ≤ −1 + xi
vi ∈V
298 7. EDGE SUBRINGS AND COMBINATORIAL OPTIMIZATION

and γ ∕∈ HA . Assume vn+1 ∕∈ A, since β satisfies the inequality

∑ ∑
xi ≤ −1 + xi ,
vi ∈A vi ∈NG (A)
it follows that γ satisfies
∑ ∑
xi ≤ −1 + xi .
vi ∈A vi ∈NC(G) (A)

Hence γ ∕∈ HA . Observe that k[C(G)] is normal by Corollary 5.6.14. Thus using

Eq.(7.70) one obtains xγ ∈ ω2 and the proof of the first inequality is complete. The
second inequality follows from Proposition 7.6.2. □
PROPOSITION 7.6.5. Let G1 , G2 be two connected non bipartite graphs. If k[G1 ]
and k[G2 ] are normal, then
a(k[G1 ]) + a(k[G2 ]) ≤ a(k[G1 ∗ G2 ]).
PROOF. Choose vectors a and b in
(ℝ+ 𝒜G1 )o ∩ ℕ𝒜G1 and (ℝ+ 𝒜G2 )o ∩ ℕ𝒜G2
respectively, such that deg(xa ) = −a(k[G1 ]) and deg(xb ) = −a(k[G2 ]). Using
arguments similar to those given in the proof of Proposition 7.6.4 it follows that
(a, b) is an interior point of ℝ+ 𝒜G1 ∗G2 , which implies the asserted inequality. □

An application to graph theory

Edge subrings are useful to study matching problems in graphs. We begin with the
following membership criterion.
PROPOSITION 7.6.6 ([186]). Let G be a connected non bipartite graph with n
vertices. If k[G] is a normal domain, then a monomial x1β1 ⋅ ⋅ ⋅ xnβn belongs to k[G] if
and only if the following two conditions hold
(i) β = (β1 , . . . , βn ) is in the edge cone of G, and
i=1 βi is an even integer.
∑n
(ii)
PROOF. Set 𝒜 = 𝒜G = {α1 , . . . , αq }. Assume β ∈ ℝ+ 𝒜 and deg(xβ ) even. We
proceed by induction on deg(xβ ). Using Corollary 5.3.9 one has the isomorphism
ℤn /(α1 , . . . , αq ) ≃ ℤ2 ,
hence 2β ∈ ℝ+ 𝒜 ∩ ℤ𝒜 = ℕ𝒜 and one may write
q
∑ q
∑
2β = 2 si αi + i α i , si ∈ ℕ and i ∈ {0, 1},
i=1 i=1
by induction one may assume qi=1 si αi = 0. Therefore from the equality above one
∑
concludes that the subgraph whose edges are defined by the set {αi ∣ i = 1} is an
 7. EDGE SUBRINGS AND COMBINATORIAL OPTIMIZATION 299

edge disjoint union of cycles Z1 , . . . , Zr . By induction one may further assume that
all the Zi ’s are odd cycles. Note r ≥ 2, because deg(xb ) is even. As G is connected,
using Corollary 5.6.14 it follows that xβ ∈ k[G]. The converse is clear because k[G]
is normal. □
REMARK 7.6.7. Let us give a simple example to see that the connectivity hy-
pothesis is essential in Proposition 7.6.6. Consider the graph G = G1 ∪ G2 with six
vertices consisting of two disjoint triangles:
∙??? ∙
?? 
?? 
?? 

?
∙ ?? G2
G1
∙ ??
 ??
 ??

∙ ∙
The vector β = (1, . . . , 1) belongs to the edge cone because 2β ∈ ℕ𝒜, but xβ is not
in k[G].
Let G be a connected non bipartite graph with k[G] normal. The last results can
be used to give conditions for a graph to have a perfect matching.
PROPOSITION 7.6.8. Let G be a connected graph with n vertices. If n is even and
k[G] is normal of dimension n, then x1 ⋅ ⋅ ⋅ xn is in k[G] if and only if ∣A∣ ≤ ∣N(A)∣
for every independent set of vertices A of G.
PROOF. ⇒) Since k[G] is normal:
ℝ+ 𝒜G ∩ ℤ𝒜G = ℕ𝒜G .
Hence a = (1, . . . , 1) = (a1 , . . . , an ) is in ℝ+ 𝒜G . Using Corollary 6.1.8 we get that
the vector a satisfies the inequalities:
∑ ∑
∣ A∣ = ai ≤ ai = ∣N(A)∣
vi ∈A vi ∈N(A)

for every independent set of vertices A of G, as required.

⇐) First we use Corollary 6.1.8 to conclude that a is in the edge cone, then
apply Proposition 7.6.6 to get xa ∈ k[G]. □
COROLLARY 7.6.9 (Marriage theorem). Let G be a graph with an even number
of vertices. If G is connected and satisfies the odd cycle condition, then the following
are equivalent
(a) G has a perfect matching.
(b) ∣A∣ ≤ ∣N(A)∣ for all A independent set of vertices of G.

PROOF. The proof follows using Theorem 2.1.9 and Proposition 7.6.8. □
300 7. EDGE SUBRINGS AND COMBINATORIAL OPTIMIZATION

Exercises
7.6.10. Let G = G1 ∪G2 be a graph with two non bipartite connected components
G1 and G2 . Find a relation between the irreducible representation of the cone ℝ+ 𝒜G
and the irreducible representations of the cones ℝ+ 𝒜G1 and ℝ+ 𝒜G2 .
7.6.11. Let G be a connected non bipartite graph with n vertices and let α1 , . . . , αq
be the columns of its incidence matrix. If n is even, prove that the vector (1, . . . , 1)
is in L = ℤα1 + ⋅ ⋅ ⋅ + ℤαq .
Hint The map ϕ : ℤn → ℤ2 given by ϕ(a) = ∣a∣ is onto and ker(A) = L.
7.6.12. Let G be a connected non bipartite graph with n vertices. If n is even,
prove that the monomial x1 ⋅ ⋅ ⋅ xn is in the field of fractions of k[G].
7.6.13. Let G be a connected non bipartite graph with n vertices. If k[G] is a
normal domain, then x1 ⋅ ⋅ ⋅ xn is in k[G] if and only if (1, . . . , 1) is in the edge cone
of G and n is even.
7.6.14. Prove that the following graph does not have a perfect matching by
exhibiting an independent set A such that ∣A∣ ∕≤ ∣N(A)∣.
s
@
@ s
@ 
s
@
@s s
 
HH
@  HH
H  H
@ HHs
 HHs
@
@
@s

7.6.15. Prove that the following are the equations of the edge cone of the graph
G shown on the left.

s x1
@ − x1 ≤ 0
− x4 ≤
@
@ 0
− x5 ≤ 0
x3 sH @s x2
@
@H  x3 ≤ x1 + x2 + x4 + x5
HH x4 
@ Hs x2 ≤ x1 + x3 + x4 + x5
@ x1 + x4 + x5 ≤ x2 + x3
@s x5
@

Show that the a-invariant of k[G] is equal to −4.

 7. EDGE SUBRINGS AND COMBINATORIAL OPTIMIZATION 301

7.6.16. Let G be the graph consisting of two squares joined by an edge. Prove
that k[G] is a Gorenstein ring, its a-invariant is −5, it has a perfect matching, and
α = (1, . . . , 1) is not in the relative interior of the edge cone of G. Prove that the
monomial corresponding to (1, 1, 2, 1, 1, 2, 1, 1) generates the canonical module of
k[G].
Bibliography
[1] M. Aigner, Combinatorial Theory, Springer, 1997.
[2] A. Alcántar, Rees algebras of square-free Veronese ideals and their a-invariants, Discrete Math.
302 (2005), 7–21.
[3] M. F. Atiyah and I. G. Macdonald, Introduction to Commutative Algebra, Addison-Wesley,
Reading, MA, 1969.
[4] D. Avis and K. Fukuda, A pivoting algorithm for convex hulls and vertex enumeration of
arrangements and polyhedra, Discrete Comput. Geom. 8 (1992), no. 3, 295–313.
[5] D. Avis and K. Fukuda, Reverse search for enumeration, Discrete Appl. Math. 65 (1996), no.
1-3, 21–46.
[6] J. Bang-Jensen and G. Gutin, Digraphs. Theory, Algorithms and Applications, Springer Mono-
graphs in Mathematics, Springer, 2nd printing, 2006.
[7] M. Barile, On the arithmetical rank of the edge ideals of forests, Comm. Algebra 36 (2008), no.
12, 4678 - 4703.
[8] C. Berge, Some common properties for regularizable graphs, edge-critical graphs and B-graphs,
in “Theory and practice of combinatorics” (A. Rosa, G. Sabidussi and J. Turgeon, Eds.), North-
Holland Math. Stud. 60, North-Holland, Amsterdam, 1982, pp. 31–44.
[9] C. Berge, Hypergraphs Combinatorics of Finite Sets, Mathematical Library 45, North-Holland,
1989.
[10] D. Bertsimas and J. N. Tsitsiklis, Introduction to linear optimization, Athena Scientific, Mas-
sachusetts, 1997.
[11] A. Björner, Topological methods, Handbook of combinatorics, Vol. 1, 2, 1819–1872, Elsevier,
Amsterdam, 1995.
[12] A. Björner, M. Wachs, Shellable nonpure complexes and posets. I. Trans. Amer. Math. Soc. 348
(1996), 1299–1327.
[13] J. Blasiak, The toric ideal of a graphic matroid is generated by quadrics, Combinatorica 28
(2008), no. 3, 283–297.
[14] B. Bollobás, Modern Graph Theory, Graduate Texts in Mathematics 184 Springer-Verlag, New
York, 1998.
[15] V. Bonanzinga, C. Escobar and R. H. Villarreal, On the normality of Rees algebras associated
to totally unimodular matrices, Results Math. 41, 3/4, (2002), 258–264.
[16] P. Bouwknegt, q-identities and affinized projective varieties, I. Quadratic monomial ideals,
arXiv:math-ph/9902010.
[17] J. P. Brennan, L. A. Dupont and R. H. Villarreal, Duality, a-invariants and canonical modules of
rings arising from linear optimization problems, Bull. Math. Soc. Sci. Math. Roumanie (N.S.)
51 (2008), no. 4, 279–305.
[18] A. Brøndsted, Introduction to Convex Polytopes, Graduate Texts in Mathematics 90, Springer-
Verlag, 1983.

303
304 BIBLIOGRAPHY

[19] W. Bruns and J. Gubeladze, Normality and covering properties of affine semigroups, J. reine
angew. Math. 510 (1999), 161–178.
[20] W. Bruns and J. Gubeladze, Polytopes, rings, and K-theory, Springer Monographs in Mathe-
matics. Springer, Dordrecht, 2009.
[21] W. Bruns, J. Gubeladze and N. V. Trung, Normal polytopes, triangulations, and Koszul algebras,
J. reine angew. Math. 481 (1997), 123–160.
[22] W. Bruns and J. Herzog, On the computation of a-invariants, Manuscripta Math. 77 (1992),
201–213.
[23] W. Bruns and J. Herzog, Cohen-Macaulay Rings, Revised Edition, Cambridge University Press,
Cambridge, 1997.
[24] W. Bruns and B. Ichim, NORMALIZ 2.0, Computing normalizations of affine semigroups 2008.
Available from
http://www.math.uos.de/normaliz.
[25] W. Bruns and R. Koch, NORMALIZ, computing normalizations of affine semigroups 2003.
Available from:
ftp.mathematik.Uni-Osnabrueck.DE/pub/osm/kommalg/software.
[26] W. Bruns and R. Koch, Computing the integral closure of an affine semigroup. Effective methods
in algebraic and analytic geometry, 2000 (Kraków). Univ. Iagel. Acta Math. 39 (2001), 59–70.
[27] W. Bruns and G. Restuccia, Canonical modules of Rees algebras, J. Pure Appl. Algebra 201
(2005), 189–203.
[28] W. Bruns, A. Simis and N. V. Trung, Blow-up of straightening-closed ideals in ordinal Hodge
algebras, Trans. Amer. Math. Soc. 326 (1991), 507–528.
[29] W. Bruns, W. V. Vasconcelos and R. H. Villarreal, Degree bounds in monomial subrings, Illinois
J. Math. 41 (1997), 341–353.
[30] W. Bruns and U. Vetter, Determinantal Rings, Lecture Notes in Mathematics 1327, Springer-
Verlag, Berlin-Heidelberg-New York, 1988.
[31] B. Buchberger, An algorithmic method in polynomial ideal theory, in Recent Trends in Mathe-
matical Systems Theory (N.K. Bose, Ed.), Reidel, Dordrecht, 1985, 184–232.
[32] A. Capani, G. Niesi and L. Robbiano, CoCoA: A system for doing computations in commutative
algebra. 1998. Available from
lancelot.dima.unige.it.
[33] G. Carrá Ferro and D. Ferrarello, Cohen-Macaulay graphs arising from digraphs. (2007)
Preprint. math.AC/0703417.
[34] G. J. Chang and M. J. Jou, Survey on counting maximal independent sets, in “Proceedings of
the second Asian Mathematical Conference 1995” (S. Tangmanee and E. Schulz, Eds.), Word
Scientific, River Edge, NJ., 1998, pp. 265–275.
[35] J. Chen, S. Morey and A. Sung, The stable set of associated primes of the ideal of a graph,
Rocky Mountain J. Math. 32 (2002), 71–89.
[36] T. Christof, revised by A. Löbel and M. Stoer, PORTA: A Polyhedron Representation Transfor-
mation Algorithm, 1997. Available from:
http://www.zib.de/Optimization/Software/Porta.
[37] M. Conforti and G. Cornuéjols, Clutters that pack and the Max-Flow Min-Cut property: A
conjecture, The Fourth Bellairs Workshop on Combinatorial Optimization (W. R. Pulleyblank,
F. B. Shepherd, eds.).
[38] G. Cornuéjols, Combinatorial optimization: Packing and covering, CBMS-NSF Regional Con-
ference Series in Applied Mathematics 74, SIAM (2001).
[39] G. Cornuéjols, B. Guenin and F. Margot, The packing property, Math. Program. 89 (2000),
no. 1, Ser. A, 113–126.
 BIBLIOGRAPHY 305

[40] D. Cox, J. Little and D. O’Shea, Ideals, Varieties, and Algorithms, Springer-Verlag, 1992.
[41] D. Cox, J. Little and D. O’Shea, Using Algebraic Geometry, Graduate Texts in Mathematics
185, Springer-Verlag, 1998.
[42] V. I. Danilov, The geometry of toric varieties, Russian Math. Surveys 33 (1978), 97-154.
[43] D. Delfino, A. Taylor, W. V. Vasconcelos, R. H. Villarreal and N. Weininger, Monomial ideals
and the computation of multiplicities, Commutative ring theory and applications (Fez, 2001),
pp. 87–106, Lecture Notes in Pure and Appl. Math. 231, Dekker, New York, 2003.
[44] J. A. De Loera, B. Sturmfels and R. R. Thomas, Gröbner bases and triangulations of the second
hypersimplex, Combinatorica 15 (1995), 409–424.
[45] R. Diestel, Graph Theory, Graduate Texts in Mathematics 173, Springer-Verlag, New York,
2nd ed., 2000.
[46] G. A. Dirac, On rigid circuit graphs, Abh. Math. Sem. Univ. Hamburg 38 (1961), 71–76.
[47] A. Dochtermann and A. Engstrom, Algebraic properties of edge ideals via combinatorial topol-
ogy, Electron. J. Combin. 16 (2009), no. 2, R2.
[48] L. Doering and T. Gunston, Algebras arising from planar bipartite graphs, Comm. Algebra 24
(1996), 3589–3598.
[49] A. M. Duval, Algebraic shifting and sequentially Cohen-Macaulay simplicial complexes. Elec-
tron. J. Combin. 3 (1996), no. 1, Research Paper 21.
[50] L. A. Dupont, Rees algebras, Monomial Subrings and Linear Optimization Problems, PhD
thesis, Cinvestav–IPN. In preparation.
[51] L. A. Dupont and R. H. Villarreal, Algebraic and combinatorial properties of ideals and algebras
of uniform clutters of TDI systems, J. Comb. Optim., to appear.
[52] L. A. Dupont and R. H. Villarreal, Edge ideals of clique clutters of comparability graphs and
the normality of monomial ideals, Math. Scand., to appear.
[53] J. A. Eagon and V. Reiner, Resolutions of Stanley-Reisner rings and Alexander duality, J. Pure
& Applied Algebra 130 (1998), 265–275.
[54] E. Ehrhart, Sur une probléme de géométrie diophantine linéaire, J. reine angew. Math. 227
(1967), 1–29.
[55] E. Ehrhart, Démonstration de la loi de réciprocité du polyedre rationnel, C. R. Acad. Sci. Paris
Ser. A 265 (1967), 91–94.
[56] D. Eisenbud, Commutative Algebra with a view toward Algebraic Geometry, Graduate Texts in
Mathematics 150, Springer-Verlag, 1995.
[57] D. Eisenbud, C. Huneke and W. V. Vasconcelos, Direct methods for primary decomposition,
Invent. Math. 110 (1992), 207–235.
[58] D. Eisenbud and B. Sturmfels, Binomial ideals, Duke Math. J. 84 (1996), 1–45.
[59] R. B. Ellis and D. Ferrero, The spectra of super line multigraphs, preprint.
[60] E. Emtander, Betti numbers of hypergraphs, Comm. Algebra 37 (2009), no. 5, 1545 - 1571.
[61] P. Erdös and T. Gallai, On the minimal number of vertices representing the edges of a graph,
Magyar Tud. Akad. Mat. Kutató Int. Közl. 6 (1961), 181–203.
[62] P. Erdös, A. Hajnal and J.W. Moon, A problem in graph theory, Amer. Math. Monthly 71 (1964),
1107–1110.
[63] C. Escobar, Normal monomial subrings, unimodular matrices and Ehrhart rings, PhD thesis,
Cinvestav–IPN, 2004.
[64] C. Escobar, J. Martı́nez-Bernal and R. H. Villarreal, Relative volumes and minors in monomial
subrings, Linear Algebra Appl. 374 (2003), 275–290.
[65] C. Escobar, R. H. Villarreal and Y. Yoshino, Torsion freeness and normality of blowup rings
of monomial ideals, Commutative Algebra, Lect. Notes Pure Appl. Math. 244, Chapman &
Hall/CRC, Boca Raton, FL, 2006, pp. 69-84.
306 BIBLIOGRAPHY

[66] M. Estrada and R. H. Villarreal, Cohen-Macaulay bipartite graphs, Arch. Math. 68 (1997),
124–128.
[67] G. Ewald, Combinatorial Convexity and Algebraic Geometry, Graduate Texts in Mathematics
168, Springer-Verlag, New York, 1996.
[68] S. Faridi, Simplicial trees are sequentially Cohen-Macaulay. J. Pure Appl. Algebra 190 (2004),
121–136.
[69] S. Faridi, Cohen-Macaulay properties of square-free monomial ideals, J. Combin. Theory Ser.
A 109(2) (2005), 299-329.
[70] H. Flaschka and L. Haine, Torus orbits on G/P, Pacific J. Math. 149 (1991), 251–292.
[71] G. Fløystad and J. E. Vatne, (Bi-)Cohen-Macaulay simplicial complexes and their associated
coherent sheaves, Comm. Algebra 33 (2005), no. 9, 3121–3136.
[72] C. A. Francisco and A. Van Tuyl, Sequentially Cohen-Macaulay edge ideals. Proc. Amer. Math.
Soc. 135 (2007), 2327-2337.
[73] P. Frankl, Extremal set systems, in Handbook of Combinatorics II (R. L. Graham et. al., Eds.),
Elsevier, 1995, pp. 1293–1329.
[74] R. Fröberg, A study of graded extremal rings and of monomial rings, Math. Scand. 51 (1982),
22–34.
[75] R. Fröberg, A note on the Stanley-Reisner ring of a join and of a suspension, Manuscripta Math.
60 (1988), 89-91.
[76] R. Fröberg, On Stanley-Reisner rings, in Topics in algebra (S. Balcerzyk et. al., Eds.), Part 2.
Polish Scientific Publishers, 1990, pp. 57–70.
[77] D. R. Fulkerson, A. J. Hoffman and M.H. McAndrew, Some properties of graphs with multiple
edges, Canad. J. Math. 17 (1965), 166-177.
[78] E. Gawrilow, and M. Joswig, polymake: a framework for analyzing convex polytopes.
Polytopes—combinatorics and computation (Oberwolfach, 1997), 43–73, DMV Sem., 29,
Birkhäuser, Basel, 2000.
[79] I. Gitler, E. Reyes and R. H. Villarreal, Blowup algebras of ideals of vertex covers of bipartite
graphs, Contemp. Math. 376 (2005), 273–279.
[80] I. Gitler, E. Reyes and R. H. Villarreal, Blowup algebras of square–free monomial ideals and
some links to combinatorial optimization problems, Rocky Mountain J. Math. 39 (2009), no. 1,
71–102.
[81] I. Gitler and C. Valencia, Bounds for invariants of edge-rings, Comm. Algebra 33 (2005),
1603-1616.
[82] I. Gitler and C. Valencia, Multiplicities of edge subrings, Discrete Math. 302 (2005), 107-123.
[83] I. Gitler, C. Valencia and R. H. Villarreal, A note on the Rees algebra of a bipartite graph, J.
Pure Appl. Algebra, 201 (2005), 17–24.
[84] I. Gitler, C. Valencia and R. H. Villarreal, A note on Rees algebras and the MFMC property,
Beiträge Algebra Geom. 48 (2007), No. 1, 141-150.
[85] C. Godsil and G. Royle, Algebraic Graph Theory , Graduate Texts in Mathematics 207, Springer,
2001.
[86] M. C. Golumbic, Algorithmic graph theory and perfect graphs, second edition, Annals of
Discrete Mathematics 57, Elsevier Science B.V., Amsterdam, 2004.
[87] M. González-Sarabia and C. Renterı́a-Márquez, Evaluation codes associated to complete bipar-
tite graphs. Int. J. Algebra 2 (2008), no. 1-4, 163–170.
[88] P. Gordan, Üeber die Auflösung linearer Gleichungen mit reellen Coefficienten, Math. Ann. 6
(1873), no. 1, 23–28.
[89] S. Goto and K. Nishida, The Cohen-Macaulay and Gorenstein Rees algebras associated to
filtrations, Mem. Amer. Math. Soc. 110 (1994), no. 526, 1–134.
 BIBLIOGRAPHY 307

[90] D. Grayson and M. Stillman, Macaulay2, 1996. Available from

math.uiuc.edu.
[91] J. L. Gross and T. W. Tucker, Topological graph theory, reprint of the 1987 original Wiley, New
York, Dover Publications, Inc., Mineola, NY, 2001.
[92] J. Grossman, D. M. Kulkarni and I. Schochetman, On the minors of an incidence matrix and its
Smith normal form, Linear Algebra Appl. 218 (1995), 213–224.
[93] H. T. Hà and S. Morey, Embedded associated primes of powers of square-free monomial ideals,
J. Pure Appl. Algebra, to appear.
[94] H. T. Hà, S. Morey and R. H. Villarreal, Cohen-Macaulay admissible clutters, J. Commut.
Algebra 1 (2009), no. 3, 463–480.
[95] H. T. Hà and A. Van Tuyl, Monomial ideals, edge ideals of hypergraphs, and their graded Betti
numbers, J. Algebraic Combin. 27(2008), 215–245.
[96] F. Harary, Graph Theory, Addison-Wesley, Reading, MA, 1972.
[97] B. Hedman, The maximum number of cliques in dense graphs, Discrete Math. 54 (1985),
161–166.
[98] J. Herzog, Generators and relations of abelian semigroups and semigroup rings, Manuscripta
Math. 3 (1970), 153–193.
[99] J. Herzog, Alexander duality in commutative algebra and combinatorics, Algebra Colloq. 11
(2004), no. 1, 21–30.
[100] J. Herzog and T. Hibi, Componentwise linear ideals, Nagoya Math. J. 153 (1999), 141–153.
[101] J. Herzog and T. Hibi, Discrete polymatroids, J. Algebraic Combin. 16(3) (2002), 239–268.
[102] J. Herzog and T. Hibi, Distributive lattices, bipartite graphs and Alexander duality, J. Algebraic
Combin. 22 (2005), no. 3, 289–302.
[103] J. Herzog, T. Hibi and N. V. Trung, Symbolic powers of monomial ideals and vertex cover
algebras, Adv. Math. 210 (2007), 304–322.
[104] J. Herzog, T. Hibi and N. V. Trung, Vertex cover algebras of unimodular hypergraphs, Proc.
Amer. Math. Soc. 137 (2009), 409-414.
[105] J. Herzog, T. Hibi, N. V. Trung and X. Zheng, Standard graded vertex cover algebras, cycles
and leaves, Trans. Amer. Math. Soc. 360 (2008), 6231-6249.
[106] J. Herzog, T. Hibi and M. Vladoiu, Ideals of fiber type and polymatroids, Osaka J. Math. 42
(2005), 1–23.
[107] J. Herzog, T. Hibi and X. Zheng, Dirac’s theorem on chordal graphs and Alexander duality,
European J. Combin. 25 (2004), 949–960.
[108] J. Herzog, A. Simis and W. V. Vasconcelos, Koszul homology and blowing-up rings, in Com-
mutative Algebra, Proceedings: Trento 1981 (S. Greco and G. Valla, Eds.), Lecture Notes in
Pure and Applied Math. 84, Marcel Dekker, New York, 1983, pp. 79-169.
[109] J. Herzog, A. Simis and W. V. Vasconcelos, Arithmetic of normal Rees algebras, J. Algebra 143
(1991), 269–294.
[110] T. Hibi, Union and glueing of a family of Cohen-Macaulay partially ordered sets, Nagoya Math.
J. 107 (1987), 91–119.
[111] T. Hibi, Algebraic Combinatorics on Convex Polytopes, Carslaw Publications, Australia, 1992.
[112] T. Hibi and H. Ohsugi, Normal polytopes arising from finite graphs, J. Algebra 207 (1998),
409–426.
[113] M. Hochster, Rings of invariants of tori, Cohen-Macaulay rings generated by monomials, and
polytopes, Ann. of Math. 96 (1972), 318–337.
[114] M. Hochster, Cohen-Macaulay rings, combinatorics and simplicial complexes, in Ring Theory
II, Lecture Notes in Pure and Applied Math. 26, Marcel Dekker, New York, 1977, pp. 171–223.
308 BIBLIOGRAPHY

[115] C. Huneke, A. Simis and W. V. Vasconcelos, Reduced normal cones are domains, Contemp.
Math. 88 (1989), 95–101.
[116] C. Huneke and I. Swanson, Integral Closure of Ideals Rings, and Modules, London Math. Soc.,
Lecture Note Series 336, Cambridge University Press, Cambridge, 2006.
[117] N. Jacobson, Basic Algebra I, Second Edition, W. H. Freeman and Company, New York, 1996.
[118] N. Jacobson, Basic Algebra II, W. H. Freeman and Company, New York, 1989.
[119] M. Katzman, Bipartite graphs whose edge algebras are complete intersections, J. Algebra 220
(1999), 519–530.
[120] B. Korte and J. Vygen, Combinatorial Optimization Theory and Algorithms, Algorithms and
Combinatorics 21, Springer-Verlag, Berlin, 2000.
[121] D. König, Graphen und Matrizen, Mat. Fiz. Lapok 30 (1931), 116–119.
[122] M. Kummini, Regularity, depth and arithmetic rank of bipartite edge ideals, J. Algebraic
Combin., to appear.
[123] J. P. S. Kung, Basis-exchange properties, Theory of matroids, 62–75, Encyclopedia Math. Appl.,
26, Cambridge Univ. Press, Cambridge, 1986.
[124] S. Lang, Calculus of Several Variables, Addison-Wesley, 1973.
[125] A. Lehman, On the width-length inequality and degenerate projective planes, in Polyhedral
Combinatorics (W. Cook and P. Seymour Eds.) DIMACS Series in Discrete Mathematics and
Theoretical Computer Science 1, Amer. Math. Soc., 1990, pp. 101-105
[126] M. Llabres and F. Rossello, A new family of metrics for biopolymer contact structures, Comput.
Biol. Chem. 28 (1) (2004), 21–37.
[127] C. L. Lucchesi and D. H. Younger, A minimax theorem for directed graphs, J. London Math.
Soc. (2) 17 (1978), no. 3, 369–374.
[128] G. Lyubeznik, Set theoretic intersections and monomial ideals, PhD thesis, Columbia University,
1984.
[129] G. Lyubeznik, The minimal non-Cohen-Macaulay monomial ideals, J. Pure Applied Algebra
51 (1988), 261–266.
[130] J. Martı́nez-Bernal and R. H. Villarreal, Toric ideals generated by circuits, Algebra Colloq., to
appear.
[131] H. Matsumura, Commutative Algebra, Benjamin-Cummings, Reading, MA, 1980.
[132] H. Matsumura, Commutative Ring Theory, Cambridge Studies in Advanced Mathematics 8,
Cambridge University Press, 1986.
[133] E. Miller and B. Sturmfels, Combinatorial Commutative Algebra, Graduate Texts in Mathemat-
ics 227, Springer, 2004.
[134] B. Mohar and C. Thomassen, Graphs on Surfaces, Johns Hopkins University Press, Baltimore,
Maryland, 1997.
[135] M. Morales and A. Thoma, Complete intersection lattice ideals, J. Algebra 284 (2005), 755–770.
[136] S. Morey, Stability of associated primes and equality of ordinary and symbolic powers of ideals,
Comm. Algebra 27 (1999), 3221–3231.
[137] S. Morey, Depths of powers of the edge ideal of a tree, Comm. Algebra, to appear.
[138] S. Morey, E. Reyes and R. H. Villarreal, Cohen-Macaulay, shellable and unmixed clutters with
a perfect matching of König type, J. Pure Appl. Algebra 212(7)(2008), 1770-1786.
[139] J. R. Munkres, Topological results in combinatorics, Michigan Math. J. 31 (1984), no. 1,
113–128.
[140] M. Newman, Integral Matrices, Pure and Applied Mathematics 45 , Academic Press, New
York, 1972.
[141] J. Oxley, Matroid Theory, Oxford University Press, Oxford, 1992.
[142] M. D. Plummer, Some covering concepts in graphs, J. Combinatorial Theory 8 (1970), 91–98.
 BIBLIOGRAPHY 309

[143] C. Polini, B. Ulrich and W. V. Vasconcelos, Normalization of ideals and Briançon-Skoda

numbers, Math. Research Letters 12 (2005), 827–842.
[144] G. Ravindra, Well-covered graphs, J. Combinatorics Information Syst. Sci. 2 (1977), no. 1,
20–21.
[145] G. Reisner, Cohen-Macaulay quotients of polynomial rings, Adv. in Math. 21 (1976), 31–49.
[146] E. Reyes, Toric ideals and affine varieties, blowup algebras and combinatorial optimization
problems, PhD thesis, Cinvestav–IPN, 2006.
[147] R. T. Rockafellar, The elementary vectors of a subspace of RN , in Combinatorial Mathematics
and its Applications, Proc. Chapel Hill Conf., Univ. North Carolina Press, 1969, pp. 104–127.
[148] R. T. Rockafellar, Convex Analysis, Princeton University Press, 1970.
[149] A. Schrijver, On total dual integrality, Linear Algebra Appl. 38 (1981), 27–32.
[150] A. Schrijver, Theory of Linear and Integer Programming, John Wiley & Sons, New York, 1986.
[151] A. Schrijver,Combinatorial Optimization, Algorithms and Combinatorics 24, Springer-Verlag,
Berlin, 2003.
[152] P. D. Seymour, The matroids with the max-flow min-cut property, J. Combinatorial Theory Ser.
B 23 (1977), 189-222.
[153] P. D. Seymour, Decomposition of regular matroids, J. Combin. Theory Ser. B 28 (1980), no. 3,
305–359.
[154] A. Simis, On the Jacobian module associated to a graph, Proc. Amer. Math. Soc. 126 (1998),
989-997.
[155] A. Simis and W. V. Vasconcelos, Krull dimension and integrality of symmetric algebras,
Manuscripta Math. 61 (1988), 63–75.
[156] A. Simis, W. V. Vasconcelos and R. H. Villarreal, On the ideal theory of graphs, J. Algebra, 167
(1994), 389–416.
[157] A. Simis, W. V. Vasconcelos and R. H. Villarreal, The integral closure of subrings associated to
graphs, J. Algebra 199 (1998), 281–289.
[158] A. Simis and R. H. Villarreal, Constraints for the normality of monomial subrings and bira-
tionality, Proc. Amer. Math. Soc. 131 (2003), 2043–2048.
[159] D. E. Smith, On the Cohen-Macaulay property in commutative algebra and simplicial topology,
Pacific J. Math. 141 (1990), 165–196.
[160] A. Soleyman Jahan and X. Zheng, Pretty clean monomial ideals and linear quotients. (2007)
Preprint. arXiv:0707.2914v1
[161] E. Sperner, Ein satz über untermengen einer endlichen menge, Math. Z 27 (1928), 544–548.
[162] R. Stanley, The upper bound conjecture and Cohen-Macaulay rings, Stud. Appl. Math. 54
(1975), 135–142.
[163] R. Stanley, Hilbert functions of graded algebras, Adv. Math. 28 (1978), 57–83.
[164] R. Stanley, Balanced Cohen-Macaulay complexes, Trans. Amer. Math. Soc. 249 (1979), 139–
157.
[165] R. Stanley, Enumerative Combinatorics I, Wadsworth-Brooks/Cole, Monterey, California,
1986.
[166] R. Stanley, Combinatorics and Commutative Algebra, Birkhäuser Boston, 2nd ed., 1996.
[167] I. Stewart and D. Tall, Algebraic Number Theory, Chapman and Hall Mathematics Series, 1979.
[168] B. Sturmfels, Gröbner bases and Stanley decompositions of determinantal rings, Math. Z. 205
(1990), 137–144.
[169] B. Sturmfels, Gröbner bases of toric varieties, Tôhoku Math. J. 43 (1991), 249–261.
[170] B. Sturmfels, Gröbner Bases and Convex Polytopes, University Lecture Series 8, American
Mathematical Society, Rhode Island, 1996.
[171] S. Sullivant, Combinatorial symbolic powers, J. Algebra 319(1) (2008), 115-142.
310 BIBLIOGRAPHY

[172] B. Toft, Colouring, stable sets and perfect graphs, in Handbook of Combinatorics I (R. L. Graham
et. al., Eds.), Elsevier, 1995, pp. 233–288.
[173] K. Truemper, Matroid Decomposition, Academic Press, 1992.
[174] C. Valencia, Studies on graded rings associated to graphs, PhD thesis, Cinvestav–IPN, 2003.
[175] C. Valencia and R. H. Villarreal, Canonical modules of certain edge subrings, European J.
Combin. 24(5) (2003), 471–487.
[176] C. Valencia and R. H. Villarreal, Explicit representations of the edge cone of a graph, Int.
Journal of Contemp. Math. Sciences, Vol. 1, 2006, no. 1–4, 53–66.
[177] A. Van Tuyl, Sequentially Cohen-Macaulay bipartite graphs: vertex decomposability and reg-
ularity, preprint.
[178] A. Van Tuyl and R. H. Villarreal, Shellable graphs and sequentially Cohen-Macaulay bipartite
graphs, J. Combin. Theory Ser. A 115 (2008), no.5, 799-814.
[179] W. V. Vasconcelos, Arithmetic of Blowup Algebras, London Math. Soc., Lecture Note Series
195, Cambridge University Press, Cambridge, 1994.
[180] W. V. Vasconcelos, Computational Methods in Commutative Algebra and Algebraic Geometry,
Springer-Verlag, 1998.
[181] W. V. Vasconcelos, Integral Closure, Springer Monographs in Mathematics, Springer, New
York, 2005.
[182] R. H. Villarreal, Rees algebras and Koszul homology, J. Algebra 119 (1988), 83–104.
[183] R. H. Villarreal, Cohen-Macaulay graphs, Manuscripta Math. 66 (1990), 277–293.
[184] R. H. Villarreal, Rees algebras of edge ideals, Comm. Algebra 23 (1995), 3513–3524.
[185] R. H. Villarreal, Normality of subrings generated by square free monomials, J. Pure & Applied
Algebra 113 (1996), 91–106.
[186] R. H. Villarreal, On the equations of the edge cone of a graph and some applications, Manuscripta
Math. 97 (1998), 309–317.
[187] R. H. Villarreal, Monomial Algebras, Monographs and Textbooks in Pure and Applied Mathe-
matics 238, Marcel Dekker, Inc., New York, 2001.
[188] R. H. Villarreal, Monomial Algebras and Polyhedral Geometry, in Handbook of Algebra, Vol.
3 (M. Hazewinkel Ed.), Elsevier, Amsterdam, 2003, pp. 257–314.
[189] R. H. Villarreal, Normality of semigroups with some links to graph theory, Discrete Math. 302
(2005), 267-284.
[190] R. H. Villarreal, Unmixed bipartite graphs, Rev. Colombiana Mat. 41(2) (2007), 393–395.
[191] R. H. Villarreal, Normalization of monomial ideals and Hilbert functions, Proc. Amer. Math.
Soc. 136 (2008), 1933-1943.
[192] R. H. Villarreal, Rees cones and monomial rings of matroids, Linear Algebra Appl. 428 (2008),
2933-2940.
[193] R. H. Villarreal, Rees algebras and polyhedral cones of ideals of vertex covers of perfect graphs,
J. Algebraic Combin. 27 (2008), no. 3, 293-305.
[194] R. Webster, Convexity, Oxford University Press, Oxford, 1994.
[195] D. J. A. Welsh, Matroids: Fundamental Concepts, in Handbook of Combinatorics, Vol. 1, 2,
Elsevier, Amsterdam, 1995, pp. 481–526.
[196] N. L. White, The basis monomial ring of a matroid, Adv. in Math. 24 (1977), 292–297.
[197] N. L. White, A unique exchange property for bases, Linear Algebra Appl. 31 (1980), 81–91.
[198] S. Wolfram, Mathematica. A System for Doing Mathematics by Computer, Addison Wesley,
1989.
[199] R. Woodroofe, Vertex decomposable graphs and obstructions to shellability, Proc. Amer. Math.
Soc. 137 (2009), 3235-3246.
 BIBLIOGRAPHY 311

[200] G. M. Ziegler, Lectures on Polytopes, Graduate Texts in Mathematics 152, Springer-Verlag,

1994.
Index of Notation

α0 (G), vertex covering number, 46 PF , toric ideal, 145

α1 (G), edge covering number, 48 Q(A), covering polyhedron, 166
(I : J), ideal quotient, 11 R/I(G), edge-ring, 49
(I : x), ideal quotient, 11 R[It], Rees algebra, 2
(N1 : N2 ), ideal quotient, 11 R(r) , rth Veronese subring, 148
A(𝒫 ), Ehrhart ring, 2, 154 R s (I), symbolic Rees algebra, 166
At , transpose of A, 95 S ( f, g), the S -polynomial, 141
Ao , interior of A, 94 S −1 (M), localization, 10
B[F], subring generated by F, 30 T (M), torsion subgroup, 110
C(G), cone of a graph G, 245 V(G), VG , vertex set, 41
C ∗ , dual cone, 98 Ass(I), associated primes, 11
F(M, t), Hilbert series, 34 AssR (M), associated primes, 11
G ∖ {v}, deletion of a vertex, 44 ∆q , q-skeleton of ∆, 27
G1 ∗ G2 , join of two graphs, 296 ∆r (A), g.c.d. of i × i minors, 127
H(y, c), hyperplane, 94 ∆I , Stanley-Reisner complex, 23
H + (y, c), halfspace, 94 Spec(R), spectrum of a ring, 10
HA , supporting hyperplane, 256 Supp(M), support of M, 11
I(G), edge ideal, 49 ϒ(G), minimal covers, 85
I ∨ , Alexander dual of I, 65 ∥ x∥, Euclidean norm, 94
I∆ , Stanley-Reisner ideal, 23 α+ , positive part of a vector, 134
Ic , ideal of covers, 65 α− , negative part of a vector, 134
K, a field, 139 ht (I), height of an ideal, 24
K[F], monomial subring, 144 ann (M), annihilator of M, 11
K[x1 , . . . , xn ], polynomial ring, 1 β′0 (G), 87
K ∗ = K ∖ {0}, 140 β0 (G), vertex indep. number, 46
M(−a), shift, 35 β1 (G), edge indep. number, 46
M𝔭 , localization at a prime, 10 χ(G), chromatic number, 43
NG (A), neighbor set, 45 `(L), length of a module, 33
NG (v), neighbor set of v, 66 gcd, greatest common divisor, 149
P, polytope, 94, 132 ⟨S ⟩, induced subgraph, 42, 256
312
 INDEX OF NOTATION 313

⟨ , ⟩, usual inner product, 94 𝒜′ = {(vi , 1)}qi=1 ∪ {ei }ni=1 , 124

lcm, least common multiple, 141 𝒜 = {v1 , . . . , vq }, 124
⌊n⌋, integer part of n, 56 𝒢P , Graver basis, 146
lk(F), link of a face, 26 𝒦n , complete graph, 43
log(F), 144 𝒦m,n , complete bipartite, 44
log( f ), 144 ℳn , monomials of R, 139
ℕ = {0, 1, 2, . . . }, 124 𝒬6 , clutter, 172
ℕΓ, semigroup generated by Γ, 115 𝒵 (M), zero divisors, 12
ℕ𝒜′ , Rees semigroup of 𝒜, 124 Soc(M), socle of M, 31
ℝ+ v, 96 char(R), characteristic of R, 26
ℝ+ 𝒜′ , Rees cone, 124 conv(A), convex hull of A, 94
ℝ𝒜, linear subspace, 107 deg(v), degree of a vertex, 42
ℤ𝒜, subgroup generated by 𝒜, 124 depth(M), depth of a module, 25
1, the vector (1, . . . , 1), 118 dim(M), Krull dimension, 24
𝒫 , polytope, 94 frank(G), primitive cycles, 209
𝔪 = (x1 , . . . , xn ), max. ideal, 15 grI (R), associated graded ring, 166
ν(M), number of generators, 23 in(I), initial ideal, 140
ω(G), clique number, 43 in( f ), leading term of f , 140
I i , integral closure, 171 mgrade(I), monomial grade, 179
ℕ𝒜, normalization, 124 nil(R), nilradical, 11
∂(A), relative boundary, 94 rank(G), cycle rank, 208
rad (I), radical of I, 11 rb(A), relative boundary, 94
∣ x∣, diamond norm, 103 ri(A) relative interior, 94
c0 (G), number bip. comp., 255 star(F), star of a face, 32
e′ (G), 85 supp(α), support of a vector, 134
e(M), multiplicity of M, 34 supp(xa ), support of xa , 16
ei , ith unit vector, 108 trdeg, transcendence degree, 147
f (∆), the f -vector of ∆, 36 type(M), type of a module, 31
f →F g, reduction w.r.t F, 140 vol(P), relative volume, 108
k or K, a field, 8 H̃q (∆; k), reduced homology, 20
k, a field, 14
C-M, Cohen-Macaulay, 51
k[G], edge subring, 205
k[∆], Stanley-Reisner ring, 23
k[∆G ], face ring, 58
k[x], polynomial ring, 14
x ⋅ y, inner product of x and y, 94
xa , monomial, 144
Cn , cycle, 43
χ(∆), Euler characteristic, 21
ℝ+ = { x ∈ ℝ∣ x ≥ 0}, 95
ℝ+ 𝒜, cone generated by 𝒜, 96
Index

a-invariant, 34 ascending chain condition, 10

acyclic associated
digraph, 70 graded ring, 166
graph, 43 matrix
oriented graph, 216 of a monomial subring, 146
adjacent, 42 prime
admissible grading, 149 of a module, 11
affine of an ideal, 11
K-algebra, 30, 164 atomic cycle, 225
combination, 93
B-graph, 52
hyperplane, 94
basis
map, 93
monomial ideal, 191
space, 93
monomial ring, 191
generated by a set, 93
of a matroid, 137
space generated by a set, 295 binary clutter, 181
affinely independent, 104 binomial, 145
Alexander dual ideal, 145
of a graph, 85 of a closed walk, 206
of an ideal, 79 bipartite graph, 43
algebra, 29 bipartition, 43
finite, 142 birational extension, 163
finitely generated, 30, 142 block, 43
of covers, 270 blocker
of finite type, 142 of a graph, 85
annihilator, 11 blocking polyhedron, 280
arc, 201 blowup algebras, 166
arcwise connected space, 201 bow tie, 233
arrows, 131 bridge, 43
Artinian Bruns, Winfried, 116
ring, 34 Buchberger’s algorithm, 141
314
 INDEX 315

canonical module, 150 oriented graph, 216

Carathéodory’s theorem, 103 composition series, 33
catenary ring, 50 computer algebra systems
cellular embedding, 224 PORTA, 100
characteristic cone
cone, 106 finitely generated, 96
of a ring, 26 generated by a set, 255
vector, 166 over a graph, 245
chord, 64, 207 pointed, 117
chordal graph, 64 Conforti and Cornuéjols, 180
is shellable, 74 connected component, 42
chromatic number, 43 constructible complex, 28
circuit, 135 convex
of a graph, 223 combination, 94
of a matroid, 137 cone, 95
of a toric ideal, 147 hull, 94
clique, 43 polyhedron, 95
number, 43 set, 94
closed surface, 201 cover complexity, 85
clutter, 47, 80, 165 covering number, 179
ideal, 166 edge, 48
matrix, 118 vertex, 46
CoCoA, 5, 196, 200, 286 cross-free family, 132
codimension cutpoint, 43, 47
of a module, 24 cutvertex, 43
of an ideal, 24 cycle, 43
Cohen-Macaulay basis, 208
graph, 51 rank, 208
ideal, 25 space, 208
module, 25 vector, 208
ring, 25
coloration, 48 Danilov-Stanley formula, 2, 150
complement of a graph, 63 decomposition theorem
complementary complex, 2, 58 for polyhedra, 99
complementary slackness, 106 degree
complete of a module, 34
bipartite graph, 44 of a vertex, 42
graph, 43 dependent set
ideal, 171 of a matroid, 137
intersection, 36, 210 depth
graph, 213 lemma, 29
316 INDEX

of a module, 25 integral vector, 135

diadic clutter, 177 vector, 134
Dickson’s lemma, 14 elimination order, 140
digraph, 47, 131 embedded
dimension graph, 202
of a set, 93 prime, 11
of a simplicial complex, 19 embedding of a graph, 202
directed epimorphism, 9
cut, 131 Euler
graph, 47, 215 characteristic, 21
discrete reduced, 21
graph, 43 circuit, 48
distance, 44 formula, 203
dual trail, 48
cone, 98 even cycle, 43
problem, 104 exact sequence, 9, 88
duality short, 9, 88
in Stanley-Reisner ideals, 65 exchange property
theorem bases, 137
for cones, 98
for linear programming, 104 f-vector
of a complex, 36
edge, 41 face, 19, 95, 201
cone, 2, 255 ideal, 14, 52
cover, 48 of an embedding, 225
covering number, 48 ring, 23
critical graph, 50 facets
generator, 206 of a simplicial complex, 24
ideal, 7 facial
of a clutter, 81 cycles, 203
of a graph, 49 Farkas Lemma, 96
of a hypergraph, 51 feasible solutions, 106
polytope, 241 finite
ring, 49 basis theorem, 99
subring, 2, 205, 241 length, 33
Ehrhart finitely generated
polynomial, 157, 195 semigroup, 115
reciprocity law, 158 flag complexes, 64
ring, 2, 154 foliation, 213
series, 158 forest, 43
elementary forms, 22
 INDEX 317

free vertex property, 82 of an ideal, 195

full semigroup, 123 polynomial, 34
fundamental theorem of a face ring, 36
of linear inequalities, 96 series, 34
of a face ring, 36
genus of a graph, 224 Hilbert-Serre theorem, 34
Gordan’s lemma, 116 Hochster
Gorenstein configuration, 245
ideal, 31 theorem, 150
ring, 31 Hoffman and Kruskal, 125
Gröbner basis, 140 homeomorphic graphs, 204
of toric ideals, 146 homogeneous
reduced, 140 element, 22
graded ideal, 22
algebra, 23 ring, 34
ideal, 22 subring, 151
map, 22 homomorphism
module, 22 integral, 143
graph, 41 of algebras, 30
2-connected, 43 of graphs, 42
connected, 42 of rings, 26
connected components, 42 hypergraph, 47
embedding, 202
invariant, 42 Ichim, Bogdan, 116
plane graph, 202 ideal, 8
Graver basis, 146 face, 52
finitely generated, 8
H-configuration, 245 of relations, 30
H-path, 203 of vertex covers, 65, 188
h-vector quotient, 11
of a standard algebra, 37 image, 9
Hamilton cycle, 294 improper face
Heger’s theorem, 128 of a polyhedral set, 95
height of an ideal, 24 incidence matrix, 218
Herzog, Jürgen, 62 of a clutter, 47, 166
Hilbert of a digraph, 131
basis, 116 is totally unimodular, 134
minimal, 117 of a simple graph, 209
basis theorem, 10 oriented graph, 215
coefficients, 199 Smith form, 221
function, 34 independence
318 INDEX

number vertex, 42
edge, 179 isomorphic graphs, 42
complex, 2 isomorphism, 9
of a clutter, 81
of a graph, 58 join
number of graphs, 296
edge, 46 of simplicial complexes, 31
smallest, 87
König
vertex, 46 property, 179
independent edges, 178 Theorem, 46
independent set kernel, 9
of edges, 44 Koch, Robert, 116
of vertices, 46, 51 Krull
induced subgraph, 42, 74, 256 dimension, 24
initial principal ideal theorem, 50
ideal, 140 Kuratowski’s theorem, 204
integral
closure, 148 lattice, 18
of a domain, 143 distributive, 18
of a monomial ideal, 171 of monomial ideals, 18
of a ring, 143 point, 107
of a semigroup, 124 polytope, 107
of monomial subrings, 148 simplex, 112
extension, 143 lattice polytope, 154
hull, 105 leading
polyhedron, 105 coefficient, 140
integrally closed monomial, 140
ideal, 171 term, 140
ring, 144 Lehman theorem, 179
semigroup, 124 length of a module, 33
invariant factors lex order, 140
of a matrix, 127 lexicographical order, 147
irreducible linear
cover, 270 programming, 104
of a graph, 272 quotients, 75
representation, 102 variety, 93
of a cone, 102 link of a face, 26
irredundant representation, 98 local ring, 22
irrelevant maximal ideal, 30 localization, 10
isolated at a prime, 10
prime, 11 loop, 42
 INDEX 319

marriage theorem, 46, 266 ring, 14

a generalization, 299 subring, 144
matching, 44 subring of a graph, 205
Mathematica, 5, 287, 293 monomials
matroid, 137 multiplicative semigroup, 144
algebraic, 150 monomorphism, 9
bases of, 137 multigraph, 42
basis monomial ideal, 191 multiplicity
basis monomial ring, 191 of a face ring, 38
rank of, 137 of a graded module, 34, 156
representable, 137 of a graph, 85
max-flow min-cut, 121, 176 of an edge-ring, 85
maximal of Ehrhart rings, 157
ideal, 10
independent cover, 52 Nakayama’s lemma
independent set, 46 general version, 22
MFMC, 121, 176 graded version, 22
min-max problem, 179 neighbor set, 267
minimal of a subset, 45
generating set, 116 of a vertex, 66
prime nilpotent element, 11
of a ring, 10 nilradical, 11, 175
of a module, 11 Noetherian
of edge ideal, 52 module, 9
vertex cover ring, 8
of a clutter, 81, 166 normal
of a graph, 45 ideal, 171
minimally non-normal, 191, 250 ring, 144
minimum genus embeddings, 225 semigroup, 124
minor Normaliz, 5, 109, 116, 117, 122, 164,
of a clutter, 176 168–170, 172, 196, 286
of a graph, 205 normalization, 148
of a matrix, 127 of a ring, 143
of an ideal, 81, 176 of a semigroup, 124
module, 8 of an edge subring, 235
of finite length, 33 of monomial subrings, 148
of fractions, 10 normalized
monomial, 14 degree, 148
grade, 179 grading, 206
ideal, 14 volume, 112, 157
order, 139 normally torsion free, 84
320 INDEX

objective function, 104 primal problem, 104

odd cycle condition, 240 primary
one-way cut-incidence matrix, 131 decomposition
optimal irredundant, 12
solution, 104 of a module, 12
value, 104 of an ideal, 12
order complex, 49 of monomial ideals, 16
oriented ideal, 12
cycle, 216 submodule, 12
graph, 47 prime
outerplanar graph, 204 avoidance
general version, 13
path, 42 ideal, 8
PCP, 212 primitive
perfect graph, 43 binomial, 146
perfect matching, 46 cycle, 207
Pick’s Formula, 159 cycle property, 212
pointed projective space, 38
cone, 117 proper face
polyhedron, 105 of a polyhedral set, 95
polarization, 62 pure
polygonal arc, 202 shellable complex, 71
polyhedral simplicial complex, 26
cone, 95
set, 95 quadrangulation, 203
face, 95 quasi-ideal, 178
facet, 95
polyhedron, 95 radical of an ideal, 11
integral, 105 rank of a matroid, 137
pointed, 105 rational
Polyprob, 109 hyperplane, 94
polytopal subring, 154 polyhedron, 95
polytope, 94 reciprocity law, 158
lattice, 154 reduced
PORTA, 5, 100, 285–287, 289–291 ring, 11
poset, 18 simplicial homology, 20
positively graded algebra, 30 reduction of a polynomial, 140
PP, 179 Rees
presentation algebra, 2, 166
ideal, 30, 145 cone, 118, 124, 167
of a graded algebra, 30 of a graph, 267, 270
 INDEX 321

of a matroid, 193 with respect to C, 282

quasi-ideal, 178 Simis cone, 168
semigroup, 124 Simis, Aron, 168
regions, 201 simple
regular arc, 201
element, 12 closed arc, 201
graph, 48 module, 33
sequence, 25 simplex
relative lattice, 112
boundary, 94 of a simplicial complex, 19
points, 94 oriented, 19
interior, 94 unimodular, 112
point, 94 simplicial complex, 19
volume, 108, 157 Cohen-Macaulay, 25
of a simplex, 112 constructible, 28
residue field, 31 sink, 70
revlex order, 140 skeleton, 87
ring, 7 of a complex, 27
extension, 30, 142 Smith normal form, 107, 128
graph, 211 socle
of a module, 31
S-polynomial, 141 source, 70
saturation, 142 spanning
semigroup subgraph, 42
affine, 124 tree
finitely generated, 115 existence, 217
ring, 145 spectrum of a ring, 10
sequentially Cohen-Macaulay square, 43
complex, 75 square-free monomial, 1, 15
graph, 77 stability number, 53
module, 78 stable set, 46
set covering polyhedron, 118, 166 standard
shellable algebra, 30
clutter, 81 grading, 22
complex, 28 Stanley-Reisner
nonpure, 32, 71 complex, 23
graph, 71 ideal, 23
shelling, 28 ring, 23
shift star, 44
in the graduation, 35 of a face, 32
polyhedron, 276 strongly normal semigroup, 124
322 INDEX

Sturmfels, Bernd, 248 number, 53

subdivision of a graph, 204 tree, 43, 295
subgraph, 42 spanning, 295
submodule, 9 triangle, 43
support triangulated graph, 64
of a module, 11 triangulation, 203
of a monomial, 16, 222 two-way cut-incidence matrix, 131
of a vector, 134, 256 type of a module, 31
supporting hyperplane, 95
suspension of a graph, 91, 249 unbounded face, 202
symbolic unicyclic graph, 232
Rees algebra, 166 unimodular
system of parameters covering
homogeneous, 37 weakly, 113
with support in 𝒜, 113
t-unimodular, 128 matrix, 128, 221
TDI, 120, 130 simplex, 112
blowup algebras, 176 universal Gröbner basis, 146, 147
for graphs, 180 unmixed
iff MFMC, 121 clutter, 188
implies integral, 130 graph, 51
term order, 139 ideal, 25
terminal cycle, 232 Vasconcelos, Wolmer, 175
terms, 139 vector matroid, 137
toric of a graph, 215
ideal, 145 of an oriented graph, 215
of an edge subring, 206 Veronese subring, 148
of an oriented graph, 215 square-free, 151
torsion vertex, 100
free, 110 cover, 45
subgroup, 110 minimal, 45
totally covering number, 46, 179
dual integral, 120, 130 critical
unimodular, 125 clutter, 186
tournament, 47 critical graph, 50
transcendence vertex decomposable, 75, 80
basis, 147 volume
degree, 147 of a polytope, 108
transitive digraph, 70 of a simplex, 112
transversal
matroid, 192 walk, 42
 INDEX 323

closed, 42
well covered graph, 67

Yoshino, Yuji, 175

zero divisor, 12