Springer Proceedings in Mathematics & Statistics

Alexander M. Kasprzyk
Benjamin Nill   Editors

Interactions
with Lattice
Polytopes
Magdeburg, Germany, September 2017
Springer Proceedings in Mathematics &
Statistics

Volume 386
This book series features volumes composed of selected contributions from
workshops and conferences in all areas of current research in mathematics and
statistics, including data science, operations research and optimization. In addition
to an overall evaluation of the interest, scientific quality, and timeliness of each
proposal at the hands of the publisher, individual contributions are all refereed to the
high quality standards of leading journals in the field. Thus, this series provides the
research community with well-edited, authoritative reports on developments in the
most exciting areas of mathematical and statistical research today.

More information about this series at https://link.springer.com/bookseries/10533

Alexander M. Kasprzyk · Benjamin Nill
Editors

Interactions with Lattice

Polytopes
Magdeburg, Germany, September 2017
Editors
Alexander M. Kasprzyk Benjamin Nill
School of Mathematical Sciences Fakultät für Mathematik
University of Nottingham Otto-von-Guericke-Universität Magdeburg
Nottingham, UK Magdeburg, Sachsen-Anhalt, Germany

ISSN 2194-1009 ISSN 2194-1017 (electronic)

Springer Proceedings in Mathematics & Statistics
ISBN 978-3-030-98326-0 ISBN 978-3-030-98327-7 (eBook)
https://doi.org/10.1007/978-3-030-98327-7

Mathematics Subject Classification: 14M25, 52B20, 05E10, 06A07, 14J33, 14J45, 14M15, 14M27,
14N15, 52B12, 52B40, 53D05

© Springer Nature Switzerland AG 2022

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

This Springer imprint is published by the registered company Springer Nature Switzerland AG
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
Preface

This volume contains original research and survey articles highlighting interdisci-
plinary connections between a diverse range of topics. The common points of interest
are lattice polytopes. Lattice polytopes are fundamental combinatorial objects—
convex polytopes whose vertices have integer coordinates—with many beautiful and
deep connections across modern mathematics. Topics considered include: algebraic
geometry, mirror symmetry, symplectic geometry, discrete geometry, the geometry
of numbers, and algebraic combinatorics.
The study of lattice polytopes continues to open up fertile and unforeseen inter-
actions. In order to enhance this exchange of ideas, the workshop Interactions with
Lattice Polytopes took place 14–16 September, 2017, at the Otto-von-Guericke-
Universität Magdeburg, Germany. There were 15 talks given by world-leading
experts from several different backgrounds, elaborating upon the theme of appli-
cations of lattice polytopes. Many of the presented results can be found in this
volume.
Contributions to this volume contain original as well as expository research arti-
cles that illustrate some of the varied topical approaches and settings where lattice
polytopes play an important role. This volume should be particularly beneficial to
researchers and graduate students interested in learning more about the multifaceted
use of lattice polytopes across a broad range of active research areas.
This book relies deeply on the enthusiasm and engagement of the diverse and
collegial lattice polytope community. We are extremely grateful to the contribu-
tors for their high-quality articles, and to the anonymous referees for their careful
work. We would like to express our gratitude to everyone involved for their
patience and assistance. We are also thankful for logistical and financial support
from: the Otto-von-Guericke-Universität Magdeburg; the Research Training Group
Mathematical Complexity Reduction, funded by the Deutsche Forschungsgemein-
schaft (DFG, German Research Foundation)—314838170, GRK 2297 MathCoRe;

v
vi Preface

and the Engineering and Physical Sciences Research Council (EPSRC) Fellowship
EP/N022513/1.

Nottingham, UK Alexander M. Kasprzyk

Magdeburg, Germany Benjamin Nill
October 2021
Contents

1 Difference Between Families of Weakly and Strongly Maximal

Integral Lattice-Free Polytopes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
Gennadiy Averkov
2 On the Fine Interior of Three-Dimensional Canonical Fano
Polytopes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
Victor Batyrev, Alexander Kasprzyk, and Karin Schaller
3 Lattice Distances in 3-Dimensional Quantum Jumps . . . . . . . . . . . . . . 49
Mónica Blanco
4 Flag Matroids: Algebra and Geometry . . . . . . . . . . . . . . . . . . . . . . . . . . 73
Amanda Cameron, Rodica Dinu, Mateusz Michałek,
and Tim Seynnaeve
5 Classification of Minimal Polygons with Specified Singularity
Content . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
Daniel Cavey and Edwin Kutas
6 On the Topology of Fano Smoothings . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
Tom Coates, Alessio Corti, and Genival da Silva Jr.
7 Computing Seshadri Constants on Smooth Toric Surfaces . . . . . . . . . 157
Sandra Di Rocco and Anders Lundman
8 The Characterisation Problem of Ehrhart Polynomials
of Lattice Polytopes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
Akihiro Higashitani
9 The Ring of Conditions for Horospherical Homogeneous
Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197
Johannes Hofscheier
10 Linear Recursions for Integer Point Transforms . . . . . . . . . . . . . . . . . 221
Katharina Jochemko

vii
viii Contents

11 Schubert Calculus on Newton–Okounkov Polytopes . . . . . . . . . . . . . . 233

Valentina Kiritchenko and Maria Padalko
12 An Eisenbud–Goto-Type Upper Bound for the Castelnuovo–
Mumford Regularity of Fake Weighted Projective Spaces . . . . . . . . . 251
Bach Le Tran
13 Toric Degenerations in Symplectic Geometry . . . . . . . . . . . . . . . . . . . . 263
Milena Pabiniak
14 On Deformations of Toric Fano Varieties . . . . . . . . . . . . . . . . . . . . . . . . 287
Andrea Petracci
15 Polygons of Finite Mutation Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 315
Thomas Prince
16 Orbit Spaces of Maximal Torus Actions on Oriented
Grassmannians of Planes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 335
Hendrik Süß
17 The Reflexive Dimension of (0, 1)-Polytopes . . . . . . . . . . . . . . . . . . . . . 351
Akiyoshi Tsuchiya
Contributors

Gennadiy Averkov Brandenburgische Technische Universität Cottbus-

Senftenberg, Fakultät 1, Cottbus, Germany
Victor Batyrev Mathematisches Institut, Universität Tübingen, Tübingen,
Germany
Mónica Blanco Department of Mathematics, Statistics and Computation, Univer-
sity of Cantabria, Santander, Cantabria, Spain
Amanda Cameron Max Planck Institute for Mathematics in the Sciences, Leipzig,
Germany;
Department of Mathematics and Computer Science, Eindhoven University of
Technology, Eindhoven, The Netherlands
Daniel Cavey School of Mathematical Sciences, University of Nottingham,
Nottingham, UK
Tom Coates Department of Mathematics, Imperial College London, London, UK
Alessio Corti Department of Mathematics, Imperial College London, London, UK
Rodica Dinu University of Konstanz, Fachbereich Mathematik und Statistik,
Konstanz, Germany
Akihiro Higashitani Department of Pure and Applied Mathematics, Graduate
School of Information Science and Technology, Osaka University, Osaka, Japan
Johannes Hofscheier School of Mathematical Sciences, University of Nottingham,
Nottingham, UK
Katharina Jochemko Department of Mathematics, KTH Royal Institute of Tech-
nology, Stockholm, Sweden
Alexander Kasprzyk School of Mathematical Sciences, University of Nottingham,
Nottingham, UK

ix
x Contributors

Valentina Kiritchenko Laboratory of Algebraic Geometry and Faculty of Mathe-

matics, National Research University Higher School of Economics, Moscow, Russia;
Institute for Information Transmission Problems, Moscow, Russia
Edwin Kutas Mathematics Institute, University of Warwick, Coventry, UK
Anders Lundman Department of Mathematics, KTH Royal Institute of Tech-
nology, Stockholm, Sweden
Mateusz Michałek University of Konstanz, Fachbereich Mathematik und Statistik,
Konstanz, Germany
Milena Pabiniak Mathematisches Institut, Universität zu Köln, Köln, Germany
Maria Padalko SimilarWeb, Tel Aviv, Israel
Andrea Petracci Dipartimento di Matematica, Università di Bologna, Bologna,
Italy
Thomas Prince Mathematical Institute, University of Oxford, Oxford, UK
Sandra Di Rocco Department of Mathematics, KTH Royal Institute of Tech-
nology, Stockholm, Sweden
Karin Schaller Mathematisches Institut, Freie Universität Berlin, Berlin, Germany
Tim Seynnaeve Mathematisches Institut, Universität Bern, Bern, Switzerland
Genival da Silva Jr. Department of Mathematics, Imperial College London,
London, UK
Hendrik Süß Friedrich-Schiller-Universität Jena, Institut für Mathematik, Jena,
Germany
Bach Le Tran School of Mathematics, University of Edinburgh, James Clerk
Maxwell Building, Peter Guthrie Tait Road, Edinburgh, UK
Akiyoshi Tsuchiya Graduate School of Mathematical Sciences, University of
Tokyo, Tokyo, Japan
Chapter 1
Difference Between Families of Weakly
and Strongly Maximal Integral
Lattice-Free Polytopes

Gennadiy Averkov

Abstract A d-dimensional closed convex set K in Rd is said to be lattice-free if the

interior of K is disjoint with Zd . We consider the following two families of lattice-free
polytopes: the family Ld of integral lattice-free polytopes in Rd that are not properly
contained in another integral lattice-free polytope and its subfamily Md consisting
of integral lattice-free polytopes in Rd which are not properly contained in another
lattice-free set. It is known that Md = Ld holds for d ≤ 3 and, for each d ≥ 4, Md is
a proper subfamily of Ld . We derive a super-exponential lower bound on the number
of polytopes in Ld \Md (with standard identification of integral polytopes up to affine
unimodular transformations).

Keywords Egyptian fraction · Hollow polytope · Lattice-free set · Lattice

polytope · Maximality

1.1 Introduction

By |X | we denote the cardinality of a finite set X . Let N be the set of all positive
integers and let d ∈ N be the dimension. Elements of Zd are called integral points
or integral vectors. We call a polyhedron P ⊆ Rd integral if P is the convex hull
of P ∩ Zd . Let Aff(Zd ) be the group of affine transformations A : Rd → Rd satis-
fying A(Zd ) = Zd . We call elements of Aff(Zd ) affine unimodular transformations.
For a family X of subsets of Rd , we consider the family of equivalence classes
  
X/ Aff(Zd ) := A(X ) : A ∈ Aff(Zd ) : X ∈ X

with respect to identification of the elements of X up to affine unimodular transforma-

tions. A subset K of Rd is called lattice-free if K is closed, convex, d-dimensional and

G. Averkov (B)
Brandenburgische Technische Universität Cottbus-Senftenberg, Fakultät 1, Platz der Deutschen
Einheit 1, 03046 Cottbus, Germany
e-mail: [email protected]

© Springer Nature Switzerland AG 2022 1

A. M. Kasprzyk and B. Nill (eds.), Interactions with Lattice Polytopes,
Springer Proceedings in Mathematics & Statistics 386,
https://doi.org/10.1007/978-3-030-98327-7_1
2 G. Averkov

the interior of K contains no points from Zd . A set K is called maximal lattice-free

if K is lattice-free and is not a proper subset of another lattice-free set.
Our objective is to study the relationship between the following two families of
integral lattice-free polytopes:
1. The family Ld of integral lattice-free polytopes P in Rd such that there exists
no integral lattice-free polytope properly containing P. We call elements of Ld
weakly maximal integral lattice-free polytopes.
2. The family Md of integral lattice-free polytopes P in Rd such that there exists
no lattice-free set properly containing P. We call the elements of Ld strongly
maximal integral lattice-free polytopes.
The family Ld has applications in mixed-integer optimization, algebra and alge-
braic geometry; see [1, 3, 4, 13], respectively. In [2, 11] it was shown that Ld is
finite up to affine unimodular transformations:
Theorem 1 ([2, Theorem 2.1], [11, Corollary 1.3]) Ld / Aff(Zd ) is finite.
Several groups of researchers are interested in enumeration of Ld , up to affine
unimodular transformations, in fixed dimensions. This requires understanding geo-
metric properties of Ld . Currently, no explicit description of Ld is available for
dimensions d ≥ 4 and, moreover, it is even extremely hard to decide if a given
polytope belongs to Ld . A brute-force algorithm based on volume bounds for Ld
(provided in [11]) would have doubly exponential running time in d. In con-
trast to Ld , its subfamily Md is easier to deal with. Lovász’s characterization
[9, Proposition 3.3] of maximal lattice-free sets leads to a straightforward geomet-
ric description of polytopes belonging to Md . This characterization can be used to
decide whether a given polytope is an element of Md in only exponential time in d.
Thus, while enumeration of Md in fixed dimensions is a hard task, too, enumeration
of Ld is even more challenging.
For a given dimension d, it is a priori not clear whether or not Md is a proper
subset of Ld . Recently, it has been shown that the inequality Md = Ld holds if and
only if d ≤ 3. The equality Md = Ld is rather obvious for d ∈ {1, 2}, as it is not
hard to enumerate Ld in these very small dimensions and to check that every element
of Ld belongs to Md . Starting from dimension three, the problem gets very difficult.
Results in [1, 2] establish the equality M3 = L3 and enumerate L3 , up to affine
unimodular transformations. As a complement, in [11, Theorem 1.4] it was shown
that for all d ≥ 4 there exists a polytope belonging to Ld but not to Md .
While Theorem 1.4 in [11] shows that Ld and Md are two different families, it
does not provide information on the number of polytopes in Ld that do not belong
to Md . Relying on a result of Konyagin [6], we will show that, asymptotically, the
gap between Ld and Md is very large.
For a1 , . . . , ad > 0, we introduce

1 1
κ(a) := κ(a1 , . . . , ad ) = + ··· + .
a1 ad
1 Difference Between Families of Weakly and Strongly … 3

Reciprocals of positive integers are sometimes called Egyptian fractions. Thus, if a ∈

Nd , then κ(a) is a sum of d Egyptian fractions. We consider the set
 
Ad := (a1 , . . . , ad ) ∈ Nd : a1 ≤ · · · ≤ ad , κ(a1 , . . . , ad ) = 1

of all different solutions of the Diophantine equation

κ(x1 , . . . , xd ) = 1

in the unknowns x1 , . . . , xd ∈ N. The set Ad represents possible ways to write 1 as

a sum of d Egyptian fractions. It is known that Ad is finite. Our main result allows
is a lower bound on the cardinality of (Ld \Md )/ Aff(Zd ):
   
Theorem 2 (Ld+5 \Md+5 )/ Aff(Zd+5 ) ≥ Ad .

The proof of Theorem 2 is constructive. This means that, for every a ∈ Ad , we

generate an element in Pa ∈ Ld+5 \Md+5 such that for two different elements a and b
of Ad , the respective polytopes Pa and Pb do not coincide up to affine unimodular
transformations. The proof
 of Theorem 2 is inspired by the construction in [11].
Using lower bounds on Ad  from [6], we obtain the following asymptotic estimate:
    
Corollary 3 ln ln Ld \Md / Aff(Zd ) =  lndd , as d → ∞.

Note 4 We view the elements of Rd as columns. By o we denote the zero vector

and by e1 , . . . , ed the standard basis of Rd . If x ∈ Rd and i ∈ {1, . . . , d}, then xi
denotes the i-th component of x. The relation a ≤ b for a, b ∈ Rd means ai ≤ bi for
every i ∈ {1, . . . , d}. The relations ≥, > and < on Rd are introduced analogously.
The abbreviations aff, conv, int and relint stand for the affine hull, convex hull,
interior and relative interior, respectively.

1.2 An Approach to Construction of Polytopes in Ld \Md

We will present a systematic approach to construction of polytopes in Ld \Md , but

first we discuss general maximal lattice-free sets.

Definition 5 Let P be a lattice-free polyhedron in Rd . We say that a facet F of P

is blocked if the relative interior of F contains an integral point.

Maximal lattice-free sets can be characterized as follows:

Proposition 6 ([9, Proposition 3.3]) Let K be a d-dimensional closed convex subset
of Rd . Then the following conditions are equivalent:
1. K is maximal lattice-free;
2. K is a lattice-free polyhedron such that every facet of K is blocked.
4 G. Averkov

It can happen that some facets of a maximal lattice-free polyhedron are more than
just blocked. We introduce a respective notion. Recall that the integer hull K I of a
compact convex set K in Rd is defined by

K I := conv(K ∩ Zd ).

Definition 7 Let P be a d-dimensional lattice-free polyhedron in Rd . A facet F

of P is called strongly blocked if FI is (d − 1)-dimensional and Zd ∩ relint FI = ∅.
The polyhedron P is called strongly blocked if all facets of P are strongly blocked.
The following proposition extracts the geometric principle behind the construction
from [11, Sect. 3]. (Note that arguments in [11, Sect. 3] use an algebraic language.)
Proposition 8 Let P be a strongly blocked lattice-free polytope in Rd . Then PI ∈
Ld . Furthermore, if P is not integral, then PI ∈
/ Md .
Proof In order to show PI ∈ Ld it suffices to verify that, for every z ∈ Zd such
that conv(PI ∪ {z}) is lattice-free, one necessarily has z ∈ PI . If z ∈
/ PI , then z ∈
/ P
and so, for some facet F of P, the point z and the polytope P lie on different sides
of the hyperplane aff F. Then ∅ = Zd ∩ relint FI ⊆ int(conv(P ∪ {z})), yielding a
contradiction to the choice of z. Thus, for every facet F of P, z and P lie on the
same side of aff F. It follows z ∈ P. Hence z ∈ P ∩ Zd ⊆ PI .
If P is not integral, then PI ∈
/ Md since PI  P and P is lattice-free. 

1.3 Lattice-Free Axis-Aligned Simplices

For a ∈ Rd>0 , the d-dimensional simplex

T (a) := conv{o, a1 e1 , . . . , ad ed }.

is called axis-aligned. The proof of the following proposition is straightforward.

Proposition 9 For a ∈ Rd>0 , the following statements hold:
1. the simplex T (a) is a lattice-free set if and only if κ(a) ≥ 1;
2. the simplex T (a) is a maximal lattice-free set if and only if κ(a) = 1.
We introduce transformations which preserve the values of κ. The transformations
arise from the following trivial identities for t > 0:

1 1 1
= + , (1.1)
t t + 1 t (t + 1)
1 1 1 1
= + + , (1.2)
t t + 2 t (t + 2) t (t + 2)
1 2 1
= + . (1.3)
t 3t 3t
1 Difference Between Families of Weakly and Strongly … 5

Consider a vector a ∈ Rd>0 . By (1.1), if t is a component of a, we can replace this

component with two new components t + 1 and t (t + 1) to generate a vector b ∈
>0 satisfying κ(b) = κ(a). Identities (1.2) and (1.3) can be applied in a similar
Rd+1
fashion. For every d ∈ N, with the help of (1.1)–(1.3), we introduce the following
maps:
⎛ ⎞
a1
⎜ .. ⎟
⎜ . ⎟
⎜ ⎟
φd : Rd>0 → Rd+1
>0 , φd (a) := ⎜ ad−1 ⎟ , (1.4)
⎜ ⎟
⎝ ad + 1 ⎠
ad (ad + 1)
⎛ ⎞
a1
⎜ .. ⎟
⎜ . ⎟
⎜ ⎟
⎜ a ⎟
⎜ d−1 ⎟
ψd : Rd>0 → Rd+3
>0 , ψd (a) := ⎜ a + 3 ⎟,
⎜ d ⎟
⎜ ad (ad + 1) ⎟
⎜ ⎟
⎝(ad + 1)(ad + 3)⎠
(ad + 1)(ad + 3)
⎛ ⎞
a1
⎜ .. ⎟
⎜ . ⎟
⎜ ⎟
ξd : Rd>0 → Rd+1
>0 ξd (a) := ⎜ad−1 ⎟ . (1.5)
⎜3 ⎟
⎝ ad ⎠
2
3ad

The map φd replaces the component ad by two other components based on (1.1),
while ξd replaces ad based on (1.3). The map ψd acts by replacing the component ad
based on (1.1) and then replacing the component ad + 1 based on (1.2). Identities
(1.1)–(1.3) imply

κ(φd (a)) = κ(ψd (a)) = κ(ξd (a))) = κ(a). (1.6)

Lemma 10 Let P = T (ξd (a)), where a ∈ Ad and d ≥ 2. Then P is a strongly

blocked lattice-free (d + 1)-dimensional polytope. Furthermore, if ad is odd, P is
not integral.

Proof In this proof, we use the all-ones vector

⎛ ⎞
1
⎜ ⎟
1d := ⎝ ... ⎠ ∈ Rd .
1
6 G. Averkov

dFor 1the sake

d of1 brevity we introduce the notation t := ad . One has 1 = κ(a) =
i=1 ai ≥ i=1 t = d
t
, which implies t ≥ d ≥ 2. By (1.6), one has κ(ξd (a)) = 1
and so, by Proposition 9, P is maximal lattice-free.
If t is even, the polytope P is integral and hence every facet of P is integral, too. In
view of Proposition 6, integral maximal lattice-free polytopes are strongly blocked,
and so we conclude that P is strongly blocked.
Assume that t is odd, then the polytope P has one non-integral vertex. In this case,
we need to look at facets of P more closely, to verify that P is strongly blocked. We
consider all facets of P.
 
1. The facet F = conv o, a1 e1 , . . . , ad−1 ed−1 , 3ted+1 is a d-dimensional integral
integral axis-aligned simplex. Since

κ(a1 , . . . , ad−1 , 3t) < 1,

the integral point e1 + · · · + ed−1 + ed+1 is in the relative interior of F. Hence, F

is strongly blocked.  
2. The facet F = conv o, a1 e1 , . . . , ad−1 ed−1 , 23 ted contains the d-dimensional
integral axis-aligned simplex
 3t − 1 
G := conv o, a1 e1 , . . . , ad−1 ed−1 , ed ,
2
as a subset. In view of t ≥ 2, we have
 3t − 1 
κ a1 , . . . , ad−1 , < 1,
2
which implies that the integral point e1 + · · · + ed is in the relative interior of G.
It follows that F is strongly
 blocked. 
3. The facet F := conv a1 e1 , . . . , ad−1 ed−1 , 23 ted , 3ted+1 contains the integral d-
dimensional simplex
 3t − 1 
G := conv a1 e1 , . . . , ad−1 ed−1 , ed + ed+1 , 3ted+1 .
2
as a subset. It turns out that 1d+1 is the relative interior of G, because 1d+1 is a
convex combination of the vertices of relint G, with positive coefficients. Indeed,
the equality


d−1
1   3t − 1   
1d+1 = ai ei + λ ed + ed+1 + μ 3ted+1
a
i=1 i
2

holds for λ = 2
3t−1
and μ = t−1
t (3t−1)
, where
1 Difference Between Families of Weakly and Strongly … 7


d−1
1
+ λ + μ = 1.
a
i=1 i

4. It remains to consider faces F with the vertex set

 
3
o, a1 e1 , . . . , ad ed , ted , 3ted+1 \{ai ei },
2

where i ∈ {1, . . . , d + 1}. Without loss of generality, let i = 1 so that

 
3
F = conv o, a2 e2 , . . . , ted , 3ted+1 .
2

This facet contains the integral d-dimensional simplex

 3t − 1 
G := conv o, a2 a2 , . . . , ad−1 ed−1 , ed + ed+1 , 3ted+1 .
2
Similarly to the previous case, one can check that e2 + · · · + ed+1 is an integral
point in the relative interior of G. Consequently, F is strongly blocked. 

1.4 Proof of the Main Result

For d ≥ 4, Nill and Ziegler [7] construct one vector a ∈ Rd>0 with T (a) I ∈ Ld \Md .
We generalize this construction and provide many further vectors a with the above
properties. We will also need to verify that for different choices of a, we get essentially
different polytopes T (a) I .
Lemma 11 Let P and Q be d-dimensional strongly blocked lattice-free polytopes
such that for their integral hulls the equality Q I = A(PI ) holds for some A ∈
Aff(Zd ). Then Q = A(P).
Proof Since A is an affine transformation, we have

A(PI ) = A(conv(P ∩ Zd )) = conv A(P ∩ Zd ).

Using A ∈ Aff(Zd ), it is straightforward to check the equality A(P ∩ Zd ) = A(P) ∩

Zd . We thus conclude that A(PI ) = A(P) I . The assumption Q I = A(PI ) yields Q I =
A(P) I . Since P is strongly blocked lattice-free, A(P) too is strongly blocked lattice-
free. We thus have the equality Q I = A(P) I for strongly blocked lattice-free poly-
topes Q and A(P). To verify the assertion, it suffices to show that a strongly blocked
lattice-free polytope Q is uniquely determined by the knowledge of its integer
hull Q I . This is quite easy to see. For every strongly blocked facet G of Q I , the
affine hull of G contains a facet of Q. Conversely, if F is an arbitrary facet of Q,
8 G. Averkov

then G = FI is a strongly blocked facet of Q I . Thus, the knowledge of Q I allows

to determine affine hulls of all facets of Q. In other words, Q I uniquely determines
a hyperplane description of Q. 

Lemma 12 Let a, b ∈ Rd>0 be such that the equality T (b) = A(T (a)) holds for
some A ∈ Aff(Zd ). Then a and b coincide up to permutation of components.

Proof We use induction on d. For d = 1, the assertion is trivial. Let d ≥ 2. One

of the d facets of T (a) containing o is mapped by A to a facet of T (b) that con-
tains o. Without loss of generality we can assume that the facet T (a1 , . . . , ad−1 ) × {0}
of T (a) is mapped to the facet T (b1 , . . . , bd−1 ) × {0} of T (b). By the inductive
assumption, (a1 , . . . , ad−1 ) and (b1 , . . . , bd−1 ) coincide up to permutation of com-
ponents. Since unimodular transformations
d preserve the volume, T (a) and T (b) have
 d
the same volume. This means, i=1 ai = i=1 bi . Consequently, ad = bd and we
conclude that a and b coincide up to permutation of components.

Proof (Proof of Theorem 2) For every a ∈ Ad , we introduce the (d + 5)-dimensional

integral lattice-free polytope
Pa := T (η(a)) I ,

where
η(x) := ξd+4 (ψd+1 (φd (x)))

and the functions ξd+4 , ψd+1 and φd are defined by (1.4)–(1.5).

By (1.6) for each a ∈ Ad , we have κ(η(a)) = 1. For a ∈ Ad the last component
of φd (a) is even. This implies that the last component of ψd+1 (φd (a)) is odd. Thus,
by Lemma 10, T (η(a)) is strongly blocked lattice-free polytope which is not integral.
Let a, b ∈ Ad be such that the polytopes Pa and Pb coincide up to affine uni-
modular transformations. Then, by Lemma 11, T (η(a)) and T (η(b)) coincide up to
affine unimodular transformations. But then, by Lemma 12, η(a) and η(b) coincide
up to permutations. Since the components of a and b are sorted in the ascending
order, the components of η(a) and β(b) too are sorted in the ascending order. Thus,
we arrive at the equality η(a) = η(b), which implies a = b.
In view of Proposition 8, each Pa with a ∈ Ad belongs to Ld but not to Md . Thus,
the equivalence classes of the polytopes Pa with a ∈ Ad with respect to identification
up to affine unimodular transformations form a subset of (Ld+5 \Md+5 )/ Aff(Zd+5 )
of cardinality |Ad |. This yields the desired assertion. 

Proof (Proof of Corollary 3) The assertion is a direct consequence of Theorem 2

and the asymptotic estimate
 
d
ln ln |Ad | = 
ln d

of Konyagin [6, Theorem 1] (see also [5, Corollary 1.2]). 

1 Difference Between Families of Weakly and Strongly … 9

Remark 13 In view of the asymptotic upper bound ln ln |Ad | = O(d), determined

with different degrees of precision in [8, 10] and [12, Theorem 2], the lower bound
of Konyagin is optimal up to the logarithmic factor in the denominator.

Since all known elements of Ld are of the form PI , for some strongly blocked
lattice-free polytope P, we ask the following:
Question 14 Do there exist polytopes L ∈ Ld which cannot be represented as L =
PI for any strongly blocked lattice-free polytope P?
If there is a gap between the families Ld and the family
 
PI : P ⊆ Rd strongly blocked lattice-free polytope ,

then it would be interesting to understand how irregular the polytopes from this gap
can be. For example, one can ask the following:

Question 15 Do there exist polytopes L ∈ Ld with the property that no facet of L

is blocked?

Acknowledgements I would like to thank Christian Wagner for valuable comments and Christian
Elsholtz for pointing to [5, 8, 10].

References

1. Averkov, G., Krümpelmann, J., Weltge, S.: Notions of maximality for integral lattice-free
polyhedra: the case of dimension three. Math. Oper. Res. 42(4), 1035–1062 (2017)
2. Averkov, G., Wagner, C., Weismantel, R.: Maximal lattice-free polyhedra: finiteness and an
explicit description in dimension three. Math. Oper. Res. 36(4), 721–742 (2011)
3. Blanco, M., Haase, C., Hofmann, J., Santos, F.: The finiteness threshold width of lattice poly-
topes. Trans. Am. Math. Soc. Ser. B 8, 399–419 (2021)
4. Del Pia, A., Weismantel, R.: Relaxations of mixed integer sets from lattice-free polyhedra.
Ann. Oper. Res. 240(1), 95–117 (2016)
5. Elsholtz, C.: Egyptian fractions with odd denominators. Q. J. Math. 67(3), 425–430 (2016)
6. Konyagin, S.V.: Double exponential lower bound for the number of representations of unity
by Egyptian fractions. Math. Notes 95(1–2), 277–281 (2014). Translation of Mat. Zametki 95
(2014), no. 2, 312–316
7. Lagarias, J.C., Ziegler, G.M.: Bounds for lattice polytopes containing a fixed number of interior
points in a sublattice. Canad. J. Math. 43(5), 1022–1035 (1991)
8. Landau, E.: Über die Klassenzahl der binären quadratischen Formen von negativer Discrimi-
nante. Math. Ann. 56(4), 671–676 (1903)
9. Lovász, L.: Geometry of numbers and integer programming. In: Mathematical Programming,
Tokyo (1988). Mathematics Applied (Japanese Ser.), vol. 6, pp. 177–201. SCIPRESS, Tokyo
(1989)
10. Newman, M.: A bound for the number of conjugacy classes in a group. J. London Math. Soc.
43, 108–110 (1968)
11. Nill, B., Ziegler, G.M.: Projecting lattice polytopes without interior lattice points. Math. Oper.
Res. 36(3), 462–467 (2011)
10 G. Averkov

n 1
12. Sándor, C.: On the number of solutions of the Diophantine equation i=1 xi = 1. Period.
Math. Hungar. 47(1–2), 215–219 (2003)
13. Treutlein, J.: 3-dimensional lattice polytopes without interior lattice points. Ph.D. thesis, Eber-
hart Karls Universität Tübingen (2010)
Chapter 2
On the Fine Interior of
Three-Dimensional Canonical Fano
Polytopes

Victor Batyrev, Alexander Kasprzyk, and Karin Schaller

Abstract The Fine interior FI of a d-dimensional lattice polytope  is a ratio-

nal subpolytope of  which is important for constructing minimal birational mod-
els of non-degenerate hypersurfaces defined by Laurent polynomials with Newton
polytope . This paper presents some computational results on the Fine interior of
all 674,688 three-dimensional canonical Fano polytopes.

Keywords Lattice polytope · Fine interior · Hypersurface

2.1 Introduction

Let M ∼ = Zd be a free abelian group of rank d. We set MQ := M ⊗ Q and denote by N

the dual group Hom(M, Z) in the dual Q-linear vector space NQ := Hom(M, Q).
Let ·, · : MQ × NQ → Q be the natural pairing.
A convex compact d-dimensional polytope  ⊆ MQ is called lattice d-tope if all
vertices of  belong to the lattice M ⊆ MQ , i.e.,  equals the convex hull conv( ∩
M) of all lattice points in . The usual interior ◦ of  is the complement  \
∂, where ∂ is the boundary of . Another interior of a lattice polytope  was
introduced by Fine [3, 13, 15, 20]:
Definition 1 Let  ⊆ MQ be a lattice d-tope. Denote by ord the piecewise linear
function NQ → Q with

V. Batyrev
Mathematisches Institut, Universität Tübingen, Auf der Morgenstelle 10, 72076 Tübingen,
Germany
e-mail: [email protected]
A. Kasprzyk
School of Mathematical Sciences, University of Nottingham, Nottingham NG7 2RD, UK
e-mail: [email protected]
K. Schaller (B)
Mathematisches Institut, Freie Universität Berlin, Arnimallee 3, 14195 Berlin, Germany
e-mail: [email protected]

© Springer Nature Switzerland AG 2022 11

A. M. Kasprzyk and B. Nill (eds.), Interactions with Lattice Polytopes,
Springer Proceedings in Mathematics & Statistics 386,
https://doi.org/10.1007/978-3-030-98327-7_2
12 V. Batyrev et al.

ord (y) := minx, y (y ∈ NQ ).

x∈

Then the convex subset


FI := {x ∈ MQ | x, n ≥ ord (n) + 1}
n∈N \{0}

is called the Fine interior of .

One can show that only finitely many linear inequalities x, n ≥ ord (n) + 1
are necessary to define FI . Therefore, FI is a convex hull of finitely many rational
points p ∈ MQ . Moreover, any lattice point p ∈ ◦ ∩ M in the usual interior of 
is contained in FI . Therefore, FI contains the convex hull of  ∩ M, i.e., we get
the inclusion conv(◦ ∩ M) ⊆ FI . In particular, FI is non-empty if ◦ ∩ M is
non-empty. Moreover, for any lattice polytope  of dimension d ≤ 2 one has the
equality conv(◦ ∩ M) = FI [3]. The Fine interior FI of a lattice polytope  of
dimension d ≥ 3 may happen to be strictly larger than the convex hull conv(◦ ∩ M).
The simplest famous example of such a situation is due to M. Reid. Other similar
examples based on hollow 3-topes can be found in Sect. 2.7:

Example 2 ([20, Example 4.15]) Let M ⊆ Q4 be 3-dimensional affine lattice

defined by
 
4 
4

M := (m 1 , m 2 , m 3 , m 4 ) ∈ Z  4
m i = 5, im i ≡ 0 (mod 5) .
i=1 i=1

Consider the M-lattice 3-tope  ⊆ MQ defined as the convex hull of 4 lattice points

(5, 0, 0, 0), (0, 5, 0, 0), (0, 0, 5, 0), and (0, 0, 0, 5) ∈ M.

Then conv(◦ ∩ M) = ∅, but FI is the 3-dimensional M-rational simplex

conv{(2, 1, 1, 1), (1, 2, 1, 1), (1, 1, 2, 1), (1, 1, 1, 2)}

and FI ∩ M is empty.

In this paper, we are interested in lattice d-topes  ⊆ MQ obtained as Newton

polytopes of Laurent polynomials f  in d variables x1 , . . . , xd , i.e.,

f  (x) = am x m ,
m∈∩M

where am ∈ C are sufficiently general complex numbers. The importance of the Fine
interior is explained by the following theorem [3, 15, 20]:
2 On the Fine Interior of Three-Dimensional … 13

Theorem 3 Let Z ⊆ Td be a non-degenerate affine hypersurface in the

d-dimensional algebraic torus Td defined by a Laurent polynomial f  with New-
ton d-tope . Then the following conditions are equivalent:
1. a smooth projective compactification V of Z has non-negative Kodaira dimen-
sion, i.e., κ ≥ 0;
2. Z is birational to a minimal model S with abundance;
3. the Fine interior FI of  is non-empty.
Remark 4 By well known results of Khovanskii [14], one has vanishing of the
cohomology groups
h i (OV ) = 0 (1 ≤ i ≤ d − 2)

and the equation h d−1 (OV ) = |◦ ∩ M|. The numbers h i (OV ) are birational
invariants of Z ; they do not depend on a smooth projective compactification V
of Z . In particular, the number |◦ ∩ M| is the geometric genus pg of the affine
hypersurface Z ⊆ Td .
Smooth projective compactifications of non-degenerate hypersurfaces in Td can
be obtained using the theory of toric varieties [14].
Let  ⊆ MQ be a lattice d-tope. We consider the normal fan   of  in the dual
space NQ , i.e.,   := {σ θ | θ  }, where σ θ is the cone generated by all inward-
pointing facet normals of facets containing the face θ   of . One has dim(σ θ ) +
dim(θ ) = d for any face θ  . We denote by X  the normal projective toric variety
constructed via the normal fan   . In particular, the above function ord : NQ → Q
is a piecewise linear function with respect to this fan defining an ample Cartier divisor
on X  . In particular, the cones σ θ ∈   are defined as
  

σ θ = y ∈ NQ  ord (y) = x, y for all x ∈ θ .

Remark 5 Using the normal fan   , one can compute the fundamental group
π1 (V ) of a smooth projective birational model V of a non-degenerate affine hyper-
surface Z (given as in Theorem 3). The fundamental group π1 (V ) does not depend
on the choice of the smooth birational model and it is isomorphic to the quotient of
the lattice N modulo the sublattice N  generated by all lattice points in (d − 1)-
dimensional cones σ θ of the normal fan   [4].
Example 6 The minimal model S of a non-degenerate affine surface Z defined
by a Laurent polynomial with the Newton polytope  from Example 2 is a Godeaux
surface. It is a surface of general type with pg = q = 0, K 2 = 1, and π1 (S ) ∼
=
Z/5Z.
Definition 7 A lattice d-tope  is called canonical Fano d-tope if |◦ ∩ M| = 1.
Up to a shift by a lattice vector, we will assume without loss of generality that 0 ∈ M
is the single lattice point in the interior ◦ of the canonical Fano d-tope , i.e.,
◦ ∩ M = {0}.
14 V. Batyrev et al.

All canonical Fano 3-topes have been classified [16]. There exists exactly 674,688
canonical Fano 3-topes . The aim of this paper is to present computational results
of their Fine interiors FI and some related combinatorial invariants. These data
are important for computing minimal smooth projective surfaces S with pg = 1
and q = 0 which are birational to affine non-degenerate hypersurfaces Z ⊆ T3 ∼ =
(C× )3 .
The simplest description of the minimal surface S has been obtained when  is
a reflexive 3-tope [5].

Definition 8 A lattice d-tope  ⊆ MQ containing the origin 0 ∈ M in its interior is

called reflexive if the dual polytope

∗ := {y ∈ N | x, y ≥ −1 for all x ∈ } ⊆ NQ

is a lattice polytope.

There exist 4,319 reflexive 3-topes, classified by Kreuzer and Skarke [17]. They
form a small subset in the list of all 674,688 canonical Fano 3-topes [16]. Reflex-
ive 4-topes are also classified by Kreuzer and Skarke [18]. There exist 473,800,776
reflexive 4-topes, but the complete list of all canonical Fano 4-topes is unknown and
expected to be much bigger.
If  is a reflexive d-tope, then X  is a Gorenstein toric Fano d-fold and the Zariski
closure Z  in X  is a Gorenstein Calabi-Yau (d − 1)-fold. If d = 3, then Z  is a
K 3-surface with at worst finitely many Du Val singularities of type Ak . The mini-
mal surface S is a smooth K 3-surface which is obtained as the minimal (crepant)
desingularization of Z  [5].
One motivation for the present paper is due to Corti and Golyshev, who have
found 9 interesting examples of canonical Fano 3-simplices  such that the affine
surfaces Z are birational to elliptic surfaces of Kodaira dimension κ = 1 [11].
The computation of the Fine interior FI for all canonical Fano 3-topes  ⊆ MQ
has shown that the dimension of the Fine interior FI has only three values: 0, 1, and 3.
It is rather surprising that there are no canonical Fano 3-topes  with dim(FI ) = 2.
The condition dim(FI ) = 0 holds if and only if FI equals the lattice point 0 ∈
M. There exist exactly 665,599 canonical Fano 3-topes with FI = {0}, where 0 ∈
M is the only interior lattice point of . These polytopes are characterized in [3,
Proposition 3.4] by the condition that 0 ∈ N is an interior lattice point of the lattice 3-
tope
[∗ ] := conv(∗ ∩ N ).

Remark 9 If  is a canonical Fano 3-tope, then FI = {0} if and only if the non-
degenerate affine surface Z is birational to a K 3-surface [3, Theorem 2.26].

The case dim(FI ) = 1 splits in two subcases. There exists exactly 20 canonical
Fano 3-topes  such that 0 ∈ M is the midpoint of the Fine interior FI . Therefore,
we call this Fine interior symmetric. Canonical Fano 3-topes with 1-dimensional sym-
metric Fine interior are characterized by the condition that [∗ ] is a reflexive 2-tope.
2 On the Fine Interior of Three-Dimensional … 15

The Fine interior of the remaining 9,020 canonical Fano 3-topes with dim(FI ) = 1
contains 0 ∈ M as a vertex. Therefore, we call this Fine interior asymmetric. Canon-
ical Fano 3-topes with 1-dimensional asymmetric Fine interior are combinatorially
characterized by the condition that 0 ∈ N is contained in the relative interior of a
facet  [∗ ] of the lattice 3-tope [∗ ]. The minimal surfaces S corresponding
to canonical Fano 3-topes with 1-dimensional Fine interior (symmetric and asym-
metric) are elliptic surfaces of Kodaira dimension κ = 1.
There exist exactly 49 canonical Fano 3-topes with dim(FI ) = 3. These poly-
topes are characterized by the condition that 0 ∈ N is a vertex of the lattice 3-
tope [∗ ]. The surfaces S corresponding to canonical Fano 3-topes  with 3-
dimensional Fine interior FI are of general type (i.e., S has maximal Kodaira
dimension κ = dim(S ) = 2).
Remark 10 The Fine interior computations were done using
    

FI = x ∈ MQ  x, n ≥ ord (n) + 1 ,
θ n∈H(σ θ )

where H(σ θ ) denotes the set of all irreducible elements in the monoid σ θ ∩ N .
It is the minimal generating set of the monoid σ θ ∩ N and is called Hilbert basis
of σ θ ∩ N .
In the next sections we consider examples and discuss additional properties of
canonical Fano 3-topes  in dependence of their Fine interiors FI . All computations
were done using the Graded Ring Database [8], including the data of all 674,688
canonical Fano 3-topes, and Magma [7]. Therefore, all statements have been checked
by computer calculations. The canonical Fano 3-topes used as examples in this paper
appear with an ID that is the example’s ID in the Graded Ring Database.1

2.2 Almost Reflexive Polytopes of Dimension 3 and 4

Definition 11 A canonical Fano d-tope  ⊆ MQ is called almost reflexive if the

convex hull of all N -lattice points in the dual polytope ∗ is reflexive.
It is easy to show the following statement:
Proposition 12 If a canonical Fano d-tope  is almost reflexive, then

 F I = {0}.

Proof If [∗ ] is reflexive, then  = (∗ )∗ is contained in the dual reflexive poly-
tope [∗ ]∗ . Therefore, the Fine interior of  is contained in the Fine interior of the
reflexive polytope [∗ ]∗ and ([∗ ]∗ )FI = {0}. Thus, FI = {0}. 

1 http://www.grdb.co.uk/forms/toricf3c.
16 V. Batyrev et al.

The converse statement is not true in general for d ≥ 5, but there exist many
equivalent characterizations of reflexive and almost reflexive d-topes among canon-
ical Fano d-topes if d = 3 or d = 4.
Let us recall some combinatorial invariants of arbitrary lattice d-topes.

Definition 13 The Ehrhart power series of an arbitrary lattice d-tope  ⊆ MQ is

defined as 
P (t) := |k ∩ M| t k ,
k≥0

where |k ∩ M| denotes the number of lattice points in the k-th dilate k of .

This Ehrhart series is a rational function of the form

ψd ()t d + · · · + ψ1 ()t + ψ0 ()

P (t) = ,
(1 − t)d+1

where ψi () are non-negative integers for all 0 ≤ i ≤ d [22] such that ψ0 () = 1
d
and ψ1 () = | ∩ M| − d − 1. Moreover, i=0 ψi () = v(), where v() :=
d! · vol() denotes the normalized volume of .
One has the following characterization of reflexive d-topes:

Proposition 14 ([6, Theorem 4.6]) A canonical Fano d-tope  is reflexive if and

only if
ψi () = ψd−i () (0 ≤ i ≤ d).

The Ehrhart reciprocity implies that the power series


Q  (t) := |(k)◦ ∩ M| t k
k≥1

is a rational function

ϕd+1 ()t d+1 + · · · + ϕ2 ()t + ϕ1 ()t + ϕ0 ()

Q  (t) = ,
(1 − t)d+1

where ϕ0 () = 0 and ϕ1 () = |◦ ∩ M|. Using Serre duality, one obtains [12,
Sect. 4, 5.11]
ϕi () = ψd+1−i () (1 ≤ i ≤ d + 1),

i.e., in particular
ψd () = ϕ1 () = |◦ ∩ M|

and
ψd−1 () = ϕ2 () = |2◦ ∩ M| − (d + 1)|◦ ∩ M|.
2 On the Fine Interior of Three-Dimensional … 17

Therefore, the lattice d-tope  is a canonical Fano d-tope if and only if ψd () = 1.
Moreover,
ψd−1 () = |(2)◦ ∩ M| − (d + 1)

if  is a canonical Fano d-tope.

Applying the above equations, one immediately obtains the following criterion
for reflexivity of canonical Fano d-topes in the case d = 3, 4:
Proposition 15 Let  ⊆ MQ be a canonical Fano d-tope with d ∈ {3, 4}. Then
for d = 3, one has

t 3 + (|(2)◦ ∩ M| − 4)t 2 + (| ∩ M| − 4)t + 1

P (t) =
(1 − t)4

and for d = 4, one obtains

t 4 + (|(2)◦ ∩ M| − 5)t 3 + ψ2 ()t 2 + (| ∩ M| − 5)t + 1

P (t) = .
(1 − t)5

In particular,  is reflexive if and only if

| ∩ M| = |(2)◦ ∩ M| .

Proposition 16 Let  ⊆ MQ be a canonical Fano d-tope with d ∈ {3, 4} such

that 0 ∈ N is an interior lattice point of [∗ ]. Then [∗ ] is reflexive, i.e.,  is
almost reflexive.
Proof Let n ∈ N be an interior lattice point of [∗ ]. Then x, n ≥ 0 for all x ∈
 ∩ M because

∗ = {y ∈ NQ | x, y ≥ −1 for all x ∈ }

and x, n is an integer. Since 0 ∈ ◦ ∩ M, MQ is the set of all non-negative Q-linear

combinations of all lattice points in  ∩ M. This implies x  , n ≥ 0 for all x  ∈
MQ , i.e., n = 0. Therefore, [∗ ] has only one interior lattice point 0 ∈ N , i.e., [∗ ]
is a canonical Fano d-tope.
It is clear that [∗ ] is contained in the interior of 2[∗ ]. Therefore, we have [∗ ] ∩
N ⊆ (2[∗ ])◦ ∩ N . On the other hand, for any lattice point n ∈ (2[∗ ])◦ , x, n >
−2 for all x ∈  ∩ M. Since x, n is an integer, n ∈ ∗ ∩ N , i.e.,

[∗ ] ∩ N = (2[∗ ])◦ ∩ N .

Using Proposition 15, [∗ ] is reflexive. 

Corollary 17 Let  ⊆ MQ be a canonical Fano d-tope with d ∈ {3, 4} such that 0 ∈
N is an interior lattice point of [∗ ]. Then [∗ ]∗ is the smallest (referring to
inclusion) reflexive polytope containing .
18 V. Batyrev et al.

Proof Let  ⊆ MQ be a reflexive d-tope such that  ⊆  . Then ( )∗ ⊆ ∗ .

Since ( )∗ is a lattice polytope, it is contained in [∗ ]. Thus, [∗ ]∗ is contained
in (( )∗ )∗ =  . 

Remark 18 If  is a reflexive d-tope, then [2◦ ] = . If  is a canonical Fano d-

tope with d ∈ {3, 4} such that FI = {0} and  is contained in a reflexive d-tope  ,
then [2◦ ] is contained in [(2 )◦ ] =  . Therefore, [2◦ ] is contained in the small-
est reflexive polytope [∗ ]∗ containing , i.e.,

[2◦ ] ⊆ [∗ ]∗ .

Computations showed that among all 665,599 canonical Fano 3-topes  with FI =
{0} there exist exactly 211,941 canonical Fano 3-tops such that [2◦ ] is reflexive.
For the remaining canonical Fano 3-topes  the lattice 3-topes [2◦ ] are larger
than , but are not equal to the reflexive hull [∗ ]∗ .

Remark 19 Let  be an almost reflexive 3-tope. We denote by τ () the lattice d-

tope [2◦ ] . If τ () is not reflexive, then it is almost reflexive and we can consider
the larger lattice d-tope τ 2 () := τ (τ ()) ⊆ [∗ ]∗ . After at most five steps, τ k ()
is equal to the reflexive hull [∗ ]∗ of .

In dimension 4, the situation is comparable:

Example 20 Let  ⊆ R4 be the almost reflexive 4-tope defined by the inequalities

xi ≥ −1 (1 ≤ i ≤ 4), x1 ≤ 2, and x1 + x2 + x3 + x4 ≤ 1. Then  F I = {0} and the
smallest reflexive 4-tope containing  is the 4-simplex [∗ ]∗ defined by the inequal-
ities xi ≥ −1 (1 ≤ i ≤ 4) and x1 + x2 + x3 + x4 ≤ 1. It is easy to see that τ () is
not the reflexive 4-tope [∗ ]∗ because the vertex (4, −1, −1, −1) ∈ vert([∗ ]∗ ) is
not in 2◦ . However, τ 2 () = [∗ ]∗ .

2.3 Canonical Fano 3-Topes with FI = {0}

We note that the set of all reflexive 3-topes forms a rather small part of the set of all
canonical Fano 3-topes. The majority of the canonical Fano 3-topes belong to the
subset of almost reflexive 3-topes. The proof of the following statement is based on
the result of Skarke [21] and the explanations in the previous section.

Proposition 21 A canonical Fano 3-tope  is almost reflexive if one of the following

equivalent conditions is satisfied:
2 On the Fine Interior of Three-Dimensional … 19

1. FI = {0};
2. 0 ∈ N is an interior lattice point of [∗ ];
3.  is contained in some reflexive 3-tope;
4. τ k () is the reflexive 3-tope [∗ ]∗ for some sufficiently large k (1 ≤ k ≤ 5);
5. the lattice 3-tope [2◦ ] has exactly one interior lattice point;
6. the non-degenerate affine hypersurface Z defined by a Laurent polynomial with
Newton polytope  is birational to a smooth K 3-surface.

Computations show that there exist exactly 665,599 almost reflexive canonical
Fano 3-topes. The set of almost reflexive 3-topes includes all 4,319 reflexive 3-topes.
We have shown that for any almost reflexive 3-tope , the reflexive polytope ref :=
[∗ ]∗ is the smallest reflexive 3-tope containing . We call ref the reflexive hull
of . Thus we obtain a natural surjective map  → ref from the set of almost
reflexive 3-topes to the set of reflexive 3-topes, which is the identity on the set of
reflexive 3-topes. The minimal surface S is a K 3-surface if and only if  is an
almost reflexive 3-tope. If  is an almost reflexive 3-tope, but not reflexive, then the
minimal surface S is a crepant desingularization of the Zariski closure of Z in the
Gorenstein toric Fano threefold X ref defined by the reflexive hull of .
A generalization of the reflexive hull of almost reflexive 3-topes for arbitrary
lattice d-topes with non-empty Fine interior can be obtained using the notion of the
support of the Fine interior FI .

Definition 22 Let  ⊆ MQ be a lattice d-tope with FI = ∅. Then the set

supp(FI ) := {n ∈ N | there exists x ∈ FI with x, n = ord (n) + 1}

is called support of the Fine interior of .

Example 23 If  is a reflexive d-tope, then the support of the Fine interior of  is

the set of all non-zero lattice points in ∗ ∩ N .

Remark 24 It is easy to show that one always has


FI = {x ∈ MQ | x, n ≥ ord (n) + 1}.
n∈supp(FI )

Definition 25 Let  ⊆ MQ be a lattice d-tope with FI = ∅. Then the rational

polytope 
can := {x ∈ MQ | x, n ≥ ord (n)}
n∈supp(FI )

contains  and is called the canonical hull of .

Example 26 If  is an almost reflexive 3-tope, then supp(FI ) is the set (∗ ∩

N ) \ {0} of boundary lattice points in the reflexive 3-tope [∗ ] and the canonical
20 V. Batyrev et al.

v3

v3
v1

v1 v2

v2
v5

v4

v4
(a) ID 547386. (b) ID 547385.

Fig. 2.1 Canonical Fano 3-topes  with FI ={0}. Shaded faces are occluded and the Fine
interior {0} is shown in grey with a double border. The whole polytope is the canonical hull
can as well as the reflexive hull ref and the grey coloured polytope is . a Reflexive polytope
 = conv{v1 , v2 , v3 , v4 } with v1 = (1, 0, 0), v2 = (0, 1, 0), v3 = (0, 0, 1), v4 = (−1, −1, −1), and
ref = can = . All facets of  have lattice distance 1 to the origin. b Almost reflexive polytope
 = conv{v1 , v2 , v3 , v4 } with v1 = (1, 0, 0), v2 = (0, 1, 0), v3 = (0, 0, 1), v4 = (−1, −1, −2), and
ref = can = conv{v1 , v2 , v3 , v4 , v5 } with v5 = (0, 0, −1) reflexive. The dark grey coloured facet
of  has lattice distance 2 and all other facets have lattice distance 1 to the origin

hull can equals the reflexive hull ref of the polytope , i.e., can = ref = [∗ ]∗ .
In particular, in this case can is always a lattice 3-tope.
There exists a smooth projective toric variety X  defined by a fan  whose 1-
dimensional cones are generated by all lattice vectors from the finite set supp(FI ).
Then the minimal surface S is a K 3-surface which is the Zariski closure of Z
in X  [3].

Example 27 Let us consider the (almost) reflexive canonical Fano 3-tope  =

conv{v1 , v2 , v3 , v4 } ⊆ MQ (ID 547386, Fig. 2.1a) with vertices

v1 := (1, 0, 0), v2 := (0, 1, 0), v3 := (0, 0, 1), and v4 := (−1, −1, −1)

and FI = {0}. Moreover,

ref = conv(2◦ ∩ M) = conv( ∩ M) = 

and
can = [∗ ]∗ = (∗ )∗ = 

because  is reflexive, i.e., ref = can =  reflexive (Fig. 2.1a).

2 On the Fine Interior of Three-Dimensional … 21

Example 28 Consider the almost reflexive canonical Fano 3-tope  ⊆ MQ

(ID 547385, Fig. 2.1b) with vertices

v1 := (1, 0, 0), v2 := (0, 1, 0), v3 := (0, 0, 1), and v4 := (−1, −1, −2)

and FI = {0}. Moreover,

ref = conv(( ∩ M) ∪ {v5 }) = conv{v1 , v2 , v3 , v4 , v5 }

and
can = [∗ ]∗ = conv{v1 , v2 , v3 , v4 , v5 }

with v5 := (0, 0, −1) because  is almost reflexive, i.e., ref = can =  reflex-
ive (Fig. 2.1b).

2.4 Asymmetric Fine Interior of Dimension 1

There exist exactly 9,020 canonical Fano 3-topes  with 1-dimensional Fine interior
such that 0 ∈ N belongs to a facet  [∗ ] of the lattice 3-tope [∗ ]. This class of
canonical Fano 3-topes is characterized by the property that the lattice 3-tope [2◦ ]
has exactly 2 interior lattice points.
The corresponding minimal surfaces S are simply connected (i.e., have triv-
ial fundamental group π1 (S )) elliptic surfaces of Kodaira dimension κ = 1. We
observed that the facet  [∗ ] is a reflexive 2-tope corresponding to one of the
three types pictured in Fig. 2.2. All N -lattice points on the boundary of belong
to supp(FI ). It was checked that for all these 3-topes  the canonical hull can is
again a lattice 3-tope. Moreover, the Fine interior FI is contained in the ray gener-
ated by the primitive lattice vector v ∈ M which is the primitive inward-pointing
facet normal of , i.e., x, y = 0 for all x ∈ FI , y ∈ . The lattice point 0 ∈ M is
a vertex of FI . More precisely, one has

FI = conv{0, λv },

where λ ∈ {1/2, 2/3}. The primitive lattice vector v is the unique interior lattice
point on a reflexive facet θ+   of  of one of the three possible types pictured in
Fig. 2.2. These three reflexive polygons θ+ are characterized by the condition that the
dual reflexive polygons θ+∗ are obtained from θ+ (Fig. 2.3) by enlarging the lattice Z2
in the following ways: Z2 + Z(1/3, 2/3) (Fig. 2.3a), Z2 + Z(1/2, 0) (Fig. 2.3b), and
Z2 + Z(1/2, 1/2) (Fig. 2.3c). Moreover, the reflexive facet θ+ of  is isomorphic to
the facet of [∗ ]. The projection M → M/Zv of  or of θ+ along v is a reflexive
polygon of one of the three types pictured in Fig. 2.3, which is dual to θ+ and . The
lattice vector v defines a character of the 3-dimensional torus χ : T3 → C× . For
almost all α ∈ C× , the fiber χ −1 (α) is an affine elliptic curve defined by a Laurent
22 V. Batyrev et al.

(a) (b) (c)

Fig. 2.2 Reflexive Facets of  Containing ±v . Three types of reflexive facets θ±  
of  containing ±v for all 9,020 + 20 canonical Fano 3-topes  with dim(FI ) = 1. Ver-
tices are coloured black, boundary points that are not vertices grey, and the origin light grey.
a conv{(1, 0), (0, 1), (−1, −1)}. b conv{(1, 0), (−1, 1), (−1, −1)}. c conv{(±1, 0), (0, ±1)}

(a) (b) (c)

Fig. 2.3 Reflexive Projection Polytopes. Three types of reflexive polytopes obtained via a pro-
jection of  along ±v for all 9,020 + 20 canonical Fano 3-topes  with dim(FI ) = 1. Ver-
tices are coloured black, boundary points that are not vertices grey, and the origin light grey.
a conv{(−1, 2), (−1, −1), (2, −1)}. b conv{(−2, −1), (0, 1), (2, −1)}. c conv{(±1, ±1)}

polynomial with the reflexive Newton polytope ∗ ∼ = θ+∗ of one of the three types
pictured in Fig. 2.3 with the distribution shown in Table 2.1. So χ defines birationally
an elliptic fibration.

Table 2.1 Distribution of the Reflexive Facets of  Containing ±v . Table contains: Type of
the reflexive facet θ± containing ±v , type of the dual reflexive facet θ±∗ , the enlarged lattice used
to obtain θ±∗ from θ± , the number of canonical Fano 3-topes asym := { | 1-dim. asym. FI }, and
the number of canonical Fano 3-topes sym := { | 1-dim. sym. FI } with respect to the facet type
of θ± pictured in Fig. 2.2
θ± θ±∗ Enlarged lattice #asym #sym
Figure 2.2a Figure 2.3a Z2 + 3,038 7
Z(1/3, 2/3)
Figure 2.2b Figure 2.3b Z2 + Z(1/2, 0) 4,663 9
Figure 2.2c Figure 2.3c Z2 + 1,319 4
Z(1/2, 1/2)
2 On the Fine Interior of Three-Dimensional … 23

v1 v2

v3
v5

v1

v2
v5
v3 v4 v4

(a) ID 547324 (b) ID 547323

Fig. 2.4 Canonical Fano 3-topes with Asymmetric Fine Interior of Dimension 1 Shaded
faces are occluded. The Fine interior and the origin are shown in grey, with a double border
around the origin. The facet θ+ is grey dotted. a The whole polytope is  = conv{v1 , v2 , v3 , v4 }
with v1 = (2, 3, 8), v2 = (1, 0, 0), v3 = (0, 1, 0), and v4 = (−1, −1, −1). Moreover, FI =
conv{(0, 0, 0), (1/2, 1/2, 1)}, θ+ = conv{v1 , v2 , v3 }, and can = conv{v1 , v2 , v3 , v4 , v5 } with v5 =
(0, 1, 4). b The whole polytope is  = conv{v1 , v2 , v3 , v4 } with v1 = (−1, 1, −2), v2 =
(1, −2, 3), v3 = (1, 0, 0), and v4 = (−2, 5, −3). Moreover, FI = conv{(0, 0, 0), (0, 2/3, 0)}
and θ+ = conv{v2 , v3 , v4 }, and can = conv{v1 , v2 , v3 , v4 , v5 } with v5 = (−2, 4, −3)

Example 29 Let  ⊆ MQ be a canonical Fano 3-tope given as the convex hull of

v1 := (2, 3, 8), v2 := (1, 0, 0), v3 := (0, 1, 0), and v4 := (−1, −1, −1)

(ID 547324, Fig. 2.4a, Tables 2.2 and 2.4). Then

FI = conv{(0, 0, 0), (1/2, 1/2, 1)} = conv{0, 1/2 · v },

where v = (1, 1, 2). One has v1 + 2v2 + v3 = 4v . Therefore, v is the interior lat-
tice point of the reflexive facet θ+ of  with vertices v1 , v2 , v3 and the images v 1 , v 2 , v 3
of v1 , v2 , v3 in M/Zv are vertices of the dual reflexive triangle θ+∗ (Fig. 2.3b) satis-
fying the relation
v 1 + 2v 2 + v 3 = 0.

To compute the canonical hull can of , we obtain supp(FI ) = {si | 1 ≤ i ≤

18} with s1 := (−1, −1, 1), s2 := (−1, −1, 2), s3 := (−1, −1, 3), s4 := (−1, 0, 1),
s5 := (−1, 0, 2), s6 := (−1, 1, 0), s7 := (−1, 1, 1), s8 := (−1, 2, 0), s9 := (−1, 3,
−1), s10 := (0, −1, 1), . . ., s18 := (−2, −2, 1), which leads to

can = conv{v1 , v2 , v3 , v4 , v5 }

with v5 := (0, 1, 4) (Fig. 2.4a).

24 V. Batyrev et al.

Table 2.2 9 Canonical Fano 3-topes with Asymmetric Fine Interior of Dimension 1. Table
contains: vertices vert() of , vertices vert(FI ) of the Fine interior FI , unique primitive lattice
point v ∈ θ+ in the reflexive facet θ+  , and weights (wi )0≤i≤3 of the weighted projective 3-
space P(w0 , . . . , w3 ) appearing in [11]
ID vert() vert(FI ) v (wi )0≤i≤3
547324 (2, 3, 8), (1, 0, 0), (0, 1, 0), (−1, −1, −1) 0, 1/2 · v (1, 1, 2) (1, 5, 6, 8)
547323 (−1, 1, −2), (1, −2, 3), (1, 0, 0), (−2, 5, −3) 0, 2/3 · v (0, 1, 0) (1, 4, 7, 9)
547311 (−1, 4, 2), (−1, −1, 0), (0, 0, −1), (2, 0, 1) 0, 2/3 · v (0, 1, 1) (2, 5, 8, 9)
547490 (1, 2, 4), (1, 0, 0), (1, −2, 3), (−1, 1, −2) 0, 1/2 · v (0, 1, 0) (1, 5, 8, 14)
547321 (1, −2, 3), (0, 1, 0), (1, 0, 0), (−6, 3, −8) 0, 1/2 · v (−1, 1, −2) (3, 7, 8, 10)
547305 (0, 1, 0), (1, 0, 0), (1, 2, 4), (−4, −6, −7) 0, 2/3 · v (−1, −1, −1) (4, 7, 9, 10)
547526 (1, 0, 0), (0, 1, 0), (−2, 1, 5), (2, −4, −9) 0, 2/3 · v (1, −1, −3) (5, 9, 8, 11)
547454 (2, 1, 7), (1, 0, 0), (0, 1, 0), (−2, −3, −3) 0, 1/2 · v (0, 0, 1) (3, 7, 8, 18)
547446 (0, 1, 1), (−6, 7, −15), (1, −2, 3), (1, 0, 0) 0, 1/2 · v (−1, 1, −2) (5, 8, 9, 22)

Table 2.3 9 Canonical Fano 3-topes with Asymmetric Fine Interior of Dimension 1. Table
contains: primitive inward-pointing facet normals (n i )1≤i≤4 of , vertices vert(θ+ ) of the reflexive
facet θ+  , and primitive inward-pointing facet normal n θ+ of the reflexive facet θ+  
ID (n i )1≤i≤4 vert(θ+ ) n θ+
547324 (−2, −2, 1), (−1, −1, 3), (−1, 3, −1), (7, −3, −1) (2, 3, 8), (1, 0, 0), (0, 1, 0) (−2, −2, 1)
547323 (−3, −3, −2), (−1, 0, 1), (−1, 6, 4), (17, 3, −5) (1, −2, 3), (1, 0, 0), (−2, 5, −3) (−3, −3, −2)
547311 (−1, −1, 1), (−1, 2, 1), (1, 2, −5), (7, −2, 5) (−1, 4, 2), (−1, −1, 0), (2, 0, 1) (1, 2, −5)
547490 (−2, −2, 1), (−1, 0, 0), (−1, 6, 4), (23, 2, −8) (1, 2, 4), (1, 0, 0), (−1, 1, −2) (−2, −2, 1)
547321 (−3, −3, −2), (−2, −2, 1), (−1, 3, 2), (9, −5, −8) (0, 1, 0), (1, 0, 0), (−6, 3, −8) (−2, −2, 1)
547305 (−7, −7, 11), (−2, −2, 1), (−1, 2, −1), (7, −3, −1) (0, 1, 0), (1, 2, 4), (−4, −6, −7) (7, −3, −1)
547526 (−5, −5, −2), (−3, −3, 1), (−1, 2, −1), (25, −8, 10) (1, 0, 0), (0, 1, 0), (2, −4, −9) (−3, −3, 1)
547454 (−7, −7, 2), (−1, −1, 2), (−1, 1, 0), (7, −2, −2) (2, 1, 7), (0, 1, 0), (−2, −3, −3) (7, −2, −2)
547446 (−9, 21, 14), (−5, −3, −2), (−1, −1, 0), (9, 1, −3) (0, 1, 1), (−6, 7, −15), (1, −2, 3) (9, 1, −3)

Example 30 Let  ⊆ MQ be a canonical Fano 3-tope given as the convex hull of

v1 := (−1, 1, −2), v2 := (1, −2, 3), v3 := (1, 0, 0), and v4 := (−2, 5, −3)

(ID 547323, Fig. 2.4b, Tables 2.2 and 2.4). Then (Table 2.3)

FI = conv{(0, 0, 0), (0, 2/3, 0)} = conv{0, 2/3 · v },

where v = (0, 1, 0). One has v2 + v3 + v4 = 3v . Therefore, v is the interior lat-
tice point of the reflexive facet θ+ of  with vertices v2 , v3 , v4 and the images v 2 , v 3 , v 4
of v2 , v3 , v4 in M/Zv are vertices of the dual reflexive triangle θ+∗ (Fig. 2.3a) satis-
fying the relation
v 2 + v 3 + v4 = 0.
 Table 2.4 9 Canonical Fano 3-topes with Asymmetric Fine Interior of Dimension 1. Table contains: vertices vert( ) of the reflexive facet  [∗ ],
support supp(FI ) of the Fine interior FI , and vertices vert(can ) of the canonical hull can
ID vert( ) supp(FI ) vert(can )
547324 (−1, 3, −1), (−1, −1, 1), (1, −1, 0) (−2, −2, 1), (−1, −1, 1), S1 vert(), (0, 1, 4)
547323 (−1, 0, 1), (−1, 0, 0), (2, 0, −1) (−3, −3, −2), (−1, 0, 0), S2 vert(), (−2, 4, −3)
547311 (−1, −1, 1), (0, 1, −1), (1, 0, 0) (−1, −1, 1), (−1, 0, 1), S3 vert(), (−1, 2, 0)
547490 (−1, 0, 0), (−1, 0, 1), (3, 0, −1) (−2, −2, 1), (−1, 0, 0), S4 vert(), (1, −1, 4)
547321 (−1, −1, 0), (−1, 3, 2), (1, −1, −1) (−2, −2, 1), (−1, −1, 0), (−1, 0, 0), (−1, 1, 1), (−1, 3, 2), vert(), (1, 0, 1), (0, −3, 4)
(0, −1, −1), (1, −1, −1)
547305 (−1, 2, −1), (1, −1, 0), (0, −1, 1) (−1, −1, 1), (−1, 0, 0), (−1, 2, −1), (0, −1, 1), (1, −1, 0), vert(), (0, −2, −3), (1, 2, 2)
2 On the Fine Interior of Three-Dimensional …

(7, −3, −1)

547526 (−1, −1, 0), (−1, 2, −1), (2, −1, 1) (−3, −3, 1), (−1, −1, 0), (−1, 2, −1), (0, −1, 0), (2, −1, 1) vert()\{(−2, 1, 5)}, (0, 1, 3), (−3, 1, 6)
547454 (−1, 1, 0), (0, −1, 0), (2, −1, 0) (−1, −1, 1), (−1, −1, 2), S5 vert(), (2, 1, 2)
547446 (−1, −1, 0), (0, 2, 1), (2, 0, −1) (−1, −1, 0), (−1, 0, 0), (0, 2, 1), (1, 1, 0), (2, 0, −1), (9, 1, −3) vert(), (1, 0, −1), (1, 0, 3)
where S1 := (−1, −1, 2), (−1, −1, 3), (−1, 0, 1), (−1, 0, 2), (−1, 1, 0), (−1, 1, 1), (−1, 2, 0), (−1, 3, −1), (0, −1, 1), (0, −1, 2), (0, 0, 1),
(0, 1, 0), (1, −1, 0), (1, −1, 1), (1, 0, 0), (2, −1, 0)
S2 := (−1, 0, 1), (−1, 1, 1), (−1, 2, 2), (−1, 3, 2), (−1, 4, 3), (−1, 6, 4), (0, 1, 1), (0, 2, 1), (0, 3, 2), (0, 5, 3), (1, 1, 0), (1, 2, 1),
(1, 4, 2), (2, 0, −1), (2, 1, 0), (2, 3, 1), (3, 2, 0), (4, 1, −1)
S3 := (−1, 1, 1), (−1, 2, 1), (0, 0, 1), (0, 1, −1), (0, 1, 0), (0, 1, 1), (1, 0, 0), (1, 0, 1), (1, 2, −5)
S4 := (−1, 0, 1), (−1, 1, 1), (−1, 2, 2), (−1, 3, 2), (−1, 4, 3), (−1, 6, 4), (0, 1, 1), (0, 2, 1), (0, 3, 2), (0, 5, 3), (1, 0, 0), (1, 1, 0),
(1, 2, 1), (1, 4, 2), (2, 1, 0), (2, 3, 1), (3, 0, −1), (3, 2, 0), (4, 1, −1)
S5 := (−1, 0, 1), (−1, 1, 0), (0, −1, 0), (0, −1, 1), (1, −1, 0), (2, −1, 0), (7, −2, −2)
25
26 V. Batyrev et al.

To compute the canonical hull can of , we obtain supp(FI ) = {si | 1 ≤ i ≤

20} with s1 := (−3, −3, −2), s2 := (−1, 0, 0), s3 := (−1, 0, 1), s4 := (−1, 1, 1),
s5 := (−1, 2, 2), s6 := (−1, 3, 2), s7 := (−1, 4, 3), s8 := (−1, 6, 4), s9 := (0, 1, 1),
s10 := (0, 2, 1), . . ., s20 := (4, 1, −1), which leads to

can = conv{v1 , v2 , v3 , v4 , v5 }

with v5 := (−2, 4, −3) (Fig. 2.4b).

Remark 31 The detailed information about a small selection of the 9,020 canonical
Fano 3-topes with dim(FI ) = 1 and 0 ∈ vert(FI ) can be found in Tables 2.2, 2.3,
and 2.4.

2.5 Symmetric Fine Interior of Dimension 1

There exist exactly 20 canonical Fano 3-topes  such that 0 is the center of 1-
dimensional Fine interior FI . In this case, S is an elliptic surface of Kodaira
dimension κ = 1 with non-trivial fundamental group π1 (S ) of order 2 or 3. Com-
putations show that one always has  = can and

FI = conv{−λv , λv }

with λ = 1
2
if and only if |π1 (S )| = 2 and

FI = conv{−μv , μv }

with μ = 23 if and only if |π1 (S )| = 3. The primitive lattice vectors ±v are the
two unique interior lattice points in two reflexive facets θ±   of one of the three
possible types pictured in Fig. 2.2. The reflexive facets θ± of  are isomorphic to
the facet of [∗ ]. The projections M → M/Z(±v ) of  or of θ± along ±v
reveal a reflexive polygon of one of the three types pictured in Fig. 2.3, which is
dual to θ± and . The lattice vector v defines a character of the 3-dimensional
torus χ : T3 → C× . For almost all α ∈ C× , the fiber χ −1 (α) is an affine elliptic
curve defined by a Laurent polynomial with the reflexive Newton polytope ∗ ∼ = θ±∗
of one of the three types pictured in Fig. 2.3 with the distribution shown in Table 2.1.
So χ defines birationally an elliptic fibration. The vertex sets of  and these reflexive
facets are related via vert() = vert(θ+ ) ∪ vert(θ− ). Moreover, every edge of  is
either an edge of θ+ or θ− of these two facets or it is parallel to v .
Example 32 Let  ⊆ MQ be a canonical Fano 3-tope given as the convex hull

v1 := (0, 1, 0), v2 := (2, 1, 1), v3 := (−2, −3, −5), and v4 := (2, 1, 9)

(ID 547393, Fig. 2.5b, Tables 2.5 and 2.6). Then

2 On the Fine Interior of Three-Dimensional … 27

v4 v1

v2
v3

v2

v1

v3 v4

(a) ID 547393 (b) ID 547409

Fig. 2.5 Canonical Fano 3-topes with Symmetric Fine Interior of Dimension 1. Shaded faces
are occluded. The Fine interior and the origin are shown in grey with a double border around
the origin. The facets θ± are grey dotted. a The whole polytope is  = conv{v1 , v2 , v3 , v4 }
with v1 = (1, 0, 0), v2 = (2, 1, 1), v3 = (−2, −3, −5), and v4 = (2, 1, 9). Moreover, FI =
conv{(0, 0, −1/2), (0, 0, 1/2)}, θ+ = conv{v1 , v3 , v4 }, θ− = conv{v1 , v2 , v3 }, and can = . b
The whole polytope is  = conv{v1 , v2 , v3 , v4 } with v1 = (−4, 2, 9), v2 = (1, 0, 0), v3 =
(0, 1, 0), and v4 = (7, −6, −18). Moreover, FI = conv{(−2/3, 2/3, 2), (2/3, −2/3, −2)}, θ+ =
conv{v1 , v2 , v3 }, θ− = conv{v1 , v3 , v4 }, and can = 

FI = conv{(0, 0, −1/2), (0, 0, 1/2)} = conv{−λv , λv }

with λ = 21 , where v = (0, 0, 1). One has 2v1 + v3 + v4 = 4v and 2v1 + v2 +
v3 = 4(−v ). Therefore, v is the interior lattice point of the reflexive facet θ+ = θ134
of  and −v is the interior lattice point of the reflexive facet θ− = θ123 of 
(Fig. 2.2b). The images v1 , v 3 , v 4 of v1 , v3 , v4 in M/Zv and the images v 1 , v 2 , v 3
of v1 , v2 , v3 in M/Z(−v ) are vertices of the dual reflexive triangle θ±∗ (Fig. 2.3b)
satisfying the relation
2v 1 + v3 + v4 = 0

and
2v 1 + v 2 + v 3 = 0,

respectively.
To compute the canonical hull can of , we obtain supp(FI ) = {si | 1 ≤ i ≤ 6}
with s1 := (−1, −2, 2), s2 := (−1, 1, 0), s3 := (0, −1, 0), s4 := (1, −1, 0), s5 :=
(2, −1, 0), and s6 := (9, −2, −2), which leads to can = .
28 V. Batyrev et al.

Table 2.5 20 Canonical Fano 3-topes with Symmetric Fine Interior of Dimension 1. Table
contains: vertices vert() of 
ID vert()
547393 (0, 1, 0), (2, 1, 1), (−2, −3, −5), (2, 1, 9)
547409 (−4, 2, 9), (1, 0, 0), (0, 1, 0), (7, −6, −18)
547461 (0, 1, 0), (2, 1, 1), (−2, −3, −5), (0, 1, 4)
544442 (1, 0, 0), (0, 1, 0), (3, −6, 8), (1, −4, 4), (−5, 6, −12)
544443 (−1, −2, 0), (3, −6, 8), (0, 1, 0), (1, 0, 0), (−3, 4, −8)
544651 (−4, 1, −3), (4, −2, 3), (0, 1, 0), (1, −2, 3), (−1, 1, −3)
544696 (5, −4, −15), (1, 0, 0), (0, 1, 0), (−4, 2, 9), (−3, 1, 6)
544700 (−2, −3, −3), (0, 1, 0), (1, 0, 0), (−1, −4, −6), (2, 5, 9)
544749 (−6, −5, −8), (0, 1, 0), (1, 0, 0), (−2, −1, 0), (3, 2, 4)
520925 (0, 1, 0), (0, 0, 1), (−2, −1, 0), (−2, 0, −1), (8, 2, 3), (−2, −3, −2)
520935 (3, 4, 6), (2, 1, 2), (−3, −2, −2), (1, 0, 0), (0, 1, 0), (−6, −5, −8)
522056 (−1, −1, 0), (0, 1, 0), (1, 0, 0), (−1, −1, −3), (−5, −3, −6), (6, 4, 9)
522059 (2, 5, 6), (−2, −3, −3), (0, 1, 0), (1, 0, 0), (−1, −4, −6), (0, 1, 3)
522087 (1, 0, −3), (1, 0, 0), (0, 1, 0), (−4, 2, 9), (−3, 1, 6), (5, −4, −12)
522682 (2, 1, 4), (−3, −2, −4), (−2, −3, −4), (1, 2, 4), (1, 0, 0), (0, 1, 0)
522684 (−2, −1, −4), (3, 2, 4), (−2, −1, 0), (1, 0, 0), (0, 1, 0), (−4, −3, −4)
526886 (−3, 4, −6), (1, 0, 0), (0, 1, 0), (3, −6, 8), (0, 1, −2), (2, −5, 6)
439403 (1, 2, 2), (−1, 0, 0), (−1, 1, −1), (1, 0, 0), (−1, −2, −2), (1, 1, 3), (1, −3, −1)
275525 (4, 1, 2), (0, 1, 0), (−2, −1, 0), (1, 1, 2), (−3, −1, −2), (−2, −1, −2), (1, 1, 0), (1, −1, 0)
275528 (−1, 0, −1), (−3, −2, 1), (−2, −1, 2), (0, −1, 0), (0, 1, 0), (1, 0, 1), (2, 1, −2), (3, 2, −1)

Example 33 Let  ⊆ MQ be a canonical Fano 3-tope given as the convex hull

v1 := (−4, 2, 9), v2 := (1, 0, 0), v3 := (0, 1, 0), and v4 := (7, −6, −18)

(ID 547409, Fig. 2.5b, Tables 2.5 and 2.6). Then

FI = conv{(−2/3, 2/3, 2), (2/3, −2/3, −2)} = conv{−μv , μv }

with μ = 23 , where v = (1, −1, −3). One has v1 + v2 + v3 = −3v and v1 + v3 +

v4 = −3(−v ). Therefore, v is the interior lattice point of the reflexive facet θ+ =
θ123 of  and −v is the interior lattice point of the reflexive facet θ− = θ134 of 
(Fig. 2.2b). The images v 1 , v 2 , v 3 of v1 , v2 , v3 in M/Zv and the images v 1 , v 3 , v 4
of v1 , v3 , v4 in M/Z(−v ) are vertices of the dual reflexive triangle θ±∗ (Fig. 2.3b)
satisfying the relation
v 1 + v 2 + v3 = 0,

and
v 1 + v 3 + v 4 = 0,

respectively.
Table 2.6 20 Canonical Fano 3-topes with Symmetric Fine Interior of Dimension 1. Table contains: vertices vert(FI ) of the Fine interior FI , unique
primitive lattice points ±v ∈ θ± in the reflexive facets θ±  , vertices vert(θ± ) of the reflexive facets θ±  , and support supp(FI ) of the Fine interior FI
(here: can = )
ID vert(FI ) ±v vert(θ± ) supp(FI )
547393 ±1/2 · v ±(0, 0, 1) (0, 1, 0), (2, 1, 1), (−2, −3, −5) (−1, −2, 2), (−1, 1, 0), (0, −1, 0), (1, −1, 0), (2, −1, 0), (9, −2, −2)
(0, 1, 0), (−2, −3, −5), (2, 1, 9)
547409 ±2/3 · v ±(1, −1, −3) (−4, 2, 9), (1, 0, 0), (0, 1, 0) (−3, −3, −1), (−1, −1, 0), (−1, 2, −1), (2, −1, 1), (15, −3, 7)
(−4, 2, 9), (0, 1, 0), (7, −6, −18)
547461 ±1/2 · v ±(0, 0, 1) (0, 1, 0), (2, 1, 1), (−2, −3, −5) (−3, 6, −2), (−1, −2, 2), (−1, 1, 0), (0, −1, 0), (1, −1, 0), (2, −1, 0)
(2, 1, 1), (−2, −3, −5), (0, 1, 4)
544442 ±1/2 · v ±(1, −1, 2) (0, 1, 0), (1, −4, 4), (−5, 6, −12) (−2, −2, −1), (−1, −1, 0), (−1, 1, 1), (1, −1, −1), (3, −1, −2), (10, −2, −5)
2 On the Fine Interior of Three-Dimensional …

(1, 0, 0), (0, 1, 0), (3, −6, 8)

544443 ±1/2 · v ±(1, −1, 2) (−1, −2, 0), (0, 1, 0), (−3, 4, −8) (−2, −2, −1), (−1, −1, 0), (−1, 1, 1), (1, −1, −1), (3, −1, −2), (6, −2, −3)
(3, −6, 8), (0, 1, 0), (1, 0, 0)
544651 ±2/3 · v ±(1, 0, 0) (−4, 1, −3), (0, 1, 0), (1, −2, 3) (−3, −3, 1), (0, −1, −1), (0, −1, 0), (0, 2, 1), (3, −3, −4)
(4, −2, 3), (0, 1, 0), (−1, 1, −3)
544696 ±2/3 · v ±(1, −1, −3) (1, 0, 0), (0, 1, 0), (−4, 2, 9) (−3, −3, −1), (−3, 12, −4), (−1, −1, 0), (−1, 2, −1), (2, −1, 1)
(5, −4, −15), (1, 0, 0), (−3, 1, 6)
544700 ±2/3 · v ±(1, 2, 3) (−2, −3, −3), (0, 1, 0), (−1, −4, −6) (−3, −3, 2), (−1, −1, 1), (−1, 2, −1), (2, −1, 0), (3, −3, 2)
(0, 1, 0), (1, 0, 0), (2, 5, 9)
544749 ±1/2 · v ±(1, 1, 2) (−6, −5, −8), (0, 1, 0), (1, 0, 0) (−2, −2, 3), (−1, −1, 1), (−1, 1, 0), (−1, 3, −1), (1, −1, 0), (2, −2, −1)
(0, 1, 0), (−2, −1, 0), (3, 2, 4)
(continued)
29
 30

Table 2.6 (continued)

ID vert(FI ) ±v vert(θ± ) supp(FI )
520925 ±1/2 · v ±(2, 1, 1) (−2, −1, 0), (−2, 0, −1), (−2, −3, −2) (−1, −1, 3), (0, −1, 1), (0, 1, −1), (1, −2, −2), (1, −1, −1), (1, 0, 0)
(0, 1, 0), (0, 0, 1), (8, 2, 3)
520935 ±1/2 · v ±(1, 1, 2) (1, 0, 0), (0, 1, 0), (−6, −5, −8) (−2, −2, 3), (−1, −1, 1), (−1, 1, 0), (−1, 3, −1), (0, 4, −3), (1, −1, 0)
(3, 4, 6), (2, 1, 2), (−3, −2, −2)
522056 ±2/3 · v ±(2, 1, 3) (0, 1, 0), (−1, −1, −3), (−5, −3, −6) (−3, 6, −1), (−1, −1, 1), (−1, 2, 0), (0, −3, 2), (2, −1, −1)
(−1, −1, 0), (1, 0, 0), (6, 4, 9)
522059 ±2/3 · v ±(1, 2, 3) (−2, −3, −3), (0, 1, 0), (−1, −4, −6) (−3, 3, −2), (−1, −1, 1), (−1, 2, −1), (2, −1, 0), (3, −3, 2)
(2, 5, 6), (1, 0, 0), (0, 1, 3)
522087 ±2/3 · v ±(1, −1, −3) (1, 0, 0), (0, 1, 0), (−4, 2, 9) (−3, −3, −1), (−1, −1, 0), (−1, 2, −1), (2, −1, 1), (9, 0, 4)
(1, 0, −3), (−3, 1, 6), (5, −4, −12)
522682 ±1/2 · v ±(1, 1, 2) (−3, −2, −4), (−2, −3, −4), (1, 0, 0), (0, 1, 0) (−2, −2, 1), (−2, −2, 3), (−1, −1, 1), (−1, 1, 0), (1, −1, 0), (1, 1, −1)
(2, 1, 4), (1, 2, 4), (1, 0, 0), (0, 1, 0)
522684 ±1/2 · v ±(1, 1, 2) (−2, −1, −4), (1, 0, 0), (−4, −3, −4) (−2, 2, 1), (−1, −1, 1), (−1, 1, 0), (−1, 3, −1), (1, −1, 0), (2, −2, −1)
(3, 2, 4), (−2, −1, 0), (0, 1, 0)
526886 ±1/2 · v ±(1, −1, 2) (−3, 4, −6), (0, 1, −2), (2, −5, 6) (−2, −2, −1), (−1, −1, 0), (−1, 1, 1), (0, 4, 3), (1, −1, −1), (3, −1, −2)
(1, 0, 0), (0, 1, 0), (3, −6, 8)
439403 ±1/2 · v ±(0, 1, 1) (−1, 1, −1), (1, 0, 0), (−1, −2, −2), (1, −3, −1) (−2, −1, 3), (−1, −1, 1), (−1, 0, 0), (1, 0, 0), (1, 1, −1), (2, −1, −1)
(1, 2, 2), (−1, 0, 0), (−1, 1, −1), (1, 1, 3)
275525 ±1/2 · v ±(1, 0, 0) (0, 1, 0), (−2, −1, 0), (1, 1, 2), (−3, −1, −2) (−2, 0, 3), (0, −1, 0), (0, −1, 1), ±(0, 1, −1), (0, 1, 0), (2, −2, −1)
(4, 1, 2), (−2, −1, −2), (1, 1, 0), (1, −1, 0)
275528 ±1/2 · v ±(1, 1, −1) (−3, −2, 1), (−2, −1, 2), (0, −1, 0), (1, 0, 1) (−1, 1, 0), (−1, 2, −1), (0, −1, −1), (0, 1, 1), (1, −2, 1), (1, −1, 0)
(−1, 0, −1), (0, 1, 0), (2, 1, −2), (3, 2, −1)
V. Batyrev et al.
2 On the Fine Interior of Three-Dimensional … 31

To compute the canonical hull can of , we obtain supp(FI ) = {si | 1 ≤ i ≤

5} with s1 := (−3, −3, −1), s2 := (−1, −1, 0), s3 := (−1, 2, −1), s4 := (2, −1, 1),
and s5 := (15, −3, 7), which leads to can = .

Remark 34 Information about all 20 canonical Fano 3-topes with dim(FI ) = 1

◦
and 0 ∈ (FI ) can be found in Tables 2.5 and 2.6.

2.6 Fine Interior of Dimension 3

There exist 49 canonical Fano 3-topes  such that dim(FI ) = 3. Exactly 3 of

these polytopes  define minimal surface S with non-trivial fundamental group of
order 2 and K 2 = 2. For these 3 polytopes one has  = can . The surfaces S were
investigated by Todorov [24] as well as Catanese and Debarre [10].
The remaining 46 canonical Fano 3-topes  define simply connected minimal sur-
faces S with K 2 = 1. These surfaces were investigated by Kanev [19], Catanese [9],
and Todorov [23]. Among these 46 canonical Fano 3-topes there exist exactly 26
polytopes  such that  = can .
Example 35 ([19]) Let M ⊆ Q4 be the 3-dimensional affine lattice defined by

M := {(m 1 , m 2 , m 3 , m 4 ) ∈ Z4 | m 1 + m 2 + m 3 + 2m 4 = 6, m 2 + 2m 3 ≡ 0 (mod 3)}

and  ⊆ MQ be the convex hull of 4 lattice points

(6, 0, 0, 0), (0, 6, 0, 0), (0, 0, 6, 0), and (0, 0, 0, 3) ∈ M.

Then ( )FI is the 3-dimensional rational simplex

conv{(2, 1, 1, 1), (1, 2, 1, 1), (1, 1, 2, 1), (1, 1, 1, 3/2)}

and ( )FI ∩ M = {(2, 1, 1, 1)}.

The canonical Fano 3-tope  is the Newton polytope of the μ3 -cyclic quotient Z 
of the projective surface of degree 6 defined by the polynomial z 16 + z 26 + z 36 + z 43 = 0
in the weighted projective space P(1, 1, 1, 2), where the cyclic group μ3 acts
via (z 1 : z 2 : z 3 : z 4 ) → (z 1 : εz 2 : ε2 z 3 : z 4 ). The single interior lattice point in 
corresponds to the monomial z 12 z 2 z 3 z 4 . The surface Z  has 3 cyclic quotient singu-
larities of type A2 . The minimal desingularization S of Z  is a simply connected
surface of general type with K 2 = 1.
One can identify  with the canonical Fano 3-simplex  given as the convex
hull of

v1 := (1, 0, 0), v2 := (−2, −4, −5), v3 := (1, 2, 4), and v4 := (1, 4, 2)

32 V. Batyrev et al.

v3

v4

v4
v1
v3 v2

v1
v2

(a) ID 547444 (b) ID 547465

Fig. 2.6 Canonical Fano 3-topes with Fine Interior of Dimension 3. Shaded
faces are occluded. The Fine interior and the origin are shown in grey with a
double border around the origin. a The whole polytope is  = conv{v1 , v2 , v3 , v4 }
with v1 = (1, 0, 0), v2 = (−2, −4, −5), v3 = (1, 2, 4), and v4 = (1, 4, 2). More-
over, FI = conv{(0, 0, 0), (−1/2, −1, −3/2), (0, −1/3, −2/3), (0, 1/3, −1/3)}
and can = . b The whole polytope is  = conv{v1 , v2 , v3 , v4 } with v1 =
(−3, −2, −2), v2 = (1, 0, 0), v3 = (1, 3, 1), and v4 = (1, 1, 3). Moreover, FI =
conv{(0, 0, 0), (−1, −1/2, −1/2), (0, 3/4, 1/4), (0, 1/4, 3/4)} and can = 

(ID 547444, Fig. 2.6a, Tables 2.7, 2.8, and 2.9). The primitive inward-pointing facet
normals of the facets θ124 , θ234 , θ123 , and θ134   of this simplex  are

n 1 := (−2, −1, 2), n 2 := (5, −1, −1), n 3 := (−1, 2, −1), and n 4 := (−1, 0, 0),

respectively. They satisfy the relation

n 1 + n 2 + n 3 + 2n 4 = 0.

To compute the canonical hull can of , we obtain supp(FI ) = {si | 1 ≤ i ≤ 6}

with s1 := (−2, −1, 2), s2 := (−1, 0, 0), s3 := (−1, 2, −1), s4 := (1, 1, −1), s5 :=
(3, 0, −1), and s6 := (5, −1, −1), which leads to can = .

Example 36 ([24]) Let M ⊆ Q4 be the 3-dimensional affine lattice defined by

M := {(m 1 , m 2 ,m 3 , m 4 ) ∈ Z4 |
m 1 + m 2 + 2m 3 + 2m 4 = 8, 3m 2 + m 3 + 3m 4 ≡ 0 (mod 4)}

and  ⊆ MQ be the convex hull of 4 lattice points

(8, 0, 0, 0), (0, 8, 0, 0), (0, 0, 4, 0), and (0, 0, 0, 4) ∈ M.

2 On the Fine Interior of Three-Dimensional … 33

Table 2.7 49 Canonical Fano 3-topes with Fine Interior of Dimension 3 Table contains: ver-
tices vert() of 
ID vert()
547444 (1, 0, 0), (−2, −4, −5), (1, 2, 4), (1, 4, 2)
547465 (−3, −2, −2), (1, 0, 0), (1, 3, 1), (1, 1, 3)
547524 (0, 2, 1), (−2, −3, −5), (2, 1, 1), (0, 0, 1)
547525 (0, 0, 1), (0, 1, 0), (2, 1, 1), (−2, −5, −7)
545317 (−3, 4, −6), (0, 1, 0), (1, 0, 0), (1, −2, 4), (3, −5, 6)
545932 (0, −1, −1), (1, −1, −3), (−2, 1, 5), (1, 0, 0), (1, 2, −2)
546013 (3, −5, 6), (1, −2, 4), (1, 0, 0), (−1, 1, −2), (−1, 3, −2)
546062 (0, 1, 3), (−2, 1, −1), (0, 1, 0), (1, 0, 0), (−1, −2, −2)
546070 (0, −2, −3), (0, 2, 1), (−2, −3, −5), (2, 1, 1), (0, 0, 1)
546205 (1, 2, −2), (−1, 0, 2), (1, 0, 0), (−2, 1, 5), (1, −1, −3)
546219 (1, 1, 1), (−3, −2, −2), (1, 0, 0), (1, 3, 1), (−1, −1, 1)
546663 (2, −3, −1), (1, 0, 0), (0, 1, 0), (0, 0, 1), (−2, −3, −3)
546862 (1, 0, 0), (0, 1, 0), (−2, 1, 5), (1, −1, −3), (1, 2, −2)
546863 (−1, −1, 1), (1, 3, 1), (0, 0, 1), (1, 0, 0), (−3, −2, −2)
547240 (−1, 1, −2), (0, 1, 0), (1, 0, 0), (1, −2, 4), (3, −5, 6)
547246 (0, −2, −3), (−2, −3, −5), (2, 1, 1), (0, 1, 0), (0, 0, 1)
532384 (1, −1, −3), (−2, 1, 5), (1, 0, 0), (1, −1, −2), (0, −1, −1), (1, 2, −2)
532606 (0, −1, 2), (−1, −1, 0), (0, 1, 0), (1, 0, 0), (2, 2, −3), (−2, 0, −3)
533513 (−1, 1, 2), (1, 0, 0), (0, 1, 0), (1, 1, 2), (−1, −2, −4), (−2, −3, −4)
534667 (1, 0, 3), (−1, −1, −1), (0, 1, 0), (1, 0, 0), (−1, −1, 0), (5, 2, 3)
534669 (1, 3, 0), (5, 3, 2), (−1, −1, −1), (0, 0, 1), (1, 0, 0), (−1, −1, 0)
534866 (−1, −1, −3), (1, 0, 0), (0, 1, 0), (1, 1, 1), (−1, −1, 0), (−3, −5, −3)
535952 (3, −5, 6), (1, −2, 4), (1, 0, 0), (0, 1, 0), (−1, 1, −2), (−1, 2, −2)
536013 (0, 1, 1), (0, 0, 1), (0, 1, 0), (2, 1, 1), (−2, −3, −5), (0, −2, −3)
536498 (1, 2, −2), (1, −1, −2), (1, 0, 0), (0, 1, 0), (−2, 1, 5), (1, −1, −3)
537834 (0, 0, 1), (1, 0, 0), (0, 1, 0), (−2, 1, 5), (1, −1, −3), (1, 2, −2)
538356 (−2, −3, −3), (−1, −3, −1), (0, 0, 1), (0, 1, 0), (1, 0, 0), (−1, −1, −3)
539063 (−1, 1, −1), (1, 1, 3), (−3, −2, −2), (1, 0, 0), (0, 1, 0), (1, 1, 2)
539304 (1, 0, 1), (−3, −1, −2), (1, 1, 2), (−2, −1, 0), (1, 0, 0), (1, 2, 0)
539313 (1, −1, −2), (1, 1, −1), (−1, 2, 2), (1, −1, −3), (−2, 1, 5), (1, 0, 0)
540602 (0, 0, 1), (1, 0, 0), (−2, 1, 5), (1, −1, −3), (−1, 2, 2), (1, 1, −1)
540663 (1, 0, 0), (0, 1, 0), (1, 1, 2), (−3, −1, −2), (1, 1, 1), (−3, −2, 0)
474457 (−1, 2, −3), (1, 0, 2), (0, 0, 1), (0, 1, 0), (1, 0, 0), (−1, −1, 0), (−3, −2, −3)
481575 (3, 2, 4), (−1, −1, −2), (−3, −1, −2), (−2, −1, 0), (0, 1, 0), (1, 0, 0), (0, 0, −1)
483109 (3, 0, 2), (1, −2, −2), (0, 0, −1), (−1, −1, 0), (1, 1, 1), (0, 1, 0), (−1, 0, 0)
490478 (1, −1, −2), (1, 1, −1), (−1, 2, 2), (1, −1, −3), (−2, 1, 5), (1, 0, 0), (−1, 0, 2)
490481 (−3, −2, 0), (−5, −3, −2), (1, 0, 0), (0, 1, 0), (1, 1, 2), (−1, −1, −1), (2, 1, 1)
490485 (−1, −1, 0), (1, 2, 0), (1, 0, 0), (−2, −1, 0), (1, 1, 2), (−3, −1, −2), (1, 0, 1)
490511 (1, 0, 0), (0, 1, 0), (−2, −1, 0), (1, 1, 2), (2, 1, 1), (1, 0, 1), (−5, −2, −4)
495687 (0, 0, −1), (1, 1, −1), (−1, 2, 2), (1, −1, −3), (−2, 1, 5), (1, 0, 0), (0, 0, 1)
499287 (1, 1, 1), (−1, −1, −3), (1, 0, 0), (0, 1, 0), (0, 0, 1), (−1, −3, −1), (−2, −3, −3)
499291 (−1, −1, −1), (−1, −1, −3), (1, 0, 0), (0, 1, 0), (0, 0, 1), (−1, −3, −1), (−2, −3, −3)
499470 (1, 0, 0), (0, 1, 0), (−2, −1, 0), (1, 1, 2), (0, 0, 1), (−5, −2, −4), (2, 1, 1)
501298 (3, −6, 8), (−1, 1, −2), (1, −2, 3), (0, 1, 0), (1, 0, 0), (0, 1, −1), (3, −5, 6)
501330 (1, 0, 0), (0, 1, 0), (−2, −1, 0), (1, 1, 2), (1, 1, 1), (0, 0, 1), (−5, −2, −4)
354912 (3, 1, 2), (1, 0, 0), (0, 1, 0), (−2, −1, 0), (1, 1, 2), (2, 1, 1), (1, 0, 1), (−5, −2, −4)
372528 (2, 1, 1), (−1, −1, −1), (1, 1, 2), (0, 1, 0), (1, 0, 0), (−5, −3, −2), (−3, −2, 0), (1, 1, 0)
372973 (−5, −2, −4), (1, 0, 1), (2, 1, 1), (1, 1, 2), (−2, −1, 0), (0, 1, 0), (1, 0, 0), (2, 1, 2)
388701 (1, 1, 1), (−2, −3, −3), (−1, −3, −1), (0, 0, 1), (0, 1, 0), (1, 0, 0), (−1, −1, −3), (−1, −1, −1)
34 V. Batyrev et al.

Table 2.8 49 Canonical Fano 3-topes with Fine Interior of Dimension 3. Table contains: ver-
tices vert(FI ) of the Fine interior FI
ID vert(FI )
547444 0, (−1/2, −1, −3/2), (0, −1/3, −2/3), (0, 1/3, −1/3)
547465 0, (−1, −1/2, −1/2), (0, 3/4, 1/4), (0, 1/4, 3/4)
547524 0, (0, 1/2, 0), (1/3, 1/3, 0), (−1/3, −1/3, −1)
547525 0, (0, 0, −1/2), (1/3, 0, −1/3), (−1/3, −1, −5/3)
545317 0, (1, −3/2, 2), (2/3, −2/3, 1), (1/2, −1/2, 1), (2/3, −1, 5/3)
545932 0, (−1/2, 1/2, 3/2), (0, 1/3, 2/3), (0, 2/3, 1/3)
546013 0, (1, −3/2, 2), (0, 1/2, 0), (1/2, −1/4, 1/2), (1/2, −3/4, 3/2)
546062 0, (−1/2, −1/2, −1/2), (−2/3, 0, −1/3), (−1/3, 0, 1/3)
546070 0, (0, 1/2, 0), (1/2, 1/4, 0), (0, −1/2, −1), (−1/2, −3/4, −3/2)
546205 0, (−1/2, 1/2, 3/2), (0, 1/3, 2/3), (0, 2/3, 1/3)
546219 0, (−1, −1/2, −1/2), (−1/3, 1/3, 0), (−2/3, −1/3, 0)
546663 0, (0, −1/2, 0), (1/3, −1, −1/3), (−1/3, −1, −2/3)
546862 0, (−1/2, 1/2, 3/2), (0, 1/3, 2/3), (0, 2/3, 1/3)
546863 0, (−1, −1/2, −1/2), (−1/3, 1/3, 0), (−2/3, −1/3, 0)
547240 0, (1, −3/2, 2), (2/3, −2/3, 1), (1/2, −1/2, 1), (2/3, −1, 5/3)
547246 0, (0, 0, −1/2), (1/3, 0, −1/3), (0, −1/2, −1), (−1/3, −2/3, −4/3)
532384 0, (−1/2, 1/2, 3/2), (0, 1/3, 2/3), (0, 2/3, 1/3)
532606 0, (0, 1/2, −1/2), (1/3, 2/3, −1), (−1/3, 1/3, −1)
533513 0, (−1/2, −1/2, −1), (−1/2, 0, 0), (−1/3, 0, −1/3), (−2/3, −2/3, −1)
534667 0, (1/2, 1/2, 1/2), (4/3, 2/3, 1), (2/3, 1/3, 1)
534669 0, (1/2, 1/2, 1/2), (4/3, 1, 2/3), (2/3, 1, 1/3)
534866 0, (0, −1/2, −1/2), (−1/3, −2/3, −1), (−2/3, −4/3, −1)
535952 0, (1, −3/2, 2), (2/3, −2/3, 1), (1/2, −1/2, 1), (2/3, −1, 5/3)
536013 0, (0, 0, −1/2), (1/3, 0, −1/3), (0, −1/2, −1), (−1/3, −2/3, −4/3)
536498 0, (−1/2, 1/2, 3/2), (0, 1/3, 2/3), (0, 2/3, 1/3)
537834 0, (−1/2, 1/2, 3/2), (0, 1/3, 2/3), (0, 2/3, 1/3)
538356 0, (0, −1/2, −1/2), (−1/3, −2/3, −1), (−1/3, −1, −2/3), (−1/2, −1, −1)
539063 0, (−1, −1/2, −1/2), (−2/3, 0, −1/3), (−1/3, 0, 1/3)
539304 0, (0, 1/2, 0), (−1/2, 0, 0), (0, 1/3, 1/3), (−2/3, 0, −1/3)
539313 0, (−1/2, 1/2, 3/2), (0, 1/3, 2/3), (0, 1/2, 1/2), (−1/3, 2/3, 1)
540602 0, (−1/2, 1/2, 3/2), (0, 1/3, 2/3), (0, 1/2, 1/2), (−1/3, 2/3, 1)
540663 0, (−1/2, 0, 0), (−1, −1/2, 0), (−1/3, 0, 1/3), (−1, −1/3, −1/3)
474457 0, (0, 0, −1/2), (−1/3, 1/3, −1), (−2/3, −1/3, −1)
481575 0, (−1/2, 0, 0), (1/2, 1/2, 1), (0, 1/3, 1/3), (−1/3, 0, 1/3)
483109 0, (0, −1/2, 0), (2/3, −1/3, 1/3), (1/3, −2/3, −1/3)
490478 0, (−1/2, 1/2, 3/2), (0, 1/3, 2/3), (0, 1/2, 1/2), (−1/3, 2/3, 1)
490481 0, (−1/2, 0, 0), (−1, −1/2, 0), (−1/3, 0, 1/3), (−4/3, −2/3, −1/3)
490485 0, (0, 1/2, 0), (−1/2, 0, 0), (0, 1/3, 1/3), (−2/3, 0, −1/3)
490511 0, (−3/2, −1/2, −1), (−1/2, 0, 0), (−2/3, 0, −1/3), (−1, −1/3, −1/3)
495687 0, (−1/2, 1/2, 3/2), (0, 1/3, 2/3), (0, 1/2, 1/2), (−1/3, 2/3, 1)
499287 0, (0, −1/2, −1/2), (−1/3, −2/3, −1), (−1/3, −1, −2/3), (−1/2, −1, −1)
499291 0, (0, −1/2, −1/2), (−1/3, −2/3, −1), (−1/3, −1, −2/3), (−1/2, −1, −1)
499470 0, (−3/2, −1/2, −1), (−1/2, 0, 0), (−2/3, 0, −1/3), (−1, −1/3, −1/3)
501298 0, (1/2, −1/2, 1), (2/3, −2/3, 1), (1, −3/2, 2), (1, −5/3, 7/3)
501330 0, (−3/2, −1/2, −1), (−1/2, 0, 0), (−2/3, 0, −1/3), (−1, −1/3, −1/3)
354912 0, (−3/2, −1/2, −1), (−1/2, 0, 0), (−2/3, 0, −1/3), (−1, −1/3, −1/3)
372528 0, (−1/2, 0, 0), (−1, −1/2, 0), (−1/3, 0, 1/3), (−4/3, −2/3, −1/3)
372973 0, (−3/2, −1/2, −1), (−1/2, 0, 0), (−2/3, 0, −1/3), (−1, −1/3, −1/3)
388701 0, (0, −1/2, −1/2), (−1/3, −2/3, −1), (−1/3, −1, −2/3), (−1/2, −1, −1)
Table 2.9 49 Canonical Fano 3-topes with Fine Interior of Dimension 3. Table contains: support supp(FI ) of the Fine interior FI , vertices vert(can ) of
the canonical hull can , and order of fundamental group |π1 (S )| of the minimal model S
ID supp(FI ) vert(can ) |π1 (S )|
547444 (−2, −1, 2), (−1, 0, 0), (−1, 2, −1), (1, 1, −1), (3, 0, −1), (5, −1, −1) vert() 1
547465 (−1, −1, 3), (−1, 0, 0), (−1, 0, 1), (−1, 0, 2), (−1, 1, 0), (−1, 1, 1), (−1, 2, 0), (−1, 3, −1), vert() 2
(2, −1, −1)
547524 (−1, −2, 2), (−1, 1, 0), (−1, 2, −1), (0, 0, −1), (0, 1, −1), (0, 2, −1), (1, 0, −1), (1, 1, −1), vert(), (0, −1, −1) 1
(2, 0, −1), (3, 0, −1)
2 On the Fine Interior of Three-Dimensional …

547525 (−1, −2, 2), (−1, 2, −1), (0, −1, 0), (0, 0, −1), (0, 1, −1), (1, −1, −1), (1, −1, 0), (1, 0, −1), vert(), (1, 1, 1), (−1, −2, −3) 1
(1, 1, −1), (2, −1, −1), (2, −1, 0), (2, 0, −1), (3, −1, −1), (3, −1, 0), (3, 0, −1), (4, −1, −1),
(4, 0, −1), (5, −1, −1), (6, −1, −1)
545317 (−2, −2, −1), (−1, −1, 0), (−1, 2, 2), (1, −1, −1), (1, 2, 1), (3, 2, 0) vert() 1
545932 (−2, −1, −1), (−1, 0, 0), (−1, 2, −1), (0, 1, 0), (1, 1, 0), (2, 0, 1), (3, 0, 1), (5, −1, 2) vert(), (1, −1, −2), (1, 0, −3) 1
546013 (−2, −2, −1), (−1, 0, 1), (−1, 2, 2), (0, 1, 1), (1, 0, 0), (1, 2, 1), (2, 1, 0), (3, 0, −1), (3, 2, 0) vert() 2
546062 (−1, −1, 0), (−1, −1, 1), (−1, −1, 2), (−1, 0, 0), (−1, 0, 1), (−1, 1, 0), (−1, 2, −1), (0, −1, 0), vert() 1
(2, 1, −1)
(continued)
35
 36

Table 2.9 (continued)

ID supp(FI ) vert(can ) |π1 (S )|
546070 (−1, −2, 2), (−1, 2, −1), (0, 0, −1), (0, 1, −1), (0, 2, −1), (1, 0, −1), (1, 1, −1), (2, 0, −1), (3, 0, −1) vert() 2
546205 (−2, −1, −1), (−1, 0, 0), (−1, 2, −1), (−1, 3, −1), (0, 1, 0), (1, 1, 0), (1, 2, 0), (2, 0, 1), vert() 1
(3, 0, 1), (3, 1, 1), (5, −1, 2), (5, 0, 2), (7, −1, 3)
546219 (−1, −1, 3), (−1, 0, 0), (−1, 0, 1), (−1, 0, 2), (−1, 1, −1), (−1, 1, 0), (−1, 1, 1), (−1, 2, 0), vert() 1
(0, 0, −1), (2, −1, −1)
546663 (−1, −1, −1), (−1, −1, 0), (−1, −1, 1), (−1, −1, 2), (−1, 0, −1), (0, −1, −1), (0, −1, 0), vert(), (−1, −1, −1) 1
(0, −1, 1), (0, 0, −1), (1, −1, −1), (1, −1, 0), (1, 0, −1), (1, 2, −2), (2, −1, −1), (2, −1, 0),
(2, 0, −1), (3, −1, −1)
546862 (−2, −1, −1), (−1, 0, 0), (−1, 2, −1), (−1, 3, −1), (0, 1, 0), (0, 4, −1), (1, 1, 0), (1, 2, 0), vert(), (0, 0, 1) 1
(2, 0, 1), (2, 3, 0), (3, 0, 1), (3, 1, 1), (4, 2, 1), (5, −1, 2), (5, 0, 2), (6, 1, 2), (7, −1, 3),
(8, 0, 3), (10, −1, 4)
546863 (−1, −1, 3), (−1, 0, 0), (−1, 0, 1), (−1, 0, 2), (−1, 1, −1), (−1, 1, 0), (−1, 1, 1), (−1, 2, 0), vert()\{(0, 0, 1)}, (1, 1, 1) 1
(0, 0, −1), (2, −1, −1)
547240 (−2, −2, −1), (−1, −1, 0), (−1, 0, 1), (−1, 2, 2), (0, −1, 0), (0, 1, 1), (1, −1, −1), (1, 0, 0), vert(), (0, 1, −1), (0, 0, 1) 1
(1, 2, 1), (2, −1, −1), (2, 1, 0), (3, 0, −1), (3, 2, 0)
547246 (−1, −2, 2), (−1, 2, −1), (0, −1, 0), (0, 0, −1), (0, 1, −1), (0, 2, −1), (1, −1, −1), (1, −1, 0), vert(), (1, 1, 1), (−1, −1, −2) 1
(1, 0, −1), (1, 1, −1), (2, −1, −1), (2, −1, 0), (2, 0, −1), (3, −1, −1), (3, 0, −1), (4, −1, −1)
(continued)
V. Batyrev et al.
Table 2.9 (continued)
ID supp(FI ) vert(can ) |π1 (S )|
532384 (−2, −1, −1), (−1, 0, 0), (−1, 2, −1), (0, 1, 0), (1, 1, 0), (2, 0, 1), (3, 0, 1), (5, −1, 2) vert(), (1, 0, −3) 1
532606 (−1, −1, −1), (−1, 1, 0), (−1, 2, 1), (0, −1, −1), (0, 1, 0), (1, −2, 0), (1, −1, −1), (2, −1, −1) vert(), (0, −1, 1) 1
533513 (−1, −1, 1), (−1, 0, 0), (−1, 1, 0), (−1, 2, −1), (0, −1, 0), (0, 1, −1), (0, 3, −2), (2, −2, 1) vert(), (1, 1, 1), (0, −1, −1) 1
534667 (−1, −1, 2), (−1, −1, 3), (−1, 0, 2), (−1, 1, 1), (−1, 2, 0), (0, −1, 1), (0, −1, 2), (0, 0, 1), vert() 1
(0, 1, 0), (1, −2, −1), (1, −1, 0), (1, −1, 1), (1, 0, 0), (2, −1, −1), (2, −1, 0)
534669 (−1, 0, 2), (−1, 1, 1), (−1, 2, −1), (−1, 2, 0), (0, 0, 1), (0, 1, −1), (0, 1, 0), (1, −1, −2), vert() 1
(1, 0, −1), (1, 0, 0), (2, −1, −1), (2, −1, 0)
534866 (−2, 1, 1), (−1, −1, 1), (−1, 0, 0), (−1, 1, −1), (−1, 2, −2), (0, −1, 0), (0, 0, −1), (0, 1, −2), vert() 1
2 On the Fine Interior of Three-Dimensional …

(1, −1, −1), (1, −1, 0), (1, 0, −2), (1, 0, −1), (2, −1, −2), (2, −1, −1), (2, −1, 0)
535952 (−2, −2, −1), (−1, −1, 0), (−1, 0, 1), (−1, 2, 2), (0, 1, 1), (1, −1, −1), (1, 0, 0), (1, 2, 1), vert() 1
(2, 1, 0), (3, 0, −1), (3, 2, 0)
536013 (−1, −2, 2), (−1, 2, −1), (0, −1, 0), (0, 0, −1), (0, 1, −1), (0, 2, −1), (1, −1, 0), (1, 0, −1), vert() 1
(1, 1, −1), (2, −1, 0), (2, 0, −1), (3, 0, −1)
536498 (−2, −1, −1), (−1, 0, 0), (−1, 2, −1), (0, 1, 0), (1, 1, 0), (1, 2, 0), (2, 0, 1), (2, 3, 0), vert() 1
(3, 0, 1), (3, 1, 1), (4, 2, 1), (5, −1, 2), (5, 0, 2), (6, 1, 2), (7, −1, 3), (8, 0, 3), (10, −1, 4)
537834 (−2, −1, −1), (−1, 0, 0), (−1, 2, −1), (−1, 3, −1), (0, 1, 0), (0, 4, −1), (1, 1, 0), (1, 2, 0), vert() 1
(2, 0, 1), (2, 3, 0), (3, 0, 1), (3, 1, 1), (4, 2, 1), (5, −1, 2), (5, 0, 2), (6, 1, 2), (7, −1, 3),
(8, 0, 3), (10, −1, 4)
(continued)
37
 38

Table 2.9 (continued)

ID supp(FI ) vert(can ) |π1 (S )|
538356 (−2, 1, 1), (−1, −1, −1), (−1, −1, 0), (−1, −1, 1), (−1, 0, −1), (−1, 0, 0), (−1, 1, −1), vert(), (−1, −1, −1) 1
(0, −1, −1), (0, −1, 0), (0, 0, −1), (1, −1, −1), (1, −1, 0), (1, 0, −1), (2, −1, −1), (2, −1, 0),
(2, 0, −1), (3, −1, −1)
539063 (−1, −1, 1), (−1, 0, 0), (−1, 0, 1), (−1, 0, 2), (−1, 1, 0), (−1, 1, 1), (−1, 2, 0), (−1, 3, −1), vert() \ {(0, 1, 0), (1, 1, 2)}, (1, 1, 1) 1
(0, −1, 0), (2, −1, −1)
539304 (−1, 0, 0), (−1, 0, 1), (−1, 0, 2), (−1, 1, 0), (−1, 1, 1), (−1, 2, 0), (−1, 2, 1), (−1, 3, 0), vert(), (−2, −1, −1) 1
(0, 1, −1), (0, 1, 0), (2, −2, −1)
539313 (−2, −1, −1), (−1, 0, 0), (−1, 2, −1), (0, 1, 0), (1, −1, 1), (1, 1, 0), (1, 2, 0), (2, 0, 1), vert(), (−1, 1, 2) 1
(2, 3, 0), (3, 0, 1), (3, 1, 1), (4, 2, 1), (5, 0, 2), (6, 1, 2)
540602 (−2, −1, −1), (−1, 0, 0), (−1, 2, −1), (−1, 3, −1), (0, 1, 0), (0, 4, −1), (1, −1, 1), (1, 1, 0), vert(), (−1, 1, 2) 1
(1, 2, 0), (2, 0, 1), (2, 3, 0), (3, 0, 1), (3, 1, 1), (4, 2, 1), (5, 0, 2), (6, 1, 2)
540663 (−1, −1, 1), (−1, −1, 2), (−1, 0, 0), (−1, 0, 1), (−1, 0, 2), (−1, 1, 0), (−1, 1, 1), (−1, 2, −1), vert(), (−1, 0, −1) 1
(−1, 2, 0), (−1, 2, 1), (0, −1, 0), (0, −1, 1), (2, −2, −1)
474457 (−2, 1, 2), (−1, −1, 0), (−1, 0, 0), (−1, 1, 0), (−1, 2, 0), (1, −1, −1), (1, 0, −1), (2, −1, −1) vert() 1
481575 (−1, −1, 1), (−1, 0, 1), (−1, 1, 0), (−1, 2, 0), (−1, 3, −1), (0, −1, 1), (0, 1, 0), (2, −2, −1) vert(), (−1, −1, −1) 1
483109 (−1, −1, 1), (0, −1, 0), (0, −1, 1), (0, 2, −1), (1, −1, −1), (1, −1, 0), (1, −1, 1), (1, 0, −2), vert() 1
(1, 0, −1), (1, 0, 0), (1, 0, 1)
(continued)
V. Batyrev et al.
Table 2.9 (continued)
ID supp(FI ) vert(can ) |π1 (S )|
490478 (−2, −1, −1), (−1, 0, 0), (−1, 2, −1), (0, 1, 0), (1, −1, 1), (1, 1, 0), (1, 2, 0), (2, 0, 1), vert() 1
(3, 0, 1), (3, 1, 1), (5, 0, 2)
490481 (−1, −1, 2), (−1, −1, 3), (−1, 0, 1), (−1, 0, 2), (−1, 1, 0), (−1, 1, 1), (−1, 2, −1), (−1, 2, 0), vert() 1
(0, −1, 0), (0, −1, 1), (0, −1, 2), (2, −2, −1)
490485 (−1, 0, 0), (−1, 0, 1), (−1, 0, 2), (−1, 1, 0), (−1, 1, 1), (−1, 2, 0), (−1, 2, 1), (0, 1, −1), vert() 1
(0, 1, 0), (2, −2, −1)
490511 (−1, −1, 2), (−1, 0, 1), (−1, 1, 0), (−1, 1, 1), (−1, 2, 0), (−1, 3, 0), (0, −1, 0), (0, 1, −1), (2, −2, −1) vert(), (2, 1, 2) 1
2 On the Fine Interior of Three-Dimensional …

495687 (−2, −1, −1), (−1, 0, 0), (−1, 2, −1), (−1, 3, −1), (0, 1, 0), (0, 4, −1), (1, −1, 1), (1, 1, 0), vert() 1
(1, 2, 0), (2, 0, 1), (2, 3, 0), (3, 0, 1), (3, 1, 1), (4, 2, 1)
499287 (−2, 1, 1), (−1, −1, 1), (−1, 0, 0), (−1, 1, −1), (0, −1, 0), (0, 0, −1), (1, −1, −1), (1, −1, 0), vert(), (−1, −1, −1) 1
(1, 0, −1), (2, −1, −1), (2, −1, 0), (2, 0, −1), (3, −1, −1)
499291 (−2, 1, 1), (−1, −1, −1), (−1, −1, 0), (−1, −1, 1), (−1, 0, −1), (−1, 0, 0), (−1, 1, −1), vert() 1
(0, −1, −1), (0, −1, 0), (0, 0, −1), (1, −1, −1), (1, −1, 0), (1, 0, −1), (2, −1, −1),
(2, −1, 0), (2, 0, −1), (3, −1, −1)
499470 (−1, −1, 2), (−1, 0, 1), (−1, 1, 0), (−1, 1, 1), (−1, 2, −1), (−1, 2, 0), (−1, 3, −1), (−1, 3, 0), vert() 1
(0, −1, 0), (0, 1, −1), (2, −2, −1)
(continued)
39
 40

Table 2.9 (continued)

ID supp(FI ) vert(can ) |π1 (S )|
501298 (−2, −2, −1), (−1, −1, 0), (−1, 0, 1), (−1, 2, 2), (0, −1, 0), (0, 1, 1), (1, −1, −1), (1, 0, 0), vert() 1
(1, 2, 1), (2, −1, −1), (2, 1, 0), (3, −1, −2), (3, 0, −1), (4, −1, −2), (5, 0, −2), (6, −1, −3)
501330 (−1, −1, 1), (−1, −1, 2), (−1, 0, 0), (−1, 0, 1), (−1, 1, 0), (−1, 1, 1), (−1, 2, −1), (−1, 2, 0), vert() 1
(−1, 3, −1), (−1, 3, 0), (0, −1, 0), (0, 1, −1), (2, −2, −1)
354912 (−1, −1, 2), (−1, 0, 1), (−1, 1, 1), (−1, 2, 0), (−1, 3, 0), (0, −1, 0), (0, 1, −1), (2, −2, −1) vert() 1
372528 (−1, 0, 1), (−1, 0, 2), (−1, 1, 0), (−1, 1, 1), (−1, 2, −1), (−1, 2, 0), (0, −1, 0), (0, −1, 1), vert() 1
(0, −1, 2), (2, −2, −1)
372973 (−1, −1, 2), (−1, 0, 1), (−1, 1, 0), (−1, 1, 1), (−1, 2, 0), (−1, 3, 0), (0, −1, 0), (0, 1, −1), (2, −2, −1) vert() 1
388701 (−2, 1, 1), (−1, −1, 1), (−1, 0, 0), (−1, 1, −1), (0, −1, 0), (0, 0, −1), (1, −1, −1), (1, −1, 0), vert() 1
(1, 0, −1), (2, −1, −1), (2, −1, 0), (2, 0, −1), (3, −1, −1)
V. Batyrev et al.
2 On the Fine Interior of Three-Dimensional … 41

Then ( )FI is the 3-dimensional rational simplex

conv{(3, 1, 1, 1), (1, 3, 1, 1), (1, 1, 2, 1), (1, 1, 1, 2)}

and ( )FI ∩ M = {(1, 1, 2, 1)}.

The canonical Fano 3-tope  is the Newton polytope of the μ4 -cyclic quotient Z 
of the projective surface of degree 8 defined by the polynomial z 18 + z 28 + z 34 + z 44 = 0
in the weighted projective space P(1, 1, 2, 2), where the cyclic group μ4 acts
via (z 1 : z 2 : z 3 : z 4 ) → (z 1 : i 3 z 2 : i z 3 : i 3 z 4 ). The single interior lattice point in this
lattice simplex  corresponds to the monomial z 1 z 2 z 32 z 4 . The projective surface Z 
has two Gorenstein cyclic quotient singularities of type A3 . The minimal desingu-
larization S of Z  is a surface of general type with K 2 = 2 and fundamental
group π1 (S ) of order 2.
One can identify  with the canonical Fano 3-simplex  given as the convex
hull of

v1 := (−3, −2, −2), v2 := (1, 0, 0), v3 := (1, 3, 1), and v4 := (1, 1, 3)

(ID 547465, Fig. 2.6b, Tables 2.7, 2.8, and 2.9). The primitive inward-pointing facet
normals of the facets θ123 , θ124 , θ234 , θ134   of this simplex  are

n 1 := (−1, −1, 3), n 2 := (−1, 3, −1), n 3 := (−1, 0, 0), and n 4 := (2, −1, −1),

respectively. They satisfy the relation

n 1 + n 2 + 2n 3 + 2n 4 = 0.

To compute the canonical hull can of , we obtain supp(FI ) = {si | 1 ≤

i ≤ 9} with s1 := (−1, −1, 3), s2 := (−1, 0, 0), s3 := (−1, 0, 1), s4 := (−1, 0, 2),
s5 := (−1, 1, 0), s6 := (−1, 1, 1), s7 := (−1, 2, 0), s8 := (−1, 3, −1), and s9 :=
(2, −1, −1), which leads to can = .

Remark 37 Information about all 49 canonical Fano 3-topes with dim(FI ) = 3

can be found in Tables 2.7, 2.8, and 2.9.

2.7 Hollow 3-Topes with Non-empty Fine Interior

A lattice polytope  ⊆ MQ is called hollow if it has no interior lattice points in its

relative interior, i.e., ◦ ∩ M = ∅. By [25, Theorem 1.3], any 3-dimensional hol-
low lattice polytope can be projected to the unimodular 1-simplex, to the double
unimodular 2-simplex, or is an exceptional hollow 3-tope, whereas up to unimod-
ular transformation there exist only a finite number of these. This theorem implies
that a hollow 3-tope with non-empty Fine interior has to be exceptional because the
42 V. Batyrev et al.

unimodular 1-simplex and the double unimodular 2-simplex have empty Fine inte-
rior. Treutlein has found 9 maximal exceptional hollow polytopes, which was not
an complete list. Averkov et al. [1, 2] have found the complete list consisting of 12
maximal exceptional hollow 3-topes i (1 ≤ i ≤ 12) (Table 2.10, Fig. 2.7). Com-
putations show that exactly 9 of 12 maximal exceptional hollow 3-topes i have
non-empty Fine interior iFI (Table 2.10). Moreover, no one of these 9 polytopes
contains a proper lattice 3-subpolytope with non-empty Fine interior. Thus, there
exist exactly 9 hollow 3-topes i with non-empty Fine interior iFI .
It is remarkable that all minimal surfaces Si corresponding to these 9 hollow 3-
topes i have non-trivial fundamental group π1 (S ) of order 2, 3, or 5 (Table 2.10).
There exist exactly 5 hollow 3-topes i with 0-dimensional Fine interior iFI = {R},
where R ∈ 21 M \ M is a rational point (Table 2.10). The normal fans  i of these 5
hollow polytopes i define 5 toric Fano threefolds X  i with at worst canonical
singularities (Table 2.11). These Fano threefolds can be obtained as quotients of
Gorenstein toric Fano threefolds X  in the following 5 ways:

1. P(1, 1, 2, 4) with a μ2 -action given by

(x0 , x1 , x2 , x3 ) → (x0 , −x1 , −x2 , −x3 );

2. P3 with a μ4 -action given by

(x0 , x1 , x2 , x3 ) → (x0 , i x1 , −x2 , −i x3 );

3. {x1 x2 − x3 x4 = 0} ⊆ P(2, 1, 1, 1, 1) with a μ2 -action given by

(x0 , x1 , x2 , x3 , x4 ) → (−x0 , −x1 , −x2 , x3 , x4 );

4. P1 × P(1, 1, 2) with a μ2 -action given by

(x0 , x1 , y0 , y1 , y2 ) → (x0 , −x1 , y0 , −y1 , −y2 );

5. P1 × P1 × P1 with a μ2 -action given by

(x0 , x1 , y0 , y1 , z 0 , z 1 ) → (x0 , −x1 , y0 , −y1 , z 0 , −z 1 ).

In addition, Table 2.12 contains the support supp(iFI ) of the Fine interior iFI and
the vertices of the canonical hull ican for all 9 hollow polytopes i with non-empty
Fine interior iFI .

Acknowledgements AK is supported by EPSRC Fellowship EP/N022513/1, The Combinatorics

of Mirror Symmetry. The computations that underpin this work were performed on the Imperial
College London HPC cluster. We thank Andy Thomas, Matt Harvey, and the Imperial College
Research Computing Service team for invaluable technical assistance.
Table 2.10 12 Maximal Hollow 3-topes. Table contains: index i of the maximal hollow 3-tope i , vertices vert(i ) of i , lattice width w(i ) of i ,
dimension dim(iFI ) of Fine interior iFI , vertices vert(iFI ) of iFI , and order of fundamental group |π1 (S )| of the minimal model Si
i vert(i ) w(i ) dim(iFI ) vert(iFI ) |π1 (S )|
1 (0, 0, 0), (6, 0, 0), (3, 3, 0), (4, 0, 2) 2 −1 ∅ 1
2 (0, 0, 0), (4, 0, 0), (0, 4, 0), (2, 0, 2) 2 −1 ∅ 1
3 (0, 0, 0), (3, 0, 0), (0, 3, 0), (3, 0, 3) 3 −1 ∅ 1
4 (0, 0, 0), (4, 0, 0), (2, 4, 0), (3, 0, 2) 2 0 1/2 · (5, 1, 2) 2
5 (0, 0, 0), (2, 2, 0), (1, 1, 2), (3, −1, 2) 2 0 1/2 · (3, 1, 2) 2
2 On the Fine Interior of Three-Dimensional …

6 (0, 0, 0), (2, 2, 0), (4, 0, 0), (2, −2, 0), (3, 1, 2) 2 0 1/2 · (5, 1, 2) 2
7 (0, 0, 0), (1, 1, 0), (2, −2, 0), (3, −1, 0), (1, −1, 2), (2, 0, 2) 2 0 1/2 · (3, −1, 2) 2
8 (0, 0, 0), (1, 1, 0), (1, −1, 0), (2, 0, 0), (1, −1, 2), (2, 0, 2), 2 0 1/2 · (3, −1, 2) 2
(2, −2, 2), (3, −1, 2)
9 (0, 0, 0), (3, 0, 0), (1, 3, 0), (2, 0, 3) 3 1 (4/3, 1, 1), (5/3, 1, 1) 3
10 (0, 0, 0), (1, 2, 0), (1, −1, 0), (3, 0, 0), (2, 1, 3) 3 1 (4/3, 2/3, 1), (5/3, 1/3, 1) 3
11 (0, 0, 0), (1, 1, 0), (3, 0, 0), (2, −1, 0), (4, 1, 3), (2, 2, 3) 3 1 (5/3, 2/3, 1), (7/3, 1/3, 1) 3
12 (−1, 0, 0), (0, 1, −2), (1, 2, 1), (2, −2, −1) 3 3 (1/5, 1/5, −2/5), (2/5, 2/5, −4/5), (3/5, 3/5, −1/5), 5
(4/5, −1/5, −3/5)
43
44 V. Batyrev et al.

P4 P4

P4

P2
P2
P1
P1
P3
P2
P3 P3 P1

(a) 1. (b) 2. (c) 3.

P4 P4
P3
P5

P4
P1
P2 P1 P1
P3
P3 P2 P2

(d) 4. (e) 5. (f) 6.

P4
P5
P6
P6
P5 P8
P7
P3
P2
P4 P3 P2
P1 P1 P4

P2 P3 P1

(g) 7. (h) 8. (i) 9.

P5
P3
P6 P5

P1

P4
P2
P2 P2
P3 P3
P1 P1 P4 P4
(j) 10 . (k) 11 . (l) 12 .

Fig. 2.7 12 Maximal Hollow 3-topes. Shaded faces are occluded. The Fine interior is shown in
grey with double borders around its vertices
Table 2.11 5 Hollow 3-topes with 0-dimensional Fine Interior. Table contains: index i of the maximal hollow 3-tope i , rays of the normal fan  i
corresponding to i , ID of the canonical Fano 3-tope  such that  i ∼ =  , rays of the spanning fan  , ID of the reflexive canonical Fano 3-tope 
used to construct the Gorenstein toric Fano threefold X  to obtain the toric Fano threefold X  with at worst canonical singularities as a μ2 quotient, and
reference to the corresponding Gorenstein toric Fano threefold X  including the precise μ2 -action on Sect. 2.7
i  i (1) ID( )  (1) ID( ) X  

4 (2, −1, −3), (0, 0, 1), (0, 1, 0), (−2, −1, −1) 547354 (−2, −3, −5), (2, 1, 1), (0, 1, 0), (0, 0, 1) 547363 (i)
2 On the Fine Interior of Three-Dimensional …

5 (1, −1, 0), (1, 1, −1), (−1, 1, 2), (−1, −1, −1) 547364 (0, 0, 1), (0, 2, 1), (2, 1, 0), (−2, −3, −2) 547367 (ii)
6 (0, 0, 1), (1, 1, −2), (1, −1, −1), (−1, −1, 0), (−1, 1, −1) 544353 (1, 0, 0), (0, 1, 0), (−2, −1, 0), (1, 1, 2), (−3, −1, −2) 544357 (iii)
7 (0, 0, 1), (−1, 1, −1), (1, −1, −1), (1, 1, 0), (−1, −1, 0) 544310 (−1, −1, −2), (1, 1, 2), (−2, −1, 0), (0, 1, 0), (1, 0, 0) 544342 (iv)
8 (−1, 1, 1), (0, 0, −1), (−1, −1, 0), (1, 1, 0), (1, −1, −1), (0, 0, 1) 520134 (1, 0, 0), (0, 1, 0), (1, 1, 2), (−1, 0, 0), (0, −1, 0), 520140 (v)
(−1, −1, −2)
45
46 V. Batyrev et al.

Table 2.12 9 Hollow 3-topes with Non-empty Fine Interior. Table contains: index i of the
maximal hollow 3-tope i , support supp(iFI ) of iFI , and vertices vert(ican ) of the canonical
hull ican
i supp(iFI ) vert(ican )
4 (−2, −1, −1), (0, −1, −2), (2, −1, −3), (0, 0, 1), (0, 0, −1), (0, 1, 0) vert(i )
5 (1, −1, 0), (1, 1, −1), (0, 0, 1), (0, 0, −1), (−1, −1, −1), (−1, 1, 2) vert(i )
6 (1, 1, −2), (1, −1, −1), (−1, −1, 0), (−1, 1, −1), (0, 0, 1), (0, 0, −1) vert(i )
7 (1, 1, 0), (1, −1, −1), (−1, −1, 0), (−1, 1, −1), (0, 0, 1), (0, 0, −1) vert(i )
8 (1, 1, 0), (1, −1, −1), (−1, −1, 0), (0, 0, 1), (0, 0, −1), (−1, 1, 1) vert(i )
9 (0, −1, −1), (0, 0, 1), (3, −1, −2), (0, 1, 0), (−3, −2, −1) vert(i )
10 (−1, 2, −1), (1, 1, −1), (−1, −1, 0), (2, −1, −1), (0, 0, 1) vert(i )
11 (1, −1, 0), (0, 0, 1), (−1, −2, 1), (−1, 1, 0), (1, 2, −2) vert(i )
12 (1, 1, 1), (1, −1, 0), (−2, −1, 1), (0, 1, −2) vert(i )

References

1. Averkov, G., Krümpelmann, J., Weltge, S.: Notions of maximality for integral lattice-free
polyhedra: the case of dimension three. Math. Oper. Res. 42(4), 1035–1062 (2017)
2. Averkov, G., Wagner, C., Weismantel, R.: Maximal lattice-free polyhedra: finiteness and an
explicit description in dimension three. Math. Oper. Res. 36(4), 721–742 (2011)
3. Batyrev, V.: The stringy Euler number of Calabi-Yau hypersurfaces in toric varieties and the
Mavlyutov duality. Pure Appl. Math. Q. 13(1), 1–47 (2017)
4. Batyrev, V., Kreuzer, M.: Integral cohomology and mirror symmetry for Calabi-Yau 3-folds.
In: Mirror symmetry. V, AMS/IP Studies in Advanced Mathematics, vol. 38, pp. 255–270.
American Mathematical Society, Providence, RI (2006)
5. Batyrev, V.V.: Dual polyhedra and mirror symmetry for Calabi-Yau hypersurfaces in toric
varieties. J. Algebraic Geom. 3(3), 493–535 (1994)
6. Beck, M., Robins, S.: Computing the continuous discretely. Undergraduate Texts in Mathe-
matics, 2nd edn. Springer, New York (2015). Integer-point enumeration in polyhedra, With
illustrations by David Austin
7. Bosma, W., Cannon, J., Playoust, C.: The Magma algebra system. I. The User Language, pp.
235–265 (1997). Computational algebra and number theory (London, 1993)
8. Brown, G., Kasprzyk, A.M.: Graded Ring Database. http://www.grdb.co.uk
9. Catanese, F.: Surfaces with K 2 = pg = 1 and their period mapping. In: Algebraic Geometry
(Proc. Summer Meeting, Univ. Copenhagen, Copenhagen, 1978). Lecture Notes in Mathemat-
ics, vol. 732, pp. 1–29. Springer, Berlin (1979)
10. Catanese, F., Debarre, O.: Surfaces with K 2 = 2, pg = 1, q = 0. J. Reine Angew. Math. 395,
1–55 (1989)
11. Corti, A., Golyshev, V.: Hypergeometric equations and weighted projective spaces. Sci. China
Math. 54(8), 1577–1590 (2011)
12. Danilov, V.I., Khovanskiı̆, A.G.: Newton polyhedra and an algorithm for calculating Hodge-
Deligne numbers. Izv. Akad. Nauk SSSR Ser. Mat. 50(5), 925–945 (1986)
13. Fine, J.: Resolution and completion of algebraic varieties. Ph.D. thesis, University of Warwick
(1983)
14. Hovanskiı̆, A.G.: Newton polyhedra, and the genus of complete intersections. Funktsional.
Anal. i Prilozhen. 12(1), 51–61 (1978)
15. Ishii, S.: The minimal model theorem for divisors of toric varieties. Tohoku Math. J. 51(2),
213–226 (1999)
16. Kasprzyk, A.M.: Canonical toric Fano threefolds. Canad. J. Math. 62(6), 1293–1309 (2010)
2 On the Fine Interior of Three-Dimensional … 47

17. Kreuzer, M., Skarke, H.: Classification of reflexive polyhedra in three dimensions. Adv. Theor.
Math. Phys. 2(4), 853–871 (1998)
18. Kreuzer, M., Skarke, H.: Complete classification of reflexive polyhedra in four dimensions.
Adv. Theor. Math. Phys. 4(6), 1209–1230 (2000)
19. Kynev, V.I.: An example of a simply connected surface of general type for which the local
Torelli theorem does not hold. C. R. Acad. Bulgare Sci. 30(3), 323–325 (1977)
20. Reid, M.: Young person’s guide to canonical singularities. In: Algebraic Geometry, Bowdoin,
1985 (Brunswick, Maine, 1985), Proceedings of Symposia in Pure Mathematics, vol. 46, pp.
345–414. American Mathematical Societ, Providence, RI (1987)
21. Skarke, H.: Weight systems for toric Calabi-Yau varieties and reflexivity of Newton polyhedra.
Modern Phys. Lett. A 11(20), 1637–1652 (1996)
22. Stanley, R.P.: Decompositions of rational convex polytopes. Ann. Discrete Math. 6, 333–342
(1980)
23. Todorov, A.N.: Surfaces of general type with pg = 1 and (K , K ) = 1. I. Ann. Sci. École Norm.
Sup. 13(1, 4), 1–21 (1980)
24. Todorov, A.N.: A construction of surfaces with pg = 1, q = 0 and 2 ≤ (K 2 ) ≤ 8. Counterex-
amples of the global Torelli theorem. Invent. Math. 63(2), 287–304 (1981)
25. Treutlein, J.: 3-dimensional lattice polytopes without interior lattice points. Ph.D. thesis, Eber-
hart Karls Universität Tübingen (2010)
Chapter 3
Lattice Distances in 3-Dimensional
Quantum Jumps

Mónica Blanco

Abstract For Q a lattice polytope and x ∈ / Q a lattice point, we say that (Q, x) is
a quantum jump if conv(Q ∪ {x}) contains exactly one more lattice point than Q.
Usually this can only happen when the lattice distance between x and Q is somehow
bounded. In this paper I collect several results and information on the bound for
that distance in 3-dimensional quantum jumps, and the consequences on the distance
between the boundary of a polytope and its interior lattice points.

Keywords Lattice polytope · Lattice distance · Quantum jump · Interior points

3.1 Introduction

Throughout my research on classifying lattice 3-polytopes by their number of lattice

points [2–4] there has been a recurrent situation: suppose there is a lattice polytope Q,
and a lattice point x ∈/ Q, what can I say about x with respect to Q so that conv(Q ∪
{x}) does not contain more lattice points other than those of Q and x? Usually the
answer had to do with the distance from x to Q being bounded.
In order to explain things more formally we need to introduce notation and some
basic definitions. A lattice point is an element of Zd , and a lattice polytope is the
convex hull of finitely many lattice points. We write lattice d-polytope if the polytope
is d-dimensional. Two polytopes P and Q are equivalent, or unimodularly equivalent,
if there exists a unimodular transformation that maps one to the other. That is, if there
exists an affine map t : Rd → Rd such that t (Zd ) = Zd and t (P) = Q. The size of a
lattice polytope P ⊂ Rd is the number of lattice points in it. An affine functional f :
Rd → R is integer if f (Zd ) ⊆ Z and it is primitive if f (Zd ) = Z. The (lattice) width
of a lattice d-polytope P ⊂ Rd with respect to an integer functional f is the length
of f (P) ⊂ R, and the width of P is the minimum among those, for f non-constant.

M. Blanco (B)
Department of Mathematics, Statistics and Computation, University of Cantabria, Avda. Los
Castros s/n, 39005 Santander, Cantabria, Spain
e-mail: [email protected]

© Springer Nature Switzerland AG 2022 49

A. M. Kasprzyk and B. Nill (eds.), Interactions with Lattice Polytopes,
Springer Proceedings in Mathematics & Statistics 386,
https://doi.org/10.1007/978-3-030-98327-7_3
50 M. Blanco

Finally, for S ⊂ Rd , we denote by conv(S) and aff(S) the convex and affine hulls
of S. In particular, an affine subspace S ⊂ Rd is lattice if aff(S ∩ Zd ) = S.
Let us now introduce the main two definitions in this paper:

Definition 1 The lattice distance between a lattice hyperplane H ⊂ Rd and a lat-

tice point x ∈ Zd is dist(x, H ) := | f (x)|, where f is a primitive functional with
f (H ) = 0.

Definition 2 Let Q ⊂ Rd be a lattice polytope, not necessarily full-dimensional,

and let x ∈ Zd \ Q. We say that the pair (Q, x) is a quantum jump if
   
conv Q ∪ {x} ∩ Zd = Q ∩ Zd ∪ {x}.

More generally, let Q, R ⊂ Rd be lattice polytopes, not necessarily full-

dimensional, with Q ∩ R = ∅. We say that the pair (Q, R) is a quantum union if
     
conv Q ∪ R ∩ Zd = Q ∩ Zd ∪ R ∩ Zd .

That is, if the lattice points of conv(Q ∪ R) are either in Q or in R.

The name of quantum jump was first used by Bruns, Gubeladze, and Michałek [5].
Notice that they restrict the concept of quantum jump (Q, x) for when both Q
and conv(Q ∪ {x}) are full-dimensional and normal. Remember that a lattice d-
polytope Q is normal if, for all k ∈ N, every lattice point in k Q can be written as the
sum of k lattice points in Q.
Now, if we want to take a look at the distance of quantum jumps, we first need to
define the distance between a point and a polytope. Following Definition 1, they are
well and naturally defined the following distances:
Definition 3 1. Let Q ⊂ Rd be a lattice (d − 1)-polytope, and let x ∈ Zd \ aff(Q),
then
dist(x, Q) := dist(x, aff(Q)).

2. Let H1 , H2 ⊂ Rd be parallel lattice hyperplanes (H1 ∩ H2 = ∅), then

dist(H1 , H2 ) := dist(x, H2 ), for any x ∈ H1 .

3. Let 1 , 2 ⊂ R3 be lattice lines such that aff(1 ∪ 2 ) = R3 , then

dist(1 , 2 ) := dist(H1 , H2 ),

where H1 , H2 are the unique pair of parallel lattice hyperplanes such that i ⊂ Hi .
4. Let s1 , s2 ⊂ R3 be lattice segments such that aff(s1 ∪ s2 ) = R3 , then

dist(s1 , s2 ) := dist (aff(s1 ), aff(s2 )) .

3 Lattice Distances in 3-Dimensional Quantum Jumps 51

Q Q Q

x x x

Fig. 3.1 The three figures show different facets of a lattice polygon Q, the hyperplane they are
contained in, and the lattice point x. Only the facet in the middle figure is visible from x

Notice that the width and the distance are heavily related. In broad terms, the dis-
tance between two lower-dimensional objects R1 , R2 ⊂ Rd with aff(R1 ∪ R2 ) = Rd
is the width of conv(R1 ∪ R2 ) with respect to a specific functional that is determined
by the relative position between R1 and R2 . In general, if R := conv(R1 ∪ R2 ) ⊂ Rd
is not full-dimensional, the distance between R1 and R2 is measured in the lat-
tice aff(R) ∩ Zd ∼ = Zdim(R) . In the case of lattice segments, we call (lattice) length of
a segment the distance between its two endpoints (vertices). Notice that a lattice seg-
ment of length k has exactly k + 1 lattice points. We say that a segment is primitive
if it has length one.
Now, the distance that is not necessarily well-defined is the distance from a point
to a full-dimensional polytope. This notion will be written in terms of the distance
to the visible facets (see Fig. 3.1):
Definition 4 Let Q ⊂ Rd be a lattice d-polytope, F ⊂ Q a facet of Q and x ∈
Zd \ Q. F is visible from x if aff(F) strictly separates x from Q.

Definition 5 Let Q ⊂ Rd be a lattice d-polytope and let x ∈ Zd \ Q.

1. The minimum distance between x and Q is

dx (Q) := min {dist(x, aff(F)) | F facet visible from x} .

2. The maximum distance between x and Q is

Dx (Q) := max {dist(x, aff(F)) | F facet visible from x} .

See Fig. 3.2 for a 2-dimensional example of the maximum and minimum distances
between a point and a polygon. Notice that Dx (Q) is the height of x over Q as in [5,
Definition 4.1].
Let us see what we know about the distance of quantum jumps in each dimension.
For this, we can also think of a quantum jump as follows: any d-dimensional quantum
jump is of the form (P v , v), for P ⊂ Rd a lattice d-polytope, v ∈ vert(P) a vertex of P
and P v := conv(P ∩ Zd \ {v}). Notice that the dimension of P v can be d or d − 1.
For example, the lattice distance in quantum jumps of dimension ≤ 2 is always
one:
52 M. Blanco

Fig. 3.2 A lattice

polygon Q, a lattice point x, G
and the two facets of Q
visible from x, F and G. In Q
this case, dist(x, F) = 2
and dist(x, G) = 1. x
Hence dx (Q) = 1 F
and Dx (Q) = 2

Lemma 6 Let P be a lattice polytope of dimension d ∈ {1, 2}, and let v be a vertex
of P.
If P v is of dimension d − 1, then dist(v, P v ) = 1 and, if P v is of dimension d, we
have dv (P v ) = Dv (P v ) = 1.
To understand the idea in general: let P v be d-dimensional. For any (d − 1)-
dimensional face of P v that is visible from v, chose S an empty (d − 1)-dimensional
simplex in it. Since (P v , v) is a quantum jump, so is (S, v), which implies that the
convex hull of S and v is an empty d-simplex. Remember that an empty simplex of
dimension d is a lattice d-polytope with d + 1 vertices and such that those vertices
are its only lattice points.
In the cases of d = 1, 2, any empty simplex has to be unimodular, hence the vertex-
facet distance (lattice distance between a vertex and the only facet that does not
contain it) is always 1. In dimension 3 things get more complicated since we have
empty tetrahedra of arbitrarily high volume, and hence arbitrarily high vertex-facet
distance (e.g. Reeve tetrahedra [8]). That is, quantum jumps between a unimodular
triangle and a lattice point that is at arbitrarily high lattice distance from it.
In Sect. 3.2 of this paper I put together some information on the lattice dis-
tance of 3-dimensional quantum jumps (Q, x) that derives partially from previous
research [2–4]. We distinguish when Q is 2 or 3-dimensional:
1. If Q is 2-dimensional (Sect. 3.2.1) it so happens that the classifications of lattice 3-
polytopes of size 5 and 6 [2, 3], together with a suitable classification of lattice
polygons, give all the information there is to know about the distance from Q
to x. It can be summarized as follows:
Theorem 7 (see Corollary 14) Let Q ⊂ R3 be a lattice polygon, and let x ∈ Z3 \
aff(Q) such that (Q, x) is a quantum jump. Then, the lattice distance from x to Q
is at most 3 unless Q is a lattice triangle of width one, in which case the distance is
unbounded.
2. As a direct consequence of the results of the previous section, in Sect. 3.2.2 we
have the following result on the distance of a quantum union of lattice segments:
Theorem 8 (see Corollary 16) Let s, t ⊂ R3 be lattice segments with aff(s ∪ t) =
R3 such that (s, t) is a quantum union. Then, the lattice distance from s to t is one,
unless both s and t are primitive, in which case it is unbounded.
3 Lattice Distances in 3-Dimensional Quantum Jumps 53

3. For the case of Q being 3-dimensional (Sect. 3.2.3), Example 18 shows that there
exist quantum jumps of this type at arbitrarily high distance. We also look at
all the lattice 3-polytopes P of size 11 and width > 1 (database of [4]), and for
each vertex v of P such that P v is 3-dimensional we compute the minimum and
maximum distances from v to P v . Looking at the numbers one can easily see that
there is no hope in trying to bound these distances, having very high numbers for
both the minimum and the maximum distances.
Finally, in Sect. 3.3 I use all the information gathered in the previous section to
study the distance between the boundary of a polytope and its interior. More precisely,
for P ⊂ R3 a lattice 3-polytope we look at the distance between a lattice point or
segment in ∂ P (the boundary of P) and the inner lattice polytope of P, which
is I P := conv{int(P) ∩ Z3 }. Notice that I P together with a point (or segment) of the
boundary is always a quantum jump (or union). The definition of inner polytope also
applies to rational polytopes.
We only look at inner polytopes I P of size ≥ 3 (see Remark 20), and we separate
cases according to its dimension:
1. I P has dimension 1, that is, it is a lattice segment of length ≥ 2. In this case we
look at the distance between I P and a segment in the boundary. By Corollary 16,
this distance must be one, leading to:
Theorem 9 (see Theorem 21) The projection of P in the direction of the segment I P
is a reflexive polygon (polygon with a unique interior lattice point).
2. I P has dimension 2. In Sect. 3.3.2 we prove a specific property that a polygon
has to satisfy in order to appear as the inner polygon of a 3-dimensional lattice
polytope (see Theorem 23). Together with the results of Corollary 14 we obtain:
Corollary 10 (see Corollary 24) For I P of dimension 2 and size ≥ 12, the distance
from any boundary point of P to I P is at most 1.
3. I P has dimension 3. Again we look at the classification of lattice 3-polytopes
of size ≤ 11 and width > 1 [4], take the polytopes with 3-dimensional inner
polytope, and look at the minimum and maximum distances from any vertex
to the inner polytope. In Sect. 3.3.3 we simply collect some information on the
numbers obtained, without exploring it further. This time the values look more
promising, since the largest value that appears is a maximum distance of 6, and
in very high proportion the maximum and minimum distances are 1.
For future work one could try and complete the results on distances between a
lattice point of the boundary of a polytope, and its 2 or 3-dimensional inner polytope.
For the inner polytope of dimension 2, it is left to explore the cases when I P has up
to 11 lattice points. This seems perfectly doable with the help of the classification
of polygons of Proposition 11, together with the results of Sect. 3.2.1. On the other
hand, for I P of dimension 3, one would have to identify in the used database all the
polytopes that yield maximum and minimum distances equal to 1 and try to derive
the properties they have as opposed to those that yield larger distances. One would
have then to try and extend this to lattice 3-polytopes of size larger than 11.
54 M. Blanco

Δ2 T1 T2

o o o

F1 (k) F2 (k)

o (k, 0) o (k, 0)

F3 (k) F4 (k  , k)

o (k, 0) (−k  , 0) o (k − k  , 0)

Fig. 3.3 The complete classification of lattice polygons that do not contain a unit square

3.2 Distances in 3-Dimensional Quantum Jumps or Unions

Let Q ⊂ Z3 be a lattice polytope and let x ∈ Z3 \ Q. We study the distance between x

and Q, provided that (Q, x) is a quantum jump.

3.2.1 Quantum Jumps ( Q, x) with Q of Dimension 2

We first see at what distance can a lattice point be from a lattice polygon, so that
they form a quantum jump. For this, we first classify lattice polygons in a way that
is suitable distance-wise. In the following lemma, we call unit square any lattice
polygon unimodularly equivalent to [0, 1]2 .
Proposition 11 Let Q ⊂ R2 be a lattice polygon. Then Q either contains a unit
square or is equivalent to one of the following configurations:
1. 2 , the unimodular triangle;
2. T1 := conv{(1, 0), (0, 1), (−1, −1)}, the unique terminal triangle;
3. T2 := conv{(2, 0), (0, 1), (−1, −2)}, a clean triangle with three non-collinear
interior lattice points;
4. F1 (k) := conv{(0, 0), (0, 1), (k, 0)}, for k ≥ 2;
5. F2 (k) := conv{(0, 1), (0, −1), (k, 0)}, for k ≥ 2;
6. F3 (k) := conv{(−1, −1), (0, 1), (k, 0)}, for k ≥ 2; or
7. F4 (k  , k) := conv{(0, 1), (0, −1), (−k  , 0), (k − k  , 0)}, for 0 < k  < k.
See Fig. 3.3 for a depiction of the polygons of Proposition 11. For its proof, let us
first establish the following notation.
3 Lattice Distances in 3-Dimensional Quantum Jumps 55

Fig. 3.4 The situation in the y

proof of Proposition 11
when Q contains three
collinear lattice points in a
facet
(−1, 1) (1, 1)

T x

Remark 12 Let P ⊂ Rd be a polytope, and let R  P. If a point x ∈ Rd is not a

point of P, then P ∩ C x (R) = ∅, where C x (R) := x − R≥0 R − x . This fact fol-
lows trivially from convexity of polytopes. We use it whenever we want to determine
a polytope P, and we know R a subset of P and x a point not in P.

Proof (Proof of Proposition 11) For Q with 3 lattice points, we have Q ∼ = 2 . If Q

has 4 lattice points, then Q ∼= T1 , Q ∼
= F1 (2) or Q is equivalent to the unit square.
So assume for the rest of the proof that Q has size at least 5. In particular we know
that Q has 3 collinear lattice points and we can assume, without loss of generality,
that these are (−1, 0), (0, 0) and (1, 0).
If Q contains a unit square, we have finished. Assume for the rest of the proof
that Q does not contain a unit square. That is, we have Q a lattice polygon of
size ≥ 5, containing the lattice points (−1, 0), (0, 0) and (1, 0), and not containing
a unit square. Since Q is 2-dimensional, it has some lattice point outside of the
line  := {y = 0} and, by Lemma 6, we can choose one in either + := {y = 1}
or − := {y = −1}. Without loss of generality, we assume that the point (0, 1) ∈ +
is in Q. That is, the triangle T := conv{(−1, 0), (1, 0), (0, 1)} ⊂ Q.
Let us now distinguish the cases according to whether Q has three collinear lattice
points in a facet or not.

1. Suppose the three collinear points (−1, 0), (0, 0) and (1, 0) are in a facet of Q.
Then Q ⊂ {y ≥ 0}. Since Q does not contain a unit square, the points (−1, 1)
and (1, 1) cannot be in Q. Moreover, by Remark 12 this implies that no point in
the affine cones C(−1,1) (T ) and C(1,1) (T ) is in Q. See Fig. 3.4. That means that
the only lattice points that can lie in Q \ T are in the following sets:
   
A := (i, 0), i ∈ Z \ [−1, 1] , B := (0, j), j ∈ Z, j ≥ 2

Q can contain points of A or points of B, but in order for (1, 1) and (−1, 1) not
to be in Q, it cannot contain points of A and B at the same time. Adding points
of A to T gives rise to polygons of the type F1 (k), and adding points of B gives
rise to F2 (k).
2. If no facet of Q contains three collinear points, then the origin, which is in the
relative interior of the segment conv{(−1, 0), (1, 0)} ⊂ Q, must be an interior
point of Q. So Q must contain some lattice point in {y < 0} and, by Lemma 6,
56 M. Blanco

Fig. 3.5 The situation in the y

proof of Proposition 11
when Q does not contains T
three collinear lattice points
in a facet o x

(−3, −1) (−2, −1) (0, −1)

it contains some point of − . Let us denote this point by p− . We can assume

without loss of generality that (−2, 0), (2, 0) ∈ / conv(T ∪ { p− }), or else we
can simply choose a different triple of collinear lattice points in {y = 0}. That
is, p− ∈ {±3, ±2, ±1, 0} × {−1}. By symmetry of the already established points
with respect to the line {x = 0}, we can assume that p− ∈ {x ≤ 0}. Also we know
that p− = (−1, −1) since Q does not contain a unit square. The three remain-
ing possibilities p− ∈ {(−3, −1), (−2, −1), (0, −1)} are depicted in Fig. 3.5.
Let T  := conv(T ∪ { p− }) ⊆ Q, and let us study the three options for p− .

a. p− = (−3, −1). We can apply the unimodular transformation (x, y) → (x −

y + 1, y) so that T  is mapped to F3 (2) and assume now that F3 (2) ⊆ Q. See
the left-most picture in Fig. 3.6. The lattice points (−1, 0), (0, −1) and (1, 1)
cannot lie in Q, or else it would contain a unit square. This already implies that
the cones C(−1,0) (F3 (2)), C(0,−1) (F3 (2)) and C(1,1) (F3 (2)) do not intersect Q.
With that, the only lattice points that can be in Q are the points (r, 0), with r >
2. We can take as many as wanted and this gives rise to configurations F3 (k).
b. p− = (−2, −1). Again we apply the same unimodular transformation so
that T  is, in this case, mapped to F2 (2) ⊆ Q. In order for Q not to have unit
squares, no point in {−1, 1}2 can lie in Q which, after removing the corre-
sponding lattice cones, leaves the following possibilities for further points
of Q:
A := {(−1, 2)}, B := {(−1, −2)},

C := {(r, 0), r ∈ Z, r ≥ 3}, D := {(s, 0), s ∈ Z, s ≤ −1}

The points in A and B cannot be in Q at the same time, and each gives rise
to a configuration equivalent to T2 . The points in C or D cannot be in Q at
the same time as the points in A or B. If Q has points of D, we have configu-
rations F4 (k  , k) and, if Q only has points of C we get configurations F2 (k).
c. p− = (0, −1). In this case, T  = F4 (1, 2). After excluding the points in the
cones with apex in {−1, 1}2 , Q can have other lattice points in:

A := {(r, 0), r ∈ Z \ [−1, 1]}, B := {(0, s), s ∈ Z \ [−1, 1]}

Q cannot have points of A and B at the same time, and adding to T  points
of either A or B gives rise to configurations equivalent to F4 (k  , k). 
3 Lattice Distances in 3-Dimensional Quantum Jumps 57

y y y

(−1, 2) (0, s)

(r, 0) x (s, 0) (r, 0) x (r, 0) x

(−1, −2)

Fig. 3.6 The three possible polygons T  in the proof of Proposition 11. In each figure, the dark
gray area is the polygon T  (or equivalent). The light gray area is the union of the cones that do not
intersect Q. Black dots are lattice points of T  , crosses are lattice points that cannot be in Q, and
white dots are the possible lattice points of Q \ T 

Let us now take that classification and see what are the conditions on the coordi-
nates of a lattice point x ∈ Z3 so that a polygon Q and x ∈/ aff(Q) form a quantum
jump.

Lemma 13 Let Q ⊂ R2 × {0} be a lattice polygon and let x = (a, b, c) ∈ Z3 be a

lattice point with c = 0 and such that (Q, x) is a quantum jump. Then:

1. if Q = 2 , then at least one of the following happens:

i. a ≡ 1 (mod c) and gcd(b, c) = 1;
ii. b ≡ 1 (mod c) and gcd(a, c) = 1;
iii. a + b ≡ 0 (mod c) and gcd(a, c) = 1;
2. if Q contains a unit square, then c = ±1;
3. if Q = T1 , then c = ±1, or c = ±3 and a ≡ −b ≡ ±1 mod 3;
4. if Q = F1 (k), for k ≥ 2, then b ≡ 1 mod c and gcd(a, c) = 1;
5. if Q = F3 (k), for k ≥ 2, then c = ±1; or
6. if Q = T2 , F2 (k), F4 (k  , k), for k > k  > 0, then c = ±1, or c = ±2 and a ≡
b ≡ 1 mod 2.

For the purpose of simplifying notation in Lemma 13 and its proof, let us denote by
Q (resp. R) both the lattice polygon in R2 and its embedding Q × {0} (resp. R × {0})
in R3 .

Proof Part 1 of the statement follows from the classification of empty tetrahedra [9],
which states that a lattice tetrahedron is empty if one of the three pairs of opposite
edges are at lattice distance one. It is also required that these opposite edges are
primitive segments (gcd condition in the statement).
In each of the cases 2–6, we choose a subpolygon R of Q of size 4 or 5:
2. R is the unit square in Q;
3. R := Q of size 4;
4. R := F1 (2) ⊆ Q of size 4;
5. R := F3 (2) ⊆ Q, of size 5;
58 M. Blanco

6. R := F2 (2) ⊆ Q or R := F4 (1, 2) ⊆ Q, of size 5.

Since (Q, x) is a quantum jump, and R ⊆ Q, so is (R, x). That is, the poly-
tope P := conv(R ∪ {x}) is of size 5 or 6. Let us find the possible equivalences of P
in the classification of lattice 3-polytopes of size 5 or 6 [2, 3]. Notice that in all the
cases, P is a pyramid over a known polygon with apex x, so it suffices to find these
in the mentioned classification:
2. P ∼ = conv{(0, 0, 0), (1, 0, 0), (0, 1, 0), (1, 1, 0), (0, 0, 1)} (the unique configu-
ration of signature (2, 2) in [2]);
3. P ∼ = conv{(0, 0, 0), (1, 0, 0), (0, 1, 0), (−1, −1, 0), (0, 0, 1)} or
P∼ = conv{(0, 0, 0), (1, 0, 0), (0, 1, 0), (−1, −1, 0), (1, 2, 3)} (the two configu-
rations of signature (3, 1) in [2]);
4. P ∼ = conv{(0, 0, 0), (1, 0, 0), (−1, 0, 0), (0, 1, 0), ( p, q, 1)}, with gcd( p, q) =
1 (the configurations of signature (2, 1) in [2]);
5. P ∼ = conv{(−1, −1, 0), (1, 0, 0), (0, 0, 0), (0, 1, 0), (0, 2, 0), (0, 0, 1)} (a pyra-
mid of width one in [3]);
6. P ∼ = conv{(−1, 0, 0), (1, 0, 0), (0, 0, 0), (0, 1, 0), (0, 2, 0), (0, 0, 1)},
P∼ = conv{(−1, 0, 0), (1, 0, 0), (0, 0, 0), (0, 1, 0), (0, 2, 0), (1, 1, 2)},
P∼ = conv{(−1, 1, 0), (1, 1, 0), (0, 0, 0), (0, 1, 0), (0, 2, 0), (0, 0, 1)}, or
P∼ = conv{(−1, 1, 0), (1, 1, 0), (0, 0, 0), (0, 1, 0), (0, 2, 0), (1, 0, 2)} (two pyra-
mids of width one, and configurations A.1 and A.2 in [3]).
It is left to the reader to see that adding all the lattice points in Q \ R does not
put any further restrictions on the coordinates of x. That is, for each value of x so
that (R, x) is a quantum jump, we also have that (Q, x) is a quantum jump. Notice
that the different possibilities for the values of a, b and c that appear in the statement
in each of the cases, appear by applying to P all the unimodular transformations
in R3 that are automorphisms of Q. 

In terms of the distance from x to Q, which in Lemma 13 is the value |c|, we have
the following result.

Corollary 14 Let Q ⊂ R3 be a lattice polygon and let x ∈ Z3 \ aff(Q) be a lattice

point such that (Q, x) is a quantum jump. Then exactly one of the following happens:
1. Q contains a unit square or Q = ∼ F3 (k), and dist(x, Q) = 1;
2. Q ∼
= T2 , F2 (k) or F4 (k  , k), and dist(x, Q) = 1 or 2;
3. Q ∼
= T1 and dist(x, Q) = 1 or 3; or
4. Q ∼
= 2 or F1 (k), and the distance from x to Q is unbounded.
The polygons 2 , Ti and Fi (k) are as in Proposition 11, for 0 < k  < k.

Notice that the only cases when the distance is unbounded are Q ∼
= 2 or Q ∼
=
F1 (k), that is, when Q is a triangle of width one.
3 Lattice Distances in 3-Dimensional Quantum Jumps 59

3.2.2 Quantum Unions of Lattice Segments

The two cases where the distance is unbounded in the previous section also have in
common that all the lattice points are along two lattice segments. Let us think about
them as quantum unions of lattice segments.

Remark 15 1. In the case of Q = 2 , we have that (Q, x) is a quantum jump

if P := conv(Q ∪ {x}) is an empty tetrahedron. We can also write it as P =
conv(s1 ∪ s2 ), with (s1 , s2 ) a quantum union of primitive segments, where s1 is
an edge from x and one of the three vertices of 2 , and s2 is the opposite edge.
Notice that in this case there are three possible choices for the pair of primitive
segments.
2. In the case Q = F1 (k), we have that (Q, x) is a quantum jump if (s1 (k), s2 ) is
a quantum union between the lattice segment s1 (k) := conv{(0, 0, 0), (k, 0, 0)}
and the primitive segment s2 := conv{(0, 1, 0), x}.

Cases 1 and 4 of Lemma 13, reformulated in terms of the distance between seg-
ments that form a quantum union, are as follows:

Corollary 16 Let s, t ⊂ R3 be lattice segments such that aff(s ∪ t) = R3 and such

that (s, t) is a quantum union. Then:
1. if one of s or t is not primitive, then dist(s, t) = 1;
2. if both s and t are primitive, the distance dist(s, t) can be arbitrarily high, but
one of the following distances must be one:
   
dist(s, t), dist conv{s1 , t1 }, conv{s2 , t2 } , dist conv{s1 , t2 }, conv{s2 , t1 }

where si , ti ∈ Z3 are the end-points of s and t, respectively.

3.2.3 Quantum Jumps ( Q, x) with Q of Dimension 3

For the case when Q ⊂ R3 is a lattice 3-polytope, remember that we defined the
distance from a point x ∈ Z3 \ Q to Q in terms of the distance to the facets of Q
that are visible from x (Definition 5). In particular, one can study the distance from x
to Q, for (Q, x) a quantum jump, by combining the results of the previous section
on the facets of Q that are visible from x.

Remark 17 Let Q ⊂ R3 be a lattice 3-polytope. For each facet F of Q, let HF− be

the open halfspace from which the facet F is visible, and denote by HF+ = R3 \ HF−
the closed halfspace with Q ⊂ HF+ . Then subdivide R3 \ Q into the regions
 
RI := HF− ∩ HF+ , I = ∅,
F∈I F ∈I
/
60 M. Blanco

Fig. 3.7 The subdivision of

the space of Remark 17 for a 14 124 12
lattice polygon. The facets of
the polygon are labeled 1 to
4 and each semi-open region
of the space is labeled by the 134 1
facets that are visible from it. 4 4 1
Each region can see one, two
or three facets 3 2 2
34

3 23

so that, for all x ∈ RI , the facets of Q that are visible from x are exactly those of I.
Notice that the closures of the regions RI are rational polytopes or polyhedra. See
Fig. 3.7 for a 2-dimensional example of this subdivision of the space.
Let RI be one of those regions and suppose that x ∈ RI ∩ Z3 is such that (Q, x)
is a quantum jump. We can have three different types of situations.
1. If RI is bounded (or if RI ∩ Z3 is finite), the distance of x to Q is automatically
bounded.
2. If RI ∩ Z3 has infinitely many points (in particular RI is unbounded), and some
facet of I is not a triangle of width one, then the distance of x to Q is bounded
by the results of the previous section (Corollary 14).
3. Finally, if RI ∩ Z3 has infinitely many points and all the facets in I are triangles
of width one, the distance from x to Q may not be bounded. Notice that, even in
this last case the distance from x to Q could still be bounded by combining the
restrictions given for the coordinates of x as in parts 1 and 4 of Lemma 13 for all
the different facets of I.

For instance, we can find arbitrarily high distance in these types of quantum jumps.

Example 18 Let h ∈ Z, h > 0. Let:

1. x = (0, 0, 0);
2. F := conv{(1, 0, h), (0, 1, h), (1, 1, h)}, a unimodular triangle in {z = h};
3. Q ⊂ R3 be a lattice 3-polytope such that Q ⊂ C, for C := R≥0 (F) = {xh ≤
z} ∩ {yh ≤ z} ∩ {(x + y)h ≥ z} (the triangular cone of F with apex at the origin)
and such that x ∈
/ Q, F ⊂ Q.
Then F is a facet of Q, it is the only facet that is visible from x, and Q and x
are such that (Q, x) is a quantum jump with Dx (Q) = dx (Q) = dist(x, F) = h.
If moreover Q is contained in C x := C ∩ {x ≤ 1} or C y := C ∩ {y ≤ 1}, then Q
and conv(Q ∪ {x}) are polytopes of width one.
3 Lattice Distances in 3-Dimensional Quantum Jumps 61

Fig. 3.8 The regions C and C y of Example 18

We can also choose Q of size n, for any n ≥ 4, since the regions C, C x or C y have
infinitely many lattice points. See Fig. 3.8 for a depiction of the polyhedral regions C
and C y .

Even though it is clear that arbitrarily bad examples can occur, how often does
this happen? Are they rare or does the general picture look bad? For this, we look at
our database of lattice 3-polytopes of size ≤ 11 and width > 1 [4]. For each of those
polytopes P ⊂ R3 and for each v ∈ vert(P) such that P v is full dimensional, we look
at the distance in the quantum jump (P v , v). Notice that our database contains all the
information on the types of quantum jumps when P v is of size ≤ 10 and extends to
a polytope P of width > 1. That is, of size ≤ 10, we do not have the information on
polytopes of width one that extend to polytopes of width one, which are infinitely
many for each size (and no enumeration exists).
For each quantum jump (P v , v) we compute the values dv (P v ) and Dv (P v ) and
store the following vectors:
 
1. d P := dv (P v ) v∈vert(P),P v full-dimensional
 
2. D P := Dv (P v ) v∈vert(P),P v full-dimensional
We separate the 216, 453 polytopes of our database in three different groups. Notice
that the entries of each vector d P and D P are positive integers.
1. d P = (1, 1, . . . , 1) = D P . This is the best case scenario we can find, since every
vertex v of P, with P v full-dimensional, is at distance one from all the facets
of P v that are visible from v. However, only 5,796 polytopes (about 2.7%) fall
into this category.
2. d P = (1, 1, . . . , 1), D P = (1, 1, . . . , 1). In this case, things are not as nice, but
we still have that every vertex v of P, with P v full-dimensional, is at distance one
62 M. Blanco

from at least one facet of P v that is visible from v. In this category we have 77,443
polytopes (∼35.8%).
3. d P , D P = (1, 1, . . . , 1). This is the worst case we can have, in which some
vertex v of P, with P v full-dimensional, is at distance larger than one from all
the facets of P v that are visible form v. This is the case for most of the polytopes
in our database: 133,214 polytopes, or ∼61.5% of the total.
In terms of the magnitudes of the entries, we have that the largest entries in the
vectors d P and D P , for each n the size of P, are:

n 5 6 7 8 9 10 11
max dv (P v ) 5 7 13 19 25 31 37
max Dv (P v ) 7 13 19 25 31 37 43

Notice that the maximum values for dv (P v ) and Dv (P v ), for P of size n, are 6(n −
5) + 1 and 6(n − 4) + 1, respectively (for n = 5 in the first case). This has to do with
the fact that, as h grows (see Example 18) we need more lattice points to construct a
polytope of width > 1 that yields a vertex at distance h.
The average values of the dv (I P ) and Dv (I P ) are, respectively, 1.42 and 3.35.

Remark 19 If we were to follow the lines of Sect. 3.2.1, we would want to have,
in this section, an irredundant list of lattice 3-polytopes Q, and the maximum and
minimum distances a point x can be from Q, for (Q, x) a quantum jump.
However, we need to consider that we have 216,453 polytopes and that, for each of
those polytopes P and each vertex v of P we have a different polytope P v . Organizing
the information on the distances with no redundancies among the P v does not seem
to be worth undertaking, in light of the distances that appear and the arguments made.

3.3 Distance from the Boundary to the Inner Polytope

Let P ⊂ R3 be a lattice 3-polytope with I P = ∅.

Remark 20 For I P of size 1 or 2, the classification of lattice 3-polytopes with 1

and 2 interior lattice points was completed, respectively, by Kasprzyk [7] and by
Balletti and Kasprzyk [1].
1. The 3-dimensional distances that can be measured in the case of I P consisting of
one lattice point are the distances between this point and the facets of P.
2. In the case of I P having two lattice points, we would have to look at the distance
between I P and a non-coplanar lattice segment in the boundary.
In these two situations the distance is a priori unbounded if we look at it locally:
we can have a quantum jump between a unimodular triangle and a point in the first
case, and a quantum union of primitive segments in the second, at arbitrarily high
distance (see Corollaries 14 and 16). There will be a bound following from the fact
3 Lattice Distances in 3-Dimensional Quantum Jumps 63

Fig. 3.9 A polytope P

having inner polytope I P of
dimension 1
IP

that there are only finitely many lattice 3-polytopes with 1 and 2 interior lattice points,
but the author does not believe it is worth exploring the more than 23 million such
polytopes.

So let P be such that I P has size at least three. Let S ⊂ ∂ P be a lattice point
or primitive segment in the boundary of P, we look at the distance between S
and I P , relying on the fact that (I P , S) is a quantum jump (or union). This hap-
pens because conv(I P ∪ S) \ S ⊂ int(P), and the only interior lattice points of P
are those of I P .

3.3.1 Inner Polytope of Dimension 1

If I P is a lattice segment (see Fig. 3.9), and since I P has size at least 3, by Corol-
lary 16, the distance from I P to any lattice segment in the boundary must be one. A
consequence of this is the following result1 :

Theorem 21 (Averkov–Balletti–Blanco–Nill–Soprunov) Let P ⊂ R3 be a lattice 3-

polytope with I P a lattice segment of lattice length k (k + 1 collinear lattice points),
for k ≥ 2. If π : R3 → R2 is the lattice projection that maps I P to the origin
then π(P) is a reflexive polygon.

Proof Since k > 0, the projection π is well defined and unique, up to unimodular
transformation. Because the k + 1 collinear lattice points are in the interior of P,
their projection, i. e. the origin, is an interior point of π(P). Let e be an edge of π(P),
then there exists a lattice segment e in the boundary of P such that π(e ) = e. Take
the following polytope Re := conv(I P ∪ e ) ⊂ P. Since e ⊂ ∂ P and I P ⊂ int(P),
then Re cannot contain any extra lattice points. That is, it is the quantum union
of two lattice segments. By Corollary 16, and since I P is not primitive, the distance
between I P and e must be one. In the projection, this directly implies that the distance
from the edge e and the origin (the respective projections of the segments) is one.
Hence π(P) is reflexive. 

1Discussed in the Oberwolfach mini-workshop Lattice polytopes: Methods, Advances and Appli-
cations, September 2017.
64 M. Blanco

This result can help, for example, in the full classification of lattice 3-polytopes P
with I P a lattice segment. The projection has 16 possibilities: the 16 reflexive poly-
gons. For one such Q fixed, all the lattice points in P must be in π −1 (Q).

3.3.2 Inner Polytope of Dimension 2

For P having inner polytope I P of dimension 2 (see Fig. 3.10), our main result resides
in proving a specific property that a polygon must have so that it can actually appear
as the inner polytope of a lattice 3-polytope. For this, let us introduce the concept
of front:
Definition 22 Let Q ⊂ Z2 be a lattice polygon and let v be a vertex of Q. A front
of Q from v is a facet of the polygon Q v := conv(Q \ {v} ∩ Z2 ) that is visible from v.
See Fig. 3.11 for an example of the fronts of a polygon.
Theorem 23 If P is a lattice 3-polytope with I P of dimension 2, then the fronts of I P
have length ≤ 8.

Proof Let F ⊂ I P be the longest front of I P , of length  > 0. We can assume with-
out loss of generality that I P ⊂ {z = 0},  ≥ 3, that v := (0, 1, 0) is a vertex of I P

Fig. 3.10 A polytope P

having inner polytope I P of
dimension 2
IP
P

Fig. 3.11 Four copies of a

polygon Q, each showing
the fronts of Q from a
Q Q
specific vertex

Q Q
3 Lattice Distances in 3-Dimensional Quantum Jumps 65

Fig. 3.12 The setting of the y

proof of Theorem 23

r1 r2 ()
v0
v
T
x
o (, 0)

and F := conv{(0, 0, 0), (, 0, 0)} (that is, F is a front of I P from v). In particular, we
have that T := F1 () × {0} = conv(F ∪ {v}) ⊆ I P (for F1 () as in Proposition 11)
and that the only lattice points of I P in {y ≥ 0} are those of F and v. We need to
prove that  ≤ 8.
The intersection of P with the plane {z = 0} is the rational polygon P0 := P ∩
{z = 0}. We have that I P = conv(relint(P0 ) ∩ Z3 ). That is, the inner polytope of P
coincides with the relative inner polygon of P0 . For now let us identify R2 × {0}
and R2 in the trivial way, so from now on we simply say interior of P0 for the
relative interior of it embedded in the space R3 .
The vertex v = (0, 1) of I P is an interior point of P0 , so for any line passing
through v there must be a vertex of P0 in each of the open halfspaces determined by
this line. In particular, there must be a vertex of P0 in the open halfspace {y > 1}. Let
us denote this vertex by v0 and consider the rational polygon T  := conv(T ∪ {v0 }).
Since T ⊆ I P ⊂ int(P0 ) and v0 ∈ ∂ P0 , then T  \ {v0 } ⊂ int(P0 ). That is, the only
lattice points of T  \ {v0 } are those of T . In particular, (−1, 1), (1, 1) ∈ / T  \ {v0 },
which implies that v0 ∈ / C(−1,1) (T ) ∪ C(1,1) (T ) (see Remark 12).
This in turn implies that v0 must lie in the open rational triangle R determined
by the hyperplanes {y = 1}, r1 := aff{(0, 0), (−1, 1)} = {x + y = 0} and r2 () :=
aff{(, 0), (1, 1)} = {x + ( − 1)y = } (see Fig. 3.12).
That is,
  

− 
v0 ∈ R = int conv (−1, 1), (1, 1), , ,
−2 −2

which is well defined for  ≥ 3.

Observe that R ⊂ R3 for all  > 3, and that R3 ∩ Z2 = ∅. That is, there is no
lattice point in R . In particular, v0 ∈
/ Z3 , and the only possibility is that v0 is the
intersection of a primitive segment uw := conv{u, w} ⊂ P with the plane {z = 0},
with neither u nor w in this plane (and one in each of the halfspaces). This segment uw
is contained in an edge of P, although it is not necessarily equal to it.
In order to find out more about the coordinates of v0 , we need to know the dis-
tances du := dist(u, H ) and dw := dist(w, H ) in the full-dimensional polytope P,
for H := aff{P0 } = {z = 0}.
For this, let us look at a 3-dimensional proper subpolytope of P:
66 M. Blanco

K := conv{(0, 0, 0), (1, 0, 0), (2, 0, 0), (0, 1, 0), u, w} ⊂ P

So far, the information we have is that K = conv(K 0 ∪ {u, w}), for K 0 := K ∩ {z =

0} = conv{o, (2, 0, 0), (0, 1, 0), v0 } and u and w lying one in {z > 0} and the other
in {z < 0}, the edge uw being primitive and cutting the plane {z = 0} at the rational
point v0 ∈ R . Let us prove the following properties of K :
1. K has size 6. On one hand, K ∩ {z = 0} = K 0 does not contain more lattice points
other than (0, 0, 0), (1, 0, 0), (2, 0, 0) and (0, 1, 0), since K 0 ⊂ T  and T  ∩ Z3 =
T ∩ Z3 . On the other hand, if K contains an extra lattice point other than those
four and u or w, this lattice point would have to lie outside of uw (which is a
primitive edge) and outside of {z = 0}. Since K \ uw ⊂ int(P), this would be an
interior lattice point of P outside of the plane {z = 0}, which is impossible by
hypothesis.
2. K has width > 1. Let f : R3 → R be a linear primitive functional. If f is not
constant in the line {y = 0 = z}, then the width of K with respect to f is > 1
since f will take three different values in the points (0, 0, 0), (1, 0, 0), (2, 0, 0) ∈
K . Take now f to be constant in that line. Then the width of K with respect to f
is the width of π(K ) with respect to f  , for π the lattice projection π : R3 →
R2 , π(x, y, z) = (y, z), and f  the primitive functional f  : R2 → R, f  (y, z) =
f (π −1 (y, z)). Notice that f  is well defined because f is constant in the fibers
of the projection. But under this projection, the point (0, 1) is an interior point
of π(K ), hence the width of π(K ) with respect to any functional is > 1. See the
picture on the right in Fig. 3.13.
That is, K is a lattice 3-polytope of size 6 and width > 1, and the classification
of such polytopes appears in [3]. However, not all of these polytopes are a possible
candidate for K . To narrow the possibilities, we can figure out the oriented matroid
of our configuration K , since the classification in [3] is also organized according to
this combinatorial information. Remember that the oriented matroid of a finite set of
points is the information recording the affine dependencies, in particular coplanarities
and collinearities between the points (see [6] for information on oriented matroids).
For this, one extra thing that we can notice is that v cannot be a vertex of K . Suppose
otherwise, then the polytope K  := K (0,1,0) = conv{o, (1, 0, 0), (2, 0, 0), u, w} is a
polytope of size 5 with three collinear lattice points. That is, K  is the convex hull

Fig. 3.13 The intersection K 0 of K with the plane {z = 0}, and the projection of K under π
3 Lattice Distances in 3-Dimensional Quantum Jumps 67

of a lattice segment of length 2 and a primitive lattice segment uw. By Corollary 16,
the lattice distance between these two lattice segments must be one. Taking again
projection π , this is equivalent to the segment π(uw) being at distance one from the
origin. Which is impossible since (0, 1) is a lattice point strictly in between π(uw)
and the origin (see the picture in the right of Fig. 3.13). In particular, v is not a vertex
of K 0 and the picture on the left of Fig. 3.13 is not accurate.
That is, the oriented matroid of the six lattice points of K can be described as
follows:
1. four of them are vertices (o, (2, 0, 0), u and w);
2. one non-vertex point is in an edge ((1, 0, 0) = 21 ((0, 0, 0) + (2, 0, 0)));
3. the hyperplane containing the three collinear points and the other non-vertex point
({z = 0}) leaves the remaining two vertices (u and w) strictly in opposite sides of
it.
Notice that there are four different types (or orbits) of points: the endpoints of
the collinearity (o and (2, 0, 0)), the middle point of the collinearity ((1, 0, 0)),
the other two vertices (u and w), and the remaining non-vertex point (v). This
sixth point v has three different possibilities, in terms of the oriented matroid.
The three possibilities for v are: (I) it is in the relative interior of one of the
facets conv{o, u, w} or conv{(2, 0, 0), u, w}; (II) it is in the relative interior of the
triangle conv{(1, 0, 0), u, w}; or (III) it lies in the interior of one of the tetrahe-
dra conv{o, (1, 0, 0), u, w} or conv{(1, 0, 0), (2, 0, 0), u, w}. These three options
are shown in Fig. 3.14. For each of these three cases, the oriented matroid is fully
described. Without going into details of how the oriented matroids are represented and
classified in [3], one can derive that the oriented matroid of the three options (I), (II)
and (III) are, respectively, oriented matroids 3.6, 3.8 and 4.11 as encoded in [3,
Fig. 1].
In [3, Tables 8 and 9] we can see that the only lattice 3-polytopes of size 6,
width > 1 and with one of the three specified oriented matroids are B.7 (oriented
matroid 3.8), C.1 (oriented matroid 3.6), and F.13 to F.17 (oriented matroid 4.11).
The following 3 × 6 matrices have, as columns, the six lattice points of each of those
seven polytopes:

Fig. 3.14 The three possibilities (I), (II), and (III) in the proof of Theorem 23
68 M. Blanco
⎛ ⎞
0 1 0 −2 1 4
⎛ ⎞ F.13 ⎝ 0 1 0 −1 0 1 ⎠ ⎛ ⎞
0 1 0 −1 0 0 0 1 1 −2 0 2 0 1 −1 0 1 2
B.7 0 0 1 −1 0 2 ⎠
⎝ ⎛ ⎞ F.16 0 0 −2 0 3 6 ⎠
⎝
0 0 0 0 1 −1 0 0 −1 1 1 1 0 0 −1 1 1 1
F.14 0 0 −1 2 0 −2 ⎠
⎝
⎛ ⎞ 0 1 −1 1 0 −1 ⎛ ⎞
0 1 0 −1 0 −1 0 1 1 0 −1 −2
⎛ ⎞
C.1 0 0 1 −1 0 0 ⎠
⎝ 0 0 1 1 −1 −3 F.17 0 0 3 0 −2 −4 ⎠
⎝
000 0 1 2 F.15 0 0 0 2 −1 −4 ⎠
⎝ 0 0 1 1 −1 −3
0 1 0 1 −1 −3

That is, our polytope K must be equivalent to one of them, say K̃ , and let t :
R3 → R3 be any unimodular transformation that maps K to K̃ . Then t will send the
edge conv{(0, 0, 0), (2, 0, 0)} to the unique collinearity of three lattice points in K̃ ,
and v = (0, 1, 0) to the only non-vertex of the remaining lattice points.
Since unimodular transformations preserve distances, we have that {du , dw } =
{du , dw }, for du := dist(t (u), t (H )) and dw := dist(t (w), t (H )). Moreover, we can
assume without loss of generality that du ≤ dw . Then:
⎧
⎪
⎨(1, 1) if K ∼
= B.7, C.1, F.13, F.15
(du , dw ) = (1, 2) ∼ F.14, F.17
if K =
⎪
⎩
(1, 3) if K ∼
= F.16

The distance in our original coordinates of K , since H = {z = 0}, is measured on

the z-coordinate of the points. That is, let z u and z w be the respective z-coordinates of u
and w, we have that du = |z u | and dw = |z w |. Without loss of generality z u > 0 > z w :
⎧
⎪
⎨(1, −1) if K ∼
= B.7, C.1, F.13, F.15
(z u , z w ) = (1, −2) if K ∼
= F.14, F.17
⎪
⎩
(1, −3) if K ∼
= F.16

Let us now see that the denominator of the rational coordinates of v0 can only
be 2, 3 or 4:
v0 = (1 − λ)u + λw, for some λ ∈ [0, 1]

where λ is such that the z-coordinate of v0 is 0:

zu du
0 = (1 − λ)z u + λz w =⇒ λ = =
zu − zw du + dw

That is, λ ∈ 21 Z, 13 Z or 41 Z, hence

  2  2
1 2 1 1
v0 = (a, b, 0), for (a, b) ∈ Z ∪ Z ∪ Z .
2 3 4
3 Lattice Distances in 3-Dimensional Quantum Jumps 69

(−3, 3)

R3

(−1, 1) (1, 1)

(−2, 2)  
−5 5
,
3 3

R4 R5

(−1, 1) (1, 1) (−1, 1) (1, 1)

 
−3 3  
, −7 7
2 2 ,
R6 5 5 R7

(−1, 1) (1, 1) (−1, 1) (1, 1)

   
−4 4
,
3 3 R8 −9 9
, R9
7 7

(−1, 1) (1, 1) (−1, 1) (1, 1)

Fig. 3.15 The regions R3 to R9 , with the points of the lattices L 2 , L 3 and L 4 contained in them.
Large squares are points of L 2 \ Z2 , medium squares are the points of L 3 \ Z2 , and small squares
the points of L 4 \ L 2 . Black dots and crosses represent points of Z2 in ∂ R and R2 \ R , respectively

Remember also that v0 must lie in the open triangle R . To prove the statement of
the theorem it remains to see that the intersection of R with any of the lattices L 2 , L 3
 2
or L 4 , for L i := 1i Z × {0} is empty for  ≥ 9. This is true since R9 does not contain
any point of those lattices, and since R ⊆ R9 for  ≥ 9. To help the reader visualize
this we have drawn in Fig. 3.15 all the regions R3 to R9 with the possible positions
for the point v0 . 
70 M. Blanco

As a consequence of the theorem, we find that polygons that do not contain a

unit square can only be inner polytopes of lattice 3-polytopes if they have few lattice
points:

Corollary 24 Let P ⊂ R3 be a lattice 3-polytope with I P of dimension 2. Then

exactly one of the following happens:
1. IP contains a unit square and all its fronts are of length ≤ 8;
2. IP ∼
= T1 ;
3. IP ∼
= F3 (k), for 2 ≤ k ≤ 8;
4. IP ∼
= T2 , F2 (k) or F4 (k  , k), for 0 < k  < k ≤ 8; or
5. IP ∼
= 2 or I P ∼ = F1 (k), for 2 ≤ k ≤ 8.
In particular, in cases 1–4 any lattice point in the boundary of P is at distance at
most 1, 3, 1 and 2, respectively, from I P .

Proof The first part of the statement follows from Proposition 11 and Theorem 23,
considering that the longest fronts in F1 (k), F2 (k), F3 (k) and F4 (k  , k) have length k.
The second part follows from Corollary 14. 

Remark 25 In case 5, the distance of any boundary lattice point of P to I P will also
be bounded since there are only finitely many lattice 3-polytopes with those particular
polygons as inner polytopes. However, this bound can only be found globally, and
not locally, since the distance from a single lattice point to I P is a priori unbounded
(see Corollary 14).

3.3.3 Inner Polytope of Dimension 3

Now there is only left the case where I P is 3-dimensional. In particular, these are
quantum jumps of the type considered in Sect. 3.2.3, and we can apply the results
of Sect. 3.2.1 as explained in Remark 17. However, notice that we cannot use the
results of Sect. 3.3.2, since they heavily rely on I P being 2-dimensional.
We do the same as we did in Sect. 3.2.3: we check our database of lattice 3-
polytopes of size ≤ 11, width > 1 and 3-dimensional inner polytope, of which
there are 15,763 polytopes [4]. In this case, since a polytope with interior lattice
points cannot have width one, we are not losing cases by having only polytopes
of width > 1, but we only have polytopes with at most 11 lattice points in total.
Since I P is 3-dimensional, it has at least size 4, and since vert(P) ⊂ P \ I P , then I P
has at most 7 lattice points. That is, our database contains the information on I P of
size k ∈ {4, 5, 6, 7}, with P of size n ∈ {k + 4, . . . , 11} and with n − k ≤ 7 lattice
points in the boundary. In particular, both P and I P are very clean (few points in the
boundary) among the polytopes being checked.
3 Lattice Distances in 3-Dimensional Quantum Jumps 71

We consider the following vectors:

 
1. d P := dv (I P ) v∈vert(P)
 
2. D P := Dv (I P ) v∈vert(P)
In this case, the results we get are much more hopeful, since most of the polytopes P
have the distance from vertices to I P all ones:
1. 8,786 polytopes (∼55.74%) have d P = (1, 1, . . . , 1) = D P . That is, every ver-
tex v of P is at distance one from all the facets of I P that are visible from v.
2. 5,804 polytopes (∼36.82%) have d P = (1, 1, . . . , 1), D P = (1, 1, . . . , 1). In this
case, every vertex v of P is at distance one from at least one facet of I P that is
visible from v.
3. 1,173 polytopes (∼7.44%) have d P , D P = (1, 1, . . . , 1). That is, there exists a
vertex v of P that is at distance larger than one from all the facets of I P that are
visible form v.
Moreover, the values of the distances are much smaller. The maximum and minimum
values for each size n are as follows:

n 8 9 10 11
max Dv (I P ) 3 4 5 6
max dv (I P ) 3 3 3 4

and the average values of the Dv (I P ) and dv (I P ) are, respectively, 1.12 and 1.02.

Remark 26 Following the reasonings of Remark 19, in this case we would want
to have a list of polytopes Q and the maximum distance we can have a point x
so that (Q, x) is a quantum jump and there exists a polytope P such that Q = I P
and x ∈ ∂ P. From our database we are only considering 15,763 polytopes and, for
each of those polytopes P, we have exactly one polytope I P .
Putting together all the equivalent inner polytopes, we find out that there are
only 39 equivalence classes of inner polytopes. Moreover, around 9,000 polytopes in
the database (more than half) have the unimodular tetrahedron as its inner polytope.
The maximum and minimum distances for I P of size k are as follows:

k 4 5 6 7
max Dv (I P ) 4 4 5 6
max dv (I P ) 4 3 2 2

Altogether, it seems that we could find manageable bounds for the distance in
quantum jumps (I P , v), although further work is required.

Acknowledgements I would like to thank the referee for the useful comments and suggestions.
72 M. Blanco

References

1. Balletti, G., Kasprzyk, A.M.: Three-dimensional lattice polytopes with two interior lattice points
(2016). arXiv:1612.08918 [math.CO]
2. Blanco, M., Santos, F.: Lattice 3-polytopes with few lattice points. SIAM J. Discrete Math.
30(2), 669–686 (2016)
3. Blanco, M., Santos, F.: Lattice 3-polytopes with six lattice points. SIAM J. Discrete Math. 30(2),
687–717 (2016)
4. Blanco, M., Santos, F.: Enumeration of lattice 3-polytopes by their number of lattice points.
Discrete Comput. Geom. 60(3), 756–800 (2018)
5. Bruns, W., Gubeladze, J., Michałek, M.: Quantum jumps of normal polytopes. Discrete Comput.
Geom. 56(1), 181–215 (2016)
6. De Loera, J.A., Rambau, J., Santos, F.: Triangulations. Algorithms and Computation in Mathe-
matics, vol. 25. Springer, Berlin (2010). Structures for algorithms and applications
7. Kasprzyk, A.M.: Canonical toric Fano threefolds. Canad. J. Math. 62(6), 1293–1309 (2010)
8. Reeve, J.E.: On the volume of lattice polyhedra. Proc. Lond. Math. Soc. 3(7), 378–395 (1957)
9. White, G.K.: Lattice tetrahedra. Can. J. Math. 16, 389–396 (1964)
Chapter 4
Flag Matroids: Algebra and Geometry

Amanda Cameron, Rodica Dinu, Mateusz Michałek, and Tim Seynnaeve

Abstract Matroids are ubiquitous in modern combinatorics. As discovered by

Gel’fand, Goresky, MacPherson and Serganova there is a beautiful connection
between matroid theory and the geometry of Grassmannians: representable matroids
correspond to torus orbits in Grassmannians. Further, as observed by Fink and Speyer,
general matroids correspond to classes in the K -theory of Grassmannians. This yields
in particular a geometric description of the Tutte polynomial. In this review we
describe all these constructions in detail, and moreover we generalise some of them
to polymatroids. More precisely, we study the class of flag matroids and their rela-
tions to flag varieties. In this way, we obtain an analogue of the Tutte polynomial for
flag matroids.

Keywords Flag matroids · Tutte polynomial · Torus orbits · Grassmannians ·

K-theory of flag varieties

A. Cameron
Max Planck Institute for Mathematics in the Sciences, Inselstr. 22, 04103 Leipzig, Germany
e-mail: [email protected]
Department of Mathematics and Computer Science, Eindhoven University of Technology, PO
Box 513, 5600 Eindhoven, The Netherlands
R. Dinu · M. Michałek (B)
University of Konstanz, Fachbereich Mathematik und Statistik, Fach D 197, 78457 Konstanz,
Germany
e-mail: [email protected]
R. Dinu
e-mail: [email protected]
T. Seynnaeve
Mathematisches Institut, Universität Bern, Alpeneggstrasse 22, 3012 Bern, Switzerland
e-mail: [email protected]

© Springer Nature Switzerland AG 2022 73

A. M. Kasprzyk and B. Nill (eds.), Interactions with Lattice Polytopes,
Springer Proceedings in Mathematics & Statistics 386,
https://doi.org/10.1007/978-3-030-98327-7_4
74 A. Cameron et al.

4.1 Introduction

The aim of this article is to present beautiful interactions among matroids and alge-
braic varieties. Apart from discussing classical results, we focus on a special class
of polymatroids known as flag matroids. The ultimate result is a definition of a Tutte
polynomial for flag matroids. Our construction is geometric in nature and follows
the ideas of Fink and Speyer for ordinary matroids [16]. The audience we are aim-
ing at is the union of combinatorists, algebraists and algebraic geometers, not the
intersection.
Matroids are nowadays central objects in combinatorics. Just as groups abstract
the notion of symmetry, matroids abstract the notion of independence. The interplay
of matroids and geometry is in fact already a classical subject [21]. Just one of such
interactions (central for our article) is the following set of associations:

matroids/flag matroids/polymatroids → lattice polytopes → toric varieties.

We describe these constructions in detail. They allow to translate results in com-

binatorics to and from algebraic geometry. As an example we discuss two ideas due
to White:
1. combinatorics of basis covers translates to projective normality of (all maximal)
torus orbit closures in arbitrary Grassmannians—Theorem 40;
2. White’s conjecture about basis exchanges (Conjecture 38) is a statement about
quadratic generation of ideals of toric subvarieties of Grassmannians.
Although the idea to study a matroid through the associated lattice polytope is
certainly present already in the works of White and Edmonds, the importance of
this approach was only fully discovered by Gel’fand, Goresky, MacPherson and
Serganova [21]. The construction of associating a toric variety to a lattice polytope
can be found in many sources, we refer the reader e.g. to [9, 17, 49].
The object we focus on is one of the main invariants of a matroid: the Tutte
polynomial. It is an inhomogeneous polynomial in two variables. On the geometric
side it may be interpreted as a cohomology class (or a class in K -theory or in Chow
ring) in a product of two projective spaces.
The applications of algebro-geometric methods are currently flourishing. A
beautiful result of Huh confirming a conjecture by Read on unimodality of chro-
matic polynomials of graphs is based on Lefschetz theorems [28]. This led fur-
ther to a proof of the general Rota-Heron-Welsh conjecture [1]—which we state in
Theorem 93. Although the latter proof is combinatorial in nature, the authors were
inspired by geometry, in particular Lefschetz properties. We would like to stress that
the varieties and Chow rings studied by Adiprasito, Huh and Katz are not the same
as those we introduce in this article. Still, as the focus of both is related to the Tutte
polynomial it would be very interesting to know if their results can be viewed in the
setting discussed here.
4 Flag Matroids: Algebra and Geometry 75

We finish the article with a few open questions. As the construction of the Tutte
polynomial we propose is quite involved it would be very nice to know more direct,
combinatorial properties and definitions.

Note 1 Ewill always denote a finite set of cardinality n. P(E) is the set of all subsets
of E, and Ek is the set of all subsets of E of cardinality k. We use [n] as a shorthand
notation for the set {1, 2, . . . , n}. We will denote the difference of two sets X and Y
by X − Y . This does not imply that Y ⊆ X . If Y is a singleton {e}, we write X − e
instead of X − {e}.

4.2 Matroids: Combinatorics

For a comprehensive monograph on matroids we refer the reader to [43].

4.2.1 Introduction to Matroids

There exist many cryptomorphic definitions of a matroid—it can be defined in terms

of its independent sets, or its rank function, or its dependent sets, amongst others.
One of the most relevant definitions for us is that of the rank function:

Definition 2 A matroid M = (E, r ) consists of a ground set E and a rank

function r : P(E) → Z≥0 such that, for X, Y ∈ P(E), the following conditions hold:
R1. r (X ) ≤ |X |;
R2. (monotonicity) if Y ⊆ X , then r (Y ) ≤ r (X ); and
R3. (submodularity) r (X ∪ Y ) + r (X ∩ Y ) ≤ r (X ) + r (Y ).

We write r (M) for r (E). When r (X ) = |X |, we say that X is independent,

and dependent otherwise. A minimal dependent set is called a circuit. A matroid
is connected if and only if any two elements are contained in a common circuit. It
can be shown that “being contained in a common circuit” is an equivalence relation
on E; the equivalence classes are called connected components.
If |X | = r (X ) = r (M) we call X a basis of M. We can use bases to provide an
alternative set of axioms with which to define a matroid. We present this as a lemma,
but it can just as well be given as the definition. The diligent reader can check that
each set of axioms implies the other.
Lemma 3 A matroid M = (E, B) can be described by a set E and a collection of
subsets B ⊆ P(E) such that:
76 A. Cameron et al.

B1. B = ∅; and
B2. (basis exchange) if B1 , B2 ∈ B and e ∈ B1 − B2 , there exists f ∈ B2 − B1 such
that (B1 − e) ∪ f ∈ B.
A reader new to matroid theory should not be surprised by the borrowed termi-
nology from linear algebra: matroids were presented as a generalisation of linear
independence in vector spaces in the paper by Whitney [57] initiating matroid the-
ory. Matroids also have a lot in common with graphs, thus explaining even more
of the terminology used. For instance, very important matroid operations are that
of minors. These are analogous to the graph operations of the same names. As there,
deletion is very simple, while contraction requires a bit more work.

Definition 4 (Deletion and Contraction)

1. We can remove an element e of a matroid M = (E, r M ) by deleting it. This yields
a matroid M\e = (E − e, r M\e ), where r M\e (X ) = r M (X ) for all X ⊆ E − e.
2. We can also remove an element e of a matroid M = (E, r M ) by contracting it. This
gives a matroid M/e = (E − e, r M/e ) where r M/e (X ) = r M (X ∪ e) − r M ({e}) for
all X ⊆ E − e.

Remark 5 More generally, if M = (E, r M ) is a matroid and S is a subset of E, we

can define the deletion M\S (resp. contraction M/S) by deleting (resp. contracting)
the elements of S one by one. We have that r M\S (X ) = r M (X ) for all X ⊆ E − S
and r M/S (X ) = r M (X ∪ S) − r M (S) for all X ⊆ E − S.

We will now give two examples of classes of matroids which show exactly the
relationship matroids have with linear algebra and graph theory. The first one plays
a central role in our article.

Definition 6 Let V be a vector space, and φ : E → V a map that assigns to every

element in E a vector of V . For every subset X of E, define r (X ) to be the dimen-
sion of the linear span of φ(X ). We have that (E, r ) is a matroid, which we say
is representable.

Remark 7 Our definition differs slightly from the one found in literature: typically
one identifies E with φ(E). Our definition does not require φ to be injective; we can
take the same vector several times. We also note that the matroid represented by φ :
E → V only depends on the underlying map φ : E → P(V ), assuming φ(E) ⊂
V \{0}.

If V is defined over a field F, we say that M is F-representable. We can describe

the bases of a representable matroid: X ⊆ E is a matroid basis if and only if φ(X )
is a vector space basis of the linear span of φ(E).

Example 8 (The non-Pappus matroid) Here is an example of a non-representable

matroid: consider the rank-3 matroid R on [9], whose bases are all 3-element subsets
of [9] except for the following (see also Fig. 4.1):
4 Flag Matroids: Algebra and Geometry 77

Fig. 4.1 The non-Pappus 4 5 6

matroid

8
7 9

1 2 3

{1, 2, 3}, {4, 5, 6}, {1, 5, 7}, {1, 6, 8}, {2, 4, 7}, {2, 6, 9}, {3, 4, 8}, {3, 5, 9}.

If R were representable over a field F, there would be a map [9] → P2F : i → pi

such that pi , p j , pk are collinear if and only if {i, j, k} is not a basis of R. Now, the
classical Pappus’ Theorem precisely says that this is impossible: if the non-bases are
all collinear, then so are p7 , p8 , p9 .

Definition 9 Let G = (V, E) be a graph. The graphic (or cycle) matroid M of G is

formed by taking E(M) = E(G), and setting the rank of a set of edges equal to the
cardinality of the largest spanning forest contained within it.

It is easy to see that connectedness of the graph G is not equivalent to connectedness

of the matroid M(G). However, there is a correspondence with higher graph connec-
tivity. We say that a graph G is k-connected if we cannot disconnect G by removing
less than k vertices. Now, M(G) is connected if and only if G is 2-connected.

4.2.2 The Tutte Polynomial

Further matroid definitions will be given later, but we have covered enough to give
the major object of our interest in this paper, namely the Tutte polynomial. This
is the most famous matroid (and graph) invariant, and, like matroids themselves,
has multiple definitions. These will be mentioned where relevant. Here, we give
the corank-nullity formula, two terms which will be defined below.

Definition 10 Let M = (E, r ) be a matroid with ground set E and rank function r :
P(E) → Z≥0 . The Tutte polynomial of M is

TM (x, y) = (x − 1)r (M)−r (S) (y − 1)|S|−r (S) .
S⊆E

The term r (M) − r (S) is called the corank, while the term |S| − r (S) is called
the nullity. Readers familiar with matroid theory should be careful not to confuse
a mention of corank with dual rank, given the usual naming convention of dual
objects. By identifying the rank function of a matroid with the connectivity function
78 A. Cameron et al.

of a graph in an appropriate way, one can pass between this formula and the original
formulation of the Tutte polynomial which was given for graphs.

Example 11 For the (matroid of the) complete graph K 4 , there are four subsets with
three elements of rank 2 and all the other subsets with three elements have rank 3.
In this case, the Tutte polynomial is

TM(K 4 ) (x, y) = x 3 + 3x 2 + 2x + 4x y + 2y + 3y 2 + y 3 .

Readers interested in seeing what the Tutte polynomial looks like for a range of
different classes of matroids should consult [36].
The prevalence of the Tutte polynomial in the literature is due to the wide range
of applications it has. The simplest of these occurs when we evaluate the polynomial
at certain points, these being called Tutte invariants. For instance, T (1, 1) gives the
number of bases in the matroid (or the number of spanning trees in a graph). In this
way we can also count the number of independent sets in a matroid or graph, and
the number of acyclic orientations of a graph, as well as some other such quantities.
Beyond numerics, the Tutte invariants also include other well-known polynomials,
appearing in graph theory (the chromatic polynomial, concerned with graph colour-
ings; see also Theorem 93) and network theory (the flow and reliability polynomials).
Extending to further disciplines, one can find multivariate versions of the Tutte poly-
nomial which specialise to the Potts model [55] from statistical physics and the Jones
polynomial [52] from knot theory. In this paper, we will be looking at the classical
Tutte polynomial from an algebraic point of view.
We noted that there are multiple definitions of the Tutte polynomial. One is both so
useful and attractive that we would be remiss to not include it. It states that, instead of
calculating the full sum above, we can instead simply form a recurrence over minors
of our matroid, which can lead to faster calculations. Note that a coloop is an element
of E which is in every basis of M, while a loop is an element which is in no basis.

Lemma 12 ([7]) Let TM (x, y) be the Tutte polynomial of a matroid M = (E, r ).

Then the following statements hold:
1. TM (x, y) = x TM/e (x, y) if e is a coloop;
2. TM (x, y) = yTM\e (x, y) if e is a loop;
3. TM (x, y) = TM\e (x, y) + TM/e (x, y) if e is neither a loop nor a coloop.

The Tutte polynomial is in fact universal for such formulae: any formula for
matroids (or graphs) involving just deletions and contractions will be an evaluation
of the Tutte polynomial. There are numerous proofs of this in the literature, and also
extensions to related classes of objects. One such reference is [13, Sect. 4].
4 Flag Matroids: Algebra and Geometry 79

4.2.3 The Base Polytope

We will now give two more axiom systems for matroids. The first one, via base
polytopes, will play a fundamental role in this paper.
We first define what the base polytope of a matroid is: let B be the set of bases of a
matroid M = (E, r ). We work in the vector space R E = {(ri | i ∈ E)}, where ri ∈ R.
For a set U ⊆ E, eU ∈ R E is the indicator vector of U , that is, eU is the sum of the
unit vectors ei , for all i ∈ U . Note that e{i} = ei .

Definition 13 The base polytope of M is

P(M) = conv{e B | B ∈ B}.

Note that this is always a lattice polytope. Its dimension is equal to |E| minus the
number of connected components of the matroid [15, Proposition 2.4]. We also note
that the vertices of P(M) correspond to the bases of M. In particular: given P(M) ⊂
R E , we can recover M.
The following theorem gives a characterisation of which lattice polytopes appear
as the base polytope of a matroid. It can be used as an axiom system to define
matroids:

Theorem 14 ([12], see also [21, Theorem 4.1]) A polytope P ⊂ R E is the base
polytope of matroid on E if and only if the following two conditions hold:
P1. every vertex of P is a 0, 1-vector; and
P2. every edge of P is parallel to ei − e j for some i, j ∈ E.

More generally the description of faces of matroid base polytopes is provided in [15,
30]. The base polytope is a face of the independent set polytope of M, which is the
convex hull of indicator vectors of the independent sets of M.

4.2.4 Definition via Gale Orderings

We move on to another axiom system: via Gale orderings. This definition is orginally
due to Gale [20]; our formulation is based on lecture notes by Reiner [45].
Definition 15 Let ω be a linear ordering
 on E, which we will denote by ≤. Then
the dominance ordering ≤ω on Ek , also called Gale ordering, is defined as follows.
 
Let A, B ∈ Ek , where

A = {i 1 , . . . , i k }, i 1 < · · · < i k

and
B = { j1 , . . . , jk }, j1 < · · · < jk .
80 A. Cameron et al.

Then we set
A ≤ω B if and only if i 1 ≤ j1 , . . . , i k ≤ jk .
 
Theorem 16 ([20]) Let B ⊆ Ek . We have that B is the set of bases of a matroid if
and only if for every linear ordering ω on E, the collection B has a maximal element
under the Gale ordering ≤ω (i.e. there is a unique member A ∈ B such that B ≤ω A,
for all B ∈ B).

In Sect. 4.6.1, we will introduce a generalisation of matroids, called flag matroids,

via Gale orderings. This will generalise the above characterisation of matroids. In
this paper, they will arise quite naturally when generalising our geometric description
of matroids given in Sect. 4.5.

4.2.5 The Matroid Union Theorem

Next we present one of the central theorems in matroid theory.

Theorem 17 (The matroid union theorem) Let M1 , . . . , Mk be matroids on the same
ground set E with respective families of independent sets l1 , . . . , lk and rank func-
tions r1 , . . . , rk . Let

k
l := {I ⊂ E : I = Ii for Ii ∈ li }.
i=1

Then l is also a family of independent sets for a matroid, known as the union
of M1 , . . . , Mk . Further, the rank of any set A ⊂ E for the union matroid is given
by:
k
r (A) = min{|A\B| + ri (B)}.
B⊂A
i=1

The proof can be found e.g. in [43, 12.3.1]. The following corollary is essentially
due to Edmonds.
Corollary 18 Let M1 , . . . , Mk be matroids on a ground set E with rank functions
respectively r1 , . . . , rk . E can be partitioned into independent
k sets, one for each
matroid, if and only if for all subsets A ⊂ E we have |A| ≤ i=1 ri (A).

Proof The implication ⇒ is straightforward.

For the other implication let U be a matroid that is the union of M1 , . . . , Mk . We
compute the rank of E in U , applying the matroid union Theorem 17:
 

k
rU (E) = min |E| − |B| + ri (B) .
i=1
4 Flag Matroids: Algebra and Geometry 81

k
By assumption, for any B ⊂ E we have |E| − |B| + i=1 ri (B) ≥ |E|. Further,
equality holds for B = ∅. Hence, rU (E) = |E|. Thus, E is an independent set of U .
By definition it is a union of k independent sets, one in each of the Mi ’s. 

4.3 Polymatroids: Combinatorics

Consider what happens if we drop one of the rank axioms, namely that which
states r (X ) ≤ |X |. What object do we get, and what relation does it have to matroids?
This object was originally studied by Edmonds [12] (although in a different guise,
see Definition 23), and dubbed a polymatroid. The class of polymatroids includes,
naturally, the class of matroids, and is greatly important in the field of combinatorial
optimisation.

Definition 19 A polymatroid M = (E, r ) consists of a ground set E and a rank

function r : P(E) → Z≥0 . The rank function r satisfies conditions R2 (monotonicity)
and R3 (submodularity) of Definition 2, while condition R1 is relaxed to r (∅) = 0.
A polymatroid is called a k-polymatroid if all singletons have rank at most k. In
particular, a matroid is a 1-polymatroid.

Remark 20 As we assume that our rank function take only integral values, the object
we defined is sometimes referred to in the literature as a discrete polymatroid [27].

One vital difference between matroids and polymatroids is that polymatroids do

not have well-defined properties of deletion and contraction.
One problem behind this is the following. Take any element e in the ground set of
a given matroid. All the bases not containing e are bases in the deletion of e, while
all the bases containing e exactly correspond to bases in the contraction of e. There is
no such partition of bases among minors in a polymatroid. In consequence, the Tutte
polynomial is not directly applicable to polymatroids. In restricted cases, this can be
somewhat solved: this is done by Oxley and Whittle [44] for 2-polymatroids, where
the corank-nullity polynomial is still universal for a form of deletion-contraction
recurrence. In [11], the authors strengthen the notion of “deletion-contraction invari-
ant” to more general combinatorial objects via the use of coalgebras, and compare
their results to that of Oxley and Whittle. In their strengthening, the corank-nullity
polynomial is indeed still universal, and furthermore, out of the polynomials found
in [44], the corank-nullity one is optimal, under the norms used in [11].
Cameron and Fink [8] construct a version of the Tutte polynomial for all poly-
matroids which specialises to an evaluation of the classical Tutte polynomial when
applied to a matroid. This will be discussed below. In order to describe it, we first
have to explain bases and base polytopes for polymatroids.

Definition 21 Let M = (E, r ) be a polymatroid. An integer vector x ∈ Z≥0 E

is called
an independent vector if x · eU ≤ r (U ) for all U ⊆ E. If in addition x · e E = r (E),
then x is called a basis.
82 A. Cameron et al.

Analogously to the matroid case, we can give an axiom system for polymatroids in
terms of their bases.
Lemma 22 ([27, Theorem 2.3]) A nonempty finite set B ⊂ Z≥0
E
is the set of bases
of a polymatroid on E if and only if B satisfies
1. all u ∈ B have the same modulus (sum of entries); and
2. if u = (u 1 , . . . , u n ) and v = (v1 , . . . , vn ) belong to B with u i > vi then there
is j ∈ E with u j < v j such that u − ei + e j ∈ B.
Definition 23 Let M = (E, r ) be a polymatroid. Let I ⊆ Z≥0 E
be the set of indepen-
dent vectors, and B ⊆ Z≥0 be the set of bases. We have the independent set polytope,
E

which is also referred to as the extended polymatroid of r :

E P(M) = conv I = {x ∈ R≥0

E
| x · eU ≤ r (U ) for all U ⊆ E}.

This is in fact what was originally defined to be a polymatroid, by Edmonds [12].

We also have the polymatroid base polytope:

P(M) = conv B = E P(M) ∩ {x ∈ R E | x · e E = r (E)}.

As before, the base polytope is a face of the extended polymatroid. When the
polymatroid considered is a matroid, these definitions coincide exactly with those
from Sect. 4.2.3.
Theorem 14 generalises to the case of polymatroids, giving us another equivalent
definition of polymatroids in terms of their base polytopes:
Theorem 24 ([27, Theorem 3.4]) A polytope P ⊂ Rn is the base polytope of a
polymatroid on [n] if and only if the following two conditions hold:
1. every vertex of P has coordinates in Z≥0 ; and
2. every edge of P is parallel to ei − e j , for some i, j ∈ [n].
If M is a polymatroid, then the bases (resp. independent vectors) of M are precisely
the lattice points of P(M) (resp. E P(M)). The following proposition describes
which bases of M correspond to vertices of P(M).
Proposition 25 ([12], see also [27, Proposition 1.3]) Let M = ([n], r ) be a poly-
matroid and assign an ordering S to the ground set [n]. Let Si be the first i elements
according to this ordering. Every possible S corresponds (not necessarily uniquely)
to a vertex of P(M); x = x S , where x = (x1 , . . . xn ), and

xi = r (Si ) − r (Si−1 ).

In particular, a polymatroid base polytope has at most n! vertices.

We finish by slightly generalising the ideas of White [56]. As we will see later in
Theorem 40 the statement below has geometric consequences. It was proven in [46,
Corollary 46.2c] using different methods.
4 Flag Matroids: Algebra and Geometry 83

Theorem 26 Let M1 , . . . , Mk be polymatroids on a ground set E with respec-

tive polytopes P(M1 ), . . . , P(Mk ). Then every lattice point p ∈ P(M1 ) + · · · +
P(Mk ) is a sum p = s1 + · · · + sk , where each si is a lattice point of P(Mi ).
Proof Proceeding by induction on k, it is enough to prove the theorem for k = 2.
Let us choose r large enough, so that M1 and M2 are r -polymatroids. We
define a matroid M̃1 (resp. M̃2 ) on E × [r ] as follows. Let π1 : E × [r ] → E be
the projection. A subset A ⊂ E × [r ] is independent in M̃1 (resp. M̃2 ) if and only
if for every subset B ⊂ π1 (A) we have r1 (B) ≥ |A ∩ (B × [r ])| (resp. r2 (B) ≥
|A ∩ (B × [r ])|), where r1 (resp. r2 ) is the rank function of the polymatroid M1
(resp. M2 ). Intuitively, an independent set in M̃ j is an independent set I in M j where
we replace one point in E by as many points as the rank function dictates. We have
natural surjections, P( M̃ j ) → P(M j ) and P( M̃1 ) + P( M̃2 ) → P(M1 ) + P(M2 ),
coming from the projection π1 . Thus, it is enough to prove the statement for two
matroids. From now on we assume that M1 and M2 are matroids.  
p ∈ P(M
Let 1 ) + P(M2 ) be a lattice point. We know that p= i λi ti + j μjqj
with λi = 1, μ j = 1, for λi , μ j ∈ Q≥0 , and that ti (resp. q j ) are lattice points
of P(M1 ) (resp. P(M2 )). After clearing the denominators we have
 
dp = λi ti + μj q j ,
i j

 
where λi = d, μj = d and λi , μj ∈ Z≥0 .
By restricting the set E we may assume that all coordinates of p = ( p1 , . . . , pn )
are nonzero (i.e. pi ∈ {1, 2}), where we identify E with [n]. Dually, by contracting
the elements of E that belong to all bases corresponding to any ti and q j , we may
assume p = (1, . . . , 1).
We want to prove that the ground set E is covered by a basis of M1 and a basis
of M2 . Hence, by Corollary 18 it is sufficient to prove the following:
For any A ⊂ E we have |A| ≤ r M1 (A) + r M2 (A).

We define a matroid N1 (resp. N2 ) on the ground set E N := {(i, j) : i ∈ E, 1 ≤

j ≤ d}. In other words we replace any point of E by d equivalent points. A sub-
set {(i 1 , j1 ), . . . , (i s , js )} ⊂ E N is independent in N1 (resp. N2 ) if only if
1. all i q ’s are distinct, and
2. {i 1 , . . . , i s } is an independent set in M1 (resp. M2 ).
We have a natural projection π : E N → E given by forgetting the second coordinate.
We note that r N j (π −1 (A)) = r M j (A) for j = 1, 2. As the point dp is decomposable
we know that the set E N can be covered by d bases of N1 and d bases of N2 . Hence,
for any B ⊂ E N we have: |B| ≤ dr N1 (B) + dr N2 (B). Applying this to π −1 (A) we
obtain:
 
d|A| = |π −1 (A)| ≤ d · r N1 (π −1 (A)) + d · r N2 (π −1 (A)) = d · r M1 (A) + r M2 (A) .

After dividing by d we obtain the statement we wanted to prove. 

84 A. Cameron et al.

4.3.1 The Tutte Polynomial for Polymatroids

The Tutte polynomial for polymatroids is not nearly as well-studied as in the matroid
case. We gave two examples of where it was considered in certain classes of poly-
matroids. We will now go into detail about one suggestion how to construct Tutte
polynomial in full generality.
As mentioned, Cameron and Fink [8] form a polynomial having Tutte-like prop-
erties for polymatroids, which specialises to an evaluation of the Tutte polynomial
when applied only to matroids. This is a construction which takes a polytopal, lattice-
point-counting, approach as opposed to a straight combinatorial one. It is motivated
by an alternative definition of the Tutte polynomial to those we have discussed so
far.

Definition 27 Take a matroid M = (E, r ), and give E some ordering. Let B be a

basis of M.
1. We say that e ∈ E − B is externally active with respect to B if e is the smallest
element in the unique circuit contained in B ∪ e, with respect to the ordering
on E.
2. We say that e ∈ B is internally active with respect to B if e is the smallest element
in the unique cocircuit in (E − B) ∪ e.

A cocircuit is a minimal set among sets intersecting every basis. We will not be
using this notion again in the article.
We will denote the number of internally active elements with respect to B
with I (B) and the number of externally active elements by E(B). Then we have
the following result.

Theorem 28 ([53]) 
TM (x, y) = x I (B) y E(B) .
B∈B

Activity was generalised to hypergraphs by Kálmán in [29], where he proved

that a formula similar to the one above does not hold for hypergraphs. This is
due to the above sum not being independent of the edge ordering chosen, as is
the case for matroids. The one-variable specialisations are, however, consistent.
That is, T (x, 0), T (0, y) can be written in terms of activity generating functions
for hypergraphic polymatroids. In [8], the authors show that this behaviour extends
to all polymatroids given their own Tutte-like polynomial for polymatroids. Their
construction is as follows.
Let  be the standard simplex in R E of dimension equal to |E| − 1, and ∇ be
its reflection through the origin. Construct the polytope given by the Minkowski
sum P(M) + u + t∇ where M = (E, r ) is any polymatroid and u, t ∈ Z≥0 .
By [35, Theorem 7], the number of lattice points inside the polytope is a poly-
nomial in t and u, of degree dim(P(M) + u + t∇) = |E| − 1. This polynomial is
written in the form
4 Flag Matroids: Algebra and Geometry 85

u t
Q M (t, u) := #(P(M) + u + t∇) = ci j .
i, j
j i

Changing the basis of the vector space of rational polynomials gives the polynomial

Q M (x, y) = ci j (x − 1)i (y − 1) j
ij

As mentioned, this specialises to the Tutte polynomial:

Theorem 29 ([8, Theorem 3.2]) Let M = (E, r ) be a matroid. Then we have

x |E|−r (M) y r (M) x +y−1 x +y−1

Q M (x, y) = · TM ,
x +y−1 y x

as an identification of rational functions.

Formulae can be given for Q M under the polymatroid generalisation of many

standard matroid operations such as direct sum and duality, and in many cases the
results are versions of those true for the Tutte polynomial. Most importantly, this
polynomial also satisfies a form of deletion-contraction recurrence.

Theorem 30 ([8, Theorem 5.6]) Let M = (E, r ) be a polymatroid and take a ∈

E(M). Let Nk be the convex hull of { p ∈ P(M) | pa =k} for some k ∈ {0, . . . , r (M)},
and Q N the polynomial formed by replacing the M in the above definition by N.
Then

Q M (x, y) = (x − 1)Q M\a (x, y) + (y − 1)Q M/a (x, y) + Q Nk (x, y).
k

Remark 31 We mention another approach of generalising the Tutte polynomial to

polymatroids. In the proof of Theorem 26, we explained how to associate to an r -
polymatroid M on E a matroid M on E × [r ]. We could define the Tutte polynomial
of M to be simply the usual Tutte polynomial of M. Of course, the result might
depend on the chosen r . Still, it is natural to ask if there is any relation between this
construction and the one described above.
One possible strategy for getting rid of this dependence on r is as follows: There is
an action of Sr permuting the second factor of E × [r ], and the sum in Definition 10 is
constant on orbits, so the whole polynomial can be enriched to an Sr -representation.
Our hope is that as r increases, this representation stabilises in some way, removing
the dependency on r .
86 A. Cameron et al.

4.4 Flag Varieties: Geometry

In this section, we always work over the field of complex numbers.

4.4.1 Representations and Characters

We begin by fixing notation regarding the representation theory of G L n . We refer the

reader to [50, Chap. 4] for a brief introduction to the representation theory of G L n ,
or to [19] for a more detailed account.
The (polynomial) irreducible representations of G L n are in bijection with Young
diagrams λ with at most n rows. We write λ = (a1 , . . . , ak ) for the Young diagram
with rows of length a1 ≥ a2 ≥ . . . ≥ ak > 0. The associated G L n -representation is
called a Weyl module of highest weight λ, and will be denoted by Sλ V (here V refers
to the natural representation of G L n , i.e. an n dimensional vector space with the
linear G L n = G L(V ) action). The usual construction of Weyl modules goes via
Young symmetrisers, which we will not recall here. For readers not familiar with
them, it suffices for now to know that S(a) V = S a V , the a-th symmetric power of the
naturalrepresentation, and that S(1,...,1) V (where 1 appears a times) is the exterior
power a V . We will give a full description later: see after Example 33.
In this article we will be interested in the action of a maximal torus T ⊂ G L(V )
on flag varieties. Such a torus T  (C∗ )n may be identified with the diagonal non-
degenerate matrices, after fixing a basis of V . We recall that a torus T acting on any
vector space W induces a weight decomposition:

W = Wc ,
c∈Zn

where (t1 . . . , tn ) ∈ T acts on v ∈ Wc by scaling as follows:

(t1 , . . . , tn )v = t1c1 · · · tncn v.

In particular, an action of the torus T on a one-dimensional vector space C may be

identified with a lattice point in Zn . We call Zn = M the lattice of characters of T .
An element (a1 , . . . , an ) ∈ M is a character identified with the map:

T  (t1 , . . . , tn ) → t1a1 · · · tnan ∈ C∗ .

Any irreducible G L(V )-representation W = Sc1 ,...,cn V decomposes as above

under the action of T with a one-dimensional component Wc1 ,...,cn ; moreover all
other components have a lexicographically smaller weight. This explains the name
“Weyl module of highest weight (c1 , . . . , cn ).”
4 Flag Matroids: Algebra and Geometry 87

4.4.2 Grassmannians

A basic example of a flag variety is a Grassmannian G(k, V ), which as a set param-

eterises k-dimensional subspaces of an n-dimensional space V . A point in G(k, n)
can be represented by a full-rank k × n matrix A, where our k-dimensional subspace
is the row span of A. Two matrices A and B represent the same point in G(k, n) if
and only if they are the same up to elementary row operations.
G(k, n) can be realised as an algebraic variety as follows:

k
G(k, n) = G(k, V ) = {[v1 ∧ . . . ∧ vk ] ⊂ P( V )}.

Here, v1 , . . . , vk are the rows of the aforementioned matrix A, and thus a point
of G(k, n) is identified with the subspace v1 , . . . , vk . The embedding presented
above is known as the Plücker embedding and the Grassmannian is defined by
quadratic polynomials known as Plücker relations [34]. Explicitly, in coordinates,
the map associates to the matrix A the value of all k × k minors. We refer the readers
not familiar with algebraic geometry, and in particular Grassmannians, to a short
introduction in [39, Chap. 5].
The Plücker embedding may be identified with a very ample line bundle on G(k, n),
which we will denote by O(1). Other very ample line bundles on G(k, n) are the d-th
tensor powers O(d). They can be realised as a composition
 of the Plücker embedding
with the d-th Veronese map P( k V ) → P(Symd k V ).

Remark 32 A reader not familiar at all with very ample line bundles may think
about them as maps into projective spaces. Let us present this with the example
of the projective space Pn (which also equals G(1, n + 1)). We have an identity
map Pn → Pn , which corresponds to O(1). The r -th Veronese map embeds Pn in a
larger projective space P( n )−1 by evaluating on a point all degree r monomials. The
n+r

associated map is given by O(r ). For n = 1 and r = 2 we get:

P1  [x : y] → [x 2 : x y : y 2 ] ∈ P2 .

It will follow from Proposition 34 that the embedding of G(k, n) by O(d) spans
a projectivisation of an irreducible representation Vλ0 of G L n . The Young dia-
gram λ0 = (d, . . . , d) consists of k rows of length d.

4.4.3 Flag Varieties

More generally, for any irreducible representation Vλ of G L n the projective space

P(Vλ ) contains a unique closed orbit, known as a homogeneous variety or more
precisely a flag variety. To describe it, let us fix a sequence of s positive integers 0 <
k1 < · · · < ks < n. The flag variety is defined as follows:
88 A. Cameron et al.

Fl(k1 , . . . , ks ; n) = {V1 ⊂ · · · ⊂ Vs ⊂ V : dim Vi = ki }

⊂ G(k1 , V ) × · · · × G(ks , V ).

Hence, flag varieties are in bijection with (nonempty) subsets of {1, . . . , n − 1},
while Grassmannians correspond to singletons.
From now on we will abbreviate the tuple (k1 , . . . , ks ) to k, and the flag vari-
ety Fl(k1 , . . . , ks ; n) to Fl(k, n). A point in Fl(k, n) can be represented by a full-
rank n × n-matrix A: the row span of the first di rows is Vi . (Although note that
only the first ks rows of the matrix are relevant.) As with Grassmannians, different
matrices can represent the same point in Fl(k, n). More precisely, if we partition the
rows of A into blocks of size k1 , k2 − k1 , . . . , n − ks , then we are allowed to do row
operations on A, with the restriction that to a certain row we can only add a multiple
of a row in the same block or a block above. Another way to think about this is the
following: let Pk ⊂ G L n (C) be the parabolic subgroup of all invertible matrices A
with Ai j = 0 if i ≤ kr < j, for some r . Then two n × n matrices represent the same
flag if and only if they are the same up to left multiplication with an element of Pk .
Hence Fl(k, n) can also be described as the quotient Pk \G L n (C) (a homogeneous
variety).
Just as for Grassmannians we may study different embeddings of flag varieties.
The natural one is given by the containment

k1
ks
k1
ks
G(k1 , V ) × · · · × G(ks , V ) ⊂ P( V ) × · · · × P( V ) ⊂ P( V ⊗ ··· ⊗ V ),

 
where the last map is the Segre embedding. The representation k1 V ⊗ · · · ⊗ ks V
in general is reducible—a precise decomposition is known by Pieri’s rule (or more
generally by the Littlewood-Richardson rule), see for example [19, Proposition
15.25]. As we will prove below, the flag variety spans an irreducible representation
with the corresponding Young diagram with s columns of lengths ks , ks−1 , . . . , k1
respectively.
Other embeddings can be obtained as follows. We replace the Segre map by the
Segre–Veronese, i.e. we first re-embed a Grassmannian with a Veronese map. Thus,
a flag variety with an embedding can be specified by a function:

f : {1, . . . , n − 1} → Z≥0 .

To abstractly obtain a flag variety from f we first consider a subset {a ∈ [n − 1] :

f (a) > 0}. The Segre–Veronese map is specified by the values of the function f —
the case f (a) = 1 corresponds to the usual Segre. The irreducible representation
we obtain has an associated Young diagram with f (a) columns of length a. Before
proving all these statements we present an example.
Example 33 Let us fix n = 4 and a function f that on 1, 2, 3 takes values 2, 1, 0
respectively. The corresponding flag variety equals Fl(1, 2; 4), i.e. it parameterises
one-dimensional subspaces, l, inside two-dimensional subspaces, S, in a fixed four-
4 Flag Matroids: Algebra and Geometry 89

dimensional space, V . The line l corresponds to a point in a projective space P(V ) =

G(1, V ) and the space S to a point in G(2, V ). If we suppose l = [v1 ], then we may
always find v2 ∈ V such that S = v1 , v2 . Hence

Fl(1, 2) = {[v1 ] × [v1 ∧ v2 ] ∈ P(V ) × G(2, V )}.

We now pass to the embedding. As f (1) = 2 we have to consider the second Veronese
map P(V ) → P(S 2 (V )) given by [v] → [v · v]. We obtain

2
Fl(1, 2) = {[v1 · v1 ] × [v1 ∧ v2 ] ∈ P(S 2 V ) × G(2, V )} ⊂ P(S 2 V ⊗ V ).

By Pieri’s rule, we have a decomposition of G L(V ) representations:

2
S2 V ⊗ V = S3,1 V ⊕ S2,1,1 V.

Hence, S3,1 (V ) corresponds to the Young diagram with the first row of length three
and the second row of length one. We note that this diagram indeed has 2 columns
of length 1, 1 column of length 2, and 0 columns of length 3.
The flag variety is always contained in the lexicographically-first (highest weight)
irreducible component—cf. Proposition 34 below; in our example this is S3,1 (V ). In
particular, we may realise the representation S3,1 (V ) as a linear span of the affine
cone over the flag variety:

 
2
2) = {(v1 · v1 ) ⊗ (v1 ∧ v2 ) : v1 , v2 ∈ V } ⊂ S 2 V ⊗
Fl(1, V.

For readers not familiar with the construction of Sλ V , this can be taken as a def-
inition. For a proof that this definition is equivalent to the usual construction, see
Proposition 34 below.

The function f is often represented on a Dynkin diagram as shown in Fig. 4.2.

For Example 33 this would be as shown in Fig. 4.3.
We note that if f is positive we obtain a complete flag variety, i.e. the variety
parametrizing complete flags:

Fig. 4.2 The Dynkin

diagram for the function f f(1) f(2)
... f(n-2) f(n-1)

n-1

Fig. 4.3 The Dynkin

2 1 0
diagram for Example 33
90 A. Cameron et al.

V1 ⊂ V2 ⊂ · · · ⊂ Vn−1 ⊂ Cn .

The complete flag variety maps to any other flag variety, simply by forgetting the
appropriate vector spaces. We note that all our constructions are explicit and only
use exterior (for Grassmannians), symmetric (for Veronese) and usual (for Segre)
tensor products, as in Example 33.
We are now ready to prove a special case of the Borel-Weil-Bott Theorem relating
representations and embeddings of flag varieties.
Proposition 34 (Borel-Weil) Any flag variety Fl(k1 , . . . , ks ; n) with an embedding
given by a function f spans the irreducible G L(V )-representation Sλ V , where the
Young diagram λ has f ( j) columns of length j.

Proof Fix a basis e1 , . . . , en of V . Let us consider the flag of subspaces

e1 , . . . , ek1  ⊂ e1 , . . . , ek2  ⊂ · · · ⊂ e1 , . . . , eks 

and the corresponding point p ∈ Fl(k1 , . . . , ks ). Under the embedding specified by f

it is mapped to

(e1 ∧ · · · ∧ ek1 )◦ f (k1 ) ⊗ · · · ⊗ (e1 ∧ · · · ∧ eks )◦ f (ks )

k1
ks
⊂ S f (k1 ) ( V ) ⊗ · · · ⊗ S f (ks ) ( V ).

The G L(V )-decomposition of the ambient space is highly non-trivial. However,

looking directly at the T decomposition we see that, up to scaling, the image of p
is the unique lexicographically-highest vector. Hence, in particular, the image of p
belongs to Sλ V , as all other G L(V )-representations appearing in the decomposition
have strictly smaller highest weights. Furthermore, the flag variety is an orbit under
the G L(V )-action—one can explicitly write a matrix mapping any flag to any other
given flag. Thus, if one point is contained in the irreducible representation, the whole
variety must be contained in it.
It remains to show that the span of the flag variety is indeed the whole irreducible
representation. This is true, as the flag variety is G L(V )-invariant, and thus its linear
span is a representation of G L(V ). As Sλ V is irreducible, the linear span must
coincide with it. 

The above theorem may be regarded as a realisation of irreducible G L(V )-

representations as spaces of sections of a very ample line bundle on a flag variety.
A more general Borel-Weil-Bott theorem provides not only a description of global
sections—zeroth cohomology—but also higher, arbitrary cohomology.
Later on, we will denote a flag variety together with the embedding given by f
as Fl(k; n), where k = (k1 , . . . , ks ) satisfies 0 < k1 ≤ . . . ≤ ks < n and has f (a)
entries equal to a, for all a ∈ [n − 1]. For example, the embedding of Example 33
will be written as Fl(1, 1, 2; 4).
4 Flag Matroids: Algebra and Geometry 91

4.5 Representable Matroids: Combinatorics and Geometry

Let us consider a representable matroid M given by n = |E| vectors spanning a k-

dimensional vector space V . By fixing a basis of V we may represent this matroid
as a k × n matrix A. On the other hand the matrix A may be regarded as defining
a k-dimensional subspace of an n-dimensional vector space, i.e. a point in G(k, n).
Since applying elementary row operations to A does not change which of the maximal
minors of A vanish, the matroid M only depends on the k-dimensional subspace,
and not on the specific matrix A representing our subspace. In this way we have
associated to any point p ∈ G(k, n) a representable rank-k matroid M p on [n].
Remark 35 In the literature, the correspondence between points in G(k, n) and
vector arrangements in Ck is known as the Gel’fand-MacPherson correspondence.
The way we just constructed it is very explicit, but has the disadvantage of not being
canonical (it depends on a chosen basis of Cn ). There are several ways to fix this.
One way of obtaining a more intrinsic construction is to replace the Grassman-
nian G(k, n) by the Grassmannian G(n − k, n). If A represents a surjective linear
map from Cn to V , then to A one can associate the (n − k)-dimensional kernel of
this map, i.e. a point in G(n − k, n). As this construction requires dual matroids we
decided to present the one above in coordinates.
A different (but closely related) intrinsic construction would be to define G(k, Cn )
as the space of k-dimensional quotients of Cn (instead of k-dimensional subspaces).
Then A maps the standard basis of Cn to n vectors in a smaller space V ∈ G(k, Cn );
these n vectors represent a matroid.
A third solution would be to talk about hyperplane arrangements instead of vector
arrangements.
The vector space Cn comes with the action of a torus T = (C∗ )n . We have associ-
ated a point p ∈ G(k, Cn ) to a representation of a matroid. If we change the represen-
tation by rescaling the vectors we do not change the matroid and the associated point
belongs to the orbit T p. Hence, the intrinsic properties of the matroid M p should be
related to the geometry of T p—a feature we will examine in detail throughout the
article. The closure T p is a projective toric variety. For more information about toric
geometry we refer to [9, 17, 38, 49].
Remark 36 Of course it can happen that different torus orbits give rise to the same
matroid: there are only finitely many matroids on [n], but if 1 < k < n − 1 there
are infinitely many torus orbits in G(k, n). In fact, the set of all points in G(k, n)
giving rise to the same matroid forms a so-called thin Schubert cell or matroid
stratum, which typically is a union of infinitely many torus orbits. Thin Schubert
cells were first introduced in [21]. Thin Schubert cells are badly behaved in general:
for fixed k ≥ 3 the thin Schubert cells of G(k, n) exhibit arbitrary singularities if n
is large enough. This is a consequence of Mnëv’s theorem [41]. See [31, Sect. 1.8]
for a more detailed discussion.
Theorem 37 ([21]) The lattice polytope representing the projective toric variety T p
described above is equal to P(M p ).
92 A. Cameron et al.

Proof Let A be the matrix whose rows span the space corresponding to p. The
parameterisation of T p is given by:

k
φ : T → P( Cn ).

The coordinates of the ambient space are indexed by k-element subsets of the n
columnsof the matrix A. The Plücker coordinate indexed by I of φ(t1 , . . . , tn )
equals i∈I ti times the k × k minor of A determined by I , which we will denote
by det(A I ). In other words, the map φ in Plücker coordinates is given as follows:

φ(t1 , . . . , tn ) = (det(A I ) · ti ) I ∈([n]) .
k
i∈I

The I -th coordinate is nonzero if and only if I is a basis of M p . Hence, the ambient
space of T p has coordinates indexed by basis elements of M p . After restricting
to this ambient space and composing with the isomorphism inverting the nonzero
minors det(A I ), our map can be written as
 

φ(t1 , . . . , tn ) = ti .
i∈I I ∈P(M p )

This is exactly the construction of the toric variety represented by P(M p ). 

It is a major problem to provide the algebraic equations of T p. This is equivalent to

finding integral relations among the basis of a matroid. We point out that matroids
satisfy a ‘stronger’ property then one could expect from the basis exchange axiom
B2 of Lemma 3. Precisely, for any two bases B1 , B2 ∈ B and a subset A ⊂ B1 − B2 ,
there exists A ⊂ B2 − B1 such that (B1 − A) ∪ A and (B2 − A ) ∪ A are in B [26].
This exactly translates to a binomial quadric (degree 2 polynomial) in the ideal of T p:
x B1 x B2 − x(B1 −A)∪A x(B2 −A )∪A , where, as in the proof of Theorem 37, we label each
coordinate by a basis of the matroid. Further, if |A| = 1 we obtain special quadrics
corresponding to exchanging one element in a pair of bases. The following conjecture
due to White provides a full set of generators for any matroid M.
Conjecture 38 The ideal of the toric variety represented by P(M) is generated by
the special quadrics corresponding to exchanging one element in a pair of bases.
We note that it is unknown whether the ideal of this toric variety is generated by
quadrics, or that all quadrics are spanned by the special quadrics described above.
However, it is known that the special quadrics define the variety as a set (or more
precisely as a projective scheme) [32, 33].
The combinatorial methods can be used to prove geometric properties of torus
orbit closures in Grassmannians. Below we recall a definition of a normal lattice
polytope.
4 Flag Matroids: Algebra and Geometry 93

Definition 39 A lattice polytope P, containing 0 and spanning (as a lattice) the

lattice N , is normal if and only if for any k ∈ Z≥0 and any p ∈ k P ∩ N we have p =
p1 + · · · + pk for some pi ∈ P ∩ N .
Normality of a polytope is a very important notion as it corresponds to projective
normality of the associated toric variety [9, Chap. 2], [49] (less formally, the associ-
ated toric variety is not very singular and is embedded in a particularly nice way in
the projective space).
Theorem 40 (White) For any matroid M the polytope P(M) is normal. In partic-
ular, any torus orbit closure in any Grassmannian is projectively normal.
Proof This is a special case of Theorem 26, where we take all Mi equal to M. 

4.6 Introduction to Flag Matroids

In Sect. 4.5 we explained a correspondence between torus orbits in a Grassman-

nian (geometric objects) and representable matroids (combinatorial objects). We
will generalise this correspondence in different ways. For instance, on the geometry
side, we can replace Grassmannians with flag varieties. On the combinatorics side,
this naturally leads to the notion of a (representable) flag matroid. Flag matroids first
arose as a special case of the so-called Coxeter matroids, introduced by Gel’fand
and Serganova [22, 23]. In this section we first give a combinatorial introduction
to flag matroids. Afterwards, we explain how they are related to flag varieties. The
exposition is largely based on [3, Chap. 1].

4.6.1 Flag Matroids: Definition

We start by defining flag matroids in the way they are usually defined in the literature:
using Gale orderings.
Definition 41 Let 0 < k1 ≤ . . . ≤ ks < n be natural numbers. Let k = (k1 , . . . , ks ).
A flag F of rank k on E is an increasing sequence

F1 ⊆ F2 ⊂ · · · ⊆ Fs

of subsets of E such that |F i | = ki for all i. The set of all such flags will be denoted
by F Ek .
Let ω be a linear ordering on E. We can extend the Gale ordering ≤ω to flags:

(F 1 , . . . , F s ) ≤ω (G 1 , . . . , G s ) if and only if F i ≤ω G i for all i.

94 A. Cameron et al.

Definition 42 A flag matroid of rank k on E is a collection F of flags in F Ek , which

we call bases, satisfying the following property: for every linear ordering ω on E,
the collection F contains a unique element which is maximal in F with respect to
the Gale ordering ≤ω .
If F is a flag matroid, the collection {F i | F ∈ F } is called the i-th constituent of F .
This is clearly a matroid (of rank ki ).
Remark 43 In the literature it is usually required that we have strict inequali-
ties 0 < k1 < · · · < ks < n. From a combinatorial point of view this does not make
a difference, but when we later consider flag matroid polytopes this restriction would
appear artificial. This is also the reason why in Sect. 4.4.3 we did not just consider
flag varieties, but also their Veronese re-embeddings.
Next, we want to describe which tuples of matroids can arise as the constituents of
a flag matroid. In order to give this characterisation, we first need to recall matroid
quotients.

4.6.2 Matroid Quotients

Definition 44 Let M and N be matroids on the same ground set E. We say that N
is a quotient of M if one of the following equivalent statements holds:
1. every circuit of M is a union of circuits of N ;
2. if X ⊆ Y ⊆ E, then r M (Y ) − r M (X ) ≥ r N (Y ) − r N (X );
3. there exists a matroid R and a subset X of E(R) such that M = R\X and N =
R/ X ;
/ B, there is a basis B  of N with B  ⊆ B and such
4. for all bases B of M and all x ∈

that {y : (B − y) ∪ x ∈ B(N )} ⊆ {y : (B − y) ∪ x ∈ B(M)}.

For the equivalence of 1, 2 and 3, we refer to [43, Proposition 7.3.6]. Part 4 is left to
the reader.
Here are some basic properties of matroid quotients:
Proposition 45 Let N be a quotient of M.
1. Every basis of N is contained in a basis of M, and every basis of M contains a
basis of N .
2. rk(N ) ≤ rk(M) and in case of equality N = M.

Proof Both statements can be easily deduced by plugging in Y = E or X = ∅ in

Definition 44 2. 

The next result will be essential for defining representable flag matroids. It also
explains where the term “matroid quotient” comes from—below we think of W as a
vector space quotient of V .
4 Flag Matroids: Algebra and Geometry 95

Proposition 46 ([6, Proposition 7.4.8 (2)]) Let V and W be vector spaces and ψ :
E → V be a map. Furthermore, let f : V → W be a linear map. Consider the
matroid M represented by ψ, and the matroid N represented by f ◦ ψ. Then N is a
matroid quotient of M.

Example 47 If R is a representable matroid on E and X is a subset of E,

then M := R\X and N := R/ X are representable matroids, and there is a linear
map as in Proposition 46. Indeed, if R is represented by ψ : E → V , then con-
sider the projection π : V → V /ψ(X ). It isnot hard to see that M is represented
by ψ  E−X and that N is represented by π ◦ ψ  E−X .

Example 48 The converse of Proposition 46 is false: we now give an example (taken

from [3, Sect. 1.7.5]) of two representable matroids M and N such that N is a quotient
of M, but there is no map as in Proposition 46.
Let M be the rank-3 matroid on [8] represented by the following matrix
⎛ ⎞
10101101
⎝0 1 1 0 2 2 2 1⎠
00011211

and let N be the rank-2 matroid on [8] whose bases are all 2-element subsets except
for {2, 6} and {3, 5}. It is easy to see that N is a representable matroid: just pick
six pairwise independent vectors in the plane, and map 2 and 6, as well as 3 and 5,
to the same vector. Now N is a matroid quotient of M, since the matroid R from
Example 8 satisfies M = R\9 and N = R/9. However, it is not possible to find
representations V (resp. W ) of M (resp. N ) such that there is a map f : V → W as
in Proposition 46. Roughly speaking, the problem is that the “big” matroid R is not
representable. For a more precise argument, see [3, Sect. 1.7.5].

4.6.3 Representable Flag Matroids

We now come to the promised characterisation of constituents of flag matroids. In

fact, it will turn out we can use it as an alternative definition of flag matroids.
A collection (M1 , . . . , Ms ) of matroids is called concordant if, for every
pair (Mi , M j ), either Mi is a quotient of M j or vice versa. Note that this is equivalent
to the fact that they can be ordered in such a way that Mi is a quotient of Mi+1 ,
because “being a quotient of” is transitive.
Theorem 49 ([3, Theorem 1.7.1]) A collection F of flags in F Ek is a flag matroid if
and only if the following three conditions hold:
1. every constituent Mi := {F i | F ∈ F } is a matroid;
2. the matroids M1 , . . . , Ms are concordant;
3. every flag B1 ⊆ . . . ⊆ Bs , with Bi a basis of Mi , is in F .
96 A. Cameron et al.

In other words, flag matroids on E are in one-to-one correspondence with tuples of

concordant matroids on E.
We can now define representable flag matroids: let 0  V1 ⊆ · · · ⊆ Vs  Cn
be a flag of subspaces of Cn . Then, viewing Vi as a point in G(ki , n), it holds
that (MV1 , . . . , MVs ) is a concordant collection of matroids by Proposition 46. Indeed:
if Vi ⊆ V j , we can pick a k j × n matrix A j representing V j such that the first ki rows
of A j span Vi . Then the columns of Ai are obtained from the columns of A j by
deleting the last k j − ki entries.
Definition 50 The representable flag matroid F (V1 ⊆ · · · ⊆ Vs ) is the unique flag
matroid whose constituents are MV1 , . . . , MVs .

Remark 51 Example 48 shows that it can happen that all constituents of a flag
matroid are representable matroids, but still the flag matroid is not representable
(because the matroid representations are “not compatible”).

4.6.4 Flag Matroid Polytopes

Definition 52 Given a flag F on [n], we can identify each constituent with a 0, 1-

vector and then add them all up to a vector e F ∈ Zn≥0 . In this way we have identified
the flags in [n] of rank k with integer vectors with k1 entries equal to m, k2 − k1
entries equal to m − 1, …, ki+1 − ki entries equal to m − i, …, ks − ks−1 entries
equal to 1, and n − ks entries equal to 0. We will refer to such vectors as rank-k
vectors. In other words, if we think of k as a partition of length s, we can consider
the conjugate partition k∗ of length ≤ n. Then a rank-k vector is a vector v ∈ Zn≥0
obtained from k∗ by adding 0’s and permuting the entries.

Definition 53 The base polytope of a flag matroid F on [n] is the convex hull of
the set {e F | F ∈ F } ⊂ Rn .

Example 54 Let F be the rank (1, 2) flag matroid on [3] whose bases are 1 ⊆
12, 1 ⊆ 13, 2 ⊆ 12 and 3 ⊆ 13. Then its base polytope is the convex hull of the
points (2, 1, 0), (2, 0, 1), (1, 2, 0), (1, 0, 2). Its constituents are the uniform rank 1
matroid on [3], and the rank 2 matroid with bases 12 and 13. The base polytope of this
flag matroid is depicted in Fig. 4.4. F is a representable flag matroid: a representing
flag is for example e1 + e2 + e3  ⊂ e1 , e2 + e3  ⊂ C3 .

Theorem 55 ([3, Theorem 1.11.1]) A lattice polytope P ⊂ Rn is the base polytope

of a rank-k flag matroid on [n] if and only if the following two conditions hold:
1. every vertex of P is a rank-k vector;
2. every edge of P parallel to ei − e j for some i, j ∈ [n].

Theorem 56 ([3, Corollary 1.13.5]) The polytope of a flag matroid is the Minkowski
sum of the matroid polytopes of its constituent matroids.
4 Flag Matroids: Algebra and Geometry 97

Fig. 4.4 A flag matroid base (1,0,2)

polytope

(2,0,1)

(2,1,0) (1,2,0)

Thus, each flag matroid defines a polytope that is a base polytope of a polymatroid.
Remark 57 It follows from Theorem 26 and the previous theorem that the lat-
tice points of a flag matroid polytope correspond to tuples (not necessarily
flags) (B1 , . . . , Bs ), where Bi is a basis of Mi . For example point (1, 1, 1) in Fig. 4.4
corresponds to a basis of a polymatroid, but not to a flag, i.e. not to a basis of the flag
matroid.

4.6.5 Flag Matroids and Torus Orbits

Consider the flag variety Fl(k, n), as described in Sect. 4.4. The action of the
torus T = (C∗ )n on Cn induces an action of T on Fl(k, n). A point p ∈ Fl(k, n)
gives rise to a representable flag matroid M on [n], as in Definition 50. All points
in the orbit T p give rise to the same flag matroid. This last statement follows easily
from the analogous fact for matroids and the fact that a flag matroid is determined
by its constituent matroids. The analogue of Theorem 37 holds:
Theorem 58 The lattice polytope representing the toric variety T p is equal to the
flag matroid polytope of M.
Proof The proof is a straightforward generalisation of the proof of Theorem 37, with
the parameterisation of T p given by:

k1
ks
φ : T → P( Cn ) × · · · × P( Cn ).

4.7 Representable Polymatroids

Representable polymatroids generalise representable matroids, replacing vectors

with subspaces. While they do not appear as frequently in pure mathematics as
matroids, one of the possible references for an interested reader is [4, Sect. 3].
We start by giving a precise definition.
Definition 59 Let V be a vector space and denote by S(V ) the set of all subspaces
of V . Suppose we have a map φ : E → S(V ) and assume, without loss of generality,
98 A. Cameron et al.


that e∈E φ(e) = V . The representable polymatroid M(φ) is  defined by the rank
function: for A ⊆ E, define r (A) as the dimension of φ(A) := a∈A φ(a).

Example 60 Consider the map φ : [3] → S(C3 ) defined by φ(1) = e1 , e2 

and φ(2) = φ(3) = e1 , e3 . Then M(φ) is a rank 3 polymatroid on [3] with 5
bases: (2, 1, 0), (2, 0, 1), (1, 2, 0), (1, 0, 2), (1, 1, 1). Its base polytope is given in
Fig. 4.4.

Now we want to define the polymatroid analogue of a matroid associated to a

subspace. For this, we fix a number r ∈ Z>0 , and we consider the vector space Cr n ,
with a fixed basis indexed by [r ] × [n]. If we consider a full rank k × r n matrix A,
it represents a k-dimensional subspace of Cr n . Moreover, the r columns indexed
by (i, j) for some fixed j span a subspace W j of Ck . Then the map [n] → S(Ck ) :
j → W j defines a representable polymatroid. Note that this is an r -polymatroid, and
that every representable r -polymatroid can be obtained in this way.
As before, applying elementary row operations to A does not change the poly-
matroid. Hence, we have described how to associate an r -polymatroid M(W ) to a
subspace W of Cr n .
Remark 61 If we defined Grassmannians via quotients instead of subspaces, we
could present this story more invariantly, as follows: Let V be a vector space of dimen-
sion r n, with a fixed collection of n independent r -dimensional subspaces V1 , . . . Vn .
Then any k-dimensional quotient φ : V → W gives rise to an r -polymatroid via the
map [n] → S(W ) : i → φ(Vi ).
We consider the action of T = (C∗ )n on Cr n , where t = (t1 , . . . , tn ) acts by
multiplying the (i, j)-th coordinate by t j . If we consider the action of T = (C∗ )n
on G(k, Cr n ) induced by the action on Cr n , then every point in the orbit T p gives the
same polymatroid. This is very similar to the situation for representable matroids, so
it should be no surprise that we also have an analogue of Theorem 37 in this setting:
Theorem 62 The lattice polytope representing the projective toric variety T p
described above is equal to the base polytope of the polymatroid M.

Proof It is a straightforward generalisation of the proof of Theorem 37. 

4.7.1 Comparison Between Polymatroids and Flag Matroids

A flag matroid polytope is a special case of a polymatroid polytope, by Theorem 55.

It is tempting to think about flag matroids as a special case of polymatroids, but we
need to be careful when doing this: the notion of a basis of a flag matroid is not
compatible with the notion of a basis for a polymatroid. More precisely, the bases
of a polymatroid are all lattice points of its base polytope, while the bases of a flag
matroid are only the vertices of the associated polytope—cf. Remark 57 and Fig. 4.4.
For a flag matroid F , let M(F ) be the associated polymatroid.
4 Flag Matroids: Algebra and Geometry 99

The definitions of representable flag matroid and representable polymatroid look

unrelated at first sight, but there is a connection.
Proposition 63 If F is a representable flag matroid, then M(F ) is a representable
polymatroid.

Proof For this proof we will use the more intrinsic definition of representable
flag matroids and polymatroids using quotients. Let F be the flag matroid repre-
sented
 by the flag of quotients Cn → V1 → · · · → Vr . Then the quotient Ckn →
V1 · · · Vr represents a polymatroid M on [n]. We need to argue that F and M
have the same base polytope.
Bases of F correspond to flags [n]  F 1 ⊇ . . . ⊇ F r such that F i gives a basis
of Vi . On the other hand, choosing a basis of M corresponds to choosing for every i
a subset F i  [n] such that F i gives a basis of Vi . From this it follows that every
vertex of P(F ) is a lattice point of P(M), and (using Remark 57) that every lattice
point of P(M) is a lattice point of P(F ). So P(M) = P(F ) as desired. 

Remark 64 Thinking again about flag varieties in terms of subspaces, we have that
if F is represented by a flag V1 ⊆ . . . ⊆ Vr  Cn , then M(F ) is represented by a
subspace V1 ⊕ · · · ⊕ Vr  Cr n . Geometrically, this construction corresponds to an
algebraic map 
Fl(k1 , . . . , kr ; n) → G( ki , r n).
i

Remark 65 The converse of Proposition 63 is not true: given a representable poly-

matroid that is also a flag matroid, it is not always a representable flag matroid.
One way to construct a counterexample is as follows. If M1 , M2 are representable
matroids with corresponding base polytopes P1 , P2 , then P1 + P2 corresponds to a
representable polymatroid, by the same argument as in the proof of Proposition 63.
However, Example 48 gives an example of two such (concordant) matroids such
that P1 + P2 corresponds to a flag matroid that is not representable.

4.8 Equivariant K -theory

We have presented a correspondence between representable matroids and torus orbits

in Grassmannians, and generalisations of this correspondence to representable flag
matroids and representable polymatroids. We would like to drop the word “repre-
sentable” from all of those. As we will see, one way to do this is by replacing “torus
orbits” with “classes in equivariant K -theory”. This was done for matroids by Fink
and Speyer [16]. In this section, we review their construction, and consider gen-
eralisations to flag matroids. We start with an introduction to non-equivariant and
equivariant K -theory.
100 A. Cameron et al.

4.8.1 A Very Brief Introduction to K -theory

This section is based on [18, Sect. 15.1].

Let X be an algebraic variety. We define K 0 (X ) to be the free abelian group gen-
erated by vector bundles on X , subject to relations [E] = [E  ] + [E  ] whenever E 
is a subbundle of E, with quotient bundle E  = E/E  . The group K 0 (X ) inherits a
ring structure from the tensor product: [E] · [F] = [E ⊗ F].
Similarily, we can define K 0 (X ) to be the free abelian group generated by iso-
morphism classes of coherent sheaves on X , subject to relations [A] + [C] = [B]
whenever there is a short exact sequence 0 → A → B → C → 0. There is an inclu-
sion K 0 (X ) → K 0 (X ). From now on, we will always assume that X is a smooth
variety. In this case, the inclusion is an isomorphism, allowing us to identify K 0 (X )
and K 0 (X ).
Let f : X → Y be a map of (smooth) varieties. Then there is a pullback map f ∗ :
K (Y ) → K 0 (X ) defined by f ∗ [E] = [ f ∗ E] (where E is a vector bundle on Y ).
0

 map, there is also a pushforward map f ∗ : K 0 (X ) → K 0 (Y ) given

If f is a proper
by f ∗ [A] = (−1)i [R i f ∗ A]. Here R i f ∗ are right derived functors of the push-
forward. An interested reader is advised to find the details in [18, Sect. 15]. In this
paper, we will not be using the formal definitions of K 0 (X ), f ∗ or f ∗ . Instead, we will
refer to explicit descriptions of those in the cases that we need, each time providing
a theorem we build upon.

Remark 66 In all the cases we study the ring K 0 (X ) is isomorphic to the cohomol-
ogy ring and to the Chow ring (after tensoring with Q). Note however that the map
from K 0 (X ) to the Chow ring is nontrivial and given by the Chern character.

Example 67 Consider the projective space Pn . The (rational) Chow ring is A(Pn ) =
Q[H ]/(H n+1 ). Here one should think about H as a hyperplane in Pn and H k as a codi-
mension k projective subspace. The most important n line bundle is O(1). The Chern
character ch : K 0 (Pn ) → A(Pn ) sends [O(1)] to i=0 H i /i!. Note that K 0 (Pn ) can
be written as Z[α]/(α n+1 ), where α = 1 − [O(−1)] is the class of the structure sheaf
of a hyperplane. As a special case, the K -theory of a point is Z.

If X is a smooth variety equipped with an action of a torus T , we can define

its equivariant K -theory K T0 (X ) ∼
= K 0T (X ). The construction and properties are
exactly the same as in the previous paragraphs, if we replace “vector bundles” and
“coherent sheaves” by “T-equivariant vector bundles” and “T-equivariant coherent
sheaves”.
For later reference, we describe the equivariant K -theory of a point: K T0 ( pt) =
Z[Char(T )], where Char(T ) = Hom(T, C∗ ) is the lattice of characters of T .
Here Z[Char(T )] is the group ring of Char(T ), i.e. as a module over Z it has a
basis given by Char(T ), and multiplication is induced from addition in Char(T ). It
is the ring of Laurent polynomials in dim T variables. Explicitly, a T -equivariant
sheaf on pt is just a vector space W with a T -action. We may decompose W =
⊕c∈Char(T ) Wc as in Sect. 4.4.1. The corresponding element of Z[Char(T )] is the
4 Flag Matroids: Algebra and Geometry 101


character (also called Hilbert series) Hilb(W ) := c∈Char(T ) (dim Wc )c. We point
out that even for infinite-dimensional T -modules, Hilb(W ) makes sense as a formal
power series, as long as Wc is finite-dimensional for all c.
We finish this section by describing the relation between ordinary and T -
equivariant K -theory:
Theorem 68 ([37, Theorem 4.3]) Let X be a smooth projective variety with an
action of a torus T . Let S ⊆ T be a subtorus. Then the natural map

K T0 (X ) ⊗Z[Char(T )] Z[Char(S)] → K S0 (X )

is an isomorphism. In particular, taking S to be the trivial group, the natural map

K T0 (X ) ⊗Z[Char(T )] Z → K 0 (X )

is an isomorphism.
We note that the map Z[Char(T )] → Z above is given in coordinates by sending
each generator ti of T to 1.

4.8.2 Explicit Construction via Equivariant Localisation

Let X be a smooth projective variety over C, and T a torus acting on it. If X has only
finitely many torus-fixed points, we can use the method of equivariant localisation
to give an explicit combinatorial description of classes in K T0 (X ). Our exposition
here is largely based on the one in [16]. The following theorem is central to our
discussion.
Theorem 69 ([42, Theorem 3.2], [16, Theorem 2.5] and references therein) If X is
a smooth projective variety with a torus action, then the restriction map K T0 (X ) →
K T0 (X T ) is an injection.
From now on we will always assume that X has only finitely many torus-fixed points.
In this case K T0 (X T ) is simply the ring of functions from X T to Z[Char(T )]. In other
words, we can describe a class in K T0 (X ) just by giving a finite collection of Laurent
polynomials in Z[Char(T )].
Remark 70 In the literature, a variety X for which K T0 (X ) is a free Z[Char(T )]-
module, and has a Z[Char(T )]-basis that restricts to a Z-basis of K 0 (X ), is
called equivariantly formal. This notion was first introduced in [24]. In [2, Sect. 2.4],
it is noted that smooth projective varieties with finitely many T -fixed points are
equivariantly formal.
We now explicitly describe the class of a T -equivariant coherent sheaf on X . We will
do this under the following additional assumption (which is not essential but makes
notation easier and will hold for all varieties of interest):
102 A. Cameron et al.

Definition 71 A finite-dimensional representation of T is called contracting if all

characters lie in an open halfspace, or equivalently if the characters generate a pointed
cone (see Sect. 4.8.3). The action of T on a variety X is contracting, if for every torus-
fixed point x ∈ X , there exists an open neighbourhood Ux isomorphic to A N such
that the action of T on Ux is a contracting representation.
Let E be a T -equivariant coherent sheaf on X . We will construct a map [E]T :
X T → Z[Char(T )]. For every x ∈ X T , we have an open neighbourhood Ux as in
Definition 71. Let χ1 , . . . , χ N be the characters by which T acts on Ux (so O(Ux )
is a polynomial ring multigraded by T in the sense of [40, Definition 8.1], with
characters χ1−1 , . . . , χ N−1 ). Our sheaf E, restricted to Ux , corresponds to a graded,
finitely generated O(Ux )-module E(Ux ).
Since E(Ux ) is a graded module over the polynomial ring O(Ux ), which is multi-
graded by T , it follows from [40, Theorem 8.20] that E(Ux ) is a T -module, and its
Hilbert series is of the form
K (E(Ux ), t)
N −1
, (4.1)
i=1 (1 − χi )

for some K (E(Ux ), t) ∈ Z[Char(T )].

Definition 72 For E a T -equivariant coherent sheaf on X , we define [E]T to be the
map that sends x ∈ X T to K (E(Ux ), t) ∈ Z[Char(T )], the numerator in (4.1).
Theorem 73 ([16, Theorem 2.6]) The map [E]T defined above is the image of the
class of E under the injection K T0 (X ) → K T0 (X T ) of Theorem 69.
Example 74 Let X = Pn , equipped with the natural torus action t · [a0 : . . . : an ] =
[t0−1 a0 : . . . : tn−1 an ]. Then O(d) is a T -equivariant sheaf. The torus action on Pn
has n + 1 fixed points, namely pi = [0 : . . . : 1 : . . . : 0], where the 1 is at position i.
We use equivariant localisation to describe the class [O(d)]T .
Every pi has an open neighbourhood Ui = Spec Ai , where

Ai = C[x0 , . . . , xˆi , . . . , xn ]

is multigraded by T via deg(x j ) = ti−1 t j . The A i -module O(d)(Ui ) is a copy of Ai

generated in degree tid . So its Hilbert series is tid / j (1 − ti−1 t j ). Hence [O(d)]T can
be represented by the map (Pn )T → Z[Char(T )] : pi → tid .
We can describe the image of the map from Theorem 69 explicitly, if we impose
an additional condition on X .
Theorem 75 ([54, Corollary 5.12], [16, Theorem 2.9] and references therein) Sup-
pose X is a projective variety with an action of a torus T , such that X has finitely
many T -fixed points and finitely many 1-dimensional T -orbits, each of which has
closure isomorphic to P1 . Then a map f : X T → Z[Char(T )] is in the image of the
map K T0 (X ) → K T0 (X T ) of Theorem 69 if and only if the following condition holds:
For every one-dimensional orbit, on which T acts by character χ and for which x
and y are the T -fixed points in the orbit closure, we have
4 Flag Matroids: Algebra and Geometry 103

f (x) ≡ f (y) mod 1 − χ .

Example 76 (Example 74 continued) Note that Pn has only finitely many one-
dimensional torus orbits: for every pair pi , p j of T -fixed points, there is a unique T -
orbit whose closure contains pi and p j . Furthermore, T acts on this orbit with
character t −1 d d −1
j ti . We see that ti ≡ t j mod 1 − t j ti , so that the class [O(d)] indeed
T

fulfills the condition of Theorem 75.

We also can describe pullback and pushforward in the language of equivariant
localisation. Let π : X → Y be a T -equivariant map of smooth projective varieties
with finitely many T -fixed points, then for [E]T ∈ K T0 (Y ), its pullback can be com-
puted by
(π ∗ [E]T )(x) = [E]T (π(x)) (4.2)

for x ∈ X T .
Describing the pushforward of [F]T ∈ K T0 (X ) is a bit more complicated. Sup-
pose that X and Y are contracting. For every point x ∈ X T (resp. y ∈ Y T ), we
pick as before an open neighbourhood Ux (resp. Vy ) on which T acts by charac-
ters χ1 (x), . . . , χr (x) (resp. η1 (y), . . . , ηs (y)). Then the pushforward of [F]T is
determined by the formula

(π∗ [F]T )(y)  [F]T (x)

 =  . (4.3)
(1 − η j (y)−1 ) (1 − χi (x)−1 )
x∈π −1 (y)∩X T

Remark 77 We can use Theorem 68 to obtain a description of the ordinary K -theory

using equivariant localisation. However, one should be careful when using this for
computations in practice. Here is a toy example: let X = P2 with the usual action
of (C∗ )2 . Then X T = {[1 : 0], [0 : 1]}, and we can write the elements of K T0 (X T ) ∼ =
Maps(X T , Z[t0± , t1± ])  f as pairs ( f ([1 : 0]), f ([0 : 1])). Then (t0 − t1 , 0) satisfies
the condition from Theorem 75, hence it gives a class in K T0 (X ). It is tempting to do
the following computation in K 0 (X ) ∼ = K T0 (X ) ⊗Z[Char(T )] Z:

(t0 − t1 , 0) ⊗ 1 = (1, 0) ⊗ (1 − 1) = 0

but this is wrong! Indeed, (1, 0) does not satisfy the condition from Theorem 75,
hence is not in K T0 (X ). In fact, one can check that (t0 − t1 , 0) is the equivariant class
of the torus-fixed point [1 : 0] ∈ P2 .

4.8.3 A Short Review on Cones and Their Hilbert Series

For more details about the topic of this subsection we refer to [9, Sect. 1.2] and [48,
convex polyhedral rational cone is a subset of Rn of the
Sect. 4.5]. Recall that a
form C = cone(S) := { u∈S λu u | λu ∈ R≥0 }, where S ⊂ Zn ⊂ Rn is finite. A cone
104 A. Cameron et al.

is called pointed if it does not contain a line. If C is a pointed rational cone, then
every one-dimensional face ρ contains a unique lattice point uρ that is closest to the
origin. It is not hard to see that M G(C) := {uρ | ρa one-dimensional face ofC} is
a minimal generating set of C. If the minimal generators are linearly independent
over R, we call C simplicial. If they are part of a Z-basis of Zn , we call C regular.
For a pointed cone C in Rn , we define its Hilbert series Hilb(C) by:

Hilb(C) := ta .
a∈C∩Zn

Thisis always a rational function [48, Theorem 4.5.11] whose denominator is equal
to u∈M G(C) (1 −tu ). If C is a regular cone, then its Hilbert series is easy to com-
pute: Hilb(C) = u∈M G(C) 1/(1 − tu ). If C is a simplicial cone, we can compute
 Hilbert series as follows. First compute the finite set DC := {b ∈ C ∩ Z : b =
n
its
b∈M G(C) λu u | 0 ≤ λu < 1}. Then

⎛ ⎞
  1
Hilb(C) = ⎝ tb ⎠ .
b∈DC u∈M G(C)
1 − tu

For a general rational polyhedral cone, we can compute its Hilbert series by
triangulating it.

4.8.4 Matroids and the K -theory of Grassmannians

In this subsection we compute the class in equivariant K -theory of a torus orbit closure
in a Grassmannian. We then note that this class only depends on the underlying
matroid, and give a combinatorial algorithm to get the class in K -theory directly
from the matroid. This algorithm can then be used as a definition to associate a class
in K -theory to an arbitrary (not necessarily representable) matroid. This was first
done by Fink and Speyer [16].
Let us first fix the following sign conventions. The torus T = (C∗ )n acts on Cn as
follows: t · (x1 , . . . , xn ) = (t1−1 x1 , . . . , tn−1 xn ). The action of T on G(k, n) is induced
from this action. Explicitly, if p ∈ G(k, n) has Plücker coordinates [PI ] I ∈([n]) , then t ·
 k
p has Plücker coordinates [( i∈I ti−1 )PI ] I ∈([n]) .
k
We begin by describing the T -equivariant K -theory of the Grassmannian G(k, n)
using equivariant localisation.
The torus-fixed points of G(k, n) are easy to describe: for every size k sub-
set I ⊂ [n], we define the k-plane VI = span({ei | i ∈ I }) ⊂ Cn , and denote the
corresponding point in G(k, n) by p I. In Plücker coordinates, p I is given by PJ = 0
if J = I . It is easy to see that the nk points p I are precisely the torus-fixed points
of G(k, n).
4 Flag Matroids: Algebra and Geometry 105

We can also describe the one-dimensional torus orbits: there is a (unique) one-
dimensional torus orbit between p I and p J if and only if |I ∩ J | = k − 1. In this
case, we write I − J = {i}, J − I = { j}. If we identify the one-dimensional orbit
from p I to p J with A1 \0 in such a way that the origin corresponds to the torus-fixed
point p I (and so p J corresponds to the point at infinity), then T acts on the orbit with
character t −1
j ti .
Let us now check that the action of T is contracting. We fix a torus-fixed point p I ,
and consider the open neighbourhood U I given by PI = 1. Then U I ∼ = Ak(n−k) .
PI −i∪ j
For p ∈ U I , we will denote its coordinates with (u i, j )i∈I, j ∈I / , where u i, j = PI .
Then t · p has coordinates (t −1j ti u i, j )i∈I, j ∈I
/ . Thus, T acts on this space with char-
−1 a1
acters t j ti , where i ∈ I, j ∈
/ I . Identifying  t1 · · · tn with (a1 , . . . , an ), all these
an

points lie in the open halfspace defined by i∈I ai > 0.

Example 78 We compute the class of O(1). The sheaf O(1) on G(k, n) was already
(nk)−1 via the Plücker embed-
k ∨ of O(1) on P
mentioned in Sect. 4.4.2: it is the pullback
ding. We can also describe O(1) as S , where S is the tautological vector bundle
on G(k, n) whose fiber over a point is the corresponding k-plane.
We can apply Theorem 68 to the result from Example 74 to replace the torus
action there with a different torus action, induced from the action on the Plücker
coordinates. By applying pullback formula (4.2), we find that the class [O(1)]T in
equivariant K -theory is the map

[O(1)]T : Gr (k, n)T → Z[Char(T )] : p I → ti1 · · · tik ,

where we wrote I = {i 1 , . . . , i k }.

Let p be a point in Gr (k, n) and M = M p be the corresponding matroid. Then T p

is a closed subvariety of Gr (k, n); in particular, it is given by a coherent sheaf.
We want to compute its class in T -equivariant K -theory, which is a map [T p]T :
Gr (k, n)T → Z[Char(T )]. As above, let p I ∈ Gr (k, n)T be the torus-invariant point
given by PJ = 0 for J = I , and let U I be the affine open neighbourhood U I of p I
defined by PI = 1.
If I is not a basis of M, then T p does not intersect U I , hence [T p]T ( p I ) = 0.
Hence, we will assume that I is a basis of M, i.e. that p ∈ U I .
The coordinate ring of T p ∩ U I is isomorphic to C[si−1 s j ], where si−1 s j is a gen-
erator if and only if (I − i) ∪ j is a basis of M. We will denote this ring by R M,I . This
ring should be viewed as a T -module, with t · si−1 s j = ti−1 t j si−1
s j . The
 Hilbert series
−1
of R M,I is a rational function with denominator dividing i∈I j ∈I / (1 − ti t j ).
Thus, by (4.1), 
[T p]T ( p I ) = Hilb(R M,I ) (1 − ti−1 t j ). (4.4)
i∈I j ∈I
/

Definition 79 For any lattice polytope P and v a vertex of P, we define conev(P)

to be the cone spanned by all vectors of the form u − v with u ∈ P. For I ∈ [n]k
,
106 A. Cameron et al.

we write cone I (M) := conee I (P(M)) if I is a basis of M, and cone I (M) := ∅

otherwise.

Since cone I (M) is the positive real span of all vectors e J − e I , where J ∈ B(M),
we find that Hilb(R M,I ) = Hilb(cone I (M)). So (4.4) can also be written as

[T p]T ( p I ) = Hilb(cone I (M)) (1 − ti−1 t j ). (4.5)
i∈I j ∈I
/

We note that (4.5) only depends on the matroid M and not on the chosen point p
or even the torus orbit T p. Moreover, the formulas make sense even for non-
representable matroids. Thus we can use them as a definition for the class in K -theory
for a matroid:
Definition 80 ([16]) For any rank k matroid M on [n], we define y(M) :
Gr (k, n)T → Z[Char(T )] by

y(M)( p I ) = Hilb(cone I (M)) (1 − ti−1 t j ).
i∈I j ∈I
/

Theorem 81 ([16, Proposition 3.3]) The class y(M) ∈ K T0 (Gr (k, n)T ) satisfies the
condition of Theorem 75, and hence defines a class in K T0 (Gr (k, n)).

4.8.5 Flag Matroids and the K -theory of Flag Varieties

In this section, we generalize the results from the previous section replacing
“matroids” by “flag matroids” and “Grassmannians” by “flag varieties”.
We first describe the equivariant K -theory of a flag variety Fl(k, n). The torus-
fixed points are given as follows: for every (set-theoretic) flag F = (F1 ⊆ . . . ⊆ Fs )
of rank k on [n], we define a (vector space) flag V1 ⊆ . . . ⊆ Vs by Vi = span({e j | j ∈
Fi }). We will denote the corresponding point in Fl(k, n) by p F . The Plücker coor-
dinates of p F are given by PS = 1 if S is a constituent of F and PS = 0 otherwise.
Here, the Plücker coordinates
 of a point in Fl(k, n) are the ones induced from the
embedding Fl(k, n) → G(ki , n).
We can also describe the one-dimensional torus orbits: let p F be a torus-fixed
point. We define S(F) to be the set of all pairs (i, j) ∈ [n] × [n] for which there
exists an such that i ∈ F and j ∈ / F . For every (i, j) ∈ S(F), we define a new
flag F  = Fi→ j by switching the roles of i and j. More precisely: if i ∈ Fr but j ∈
/ Fr ,
then Fr := (Fr − i) ∪ j; in any other case Fr := Fr . Then there is a unique one-
dimensional torus orbit between p F and p F  , and all one-dimensional torus orbits
arise in this way. The torus T acts on this orbit with character t −1
j ti .

Lemma 82 The action of T on Fl(k, n) is contracting.

4 Flag Matroids: Algebra and Geometry 107

Proof For every torus-fixed point p F , we consider the open neighbourhood U F

given
s by PFr = 0 for all r . Then U F ∼ = A N , where N = dim(Fl(k, n)) =
i=1 ki (ki+1 − ki ) (here ks+1 := n). We will denote the coordinates of a point q
P
in U F by (u i, j )(i, j)∈S(F) , where u i, j = FPr −i∪
F
j
for any r which satisfies i ∈ Fr
r
and j ∈ / Fr . Then t · q has coordinates (t −1
j ti u i, j )(i, j)∈S(F) . So T acts on U F with
−1
characters t j ti , (i, j) ∈ S(F). As before, identifying t1a1 · · · tnan with (a1 , . . . , an ),
 
all these characters lie on the open halfspace rs =1 i∈Fr ai > 0. 
Let p be a point in Fl(k, n), and let F be the corresponding flag matroid. We
want to compute the T -equivariant class of T p, which is a map [T p]T : Fl(k, n)T →
Z[Char(T )]. We fix a point p F ∈ Fl(k, n)T , and consider the affine neighbour-
hood U F described above.
If F is not a basis of F , then T p does not intersect U F , hence [T p]T ( p F ) = 0.
Thus, we will assume that F is a basis of F , i.e. that p ∈ U F .
The coordinate ring of T p ∩ U F is isomorphic to k[si−1 s j ], where si−1 s j is a
generator if and only if Fi→ j ∈ F . We will denote this ring by RF ,F . This ring should
be viewed as a T -module, with t · si−1 s j = ti−1 t j si−1 s j . 
The Hilbert series of this T -
module is a rational function with denominator dividing (i, j)∈S(F) (1 − ti−1 t j ). Thus,
by (4.1), 
[T p]T ( p F ) = Hilb(RF ,F ) (1 − ti−1 t j ). (4.6)
(i, j)∈S(F)

Definition 83 We define cone F (F ) to be the cone conee F (P(F )), as in Definition 79.
As before, we find that Hilb(RF ,F ) = Hilb(cone F (F )). Hence, (4.6) can also be
written as 
[T p]T ( p F ) = Hilb(cone F (F )) (1 − ti−1 t j ). (4.7)
(i, j)∈S(F)

As before, (4.7) only depends on the flag matroid F and not on the chosen point p
or even the torus orbit T p. Moreover the formulas make sense even for non-
representable flag matroids. Thus we can use them as a definition for the class
in K -theory for a flag matroid:
Definition 84 For any rank k flag matroid F on [n], we define y(F ) : Fl(k, n)T →
Z[Char(T )] by

y(F )( p F ) = Hilb(cone F (F )) (1 − ti−1 t j ).
(i, j)∈S(F)

Proposition 85 The class y(F ) ∈ K T0 (Fl(k, n)T ) satisfies the condition of

Theorem 75, and hence defines a class in K T0 (Fl(k, n)).
Proof The proof is a straightforward generalisation of the proof of [16, Proposition
3.3]. 
108 A. Cameron et al.

Fig. 4.5 The class

in K -theory of a flag matroid

Example 86 Let F be the flag matroid of Example 54. We first com-

pute y(F )( p F ), where F is the flag 2 ⊆ 12 (so e F = (1, 2, 0)). From Fig. 4.4,
it is clear that cone F (F ) = cone((1, −1, 0), (0, −1, 1)), which has Hilbert
1
series (1−t −1 t )(1−t −1
t )
. Furthermore, we have S(F) = {(2, 1), (2, 3), (1, 3)}. We find
2 1 2 3

that y(F )( p F ) = 1 − t1−1 t3 . We can do the same for the other torus-fixed points. The
result is summarised in Fig. 4.5.

4.8.6 The Tutte Polynomial via K -theory

In [16], a geometric description of the Tutte polynomial of a matroid is given. Con-

sider the following diagram, where all the maps are natural projections or inclusions:
Fl(1, k, n − 1; n)
πk

G(k, n) π1(n−1) Fl(1, n − 1; n)

Pn−1 × Pn−1  G(1, n) × G(n − 1, n)

Generalising Example 67, one can show that K 0 (Pn−1 × Pn−1 ) ∼

= Z[α, β]/(α n , β n ),
where α and β are the structure sheaves of hyperplanes.
Theorem 87 ([16, Theorem 7.1]) The following equality holds:

(π1(n−1) )∗ πd∗ (Y (M) · [O(1)]) = TM (α, β),

where Y (M) is the class associated to the matroid M in the non-equivariant K-theory
of the Grassmannian.
In other words, the Tutte polynomial of a matroid can be viewed as a Fourier-Mukai
transform of its associated class in K -theory.
We can now generalize this construction to get a definition of the Tutte polynomial
of a flag matroid.

Definition 88 Consider the following diagram.

4 Flag Matroids: Algebra and Geometry 109

Fl(1, k, n − 1; n)
πk

Fl(k; n) π1(n−1) Fl(1, n − 1; n)

Pn−1 × Pn−1

Let F be a flag matroid on [n] of rank k, and let Y (F ) ∈ K 0 (Fl(k; n)) be its class in
non-equivariant K -theory, as in Definition 84. Then the K -theoretic Tutte polynomial
of F is defined to be

TF (α, β) := (π1(n−1) )∗ πd∗ (Y (F ) · [O(1)]).

We computed the Tutte polynomial for some small examples using Sage [51],
Macaulay2 [25], and Normaliz [5]. Our code is available at [47]. The program first
computes the equivariant class (π1(n−1) )∗ πd∗ (y(M) · [O(1)]) ∈ K T0 (Pn−1 × Pn−1 )
using equivariant localisation, and then computes the underlying non-equivariant
class.

Remark 89 After we finished the article Christopher Eur implemented the algorithm
in Macaulay2 [14].

Example 90 We consider again the flag matroid of Examples 54 and 86. We

first compute y(F ) · [O(1)], which is displayed in Fig. 4.6. The two projec-
tions from Fl(1, k, 2; 3) = Fl(1, 1, 2, 2; 3) to Fl(1, 2; 3) are isomorphisms, hence
pulling back and pushing forward along them does nothing. Next we need to push our
class X = y(F ) · [O(1)] ∈ K T0 (Fl(1, 2; 3)) to a class Y ∈ K T0 (Pn−1 × Pn−1 ), using
formula (4.3).
The T -fixed points of Pn−1 × Pn−1 are given by pairs p = ( , H ), where ∈
G(1, 3)T = {e1 , e2 , e3 } and H ∈ G(2, 3)T = {e1 , e2 , e1 , e3 , e2 , e3 }.
If ⊂ H , then Y ( p) = 0. If ⊂ H , then p ∈ Fl(1, 2; 3) ⊂ P2 × P2 . Since we
are pushing forward along an embedding, the formula (4.3) has a simple form: we
can find characters χ1 , χ2 , χ3 , η and open neighbourhoods p  U1 ⊂ Fl(1, 2; 3)
and p  U2 ⊂ P2 × P2 , such that T acts on U1 with characters χ1 , χ2 , χ3 , and on U2
with characters χ1 , χ2 , χ3 , η. Then (4.3) becomes:

Y ( p) = (1 − η−1 )X ( p).

Fig. 4.6 y(F ) · [O(1)]

110 A. Cameron et al.

Consider for example p = (e1 , e1 , e2 ). Then p ∈ Fl(1, 2; 3) has an open
neighbourhood where T acts by characters t2 t3−1 , t1 t3−1 , t1 t2−1 , while p ∈ P2 × P2
has an open neighbourhood where T acts by characters t2 t3−1 , t1 t3−1 , t1 t3−1 , t1 t2−1 . We
compute that

Y ((e1 , e1 , e2 )) = t12 t2 (1 − t1−1 t3 )(1 − t1−1 t3 ) = t2 (t1 − t3 )2 .

Similarily, we find that

Y ((e1 , e1 , e3 )) = t3 (t1 − t2 )2

Y ((e3 , e1 , e3 )) = t3 (t1 − t2 )(t3 − t2 )
Y ((e2 , e1 , e2 )) = t2 (t1 − t3 )(t2 − t3 )

and Y ( p) = 0 in all other cases.

Finally, we need to find the underlying class in non-equivariant K -theory. This
is quite tedious to do by hand, so we just refer to the algorithm provided at [47] for
this. In the end, we find that

TF (x, y) = x 2 y 2 + x 2 y + x y 2 + x 2 + x y.

Example 91 As another example, consider the uniform flag matroid U(2,3);5

of rank (2, 3) on [5] (that is, the constituents of U(2,3);5 are the uniform
matroids U2,5 and U3,5 ). Using our program, we find that its K -theoretic Tutte poly-
nomial TU(2,3):5 (x, y) equals

x 3 y 3 + 2x 3 y 2 + 2x 2 y 3 + 3x 3 y + 8x 2 y 2 + 3x y 3 + 4x 3 +
8x 2 y + 8x y 2 + 4y 3 + 2x 2 + 4x y + 2y 2 .

After the preprint of the article appeared Christopher Eur found an example of
a K -theoretic Tutte polynomial TU(1,3):5 (x, y) with a strictly negative coefficient.

4.9 Open Problems

Our definition of the Tutte polynomial of a flag matroid is admittedly quite involved.
It is natural to wonder whether there is an easier definition, avoiding geometry:
Problem 92 Is there a purely combinatorial description of the K -theoretic Tutte
polynomial of a flag matroid? In particular, can one obtain the K -theoretic Tutte
polynomial from the Tutte polynomials of the constituents?
For matroids, one can define the characteristic polynomial (also called chromatic
polynomial, as it generalises the chromatic polynomial of a graph) by
4 Flag Matroids: Algebra and Geometry 111

χ M (λ) = (−1)r (M) TM (1 − λ, 0).

In 2015, Adiprasito, Huh and Katz proved the following conjecture by Rota-Heron-
Welsh:
Theorem 93 ([1]) Let wi (M) be the absolute value of the coefficient of λr (M)−i in
the characteristic polynomial of M. Then the sequence wi (M) is log-concave:

wi−1 (M)wi+1 (M) ≤ wi (M)2 for all 1 ≤ i < r (M).

Since we now have a definition for the Tutte polynomial of a flag matroid, we can
define the characteristic polynomial of a rank k flag matroid F by

χF (λ) = (−1)r (F ) TF (1 − λ, 0),


where r (F ) := |k| := ki .
Conjecture 94 Theorem 93 holds for the characteristic polynomial of an arbitrary
flag matroid.
In Examples 90 and 91, the characteristic polynomials are −λ2 + 2λ − 1 and 4λ3 −
14λ2 + 16λ − 6, respectively. Thus, we see that Conjecture 94 holds for these exam-
ples.
Flag matroids are a special class of Coxeter matroids. Hence, another possible
direction of research would be:
Problem 95 Explore how our constructions and results could be generalised to arbi-
trary Coxeter matroids.
Flag matroids can also be viewed as a special class of polymatroids. In particular,
we can apply the construction of Sect. 4.3.1 to them.
Problem 96 Is there a connection between the Tutte polynomial of a polymatroid,
as defined by Cameron and Fink, and our construction of the Tutte polynomial of
flag matroid? How about the construction from Remark 31?
Next, we note that by Sect. 4.7 we know how to associate to a rank k representable
r -polymatroid on [n] a class in K T0 (G(k, r n)).
Problem 97 Can we associate a class in K T0 (G(k, r n)) to a non-representable
r -polymatroid? Can this be used to define a K -theoretic Tutte polynomial for
r -polymatroids?
Another possible generalisation, brought to our attention by Alastair Craw after we
wrote the article, would be the class of quiver flag varieties [10]. We have not yet
pursued this direction.
Finally, we could apply the construction of Sect. 4.8.6 to any subvariety of a
Grassmannian (or even a flag variety), not just to torus orbits. It could be interesting
to study the properties of this invariant.
112 A. Cameron et al.

Acknowledgements We would like to thank Christopher Eur for useful comments and sharing
with us interesting examples. We thank Alastair Craw and the anonymous reviewers for their careful
reading and suggestions to improve the manuscript. The third author was supported by the Polish
National Science Centre grant no. 2015/19/D/ST1/01180.

References

1. Adiprasito, K., Huh, J., Katz, E.: Hodge theory for combinatorial geometries. Ann. Math.
188(2), 381–452 (2018)
2. Anderson, D.: Introduction to equivariant cohomology in algebraic geometry. In: Contributions
to Algebraic Geometry, EMS Series of Congress Reports, pp. 71–92. European Mathematical
Society, Zürich (2012)
3. Borovik, A.V., Gel’fand, I.M., White, N.: Coxeter Matroids, Progress in Mathematics, vol.
216. Birkhäuser Boston Inc, Boston, MA (2003)
4. Brändén, P.: Obstructions to determinantal representability. Adv. Math. 226(2), 1202–1212
(2011)
5. Bruns, W., Ichim, B.: Normaliz: algorithms for affine monoids and rational cones. J. Algebra
324(5), 1098–1113 (2010)
6. Brylawski, T.: Constructions. In: Theory of Matroids, The Encyclopedia of Mathematics and
its Applications, vol. 26, pp. 127–223. Cambridge University Press, Cambridge (1986)
7. Brylawski, T., Oxley, J.: The Tutte polynomial and its applications. In: Matroid Applications, the
Encyclopedia of Mathematics and Its Applications, vol. 40, pp. 123–225. Cambridge University
Press, Cambridge (1992)
8. Cameron, A., Fink, A.: The Tutte polynomial via lattice point counting (2018).
arXiv:1802.09859
9. Cox, D.A., Little, J.B., Schenck, H.K.: Toric Varieties, Graduate Studies in Mathematics, vol.
124. American Mathematical Society, Providence, RI (2011)
10. Craw, A.: Quiver flag varieties and multigraded linear series. Duke Math. J. 156(3), 469–500
(2011)
11. Dupont, C., Fink, A., Moci, L.: Universal Tutte characters via combinatorial coalgebras. Algebr.
Comb. 1(5), 603–651 (2018)
12. Edmonds, J.: Submodular functions, matroids, and certain polyhedra. In: Combinatorial Struc-
tures and their Applications (Proc. Calgary Internat. Conf., Calgary, Alta., 1969), pp. 69–87.
Gordon and Breach, New York (1970)
13. Ellis-Monaghan, J.A., Merino, C.: Graph polynomials and their applications I: the Tutte poly-
nomial. In: Structural Analysis of Complex Networks, pp. 219–255. Birkhäuser/Springer, New
York (2011)
14. Eur, C.: Webpage. https://github.com/chrisweur
15. Feichtner, E.M., Sturmfels, B.: Matroid polytopes, nested sets and Bergman fans. Port. Math.
(N.S.) 62(4), 437–468 (2005)
16. Fink, A., Speyer, D.E.: K -classes for matroids and equivariant localization. Duke Math. J.
161(14), 2699–2723 (2012)
17. Fulton, W.: Introduction to Toric Varieties, Annals of Mathematics Studies, vol. 131. Princeton
University Press, Princeton, NJ (1993). The William H. Roever Lectures in Geometry
18. Fulton, W.: Intersection theory, Ergebnisse der Mathematik und ihrer Grenzgebiete. 3. Folge.
A Series of Modern Surveys in Mathematics [Results in Mathematics and Related Areas. 3rd
Series. A Series of Modern Surveys in Mathematics], vol. 2, 2nd edn. Springer, Berlin (1998)
19. Fulton, W., Harris, J.: Representation Theory, Graduate Texts in Mathematics, vol. 129.
Springer, New York (1991). A first course, Readings in Mathematics
20. Gale, D.: Optimal assignments in an ordered set: an application of matroid theory. J. Comb.
Theory 4, 176–180 (1968)
4 Flag Matroids: Algebra and Geometry 113

21. Gel’fand, I.M., Goresky, R.M., MacPherson, R.D., Serganova, V.V.: Combinatorial geometries,
convex polyhedra, and Schubert cells. Adv. Math. 63(3), 301–316 (1987)
22. Gel’fand, I.M., Serganova, V.V.: Combinatorial geometries and the strata of a torus on homo-
geneous compact manifolds. Uspekhi Mat. Nauk 42(2(254)), 107–134, 287 (1987)
23. Gel’fand, I.M., Serganova, V.V.: On the general definition of a matroid and a greedoid. Dokl.
Akad. Nauk SSSR 292(1), 15–20 (1987)
24. Goresky, M., Kottwitz, R., MacPherson, R.: Equivariant cohomology, Koszul duality, and the
localization theorem. Invent. Math. 131(1), 25–83 (1998)
25. Grayson, D.R., Stillman, M.E.: Macaulay2, a software system for research in algebraic geom-
etry. http://www.math.uiuc.edu/Macaulay2/
26. Greene, C.: A multiple exchange property for bases. Proc. Am. Math. Soc. 39, 45–50 (1973)
27. Herzog, J., Hibi, T.: Discrete polymatroids. J. Algebr. Combin. 16(3), 239–268 (2002, 2003)
28. Huh, J.: Milnor numbers of projective hypersurfaces and the chromatic polynomial of graphs.
J. Am. Math. Soc. 25(3), 907–927 (2012)
29. Kálmán, T.: A version of Tutte’s polynomial for hypergraphs. Adv. Math. 244, 823–873 (2013)
30. Kim, S.: Flag enumerations of matroid base polytopes. J. Combin. Theory Ser. A 117(7),
928–942 (2010)
31. Lafforgue, L.: Chirurgie Des Grassmanniennes. CRM Monograph Series, vol. 19. American
Mathematical Society, Providence, RI (2003)
32. Lasoń, M.: On the toric ideals of matroids of a fixed rank. Selecta Math. (N.S.) 27(2), Paper
No. 18, 17 (2021)
33. Lasoń, M., Michałek, M.: On the toric ideal of a matroid. Adv. Math. 259, 1–12 (2014)
34. Manivel, L.: Symmetric functions, Schubert polynomials and degeneracy loci. SMF/AMS Texts
and Monographs, vol. 6. American Mathematical Society, Providence, RI; Société Mathéma-
tique de France, Paris (2001). Translated from the 1998 French original by John R. Swallow,
Cours Spécialisés [Specialized Courses], 3
35. McMullen, P.: Lattice invariant valuations on rational polytopes. Arch. Math. (Basel) 31(5),
509–516 (1978/79)
36. Merino, C., Ramírez-Ibáñez, M., Rodríguez-Sánchez, G.: The Tutte polynomial of some
matroids. Int. J. Comb. Art. ID 430859, 40 (2012)
37. Merkur’ev, A.S.: Comparison of the equivariant and the standard K -theory of algebraic vari-
eties. Algebra i Analiz 9(4), 175–214 (1997)
38. Michałek, M.: Selected topics on toric varieties. In: The 50th anniversary of Gröbner bases.
Adv. Stud. Pure Math. Math. Soc. Japan, Tokyo 77, 207–252 (2018)
39. Michałek, M., Sturmfels, B.: Invitation to nonlinear algebra, vol. 211. American Mathematical
Society (2021)
40. Miller, E., Sturmfels, B.: Combinatorial Commutative Algebra, Graduate Texts in Mathematics,
vol. 227. Springer, New York (2005)
41. Mnëv, N.E.: The universality theorems on the classification problem of configuration varieties
and convex polytopes varieties. In: Topology and geometry—Rohlin Seminar. Lecture Notes
in Mathematics, vol. 1346, pp. 527–543. Springer, Berlin (1988)
42. Nielsen, H.A.: Diagonalizably linearized coherent sheaves. Bull. Soc. Math. France 102, 85–97
(1974)
43. Oxley, J.: Matroid theory. Oxford Graduate Texts in Mathematics, vol. 21, 2nd edn. Oxford
University Press, Oxford (2011)
44. Oxley, J., Whittle, G.: A characterization of Tutte invariants of 2-polymatroids. J. Combin.
Theory Ser. B 59(2), 210–244 (1993)
45. Reiner, V.: Lectures on Matroids and Oriented Matroids (2005). https://www-users.math.umn.
edu/~reiner/Talks/Vienna05/Lectures.pdf
46. Schrijver, A.: Combinatorial optimization. Polyhedra and efficiency. Algorithms and Combi-
natorics, vol. C, 24. Springer, Berlin (2003). Disjoint paths, hypergraphs, Chapters 70–83
47. Seynnaeve, T.: Webpage. https://mathsites.unibe.ch/seynnaeve/index.html
48. Stanley, R.P.: Enumerative combinatorics. Cambridge Studies in Advanced Mathematics, vol. 1,
49. Cambridge University Press, Cambridge (1997). With a foreword by Gian-Carlo Rota,
Corrected reprint of the 1986 original
114 A. Cameron et al.

49. Sturmfels, B.: Gröbner Bases and Convex Polytopes. University Lecture Series, vol. 8. Amer-
ican Mathematical Society, Providence, RI (1996)
50. Sturmfels, B.: Algorithms in Invariant Theory. Texts and Monographs in Symbolic Computa-
tion, 2nd edn. Springer, Wien, NewYork, Vienna (2008)
51. The Sage Developers: SageMath, the Sage Mathematics Software System (Version 8.4.0)
(2018). http://www.sagemath.org
52. Thistlethwaite, M.B.: A spanning tree expansion of the Jones polynomial. Topology 26(3),
297–309 (1987)
53. Tutte, W.T.: A contribution to the theory of chromatic polynomials. Canad. J. Math. 6, 80–91
(1954)
54. Vezzosi, G., Vistoli, A.: Higher algebraic K -theory for actions of diagonalizable groups. Invent.
Math. 153(1), 1–44 (2003)
55. Welsh, D.J.A., Merino, C.: The Potts Model and the Tutte Polynomial. pp. 1127–1152 (2000).
Probabilistic techniques in equilibrium and nonequilibrium statistical physics
56. White, N.L.: The basis monomial ring of a matroid. Adv. Math. 24(3), 292–297 (1977)
57. Whitney, H.: On the abstract properties of linear dependence. Am. J. Math. 57(3), 509–533
(1935)
Chapter 5
Classification of Minimal Polygons with
Specified Singularity Content

Daniel Cavey and Edwin Kutas

Abstract It is known that there are only finitely many mutation-equivalence classes
with a given singularity content, and each of these equivalence classes contains
only finitely many minimal polygons. We describe an efficient algorithm to clas-
sify these minimal polygons. To illustrate this algorithm we compute all mutation-
equivalence classes of Fano polygons with basket of singularities given by (i) B =
{m 1 × 13 (1, 1), m 2 × 16 (1, 1)} and (ii) B = {m × 15 (1, 1)}.

Keywords Fano variety · qG-deformation · Mutation · Singularity content

5.1 Introduction

Let N be a lattice. A polytope P in NR = N ⊗ R is a set of the form

  
P= λu u : λu ≥ 0 and λu = 1 ,
u∈S u∈S

where S ⊂ NR is a finite set of points. A Fano polytope is a full-dimensional convex

polytope such that the vertices vert(P) ⊂ N are all primitive, and that the origin
lies in the strict interior of P. Polytopes are considered up to GL(N )-equivalence.
When N is a rank-two lattice, P is known as a Fano polygon.
The span of each face E of a Fano polygon P, that is R≥0 E, defines a cone. Equiv-
alently this is the polyhedral cone whose primitive generating vertices are given by
the endpoints of E. This describes a fan in NR corresponding to P, which in turn
determines a toric del Pezzo surface X P . Many properties of X P have combinato-

D. Cavey (B)
School of Mathematical Sciences, University of Nottingham, Nottingham NG7 2RD, UK
e-mail: [email protected]
E. Kutas
Mathematics Institute, University of Warwick, Coventry CV4 7AL, UK
e-mail: [email protected]

© Springer Nature Switzerland AG 2022 115

A. M. Kasprzyk and B. Nill (eds.), Interactions with Lattice Polytopes,
Springer Proceedings in Mathematics & Statistics 386,
https://doi.org/10.1007/978-3-030-98327-7_5
116 D. Cavey and E. Kutas

rial analogues in the Fano polygon P; examples include the singularities and the
anticanonical degree (−K X P )2 . Toric geometry can be further studied in [6, 8].
The dual lattice of N is M = Hom(N , Z) with the pairing ·, · : N × M → Z.
The lattice length of a line segment E ⊂ NR is given by the value |E ∩ N | − 1.
The lattice height of a line segment is given by the lattice distance from the origin:
that is, given the unique primitive inward pointing normal n E of E belonging to M,
the height is given by |v, n E |, for any v ∈ E.
The motivation for the paper comes from an attempt to classify Fano varieties using
mirror symmetry. An understanding of mirror symmetry can be gained from Coates–
Corti–Galkin–Golyshev–Kasprzyk [4]. Mirror symmetry suggests that classifying
Fano polytopes could help to classify Fano varieties.
More specifically we study Fano polytopes up to mutation-equivalence. A muta-
tion is a combinatorial operation on a Fano polygon P ⊂ NR introduced by Akhtar–
Coates–Galkin–Kasprzyk [2], and is described in Sect. 5.2. An important mutation
invariant of Fano polygons known as singularity content was introduced in [3] by
Akhtar–Kasprzyk. This is a combinatorial property of P that describes the singu-
larities of the corresponding Fano toric variety X P . In this setting Fano varieties
are considered up to qG-deformation, see Coates–Corti–Kasprzyk et al. and Kollár–
Shepherd-Barron [1, 12].

Definition 1 ([1]) A del Pezzo surface with cyclic quotient singularities that admits
a qG-deformation to a normal toric del Pezzo surface is said to be of class TG.

The reason we consider Fano polytopes and Fano varieties up to their respective
equivalence classes lies in the following conjecture.

Conjecture 2 ([1, Conjecture A]) There exists a bijective correspondence between

the set of mutation-equivalence classes of Fano polygons and the set of qG-
deformation equivalence classes of locally qG-rigid class TG del Pezzo surfaces
with cyclic quotient singularities.

Recent results from Corti–Heuberger [5] and Kasprzyk–Nill–Prince [11] support

this conjecture.

Theorem 3 ([11, Theorem 1.2]) There are precisely ten mutation-equivalence

classes of Fano polygons such that the toric del Pezzo surface X P has only T-
singularities. They are in bijective correspondence with the ten families of smooth
del Pezzo surfaces.

Theorem 4 ([5, 11]) There are precisely 29 qG-deformation families of del Pezzo
surfaces with m ≥ 1 singular points of type 13 (1, 1). Precisely 26 of these are of class
TG, and furthermore are in bijective correspondence with 26 mutation-equivalence
classes of Fano polygons with singularity content (n, {m × 13 (1, 1)}), where m ≥ 1.

Within this context, the aim is to classify Fano polygons with a given singu-
larity content. Assuming Conjecture 2 holds, this is equivalent to a classification
of del Pezzo surfaces admitting a toric degeneration which have the singularities
5 Classification of Minimal Polygons with Specified Singularity Content 117

described by the specified singularity content. We will use our work to provide more
examples for understanding Conjecture 2. The main result of this paper is an efficient
algorithm (described in Sect. 5.4) to produce representations (called minimal poly-
gons) for the mutation-equivalence classes in a classification of Fano polygons with
a specified singularity content in the case where the maximal height of the edges of
each Fano polygon is given by an edge representing a non qG-smoothable singular-
ity. As a corollary to the algorithm the following classifications (derived in Sects. 5.5
and 5.6 respectively) have been completed:
Theorem 5 There are precisely 14 mutation-equivalence classes of Fano polygons
with singularity content (n, {m 1 × 13 (1, 1), m 2 × 16 (1, 1)}) with m 1 ≥ 0, m 2 > 0.
Theorem 6 There are precisely 12 mutation-equivalence classes of Fano polygons
with singularity content (n, {m × 15 (1, 1)}) with m > 0.
The reason for computing these classifications is that they are both in some sense
the simplest cases after the 13 (1, 1) classification of Theorem 4. A 16 (1, 1) singularity,
like a 13 (1, 1), is represented by an edge of a Fano polygon P of height 3. This
is the smallest possible height for an edge representing any singularity that is not
smoothable by a qG-deformation. Whereas 5 is the smallest value for r after 3 for
which a r1 (1, 1) singularity is not smoothable by a qG-deformation.
If one of the Fano polygons appearing in the classifications is (via mutation if
necessary) a triangle, then the corresponding toric variety will be the quotient of a
weighted projective space as described in [9]. In particular let ρ0 , ρ1 , ρ2 ∈ N be the
primitive generators of the rays for the fan of a Fano polygon P. Suppose that ρi
satisfy the equation
λ0 ρ0 + λ1 ρ1 + λ2 ρ2 = 0,

and span the lattice N . Then X P = P(λ0 , λ1 , λ2 ).

5.2 Mutations of Fano Polygons and Singularity Content

5.2.1 Mutations

Recall the definition of the Minkowski sum of lattice polygons.

Definition 7 Let P, Q ⊂ NR be two lattice polytopes. Define the Minkowski sum
of P and Q by
P + Q = { p + q : p ∈ P, q ∈ Q}.

By convention P + ∅ = ∅.
Let P ⊂ NR be a Fano polygon, and E be an edge of P. Consider the primitive
inward pointing normal n E ∈ M of this edge. This vector acts as a grading function
on P. For h ∈ Z, define
118 D. Cavey and E. Kutas

ωh (P) = conv{v ∈ N ∩ P : n E (v) = h}.

Note that ωh (P) may be empty (indeed it will be for infinitely many values of h)
and that ω−r E (P) = E, where r E is the height of E. Choose v E to be a primitive
vector of N such that n E (v E ) = 0. Note this is uniquely defined up to sign in the two
dimensional case. Set F = conv{0, v E }; a line of lattice length 1 and height 0, that
is parallel to E.
Definition 8 For all h < 0, suppose that there exists G h ⊂ NR such that

{v ∈ vert(P) : n E (v) = h} ⊆ G h + |h|F ⊆ ωh (P).

In the case ωh (P) = ∅ the inclusion holds taking G h = ∅. Then define the mutation
of P with respect to n E , F and G h to be
  
mut (n E ,F) (P) = conv Gh ∪ (ωh (P) + h F) ⊂ NR .
h<0 h≥0

Example 9 Consider the Fano polygon P = conv{(1, 0), (0, 1), (−5, −1)} corre-
sponding to the weighted projective space P(1, 1, 5). This polygon will be used
as a running
 example throughout
the paper. Mutate P with respect to the edge
E = conv (1, 0), (0, 1) . The primitive inner pointing normal is given by
n E = (−1, −1) ∈ M. This describes
a grading on the points of N as shown in Fig. 5.1.
Set F = conv 0, (1, −1) , a primitive slice at height 0. Choose G −1 = {(0, 1)}
which satisfies the required inclusion:
 
(0, 1), (1, 0) ⊆ G −1 + F ⊆ conv (0, 1), (1, 0) .

For h < −1, ωh (P) = ∅, so trivially choose G h = ∅. Calculating the mutation of P

with respect to the primitive inner point normal n E , the factor F and the polygon G −1
obtain:

-1

6 5 4 3 2 1 0

Fig. 5.1 The grading induced by n E = (−1, −1) ∈ M

5 Classification of Minimal Polygons with Specified Singularity Content 119

Fig. 5.2 The polygon Q

associated with P(1, 5, 36)

Q = mut (n E ,F) (P)

         
= conv G −1 ∪ ω0 (P) ∪ ω1 (P) + F ∪ ω2 (P) + 2F ∪ · · · ∪ ω6 (P) + 6F


= conv (0, 1), (−5, −1), (1, −7) .

Q corresponds to the toric variety P(1, 5, 36). Informally the mutation subtracts
one copy of F from P along the edge E, and adds six copies of F at the opposite
vertex of P which is at (−5, −1). This is illustrated in Fig. 5.2.
Note mut (n E ,F) (P) is independent of the choice for G h . If there is no possible
choice of G h , then the mutation with respect to n E does not exist.
Lemma 10 ([11, Lemma 2.3]) Let E be an edge of a Fano polygon P with primitive
inner normal vector n E ∈ M. Then P admits a mutation with respect to n E if and
only if
|E ∩ N | − 1 ≥ r E .

Applying Lemma 10 to the polygon P in Example 9 gives that the edge conv{(0, 1),
(−5, −1)} does not admit a mutation: it has lattice length 1 and height 5.
There are a number of additional properties of mutations.
1. The choice of v E is not important: mut (n E ,F) (P) is isomorphic to mut (n E ,−F) (P)
via a GL(N )-equivalence.
2. Mutation is invertible: If Q = mut (n E ,F) (P), then P = mut (−n E ,F) (Q).
3. P is a Fano polytope if and only if mut (n E ,F) (P) is a Fano polytope
[2, Proposition 2].
Definition 11 Let P, Q ⊂ NR be two Fano polygons. Then P and Q are mutation-
equivalent if there exists a finite sequence of polygons P0 , P1 , . . . , Pn such that
P0 ∼
= P, Pn ∼ = Q and, Pi+1 = mut(ni ,Fi ) (Pi ) for some n i and Fi , for all
i ∈ {0, . . . , n − 1}.
120 D. Cavey and E. Kutas

Mutation-equivalence defines an equivalence relation.

5.2.2 Singularity Content

Recall the definition of singularity content introduced in [3].

Consider the action of μ R , the cyclic group of order R, on C2 by
(x, y) → ( a x,  b y) where  is an Rth root of unity. Recall that a quotient singularity
R
1
(a, b) is given by the germ of the origin of Z = Spec(C[x, y]μ R ).
A cyclic quotient singularity is a quotient singularity R1 (a, b) such that
gcd(R, a) = 1 and gcd(R, b) = 1. In this situation, set k = gcd(a + b, R) so that
a + b = k c̃ and R = kr , and let c = c̃a −1 . The cyclic quotient singularity can be
written as kr1 (1, kc − 1).

Definition 12 A cyclic quotient singularity 1

kr
(1, kc − 1) is a T-singularity if r | k.

For an arbitrary T-singularity, let k = nr , so the singularity can be written in the

form nr12 (1, nr c − 1). If n = 1, the singularity is known as a primitive T-singularity.
Kollár–Shepherd-Barron [12] show a cyclic quotient singularity is a T-singularity if
and only if it admits a qG-smoothing.

Definition 13 A cyclic quotient singularity 1

kr
(1, kc − 1) is an R-singularity if
k < r.

A singularity is an R-singularity if and only if it is rigid under qG-deformation.

Consider an arbitrary cyclic quotient singularity kr1 (1, kc − 1). By the Euclidean
algorithm there exists unique non-negative integers n and
k0 such that k = nr + k0 . If k0 = 0, then the singularity is a T-singularity and
is qG-smoothable. If k0 > 0, then the singularity is qG-deformation equivalent to
a k10 r (1, k0 c − 1) cyclic quotient singularity.

Definition 14 ([3]) Let σ = kr1 (1, kc − 1) be a cyclic quotient singularity, and

k = nr + k0 . The residue of σ is given by

∅, if k0 = 0,
res(σ ) = 1
k0 r
(1, k0 c − 1), otherwise.

The singularity content of σ is the pair

 
SC(σ ) = n, res(σ ) .

This discussion of T-singularities and R-singularities has a natural description in

the language of cones. Each cone of a Fano polygon corresponds to a singularity
 of the
corresponding toric variety. Specifically a cone C defines an edge E = conv u 1 , u 2 ,
5 Classification of Minimal Polygons with Specified Singularity Content 121

where u 1 and u 2 are primitive

generators for the rays of C. Choose a point v in the
interior of conv 0, u 1 , u 2 such that

conv 0, u i , v ∩ N = {0, u i , v}, for some i ∈ {1, 2}.

Expressing v in terms of u 1 and u 2 , obtain v = pu 1 + qu 2 where p = a/R and

q = b/R. The cone corresponds to the singularity σC = R1 (a, b) on X P .
Let C ⊂ NR be a cone with generating rays described by the primitive lattice
points p0 and p1 . Consider E = conv{ p0 , p1 }. Let r be the height of E and k be
the lattice length. By the Euclidean algorithm k = nr + k0 . Divide C into separate
sub-cones C0 , . . . , Cn where C1 , . . . , Cn (known as T-cones) have lattice length r ,
and C0 has lattice length k0 and is known as an R-cone. Importantly T-cones cor-
respond to T-singularities on X P and R-cones to R-singularities. The decomposi-
tion of C into sub-cones is analogous to the fact that the cyclic quotient singu-
larity σC = kr1 (1, kc − 1) can be qG-smoothed to the cyclic quotient singularity
1
k0 r
(1, k0 c − 1) which is described by the sub-cone C0 . This discussion is indepen-
dent of the choice of decomposition of C.

Definition 15 Let P ⊂ NR be a Fano polygon. Label the edges of P in clock-

wise order E 1 , . . . E k . Each edge E i describes a cone which corresponds to a
cyclic quotient singularity
 σi . Define
 thesingularity content of an edge to be
k
SC(E i ) = SC(σi ) = n i , res(σi ) . Set n = i=1 n i and B = {res(σ1 ), . . . , res(σk )},
where B is an ordered set known as the basket of R-singularities. Then the singularity
content of P is defined to be  
SC(P) = n, B .

Example 16 Consider P = conv (0, 1), (1, 0), (−5, −1) from Example 9. The
edges conv (1, 0), (0, 1) and conv (1, 0), (−5, −1) are both of length 1 and
height 1 and so contribute a single T-cone each. The final edge conv (1, 0), (0, 1)
is of height 5 and length 1 and so is a single R-cone. It remains to calcu-
late the singularity corresponding to this edge. Using the earlier notation, the
only choice for v is (−1, 0), and finding the unique solution to the equation
(−1, 0) = p(0, 1) + q(−5, −1) as p = q = 15 shows that the edge represents a 15 (1, 1)
R-singularity. Therefore SC(P) = (2, { 15 (1, 1)}).

Proposition 17 ([3, Proposition 3.6]) The singularity content is an invariant of Fano

polygons under mutation.
 
Indeed in Example 9, it is routine to check that SC(Q) = 2, { 15 (1, 1)} which is
the same as the singularity content for P.
122 D. Cavey and E. Kutas

5.2.3 Hirzebruch–Jung Continued Fractions and

Applications to Algebraic Geometry

There is information about the del Pezzo surface X P corresponding to a polygon P

written into the singularity content; X P is qG-deformation
 equivalent
 to a del Pezzo
surface X such that the topological Euler number χ X \Sing(X ) = n and the sin-
gular points are given by Sing(X ) = B. The anticanonical degree and Hilbert series
of X P are totally determined by the singularity content. See [1, 3].

Definition 18 Let p, q ∈ Z0 be coprime. Then the Hirzebruch–Jung continued frac-

tion of p/q is given by the continued fraction of the form:

p 1
= a1 −
q a2 − a −1 1
..
3

.
= [a1 , . . . , ak ].

Given a cyclic quotient singularity σ = R1 (1, a − 1), construct the variety

Z = Spec(C[x, y]μ R ) as in the definition of quotient singularity. Information about
a minimal resolution of Z can be calculated from the Hirzebruch–Jung continued
fraction of R/(a − 1). Consider the minimal resolution π : Y → Z with


k
KY = π ∗ K Z + di E i .
i=1

Let [a1 , . . . , akσ ] be the Hirzebruch–Jung continued fraction of R/(a − 1). Then
the values −ai are the self-intersection numbers of the exceptional divisors E i . Addi-
tionally define αi , βi for i ∈ {1, . . . , kσ } by:

α1 = βkσ = 1,
αi
= [ai−1 , . . . , a1 ], for i ∈ {2, . . . , kσ },
αi−1

βi
= [ai+1 , . . . , akσ ], for i ∈ {1, . . . , kσ − 1}.
βi+1

Note that the αi are in increasing order, and the βi are in decreasing order. The
discrepancy of E i is given by di = −1 + (αi + βi )/R. For further reading on minimal
resolutions, see Reid [15].

Proposition 19 ([3, Proposition 3.3, Corollary 3.5]) Let P be a Fano polygon and
let X P be the corresponding toric surface. Suppose P has singularity content (n, B).
Then
5 Classification of Minimal Polygons with Specified Singularity Content 123

(−K X P )2 = 12 − n − Aσ ,
σ ∈B

k σ 2 kσ −1
where Aσ = kσ + 1 − i=1 di ai + 2 i=1 di di+1 . Furthermore the anticanonical
Hilbert series of X P admits a decomposition
 
1 + (−K X P )2 − 2 t + t 2 
Hilb(X P , −K X P ) = + Q σ (t),
(1 − t)3 σ ∈B

 R−1
where Q R1 (1,a−1) (t) = 1
1−t R
(δai − δ0 )t i−1 is the Riemann–Roch contribution
i=1
 j
coming from the singularity R1 (1, a − 1) and δ j = R1 ∈μ R ,=1 (1−)(1− a−1 ) are the

Dedekind sums.

Example 20 Recall from Example 16 that P = conv (0, 1), (1, 0), (−5, −1) has
 1 
singularity content 2, { 5 (1, 1)} . The Hirzebruch–Jung continued fraction of the
cyclic quotient singularity 15 (1, 1) is simply [5], so d1 = − 35 and A 15 (1,1) = 15 .
2
+t 3
Also Q 15 (1,1) = t−2t
5(1−t 5 )
. Therefore the anticanonical degree and Hilbert series of
X = P(1, 1, 5) are given by
49
(−K X )2 = ,
5

  1 + 8t + 2t 3 − 2t 4 − 8t 6 − t 7
Hilb X, −K X = .
(1 − t 5 )(1 − t)3

More generally for a polygon P with n primitive T-singularities and basket of

singularities B = {m × 15 (1, 1)}:

1
(−K X P )2 = 12 − n − m,
5

Hilb(X P , −K X P ) =
−t 7 + (n − 10)t 6 + (m − 1)t 5 − 2mt 4 + 2mt 3 + (1 − m)t 2 + (10 − n)t + 1
.
(1 − t)3 (1 − t 5 )

Hirzebruch–Jung fractions can be further studied in [8, 14].

5.3 Minimal Fano Polygons

Mutation-equivalence classes raise the issue about our choice of representative of

a mutation-equivalence class of Fano polygons. This leads to the definition of a
124 D. Cavey and E. Kutas

minimal polygon from [11]. For a polygon P, the notation ∂ P denotes the boundary
of P.

Definition 21 Let P ⊂ NR be a Fano polygon. The polygon P is minimal if

|∂ P ∩ N | ≤ |∂ Q ∩ N |, for all Q = mut (n,F) (P).

For an edge E of P, let n E ∈ M be the primitive inner pointing normal of E.

E
Define h min = min{n E (v) : v ∈ P} and h max
E
= max{n E (v) : v ∈ P}. Then [11] tells
us P is minimal if and only if |E ∩ N | − 1 ≥ |h min E
| implies |h min
E
| ≤ h max
E
for all
edges E of P.
Given a Fano polygon P, find a minimal representative of the mutation-equivalence
class by calculating all possible mutations of P. If none of the Fano polygons obtained
via these mutations have fewer boundary points, then P is minimal. Otherwise choose
one of the mutations of P that have fewer boundary points as our new representa-
tive. Repeat this inductively, to obtain a minimal representative. The process must
terminate since the number of boundary points of P is finite and non-negative.
A minimal representative of a mutation-equivalence class is not necessarily
unique. We will always choose the representative of any equivalence class of Fano
polygons to be minimal.

Example 22 In Example 9, |∂ P ∩ N | = 3 and |∂ Q ∩ N | = 8. We know there is

only one other existing mutation of P. It is routine to check that this remaining
mutation does not have fewer boundary points than P. It follows that P is minimal.

5.4 Algorithm to Calculate Minimal Polygons with Given

Basket

5.4.1 Special Facets

We require the notion of a special facet introduced by Øbro [13].

Definition 23 Let P ⊂ NR be a Fano polygon. An edge E of P is a special facet if


v ∈ R≥0 E.
v∈vert(P)

Example 24 Considering P = conv{(0, 1), (1, 0), (−5, −1)}, calculate that:

v = (0, 1) + (1, 0) + (−5, −1) = (−4, 0).
v∈vert(P)

So P has a unique special facet given by F = conv{(0, 1), (−5, −1)}.

5 Classification of Minimal Polygons with Specified Singularity Content 125

By the definition of Fano polygon, 0 ∈ int(P). Therefore the union of all cones
obtained from a Fano polygon P is equal to NR , so P has at least one special facet.
We use a result from [10] which is derived from a proof in [7].

Lemma 25 Let P be a Fano polygon. Let F be a special facet of P of height h and

with inward pointing normal n F ∈ M. Then

vert(P) ⊂ (a, b) ∈ NR : −h(h + 1) ≤ n F , (a, b) ≤ h .

5.4.2 Description of Algorithm

Define the maximal local index of a Fano polygon P by

m P = max (height(E) : E ∈ F (P)) ,

where F (P) is the set of edges of P. Similarly define m B to be the maximum height
among the cones representing the R-singularities of P.
The classification of Fano polygons with a given basket of singularities B up to
mutation-equivalence is split into two cases:
1. m P = m B
2. m P > m B
The proof of [11, Theorem 6.3] efficiently tackles Case 2. Note the polygons
this proof outputs are not necessarily minimal. It remains to deal with Case 1. An
algorithm to compute this classification has been completed in [10]. However tack-
ling classifications beyond the case of polygons with only 13 (1, 1) R-singularities is
inefficient.
The main result of this paper is an efficient algorithm to tackle Case 1. The
algorithm is as follows: start with only a single vertex (a, l F ) = v ∈ F where F is
the special facet. We can assume that the other endpoint of F is (b, l F ), where b < a.
Since m B = m P we have a bound on the heights on all edges. Given a point v1 consider
the set of points

S = {v2 : v2 − v1 is primitive and height(E (v1 ,v2 ) ) = h}

where E (v1 ,v2 ) is the line segment from v1 to v2 . Note S is a subset of a line L h . Given
vertices v1 , v2 , . . . vk consider the lines L 1 , . . . L m B then pick all the valid points u
on these lines that give us an edge E (vk ,u) that respects convexity and defines either
a T-singularity or a residual singularity of B. Lemma 25 gives a bound on how low
these lines can go. This often suffices as bound, however it is possible that either
1. there exists h such that L h is horizontal, or
2. there exists u ∈ L h such that the E (vk ,u) is horizontal.
126 D. Cavey and E. Kutas

In Case 1 L h never touches the lower bound. In a similar fashion in Case 2 we

are uncertain how far we can extend the edge E (v,u) by adding T-cones resulting in
infinitely many points of the form vk+1 = u + n(u − vk ) where n is chosen suitably
from Z≥0 .
We start by showing how to deal with Case 2. Suppose
vk+1 = u + n(u − vk ) ∈ vert(P). Once the construction of P is complete then
there must be another vertex vk+2 ∈ vert(P), where subscripts are considered mod-
ulo # vert(P), defining an edge E (vk+1 ,vk+2 ) . Consider replacing n by n + h, which has
the same effect as a shear transformation parallel to the line L h . Having done this,
all new choices of vk+2 equal the image of the old vk+2 under the same shearing.
In particular as n → ∞, the lines L h tend towards lines going through the origin.
This means that at suitably large values of n, since l F > 0, the choices of vk+2 will
end up violating convexity. Hence there is only a finite choice along each L h . Note
that while this is quite computationally intensive, this happens at most once per a
polygon.
The method to deal with Case 1 is similar. Similarly to the previous case we
fix L h and consider points vk+1 on it. From the point vk+1 we construct the lines L h
we then pick the next vertex vk+2 . At this point we exclude the cases where vk+1
and vk+2 lie on a horizontal line as this puts us in Case 2. As before we see that that
as the x-coordinate of vk+1 tends towards −∞, then the x-coordinate of vk+2 tends
towards ∞. Hence once this cutoff occurs there are a finite set of possible points
for vk+1 .

Theorem 26 The algorithm gives a complete classification for Fano polygons with
a specific basket of singularities B.

Proof It suffices to show that at each stage there are only finitely many choices of
vertices, this is clear from the above description. There are also only finitely many
choices for inputs as up to a linear transformation 0 < a ≤ l f and l f ≤ m B , so there
are only finitely many choices of a. 

We check our output up to GL(N )-equivalence and mutation-equivalence. Two

polygons can be shown to be mutation-equivalent by explicitly calculating a sequence
of mutations between the two. Conversely a polygon P has corresponding to it
a maximally mutable Laurent polynomial f ∈ C[x1±1 , . . . , xn±1 ] as defined in [1].
The classical period of f given by
  
1 n 1 d x1 d xn
π f (t) = ...
2πi |x1 |=···=|xn |=1 1 − t f (x 1 , . . . , x n ) x 1 xn

= coeff1 ( f )t ,
k k

k≥0

is an invariant under mutation. Hence two polygons with different periods cannot
be mutation-equivalent. We have computer code in Sage [16] that efficiently imple-
ments the algorithm.
5 Classification of Minimal Polygons with Specified Singularity Content 127

It is important to compare the LDP-algorithm [10] used to calculate the 13 (1, 1)

classification of Fano polygons in Theorem 4 with this new  algorithm. The LDP-
algorithm takes as input a value for the index l P = lcm height(E) : E ∈ F (P) ,
and returns all Fano polygons P with this index. Hence when used to calculate the
case m P = m B in the 13 (1, 1) classification, edges of height 1, 2 and 3 are permitted,
and so the LDP-algorithm needs to be run for l P ∈ {3, 6}. In the 15 (1, 1) classification
of Theorem 6, the value of l F can be up to 60. All polygons P satisfying l P ≤ 16
have been classified using the LDP-algorithm but this took approximately three days
to obtain and the run time will increase at least  l P . A classification
 quadratically with
of Fano polygons with singularity content n, {m × r1 (1, 1)} , where r ≥ 7 would
be impossible to calculate using the LDP-algorithm (l P would be bounded above
by 420 if r = 7).
In comparison we had the following run times to calculate the classifications of
Theorems 5 and 6 using our algorithm:
Basket in classification Run time
(n, {m × 13 (1, 1)}) 40 s
(n, {m × 15 (1, 1)}) 8 min

Furthermore a classification of Fano polygons whose basket of singularities only

contains 17 (1, 1) R-singularities has been informally completed. The main reasons
for the difference in speed between the LDP-algorithm and our algorithm are as
follows.
1. We only look for minimal representatives for each mutation equivalence class.
This is not the case in the LDP-algorithm.
2. The LDP-algorithm is not designed to look for polygons based on the sin-
gularity content. For example in the 13 (1, 1) classification running the LDP-
algorithm for l P ∈ {3, 6} will output polygons that do not have singularity content
(n, {m × 13 (1, 1)}).

5.5 Minimal Fano Polygons

with B = {m1 × 13 (1, 1), m2 × 61 (1, 1)}

We apply our algorithm to classify all Fano polygons whose only R-singularities
are the cyclic quotient singularities 1
3
(1, 1) and 1
6
(1, 1). Set
B = {m 1 × 3 (1, 1), m 2 × 6 (1, 1)}, where m 1 ∈ Z≥0 and m 2 ∈ Z>0 . Here m 2 is
1 1

non-zero since a classification for Fano polygons with only 13 (1, 1) R-singularities
has been completed in [11].
In the 13 (1, 1) classification of [11], a bound on the number of R-singularities
is found by substituting the degree contribution A 13 (1,1) > 0 into an expression for
the anticanonical degree of the corresponding toric Fano variety from Proposition 19
since we know this value to be positive. However the degree contribution A 16 (1,1) is
128 D. Cavey and E. Kutas

negative and a similar method does not yield a bound. We show that this can be done
by a purely combinatorial argument instead.

Lemma 27 There exist no minimal Fano polygons P ⊂ NR , with m P = 3 and resid-

ual basket given by B = {m × 16 (1, 1)}, where m ≥ 3.

Proof The result for m > 3 follows from the base case m = 3.
Let P be a polygon with B = {3 × 16 (1, 1)}. By a GL(N )-translation, assume that
one of the R-singularities is given by E 1 = conv{(−1, 3), (1, 3)}. By mutating with
respect to any T-singularity lying between E 1 and a second R-singularity, assume
this second R-singulary is adjacent to E 1 , given by an edge E 2 with endpoints (1, 3)
and (a, b). The primitive inner pointing normal of E 2 is given by
b − 3 1 − a 
n E2 = , ∈M
g g

where g = gcd(b − 3, 1 − a). The height of E 2 is

3a − b
h = −n E2 · (1, 3) = .
g

Since E 2 represents a 16 (1, 1) singularity, set h = 3:

3a − b
= 3, b = 3a − 3 gcd(b − 3, 1 − a).
g

By convexity b < 3. The only remaining integer solution with a ≥ 0, is given

by (0, −3). However this point is not primitive so it can not be chosen as a ver-
tex of a Fano polygon. Hence a < 0.
Suppose the second edge from (−1, 3), denoted E 3 , is vertical. By convexity
a = −1 and (a, b) is a vertex of E 3 . But then E 3 is of height 1 so cannot represent
a 16 (1, 1) singularity and m < 3. Suppose E 3 is not vertical. Again convexity demands
that the second endpoint of E 3 has first coordinate less than −1. Then height(E 3 ) > 3
which contradicts m P = 3.
Therefore there can be no minimal Fano polygon with residual basket given by
B = {3 × 16 (1, 1)} with m P = 3. 

A similar argument shows that for a basket B = {m 1 × 13 (1, 1), m 2 × 16 (1, 1)} as
above, then m 1 + m 2 < 3.
Examples in this particular classification demonstrate a notion known as shattering
introduced by Wormleighton [17]. Let C1 = u, v, C2 = v, w be two cones in NR .
Suppose the vectors v − u, w − v are parallel. Then define the hyperplane sum of C1
and C2 to be given by C1 ∗ C2 = u, w.

Corollary 28 ([17, Corollary 2.2]) Let σ1 ∗ σ2 ∗ · · · ∗ σn = τ be a T-singularity.

Then the Riemann–Roch contributions Q σi and the degree contributions Aσi satisfy
5 Classification of Minimal Polygons with Specified Singularity Content 129

Table 5.1 The polygons with singularity content of the form (n, {m 1 × 13 (1, 1), m 2 × 16 (1, 1)})
with m 2 = 0
# vert(P) n m1 m2 (−K X )2
1.1 (−1, 3), (1, 3), (0, −1) 2 0 1 32/3
1.2 (−1, 3), (1, 3), (1, 2), (0, −1) 3 0 1 29/3
1.3 (−1, 3), (1, 3), (1, 1), (0, −1) 4 0 1 26/3
1.4 (−1, 3), (1, 3), (1, 0), (0, −1) 5 0 1 23/3
1.5 (−1, 3), (1, 3), (1, 2), (0, −1), (−1, 0) 6 0 1 20/3
1.6 (−1, 3), (1, 3), (1, 2), (0, −1), (−1, −1) 7 0 1 17/3
1.7 (−1, 3), (1, 3), (1, 0), (0, −1), (−1, 0) 8 0 1 14/3
1.8 (−1, 3), (1, 3), (1, 0), (−1, −1) 8 0 1 14/3
1.9 (−1, 3), (1, 3), (1, 0), (0, −1), (−1, −1) 9 0 1 11/3
1.10 (−1, 3), (1, 3), (1, 2), (−1, −4) 10 0 1 8/3
1.11 (−1, 3), (1, 3), (1, −1), (−1, −3) 11 0 1 5/3
1.12 (−1, 3), (1, 3), (5, −1), (−5, −1) 12 0 1 2/3
1.13 (−1, 1), (1, 1), (5, −1), (−5, −1) 12 0 2 4/3
1.14 (−1, 3), (1, 3), (1, −1), (−1, −2) 9 1 1 2

Q σ1 + · · · + Q σn = 0,

lattice length(τ )
A σ1 + · · · + A σn = A τ = .
lattice height(τ )

Consider a T-cone C = cone (−2, 3), (1, 3) of height 3. By adding an additional
ray given by primitive generating vector (−1, 3) decompose C into two sub-cones C1
and C2 representing a 13 (1, 1) and a 16 (1, 1) R-singularity respectively. By Corollary 28

Q 13 (1,1) + Q 16 (1,1) = 0, and A 13 (1,1) + A 16 (1,1) = 1.

Knowing A 13 (1,1) = 5
3
and Q 31 (1,1) = − 3(1−t
t
3 ) , derive:

2 t
A 16 (1,1) = − , and Q 16 (1,1) = .
3 3(1 − t 3 )

By Proposition 19 and Lemma 27, calculate (−K X P )2 = 12 − n − 53 m 1 + 23 m 2 .

Since we are interested in Fano polygons, (−K X P )2 > 0, so n ≤ 13. The algorithm
can be run a finite number of times to get the desired classification.
The results for the classification of polygons with singularity content of the
form (n, {m 1 × 13 (1, 1), m 2 × 16 (1, 1)}) with m 2 = 0, up to mutation-equivalence
is given in Table 5.1. All the polygons listed arose in Case 1 with the exception of
polygon 1.12. These polygons are illustrated in Fig. 5.3.
130 D. Cavey and E. Kutas

1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9

1.10 1.11 1.12 1.13 1.14

Fig. 5.3 Minimal representatives of mutation-equivalence classes of Fano polygons with singularity
content (n, {m 1 × 13 (1, 1), m 2 × 16 (1, 1)}), where m 1 ≥ 0, m 2 > 0

Recall that the maximally mutable Laurent polynomial of a polygon P is a poly-

nomial f such that Newt( f ) = P, and that the mutations of f remain Laurent poly-
nomials. The periods of maximally mutable Laurent polynomials are mutation invari-
ants by [2, Lemma 1]. The maximally mutable Laurent polynomials of 1.7 and 1.8
are given respectively by:

y3 y2 y 1 1
f = x y 3 + 3x y 2 + ay 3 + 3x y + by 2 + + x + cy + 3 + 3 + + ,
x x x y x
3 2
y y y 1 1
g = x y 3 + 3x y 2 + dy 3 + 3x y + ey 2 + +x + fy +4 +6 +4 + .
x x x x xy

Calculating the corresponding periods of f and g obtain:

π f = 1 + (2a + 2)x 2 + (3b + 36)x 3 + (6a 2 + 24a + 4c + 186)x 4 +

(20ab + 360a + 60b + 760)x 5 + · · · ,
πg = 1 + 14x 2 + 6ax 3 + 546x 4 + (420a + 30b)x 5 + · · · .

It is easy to see that these periods are not equal and hence the polygons cannot be
mutation-equivalent. All other Fano polygons in this classification have pairwise
distinct singularity content, hence are not mutation equivalent.
5 Classification of Minimal Polygons with Specified Singularity Content 131

5.6 Minimal Fano Polygons with B = {m × 15 (1, 1)}

We find all Fano polygons with singularity content of the form (n, {m × 15 (1, 1)})
with m > 0. Similarly to Sect. 5.5, bounds on n and m are required to ensure the
complete classification.

Lemma 29 There exist no minimal Fano polygons P ⊂ NR , with m P = 5 and resid-

ual basket given by B = {m × 15 (1, 1)}, where m ≥ 3.

Proof Similarly to the proof of Lemma 27, assume the existence of a Fano poly-
gon P with three 15 (1, 1) singularities, and perform a combination of GL(N )-
translations and mutations so that one of the R-singularities is represented by
the edge E 1 = conv{(−3, 5), (−2, 5)}, and another by E 2 = conv{(−2, 5), (a, b)},
where (a, b) = (−3, 5). We show you can always mutate P so that the third R-
singularity is represented by E 3 = conv{(−3, 5), (c, d)}, where (c, d) = (−3, 5),
without disrupting the original two 15 (1, 1) singularities sitting adjacently.
Study the possible T-cones that when mutated with respect to would separate the
adjacent R-singularities. Calculate the line of points from (−2, 5) that give an edge
at height 5, since (a, b) must lie on this line in order for E 2 to define a 15 (1, 1) R-
singularity. Unlike Lemma 27, since we are only interested in 15 (1, 1) singularities,
assume that the inner pointing normal n E2 = (b − 5, −2 − a) is primitive. This line
of points on which (a, b) lies, provides a bound on where (c, d) can lie by convexity.
Convexity also determines that d ≤ 5. Furthermore since P is Fano, the origin (0, 0)
must lie in the interior which further bounds the region (c, d) lies in. Finally since
we are only interested in the case where the prospective T-singularity would disrupt
the adjacent R-cones when mutated with respect to we obtain a final bound on the

Table 5.2 The polygons with singuarity content of the form (n, {m × 15 (1, 1)}) with m > 0
# vert(P) n m (−K X )2
2.1 (−3, 5), (−2, 5), (1, −2) 2 1 49/5
2.2 (−3, 5), (−2, 5), (−1, 3), (1, −2) 3 1 44/5
2.3 (−3, 5), (−2, 5), (−1, 3), (1, −2), (−2, 3) 4 1 39/5
2.4 (−3, 5), (−2, 5), (−1, 3), (1, −2), (−1, 1) 5 1 34/5
2.5 (−3, 5), (−2, 5), (0, 1), (1, −2), (−1, 1) 6 1 29/5
2.6 (−3, 5), (−2, 5), (0, 1), (1, −2), (0, −1) 7 1 24/5
2.7 (−3, 5), (−2, 5), (1, −1), (0, −1) 7 1 24/5
2.8 (−3, 5), (−2, 5), (1, −1), (1, −2), (0, −1) 8 1 19/5
2.9 (−3, 5), (−2, 5), (1, −1), (1, −3) 9 1 14/5
2.10 (−3, 5), (−2, 5), (2, −3), (2, −5) 10 1 9/5
2.11 (−3, 5), (−2, 5), (4, −1), (−3, −1) 11 1 4/5
2.12 (−3, 5), (−2, 5), (3, −5), (2, −5) 10 2 8/5
132 D. Cavey and E. Kutas

region in which (c, d) can lie. It is then possible to exhaustively check that none of
the primitive lattice points in this region give the second vertex of a T-cone.
Hence assume that the three R-cones lie adjacently. Calculating the points (c, d)
so that E 3 is height 5 and comparing with the possible choices of (a, b), note that
there are no choices of (a, b) and (c, d) that maintain convexity.
Therefore
 there can
be no minimal Fano polygon with residual basket given
by B = 3 × 15 (1, 1) with m P = 5. 
The above method extends very nicely to a combinatorial proof that
Lemma 30 There exist no Fano polygons P ⊂ NR with m P = p and residual basket
given by B = {m × 1p (1, 1)}, where m ≥ 3, p is odd and p ≥ 3.

2.1 2.2 2.3 2.4 2.5 2.6

2.7 2.8 2.9 2.10 2.11

2.12

Fig. 5.4 Minimal representatives of mutation-equivalence classes of Fano polygons with singularity
content (n, {m × 15 (1, 1)}) where m > 0
5 Classification of Minimal Polygons with Specified Singularity Content 133

By Example 20, the anticanonical degree of the toric variety corresponding

to a Fano polygon with only R-singularities of type 15 (1, 1) is given by
(−K X P )2 = 12 − n − 15 m > 0. Therefore n < 12. We apply the algorithm to com-
plete the classification.
The results for the classification of polygons with singuarity content of the
form (n, {m × 15 (1, 1)}) with m > 0 is given in Table 5.2. All the polygons were
found in Case 1. None arose in Case 2. These polygons are illustrated in Fig. 5.4.
Similarly to Sect. 5.5 note that polygons 2.6 and 2.7 are not mutation equivalent
by looking at the periods π f , πg of their respective maximally mutable Laurent
polynomials f and g:

π f = 1 + 12x 2 + 6ax 3 + 396x 4 + (360a + 30b)x 5 + · · · ,

πg = 1 + (2c + 12)x 2 + (6c + 3d + 90)x 3 + (6c2 + 24d + 144c + 636)x 4 +
(20cd + 60c2 + 390d + 1260c + 6900)x 5 + · · · .

All other Fano polygons in the classification have pairwise distinct singularity content
and therefore belong to different mutation equivalence classes.

Acknowledgements We thank Alexander Kasprzyk for his guidance and many useful conversa-
tions. Much of the paper was written in during a visit of EK to Nottingham supported by EPRSC
Fellowship EP/NO22513/1. EK would also like to thank his doctoral advisor, Miles Reid.

References

1. Akhtar, M., Coates, T., Corti, A., Heuberger, L., Kasprzyk, A.M., Oneto, A., Petracci, A.,
Prince, T., Tveiten, K.: Mirror symmetry and the classification of orbifold del Pezzo surfaces.
Proc. Amer. Math. Soc. 144(2), 513–527 (2016)
2. Akhtar, M., Coates, T., Galkin, S., Kasprzyk, A.M.: Minkowski polynomials and mutations.
SIGMA Symmetry Integr. Geom. Methods Appl. 8, Paper 094, 17 (2012)
3. Akhtar, M., Kasprzyk, A.M.: Singularity content (2014). arXiv:1401.5458 [math.AG]
4. Coates, T., Corti, A., Galkin, S., Golyshev, V., Kasprzyk, A.M.: Mirror symmetry and Fano
manifolds. In: European Congress of Mathematics, pp. 285–300. European Mathematical Soci-
ety, Zürich (2013)
5. Corti, A., Heuberger, L.: Del Pezzo surfaces with 13 (1, 1) points. Manuscripta Math. 153(1–2),
71–118 (2017)
6. Cox, D.A., Little, J.B., Schenck, H.K.: Toric Varieties. Graduate Studies in Mathematics, vol.
124. American Mathematical Society, Providence (2011)
7. Dais, D.I., Nill, B.: A boundedness result for toric log del Pezzo surfaces. Arch. Math. (Basel)
91(6), 526–535 (2008)
8. Fulton, W.: Introduction to toric varieties. Annals of Mathematics Studies, vol. 131. Princeton
University Press, Princeton (1993). The William H. Roever Lectures in Geometry
9. Kasprzyk, A.M.: Bounds on fake weighted projective space. Kodai Math. J. 32(2), 197–208
(2009)
10. Kasprzyk, A.M., Kreuzer, M., Nill, B.: On the combinatorial classification of toric log del
Pezzo surfaces. LMS J. Comput. Math. 13, 33–46 (2010)
11. Kasprzyk, A.M., Nill, B., Prince, T.: Minimality and mutation-equivalence of polygons. Forum
Math. Sigma 5, e18, 48 (2017)
134 D. Cavey and E. Kutas

12. Kollár, J., Shepherd-Barron, N.I.: Threefolds and deformations of surface singularities. Invent.
Math. 91(2), 299–338 (1988)
13. Øbro, M.: An algorithm for the classification of smooth Fano polytopes (2007).
arXiv:0704.0049 [math.CO]
14. Reid, M.: Surface cyclic quotient singularities and Hirzebruch–Jung resolutions. http://
homepages.warwick.ac.uk/~masda/surf/more/cyclic.pdf
15. Reid, M.: Young person’s guide to canonical singularities. In: Algebraic Geometry, Bowdoin,
1985 (Brunswick, Maine, 1985), Proceedings of Symposia in Pure Mathematics, vol. 46, pp.
345–414. American Mathematical Society, Providence, RI (1987)
16. The Sage Developers: SageMath, the Sage Mathematics Software System (Version 8.4.0)
(2018). http://www.sagemath.org
17. Wormleighton, B.: Reconstruction of singularities on orbifold del Pezzo surfaces from their
Hilbert series. Comm. Algebra 48(1), 119–140 (2020)
Chapter 6
On the Topology of Fano Smoothings

Tom Coates, Alessio Corti, and Genival da Silva Jr.

Abstract Suppose that X is a Fano manifold that corresponds under Mirror Sym-
metry to a Laurent polynomial f , and that P is the Newton polytope of f . In this
setting it is expected that there is a family of algebraic varieties over the unit disc
with general fiber X and special fiber the toric variety defined by the spanning fan
of P. Building on recent work and conjectures by Corti–Hacking–Petracci, who con-
struct such families of varieties, we determine the topology of the general fiber from
combinatorial data on P. This provides evidence for the Corti–Hacking–Petracci
conjectures, and verifies that their construction is compatible with expectations from
Mirror Symmetry.

Keywords Fano manifolds · Mirror symmetry · Betti numbers · Vanishing cycles

6.1 Introduction

There has been recent interest in the classification of Fano manifolds via Mirror
Symmetry [1, 7, 8]. For us, an n-dimensional Fano manifold X corresponds under
Mirror Symmetry to a Laurent polynomial f ∈ C[x1±1 , . . . , xn±1 ] if the regularized
quantum period of X , which is a generating function for certain genus-zero Gromov–
Witten invariants of X , coincides with classical period π f of f :

1 1 d x1 d xn
π f (t) = ··· .
(2π i)n S 1 ×···×S 1 1 − t f x1 xn

T. Coates (B) · A. Corti · G. da Silva Jr.

Department of Mathematics, Imperial College London, 180 Queen’s Gate, London
SW7 2AZ, UK
e-mail: [email protected]
A. Corti
e-mail: [email protected]
G. da Silva Jr.
e-mail: [email protected]
© Springer Nature Switzerland AG 2022 135
A. M. Kasprzyk and B. Nill (eds.), Interactions with Lattice Polytopes,
Springer Proceedings in Mathematics & Statistics 386,
https://doi.org/10.1007/978-3-030-98327-7_6
136 T. Coates et al.

If a Fano manifold X corresponds under Mirror Symmetry to a Laurent polynomial f

then it is expected that there is a degeneration X → , where  ⊂ C is a disc, such
that the general fiber is X and the special fiber is the toric variety X f defined by
the spanning fan of the Newton polytope of f . In general, X f will be singular. It is
natural to ask whether one can determine the Betti numbers of the Fano manifold X
from its mirror partner f . We will show that, in dimension 3, the answer to this
question is ‘yes’.
The key new ingredient presented here is work by Corti–Hacking–Petracci [10],
in preparation and still partly conjectural, which goes a long way towards establish-
ing the expected picture described above in dimension 3. Corti–Hacking–Petracci
construct, given a 3-dimensional reflexive polytope P decorated (as described below)
with ‘decomposition data’, a smoothing X →  of the toric variety X P defined by
the spanning fan of P. The general fiber of this smoothing is a three-dimensional
Fano manifold X . The decomposition data also determine a Laurent polynomial f
with Newton polytope P, so that X f = X P . In this paper we show that the Betti
numbers of X depend on the decomposition data only via f . That is, for each of
the (many) choices of decomposition data that give rise to the same Laurent polyno-
mial f , the Betti numbers of the corresponding smoothing are the same. Furthermore,
these coincide with the Betti numbers of the Fano manifold that corresponds to f
under Mirror Symmetry.
We proceed as follows. Corti–Hacking–Petracci construct, given a choice of
decomposition data for a reflexive polytope P, a toric partial resolution π : Y → X P
and a family Y →  with special fiber Y and fiber over a general point t ∈  a
weak Fano manifold Yt . Contracting finitely many (−1, −1) curves in Yt gives a
resolution πt : Yt → X t , where X t is a Fano variety with ordinary double points, and
this fits into a diagram

Yt Y
πt π (6.1)
Xη Xt XP

where the arrow A B means that A is the general fiber in a family over 
with special fiber B. Here X η X t is Namikawa’s smoothing of Fano vari-
eties with ordinary double points. The Fano variety X η is our desired smoothing
of X P , and the diagram above allows us to compute its Betti numbers. The central
fiber Y is a toric variety, so we know its cohomology groups explicitly; we can com-
pute the Betti numbers of Yt by analysing the vanishing cycles of the degeneration
Yt Y and, since the left-hand part of the diagram is a conifold transition
from Yt to X η , this determines the Betti numbers of X η .
We begin by reviewing the cohomology of toric varieties and the vanishing cycle
exact sequence. We then explain in Sect. 6.3 how to compute the Betti numbers of the
smoothing X from the decomposition data, and give examples in Sect. 6.4. In Sect. 6.5
we prove that the Betti numbers of X depend on the decomposition data only via
6 On the Topology of Fano Smoothings 137

the Laurent polynomial f determined by those data, and in Sect. 6.6 we verify that
these Betti numbers of X coincide with the Betti numbers of the Fano manifold that
corresponds to f under Mirror Symmetry.

6.2 Cohomology and Vanishing Cycles

6.2.1 The Cohomology of Toric Varieties

We will compute the Betti numbers of the fiber Y in diagram (6.1) using the fact that
it is a toric variety.1
Theorem 1 (cf. [16, Proposition 3.5.3]) Let  be a complete fan in a three-
dimensional lattice and let X  be the toric threefold defined by . Let di denote
the number of i-dimensional cones in , and let bi denote the ith Betti number
of X  . Then:

d1 − d2 + d3 = 2 b2 = rk Pic(X  )
b0 = b6 = 1 b3 = rk Pic(X  ) − d2 + 2d1 − 3
b1 = b5 = 0 b4 = d1 − 3

and the Euler characteristic of X is d3 .

6.2.2 The Vanishing Cycle Exact Sequence

We will compute the Betti numbers of the fiber Yt in diagram (6.1) by analysing the
vanishing cycles for the degeneration Yt Y . Consider a complex variety Y,
a disc  ⊂ C, and a projective morphism f : Y → . Let ∗ =  \ {0} be the
punctured disc, and
i0 j0
{0} − − ∗
→←

be the natural inclusions. Denote the fiber over t ∈ ∗ by Yt , and the fiber over 0 ∈ 
by Y . Choose a universal covering map p0 :  ∗ → ∗ , and consider the diagram

Y
i
Y
j
Y\Y
p

Y \Y

j0 p0
{0}
i0
 ∗ ∗ .


1 We learned the statement of Theorem 1 from Andrea Petracci.

138 T. Coates et al.

Let S be a stratification for Y and suppose that F• ∈ DbS (Y). The nearby sheaf is
defined [14, Exposé I] to be the complex

ψ f F• = i ∗ R( j ◦ p)∗ ( j ◦ p)∗ F• .

By adjunction, there is a natural map

i ∗ F• → ψ f F• .

The sheaf of vanishing cycles φ f F• is the cone on this map (ibid.), and there is a
distinguished triangle:
+1
i ∗ F• → ψ f F• → φ f F• −→ (6.2)

Consider now the cohomology sheaves Hi (ψ f F• ) and Hi (φ f F• )—these are com-

plexes of sheaves on Y0 —and their stalks at y ∈ Y0 . Embed Y into an affine space and
let B(y, ) be the open ball of radius  around y. Then for sufficiently small  > 0
and for all t ∈ ∗ such that |t| < , we have

Hi (ψ f F• ) y ∼
= Hi (F f,y , F• )

where F f,y is the Milnor fiber of f at y, F f,y := B(y, ) ∩ Y ∩ f −1 (t). Similarly,

Hi (φ f F• ) y ∼
= Hi+1 (B(y, ) ∩ Y, F f,y ; F• ).

Now consider the constant sheaf QY as a complex concentrated in degree 0. Taking

stalks of the hypercohomology of the distinguished triangle in (6.2), we get

· · · → H i (Y, Q) → H i (Yt , Q) → Hvi (Yt , Q) → H i+2 (Y, Q) → · · · (6.3)

where Hvi (Yt , Q) is the subspace in H i (Yt , Q) generated by vanishing cycles, that
is, cycles in the kernel of the natural map H i (Yt , Q) → H i (Y0 , Q).

6.3 Smoothing Toric Fano Threefolds

Every three-dimensional Gorenstein toric Fano variety is the toric variety X P defined
by the spanning fan of a three-dimensional reflexive polytope P. This gives a one-to-
one correspondence between three-dimensional Gorenstein toric Fano varieties up to
isomorphism and three-dimensional reflexive polytopes up to GL(3, Z)-equivalence.
In general such a toric variety X P is singular. As mentioned in the Introduction, Corti–
Hacking–Petracci construct, starting from a three-dimensional reflexive polytope P
decorated with some additional data, a smoothing X →  of X P . In this section we
describe their construction and the additional data required.
6 On the Topology of Fano Smoothings 139

Definition 2 Let n be an integer such that n ≥ −1. An An triangle is a lattice polygon

that is Z2  GL(Z2 )-equivalent to the polygon in Q2 with vertices (0, 0), (0, 1),
and (n + 1, 0)

Example 3 The standard two-dimensional simplex is an A0 -triangle.

Example 4 An A−1 -triangle is a line segment of unit length.

Definition 5 Let F be a lattice polygon. An admissible Minkowski decomposition

of F is a Minkowski decomposition F = F1 + · · · + Fk of F as a sum of lattice
polygons F j such that each F j is an An -triangle for some n ≥ −1. (Here n can
depend on j.) We consider admissible Minkowski decompositions of F that differ
by reordering and translation of the summands to be equivalent.

We now introduce certain polyhedral decompositions of lattice polygons. Suppose

that F = F1 + · · · + Fk is an admissible Minkowski decomposition of the lattice
polygon F. Recall that the Cayley polytope C F1 ,...,Fk is the convex hull of

F1 + e1 , F2 + e2 , . . . , Fk + ek

in L ⊕ Ze1 ⊕ · · · ⊕ Zek , where L is the lattice containing F. The map

  
F −→ C F1 ,...,Fk ∩ L R × k1 , · · · , k1
(6.4)
v −→ k1 (v + e1 + · · · + ek )

is bijective, and this allows us to obtain polyhedral decompositions of F from poly-

hedral decompositions of the Cayley polytope C F1 ,...,Fk .

Definition 6 Let F = F1 + · · · + Fk be an admissible lattice Minkowski decom-

position as above. A regular fine mixed subdivision of F is a subdivision of F
induced, via (6.4), by a regular unimodular triangulation of the Cayley poly-
tope C F1 ,...,Fk . We say that such a subdivision is subordinate to the Minkowski decom-
position F = F1 + · · · + Fk .

There is a whole body of theory here, which we will not discuss, concerning
polyhedral subdivisions which may be neither fine nor mixed: an introduction to
these topics can be found in the extremely beautiful book by De Loera et al. [11].
We will only consider regular fine mixed subdivisions.
For the rest of this section, fix a three-dimensional reflexive polytope P. Corti–
Hacking–Petracci consider the polytope P together with decomposition data. This
is, for each facet F of P:
A. a choice of admissible Minkowski decomposition F = F1 + · · · + Fk ;
B. a choice of regular fine mixed subdivision of F subordinate to (A);
satisfying a condition that we now describe. Note that by taking cones over the regular
fine mixed subdivisions of each facet, we obtain a complete fan  that refines the
140 T. Coates et al.

Fig. 6.1 The normalisation

of a 4-valent vertex in the
toric 1-skeleton

spanning fan of P, and thus a toric crepant partial resolution Y → X P . Recall first
that irreducible toric curves in X P (respectively in Y ) correspond to two-dimensional
cones in the spanning fan of P (respectively in the fan ). Thus irreducible toric
curves in X P correspond to edges of P, and irreducible toric curves in Y correspond
to edges in the polyhedral subdivision of the boundary of P determined by the
decomposition data.

Theorem 7 (Corti–Hacking–Petracci) The singularities of the toric partial resolu-

tion Y of X P are at worst quasi-ordinary double points (qODPs).

This amounts to the statement that each polygon in the regular fine mixed subdi-
vision (B) above is either a standard 2-simplex or a quadrilateral with sides of unit
length. The curves in the toric 1-skeleton of Y , therefore, either meet at 3-valent
vertices—which are the torus-fixed points on Y corresponding to the cones over
the 2-simplices—or at 4-valent vertices, which are the torus-fixed points on Y cor-
responding to the cones over the quadrilaterals. The 3-valent vertices are smooth
points on Y , and the 4-valent vertices are the qODPs.
Let e denote the irreducible toric curve in X P determined by the edge e of P,
and let  e denote the set of irreducible toric curves in Y that map dominantly to e
under the partial resolution Y → X P . The fact that each polygon in the subdivision
(B) above is either a standard 2-simplex or a quadrilateral with sides of unit length
implies that | e | =
(e), the lattice length of the edge e.
Let  denote the toric 1-skeleton of Y . Consider the partial normalisation  → 
constructed by normalising each 4-valent vertex as shown in Fig. 6.1.
The partial normalisation  consists of rational curves that meet at either bivalent
or trivalent vertices. Let e denote an edge of P, and e denote the corresponding toric
curve in X P . We define two partitions of the set  e , as follows. Let p ∈ X P be one of
the endpoints of e , and F ⊂ P the corresponding face of P. Consider the part  p
of the partially-normalised toric 1-skeleton  that lies over p; this consists of the
dual graph to the polyhedral subdivision (B) of F, partially normalised as described
in Fig. 6.1. The components of the partially-normalised dual graph define a partition
of 
e . There is one such partition for each of the two endpoints p of e : let us denote
them by e and  e (the order will not matter). The condition that Corti–Hacking–
Petracci impose on their decomposition data is: for each edge e of P such that the
dual edge e has lattice length
(e ) equal to 1 and for each pair T ∈ , T ∈  , we
have |T ∩ T | ≤ 1.
 6 On the Topology of Fano Smoothings 141

6.3.1 Computing the Betti Numbers of the Smoothing

Corti–Hacking–Petracci prove:
Theorem 8 Let P be a 3-dimensional reflexive polytope and let X P be the toric
Fano threefold associated to the spanning fan of P. Fix decomposition data for P as
above, and let π : Y → X P be the associated crepant toric partial resolution. Then:
1. Y is unobstructed and smoothable;
2. if Yt is a general smoothing of Y , then Yt is a weak Fano threefold and the
anticanonical morphism πt : Yt → X t , where X t = Proj R(Yt , −K Yt ), contracts
a finite number of disjoint nonsingular rational curves, each with normal bun-
dle O(−1) ⊕ O(−1);
3. X t is a Fano threefold with ordinary nodes as singularities and it is a deformation
of X P .
A theorem of Namikawa [18, Theorem 11] now implies that X t is smoothable. It
follows that X P is smoothable. Let X η denote a generic smoothing of X t . Our goal
is to compute the Betti numbers of X η .
We begin by analysing the topology of Yt . The vanishing cycles for the degen-
eration Yt  Y are three-dimensional spheres. The sheaf of vanishing cycles φ f
from (6.2), which is concentrated at the nodes of Y , therefore has stalk at each node
equal to Q concentrated in degree three, and the vanishing cycle exact sequence (6.3)
gives

b0 (Yt ) = b0 (Y ) b4 (Yt ) = b2 (Y )
b1 (Yt ) = b1 (Y ) b5 (Yt ) = b1 (Y )
(6.5)
b2 (Yt ) = b2 (Y ) b6 (Yt ) = b0 (Y )
b3 (Yt ) = b3 (Y ) − b4 (Y ) + b2 (Y ) + k

where k is the number of nodes on Y . Note that H2 (Yt ) is canonically identified

with H2 (Y ); also b5 (Yt ) = b1 (Yt ) = 0 by Poincaré duality and Kodaira vanishing.
Passing from Yt to X η is an example of a conifold transition [5], and therefore

b0 (X η ) = b0 (Yt ) = 1 b4 (X η ) = b2 (Yt ) − l
b1 (X η ) = b1 (Yt ) = 0 b5 (X η ) = b1 (Yt ) = 0
(6.6)
b2 (X η ) = b2 (Yt ) − l b6 (X η ) = b0 (Yt ) = 1
b3 (X η ) = b3 (Yt ) + 2m − 2l

where l is the dimension of the subspace L of H2 (Yt ) spanned by the classes of

curves that are contracted by πt : Yt → X t , and m is the number of nodes on X t .
In view of Theorem 1, computing the Betti numbers of the smoothing X η comes
down to computing the integers k, l, and m. We have seen that k is the number of
quadrilaterals in the polyhedral decomposition of the boundary of P determined by
142 T. Coates et al.

the decomposition data. Corti–Hacking–Petracci conjecture the values of l and m,

as follows. Let us identify L ⊂ H2 (Yt ) as a subspace of H2 (Y ) via the canonical
isomorphism H2 (Yt ) ∼ = H2 (Y ) just discussed. Recall that an edge e of P determines
a toric rational curve e in X P : this is the toric subvariety of X P defined by the cone
over e. The set  e of toric curves that map dominantly to e under the map Y → X P is
indexed by the
(e) line segments that subdivide e. If
(e) ≥ 2 then e contains interior
lattice points: such a lattice point v then corresponds to a toric surface Sv ⊂ Y that
projects to e . Let us denote the homology class2 of a fiber of Sv → e by Fv . For
a pair of toric curves C, C ∈  e , let c, c denote the corresponding line segments
and (c, c ) ⊂ P denote the relative interior of the convex hull of c and c . Recall

the two partitions of  e defined in the discussion around Fig. 6.1, and define n c,c
by starting with
(e) and subtracting one for each partition that has c and c in the
same part. The set of exceptional curves for πt : Yt → X t is conjecturally indexed by
edges e of P such that
(e) ≥ 2 and pairs of distinct elements c, c ∈  e : it contains
precisely n c,c curves in the homology class

Fv
v∈(c,c )

and no others. Note that n c,c here is non-negative; it can be zero. This conjecture
determines the subspace L ⊂ H2 (Yt ) spanned by exceptional curves, and thus deter-
mines l = dim L. For m, suppose that the elements of the two partitions of  e have
sizes a1 , a2 , . . . and b1 , b2 , . . . respectively. Set





(e) ai bi
Ne =
(e ) − − . (6.7)
2 i
2 i
2

Then
m= Ne . (6.8)
e :
(e)≥2

6.3.2 Minkowski Polynomials and Smoothings

An admissible Minkowski decomposition of a three-dimensional reflexive poly-

tope P determines a Minkowski polynomial [2]. This is a Laurent polynomial with
Newton polytope P. In the notation of the discussion in Sect. 6.3, it depends on the
choices (A) of admissible Minkowski decomposition of each facet of P, but not on
the choices (B) of regular fine mixed subdivision.

2 The fiber Fv is homologous in Y to the sum of toric curves corresponding to 2-dimensional cones
in  that contain the ray spanned by v and that lie entirely on one side of the hyperplane defined
by the edge e. The choice of side does not matter here, as the resulting sums are homologous.
6 On the Topology of Fano Smoothings 143

It is known that Minkowski polynomials provide mirrors for three-dimensional

Fano manifolds [8]. As mentioned in the Introduction, if a Fano manifold X is mirror
to a Laurent polynomial f with Newton polytope P, it is expected that there is a
degeneration X →  with general fiber X and special fiber X P . For this expectation
to be compatible with the results and conjectures of Corti–Hacking–Petracci, there-
fore, the Betti numbers of the Corti–Hacking–Petracci smoothings X η must depend
only on the Minkowski polynomial. That is, the Betti numbers must depend only on
decomposition data only through the choices (A) of admissible Minkowski decom-
positions of facets: they must be independent of the choices (B) of regular fine mixed
subdivision. This is not obvious from the construction; we prove it in Sect. 6.5.

6.4 Examples

6.4.1 Cube

Consider the cube P centered at the origin with vertices (±1, ±1, ±1). This has six
non-simplicial facets and twelve edges of length two; thus the toric variety X P defined
by the spanning fan of P has six singular points and twelve curves of transverse A1
singularities. These are arranged as on the right-hand side of Fig. 6.2, with the singular
points at the vertices of the octahedron and the singular curves as the edges.
Each facet F of P is a square with side-length two; this has a unique admissible
Minkowski decomposition, as a Minkowski sum of four line segments, which in turn
leads to the unique fine mixed subdivision of F shown in Fig. 6.3.
The fan  that defines the partial resolution Y of X P is therefore obtained by taking
cones over the polyhedral decomposition of the boundary of P shown in Fig. 6.4.
The variety Y contains 24 ordinary double points and 48 toric curves, arranged as on
the right-hand side of Fig. 6.4, with the singular points at the vertices and the toric

Fig. 6.2 The cube P and a

schematic picture of the toric
variety X P

Fig. 6.3 The unique fine

mixed subdivision of the
facet F
144 T. Coates et al.

Fig. 6.4 The polyhedral

subdivision and a schematic
picture of Y

curves (along which Y is non-singular) as the edges. Furthermore each edge in the
dual polygon P  has length 1, and each partition e of  e is into singleton sets, so
the polyhedral decomposition satisfies the condition to be decomposition data.
Applying Theorem 1 gives

b0 (Y ) = 1 b4 (Y ) = 23
b1 (Y ) = 0 b5 (Y ) = 0
b2 (Y ) = 4 b6 (Y ) = 1
b3 (Y ) = 5

There are k = 24 quadrilaterals in the polyhedral subdivision of the boundary of P,

and the discussion in Sect. 6.3.2 yields

b0 (Yt ) = 1 b4 (Yt ) = 4
b1 (Yt ) = 0 b5 (Yt ) = 0
b2 (Yt ) = 4 b6 (Yt ) = 1.
b3 (Yt ) = 10

From the description of the homology classes Fv in footnote 2 and

[12, Proposition 2.1], we see that l = 3, with generators for the subspace L ⊂
H2 (Yt ) ∼
= H2 (Y ) of classes of exceptional curves as shown in Fig. 6.5.
Equation (6.8) gives m = 12, and we find:

Fig. 6.5 Toric curves that

generate the subspace of
exceptional curves
6 On the Topology of Fano Smoothings 145

b0 (X η ) = 1 b4 (X η ) = 1
b1 (X η ) = 0 b5 (X η ) = 0
b2 (X η ) = 1 b6 (X η ) = 1
b3 (X η ) = 28

The only smooth Fano 3-fold with these Betti numbers is V8 . Thus X η is isomorphic
to V8 , which is consistent with the fact that the Minkowski polynomial

(1 + x)2 (1 + y)2 (1 + z)2

f = −8
x yz

defined by our decomposition data for P is a mirror to V8 .

6.4.2 A Singular Toric Variety with Two Different Smoothings

Consider the three-dimensional polytope, pictured in Fig. 6.6, with vertices

(0, 0, 1), (0, 1, −1), (1, 1, −1), (1, 0, −1), (0, −1, −1), (−1, −1, −1), (−1, 0, −1).

This polytope P is reflexive. It has six facets that are standard simplices, one non-
simplicial facet (a hexagon), and 12 edges of length 1. Thus the toric variety X P
defined by the spanning fan of P contains six non-singular toric points, a unique
singular point (at which the singularity is a cone over the del Pezzo surface of
degree 6), and 12 non-singular toric curves. These are arranged as on the right-hand
side of Fig. 6.6, with the non-singular toric points at the 3-valent vertices, the singular
point at the 6-valent vertex, and the toric curves as the edges.
The hexagonal facet F of P admits two Minkowski decompositions: as the sum of
three line segments, and as the sum of two triangles. Up to automorphism, these each
give rise to a unique fine mixed subdivision of F, as shown in Fig. 6.7. Consider first
the left-most fine mixed subdivision in Fig. 6.7. This leads to the polyhedral subdi-
vision of the boundary of P shown in Fig. 6.8. The fan  given by taking cones over

Fig. 6.6 The polytope P and a schematic picture of the toric variety X P
146 T. Coates et al.

Fig. 6.7 Two fine mixed subdivisions of the facet F

Fig. 6.8 The first polyhedral subdivision and a schematic picture of Y

this polyhedral subdivision defines a toric partial resolution Y of X P . This variety Y

has three ordinary double points, six non-singular toric points, and 15 non-singular
toric curves, arranged as on the right-hand side of Fig. 6.8: the ordinary double points
are the 4-valent vertices, the non-singular points are the 3-valent vertices, and the
toric curves are the edges. The polyhedral decomposition satisfies the condition to
be decomposition data.
Applying Theorem 1 gives

b0 (Y ) = 1 b4 (Y ) = 5
b1 (Y ) = 0 b5 (Y ) = 0
b2 (Y ) = 2 b6 (Y ) = 1
b3 (Y ) = 0

and since there are k = 3 quadrilaterals in the polyhedral subdivision of the boundary
of P, we have

b0 (Yt ) = 1 b4 (Yt ) = 2
b1 (Yt ) = 0 b5 (Yt ) = 0
b2 (Yt ) = 2 b6 (Yt ) = 1.
b3 (Yt ) = 0

The discussion in Sect. 6.3.1 implies that, assuming the conjectures of Corti–
Hacking–Petracci, there are no exceptional curves in Yt and so the smoothings X η
of X P and Yt of Y are isomorphic. Thus
6 On the Topology of Fano Smoothings 147

Fig. 6.9 The second polyhedral subdivision and a schematic picture of Y

b0 (X η ) = 1 b4 (X η ) = 2
b1 (X η ) = 0 b5 (X η ) = 0
b2 (X η ) = 2 b6 (X η ) = 1.
b3 (X η ) = 0

These are the Betti numbers of the hypersurface W1,1 of bidegree (1, 1) in P2 × P2 ,
which is consistent with the fact that the Minkowski polynomial

(1 + x)(1 + y)(1 + x y)
f = +z
x yz

defined by our decomposition data for P is a mirror to W1,1 .

One could instead consider the right-most fine mixed subdivision in Fig. 6.7.
This leads to the polyhedral subdivision of the boundary of P shown in Fig. 6.9;
again this satisfies the condition to be decomposition data. This time the toric partial
resolution Y has two ordinary double points, eight non-singular toric points, and 16
non-singular toric curves. These are arranged as on the right-hand side of Fig. 6.8,
with the ordinary double points as the 4-valent vertices, the non-singular points as
the 3-valent vertices, and the toric curves as the edges.
Theorem 1 yields

b0 (Y ) = 1 b4 (Y ) = 5
b1 (Y ) = 0 b5 (Y ) = 0
b2 (Y ) = 3 b6 (Y ) = 1
b3 (Y ) = 0

and since there are k = 2 quadrilaterals in the polyhedral subdivision of the boundary
of P, we have
148 T. Coates et al.

b0 (Yt ) = 1 b4 (Yt ) = 3
b1 (Yt ) = 0 b5 (Yt ) = 0
b2 (Yt ) = 3 b6 (Yt ) = 1.
b3 (Yt ) = 0

Once again, there are (conjecturally) no exceptional curves in Yt and so the smooth-
ings X η of X P and Yt of Y are isomorphic. Thus

b0 (X η ) = 1 b4 (X η ) = 3
b1 (X η ) = 0 b5 (X η ) = 0
b2 (X η ) = 3 b6 (X η ) = 1
b3 (X η ) = 0

These are the Betti numbers of P1 × P1 × P1 , which is consistent with the fact that
the Minkowski polynomial

(1 + x + x y)(1 + y + x y)
f = +z
x yz

defined by our decomposition data for P is a mirror to P1 × P1 × P1 .

6.4.3 An Example with Transverse A2 Singularities

Consider the three-dimensional reflexive polytope P, pictured in Fig. 6.10, with ver-
tices
(1, 0, 0), (0, 1, 0), (0, 0, 1), (−1, −1, −1), (−1, −1, 2).

This polytope has four facets that are standard simplices, two non-standard simplicial
facets, eight edges of length 1, and one edge of length 3. Thus the toric variety X P

Fig. 6.10 The polytope P and a schematic picture of the toric variety X P
6 On the Topology of Fano Smoothings 149

Fig. 6.11 The polyhedral subdivision and a schematic picture of Y

defined by the spanning fan of P contains four non-singular toric points, two orb-
ifold points, eight non-singular toric curves, and one toric curve with transverse A2
singularities. These are arranged as on the right-hand side of Fig. 6.10, with the toric
points at the vertices, the orbifold points indicated in red, and the toric curves as the
edges. The curve of A2 singularities is the edge between the two orbifold points.
Figure 6.11 shows the unique fine mixed subdivision of the boundary of P. This
defines decomposition data for P. Taking cones over the polyhedra in this decom-
position gives a fan  that defines a toric resolution Y of X P . The variety Y is
smooth, with ten toric points and fifteen toric curves arranged as on the right-hand
side of Fig. 6.11: the toric points are the vertices and the toric curves are the edges.
Note the two toric surfaces in Y that map to the curve of singularities under the
resolution Y → X P .
Applying Theorem 1 gives

b0 (Y ) = 1 b4 (Y ) = 4
b1 (Y ) = 0 b5 (Y ) = 0
b2 (Y ) = 4 b6 (Y ) = 1
b3 (Y ) = 0

and since there are no quadrilaterals in the polyhedral subdivision, we find that the
Betti numbers of Yt coincide with those of Y . Computing the subspace L ⊂ H2 (Yt ) ∼
=
H2 (Y ) of classes of exceptional curves, as in the cube example, we find that l = 2
and that generators for L are as shown in Fig. 6.12. These generators are the fibers
of the toric surfaces in Y that resolve the curve of transverse A2 singularities.
From (6.7) and (6.8) we find that (conjecturally) there are m = 3 exceptional
curves in total, and therefore:

b0 (X η ) = 1 b4 (X η ) = 2
b1 (X η ) = 0 b5 (X η ) = 0
b2 (X η ) = 2 b6 (X η ) = 1.
b3 (X η ) = 2
150 T. Coates et al.

Fig. 6.12 Toric curves that generate the subspace of exceptional curves

These are the Betti numbers of the blow-up X of P3 in a plane cubic, which is
consistent with the fact that the Minkowski polynomial

(1 + z)3
f = +x+y+z
x yz

defined by our decomposition data for P is a mirror to X .

6.5 Betti Numbers Depend Only on the Mirror Laurent

Polynomial

We have seen that decomposition data for a three-dimensional reflexive polytope P

are, for each facet F of P:
A. a choice of admissible Minkowski decomposition F = F1 + · · · + Fk ;
B. a choice of regular fine mixed subdivision of F subordinate to (A);
and that decomposition data determine a Minkowski polynomial f . In this section
we will show that the Betti numbers of the smoothing X of X P determined by the
decomposition data are independent of the choice (B). This implies, in view of the
discussion in Sect. 6.3.2, that the Betti numbers of X depend on the decomposition
data only via f . We will prove:

Theorem 9 Let P be a three-dimensional reflexive polytope. Consider two sets of

decomposition data for P, where the choices (A) are the same in each case but the
choices (B) are different. Let X denote the Corti–Hacking–Petracci smoothing of X P
determined by the first set of decomposition data, with Y → X P the toric partial res-
olution and Yt the smoothing of Y ; let X denote the smoothing of X P determined
by the second set of decomposition data, with Y → X P the toric partial resolu-
tion and Yt the smoothing of Y . Assume the Corti–Hacking–Petracci conjectures
described on Sect. 6.3.2 hold. Then:
6 On the Topology of Fano Smoothings 151

Fig. 6.13 A move of Type I

Fig. 6.14 A move of Type II

a. b2 (Y ) = b2 (Y );
b. the Betti numbers of Yt and Yt coincide;
c. the Betti numbers of X and X coincide.
The key point is that the polyhedral decompositions of the boundary of P
that define Y and Y differ by a sequence of the two types of moves depicted in
Figs. 6.13 and 6.14, or their inverses. The vertices of the outer pentagon in Fig. 6.13
are (0, 0), (1, 0), (α + 1, β), (β, α + 1), (0, 1), where α and β are positive coprime
integers, and the interior lattice point pictured is at (α, β); unless α = β = 1 then
there are other interior lattice points which are not pictured. The precise values of α
and β will not affect the analysis. The vertices of the outer quadrilateral in Fig. 6.14
are at (0, 0), (2, 0), (1, 1), and (0, 1).

6.5.1 Type I Moves

Let us analyse how b2 changes under a Type I move. Suppose first that Y1 , Z , and Y2
are three-dimensional toric varieties defined by polyhedral decompositions of the
boundary of P that differ only as shown in Fig. 6.15.3 Then Y1 and Y2 are toric
partial resolutions of X P which differ by a Type I move, and there is a diagram

Y1 Y2

f1 f2
Z

Let z ∈ Z be the torus-fixed point corresponding to the pentagon pictured, and

write Z 0 = Z \ {z}. The maps f 1 and f 2 induce isomorphisms f 1−1 (Z 0 ) → Z 0

3 The co-ordinates of the vertices and interior lattice point pictured in Fig. 6.15 are as in Fig. 6.13.
152 T. Coates et al.

1 2

Fig. 6.15 The fans for Y1 , Z , and Y2 differ only at the cones over these polygons

Fig. 6.16 The values of a 2

piecewise-linear function on
the fan for Z 0
3 1

4 5

and f 2−1 (Z 0 ) → Z 0 , and the resulting inclusions j1 : Z 0 → Y1 and j2 : Z 0 → Y2

define a diagram
Pic(Y1 ) Pic(Y2 )

j1∗ j2∗

Pic(Z 0 )

We will identify Pic(Y1 ) and Pic(Y2 ) as subspaces of Pic(Z 0 ). To give a line bundle
on Z 0 is to give a piecewise-linear function on each maximal cone in the fan for Z 0 ,
subject to the constraint that these piecewise-linear functions agree along faces. Let
us write the values of such a piecewise linear function at the vertices of the polyhedral
decomposition that we are considering as in Fig. 6.16.
Since the fan for Z 0 does not include the cone over the pentagon pictured in
Fig. 6.16, that cone does not impose any relation between the values a1 , . . . , a5 .
(There may be other relations from the part of the fan not pictured, but these will
be the same for Z , Y1 , and Y2 .) This piecewise-linear function defines a line bundle
on Y1 if and only if it is piecewise-linear on the two cones pictured on the left-hand
side of Fig. 6.15, that is, if and only if a1 + a3 = a2 + a5 . So

Pic(Y1 ) = a1 + a3 = a2 + a5 } ⊂ Pic(Z 0 ).

Similarly, the piecewise-linear function on Z 0 defines a line bundle on Y2 if and only

if it is piecewise-linear on the two cones pictured on the right-hand side of Fig. 6.15,
that is, if and only if a1 + a4 − a5 = a2 + a4 − a3 . So

Pic(Y2 ) = a1 + a4 − a5 = a2 + a4 − a3 } ⊂ Pic(Z 0 ).
6 On the Topology of Fano Smoothings 153

These are the same subspace of Pic(Z 0 ); therefore Pic(Y1 ) and Pic(Y2 ) are canonically
isomorphic. Since all maximal cones in the fan for Y1 are full-dimensional, the Picard
group of Y1 is isomorphic to H 2 (Y1 ); the same statement is true for Z and for Y2 .
Thus b2 is invariant under Type I moves.
A result of Fulton and Sturmfels [12, Proposition 1.1] implies that b4 (Yi ) is equal to
three less than the number of vertices in the polyhedral decomposition that defines Yi ,
and so b4 (Y2 ) = b4 (Y1 ) + 1. Furthermore χ (Y2 ) = χ (Y1 ) + 1—here we used The-
orem 1—and so b3 (Y1 ) = b3 (Y2 ). In summary:

b0 (Y2 ) = b0 (Y1 ) = 1 b4 (Y2 ) = b4 (Y1 ) + 1

b1 (Y2 ) = b1 (Y1 ) = 0 b5 (Y2 ) = b5 (Y1 ) = 0
b2 (Y2 ) = b2 (Y1 ) b6 (Y2 ) = b6 (Y1 ) = 1
b3 (Y2 ) = b3 (Y1 )

6.5.2 Type II Moves

An essentially identical argument shows that the Betti numbers of Y are invariant
under Type II moves. We are now in a position to prove Theorem 9.

6.5.3 Proof of Theorem 9

Since Y and Y differ by a sequence of moves of Type I and II, and their inverses,
we have that b2 (Y ) = b2 (Y ). This is part (a) of the Theorem. Furthermore b3 (Y ) =
b3 (Y ), and if the sequence of moves connecting Y to Y contains M moves of Type I
and N moves of (Type I)−1 then b4 (Y ) = b4 (Y ) + M − N , and the numbers k and k
of quadrilaterals in the polyhedral decompositions defining Y and Y satisfy k =
k + M − N . The vanishing cycle analysis (6.5) now implies (b).
To prove (c), it suffices to show that the quantities l and m occurring in Eq. (6.6)
are the same for Y and Y . This is obvious for the number of nodes m: the conjectural
formula (6.8) for m depends only on the sizes of the partitions of  e , which in turn
depends on the choices of Minkowski decomposition (A) but not on the fine mixed
subdivisions (B). It remains to show that the dimension l of the subspace L of H2 (Y )
spanned by the classes of exceptional curves in the smoothing Yt is the same as the
dimension l of the subspace L of H2 (Y ) spanned by the classes of exceptional
curves in the smoothing Yt . Let us return to the situation considered in Sect. 6.5.1,
where Y1 and Y2 are three-dimensional toric varieties that differ by a Type I move.
We showed there that H 2 (Y1 ) and H 2 (Y2 ) are isomorphic, via the inclusions
154 T. Coates et al.

H 2 (Y1 ) H 2 (Y2 )

j1∗ j2∗

H 2 (Z 0 ).

Dualising gives
H2 (Y1 ) H2 (Y2 )

j1 ∗ j2 ∗

H2 (Z ).
0

and since the subspace of H2 (Yi ) spanned by exceptional curves is pushed forward
from H2 (Z 0 ) via ji ∗ it follows that the dimension l of this subspace is also invariant
under Type I moves. Repeating this analysis for Type II moves shows that l = l , and
proves Theorem 9.

6.6 Systematic Analysis

The computation of Betti numbers described in Sect. 6.3 can be automated. The key
ingredients are as follows.
1. Algorithms for computing with lattice polyhedra and their duals. There are sev-
eral robust and well-tested implementations here, including those in Magma [4],
Sage [20], and polymake [13].
2. The Kreuzer–Skarke classification [17] of three-dimensional reflexive polytopes.
3. Altmann’s determination [3] of all Minkowski summands of a given polytope.
4. The computation of fine mixed subdivisions (Definition 6), that is, the determi-
nation of all regular triangulations of a Cayley polytope. For this we use Jörg
Rambau’s TOPCOM package [19].
5. An HPC cluster. Some of the computations involved are quite challenging.
Full source code for these computations, written in Magma, can be found at [6].
This relies in an essential way on code from the Fanosearch project [9].
There are 4319 three-dimensional reflexive polytopes, which in total admit more
than a billion decomposition data. These decomposition data give rise to 3857 distinct
Minkowski polynomials,4 which together give mirrors to the 98 three-dimensional

4 The number of Minkowski polynomials here differs slightly from the count in [2], because there
the authors required Minkowski decompositions of facets to satisfy an additional lattice condition
(ibid., Definition 7) and here we do not regard GL(3, Z)-equivalent Minkowski polynomials as the
same.
6 On the Topology of Fano Smoothings 155

Fano manifolds with very ample5 anticanonical bundle. We analysed several mil-
lion decomposition data, including at least one decomposition for each of the 3857
Minkowski polynomials. In each case we found that

the Betti numbers of the smoothing X η determined by the decomposition

data depend only on the mirror Fano manifold X , and coincide with those (∗ )
of X.

This provides evidence for the conjectural picture described in the introduction6 :
that if a Fano manifold X corresponds under Mirror Symmetry to a Laurent poly-
nomial f then there is a degeneration X →  with general fiber X and special
fiber the toric variety defined by the spanning fan of the Newton polytope of f . It
also provides evidence for the conjectures of Corti–Hacking–Petracci described in
Sect. 6.3, on the number and homology class of the exceptional curves in their res-
olution πt : Y y → X t . If these conjectures are correct then, in view of Theorem 9,
3857 of these calculations give a computer-assisted rigorous proof of (∗ ).

Acknowledgements This project has received funding from the European Research Council (ERC)
under the European Union’s Horizon 2020 research and innovation programme (grant agreement
No. 682603), and from the EPSRC Programme Grant EP/N03189X/1, Classification, Computation,
and Construction: New Methods in Geometry. We thank Paul Hacking and Andrea Petracci for
many extremely useful conversations, and thank Andy Thomas, Matt Harvey, and the Imperial
College Research Computing Service team for invaluable technical assistance. TC thanks Alexander
Kasprzyk for a number of very useful conversations about computations.

References

1. Akhtar, M., Coates, T., Corti, A., Heuberger, L., Kasprzyk, A.M., Oneto, A., Petracci, A.,
Prince, T., Tveiten, K.: Mirror symmetry and the classification of orbifold del Pezzo surfaces.
Proc. Am. Math. Soc. 144(2), 513–527 (2016)
2. Akhtar, M., Coates, T., Galkin, S., Kasprzyk, A.M.: Minkowski polynomials and mutations.
SIGMA Symmetry Integrability Geom. Methods Appl. 8, Paper 094, 17 (2012)
3. Altmann, K.: The versal deformation of an isolated toric Gorenstein singularity. Invent. Math.
128(3), 443–479 (1997)
4. Bosma, W., Cannon, J., Playoust, C.: The Magma algebra system. I. The user language, pp.
235–265 (1997). Computational algebra and number theory (London, 1993)
5. Candelas, P., Green, P.S., Hübsch, T.: Rolling among Calabi-Yau vacua. Nucl. Phys. B 330(1),
49–102 (1990)
6. Coates, T., Corti, A., Da Silva Jr., G.: Code Repository (2019). https://bitbucket.org/fanosearch/
3d_fano_smoothing
7. Coates, T., Corti, A., Galkin, S., Golyshev, V., Kasprzyk, A.M.: Mirror symmetry and Fano
manifolds. In: European Congress of Mathematics, pp. 285–300. European Mathematical Soci-
ety, Zürich (2013)

5 The seven three-dimensional Fano manifolds without very ample canonical bundle are not expected
to admit Laurent polynomial mirrors with reflexive Newton polytopes, and so fall outside the range
of the Corti–Hacking–Petracci construction.
6 Betti numbers for three-dimensional Fano manifolds can be found in [15].
156 T. Coates et al.

8. Coates, T., Corti, A., Galkin, S., Kasprzyk, A.M.: Quantum periods for 3-dimensional Fano
manifolds. Geom. Topol. 20(1), 103–256 (2016)
9. Coates, T., Kasprzyk, A.M.: Code repository (2019). https://bitbucket.org/fanosearch/magma-
core
10. Corti, A., Hacking, P., Petracci, A.: Smoothing Toric Fano Threefolds (2021). In preparation
11. De Loera, J.A., Rambau, J., Santos, F.: Triangulations. Structures for algorithms and applica-
tions. Algorithms and Computation in Mathematics, vol. 25. Springer-verlag, Berlin (2010)
12. Fulton, W., Sturmfels, B.: Intersection theory on toric varieties. Topology 36(2), 335–353
(1997)
13. Gawrilow, E., Joswig, M.: Polymake: a framework for analyzing convex polytopes. In:
Polytopes—Combinatorics and Computation (Oberwolfach, 1997), DMV SEM, vol. 29, pp.
43–73. Birkhäuser, Basel (2000)
14. Grothendieck, A., Raynaud, M., Rim, D. (eds.): Groupes de monodromie en géométrie
algébrique. I. Lecture Notes in Mathematics, Vol. 288. Springer, Berlin, New York (1972).
Séminaire de Géométrie Algébrique du Bois-Marie 1967–1969 (SGA 7 I)
15. Iskovskikh, V.A., Prokhorov, Y.G.: Fano varieties. In: Algebraic geometry, V. Encylopaedia of
Mathematical Sciences, vol. 47, pp. 1–247. Springer, Berlin (1999)
16. Jordan, A.: Homology and cohomology of toric varieties. Ph.D. thesis, Universität Konstanz
(1997)
17. Kreuzer, M., Skarke, H.: Classification of reflexive polyhedra in three dimensions. Adv. Theor.
Math. Phys. 2(4), 853–871 (1998)
18. Namikawa, Y.: Smoothing Fano 3-folds. J. Algebr. Geom. 6(2), 307–324 (1997)
19. Rambau, J.: TOPCOM: triangulations of point configurations and oriented matroids. In: Math-
ematical Software (Beijing, 2002), pp. 330–340. World Scientific Publishing, River Edge, NJ
(2002)
20. The Sage Developers: SageMath, the Sage Mathematics Software System (Version 8.4.0)
(2018). http://www.sagemath.org
Chapter 7
Computing Seshadri Constants on
Smooth Toric Surfaces

Sandra Di Rocco and Anders Lundman

Abstract In this paper we consider the problem of computing Seshadri constants

at a general point on a smooth polarized toric surface. We consider the case when
the degree of jet separation is small or the core of the associated polygon is a line
segment. Our main result is that under these hypothesis the Seshadri constant at the
general point can often be determined in terms of easily computable invariants of
the surfaces at hand. When the core of the associated polygon is a point we show
that the surface can be constructed via consecutive equivariant blow-ups of either P2
or P1 × P1 .

Keywords Toric geometry · Polytopes · Surface classification · Seshadri constants

7.1 Introduction

Let X be a complex projective variety and L be a nef line bundle on X .

Definition 1 For a point x ∈ X the Seshadri constant at x is defined as the number

L·C
(X, L; x) = inf ,
C mult x (C)

where the infimum is taken over all irreducible curves C passing through x.
The motivation for studying Seshadri constants is that they measure the local positiv-
ity of L at x. This can be seen from the Seshadri criterion which says that L is ample
if and only if (X, L; x) > 0 for all points x ∈ X [11, Theorem 1.4.13]. These con-
stants were first introduced by Demailly in relation to the Nagata conjecture [1, 4].

S. Di Rocco (B) · A. Lundman

Department of Mathematics, KTH Royal Institute of Technology, SE-100 44 Stockholm, Sweden
e-mail: [email protected]
A. Lundman
e-mail: [email protected]

© Springer Nature Switzerland AG 2022 157

A. M. Kasprzyk and B. Nill (eds.), Interactions with Lattice Polytopes,
Springer Proceedings in Mathematics & Statistics 386,
https://doi.org/10.1007/978-3-030-98327-7_7
158 S. Di Rocco and A. Lundman

In most cases the interest lies on the value of the Seshadri constant at a general point,
which is harder to compute with respect to specific points, like torus invariant points
for toric varieties. By semi-continuity the value of these constants at special points
can drop below that of general points. As a consequence the known results are often
either bounds or only valid for certain special points of certain classes of varieties,
see for example [1, 10, 15]. In this paper we approach the problem of computing
Seshadri constants at the general point on smooth polarized toric surfaces. We do so
by relating these constants to other invariants from local positivity and adjunction
theory that are easy to compute on smooth toric surfaces.
The methods used in this paper rely on the fact that a polarized toric variety (X, L)
corresponds to a convex lattice polytope PL . Under this correspondence the fixpoints
of the torus action on X correspond to the vertices of P. The Seshadri constant at these
fixpoints is equal to the lattice length of the shortest edge through the corresponding
vertex of PL , [1, 9], and it is thus easy to compute. It is known that the values
of the Seshadri constant at the general point and at the general point of all torus-
invariant subvarieties of X determine the Seshadri constant at every point of X [9,
Proposition 3.2]. For a general point there is however no exact description, even
though lately some very useful bound has been proven [9, 14].
In this note we related the Seshadri constant to the degree of jet separation, s(L, 1),
and the unnormalized spectral value, μ(L). This allows us to compute the Seshardi
constant at the general point for a large class of smooth toric surfaces. Unfortunately
not for all. For high values of either the degree of jet separation or unnormalized
spectral value the combinatorics quickly becomes intractable, see Example 40. We
will therefore concentrate on low values of these constants.
Definition 2 Let x be a point of a projective variety X with maximal ideal mx . The
degree of jet separation s(L, x) of a line bundle L at the point x ∈ X is the largest
integer k for which the natural map jxk : H 0 (X, L) → H 0 (X, L ⊗ O X /mk+1
x ) is onto.
Like Seshadri constants the degree of jet separation is a semi-continuous invariant.
In fact for smooth toric varieties s(L, x) obtains its minimum at some torus fixpoint
and its maximum at the general point [14]. We will use the notation s(L, 1) to denote
the degree of jet separation at a general point. Our first main result says that if s(L, 1)
is small on a smooth polarized toric surface, then it equals the Seshadri constant.
Theorem 3 Let (X, L) be smooth polarized toric surface with s(L, 1) ≤ 2, then

(X, L; 1) = s(L, 1).

We remark that we do not know if the bound s(L, 1) ≤ 2 in Theorem 3 is sharp.

Unfortunately the combinatorics involved in the proof gets out of hand already in the
case s(L, 1) = 3. On the one hand it was shown in [12] that if s(L, x) is constant
at all points of a smooth polarized toric variety (X, L), then (X, L; x) = s(L, x)
for all points x ∈ X . Moreover the results in [12] suggest that there might be an
integer k > 2 such that if s(L, 1) ≤ k then (X P , L P ; 1) = s(L, 1).
Example 40, due to Atsushi Ito [8], provides a smooth toric surfaces with s(L, 1) =
7 and (X, L; 1) = 7. Thus the constant k has to be lower than 7.
7 Computing Seshadri Constants on Smooth Toric Surfaces 159

Let K X denote the canonical divisor on X . A linear system of the form |K X +

sL| is called an adjoint linear system. Such systems play a prominent role in the
classification of varieties and define important invariants, see [2] for more details.
One such invariant is the unnormalized spectral value.
Definition 4
  
 dim(H 0 (X, m(K + sL)) = 0 for all integers m > 0
 X
μ(L) = sup s ∈ Q  .
 such that m(K X + sL) is an integral Cartier divisor

Observe that since any ample line bundle is big it holds that μ(L) < ∞. Rationality
of the nef and ample cone implies that μ(L) ∈ Q for toric varieties. By the toric
dictionary any big Q-Weil divisor L on a toric variety X of dimension n corresponds
to a n-dimensional polytope PL ⊂ Rn , whose vertices have rational coordinates. Any
such polytope has a unique minimal description as

PL = {x ∈ Rn : Ax ≥ b}

where b ∈ Qn and A is a matrix with integer coefficient with the entries in every row
being relatively prime. If PL is as above, then tK X + L corresponds to the polytope

PtK X +L = {x ∈ Rn : Ax ≥ b + t1},

where 1 = (1, . . . , 1)T . Here the polytope P ab K X +L is the polytope obtained by first
dilating the polytope PL by the factor b and then moving all supporting hyperplanes a
steps inwards. It follows that μ(L) is the maximum over all t such that PtK X +L is
non-empty. Following [5] we will call the polytope Pμ(L)−1 K X +L the core of PL and
denote it by core(PL ). Observe that, unless P is a point, the dimension of PL is
strictly larger than the dimension of its core. Thus if X is a toric surface, PL is a
polygon and core(PL ) is either a line segment or a point. Our second main result is
the following

Theorem 5 Let (X, L) be a smooth polarized toric surface associated to the poly-
tope PL . If core(PL ) is a line segment and μ(L)−1 < 3, then (X, L; 1) = 2μ(L)−1 .

We remark that, unlike the Seshadri constant at the general point, μ(L) is eas-
ily computable in the toric setting, see [5, Proposition 1.14]. It is worth pointing
out that for (P2 , OP2 (1)) it holds that core(PO(1) ) is a point, μ(O(1))−1 = 1/3,
while (P2 , OP2 (1); 1) = 1. Thus the assumption that core(P) is a line segment is
necessary in Theorem 5. We do not however have an example of a smooth polar-
ized toric surface (X, L) such that core(PL ) is a line segment, while (X, L; 1) =
2μ(L)−1 . Unfortunately it is apparent from the proof of Theorem 5 that the combina-
torics involved quickly grows, with increasing values of μ(L)−1 . This is further sup-
ported by Example 34 which shows that further complications will appear for higher
values of μ(L)−1 . Moreover Example 33 indicates that already the case μ(L)−1 = 3
160 S. Di Rocco and A. Lundman

will be more involved than the cases we consider. We leave it as an open question to
find an optimal bound on μ(L)−1 in Theorem 5.
Our next result makes use of Lemma 31, which says that if PL is a polygon
and core(P) is a line segment, then there exist two edges e and e
of PL which
are parallel to core(P). To state our next theorem we let K (P) be the linear space
parallel to the affine hull of core(P). Moreover we will let π : R2 → R2 /K (P)
and π ⊥ : R2 → R2 /K (P)⊥ be the natural projections, where K (P)⊥ is the orthog-
onal complement of K (P).

Theorem 6 Let (X, L) be a smooth polarized toric surface associated to the poly-
tope PL and assume that core(P) is a line segment. If π has a fiber of lattice length
at least 2μ(L)−1 that is a rational polytope or if there is a fiber of π ⊥ intersecting
both e and e
, then (X, L; 1) = 2μ(L)−1 .

For our purposes Theorem 6 serves as an important tool in proving Theorem 5.

However as the assumptions of Theorem 6 are easily checked it provides a quick
method to compute the Seshadri constant in many cases, even when μ(L)−1 ≥ 3.
Therefore we believe that it can be of independent interest.
In the final part of the paper we focus, in contrast to Theorems 5 and 6, on the
case when core(P) is a point. We show the following characterization.

Theorem 7 Let (X, L) be a smooth polarized toric surface such that core(P) is
a point, then X is given by a sequence of consecutive equivariant blow-ups of P2
or P1 × P1 .

7.1.1 Open Questions and Future Directions

This paper offers encouraging results for a class of smooth toric embeddings. Results
that open a series of future directions and related questions.
One natural question to ask is if the bounds appearing in Theorems 3 and 5 are
sharp. In the case of Theorem 3 this means asking for a minimal integer k such that
if s(L P , 1) ≤ k, then (X P , L P ; 1) = s(L P , 1). By Example 40 we know that k < 7,
but it would be interesting to find a sharp bound.
In relation to Theorem 5, we do not believe that the equality (X P , L P ; 1) =
2μ(L P )−1 will hold in more generality, whenever core(P) is a line segment. However
we are not aware of any counter example. Finding such an example would be enlight-
ening and give further insight into the relation between (X P , L P ; 1) and μ(L P ),
for such varieties.
Regarding Theorem 7, notice that as P2 and P1 × P1 are projective bundles Theo-
rem 3 implies that s(L, 1) = (X, L; 1) for these surfaces. However when consider-
ing equivariant blow-ups of these surfaces it is not clear how the relationship between
the Seshadri constant and degree of jet separation will change. It is worth pointing out
that, as shown in Example 40, it can happen that s(L, 1), (X, L; 1) and 2μ(L)−1
are all distinct if core(P) is a point. The surface appearing in Example 40 can be
7 Computing Seshadri Constants on Smooth Toric Surfaces 161

described both via six consecutive equivariant blow-ups of P2 and via five consec-
utive equivariant blow-ups of P1 × P1 . If instead X is obtained by a small number
of blow-ups of either P2 or P1 × P1 it seems reasonable to believe that (X, L; 1)
could be computed in terms of s(L, 1) or μ(L)−1 . For example the blow-up of P2 at
one fixpoint is the projective bundle F1 so s(L, 1) = (F1 , L; 1) for any ample line
bundle L on F1 . The main difficulty in proving such statements lies in the number
of different varieties achieved through consecutive equivariant blow-ups grows very
fast in the number of blow-ups performed. We leave the computation of the Seshadri
constant at the general point for these varieties as an open problem.
Another possible direction for future research is trying to achieve similar results
for higher dimensional smooth polarized toric varieties.

7.2 Background

We start with a basic example on Seshadri constants.

Example 8 It holds that (Pn , OPn (d); x) = d for all x ∈ Pn . This follows from
the fact that for any irreducible curve C and point x ∈ X , we have that O(d) · C ≥
d mult x (C) with equality when C is a line.
Another measure of local positivity is given by considering osculating spaces.
Definition 9 Let X be a projective variety and let L be a line bundle on X . For a
non-negative integer k and a smooth point x ∈ X consider the natural map

jxk : H 0 (X, L) → H 0 (X, L ⊗ O X /mk+1

X )

The projectivization of the image P(Im( jxk )) is called the kth osculating space,
Tkx (X, L) of (X, L) at x. Moreover we will say that L is k-jet spanned at x if jxk is
onto and call L k-jet spanned if jxk is onto for all points x ∈ X .
We remark that the map jxk is givenmby sending a section s to its Taylor expan-
sion around x. Thus if H 0 (X, L) = i=1 Csi , then dim(Tkx (X, L)) = rk(Jxk ) − 1,
where Jxk is the matrix whose rows are given by the partial derivatives of (s0 , . . . , sm )
of order at most k evaluated at x. We will call Jxk the matrix of k-jets at x. Thus
given sufficient knowledge about the global sections of a line bundle L, it is straight
forward to check if L is k-jet spanned at a given point x. As in the introduction we
will let s(L, x) denote the degree of jet separation of L at x, which is defined as
the maximal k such that L is k-jet spanned at x. By counting the number of partial
derivatives of order at most k it follows from the above that s(L, x) satisfies the
relation  
n + s(L, x)
≤ dim(H 0 (X, L)) < ∞. (7.1)
s(L, x)

Thus the the degree of jet separation, s(L, x) can be computed in finite time.
162 S. Di Rocco and A. Lundman

As s(L, 1) is considerably easier to compute than (X, L; 1) it is of interest to

relate the two. An important theorem in this direction is the following
Theorem 10 ([4]) Let X be a projective variety and L be a nef line bundle on X .
For any point x ∈ X it holds that

s(nL, x)
(X, L; x) = lim
n→∞ n
We have the following well-known corollary of Theorem 10, which we will use
repeatedly in this paper.
Corollary 11 Let X be a projective variety and L be a nef line bundle on X . For
any point x ∈ X it holds that

s(L, x) ≤ (X, L; x).

Proof If L is k-jet spanned line bundle, then nL is nk-jet spanned. Therefore it holds
that s(nL, x) ≥ ns(L; x) and the corollary follows by taking the limit in Theorem 10.


7.2.1 Toric Geometry

In this paper we assume that the reader is familiar with the basic properties of toric
varieties. We recommend [3] as a good general introduction. A key fact in toric
geometry is that any normal toric variety X  corresponds to a polyhedral fan .
Moreover every torus invariant subvariety of codimension d on a toric variety X 
corresponds to a d-dimensional cone in the fan . In particular any torus invariant
prime divisor Dρ corresponds to a 1-dimensional cone ρ, which is typically called
a ray of , and fixed points m σ correspond to the cones σ.
This interplay between toric and convex geometry is stronger when considering
polarized toric varieties (X, L). In fact any such variety (X, L) corresponds to a
convex lattice polytope PL . To see how this correspondence is defined let D be
a divisor having the property that O(D) = L. As the class group of a projective
toric variety
 X is generated by the torus invariant prime divisors of X , it holds
that D = ρ∈(1) aρ Dρ , where (1) denotes the set of rays of  and aρ ∈ Z. We
then define the polytope PL associated to (X, L) as

PL = {x ∈ Rn : x, ρ ≥ −aρ , ∀ρ ∈ (1)}.

Via this connection many concepts in algebraic geometry can be understood in con-
vex geometric terms and vice versa. One example is the following theorem. After
choosing a basis of the lattice Zn and hence choosing an affine patch Uσ correspond-
7 Computing Seshadri Constants on Smooth Toric Surfaces 163

ing to the vertex vσ placed at 0, then one has a convenient monomial basis for the
global sections:
Theorem 12 ([3, Theorem 4.3.3]) Let X be a projective toric variety and L a line
bundle on X, then

H 0 (X, L) ∼
= C < t1m 1 · · · tnm n > .
(m 1 ,...,m n )∈PL ∩Zn

where the monomial basis is express in terms of torus coordinates (t1 , · · · , tn ) in the
corresponding affine patch Uσ .
In particular every entry in the matrix of k-jets, Jxk is the evaluation of a monomial.
This leads to the following proposition.
Proposition 13 ([14]) Let X be a smooth polarized toric variety. Then s(L, 1) = k
if and only if there exist a polynomial of degree k + 1 vanishing on the lattice points
of PL and any other polynomial with the same property has at least degree k + 1.
We recall that a polarized toric variety (X, L) of dimension n is smooth if and only
if for every vertex v of PL the shortest lattice vectors along the edges through v form
a basis for Zn . We will call these vectors the edge-directions at v.
Next we introduce maps of fans which are the combinatorial equivalent of equiv-
ariant maps of toric varieties.

Definition 14 Let  and 

be fans in Rn . A linear map φ : Rn → Rn is called a map
of fans if for every σ
∈ 
there exist a σ ∈  such that φ(σ
∩ Zn ) ⊂ σ ∩ Zn . If φ
is a map of fans, then we often write φ : 
→  in place of φ : Rn → Rn .

We remark that a map of fans 

→  induces an equivariant morphism X 
→
X  of toric varieties [6, Theorem VI.6.1]. The most important examples of maps of
fans are obtained by letting 
be a refinement of  and φ be the identity. Recall
that 
is a refinement of  if for every σ
∈  it holds that σ
⊂ σ for some
cone σ ∈ . When 
is a refinement of , we will call  a coarsening of 
.

Example 15 Let X  be a smooth toric surface, then blowing up a torus fixpoint

of X  can be realised as a map of fans. To see this recall that the fixpoints of X 
correspond to the maximal dimensional cones in . Choosing a fixpoint xσ to blow
up corresponds to a choice of such a cone σ = cone(e1 , e2 ), where the ei :s can be
taken to be a basis of Z2 since X  is smooth. The fan 
given by replacing σ with
the cones cone(e1 , e1 + e2 ) and cone(e1 + e2 , e2 ) while keeping all other cones in ,
defines a new fan 
. The identity map id : Rn → Rn is a map of fans id : 
→ 
which induces a toric morphism π : X 
→ X  . By [6, Theorem VI.7.2], π : X 
→
X  is exactly the blow-up of X  at xσ . A corresponding construction works when
blowing-up any torus invariant subvariety of a smooth toric variety of any dimension
(Fig. 7.1).

Smooth toric surfaces are characterized by the following theorem.

164 S. Di Rocco and A. Lundman

Fig. 7.1 The fan of P2 and

the blow-up of P2 in a torus
fixpoint

Theorem 16 ([6, Theorem VI.7.5]) Let X be a smooth toric surface. Then there
exist a chain of equivariant blow-downs of X to either a Hirzebruch surface or P2 .
We next focus on Seshadri constants on toric varieties.
Definition 17 Let P be a full-dimensional polytope. Then P is called a Cayley
polytope if the vertices of P are contained in two parallel hyperplanes H0 and H1 .
Moreover if the lattice distance between H0 and H1 is k, then P is said to be of
type [P0 ∗ P1 ]k , where P0 = P ∩ H0 and P1 = P ∩ H1 . Lastly if P is a Cayley
polytope, then we call the projection onto the linear space perpendicular to H0 and H1
the defining projection of P.
In particular P is a Cayley polytope of type [P0 ∗ P1 ]1 if and only if P has lattice
width 1. In this case the defining projection projects P onto a line segment of lattice
length 1.
The two following theorems relate Seshadri constants and Cayley polytopes.
Theorem 18 ([10, Theorem 1.3]) Let P be a lattice polytope associated to the
polarized toric variety (X P , L P ), then the following are equivalent:
1. (X P , L P ; 1) = 1;
2. P = [P0 ∗ P1 ]1 ;
3. for every point p ∈ X P there exist a curve C containing p, having the property
that (C, L|C ) ∼ = (P1 , OP1 (1)).
Theorem 19 ([12]) Let P be a smooth lattice polytope associated to the polar-
ized toric variety (X P , L P ) and let k be a positive integer. Then the following are
equivalent:
1. (X P , L P ; x) = k for all points x ∈ X P ;
2. s(L P , x) = k for all points x ∈ X P ;
3. P = [P0 ∗ P1 ]k and every edge of P has lattice length at least k.
The following theorem gives bounds on the Seshadri constant at a general point
and is the most important result in establishing the new results of this paper.
Theorem 20 ([9, Theorem 3.6]) Let M be a lattice and MR = M ⊗R R. Let fur-
ther P ⊂ MR be a smooth polygon and π : M → Z be a lattice projection. Then

min{|πR (P)|π(M) , |F| M } ≤ (X P , L P ; 1) ≤ |πR (P)|π(M)

where πR is the map πR : MR → R induced by π , F is any fiber which is also a

rational polytope, and | · | M is the lattice distance with respect to the lattice M.
7 Computing Seshadri Constants on Smooth Toric Surfaces 165

7.2.2 Adjunction Theory for Toric Varieties

Given a linear system |L| on projective variety X , a linear system of the form |K X +
sL| is called an adjoint system to L. In this paper we are interested in two invariants
associated to |K X + sL|, namely the nef value and the unnormalized spectral value.

Definition 21 Let (X, L) be a polarized variety we call

τ (L) = inf {s ∈ R : K X + sL is nef } .

the nef value of L.

We remark that τ (L) equals the largest s such that K + sL is nef but not ample [2].
The second invariant of interest to us is the unnormalized spectral value and we recall
its definition from the introduction.
Definition 22 Let (X, L) be a polarized variety, then the unnormalized spectral
value of L is
  
 dim(H 0 (X, m(K + sL)) = 0 for all integers m > 0
 X
μ(L) = sup s ∈ Q  .
 such that m(K X + sL) is an integral Cartier divisor

Both the nef value and the unnormalized spectral value have convex geometric
interpretations for toric varieties. We briefly recall these interpretations.
Remark 23 Let P be a lattice polytope associated to the polarized toric vari-
ety (X P , L P ). For any s ∈ Q the polytope PsK X P +L P is denoted by P (s) . The nef
value is the minimum over all s such that the inner-normal fan of P is a refinement of
the inner-normal fan of P (1/s) . The unnormalized spectral value μ(L P ) of (X P , L P )
is the minimum over all s such that P (1/s) is non-empty. The polytope P (1/μ(L P )) ,
called the core of P, is a polytope of strictly smaller dimension than P and is denoted
by core(P). We remark that the core of a polytope and μ(L P )−1 are easily computable
using [5, Proposition 1.14]. These definitions are illustrated in Fig. 7.2, for further
details on these correspondences we refer to [5].

Definition 24 Let P ⊂ Rn be a polytope and let K (P) be the linear space parallel
to the affine span of the core of P. The projection π : Rn → Rn /K (P) is called
the natural projection associated to P.

Remark 25 If one restricts the attention to polarized toric surfaces, then there are
two possibilities, either core(P) is a point or a line segment. When the core of P is
a point the natural projection is the identity and when it is a line segment the natural
projection is the projection onto a line.
166 S. Di Rocco and A. Lundman

Fig. 7.2 A polytope P, P (1/τ (L P )) , core(P) = P (1/μ(L P )) and the natural projection

7.3 Seshadri Constants and Jet Separation

In the following section we prove Theorem 3. To make the exposition more accessible
we first prove Lemmas 27 and 28 which are key-steps in the proof of the theorem.
We further make the following notational definition.
Definition 26 Let P be a smooth polytope, then we say that P is canonically posi-
tioned if P has a vertex at the origin and an edge along each coordinate axis in the
positive direction.

Lemma 27 Let P be a smooth Cayley polygon associated to the smooth polarized

toric surface (X P , L P ), then s(L P , 1) = (X P , L P ; 1).

Proof Since P has dimension 2 the assumption that P is a Cayley polytope means
that either P ∼= k 2 = conv((0, 0), (k, 0), (0, k)) or P is the convex hull of two line
segments. If P ∼ = k 2 then s(L P , 1) = (X P , L P ; 1) = k. If instead P = [l0 , l1 ]k ,
where l0 and l1 are line segments, then there are two cases: either k = s(L P , 1)
or k = s(L P , 1). If k = s(L P , 1), then projecting onto the defining projection of P
shows that (X P , L P ; 1) ≤ s(L P , 1) which by Corollary 11 implies s(L P , 1) =
(X P , L P ; 1). If instead k = s(L P , 1), we first observe that it must hold that k >
s(L P , 1), otherwise the defining projection of P would give that (X P , L P ; 1) <
s(L P , 1) by Theorem 20, which contradicts Corollary 11. Thus without loss of gen-
erality we can assume that k > s(L P , 1) and that P is canonically positioned with
the longest defining line segment along the x-axis. The assumption k > s(L P , 1)
implies that (0, s(L P , 1) + 1) ∈ P. There are now two cases:
1. (s(L P , 1) + 1, 0) ∈
/ P, in which case the projection onto the first coordinate axis
shows that (X P , L P ; 1) = s(L P , 1) by Theorem 20 and Corollary 11.
2. (s(L P , 1) + 1, 0) ∈ P, in which case (s(L P , 1) + 1) 2 ⊆ P by convexity. This
implies that P is (s(L P , 1) + 1)-jet spanned at the general point contradicting
the definition of s(L P , 1).
From the above we conclude that s(L P , 1) = (X P , L P ; 1). 
7 Computing Seshadri Constants on Smooth Toric Surfaces 167

Fig. 7.3 Five lattice points

in P, a configuration of four
lines and open regions

Lemma 28 Let P be a smooth polygon associated to the smooth toric surface

(X P , L P ). If s(L P , 1) = 2, then (X P , L P ; 1) = 2.

Proof By Lemma 27 we can without loss of generality assume that P is not a Cayley
polytope. Assume that P is canonically positioned, then (0, 0), (1, 0), (0, 1) ∈ P by
the smoothness assumption. Furthermore we claim that the points (2, 1) and (1, 2)
must lie in P. To see why this is the case note that if either of these points fail to be
in P, then by smoothness and convexity either P = 2 2 or all lattice points of P
lie on two parallel lines at lattice distance 1 apart. In the former case s(L P , 1) =
(X P , L P ; 1) = 2 and the latter is a contradiction to s(L P , 1) = 2. From convexity
it now follows that (1, 1) ∈ P. Thus the indicated lattice points in Fig. 7.3 lie in P.
In Fig. 7.3 we also introduce four lines and enumerate some open regions delimited
by these lines.
We now observe that if either the horizontal, vertical or both diagonal lines are
supporting lines of P, then there exists a projection showing that (X P , L P ; 1) ≤ 2
by Theorem 20. Thus without loss of generality we can assume that P contains at
least one point on the other side of the horizontal and vertical line and one point
outside the diagonal strip. Those constrains can be satisfied in a few different ways
and we will end the proof by considering all possibilities.
(1) If P contains a lattice point in region 4, then convexity implies that the
points (2, 2) and (3, 2) lie in P. Furthermore if there is a point in region 4, then
the line L 2 cannot be a supporting line of P. By the smoothness assumption at the
vertex of P on the x-axis, different from the origin, it then follows that (3, 1) ∈ P.
The matrix of 3-jets evaluated at 1, for the 9 lattice points we now know lie in P,
consists of 9 independent rows and a row of zeros corresponding to the derivative ∂∂y 3 .
3

However any monomial corresponding to a point in region 4 will yield a non-zero

result when taking the derivative ∂∂y 3 and evaluating at 1. Thus we can conclude
3

that L P is 3-jet spanned at the general point if it contains a point in region 4, which
is a contradiction. We remark that this argument also takes care of the case when
168 S. Di Rocco and A. Lundman

there is a point of P on the line L 1 to the right of (3, 2). Since in that case either L 1
is a supporting line of P or P contains a point whose y-coordinate is at least 3. In
the former case (X P , L P ; 1) ≤ 2 and there is nothing to prove while in the latter
case L P is generically 3-jet spanned. Moreover by symmetry P cannot have a point
in region 2 and can be assumed not to have a point on L 2 above (2, 3).
(2) Assume that P contains a lattice point in region 5. We claim first that without
loss of generality we can assume that (3, 1) ∈ P. To see this note simply that other-
wise (3, 0) ∈ P but (3, 1) ∈ / P, which implies that P must be a Cayley polytope. By
smoothness and convexity it then follows that (3, 0) ∈ P or (3, 2) ∈ P. If (3, 2) ∈ P
we can apply the same argument as in the case when P contains a point in region 4.
Thus we can without loss of generality assume that (3, 0) ∈ P but that (3, 2) ∈ / P.
Next we note that we can assume that (2, 2) ∈ P. To see this note that otherwise one
readily checks that, (1, 2) has to be a vertex of P. Thus L 1 is a supporting line of P
and the projection onto the y-axis shows that (X P , L P ; 1) = 2. Thus we can assume
that (2, 2) ∈ P. Let J13 be the matrix of 3-jets at the general point for the convex hull
of the points we now can assume lie in P. Then J13 has nine linearly independent
rows and one row of zeros corresponding to the derivative ∂∂y 3 . Thus there are two
3

possibilities: either P contains a point above L 1 and L P is 3-jet spanned at the general
point or the projection onto the y-axis shows that (X P , L P ; 1) = 2. In conclusion
if P contains a point in region 5 and s(L P , 1) = 2, then (X P , L P ; 1) = 2. The case
when P contains a point in region 1, follows by symmetry.
By the above we can assume that all lattice points of P lie in the convex hull of L 2
and L 3 , except possibly the points (2, 0) and (0, 2). Thus if (X P , L P ; 1) = 2, then P
must contain a point in the region 3 or on one of the lines L 2 and L 3 further away from
the origin than (2, 1) or (2, 1) respectively. However convexity and smoothness then
show that neither (2, 0) nor (0, 2) can be a point in P, since if for example (2, 0) ∈ P
then (3, 1) ∈ P, which is a contradiction. Thus P is contained in the convex hull of L 2
and L 3 and the diagonal projection shows that (X P , L P ; 1) = 2. 

We are now ready to prove Theorem 3.

Proof (Proof of Theorem 3) We can assume s(L P , 1) ≥ 1, since L P is very ample.

Thus there are two cases either s(L P , 1) = 1 or s(L P , 1) = 2. In the first case we
have that s(L P , x) = 1 al all points in X and thus Theorem 19 gives the conclusion.
In the latter case Lemma 28 gives (X P , L P ; 1) = s(L P , 1). 

Corollary 29 Let P be a smooth polygon such that |P ∩ Zn | ≤ 20, then it holds

that (X P , L P ; 1) = s(L P , 1).

Proof If |P ∩ Zn | < 20, then s(L P , 1) ≤ 2, by the bound (7.1) and Theorem 3
implies the claim. If instead |P ∩ Zn | = 20, then either s(L P , 1) ≤ 2 or s(L P , 1) = 3
and P = 3 2 by the main theorem in [7]. However 3 2 corresponds to (P2 , O(3)),
so the corollary follows since s(O(3), 1) = (P2 , O(3); 1) = 3. 

The following example shows that if one considers non-complete embeddings,

then it may happen that s(L P , 1 P ) < (X P , L P , 1 P ).
7 Computing Seshadri Constants on Smooth Toric Surfaces 169

Example 30 Let P be the following smooth polygon corresponding to a non-

complete embedding. (The white point is not included.)

We claim that s(L P , 1) = 1 while (X P , L P , 1) = 2. To prove the claim note first
that s(L P , 1) = 1 since f (x, y) = x(x − 1) + y(y − 1) − x y is a degree 2 polyno-
mial vanishing at the black points. To show that (X P , L P , 1) ≥ 2 it is by Theo-
rem 10 enough to show that s(2L P , 1) = s(2L P
, 1) = 4, where P
is the complete
embedding corresponding to the polytope conv(P ∩ M). This follows from the fact
that 2P
∩ M = 2P ∩ M. Which in turn follows from the following set of inclusions
of sets.

2P ∩ M ⊇ P ∩ M + P ∩ M = P
∩ M + P
∩ M = 2P
∩ M ⊇ 2P ∩ M

Here the first and last inclusion follows by definition, while the second equality
follows by the projective normality of P
. Checking the first equality can be done
directly since a priori a lattice point p = (x, y) + (z, w) ∈ P
∩ M + P
∩ M can
fail to be in P ∩ M + P ∩ M if and only if (x, y) = (1, 1) or (z, w) = (1, 1). By
symmetry assume (x, y) = (1, 1). If (z, w) ∈ {(0, 0), (1, 1), (2, 1)}, then (x, y) −
(0, 1) ∈ P ∩ M and (z, w) + (0, 1) ∈ P ∩ M hence

p = (x, y) − (0, 1) + (z, w) + (0, 1) ∈ P ∩ M + P ∩ M.

If (z, w) ∈ {(0, 1), (1, 0)}, then (x, y) − (1, 1) ∈ P ∩ M and (z, w) + (1, 1) ∈ P ∩
M. If instead (z, w) = (1, 2), then (x, y) − (1, 0) ∈ P ∩ M and (z, w) + (1, 0) ∈
P ∩ M. Finally if (z, w) = (2, 2), then (x, y) + (1, 0) ∈ P ∩ M and (z, w) − (1, 0)
∈ P ∩ M. We conclude on the one hand that 2P ∩ M = 2P
∩ M, i.e. that (X P ,
L P , 1) ≥ 2. On the other hand it is clear that s(nL P , 1) ≤ s(nL P
, 1), since the global
sections of L P is a subspace of the global sections of L P
. Thus (X P , L P , 1) =
(X P , L P
, 1) = 2 by Theorem 10.

7.4 Seshadri Constants and Unnormalized Spectral Values

In the current section we prove Theorems 5 and 6. We start by showing the following
lemma.
Lemma 31 Let P be a polygon. If core(P) is a line segment, then

(X P , L P ; 1) ≤ 2μ(L P )−1

170 S. Di Rocco and A. Lundman

and there exist two edges e and e

which are parallel to core(P).
Proof Let aff(core(P)) be the affine hull of core(P) and K (P) be the parallel linear
space. Furthermore we let dn (aff(core(P))) denote the distance from the support-
ing hyperplane of P with normal n to aff(core(P)). By [5, Lemma 2.2] the nor-
mals of the supporting hyperplanes with dn (aff(core(P))) = μ(L P )−1 positively
span K (P)⊥ . Since aff(core(P)) is one dimensional and P is convex it then fol-
lows that on either side of aff(core(P)) there is an edge which is parallel and at
lattice distance μ(L P )−1 from aff(core(P)). Denote these edges by e and e
. Theo-
rem 20 implies that (X P , L P ; 1) ≤ |π(P)| for any projection π induced by a lattice
projection. Thus projecting onto K (P)⊥ shows that (X P , L P ; 1) ≤ 2μ(L P )−1 . 
Remark 32 We remark that the assumption that the core of P is a line segment
is a necessary condition even in the smooth setting. To see this it is enough to
consider (P2 , O(1)). It is well-known that s(O(1), 1) = (P2 , O(1); 1) = 1, which
can also be seen from Theorem 19. However it is straight forward to check that P (1/3)
is a point, so that core(P) is a point and 2μ(L P )−1 = 2/3.
We will prove Theorem 6 before proving Theorem 5.
Proof (Proof of Theorem 6) By Lemma 31 it is sufficient to show that

(X P , L P ; 1) ≥ 2μ(L P )−1 .

Assume that P is canonically positioned with the edge e of Lemma 31 along the x-
axis. By Theorem 20 it is enough to show the existence of a projection π such
that |π(P)| ≥ 2μ(L P )−1 and a fiber F of π that satisfies |F| ≥ 2μ(L P )−1 and
corresponds to a rational polytope. Choosing π to be the natural projection proves
the first part of the theorem since e and e
are at distance 2μ(P)−1 apart.
It remains to consider the case when there is a fiber of the projection onto the x-
axis containing both a point in e and a point in e
. Let π be the projection onto
the x-axis. The assumption is equivalent to requiring that there exists a fiber of π
with length exactly 2μ(P)−1 . Since e and e
are lattice line segments, the fiber can
be taken to be a lattice polytope. It remains to show that |π(P)| ≥ 2μ(P)−1 .
Define now e = (e1 , e2 , . . . , e p ) to be the sequence of distinct edges of maximal
length such that:
1. e1 is the edge sharing a vertex with e different from the origin;
2. ei shares a vertex with ei−1 ;
3. e p = e
.
Let furthermore K be the sequence defined by letting ki be the slope of the edge ei
and ki = ∞ if ei is a vertical edge. From convexity we first observe that the
sequence K can be:
1. positive and strictly monotonically increasing, not including ∞;
2. positive and strictly monotonically increasing and then negative and strictly
monotonically increasing.
7 Computing Seshadri Constants on Smooth Toric Surfaces 171

2 2

1 1

Fig. 7.4 Illustration of e when K has one and two phases

In the former case we will say that K has one phase and in the latter that K has two
phases (see Fig. 7.4). We will call the subsequence of K that is strictly positive and
monotonically increasing the first phase of K and the subsequence which is negative
and strictly monotonically increasing the second phase of K.
Consider first the case when K has one phase and let e p be the last edge in e.
By the smoothness assumption e p must pass through a lattice point on the line y =
2μ(L P )−1 − 1 and end at a vertex of the edge e
. As e
lies on the line y = μ(L P )−1
we can therefore conclude that e p has slope at most 1. Because K is strictly increasing
the same holds for all edges in e. Thus |π(e)| is larger or equal to the distance between
the x-axis and the line y = 2μ(L P )−1 , i.e. |π(P)| ≥ 2μ(L P )−1 .
We now turn to the case when K has two phases. We claim that by the smoothness
of P these two phases of the sequence K are separated by ∞, corresponding to the
slope of a vertical line. To see this let (−a, −b) and (x, y) be the edge-direction
of two adjacent edges in e, where a, b, y > 0 while x ≤ 0. Thus (−a, −b) is the
direction vector of an edge of the first phase of K and (x, y) is a connecting edge
which does not belong to this phase. Then the smoothness condition at the vertex of
intersection between these edges says that
 
 −a x 
  = bx − ay = ±1.
 −b y 

As a, b, y > 0 while x ≤ 0 it follows that x = 0 and y = a = 1. Thus the edge having

direction (x, y) is vertical. Furthermore there is a vertical edge of P along the y-
axis. Let these edges lie on the lines x = 0 and x = c. Then the lines x = μ(L P )−1
and x = c − μ(L P )−1 give halfspaces containing core(P). As core(P) is a line
segment it then follows that c > 2μ(L P )−1 . As a consequence |π(P)| > 2μ(L P )−1
and (X P , L P ; 1) = 2μ(L P )−1 . 

Example 33 Consider the smooth polygon P depicted in Fig. 7.5, which has the
property that core(P) is a line segment and μ(L P )−1 = 3.
Here the natural projection is the projection on the y-axis. The longest fibers of
that projection are the fibers between the two dashed lines. The length of any such
172 S. Di Rocco and A. Lundman

Fig. 7.5 A polytope P and

its core

core( )

fiber is 11/2. However the point 3 under the projection onto the x-axis has a fiber
containing a point in e and a point in e
. Thus Theorem 6 shows that (X P , L P ; 1) =
2μ(L P )−1 = 6.

We now prove Theorem 5.

Proof (Proof of Theorem 5) We will proceed as in the proof of Theorem 6. Therefore

we assume without loss of generality that P is canonically positioned with the edge e
of Lemma 31 along the x-axis. As P is a lattice polytope μ(L P )−1 ∈ 21 Z, since e
and e
must contain lattice points. We will proceed by considering all possible values
of μ(L P )−1 less than three.
If μ(L P )−1 = 1/2, then (X P , L P ; 1) = 2μ(L P )−1 = 1 by Theorem 6, since
both e and e
are fibers of the projection onto the y-axis.
Consider next the case μ(L P )−1 = 1. In this case the edge of P along the y-
axis either contains a vertex of both e and e
or ends at (0, 1). In the latter case the
smoothness assumption at (0, 1) implies that (1, 2) ∈ P. In either case there is a
fiber of the projection onto the x-axis with lattice length 2μ(L P )−1 , which shows
that (X P , L P ; 1) = 2μ(L P )−1 = 2.
Consider next the case μ(L P )−1 = 3/2. If either (3, 0) or (3, 1) lies in P, then by
projecting onto the y-axis it follows that (X P , L P ; 1) = 2μ(L P )−1 by Theorem 6.
Thus the vertex v of e different from the origin lies at a point (a, 1) with a < 3.
Consider now the edge f sharing the vertex v with e. By smoothness f must have
an edge-direction of the form (b, 1), with b ∈ Z. Assume now that v is at (2, 0).
If f passes through the point (0, 1) or (1, 1), then the projection onto the y-axis has
length less than 3, which is a contradiction. Hence without loss of generality we can
assume that f passes through the point (2, 1). Thus in this case x = 2 is a supporting
line of P, but then P (3/2) is empty which is a contradiction. Therefore we can assume
that f has a vertex at (1, 0) and that a = 1, but then again P (3/2) is again empty.
Consider the case μ(L P )−1 = 2. Arguing in the same way as for μ(L P )−1 =
3/2 we see that the edge f , defined as above, can be assumed to have a vertex
at either (1, 0) or (2, 0). If f has a vertex at (2, 0), then we can assume that f
passes through the point (3, 1). This is because otherwise either the fiber of 1 under
the projection onto the y-axis has length at least 2μ(L P )−1 or core(P) is empty.
7 Computing Seshadri Constants on Smooth Toric Surfaces 173

However it then follows that f has slope one and that core(P) is a point (or empty)
which is a contradiction.
Thus we can assume that f has a vertex at (1, 0). Then f can be assumed to
pass through the point (3, 1) since otherwise either the fiber of 1 has length at
least 2μ(L P )−1 or we get a contradiction to that core(P) is a line segment. In this
case let g be the edge sharing a vertex with f , other than e. By convexity the slope
of g is greater than that of f . However there must be an edge passing through a
lattice point on the line y = 2μ(L P )−1 − 1 that also shares a vertex with e
. Thus
by smoothness the only possibilities are g = e
or that g has slope 1 or 2/3. As a
consequence either f has length at least 2 of g has slope 2/3, since otherwise core(P)
would be a point (or empty). If f has length 2 the point (5, 2) lies in P. Therefore
the fiber of 2 under the projection onto the y-axis has length at least 2μ(L P )−1 ,
since (1, 2) ∈ P. If instead g has slope 2/3, then (6, 3) ∈ P. However by con-
vexity and smoothness either (2, 3) ∈ P or there is an edge h of slope 1 passing
through the lattice points (0, 1) and (1, 2). In the former case the fiber of 3 under
the projection onto the y-axis has length at least 4. In the latter case by moving the
edges e, e
, f and h we see that P (2) is a point which is a contradiction. This shows
that (X P , L P ; 1) = 2μ(L P )−1 = 4 in this case.
It remains to consider the case μ(L P )−1 = 5/2. Arguing in the same way as above
we see that we can assume that f has a vertex at either (1, 0) or (2, 0) or (3, 0). If f
has a vertex at (3, 0), then we can assume that f pass through the point (4, 1).
Since otherwise the fiber of 1 under the projection onto the y-axis has length at
least 2μ(L P )−1 or P (5/2) would be empty. Thus f can be assumed to have slope 1.
Furthermore f can be assumed to have length 1, since otherwise either (0, 2) ∈ P and
the fiber of 2 under the projection onto the y-axis has length 2μ(L)−1 or (0, 2) ∈ / P
−1
and P μ(L P ) is empty. Let g be the edge different than e sharing an edge with f .
By smoothness and convexity the edge-direction of g is (0, 1), (1, 2) or (2, 3), since
there must be an edge through a lattice point on the line y = 2μ(L P )−1 − 1 sharing
a vertex with e
. If g has edge-direction (1, 2), then the edge sharing a vertex with g
different from f must be vertical. Therefore if g has edge-direction (0, 1) or (1, 2),
we get that P (5/2) is a point or empty by moving e, e
and the vertical edges, yielding a
contradiction. Assume instead that g has edge-direction (2, 3). Then the edge sharing
the vertex (6, 4) with g must have slope 1 by smoothness and convexity, i.e. the same
slope as f which is a contradiction.
If instead f has a vertex at (2, 0), then f can be assume to pass through the
point (4, 1). Again because otherwise P (5/2) would be empty or the fiber of 1 would
have length at least 2μ(L P )−1 . If f has length at least 2, then (6, 2) ∈ P and the
fiber of 2 under the projection on the y-axis has length at least 2μ(L P )−1 . Thus we
can assume that f ends at (4, 1). Let g be the edge sharing an edge with f at (4, 1)
and let (a, b) be the edge-direction of g. Note first that b = 1, 2 since there must be
an edge of P passing through a lattice point on the line y = 2μ(L P )−1 that shares a
vertex with e
having a different slope than f .
If b = 1 then g has slope 1. In this case (1, 3) ∈ P, since otherwise P (5/2) is a
point or empty. Now either g has length at least 2 or g ends at (5, 2). In the former
case the fiber of 3 under the projection onto the y-axis has length 2μ(L P )−1 . In
174 S. Di Rocco and A. Lundman

the latter case let h be the edges sharing a vertex with g different from f . As there
must be an edge i passing through a lattice point on the line y = 2μ(L P )−1 − 1
sharing a vertex with e
, it follows from the smoothness assumption that h has edge-
direction (1, 2) and i is vertical. In this case the point (1, 4) ∈ P, since otherwise
there is an edge of P passing through the lattice points (1, 3) and (2, 4) and P (5/2)
is empty. But as (6, 4) ∈ P the fiber of 4 under the projection onto the y-axis has a
length 2μ(L P )−1 proving that (X P , L P ; 1) = 2μ(L P )−1 .
If instead b = 2, then smoothness and convexity implies that a = 3, i.e. (7, 3) ∈
P. Consider now the edge h different than f which shares a vertex with g. As there
must be an edge i through a lattice point on the line y = 2μ(L P )−1 − 1 sharing
a vertex with e
, the edge-directions of h are on the form (a, 1), with a ∈ Z. By
convexity and smoothness it then follows that a = 1 so that h has slope 1. We now
claim that (2, 3) ∈ P which proves that the fiber of 3 under the projection onto
the y-axis has length at least 2μ(L P )−1 . To see that (2, 3) ∈ P note first that by
smoothness the edge j different from e sharing a vertex with the edge along the y-
axis must contain a point on the line x = 1. If (2, 3) ∈ / P, then j must have slope 1
and end at (1, 2). However that leads to a contradiction since then P (5/2) is a point
(or empty), as can be seen by moving the edges e, e
, h and j.
Lastly we consider the case when (1, 0) is a vertex of P. In this final case we can
assume that the edge f must pass through either (3, 1) or (4, 1). Like before this is
because otherwise we either get a contradiction to that core(P) is a line segment or
the fiber of 1 under the projection onto the y-axis has length at least 2μ(L P )−1 . If f
passes through (4, 1), then either f has length at least 2 or f ends at (4, 1). By the
same argument as above the next edge g has edge-direction (2, 1), (5, 2) or (8, 3).
Thus regardless if f ends at (4, 1) or not it holds that (6, 2) ∈ P. Since (1, 2) ∈ P
it then follows that the fiber of 2 under the projection onto the y-axis has length at
least 2μ(L P )−1 .
It remains to check the case when f passes through (3, 1). In this case the
−1
point (1, 3) ∈ P, since otherwise P (μ(L P ) ) is empty. There are then two possi-
bilities either f has length at least 2 or it ends at (3, 1). It is easy to check that, like
before, the edge-directions of the next edge g is (5, 3), (3, 2) or (1, 1) by smoothness
and convexity. If f has length 1 and g has direction (1, 1) then one gets a contradic-
tion to that core(P) is a line segment. In all other cases (6, 3) ∈ P so the fiber of 3
under the projection onto the y-axis has length at least 2μ(L P )−1 . This concludes
the case μ(L P )−1 = 5/2 (Figs. 7.6, 7.7, 7.8 and 7.9). 

Example 34 Let P be the smooth polygon depicted in Fig. 7.10. It can readily be
checked that core(P) is a line segment and that μ(L P )−1 = 4. By direct computation
one can also show that s(L, 1) = 8, which implies that (X P , L P ; 1) = 8. However
the lattice length of a fiber is bounded from above by the maximal lattice length of a
line segment contained in P. It is tedious but straight forward to check that the lattice
length of any line segment in P is strictly less than 8. Thus there exists no projection
of P such that the lower bound in Theorem 20 suffices to compute (X P , L P ; 1).
7 Computing Seshadri Constants on Smooth Toric Surfaces 175

Fig. 7.6 Illustration of the

case μ(L P )−1 = 2
when (1, 0) is a vertex of P

Fig. 7.7 Illustration for the

case μ(P)−1 = 5/2
when (3, 0) is a vertex of P

Fig. 7.8 Illustration for a

special case
when μ(P)−1 = 5/2
and (2, 0) is a vertex of P

Fig. 7.9 Illustration for the

case when μ(P)−1 = 5/2
and (1, 0) is a vertex of P
176 S. Di Rocco and A. Lundman

Fig. 7.10 The smooth

polygon from Example 34

core( )

Remark 35 We are not aware of any smooth polygon P such that the core of P is
a line segment but 2μ(P)−1 = (X P , L P ; 1). However Example 34 illustrates that
if μ(L P )−1 ≥ 4, then the methods used to prove Theorem 5 will not suffice to show
that (X P , L P ; 1) = 2μ(L P )−1 .

7.5 Characterizing Polygons Whose Core is a Point

In the current section we characterize smooth polygons P such that core(P) is a

point.
Proposition 36 Assume P is a polytope such that core(P) is a point. If μ(L P ) =
τ (L P ), then X P is a Fano variety.

Proof The assumption that core(P) is a point is equivalent to H 0 (X P , μ(L P )−1 K X +

L P ) = C. Recall that for toric varieties μ(L P )−1 K X P + L P being nef is equivalent to
being globally generated, see [3, Theorem 6.3.12]. Thus we can assume that at every
point x ∈ X there exists a nowhere vanishing global section of μ(L P )−1 K X P + L P .
But as all such sections are linearly dependent it follows that μ(L P )−1 K X P + L P
has a global section s such that s(x) = 0 for all x ∈ X P . Thus μ(L P )−1 K X P + L P
is free of rank 1, i.e.

μ(L P )−1 K X P + L P = O X P ⇐⇒ −K X P = μ(L P )L P .

Thus X P is a Fano variety since L P was assumed to be ample and μ(L P ) > 0. 
7 Computing Seshadri Constants on Smooth Toric Surfaces 177

Corollary 37 Assume that P is a smooth polygon such that core(P) is a point

and μ(L P ) = τ (L P ). Then X P is a Del Pezzo surface.

Proposition 38 Assume P is a polygon such that core(P) is a point. If μ(L P ) =

τ (L P ) then X P is constructed by a sequence of toric morphisms onto a Fano variety.
Moreover this sequence of morphisms can be taken to only correspond to refinements
of the associated fans.

Proof By definition τ (L P )−1 K X + L P is a line bundle which is nef but not ample.
It follows that (X P , τ (L P )−1 K X P + L P ) corresponds to a polytope P1 whose inner-
normal fan is a coarsening of the inner-normal fan of X , since every normal of an
edge in P1 also is a normal of an edge in P. Let (X 1 , L1 ) be the polarized toric variety
and ample line bundle associated to P1 . Because the fan of X 1 is a coarsening of the
fan of X we get a toric morphism φ : X P → X 1 which contracts some torus invariant
curves on X P . Observe now that if a < τ (L P ), then (X, a −1 K X P + L P ) will yield
the same polygon as (X 1 , (a −1 − τ (L P )−1 )K X 1 + L1 ). To see this let E be a torus
invariant divisor contracted by φ, then (τ (L P )−1 K X P + L P ) · E = 0 and K X P · E <
0. Thus
     
1 1 1 1
KX P + LP · E = − KX P · E + K X P + L P · E < 0.
a a τ (L P ) τ (L P )

Therefore for any a < τ (L P ), there is no edge of the rational polygon P

associated
to the linear system (X P , a −1 K X P + L P ) which corresponds to E. Thus by the main
theorem of polytopes P
can be defined without the inequality corresponding to E.
We can apply the same procedure to P1 using the nef value of (X 1 , L1 ). Continuing
in this fashion, we obtain a sequence of toric morphisms:

XP X1 ... Xr (7.2)

Here X r has the property that μ(Lr ) = τ (Lr ). The argument for replacing K X P
with K X 1 shows that once the edge corresponding to a prime divisor has disappeared
it will not reappear later in the sequence. Thus all maps corresponds to refinements
of the associated fans. Furthermore the sequence is finite since its length is bounded
from above by the Picard number of X minus one. By the argument in the proof of
Proposition 36 it now follows that X r is a Fano variety. 

Corollary 39 Assume that P is a smooth polygon such that core(P) is a point,

then X P can be constructed by taking consecutive blow-ups of P2 or P1 × P1 .

Proof We start by showing that if X P is smooth, then the variety X r of Proposition 38

is smooth. Assume that this is not the case, then the sequence (7.2) is a resolution
of the singularities of X r , since all maps are birational
 and proper. Let f be the
composition of these maps, then K X P = f ∗ K X r + E i where the E i are the torus
invariant curves contracted by the chain of maps. Thus X r is a Fano variety with at
worse terminal singularities. Furthermore by [13, Lemma 1.17] a toric Fano variety
178 S. Di Rocco and A. Lundman

Fig. 7.11 The smooth

polygon from Example 40

of dimension d ≤ 3 with terminal singularities is smooth. Hence X r is a Del Pezzo

surface, i.e. it can be further blown-down to P2 or P1 × P1 . Furthermore as X P and X r
are both smooth, all maps in (7.2) must be blow-ups, by [6, Theorem V.6.2]. 

The following example due to Atsushi Ito shows that (X P , L P ; 1) may fail to be
an integer for smooth polarized toric surfaces. When this happens no combination
of Corollary 11 and Theorem 20 can be used to compute it.

Example 40 Let P be the smooth polygon depicted in Fig. 7.11. It is straight forward
to check that the core of P is a point, s(L P , 1) = 7 and μ(L P )−1 = 4. However
as shown by Ito [8], (X P , L P ; 1) = 15/2. Moreover this example is part of an
infinite family of smooth polarized toric surfaces having the property that the Seshadri
constant is an odd multiple of 1/2. Furthermore for any n Ito has constructed an
infinite family of smooth polarized toric n-folds such that (X P , L P ; 1) is an odd
multiple of 1/2.

Acknowledgements We are grateful to Atsushi Ito for sharing Example 40 with us and letting us
use it in the paper. The authors’ research was supported by the Swedish Research Council, grant
number 2014-4763. The research for this paper was supported by a VR grant [NT:2014-4763].

References

1. Bauer, T., Di Rocco, S., Harbourne, B., Kapustka, M., Knutsen, A., Syzdek, W., Szemberg,
T.: A primer on Seshadri constants. In: Interactions of Classical and Numerical Algebraic
Geometry. Contemporary Mathematics, vol. 496, pp. 33–70. American Mathematical Society,
Providence (2009)
2. Beltrametti, M.C., Sommese, A.J.: The Adjunction Theory of Complex Projective Varieties.
De Gruyter Expositions in Mathematics, vol. 16. Walter de Gruyter & Co., Berlin (1995)
7 Computing Seshadri Constants on Smooth Toric Surfaces 179

3. Cox, D.A., Little, J.B., Schenck, H.K.: Toric Varieties. Graduate Studies in Mathematics, vol.
124. American Mathematical Society, Providence (2011)
4. Demailly, J.P.: Singular Hermitian metrics on positive line bundles. In: Complex Algebraic
Varieties (Bayreuth, 1990). Lecture Notes in Mathematics, vol. 1507, pp. 87–104. Springer,
Berlin (1992)
5. Di Rocco, S., Haase, C., Nill, B., Paffenholz, A.: Polyhedral adjunction theory. Algebra Number
Theory 7(10), 2417–2446 (2013)
6. Ewald, G.: Combinatorial Convexity and Algebraic Geometry. Graduate Texts in Mathematics,
vol. 168. Springer, New York (1996)
7. Fulton, W., Kleiman, S., Piene, R., Tai, H.: Some intrinsic and extrinsic characterizations of
the projective space. Bull. Soc. Math. France 113(2), 205–210 (1985)
8. Ito, A.: Personal communication
9. Ito, A.: Seshadri constants via toric degenerations. J. Reine Angew. Math. 695, 151–174 (2014)
10. Ito, A.: Algebro-geometric characterization of Cayley polytopes. Adv. Math. 270, 598–608
(2015)
11. Lazarsfeld, R.: Positivity in algebraic geometry. I. Ergebnisse der Mathematik und ihrer Gren-
zgebiete. 3. Folge. A Series of Modern Surveys in Mathematics [Results in Mathematics and
Related Areas. 3rd Series. A Series of Modern Surveys in Mathematics], vol. 48. Springer,
Berlin (2004). Classical setting: line bundles and linear series
12. Lundman, A.: Local positivity of line bundles on smooth toric varieties and Cayley polytopes.
J. Symbolic Comput. 74, 109–124 (2016)
13. Nill, B.: Gorenstein toric Fano varieties. Manuscripta Math. 116(2), 183–210 (2005)
14. Perkinson, D.: Inflections of toric varieties. pp. 483–515 (2000). Dedicated to William Fulton
on the occasion of his 60th birthday
15. Szemberg, T.: An effective and sharp lower bound on Seshadri constants on surfaces with
Picard number 1. J. Algebra 319(8), 3112–3119 (2008)
Chapter 8
The Characterisation Problem of
Ehrhart Polynomials of Lattice Polytopes

Akihiro Higashitani

Abstract One of the most important invariants of a lattice polytope is the Ehrhart
polynomial. The problem of which polynomials can be Ehrhart polynomials of lattice
polytopes is now well-studied. In this survey paper, after recalling the fundamental
properties of the Ehrhart polynomials of lattice polytopes, we survey the results
on this problem including recent developments. We discuss the characterisation of
Ehrhart polynomials in several particular cases: small dimensions; small volumes;
palindromic; small degrees. We also suggest some possible further questions.

Keywords Lattice polytopes · Ehrhart polynomials · h ∗ -vector · h ∗ -polynomial

8.1 Introduction

We say that a convex polytope P ⊂ Rd is a lattice polytope if all of its vertices belong
to the standard lattice Zd . Given a lattice polytope P ⊂ Rd of dimension d, we can
associate the enumerative function n → (n P ∩ Zd ), which counts the number of
lattice points contained in the nth dilation of P. Ehrhart [11] showed that there
exists a polynomial E P (n) in n of degree d, where d is the dimension of P, such
that E P (n) = (n P ∩ Zd ) for any positive integer n. We call the polynomial E P (n)
the Ehrhart polynomial of P. The Ehrhart polynomial E P (n) has the following
properties:
1. the constant term of the polynomial E P (n) is always equal to 1;
2. the leading term of the polynomial is equal to the Euclidean volume of P.
Moreover, E P (n) also satisfies the Ehrhart-Macdonald reciprocity [27]:

(n P ◦ ∩ Zd ) = (−1)d E P (−n) for n ∈ Z>0 , (8.1)

A. Higashitani (B)
Department of Pure and Applied Mathematics, Graduate School of Information Science
and Technology, Osaka University, Osaka 565-0871, Japan
e-mail: [email protected]

© Springer Nature Switzerland AG 2022 181

A. M. Kasprzyk and B. Nill (eds.), Interactions with Lattice Polytopes,
Springer Proceedings in Mathematics & Statistics 386,
https://doi.org/10.1007/978-3-030-98327-7_8
182 A. Higashitani

where P ◦ denotes the relative interior of P. Namely, the number of lattice points
contained in the nth dilation of the interior of P can be expressed by using the Ehrhart
polynomial of P.
When formally expressed as a rational function, the generating function of E P (n)
is known to be 
∞ ∗ i
i≥0 h i t
E P (n)t =
n
,
n=0
(1 − t)d+1

and it is known that h i∗ = 0 for i ≥ d + 1 and that each h i∗ is an integer [30]. We call
the integer sequence
h ∗ (P) = (h ∗0 , h ∗1 , . . . , h ∗d )

the h ∗ -vector of P, denoted by h ∗ (P), and the polynomial


d
h ∗P (t) = h i∗ t i
i=0

the h ∗ -polynomial of P, denoted by h ∗P (t). In some literature, the h ∗ -vector or

the h ∗ -polynomial is called the δ-vector or the δ-polynomial (see, e.g., [13, 16, 17,
19]).
For the study of Ehrhart polynomials of lattice polytopes, it is often preferable to
use h ∗ -vectors (h ∗ -polynomials) instead of Ehrhart polynomials since the h ∗ -vectors
behave better than the coefficients of Ehrhart polynomials. For example, it is known
that h ∗ -vectors are always nonnegative integers [30], while the coefficients of Ehrhart
polynomials are not necessarily integers (just rational numbers) and not necessarily
nonnegative except for the leading, the second and the constant terms (see, e.g., [18]).
Given a lattice polytope, knowing its h ∗ -vector is equivalent to knowing its Ehrhart
polynomial. In fact, when h ∗ (P) = (h ∗0 , h ∗1 , . . . , h ∗d ), we see that


d  
n+d −i
E P (n) = h i∗ . (8.2)
i=0
d

We refer the reader to [7, 13] for an introduction to Ehrhart theory.

It follows from (8.1) that for a lattice polytope P with its h ∗ -vector (h ∗0 , h ∗1 , . . . ,
∗
h d ), we have

∞ d
 h ∗d−i t i+1
(n P ◦ ∩ Zd )t n = i=0
. (8.3)
n=1
(1 − t)d+1

The h ∗ -vectors of lattice polytopes have the following properties:

1. h ∗0 = 1, h ∗1 = (P ∩ Zd ) − (d + 1) and h ∗d = (P ◦ ∩ Zd );
 8 The Characterisation Problem of Ehrhart Polynomials of Lattice Polytopes 183

2. if P ◦ ∩ Zd is non-empty, i.e., h ∗d > 0, then h i∗ ≥ h ∗1 holds for i = 1, . . . , d −

1 [15]; d
3. the leading coefficient of E P (n) is equal to ( i=0 h i∗ )/d! by (8.2).
d
We call i=0 h i∗ the normalized volume of P, denoted by vol(P).
Note that h 1 ≥ h ∗d holds. In fact, since P is of dimension d, we have (∂ P ∩ Zd ) ≥
∗

d + 1, which implies that

h ∗1 = (P ◦ ∩ Zd ) + (∂ P ∩ Zd ) − d + 1 ≥ (P ◦ ∩ Zd ) = h ∗d .

Hence, we see that h ∗1 = h ∗d if and only if (∂ P ∩ Zd ) = d + 1, which implies that P

is a simplex. In particular, when h ∗1 = h ∗d = 0, P is a simplex which contains no lattice
point except for its vertices, called an empty simplex.
Let deg(P) denote the degree of the polynomial h ∗P (t), i.e., deg(P) = max{i :
∗
h i = 0}. The invariant deg(P) is called the degree of P. It is follows from (8.3) that

deg(P) = d + 1 − min{m : m P ◦ ∩ Zd = ∅}. (8.4)

In particular, deg(P) ≤ d.
The following is one of the most important unsolved problems in Ehrhart theory:
Problem 1 Characterise the sequences (or polynomials) that are the h ∗ -vectors
(or h ∗ -polynomials) of lattice polytopes.
This problem has two steps: to prove the necessity and the sufficiency. More pre-
cisely, one step is to show some conditions for an integer sequence (h ∗0 , h ∗1 , . . . , h ∗d )
to be the h ∗ -vector of some lattice polytope (e.g., h i∗ ≥ h ∗1 for each 1 ≤ i ≤ d − 1
if h ∗d > 0, h ∗1 ≥ h ∗d and so on), and the other step is to construct a concrete lattice
polytope whose h ∗ -vector coincides with a desired integer sequence.
We see that it is quite difficult to solve this problem in general. Furthermore, it
might be believed that there exists no nice solution to this problem. Thus, in this
survey paper, we will restrict to some particular cases. More concretely, we discuss
the following cases:
1. small dimensions, i.e.,dd is small—see Sect. 8.3;
2. small volumes, i.e., i=0 h i∗ is small—see Sect. 8.4;
3. palindromic, i.e., h i∗ = h ∗deg(P)−i for any i—see Sect. 8.5;
4. small degrees, i.e., deg(P) is small—see Sect. 8.6.

8.2 Preliminaries

In this section, we present the basic concepts that we will use in this paper.
1. For two lattice polytopes P, P ⊂ Rd , we say that P and P are unimodularly
equivalent if there exist f ∈ GLd (Z) and u ∈ Zd such that P = f (P) + u.
 184 A. Higashitani

2. For a lattice polytope P ⊂ Rd of dimension d, let

Pyr(P) = conv({(α, 0) ∈ Rd+1 : α ∈ P} ∪ {(0, . . . , 0, 1)}) ⊂ Rd+1 .

This new lattice polytope is said to be a lattice pyramid over P. It is not so hard
to see that for a lattice polytope P, we have [7, Theorem 2.4]
∞
 h ∗P (t)
E Pyr(P) (n)t n = .
n=0
(1 − t)d+2

In particular, we have

h ∗P (t) = h ∗Pyr(P) (t) and deg(P) = deg(Pyr(P)).

We recall the finite abelian group associated with a lattice simplex and discuss
some properties on a lattice simplex in terms of this group. Let  ⊂ Rd be a lattice
simplex of dimension d with its vertices v1 , v2 , . . . , vd+1 ∈ Zd . Let
 

d+1 
d+1
 = (x1 , x2 , . . . , xd+1 ) ∈ [0, 1)d+1 : xi vi ∈ Zd , xi ∈ Z
i=1 i=1

equipped with its addition defined by x + y = ({x1 + y1 }, . . . , {xd+1 + yd+1 }) ∈

[0, 1)d+1 for x = (x1 , . . . , xd+1 ) ∈ [0, 1)d+1 and y = (y1 , . . . , yd+1 ) ∈ [0, 1)d+1 ,
where {r } = r − r  denotes the fractional part of r ∈ R. We see that  is a
finite abelian group. In fact, one can see that 0 = (0, . . . , 0) ∈  and −x :=
({1 − x1 }, . . . , {1 − xd+1 }) ∈  for any x ∈  .
Let F (d) denote the set of unimodular equivalence classes of lattice simplices of
dimension d with a fixed vertex order, and let A(d) denote the set of finite abelian
subgroups  of [0, 1)d+1 satisfying that the sum of all entries of each element in 
is an integer. Actually, [6, Theorem 2.3] says that the correspondence

F (d) → A(d);  → 

is a bijection. In particular, a unimodular equivalence class of lattice simplices  is

uniquely determined by the finite abelian group  up to permutation of coordinates.
We can discuss h ∗ (t), deg(), vol() and whether  is a lattice pyramid in terms
of  . We fix some notation. For x = (x1 , . . . , xd+1 ) ∈ 
d+1
1. let ht(x) = i=1 xi ∈ Z≥0 ;
2. let supp(x) = {i ∈ [d + 1] : xi = 0}, where [n] = {1, . . . , n} for n ∈ Z>0 ;
3. let wt(x) = (supp(x)).
Then we have [7, Corollary 3.11]
8 The Characterisation Problem of Ehrhart Polynomials of Lattice Polytopes 185

h ∗ (t) = t ht(x) .
x∈

In particular,

deg() = max{ht(x) : x ∈  } and vol() = | |.

We see that a lattice simplex  of dimension d is not a lattice pyramid if and only if
for each i ∈ [d + 1], there is (x1 , . . . , xd+1 ) ∈  such that xi = 0.
Furthermore, we have

wt(α) = ht(α) + ht(−α) (8.5)

for any α ∈  since

 
wt(α) = 1= (αi + 1 − αi ) = ht(α) + ht(−α).
i∈supp(α) i∈supp(α)

8.3 Small Dimensions

We first consider the classification of h ∗ -polynomials of lattice polytopes in low

dimensions. Namely, we discuss whether a given integer sequence (h ∗0 , h ∗1 , . . . , h ∗d )
is a possible h ∗ -vector of some lattice polytope of dimension d for the case d is small.
In the case d = 1, the following is easy to prove.
Proposition 2 For any nonnegative integer a, the integer sequence (1, a) is the h ∗ -
vector of some lattice polytope of dimension 1.

Proof Take the lattice segment {x ∈ R : 0 ≤ x ≤ a + 1} of length a + 1. 

The case d = 2 is highly non-trivial.

Theorem 3 ([29]) Given nonnegative integers a and b, the integer sequence (1, a, b)
is the h ∗ -vector of a lattice polytope of dimension 2 if and only if (a, b) satisfies one
of the following conditions:
1. b = 0;
2. 0 < b ≤ a ≤ 3b + 3;
3. (a, b) = (7, 1).
The “only if” part is due to [29]. The “if” part is easy to see. In fact, we can construct
lattice polygons for any (a, b) satisfying 1, 2 or 3: see Fig. 8.1. Note that lattice
polygons in the Case 1 can be given by Proposition 2 and lattice pyramid construction.
For the case d = 3, the characterisation of h ∗ -vectors is wide open, with the
exception of a few partial results: the result in the case of palindromic h ∗ -polynomials
186 A. Higashitani

Fig. 8.1 Construction of lattice polygons whose h ∗ -vectors are one of 1, 2 or 3 in Theorem 3

will appear in Sect. 8.5 and the result in the case h ∗3 = 0 will appear in Sect. 8.6. The
characterisation in the case h ∗3 = 1 can be essentially obtained from [23] and the
case h ∗3 = 2 is from [3].
Moreover, by the series of the papers [8–10] by Blanco–Santos, an algorithm
which provides a complete list of all lattice polytopes of dimension 3 up to unimodular
equivalence has been given. This enables us to get the complete list of h ∗ -vectors
of lattice polytopes of dimension 3 containing only a few lattice points. In [10],
the list of lattice polytopes P ⊂ R3 of dimension 3 with (P ∩ Z3 ) ≤ 11 is given
up to unimodular equivalence. As a by-product, we also know the characterisation
of h ∗ -vectors (h ∗0 , h ∗1 , h ∗2 , h ∗3 ) with h ∗1 ≤ 7.
The contribution of the papers [8–10] naturally suggests the following problem:
Problem 4 Characterise the possible h ∗ -vectors of lattice polytopes of dimension 3.
In fact, Balletti [1, Conjecture 8.7] predicts some necessary inequalities for h ∗ -
vectors (h ∗0 , h ∗1 , h ∗2 , h ∗3 ) of lattice polytopes of dimension 3.

8.4 Small Volumes

Next, we consider the case of small normalized volumes. Namely, we discuss whether
a given integer sequence (h ∗0 , h ∗1 , . . . , h ∗d ) is a possible h ∗ -vector of some lattice
d
polytope of dimension d for the case i=0 h i∗ is small.
8 The Characterisation Problem of Ehrhart Polynomials of Lattice Polytopes 187

d
In the case i=0 h i∗ = 1, since each h i∗ is nonnegative and h ∗0 = 1, we conclude
that h 0 = 1 and h 1 = · · · = h ∗d = 0. This is the h ∗ -vector of a unit simplex of dimen-
∗ ∗

sion d, where a unit simplex of dimension d is the convex hull of the origin of Rd
d of R
d
and the unit vectors .
∗
For the case i=0 h i ≥ 2, we recall two well-known inequalities on h ∗ -vectors.
Let P ⊂ Rd be a lattice polytope of dimension d with degree s. One inequality is

h ∗0 + h ∗1 + · · · + h i∗ ≤ h ∗s + h ∗s−1 + · · · + h ∗s−i for i = 0, 1, . . . , s. (8.6)

This is proved by Stanley [31]. Another one is

h ∗d + h ∗d−1 + · · · + h ∗d−i ≤ h ∗1 + h ∗2 + · · · + h i+1

∗
for i = 0, 1, . . . , d − 1. (8.7)

This appears in the work of Hibi [15]. Note that these inequalities (8.6) and (8.7) are
vastly generalized by Stapledon [32, 33]. All inequalities in (8.6) hold with equality
for any i if and only if h i∗ = h ∗s−i for i = 0, 1, . . . , s. This is equivalent to the lattice
polytope being Gorenstein. See Sect. 8.5. All inequalities in (8.7) hold with equality
∗
for any i if and only if h i+1 = h ∗d−i for i = 0, 1, . . . , d − 1. This is called shifted
symmetric, introduced in [19]. A lattice polytope P has a shifted symmetric h ∗ -vector
if and only if P is a lattice simplex all of whose facets have normalized volume 1
(see [19, Theorem 2.1]). d
A characterisation in the cases of i=0 h i∗ = 2 and 3 is given as follows:
Theorem 5 ([17, Theorem 0.1]) Given a sequence (h ∗0 , h ∗1 , . . . , h ∗d ) of nonnegative
d
integers, where h ∗0 = 1 and i=0 h i∗ ≤ 3, there exists a lattice polytope of dimen-
sion d whose h ∗ -vector coincides with (h ∗0 , h ∗1 , . . . , h ∗d ) if and only if (h ∗0 , h ∗1 , . . . , h ∗d )
satisfies the inequalities (8.6) and (8.7).
The “only if” part directly follows from the works by Stanley and Hibi. The
essential part of this theorem is the “if” part.
dThe ∗following example shows that Theorem 5 is no longer true for the case
i=0 h i = 4.

Example 6 ([17, Example 1.2]) We consider the sequence (1, 0, 1, 0, 1, 1, 0, 0).

Then we see that this cannot be the h ∗ -vector of any lattice polytope of dimen-
sion 7. Suppose, on the contrary, that there exists a lattice polytope P with its h ∗ -
vector (1, 0, 1, 0, 1, 1, 0, 0). Since h ∗1 = 0, we know that P should be a simplex.
Let  P = {0, α, β, γ } be as in Sect. 8.2, where ht(α) = 2, ht(β) = 4 and ht(γ ) = 5.
Since −γ ∈  P and wt(γ ) ≤ 8 together with ht(−γ ) = wt(γ ) − 5 ∈ {2, 4, 5} by
(8.5), we see that wt(γ ) = 7 and −γ = α. Thus, we also see that −β = β. This
implies that wt(β) = 8 by (8.5). Hence, supp(γ ) = supp(α) = [8] \ {i} for some i ∈
[8] and supp(β) = [8]. Then one has i ∈ supp(α + β). This means that α + β = β,
which means that α = 0, a contradiction.
d
Before considering the statement in the case i=0 h i∗ = 4, we introduce the fol-
lowing notation. For a sequence (h 0 , h 1 , . . . , h d ) of nonnegative integers with h ∗0 = 1
∗ ∗ ∗
d
and i=0 h i∗ = m, let i 1 , . . . , i m−1 be the positive integers such that
 188 A. Higashitani


d
h i∗ t i = 1 + t i1 + · · · + t im−1 ,
i=0

where 1 ≤ i 1 ≤ · · · ≤ i m−1 ≤ d. For example, for a sequence (1, 0, 2, 3, 0, 1, 0), we

have i 1 = i 2 = 2, i 3 = i 4 = i 5 = 3 and i 6 = 5. Notice that the sequence (h ∗0 , h ∗1 , . . . ,
h ∗d ) and the positive integers i 1 , . . . , i m−1 have the same information. Actually, we
can rephrase the inequalities (8.6) and (8.7) as follows:
Proposition 7 ([20, Proposition 2.2]) With notation as above:
1. the inequalities i j + i m− j−1 ≥ i m−1 , where 1 ≤ j ≤ m − 2, are equivalent to the
inequalities (8.6);
2. the inequalities i j + i m− j ≤ d + 1, where 1 ≤ j ≤ m − 1, are equivalent to the
inequalities (8.7).
d
A characterisation in the case of i=0 h i∗ = 4 is given as follows:
Theorem 8 ([16, Theorem 5.1]) Let 1 + t i1 + t i2 + t i3 be a polynomial with 1 ≤
i 1 ≤ i 2 ≤ i 3 ≤ d. Then there exists a lattice polytope of dimension d whose h ∗ -
polynomial equals 1 + t i1 + t i2 + t i3 if and only if the triple (i 1 , i 2 , i 3 ) satisfies:
1. i 1 + i 2 ≥ i 3 , i 1 + i 3 ≤ d + 1 and 2i 2 ≤ d + 1 hold;
2. either 2i 2 ≤ i 1 + i 3 or i 2 + i 3 ≤ d + 1 holds.
We remark that as mentioned in Proposition 7, the first inequality in Condition 1
is equivalent to (8.6) and the second and third
d inequalities are equivalent to (8.7).
Condition 2 is the special one for the case i=0 h i∗ = 4.

Example 9 (Example 6 continued) Consider the sequence (1, 0, 1, 0, 1, 1, 0, 0)

again. For the above notation, we have i 1 = 2, i 2 = 4, i 3 = 5 and d + 1 = 8. Then
Condition 1 in Theorem 8 are satisfied, while both 2i 2 > i 1 + i 3 and i 2 + i 3 > d + 1
hold. Hence Condition 2 is not satisfied.
d
Remark 10 ([16, Remark 5.3]) For the case of i=0 h i∗ ≤ 4, we see from the proofs
∗
of Theorems 5 and 8 thatdall the possible h -vectors can be obtained by lattice sim-
plices. However, when i=0 h i∗ = 5, the h ∗ -vector (1, 3, 1) cannot be obtained from
any simplex, while it is a possible h ∗ -vector of a lattice polygon (see Theorem 3).

This remark implies that for the complete characterisations with larger normalized
volumes, we have to see the h ∗ -vectors of not only lattice simplices but also non-
simplices. Hence, for the further investigation, it is natural to think of the h ∗ -vectors
of lattice simplices as a first step.

Theorem 11 ([20, Theorem 1.1]) Let P be a lattice simplex dof dimension d and
let h ∗ (P) = (h ∗0 , h ∗1 , . . . , h ∗d ) be its h ∗ -vector. Suppose that i=0 h i∗ = p is an odd
d
prime. Let i 1 , . . . , i p−1 be the positive integers such that i=0 h i∗ t i = 1 + t i1 + · · · +
t i p−1 , where 1 ≤ i 1 ≤ · · · ≤ i p−1 ≤ d.Then:
 8 The Characterisation Problem of Ehrhart Polynomials of Lattice Polytopes 189

1. i 1 + i p−1 = i 2 + i p−2 = · · · = i ( p−1)/2 + i ( p+1)/2 ≤ d + 1; and

2. i k + i
≥ i k+
for k,
≥ 1 with k +
≤ p − 1.
By using these necessary conditions for integer sequences to be the h ∗ -vector of
lattice simplices, we can prove the following:
Theorem 12 ([20, Theorem 1.2]) Let 1 + t i1 + t i2 + t i3 + t i4 be a polynomial with 1
≤ i 1 ≤ · · · ≤ i 4 ≤ d. Then there exists a lattice simplex of dimension d whose h ∗ -
polynomial equals 1 + t i1 + t i2 + t i3 + t i4 if and only if (i 1 , i 2 , i 3 , i 4 ) satisfies i 1 +
i 4 = i 2 + i 3 ≤ d + 1 and i k + i
≥ i k+
for k,
≥ 1 with k +
≤ 4.
Theorem 13 ([20, Theorem 1.3]) Let 1 + t i1 + t i2 + t i3 + t i4 + t i5 + t i6 be a poly-
nomial with 1 ≤ i 1 ≤ · · · ≤ i 6 ≤ d. Then there exists a lattice simplex of dimen-
sion d whose h ∗ -polynomial equals 1 + t i1 + t i2 + t i3 + t i4 + t i5 + t i6 if and only
if (i 1 , i 2 , i 3 , i 4 , i 5 , i 6 ) satisfies i 1 + i 6 = i 2 + i 5 = i 3 + i 4 ≤ d + 1 and i k + i
≥ i k+

for k,
≥ 1 with k +
≤ 6.
d
Remark 14 (see [20, Sect. 5.1]) Theorems 12 and 13 say that when i=0 h i∗ is 5
or 7, the necessary conditions in Theorem 11 are also sufficient for lattice sim-
plices. However, this is not true for all prime numbers. In fact, since the volume
of a lattice polytope containing a unique lattice point in its interior has an upper
bound (see, e.g., [26]), if p is a sufficiently large prime number, then the integer
sequence (1, 1, p − 3, 1) cannot be the h ∗ -vector of any lattice simplex of dimen-
sion 3, although (1, 1, p − 3, 1) satisfies all the conditions in Theorem 11.
Recently, a complete characterisation of the possible h ∗ -vectors (h ∗0 , h ∗1 , . . . , h ∗d )
d
of lattice polytopes with i=0 h i∗ = 5 has been given.
Theorem 15 ([35, Theorem 0.4]) Let 1 + t i1 + t i2 + t i3 + t i4 be a polynomial with 1
≤ i 1 ≤ · · · ≤ i 4 ≤ d. Then there exists a lattice polytope of dimension d whose h ∗ -
polynomial equals 1 + t i1 + t i2 + t i3 + t i4 if and only if
1. (i 1 , i 2 , i 3 , i 4 ) satisfies the conditions in Theorem 12; or
2. 1 + t i1 + t i2 + t i3 + t i4 = 1 + 3t + t 2 or 1 + t + 3t 2 or 1 + t + t 2 + 2t 3 .
Note that the polynomials 1 + 3t + t 2 and 1 + t + 3t 2 and 1 + t + t 2 + 2t 3 are
the h ∗ -polynomials of some lattice polytopes respectively (see [35, Example 2.4] for
a more detailed construction of those lattice polytopes), but these are never the h ∗ -
polynomials of lattice simplices.
The proof of Theorem 15 relies on the following theorem, which is a recent, major
contribution to Ehrhart theory:
Theorem 16 ([22, Theorem 1.3]) If a lattice polytope P is spanning then the h ∗ -
vector of P has no gap, i.e., h i∗ > 0 for any i = 0, 1, . . . , deg(P).
Here, we say that a lattice polytope P ⊂ Rd is spanning if the affine lattice generated
by P ∩ Zd , denoted by  P , is equal to Zd , Note that the converse of Theorem 16 is not
true in general. In fact, the lattice simplex P = conv({e1 , e2 , ±(e1 + e2 + 2e3 )}) ⊂
R3 has the h ∗ -vector (1, 1, 2, 0), while this simplex is not spanning since e3 ∈ Z3 is
not contained in  P .
The key lemma for the proof of Theorem 15 is the following:
190 A. Higashitani

Lemma 17 ([35, Theorem 1.1]) Let P be a lattice polytope which is not a simplex.
Suppose that vol(P) is prime. Then P is spanning.
Note that this lemma is not true for non-prime case. For example, let

P = conv({0, e1 , e2 , e1 + e2 , e3 , e1 + e2 + e3 + 2e4 }) ⊂ R4 .

Then P is a lattice polytope with vol(P) = 4. Moreover, P has 6 vertices, i.e., P is

not a simplex, and P ∩ Z4 equals the set of vertices. Hence, we observe that e4 ∈
/ P .
Proof (Sketch of proof of Theorem 15) For the “if” part, we may construct the lattice
polytopes whose h ∗ -polynomials are 1 + 3t + t 2 , 1 + t + 3t 2 and 1 + t + t 2 + 2t 3 ,
respectively.
For the “only if” part, we consider the h ∗ -polynomial 1 + t i1 + · · · + t i4 of the
lattice polytope P which is not a simplex. Then it follows from Theorem 16 and
Lemma 17 that the h ∗ -polynomial has no gap. Thus, the possible h ∗ -polynomials are
the following polynomials:

1 + 4t, 1 + 3t + t 2 , 1 + t + 3t 2 , 1 + 2t + 2t 2 , 1 + t + 2t 2 + t 3 , 1 + t + t 2 + 2t 3 , 1 + t + t 2 + t 3 + t 4 .

Note that 1 + 2t + t 2 + t 3 violates the inequality (8.6). Since 1 + 4t, 1 + 2t +

2t 2 , 1 + t + 2t 2 + t 3 and 1 + t + t 2 + t 3 + t 4 are the possible h ∗ -polynomials of
lattice simplices by Theorem 12, we conclude the desired assertion. 
As a future problem, we suggest the following:
Problem 18 Characterise the h ∗ -polynomials of lattice polytopes with vol(P) = 7.
Towards this characterisation, the similar idea to the proof of Theorem 15 enables us
to deduce that we must consider the following polynomials:

1 + 2t + 3t 2 + t 3 , 1 + t + 3t 2 + 2t 3 , 1 + t + 2t 2 + 3t 3 , 1 + t + t 2 + 4t 3 ,
1 + t + t 2 + 3t 3 + t 4 , 1 + t + 3t 2 + t 3 + t 4 , 1 + t + t 2 + t 3 + 2t 4 + t 5 .

More precisely, we may determine the existence or non-existence of a lattice polytope

whose h ∗ -polynomial is one of them.
The following is also a natural problem to try as a next step:
Problem 19 Characterise the h ∗ -polynomials of lattice simplices with vol(P) = 6.

8.5 Palindromic
s
We say that a polynomial i=0 ai t i of degree s is palindromic if ai = as−i for i =
∗
0, 1, . . . , s. Next, we consider the case of palindromic hs -polynomials. Namely,
we discuss whether a given palindromic polynomial i=0 h i∗ t i of degree s is a
possible h ∗ -polynomial of some lattice polytope with degree s.
 8 The Characterisation Problem of Ehrhart Polynomials of Lattice Polytopes 191

Recall that a lattice polytope P ⊂ Rd is reflexive if the origin of Rd is contained

in the interior of P and the polar polytope

P ∨ = {x ∈ Rd : x, y ≤ 1 for any y ∈ P}

is also a lattice polytope, where ·, · denotes the usual inner product of Rd . Reflexivity
of lattice polytopes is characterised as follows:
Proposition 20 (cf. [5, 14]) Let P be a lattice polytope of dimension d, let
(h ∗0 , h ∗1 , . . . , h ∗d ) be its h ∗ -vector and let E P (n) = ad n d + ad−1 n d−1 + · · · + 1 be
its Ehrhart polynomial. Then the following four conditions are equivalent:
1. P is a reflexive polytope;
2. deg(P) = d and h ∗P (t) is palindromic, i.e., h i∗ = h ∗d−i for i = 0, 1, . . . , d;
3. the functional equation E P (n) = (−1)d E P (−n − 1) holds;
4. we have dad = 2ad−1 .
We also recall that a lattice polytope P ⊂ Rd is called Gorenstein of index

if
P is unimodularly equivalent to a reflexive polytope. Since (
P ∩ Zd ) = 1,
we see that deg(P) = dim P + 1 −
by (8.4). The following easily follows from
Proposition 20.
Proposition 21 Let P be a lattice polytope of dimension d with its h ∗ -vector
(h ∗0 , h ∗1 , . . . , h ∗d ). Then the following three conditions are equivalent:
1. P is a Gorenstein polytope of index
;
2. deg(P) = s and h ∗P (t) is palindromic, i.e., h i∗ = h ∗s−i for i = 0, 1, . . . , s, where
s = d + 1 −
;
3. the functional equation E P (n) = (−1)d E P (−n −
) holds.
Reflexive polytopes together with Gorenstein polytopes are particularly impor-
tant in many areas, e.g., commutative algebra, toric geometry and mirror symmetry
(e.g., [4, 5, 31] and so on). Hence, many researchers are quite interested in reflexive
polytopes and they have been intensively studied. On the characterisation of palin-
dromic h ∗ -polynomials, i.e., the h ∗ -polynomials of Gorenstein polytopes, the case
of 2-dimensional reflexive polytopes is well known:
Proposition 22 (See e.g. [28]) Given an integer a, the sequence (1, a, 1) is the h ∗ -
vector of a reflexive polytope of dimension 2 if and only if a ∈ {1, 2, . . . , 7}.
The case of Gorenstein polytopes of degree 2 is also known, which is highly
non-trivial:
Theorem 23 ([4, Theorem 2.10]) Given an integer a, the polynomial 1 + at + t 2 is
the h ∗ -polynomial of a Gorenstein polytope of degree 2 if and only if a ∈ {0, 1, . . . , 7}.
The following is one natural analogue of Theorem 23.
Theorem 24 ([21, Corollary 1.1]) Given an integer a, the polynomial 1 + at k + t 2k
is the h ∗ -polynomial of a Gorenstein polytope of degree 2k with k ≥ 2 if and only if
one of the following conditions holds:
 192 A. Higashitani

1. a ∈ {0, 1, 2, 4, 6};
2. there exist
≥ 4 and m ≥ 1 such that a = 2
− 2 and k = 2
−3 m;
3. there exist
≥ 3 and m ≥ 1 such that a = 3
− 2 and k = 3
−2 m.
The case of 3-dimensional reflexive polytopes is also known:
Proposition 25 ([24]) Given an integer a, the integer sequence (1, a, a, 1) is the h ∗ -
vector of a reflexive polytope of dimension 3 if and only if a ∈ {1, 2, . . . , 35} \
{33, 34}.
Similar to this theorem, the necessary and sufficient condition for the integer
sequence (1, a, b, a, 1) to be the h ∗ -vector of a reflexive polytope of dimension 4 is
also known by [25]. (There are 14, 373 possible h ∗ -vectors.)
Since we already know the possible h ∗ -vectors of reflexive polytopes of dimen-
sion 3, the following is a natural problem to try as a next step:
Problem 26 Characterise the h ∗ -polynomials of Gorenstein polytopes of degree 3.

8.6 Small Degrees

Next, weconsider the case of small degrees. Namely, we discuss whether a given poly-
nomial i=0s
h i∗ t i is a possible h ∗ -polynomial of some lattice polytope of degree s
when s is small.
In general, we see the following:
d
Proposition 27 If a given polynomial i=0 h i∗ t i is the h ∗ -polynomial of some lattice
polytope of dimension d, then it is also the h ∗ -polynomial of some lattice polytope
of dimension d + m for any m ∈ Z≥0 .

Proof Take the m times iterated lattice pyramids. 

Thus, we have the inclusion

{h ∗ -polynomials of lattice polytopes of dimension d}

⊂ {h ∗ -polynomials of lattice polytopes of degree at most d}.

In the case degree at most 2, we see that there is a difference between these two
sets. More precisely, the following is known. It looks quite similar to Theorem 3 but
slightly different.
Theorem 28 ([34, Theorem 2]) Let P be a lattice polytope of degree at most 2. Then
its h ∗ -polynomial 1 + h ∗1 t + h ∗2 t 2 satisfies one of the following conditions:
1. h ∗2 = 0;
2. h ∗1 ≤ 3h ∗2 + 3 and h ∗2 > 0; or
3. (h ∗1 , h ∗2 ) = (7, 1).
 8 The Characterisation Problem of Ehrhart Polynomials of Lattice Polytopes 193

Note that the conditions in Theorem 28 are also sufficient. Namely, it is proved
in [12, Proposition 1.10] that given nonnegative integers (h ∗1 , h ∗2 ) satisfying one of
the conditions in Theorem 28, there exists a lattice polytope P such that h ∗P (t) = 1 +
h ∗1 t + h ∗2 t 2 . Those polytopes can be constructed by lattice polytopes of dimension
at most 3. Hence, the possible h ∗ -polynomials of degree at most 2 are completely
characterised.
In Theorem 28, the condition “h ∗2 ≤ h ∗1 ” is missing, while b ≤ a appears in
Theorem 3. The inequality b ≤ a comes from h ∗1 ≥ h ∗d (see Introduction) and b ≤ a
is a particular condition for d = 2. Hence, Theorem 3 essentially characterises the
possible h ∗ -polynomials of lattice polytopes of not only dimension 2 but also degree
at most 2.
The following is natural to try as a further problem in addition to Problem 4.
Problem 29 Characterise the h ∗ -polynomials of lattice polytopes of degree ≤ 3.

8.7 Universal Inequalities

Finally, we introduce a new kind of condition on h ∗ -vectors of lattice polytopes.

Recently, the following theorem has been proved:
Theorem 30 ([2, Theorem 1.4]) Let P be a lattice polytope and let h ∗P (t) =
 ∗ i ∗ ∗ ∗
i≥0 h i t . Assume that h 3 = 0. Then (h 1 , h 2 ) satisfies one of the following con-
ditions:
1. h ∗2 = 0;
2. h ∗1 ≤ 3h ∗2 + 3 and h ∗2 > 0; or
3. (h ∗1 , h ∗2 ) = (7, 1).
Theorem 30 is a generalization of Theorem 28 since deg(P) ≤ 2 is equivalent
to h ∗3 = h ∗4 = · · · = 0. Theorem 30 says that the conditions in Theorem 28 is valid
independently of both the dimension and the degree of lattice polytopes. An inequal-
ity is therefore called universal if it does not depend on the dimension or the degree
of the lattice polytopes. (This terminology is suggested in [2].)

Example 31 ([2, Examples 1.5, 1.6])

1. Theorem 30 is no longer true if h ∗3 = 0. In fact, there is a lattice polytope of
dimension 5 whose h ∗ -polynomial equals 1 + 8t + t 2 + 8t 3 , while Theorem 28
says that there is no lattice polytope of degree 2 whose h ∗ -polynomial equals 1 +
8t + t 2 .
2. Theorem 30 cannot be generalized into the case h ∗3 < h ∗1 . In fact, there is a lattice
polytope of dimension 9 whose h ∗ -polynomial equals 1 + 17t + 4t 2 + 14t 3 +
8t 4 + 10t 5 + 12t 6 + 6t 7 , while Theorem 28 says that there is no lattice polytope
of degree 2 whose h ∗ -polynomial equals 1 + 17t + 4t 2 .

As a further investigation for this direction, we suggest the following:

194 A. Higashitani

Problem 32 Find other universal inequalities for h ∗ -vectors.

Acknowledgements The author would like to be grateful to Christopher Borger, Johannes Hof-
scheier, Alexander Kasprzyk and Benjamin Nill for the organization of and the invitation to the
wonderful conference “Interactions with Lattice Polytopes”, held in September 14–16 2017 at
Magdeburg. The author would also like to thank the anonymous referee for a lot of helpful com-
ments.

References

1. Balletti, G.: Enumeration of lattice polytopes by their volume. Discrete Comput. Geom. 65(4),
1087–1122 (2021)
2. Balletti, G., Higashitani, A.: Universal inequalities in Ehrhart theory. Israel J. Math. 227(2),
843–859 (2018)
3. Balletti, G., Kasprzyk, A.M.: Three-dimensional lattice polytopes with two interior lattice
points (2016). arXiv:1612.08918 [math.CO]
4. Batyrev, V., Juny, D.: Classification of Gorenstein toric del Pezzo varieties in arbitrary dimen-
sion. Mosc. Math. J. 10(2), 285–316, 478 (2010)
5. Batyrev, V.V.: Dual polyhedra and mirror symmetry for Calabi-Yau hypersurfaces in toric
varieties. J. Algebraic Geom. 3(3), 493–535 (1994)
6. Batyrev, V.V., Hofscheier, J.: Lattice polytopes, finite abelian subgroups in SL(n, C) and coding
theory (2013). arXiv:1309.5312 [math.CO]
7. Beck, M., Robins, S.: Computing the continuous discretely, 2nd edn. Undergraduate Texts
in Mathematics. Springer, New York (2015). Integer-point enumeration in polyhedra, With
illustrations by David Austin
8. Blanco, M., Santos, F.: Lattice 3-polytopes with few lattice points. SIAM J. Discrete Math.
30(2), 669–686 (2016)
9. Blanco, M., Santos, F.: Lattice 3-polytopes with six lattice points. SIAM J. Discrete Math.
30(2), 687–717 (2016)
10. Blanco, M., Santos, F.: Enumeration of lattice 3-polytopes by their number of lattice points.
Discrete Comput. Geom. 60(3), 756–800 (2018)
11. Ehrhart, E.: Sur les polyèdres rationnels homothétiques à n dimensions. C. R. Acad. Sci. Paris
254, 616–618 (1962)
12. Henk, M., Tagami, M.: Lower bounds on the coefficients of Ehrhart polynomials. Eur. J.
Combin. 30(1), 70–83 (2009)
13. Hibi, T.: Algebraic Combinatorics on Convex Polytopes. Carslaw Publications, Glebe (1992)
14. Hibi, T.: Dual polytopes of rational convex polytopes. Combinatorica 12(2), 237–240 (1992)
15. Hibi, T.: A lower bound theorem for Ehrhart polynomials of convex polytopes. Adv. Math.
105(2), 162–165 (1994)
16. Hibi, T., Higashitani, A., Li, N.: Hermite normal forms and δ-vectors. J. Combin. Theory Ser.
A 119(6), 1158–1173 (2012)
17. Hibi, T., Higashitani, A., Nagazawa, Y.: Ehrhart polynomials of convex polytopes with small
volumes. Eur. J. Combin. 32(2), 226–232 (2011)
18. Hibi, T., Higashitani, A., Tsuchiya, A., Yoshida, K.: Ehrhart polynomials with negative coef-
ficients. Graphs Combin. 35(1), 363–371 (2019)
19. Higashitani, A.: Shifted symmetric δ-vectors of convex polytopes. Discrete Math. 310(21),
2925–2934 (2010)
20. Higashitani, A.: Ehrhart polynomials of integral simplices with prime volumes. Integers 14,
Paper No. A45, 15 (2014)
21. Higashitani, A., Nill, B., Tsuchiya, A.: Gorenstein polytopes with trinomial h ∗ -polynomials.
Beitr. Algebra Geom. 62(3), 667–685 (2021)
8 The Characterisation Problem of Ehrhart Polynomials of Lattice Polytopes 195

22. Hofscheier, J., Katthän, L., Nill, B.: Ehrhart theory of spanning lattice polytopes. Int. Math.
Res. Not. IMRN 19, 5947–5973 (2018)
23. Kasprzyk, A.M.: Canonical toric Fano threefolds. Canad. J. Math. 62(6), 1293–1309 (2010)
24. Kreuzer, M., Skarke, H.: Classification of reflexive polyhedra in three dimensions. Adv. Theor.
Math. Phys. 2(4), 853–871 (1998)
25. Kreuzer, M., Skarke, H.: Complete classification of reflexive polyhedra in four dimensions.
Adv. Theor. Math. Phys. 4(6), 1209–1230 (2000)
26. Lagarias, J.C., Ziegler, G.M.: Bounds for lattice polytopes containing a fixed number of interior
points in a sublattice. Canad. J. Math. 43(5), 1022–1035 (1991)
27. Macdonald, I.G.: Polynomials associated with finite cell-complexes. J. Lond. Math. Soc. 2(4),
181–192 (1971)
28. Poonen, B., Rodriguez-Villegas, F.: Lattice polygons and the number 12. Amer. Math. Monthly
107(3), 238–250 (2000)
29. Scott, P.R.: On convex lattice polygons. Bull. Austral. Math. Soc. 15(3), 395–399 (1976)
30. Stanley, R.P.: Decompositions of rational convex polytopes. Ann. Discrete Math. 6, 333–342
(1980)
31. Stanley, R.P.: On the Hilbert function of a graded Cohen-Macaulay domain. J. Pure Appl.
Algebra 73(3), 307–314 (1991)
32. Stapledon, A.: Inequalities and Ehrhart δ-vectors. Trans. Amer. Math. Soc. 361(10), 5615–5626
(2009)
33. Stapledon, A.: Additive number theory and inequalities in Ehrhart theory. Int. Math. Res. Not.
IMRN 5, 1497–1540 (2016)
34. Treutlein, J.: Lattice polytopes of degree 2. J. Combin. Theory Ser. A 117(3), 354–360 (2010)
35. Tsuchiya, A.: Ehrhart polynomials of lattice polytopes with normalized volumes 5. J. Comb.
10(2), 283–290 (2019)
Chapter 9
The Ring of Conditions
for Horospherical Homogeneous Spaces

Johannes Hofscheier

Abstract These are notes of a five talks lecture series during the “Graduate Summer
School in Algebraic Group Actions”, at McMaster University, June 11th–15th, 2018.
The aim of this lecture series is to introduce the ring of conditions of a spherical homo-
geneous space with a special emphasis on the horospherical case, i.e., homogeneous
spaces with respect to a connected complex reductive group which are torus bundles
over a flag variety. In these notes, we start with an example from enumerative geom-
etry which naturally yields first instances of spherical varieties. We continue with
a recollection of the necessary background on reductive groups needed throughout
the rest of the manuscript. After that we introduce spherical varieties: we discuss the
Luna–Vust theory of spherical embeddings and explain the complete combinatorial
description of horospherical varieties (an important subfamily of spherical varieties).
We conclude with the definition of the ring of conditions of spherical homogeneous
spaces and give an explicit description for the horospherical case.

Keywords Spherical variety · Linear algebraic group · Enumerative geometry ·

Ring of conditions

9.1 Motivation

In the following, some elementary knowledge of algebraic geometry is expected

from the reader. Introductory texts which cover the required topics are, e.g., [9, 23,
24, 27]. Parts of this manuscript follow the lecture notes [11] by Kiritchenko.
First examples of spherical varieties emerged from enumerative geometry such as
Grassmannians. It turns out that many enumerative problems reduce to intersection
theoretic questions on algebraic varieties equipped with a “good” transitive (or almost
transitive) action of an algebraic group. Here is a classical example:

J. Hofscheier (B)
School of Mathematical Sciences, University of Nottingham, Nottingham NG7 2RD, UK
e-mail: [email protected]

© Springer Nature Switzerland AG 2022 197

A. M. Kasprzyk and B. Nill (eds.), Interactions with Lattice Polytopes,
Springer Proceedings in Mathematics & Statistics 386,
https://doi.org/10.1007/978-3-030-98327-7_9
198 J. Hofscheier

Problem 1 How many lines in 3-space C3 intersect 4 given lines in general position?

Recall the general trick how to rephrase an affine geometric question into a linear
one: Suppose X is an affine geometric object in Cn . Introduce one further dimension
and consider the linear span of X regarded as a subset of the affine hyperplane {xn+1 =
1} where xn+1 denotes the additional coordinate.
Hence, the above question reduces to a problem in the Grassmannian Gr(2, 4) (2-
planes in C4 ). This algebraic variety admits a transitive action by GL4 (the general lin-
ear group of invertible 4 × 4-matrices with complex entries). Indeed, let e1 , . . . , e4 be
the standard basis in C4 and consider the natural action of GL4 on C4 . Clearly GL4 acts
transitively on planes in C4 and the stabilizer P of the coordinate plane span{e1 , e2 }
is given by   
AC
P= : A, B ∈ GL2 , C ∈ Mat(2 × 2) .
0 B

Hence Gr(2, 4) ∼ = GL4 /P is a homogeneous space under GL4 and P is an example

of a parabolic subgroup (see definition below).
Note that by the transition to Gr(2, 4), we implicitly consider the lines as sitting in
the projective 3-space P3 and intersections are taken in the projective sense. Indeed,
two parallel lines do not intersect in the affine 3-space, but their corresponding 2-
planes do. This corresponds to the fact that two parallel lines intersect at infinity
in P3 .
Let us recall the crucial ideas of Schubert’s solution to Problem 1. To solve it, he
developed the calculus of “conditions” (see [22]), which has since become known
as Schubert Calculus. Examples of conditions are:
1. for a given point a, denote by σa the condition that a line contains a;
2. for a given line , denote by σ the condition that a line intersects ; or
3. for a plane , denote by σ the condition that a line is contained in .
Schubert’s brilliant idea was that conditions can be added and multiplied and this
corresponds to logical “or” and “and” operations on the conditions, e.g., σ1 + σ2 is
the condition that a line intersects line 1 or line 2 while σ1 · σ2 is the condition that
a line intersects both lines 1 and 2 . So, for instance, Problem 1 can be reformulated
to: What is σ1 · · · σ4 where i are four lines in general position? In particular, we
can reformulate the problem in an algebraic equation and obtain

σ1 · σ2 · σ3 · σ4 = (σ1 · σ2 ) · (σ3 · σ4 ) =?.

So we have to understand the conditions σ1 · σ2 and σ3 · σ4 . By using some
heuristics, Schubert came to the conclusion that “perturbations” of the condi-
tion σ1 · · · σ4 do not change the answer (the conservation of number principle),
i.e., we are allowed to move the lines i . In particular, we may assume that 1 , 2
lie on a plane, and so do 3 and 4 (see Fig. 9.1). From that it straightforwardly fol-
lows that a line intersects both 1 and 2 if and only if it is either contained in the
9 The Ring of Conditions for Horospherical Homogeneous Spaces 199

Fig. 9.1 The conservation of number principle implies that we are allowed to move the lines i

plane  spanned by 1 and 2 (recall that we take intersections in P3 ) or it contains

the intersection point of 1 and 2 . Using Schubert calculus this means

σ1 · σ2 = σa + σ and σ3 · σ4 = σa  + σ

where a is the intersection point of 1 and 2 and  is the plane spanned by 1 and 2
and similarly for 3 , 4 .
Thus, we get

σ1 · σ2 · σ3 · σ4 = (σa + σ ) · (σa  + σ ) = σa · σa  + σa · σ + σa  · σ + σ · σ .

Clearly there is exactly one line passing through both a and a  and there is exactly
one line contained in both  and  . On the other hand, as a is not contained in  ,
the condition σa · σ is not satisfied by any line, and similarly for σa  · σ .
We obtain

σ1 · σ2 · σ3 · σ4 = σa · σa  + σa · σ + σa  · σ + σ · σ = 2.

       
=1 =0 =0 =1

Of course, we haven’t given a precise explanation yet and in his fifteenth problem
Hilbert asked for a rigorous foundation of Schubert Calculus. Our goal will be to
understand De Concini’s and Procesi’s solution to Hilbert’s problem. For that, we
also need to understand spherical geometry, a topic which is exciting in its own right.
200 J. Hofscheier

9.2 Linear Algebraic Groups: A Crash Course

The classical books by Borel, Humphreys and Springer [2, 10, 25] are excellent refer-
ences for what follows. A more modern and accessible book is [18]. For convenience,
we work over the field of complex numbers C.
An algebraic variety G is called an algebraic group if G is a group and the
maps G × G → G, (g, h) → gh and G → G, g → g −1 are morphisms of algebraic
varieties. The Lie algebra of G, usually denoted by g, is the tangent space Te G at
the identity element e ∈ G equipped with a binary operation [·, ·] called the Lie
bracket. Important examples of algebraic groups are GLn (=the general linear
group of invertible n × n-matrices with complex entries), SLn (=the special linear
group of n × n-matrices with complex entries and determinant 1), abelian varieties
(=complete connected algebraic groups) or elliptic curves (=1-dimensional abelian
varieties). We will work with linear algebraic groups, i.e., Zariski closed subgroups
of GLn . If G ⊆ GLn is a linear algebraic group, then Te G = g ⊆ gln = Te GLn =
{(n × n) − matrices}, and the Lie bracket can be defined as the commutator of matri-
ces
[A, B] := AB − B A.

Remark 2 If one replaces “algebraic varieties” and “morphisms of algebraic vari-

eties” by “smooth manifolds” and “smooth maps”, one obtains the definition of a
Lie group.

Exercise 3 Let G be an algebraic group.

1. Show that only one irreducible component of G can pass through e. This com-
ponent is called the identity component of G, usually denoted by G ◦ .
2. Show that G ◦ is a normal subgroup of finite index in G, whose cosets are the
connected as well as irreducible components of G.

Exercise 4 Which of the following algebraic groups are linear?

1. (Cn , +),
2. An elliptic curve,
3. PGLn .

From now on all algebraic groups are assumed to be linear, unless stated otherwise.

Definition 5 An element g ∈ G ⊆ GLn is called semisimple if the matrix g is diag-

onalizable. It is called unipotent if all eigenvalues of g are equal to 1. (This definition
is independent of the choice of the embedding G ⊆ GLn .)

Exercise 6 Let π : G → GLn be a (rational) representation of an algebraic group G,

i.e., π is a morphism of algebraic groups. Show that:
1. If G = (C∗ )n , then any matrix in π(G) is diagonalizable.
2. If G = Cn , then any matrix in π(G) is unipotent.
9 The Ring of Conditions for Horospherical Homogeneous Spaces 201

Exercise 7 (Jordan decomposition) Show that every element g ∈ G has a unique

decomposition g = gs gu , where gs ∈ G is semisimple, gu ∈ G is unipotent, and gs
and gu commute.
The radical, denoted by R(G), of an algebraic group G is the identity component
of its maximal normal solvable subgroup. The unipotent radical, denoted by Ru (G),
is the set of unipotent elements in the radical of G.
Definition 8 An algebraic group G is reductive if Ru (G) = {e}. It is semisimple
if R(G) = {e}.
Theorem 9 (Characterization of reductive groups) Let G be an algebraic group.
The following conditions are equivalent:
1. G is reductive;
2. R(G) is a torus;
3. G ◦ = T · S, where T is a torus and S is a connected semisimple subgroup;
4. any finite-dimensional rational representation of G is completely reducible
(recall: a rational representation of G in a vector space V is a homomor-
phism G → GL(V ) of algebraic groups);
5. G admits a faithful finite-dimensional completely reducible rational representa-
tion;
6. the Lie algebra of G admits a direct sum decomposition g = h ⊕ ih where h is
the Lie algebra of a maximal compact real Lie subgroup of G.
Exercise 10 Which of the following groups are reductive?
1. Cn ,
2. GLn ,
3. An elliptic curve.
Exercise 11 Show that an algebraic group G is reductive if and only if G does not
contain a normal subgroup isomorphic to Cn .
A character of an algebraic group G is a homomorphism of algebraic groups
χ : G → C∗ and the set of all characters gives the character group of G, i.e., X(G) :=
{χ : G → C∗ character}.
An algebraic torus is an algebraic group T that is isomorphic to

(C∗ )n = {(z 1 , . . . , z n ) ∈ Cn : z i = 0}.

If G is an algebraic group, then a maximal element of the set

{H ⊆ G closed subgroup, H an algebraic torus}

(which is ordered by inclusion) is called a maximal torus of G.

Theorem 12 In an algebraic group, any two maximal tori are conjugated.
The dimension of T is called the rank of G. The character lattice X(T ) of T is also
called the weight lattice of G, and its elements are called weights of G.
202 J. Hofscheier

Exercise 13 Find a maximal torus of the following groups:

1. GLn ,
2. SLn ,
3. SOn .
The set {H ⊆ G closed connected solvable subgroup} is partially ordered by
inclusion. A maximal element of this set is called a Borel subgroup.
Exercise 14 Show that the upper-triangular invertible matrices form a Borel sub-
group in GLn and that any two Borel subgroups are conjugated. (Hint: Use the
Lie–Kolchin theorem [10, Theorem 17.6].)
Theorem 15 In an algebraic group, any two Borel subgroups are conjugated.
Definition 16 A (Zariski) closed subgroup P ⊆ G is called parabolic if P contains
a Borel subgroup of G.
Exercise 17 Let G be a linear algebraic group, B ⊆ G a Borel subgroup and T ⊆ G
a maximal torus.
1. Show that up to conjugation T ⊆ B.
2. Show that restricting characters from B to T induces an isomorphism of character
lattices X(B) ∼
= X(T ).
The Weyl group W of G is defined as NG (T )/C G (T ), where NG (T ) and C G (T )
denote the normalizer and centralizer, respectively, of a maximal torus T ⊆ G. The
Weyl group acts on T by conjugation: if w = nC G (T ) for n ∈ NG (T ), then w(t) :=
(ntn −1 ) for t ∈ T .
Theorem 18 If G is a connected reductive group, then C G (T ) = T for any maximal
torus T ⊆ G. In particular, the Weyl group is given by W = NG (T )/T .
Exercise 19 If G is a connected reductive group, T ⊆ G a maximal torus, and
B ⊆ G a Borel subgroup with T ⊆ B, show that for any w ∈ W the double
coset B ẇ B is independent of the choice of a representative ẇ ∈ NG (T ). Thus, by
abuse of notation, we will denote this double coset by Bw B.
Theorem 20 (Bruhat decomposition) If G is a connected reductive group, T ⊆ G
a maximal torus, and B ⊆ G a Borel subgroup with T ⊆ B, then there is a disjoint
union, i.e., Bw B = Bw B if and only if w = w in W ,

G= Bw B
w∈W

In particular,

G/B = Bw B/B.
w∈W
9 The Ring of Conditions for Horospherical Homogeneous Spaces 203

Exercise 21 You may want to start with n = 3 or n = 4 in the following problems.

1. Explicitly compute the Bruhat decomposition of GLn (take T to be the maximal
torus of diagonal matrices and B the Borel subgroup of upper triangular matrices).
2. Classify all parabolic subgroups in GLn (up to conjugation). (Hint: There is a
relationship between parabolic subgroups in GLn and flags in Cn . Recall that a flag
is an increasing sequence of subspaces of Cn , i.e., {0} = V0  V1  . . .  Vk =
Cn . The dimensions di := dim Vi yield an increasing sequence of integers 0 =
d0 < d1 < · · · < dk = n, called the signature of the flag.)

Exercise 22 Let G be a connected reductive group, T ⊆ G a maximal torus, B ⊆ G

a Borel subgroup with T ⊆ B and P ⊆ G a parabolic subgroup with B ⊆ P. Show
that

G/P = Bw P/P
w∈W/W P

where W P = N P (T )/T is the Weyl group of P.

The closure of B-orbits in G/P are the Schubert varieties (denoted by X (w)).
They play an important role in the study of G/P. The dimension of X (w) equals
the length l(w) of w (i.e., the minimal number of factors needed to write w as a
product of simple reflections). In particular, there exists a unique element w0 of
maximal length in W/W P .

Example 23 Let G = SLn and T be the maximal torus of diagonal matrices con-
tained in the Borel subgroup B of upper-triangular matrices. The Lie algebra g = sln
(i.e., the tangent space Te SLn equipped with the Lie bracket [·, ·]) is the set of
traceless matrices in Mat(n × n) equipped with the commutator bracket [A, B] =
AB − B A. Furthermore, the Lie algebra t of T coincides with the subspace of
diagonal matrices in sln . Observe that the Lie bracket induces a map ad : t →
End(g); A → [A, ·] which is a representation of Lie algebras, i.e., ad([A, B]) =
ad(A) ad(B) − ad(B) ad(A) for any A, B ∈ t (check this!). Let ε1 , . . . , εn be the
linear forms in t∗ induced by the diagonal entries, i.e., εi (diag(t1 , . . . , tn )) = ti and
set εi j = εi − ε j . It is straightforward to show that the Lie algebra decomposes into
eigenspaces as follows

and R = {εi j : 1 ≤ i, j ≤ n, i = j}. If b is the Lie algebra of B, then


b=t⊕ gα
α∈R +
204 J. Hofscheier

Fig. 9.2 Illustration of the

root system A2 . We identify
the hyperplane {x + y + z =
0} ⊆ R3 with R2 via the 2
basis obtained by applying
the Gram–Schmidt algorithm
to the basis (α1 , α2 ) 2
where R3 and R2 are
equipped with the usual 1
euclidian scalar products
1

where R + = {εi j : i < j} and this set is called the set of positive roots. The set
of simple roots S = {αi := εi,i+1 : i = 1, . . . , n − 1} (cf. Fig. 9.2) forms a basis of t∗
(check this!) and induces an isomorphism t∗ ∼ = {(x1 , . . . , xn ) ∈ Rn : x1 + . . . + xn =
0} ⊆ R . To any simple root αi one associates a reflection si , namely the linear
n

transformation on Rn which swaps the coordinates with index i and (i + 1). We

identify si with an element in W = NG (T )/T :
⎛ ⎞
Ii−1
⎜ 0 1 ⎟
⎜ ⎟ T ∈ W.
⎝ −1 0 ⎠
In−i−1

Then W is generated by the si , i.e., W = si : i = 1, . . . , n − 1 (check this!). It

straightforwardly follows that W is isomorphic to the group Sn of permutations on
the coordinates of Rn via si → (i, i + 1) (transposition swapping i with i + 1).

In general, the Lie bracket induces a natural representation ad : t → End(g); x →

[x, ·]. There is a set of linear forms, called roots, R ⊆ t∗ such that

g=t⊕ gα
α∈R

where gα denotes the linear subspace of eigenvectors of weight α, i.e., the set of
vectors x ∈ g such that [h, x] = α(h)x for all h ∈ t. The Lie algebra b of B can be
written as 
b=t⊕ gα
α∈R+
9 The Ring of Conditions for Horospherical Homogeneous Spaces 205

for some subset R + ⊆ R, called the set of positive roots. There exists a unique basis S
contained in R + such that all positive roots are linear combinations of elements in S
with nonnegative integer coefficients. The elements of S are called simple roots.
The following fundamental theorem on parabolic subgroups can be found in any
introductory text on algebraic groups. Recall the definition of the Weyl group: W =
NG (T )/T . Let R be the set of roots and let S be the set of simple roots induced by
the choice of the Borel subgroup B.

Theorem 24 The assignment I → PI = w∈W I Bw B induces a bijection between
subsets of the set of simple roots S and parabolic subgroups of G which contain B
(here W I denotes the group generated by the simple reflections sα for α ∈ I ).

9.3 Spherical Varieties

Recall the definition of a toric variety:

Definition 25 Let T be an algebraic torus. A normal irreducible T -variety is
called toric variety if it contains an open dense T -orbit.
Spherical varieties can be thought of as a generalization of toric varieties where
one allows also non-abelian group actions. Unfortunately, the straightforward gen-
eralization does not work:
Definition 26 (Incorrect definition) Let G be a connected linear algebraic group. A
normal irreducible G-variety is called spherical if it contains an open dense G-orbit.
Exercise 27 Show that the “incorrect definition” of spherical varieties does not
imply finiteness of the number of orbits, a property one would expect from a gener-
alization of toric varieties. (Hint: Consider the action of GLn on the space of (n × n)-
matrices by left multiplication. Show that the GLn -orbits are classified by matrices
in reduced row echelon form. If n ≥ 2, deduce that, although there is an open GLn
orbit, the number of GLn -orbits is infinite.)
So the definition of spherical varieties is more subtle: Let G be a connected
reductive complex linear algebraic group (this assumption has several implications
which make this choice important: finite generation properties, good representation
theory, cf. Theorem 9).
Definition 28 A normal irreducible G-variety is said to be spherical if it contains
an open orbit under the action of a Borel subgroup of G. (In particular, it contains
an open G-orbit.)
Example 29 Examples of spherical varieties are toric varieties (a Borel subgroup
of (C∗ )n is (C∗ )n itself).
Another point of view on spherical varieties, important to the theory, is as fol-
lows: First consider the open G-orbit which is a homogeneous space G/H for some
subgroup H of G. Then consider the embedding of G/H in X . So we make the
following definitions:
206 J. Hofscheier

Definition 30 A closed subgroup H ⊆ G is called spherical if G/H has a dense

open orbit for a Borel subgroup of G. In this case, G/H is called a spherical homo-
geneous space.
Recall that in Exercise 27, we have seen that an open G-orbit does not guarantee
the finiteness of G-orbits, a property one would expect from a generalization of
toric varieties. It is interesting that one can use this property as a characterization of
spherical homogeneous spaces:
Theorem 31 ([1]) A homogeneous G-space O is spherical if and only if any G-
variety X with an open orbit equivariantly isomorphic to O has finitely many G-
orbits.
A morphism ϕ : X → Y of G-varieties is called equivariant if ϕ(g · x) = g · ϕ(x)
for any g ∈ G and all x ∈ X .
Definition 32 Suppose G/H is a spherical homogeneous space. An equivariant
open embedding of G/H into a normal irreducible G-variety X is called a spherical
embedding, and X is called a spherical variety.
In particular, the description of spherical varieties splits into two parts:
1. Classify all spherical homogeneous spaces G/H .
2. For a fixed spherical homogeneous space, classify all G-equivariant open embed-
dings G/H
→ X into normal irreducible G-varieties.
Historically, the second problem has been answered first by the work of Luna and
Vust [17]. Only recently, the first problem has been answered by work of several
researchers (see [3, 6, 15, 16]).
Exercise 33 Show the following statements:
1. A closed subgroup H ⊆ SL2 is spherical if and only if dim H > 0.
2. Table 9.1 shows a list of all spherical subgroups of SL2 (up to conjugation).
(Although spherical varieties with an SL2 -action seem to be special, they actually
play a crucial role in the development of spherical varieties. See, for example, [17]
or [8, 14].) Hints: If this is too hard, then verify explicitly that the subgroups given
in Table 9.1 are spherical:
a. SL2 /B = ∼ P1 where SL2 naturally acts on P1 ,
b. it is enough to show that U1 is spherical (why?) and to do that consider the
natural action of SL2 on A2 ,
c. consider the natural action of SL2 on P1 × P1 to show that T is a spherical
subgroup of SL2 ,
d. consider the natural action of SL2 on the symmetric (2 × 2)-matrices to show
that N is a spherical subgroup of SL2 .

Exercise 34 A closed subgroup H ⊆ G is called horospherical if it contains a max-

imal unipotent subgroup U of G. Show that horospherical subgroups are spherical.
In particular, flag varieties are spherical. (Hint: Use the Bruhat decomposition.)
9 The Ring of Conditions for Horospherical Homogeneous Spaces 207

Table 9.1 Classification of spherical subgroups of SL2 (up to conjugation)

H Details
SL2
B Borel subgroup
  
ξ ∗
Uk = : ξ ∈ μk k ∈ N, μk group of kth roots of unity
0 ξ −1
T Maximal torus
N Normalizer of a maximal torus

9.3.1 The Luna–Vust Theory of Spherical Embeddings

Recall that for a fixed algebraic torus T , all toric embeddings T

→ X into a normal
irreducible T -variety can be combinatorially described by polyhedral fans in the
vector space Hom(X(T ), Q). A similar description exists for spherical embeddings
which we now explain. This is called the Luna–Vust theory of spherical embeddings.
Good references for this theory are [13, 17]:
Let G be a connected reductive complex algebraic group and fix a spherical
subgroup H ⊆ G. Let B be a Borel subgroup of G and T a maximal torus of G
contained in B. We now explain how all spherical embeddings G/H
→ X can be
described combinatorially.

Definition 35 The combinatorial objects needed in the Luna–Vust theory are listed
in Table 9.2. The rank of M is also called the rank of the spherical homogeneous
space G/H , i.e., rk(G/H ) = rk(M). Let N := Hom(M, Z) be the dual lattice
of M and note that we have a dual pairing ·, · : N × M → Z. Furthermore,
set MQ := M ⊗ Q and NQ = Hom(M, Q). Recall, that in our context a valuation
is a map ν : C(G/H )∗ = C(G/H ) \ {0} → Q which satisfies:
1. ν( f 1 + f 2 ) ≥ min{ν( f 1 ), ν( f 2 )} whenever f 1 , f 2 , f 1 + f 2 ∈ C(G/H )∗ ;
2. ν( f 1 f 2 ) = ν( f 1 ) + ν( f 2 ) for all f 1 , f 2 ∈ C(G/H )∗ ; and
3. ν(C∗ ) = 0.
A valuation ν is called G-invariant if ν(g · f ) = ν( f ) for all g ∈ G and f ∈
C(G/H )∗ .

As the set of B-semi-invariant rational functions on G/H will appear frequently

below, we introduce the notation C(G/H )(B) for it.

Lemma 36 As G/H has an open B-orbit, a B-semi-invariant rational function f

is determined by its weight χ f up to a scalar multiple. Said in other words: For
any χ ∈ M, there is f χ ∈ C(G/H )(B) (unique up to a scalar multiple) such that b ·
f χ = χ (b) f χ .
208 J. Hofscheier

Table 9.2 The Luna–Vust data

Object Definition
Weight lattice M = {χ ∈ X(B) :
∃ f ∈ C(G/H )∗ , B-semi-invariant
with b · f = χ(b) f for b ∈ B}
Colors D = {B-invariant prime divisors in G/H }
Valuation cone V = {ν : C(G/H )∗ → Q, G-invariant
valuation}

Table 9.3 The “Luna–Vust data” for the toric case

Object Toric case
Weight lattice M = X(T )
Colors D=∅
Valuation cone V = NQ

The following interpretation of a valuation ν : C(G/H ) → Q (invariant or not)

will be useful:

ρ : {ν : C(G/H ) → Q valuation} → NQ ; ν → [χ → ν( f χ )].

Theorem 37 ([4]) The map ρ|V : V

→ NQ is injective and its image is a polyhedral
cone whose dual cone is simplicial and not necessarily full-dimensional.

Any color D ∈ D induces a valuation ν D and by abuse of notation, we will

write ρ(D) := ρ(ν D ). In general, the map ρ|D : D → NQ is not that well-behaved
(see Exercise 38).

Exercise 38 Find the “Luna–Vust data” for the spherical homogeneous spaces from
Exercise 33. In particular, you should see the following phenomena:
1. SL2 /T : the map ρ : D → N might not be injective;
2. SL2 /N : the image of a color ρ(D) might not be primitive in N;
3. SL2 /B: the image of a color might even be zero, i.e., ρ(D) = 0.

Example 39 The “Luna–Vust data” of the toric case is listed in Table 9.3.

Example 40 Consider the natural action of SL2 on C2 . Let B be the Borel subgroup
of upper triangular matrices, T the maximal torus of diagonal matrices and U the
unipotent radical of B. Denote the standard basis of C2 by e1 , e2 . Then SL2 /U ∼
=
SL2 ·e1 = C2 \ {0} and B · e2 = C × C∗ is the open B-orbit. The rational functions
on SL2 /U are given by C(x, y) = C(A2 ). It is straightforward to verify that

C(SL2 /U )(B) ∼
= {cy k : c ∈ C, k ∈ Z}.
9 The Ring of Conditions for Horospherical Homogeneous Spaces 209

Fig. 9.3 Illustrating the “Luna–Vust data” of SL2 /U

Hence, M = Zω where ω is the fundamental weight of SL2 induced by the diagonal

elements of T . Clearly D = div(y) is the only B-stable prime divisor in SL2 /U ,
i.e., we only have one color D = {D} and ρ(D) = α̌|M where α̌ denotes the coroot
associated to the simple root α of SL2 (with respect to our choice of Borel). If we
consider the blowup of A2 at the origin, we obtain an exceptional SL2 -invariant
divisor E which induces an SL2 -invariant valuation ν E such that ρ(ν E ) = α̌|M . In
particular, the ray Q≥0 α̌ is contained in the valuation cone V. If we consider the
spherical embedding SL2 /U
→ P2 , we see that the line at infinity induces a G-
invariant valuation ν with ρ(ν) = −α̌|M , and thus V = NQ .
One usually illustrates the combinatorial data in a picture (see Fig. 9.3).

Definition 41 We introduce the following “colored” extensions of the notions of

polyhedral cone, face and fan from the toric case.
1. A colored cone is a pair (C, F ) with
a. F ⊆ D,
b. C ⊆ NQ convex cone generated by ρ(F ) and finitely many elements of V ∩
N,
c. the relative interior of C intersects V,
d. C contains no lines and 0 ∈ / ρ(F ).
Such colored cones are usually called strictly convex colored cones, but as we are
only interested in strictly convex cones, we will omit the specifier and just speak
of colored cones.
2. A colored face of a colored cone (C, F ) is a pair (C , F  ) such that
a. C is a face of C (in the usual sense),
b. the relative interior of C intersects V,
c. F  = {D ∈ F : ρ(D) ∈ C }.
3. A colored fan is a finite set  of colored cones with the following properties:
a. every colored face of a colored cone of  is in ,
b. for all ν ∈ V, there exists at most one colored cone (C, F ) ∈  such that ν
is in the relative interior of C.

4. The support of a color fan  is the set (C,F ) (C ∩ V) ⊆ V where (C, F ) runs
through all elements in .

Let us explain how to associate a colored fan  X to a spherical embedding

G/H
→ X .
210 J. Hofscheier

Fig. 9.4 The colored fans of SL2 /U

→ A2 and SL2 /U
→ P2

Theorem 42 X is covered by finitely many G-invariant open subvarieties of X

containing a unique closed G-orbit (such varieties are called simple embeddings).
Let X  ⊆ X be an open G-invariant subvariety which is a simple embedding
and denote the G-invariant divisors of X  by X 1 , . . . , X m  . Let F  be the set of
colors D ∈ D whose closure in X contain the closed orbit of X  . We define C to be the
cone in NQ generated by ρ(D) for D ∈ F  and ρ(X i ) := ρ(ν X i ) for i = 1, . . . , m  .
Then (C , F  ) is a colored cone in NQ . Moreover the set of colored cones constructed
this way (together with their colored faces) forms a colored fan, which we denote
by  X .
Theorem 43 (Luna–Vust) The map X →  X is a bijection from the isomorphism
classes of spherical G/H -embeddings and the set of colored fans.
Example 44 (Example 40 continued) Clearly SL2 /U
→ A2 is a simple spherical
embedding (the origin is the only closed SL2 -orbit). On the other hand the spheri-
cal embedding SL2 /U
→ P2 is not simple (indeed we can cover it with an affine
chart A2 and the complement of the unique SL2 fixed point). The corresponding col-
ored fans are illustrated in Fig. 9.4 (understand how to get them and which colored
fan corresponds to which spherical embedding). Note that the circle means that the
cone Q≥0 is “colored” by the unique color of SL2 /U , i.e., (Q≥0 , {D}).

Exercise 45 Use the Luna–Vust theory to classify all spherical embeddings of

SL2 /T and SL2 /N . Draw the corresponding colored fans. Hint: You should find 2
in both cases.
Many results about spherical varieties are known. Unfortunately, due to lack of
time, we won’t be able to dig any deeper.
Theorem 46 A list of selected results:
1. The number of B-orbits is finite;
2. X is complete if and only if any ν ∈ V is contained in some colored cone of  X ;
3. there is a bijective correspondence between G-orbits in X and colored cones
in  X ;
4. a combinatorial smoothness criterion;
5. combinatorial descriptions of the Picard group and the divisor class group;
6. ampleness criterion for divisors;
and many more . . .
To learn more about the features of spherical varieties, the interested reader is
encouraged to consult [26] for further reading.
9 The Ring of Conditions for Horospherical Homogeneous Spaces 211

9.3.2 The Classification of Spherical Homogeneous Spaces

The classification of spherical homogeneous spaces G/H turns out to be quite hard.
Luna’s brilliant insight in spherical varieties made it possible to find such a description
from Wasserman’s list of certain spherical varieties of rank 2 [28]. Inspired by it,
Luna [16] formulated a conjectural description and proved it for spherical varieties
of type A. Only recently Luna’s Conjecture was proven in general with the combined
efforts of several researchers [3, 6, 15]. Unfortunately, time does not permit to give
more details on this exciting topic, instead we will see a complete answer for an
interesting special case.

9.3.3 The Complete Picture in the Horospherical Case

Recall from Exercise 34 that a closed subgroup H ⊆ G is called horospherical

if it contains a maximal unipotent subgroup of G. An exceedingly well written
introduction to horospherical varieties can be found in [19, 20] by Pasquier.
Fix a maximal unipotent subgroup U ⊆ G, a Borel subgroup U ⊆ B of G and a
maximal torus T ⊆ B.
Let us list some fundamental properties of horospherical subgroups. We refer
to [19] for further details and references.

Proposition 47 For a horospherical subgroup H ⊆ G with U ⊆ H , the following

statements hold:
1. the normalizer P := NG (H ) is a parabolic subgroup containing B. Let I ⊆ S
be the unique set of simple roots such that P = PI (see Theorem 24);
2. M =  {χ ∈ X(P) : χ | H = 1} ⊆ {χ ∈ X(T ) : α̌, χ  = 0 for all α ∈ I };
3. H = χ∈M ker(χ );
4. D = {Dα := Bw0 sα P/H : α ∈ S \ I } where w0 ∈ W is the longest element in
the Weyl group W = NG (T )/T and sα denotes the simple reflection associated
to the simple root α;
5. P−w0 (I ) coincides with the stabilizer of the open B-orbit in G/H ;
6. V = NQ .

Theorem 48 ([13, Theorem 6.1]) If H ⊆ G is a spherical subgroup, then NG (H )/H

is diagonalizable. In particular, if H contains U , then P/H is a torus where
P := NG (H ).

Now, we come to an important geometric characterization of horospherical homo-

geneous spaces.
Recall that a continuous surjective map p : E → X of topological spaces is called
a fiber bundle with fiber F (another topological space) if X can be covered with open
subsets U such that there are homeomorphisms ϕ : p −1 (U ) → U × F in such a way
that p agrees with the projection onto the first factor (see Fig. 9.5). It is said to be
212 J. Hofscheier

Fig. 9.5 Local trivialization

of fiber bundles

Fig. 9.6 Quotient by free

torus action is locally trivial
in Zariski topology

a principal -bundle, for  a topological group, if in addition E is equipped with a

continuous -action  × E → E preserving the fibers of p, i.e., if y ∈ p −1 (x) for
some x ∈ X , then γ · y ∈ p −1 (x) for any γ ∈ , and acts freely and transitively on
them.
Here is a crucial observation on principal torus bundles in algebraic geometry.

Lemma 49 Let the torus S act freely on the normal irreducible variety E with good
geometric quotient p : E → E/S. Then for each y ∈ E/S there exists an affine open
neighbourhood U ⊆ E/S of x such that the diagram in Fig. 9.6 commutes and the
upper left isomorphism is S-equivariant.

Exercise 50 Show that if an algebraic torus S acts freely on a normal irreducible

variety E with good geometric quotient p : E → E/S, then p admits Zariski open
trivializations (i.e., prove Lemma 49). What if we replace S by a connected reduc-
tive G? Hint: For the second part of the question, you may want to consider the
morphism φ : X → Y ; (A, B) → (det(A), tr(AB), det(B)) where X = {(A, B) ∈
Mat(2 × 2, C)2 : det(AB − B A) = 0, tr(A) = tr(B) = 0} and Y = {y ∈ C3 :
4y1 y3 − y22 = 0}.

Let p : E → X be a -principal bundle. Suppose that E and X are G-spaces for

another topological group G. Then p : E → X is an equivariant principal -bundle
if p is equivariant (i.e., p(g · y) = g · p(y) for any g ∈ G and y ∈ E) and the two
actions by  and G commute (usually one assumes that  acts from the right while G
acts from the left).

Proposition 51 If H ⊆ G is a closed subgroup, then the following statements are

equivalent:
1. H is horospherical, i.e., contains the unipotent radical of a Borel subgroup;
2. G/H is an (algebraic) equivariant principal torus bundle over a flag variety G/P
where G naturally acts on G/H resp. G/P by left translations (the dimension of
the torus
 fiber coincides with the rank of G/H );
3. H = χ∈M ker χ for some parabolic subgroup P of G and some sublattice M ⊆
X(P).
9 The Ring of Conditions for Horospherical Homogeneous Spaces 213

Furthermore, P = NG (H ) and P = T H = B H for all maximal tori T of B con-

tained in P and all Borel subgroups B of G contained in P.
Proof (1) ⇒ (2): By Theorem 48, S := P/H = NG (H )/H is an algebraic torus. It
acts on G/H by right-translations, i.e.,

S × G/H → G/H ; ( p H, x H ) → x p −1 H .

It is straightforward to check that S acts freely on G/H , and thus the result follows
by Lemma 49.
(2) ⇒ (1): Suppose that the fibers of the torus bundle p : G/H → G/P are
isomorphic to the algebraic torus T . As the two actions by G and T commute, the
morphisms ϕt : G/H → G/H ; x H → t · x H for t ∈ T are G-equivariant automor-
phisms. It follows by [26, Proposition 1.8] that we may consider T as a subgroup
of NG (H )/H . Set N := NG (H ). Let T  be the preimage of T under the natural pro-
jection map N → N /H . Note that N → N /H is a morphism of algebraic groups
and that T  is a closed subgroup of G. Since p −1 (x P) ∼
= T for any x P ∈ G/P, it fol-
lows that a conjugate of P is contained in T, and thus it contains a maximal unipotent
subgroup U . As the natural projection morphism of algebraic groups T  → T maps
unipotent elements on unipotent elements, it follows that U is in its kernel which
implies that U ⊆ H .
(1) ⇔ (3) straightforwardly follows from Proposition 47 (3). 
Exercise 52 Show that SL2 /Uk is indeed a torus bundle over SL2 /B.
Exercise 53 Use the Luna–Vust theory to classify all spherical embeddings of
SL2 /U where U ⊆ SL2 is a maximal unipotent subgroup. Draw the correspond-
ing colored fans. Hint: You should find 6.
Proposition 54 ([20, Proposition 1.6]) The assignment which associates to a
horospherical subgroup H ⊆ G the pair (M, I ) (see Proposition 47) induces a
bijection between horospherical subgroups of G and pairs (M, J ) where J ⊆ S
and M ⊆ X(T ) is a sublattice such that α̌, χ  = 0 for any α ∈ J and all χ ∈ M.
The horospherical
 subgroup associated to a pair (M, J ) as in Proposition 54 is
given by H = χ∈M ker χ where M ⊆ X(PJ ).
Exercise 55 Use the combinatorial description of horospherical subgroups to clas-
sify those contained in SL2 .
A colored fan  is called toroidal if F = ∅ for any (C, F ) ∈ . Observe that
in the horospherical case toroidal fans coincide with fans in the toric sense. In this
special case, we have the following explicit construction of horospherical toroidal
varieties:
Proposition 56 ([20, Examples 1.13 (2)]) If H ⊆ G is a horospherical subgroup
containing U and  is a toroidal fan, then the corresponding spherical embed-
ding is G-equivariantly isomorphic to G × P X  where X  denotes the toric variety
corresponding to the fan  (with acting torus P/H where P = NG (H )).
214 J. Hofscheier

In the situation of Proposition 56, recall that P acts on G × X  by p · (g, x) :=

(gp −1 , p H · x) with good geometric quotient G × P X  = (G × X  ) /P.

9.4 The Ring of Conditions of a Horospherical Variety

A good reference for the ring of conditions is the classical paper by De Concini and
Procesi [7].
Let G be a connected complex algebraic group and H ⊆ G a closed subgroup
(not necessarily spherical). Consider the homogeneous space G/H .
Recall that two subvarieties X, Y ⊆ G/H are said to intersect properly if either
X ∩ Y = ∅ or each irreducible component of the intersection X ∩ Y has dimen-
sion dim(X ) + dim(Y ) − dim(G/H ). They are said to intersect transversally if
the intersection X ∩ Y is smooth and has pure dimension dim(X ) + dim(Y ) −
dim(G/H ).

Theorem 57 (Kleiman’s transversality theorem [12, Corollary 4]) Let X, Y ⊆ G/H

be two irreducible subvarieties. The left translate of X by g ∈ G we denote by g X .
1. There exists a dense open subset U ⊆ G such that g X and Y intersect properly
for each g ∈ U .
2. If X, Y are smooth, then there exists a dense open subset U ⊆ G such that g X
and Y intersect transversally for any g ∈ U .
In particular, if X and Y have complementary dimensions (but are not necessarily
smooth), the intersection g X ∩ Y consists of finitely many points and this number is
constant for generic g ∈ G.

Remark 58 There is a slight strengthening of Kleiman’s transversality theorem

in [7, Sect. 6.1].

Recall that the free abelian group Zk (G/H ) = ZX , where the sum is over
X ⊆G/H
closed irreducible subvarieties of codimension k, is said to be the group of algebraic
cycles of codimension k. Theorem 57 makes it possible to introduce an intersection
pairing between groups of algebraic cycles of complementary codimensions

Zk (G/H ) and Zdim(G/H )−k (G/H ).

It is enough to define it for irreducible cycles and then extend bilinearly:

Zk (G/H ) × Zdim(G/H )−k (G/H ) → Z;

(X, Y ) → (X · Y ) := #(g X ∩ Y ) (for generic g ∈ G).

Here X, Y ⊆ G/H are assumed to be irreducible subvarieties.

9 The Ring of Conditions for Horospherical Homogeneous Spaces 215

Definition 59 Two algebraic cycles X, Y ∈ Zk (G/H ) are said to be equivalent,

i.e., X ∼ Y , if for any algebraic cycle of complementary codimension
Z ∈ Zdim(G/H )−k (G/H ) the intersection products are the same (X · Z ) = (Y · Z ).
We denote the group of equivalence classes by C k (G/H ) := Zk (G/H )/ ∼ and
consider it as the “group of conditions of dimension dim(G/H ) − k”.

Clearly the intersection pairing factors through the equivalence relation, so

that we obtain an intersection pairing on the groups of conditions: C k (G/H ) ×
dim(G/H ) k
C dim(G/H )−k (G/H ) → Z. So far C ∗ (G/H ) := k=0 C (G/H ) is only a group,
but we want to introduce a product on it, so that it becomes a ring. Again, it is
enough to define a product structure for classes of irreducible cycles X ∈ Zk (G/H )
and Y ∈ Zl (G/H ) and then extend bilinearly. Here is a naive approach:
Definition 60 Define the intersection product of [X ] and [Y ] where X, Y ⊆ G/H
are two irreducible subvarieties by [X ] · [Y ] := [g X ∩ Y ] for generic g ∈ G.
Unfortunately this definition of an intersection product may not be well-defined
in general (see Exercise 61).
Exercise 61 Show that the naive definition of an intersection product of two
irreducible subvarieties X, Y ⊆ G/H is not well-defined in general. (Hint: Con-
sider G = (C3 , +) acting on A3 by translations. Let H = {0} and compute the inter-
section product of X = {y = 0}, Y = {x = yz} ⊆ C3 .)

Proposition 62 ([7]) For a flag variety G/P, the intersection product in

Definition 60 is well-defined and the ring C ∗ (G/P) can be identified with the Chow
ring A∗ (G/P) or with the cohomology ring H ∗ (G/P, Z).

Led by this observation, De Concini and Procesi showed the remarkable fact that
the intersection product of Definition 60 is well-defined on spherical homogeneous
spaces. Let C be the set of smooth (or complete) spherical embeddings G/H
→
X . This set C admits the partial ordering defined such that a spherical embed-
ding G/H
→ X 1 is greater than G/H
→ X 2 if there exists an equivariant mor-
phism X 1 → X 2 . De Concini’s and Procesi’s idea is to show that for any X ∈ C
and any algebraic cycle Y ⊆ G/H , there is an X  ∈ C with X ≤ X  such that the
closure Y of Y in X  intersects the boundary of the open G-orbit in X  properly.
The existence of such a “good compactification” ensures that if one considers the
embedding X  then we may always assume (up to generic translations by G) that the
intersection with Y takes place in the open G-orbit G/H . To get an isomorphism of
rings, we have to consider “good compactifications” of all algebraic cycles at once.

Theorem 63 ([7, Sect. 6.3]) The intersection product from Definition 60 is well-
defined on a spherical homogeneous space G/H and there is a canonical isomor-
phism of graded rings

C ∗ (G/H ) = lim A∗ (X  ) = lim H ∗ (X  , Z)

−→ −→
X X
216 J. Hofscheier

where the limit is taken over complete (or equivalently smooth) spherical embed-
dings G/H
→ X  .

Remark 64 Any complete spherical embedding is dominated by a smooth projective

toroidal one, and thus they form a cofinal set.

Exercise 65 Explicitly compute the ring of conditions for some spherical homoge-
neous spaces SL2 /H where H ⊆ SL2 is a spherical subgroup. (Hint: In this case,
the rank of the spherical homogeneous space is bounded by 1, and thus there are
only finitely many spherical embeddings, so that we can straightforwardly compute
the direct limit of cohomology rings.)

Exercise 66 Use the ring of conditions of Gr(2, 4) to solve the “4-lines problem”.

9.4.1 The Horospherical Case

From now on let H ⊆ G be a horospherical subgroup containing the unipotent rad-

ical U of a Borel subgroup B. Set P := NG (H ) which is a parabolic subgroup
containing B.
Any character α ∈ X(P) induces an action of P on the affine line Cα by p · x =
α( p)x. We obtain an action of P on G × Cα by p · (g, x) = (gp −1 , α( p)x). The geo-
metric quotient by this action exists and is denoted by G × P Cα , i.e., G × P Cα =
(G × Cα )/P. For the equivalence classes in G × P Cα we write g  x. The projec-
tion map G × P Cα → G/P; g  x → g P yields an equivariant line bundle on G/P
where G acts on the left, i.e., g  · (g  x) = (g  g)  x. We write δ(α) := G × P Cα
and note that these bundles are usually called homogeneous fiber bundles (see [26,
Section 2.1]). If we compose the map X(P) → Pic(G/P) with the inclusion M ⊆
X(P), we obtain a map δ : M → Pic(G/P).
The following statement combinatorially describes the cohomology ring of
smooth projective toroidal horospherical varieties. It is a special case of a more
general result.
Theorem 67 ([21, Theorem 1.2]) Let X  be a smooth projective toroidal horospher-
ical variety defined by an (uncolored) fan  with rays ρ1 , . . . , ρn . Let v1 , . . . , vn ∈ N
be the primitive vectors along the rays ρi . Then the cohomology ring H ∗ (X  , Q) is
isomorphic as an H ∗ (G/P, Q)-algebra to the quotient of H ∗ (G/P, Q)[x1 , . . . , xn ]
by the sum of ideals
 
x j1 · · · x jk : ρ j1 , . . . , ρ jk do not span a cone of  +
 
n
c1 (δ(m)) − vi , mxi : m ∈ M ,
i=1

where c1 (δ(m)) ∈ H 2 (G/P, Z) denotes the first Chern class of the line bundle δ(m).
9 The Ring of Conditions for Horospherical Homogeneous Spaces 217

Note that the first ideal in the sum of ideals in Theorem 67 corresponds to the Stanley-
Reisner ideal of the toric variety. The challenge is to find a good description of the
ring in Theorem 67 as we want to take the direct limit over all smooth projective
toroidal fans .
The following approach is inspired by [5]. To keep notation simple, set MQ =
M ⊗ Q and NQ = HomZ (M, Q). Let  be a smooth projective toroidal fan in NQ .
A map f : NQ → Q is piecewise polynomial if for any σ ∈ , the map f |σ : σ → Q
extends to a polynomial function on the linear space spanQ {σ }, i.e., a piece-
wise polynomial function f on  is a collection of compatible polynomial func-
tions f σ : σ → Q. In particular, such a function is continuous. We denote by R
the set of all piecewise polynomial functions on  which is a ring under pointwise
addition and multiplication. Let S ∗ (MQ ) be the symmetric algebra of the Q-vector
space MQ . Recall that S ∗ (MQ ) can be naturally identified with the polynomial
functions on NQ . Note that R is a positively graded Q-algebra with graded subal-
gebra S ∗ (MQ ). Indeed, any piecewise polynomial function f = ( f σ )σ ∈ uniquely
decomposes into a sum of homogeneous piecewise polynomial functions.
Exercise 68 Let  be a smooth projective toroidal fan in NQ . Show that for any
ray ρ there is a piecewise linear function ϕρ on  which vanishes on all the other rays
and satisfies ϕ(u ρ ) = 1 where u ρ is the primitive ray generator in N of the ray ρ.
Let us write (1) for the set of rays of a fan  and u ρ for the primitive generator
in N of the ray ρ ∈ (1).
Lemma 69 If  is a smooth projective toroidal fan, then {ϕρ : ρ ∈ (1)} (where ϕρ
is defined in Exercise 68) forms a basis of R
1
the space of piecewise linear functions
on .
Exercise 70 Let  be a smooth projective toroidal fan. Show that R is isomorphic
to the Stanley-Reisner algebra R , i.e., the quotient ring of Q[Tρ : ρ ∈ (1)] by the
k
relations i=1 Tρi = 0 whenever ρ1 , . . . , ρk are distinct rays which do not generate a
k
cone of . (Hint: Clearly i=1 ϕρi = 0 whenever ρ1 , . . . , ρk do not generate a cone
of . Therefore, there is a unique algebra homomorphism from R to R , which
sends Tρ to ϕρ . Show that this map is an isomorphism.)
We can now reformulate Theorem 67:
Proposition 71 Let X  be a smooth projective toroidal horospherical variety
defined by an (uncolored) fan . Then the cohomology ring H ∗ (X  , Q) is iso-
morphic as an H ∗ (G/P, Q)-algebra to the quotient of H ∗ (G/P, Q) ⊗ R by the
ideal
 

c1 (δ(m)) ⊗ 1 − u ρ , m1 ⊗ ϕρ : m ∈ M =
ρ∈(1)
 
c1 (δ(m)) ⊗ 1 − 1 ⊗ ·, m : m ∈ M ,

where ·, m ∈ S ∗ (MQ ) is a (piecewise) linear function on .

218 J. Hofscheier

Proof The statement is a reformulation of Theorem 67 except the equality of the two
ideals which remains to be shown. Recall from Lemma 69 that the set of piecewise
linear functions {ϕρ : ρ ∈ (1)} (where ϕρ are defined in Exercise 68) forms a basis
of R1
. Then the (piecewise) linear function ·, m!
for m ∈ M can be expressed as a
linear combination in this basis, namely ·, m = ρ∈(1) u ρ , mϕρ . 

We denote by R the set of all piecewise

 polynomial functions on smooth projective
toroidal fans in NQ , i.e., R =  R where the union is taken over all smooth
projective toroidal fans .

Theorem 72 We have that

" #
C ∗ (G/H ) ⊗ Q ∼
= H ∗ (G/P, Q) ⊗ R / c1 (δ(m)) ⊗ 1 − 1 ⊗ ·, m : m ∈ M ,

where ·, m ∈ S ∗ (MQ ) is a piecewise linear function on any smooth projective

toroidal fan.

Proof For convenience, let us write A := H ∗ (G/P, Q).

By Theorem 63, we have
 
C (G/H ) ⊗ Q = lim H (X , Z) ⊗ Q = lim H ∗ (X  , Q)
∗ ∗ 
−→ −→
X X

where the limit is taken over all smooth projective toroidal embeddings of G/H
which is a directed set. Indeed, for any two smooth projective toroidal embeddings
with corresponding fans 1 , 2 , we can find a third smooth projective toroidal fan 
which refines both fans 1 and 2 . We introduce the relation     whenever  
refines . Suppose     , so that we obtain an equivariant map X   → X  .
Our goal is to understand how the representation of cohomology rings given in
Proposition 71 behaves under this map. By Proposition 71, the cohomology rings
(as A-algebras) are generated by classes of divisors, so that the map corresponding
to X   → X  is given by pulling back divisors which in turn induces the natural inclu-
sion R ⊆ R  . Let I := c1 (δ(m)) ⊗ 1 − 1 ⊗ ·, m : m ∈ M ⊆ A ⊗ R . Simi-
larly define I  in A ⊗ R  . As I ⊆ I  , we obtain the natural map μ,  : (A ⊗
R )/I → (A ⊗ R  )/I  . Then ((A ⊗ R )/I , μ,  ) is the direct system yield-
ing the direct limit lim H ∗ (X  , Q). Moreover, we obtain two more direct systems,
−→
namely (I , I ⊆ I  ) and (A ⊗ R , A ⊗ R ⊆ A ⊗ R  ) (for     ). Indeed,
we obtain a direct system of exact sequences:

0 → I → A ⊗ R → (A ⊗ R )/I → 0.

The statement follows by the fact that taking direct limits in the category of modules
is an exact functor, lim A ⊗ R = A ⊗ R, and lim I = I , where I denotes the ideal
−→ −
→
in the statement. 
9 The Ring of Conditions for Horospherical Homogeneous Spaces 219

Acknowledgements We want to express our gratitude to Megumi Harada and Adam Van Tuyl
for organizing the exciting “Graduate Summer School in Algebraic Group Actions” at McMaster
University and their great support. We also want to thank the participants of the summer school for
their constructive feedback. Finally, we are grateful to the anonymous referee for several valuable
suggestions and remarks.

References

1. Akhiezer, D.N.: Actions with a finite number of orbits. Funktsional. Anal. i Prilozhen. 19(1),
1–5, 96 (1985)
2. Borel, A.: Linear algebraic groups. Graduate Texts in Mathematics, vol. 126, 2nd edn. Springer,
New York (1991)
3. Bravi, P., Pezzini, G.: Primitive wonderful varieties. Math. Z. 282(3–4), 1067–1096 (2016)
4. Brion, M.: Vers une généralisation des espaces symétriques. J. Algebra 134(1), 115–143 (1990)
5. Brion, M.: Piecewise polynomial functions, convex polytopes and enumerative geometry. In:
Parameter spaces (Warsaw, 1994), Banach Center Publications, vol. 36, pp. 25–44. Polish
Academy of Sciences Institute of Mathematics, Warsaw (1996)
6. Cupit-Foutou, S.: Wonderful varieties: a geometrical realization (2009). arXiv:0907.2852
7. De Concini, C., Procesi, C.: Complete symmetric varieties. II. Intersection theory. In: Algebraic
Groups and Related Topics (Kyoto/Nagoya, 1983), Advanced Studies in Pure Mathematics,
vol. 6, pp. 481–513. North-Holland, Amsterdam (1985)
8. Gagliardi, G., Hofscheier, J.: Gorenstein spherical Fano varieties. Geom. Dedicata 178, 111–
133 (2015)
9. Hartshorne, R.: Algebraic Geometry. Springer, New York, Heidelberg (1977). Graduate Texts
in Mathematics, No. 52
10. Humphreys, J.E.: Linear Algebraic Groups. Springer, New York, Heidelberg (1975). Graduate
Texts in Mathematics, No. 21
11. Kiritchenko, V.: Geometry of spherical varieties (2009). Course notes. https://users.mccme.ru/
valya/sph1.pdf
12. Kleiman, S.L.: The transversality of a general translate. Compositio Math. 28, 287–297 (1974)
13. Knop, F.: The Luna-Vust theory of spherical embeddings. In: Proceedings of the Hyderabad
Conference on Algebraic Groups (Hyderabad, 1989), pp. 225–249. Manoj Prakashan, Madras
(1991)
14. Knop, F.: Localization of spherical varieties. Algebra Number Theory 8(3), 703–728 (2014)
15. Losev, I.V.: Uniqueness property for spherical homogeneous spaces. Duke Math. J. 147(2),
315–343 (2009)
16. Luna, D.: Variétés sphériques de type A. Publ. Math. Inst. Hautes Études Sci. 94, 161–226
(2001)
17. Luna, D., Vust, T.: Plongements d’espaces homogènes. Comment. Math. Helv. 58(2), 186–245
(1983)
18. Malle, G., Testerman, D.: Linear algebraic groups and finite groups of Lie type. In: Cambridge
Studies in Advanced Mathematics, vol. 133. Cambridge University Press, Cambridge (2011)
19. Pasquier, B.: Fano horospherical varieties. Ph.D. thesis, Université Joseph-Fourier-Grenoble I
(2006)
20. Pasquier, B.: Variétés horosphériques de Fano. Bull. Soc. Math. France 136(2), 195–225 (2008)
21. Sankaran, P., Uma, V.: Cohomology of toric bundles. Comment. Math. Helv. 78(3), 540–554
(2003)
22. Schubert, H.: Kalkül der abzählenden Geometrie. Springer, Berlin-New York: Reprint of the
1879 original. With an introduction by Steven L, Kleiman (1979)
23. Shafarevich, I.R.: Basic algebraic geometry. 1, Russian edn. Springer, Heidelberg (2013).
Varieties in projective space
220 J. Hofscheier

24. Shafarevich, I.R.: Basic Algebraic Geometry. 2, 3rd edn. Springer, Heidelberg (2013). Schemes
and complex manifolds, Translated from the 2007 third Russian edition by Miles Reid
25. Springer, T.A.: Linear Algebraic Groups, 2nd edn. Modern Birkhäuser Classics. Birkhäuser
Boston Inc, Boston, MA (2009)
26. Timashev, D.A.: Homogeneous spaces and equivariant embeddings. Encyclopaedia of Mathe-
matical Sciences, vol. 138. Springer, Heidelberg (2011). Invariant Theory and Algebraic Trans-
formation Groups, 8
27. Vakil, R.: The rising sea: foundations of algebraic geometry (2017). Course notes.
http://math.stanford.edu/∼vakil/216blog/FOAGnov1817public.pdf
28. Wasserman, B.: Wonderful varieties of rank two. Transform. Groups 1(4), 375–403 (1996)
Chapter 10
Linear Recursions for Integer Point
Transforms

Katharina Jochemko

Abstract We consider the integer point transform


σ P (x) = xm ∈ C[x1±1 , . . . , xn±1 ]
m∈P∩Zn

of a polytope P ⊂ Rn . We show that if P is a lattice polytope then for any polytope Q

the sequence {σk P+Q (x)}k≥0 satisfies a multivariate linear recursion that only depends
on the vertices of P. We recover Brion’s Theorem and by applying our results to
Schur polynomials we disprove a conjecture of Alexandersson (2014).

Keywords Lattice Polytops · Integer point transforms · Valuations · Brison’s

Theorem · Schur polynomials

10.1 Introduction

A polytope is the convex hull of finitely many points in Rn . A polytope is a lattice

polytope if all its vertices lie in the integer lattice Zn . The integer point transform
of a polytope P is defined by

σ P (x) = xm ∈ C[x1±1 , . . . , xn±1 ] ,
m∈P∩Zn

where xm denotes x1m 1 · · · xnm n for all m ∈ Zn . In this note we study sequences
{σk P (x)}k≥0 of integer point transforms of integer dilates of polytopes P and rel-
atives. We prove the following linear recursion.
Theorem 1 Let Q be a polytope in Rn and let P be a lattice polytope with vertex
set vert(P) = {v1 , . . . , vr }. Then the sequence {σk P+Q (x)}k≥0 satisfies the linear

K. Jochemko (B)
Department of Mathematics, KTH Royal Institute of Technology, SE–100 44 Stockholm, Sweden
e-mail: [email protected]
© Springer Nature Switzerland AG 2022 221
A. M. Kasprzyk and B. Nill (eds.), Interactions with Lattice Polytopes,
Springer Proceedings in Mathematics & Statistics 386,
https://doi.org/10.1007/978-3-030-98327-7_10
222 K. Jochemko

recursion
 
σ(k+r )P+Q (x) = (−1)1+|I | x i∈I vi
σ(k+r −|I |)P+Q (x)
∅= I ⊆[r ]

with characteristic polynomial


χ P;Q (X ) := (X − xv ) .
v∈vert(P)

If Q is a lattice polytope, then χ P;Q is minimal.

In particular, the recursion only depends on the vertices of P. This improves results
by Alexandersson [2] where it was assumed that P has the integer decomposition
property and Q = {0}.
Employing classical results from valuation theory, in Sect. 10.2 we first prove a
recursion for indicator functions of dilated polytopes. Then, in Sect. 10.3, we apply
these results to integer point transforms and prove Theorem 1. We recover Brion’s
Theorem in Sect. 10.4 and by applying our results to Schur polynomials we disprove
a conjecture of Alexandersson [1] in Sect. 10.5.

10.2 Characteristic Functions and Valuations

In this section we prove a linear recursion for indicator functions of integer dilates of
a polytope P. Let P denote the set of polytopes in Rn and let G be an abelian group.
A valuation is a map ϕ : P → G such that ϕ(∅) = 0 and

ϕ(P ∪ Q) = ϕ(P) + ϕ(Q) − ϕ(P ∩ Q) ,

for all P, Q ∈ P such that also P ∪ Q ∈ P. The volume, the number of lattice points
inside a polytope and the integer point transform are examples of valuations. It was
shown by Volland [19] that every valuation satisfies the inclusion-exclusion property.
That is, for polytopes P, P1 , . . . , Pr such that P = P1 ∪ · · · ∪ Pr

ϕ(P) = (−1)|I |+1 ϕ(PI ) ,
∅= I ⊆[r ]


 PI := i∈I Pi . Stronger even, it follows from a result of Groemer
where  [9], that
if αi 1 Pi = 0 for polytopes P1 , . . . , Pm and some α1 , . . . , αm ∈ Z then i αi ϕ(Pi )
= 0 where1 P denotes the indicator function for every polytope P. A function of
the form αi 1 Pi is called a polytopal simple function. By Groemer’s result, every
valuation uniquely defines a homomorphism from the abelian group of polytopal
simple functions to G, that is, every polytope can be identified with its indicator
10 Linear Recursions for Integer Point Transforms 223

function. For valuations on lattice polytopes this was proved by McMullen [12]. It
is well-known that for every affine linear map T : Rn → Rm

1 P → 1T (P) (10.1)

defines a valuation (see, e.g., [4, Chap. 8]). Using this push forward map we obtain
the following recursion on indicator functions.
Theorem 2 Let P be a polytope in Rn with vertex set vert(P) = {v1 , . . . , vr }. Then

1(k+r )P = (−1)1+|I | 1 Q k,r
I
∅= I ⊆[r ]


I = (k + r − |I |)P +
for all k ≥ 0 where Q k,r i∈I vi .
Proof We first assume that P is the (d − 1)-dimensional standard simplex d−1 =
{x ∈ Rd : x1 + · · · + xd = 1, x1 , . . . , xd ≥ 0}. Its (k + d)th dilate is given by

(k + d)d−1 = {x ∈ Rd : x1 + · · · + xd = d + k, x1 , . . . , xd ≥ 0}

For all I ⊆ [d], let

PI = (k + d)d−1 ∩ {x ∈ Rd : xi ≥ 1 for all i ∈ I } .


I = i∈I P{i} for all ∅ = I ⊆ [r ]. As in [15] we observe that (k + d)d−1 =
Then P
P∅ = i∈[d] P{i} for all k ≥ 0. Therefore, by inclusion-exclusion,

1(k+d)d−1 = (−1)1+|I | 1 PI
∅= I ⊆[d]

 we finish the proof of this case by observing that PI = (k + d − |I |)d−1 +

and
i∈I ei .
For the general case, we recall that every polytope is an affine linear projection
of a standard simplex and, thus, the claim follows by applying the push forward
map (10.1). 
For fixed Q ∈ P, 1 P → 1 P+Q defines a valuation (see, e.g., [16]) where
P + Q = { p + q : p ∈ P, q ∈ Q} is the Minkowski sum. The family of all poly-
topal simple functions forms an algebra where the multiplicative structure is given
by the Minkowski sum of polytopes: 1 P  1 Q := 1 P+Q for all polytopes P and Q.
Another proof of Theorem 2 can be obtained from the following result which was
proved in [11]. See also [14] for related material.
Theorem 3 ([11, Lemma 5]) Let P be a polytope with vertex set vert(P) =
{v1 , . . . , vr }. Then    
1 P − 1v1  · · ·  1 P − 1vr = 0 . (10.2)
224 K. Jochemko

Proof (2nd proof of Theorem 2) The proof follows from Theorem 3 by expanding
Eq. (10.2) and multiplying both sides with 1k P . 

By the discussion above, Theorem 2 is equivalent to the following.

Theorem 4 Let P be a polytope in Rn with vertex set vert(P) = {v1 , . . . , vr }, and
let ϕ : P → G be a valuation. Then

ϕ((k + r )P) = (−1)1+|I | ϕ(Q k,r
I )
∅= I ⊆[r ]


I = (k + r − |I |)P +
for all k ≥ 0 where Q k,r i∈I vi .

10.3 A Multivariate Recursion

A sequence a = {ak }k≥0 of elements in C(x1 , . . . , xn ) satisfies a linear recursion of

order d ≥ 1 if there are c1 , . . . , cd ∈ C(x1 , . . . , xn ), cd = 0, such that


d
ak = c j ak− j
j=1

d all k ≥d−
for d. The corresponding characteristic polynomial χc is defined as X d −
j=1 c j X
j
∈ C(x1 , . . . , xn )[X ]. The polynomial χc is called minimal if for every
vector c = (c1 , . . . , cd  ) corresponding to a linear recursion of a we have χc |χc .


Since C(x1 , . . . , xn )[X ] is a principal ideal domain a uniquely determined minimal

polynomial exists.
We are now ready to proof Theorem 1.
Proof (Proof of Theorem 1) Let r = |vert(P)| be the number of vertices of P. Since
the maps P → P + Q and also P → σ P (x) define valuations, by Theorem 4

σ(k+r )P+Q (x) = (−1)1+|I | σ(k+r −|I |)P+i∈I vi +Q (x)
∅= I ⊆[r ]
 
= (−1)1+|I | x i∈I vi
σ(k+r −|I |)P+Q (x) ,
∅= I ⊆[r ]

where the last equation follows by observing that σ P+v (x) = xv σ P (x) for all v ∈ Zn .
We observe that χ P;Q is the characteristic polynomial of this linear recursion.
Now let Q be a lattice polytope and suppose that χ P;Q is not minimal. Then, for
some vertex u of P, {σk P+Q (x)}k≥0 satisfies a linear recursion with characteristic
polynomial v∈vert(P)\{u} (X − xv ). That is
10 Linear Recursions for Integer Point Transforms 225

|vert(P)|−1

σ(k+r )P+Q (x) + (−1) j e j ({xv : v ∈ vert(P) \ {u}})σ(k+r − j)P+Q (x) = 0
j=1

where e j denotes the jth elementary symmetric polynomial in |vert(P)| − 1 vari-

ables. Now let v be a vertex of Q such that u + v is a vertex of P + Q. Then (k +
r )u + v is a vertex of (k + r )P + Q and thus x (k+r )u+v appears as a summand
in σ(k+r )P+Q (x). However, it does not appear in
e j ({xv : v ∈ vert(P) \ {u}})σ(k+r − j)P+Q (x) for any 1 ≤ j ≤ |vert(P)| − 1. To see
that, it suffices to argue that (k + r )u + v is not contained in (k + r − j)P + Q +
 j
l=1 vl for any choice of v1 , . . . , v j ∈ vert(P) \ {u}. For that, let  : R → R be
n

a linear functional such that (u) > ( p) for  all p = u in P and (v) > (q) for
j
all q = v in Q. Then ((k + r − j) p + q + l=1 vl )=(k + r − j)( p) + (q) +
j
l=1 (vl ) < (k + r − j)(u)
+ (v) + j(u) = ((k + r )u + v) for all p ∈ P and q ∈ Q and the conclusion fol-
lows. 
Every linear map f : Rn → Rl with the property that f (Zn ) ⊆ Zl induces an
algebra homomorphism



f¯ : C x1±1 , . . . , xn±1 → C x1±1 , . . . , xl±1
xm → x f (m)

As a consequence of Theorem 1 we therefore obtain the following.

Proposition 5 Let Q be a polytope in Rn and P be a lattice polytope with
vertex set vert(P), and let f : Rn → Rl a linear map such that f (Zn ) ⊆ Zl .
Then { f¯(σk P+Q (x))}k≥0 satisfies a linear recursion with characteristic polynomial

(X − x f (v) ).
f
χ P;Q (X ) :=
v∈vert(P)

The following two examples show that the minimality of a characteristic polyno-
mial is not necessarily preserved under affine transformations or taking Minkowski
sums.

Example 6 If Q in Theorem 1 is not a lattice polytope then χ P;Q is not necessar-

ily minimal. A counterexample is given by the lattice segment P = [0, 1] and the
point Q = {(0.5, 0.5)} in R2 . In that case σk P+Q ≡ 0 is constant.

Example 7 (Ehrhart polynomials) For f : Rn → R0 and f ≡ 0 we obtain

f¯(σk P (x)) = |k P ∩ Zn | and thus recover the Ehrhart function counting lattice points
in integer dilates of P. If P is a lattice polytope then this function is known to agree
with a polynomial of degree dim P [7]. Therefore the order of the minimal polyno-
mial of the sequence is dim P as was demonstrated in [15] and is thus in general
smaller than |vert(P)|.
226 K. Jochemko

These examples motivate the following question.

Question 8 What are necessary and sufficient conditions on Q and on f that guar-
f
antee that χ P;Q is minimal?

10.4 Brion’s Theorem

In this section we provide a proof of Brion’s Theorem using the recursion given
in Theorem 1. For a polytope P ⊆ Rn and a vertex v of P the tangent cone Kv is
defined as {v + w : v + εw ∈ P for 0 < ε  1}. If the polytope P has rational edge
directions, in particular, if it is a lattice polytope, then the integer point transform
of Kv is a rational function.
Theorem 9 (Brion’s Theorem [6]) Let P be a lattice polytope with vertex set
vert(P). Then 
σ P (x) = σKv (x)
v∈vert(P)

as rational functions.
The following is an immediate consequence of [5, Lemma 13.5.]

Lemma 10 ([5]) Let u 1 , . . . , u k ∈ Zn such that the cone K := cone(u 1 , . . . , u k )

generated by u 1 , . . . , u k is pointed. Then

σK (x) = xm
m∈K∩Zn

is a rational function and converges absolutely for all x

in {x ∈ Cn : |xu i | < 1 for i = 1, . . . , k}.

A further ingredient for our proof of Brion’s Theorem is the following well-known
result (see, e.g., [17, Chap. 5]).
Lemma 11 Let {an }n∈N be a sequence of elements of a field K that satisfy a linear
recursion of order d with characteristic polynomial


d
(X − ri ).
i=1

If all roots r1 , . . . , rd are distinct then there are α1 , . . . , αd ∈ K such that


d
an = αi rin , for all n ∈ N.
i=1
10 Linear Recursions for Integer Point Transforms 227

Proof (Proof of Theorem 9) By Theorem 1 and Lemma 11 there exist

cv ∈ C(x1 , . . . , xn ) for all v ∈ vert(P) such

σk P (x) = cv xkv (10.3)
v∈vert(P)

for all k ≥ 0. Our goal is to show that cw · xw = σKw (x) as rational functions for
all w ∈ vert(P), or, equivalently, that cw equals the integer point transform of the
0 of the translated polytope P − w. Equation (10.3) is
tangent cone K̃0 of the vertex
equivalent to σk(P−w) (x) = v∈vert(P) cv xk(v−w) . As k goes to infinity σk(P−w) (x) con-
verges absolutely to σK̃0 (x) on WK̃0 = {x ∈ Cn : |xv−w | < 1 for all

v ∈ vert(P) \ {w}} by Lemma 10. On the other hand, v∈vert(P) cv xk(v−w) converges
to cw . Thus σK̃0 (x) and cw coincide on WK̃0 and are therefore the same as rational
functions. 

10.5 Schur Polynomials

In this section we apply our results to Schur polynomials.

A partition is a vector λ = (λ1 ≥ λ2 ≥ · · · ≥ λn ) of weakly decreasing nonneg-
ative integers. The number of strictly positive entries λi is called the length of λ . A
partition μ is smaller than a partition λ with respect to the inclusion order if μi ≤ λi
for all i. The partition μ is an partition λ
smaller than with respect to the domination
n k k
order, denoted λ  μ , if i=1 λi = i=1 μi and i=1 λi ≥ i=1 μi for all k. A
skew Young diagram of shape λ /μ μ isan axes-parallel arrangement of unit squares in
the plane centered at the coordinates (i, j) ∈ Z2 : μi < j ≤ λi . A semi-standard
Young tableau is a Young diagram together with a filling of the boxes with natural
numbers such that the numbers are strictly increasing in each column and weakly
increasing in every row. Let Tλn /μμ denote the set of semi-standard Young tableaux
filled with numbers in [n] = {1, 2, . . . , n}. For every T in Tλn /μμ let w(T ) be the vec-
tor t = (t1 , . . . , tn ) where ti is the number of boxes filled with i. The vector w(T )
is called the weight of T . The Kostka coefficient Kλ /μμ,w equals the number of
tableaux of shape λ /μ μ with weight w. In particular, Kλ /μμ,w > 0 if and only if there
is T ∈ Tλn /μμ with w(T ) = w. The skew Schur polynomial of shape λ /μ μ is defined
as 
sλ/μμ (x) = xw(T ) ∈ C[x1±1 , . . . , xn±1 ] .
T ∈Tλn /μμ

In [1] Alexandersson proved the following recursion for Schur polynomials.

Theorem 12 ([1, Theorem 1]) Let n be a natural number and let κ , λ , μ , ν be
λ ⊇ ν + kμ
partitions of length at most n such that λ ⊇ μ and κ + kλ μ for some positive
∞
integer k. Then there is a natural number r such that the sequence {sκ +lλλ/νν +lμμ (x)}l=r
satisfies a linear recursion with characteristic polynomial
228 K. Jochemko

Fig. 10.1 Gelfand–Tsetlin patterns

  
χ (X ) = X − xw(T ) .
T ∈Tλn /μμ

Furthermore, in [1] the following conjecture concerning the minimal polynomial

was stated. For every vector w let w denote the vector obtained from w by rearranging
its coordinates in non-increasing order.
Conjecture 13 ([1, Conjecture 25]) Let κ , λ , μ , ν be as in Theorem 12 and let

W = {w ∈ Nn : Kλ /μμ,w > 0 and w  λ − μ } .

∞
Then, for sufficiently large r , {sκ+lμ/λ+lν (x)}l=r satisfies a linear recursion with min-
imal polynomial  
χ (X ) = X − xw .
w∈W

We use Theorem 1 and a well-known correspondence between elements in Tλn /μμ

and lattice points in the Gelfand–Tsetlin-Polytope GLλ /μμ to improve Theorem 12
and to given an example in which the polynomial in Conjecture 13 is not minimal
thus refuting the conjecture.
There is a one-to-one correspondence between semi-standard Young tableaux
and Gelfand–Tsetlin patterns. A Gelfand–Tsetlin pattern is a rectangular array of
nonnegative real numbers {xi, j } i=1,...,n+1 arranged as in Fig. 10.1 such that the entries
j=1,...,n
are weakly increasing in north-east and south-east direction, that is xi, j ≤ xi+1, j+1
for all i, j and xi, j ≤ xl, j for all i > l. For fixed top and bottom rows the family of
Gelfand–Tsetlin patterns forms a polytope, the Gelfand–Tsetlin polytope, which
belongs to the class of marked order polytopes introduced by Ardila, Bliem and
Salazar [3]. There is a well-known one-to-one correspondence between elements
of Tλn /μμ and integer valued Gelfand–Tsetlin patterns with top row λ and bottom
row μ , that is, lattice points in the corresponding Gelfand–Tsetlin polytope GLλ /μμ .
Via this correspondence, the weight function can be represented as a linear function
 Gelfand–Tsetlin pattern x = {xi, j } the ith coordinate of
on GLλ /μμ , namely for every
the weight w(x) equals nk=1 (xi,k − xi+1,k ) for all 1 ≤ i ≤ n. Further details may
be found in [18]. It follows that

sλ /μμ (x) = xw( p) ,
p

where p is over all lattice points in GLλ /μμ .

10 Linear Recursions for Integer Point Transforms 229

As a corollary of Theorem 1 we obtain the following.

Corollary 14 Let n be a natural number and let κ , λ , μ , ν be partitions of length at
most n such that λ ⊇ μ and κ + kλ λ ⊇ ν + kμ μ for some positive integer k. Let V be the
∞
set of vertices of GLλ /μμ . Then there is an integer r  0 such that {sκ +lλλ/νν +lμμ (x)}l=r
satisfies a linear recursion with characteristic polynomial
 
χ (X ) = X − xw(v) .
v∈V

Proof Let f = (λ λ, μ) and g = (κκ , ν ). Then there is an r  0 such that if f i < f j

then r f i + gi < r f j + g j for all i = j. In particular, one can find a permutation σ ∈
S2n such that

f σ (1) ≤ f σ (2) ≤ · · · ≤ f σ (2n) and r f σ (1) + gσ (1) ≤ · · · ≤ r f σ (2n) + gσ (2n) .

Then, by Theorem [8, Theorem 2.10],

GLκ +lλλ/νν +lμμ = GLκ +rλλ/νν +rμμ + (l − r )GLλ /μμ

for all l ≥ r . The claim now follows from Proposition 5 since the weight function w
is linear. 

Since typically there are more lattice points in GLλ /μμ than vertices, Corollary 14
shows that the characteristic polynomial given in Theorem 12 is in general not mini-
mal. The next example shows that also the polynomial given in Conjecture 13 is not
minimal in general, thus refuting it.

Example 15 Let n = 3, λ = (5, 3, 1) and μ = (3, 0, 0). Consider the skew Young
tableau T and its corresponding Gelfand–Tsetlin pattern p depicted in Fig. 10.2.
Then
w(T ) = w( p) = (4, 2, 0)  (3, 2, 1) = λ − μ .

2 3 1 3 5
0 1 4
2 3 3 0 0 3
0 0 3
3

Fig. 10.2 The skew Young tableau T in Example 15 and its corresponding Gelfand–Tsetlin pat-
tern p
230 K. Jochemko

From the face structure studied in [10, 13] it follows that the coordinates of any
vertex of GLλ /μμ are in the set {0, 1, 3, 5}. Let x = {xi, j } be a Gelfand–Tsetlin pattern
that is a vertex of GLλ/μμ . Then x4,1 , x4,2 and x3,1 are 0. Furthermore, x2,1 ∈ {0, 1}.
If x2,1 = 0, then the sum of entries of the first row of x is odd and the sum of entries
of the second is even, therefore w(x)1 is odd and w(x) = (4, 2, 0). On the other hand,
if x2,1 = 1, then x3,2 ∈ {1, 3} and in that case w(x)2 is odd and again w(x) = (4, 2, 0).
In summary, (4, 2, 0) ∈ W is not the weight of a vertex of GLλ /μμ and therefore
   
X − xw  X − xw(v) .
w∈W v∈V

Therefore, by Corollary 14, w∈W (X − xw ) cannot be the minimal polynomial.

Remark 16 To verify the counterexample given in Example 15, one can also use [2,
Proposition 6].

Acknowledgements The author would like to thank Per Alexandersson and Raman Sanyal for
inspiring and fruitful discussions and many helpful comments. The author was partially supported
by a Hilda Geiringer Scholarship at the Berlin Mathematical School, the Knut and Alice Wallen-
berg Foundation and a Microsoft Research Fellowship of the Simons Institute for the Theory of
Computing.

References

1. Alexandersson, P.: Stretched skew Schur polynomials are recurrent. J. Combin. Theory Ser. A
122, 1–8 (2014)
2. Alexandersson, P.: Polynomials defined by tableaux and linear recurrences. Electron. J. Com-
bin. 23(1), Paper 1.47, 24 (2016)
3. Ardila, F., Bliem, T., Salazar, D.: Gelfand-Tsetlin polytopes and Feigin-Fourier-Littelmann-
Vinberg polytopes as marked poset polytopes. J. Combin. Theory Ser. A 118(8), 2454–2462
(2011)
4. Barvinok, A.: A course in convexity, Graduate Studies in Mathematics, vol. 54. American
Mathematical Society, Providence (2002)
5. Barvinok, A.: Integer points in polyhedra. Zurich Lectures in Advanced Mathematics. European
Mathematical Society (EMS), Zürich (2008)
6. Brion, M.: Points entiers dans les polyèdres convexes. Ann. Sci. École Norm. Sup. (4) 21(4),
653–663 (1988)
7. Ehrhart, E.: Sur les polyèdres rationnels homothétiques à n dimensions. C. R. Acad. Sci. Paris
254, 616–618 (1962)
8. Fang, X., Fourier, G., Pegel, C.: The Minkowski property and reflexivity of marked poset
polytopes. Electron. J. Combin. 27(1), Paper No. 1.27, 19 (2020)
9. Groemer, H.: On the extension of additive functionals on classes of convex sets. Pacific J. Math.
75(2), 397–410 (1978)
10. Jochemko, K., Sanyal, R.: Arithmetic of marked order polytopes, monotone triangle reciprocity,
and partial colorings. SIAM J. Discrete Math. 28(3), 1540–1558 (2014)
11. Lawrence, J.: Three rings of polyhedral simple functions. J. Res. Natl. Inst. Stand. Technol.
111(2), 127–134 (2006)
12. McMullen, P.: Valuations on lattice polytopes. Adv. Math. 220(1), 303–323 (2009)
10 Linear Recursions for Integer Point Transforms 231

13. Pegel, C.: The face structure and geometry of marked order polyhedra. Order 35(3), 467–488
(2018)
14. Pukhlikov, A.V., Khovanskiı̆, A.G.: Finitely additive measures of virtual polyhedra. Algebra i
Analiz 4(2), 161–185 (1992)
15. Sam, S.V.: A bijective proof for a theorem of Ehrhart. Amer. Math. Monthly 116(8), 688–701
(2009)
16. Schneider, R.: Convex bodies: the Brunn-Minkowski theory. Encyclopedia of Mathematics and
its Applications, vol. 151, expanded edn. Cambridge University Press, Cambridge (2014)
17. Spiegel, M.R.: Calculus of Finite Differences and Difference Equations. Schaum’s Outline
Series. McGraw-Hill, New York (1971)
18. Stanley, R.P.: Enumerative combinatorics, vol. 2. Cambridge Studies in Advanced Mathematics,
vol. 62. Cambridge University Press, Cambridge (1999). With a foreword by Gian-Carlo Rota
and appendix 1 by Sergey Fomin
19. Volland, W.: Ein Fortsetzungssatz für additive Eipolyederfunktionale im euklidischen Raum.
Arch. Math. (Basel) 8, 144–149 (1957)
Chapter 11
Schubert Calculus on Newton–Okounkov
Polytopes

Valentina Kiritchenko and Maria Padalko

Abstract A Newton–Okounkov polytope of a complete flag variety can be turned

into a convex geometric model for Schubert calculus. Namely, we can represent
Schubert cycles by linear combinations of faces of the polytope so that the intersection
product of cycles corresponds to the set-theoretic intersection of faces (whenever
the latter are transverse). We explain the general framework and survey particular
realizations of this approach in types A, B and C.

Keywords Schubert calculus · Mitosis · Gelfand-Zetlin polytopes

11.1 Introduction

Theory of Newton–Okounkov convex bodies [12, 19] allows us to apply ideas of

toric geometry in the non-toric setting. In this paper, we explore non-toric applica-
tions of polytope rings (see Sect. 11.2 for a definition) introduced by Khovanskii and
Pukhlikov [25]. With a convex polytope P ⊂ Rd , they associated a graded commu-
tative ring (the polytope ring):
 d
RP = R iP
i=0

that has Poincaré duality. The polytope rings were originally used to give a convenient
functorial description of the cohomology rings of smooth toric varieties. In this
case, P is always a simple lattice polytope, that is, all vertices of P belong to Zd ⊂ Rd ,

V. Kiritchenko (B)
Laboratory of Algebraic Geometry and Faculty of Mathematics, National Research University
Higher School of Economics, Usacheva str. 6, 119048 Moscow, Russia
Institute for Information Transmission Problems, Moscow, Russia
e-mail: [email protected]
M. Padalko
SimilarWeb, Tel Aviv, Israel

© Springer Nature Switzerland AG 2022 233

A. M. Kasprzyk and B. Nill (eds.), Interactions with Lattice Polytopes,
Springer Proceedings in Mathematics & Statistics 386,
https://doi.org/10.1007/978-3-030-98327-7_11
234 V. Kiritchenko and M. Padalko

and only d edges meet at every vertex of P. In [10], Kaveh noted that polytope
rings can also be used for a partial description of the cohomology rings of spherical
varieties. In this case, P is still a lattice polytope but not necessarily simple.
For simple polytopes, every face  ⊂ P can be naturally identified with an ele-
ment x ∈ R P so that
x x = x∩

for any two transverse faces  and   . This is no longer true for non-simple polytopes,
that is, individual faces of P do not have natural counterparts in R P . However, it is
still possible to identify every element of R P with a linear combination of faces
of P so that the product in the polytope ring corresponds to the intersection of faces.
In [14], the first author, Smirnov and Timorin developed a general framework for
such calculus on polytopes, and studied its applications to Schubert calculus on
Gelfand–Zetlin polytopes in type A. In this paper, we mainly consider applications
to Schubert calculus in types B and C.
Representation theory of classical groups is a source of several interesting fami-
lies of lattice convex polytopes. For S L n (C) (type A), there is a well-known family
of Gelfand–Zetlin (GZ) polytopes G Z λ . Here λ := (λ1 , . . . , λn ) ∈ Zn runs through
dominant weights of S L n (C), that is, λ1 ≥ λ2 ≥ · · · ≥ λn . Originally, GZ polytopes
were constructed using representation theory, namely, lattice points in the polytope
G Z λ parameterize the vectors in a special basis in the irreducible representation Vλ
of S L n (C) with the highest weight λ (see [21] for a survey on GZ bases). In con-
vex geometric terms, the GZ polytope G Z λ ⊂ Rd , where d := n(n−1) 2
, is defined as
the set of all points (z 11 , z 21 , . . . , z n−1
1
; z 12 , . . . , z n−2
2
; . . . ; z 1n−1 ) ∈ Rd that satisfy the
following interlacing inequalities:

λ1 λ2 λ3 ... λn
z 11 z 21 ... 1
z n−1
z 12 ... 2
z n−2
.. .. (G Z A )
. .
z 1n−2 z 2n−2
z 1n−1

where the notation

a b
c

means a ≥ c ≥ b (the table encodes 2d inequalities). Figure 11.1 shows the 3-

dimensional GZ polytope for n = 3 and λ = (3, 0, −3). Note that G Z polytopes
are not simple.
GZ polytopes in types B, C and D were defined in [1] (see Sect. 11.2.2 for defini-
tions in types B and C) and are related to representation theory of S O2n+1 (C), Sp2n
and S O2n (C), respectively. They are special cases of string polytopes introduced
by Berenstein–Zelevinsky and Littelmann [20]. There are other families of poly-
11 Schubert Calculus on Newton–Okounkov Polytopes 235

Fig. 11.1 GZ polytope in

type A for n = 3
and λ = (3, 0, −3)

topes in representation theory such as Nakashima–Zelevinsky polyhedral realiza-

tions of crystal bases and Feigin–Fourier–Littelmann–Vinberg polytopes. They have
representation-theoretic meaning similar to that of string polytopes but are not com-
binatorially equivalent to the latter. All these polytopes were exhibited as Newton–
Okounkov polytopes of complete flag varieties for certain geometric valuations [3,
4, 6, 11, 17] (see Sect. 11.2.3 for more details).
For G = S L n (C), the complete flag variety G/B (here B ⊂ G denotes the sub-
group of upper-triangular matrices) can be thought of as a variety of complete flags
of subspaces ({0} ⊂ V 1 ⊂ V 2 ⊂ · · · ⊂ V n−1 ⊂ Cn ) where dim V i = i, and there
are no gaps. There are similar descriptions of complete flag varieties G/B for other
classical groups G (see Sect. 11.2.3). Recall that globally generated line bundles L λ
on G/B are in bijective correspondence with irreducible representations Vλ of G so
that H 0 (L λ , G/B)  Vλ∗ [2, Proposition 1.4.5]. Here λ runs through the dominant
weights of G. We denote by degλ (G/B) the degree of the image of G/B under the
map G/B → P(Vλ ) = P(H 0 (L λ , G/B)∗ ).
In [10], polytope rings of string polytopes were identified with the cohomology
rings of complete flag varieties. More generally, string polytope in this description
can be replaced with any linear family (in the sense of [13]) of convex polytopes Pλ
parameterized by the dominant weights λ whenever the following identity holds:

Vol(Pλ ) = d! degλ (G/B) (11.1)

where d := dim G/B. We regard both sides of this identity as polynomials in λ. In

particular, polytopes Pλ yield an analog of Kushnirenko’s theorem for G/B.
Since Newton–Okounkov polytopes of line bundles on G/B by construction sat-
isfy identity (11.1) they can be used to model Schubert calculus. Recall that the coho-
mology ring H ∗ (G/B, Z) has a special basis of Schubert cycles [X w ] with striking
u
positivity properties. Namely, theustructure constants (i.e., the coefficients cvw in the
decomposition [X w ][X v ] = u cvw [X u ]) are always non-negative. However, no enu-
merative meaning (in the spirit of Littlewood–Richardson rule for Grassmannians)
of these coefficients is known. Polytope rings provide a new framework for combina-
torial interpretation of structure constants. An important task is to find presentations
236 V. Kiritchenko and M. Padalko

of Schubert cycles in polytope rings by linear combinations of faces with positive

coefficients. Another task is to find Newton–Okounkov polytopes for which these
presentations have especially simple combinatorics. It is tempting to use Grossberg–
Karshon cubes [7, 8] since they are combinatorial cubes. However, there are might
be issues with positivity, that is, some Schubert cycles will be represented by linear
combinations of faces with negative coefficients (see Example 10).
There is an algorithm (geometric mitosis) for finding positive presentations of
Schubert cycles by faces using convex geometric analogs of Demazure operators from
representation theory [15, 16]. In the present paper, we describe geometric mitosis in
more combinatorial terms, outline its applications and formulate conjectures. For GZ
polytopes in type A, this algorithm reduces to Knutson–Miller mitosis on pipe dreams
and was used in [14]. In types B and C, geometric mitosis reduces to a different
combinatorial rule that conjecturally yields presentations of Schubert cycles by faces
of GZ polytopes in respective types. In particular, 4-dimensional GZ polytope in
type C2 can be used to model Schubert calculus on the variety of isotropic flags
in C4 [24]. Another convex geometric model for the same flag variety was constructed
in [9] using a different string polytope in type C2 .

11.2 Preliminaries

In this section, we recall the definitions of polytope rings, GZ polytopes and flag
varieties in types B and C. We discuss the relationship between the polytope rings
of GZ polytopes and cohomology rings of flag varieties. We also define Newton–
Okounkov polytopes of flag varieties.

11.2.1 Polytope Ring

Let L ⊂ Rd be a lattice, and P ⊂ Rd a convex polytope whose vertices lie in L.

We say that P is a lattice polytope with respect to L. By the standard lattice Zd
we mean the lattice {(x1 , . . . , xd ) ∈ Rd | xi ∈ Z for all i = 1, . . . , d}. We choose the
translation invariant volume form on Rd so that the covolume of L is 1.
Recall that two convex polytopes P and Q are called analogous if they have the
same normal fan, i.e. there is a one-to-one correspondence between the faces of P
and the faces of Q such that any linear functional, whose restriction to P attains its
maximal value at a given face F ⊆ P has the property that its restriction to Q attains
its maximal value at the corresponding face of Q.
Denote by S P the set of all polytopes analogous to P. This set can be endowed
with the structure of a commutative semigroup using Minkowski sum

P1 + P2 = {x1 + x2 ∈ Rd | x1 ∈ P1 , x2 ∈ P2 }
11 Schubert Calculus on Newton–Okounkov Polytopes 237

It is not hard to check that this semigroup has cancelation property. We can also
multiply polytopes in S P by positive real numbers using dilation:

λP = {λx | x ∈ P}, λ ≥ 0.

Hence, we can embed the semigroup of convex polytopes into its Grothendieck
group V P , which is a real vector space. The elements of V P are called virtual polytopes
analogous to P.
On the vector space V P , there is a homogeneous polynomial vol P of degree d,
called the volume polynomial. It is uniquely characterized by the property that its
value vol P (Q) on any convex polytope Q ∈ S P is equal to the volume of Q.
Let  P be a lattice in V P generated by some lattice polytopes (with respect to L)
analogous to P (we do not assume that  P contains all lattice polytopes analogous
to P). The symmetric algebra Sym( P ) of  P can be thought of as the ring of differ-
ential operators with constant integer coefficients acting on R[V P ], the space of all
polynomials on V P . If D ∈ Sym( P ) and ϕ ∈ R[V P ], then we write Dϕ ∈ R[V P ] for
the result of this action. Define A P as the homogeneous ideal in Sym( P ) consisting
of all differential operators D such that D vol P = 0. Set R P = Sym( P )/A P . This
ring is called the polytope ring associated with the polytope P and the lattice  P .

Example 1 Let L = Zd be the standard lattice, and P an integrally simple lattice

polytope (that is, only d edges meet at every vertex of P, and primitive vectors on
these edges span L over Z). Let  P be the lattice in V P generated by all lattice
polytopes (with respect to L) analogous to P. Then the ring R P is isomorphic to the
Chow (or cohomology) ring H ∗ (X P , Z) of the smooth toric variety X P associated
with the normal fan of P [25].
When P is simple, every facet  ⊂ P defines a differential operator ∂ ∈ R P
(see [14, Sect. 2.3] for the details). Recall that the closures of torus orbits in X P are
in bijective correspondence with faces of P. They also give a generating set in the
cohomology ring H ∗ (X P , Z). Every face F = 1 ∩ · · · ∩ k can be identified with
the operator [F] = ∂1 · · · ∂k ∈ R P . Using linear relations between ∂ in R P we can
compute products in H ∗ (X P , Z) by intersecting faces of P.
For instance, if P ⊂ R2 is the trapezoid with vertices (0, 0), (1, 0), (0, 1), (1, 2),
then the corresponding toric variety X P is the blow-up of CP2 at one point. The
edge 1 = {x = 0} corresponds to the exceptional divisor E ⊂ X P . The other edges
are 2 = {y − x = 1}, 3 = {x = 1} and 4 = {y = 0}. There are two linear rela-
tions between ∂i . Namely, the parallel translations along x and y axes do not
change the area of P, hence, ∂1 + ∂2 = ∂3 and ∂2 = ∂4 . In particular, the iden-
tity [E]2 = −[ pt] in H ∗ (X P , Z) can be obtained as follows:

[1 ]2 = [1 ]([3 ] − [2 ]) = [1 ∩ 3 ] − [1 ∩ 2 ] = −[ pt].

Example 2 Let L = Zd , and P the GZ polytope in type A corresponding to a strictly

dominant λ = (λ1 , . . . , λn ) (that is, λ1 > λ1 > · · · > λn ). Let  P be the lattice in V P
generated by all GZ polytopes Pλ for all dominant λ. Then the ring RG Z := R P is
238 V. Kiritchenko and M. Padalko

isomorphic to the cohomology ring H ∗ (G L n (C)/B, Z) of the complete flag variety

in type A [10].
Since the GZ polytope is not simple, there is no correspondence between individ-
ual faces of P and elements of H ∗ (G L n (C)/B, Z). However, it is possible to identify
every element of H ∗ (G L n (C)/B, Z) with a linear combination of faces of P (see [14,
Sect. 2] for more details). Again, we can compute all products in H ∗ (G L n (C)/B, Z)
by intersecting faces of P (see [14, Sect. 2.4] for an example of such computations).

In what follows, L will be a sublattice of 12 Zd := {(x1 , . . . , xd ) | 2xi ∈ Z for all i =

1, . . . , d}. We always compute volumes of faces of P with respect to the lattice L.
More precisely, if F ⊂ P is a face, and RF is its affine span then the volume of the
face is computed using the volume form on RF normalized so that the covolume
of L ∩ RF is 1.

11.2.2 GZ Polytopes in Types B and C

Let λ = (λ1 , . . . , λn ) be a non-increasing collection of non-negative integers. Put d =

n 2 . Denote coordinates in Rd by (x11 , . . . , xn1 ; y11 , . . . , yn−1
1
; . . . ; x1n−1 , x2n−1 , y1n−1 ;
x1 ). For every λ, define the symplectic GZ polytope SG Z λ ⊂ Rd for Sp2n (C) by the
n

following interlacing inequalities:

λ1 λ2 λ3 ... λn 0
x11 x21 ... xn1
y11 y21 . . . yn−1
1
0
x12 ... 2
xn−1
y12 . . . yn−2
2
0 G ZC
.. .. ..
. . .
x1n−1 x2n−1
n−1
y1 0
x1n

Again, every coordinate in this table is bounded from above by its upper left neighbor
and bounded from below by its upper right neighbor (the table encodes 2d inequal-
ities). We regard SG Z λ as a lattice polytope with respect to the standard lattice Zd .
Roughly speaking, SG Z λ is the polytope defined using half of the GZ pattern (G Z A )
for S L 2n (C).

Example 3 The polytope SG Z λ ⊂ R4 for Sp4 (C) is given by 8 inequalities:

λ1 ≥ x11 ≥ λ2 ; λ2 ≥ x21 ≥ 0; x11 ≥ y11 ≥ x21 ; y11 ≥ x12 ≥ 0.

It is not hard to compute the volume polynomial of SG Z λ :

11 Schubert Calculus on Newton–Okounkov Polytopes 239

1
vol SG Z (λ1 , λ2 ) = λ1 λ2 (λ1 − λ2 )(λ1 + λ2 ).
6
This volume times 4! is equal to the degree degλ (Sp4 (C)/B) of the isotropic flag
variety.

The polytope ring R SG Z defined by the family of symplectic GZ polytopes is

isomorphic to the cohomology ring H ∗ (Sp2n (C)/B, Z). Indeed, by [10] it is iso-
morphic to the subring of H ∗ (Sp2n (C)/B, Z) generated by the first Chern classes
of line bundles L λ corresponding to the weights of Sp2n (C). Since the torsion index
of Sp2n (C) is 1, this subring coincides with the whole ring (see [26] for the details
on torsion indices of classical groups).
The odd orthogonal GZ polytope OG Z λ ⊂ Rd for S O2n+1 (C) is defined using
the same pattern (G Z C ) but a different lattice L B ⊂ Rd . Namely, L B consists
of all points (x11 , . . . , xn1 ; y11 , . . . , yn−1
1
; . . . ; x1n−1 , x2n−1 , y1n−1 ; x1n ) ∈ 21 Zd such that
all coordinates except for xn , xn−1 ,…, x1n are integer. Lattice points SG Z λ ∩ Zd
1 2

and SG Z λ ∩ L B parameterize basis vectors in irreducible representations of Sp2n (C)

and S O2n+1 (C), respectively (see [20, Sect. 6] for more details).

Remark 4 Family of odd orthogonal GZ polytopes (as defined in [1, 20]) consists of
two subfamilies parameterized by integer and half-integer λ. The group S O2n+1 (C)
is not simply connected, and half-integer weights correspond to the characters of
the maximal torus in the universal cover Spin(2n + 1). If we define the polytope
ring R SG Z using the first subfamily we get a subring of H ∗ (S O2n+1 /B, Z) generated
by the first Chern classes of line bundles L λ corresponding to the characters λ of the
maximal torus in S O2n+1 (C).

Example 5 The polytope OG Z λ ⊂ R4 for Sp4 (C) is given by the same 8 inequali-
ties as in Example 3. However, its volume polynomial is computed using a different
volume form chosen so that the covolume of L B is 1. Since Z4 ⊂ L B has index 4,
we get vol OG Z = 4 vol SG Z .
There is an exceptional isomorphism Sp4 (C)/±1  S O5 (C). In particular, flag
varieties in types B2 and C2 are the same. This isomorphism takes the dominant
weight λ = (λ1 , λ2 ) of Sp4 (C) to the dominant weight  λ = (λ1 + λ2 )/2, (λ1 −
λ2 )/2) of S O5 (C). This agrees with the identity vol(SG Z λ ) = vol(OG Zλ ).

11.2.3 Newton–Okounkov Polytopes of Flag Varieties

We recall a definition of Newton–Okounkov convex bodies in the case of flag vari-

eties. We refer the reader to [12, 19] for definitions in the more general setting.
Recall that the complete flag variety S L n (C)/B is defined as the variety of com-
plete flags of subspaces M • = ({0} ⊂ V 1 ⊂ V 2 ⊂ · · · ⊂ V n−1 ⊂ Cn ). We define
S On (C)/B and Sp2n /B as subvarieties of orthogonal and isotropic flags in S L n (C)/B
and S L 2n /B, respectively. A complete flag M • in Cn is orthogonal if V i is orthogonal
240 V. Kiritchenko and M. Padalko

to to V n−i with respect to a non-degenerate symmetric bilinear form fixed by S On (C).

Let ω be a non-degenerate skew-symmetric bilinear form fixed by Sp2n (C). A
complete flag M • in C2n is called isotropic if the restriction of ω to V n is zero,
and V 2n−i = {v ∈ C2n | ω(v, u) = 0 for all u ∈ V i }.
Every flag variety X of dimension d has an open dense subset C (open Schubert
cell) isomorphic to the affine space Cd . It can be constructed as follows. Fix a com-
plete flag F • := (F 1 ⊂ F 2 ⊂ · · · ⊂ F n−1 ⊂ Cn ) such that F • ∈ X (this amounts to
fixing a Borel subgroup B ⊂ G). Also fix a basis e1 ,…, en in Cn compatible with F •
(or a maximal torus in B), that is, F i = e1 , . . . , ei . The open Schubert cell C with
respect to F • is defined as the set of all flags M • that are in general position with
the standard flag F • , i.e., all intersections M i ∩ F j are transverse. Let x1 , …, xd be
coordinates on the open Schubert cell C.

Example 6 In type A, we can identify the open Schubert cell C with an affine
space Cd (for d = n(n − 1)/2) by choosing for every flag M • a basis v1 ,…, vn in Cn
of the form:
v1 = en + x1n−1 en−1 + · · · + x11 e1 ,

v2 = en−1 + x2n−2 en−2 + · · · + x21 e1 , . . . , vn−1 = e2 + xn−1

1
e1 , vn = en ,

so that M i = v1 , . . . , vi . Such a basis is unique, hence, the coefficients (x ij )i+ j<n

are coordinates on the open cell. In other words, every flag M • ∈ C gets identified
with a triangular matrix: ⎛ 1 1 ⎞
x1 x2 . . . xn−1
1
1
⎜ x12 x22 . . . 1 0⎟
⎜ ⎟
⎜ .. .. ⎟
⎜ . . 0⎟ (F F L V ).
⎜ n−1 ⎟
⎝x 1 ... 0 0 ⎠
1
1 0 0 0

Similar coordinates can be introduced on flag varieties in other types.

Let V ⊂ C(X ) = C(x1 , . . . , xd ) be a finite-dimensional subspace of rational func-

tions on X . Our main examples are spaces of global sections H 0 (L λ , X )  Vλ∗ of line
bundles on X . We fix a section s0 ∈ H 0 (L λ , X ), and identify sections s ∈ H 0 (L λ , X )
with rational functions f = ss0 ∈ C(X ).

Example 7 (Example 6 continued) If

λ = (1, . . . , 1, 0 . . . , 0),
     
k n−k

then Vλ∗ can be identified with the subspace of C(x ij )i+ j<n spanned by the minors of
the n × k matrix formed by the first k columns of the matrix (F F L V ). These minors
are exactly the Plücker coordinates of the Grassmannian G(k, n) in the Plücker
embedding. The map X → H 0 (L λ , X )∗ is the composition of the projection X →
11 Schubert Calculus on Newton–Okounkov Polytopes 241

G(k, n) (obtained by forgetting all subspaces in the flag M • except for the V k ) and
the Plücker embedding of G(k, n).
To assign the Newton–Okounkov convex body to V we need an extra ingredient.
Choose a translation-invariant total order on the lattice Zd (e.g., we can take the
lexicographic order). Consider a map

v : C(x1 , . . . , xd ) \ {0} → Zd ,

that behaves like the lowest order term of a polynomial, namely: v( f + g) ≥

min{v( f ), v(g)} and v( f g) = v( f ) + v(g) for all nonzero f, g. Recall that maps
with such properties are called valuations.
Definition 8 The Newton–Okounkov convex body v (X, V ) is the closure of the
convex hull of the set
∞  
v( f )
| f ∈ V k ⊂ Rd .
k=1
k

By V k we denote the subspace spanned by the kth powers of the functions from V .
Example 9 Using coordinates of Example 6 we can define the valuation v as fol-
lows. Order the coefficients (x ij )i+ j<n of the matrix (F F L V ) by starting from col-
umn (n − 1) and going from top to bottom in every column and from right to left
along columns. Then v (X, Vλ∗ ) coincides with the Feigin–Fourier–Littelmann–
Vinberg polytope F F L V (λ) [17]. Moreover, the inclusion F F L V (λ) ⊂ v (X, Vλ∗ )
follows from a straightforward computation of the valuation v on the minors of the
matrix (F F L V ) (see [17, Example 2.9] for more details).
Different valuations might yield different Newton–Okounkov convex bodies. In
particular, GZ polytopes can also be obtained as Newton–Okounkov polytopes of
flag varieties [6, 11]. Okounkov made the first explicit computation of this kind,
namely, he exhibited symplectic GZ polytopes as Newton–Okounkov polytopes of
the isotropic flag varieties [22].

11.3 Geometric Mitosis

In [15], convex geometric analogs of Demazure (or divided difference) operators

are defined on convex polytopes and used to construct DDO polytopes that have the
same properties as Newton–Okounkov polytopes of flag varieties. In [16], operations
on faces of a DDO polytope (geometric mitosis) are defined that yield positive pre-
sentations of Schubert cycles by faces. Here we define the same operations in more
combinatorial terms using a vertex cone instead of a DDO polytope. We refer the
reader to [15, Theorem 3.6], [16, Proposition 2.5] for connections with representation
theory and convex geometry.
242 V. Kiritchenko and M. Padalko

Fig. 11.2 Mitosis as explained in Example 10

Example 10 Figure 11.2 illustrates the idea of mitosis in the simplest example. The
trapezoid and rectangle on the left picture have the same number of lattice points
with given sum of coordinates. The same is true for the right picture. However, the
trapezoid on the right picture becomes a virtual polytope (in particular, lattice points
marked with circles have to be counted with the zero coefficient) while the rectangle
remains a true polytope. There is a price to pay: the left vertical edge of the trapezoid
corresponds to two edges of the rectangle (that is, a single edge of the trapezoid has
the same number of lattice points as two edges of the rectangle). In short, mitosis
preserves positivity at the cost of more involved combinatorics.
Consider a vector space with the direct sum decomposition

Rd = Rd1 ⊕ · · · ⊕ Rdr ,

and choose coordinates x = (x11 , . . . , xd11 ; . . . ; x1r , . . . , xdrr ) with respect to this
decomposition. Let C ⊂ Rd be a convex polyhedral cone with the vertex at the ori-

gin 0. Assume that C is given by inequalities either of type x ij ≤ ax ij  where a > 0
and i = i  or of type 0 ≤ x ij . In what follows, we use the bijective correspondence

between facets of C and inequalities, namely, every inequality x ij ≤ ax ij  defines the

facet H (i, j; i  , j  ) given by the equation x ij = ax ij  , and every inequality 0 ≤ x ij
defines the facet H (0, 0; i, j) given by the equation x ij = 0.
In addition, assume that C does not contain any rays parallel to the x ij -axis
unless j = 1. Then the geometric mitosis of [15, Sect. 5.1] can be defined on faces
of C. Below we describe the resulting mitosis operations M1 ,…, Mr from a combi-
natorial viewpoint.
Let  be a face of the cone C of codimension . The ith mitosis operation Mi
applied to  will produce a collection Mi () (possibly empty) of faces of C. Choose
a minimal subset of facets H1 ,…, H of C such that  = H1 ∩ · · · ∩ H . If none
of these facets coincides with H ( p, q; i, di ) for some p and q, then Mi () = ∅.
Otherwise, let s be the smallest number such that the subset {H1 , . . . , H } contains
facets of type H (·, ·; i, j) for all j = s, s + 1, …, di . For brevity, we label these
facets by H + (i, s), H + (i, s + 1), …, H + (i, di ). For every j = s + 1, s + 2,…, di ,
we now label by H+ (i, j) the facet of type H (i, j; ·, ·). If there are two such
11 Schubert Calculus on Newton–Okounkov Polytopes 243

p
facets H (i, j; p, q) and H (i, j; p  , q  ), and xq ≤ xq  everywhere on  then we
p

put H+ (i, j) := H (i, j; p, q).

Let Ji () ⊂ {s, s + 1, . . . , di } consist of indices j such that H+ (i, j) ∈
/ {H1 , . . . ,
H }. For every j ∈ Ji (), we define an offspring  j ∈ Mi () as the intersection of
facets
 j = H1 ( j) ∩ H2 ( j) ∩ · · · ∩ H −1 ( j),

where the set {H1 ( j), . . . , H −1 ( j)} is obtained from the set {H1 , . . . , H } by the
following rule. First, remove the facet H + (i, j). Second, for every k ∈ Ji () such
that k > j replace the facet H + (i, k) by the facet H+ (i, k). Note that dim  j =
dim  + 1.
Definition 11 The ith mitosis operation Mi sends  to

Mi () = { j | j ∈ Ji ()}.

11.3.1 Type A: GZ Polytopes

Let C be the vertex cone of the GZ polytope in type A for the vertex a =
(λ2 , . . . , λn ; λ3 , . . . , λn ; . . . ; λn ) (see table (G Z A )). After an affine change of coor-
dinates x = z − a the inequalities that define C can be written as follows:

0 ≤ x11 ; 0 ≤ x21 ≤ x12 ; . . . ; 0 ≤ xn−1

1
≤ xn−2
2
≤ · · · ≤ x1n−1 .

The cone C has d = n(n−1) 2

facets: H (0, 0; 1, i) for 1 ≤ i ≤ (n − 1) and H (i −
1, j + 1; i, j) for 2 ≤ i ≤ (n − 1), 1 ≤ j ≤ n − i. In particular, we have the follow-
ing identifications of facets:

H (0, 0; 1, i) = H + (1, i), H (i − 1, j + 1; i, j) = H + (i, j) = H+ (i − 1, j + 1).

It is convenient to encode a face  of C by an n × n table (pipe dream) filled with + as

follows. The table contains + in cell (i, i + j) iff  ⊂ H (i − 1, j + 1; i, j) and i ≥ 2
or  ⊂ H (0, 0; i, j) and i = 1. In particular, only cells above the main diagonal
might have +. In this notation, mitosis operations M1 , M2 applied to the vertex 0
produce the following faces (only cells (1, 2), (1, 3) and (2, 3) of 3 × 3 tables are
shown since the other cells never contain +):

{0} = + + −→ + −→ + −→
M1 M2 M1
=C
+ +
244 V. Kiritchenko and M. Padalko
⎧ ⎫
⎨ ⎬
{0} −→ + + −→ +
M2 M1 M2
, −→ C
⎩ +⎭

For arbitrary n, the mitosis operations M1 ,…, Mn−1 encoded by tables coincide
with Knutson–Miller mitosis on pipe dreams [18] after reflecting tables in a vertical
line. Instead of the vertex cone C we could take the GZ polytope in type A and
consider mitosis operations on faces that contain the vertex a (so called Kogan
faces). Geometric meaning of the resulting collections of faces is described in [14,
Theorem 5.1, Corollary 5.3]. In particular, the following analog of Kushnirenko’s
theorem holds.
Recall that Schubert subvarieties X w are labeled by the elements of the Weyl
group of G, namely, X w is the closure of the B-orbit Bw B/B, where w is an element
of the Weyl group of G. The Weyl group of G = S L n (C) is the symmetric group Sn .
By s1 ,…, sn−1 we denote the elementary transpositions.
Theorem 12 ([14, Theorem 5.4]) Let X w ⊂ S L n (C)/B be the Schubert subvariety
corresponding to a permutation w ∈ Sn . Let w = s j1 . . . s j be a reduced decompo-
sition of a permutation w ∈ Sn such that ( j1 , . . . , j ) is a subword of (1; 2, 1; 3, 2,
1; . . . ; n − 1, . . . , 1). Let Sw ⊂ G Z λ be the set of all faces produced from the ver-
tex a ∈ G Z λ by applying successively the operations Mn− j ,…, Mn− j1 :

Sw = Mn− j1 · · · Mn− j (a).

Then 
degλ (X w ) = ! Vol().
∈Sw

This implies that the Schubert cycle [X w ] (that is, the cohomology class of X w
in H ∗ (S L n (C)/B, Z)) in the polytope ring RG Z  H ∗ (S L n (C)/B, Z) is represented
by the sum of faces in Sw .

Example 13 For n = 3, we have [X s1 s2 ] = M2 M1 (a) and [X s2 s1 ] = M1 M2 (a). Since

the faces in these two presentations are transverse and their intersection consists of
two edges M1 (a) and M2 (a) we get the identity: [X s1 s2 ] · [X s2 s1 ] = [X s1 ] + [X s2 ] (see
Fig. 11.3).

11.3.2 Type C2−3 : DDO Polytopes

In [16, Example 2.9], the following family of DDO polytopes in R4 = R2 ⊕ R2 is

considered:
0 ≤ x11 ≤ λ1 , x12 ≤ x11 + λ2 , x21 ≤ 2x12 ,
11 Schubert Calculus on Newton–Okounkov Polytopes 245

Fig. 11.3 Faces M1 (a), M2 (a), M2 M1 (a), M1 M2 (a), M1 M2 M1 (a) of the GZ polytope in
type A, n = 3

x21
x21 ≤ x12 + λ2 , 0 ≤ x22 ≤ λ2 , x22 ≤
2
(these polytopes can also be realized as Newton–Okounkov polytopes of the isotropic
flag variety Sp4 /B [16, Proposition 4.1]). The vertex cone C of the vertex 0 is given by
4 homogeneous inequalities: 0 ≤ x11 , 0 ≤ 2x22 ≤ x21 ≤ 2x12 . It is convenient to encode
a face  of C by a (2n − 1) × n table (skew pipe dream) for n = 2 filled with +
as follows (see Sect. 11.3.3 for the general definition of skew pipe dreams). The
table contains + in cell (3 − i, i) (for i = 1, 2) iff  ⊂ H (0, 0; i, i), + in cell (2, 2)
iff  ⊂ H (2, 2; 1, 2) and + in cell (3, 2) iff  ⊂ H (1, 2; 2, 1). There are two mitosis
operations M1 and M2 .

+ + + +
M1 M2 M1 M2
{0} = + + −→ + −→ + −→ −→ =C
+ +
⎧ ⎫
+ ⎨ + +⎬
M2 M1 M2
{0} −→ + + −→ , + −→
⎩ ⎭
+
⎧ ⎫
⎨ ⎬
M2 M1
−→ + , , + −→ C
⎩ ⎭
+

The Weyl group of G = Sp4 (C) is the dihedral group D4 . By s1 , s2 we denote

simple reflections that generate D4 so that s2 corresponds to the longer root. By [16,
Corollary 3.6] we have
Proposition 14 Let X w ⊂ Sp4 (C)/B be the Schubert subvariety corresponding to
a permutation w ∈ D4 . Let w = s j1 . . . s j be a reduced decomposition of a permu-
tation w ∈ D4 such that ( j1 , . . . , j ) is a subword of (1, 2; 1, 2). Let Sw ⊂ G Z λ be
the set of all faces produced from the vertex a ∈ G Z λ by applying successively the
operations M j ,…, M j1 :
246 V. Kiritchenko and M. Padalko

Sw = M j1 · · · M j (0).

Then 
degλ (X w ) = ! Vol().
∈Sw

This example can be extended to DDO polytopes for Sp2n . For n = 3 and the
DDO polytope for (s3 s2 s1 )3 (where s3 is the simple reflection with respect to the
longer root) this was done in [23]. The corresponding family of DDO polytopes
in R9 = R3 ⊕ R3 ⊕ R3 is given by inequalities:

0 ≤ x11 ≤ λ1 ; x12 ≤ λ2 + x11 ; x13 ≤ λ3 + x12 ;

0 ≤ x21 ≤ min{x12 , λ2 }; x22 ≤ min{λ3 + x21 + x13 , 2x13 };

 
1
x23 ≤ min x21 + λ3 , x22 ; x31 ≤ min{x22 , x13 + λ3 , λ3 + x22 − x23 };
2
 
1 2
x32 ≤ min{x31 , x23 + λ3 , 2x23 }; 0 ≤ x33 ≤ min x 3 , λ3 .
2

In particular, the vertex cone at 0 is not simplicial. It is defined by 10 inequalities:

1 2 1 1 1 2
0 ≤ x11 ; 0 ≤ x21 ≤ x12 ; 0 ≤ x33 ≤ x ≤ x3 ≤ x2 ≤ x13 ;
2 3 2 2
1 2 1
x3 ≤ x23 ≤ x22 .
2 2
An analog of Proposition 14 follows easily from [16, Corollary 3.6]. However, com-
binatorics of mitosis becomes more involved as analogs of pipe dreams in this case
have a loop.
Recently, Fujita identified DDO polytopes with certain Nakashima–Zelevinsky
polyhedral realizations of crystal bases [5, Theorem 4.1]. In particular, there are
explicit inequalities for these polytopes in types A, B, C, D and G 2 [5, Example 4.3].
In type A, they coincide with the GZ polytope and in type C2−3 with the polytopes
described in this section. It would be interesting to apply geometric mitosis to these
polytopes in the other cases.

11.3.3 Type C: GZ Polytopes

The combinatorics of C2 example from Sect. 11.3.2 can be extended to Cn in a

different way by using the string cone C for the reduced decomposition w0 =
11 Schubert Calculus on Newton–Okounkov Polytopes 247

(sn sn−1 . . . s2 s1 s2 . . . sn−1 sn ) . . . (s2 s1 s2 )(s1 ) of the longest element in the Weyl group
(here s1 corresponds to the longer root). The corresponding string polytope coincides
with the symplectic GZ polytope after a unimodular change of coordinates [20, Sect.
6]. The cone C is simplicial and is given by d = n 2 inequalities:

0 ≤ x2i ≤ x4i−1 ≤ x6i−2 ≤ · · · ≤ x2i−2

2
≤ xi1 ≤ x2i−3
2
≤ · · · ≤ x5i−2 ≤ x3i−1 ≤ x1i

for all i = 1,…, n. We define symplectic mitosis as the geometric mitosis associ-
ated with the cone C. Combinatorics of the symplectic mitosis is quite simple and
described in detail in [16, Sect. 5.2] using skew pipe dreams. However, arguments
of [16, Corollary 3.6] do not directly yield presentations for Schubert cycles since
the symplectic GZ polytope does not satisfy the necessary conditions. Still computa-
tions for n = 2, 3 suggest that the collections of faces of the symplectic GZ polytope
obtained using symplectic mitosis do represent the corresponding Schubert cycles
in the polytope ring R SG Z . Below we describe a bijection between faces of C and
faces of SG Z λ that we used.
Let v be the vertex of SG Z λ given by equations λs = x ij = ylk for all triples λs , x ij
and ylk such that s = i + j − 1 = k + l. We now define a bijection between those
facets of Pλ that contain v and skew pipe dreams of size n with exactly one +. Recall
that a skew pipe dream of size n is a (2n − 1) × n table whose cells are either empty or
filled with +. Only cells (i, j) with n − j < i < n + j are allowed to have + (see [16,
Sect. 5.2] for more details on skew pipe dreams). Put yi0 := λi for i = 1,…, n. The
facet given by equation x ij = y i−1 j corresponds to the skew pipe dream with + in
cell (i + j − 1, n − i + 1). The facet given by equation y ij = x ij+1 corresponds to
the skew pipe dream with + in cell (2n − i − j + 1, n − i + 1). In what follows,
we denote by H(k,l) the facet whose skew pipe dream under this correspondence
contains + in cell (k, l).
This correspondence between facets and skew pipe dreams with a single + extends
to all faces of the symplectic GZ polytope that contain the vertex v. Namely, the
face Hk1 ,l1 ∩ · · · ∩ Hks ,ls obtained as the intersection of s facets corresponds to the
skew pipe dream that has + precisely in cells (k1 , l1 ),…, (ks , ls ). In particular, the
vertex v corresponds to the skew pipe dream D0 that has + in all (fillable) cells. In
what follows, we denote by FD the face corresponding to a skew pipe dream D.
We now formulate a conjecture. Let w be an element of the Weyl group
of G = Sp2n . Choose a reduced decomposition w = s j1 . . . s j such that it is a sub-
word of (sn sn−1 . . . s2 s1 s2 . . . sn−1 sn ) . . . (s2 s1 s2 )(s1 ).

Conjecture 15 Define the set Sw of faces of the symplectic GZ polytope as follows:

Sw = {FD | D ∈ Mn+1− j · · · Mn+1− j1 (D0 )}

where Mi denotes the ith symplectic mitosis operation. Then


degλ (X w ) = ! Vol(F).
F∈Sw
248 V. Kiritchenko and M. Padalko

This conjecture is verified in the case n = 2 and for certain w in the case n = 3 [24].
Note that the bijection between faces of SG Z λ that contain the vertex v and faces
of the string cone C does not come from the unimodular change of coordinates that
identifies the string polytope and the symplectic GZ polytope. There are might be
piecewise linear transformations (such as the ones used in [17, Sect. 5.2]) that yield
scissor congruence of unions of faces of SG Z λ and faces of another polytope for
which geometric meaning of symplectic mitosis is more transparent.

11.3.4 Type B: GZ Polytopes

Note that the Weyl groups of Sp2n (C) and S O2n+1 (C) are the same. Since the GZ
polytopes for both groups differ only by lattices symplectic mitosis is also a natu-
ral tool for finding presentations of Schubert cycles by faces of OG Z λ in type B.
However, coefficients will be rational rather than integer (with powers of 2 in denom-
inator) because the torsion index of S O2n+1 (C) is a power of 2. Note also that the
volumes of faces of both SG Z λ and OG Z λ should be computed with respect to their
lattices. The difference is already visible in the case n = 2 (see Example 5).

Acknowledgements The study has been partially funded by the Russian Academic Excellence
Project ‘5–100’.

References

1. Berenstein, A.D., Zelevinsky, A.V.: Tensor product multiplicities and convex polytopes in
partition space. J. Geom. Phys. 5(3), 453–472 (1988)
2. Brion, M.: Lectures on the geometry of flag varieties. In: Topics in Cohomological Studies of
Algebraic Varieties, Trends in Mathematics, pp. 33–85. Birkhäuser, Basel (2005)
3. Fang, X., Fourier, G., Littelmann, P.: Essential bases and toric degenerations arising from
birational sequences. Adv. Math. 312, 107–149 (2017)
4. Feigin, E., Fourier, G., Littelmann, P.: Favourable modules: filtrations, polytopes, Newton-
Okounkov bodies and flat degenerations. Transform. Groups 22(2), 321–352 (2017)
5. Fujita, N.: Polyhedral realizations of crystal bases and convex-geometric Demazure operators.
Selecta Math. (N.S.) 25(5), Paper No. 74 (2019)
6. Fujita, N., Oya, H.: A comparison of Newton-Okounkov polytopes of Schubert varieties. J.
Lond. Math. Soc. (2) 96(1), 201–227 (2017)
7. Harada, M., Yang, J.J.: Newton-Okounkov bodies of Bott-Samelson varieties and Grossberg-
Karshon twisted cubes. Michigan Math. J. 65(2), 413–440 (2016)
8. Harada, M., Yang, J.J.: Singular string polytopes and functorial resolutions from Newton-
Okounkov bodies. Illinois J. Math. 62(1–4), 271–292 (2018)
9. Ilyukhina, M.: Schubert Calculus and Geometry of a String Polytope for the Group sp4 . Moscow
State University, Diploma (2012)
10. Kaveh, K.: Note on cohomology rings of spherical varieties and volume polynomial. J. Lie
Theory 21(2), 263–283 (2011)
11. Kaveh, K.: Crystal bases and Newton-Okounkov bodies. Duke Math. J. 164(13), 2461–2506
(2015)
11 Schubert Calculus on Newton–Okounkov Polytopes 249

12. Kaveh, K., Khovanskii, A.G.: Newton-Okounkov bodies, semigroups of integral points, graded
algebras and intersection theory. Ann. of Math. (2) 176(2), 925–978 (2012)
13. Kaveh, K., Villella, E.: On a notion of anticanonical class for families of convex polytopes
(2018). arXiv:1802.06674 [math.AG]
14. Kirichenko, V.A., Smirnov, E.Y., Timorin, V.A.: Schubert calculus and Gelfand-Tsetlin poly-
topes. Uspekhi Mat. Nauk 67(4(406)), 89–128 (2012)
15. Kiritchenko, V.: Divided difference operators on polytopes. In: Schubert calculus—Osaka
2012, Advanced Studies in Pure Mathematics, vol. 71, pp. 161–184. Mathematics Society
Japan, [Tokyo] (2016)
16. Kiritchenko, V.: Geometric mitosis. Math. Res. Lett. 23(4), 1071–1097 (2016)
17. Kiritchenko, V.: Newton-Okounkov polytopes of flag varieties. Transform. Groups 22(2), 387–
402 (2017)
18. Knutson, A., Miller, E.: Gröbner geometry of Schubert polynomials. Ann. of Math. (2) 161(3),
1245–1318 (2005)
19. Lazarsfeld, R., Mustaţă, M.: Convex bodies associated to linear series. Ann. Sci. Éc. Norm.
Supér. (4) 42(5), 783–835 (2009)
20. Littelmann, P.: Cones, crystals, and patterns. Transform. Groups 3(2), 145–179 (1998)
21. Molev, A.I.: Gelfand-Tsetlin bases for classical Lie algebras. In: Handbook of Algebra. vol. 4,
Handbook of Algebra, vol. 4, pp. 109–170. Elsevier/North-Holland, Amsterdam (2006)
22. Okounkov, A.: Multiplicities and Newton Polytopes. In: Kirillov’s seminar on representation
theory, American Mathematical Society Translations: Series 2, vol. 181, pp. 231–244. American
Mathematical Society, Providence, RI (1998)
23. Padalko, M.: Mitosis for symplectic flag varieties. B.Sc. Thesis, National Research University
Higher School of Economics (2015)
24. Padalko, M.: Schubert calculus and symplectic Gelfand–Zetlin polytopes. M.Sc. Thesis,
National Research University Higher School of Economics (2017)
25. Pukhlikov, A.V., Khovanskiı̆, A.G.: The Riemann-Roch theorem for integrals and sums of
quasipolynomials on virtual polytopes. Algebra i Analiz 4(4), 188–216 (1992)
26. Totaro, B.: The torsion index of the spin groups. Duke Math. J. 129(2), 249–290 (2005)
Chapter 12
An Eisenbud–Goto-Type Upper Bound
for the Castelnuovo–Mumford Regularity
of Fake Weighted Projective Spaces

Bach Le Tran

Abstract We will give an upper bound for the k-normality of very ample lattice
simplices, and then give an Eisenbud–Goto-type bound for some special classes of
projective toric varieties.

Keywords k-normality · Toric variety · Very ample lattice simplex ·

Einsenbud-Goto conjecture · Castelnuovo-Mumford regularity

12.1 Introduction

The study of the Castelnuovo–Mumford regularity for projective varieties has been
greatly motivated by the Eisenbud–Goto conjecture [7] which asks for any irreducible
and reduced variety X , is it always the case that

reg(X ) ≤ deg(X ) − codim(X ) + 1?

The Eisenbud–Goto conjecture is known to be true for some particular cases. For
example, it holds for smooth surfaces in characteristic zero [13], connected reduced
curves [8], etc. Inspired by the conjecture, there are also many attempts to give an
upper bound for the Castelnuovo–Mumford regularity for various types of algebraic
and geometric structures [5, 12, 15, 20].
For toric geometry, suppose that (X, L) is a polarized projective toric varieties such
that L is very ample. Then there is a corresponding
 very ample lattice polytope P :=
PL associated to L such that (X, L) = m∈P∩M C · χ m [4, Sect. 5.4]. Therefore,
by studying the k-normality of P (cf. Definition 2), we can obtain the k-normality
and also the regularity of the original variety (X, L). For the purpose of this article,

B. L. Tran (B)
School of Mathematics, University of Edinburgh, James Clerk Maxwell Building, Peter Guthrie
Tait Road, Edinburgh EH9 3FD, UK
e-mail: [email protected]

© Springer Nature Switzerland AG 2022 251

A. M. Kasprzyk and B. Nill (eds.), Interactions with Lattice Polytopes,
Springer Proceedings in Mathematics & Statistics 386,
https://doi.org/10.1007/978-3-030-98327-7_12
252 B. Le Tran

we will focus on the case that X is a fake weighted projective d-space and PL a d-
simplex.
For any fake weighted projective d-space X embedded in Pr via a very ample line
bundle, Ogata [17] gives an upper bound for its k-normality:
 
dim X
k X ≤ dim X + − 1.
2

In this article, we will improve Ogata’s bound by giving a new upper bound for
the k-normality of very ample lattice simplices and show that
 
dim X
reg(X ) ≤ deg(X ) − codim(X ) + . (12.1)
2

Recently, McCullough and Peeva showed some counterexamples to the Eisenbud–

Goto conjecture and that the difference reg(X ) − deg(X ) + codim(X ) can be arbi-
trary large [14, Counterexample 1.8]. However, for any fake weighted projective
space X embedded in Pr via a very ample line bundle, it follows from (12.1)
that reg(X ) − deg(X ) + codim(X ) is bounded above by dim(X )/2. Furthermore,
we will show that the Eisenbud–Goto conjecture holds for any projective toric vari-
ety corresponding to a very ample Fano simplex.

12.2 Background Material

12.2.1 Toric Varieties and Lattice Simplices

We begin this section by recalling the definition of the Castelnuovo–Mumford reg-

ularity:
Definition 1 Let X ⊆ Pr be an irreducible projective variety and F a coherent sheaf
over X . We say that F is k-regular if

Hi (X, F (k − i)) = 0

for all i > 0. The regularity of F , denoted by reg(F ), is the minimum number k such
that F is k-regular. We say that X is k-regular if the ideal sheaf I X of X is k-regular
and use reg(X ) to denote the regularity of X (or of I X ).
As the main object of the article is to find an upper bound for k-normality of very
ample lattice simplices, it is important for us to revisit the definition of k-normality
of lattice polytopes.
Definition 2 A lattice polytope P is k-normal if the map
12 An Eisenbud–Goto-Type Upper Bound for the Castelnuovo–Mumford … 253

P ∩ M + ·
· · + P ∩ M → k P ∩ M
k times

is surjective. The k-normality of P, denoted by k P , is the smallest positive integer k P

such that P is k-normal for all k ≥ k P .
Suppose now that X is a fake weighted projective d-space embedded in Pr via
a very ample line bundle. Then the polytope P corresponding to the embedding
is a very ample lattice d-simplex. Furthermore, codim(X ) = |P ∩ M| − (d + 1),
where M is the ambient lattice, and deg(X ) = Vol(P), the normalized volume of P.
We have a combinatorial interpretation of reg(X ) in terms of k P and deg(P) [21,
Proposition 4.1.5] as follows:

reg(X ) = max{k P , deg(P)} + 1. (12.2)

From this, we obtain a combinatorial form of the Eisenbud–Goto conjecture: for very
ample lattice polytope P ⊂ MR , is it always true that

max{deg(P), k P } ≤ Vol(P) − |P ∩ M| + d + 1?

The first inequality was confirmed to be true recently [11, Proposition 2.2]; namely,

deg(P) ≤ Vol(P) − |P ∩ M| + d + 1. (12.3)

Therefore, in order to verify the Eisenbud–Goto conjecture for the polarized toric
variety (X, L), it suffices to check if

k P ≤ Vol(P) − |P ∩ M| + d + 1. (12.4)

12.2.2 Ehrhart Theory

We now recall some basic facts about Ehrhart theory of polytopes and the definition
of their degree.
Let P be a lattice polytope of dimension d. We define ehr P (k) = |k P ∩ M|, the
number of lattice points in k P. Then from Ehrhart’s theory [6, 19],
∞
h ∗P (t)
Ehr P (t) = ehr P (k)t k =
k=0
(1 − t)d+1

for some polynomial h ∗P ∈ Z≥0 [t] of degree less than or equal to d. Let h ∗P (t) =

d ∗ i
i=0 h i t . We have
254 B. Le Tran

d
h ∗0 = 1, h ∗1 = |P ∩ M| − d − 1, h ∗d = |P 0 ∩ M|, and h i∗ = Vol(P).
i=0

Definition 3 ([1, Remark 2.6]) Let P be a lattice polytope of dimension d. We define

the degree of P, denoted by deg(P), to be the degree of h ∗P (t). Equivalently,

d if |P 0 ∩ M| = 0,
deg(P) = 
min i ∈ Z≥0 |(k P) ∩ M = ∅ for all 1 ≤ k ≤ d − i otherwise.
0

12.3 k-Normality of Very Ample Simplices

The following lemma by Ogata is crucial to the main result of this article:
Lemma 4 ([17, Lemma 2.1]) Let P = conv(v0 , . . . , vd ) be a very ample lattice n-
simplex. Suppose that k ≥ 1 is an integer and x ∈ k P ∩ M. For any i = 0, . . . , d,
we have

2k−1
x + (k − 1)vi = uj
j=1

for some u j ∈ P ∩ M.
Using the ideas in [17, Lemma 2.5], we generalize the above lemma as follows.
Lemma 5 Suppose that P = conv(v0 , . . . , vd ) is a very ample
d-simplex. Let k ∈
d
N≥1 . Then for any x ∈ k P ∩ M, a0 , . . . , ad ∈ Z≥0 such that i=0 ai = k − 1, we
have
d
2k−1
ai vi + x = ui
i=0 i=1

for some u i ∈ P ∩ M.
Proof We will use induction in this proof. The case k = 1 is trivial. Suppose that
the lemma holds for k = s − 1. We will now
show that it holds for k = s; i.e., for
d
any x ∈ s P ∩ M, a1 , . . . , ad ∈ Z≥0 such that i=0 ai = s − 1, we have

d
2s−1
ai vi + x = ui (12.5)
i=0 i=1

for some u i ∈ P ∩ M. Without loss of generality, we can take a0 to be positive and

move v0 to the origin. By Lemma 4,

2s−1
(s − 1)v0 + x = wi
i=1
12 An Eisenbud–Goto-Type Upper Bound for the Castelnuovo–Mumford … 255

2s−1
for some wi ∈ P ∩ M. Since v0 = 0, we can write x = i=1 wi . If wi + w j ∈ P ∩
M for any i = j, then we can let ti = w
2i−1 + w2i for i = 1, . . . , s − 1 and have x =
s
t1 + · · · + ts−1 + w2s−1 , which lies in i=1 P ∩ M. Therefore,

d
d
s−1
ai vi + x = ai vi + ti + w2s−1 ,
i=0 i=0 i=1

which satisfies (12.5). Conversely, without loss of generality, suppose that w1 + w2 ∈

/
P ∩ M. Then since x = w1 + w2 + (w3 + · · · + w2s−1 ) ∈ s P ∩ M, we have y :=
w3 + · · · + w2s−1 ∈ (s − 1)P ∩ M and v0 + x = w1 + w2 + y. Using the induction
hypothesis,
d
2(s−1)−1
(a0 − 1)v0 + ai vi +y = wi
i=1 i=1
  

d
a0 −1+ i=1 ai =s−2

for some wi ∈ P ∩ M. It follows that

d
d
ai vi + x = v0 + x + (a0 − 1)v0 + ai vi
i=0 i=1

d
= w1 + w2 + y + (a0 − 1)v0 + ai vi
i=1

2(s−1)−1
= w1 + w2 + wi .
i=0

The conclusion follows. 

Now define the invariants d P and ν P as in [21, Definition 2.2.8]:
Definition 6 Let P be a lattice polytope with the set of vertices V = {v0 , . . . , vn−1 }.
We define d P to be the smallest positive integer such that for every k ≥ d P ,

(k + 1)P ∩ M = P ∩ M + k P ∩ M.

We also define ν P to be the smallest positive integer such that for any k ≥ ν P ,

(k + 1)P ∩ M = V + k P ∩ M.

Notice that for P an n-simplex, d P ≤ ν P ≤ n − 1.

Proposition 7 Let P = conv(v0 , . . . , vd ) be a very ample d-simplex. Then

k P ≤ ν P + d P − 1.
256 B. Le Tran

Proof For any k ≥ d P + ν P − 1 and p ∈ k P ∩ M, by the definition of d P and ν P ,

we have
ν
P −d P
d
p=x+ ui + ai vi (12.6)
i=1 i=0

d
for some x ∈ d P P ∩ M, u i ∈ P ∩ M, i=0 ai = k − ν P . By assumption, k − ν P ≥
d P − 1, so it follows from Lemma 5 that

d d P +k−ν
P
x+ ai vi = u i (12.7)
i=0 i=1

for some u i ∈ P ∩ M. Substitute (12.7) into (12.6), we have

ν
P −d P d P +k−ν
P
p= ui + u i .
i=1 i=1

The conclusion follows. 

Remark 8 This bound is stronger than [17, Proposition 2.4] since ν P ≤ d [21,
Proposition 2.2] and d P ≤ d/2 [17, Proposition 2.2].

12.4 Eisenbud–Goto-Type Upper Bound for Very Ample

Simplices

Suppose that P is a very ample simplex. If P is unimodularly equivalent to the stan-

dard simplex d = conv(0, e1 , . . . , ed ) then (12.4) holds. Now consider the case P
is not unimodularly equivalent to d .
The following lemma is a rewording of [9, Proposition IV.10] to fit our purpose.
We provide a proof for the sake of completeness.
Lemma 9 Let V = {v0 , . . . , vd } and suppose that P = conv(V) is a lattice simplex
not unimodularly equivalent to d . Then deg(P) ≥ ν P .

Proof Since ν P ≤ d, it suffices to show that for any d ≥ k ≥ deg(P),

V + k P ∩ M  (k + 1)P ∩ M.

d
Indeed, any x ∈ (k + 1)P ∩ M can be written as x = i=0 ai vi such that ai ≥ 0

d
d
and i=0 ai = k + 1. If ai < 1 for all i, then d > k and the point i=0 (1 − ai )vi
is an interior lattice point of (d − k)P, a contradiction since d − k ≤ d − deg(P).
Hence, ai ≥ 1 for some i, say a0 ≥ 1. Then
12 An Eisenbud–Goto-Type Upper Bound for the Castelnuovo–Mumford … 257
 

d
d
x = v0 + (a0 − 1)v0 + ai vi = v0 + (a0 − 1)v0 + ai vi ∈ V + k P ∩ M.
i=1 i=1

Hence, k ≥ ν P . The conclusion follows. 

Proposition 10 Let P = conv(v0 , . . . , vd ) be a very ample simplex. Then
 
d
k P ≤ Vol(P) − |P ∩ M| + d + .
2

Proof From Proposition 7, (12.3), and Lemma 9,

k P ≤ d P + ν P − 1 ≤ d P + deg(P) − 1
≤ d P + Vol(P) − |P ∩ M| + d.

By [17, Proposition 2.2], d P ≤ d2 . Therefore, since k P , Vol(P), and |P ∩ M| are all

integers,  
d
k P ≤ Vol(P) − |P ∩ M| + d + .
2


Remark 11 We show some cases that the result of Proposition 10 is stronger
than [17, Proposition 2.4]:
1. Vol(P) ≤ |P ∩ M| + 2. In this case,
   
d d
Vol(P) − |P ∩ M| + d + ≤d+ − 2.
2 2

Example 12 Let d be the standard d-simplex. Then

     
d d d
Vol(d ) − |d ∩ M| + d + = 1 − (d + 1) + d + = .
2 2 2
d 
This is clearly a better bound compared to d + 2
− 1.
2. P 0 ∩ M = ∅ or equivalently deg(P) ≤ d − 1. Indeed, in this case,
 
d
k P ≤ d P + deg(P) − 1 ≤ + d − 2.
2

We will show in next section that this is the only case that we still need to consider
in order to verify the Eisenbud–Goto conjecture for very ample simplices.
Example 13 Consider P = 2d for d ≥ 4, where d is the standard d-simplex.
Then deg(P) = 2 and by Proposition 7,
258 B. Le Tran
   
d d
k P ≤ dP + 1 ≤ +1< + d − 1.
2 2

Theorem 14 Suppose that X is a fake weighted projective space embedded in Pr

via a very ample line bundle. Then

dim(X )
reg(X ) ≤ deg(X ) − codim(X ) + .
2

Proof Let P be the corresponding polytope of the embedding. From (12.2), (12.3),
and Proposition 10, it follows that
   
d d
reg(X ) ≤ Vol(P) − |P ∩ M| + d + + 1 = deg(X ) − codim(X ) + .
2 2

12.5 Eisenbud–Goto Conjecture for Non-hollow Very

Ample Simplices

In this section, we will improve the bound of k-normality for non-hollow very ample
simplices.

Definition 15 A lattice polytope is hollow if it has no interior lattice points.

We now show that the inequality (12.4) holds for non-hollow very ample simplices.
Proposition 16 Let P ⊆ MR be a non-hollow very ample lattice d-simplex. Then

k P ≤ Vol(P) − |P ∩ M| + d + 1.

Proof We will consider two cases, namely |P ∩ M| = d + 2 and |P ∩ M| ≥ d + 3.

For the first case, we have the following lemma:

Lemma 17 Suppose that P = conv(v0 , . . . , vd ) is a very ample lattice d-simplex

with u is the only lattice point beside the vertices. Then P is normal. 

Proof Assume that d P ≥ 2. Then there exists a point p ∈ d P P ∩ M such that p

cannot be written as p = x + w for some x ∈ (d P − 1)P ∩ M and w ∈ P ∩ M.
Since P is a simplex, u and p can be uniquely written as

d
d
p= λi vi , λi = d P
i=0 i=0

and
12 An Eisenbud–Goto-Type Upper Bound for the Castelnuovo–Mumford … 259

d
d
u= λi∗ vi , λi∗ = 1,
i=0 i=0

respectively. It follows from the condition of p that λi < 1 for all 0 ≤ i ≤ d and
there exists 0 ≤ i ≤ d such that λi < λi∗ , say i = 0. By Lemma 4,

d
p + (d P − 1)v1 = ai vi + bu
i=0

d
d
for some ai , b ∈ Z≥0 such that b + ai = 2d P − 1. Replacing p by λi vi

d i=0 i=0
and u by i=0 λi∗ vi yields

λ0 = a0 + bλ∗0
λ1 + d P − 1 = a1 + bλ∗1
λ2 = a2 + bλ∗2
..
.
λd = ad + bλ∗d .

Since λ0 < λ∗0 and λi < 1 for all 0 ≤ i ≤ d, it follows that a0 = a2 = · · · = ad = 0

and b = 0. Then p = d P v1 , a contradiction to the choice of p. Therefore, P is normal.


As a consequence, 1 = k P ≤ Vol(P) − |P ∩ M| + d + 1 = Vol(P) − 1. Now we

consider the case |P ∩ M| ≥ d + 3. By the hypothesis, |P ∩ M| − (d + 1) ≥ 2.
Consider the Ehrhart vector h ∗ = (h ∗0 , · · · , h ∗d ) of P. We have

h ∗0 = 1
h ∗1 = |P ∩ M| − d − 1 ≥ 2
h ∗d = |P 0 ∩ M| ≥ 2.

By [10, Theorem 1.1], 2 ≤ h ∗1 ≤ h i∗ for all 1 ≤ i < d. Therefore,

Vol(P) − |P ∩ M| + d + 1 = h ∗0 + h ∗2 + · · · + h ∗d ≥ 1 + 2(d − 1) = 2d − 1.

By [17, Proposition 2.4],

 
d
kP ≤ + d − 1 ≤ 2d − 1 ≤ Vol(P) − |P ∩ M| + d + 1
2

for all d ≥ 3. The conclusion follows. 

260 B. Le Tran

Let us now recall the definition of Fano polytopes:

Definition 18 A Fano polytope is a convex lattice polytope P ⊆ MR such that P 0 ∩
M = {0} and each vertex v of P is a primitive point of M.
From Proposition 16, we obtain the following corollary:

Corollary 19 The Eisenbud–Goto conjecture holds for any projective toric variety
corresponding to a very ample Fano simplex.

12.6 Final Remarks

We start with a remark that Proposition 7 fails in general.

Example 20 ([3]) Consider the polytope P which is the convex hull of the vertices
given by the columns of the following matrix
⎛ ⎞
0100101 1
M = ⎝0 0 1 0 0 1 1 1 ⎠
0 0 0 1 1 1 s s+1

with s ≥ 4. Then d P = ν P = 2, and by [2, Theorem 3.3], k P = s − 1. It is clear

that k P > d P + ν P − 1 for all s ≥ 6.
Furthermore, it can be shown that P cannot be covered by very ample simpli-
cies [21, Proposition 4.3.3]; hence, it is very unlikely that we can apply Proposition 7
to find a bound of the k-normality of generic very ample polytopes.

12.6.1 Hollow Very Ample Simplices

Finally, we would love to see a classification of hollow very ample lattice simplices.
For dimension 2, Rabinotwiz [18, Theorem 1] showed that any such simplex is
unimodularly equivalent to either T p,1 := conv(0, ( p, 0), (0, 1)) for some p ∈ N
or T2,2 = conv(0, (2, 0), (0, 2)). Now we will show a way to obtain some hollow
very ample simplices in any dimension with arbitrary volume.
We recall the definition of lattice pyramids as in [16]:
Definition 21 Let B ⊆ Rk be a lattice polytope with respect to Zk . Then conv(0, B ×
{1}) ⊆ Rk+1 is a lattice polytope with respect to Zk+1 , called the (1-fold) standard
pyramid over B. Recursively, we define for l ∈ N≥1 in this way the l-fold standard
pyramid over B. As a convention, the 0-fold standard pyramid over B is B itself.

Proposition 22 Let P be a lattice polytope. Then the 1-fold pyramid over P is very
ample if and only if P is normal.
12 An Eisenbud–Goto-Type Upper Bound for the Castelnuovo–Mumford … 261

Proof Let Q = conv(0, P × {1}) be the 1-fold pyramid over P. Then it is easy to
see that if P is normal then so is Q. Now suppose that Q is very ample. We have k Q ≥
k P [21, Lemma 4.2.2] and each lattice point of k Q Q ∩ M sits in (t P ∩ M) × {t} for
some 0 ≤ t ≤ k Q . In particular, suppose that (x, t) ∈ (t P ∩ M) × {t} ⊆ k Q Q ∩ M.
Then
t
(x, t) = (u i , 1) + (k Q − t)0
i=1

t
for some u i ∈ P ∩ M. It follows that x = i=1 u i . Hence, P is t-normal for all k Q ≥
t ≥ 1. Since k Q ≥ k P , it follows that P is normal. The conclusion follows. 

From Proposition 22, if we take any (d − 2)-fold pyramid over either T p,1 with p ∈
Z≥1 or T2,2 , which are all normal, then we obtain a hollow normal (hence very
ample) d-simplex with normalized volume p. The Eisenbud–Goto conjecture holds
for these.
Example 23 We give here an example to demonstrate the case that if Q is very
ample but not normal then the 1-fold pyramid over Q is not very ample. Let Q be
the convex polytope given by taking s = 4 in Example 20. Then Q is very ample;
however, the 1-fold pyramid of Q, which is given by the convex hull of
⎛ ⎞
0 0 1 0 0 1 0 11
⎜0 0 0 1 0 0 1 1 1⎟
⎜ ⎟,
⎝0 0 0 0 1 1 1 4 5⎠
0 1 1 1 1 1 1 11

is not very ample.

Acknowledgements We would like to thank Milena Hering for reading the drafts of this article
and for some valuable suggestions.

References

1. Batyrev, V., Nill, B.: Multiples of lattice polytopes without interior lattice points. Mosc. Math.
J. 7(2), 195–207, 349 (2007)
2. Beck, M., Delgado, J., Gubeladze, J., Michał ek, M.: Very ample and Koszul segmental fibra-
tions. J. Algebraic Combin. 42(1), 165–182 (2015)
3. Bruns, W., Gubeladze, J.: Polytopes, Rings, and K -Theory. Springer Monographs in Mathe-
matics. Springer, Dordrecht (2009)
4. Cox, D.A., Little, J.B., Schenck, H.K.: Toric Varieties. Graduate Studies in Mathematics, vol.
124. American Mathematical Society, Providence (2011)
5. Derksen, H., Sidman, J.: A sharp bound for the Castelnuovo-Mumford regularity of subspace
arrangements. Adv. Math. 172(2), 151–157 (2002)
6. Ehrhart, E.: Sur les polyèdres rationnels homothétiques à n dimensions. C. R. Acad. Sci. Paris
254, 616–618 (1962)
262 B. Le Tran

7. Eisenbud, D., Goto, S.: Linear free resolutions and minimal multiplicity. J. Algebra 88(1),
89–133 (1984)
8. Giaimo, D.: On the Castelnuovo-Mumford regularity of connected curves. Trans. Amer. Math.
Soc. 358(1), 267–284 (2006)
9. Hering, M.S.: Syzygies of toric varieties. ProQuest LLC, Ann Arbor, MI (2006). Thesis
(Ph.D.)–University of Michigan
10. Hibi, T.: A lower bound theorem for Ehrhart polynomials of convex polytopes. Adv. Math.
105(2), 162–165 (1994)
11. Hofscheier, J., Katthän, L., Nill, B.: Ehrhart theory of spanning lattice polytopes. Int. Math.
Res. Not. IMRN 19, 5947–5973 (2018)
12. Kwak, S.: Castelnuovo regularity for smooth subvarieties of dimensions 3 and 4. J. Algebraic
Geom. 7(1), 195–206 (1998)
13. Lazarsfeld, R.: A sharp Castelnuovo bound for smooth surfaces. Duke Math. J. 55(2), 423–429
(1987)
14. McCullough, J., Peeva, I.: Counterexamples to the Eisenbud-Goto regularity conjecture. J.
Amer. Math. Soc. 31(2), 473–496 (2018)
15. Miyazaki, C.: Sharp bounds on Castelnuovo-Mumford regularity. Trans. Amer. Math. Soc.
352(4), 1675–1686 (2000)
16. Nill, B.: Lattice polytopes having h ∗ -polynomials with given degree and linear coefficient. Eur.
J. Combin. 29(7), 1596–1602 (2008)
17. Ogata, S.: k-normality of weighted projective spaces. Kodai Math. J. 28(3), 519–524 (2005)
18. Rabinowitz, S.: A census of convex lattice polygons with at most one interior lattice point. Ars
Combin. 28, 83–96 (1989)
19. Stanley, R.P.: Decompositions of rational convex polytopes. Ann. Discrete Math. 6, 333–342
(1980)
20. Sturmfels, B.: Equations defining toric varieties. In: Algebraic Geometry—Santa Cruz 1995,
Proceedings of Symposia in Pure Mathematics, vol. 62, pp. 437–449. American Mathematical
Society, Providence (1997)
21. Tran, B.L.: On k-normality and regularity of normal projective toric varieties. Ph.D. thesis,
University of Edinburgh (2018)
Chapter 13
Toric Degenerations in Symplectic
Geometry

Milena Pabiniak

Abstract Toric degeneration in algebraic geometry is a process of degenerating a

given projective variety into a toric one. Then one can obtain information about
the original variety via analyzing the toric one, which is a much easier object to
study. Harada and Kaveh described how one incorporates a symplectic structure into
this process, providing a very useful tool for solving certain problems in symplectic
geometry. Below we present two applications of this method: questions about the
Gromov width, and cohomological rigidity problems.

Keywords Symplectic toric manifold · Bott manifold · Toric degeneration ·

Gromov width

13.1 Introduction

Manifolds and algebraic varieties equipped with a group action are usually better
understood as a presence of an action is a sign of certain symmetries. In particular,
toric varieties form a very well understood class of varieties. These are varieties which
contain an algebraic torus TCn := (C∗ )n as a dense open subset and are equipped with
an action of TCn which extends the usual action of TCn on itself. For more about toric
varieties see, for example, [5, 12]. To understand a given projective variety X one
can try to “degenerate” it to a toric one, i.e., form a family of varieties with generic
member X and one special member some toric variety X 0 . The varieties X and X 0
are closely related and thus one can obtain information about X by studying X 0 .
Moreover, such a degeneration gives a map from X to X 0 which, in certain situations,
preserves some special structures X and X 0 might be equipped with (for example: a
symplectic structure).
One can use the method of toric degenerations to solve problems in symplectic
geometry. In this work we discuss the following two applications:

M. Pabiniak (B)
Mathematisches Institut, Universität zu Köln, Weyertal 86-90, 50931 Köln, Germany
e-mail: [email protected]

© Springer Nature Switzerland AG 2022 263

A. M. Kasprzyk and B. Nill (eds.), Interactions with Lattice Polytopes,
Springer Proceedings in Mathematics & Statistics 386,
https://doi.org/10.1007/978-3-030-98327-7_13
264 M. Pabiniak

1. calculating lower bounds for the Gromov width, i.e. trying to find the largest ball
which can be symplectically embedded into a given symplectic manifold;
2. constructing symplectomorphisms needed for a cohomological rigidity problem
for symplectic toric manifolds. This problem is about checking whether any two
symplectic toric manifolds with isomorphic integral cohomology rings (via an
isomorphism preserving the class of symplectic form) are symplectomorphic.
Recall that an 2n-dimensional symplectic manifold (M, ω) equipped with an
effective Hamiltonian action of an n-dimensional torus T = (S 1 )n is called a symplec-
tic toric manifold. The action being Hamiltonian means that there exists a moment
map1 μ : M → Rn . Such a manifold can be given a complex structure interacting
well with the symplectic one so that one calls ω a Kähler form and (M, ω) a Kähler
manifold. In particular, symplectic toric manifolds are toric varieties in the sense of
algebraic geometry. A theorem of Delzant states that we have a bijection2

{2n-dim compact symplectic

{rational and smooth polytopes in Rn }
toric manifolds}
⇐⇒ up to translations and
up to equivariant
GL(n, Z) transformations.
symplectomorphisms

In this bijection, a manifold corresponds to an image of its moment map, therefore

the associated polytope is often called a moment polytope or a moment image. Not
much is known about a classification of symplectic toric manifolds up to symplec-
tomorphism. The cohomological rigidity mentioned in the second bullet above asks
if such classification might be given by the integral cohomology rings.
In Sects. 13.3 and 13.4 respectively we describe the above problems in detail
and explain how one can use toric degenerations to solve problems of this type. In
particular we prove (rather, outline the proofs of) the following two results, obtained
in projects joint with I. Halacheva, X. Fang, P. Littelmann, and S. Tolman. As to
apply a toric degeneration to (M, ω) one needs ω to be an integral symplectic form,
in both theorems we demand that the symplectic form is integral up to scaling, i.e.
that a[ω] ∈ H∗ (M; Z) for some a ∈ R. To simplify the exposition we slightly abuse
the notation and given a map F defined on H∗ (M; Z) we use F to also denote the
map induced by F on H∗ (M; Z) ⊗Z R.
Theorem 1 ([11, 14]) Let K be a compact connected simple Lie group. The Gromov
width of a coadjoint orbit Oλ through a point λ, integral up to scaling, equipped with
the Kostant–Kirillov–Souriau symplectic form, is at least
   
min{  λ, α ∨  ; α ∨ a coroot and λ, α ∨ = 0}. (13.1)

1 A moment map is a T -invariant map μ : M → Lie(T )∗ = ∼ Rn such that for every X ∈ Lie(T ) it

holds that ι X  ω = dμ where X denotes the vector field on M induced by X and μ X : M → R is
X

defined by μ X ( p) = μ( p), X ..
2 Recall that a polytope in Rn is called rational if the directions of its edges are in Zn . It is called

smooth if for every vertex the primitive vectors in the directions of edges meeting at that vertex
form a Z-basis for Zn .
13 Toric Degenerations in Symplectic Geometry 265

Theorem 2 ([27]) Let (M, ω M ) and (N , ω N ) be Bott manifolds with symplectic

forms integral up to scaling. Moreover, assume that H∗ (M; Q) and H∗ (N ; Q) are iso-
morphic to Q[x1 , . . . , xn ]/x12 , . . . , xn2 . For any ring isomorphism F : H∗ (M; Z) →
H∗ (N ; Z) with F([ω M ]) = [ω N ], there exists a symplectomorphism f : (N , ω N ) →
(M, ω M ) such that the map H∗ ( f ) : H∗ (M; Z) → H∗ (N ; Z) induced by f on inte-
gral cohomology rings is exactly F.

There are other applications of toric degenerations in symplectic geometry. For

example, one can obtain information about Ginzburg–Landau potential function on X
from that of X 0 and thus detect some non-displaceable Lagrangians of X , see [25].

13.2 Toric Degenerations

A toric degeneration of a projective variety X is a flat family π : X → C with

generic fiber X and one special fiber X 0 = π −1 (0), a (not necessarily normal) toric
variety. A construction of such a degeneration of a projective variety X , equipped with
a very ample line bundle satisfying certain conditions, can be found in Anderson [1,
Theorem 2].

Example 3 Using the Plücker embedding,3 view X = Gr (2, C4 ), the Grassman-

nian of 2-planes in C4 , as a subset of CP5 with coordinates {xi j ; 1 ≤ i < j ≤ 4},
consisting of points satisfying

x12 x34 − x13 x24 + x14 x23 = 0.

Consider the subset X ⊂ CP5 × C consisting of points satisfying

x12 x34 − x13 x24 + t x14 x23 = 0,

where t denotes the coordinate in C. Let π : X → C be the restriction to X of

the projection onto the second factor. This family constitutes a toric degeneration
of Gr (2, C4 ). In fact, {xi j } form a SAGBI basis of the homogeneous coordinate ring
of X and this ensures the flatness [8, Theorem 15.17]. Clearly π −1 (1) is Gr (2, C4 ).
Moreover, performing a change of coordinates, one can show that π −1 (t) for t = 0
is also bihomolomorhpic to Gr (2, C4 ). The central fiber, π −1 (0), is described by the
binomial ideal x12 x34 − x13 x24  and thus is a toric variety.

Harada and Kaveh [16] enriched the construction of Anderson by incorporating a

symplectic structure. They start with a smooth projective variety X , of complex

3Recall that the Plücker embedding sends a Grassmannian spanned by vectors v, w ∈ C4 to a

point [x12 : . . . : x34 ] ∈ CP5 with xi j equal to the determinant of the 2 × 2 minor of [v T , w T ]
spanned by rows i and j.
266 M. Pabiniak

dimension n, equipped with a very ample line bundle L, with some fixed Her-
mitian structure. Let L := H0 (X, L) denote the vector space of holomorphic sec-
tions, L : X → P(L ∗ ) the Kodaira embedding and ω = ∗L (ω F S ) the pull back of
the Fubini–Study form, i.e., of the standard symplectic structure on complex projec-
tive spaces. Then (X, ω) is a Kähler manifold. With this data they construct (under
certain assumptions) not only a flat family π : X → C but also a Kähler structure ω̃
on (the smooth part of) X so that (π −1 (1), ω̃|π −1 (1) ) is symplectomorphic to (X, ω).
Moreover, the special fiber X 0 = π −1 (0) obtains a Kähler form, the restriction of ω̃,
defined on its smooth part U0 := (X 0 )smooth , and thus it also obtains a divisor. If X 0 is
normal, then the polytope associated to X 0 and this divisor by the usual procedure of
toric algebraic geometry (see, for example, [5, Chap. 4]) is the closure of the moment
image of the (non-compact) symplectic toric manifold (U0 , ω̃|U0 ). As we will see, this
polytope can be computed by analyzing the behaviour of the holomorphic sections
of L. Here are more details of this procedure.
Denote by L m the image of span  f 1 · . . . · f m ; f i ∈ L in H0 (X, L⊗m ) and by
R = C[X ] = ⊕m≥0 L m the homogeneous coordinate ring of X with respect to the
embedding L . An important ingredient of the construction is a choice of a val-
uation with one dimensional leaves, ν : C(X ) \ {0} → Zn , from the ring C(X ) of
rational functions on X . A precise definition of a general valuation can be found, for
example, in [16, Definition 3.1]. In this paper we only use valuations induced by a
flag of subvarieties and a special case of these, called lowest/highest term valuations
associated to a coordinate system.
Example 4 (Lowest/highest term valuations [16, Example 3.2]) Fix a smooth
point p ∈ X and let (u 1 , . . . , u n ) be a system of coordinates in a neighborhood
of p, meaning that u 1 , . . . , u n are regular functions at p, vanishing at p, and such
that their differentials du 1 , . . . , du n are linearly independent  at p. Then any reg-
ular function at p can be represented as a power series α∈Zn≥0 cα u α . Here by u α ,
with α = (α1 , . . . , αn ) ∈ Zn≥0 , we mean u α1 1 · . . . · u αn n . Choose and fix a total order >
on Zn respecting the addition, for example the lexicographic order. Define a map ν
from the set of functions regular at p to Zn by
  
ν cα u α = min{α; cα = 0},
α∈Zn≥0

and extend it to C(X ) \ {0} by setting ν( f /g) = ν( f ) − ν(g). Then ν is a valuation

with one dimensional leaves, called a lowest term valuation. If one uses max instead
of min in the definition of ν, one obtains a highest term valuation.
Example 5 (Valuations induced by a flag of subvarieties [16, Example 3.3]) Take a
flag of normal subvarieties (called a Parshin point) of X

{ p} = Yn ⊂ . . . ⊂ Y0 = X,

with dimC (Yk ) = n − k and Yk non-singular along Yk+1 . By the non-singularity

assumption there exists a collection of rational functions u 1 , . . . , u n on X such
13 Toric Degenerations in Symplectic Geometry 267

that u k|Yk−1 is a rational function on Yk−1 which is not identically zero and which
has a zero of first order on Yk . Then the lowest term valuation with respect to
the lexicographic order can alternatively be described in the following way: for
any f ∈ C(X ), f = 0, the valuation v( f ) = (k1 , . . . , kn ) where k1 is the order of
vanishing of f on Y1 , k2 is the order of vanishing of f 1 := (u −k
1
1
f )|Y1 on Y2 , etc.
Given such X , L, and ν we form a semigroup S = S(ν, L), in the following way.
Fix a non-zero element h ∈ L and use it to identify L with a subspace of C(X ) by
mapping f ∈ L to f / h ∈ C(X ). Similarly identify L m with a subspace of C(X ) by
sending f ∈ L m to f / h m ∈ C(X ). As any valuation satisfies that ν( f g) = ν( f ) +
ν(g), the set
S = S(ν, L) = {(m, ν( f / h m )) | f ∈ L m \ {0} }
m≥0

is a semigroup with identity (i.e. a monoid). If S is finitely generated, one can con-
struct a toric degeneration whose special fiber is a toric variety Proj C[S] (which is
normal if S is saturated). Moreover we obtain an Okounkov body
 

=
(S) = conv {x/m | (m, x) ∈ S} ⊂ Rn .
m>0

Note that if S is finitely generated, then

is a rational convex polytope. The toric
variety corresponding to
is the normalization of Proj C[S].4
In the following theorem we rephrase several results from [16].
Theorem 6 ([16]) Let L be a very ample Hermitian line bundle on a smooth n-
dimensional projective variety X and ω = ∗L (ω F S ) the induced symplectic form.
Let ν : C(X ) \ {0} → Zn be a valuation with one dimensional leaves, and such that
the associated semigroup S is finitely generated. Then
• There exists a toric degeneration π : X → C with generic fiber X and special
fiber X 0 := Proj C[S], and a Kähler structure ω̃ on (the smooth part of) X such that
(π −1 (1), ω̃|π −1 (1) ) is symplectomorphic to (X, ω) and the closure of the moment
image of symplectic toric manifold (U0 , ω̃|U0 ), where U0 := (X 0 )smooth , is the
Okounkov body
(S). The set U0 contains the preimage of the interior of
(S).
• Moreover, there exists a surjective continuous map φ : X → X 0 that restricts to a
symplectomorphism from (φ −1 (U0 ), ω) to (U0 , ω̃|U0 ).
In particular, if X 0 = Proj C[S] built from S is smooth (thus also normal), then φ −1
(U0 ) = X and therefore φ provides a symplectomorphism between (X, ω) and the
symplectic toric manifold (X
(S) , ω
(S) ) associated to
(S) via Delzant’s construc-
tion.

4 Recall that for a graded algebra A = ⊕∞ j=0 A j the set Proj A is the set of homogeneous prime ideals
in A that do not contain all of A+ := ⊕∞ j=1 A j . The topology on Proj A is defined by setting the
closed sets to be V (I ) := {J ; J ⊂ I is a homogeneous prime ideal of A not containing all of A+ },
for some homogeneous ideal I of A. For more details see, for example [17, II.2], [9, III.2], and [5,
Chap. 7].
268 M. Pabiniak

Checking whether S is finitely generated is a very difficult problem. However, it

was observed by Kaveh [20] that even if S is not finitely generated one can still form
a (not flat) family with generic fiber X and special fiber (C∗ )n . Even though such a
construction provides much less information about X , it still suffices for the purpose
of finding lower bounds on the Gromov width. We describe this idea in Sect. 13.3.

13.3 Gromov Width

The Gromov width of a 2n-dimensional symplectic manifold (X, ω) is the supremum

of the set of the positive real numbers a such that the ball of capacity a (radius πa ),

 a    n


Ba2n = B 2n = (x1 , y1 , . . . , xn , yn ) ∈ R2n  π (xi2 + yi2 ) < a ⊂ (R2n , ωst ),
π
i=1

can be symplectically embedded in (X, ω). Here ωst = j d x j ∧ dy j denotes the
standard symplectic form on R2n . This question was motivated by the Gromov non-
squeezing theorem which states that a ball B 2n (r ) ⊂ (R2n , ωst ) cannot be symplec-
tically embedded into B 2 (R) × R2n−2 ⊂ (R2n , ωst ) unless r ≤ R.
J -holomorphic curves give obstructions to ball embeddings, while Hamiltonian
torus actions can lead to constructions of such embeddings (by extending a Darboux
chart using the flow of the vector field induced by the action).
This is why toric degenerations provide a useful tool for finding lower bounds on
the Gromov width. Given a toric degeneration of (X, ω), as described in Theorem 6,
one can use the toric action on X 0 to construct embeddings of balls into a smooth
symplectic toric manifold (U0 , ω̃|U0 ), where U0 = (X 0 )smooth . Postcomposing such
embedding with the symplectomorphism φ −1 produces a symplectic embedding
into (X, ω).
Moreover, many embeddings of balls into symplectic toric manifolds can be read
off from the associated (by the Delzant classification theorem) polytope. Identify
the dual of the Lie algebra of the compact torus T with Euclidean space using the
convention that S 1 = R/Z, i.e. the lattice of t∗ is mapped to Zdim T ⊂ Rdim T . With
this convention, the moment map for the standard (S 1 )n action on (R2n , ωst ) maps Ba2n
onto an n-dimensional simplex of size a, closed on n sides
⎧ ⎫
⎨ 
n ⎬
Sn (a) := (x1 , . . . , xn ) ∈ Rn | 0 ≤ x j < a, xj < a .
⎩ ⎭
j=1

Moreover, if the moment image contains an open simplex of size a, then for any ε > 0
a ball of capacity a − ε can be embedded into the given symplectic toric manifold:
see [28, Lemma 5.3.1] and, independently, [26, Propositions 2.1 and 2.4].
13 Toric Degenerations in Symplectic Geometry 269

Proposition 7 ([24, Proposition 1.3] and [26, Proposition 2.5]) For any connected,
proper (not necessarily compact) symplectic toric manifold U of dimension 2n, with
a momentum map μ, the Gromov width of U is at least

sup{a > 0 | ∃ ∈ GL(n, Z), x ∈ Rn , such that (int Sn (a)) + x ⊂ μ(U )}.

The appearance of and x comes from the facts that the identification t∗ ∼ = Rdim T
depends on a splitting of T into (dim T ) circles, and that a translation of a moment
map also provides a moment map.
The above results lead to the following method for finding lower bounds on the
Gromov width.
Corollary 8 Let X be a smooth projective variety of complex dimension n, L an
ample line bundle on X , and ω = ∗L (ω F S ) ∈ H2 (X, Z) an integral Kähler form
obtained using the Kodaira embedding L : X → P(L ∗ ). Suppose that there exists
a valuation ν giving a finitely generated and saturated semigroup S = S(ν, L). Let

be the associated Okounkov body. The Gromov width of (X, ω) is at least

sup{a > 0 | ∃ ∈ GL(n, Z), x ∈ Rn , such that (int Sn (a)) + x ⊂

}.

Proof By the result of [16] cited here as Theorem 6, there exists a toric degenera-
tion of (X, ω) to a normal toric variety X 0 = Proj C[S], and a surjective continuous
map φ : X → X 0 whose appropriate restriction is a symplectomorphism. The sub-
set U := φ −1 (U0 ) of X inherits a toric action whose moment image contains int
,
the interior of
(recall that a moment map sends singular points of a toric variety to
the boundary of the moment polytope). The corollary follows from Proposition 7. 
In fact one does not need S to be saturated. The same corollary holds even if X 0
is not a normal toric variety. This is because a normalization map for X 0 induces a
biholomorphism between (X 0 )smooth and an appropriate subset of the normalization
of X 0 .
It is, however, necessary that S is finitely generated for a toric degeneration to
exist. Otherwise one can still form a family of manifolds, but one cannot guarantee
that this family is flat, and thus X and X 0 are no longer so strongly related. As we
already mentioned, Kaveh in [20] observed that such a (not necessarily flat) family,
with X 0 = (C∗ )n , still provides information about the Gromov width of (X, ω).
To state this result we need additional notation. In the notation of Sect. 13.2, for
any m ∈ Z>0 let

1
Am := {ν( f / h m ) | f ∈ L m \ {0} } ⊂ Zn ,
m = conv(Am ).
m

Note that
= ∪m>0
m . Fix m and let r = rm denote the number of elements
in Am = {β1 , . . . , βr }. From these data we form a symplectic form, ωm , on (C∗ )n
using a standard procedure: ωm is the pull back of the Fubini–Study form on CPr −1 via
the map m : (C∗ )n → CPr −1 , u → (u β1 c1 , . . . , u βr cr ), where c = [(c1 , . . . , cr )] is
270 M. Pabiniak

some element in CPr −1 with all ci = 0. (In [20] the elements ci come from coeffi-
cients of leading terms of elements in appropriately chosen basis of L m . One also
needs that the differences of elements in Am span Zn , which, by [20, Remark 5.6],
is always true for lowest term valuations.)
Kaveh proved that:
1. for every m > 0 there exists an open subset U ⊂ X such that (U, ω) is symplec-
tomorphic to ((C∗ )n , m1 ωm ) [20, Theorem 10.5];
2. the Gromov width of ((C∗ )n , m1 ωm ) is at least Rm , where Rm is the size of the
largest open simplex that fits in the interior of
m = m1 conv (Am ) [20, Corollary
12.3].
This leads to the following corollary.
Corollary 9 ([20, Corollary 12.4]) Let X be a smooth projective variety of dimen-
sion n, L an ample line bundle on X , and ω = ∗L (ω F S ) ∈ H 2 (X, Z) an integral
Kähler form. Let ν be a lowest term valuation on C(X ), with values in Zn , and

the associated Okounkov body. The Gromov width of (X, ω) is at least R, where R
is the size of the largest open simplex that fits in the interior of
.

13.3.1 Results About Coadjoint Orbits

The methods for finding the Gromov width described in Corollaries 8 and 9 have
been used in [11, 14] for coadjoint orbits of compact Lie groups.
Recall that given a compact Lie group K each orbit O ⊂ k∗ := (Lie K )∗ of the
coadjoint action of K on k∗ is naturally a symplectic manifold. Namely it can be
equipped with the Kostant–Kirillov–Souriau symplectic form ω K K S defined by:

ωξK K S (X # , Y # ) = ξ, [X, Y ], ξ ∈ O ⊂ k∗ , X, Y ∈ k,

where X # , Y # are the vector fields on k∗ induced by X, Y ∈ k via the coadjoint

action of K . For more details see, for example, [7, Sect. 21.5, Homeworks 16 and
17]. Coadjoint orbits are in bijection with points in a positive Weyl chamber as
every coadjoint orbit intersects such a chamber in a single point. An orbit is called
generic (resp. degenerate) if this intersection point is an interior point of the chamber
(resp. a boundary point). For example, when K = U(n, C) is the unitary group, a
coadjoint orbit can be identified with the set of Hermitian matrices with a fixed set
of eigenvalues. The orbit is generic if all eigenvalues are different, and in this case it
is diffeomorphic to the manifold of complete flags in Cn .
It has been unofficially conjectured5 that the Gromov width of (Oλ , ω K K S ) of K ,
through a point λ in a positive Weyl chamber should be given by the following neat

5 During the work on the project [18], about complex Grassmannians, Karshon and Tolman looked
at various examples of other coadjoint orbits and got the impression that the above value might be
the Gromov width of all coadjoint orbits. They never formulated this expectation formally as their
13 Toric Degenerations in Symplectic Geometry 271

formula, expressed entirely in the Lie-theoretical language

   
min{  λ, α ∨  ; α ∨ a coroot and λ, α ∨ = 0}.

For example, as {eii − e j j ; i = j} forms a root system for the unitary group U(n, C),
the Gromov width of its coadjoint orbit Oλ passing through a point

λ = diag(λ1 , . . . , λn ) ∈ u(n)∗ ,

integral up to scaling, is equal to min{|λi − λ j |; i, j ∈ {1, . . . , n}, λi = λ j }. Here

we identified u(n) and u(n)∗ with the set of n × n Hermitian matrices.
This conjecture was motivated by the computation of the Gromov width of com-
plex Grassmannians, i.e. degenerate coadjoint orbits of U(n, C), done by Karshon
and Tolman [18], and independently by Lu [23]. Later, using holomorphic tech-
niques, Zoghi [29] showed that the above formula provides an upper bound for
the Gromov width for generic indecomposable6 orbits of U(n, C). This result was
generalized to all coadjoint orbits by Caviedes [2]. The fact that this formula also pro-
vides a lower bound was proved using explicit Hamiltonian torus actions by several
authors: [29] gives a proof for generic indecomposable orbits of U(n, C) using the
standard action of the maximal torus, Lane [21] proves this lower bound for generic
orbits of the exceptional group G2 , and [26] settled the case of U(n, C), SO(2n, C)
and SO(2n + 1, C) orbits7 using the Gelfand–Tsetlin torus action.

13.3.2 A Sketch of the Proof of Theorem 1

The first usage of toric degenerations in Gromov width problems appeared in [14],
where the generic orbits of the symplectic group Sp(n) = U(n, H) are considered.
Then it was used in [11] to prove that the formula (13.1) is a lower bound for the
Gromov width of any coadjoint orbit of any compact connected simple Lie group K ,
passing through a point in the Weyl chamber, integral up to scaling, i.e. to prove
Theorem 1.
The rationality assumption comes from the fact that the toric degeneration method
can be applied only to the orbits passing through an integral point λ of a positive Weyl

conjecture, but they shared this idea with other mathematicians in private communications. This is
how this value became to be known as the expected Gromov width for coadjoint orbits.
6 A coadjoint orbit through a point λ in the interior of a chosen positive Weyl chamber is called

indecomposable in [29] if there exists a simple positive root α such that for any positive root α 
there exists a positive integer k such that λ, α   = kλ, α.
7 The result about SO(2n + 1, C) holds only for orbits satisfying a mild technical condition: the

point λ of intersection of the orbit and a chosen positive Weyl chamber should not belong to a
certain subset of one wall of the chamber; see [26] for more details. In particular, all generic orbits
satisfy this condition.
272 M. Pabiniak

chamber, i.e., in the language of representation theory language, through a dominant

weight.
Let G be a simply connected simple complex algebraic group and K ⊂ G be
its maximal compact subgroup. With a dominant weight λ one can associate an
irreducible representation V (λ) of G of highest weight λ. Let Cvλ be the highest
weight line and P = Pλ be the normalizer in G of this line. Then the coadjoint
orbit Oλ of K is diffeomorphic to G/P (and to K /K ∩ P) and there exists a very
ample line bundle Lλ on G/P such that the pull back of the Fubini–Study form on the
projective space P(H0 (G/P, Lλ )∗ ) = P(V (λ)) via the Kodaira embedding G/P →
P(H0 (G/P, Lλ )∗ ) is exactly the Kostant–Kirillov–Souriau symplectic form ω K K S
on Oλ (see for example [2, Remark 5.5]). Thus for integral λ’s one can try to construct
toric degenerations of projective variety G/P with line bundle Lλ and obtain some
lower bounds for the Gromov width of the orbit Oλ . Rescaling of symplectic forms
allows to extend such a result to orbits Oaλ , for any a ∈ R>0 .
It remains to discuss how one can construct desired toric degenerations.
A great advantage of working with coadjoint orbits of a complex algebraic group G
is that a lot of information can be obtained from studying representations of G. This
leads to a beautiful interplay between symplectic geometry and representation theory.
A reduced decomposition of the longest word in the Weyl group, w0 = siα1 · . . . · siα N
provides the following items (defined below) related in an interesting way:
1. a valuation νw0 ;
2. a string parameterization of a crystal basis of Vλ∗ .
We continue to denote by λ a dominant weight (i.e. an integral element in a positive
Weyl chamber of g∗ ) and by Vλ the finite dimensional irreducible representation of G
with highest weight λ. Let Vλ∗ denote the dual representation. One often seeks for a
basis of Vλ∗ consisting of elements which behave nicely under the action of Kashiwara
operators. A crystal basis is a basis whose elements are permuted under the Kashiwara
operators. Given a crystal basis one can form a crystal graph of a given representation:
vertices are elements of the crystal basis and {0}, and edges are labelled by simple
roots and correspond to the action of Kashiwara operators. There are (not canonical)
ways of embedding such graph into R N , N = dimC G/P. A reduced decomposition
of the longest word in the Weyl group (into a composition of reflections with respect
to simple roots), w0 = sα1 · . . . · sα N , provides a way of assigning to each vertex of
the crystal graph a string of N integers (string parametrization), and thus gives such
an embedding. A convex hull of the image of string parametrization is called a string
polytope. It depends on λ and also on the chosen decomposition w0 . String polytopes
have been extensively studied in representation theory.
Moreover, a reduced decomposition w0 = siα1 · . . . · siα N defines a sequence of
Schubert subvarieties
[P] = Y N ⊂ . . . ⊂ Y0 = G/P,

where Y j denotes the Schubert variety corresponding to element siα j+1 · . . . · siα N of
the Weyl group. We denote by νw0 the highest term valuation associated with this
flag of subvarieties.
13 Toric Degenerations in Symplectic Geometry 273

A theorem of Kaveh relates these two objects.

Theorem 10 ([19]) The string parametrization for Vλ∗ = H0 (G/P, Lλ ) obtained
using the reduced decomposition w0 is the restriction of the valuation νw0 and thus
the corresponding string polytope is the Okounkov body
(νw0 ).
Detailed computations for the case of G = S L(3, C) and w0 = s1 s2 s2 are presented
in [19, Sect. 5].
Explicit descriptions of string polytopes for classical Lie groups and some well-
chosen reduced decompositions of the longest words were presented in the work of
Littelmann [22]. With a bit of work one can show that the string polytope for Vλ∗
with G = Sp(2n, C) the symplectic group (with maximal compact subgroup K =
Sp(n) = U(n, H)), described in [22], contains a simplex of size prescribed by (13.1).
Then, the result of Kaveh, [19], quoted above together with Corollary 8 prove that
the Gromov width of Sp(n) coadjoint orbit (Oλ , ω K K S ) is at least equal to the value
prescribed by (13.1), i.e. proves Theorem 1 for the case of the symplectic group.
This is exactly the argument used in [14].
Similar methods could be applied for other classical Lie groups. However, one
would need to consider each type separately, as the descriptions of string polytopes
contained in [22] depend on reduced decompositions which are different for different
group types.
To obtain a unified proof which works for all group types, in [11] we use lowest
term valuations ν arising from a system of parameters induced by an enumera-
tion {β1 , . . . , β N } of certain positive roots, also in the cases where this enumeration
does not come from a reduced decomposition of the longest word. In these cases
it might be very difficult to show that the associated semigroup S(ν) is finitely
generated (if it is) and to find an explicit description of the associated Okounkov
body. Moreover, on the representation theory side, we do not have a natural way of
obtaining a string parametrization of a crystal basis of Vλ∗ from such enumerations.
However, in [10] the authors managed to give a representation-theoretic description
of the associated semigroup S(ν) in the case when the enumeration is a good ordering
in the sense of [10]. Here is the main idea. Given such enumeration one can define
for each α ∈ Z≥0 N
subspaces V (λ)≤α and V (λ)<α of V (λ). An element α ∈ Z≥0 N
is
called essential for λ if dim V (λ)≤α /V (λ)<α = 1. It was proved in [10] that the
set {(l, α); l ∈ Z≥0 , α essential for lλ} is a semigroup which coincides with S(ν).
Moreover, building on other results from [10] concerning essential elements, one
can show that the Okounkov body associated to S(ν) contains a simplex of size
prescribed by (13.1). Then, using the result of [20] cited here as Corollary 9 (which
does not require the semigroup to be finitely generated), one proves Theorem 1. The
details of this argument are presented in [11].

13.4 Cohomological Rigidity

The following section is based on a project joint with Sue Tolman [27].
274 M. Pabiniak

Cohomological rigidity problems are problems where one tries to determine

whether the integral cohomology ring can distinguish between manifolds of cer-
tain family, and whether all isomorphisms between integral cohomology rings are
induced by maps (homeomorphisms or diffeomorphisms, depending on the setting)
between manifolds. The integral cohomology ring is too weak to distinguish a home-
omorphism type of a manifold. However, by a result of Freedman, it classifies (up to
a homeomorphism) all closed, smooth, simply connected 4-manifolds. Masuda and
Suh posed a question of whether the cohomological rigidity holds for the family
of symplectic toric manifolds. The question was studied by its authors, Choi, and
Panov. No counterexample was found and partial positive results were proved. (Inter-
ested reader is encouraged to consult a nice survey [4] and references therein.) Due
to the presence of symplectic structure, it seems natural to consider the following
symplectic variant of the above question.
Question 11 (Symplectic cohomological rigidity for symplectic toric manifolds)
1. (weak) Are two symplectic toric manifolds (M, ω M ) and (N , ω N ) necessar-
ily symplectomorphic whenever there exists an isomorphism F : H∗ (M; Z) →
H∗ (N ; Z) sending the class [ω M ] to the class [ω N ]?
2. (strong) Is any such isomorphism F : H∗ (M; Z) → H∗ (N ; Z) induced by a sym-
plectomorphism?
In [27] it is shown that weak and strong symplectic cohomological rigidity hold
for the family of Bott manifolds with rational cohomology ring isomorphic to that of
a product of copies of CP1 . Bott manifolds can be viewed as higher dimensional gen-
eralizations of Hirzebruch surfaces discussed in the example below. For the definition
see Sect. 13.4.2.
Remark 12 Strong (not symplecic) cohomological rigidity, with diffeomorphisms,
was already proved for this family by Choi and Masuda [3]. Their diffeomorphisms
usually do not preserve the complex structure. If they had, then our result would be
an immediate consequence of theirs. Indeed, if f : N → M is a biholomorphism
inducing F, then ω N and f ∗ (ω M ) are both Kähler forms on N , defining the same
cohomology class in H∗ (N ; Z), and thus in this case (N , ω N ) and (N , f ∗ (ω M )) are
symplectomorphic by the Moser’s trick.
Example 13 (Hirzebruch surfaces) Hirzebruch surfaces are CP1 bundles over CP1 .
As complex manifolds they are classified by integers (encoding the twisting of
the bundle): for each A ∈ Z we denote by H−A the bundle P(O(A) ⊕ O(0)) →
CP1 . In particular, H0 = CP1 × CP1 . They can be equipped with a symplectic
(even Kähler) structure and a toric action. A polytope corresponding to H−A in
Delzant classification is (up to GL(2, Z) action) a trapezoid with outward nor-
mals (−1, 0), (0, −1), (1, 0), (A, 1). The lengths of the edges of this trapezoid
depend on the chosen symplectic structure and can be encoded in λ = (λ1 , λ2 ) ∈
(R>0 )2 . We denote by (H−A , ωλ ) the symplectic toric manifold corresponding to
the trapezoid
(A, λ) := conv((0, 0), (λ1 , 0), (λ1 , λ2 − Aλ1 ), (0, λ2 )). For exam-
ple, Fig. 13.1 presents (H0 , ω(1,3) ) and (H−2 , ω(1,5) ).
13 Toric Degenerations in Symplectic Geometry 275

Fig. 13.1 Hirzebruch

surfaces (H0 , ω(1,3) ) 5
and (H−2 , ω(1,5) )

1 1

It was observed by Hirzebruch that H−A and H− A are diffeomorphic if and only
if A =∼A  mod 2. Moreover, the symplectic toric manifolds (H−A , ωλ ) and (H− A,
ωλ ) are (not equivariantly) symplectomorphic if and only if A ∼ =A  mod 2 and the
widths and the areas of the associated polytopes agree, i.e. λ1 = 
λ1 and λ2 − 21 Aλ1 =

λ2 − 21 Aλ1 . For example, the manifolds presented on Fig. 13.1 are symplectomor-
phic. The cohomology ring can be presented as

H∗ (H−A ; Z) = Z[x1 , x2 ]/x22 , x12 + Ax1 x2 ,

with [ωλ ] = λ1 x1 + λ2 x2 . If A = ∼A  mod 2, then the isomorphism Z[x1 , x2 ] →

Z[x1 , 
x2 ] defined by x1 →   − A)
x1 + 21 ( A x2 , x2 → 
x2 descends to an isomorphism
between H∗ (H−A ; Z) and H∗ (H− A; Z). Note that this isomorphism sends [ωλ ] =

λ1 x1 + λ2 x2 to λ1 x1 + (λ2 + A−A 2
x2 which is equal to [ωλ ] if and only if λ1 = 
λ1 )  λ1

 A
and λ2 − 2 λ1 = λ2 − 2 λ1 . Therefore, for Hirzebruch surfaces (weak) symplectic
A

cohomological rigidity holds.

To approach symplectic cohomological rigidity problem one needs a good method

of constructing symplectomorphisms. Here is where toric degenerations come into
play. By Theorem 6 a toric degeneration whose central fiber Proj C[S] is smooth
produces a symplectomorphism between the symplectic manifold one started with
and the central fiber. The main difficulty in this method of constructing symplecto-
morphisms lies in finding toric degenerations with smooth central fibers.
A great advantage of working with toric manifolds is that the sections of their line
bundles are well understood and one can form very concrete constructions of toric
degenerations.

13.4.1 Toric Degenerations for Symplectic Toric Manifolds

Let (X P , ω P ) be a symplectic toric manifold with ω P ∈ H2 (M, Z), corresponding

to a polytope P ⊂ Rn via Delzant construction. Then P is an integral polytope (i.e.
276 M. Pabiniak

with vertices in Zn ) and there exists a very ample line bundle L over X P inducing ω P .
In this situation a basis of the space of holomorphic sections of L can be identified
with the integral points of P, ([6], see also [15]). Without loss of generality we can
assume that P in a neighborhood of some vertex looks like (R≥0 )n in a neighborhood
of the origin in Rn . Then we can identify L = H0 (X P , L) with a subset of the ring of
rational functions, C(X P ), as described on Sect. 13.2, using the section corresponding
to the origin as the fixed element h:

f
f → .
section corresponding to the origin

Note 14 For simplicity of notation, given a valuation ν we will write ν(L) to denote

ν(L) := {ν( f / h); f ∈ L \ {0}}.

Similarly, let ν(L m ) := {ν( f / h m ); f ∈ L m \ {0}} for any m > 1.

We denote by f j ∈ C(X P ) the rational function coming from the section corre-
sponding to the j-th basis vector, j = 1, . . . , n. Note that f 1 , . . . , f n form a coordi-
nate system around the fixed point of X P corresponding to the origin via the moment
map. To see this, one can, for example, use the description of X P and f j ’s from [15].
Choose and fix a non-negative integer c and two elements k < l ∈ {1, . . . , n}.
Then

{u 1 = f 1 , . . . , u k−1 = f k−1 , u k = f k − flc , u k+1 = f k+1 , . . . , u n = f n }

also gives a coordinate system. Let ν be the associated lowest term valuation (as in
Example 4). The image ν(L) can be obtain by using a “sliding” operator F−ek +cel ,
defined as follows. For each affine line  in Rn in the direction of −ek + cel , with P ∩
 ∩ Zn = ∅, translate the set {P ∩  ∩ Zn } by a(−ek + cel ) with a ≥ 0 maximal non-
negative number for which a(−ek + cel ) + {P ∩  ∩ Zn } ⊂ (R≥0 )n .
Lemma 15 One obtains ν(L) by sliding the integral points of P in the direc-
tion −ek + cel , inside (R≥0 )n , i.e.

ν(L) = F−ek +cel (P ∩ Zn ).

Instead of the proof, which can be found in [27], we give the following example
which ilustrates the main idea behind the proof.
Example 16 Let (X P , ω P ) be the symplectic toric manifold corresponding to the
polytope P = conv {(0, 0), (1, 0), (1, 3), (0, 3)} ⊂ R2 . That is, X P is diffeomorphic
to CP1 × CP1 with product symplectic structure (with different rescaling of the
Fubini–Study symplectic form on each factor). Let ν be the lowest term valuation
associated to the coordinate system

u 1 = f 1 − f 22 , u 2 = f 2 .
13 Toric Degenerations in Symplectic Geometry 277

Line {(0, 2) + t (1, −2); t ∈ R} intersects P in two integral points: (1, 0) and (0, 2).
The corresponding functions are f 1 and f 22 , and one can easily calculate that

ν( f 1 ) = ν( f 22 ) = (0, 2) and ν( f 1 − f 22 ) = (1, 0).

Similarly, using the integral points on the line {(0, 3) + t (1, −2); t ∈ R} we obtain

ν( f 1 f 2 ) = ν( f 23 ) = (0, 3) and ν( f 1 f 2 − f 23 ) = ν(( f 1 − f 22 ) f 2 ) = (1, 1).

More generally, if the integral points (a, b), (a, b) + (1, −2), . . . , (a, b) + m(1, −2)
are in P (implying that b − 2m > 0), then one can use the corresponding functions to
construct functions with valuations (0, b + 2a), (0, b + 2a) + (1, −2), . . . , (0, b +
2a) + m(1, −2) = (m, 2a + b − 2m). Precisely, for any l = 0, . . . , m


l
a+ j b−2 j
f 1a f 2b−2l ( f 1 − f 22 )l = (−1)l− j f 1 f 2 and
j=0

ν( f 1a f 2b−2l ( f 1 − f 22 )l ) = (l, 2a + b − 2l).

This proves that ν(L) ⊃ F(−1,2) (P ∩ Z2 ). By [16, Proposition 3.4] the cardinality
of ν(L) is the dimension of L, that is, the number of integral points in P. Therefore

ν(L) = F(−1,2) (P ∩ Z2 ).

The polytopes P and conv(ν(L)) are presented in Fig. 13.1.

Understanding ν(L) is not enough for constructing and understanding a toric

degeneration. First of all, to construct a flat family with toric fiber π −1 (0) one
needs the associated semigroup S = S(ν) to be finitely generated. Additionally,
this toric fiber π −1 (0) = Proj C[S] is the toric variety associated to the Okounkov
body
if Proj C[S] is normal, that is, if S is saturated. Moreover, to describe the
Okounkov body one also needs to find ν(L m ) for m > 1. Note that in general L m
differs from H0 (X, L⊗m ). The following proposition describes an especially nice
situation where all these conditions simplify.

Proposition 17 Let (X, ω = ∗L (ω F S )) be a 2n dimensional projective symplectic

toric manifold associated to a smooth polytope P, with the projective embedding
induced by a very ample line bundle L. Let ν be a lowest term valuation associated
to a coordinate system of the type presented on Sect. 13.4.1, and S the induced
semigroup. Assume that there exists a smooth integral polytope
⊂ Rn such that

S = (cone ({1} ×
)) ∩ (Z × Zn ).

Then (X, ω) is symplectomorphic to the symplectic toric manifold (X

, ω
) associ-
ated to
via Delzant construction.
278 M. Pabiniak

Here cone ({1} ×

) denotes the set {(t, t x); x ∈
, t ∈ R+ } ⊂ Rn+1 .
Proof (sketch) The assumptions imply that the semigroup S is saturated and (by
Gordan’s Lemma) finitely generated. Therefore there is a toric degeneration (X, ω̃)
with generic fiber (X, ω) and the special fiber π −1 (0) = Proj C[S] which is a normal
toric variety. Moreover, the Okounkov body associated to the semigroup S is pre-
cisely
and therefore Proj C[S] equipped with the restriction of ω̃, is the symplectic
toric manifold (X
, ω
) associated to
via Delzant construction. 
Note that S=(cone
) ∩ (Z × Zn ) imply, in particular, that ν(L m ) contains “enough”
of integral points, namely that

∀ m ≥ 1 ν(L m ) = m
∩ Zn = conv(ν(L m )) ∩ Zn .

To understand better the requirement conv (ν(L m )) ∩ Zn = ν(L m ), consider the

following example.

Example 18 (“Enough” of integral points and saturation) Let (X P , ω P ) be the

symplectic toric manifold corresponding to the polytope

P = conv {(0, 0), (2, 0), (2, 2), (0, 2)} ⊂ R2 ,

that is, X P is diffeomorphic to CP1 × CP1 as in the previous example, but the
symplectic form is different. As before, let ν be the lowest term valuation associated
to the coordinate system
u 1 = f 1 − f 22 , u 2 = f 2 .

Then

ν(L) = F(−1,2) (P ∩ Z2 )
= {(0, j); j = 0, . . . , 6} ∪ {(1, 0), (1, 2)}  conv (ν(L)) ∩ Z2 .

In fact conv (ν(L)) is exactly the associated Okounkov body

(S(ν)). Indeed,

(S(ν)) must contain conv (ν(L)). Moreover, 2! vol2 (

(S(ν))) is the degree of
the Kodaira embedding L : X P → P(L ∗ ) induced by the line bundle L corre-
sponding to ω P [16, Theorem 3.9]. Thus the area of
(S(ν)) must be equal to
the area of P, which in this case is also the area of conv (ν(L)). Therefore, in
our example, ν(L) is “missing” the point (1, 1) in a sense that ν(L) =
(S(ν)) ∩
Z2 \ {(1, 1)}, and thus (1, 1, 1) ∈
/ S(ν). However, the line  {t (1, 1, 1); t ∈ R+ }
intersects S(ν): (2, 2, 2) = 2, ν( f 1 ( f 1 − f 22 ) · ( f 1 − f 22 )) ∈ {2} × ν(L 2 ). There-
fore the semigroup S(ν) is not saturated.

Let us analyse why in the above example the point (1, 1) is missing. Observe that
the parallel lines 1 := {(0, 2) + t (−1, 2); t ∈ R}, 2 := {(0, 3) + t (−1, 2); t ∈ R}
and 3 := {(0, 4) + t (−1, 2); t ∈ R} intersect P at intervals of the same length but
with, respectively, 2, 1 and 2 integral points. Therefore the intersections of 1 , 2
13 Toric Degenerations in Symplectic Geometry 279

Fig. 13.2 Illustration of

Example 18

Integral points of P ν(L)

and 3 with ν(L) = F(−1,2) (P ∩ Z2 ) also contain, respectively, 2, 1 and 2 integral

points. As a result, the points (1, 0) and (1, 2) are in ν(L), but (1, 1) is not. The
following condition is sufficient, though not necessary, to guarantee that we do not
encounter that problem and have enough of integral points (Fig. 13.2).
Lemma 19 Let λ1 , λ2 , c ∈ Z>0 and
    

= p ∈ R2  0 ≤  p, e1  ≤ λ1 , 0 ≤  p, e2  and p, e2 + Ae1 ≤ λ2 .

Assume that
λ2 − cλ1 > 0.

Then the polytope conv F(−1,c) (

∩ Z2 ) is a trapezoid of the same area as
and

(conv F(−1,c) (
∩ Z2 )) ∩ Z2 = F(−1,c) (
∩ Z2 ).

If c ≤ A then conv F(−1,c) (

∩ Z2 ) is simply
, and if c > A then it is
 
p ∈ R2  0 ≤  p, e1  ≤λ1 , 0 ≤  p, e2  and
  
p, e2 + (2c − A)e1 ≤ λ2 + (c − A)λ1 ) .

An example is illustrated in Fig. 13.3.

13.4.2 Cohomological Rigidity for Bott Manifolds

A Bott manifold is a manifold obtained as a total space of a tower of iterated bundles

with fiber CP1 and the first base space CP1 . Such a manifold naturally carry an
algebraic torus action, and can be viewed as a toric manifold. Note that 4-dimensional
Bott manifolds are exactly the Hirzebruch surfaces discussed in Example 13. For
more information about Bott manifolds see, for example, [13].
280 M. Pabiniak

Fig. 13.3 Toric degeneration of a Hirzebruch surface

The simplest example of an 2n-dimensional Bott manifold is the product of n

copies of CP1 ’s. Equipped with a product symplectic structure ω = π1∗ (a1 ω F S ) +
. . . + πn∗ (an ω F S ), for some a j ∈ R>0 , and the standard toric action8 it becomes a
symplectic toric manifold, whose Delzant polytope is a product of intervals, with
lengths depending on a j ’s. Here π j : CP1 × . . . × CP1 → CP1 denotes the projec-
tion onto the j-th factor, and ω F S stands for the Fubini–Study symplectic form. In
particular, if all a j ’s are equal, then the moment image is a hypercube.
A moment image for a general 2n-dimensional Bott manifold is combinatorially
an n-dimensional hypercube. By applying a translation and a GL(n, Z) transforma-
tion one can always arrange that the moment image is a polytope of the form
     

=
(A, λ) = p ∈ Rn   p, e j  ≥ 0 and p, e j + Aij ei ≤ λ j ∀ 1 ≤ j ≤ n ,
i

where A ∈ Mn (Z) is an n × n strictly upper-triangular integral matrix, that is Aij = 0

unless i < j, and λ ∈ (R>0 )n . Certain relation between A and λ must be satis-
fied in order for
(A, λ) to have 2n facets and be combinatorially equivalent to
a hypercube (see [27].) In that case we say that (A, λ) defines a symplectic toric Bott
manifold (M A , ωλ ) corresponding to the Delzant polytope
(A, λ). The matrix A
encodes the twisting of consecutive CP1 bundles, and thus determines a diffeomor-
phism type of M A , while λ determines the symplectic structure. By a classical result
of Danilov [6]
  
H∗ (M A ; Z) = Z[x1 , . . . , xn ]/ xi2 + Aij x j xi , (13.2)
j


with [ωλ ] = i λi xi ∈ H∗ (M A ; Z) ⊗Z R. If all coefficients λi are integral then [ωλ ]
is an integral symplectic. Note that this particular presentation of H∗ (M A ; Z) depends

8In the standard action of (S 1 )n on (CP1 )n each S 1 in (S 1 )n acts on the respective copy of CP1
by eit · [(z 0 , z 1 )] = [(z 0 , eit z 1 )].
13 Toric Degenerations in Symplectic Geometry 281

on
 A. (The  element  x j is the Poincaré dual to the preimage of facet
(A, λ) ∩
{ p, e j + i Aij ei = λ j }.)
We say that a Bott manifold is Q-trivial if H∗ (M; Q)  H∗ ((CP1 )n ; Q). For
example, observe that all Hirzebruch surfaces are Q-trivial Bott manifolds.
Recall that we want to prove Theorem 2 which says that for Q-trivial Bott mani-
folds (N , ω N ) and (M, ω M ), and any ring isomorphism F : H∗ (M; Z) → H∗ (N ; Z),
with F([ω M ]) = [ω N ], there exists a symplectomorphism f : (N , ω N ) → (M, ω M )
inducing F. The key ingredient of the proof of Theorem 2 is the following construc-
tion of symplectomorphisms, which uses toric degenerations.

Proposition 20 ([27]) Let (M, ω) and ( M,   ω) be symplectic Bott manifolds asso-

ciated to strictly upper triangular A and A  in Mn (Z) and λ and  λ in Zn , respec-
tively. Assume that there exist integers 1 ≤ k <  ≤ n so that A and A k k are of the
same parity and the isomorphism from Z[x1 , . . . , xn ] to Z[ x1 , . . . , 
xn ] that sends xk
k −Ak
xi for all i = k descends to an isomorphism from H∗ (M; Z)
A
toxk +  2   x and xi to
 
 Z) and takes λi xi to 
to H∗ ( M; λi  k ≥ 0, then M and M
xi . If Ak + A  are sym-
plectomorphic.

Proof (sketch) Without loss of generality we can assume that the polytope
(A, λ)
associated to (A, λ) is normal, that is, any integral point of m
(A, λ) can be
expressed as a sum of m integral points of
(A, λ). Indeed, if
(A, λ) is not a normal
polytope, replace (M, ω) and ( M,  ω) by (M, (n − 1) ω) and ( M, (n − 1)  ω). This
dilates the corresponding polytopes by (n − 1). For any integral polytope P ⊂ Rn
its dialate m P with m ≥ n − 1 is normal (see, for example, [5, Theorem 2.2.12]).
Obviously if (M, (n − 1) ω) and ( M,  (n − 1) ω) are symplectomorphic, then so

are (M, ω) and ( M, ω). As usually, let L denote the very ample line bundle over M
corresponding to ω and L the space of its holomorphic sections. Note that normality
implies that L m can be identified with H 0 (M, L⊗m ) because a basis for both of these
vector spaces is given by the integral points m
(A, λ) ∩ Zn .
Also without loss of generality we can assume that A k ≥ Ak . Let c = 1 (Ak +
2

A ) ≥ 0. We will work with a lowest term valuation ν associated to the following
k

coordinate system

{u 1 = f 1 , . . . , u k−1 = f k−1 , u k = f k − flc , u k+1 = f k+1 , . . . , u n = f n .}

From Lemma 15 and the normality assumption, for all m ≥ 1 we have that

ν(L m ) = F−ek +cel (m

(A, λ) ∩ Zn ).

To understand F−ek +cel (m

(A, λ) ∩ Zn ) consider the action of F−ek +cel on 2-
dimensional “slices”, that is, the intersections of m
(A, λ) with affine subspaces
which are translations of (ek , el )-planes. Such slices are either empty or are trapezoids
like in Example 16 and Corollary 19, possibly with a cut. A bit tedious computation
shows that
F−ek +cel (m
(A, λ) ∩ Zn ) = m
( A,  λ) ∩ Zn .
282 M. Pabiniak

For that computation one uses relations between A, λ, A  and λ which are implied
by the facts that
(A, λ) and
( A, λ) are combinatorially hypercubes, and by the
existence of the isomorphism described in the statement of the proposition. In partic-
ular, these relations also allow to generalize Corollary 19 (precisely: to show that the
equivalent of condition λ2 − cλ1 > 0 holds). Therefore the semigroup S associated
to the valuation ν of (M, ω) is exactly S = (cone
( A,  λ)) ∩ (Z × Zn ). Then the
claim follows from Proposition 17. 

Using Proposition 20 we show below (Corollary 23) that each Q-trivial Bott manifold
is associated to a matrix A of a particularly easy form. To explain this idea we need
few more definitions. Recall the presentation of the cohomology of symplectic Bott
manifold M A given in (13.2). We define the following special elements
 1
αk = − Akj x j ∈ H∗ (M A ; Z), yk = xk − αk ∈ H∗ (M A ; Q)
j
2

for all k. We say xk is of even (odd) exceptional type if αk = cyl for some l, where c is
an even (respectively, odd) integer. In “coordinates”, this means that Akj = 0 for j < l
and Akj = 21 Alk Alj for j > l. Note that if xk is even (resp. odd) exceptional, say αk =
myl , then one can construct an isomorphism of Proposition 20 from H∗ (M A ; Z)
to H∗ (M A; Z) for some A  with Alk equal to 0 (resp. −1). For example if xk is of
even exceptional type, i.e. αk = 2myl for some m and l, implying that Alk = −2m
and Akj = −m Alj for j = l, then one should put A lk = 0, A
ij = Aij for all i and

all j = l, and Al = Al + m Ak for all i = k. Therefore, consecutive applications of
i i i

the above proposition lead to simplifying the description of a given Bott manifold.

Corollary 21 Any symplectic toric Bott manifold, with integral symplectic form is
symplectomorphic to one for which Alk = 0 (resp. Alk = −1) whenever xk has even
(resp. odd) exceptional type and αk = myl .

In the case of Q-trivial Bott manifolds all xi have exceptional type, [3, Proposition
3.1]. Therefore, any Q-trivial symplectic toric Bott manifold with integral symplectic
form must be a product of the following standard models of Q-trivial Bott manifolds.

Example 22 (Q-trivial Bott manifold) Take n ∈ Z>0 . Let Ain = −1 for all 1 ≤
i < n, and Aij = 0 otherwise. For such upper triangular matrix A = [Aij ] and
any λ ∈ (R>0 )n , the polytope
(A, λ) is combinatorially a hypercube, thus it defines
a symplectic toric Bott manifold, which we will denote by H = H(λ1 , . . . , λn ).
Observe that
 
H∗ (H; Z) = Z[x1 , . . . , xn ]/ x12 − x1 xn , . . . , xn−1
2
− xn−1 xn , xn2 .

Consider elements yi = xi − 21 xn ∈ H∗ (H; Q) for all i < n, and yn = xn , and

note that they form a basis for H ∗ (H; Q). Moreover, as yi2 = 0 for all i, we get
that H∗ (H; Q)  Q[y1 , . . . , yn ]/ y12 , . . . , yn2 ), that is, H is Q-trivial.
13 Toric Degenerations in Symplectic Geometry 283

m
More generally, any partition of n, i=1 li = n together with λ ∈ (R>0 )n , define
a Q-trivial Bott manifold

H(λ1 , . . . , λl1 ) × . . . × H(λn−lm +1 , . . . , λn ).

Corollary 23 Each 2n-dimensional Q-trivial Bott manifold M with integral sym-

plectic form is symplectomorphic to

H(λ1 , . . . , λl1 ) × · · · × H(λn−lm +1 , . . . , λn ),

m
for some partition n = i=1 li of n and some λ1 , . . . , λn ∈ Z>0 .

The above standard model is easy enough, so that one can understand all possible
ring isomorphisms between cohomology rings and prove that they are induced by
maps on manifolds.
m m
Lemma 24 Fix n ∈ Z>0 . Let i=1 li = i=1 
li = n be partitions of n, and let λ, 
λ∈
(R>0 )n . Consider symplectic Bott manifolds

(M, ω) = H(λ1 , . . . , λl1 ) × · · · × H(λn−lm +1 , . . . , λn ),

 
( M, λ1 , . . . , 
ω) = H( λl ) × · · · × H(
λn−l +1 , . . . , 
λn ).
1 
m

 Z) such that F[ω] = [

Given a ring isomorphism F : H∗ (M; Z) → H∗ ( M; ω], there
 
exists a symplectomorphism f from ( M, ω) to (M, ω) so that H∗ ( f ) = F.

Proof (sketch) First consider the situation when

 
(M, ω) = H(λ1 , . . . , λn ) and ( M, ω) = H(
λ1 , . . . , 
λn ).

The Q-triviality assumption implies that there are exactly 2n primitive classes
in H2 (M; Z) which square to 0. A short computation shows that these are ±z 1 , . . . ,
±z n , where z n = xn and z i = 2xi − xn for all i < n. Similarly for M.  As the coho-
mology of a symplectic toric manifold is generated in degree 2, any ring isomorphism
between H∗ (M; Z) and H∗ ( M;  Z) restricts to a bijection on the set of such elements,
that is, there exists  = (1 , . . . , n ) ∈ {−1, 1}n and a permutation σ ∈ Sn such
that F(z j ) =  jz σ ( j) . Moreover, presenting [ω] (resp. [ ω]) in R-basis {z 1 , . . . , z n }
of H∗ (M; Z) ⊗Z R (resp. { z 1 , . . . ,
z n }) and recalling that the isomorphism F is to
map [ω] to [ ω], one can deduce that F acts by a permutation: F(z j ) =  z σ ( j) for some
permutation σ ∈ Sn with σ (n) = n, and that λ j =  λσ ( j) . Moreover F takes xi to xσ (i)
and it holds that Aij = A σ (i) for all i, j. If  ∈ GL(n, Z) denotes the unimodular
σ ( j)
matrix taking ei to eσ (i) , then T (
( A,  λ)) =
(A, λ); Therefore, by the Delzant
theorem, the manifolds (M, ω) and ( M,  ω) are (equivariantly) symplectomorphic,
by some symplectomorphism f . Moreover, as T maps the facet{ p, eσ ( j)  =
0} ∩
( A, λ) to the facet { p, e j  = 0} ∩
(A, λ), and { p, eσ ( j) + i A i ei  =
σ ( j)
 i
 
λσ ( j) } ∩
( A, λ) to { p, e j + i A j ei  = λ j } ∩
(A, λ), the map H∗ ( f ) induced
284 M. Pabiniak

by f on cohomology must map the Poincaré duals of preimages of these facets

accordingly. That is, H∗ ( f ) = F.
In a general case, denote by λls the ls -tuple of numbers (λl1 +···+ls−1 +1 , . . . , λl1 +···+ls ),

and define λls similarly. Again, we look at primitive elements with trivial squares.
In H∗ (M; Z) these are precisely

±xls and ± (2xi − xls ) for s = 1, . . . , m and i s−1 < i < i s .

Note that each such element is contained in some subring H∗ (H(λls ); Z) ⊆ H∗

(M; Z), and that all primitive square zero elements in H∗ (H(λls ); Z) are equal
modulo 2. Therefore F must restrict to an isomorphism from H∗ (H(λls ); Z) to

some H∗ (H( λlr ); Z) with ls = 
lr . This implies that both partitions of n are equal,
up to permutation of factors. Repeating the arguments of the previous paragraph one
can construct a symplectomorphism inducing the ring isomorphism F. 

Proof (Proof of Theorem 2) Let (M, ω), ( M,  ω) be two Q-trivial Bott manifolds
with symplectic forms integral up to scaling and let F : H∗ (M; Z) → H∗ ( M;  Z)
be a ring isomorphism such that F[ω] = [ ω]. Rescaling the symplectic forms if
necessary we can assume that both ω and  ω are integral. As the cohomology of a
symplectic toric manifold is generated in degree 2, the isomorphism F must map
 Z). Using (13.2) we see that dim H2 (M; Z) = 1 dim M, and
H2 (M; Z) to H2 ( M; 2
2   Therefore dim M = dim M.  We will denote
similarly dim H ( M; Z) = 21 dim M.
this dimension by 2n. By Corollary 23 and the assumption that the symplectic forms
are integral we have that

(M, ω) = H(λ1 , . . . , λl1 ) × · · · × H(λn−lm +1 , . . . , λn ),

 
( M, λ1 , . . . , 
ω) = H( λl ) × · · · × H(
λn−l +1 , . . . , 
λn ).
1 
m

m m
for some i=1 li = i=1 
li = n partitions of n, and some λ, 
λ ∈ (Z>0 )n . Now
 
Lemma 24 gives that there exist a symplectomorphism f from ( M, ω) to (M, ω) so
that H∗ ( f ) = F. 

Acknowledgements First of all, the author would like to thank her collaborators: Xin Fang, Iva
Halacheva, Peter Littelmann and Sue Tolman. Results contained in this manuscript were obtained
in collaboration with the above mathematicians [11, 14, 27], and therefore all of them could also
be considered as the authors of this paper. The author also thanks the organisers of the workshops
“Interactions with Lattice Polytopes” for giving her the opportunity to participate and present her
results at these workshops. The author is supported by the DFG (Die Deutsche Forschungsgemein-
schaft) grant CRC/TRR 191 “Symplectic Structures in Geometry, Algebra and Dynamics”.
13 Toric Degenerations in Symplectic Geometry 285

References

1. Anderson, D.: Okounkov bodies and toric degenerations. Math. Ann. 356(3), 1183–1202 (2013)
2. Caviedes Castro, A.: Upper bound for the Gromov width of coadjoint orbits of compact Lie
groups. J. Lie Theory 26(3), 821–860 (2016)
3. Choi, S., Masuda, M.: Classification of Q-trivial Bott manifolds. J. Symplectic Geom. 10(3),
447–461 (2012)
4. Choi, S., Masuda, M., Suh, D.Y.: Rigidity problems in toric topology: a survey. Tr. Mat.
Inst. Steklova 275(Klassicheskaya i Sovremennaya Matematika v Pole Deyatel’nosti Borisa
Nikolaevicha Delone), 188–201 (2011)
5. Cox, D.A., Little, J.B., Schenck, H.K.: Toric varieties, Graduate Studies in Mathematics, vol.
124. American Mathematical Society, Providence, RI (2011)
6. Danilov, V.I.: The geometry of toric varieties. Uspekhi Mat. Nauk 33(2(200)), 85–134, 247
(1978)
7. da Silva, A.C.: Lectures on symplectic geometry. Lecture Notes in Mathematics, vol. 1764.
Springer, Berlin (2001)
8. Eisenbud, D.: Commutative algebra. Graduate Texts in Mathematics, vol. 150. Springer, New
York (1995). With a view toward algebraic geometry
9. Eisenbud, D., Harris, J.: The geometry of schemes. In: Graduate Texts in Mathematics, vol.
197. Springer, New York (2000)
10. Fang, X., Fourier, G., Littelmann, P.: Essential bases and toric degenerations arising from
birational sequences. Adv. Math. 312, 107–149 (2017)
11. Fang, X., Littelmann, P., Pabiniak, M.: Simplices in Newton-Okounkov bodies and the Gromov
width of coadjoint orbits. Bull. Lond. Math. Soc. 50(2), 202–218 (2018)
12. Fulton, W.: Introduction to toric varieties. Annals of Mathematics Studies, vol. 131. Princeton
University Press, Princeton, NJ (1993). The William H. Roever Lectures in Geometry
13. Grossberg, M., Karshon, Y.: Bott towers, complete integrability, and the extended character of
representations. Duke Math. J. 76(1), 23–58 (1994)
14. Halacheva, I., Pabiniak, M.: The Gromov width of coadjoint orbits of the symplectic group.
Pacific J. Math. 295(2), 403–420 (2018)
15. Hamilton, M.D.: The quantization of a toric manifold is given by the integer lattice points in
the moment polytope. In: Toric topology, Contemporary Mathematics, vol. 460, pp. 131–140.
American Mathematical Society, Providence, RI (2008)
16. Harada, M., Kaveh, K.: Integrable systems, toric degenerations and Okounkov bodies. Invent.
Math. 202(3), 927–985 (2015)
17. Hartshorne, R.: Algebraic Geometry. Springer, New York, Heidelberg (1977). Graduate Texts
in Mathematics, No. 52
18. Karshon, Y., Tolman, S.: The Gromov width of complex Grassmannians. Algebr. Geom. Topol.
5, 911–922 (2005)
19. Kaveh, K.: Crystal bases and Newton-Okounkov bodies. Duke Math. J. 164(13), 2461–2506
(2015)
20. Kaveh, K.: Toric degenerations and symplectic geometry of smooth projective varieties. J.
Lond. Math. Soc. (2) 99(2), 377–402 (2019)
21. Lane, J.: A Completely Integrable System on g2 Coadjoint Orbits (2016). arXiv:1605.01676
22. Littelmann, P.: Cones, crystals, and patterns. Transform. Groups 3(2), 145–179 (1998)
23. Lu, G.: Gromov-Witten invariants and pseudo symplectic capacities. Israel J. Math. 156, 1–63
(2006)
24. Lu, G.: Symplectic capacities of toric manifolds and related results. Nagoya Math. J. 181,
149–184 (2006)
25. Nishinou, T., Nohara, Y., Ueda, K.: Toric degenerations of Gelfand-Cetlin systems and potential
functions. Adv. Math. 224(2), 648–706 (2010)
26. Pabiniak, M.: Gromov width of non-regular coadjoint orbits of U (n), S O(2n) and S O(2n + 1).
Math. Res. Lett. 21(1), 187–205 (2014)
286 M. Pabiniak

27. Pabiniak, M., Tolman, S.: Symplectic cohomological rigidity via toric degnerations. In prepa-
ration
28. Schlenk, F.: Embedding problems in symplectic geometry. In: De Gruyter Expositions in Math-
ematics, vol. 40. Walter de Gruyter GmbH & Co. KG, Berlin (2005)
29. Zoghi, M.: The Gromov Width of Coadjoint Orbits of Compact Lie Groups. ProQuest LLC,
Ann Arbor, MI (2010). Thesis (Ph.D.)–University of Toronto (Canada)
Chapter 14
On Deformations of Toric Fano Varieties

Andrea Petracci

Abstract In this note we collect some results on the deformation theory of toric
Fano varieties.

Keywords Deformations · Fano varieties · Toric varieties · Reflexive polytopes

14.1 Introduction

A Fano variety is a normal projective variety X over C such that its anticanonical
divisor −K X is Q-Cartier and ample. Fano varieties constitute the basic building
blocks of algebraic varieties, according to the Minimal Model Program. The geometry
of Fano varieties is a well studied area. In particular, moduli (and consequently
deformations) of Fano varieties constitute a very interesting and important topic in
algebraic geometry, e.g. [21, 62, 69].
Here we will concentrate on deformations and smoothings of toric Fano varieties.
These varieties occupy a prominent role in Mirror Symmetry, a large part of which
is based on the phenomenon of toric degeneration as in [17, 18, 30, 43].
Toric Fano varieties correspond to certain polytopes which are called Fano poly-
topes. The goal of this note is to present some combinatorial criteria on Fano poly-
topes which can detect whether the corresponding toric Fano variety is smoothable,
i.e. can be deformed to a smooth (Fano) variety.
Special attention is given to toric Fano threefolds with Gorenstein singularities.
These varieties correspond to the 4319 reflexive polytopes of dimension 3, which
were classified by Kreuzer and Skarke [66]. In this case, thanks to the use of the soft-
ware Magma [22], we were able to produce a lot of examples for the combinatorial
criteria discussed in this note.

A. Petracci (B)
Dipartimento di Matematica, Università di Bologna, Piazza di Porta San Donato 5,
40126 Bologna, Italy
e-mail: [email protected]

© Springer Nature Switzerland AG 2022 287

A. M. Kasprzyk and B. Nill (eds.), Interactions with Lattice Polytopes,
Springer Proceedings in Mathematics & Statistics 386,
https://doi.org/10.1007/978-3-030-98327-7_14
288 A. Petracci

14.1.1 Outline

In Sect. 14.2.1 the very classical theory of infinitesimal deformations of algebraic

varieties is recalled. In Sect. 14.2.2 we survey some properties of smoothings of
algebraic varieties. In Sect. 14.2.3 two well-studied deformation invariants for Fano
varieties are introduced.
In Sect. 14.3.1 we recall some results on the deformation theory of affine toric
varieties. We provide an example in Sect. 14.3.2.
The core of this note is Sect. 14.4. We recall the definition of Fano polytopes
in Sect. 14.4.1. In Sect. 14.4.2 we present a couple of sufficient conditions that ensure
that a toric Fano variety is non-smoothable. The rigidity of toric Fano varieties is
examined in Sect. 14.4.3. In Sects. 14.4.4 and 14.4.5 we study the smoothability of
toric Fano surfaces and toric Fano threefolds with isolated singularities; an example
is presented in Sect. 14.4.6. In Sect. 14.4.7 we present another sufficient condition
that ensures that a toric Fano threefold is non-smoothable. In Sect. 14.4.8 we include
more results on deformations of toric Fano varieties.
In Sect. 14.5 we write down the lists of the reflexive polytopes of dimension 3
which satisfy the several combinatorial conditions considered in Sect. 14.4.

14.1.2 Notation and Conventions

We work over C, but everything will hold over a field of characteristic zero with
appropriate modifications.
In Sects. 14.3 and 14.4 we assume that the reader is familiar with the basic notions
of toric geometry, which can be found in [34, 41]. All toric varieties considered here
are normal. A lattice is a finitely generated free abelian group. The letters N , N , Ñ
stand for lattices and M, M, M̃ for their duals, e.g. M = HomZ (N , Z); the duality
pairing M × N → Z and its extension MR × NR → R are denoted by ·, ·.
In a real vector space of finite dimension a polytope is the convex hull of finitely
many points, or equivalently a compact subset which is the intersection of finitely
many closed halfspaces. We refer the reader to the book [99] for the geometry of
polytopes.

14.2 Deformations

14.2.1 Infinitesimal Deformations

Let (Comp) be the category of noetherian complete local C-algebras with residue
field C. For every R ∈ (Comp) we denote by m R the maximal ideal of R. Let (Art)
be the subcategory of (Comp) whose objects are artinian, i.e. local finite C-algebras.
14 On Deformations of Toric Fano Varieties 289

A functor of Artin rings is a functor F from the category (Art) to the category of
sets such that F(C) is the set with one element. We will only consider functor of
Artin rings which satisfy some additional properties: Schlessinger’s axioms (H1)
and (H2) [91] and Fantechi–Manetti condition (L) [37, (2.9)]. We will not specify
these conditions here, but we refer the reader to [37, Sect. 2] for a quick introduction.
Precise formulations and additional details about the notions we introduce below
can be found in any reference about deformation theory, e.g. [13, 36, 50, 70, 91, 92,
95, 97].
A natural transformation (or briefly map) of functors φ : F → G is called smooth
if the lifting property in Grothendieck’s definition of formally smooth morphisms
holds, i.e. for every local surjection A  A in (Art) the natural map F(A ) →
F(A) ×G(A) G(A ) is surjective; in particular, if φ is smooth then φ(A) : F(A) →
G(A) is surjective for all A ∈ (Art). A functor F is called smooth if the map from F
to the trivial functor is smooth.
For a functor F, the set F(C[t]/(t 2 )) has a natural structure of a C-vector space,
denoted by TF and called the tangent space of F. One can prove that F is the trivial
functor if and only if TF = 0. If φ : F → G is a map, then the function φ(C[t]/(t 2 ))
is linear and denoted by Tφ : TF → TG.
If R ∈ (Comp) one can consider the functor h R = Hom(·, R) prorepresented
by R. A map h R → F is equivalent to a pro-object of F on R = lim R/mn+1 R , i.e.
←−
an element of the set lim F(R/m R ). A hull for a functor F is a ring R ∈ (Comp)
n+1
←−
together with a smooth morphism φ : h R → F such that Tφ is bijective. A hull exists
if and only if TF has finite dimension. If a hull exists, it is unique. Provided that TF
has finite dimension r , then F is smooth if and only if the hull of F is isomorphic
to C[[t1 , . . . , tr ]].
For a functor F, consider the set E made up of pairs (π, ξ ), where π : A →
A is a surjection in (Art) such that m A · (ker π ) = 0 and ξ ∈ F(A). A C-vector
space
 V is called an obstruction space for F if there exists a function ω : E →
(π,ξ )∈E ker π ⊗C V such that the two following conditions are satisfied:

1. for every (π, ξ ) ∈ E, ω(π, ξ ) ∈ ker π ⊗C V ;

2. for every (π, ξ ) ∈ E, we have that ω(π, ξ ) = 0 if and only if there exists ξ  ∈
F(A ) which maps to ξ .
There are infinitely many obstruction spaces for a functor F because any vector space
containing an obstruction space is an obstruction space. A functor F is smooth if and
only if 0 is an obstruction space for F; in this case we also say that F is unobstructed.
There is a notion of compatible obstruction spaces for a map φ : F → G: this will
be a linear map oφ from an obstruction space of F to an obstruction space of G with
some compatibility properties with respect to φ.
The following is an important smoothness criterion. Assume that φ : F → G is
a map with compatible obstruction map oφ from an obstruction space of F to an
obstruction space of G. If Tφ is surjective and oφ is injective, then φ is smooth.
Let X be a scheme of finite type over C. We denote by Def X the functor of
(infinitesimal) deformations of X . If R ∈ (Comp), a pro-object of Def X on R is
290 A. Petracci

called a formal deformation of X over R. If R is a hull for Def X , then the cor-
responding formal deformation of X over R is called the miniversal deformation
of X . We say that X is rigid if all deformations of X are trivial. If X is reduced,
then the tangent space of Def X is Ext 1 ( X , O X ); in this case X is rigid if and only
if Ext 1 ( X , O X ) = 0. If X is either normal or reduced and local complete intersec-
tion (l.c.i. for short), then Ext 2 ( X , O X ) is an obstruction space for Def X . If X is
smooth, then Hi (X, TX ) = Exti ( X , O X ) for all i ≥ 0. In particular, if X is smooth
and affine then it is rigid.
Proposition 1 If X is a smooth Fano variety, then Hi (X, TX ) = 0 for each i ≥ 2. In
particular, the infinitesimal deformations of X are unobstructed, i.e. Def X is smooth.
Proof Let n be the dimension of X . Since the anticanonical line bundle ω∨X is ample,
by Kodaira–Nakano vanishing we have Hi (X, n−1 X ⊗ ω∨X ) = 0 whenever i + n −
1 > n, i.e. i ≥ 2. We conclude because the tangent sheaf TX is isomorphic to n−1
X ⊗
ω∨X . 
Let X be a scheme of finite type over C and let Def ltX be the subfunctor of Def X
made up of the locally trivial deformations of X . The tangent space of Def ltX
is H1 (X, TX ) and H2 (X, TX ) is an obstruction space for Def ltX .
Proposition 2 Let X be a reduced scheme of finite type over C such that X is either
l.c.i. or normal. If H0 (X, Ext 1 ( X , O X )) = 0, then all deformations of X are locally
trivial, i.e. Def ltX = Def X .
Proof The local-to-global spectral sequence for Ext gives the following exact
sequence.

0 → H1 (TX ) → Ext 1 ( X , O X ) → H0 (Ext 1 ( X , O X ))

→ H2 (TX ) → Ext 2 ( X , O X )

The vanishing of H0 (Ext 1 ( X , O X )) implies that the inclusion φ : Def ltX → Def X
induces an isomorphism on tangent spaces and an injection on obstruction spaces.
Therefore φ is smooth, and consequently surjective. 
In particular, all deformations of a smooth scheme are locally trivial.
Let X be a reduced scheme of finite type over C with isolated singularities.
For each singular point x ∈ X , let Ux be an affine open neighbourhood of x such
that Ux \ {x} is smooth. Then define

X :=
Def loc Def Ux .
x∈Sing(X )

X is H (X, Ext ( X , O X )). If X is either l.c.i. or normal,

The tangent space of Def loc 0 1

then H (X, Ext ( X , O X )) is an obstruction space for Def loc

0 2
X . There is an obvious
map Def X → Def loc
X which restricts a deformation of X to a deformation of U x for
each x.
14 On Deformations of Toric Fano Varieties 291

Proposition 3 Let X be a reduced scheme of finite type over C with isolated sin-
gularities. Assume that X is either l.c.i. or normal. If H2 (X, TX ) = 0 then there
are no local-to-global obstructions for the infinitesimal deformations of X , i.e. the
map Def X → Def locX is smooth.

Proof We consider the local-to-global spectral sequence for Ext • ( X , O X ). The sec-
p,q
ond page is given by E 2 = H p (Ext q ( X , O X )). Since X has isolated singularities,
the sheaves Ext ( X , O X ) are supported on isolated points for q ≥ 1; in particu-
q
p,q
lar they do not have higher cohomology. This means that E 2 is supported on the
lines p = 0 and q = 0. Therefore, in E 2 the only non-zero differential is

d2 : H0 (Ext 1 ( X , O X )) −→ H2 (TX ).

We obtain that the bottom left corner of the third page E 3 is the following.

H3 (TX ) 0 0 0
cokerd2 0 0 0
H1 (TX ) 0 0 0
H0 (TX ) ker d2 H0 (Ext 2 ( X , O X )) H0 (Ext 3 ( X , O X ))

In E 3 the only non-zero differential is

d3 : H0 (Ext 2 ( X , O X )) −→ H3 (TX ).

The bottom left corner of the fourth page E 4 is the following.

cokerd3 0 0 0
cokerd2 0 0 0
H1 (TX ) 0 0 0
H0 (TX ) ker d2 ker d3 H0 (Ext 3 ( X , O X ))

From the fourth page on, the pieces of total degree ≤ 3 do not change any more.
Therefore we have two short exact sequences:

0 −→ H1 (TX ) −→ Ext 1 ( X , O X ) −→ ker d2 −→ 0,

0 −→ cokerd2 −→ Ext 2 ( X , O X ) −→ ker d3 −→ 0.

These can be joined to construct the following long exact sequence.

d2
0 −→ H1 (TX ) −→ Ext 1 ( X , O X ) −→ H0 (Ext 1 ( X , O X )) −→
d2
−→ H2 (TX ) −→ Ext 2 ( X , O X ) −→ H0 (Ext 2 ( X , O X ))
292 A. Petracci

So far we did not use the assumption H2 (TX ) = 0. From this vanishing, via the long
exact sequence above we deduce that the map Def X → Def loc X induces a surjection
on tangent spaces and an injection on obstruction spaces. 

14.2.2 Smoothings

Here we discuss smoothability conditions for schemes of finite type over C. We will
only consider the case of equidimensional schemes and we will refer the reader to [50,
Sect. 29] for a more general treatment, which uses the Lichtenbaum–Schlessinger
functors.
If X is a proper scheme over C, a smoothing of X is a proper flat morphism X → B
such that B is an integral scheme of finite type over C of positive dimension and
there exists a closed point b0 ∈ B such that the fibre over b0 is X and all the other
fibres are smooth. By restricting to a curve in B and normalising it, we may require
that the base B is a smooth affine curve and that the maximal ideal corresponding
to b0 is principal. We say that X is smoothable if it admits a smoothing.
For every n ≥ 0, set Sn := Spec C[t]/(t n+1 ). If X is a scheme of finite type over C
with pure dimension d, then a formal smoothing of X is a formal deformation {X n →
Sn }n of X over C[[t]] such that there exists m such that t m is in the dth Fitting ideal
of  X m /Sm . We refer the reader to [35, Sect. 20.2] for the definition and the properties
of Fitting ideals. We say that X is formally smoothable if it admits a formal smoothing.
It is clear that if X is formally smoothable, then every open subscheme of X is
formally smoothable.

Remark 4 If {X n → Sn }n is a formal deformation of X over C[[t]] and t m is in

the dth Fitting ideal of  X m /Sm , then for all n ≥ m we have that t n is in the dth Fitting
ideal of  X n /Sn .
The proof of this fact is as follows. We have O X n = O X n+1 /t n+1 O X n+1 . Since the
formation of Fitting ideals commutes with base change, we have the equality

Fittd ( X n /Sn ) = (Fittd ( X n+1 /Sn+1 ) + t n+1 O X n+1 )/t n+1 O X n+1 .

Therefore if t n ∈ Fittd ( X n /Sn ) then t n ∈ Fittd ( X n+1 /Sn+1 ) + t n+1 O X n+1 , hence t n+1 ∈
tFittd ( X n+1 /Sn+1 ) ⊆ Fittd ( X n+1 /Sn+1 ) as t n+2 = 0 in O X n+1 .

Lemma 5 Let X be a Cohen–Macaulay proper scheme over C of pure dimension d.

Let B be a smooth curve over C, b0 ∈ B be a closed point, and π : X → B be
a proper flat morphism such that the fibre over b0 is X . Let ξ be the formal mb0 -
adic completion of π at b0 , i.e. ξ = {X × B Spec O B,b0 /mn+1
b0 → Spec O B,b0 /mb0 }n .
n+1

Then:
1. if π is a smoothing of X , then ξ is a formal smoothing of X ;
2. if ξ is a formal smoothing of X , then there exists an open neighbourhood B  of b0
in B such that X × B B  → B  is a smoothing of X .
14 On Deformations of Toric Fano Varieties 293

Proof This proof comes from [14, Sect. 0E7S].

Notice that ξ does not change if we restrict π to an open neighbourhood of b0
in B. Therefore, in order to prove the statements (1) and (2) we can arbitrarily restrict
to an open neighbourhood of b0 in B. Hence we may assume that B is affine and the
maximal ideal corresponding to the point b0 is principal, generated by t ∈ O B .
We consider the set W ⊆ X made up of the points x ∈ X such that the local ring
of the fibre Xπ(x) at x is Cohen–Macaulay. By [44, 12.1.7], W is open in X. As π
is closed, B \ π(X \ W ) is an open neighbourhood of b0 in B. Therefore, if we
restrict B to an open neighbourhood of b0 in B, we may assume that all fibres of π
are Cohen–Macaulay. By [14, Lemma 02NM], we may assume that π has relative
dimension d.
Let I ⊆ OX be the dth Fitting ideal of X/B . For each n, set

Sn = Spec O B,b0 /mn+1

b0 = Spec O B /t
n+1
OB

and X n = X × B Sn ; let In ⊆ O X n be the dth Fitting ideal of  X n /Sn . Since Fitting

ideals commute with base change, we have O X n = OX /t n+1 OX and In = I O X n =
(I + t n+1 OX )/t n+1 OX .
Since π is flat of relative dimension d, the zero locus of I is the singular locus
of π . Moreover, the fibre over b0 is the closed√subset V(t). Therefore, the fibre of b0
is the unique singular fibre if and only if t ∈ I .
(1) If π is a smoothing, then there exists m such that t m ∈ I . Since Im = (I +
t m+1
OX )/t m+1 OX , this implies that t m ∈ Im . So ξ is a formal smoothing.
(2) If ξ is a formal smoothing, then t m ∈ Im = (I + t m+1 OX )/t m+1 OX for some m.
So in OX we have the equality t m = p + t m+1 q, for some p ∈ I and q ∈ OX . Writ-
ing t m (1 − tq) = p and noticing that the function 1 − tq does not vanish at the points
of X = V(t), we deduce that t m belongs to the stalk Ix of I at all points x ∈ X . This
implies that t m lies in I in an open neighbourhood U of X in X. Since π is closed,
by restricting B to B \ π(X \ U ) we have t m ∈ I . Therefore π is a smoothing. 

Proposition 6 Let X be a Cohen–Macaulay scheme proper over C.

1. If X is smoothable, then every open subscheme of X is formally smoothable.
2. Assume that X is projective and H2 (X, O X ) = 0; if X is formally smoothable,
then X is smoothable.

Proof We may assume that X is connected. Therefore X has pure dimension, say d.
(1) This follows immediately from Lemma 5 and from the fact that if X is formally
smoothable then every open subscheme of X is formally smoothable.
(2) Set d := dim X . Let ξ = {X n → Sn }n be a formal smoothing of X , where Sn
is Spec C[t]/(t n+1 ) as usual. Let m be such that t m is in the dth Fitting ideal of  X m /Sm .
As X is proper over C, the tangent space of Def X has finite dimension, there-
fore Def X has a hull R ∈ (Comp). Let η = {ηn : Yn → Spec R/mn+1 R }n be the
miniversal deformation of X . By [95, Proposition 6.51] or [92, Theorem 2.5.13],
from H2 (O X ) = 0 we deduce that η is effective, i.e. there exists a projective flat
morphism X → Spec R whose m R -adic completion is η.
294 A. Petracci

By a theorem of Artin [11, Theorem 1.6] (see also [50, Theorem 21.3]), the mor-
phism X → Spec R is algebraizable in the following sense: there exist a scheme Z
of finite type over C, a closed point z 0 ∈ Z , and a proper flat morphism X → Z ,
with fibre X over z 0 , such that R is the completion 
O Z ,z0 of the local ring of Z at z 0
and X is isomorphic, as R-schemes, to X × Z Spec R. In particular, the miniversal
deformation η is the collection {X × Z Spec O Z ,z0 /mn+1
z 0 → Spec O Z ,z 0 /mz 0 }n . The
n+1

situation is summarised in the following cartesian squares, for all n.

Yn X X
ηn

Spec R/mn+1
R Spec R = Spec 
O Z ,z0 Z

As η is miniversal, there exists a local C-algebra homomorphism

ϕ: 
O Z ,z0 = R −→ C[[t]]

such that ξ is induced by η via ϕ, i.e. X n is isomorphic to Yn ×Spec R/mn+1

R
Sn as Sn -
schemes for every n. By another theorem of Artin [10, Corollary 2.5], the map ϕ has
an algebraic approximation up to order m in the following sense: there exist a smooth
affine curve B over C with a closed point b0 ∈ B and a C-morphism f : B → Z such
that f (b0 ) = z 0 and the completion

ϕ : 
O Z ,z0 = R −→ 
O B,b0 = C[[t]]

of f b#0 : O Z ,z0 → O B,b0 satisfies the following property:

ϕ ≡ ϕ  modulo t m+1 . (14.1)

Let π be the base change X × Z B → B along f : B → Z . Let ξ  be the formal mb0 -

adic completion of π , i.e. ξ  = {X × Z Spec O B,b0 /mn+1
b0 → Spec O B,b0 /mb0 }n . The
n+1

two formal deformations ξ and ξ of X over C[[t]] are in general different, but they
coincide up to order m because of (14.1). This implies that t m is in the dth Fitting ideal
of the sheaf of Kähler differentials of X × Z Spec O B,b0 /mm+1
b0 → Spec O B,b0 /mm+1 b0 .

Therefore, ξ is a formal smoothing. By Lemma 5, up to restrict B to an open neigh-
bourhood of b0 in B, we have that π : X × Z B → B is a smoothing. 

The following theorem ensures that a projective scheme with formally smoothable
isolated singularities is smoothable, provided that some local and cohomological
conditions hold.

Theorem 7 Let X be a projective scheme over C such that:

1. X is reduced and Cohen–Macaulay;
2. X is either l.c.i. or normal;
14 On Deformations of Toric Fano Varieties 295

3. H2 (X, TX ) = 0 and H2 (X, O X ) = 0;

4. X has isolated singularities and for each singular point x ∈ X there exists an
open affine neighbourhood of x which is formally smoothable.
Then X is smoothable.

Proof Set d = dim X . Let x1 , . . . , xr be the singular points of X . Let Ui be an

affine open neighbourhood of xi in X which is formally smoothable and such
that Ui \ {xi } is smooth. Let ξi = {Ui,n → Sn }n be a formal smoothing of Ui , where Sn
is Spec C[t]/(t n+1 ) as usual.
rBy Proposition 3, from H (TX ) = 0 we deduce that the map Def X → Def X =
2 loc

i=1 Def Ui is smooth. Therefore there exists a formal deformation ξ = {X n → Sn }n

of X over C[[t]] such that for each i the restriction of ξ to Ui is ξi , i.e. for all n the
restriction of X n to Ui is Ui,n . By Remark 4 we have that ξ is a formal smoothing
of X . We conclude by Proposition 6. 

We now see some conditions that imply that a scheme is not smoothable.

Proposition 8 Let X be a singular scheme of finite type over C of pure dimension.

Assume that at least one of the following conditions holds:
1. every infinitesimal deformation of X is locally trivial;
2. the functor Def X has an artinian hull.
Then X is not formally smoothable.

Proof Set d = dim X .

(1) Let U be a singular affine open subscheme of X . Let {X n → Sn }n be a for-
mal deformation of X over C[[t]]. Let Un be the restriction of X n to U . By (1) we
get that Un is isomorphic, as Sn -scheme, to the trivial deformation U ×Spec C Sn .
Therefore Fittd (Un /Sn ) = Fittd (U/C )OUn . As U is singular, Fittd (U/C )  OU .
This implies that t n ∈
/ Fittd (Un /Sn ).
(2) Let R be the hull of Def X . Every formal deformation of X over C[[t]] is
induced by the miniversal one via a local C-algebra homomorphism f : R → C[[t]].
As every element in m R is nilpotent and C[[t]] is a domain, the homomorphism f
factors as R  R/m R = C → C[[t]]. This implies that every formal deformation
of X over C[[t]] is trivial. Using a similar argument as in (1), we can prove that X
cannot have a formal smoothing. 

The following corollary, which is a direct consequence of Propositions 2, 6 and 8,

gives some obstructions to the smoothability of a Cohen–Macaulay proper scheme.

Corollary 9 Let X be a Cohen–Macaulay scheme proper over C. Let U ⊆ X be an

open subscheme of X such that U is singular, reduced, and either l.c.i. or normal.
If H0 (U, Ext 1 (U , OU )) = 0 or Def U has an artinian hull (e.g. if Ext 1 (U , OU ) =
0), then X is not smoothable.
296 A. Petracci

14.2.3 Invariants

Here we introduce a couple of invariants for Fano varieties.

The Hilbert series of a Fano variety X is the power series defined by its anti-
plurigenera: 
Hilb(X, −K X ) := h 0 (X, −m K X )t m ∈ Z[[t]].
m≥0

The (anticanonical) degree of a Fano variety X is the positive rational num-

ber (−K X )n , where n = dim X . If X is Gorenstein, i.e. K X is Cartier, then the degree
is an integer. The degree can be recovered from the Hilbert series because, up to a
constant which depends on the dimension of X , it is the leading term of the Hilbert
polynomial of −K X .
The following proposition shows that the Hilbert series and the anticanonical
degree are deformation invariants for Fano varieties with Gorenstein log terminal
singularities.
Proposition 10 Let S be a noetherian scheme over Q and let π : X → S be a proper
flat morphism whose geometric fibres are Fano varieties with Gorenstein log terminal
singularities. Then the Hilbert series and the degree of the fibres are locally constant
on S.
Proof The morphism π is a relatively Gorenstein. Therefore, by [47, V.9.7], the dual-
ising sheaf ωπ is a line bundle on X and its restriction to each fibre X s is O X s (K X s ).
By Serre duality and Kawamata–Viehweg vanishing [63, Theorem 2.70], we
get H1 (X s , O X s (−m K X s )) = 0 for all m ≥ 0 and s ∈ S. By cohomology and base
change [48, Theorem III.12.11], for all m ≥ 0, we get that the sheaf π∗ ωπ⊗−m is
locally free and has rank h 0 (X s , O X s (−m K X s )) at the point s ∈ S. This implies that
the Hilbert series of the fibres is locally constant on S. 

14.3 Deformations of Affine Toric Varieties

14.3.1 Toric Singularities

In this section we will consider deformations of toric singularities, that is affine toric
varieties. We refer the reader to [34, 41] for an introduction to toric geometry.
If X is an affine toric variety of dimension 2, then X is a cyclic quotient surface
singularity. There is extensive literature about deformations of this kind of singular-
ities, e.g. [19, 27, 64, 88, 93, 94]. In particular, it is known that every affine toric
variety of dimension 2 is smoothable [12].
The study of the deformation theory of affine toric varieties of dimension at least 3
has been initiated by Altmann [4–8]. For example, he computed the tangent space of
the deformation functor of an affine toric variety. We will not write down the explicit
14 On Deformations of Toric Fano Varieties 297

description of Ext 1 ( X , O X ) when X is an affine toric variety, but we will mention

a consequence.

Proposition 11 (Altmann [5, Corollary 6.5.1]) If X is a Q-Gorenstein affine toric

variety which is smooth in codimension 2 and Q-factorial in codimension 3, then X
is rigid.

Corollary 12 Every isolated Q-Gorenstein toric singularity of dimension ≥ 4 is

rigid.

Now we need to do a brief detour on Minkowski sums. If F0 , F1 , . . . , Fr are

polytopes in a real vector space, their Minkowski sum is the polytope

F0 + F1 + · · · + Fr := {v0 + v1 + · · · + vr | v0 ∈ F0 , v1 ∈ F1 , . . . , vr ∈ Fr }.

When we have F = F0 + F1 + · · · + Fr , we say that we have a Minkowski decompo-

sition of the polytope F. We consider Minkowski decompositions up to translation:
for instance, we consider the Minkowski decomposition F = (v + F0 ) + (−v + F1 )
to be equivalent to F = F0 + F1 for every vector v. Moreover, in what follows we
require that the summands F j are lattice polytopes, i.e. their vertices belong to a
fixed lattice.
Altmann [5] has noticed that certain Minkowski decompositions induce deforma-
tions of affine toric varieties. In Sect. 14.3.2 we will see an example of this fact. For
the proof we refer the reader to the original reference [5] and to [71, 81].
Now let us concentrate on Gorenstein toric singularities. They are associated to
lattice polytopes of dimension one less than the dimension of the singularity. More
precisely, let F be a lattice polytope of dimension n − 1 in a lattice N of rank n − 1
and let U F be the affine toric variety associated to the cone σ F = R≥0 (F × {1})
in the lattice N := N ⊕ Z, i.e. U F = Spec C[σ F∨ ∩ M], where M = M ⊕ Z is the
dual of N and σ F∨ is the dual cone of σ F . We have that U F has dimension n and is
Gorenstein. All Gorenstein affine toric varieties without torus factors arise in this
way from a lattice polytope. The isomorphism class of U F does not change if we
change F via an affine transformation in N  GL(N , Z).
As usual in toric geometry, the geometric properties of U F can be deduced from
the combinatorial properties of F. For instance:
1. U F is smooth in codimension k if and only if all faces of F with dimension < k
are standard simplices;
2. U F is Q-factorial in codimension k if and only if all faces of F with dimension < k
are simplices.
It is always the case that U F is smooth in codimension 1 and Q-factorial in codimen-
sion 2.
If F is a segment of lattice length m + 1, then U F is the Am surface singular-
ity Spec C[x, y, z]/(x y − z m+1 ). This is an isolated hypersurface singularity, there-
fore it is very easy to write down the miniversal deformation: x y = z m+1 + tm z m−1 +
· · · + t1 over C[[t1 , . . . , tm ]]. It is clear that this singularity is smoothable.
298 A. Petracci

Fig. 14.1 A standard square, a standard triangle, an A1 -triangle and an A2 -triangle

If F is a lattice polygon, then the affine toric threefold U F has the following
properties:
1. U F has, at most, an isolated singularity if and only if the edges of F are unitary,
i.e. have lattice length 1;
2. U F is Q-factorial if and only if F is a triangle.
Now we provide some examples of lattice polygons and their corresponding toric
Gorenstein affine threefolds.

Example 13 A lattice polygon F is called a standard square (Fig. 14.1) if it is a

quadrilateral such that all its lattice points are vertices, or equivalently if it is Z2 
GL2 (Z)-equivalent to conv{(0, 0), (1, 0), (1, 1), (0, 1)} ⊆ R2 . If F is a standard
square, then U F is the ordinary double point (i.e. node) Spec C[x, y, z, w]/(x y −
zw). This singularity is clearly smoothable as it is a hypersurface singularity. Its
miniversal deformation is given by x y − zw = t over C[[t]].
A lattice polygon F is called a standard triangle if it is a triangle such that
all its lattice points are vertices, or equivalently if it is Z2  GL2 (Z)-equivalent
to conv{(0, 0), (1, 0), (0, 1)} ⊆ R2 . F is a standard triangle if and only if U F is
isomorphic to A3 .
If m ≥ 1, then a lattice polygon F is called an Am -triangle if it is a triangle such
that there are no interior lattice points and the edges have lattice lengths 1, 1, m + 1,
respectively. Equivalently, a polygon is an Am -triangle if and only if it is Z2 
GL2 (Z)-equivalent to conv{(0, 0), (m + 1, 0), (0, 1)} ⊆ R2 . If F is an Am -triangle,
then U F is the c Am -singularity Spec C[x, y, z, w]/(x y − z m+1 ). This singularity is
clearly smoothable as it is a hypersurface singularity.

Altmann [7] explicitly constructed the miniversal deformation of an isolated

Gorenstein toric singularity of dimension 3. (By Corollary 12 it is trivial to construct
the miniversal deformation of an isolated Gorenstein toric singularity of dimen-
sion ≥ 4.) A consequence of his construction is the following description of the
irreducible components of the base of the miniversal deformation.

Theorem 14 (Altmann [7]) Let F be a lattice polygon with unitary edges and let U F
be the corresponding isolated Gorenstein toric singularity of dimension 3. Let R be
the hull of Def U F . Then there exists a one-to-one correspondence between minimal
primes of R and maximal Minkowski decompositions of F. Moreover, if a minimal
prime p ⊂ R corresponds to the maximal Minkowski decomposition F = F0 + F1 +
· · · + Fr , then r = dim R/p.
14 On Deformations of Toric Fano Varieties 299

Corollary 15 Let F be a lattice polygon with unitary edges and let U F be the
associated isolated Gorenstein toric singularity of dimension 3. Then Def U F has an
artinian hull if and only if F is Minkowski indecomposable.

14.3.2 The Affine Cone over the Del Pezzo Surface

of Degree 7

Here we study an explicit example of what has been considered in Sect. 14.3.1. In
the lattice N = Z2 consider the pentagon
    

1 1 0 −1 0
F = conv , , , , ⊆ N R, (14.2)
0 1 1 0 −1

which is depicted on the left of Fig. 14.2. The toric variety associated to the face
fan of F is the smooth del Pezzo surface of degree 7, which is denoted by dP7 and
is the blow up of P2 in 2 distinct points. The anticanonical map of dP7 is a closed
embedding into P7 .
Now we put the pentagon F at height 1 in the lattice N = N ⊕ Z and we consider
the cone over it:
⎧⎛ ⎞ ⎛ ⎞ ⎛ ⎞ ⎛ ⎞ ⎛ ⎞⎫
⎨ 1 1 0 −1 0 ⎬
σ F = cone ⎝0 ⎠ , ⎝ 1⎠ , ⎝ 1 ⎠ , ⎝ 0 ⎠ , ⎝ −1 ⎠ ⊆ N R ⊕ R.
⎩ ⎭
1 1 1 1 1

The affine toric variety U F = Spec C[σ F∨ ∩ (M ⊕ Z)] is the affine cone over the anti-
canonical embedding of dP7 and has an isolated Gorenstein canonical non-terminal
singularity at the vertex of the cone.
Altmann [7, (9.1)] shows that the hull of Def U F is C[[t1 , t2 ]]/(t12 , t1 t2 ), which is a
line with an embedded point. The reduction of the miniversal deformation, i.e. the
base change to the reduction of the hull, is induced by the unique maximal Minkowski
decomposition of the pentagon F in the following way.

Fig. 14.2 The Minkowski decomposition (14.3) of the pentagon F in (14.2)

300 A. Petracci

In the lattice N we have the Minkowski decomposition

  
 

0 −1 0 0 1
F = conv , , + conv , , (14.3)
0 0 −1 0 1

which is illustrated in Fig. 14.2. Following [5, (3.4)], in the lattice Ñ = N ⊕ Ze1 ⊕
Ze2 we construct the cone
⎧⎛ ⎞ ⎛ ⎞ ⎛ ⎞ ⎛ ⎞ ⎛ ⎞⎫
⎪
⎪ 0 −1 0 0 1 ⎪
⎨⎜ ⎟ ⎜ ⎟ ⎜ ⎟ ⎜ ⎟ ⎜ ⎟⎪ ⎬
0 0 −1 0
⎟ , ⎜ ⎟ , ⎜ ⎟ , ⎜ ⎟ , ⎜1⎟ ⊆ ÑR .
σ̃ = cone ⎜ ⎝ 1 ⎠ ⎝ 1 ⎠ ⎝ 1 ⎠ ⎝ 0 ⎠ ⎝0 ⎠⎪
⎪
⎪ ⎪
⎩ ⎭
0 0 0 1 1

Notice that the the first three rays of σ̃ come from the vertices of the first summand
of F in (14.3), whereas the last two rays of σ̃ come from the vertices of the second
summand of F in (14.3). Let Ũ = Spec C[σ̃ ∨ ∩ M̃] be the affine toric variety asso-
ciated to the cone σ̃ , where M̃ denotes the dual of Ñ . One can prove that Ũ has only
an isolated terminal Gorenstein singularity. Let f 1 and f 2 be the regular functions
on Ũ associated to the characters (0, 0, 1, 0) ∈ M̃ and (0, 0, 0, 1) ∈ M̃, respectively.
The variety U F is the zero locus of the function f 1 − f 2 , i.e. we have a cartesian
diagram
UF Ũ (14.4)
π

Spec C A1C

where π is given by the function f 1 − f 2 and the bottom morphism is given by

the origin of A1C . Since f 1 − f 2 is not constant and A1C is regular of dimension
1, the morphism π is flat. The reduction of the miniversal deformation of U F is the
formal deformation of U F over C[[t]] obtained from the square (14.4) by base change
via Spec C[t]/(t n+1 ) → Spec C[t] = A1C for all n. The following proposition shows
that this is a formal smoothing.

Proposition 16 Let F be the pentagon defined in (14.2) and let U F be the cor-
responding Gorenstein toric threefold singularity. Then the collection of the base
change of π in (14.4) via Spec C[t]/(t n+1 ) → Spec C[t] = A1C for all n is a formal
smoothing of U F . In particular, U F is formally smoothable.

Proof We want to study the closed fibres of π . The fibre over the origin of A1C
is U F . Let us fix λ ∈ C \ {0} and we consider the fibre π −1 (λ) of π over the closed
point (t − λ) of A1C corresponding to λ. We consider the subcone τ1 (resp. τ2 ) of σ̃
that is generated by the first three (resp. last two) rays of σ̃ . We consider the affine
toric variety W j = Spec C[τ j∨ ∩ M̃], for j = 1, 2. We have that W1 and W2 are open
subschemes of Ũ .
14 On Deformations of Toric Fano Varieties 301

We have that W1 is the open subset of Ũ where the function f 2 does not vanish,
i.e. W1 = { f 2 = 0} ⊆ Ũ , and analogously W2 = { f 1 = 0} ⊆ Ũ . It is clear that there
is an isomorphism W1  A3 × Gm with respect to which the function f 1 |W1 becomes
a projection onto a A1 -factor in A3 and the function f 2 |W1 becomes the projection onto
the Gm -factor. There is also an isomorphism W2  A2 × G2m with respect to which
the function f 1 |W2 becomes a projection onto a Gm -factor and the function f 2 |W2
becomes a projection onto an A1 -factor.
Now π −1 (λ) ∩ W1 = { f 1 = f 2 + λ} ∩ W1 is isomorphic to A2 × Gm and π −1 (λ)
∩ W2 = { f 2 = f 1 − λ} ∩ W2 is isomorphic to A1 × G2m . Since λ = 0, it is clear
that π −1 (λ) ⊆ W1 ∪ W2 . Therefore we have proved that π −1 (λ) is smooth.
One can also show that the generic fibre of π is smooth over the generic point
of A1C . In order to prove this, it is enough to base change to the spectrum of the
field C(t) of rational function of A1C and pursue a similar argument, which deals
with toric varieties over the field C(t).
In particular, π is flat of relative dimension 3 and has Cohen–Macaulay fibres. As
in the proof of Lemma 5, from the fact that all non-special fibres of π are smooth we
can deduce that π induces a formal smoothing of U F . 

14.4 Deformations of Toric Fano Varieties

14.4.1 Fano Polytopes

Fano polytopes are the combinatorial-polyhedral avatars of toric Fano varieties.

Definition 17 A polytope P in a lattice N of rank n is called Fano if:

1. P has dimension n;
2. the origin 0 lies in the interior of P;
3. the vertices of P are primitive lattice elements of N .
If P is a Fano polytope, we denote by X P the complete toric variety associated to
the spanning fan (also called the face fan) of P.

If P is a Fano polytope, then X P is a Fano variety. All toric Fano varieties arise
in this way from a Fano polytope [34, Sect. 8.3]. The variety X P is Gorenstein,
i.e. its (anti)canonical divisor is Cartier, if and only if P is reflexive, i.e. the facets
of P lie on hyperplanes with height 1 with respect to the origin. The maximal toric
affine charts of X P (or equivalently the torus-fixed points of X P ) are in one-to-one
correspondence with the facets of P. If n is the dimension of P, for every 0 ≤ k ≤ n
there is a one-to-one correspondence between the k-dimensional torus-orbits of X P
and the (n − k − 1)-dimensional faces of P.
Fano polytopes of small dimension with specific properties have been classi-
fied [15, 16, 57–59, 65–67, 77, 78, 89, 90, 98]. We refer the reader to [60] for a
survey on the classification of Fano polytopes.
302 A. Petracci

14.4.2 Two Sufficient Conditions for Non-smoothability

It is an open problem to understand whether an arbitrary toric Fano variety is smooth-

able. Here we state a couple of conditions that forbid the smoothability. Both condi-
tions on a toric Fano variety X are based on the existence of an open affine singular
subscheme U such that U is not formally smoothable.

Theorem 18 Let N be a lattice, let P be a Fano polytope in N , and let X be the

toric Fano variety associated to the spanning fan of P. Assume that there exists a
face F of P which satisfies the following conditions:
1. for each 1-face F  of F, there exists a basis of N which contains the two vertices
of F  ;
2. each 2-face of F is a triangle;
3. there exists no basis of N which contains all the vertices of F.
Then X is not smoothable.

Proof Let U be the affine toric open subscheme of X associated to the cone
spanned by the face F. The condition (1) means that U is smooth in codimension
2. The condition (2) means that U is Q-factorial in codimension 3. Therefore U is
rigid by Proposition 11. The condition (3) implies that U is singular. Therefore, by
Corollary 9, X is not smoothable. 

If P is a reflexive polytope of dimension 3, then the theorem above applies if

there exists a triangular facet F with unitary edges and such that it is not a standard
triangle. Below we see that we can relax the condition of F being triangular to F
being Minkowski-indecomposable.

Proposition 19 Let P be a reflexive polytope of dimension 3 and let X be the toric

Fano threefold associated to the spanning fan of P. Assume that there exists a facet F
of P such that:
1. F has unitary edges;
2. F is Minkowski-indecomposable;
3. F is not a standard triangle (i.e. the vertices of F do not form a basis of the
lattice).
Then X is not smoothable.

Proof The proof is very similar to the proof of Theorem 18. Let U be the affine
toric open subscheme of X associated to the cone spanned by F. The conditions (1)
and (3) means that U has an isolated singularity. Since P is reflexive, U is Gorenstein.
By Corollary 15, from (2) we deduce that Def U has an artinian hull. Therefore, by
Proposition 8, U is not formally smoothable. By Corollary 9, X is not smoothable.

14 On Deformations of Toric Fano Varieties 303

14.4.3 Rigidity

Here we will see that if a toric Fano variety has very mild singularities then it is rigid.

Lemma 20 Let X be a toric Fano variety. Then Hi (X, TX ) = 0 for each i ≥ 1. In

particular, all locally trivial deformations of X are trivial.

Proof Set n = dim X . Consider the smooth locus j : U → X . Let D be the toric
boundary of X . The sheaves TX and ( j∗ Un−1 ⊗ O X (D))∨∨ are reflexive on X
and their restrictions to U coincide, because U is smooth and TU is isomorphic
to Un−1 ⊗ ωU∨ . Therefore, since the complement of U has codimension at least 2, by
[49, Proposition 1.6] we have that TX is isomorphic to ( j∗ Un−1 ⊗ O X (D))∨∨ . Since D
is ample, we conclude by Bott–Steenbrink–Danilov vanishing [34, Theorem 9.3.1]
(see also [24, 40, 75]). 

An immediate consequence of the lemma above is the following result.

Proposition 21 Every smooth toric Fano variety is rigid.

This result was originally proved by Bien and Brion [20]. Later de Fernex and
Hacon [38] proved the rigidity of Q-factorial terminal toric Fano varieties. The
following theorem, due to Totaro, is the most general rigidity theorem for toric
Fano varieties of which we are aware.

Theorem 22 (Totaro [96, Theorem 5.1]) A Fano toric variety which is smooth in
codimension 2 and Q-factorial in codimension 3 is rigid.

Proof By Lemma 20, H1 (TX ) = 0. By Proposition 11, the sheaf Ext 1 ( X , O X ) is

zero. From the five term exact sequence of Ext, which is written in the proof of
Proposition 2, we deduce that Ext 1 ( X , O X ) is zero. 

If P is a Fano polytope, then X P satisfies the hypotheses of this theorem if and

only if all 2-faces of P are triangles and each edge, i.e. 1-face, of P has lattice length 1
and is contained in some hyperplane which has height 1 with respect to the origin.

Corollary 23 Let X be a toric Fano variety of dimension ≥ 4. If X has isolated

singularities, then X is rigid.

In Sects. 14.4.4 and 14.4.5 we will study deformations of toric Fanos with isolated
singularities and of dimension 2 or 3.

14.4.4 Toric del Pezzo Surfaces

A del Pezzo surface is a Fano variety of dimension 2. A toric del Pezzo surface is
associated to a Fano polygon, which is a Fano polytope of dimension 2.
304 A. Petracci

Theorem 24 Every toric del Pezzo surface is smoothable.

Proof Let X be an arbitrary toric del Pezzo surface. It is well known that X
is a normal Cohen–Macaulay projective variety. By Demazure vanishing
[34, Theorem 2.9.3], H2 (O X ) = 0. By Lemma 20, H2 (TX ) = 0. Since X is normal
and of dimension 2, X has isolated singularities. By Theorem 7 it is enough to check
that the singularities of X are formally smoothable.
The singularities of X are cyclic quotient surface singularities. This kind of sin-
gularities is always smoothable; indeed, it is enough to pick the Artin component of
the base of the miniversal deformation [12]. 

Remark 25 When the canonical divisor of a normal variety X is not Cartier, flat
deformations of X are too wild for hoping to study moduli of varieties. For a
normal Q-Gorenstein non-Gorenstein variety X one should consider a subfunctor
of Def X which is made up of the deformations of X in which the canonical divi-
sor deforms well. This is the theory of Q-Gorenstein deformations, developed by
Kollár–Shepherd-Barron [64] (see also [1, 9, 45, 68]).
In the context of Q-Gorenstein deformations the analogous statement of
Theorem 24 is false: there exist non-Gorenstein toric del Pezzo surfaces which can-
not be deformed via Q-Gorenstein deformations to a smooth del Pezzo surface, e.g.
the weighted projective space P(1, 1, 3). Nonetheless, it is true that for Q-Gorenstein
deformations of del Pezzo surfaces there are no local-to-global obstructions [2,
Lemma 6]. Therefore, a del Pezzo surface is Q-Gorenstein smoothable if and only
if its singularities are Q-Gorenstein smoothable.
Since the main focus of this note is the study of deformations of Gorenstein toric
Fano threefolds, we will omit to discuss the theory of Q-Gorenstein deformations.
We refer the reader to [46, 82] for the study of toric del Pezzo surfaces which have Q-
Gorenstein smoothings.

14.4.5 Toric Fano Threefolds with Isolated Singularities

Theorem 26 Let X be a toric Fano variety of dimension 3 with isolated singularities.

Then X is smoothable if and only if its singularities are formally smoothable.

Proof By Proposition 6, if X is smoothable then its singularities are formally smooth-

able. Conversely, suppose that the singularities of X are formally smoothable. Then
we argue as in the proof of Theorem 24: X is a normal Cohen–Macaulay projective
variety with H2 (O X ) = 0, by [34, Theorem 2.9.3], and H2 (TX ) = 0, by Lemma 20.
By Theorem 7, X is smoothable. 

Corollary 27 Let P be a reflexive polytope of dimension 3 and let X be the toric

Fano threefold associated to the spanning fan of P. If each facet of P is either a
standard triangle or a standard square (see the definition in Example 13), then X is
smoothable.
14 On Deformations of Toric Fano Varieties 305

Proof By Example 13 we have that the singularities of X are at most ordinary double
points (i.e. nodes). These singularities are formally smoothable. By Theorem 26 we
conclude. 

The proof of this corollary is essentially a specific case of [39, Sect. 4.a]. The
corollary could have been deduced also from a more general result by Namikawa
according to which every Fano threefold with Gorenstein terminal singularities is
smoothable [76]. The smooth Fano threefolds which are the smoothings of the toric
Fano threefold appearing in Corollary 27 have been studied by Galkin [42].
For d ∈ {6, 7}, let dPd be the smooth del Pezzo surface of degree d; it is toric. The
complete anticanonical linear system on dPd induces a closed embedding dPd → Pd .
We consider the projective cone C(dPd ) ⊆ Pd+1 over this embedding; we have
that C(dPd ) is a toric Fano threefold with a Gorenstein canonical non-terminal iso-
lated singularity. In Sect. 14.4.6 we will see that C(dP7 ) is smoothable. In [80] it is
shown that C(dP6 ) has two smoothings (see also [56, Example 3.3]).

14.4.6 The Projective Cone over the Del Pezzo Surface of

Degree 7

Here we study the deformations of an explicit toric Fano threefold with an isolated
Gorenstein non-terminal singularity.
Fix the lattice N = Z2 . Consider the pentagon F ⊆ N R defined in (14.2), imagine
to put it into the plane N R × {1} in N R ⊕ R  R3 , and create the pyramid over it
with apex at the point (0, 0, −1): this is the polytope
⎧⎛ ⎞ ⎛ ⎞ ⎛ ⎞ ⎛ ⎞ ⎛ ⎞ ⎛ ⎞⎫
⎨ 1 1 0 −1 0 0 ⎬
P = conv ⎝0⎠ , ⎝1⎠ , ⎝1⎠ , ⎝ 0 ⎠ , ⎝−1⎠ , ⎝ 0 ⎠ (14.5)
⎩ ⎭
1 1 1 1 1 −1

in the lattice N ⊕ Z and is depicted in Fig. 14.3. It is clear that P is a Fano polytope.

Fig. 14.3 The

3-dimensional lattice
polytope P defined in (14.5)
and associated to the
projective cone over the
del Pezzo surface of degree 7
306 A. Petracci

Let X be the toric variety associated to the spanning fan of P. Then X is the
projective cone over the anticanonical embedding of the smooth del Pezzo surface
of degree 7. The affine toric variety U F considered in Sect. 14.3.2 is the affine open
toric subscheme of X associated to the pentagonal facet F of P. We have that X is a
Fano threefold with an isolated non-terminal canonical Gorenstein singularity at the
vertex of the cone.
Proposition 28 Let X be the toric Fano threefold associated to the polytope P in
(14.5), i.e. X is the projective cone over the anticanonical embedding of the smooth
del Pezzo surface of degree 7. Then X is smoothable and can be deformed to the
smooth Fano threefold P(OP2 ⊕ OP2 (1)).
Proof By Proposition 16, X has an isolated singularity which is formally smoothable.
By Theorem 26 we know that X is smoothable. We need to know to which smooth
Fano threefold X can be deformed.
From toric geometry [34, Theorem 13.4.3], we have that the anticanonical
degree (−K X )3 is the normalised volume of the polar polytope of P, which is 56 in
this case. Since X has Gorenstein canonical singularities, by Proposition 10 we have
that the anticanonical degree is preserved in the smoothing. By inspecting the list
of smooth Fano threefolds (see [53, 54, 72–74] or [55, Sect. 12]), there is a unique
smooth Fano threefold of anticanonical degree 56, namely P(OP2 ⊕ OP2 (1)). 

14.4.7 Another Sufficient Condition for Non-smoothability

In addition to the result of Proposition 19, here we present another obstruction for
the smoothability of a toric Fano threefold with Gorenstein singularities.
Theorem 29 ([79]) Let N be a lattice of rank 3, let M = HomZ (N , Z), let ·, · : M
× N → Z be the duality pairing, let P be a reflexive polytope in N , and let X be the
toric Fano threefold associated to the spanning fan of P. Assume that there are two
adjacent facets F0 and F1 of P such that:
1. both F0 and F1 are An -triangles for some integer n ≥ 1 (see the definition in
Example 13);
2. F0 ∩ F1 is a segment with n + 2 lattice points;
3. w1 , v0  = 0, where w1 ∈ M is such that F1 ⊆ {v ∈ NR | w1 , v = 1} and v0 ∈
N is the vertex of F0 which does not lie on the segment F0 ∩ F1 .
Then X is not smoothable.
With the terminology of [79], the two triangles F0 and F1 are called “two adjacent
almost-flat An -triangles”.
Proof (Sketch of the proof of Theorem 29) We refer the reader to [79] for all the
details missing here. Let Ui be the toric open affine subscheme of X associated to
the facet Fi , for each i = 0, 1. Set U = U0 ∪ U1 .
14 On Deformations of Toric Fano Varieties 307

One can show that U admits an An -bundle structure over P1 . More precisely, one
can construct a toric morphism π : U → P1 such that, for each i = 0, 1, if Vi denotes
the ith standard affine chart of P1 then π −1 (Vi ) = Ui and the restriction π |Ui : Ui →
Vi is the projection Spec C[x, y, z, w]/(x y − z n+1 ) → Spec C[w]. This An -bundle
may be non trivial, depending on the relative position of the two triangles F0 and F1 .
Set d = w1 , v0 . By [79, Proposition 3.5] there exists an isomorphism of coherent
sheaves on P1 : 
π∗ ExtO1 U (U , OU )  OP1 (− jd − j).
2≤ j≤n+1

Since d = 0, the sheaf on the right is a direct sum of negative line bundles on P1 ,
hence we have H0 (U, ExtO1 U (U , OU )) = 0. By Corollary 9, X is not smoothable.


With the same technique of the theorem above one can also construct some rigid
toric Fano threefolds with only c A1 -singularities (see [79, Theorem 1.2]). This refutes
a conjecture of Prokhorov [85] according to which every Fano threefold with com-
pound Du Val singularities is smoothable.

14.4.8 Other Methods

Here we briefly collect some other results on deformations and smoothings of toric
Fano varieties. Most of these results have been motivated by Mirror Symmetry for
Fano varieties (see [2, 3, 26, 30, 31, 61, 86, 87]).
By analysing cluster transformations of tori, Akhtar–Coates–Galkin–Kasprzyk [3]
have introduced the notion of mutation of Fano polytopes. A mutation is a combi-
natorial procedure that, under certain conditions, transforms a Fano polytope P into
another Fano polytope P  . Ilten [51] has proved that mutations of Fano polytopes
induce deformations of the corresponding toric Fano varieties; more precisely, if P
and P  are related via a mutation, then he has constructed a flat family over P1 such
that the fibre over 0 is X P and the fibre over ∞ is X P  .
Ilten, Lewis and Przyjalkowski [52] have constructed toric degenerations of
smooth Fano threefolds with Picard rank 1.
Christophersen and Ilten [29] have constructed degenerations of smooth Fano
threefolds of low degree to certain unobstructed Fano Stanley–Reisner schemes.
Since these unobstructed Fano Stanley–Reisner schemes are also degenerations of
singular toric Fano varieties, this implies the following result.

Theorem 30 (Christophersen–Ilten [28, Proposition 4.2, Theorems 5.1, 7.1]) Let X

be a toric Fano threefold with Gorenstein singularities. If (−K X )3 ∈ {4, 6, 8, 10, 12},
then X is smoothable.

Coates–Kasprzyk–Prince [32] have introduced a combinatorial gadget, called scaf-

folding, on a Fano polytope P which induces a closed embedding of the toric Fano
308 A. Petracci

variety X P into a bigger toric variety Y . Often X P is a complete intersection in the Cox
coordinates of Y , therefore it is easy to construct embedded deformations of X P in Y .
In many cases this produces smoothings of X P . For instance, Cavey and Prince [25]
have successfully applied the scaffolding method to construct deformations of toric
del Pezzo surfaces to del Pezzo surfaces with a single k1 (1, 1) singularity.
Moreover, Prince [83] has found necessary and sufficient conditions in order to
have that the ambient toric variety Y is smooth: this is the notion of cracked polytope.
He has also found a sufficient condition for a smoothing of X P to exist inside Y . Via
the scaffolding method and cracked polytopes, in [84] he constructs a degeneration
of each smooth Fano threefold with very ample anticanonical bundle and Picard
rank ≥ 2 to a Gorenstein toric Fano threefold.

14.5 Lists of Reflexive Polytopes of Dimension 3

Below we write lists of reflexive polytopes of dimension 3 which satisfy specific

properties. There are exactly 4319 reflexive polytopes of dimension 3: the classifi-
cation is due to Kreuzer and Skarke [66]. The IDs we use are numbers between 1
and 4319 and come from the Graded Ring Database [23]. All polytopes we consider
below are reflexive of dimension 3. They correspond to toric Fano threefolds with
Gorenstein singularities. We denote by X P the toric Fano threefold associated to the
spanning fan of P.
Let Ssmoothable be the set of polytopes P such that the corresponding toric Fano
threefold X P is smoothable. It is an open question to explicitly compute Ssmoothable .
Let Ssmooth be the set of polytopes which have only standard triangles as facets.
These 18 polytopes correspond to the smooth toric Fano threefolds.
Let Sisol be the set of polytopes with unitary edges such that at least one facet is
not a standard triangle. These 137 polytopes correspond to the singular toric Fano
threefolds with isolated Gorenstein singularities.
Let Snodes be the set of polytopes such that all facets are either standard triangles
or standard squares and there is at least a square facet. These 82 polytopes correspond
to the singular toric Fano threefolds with at most ordinary double points, or equiv-
alently to the singular toric Fano threefolds with Gorenstein terminal singularities.
By Corollary 27 these varieties are smoothable.
Let Slow be the set of polytopes P such that the normalised volume of the polar P ∗
of P belongs to {4, 6, 8, 10, 12}. These 220 polytopes correspond to the toric Goren-
stein Fano threefolds X such that (−K X )3 ∈ {4, 6, 8, 10, 12}.
Let Sindec be the set of polytopes which contain a facet F which has unitary edges,
is Minkowski indecomposable and is not a standard triangle. By Proposition 19 the
corresponding toric Fano threefolds are not smoothable.
Let Saft be the set of polytopes which contain a pair of adjacent almost-flat An -
triangles, for some n ≥ 1. In other words, the set Saft contains exactly all polytopes P
to which Theorem 29 applies. Therefore, the corresponding toric Fano threefolds are
not smoothable.
14 On Deformations of Toric Fano Varieties 309

Let S denote the set of all reflexive polytopes of dimension 3, i.e. the set of positive
integers not greater than 4319. We have:

Snodes ⊆ Sisol ⊆ S \ Ssmooth ,

Ssmooth ∪ Snodes ∪ Slow ⊆ Ssmoothable ,
Sindec ∪ Saft ⊆ S \ Ssmoothable .

Below we write down the elements of most of the sets mentioned above.
Ssmooth = {1, 5, 6, 7, 8, 25, 26, 27, 28, 29, 30, 31, 82, 83, 84, 85, 219, 220}
Sisol = {3, 4, 11, 12, 17, 21, 22, 23, 24, 42, 48, 49, 50, 51, 54, 68, 69, 70, 71, 72,
73, 74, 75, 76, 77, 78, 79, 80, 81, 155, 156, 158, 159, 160, 167, 168, 170, 177, 187,
188, 198, 199, 200, 201, 202, 203, 204, 205, 206, 207, 208, 209, 210, 211, 212, 213,
214, 215, 216, 217, 218, 360, 363, 364, 365, 366, 376, 377, 378, 380, 385, 403, 410,
411, 412, 413, 414, 415, 416, 417, 418, 419, 420, 421, 422, 423, 424, 425, 426, 427,
686, 688, 689, 692, 693, 694, 695, 696, 707, 710, 725, 729, 730, 731, 732, 733, 734,
735, 736, 737, 738, 739, 740, 741, 1085, 1086, 1087, 1091, 1092, 1093, 1109, 1110,
1111, 1112, 1113, 1114, 1517, 1518, 1519, 1524, 1528, 1529, 1530, 1941, 1943,
2355, 2356}
Snodes = {4, 21, 22, 23, 24, 68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81,
198, 199, 200, 201, 202, 203, 204, 205, 206, 207, 208, 209, 210, 211, 212, 213, 214,
215, 216, 217, 218, 410, 411, 412, 413, 414, 415, 416, 417, 418, 419, 420, 421, 422,
423, 424, 425, 426, 427, 729, 730, 731, 732, 733, 734, 735, 736, 737, 738, 739, 740,
741, 1109, 1110, 1111, 1112, 1113, 1114, 1528, 1529, 1530, 1943, 2356}
Slow = {1946, 2711, 2756, 2817, 3043, 3051, 3053, 3079, 3314, 3319, 3329,
3331, 3349, 3350, 3390, 3393, 3406, 3416, 3447, 3452, 3453, 3505, 3573, 3620,
3625, 3626, 3667, 3683, 3702, 3727, 3728, 3731, 3733, 3735, 3736, 3738, 3739,
3740, 3756, 3760, 3762, 3777, 3790, 3791, 3792, 3795, 3796, 3844, 3845, 3846,
3848, 3853, 3857, 3868, 3869, 3874, 3875, 3879, 3901, 3903, 3922, 3923, 3927,
3928, 3933, 3936, 3937, 3938, 3946, 3962, 3964, 3965, 3966, 3967, 3981, 3983,
3984, 3985, 3991, 3995, 4003, 4004, 4005, 4006, 4007, 4022, 4023, 4024, 4027,
4031, 4032, 4041, 4042, 4043, 4044, 4056, 4058, 4059, 4060, 4070, 4074, 4075,
4076, 4080, 4088, 4092, 4094, 4095, 4102, 4104, 4117, 4118, 4119, 4122, 4124,
4131, 4132, 4133, 4134, 4135, 4143, 4144, 4145, 4149, 4159, 4160, 4161, 4167,
4168, 4169, 4170, 4179, 4180, 4181, 4182, 4183, 4184, 4186, 4190, 4191, 4194,
4200, 4202, 4203, 4205, 4206, 4214, 4215, 4216, 4217, 4218, 4219, 4220, 4225,
4228, 4229, 4231, 4232, 4233, 4235, 4236, 4238, 4239, 4241, 4244, 4245, 4246,
4247, 4249, 4250, 4251, 4252, 4254, 4255, 4256, 4258, 4260, 4261, 4263, 4267,
4268, 4269, 4270, 4272, 4273, 4275, 4278, 4280, 4281, 4282, 4284, 4285, 4286,
4287, 4288, 4290, 4291, 4292, 4293, 4294, 4295, 4297, 4298, 4299, 4300, 4301,
4303, 4304, 4307, 4308, 4309, 4310, 4311, 4312, 4313, 4314, 4315, 4317, 4318,
4319}
Sindec = {3, 12, 17, 32, 38, 48, 49, 51, 54, 88, 91, 94, 98, 99, 100, 101, 102, 103,
105, 115, 119, 121, 134, 137, 138, 141, 142, 155, 158, 159, 170, 188, 228, 235, 239,
242, 243, 247, 248, 252, 254, 256, 260, 262, 265, 271, 278, 293, 294, 298, 299, 301,
317, 318, 330, 351, 353, 360, 378, 380, 438, 439, 440, 443, 445, 455, 468, 480, 491,
310 A. Petracci

492, 493, 497, 501, 502, 515, 525, 526, 529, 530, 532, 539, 541, 543, 546, 550, 553,
562, 570, 575, 604, 608, 609, 614, 620, 645, 650, 660, 663, 688, 744, 752, 753, 754,
756, 760, 774, 775, 776, 780, 784, 790, 791, 792, 800, 834, 841, 844, 845, 852, 856,
859, 864, 866, 887, 900, 908, 912, 914, 923, 935, 963, 979, 990, 991, 1012, 1019,
1020, 1130, 1151, 1154, 1183, 1199, 1204, 1205, 1208, 1215, 1218, 1220, 1261,
1275, 1277, 1283, 1299, 1302, 1309, 1311, 1352, 1370, 1384, 1397, 1547, 1585,
1598, 1631, 1636, 1638, 1679, 1683, 1687, 1693, 1728, 1750, 1751, 1777, 1791,
1992, 2014, 2046, 2047, 2050, 2051, 2080, 2081, 2084, 2096, 2124, 2129, 2379,
2404, 2425, 2427, 2455, 2456, 2716, 2750, 2751, 2755}
Saft = {15, 16, 36, 41, 45, 53, 58, 59, 61, 65, 66, 102, 105, 110, 111, 112, 113,
116, 117, 124, 125, 128, 135, 141, 142, 144, 146, 147, 148, 149, 152, 162, 172, 179,
183, 189, 192, 193, 197, 230, 236, 244, 248, 261, 268, 271, 272, 277, 278, 279, 280,
281, 282, 286, 288, 290, 292, 302, 310, 324, 325, 327, 331, 332, 333, 334, 335, 337,
340, 343, 347, 349, 351, 355, 356, 358, 361, 362, 386, 399, 400, 407, 443, 445, 448,
452, 453, 456, 457, 463, 467, 487, 490, 496, 497, 499, 501, 502, 505, 507, 508, 509,
511, 512, 516, 523, 540, 545, 550, 563, 569, 577, 579, 581, 582, 583, 594, 599, 600,
601, 605, 606, 617, 629, 633, 658, 670, 671, 672, 674, 679, 682, 687, 705, 760, 764,
770, 771, 780, 781, 786, 787, 792, 797, 799, 809, 811, 812, 815, 816, 824, 859, 865,
868, 873, 875, 878, 883, 884, 889, 891, 892, 893, 894, 895, 902, 905, 929, 956, 960,
965, 987, 1003, 1004, 1006, 1011, 1021, 1038, 1045, 1051, 1156, 1160, 1168, 1175,
1177, 1199, 1203, 1209, 1216, 1217, 1225, 1232, 1234, 1251, 1252, 1253, 1255,
1256, 1260, 1262, 1265, 1275, 1286, 1287, 1293, 1300, 1305, 1308, 1324, 1327,
1351, 1371, 1383, 1398, 1533, 1545, 1550, 1551, 1554, 1561, 1579, 1589, 1613,
1614, 1615, 1620, 1637, 1638, 1656, 1665, 1666, 1671, 1686, 1690, 1693, 1697,
1711, 1747, 1748, 1760, 1763, 1989, 2000, 2001, 2027, 2045, 2051, 2052, 2068,
2071, 2072, 2076, 2084, 2096, 2098, 2102, 2379, 2380, 2385, 2403, 2405, 2423,
2424, 2425, 2427, 2738, 2777, 2778, 2792, 3047, 3057, 3063, 3064}
We have |Sindec ∪ Saft | = 442. Therefore there exist at least 442 non-smoothable
toric Fano threefolds with Gorenstein singularities.

Acknowledgements I am indebted to Alessio Corti, Paul Hacking, Alexander Kasprzyk, and

Thomas Prince for many fruitful conversations about the topics of this survey. I am very grate-
ful to Tom Coates and Alexander Kasprzyk for having provided me with access to their Magma
machine and to their database of Fano polytopes [33]. Finally, I would like to thank the anonymous
referee for having done a thorough job.

References

1. Abramovich, D., Hassett, B.: Stable varieties with a twist. In: Classification of algebraic vari-
eties, EMS Series of Congress Reports, pp. 1–38. European Mathematical Society, Zürich
(2011)
2. Akhtar, M., Coates, T., Corti, A., Heuberger, L., Kasprzyk, A.M., Oneto, A., Petracci, A.,
Prince, T., Tveiten, K.: Mirror symmetry and the classification of orbifold del Pezzo surfaces.
Proc. Amer. Math. Soc. 144(2), 513–527 (2016)
14 On Deformations of Toric Fano Varieties 311

3. Akhtar, M., Coates, T., Galkin, S., Kasprzyk, A.M.: Minkowski polynomials and mutations.
SIGMA Symmetry Integr. Geom. Methods Appl. 8, Paper 094, 17 (2012)
4. Altmann, K.: Computation of the vector space T 1 for affine toric varieties. J. Pure Appl. Algebra
95(3), 239–259 (1994)
5. Altmann, K.: Minkowski sums and homogeneous deformations of toric varieties. Tohoku Math.
J. (2) 47(2), 151–184 (1995)
6. Altmann, K.: Infinitesimal deformations and obstructions for toric singularities. J. Pure Appl.
Algebra 119(3), 211–235 (1997)
7. Altmann, K.: The versal deformation of an isolated toric Gorenstein singularity. Invent. Math.
128(3), 443–479 (1997)
8. Altmann, K.: One parameter families containing three-dimensional toric-Gorenstein singular-
ities. In: Explicit birational geometry of 3-folds. London Mathematical Society Lecture Note
series, vol. 281, pp. 21–50. Cambridge University Press, Cambridge (2000)
9. Altmann, K., Kollár, J.: The dualizing sheaf on first-order deformations of toric surface singu-
larities. J. Reine Angew. Math. 753, 137–158 (2019)
10. Artin, M.: Algebraic approximation of structures over complete local rings. Inst. Hautes Études
Sci. Publ. Math. 36, 23–58 (1969)
11. Artin, M.: Algebraization of formal moduli. I. In: Global Analysis (Papers in Honor of K.
Kodaira), pp. 21–71. University Tokyo Press, Tokyo (1969)
12. Artin, M.: Algebraic construction of Brieskorn’s resolutions. J. Algebra 29, 330–348 (1974)
13. Artin, M.: Lectures on Deformations of Singularities. Tata Institute of Fundamental Research,
Bombay (1976)
14. The Stacks project (2018). https://stacks.math.columbia.edu
15. Batyrev, V.V.: Toric Fano threefolds. Izv. Akad. Nauk SSSR Ser. Mat. 45(4), 704–717, 927
(1981)
16. Batyrev, V.V.: On the classification of toric Fano 4-folds. pp. 1021–1050 (1999). Algebraic
geometry, 9
17. Batyrev, V.V.: Toric degenerations of Fano varieties and constructing mirror manifolds. In: The
Fano Conference, pp. 109–122. University Torino, Turin (2004)
18. Batyrev, V.V., Ciocan-Fontanine, I., Kim, B., van Straten, D.: Mirror symmetry and toric degen-
erations of partial flag manifolds. Acta Math. 184(1), 1–39 (2000)
19. Behnke, K., Riemenschneider, O.: Quotient surface singularities and their deformations. In:
Singularity Theory (Trieste, 1991), pp. 1–54. World Scientific Publishing, River Edge (1995)
20. Bien, F., Brion, M.: Automorphisms and local rigidity of regular varieties. Compositio Math.
104(1), 1–26 (1996)
21. Birkar, C.: Singularities of linear systems and boundedness of Fano varieties. Ann. of Math.
(2) 193(2), 347–405 (2021)
22. Bosma, W., Cannon, J., Playoust, C.: The Magma algebra system. I. The user language. pp.
235–265 (1997). Computational algebra and number theory (London, 1993)
23. Brown, G., Kasprzyk, A.M.: Graded Ring Database. http://www.grdb.co.uk
24. Buch, A., Thomsen, J.F., Lauritzen, N., Mehta, V.: The Frobenius morphism on a toric variety.
Tohoku Math. J. (2) 49(3), 355–366 (1997)
25. Cavey, D., Prince, T.: Del Pezzo surfaces with a single 1/k(1, 1) singularity. J. Math. Soc.
Japan 72(2), 465–505 (2020)
26. Cheltsov, I., Katzarkov, L., Przyjalkowski, V.: Birational geometry via moduli spaces. In: Bira-
tional Geometry, Rational Curves, and Arithmetic, Simons Symposia, pp. 93–132. Springer,
Cham (2013)
27. Christophersen, J.A.: On the components and discriminant of the versal base space of cyclic
quotient singularities. In: Singularity theory and its applications, Part I (Coventry, 1988/1989).
Lecture Notes in Math., vol. 1462, pp. 81–92. Springer, Berlin (1991)
28. Christophersen, J.A., Ilten, N.: Hilbert schemes and toric degenerations for low degree Fano
threefolds. J. Reine Angew. Math. 717, 77–100 (2016)
29. Christophersen, J.A., Ilten, N.O.: Degenerations to unobstructed Fano Stanley-Reisner
schemes. Math. Z. 278(1–2), 131–148 (2014)
312 A. Petracci

30. Coates, T., Corti, A., Galkin, S., Golyshev, V., Kasprzyk, A.M.: Mirror symmetry and Fano
manifolds. In: European Congress of Mathematics, pp. 285–300. European Mathematical Soci-
ety, Zürich (2013)
31. Coates, T., Corti, A., Galkin, S., Kasprzyk, A.M.: Quantum periods for 3-dimensional Fano
manifolds. Geom. Topol. 20(1), 103–256 (2016)
32. Coates, T., Kasprzyk, A., Prince, T.: Laurent inversion. Pure Appl. Math. Q. 15(4), 1135–1179
(2019)
33. Coates, T., Kasprzyk, A.M.: Code repository (2019). https://bitbucket.org/fanosearch/magma-
core
34. Cox, D.A., Little, J.B., Schenck, H.K.: Toric varieties. Graduate Studies in Mathematics, vol.
124. American Mathematical Society, Providence (2011)
35. Eisenbud, D.: Commutative algebra. Graduate Texts in Mathematics, vol. 150. Springer, New
York (1995). With a view toward algebraic geometry
36. Fantechi, B., Göttsche, L.: Local properties and Hilbert schemes of points. In: Fundamental
Algebraic Geometry. Mathematical Surveys and Monographs, vol. 123, pp. 139–178. American
Mathematical Society, Providence (2005)
37. Fantechi, B., Manetti, M.: Obstruction calculus for functors of Artin rings. I. J. Algebra 202(2),
541–576 (1998)
38. de Fernex, T., Hacon, C.D.: Deformations of canonical pairs and Fano varieties. J. Reine Angew.
Math. 651, 97–126 (2011)
39. Friedman, R.: Simultaneous resolution of threefold double points. Math. Ann. 274(4), 671–689
(1986)
40. Fujino, O.: Multiplication maps and vanishing theorems for toric varieties. Math. Z. 257(3),
631–641 (2007)
41. Fulton, W.: Introduction to toric varieties. Annals of Mathematics Studies, vol. 131. Princeton
University Press, Princeton (1993). The William H. Roever Lectures in Geometry
42. Galkin, S.: Small toric degenerations of Fano threefolds (2018). arXiv:1809.02705 [math.AG]
43. Gross, M., Siebert, B.: From real affine geometry to complex geometry. Ann. of Math. (2)
174(3), 1301–1428 (2011)
44. Grothendieck, A.: Éléments de géométrie algébrique. IV. Étude locale des schémas et des
morphismes de schémas. III. Inst. Hautes Études Sci. Publ. Math. (28), 255 (1966)
45. Hacking, P.: Compact moduli of plane curves. Duke Math. J. 124(2), 213–257 (2004)
46. Hacking, P., Prokhorov, Y.: Smoothable del Pezzo surfaces with quotient singularities. Compos.
Math. 146(1), 169–192 (2010)
47. Hartshorne, R.: Residues and duality. Lecture notes of a seminar on the work of A.
Grothendieck, given at Harvard 1963/64. With an appendix by P. Deligne. Lecture Notes in
Mathematics, No. 20. Springer, Berlin (1966)
48. Hartshorne, R.: Algebraic Geometry. Springer, New York (1977). Graduate Texts in Mathe-
matics, No. 52
49. Hartshorne, R.: Stable reflexive sheaves. Math. Ann. 254(2), 121–176 (1980)
50. Hartshorne, R.: Deformation theory. Graduate Texts in Mathematics, vol. 257. Springer, New
York (2010)
51. Ilten, N.O.: Mutations of Laurent polynomials and flat families with toric fibers. SIGMA
Symmetry Integr. Geom. Methods Appl. 8, Paper 047, 7 (2012)
52. Ilten, N.O., Lewis, J., Przyjalkowski, V.: Toric degenerations of Fano threefolds giving weak
Landau-Ginzburg models. J. Algebra 374, 104–121 (2013)
53. Iskovskih, V.A.: Fano threefolds. I. Izv. Akad. Nauk SSSR Ser. Mat. 41(3), 516–562, 717
(1977)
54. Iskovskih, V.A.: Fano threefolds. II. Izv. Akad. Nauk SSSR Ser. Mat. 42(3), 506–549 (1978)
55. Iskovskikh, V.A., Prokhorov, Y.G.: Fano varieties. In: Algebraic Geometry, V, Encylopaedia
of Mathematical Sciences, vol. 47, pp. 1–247. Springer, Berlin (1999)
56. Jahnke, P., Radloff, I.: Terminal Fano threefolds and their smoothings. Math. Z. 269(3–4),
1129–1136 (2011)
14 On Deformations of Toric Fano Varieties 313

57. Kasprzyk, A.M.: Toric Fano three-folds with terminal singularities. Tohoku Math. J. (2) 58(1),
101–121 (2006)
58. Kasprzyk, A.M.: Canonical toric Fano threefolds. Canad. J. Math. 62(6), 1293–1309 (2010)
59. Kasprzyk, A.M., Kreuzer, M., Nill, B.: On the combinatorial classification of toric log del
Pezzo surfaces. LMS J. Comput. Math. 13, 33–46 (2010)
60. Kasprzyk, A.M., Nill, B.: Fano polytopes. In: Strings. Gauge Fields, and the Geometry Behind,
pp. 349–364. World Scientific Publishing, Hackensack (2013)
61. Katzarkov, L., Przyjalkowski, V.: Landau-Ginzburg models—old and new. In: Proceedings of
the Gökova Geometry-Topology Conference 2011, pp. 97–124. International Press, Somerville,
MA (2012)
62. Kollár, J., Miyaoka, Y., Mori, S.: Rational connectedness and boundedness of Fano manifolds.
J. Differ. Geom. 36(3), 765–779 (1992)
63. Kollár, J., Mori, S.: Birational geometry of algebraic varieties. Cambridge Tracts in Mathemat-
ics, vol. 134. Cambridge University Press, Cambridge (1998). With the collaboration of C. H.
Clemens and A. Corti, Translated from the 1998 Japanese original
64. Kollár, J., Shepherd-Barron, N.I.: Threefolds and deformations of surface singularities. Invent.
Math. 91(2), 299–338 (1988)
65. Kreuzer, M., Nill, B.: Classification of toric Fano 5-folds. Adv. Geom. 9(1), 85–97 (2009)
66. Kreuzer, M., Skarke, H.: Classification of reflexive polyhedra in three dimensions. Adv. Theor.
Math. Phys. 2(4), 853–871 (1998)
67. Kreuzer, M., Skarke, H.: Complete classification of reflexive polyhedra in four dimensions.
Adv. Theor. Math. Phys. 4(6), 1209–1230 (2000)
68. Lee, Y., Nakayama, N.: Grothendieck duality and Q-Gorenstein morphisms. Publ. Res. Inst.
Math. Sci. 54(3), 517–648 (2018)
69. Li, C., Wang, X., Xu, C.: On the proper moduli spaces of smoothable Kähler-Einstein Fano
varieties. Duke Math. J. 168(8), 1387–1459 (2019)
70. Manetti, M.: Differential graded Lie algebras and formal deformation theory. In: Algebraic
Geometry—Seattle 2005. Part 2. Proceedings of Symposia in Pure Mathematics, vol. 80, pp.
785–810. American Mathematical Society, Providence (2009)
71. Mavlyutov, A.R.: Deformations of toric varieties via Minkowski sum decompositions of poly-
hedral complexes (2009). arXiv:0902.0967 [math.AG]
72. Mori, S., Mukai, S.: Classification of Fano 3-folds with B2 ≥ 2. Manuscripta Math. 36(2),
147–162 (1981/82)
73. Mori, S., Mukai, S.: On Fano 3-folds with B2 ≥ 2. In: Algebraic Varieties and Analytic Varieties
(Tokyo, 1981). Advanced Studies in Pure Mathematics, vol. 1, pp. 101–129. North-Holland,
Amsterdam (1983)
74. Mori, S., Mukai, S.: Erratum: “Classification of Fano 3-folds with B2 ≥ 2” [Manuscripta Math.
36(2), 147–162 (1981/82); MR0641971 (83f:14032)]. Manuscripta Math. 110(3), 407 (2003)
75. Mustaţă, M.: Vanishing theorems on toric varieties. Tohoku Math. J. (2) 54(3), 451–470 (2002)
76. Namikawa, Y.: Smoothing Fano 3-folds. J. Algebraic Geom. 6(2), 307–324 (1997)
77. Nill, B., Øbro, M.: Q-factorial Gorenstein toric Fano varieties with large Picard number. Tohoku
Math. J. (2) 62(1), 1–15 (2010)
78. Øbro, M.: An algorithm for the classification of smooth Fano polytopes (2007).
arXiv:0704.0049 [math.CO]
79. Petracci, A.: Some examples of non-smoothable Gorenstein Fano toric threefolds. Math. Z.
295(1–2), 751–760 (2020)
80. Petracci, A.: An example of mirror symmetry for Fano threefolds. In: Birational geometry and
moduli spaces. Springer INdAM Serirs, vol. 39, pp. 173–188. Springer, Cham ([2020] © 2020)
81. Petracci, A.: Homogeneous deformations of toric pairs. Manuscripta Math. 166(1–2), 37–72
(2021)
82. Prince, T.: Smoothing toric Fano surfaces using the Gross-Siebert algorithm. Proc. Lond. Math.
Soc. (3) 117(3), 617–660 (2018)
83. Prince, T.: Cracked polytopes and Fano toric complete intersections. Manuscripta Math. 163(1–
2), 165–183 (2020)
314 A. Petracci

84. Prince, T.: From cracked polytopes to Fano threefolds. Manuscripta Math. 164(1–2), 267–320
(2021)
85. Prokhorov, Y.G.: The degree of Fano threefolds with canonical Gorenstein singularities. Mat.
Sb. 196(1), 81–122 (2005)
86. Przhiyalkovskiı̆, V.V.: Weak Landau-Ginzburg models of smooth Fano threefolds. Izv. Ross.
Akad. Nauk Ser. Mat. 77(4), 135–160 (2013)
87. Przyjalkowski, V.: On Landau-Ginzburg models for Fano varieties. Commun. Number Theory
Phys. 1(4), 713–728 (2007)
88. Riemenschneider, O.: Deformationen von Quotientensingularitäten (nach zyklischen Grup-
pen). Math. Ann. 209, 211–248 (1974)
89. Sato, H.: Toward the classification of higher-dimensional toric Fano varieties. Tohoku Math.
J. (2) 52(3), 383–413 (2000)
90. Sato, H.: Smooth toric Fano five-folds of index two. Proc. Japan Acad. Ser. A Math. Sci. 82(7),
106–110 (2006)
91. Schlessinger, M.: Functors of Artin rings. Trans. Amer. Math. Soc. 130, 208–222 (1968)
92. Sernesi, E.: Deformations of algebraic schemes, Grundlehren der Mathematischen Wis-
senschaften [Fundamental Principles of Mathematical Sciences], vol. 334. Springer, Berlin
(2006)
93. Stevens, J.: On the versal deformation of cyclic quotient singularities. In: Singularity Theory
and Its Applications, Part I (Coventry, 1988/1989). Lecture Notes in Mathematics, vol. 1462,
pp. 302–319. Springer, Berlin (1991)
94. Stevens, J.: The versal deformation of cyclic quotient singularities. In: Deformations of Sur-
face Singularities. Bolyai Society Mathematical Studies, vol. 23, pp. 163–201. János Bolyai
Mathematical Society, Budapest (2013)
95. Talpo, M., Vistoli, A.: Deformation theory from the point of view of fibered categories. In:
Handbook of Moduli. Vol. III, Advanced Lectures in Mathematics (ALM), vol. 26, pp. 281–397.
International Press, Somerville (2013)
96. Totaro, B.: Jumping of the nef cone for Fano varieties. J. Algebraic Geom. 21(2), 375–396
(2012)
97. Vistoli, A.: The deformation theory of local complete intersections (1997).
arXiv:alg-geom/9703008
98. Watanabe, K., Watanabe, M.: The classification of Fano 3-folds with torus embeddings. Tokyo
J. Math. 5(1), 37–48 (1982)
99. Ziegler, G.M.: Lectures on Polytopes. Graduate Texts in Mathematics, vol. 152. Springer, New
York (1995)
Chapter 15
Polygons of Finite Mutation Type

Thomas Prince

Abstract We classify Fano polygons with finite mutation class. This classification
exploits a correspondence between Fano polygons and cluster algebras, refining the
notion of singularity content due to Akhtar and Kasprzyk. We also introduce examples
of cluster algebras associated to Fano polytopes in dimensions greater than two.

Keywords Polytopes · Toric Varieties · Mirror Symmetry

15.1 Introduction

The notion of combinatorial, or polytope, mutation was introduced by Akhtar–

Coates–Galkin–Kasprzyk [3] to describe mirror partners to Fano manifolds. Fol-
lowing Givental [17–19], Kontsevich [28], and Hori–Vafa [25], the mirror partner
to a Fano manifold consists of a complex manifold together with a holomorphic
function, the superpotential. If this mirror manifold contains a complex torus we
can write down a collection of volume preserving birational maps of this complex
torus which preserve the regularity of the superpotential. We call these rational maps
(algebraic) mutations, following [3] and work of Galkin–Usnich [16]. Combinatorial
mutation is the operation induced on the Newton polyhedra of the restriction of the
superpotential to such tori.
All the polytopes we consider are Fano, that is, polytopes which contain the origin
in the interior and such that its vertices are primitive lattice vectors. In joint work [27]
with Kasprzyk and Nill we showed that, in dimension two, the notion of polytope
mutation is compatible with the construction of a quiver and cluster algebras one can
associate to each Fano polygon.
The idea of associating a polygon with a quiver—or toric diagram—has a reason-
ably long history, particularly in the physics literature. In that setting the polygon
describes a toric Calabi–Yau singularity and the quiver is used to describe the matter

T. Prince (B)
Mathematical Institute, University of Oxford, Woodstock Road, Oxford OX2 6GG, UK
e-mail: [email protected]

© Springer Nature Switzerland AG 2022 315

A. M. Kasprzyk and B. Nill (eds.), Interactions with Lattice Polytopes,
Springer Proceedings in Mathematics & Statistics 386,
https://doi.org/10.1007/978-3-030-98327-7_15
316 T. Prince

content of a gauge theory arising on a stack of D3-branes probing the toric Calabi–
Yau singularity, (see for example [1, 5, 10, 15, 21, 22, 29] for a selection of the
literature on this subject). The construction of a quiver (and cluster algebra) from a
polygon has also been used by Gross–Hacking–Keel [20] in the study of associated
log Calabi–Yau varieties, and to study the derived category of the toric variety, or the
associated local toric Calabi–Yau as pursued, for example, in [7, 23, 24, 31, 32]. In
each setting the basic construction is the same, and we recall the version relevant to
our applications in Sect. 15.3.
Our main result, Theorem 30, is a classification of the mutation classes of polygons
which contain only finitely many polygons. This parallels a finite type result of
Mandel [30], for rank two cluster varieties. In particular we see that finite mutation
classes of polygons fall into four types An1 , for n ∈ Z≥0 , A2 , A3 , and D4 .
There is a close connection between mutation classes of Fano polygons and Q-
Gorenstein deformations of the corresponding toric varieties which is described in
detail in [2]. Following these ideas we predict the existence of a finite type param-
eter space for these deformations, together with a boundary stratification such that
each 0-stratum corresponds to a polygon in the given mutation class, and the 1-strata
corresponds to the mutation families constructed by Ilten [26].
While our main result applies in dimension two, we note that polytope mutation
is defined in all dimensions, and the construction of a quiver and cluster algebra
we provide applies to ‘compatible collection’ of mutations in any dimension, see
Definition 16. This definition is, unfortunately, less well behaved in dimensions
greater than two, but we provide an example indicating that polytope mutation can
detect known examples of cluster structures appearing on linear sections of Grass-
mannians of planes. We expect this to extend to a wide variety of other cluster
structures found in Fano manifolds and their mirror manifolds.

15.2 Quivers and Cluster Algebras

We devote this section to fixing the various conventions and notation, as well as
recalling the basic definitions. We recall the definition of cluster algebra, and in
order to address both geometric and combinatorial applications we shall adapt our
treatment from the work of Fomin–Zelevinsky [13], and the work of Fock–Goncharov
[11] and Gross–Hacking–Keel [20]. We first fix the following data:
1. N , a fixed lattice with skew-symmetric form {−, −} : N × N → Z;
2. a saturated sublattice Nu f ⊆ N , the unfrozen sublattice;
3. an index set I , |I | = rk(N ) together with a subset Iu f ⊆ I such that |Iu f | =
rk(Nu f ). For later convenience we set n := |Iu f |.
Remark 1 The requirement that the form is integral is not necessary, but is suffi-
ciently general for our applications and simplifies the exposition considerably.
Definition 2 A (labelled) seed is a pair s = (E, C), where:
1. E is a basis of N indexed by I , such that the subset indexed by Iu f is a basis
of Nu f ;
15 Polygons of Finite Mutation Type 317

2. C is a transcendence basis of F , the field of rational functions in n independent

variables over Q(xi : i ∈ I \Iu f ), referred to as a cluster. Elements of C are also
referred to as cluster variables.
Remark 3 The basis E is referred to as seed data in [11, 20]. Since we have fixed
the lattice N and skew-symmetric form {−, −} the elements of C can be identified
with coordinate functions on the seed torus TN .
Definition 4 Fix a seed s = (E, C), where E = {ei : i ∈ I } and C = {xi : i ∈ Iu f }.
Fixing an element j ∈ Iu f , the jth mutation of (E, C) is the seed (E , C  ),
where E = {ei : i ∈ I } is defined by setting

−e j if k = j,
ek =  
ek + max bk j , 0 e j otherwise,

where bkl = {ek , el }. While the cluster C  = {xi : i ∈ Iu f } is defined by setting,

 
xk = xk , if k = j, x j x j =
b b
and xk jk + xl l j . (15.1)
k such that l such that
b jk >0 b jl <0

Recall that the matrix B := (bkl )k,l∈Iu f is typically referred to as the exchange
matrix of the seed.
Definition 5 A cluster algebra is the subalgebra of F generated by the union of all
clusters obtained by mutation from a given seed.
Any skew-symmetric n × n matrix B determines a skew-symmetric form on a
(based) lattice Zn . Set N = Zn , I = Iu f = {1, . . . , n}, E to be the standard basis
on Zn , and let C = {x1 , . . . , xn }. We let A(B) denote the cluster algebra associated
to the seed (E, C).
Definition 5 is really a special case of the definition of a cluster algebra, a class
referred to as the skew-symmetric cluster algebras of geometric type. In the general
case the form {−, −} need only be skew-symmetrizable. One consequence of the
skew-symmetry of the form {−, −} is the identification of each exchange matrix
with an (unfrozen) quiver. One may assign this quiver in the obvious way, assigning
a vertex vi to each element i ∈ Iu f , and bi j arrows vi → v j , oriented according to
the sign of bi j . We may also add ‘frozen’ vertices vi for each element of i ∈ I \Iu f ,
with arrows introduced between frozen and unfrozen vertices similarly. Equivalently
we may consider the quiver associated to the extended exchange matrix, but we do
not make further use of this terminology. There is a well-known notion of quiver
mutation, going back to Fomin–Zelevinsky [13], generalising the reflection functors
of Bernstein–Gelfand–Ponomarev [6]. Mutating a seed in a skew-symmetric cluster
algebra induces a corresponding mutation of the associated quiver.
Definition 6 Given a quiver Q and an element i ∈ Iu f , the mutation of Q at vi is
the quiver mut(Q, v) obtained from Q by:
318 T. Prince

1. adding, for each subquiver v1 → v → v2 , an arrow from v1 to v2 ;

2. deleting a maximal set of disjoint two-cycles;
3. reversing all arrows incident to v.
The resulting quiver is well-defined up to isomorphism, regardless of the choice of
two-cycles in (2).
Since we shall refer to quivers frequently we shall make the following conventions.
Given a quiver Q, we define:
1. Q 0 to be the set of vertices of Q;
2. Arr(vi , v j ) to be the set of arrows from vi ∈ Q 0 to v j ∈ Q 0 ;
3. bi j to be the cardinality of Arr(vi , v j ), with sign indicating orientation.
We shall always assume Q has no vertex-loops or 2-cycles.
Given a seed s we shall also fix notation for the dual basis E of M := hom(N , Z)
and for each i ∈ I , set vi := {ei , −} ∈ M. We now define the A and X cluster varieties
defined by Fock–Goncharov [11]. Toward this, observe to a seed s we can associate
a pair of tori:

Xs = T M As = TN .

The dual pair of bases for the respective lattices define identifications of these tori
with split tori,

Xs −→ G|Im | , As −→ G|Im | .

Letting s denote the kth mutation of s, we associate birational maps μk : Xs  Xs
and μk : As  As to each seed, defined by setting

μk z n = z n (1 + z ek )−{n,ek } μk z m = z m (1 + z vk )ek ,m ,

where n ∈ N , m ∈ M, z m (resp. z n ) denotes the function on TN := Spec(C[M])

corresponding to m (resp. the function on TM corresponding to n), and −, − denotes
the canonical pairing between M and N . Note that the tori Xs and Xs are canonically
identified with TN , but the maps μk between them are not isomorphisms.
Pulling these birational maps back along the identifications with the split torus
given by the seed, the birational map μk : As  As is given by the exchange
relation (15.1). That is, this birational map is the coordinate-free expression of the
exchange relation once we identify the standard coordinates on TN with the cluster
variables xi ∈ C (including the frozen variables xn+1 . . . , xm ). We obtain schemes X
and A by gluing the seed tori As and Xs along the birational maps defined by the
mutations μk . For more details and related results we refer to [11, 20].
We conclude this section by recalling the classifications of cluster algebras of finite
type and finite mutation type.
Definition 7 A cluster algebra is said to be of finite type if it contains finitely many
clusters.
15 Polygons of Finite Mutation Type 319

Given an undirected graph G we say that a quiver Q is an orientation of G if it has

the same set of vertices and for each edge of G there is precisely one arrow between
the respective vertices. Given a simply-laced Dynkin diagram D we say that Q is
of type D if it is an orientation of the underlying graph of D.

Theorem 8 ([14]) There is a canonical bijection between the Cartan matrices of

finite type and cluster algebras (without frozen variables) of finite type. Under this
bijection, a Cartan matrix A of finite type corresponds to the cluster algebra A(B),
where B is an arbitrary skew-symmetrizable matrix with Cartan companion equal
to A.

Theorem 8 describes skew-symmetric cluster algebras with finitely many clusters.

We can ask instead for the weaker condition that only finitely many quivers appear
associated to seeds of the cluster algebra. This is the notion of finite mutation type
cluster algebra, for which a classification is also known.

Theorem 9 ([9, Theorem 6.1]) Given a quiver Q with finite mutation class, its
adjacency matrix bi j is the adjacency matrix of a triangulation of a bordered surface
or is mutation equivalent to one of eleven exceptional types.

The class of quivers coming from triangulations of surfaces is well-studied and

we make use of a combinatorial characterisation of this class of quivers via block
decomposition. A quiver Q is said to admit a block decomposition if it may be
assembled from the six blocks shown in Fig. 15.5 by identifying the vertices of quivers
shown with unfilled circles, the outlets. More precisely, we choose an injection from
a subset of the combined set of outlets O into O such that no outlet is mapped
to a vertex of the same block, including itself. We form Q by gluing the quiver
along these vertices and cancelling any two cycles formed by this process. See [12,
Definition 13.1] for further discussion and examples of this definition.

Theorem 10 ([12, Theorem 13.3]) A quiver Q given by the adjacency matrix of a

triangulation of a surface is mutation equivalent to a quiver which admits a block
decomposition.

15.3 Mutations of Polytopes

In two dimensions all combinatorial mutations are ‘tropicalisations’ of cluster muta-

tions. While this ceases to be true in higher dimensions there is a natural class of
combinatorial mutations, the edge mutations which do appear in this way. In terms
of the definition of combinatorial mutation given in [3], edge mutations are those
which have one-dimensional factor. In particular each edge mutation is obtained by
studying the effect of the following birational maps—an algebraic mutation [3]—on
the Newton polyhedra of certain Laurent polynomials. Throughout this section N
denotes an n-dimensional lattice (not necessarily related to the definition of a cluster
320 T. Prince

algebra). We recall that, working over C, if M is the lattice dual to N , the torus TM
is defined to be Spec(C[N ]).
Definition 11 Given an element w ∈ M, the weight vector, and f ∈ Ann(w), the fac-
tor, define a birational map φw, f : TM  TM sending

z n
→ z n (1 + z f )w,n .


Given a Laurent polynomial W ∈ C[N ] such that φw, f (W ) ∈ C[N ] say that W is
mutable with weight vector w and factor f .
Definition 12 (Cf. [3, pg. 12]) Fix a Fano polytope P ⊂ NQ and its dual P  ⊂
MQ , a weight vector w ∈ M, and factor f ∈ Ann(w). Define a piecewise linear
map Tw, f : MQ → MQ by setting

Tw, f : m
→ m + max(0, m, f )w.

If Tw, f (P  ) is a convex polytope then we say P admits the mutation (w, f ) and
that P mutates to (Tw, f (P  )) .
Remark 13 This definition of mutation is really a ‘dual characterisation’ of [3,
Definition 5], which encodes how the Newton polytope of a Laurent polynomial
changes under algebraic mutation.
Remark 14 In [3] the authors show that the result of applying a mutation to a Fano
polytope produces another Fano polytope, so the last dualization in Definition 12 is
well-defined.
Proposition 15 Given w ∈ M, f ∈ Ann(w) and a mutable Laurent polynomial W ∈
C[N ] we have the following identity;
    
Newt φw, fW = Tw, f Newt(W ) .

Proof The notion of combinatorial mutation is compatible with the mutation W by

construction. The interpretation of a combinatorial mutation as a piecewise linear
map is made in the proof of Proposition 4 in [3]. 
Definition 16 We define mutation data to be elements (w, f ) ∈ M ⊕ N such that w
and f are primitive, and f ∈ Ann(w). A set of mutation data {(wi , f i ) ∈ M ⊕ N :
i ∈ I }, for a finite index set I , is called compatible if

wi , f j = −w j , f i for all i, j ∈ I.

Remark 17 If dim N = 2 mutation data is automatically compatible; indeed wi ,

f j can be identified with wi ∧ w j for a suitable orientation of M.
Definition 18 To a compatible collection of mutation data E we define a quiver Q E
as follows:
15 Polygons of Finite Mutation Type 321

1. the vertex set of Q E is E;

2. between two vertices (wi , f i ) and (w j , f j ) there are wi , f j arrows, with sign
indicating orientation.
Observe that, as wi , f j is skew-symmetric, the quiver Q E contains no loops
or two cycles. Note that we can use this definition to assign a cluster algebra to a
compatible collection of mutation data. We define a rule governing how compatible
collections of mutations themselves mutate.
Definition 19 Given a compatible collection of mutation data E, let L be the sublat-
tice of M ⊕ N generated by the elements of E, and let {(wi , f i ), (w j , f j )} := wi , f j
define a skew-symmetric form on L. Fixing a pair E k = (wk , f k ) ∈ E we mutate E
to a new collection Ek as follows:
1. E k
→ −E k ;
2. E i
→ E i − max({E i , E k }, 0)E k , if i = k.
This formula is identical to the mutation of seed data given in [11]; a connection
we now make precise. Fix a compatible collection of mutations E and define a
skew-symmetric form [−, −] on Z|E| defined by setting [ei , e j ] := {θ (ei ), θ (e j )},
where θ : Z|E| → M ⊕ N is defined by sending ei
→ (wi , f i ). The following Lemma
follows immediately by comparison of the formulae for mutating seed data in a cluster
algebra with Definition 19.
Lemma 20 The operations of mutation given in Definition 19, and of mutation of
the seeds defined above, are intertwined by θ .
Example 21 In dimensions higher than two a compatible collection of mutation data
which defines a set of combinatorial mutations of a given polytope can transform
by mutation to a compatible collection of mutation data which does not define a set
of combinatorial mutations of the transformed polytope. In particular the piecewise
linear maps may fail to preserve convexity. This appears to be a important obstruc-
tion to generalising the two-dimensional theory of mutations to higher dimensional
polytopes. For example, consider the polytope
⎛⎛ ⎞ ⎛ ⎞ ⎛ ⎞ ⎛ ⎞ ⎛ ⎞ ⎛ ⎞⎞
1 0 0 0 −1 −1
P := conv⎝⎝0⎠ , ⎝1⎠ , ⎝0⎠ , ⎝ 0 ⎠ , ⎝−1⎠ , ⎝−1⎠⎠.
0 0 1 −1 0 −1

Consider mutation data (w1 , f 1 ) := (e1 , e3 ) and (w2 , f 2 ) = (e2 , e3 ). Since f 1 = f 2 ,

we have that wi , f j = 0 for all i, j ∈ {1, 2}; hence these mutations are compatible.
However, while P admits both these mutations, the composition of these two (in
either order) is the mutation corresponding to the pair (w, f ) := (e1 + e2 , e3 ), which
does not define a mutation of P.
Proposition 22 Given seed data E such that Q E is a directed simply-laced Dynkin
diagram the number of polytopes obtained by successive edge mutation is bounded
by the numbers of seeds in the cluster algebra determined by Q E . If Q E is of type An
this bound is the Catalan number Cn+1 .
322 T. Prince

In fact, compatible collections of mutations appear whenever we have a cluster

algebra with skew-symmetric exchange matrix.
Proposition 23 Every compatible collection of mutations determines and is deter-
mined by a skew-symmetric cluster algebra without frozen variables, together with a
subspace V of the kernel of the skew-symmetric form {−, −} defined by the exchange
matrix.
Proof Fix a skew-symmetric cluster algebra without frozen variables and a nom-
inated subspace V ⊂ ker{−, −}. Recall that a seed defines a basis ei of a lattice,
which we denote N
. Define M := N
/V and let p : N

→ M be the canonical projec-

→ M ⊕ hom(M, Z) defined by θ : n
→ ( p(n), {n, −}) defines
tion. The map θ : N
a compatible collection of mutation data with weight vectors in the lattice M. 
Note that N and M play dual roles to those in [11], and we insist throughout
that P ⊂ NQ . This exchange of roles explains the odd definition of M in the proof
of Proposition 23. To compare the birational maps associated to the two notions of
mutations let s be a seed of the cluster algebra determined by a compatible collection
of mutation data, and let E be the compatible collection corresponding to s. Fix an
element E k = (wk , f k ) ∈ E and consider the following diagram,

μk
As Aμk (s) (15.2)
p p

TM TM ,
φ(wk , fk )

Proposition 24 The diagram shown in (15.2) commutes.

Proof This is an exercise in writing out the definitions of the respective mutations:
see [27, Sect. 3]. 
Example 25 The del Pezzo surface of degree 5 admits a toric degeneration to a toric
surface Z with a pair of A1 singularities. Given a three-dimensional linear section X
of the Grassmannian Gr(2, 5) X admits a toric degeneration to the projective cone
over Z . The fan determined by this toric threefold is formed by taking the cones over
the faces of the reflexive polytope with PALP id 245.
In Fig. 15.1 we show a pentagon of polytopes obtained by successively mutating
the polytope shown in the top-right with respect to the mutation data

E := {(w1 , f 1 ), (w2 , f 2 )}

where,

w1 := (−1, 0, 0), f 1 := (0, 1, 1)T ,

w2 := (0, 0, −1), f 2 := (−1, 0, 0)T .
15 Polygons of Finite Mutation Type 323

Fig. 15.1 Pentagon of edge

mutations among toric
degenerations of B5

We recall that there is an A2 cluster structure on the co-ordinate ring of the Grass-
mannian, and a toric degeneration of Gr(2, 5) for each cluster chart in the dual
Grassmannian [33]. We expect that cluster structures in the mirror to a Fano variety
to be detected by such compatible collections of mutations.
Note that the polytopes we show in Fig. 15.1 are not dual to Fano polytopes.
However, recalling that B5 has Fano index 2, we can obtain a reflexive polytope by
dilating each of the polytopes shown in Fig. 15.1 by a factor of two, and translating.
In the two dimensional case, we can canonically define a maximal set of compat-
ible mutations, making use of the notion of singularity content [4].
Definition 26 (Cf. [27, Sect. 1.2]) Given a Fano polygon P ⊂ NQ with singularity
content (n, B) and m := |B| + n, we define:
1. an index set I of size m containing a subset Iu f of size n, together with functions

φu f : Iu f → {edges of P} φ f : I \Iu f → B

such that fibres φu−1f (E) contain m E := (E)/r E  elements, where (E) is the
lattice length of the edge E, and r E is the Gorenstein (or local) index of the cone
over E, while the map φ f is a bijection;
2. a lattice map ρ : Zm → M sending each basis element to the primitive, inward-
pointing normal to the edge of P defined by the cone given by the specified
functions φu f and φ;
3. a form {ei , e j } := ρ(ei ) ∧ ρ(e j ). Note that this is an integral skew-symmetric
form.
324 T. Prince

Fig. 15.2 The Fano polygon

for P2

The value m E appears in the definition of the singularity content of a two dimen-
sional cone; and is equal to the maximal number of T -cones of Gorenstein index r E
which fit inside the cone on E.
By [27, Proposition 3.17] the construction of a quiver from mutation data provided
by Definition 26 intertwines polygon and quiver mutations. We let (E P , C P ) denote
the seed associated to a Fano polygon, where E P is the standard basis ei of Zm , and C P
is the standard transcendence basis of the field of rational functions in n variables
over Q(xi : i ∈ I \Iu f ). We let Q P denote the unfrozen quiver associated to (E P , C P ).
We say a Fano polygon is of finite mutation type if it is mutation equivalent to only
finitely many Fano polygons.
Conjecture 27 The cluster algebra C P associated to a Fano polygon P, together
with a bijection between the set of frozen variables and B, is a complete mutation
invariant of the Fano polygon P.
Example 28 Consider the Fano polygon P for P2 (this is depicted in Fig. 15.2).
Computing the determinant of the inward-pointing normals we obtain the quiver Q P

•
3 3

• •
3

The mutations of this quiver are well-known, and the triple (3a, 3b, 3c) of non-zero
entries of the exchange matrix satisfy the Markov equation a 2 + b2 + c2 = 3abc.
Indeed, as the polygon P is mutated the corresponding toric surfaces are P(a 2 , b2 , c2 )
for the same triples (a, b, c). We see that in this case the mutations of the quivers
exactly capture the mutations of the polygon.
Example 29 Consider the toric surface (using the notation for these surfaces appear-
ing in [8]), X 5,5/3 associated with the Fano polygon shown in Fig. 15.3. The unfrozen
quiver associated to this surface is simply the A2 quiver:

• •

This example is important, both in this section, because it is an example of a finite

type polygon, and since a smoothing of this surface is given by 5 Pfaffian equations,
see [8, Sect. 3.3], a fact closely connected to the A2 quiver we construct here.
15 Polygons of Finite Mutation Type 325

Fig. 15.3 The Fano polygon

for X 5,5/3

15.4 Finite Type Classification

We now make use of the classification of finite type and finite mutation type cluster
algebras to establish the following result.
Theorem 30 P is of finite mutation type if and only if Q P is mutation equivalent to
a quiver of type (A1 )n , A2 , A3 , or D4 .
Remark 31 The types referred to in Theorem 30 may also be referred to as
type In , I I , I I I , and I V respectively; in analogy with Kodaira’s monodromy matri-
ces. The relationship between these matrices, log Calabi–Yau manifolds, and mon-
odromy in certain integral affine manifolds is explored by Mandel in [30].
Remark 32 We remark that all the cases which appear in Theorem 30 do occur as
(unfrozen) quivers associated to polygons. Several examples can be found in [8, p.
42] and are tabulated below.
Quiver Q P Polygon Surface
∅ = A01 9 X 6,2
A31 11 X 4,7/3
A61 12 B2,8/3
A2 7 X 5,5/3
A3 17 X 3,4
D4 5 X 4,4/3

Examples of polygons P with Q P = A2k+2

1 and A2k+1
1 for k ≥ 5 are given by the
quadrilaterals
       
−1 1 −1 1
conv , , , , and
−3 −3 k−2 k−2
       
−1 1 −1 1
conv , , , .
−3 −2 k−2 k−2

We first make two straightforward observations. First we note that the cluster
algebra C P induces a sequence of surjections:
326 T. Prince

{Clusters of C P } (15.3)

{Polygons mutation equivalent to P}

{Quivers mutation equivalent to Q P }.

The first vertical arrow follows from the fact that algebraic mutations determine
combinatorial mutations, the second from Lemma 20. For example, using this tower
of surjections in the case of a type A2 cluster algebra, we can immediately state the
following result.

Proposition 33 If a Fano polygon P has singularity content (2, B) and the primitive
inward-pointing normal vectors of the two edges corresponding to the unfrozen
variables of C P form a basis of the lattice M, then the mutation-equivalence class
of P has at most five members.

Proof The quiver associated to P is precisely an orientation of the A2 quiver. The

cluster algebra C P is well-known and its cluster exchange graph forms a pentagon.
Note however that the quiver mutation graph is trivial, as the A2 quiver mutates only
to itself. 

Proposition 24 implies that the mutation class of P has at most five elements. Note
that we do not have a non-trivial lower bound: there is only one polygon in mutation
equivalent to the polygon described in Example 29 up to GL(2, Z) equivalence. Next
observe that the sequence of surjections shown in (15.3) immediately implies that

C P finite type ⇒ P finite mutation type ⇒ C P finite mutation type.

Lemma 34 Given a Fano polygon P of finite mutation type, Q P does not contain a
Kronecker subquiver
k
Q k := { v1 v2 },

where k > 1 is the number of arrows from v1 to v2 .

Remark 35 This result is expected from results on the corresponding cluster alge-
bra. The Kronecker quiver defines a rank 2 cluster algebra which is known not to be
of finite type when k > 1. Given that P is the Newton polygon of a superpotential
which is itself a combination of cluster monomials, we expect the polygon P to grow
as we mutate.

Proof (Proof of Lemma 34) Assume there is a Q k subquiver of Q P , with ver-

tices v1 , v2 corresponding to edges E 1 E 2 of P. We define ρ : Z2 → M by mapping
the standard basis to the primitive inward normal vectors wi to E i for i ∈ {1, 2}.
15 Polygons of Finite Mutation Type 327

Fig. 15.4 Schematic

diagram of a polygon in
standard form

Let P  ⊂ Q2 be the image of P under ρ  . The resulting polygon in Q2 is shown

schematically in Fig. 15.4.
The local index of each cone in P is the integral height of the edge from the origin.
Let h i denote the local indices of E i for i ∈ {1, 2}. Note that, as h i = ei , ρ  u for
any u ∈ E i , h i is also the local index of ρ  (E i ) in P  . Mutating at v1 and v2 we denote
the new local indices,

(h 1 , h 2 ) (h 1 , h 2 ) (h 1 , h 2 ).

We first show that ρ  increases the lattice lengths of E i by a factor of k :=

|w1 ∧ w2 | for each i ∈ {1, 2}. Let u i1 and u i2 denote the vertices of E i , and f i :=
(u i1 − u i2 )/(E i ); where (E i ) denotes the lattice length of E i . Note that w1 , f 2 =
w2 , f 1 = w1 ∧ w2 for a suitable choice of orientation of M. Moreover, since – for
each i ∈ {1, 2} – ρ  (u), ei = h i is constant as u varies in E i , the edge ρ  (E i ) has

direction vector e3−i ; where {e1 , e2 } is the dual basis to {e1 , e2 }. In other words,

ρ  (u i1 ) − ρ  (u i2 ) = (ρ  (E i ))e3−i

,

and hence we have that

(ρ  (E i )) = (ρ  (E i ))e3−i


, e3−i
= ρ  (u i1 ) − ρ  (u i2 ), e3−i
= u i1 − u i2 , ρ(e3−i )
= (E i ) f i , w3−i
= ±(E i )(w1 ∧ w2 ),

where signs and orientations are chosen such that (E) is always positive. Study-
ing Fig. 15.4 note that h 1 + h 1 ≥ (ρ  (E 2 )), however—by the calculation above—
(ρ  (E 2 )) = k(E 2 ). Moreover, we have that (E 2 ) ≥ h 2 , since the Fano polygon P
admits a mutation along this edge. Hence we observe that

h 1 ≥ kh 2 − h 1 h 2 ≥ kh 1 − h 2 .
328 T. Prince

Consider the case k ≥ 3, and assume without loss of generality that h 2 ≥ h 1 . We have
that h 1 ≥ 3h 2 − h 1 ≥ 2h 2 ≥ 2h 1 . Thus in this case the values in the pair (h 1 , h 2 )
grow (at least) exponentially with mutation, and in particular take infinitely many
values.
Next consider the case k = 2. The inequalities above become,

h 1 ≥ 2h 2 − h 1 h 2 ≥ 2h 1 − h 2 ,

and we are again free to assume that h 2 ≥ h 1 . Thus h 1 ≥ 2h 2 − h 1 ≥ h 1 , and

if h 2 > h 1 , h 1 ≥ 2h 2 − h 1 > h 2 . Thus, assuming h 1 = h 2 , one can generate an
infinite increasing sequence of local indices. The only remaining case is if h :=
h 1 = h 2 = h 1 = h 2 . To eliminate this possibility observe that, since k = 2, the
edges ρ  (E 1 ), ρ  (E 2 ) must meet in a vertex with coordinates (−h, −h) (indeed,
assuming this does not hold, a mutation returns us to the previous case and one of
the above inequalities is strict). Note that the sublattice ρ  (N ) is determined by the
fact that ρ  doubles the edge lengths of E 1 and E 2 . The lattice vectors (a, a) are in
this sublattice for all a ∈ Z. Thus, by primitivity of the vertices in P, h = 1. Since
the origin is in the interior of P, mutating in one of v1 or v2 returns us to the previous
case. 

Remark 36 Proposition 34 implies all the quivers that we consider from now on
are directed graphs. Hence we refer to vertices as adjacent if they are adjacent in the
underlying graph.

As well as the non-existence of Kronecker quivers in Q P for finite mutation type

polygons P, we use heavy use of a connectedness result for quivers Q P which follows
immediately from the definition of Q P via determinants in the plane; or equivalently
from the fact the exchange matrix has rank 2 (Fig. 15.5).

Lemma 37 Given a Fano polygon P and vertices v1 , v2 , v3 of Q P such that vi

and vi+1 are not adjacent for i = 1, 2, then v1 and v3 are not adjacent.

Proof (Proof of Theorem 30) By Lemma 37, if Q P is not connected, Q P ∼ = An1 for
some n. Similarly, if Q P is of type A or D, then it must be one of A2 , A3 or D4 .
Thus we only need to show that there is no Fano polygon P of finite mutation type
such that C P is not of finite-type. However C P is of finite mutation type, and we
use the classification described in Theorems 9 and 10, following [9, 12]. In fact,
using Lemma 37, none of the eleven exceptional types can occur as Q P for a Fano
polygon P. Hence we can restrict to quivers which admit a block decomposition and
work case-by-case.
We claim that every quiver Q P associated to a Fano polygon P which admits a
block decomposition is either mutation equivalent to an orientation of a simply-laced
Dynkin diagram or to a quiver which contains a subquiver Q k for k > 1. We assume
for contradiction that Q P is the quiver associated to a Fano polygon P of finite-type
which is not mutation equivalent to a simply laced Dynkin diagram.
15 Polygons of Finite Mutation Type 329

Fig. 15.5 The blocks of a block decomposition

Fig. 15.6 Mutations of block V

Fig. 15.7 Quiver V

Block V: First observe that, since only one vertex of the block V is an outlet, the V
block quiver is a subquiver of any quiver which contains the V block in its decom-
position. However this quiver mutates to a quiver with a Q 2 subquiver as shown in
Fig. 15.6. Therefore block V never appears in a decomposition of a quiver Q P . For
later use we shall fix the following intermediate quiver, V , shown in Fig. 15.7.
Blocks IIIa and IIIb: Assume there is a type III block (a or b) connected to a quiver Q 
at a vertex v. If there is a vertex v of Q  such that v and v are not adjacent, the quiver
violates Lemma 37. In particular the vertex set of Q  must be the vertex set of a single
block. In particular, using the previous part, Q  has at most four vertices. Case by
case study shows that only the A3 and D4 types appear.
Block IV: Consider the case of a decomposition only using type IV blocks. Note that
the type IV block is itself of type D4. Consider attaching two type IV blocks. If the
blocks are attached at a single outlet the resulting quiver contradicts Lemma 37. In
fact it is easy to see that it is impossible to add additional type IV blocks to meet
this condition. If both pairs of outlets are matched there are two possible quivers
depending on the relative orientations of the arrow between the outlets, one orienta-
330 T. Prince

(a) Attaching I blocks to a IV block (b) Attaching II blocks to a IV block

Fig. 15.8 A type IV block

tion produces a Q 2 subquiver automatically, the other produces a quiver containing

the quiver V as a subquiver. Thus, for a type IV block to appear in a decomposition
of Q P it must include a type I or II block.
Now consider decompositions using type I and II blocks as well as type IV blocks.
First note there must be exactly one IV block (assuming there is at least one). Indeed,
if type IV blocks are not connected using both vertices, a non-outlet vertex of a IV
block is not adjacent to some outlet, and some non-outlet vertex of a (different)
IV block. However outlets and non-outlets of a type IV block are always adjacent,
violating Lemma 37.
Thus we must attach I and II blocks to a single type IV block. By Lemma 37 the
vertex set of the final quiver must be equal to the vertex set obtained by attaching
a single block to each outlet of the IV block. Considering these cases in turn, we
note first that attaching a type I block to cancel the arrow between the two outlets
produces a quiver mutation equivalent to D4 and therefore eliminated. For chains
type I blocks of length two, if a 3-cycle is produced, a mutation in the vertex between
the type I blocks produces the V quiver. If not, the same mutation produces a Q 2
subquiver.
Attaching a type II block along two outlets of the type IV block recovers the V
or Q 2 subquiver cases we have already seen. Attaching type II blocks to a single
outlet each we observe that every new vertex must be adjacent to both outlets of
the IV block. Hence the only case without a Q 2 subquiver is shown on the right of
Fig. 15.8, however this quiver mutates to one with a Q 2 subquiver. Attaching further
type II blocks any quiver we obtain must contradict Lemma 37.
Blocks I and II: From what we have shown above, the block decomposition of Q P
consists only of type I and type II blocks. Any connected quiver with a block decom-
position into type I blocks is a path (with possibly changing orientations), which
possibly closes up into a cycle. The only cases not violating Lemma 37 are mutation
equivalent to orientations of simply laced Dynkin diagrams.
15 Polygons of Finite Mutation Type 331

For decompositions of Q P with type I and II blocks we divide the proof into cases
indexed by the number of type II blocks. For a single type II block, we can attach a
type I block to two outlets and in this way reduce to the type III case. Attaching each
type I block to a type II block in at most one outlet, we use the fact that every new
vertex must be adjacent to at least two of the vertices of the type II block. Thus we
can obtain only two undirected graphs—the underlying graph of a type IV block or
an orientation of a tetrahedron, these cases can easily be eliminated. For example,
there is no orientation of the tetrahedron making every cycle oriented; hence after a
single mutation we obtain a quiver violating Lemma 34.
Consider the case of a pair of type II blocks. If these have disjoint vertex sets, each
outlet of a type II block cannot be adjacent to two of the outlets of the other type II
block. Thus we must cancel the arrow between these two outlets with a type I block.
However this creates a pair of 1-valent non-outlet vertices which can be eliminated
similarly to the type III case. At the other extreme, if we attach along all three outlets,
we produce two easy cases. Attaching along a pair of outlets we generate either a Q 2
subquiver or a 4-cycle. Considering the 4-cycle with two outlets v1 and v2 (on non-
adjacent corners) to meet the conditions of Lemma 37 any vertex adjacent to one
of v1 or v2 must be adjacent to the other. Moreover, if the resulting quiver contains
an arrow between v1 and v2 , a mutation at one of the non-outlet vertices gives a Q 2
subquiver. Given a vertex v adjacent to v1 and v2 , if this defines a path between them,
mutating at this node and a non-outlet in the four cycle produces a Q 2 subquiver. If v
does not lie on a path between v1 and v2 then mutating at both outlets produces a Q 2
subquiver.
Attaching the type II blocks at a single outlet, the four arrows incident to this
vertex are now fixed, so any new vertex must be adjacent to each of the remaining
four outlets by Lemma 37. However this cannot be achieved with type I blocks.
Attaching more than two type II blocks together, we can eliminate the case where
two are connected to form a 4-cycle as above. Since we can easily eliminate the case
that two type II blocks meet in three outlets, we assume that each type II block meets
every other in at most one outlet. Some pair of type II blocks must be attached in
an outlet (otherwise we can argue as in the case of type II block separated by type I
blocks). Thus, since every new vertex must be adjacent to all four outlets formed by
attaching two type II blocks, all possible quivers can be represented as an octahedron
with some orientation, see Fig. 15.9.
Considering an orientation of the octahedron; if any triangular face does not form
a cycle we can mutate to form a Q 2 subquiver. Assuming every triangle is a cycle,
and possibly mutating, the vertices adjacent to the ‘top’ of the octahedron form a
type V block subquiver. Following the same reasoning as for the type V block case
(although note that the type V block is not part of a block decomposition here) these
cases can be eliminated. 
332 T. Prince

Fig. 15.9 Octahedron of

type II blocks

Acknowledgements We thank Alexander Kasprzyk for his insights on polytope mutation, and
our many conversations about quivers. The author is supported by a Fellowship by Examination at
Magdalen College, Oxford. This work was undertaken while the author was a graduate student at
Imperial College London.

References

1. Aharony, O., Hanany, A.: Branes, superpotentials and superconformal fixed points. Nucl. Phys.
B 504(1–2), 239–271 (1997)
2. Akhtar, M., Coates, T., Corti, A., Heuberger, L., Kasprzyk, A.M., Oneto, A., Petracci, A.,
Prince, T., Tveiten, K.: Mirror symmetry and the classification of orbifold del Pezzo surfaces.
Proc. Am. Math. Soc. 144(2), 513–527 (2016)
3. Akhtar, M., Coates, T., Galkin, S., Kasprzyk, A.M.: Minkowski polynomials and mutations.
SIGMA Symmetry Integrability Geom. Methods Appl. 8, 094, 17 (2012)
4. Akhtar, M., Kasprzyk, A.M.: Singularity Content (2014). arXiv:1401.5458
5. Bergman, A., Proudfoot, N.: Moduli spaces for D-branes at the tip of a cone. J. High Energy
Phys. (3), 073, 9 (2006)
6. Bernšteı̆n, I.N., Gel fand, I.M., Ponomarev, V.A.: Coxeter functors, and Gabriel’s theorem.
Uspehi Mat. Nauk 28(2(170)), 19–33 (1973)
7. Bridgeland, T., Stern, D.: Helices on del Pezzo surfaces and tilting Calabi-Yau algebras. Adv.
Math. 224(4), 1672–1716 (2010)
8. Corti, A., Heuberger, L.: Del Pezzo surfaces with 13 (1, 1) points. Manuscripta Math. 153(1–2),
71–118 (2017)
9. Felikson, A., Shapiro, M., Tumarkin, P.: Skew-symmetric cluster algebras of finite mutation
type. J. Eur. Math. Soc. (JEMS) 14(4), 1135–1180 (2012)
10. Feng, B., Hanany, A., He, Y.H.: Phase structure of D-brane gauge theories and toric duality. J.
High Energy Phys. (8), 40, 25 (2001)
11. Fock, V.V., Goncharov, A.B.: Cluster ensembles, quantization and the dilogarithm. II. The
intertwiner. In: Algebra, arithmetic, and geometry: in honor of Yu. I. Manin, vol. I, Progress in
Mathematics, vol. 269, pp. 655–673. Birkhäuser Boston, Inc., Boston, MA (2009)
12. Fomin, S., Shapiro, M., Thurston, D.: Cluster algebras and triangulated surfaces. I. Cluster
complexes. Acta Math. 201(1), 83–146 (2008)
13. Fomin, S., Zelevinsky, A.: Cluster algebras. I. Foundations. J. Am. Math. Soc. 15(2), 497–529
(2002)
15 Polygons of Finite Mutation Type 333

14. Fomin, S., Zelevinsky, A.: Cluster algebras. II. Finite type classification. Invent. Math. 154(1),
63–121 (2003)
15. Franco, S., Hanany, A., Martelli, D., Sparks, J., Vegh, D., Wecht, B.: Gauge theories from toric
geometry and brane tilings. J. High Energy Phys. (1), 128, 40 (2006)
16. Galkin, S., Usnich, A.: Mutations of Potentials (2010). Preprint IPMU 10-0100
17. Givental, A.: A mirror theorem for toric complete intersections. In: Topological Field Theory,
Primitive Forms and Related Topics (Kyoto, 1996), Progress in Mathematics, vol. 160, pp.
141–175. Birkhäuser Boston, Boston, MA (1998)
18. Givental, A.B.: Homological geometry and mirror symmetry. In: Proceedings of the Interna-
tional Congress of Mathematicians, vol. 1, 2 (Zürich, 1994), pp. 472–480. Birkhäuser, Basel
(1995)
19. Givental, A.B.: Equivariant Gromov-Witten invariants. Internat. Math. Res. Not. 13, 613–663
(1996)
20. Gross, M., Hacking, P., Keel, S.: Birational geometry of cluster algebras. Algebr. Geom. 2(2),
137–175 (2015)
21. Hanany, A., Kazakopoulos, P., Wecht, B.: A new infinite class of quiver gauge theories. J. High
Energy Phys. (8), 054, 30 (2005)
22. Hanany, A., Vegh, D.: Quivers, tilings, branes and rhombi. J. High Energy Phys. (10), 029, 35
(2007)
23. Herzog, C.P.: Seiberg duality is an exceptional mutation. J. High Energy Phys. (8), 064, 31
(2004)
24. Hille, L., Perling, M.: Exceptional sequences of invertible sheaves on rational surfaces. Compos.
Math. 147(4), 1230–1280 (2011)
25. Hori, K., Katz, S., Klemm, A., Pandharipande, R., Thomas, R., Vafa, C., Vakil, R., Zaslow,
E.: Mirror symmetry, Clay Mathematics Monographs, vol. 1. American Mathematical Society,
Providence, RI; Clay Mathematics Institute, Cambridge, MA (2003). With a preface by Vafa
26. Ilten, N.O.: Mutations of Laurent polynomials and flat families with toric fibers. SIGMA
Symmetry Integrability Geom. Methods Appl. 8, 047, 7 (2012)
27. Kasprzyk, A.M., Nill, B., Prince, T.: Minimality and mutation-equivalence of polygons. Forum
Math. Sigma 5, e18, 48 (2017)
28. Kontsevich, M.: Lectures at ENS Paris (1998). Set of notes taken by J. Bellaiche, J.-F. Dat, I.
Martin, G. Rachinet and H. Randriambololona
29. Leung, N.C., Vafa, C.: Branes and toric geometry. Adv. Theor. Math. Phys. 2(1), 91–118 (1998)
30. Mandel, T.: Classification of rank 2 cluster varieties. SIGMA Symmetry Integrability Geom.
Methods Appl. 15, 042, 32 (2019)
31. Mukhopadhyay, S., Ray, K.: Seiberg duality as derived equivalence for some quiver gauge
theories. J. High Energy Phys. (2), 070, 22 (2004)
32. Perling, M.: Examples for exceptional sequences of invertible sheaves on rational surfaces. In:
Geometric methods in representation theory. II, Sémin. Congr., vol. 24, pp. 371–392. Society
Mathematics France, Paris (2012)
33. Rietsch, K., Williams, L.: Newton-Okounkov bodies, cluster duality, and mirror symmetry for
Grassmannians. Duke Math. J. 168(18), 3437–3527 (2019)
Chapter 16
Orbit Spaces of Maximal Torus Actions
on Oriented Grassmannians of Planes

Hendrik Süß

Abstract Motivated by Buchstaber’s and Terzić’s work on the complex Grassman-

nians G C (2, 4) and G C (2, 5) we describe the moment map and the orbit space of
the oriented Grassmannians G + R (2, n) under the action of a maximal compact torus.
Our main tool is the realisation of these oriented Grassmannians as smooth complex
quadric hypersurfaces and the relatively simple Geometric Invariant Theory of the
corresponding algebraic torus action.

Keywords Oriented Grassmannian · Complexity-one T-variety · GIT quotients

16.1 Introduction

We denote an algebraic torus (C∗ )k by TC and the corresponding compact torus

(S 1 )k ⊂ TC by T . A complex algebraic variety with a TC -action is called a TC -
variety. The complexity of a TC -variety is the minimal (complex) codimension of an
orbit. In this paper we study the T -orbit spaces of projective TC -varieties and apply
our findings to the case of oriented Grassmannians of planes and that of smooth TC -
varieties of complexity 1. Our main goal is to determine the corresponding T -orbit
spaces up to homeomorphism.
We consider the Grassmannian G + R (2, n) parametrising oriented planes in R with
2

the natural action of a maximal torus in SOn . Our main result determines the orbit
space of this action.

Theorem 1 The orbit space G +

R (2, n)/T is homeomorphic to the join

n/2−2
S n/2−1 ∗ PC .

H. Süß (B)
Friedrich-Schiller-Universität Jena, Institut für Mathematik, Ernst-Abbe-Platz 2, 07743 Jena,
Germany
e-mail: [email protected]

© Springer Nature Switzerland AG 2022 335

A. M. Kasprzyk and B. Nill (eds.), Interactions with Lattice Polytopes,
Springer Proceedings in Mathematics & Statistics 386,
https://doi.org/10.1007/978-3-030-98327-7_16
336 H. Süß

For smooth varieties with a torus action of complexity 1 we derive the following
general results on the structure of their orbit spaces.

Theorem 2 Consider a smooth projective TC -variety X of complexity 1. Then the

corresponding orbit space X/T is a topological manifold with boundary.

Theorem 3 Consider a smooth projective TC -variety X of complexity 1 with only

finitely many lower dimensional TC -orbits. Then the orbit space X/T is homeomor-
phic to a sphere.

Note that results comparable to Theorem 2 have been proved by Ayzenberg [3]
and Cherepanov [8], Theorem 3 has been proved independently, but using similar
methods, by Karshon and Tolman [15]. Their work covers the more general setting
of symplectic manifolds with Hamiltonian torus actions. In their paper they also
prove Theorem 1 for the cases of complexity 1, i.e. for n = 5, 6.
Our main tool is Geometric Invariant Theory (GIT) and its symplectic coun-
terpart in combination with the Kempf–Ness Theorem. This approach suggest to
stratify the manifold and eventually the orbit space via a polyhedral subdivision of
the momentum polytope, which encodes the variation of GIT quotients. In general
these stratifications can become arbitrarily complicated. However, in the cases con-
sidered in this paper they turn out to be almost trivial allowing us to derive concrete
results about the orbits spaces.
In Sect. 16.2 we fix our setting for compact torus actions induced by algebraic
torus actions on complex varieties and recall crucial results from Geometric Invariant
Theory. Moreover, we derive first results on the structure of orbits spaces in suitable
situations. We then apply these to the special cases of oriented Grassmannians of
planes in Sect. 16.3 and TC -varieties of complexity 1 in Sect. 16.4.
In order to distinguish between the algebraic and the topological category, we are
going to denote isomorphism of algebraic varieties by ∼ = and homeomorphisms of
topological spaces by ≈.

16.2 TC -Varieties and Their T -Orbit Spaces

Fix a linearised action of an algebraic torus TC = (C∗ )k on PCN with weights

u 0 , . . . , u N ∈ Zk , i.e. for t = (t1 , . . . , tk ) ∈ TC we have

t.(z 0 : . . . : z N ) = (t u 0 z 1 : . . . : t u N z N ),
(u j )1 (u j )k
where t u j := t1 · · · tk . Then a moment map of this action is given by

j |z j |2 u j
ν: PCN → R ; (z 0 : . . . z N ) → 
k
.
j |z j |2
16 Orbit Spaces of Maximal Torus Actions on Oriented Grassmannians … 337

For an embedded projective variety X ⊂ PCN , which is invariant under under this torus
action, a moment map of the induced torus action on X is given by the restriction μ =
ν| X . The moment image P = μ(X ) is known to be a convex polytope [2, 11]. We
start with some notions known as variation of GIT quotients with [9, 12, 16] being the
most relevant references. For a point x ∈ X the moment image (x) = μ(TC .x) ⊂ P
of its orbit closure is again a polytope and the orbit TC .x is mapped to the relative
interior ◦ (x) ⊂ (x).
For a point u ∈ P we define

X ss (u) = {x ∈ X | u ∈ (x)}, X ps (u) = {x ∈ X | u ∈ ◦ (x)}.

Hence, X ss (u) consists of those points in X whose orbit closures intersect μ−1 (u)
and X ps (u) consists of those points whose orbits intersect μ−1 (u). Equivalenty X ps (u)
is the union of closed TC -orbits in X ss (u).
Now for every u ∈ P we may consider
 
λ(u) = (x); λ◦ (u) = ◦ (x)
x,u∈(x) x,u∈◦ (x)

Since only finitely many polytopes occur as moment images of orbit closures their
intersections are again polytopes. We denote the set of all these polytopes λ(u) by .
This set is partially ordered by the face relation ≺. The polytopes λ ∈  form a
polyhedral subdivision of P and one obtains a stratification of P via their relative
interiors. 
P= λ◦ .
λ∈

For u ∈ P let us denote by λ(u) the unique element of λ ∈  such that u ∈ λ◦ .

From the definitions above it follows that X ss (u) = X ss (v) if and only if λ(u) =
λ(v), i.e. u and v are contained in the relative interior of the same element of . In this
ps
case also X ps (u) = X ps (v) holds. Hence, we may define X λss = X ss (u) and X λ =
◦
X (u) for u ∈ λ .
ps

Example 4 We consider the linear T = C∗ -action on P2C given by t.(x : y : z) =

(t x : t −1 y : z). Then the moment map is given by μ(x : y : z) = |x||x| −|y|2
2
2 +|y|2 +|z|2 . We

get P = μ(P2 ) = [−1, 1] ⊂ R. The orbits can be described as follows. We have the
fixed points (1 : 0 : 0), (0 : 1 : 0) and (0 : 0 : 1) with moment images 1, −1 and 0,
respectively. The moment images of the other TC -orbits are

μ(T.(1 : 0 : 1)) = (0, 1) ⊂ R,

μ(T.(0 : 1 : 1)) = (−1, 0) ⊂ R,
μ(T.(1 : 1 : 0)) = (−1, 1) ⊂ R,
μ(T.(α : 1 : 1)) = (−1, 1) ⊂ R, α ∈ C∗ .
338 H. Süß

Hence, in this case  is obtained by subdividing the interval [−1, 1] at the point 0,
or more formally  = {[−1, 0], [0, 1], {−1}, {0}, {1}}.
By the Kempf–Ness Theorem for rational values of u the definition of X ss (u)
coincides with the semi-stable locus of Mumford’s Geometric Invariant Theory.
Hence, there exists a categorical quotient morphism qλ : X λss → Yλ = X λss //TC where
ps
Yλ is an orbit space for the TC -action on X λ and the corresponding quotient map is
ps
given by the restriction of qλ to X λ . The occurring quotients Yu have the expected
dimension for u ∈ P ◦ , but can be lower-dimensional for elements u ∈ ∂ P. By [17,
Lemma 7.2] for u ∈ λ◦ every TC -orbit in X λ intersects μ−1 (u) in exactly one T -orbit.
ps

Hence, the restriction of qλ to μ−1 (u) induces a homeomorphisms between Yλ =

X λss //TC and the topological orbit space μ−1 (u)/T . Moreover the inclusion X λ ⊂ X γ
for γ ≺ λ induces contraction morphisms on the level of quotients pγ λ : Yλ → Yγ
forming an inverse system.
Example 5 When dim X = dim TC , i.e. if the variety is toric, then the moment
image P completely determines the variety. The moment images (x) of TC -orbit
closures are just the faces of the polytope P and the stratification of P is the decom-
position of P into the relative interiors of its faces. The preimage μ−1 (u) consists
of exactly on T -orbit with dimension equal to dimension of the face containing u
in its interior. Consequently X/T ≈ P with μ coinciding with the quotient map.
Alternatively, we may apply Proposition 12 below and obtain X/T ≈ S k−1 ∗ {pt} ≈
D k ≈ P.
Example 6 Consider a projective toric variety X corresponding to a polytope Q ⊂
Rd . Then the inclusion of a k-dimensional subtorus TC  ⊂ TC induces a surjec-
tion F : Rd → Rk . Given a moment map μ for the TC -action a corresponding moment
map μ : X → Rk is given by μ = F ◦ μ. Hence, the moment image for the TC  -
action is P := F(Q) and the stratification of P is induced by the images of the
faces of Q. More precisely, the stratification consists of the relative interiors of the
polytopes 
λ(u) = F(τ ).
τ ≺Q, u∈F(τ )

Moreover, the GIT quotients X ss (u)//TC  are again toric varieties corresponding
to the polytope F −1 (u) ∩ Q, see [13, Proposition 3.5].
Already in [10] it has been observed that orbit space of the T -action on X can be
constructed out of the inverse system of GIT quotients.
Theorem 7 ([10, Sect. 5]) We have
 

X/T ≈ λ × Yλ /∼,
λ∈

where (u, y) ∼ (u, y  ) if (u, y) ∈ γ × Yγ , (u, y  ) ∈ λ × Yλ with γ ≺ λ and pγ λ (y  )

= y.
16 Orbit Spaces of Maximal Torus Actions on Oriented Grassmannians … 339

We easily derive the following result, which turns out to be a little bit handier in
some situations.
Corollary 8 Assume that we have a compact topological space Y and with proper
surjective maps rλ : Y → Yλ being compatible with the inverse system above. Then
we have the following homeomorphism.

X/T ≈ (P × Y )/∼r .

Here, the equivalence relation is generated by

(u, y) ∼r (u, y  ) ⇔ rλ(u) (y) = rλ(u) (y  )

for u ∈ λ◦ .
Proof There is a canonical map
 

P ×Y → λ × Yλ /∼; (u, y) → [(u, rλ(u) (y)]
λ∈

This map is surjective and continuous and identifies exactly those pairs which are
equivalent under ∼r . The quotient (P × Y )/∼r is compact as (P × Y ) is and by
Theorem 7 the codomain of the map is homeomorphic to X/T , which is a Hausdorff
space. Hence, the induced continuous bijection (P × Y )/∼r → X/T is a homeo-
morphism. 
Remark 9 In [6, 7] such a Y is called a universal parameter space for the TC -obits.
In algebraic geometry a natural choice for such a dominating algebraic object Y
would be the inverse limit of the Yλ or the Chow quotient of X by TC , which can be
identified with a distinguished irreducible component of this inverse limit.
If the structure of the inverse system {Yλ }λ∈ of GIT quotients is complicated
Corollary 8 might not give much concrete information about the orbit space X/T .
However, in certain situations this structure turns out to be almost trivial allowing us
to effectively calculate the orbit space.
Definition 10 We say the TC -action on X ⊂ PCN has an almost trivial variation of
GIT if for λ ⊂ ∂ P the quotients Yλ are all isomorphic to some Y .
Example 11 If the torus action has complexity one than the quotients Yλ are smooth
algebraic curves or just a point, where the latter happens at most over the boundary
of P. The only contraction morphisms here are isomorphisms or the contraction of
a curve to a point. Hence, the definition is automatically fulfilled.
Proposition 12 Consider a TC -action on X with almost trivial variation of GIT and
only finitely many lower-dimensional TC -orbits. Then X/T is homeomorphic to the
topological join S k−1 ∗ Y .
340 H. Süß

Proof Having an almost trivial variation of GIT means that Yλ ∼ = Y for λ ⊂ ∂ P.

On the other hand, having only finitely many lower-dimensional TC -orbits implies
that the moment fibre of a boundary point u ∈ ∂ P consists of exactly one (lower-
dimensional) orbit and therefore Yλ(u) is just a point. Hence, by Corollary 8 we have

X/T ≈ (P × Y )/∼∂ ,

where the equivalence relation ∼∂ is generated by (u, y) ∼∂ (u, y  ) for u ∈ ∂ P. Now,

the claim follows from Lemma 13 below. 

Lemma 13 Consider the closed unit disc D k and the unit sphere S k−1 . Then for any
compact topological manifold Y we have

S k−1 ∗ Y ≈ (D k × Y )/∼∂ ,

where the equivalence relation ∼∂ is generated by (u, y) ∼∂ (u, y  ) for u ∈ ∂ D k .

Proof Recall that the join S k−1 ∗ Y is defined as (S k−1 × Y × [0, 1])/∼, with
the equivalence relation being generated by (s, y, 0) ∼ (s  , y, 0) and (s, y, 1) ∼
(s, y  , 1). Now the homeomorphism is given by

(S k−1 × Y × [0, 1])/∼ −→ (D k × Y )/∼∂ , [(u, y, t)] → [(tu, y)].

As a special case of Lemma 13 we may consider the situation when Y ≈ S m . Then

Lemma 13 implies (D k × S m )/∼∂ ≈ S k−1 ∗ S m , which is known to be homeomor-
phic to S m+k . The lemma below gives a slightly more general statement.

Lemma 14 For any closed H ⊂ Rk we have

((D k ∩ H ) × S m )/∼∂ ≈ S k+m ∩ (H × Rm )

where the equivalence relation ∼∂ is generated by (u, y) ∼∂ (u, y  ) for u ∈ ∂ D k .

Proof We can state the homeomorphism explicitly


((D k ∩ H ) × S m )/∼∂ → S k+m ∩ H × Rm , [(u, y)] → (u, 1 − |u|2 · y).

For every (u, y) ∈ (∂ D k ∩ H ) × S m we have (u, 1 − |u|2 · y) = (u, 0). Hence,
the map is a well-defined continuous bijection from a compact space to a Hausdorff
space and, therefore a homeomorphism. 
16 Orbit Spaces of Maximal Torus Actions on Oriented Grassmannians … 341

16.3 Oriented Grassmanians of Planes as TC -Varieties

We consider the smooth manifold G + R (2, n) parametrising oriented planes in R .

n

An oriented plane is given by an orthonormal basis (v1 , v2 ). Another orthonormal

pair (v1 , v2 ) gives rise to the same oriented plane if and only if (v1 , v2 ) = (v1 , v2 )Q φ
with 
cos φ − sin φ
Qφ = ∈ SO2 .
sin φ cos φ

Hence,
G+
R (2, n) = {(v1 , v2 ) ∈ R
n×2
| vi , v j  = δi j }/ SO2 .

A (compact) torus action on G + R (2, n) is induced by the choice of a maximal torus

in SOn via its action on the pair (v1 , v2 ). A maximal torus T is given by block diagonal
matrices of the form diag(Q φ1 , . . . , Q φk ) in the case n = 2k or diag(Q φ1 , . . . , Q φk , 1)
in the case n = 2k + 1. In the even-dimensional case the induced action on the
oriented planes is not effective as −I acts trivially. To obtain an effective action one
has to pass to the quotient T /±I . However, this does not effect the orbit structure
of the action.
It is well-known that the oriented Grassmannian of planes can be identified with
the underlying smooth manifold of the complex smooth quadric Q n−2 in Pn−1 C , see
e.g. [18, p. 280]. Indeed, the map

 : Rn×2 → Cn ; (v1 , v2 ) → w = v1 + i · v2

induces an embedding  ¯ : G+
R (2, n) → PC . This is well-defined as ((v
n
1 , v2 )Q φ ) =
e · (v1 , v2 ). Moreover the condition v1 , v2  = 0 is equivalent to ( j w2j ) = 0
iφ

and |v1 |2 /|v2 |2 = 1 is equivalent to ( j w2j ) = 0. Hence, the image of the embed-

ding in PnC is cut out by the equation j w2j = 0. A change of coordinates

z 2 j−1 = w2 j−1 + i · w2 j ; z 2 j = w2 j−1 − i · w2 j for j = 1, . . . , k

in the case n = 2k and additionally z n = wn in the case n = 2k + 1 leads to the

equivalent equation

k
z 2 j−1 z 2 j = 0 (16.1)
j=1

or

k
z n2 + z 2 j−1 z 2 j = 0, (16.2)
j=1

respectively. Now in these coordinates one easily checks that for an oriented
plane E ∈ G +R (2, n) with
342 H. Süß

¯
(E) = (z 1 : . . . : z n )

we have
 iφ1 
¯
(diag(Q φ1 , . . . , Q φk )E) = e z 1 : e
−iφ1
z 2 : . . . : eiφk z 2k−1 : e−iφ1 z 2k

in the case n = 2k and similarly

 iφ1 
¯
(diag(Q φ1 , . . . , Q φk , 1)E) = e z 1 : e−iφ1 z 2 : . . . : eiφk z 2k−1 : e−iφ1 z 2k : z 2k+1

in the case n = 2k + 1. Let e j denote the jth canonical basis vector of Zk . Then the
action of T = (S 1 )k above is induced by an algebraic torus action of TC = (C∗ )k
with weights

deg(z 2 j−1 ) = e j , deg(z 2 j ) = −e j ; j = 1, . . . , k.

and deg(z 2k+1 ) = 0 in the case n = 2k + 1.

We are now going to describe the GIT quotients in order to eventually construct
the orbit space using Corollary 8. We only describe the case of even n = 2k in detail.
The situation for n odd is very similar.
The moment map is given by

k
μ(z 1 : . . . : z n ) = n (|z 2 j−1 |2 − |z 2 j |2 )e j . (16.3)
i=1 |z i |
2
j=1

The moment image of X is the cross-polytope given as the convex hull of the
weights P = βk = conv(±e1 , . . . , ±ek ). The fixed point (1 : 0 : . . . : 0) is mapped
to e1 , (0 : 1 : . . . : 0) to −e1 and similarly for the other coordinates.

Remark 15 The proper faces of the cross-polytope P are exactly the convex hulls of
subsets of {±e1 , . . . , ±ek } where for every j at most one of e j and −e j is contained.

Lemma 16 The moment preimage of a boundary point u ∈ ∂ P consists of exactly

one T -orbit. Hence, the quotient Yλ(u) is just a single point.

Proof It follows from Remark 15 that a point (z 1 : . . . : z n ) ∈ X is mapped to the

boundary of P if and only if all the products z 2 j−1 z 2 j vanish. Indeed, assume μ(z)
lies in the convex hull conv(σ1 e1 , . . . , σn en ), where σi ∈ {−1, 1} for i = 1, . . . , n.
Then the coefficients of μ(z) in the corresponding barycentric coordinates are up to
sign the same coefficients as in (16.3). Hence, we must have that
k
|z 2 j−1 |2 − |z 2 j |2
j=1
n =1
i=1 |z i |
2

or equivalently
16 Orbit Spaces of Maximal Torus Actions on Oriented Grassmannians … 343

k

k
 
|z 2 j−1 |2 − |z 2 j |2 = |z 2 j−1 |2 + |z 2 j |2 .
j=1 j=1

This implies that for every j = 1, . . . , k either z 2 j−1 = 0 or z 2 j = 0.

Now, assume z = (z 1 : . . . : z n ) and z  = (z 1 : . . . : z n ) have the same moment
image and that z 2 j−1 z 2 j = z 2 j−1 z 2 j = 0, for j = 1, . . . k. By choosing a suitable

j |z j | =
2
representative of the homogeneous coordinates we may assume that
  2
j |z j | = 1. With these choice of homogenous coordinates |z 2 j−1 | − |z 2 j | =
2 2

|z 2 j−1 | − |z 2 j | holds for j = 1, . . . k, since μ(z) = μ(z ). For sign reasons we have
2 2

either z 2 j−1 = z 2 j−1 = 0 or z 2 j = z 2 j = 0. This implies |z 2 j | = |z 2 j | or |z 2 j−1 | =

|z 2 j−1 |, respectively. In either case we have (z 2 j−1 , z 2 j ) = (s j · z 2 j−1 , s −1
j · z 2 j ) for
∗ 
some element s j ∈ S ⊂ C . Hence, z and z lie in the same T -orbit.
1


We consider the rational map

q : Pn−1
C  Pk−2
C , (z 1 : . . . : z n )  → (z 3 z 4 : . . . : z n−1 z n ).

This map is easily seen to be invariant under the TC -action. It is well-defined on the
locus of points where at least one of the products z 2 j−1 z 2 j for j = 2, . . . , k does not
vanish. For a point z ∈ X this is equivalent to the fact that μ(z) ∈ P ◦ .

Lemma 17 For u ∈ P ◦ the map q|μ−1 (u) : μ−1 (u) → Pk−2

C is a quotient map to
the T -orbit space of the fibre.

Proof Consider z, z  ∈ X with μ(z) = μ(z  ) ∈ P ◦ and q(z) = q(z  ). By choosing a

suitable representative of the homogeneous coordinates for z and z  we may assume
that z 2 j−1 z 2 j = z 2 j−1 z 2 j for j = 2, . . . , k − 1. Then the defining equation of X ⊂
 
Pn−1
C implies j z 2 j−1 z 2 j = j z 2 j−1 z 2 j = 0. Hence, also z 1 z 2 = z 1 z 2 must hold.
 
Let us set N = i |z i |2 and N  = i |z i |2 . Assume z 2 j−1 z 2 j = 0 then μ(z) = μ(z  )
implies
|z 2 j−1 |2 − |z 2 j |2 |z 2 j−1 |2 − |z 2 j |2
= .
N N

Hence, for sign reasons we have z 2 j−1 = z 2 j−1 = 0 or z 2 j = z 2 j = 0 in this case.

Now, for each j = 1, . . . k we set s j = z 2 j−1 /z 2 j−1 or s j = z 2 j /z 2 j whichever is
defined. If they are both defined they have to coincide, since z 2 j−1 z 2 j = z 2 j−1 z 2 j .
If z 2 j−1 = z 2 j−1 = z 2 j = z 2 j = 0, then we set s j = 1. By these choices we have s.z =
z  with s = (s1 , . . . , sk ) ∈ TC . It remains to show, that s ∈ T ⊂ TC .
W.l.o.g we may assume that

|s1 | = max{|s1 |, |s1−1 |, . . . , |sk |, |sk−1 |}. (16.4)

The condition μ(z) = μ(z  ) implies

344 H. Süß

|z 1 | − |z 2 | |z  | − |z  |
= 1  2 ,
N N

Note, that by (16.4) we have |s1 |−2 ≤ NN ≤ |s1 |2 . Now, from (z 1 , z 2 ) = (s1 z 1 , s1−1 z 2 )
we obtain
|z 1 |2 − |z 2 |2 |s1 |2 |z 1 |2 − |s1 |−2 |z 2 |2
= ,
N N

which implies (N  /N − |s1 |2 )|z 1 |2 = (N  /N − |s1 |−2 )|z 2 |2 . For sign reasons this is
only possible if |s1 |2 = N  /N = 1. Now, it follows from (16.4) that |s1 | = · · · =
|sk | = 1 and s ∈ T . 

Since X ss (u) consists exactly of the orbits whose closures intersect μ−1 (u) it
follows also that q| X ss (u) is a good quotient in the sense of Geometric Invariant
Theory. Hence, it coincides with the GIT quotient.

Proof (Proof of Theorem 1) We use Corollary 8. Here, Y is just given as Y = Pn/2−2

and by Lemma 16 we have rλ(u) : Y → {pt} for u ∈ ∂ P and by Lemma 17 rλ(u) = idY .
In particular, the equivalent relation on P × Y is just given by (u, y) ∼r (u, y  )
for u ∈ ∂ P. Now, the claim follows from Lemma 13. 

We conclude this section by studying the moment images (x) of TC -orbit clo-
sures and the induced subdivision of P from Sect. 16.2. For n = 2k + 1 the convex
hull of every subset of vertices of P occurs as a moment image. For n = 2k such con-
vex hulls are moment images if an only if they are faces of P or contain at least two
pairs of opposite vertices {ei , −ei }, {e j , −e j }. Indeed, for every such polytope  a
corresponding TC -orbit is given by TC · (z 1 : z 2 : . . . : z n ) with z 2 j−1 = 0 ⇔ e j ∈ 
and z 2 j = 0 ⇔ −ei ∈  for j = 1, . . . , k. Note, that for n = 2k + 1 such (z 1 : z 2 :
. . . : z n ) fulfilling (16.2) always exist, but for n = 2k there is obviously no non-trivial
solution of (16.1) where all but one monomial vanish. In both cases, with the excep-
tion of k = 2, the induced subdivision of P is the same and coincides with the stellar
subdivision of P obtained by starring in the origin.

Remark 18 In [7] Buchstaber and Terzić introduced the notion of (2n, k)-manifold.
It’s relatively straightforward to check that for n = 2k, the axioms of this notion are
indeed fulfilled for the associated effective torus action by TC /±1. However, in
the odd case on a generic point of the hyperplane section [z n = 0] we have finite
stabilisers of order 2, which violates the conditions for a (2n, k)-manifold.

Remark 19 Note that G + R (2, 6) can be identified with G C (2, 4) as both are given by
the smooth quadric hypersurface in P5C . Hence, for this case we just rediscover the
results of [5]. Combinatorially this fact is reflected by coincidence of the moment
polytopes, i.e. the cross-polytope β3 and the hypersimplex 4,2 .
16 Orbit Spaces of Maximal Torus Actions on Oriented Grassmannians … 345

16.4 Complexity-One TC -Varieties

If the complexity of the torus action is 1 the possible GIT quotients Yλ are either single
points or isomorphic to a fixed algebraic curve Y . Hence, when applying Corollary 8
to this situation the maps pλ : Y → Yλ are either isomorphisms or contractions to a
point. Our main aim in this section is to prove that in this situation the resulting orbits
spaces are topological manifolds with boundary and even spheres if the number of
lower dimensional TC -orbits is finite.

Remark 20 In the toric case the orbit space can be identified with the moment poly-
tope. In particular, it is also a topological manifold with boundary. Hence, Theorem 2
can be seen as generalisation of this fact. On the other hand, this phenomenon is very
special to complexity 0 and 1. In higher dimensions this will almost never be the
case. For example for a smooth projective variety Y the join S n ∗ Y , which occurs
as an orbit space in the situation of Proposition 12, is a topological manifold if and
only if Y ∼
= P1C .

Proposition 21 Consider the projective d-space with a (d − 1)-torus TC acting

effectively by weights u 0 = 0, u 1 , . . . , u d ∈ Zd−1 on the coordinates z 0 , . . . , z d . Then
the orbit space PdC /T is homeormorphic to either a disc or a sphere. In particular,
it is a topological manifold with boundary.

Proof The moment image of PdC is given by the convex hull of the weights u 0 , . . . , u d .
The weights u 0 , . . . , u d are necessarily affinely dependent in Rd−1 . On the other hand
they span Rd−1 as an affine space due to the effectiveness of the torus action. Hence,
d d
there is a non-trivial choice of α j ∈ Z, such that 0 = i=0 αi u i and 0 = i=0 αi
and the coefficients are unique up to simultaneous scaling.
Set K = {i ∈ {0, . . . , d} | αi = 0}. Then P is obtained as the join Q ∗  of
the lower-dimensional polytopes Q = conv{u i }i∈K and  = conv{u i }i ∈K / of dimen-
sions m := (#K − 2) and n := (d − #K ), respectively. Here, we allow that  = ∅
and use the non-standard convention Q ∗ ∅ := Q. Note,  that  is a simplex (or
empty).
 Hence, a u ∈  has a unique representation as u= j∈K λ j u j with λ j ≥ 0
and j λ j = 1. For u ∈ Q such a representation u = j ∈K / λ j u j is unique if and
only if u ∈ ∂ Q. It follows that u ∈ P = Q ∗  has a unique such representation if
and only if u ∈ ∂ Q ∗ .
Now,Yλ(u) = μ−1 (u)/T is a point whenever u ∈ P has a unique representation
as u = j λ j u j and μ−1 (u)/T ≈ P1C otherwise. This is just a special case of Exam-
ple 6, when F : Rd → Rk is given by F(ei ) = u i for i = 1, . . . , d. Then theintersec-
tion of F −1 (u) and the standard simplex consists nof all linear combinations

n
i=1 λi ei
with non-negative coefficients, such that u = i=1 λi u i and λi = 1. The result is
a point if the linear combination is unique or a line segment if not. The correspond-
ing toric varieties are a single point and P1C , respectively. Alternatively, it not hard to
show that the non-trivial quotient morphisms X λss → Yλ = P1 are all restrictions of
the rational map
346 H. Süß
 

αi

P  P ,
d 1
(z 0 : . . . : z d ) → z 0 i
: z iαi .
i

From Corollary 8 we obtain that

 
PdC /T ≈ P × P1C / ∼,

where the equivalence relation ∼ is generated by (u, y) ∼ (u, y  ) with y, y  ∈ P1C

and u ∈ ∂ Q ∗  ⊂ ∂ P. Topologically ∂ Q can be identified with a sphere S m−1 and Q
with the cone S m−1 ∗ {pt}. Similarly we have a homeomorphism  ≈ S n−1 ∗ {pt}
for  = ∅. For the pair (P, ∂ Q ∗ ) we obtain

(P, ∂ Q ∗ ) ≈ (S m−1 ∗ pt ∗S n−1 ∗ {pt}, S m−1 ∗ (S n−1 ∗ {pt}))

≈ (S m+n−1 ∗ {pt} ∗ {pt}, S m+n−1 ∗ {pt})
≈ (D m+n ∗ {pt}, D m+n ).

Note, that the disc D m+n can be identified with the hemisphere via projection
and D m+n ∗ {pt} with the corresponding halfdisc. Now, by choosing H to be an
arbitrary halfspace it follows from Lemma 14 that the orbit space is a hemisphere.
For  = ∅ and P = Q we see directly (P, ∂ Q ∗ ) = (Q, ∂ Q) ≈ (D d−1 , S d−2 )
and we obtain PdC /T ≈ S d+1 from invoking Lemma 14 again, this time with H =
Rd−1 . 
Proof (Proof of Theorem 2) We first consider the situation of a complexity-one
torus action on the affine space Cd . Such an action is linearisable by [4]. We may
equivariantly compactify the TC -action on Cd to a TC -action on PdC . Then Cd /T is an
open subset of PdC /T . Hence, the claim follows from the observation in Proposition 21
that the orbit space PdC /T is a manifold with boundary.
To deduce the general case we consider the two situations from Lemma 23. If
contractions to a point do not occur the equivalence relation ∼r is trivial and by
Corollary 8 we have X/T ≈ P × Y which is a product of topological manifolds
with boundary.
In the second situation, we have Y ∼ = P1 . Then by [1, Theorem 5] we have an
equivariant open cover of X by copies of Cn and the result follows directly from the
consideration above. 
Remark 22 To deduce the general case from the case X = Cd in the proof of
Theorem 2 we may alternatively consider the induced TC -action on the tangent space
at a fixed point and reduce everything to this situation by applying Luna’s Slice
Theorem.

Lemma 23 ([14, Lemma 5.7]) In the situation of a complexity-one TC -action on a

smooth projective variety either
1. the GIT quotients are all isomorphic to Y , or
2. Y ∼= P1C .
16 Orbit Spaces of Maximal Torus Actions on Oriented Grassmannians … 347

Consider the following definition.

Definition 24 A k-holed n-sphere is defined as the intersection of S n ⊂ Rn+1 with k
affine closed halfspaces, such that S n is contained in none of them and their boundary
hyperplanes intersect only outside the sphere.
In other words, we remove the interiors of k disjoint closed discs from S n .
Proposition 25 Consider a smooth TC -variety X of complexity 1 with moment
polytope P and stratification . Let  be the set of λ ∈ , such that λ ⊂ ∂ P
and Yλ = {pt}. Assume that  consists of k disjoint polytopes. Then X/T is a k-
holed sphere.
Proof In our situation the equivalence  relation ∼r from Corollary 8 is generated
 ◦
by
 (u, y) ∼ r (u, y ) for u ∈ ∂ P \ λ∈  λ . Note that by the preconditions ∂ P \
◦
λ∈  λ is a k-holed sphere. Hence, by Lemma 14 the orbit space X/TC ≈ (P ×
P1 )/ ∼r is a k-holed sphere as well. 
This statement is useful to determine the orbit space in concrete situation. To
demonstrate this we look at the classification of Fano threefolds from [19]. In [20,
21] the ones within the classification which admit a TC -action of complexity 1 where
identified. In the following we determine the orbit spaces for all of them.
Theorem 26 Using the notation of the Mori-Mukai classification [19] for the Fano
threefolds with (C∗ )2 -action we obtain the orbits spaces being k-holed spheres
with k = 0 for Q, 2.24, 2.29, 2.32, 3.10 and 3.20 with k = 1 for 2.30, 2.31, 3.18,
3.21, 3.23 and 3.24, with k = 2 for 3.19, 3.22, 4.4, 4.5, 4.7 and 4.8.
Proof For every of the above Fano varieties the combinatorial data provided in [21,
Sect. 5] consists of the moment polytope P and a piecewise linear map

 : P → DivR P1C

from the polytope to the vector space of R-divisors on P1 . By [21, Sect. 3.1] we
have Yλ(u) = {pt} if and only deg (u) = 0. Now, one checks that in every case
the subset {u ∈ ∂ P | deg (u) > 0} ⊂ ∂ P consists of the interior of k disjoint facts
of P. Applying Proposition 25 gives the desired result. 
Remark 27 Note, that Q is the smooth quadric and by Sect. 16.3 coincides with
G+R (2, 5). Moreover, 2.32 is the variety of complete flags in C . Hence, we recover
3

results of [7, Proposition 8] and [15].

Proof (Proof of Theorem 3) To have finitely many lower dimensional TC -orbits
implies that the GIT quotients Yu for u ∈ ∂ P are just points. Then by Lemma 23
we conclude that Yu = P1C for every u ∈ P ◦ . By applying Proposition 12 we obtain

X/T ≈ P1 ∗ S k−1 ≈ S 2 ∗ S k−1 ≈ S k+2 .

348 H. Süß

Acknowledgements This research was supported by the program Interdisciplinary Research con-
ducted jointly by the Skolkovo Institute of Science and Technology and the Interdisciplinary Sci-
entific Center J.-V. Poncelet. In particular, I am grateful for the warm hospitality offered by Center
Poncelet. This work was also partially supported by the grant 346300 for IMPAN from the Simons
Foundation and the matching 2015–2019 Polish MNiSW fund. I would like to thank Anton Ayzen-
berg, Victor Buchstaber, Alexander Perepechko and Nigel Ray for stimulating discussions on the
subject of this paper. Finally, I want to thank Yael Karshon and Susan Tolman for making their
draft [15] available to me.

References

1. Arzhantsev, I., Perepechko, A., Süß, H.: Infinite transitivity on universal torsors. J. Lond. Math.
Soc. (2) 89(3), 762–778 (2014)
2. Atiyah, M.F.: Convexity and commuting Hamiltonians. Bull. Lond. Math. Soc. 14(1), 1–15
(1982)
3. Aı̆zenberg, A.A.: Torus actions of complexity 1 and their local properties. Tr. Mat. Inst. Steklova
302(Topologiya i Fizika), 23–40 (2018). English version published in Proc. Steklov Inst. Math.
302(1), 16–32 (2018)
4. Biał ynicki Birula, A.: Remarks on the action of an algebraic torus on k n . Bull. Acad. Polon.
Sci. Sér. Sci. Math. Astronom. Phys. 14, 177–181 (1966)
5. Buchstaber, V.M., Terzić, S.: Topology and geometry of the canonical action of T 4 on the
complex Grassmannian G 4,2 and the complex projective space CP5 . Mosc. Math. J. 16(2),
237–273 (2016)
6. Buchstaber, V.M., Terzić, S.: Toric topology of the complex Grassmann manifolds. Mosc.
Math. J. 19(3), 397–463 (2019)
7. Bukhshtaber, V.M., Terzich, S.: The foundations of (2n, k)-manifolds. Mat. Sb. 210(4), 41–86
(2019)
8. Cherepanov, V.V.: Orbit spaces of torus actions on Hessenberg varieties (2019).
arXiv:1905.02294 [math.AT]
9. Dolgachev, I.V., Hu, Y.: Variation of geometric invariant theory quotients. Inst. Hautes Études
Sci. Publ. Math. (87), 5–56 (1998). With an appendix by Nicolas Ressayre
10. Goresky, M., MacPherson, R.: On the topology of algebraic torus actions. In: Algebraic groups
Utrecht 1986, Lecture Notes in Mathematics, vol. 1271, pp. 73–90. Springer, Berlin (1987)
11. Guillemin, V., Sternberg, S.: Convexity properties of the moment mapping. Invent. Math. 67(3),
491–513 (1982)
12. Hu, Y.: The geometry and topology of quotient varieties of torus actions. Duke Math. J. 68(1),
151–184 (1992)
13. Kapranov, M.M., Sturmfels, B., Zelevinsky, A.V.: Quotients of toric varieties. Math. Ann.
290(4), 643–655 (1991)
14. Karshon, Y., Tolman, S.: Centered complexity one Hamiltonian torus actions. Trans. Amer.
Math. Soc. 353(12), 4831–4861 (2001)
15. Karshon, Y., Tolman, S.: Topology of complexity one quotients. Pacific J. Math. 308(2), 333–
346 (2020)
16. Kirwan, F.: Momentum maps and reduction in algebraic geometry. pp. 135–171 (1998). Sym-
plectic geometry
17. Kirwan, F.C.: Cohomology of quotients in symplectic and algebraic geometry, Mathematical
Notes, vol. 31. Princeton University Press, Princeton (1984)
18. Kobayashi, S., Nomizu, K.: Foundations of differential geometry. Vol. II. Interscience Tracts
in Pure and Applied Mathematics, No. 15 Vol. II. Interscience Publishers John Wiley & Sons,
Inc., New York (1969)
16 Orbit Spaces of Maximal Torus Actions on Oriented Grassmannians … 349

19. Mori, S., Mukai, S.: Classification of Fano 3-folds with B2 ≥ 2. Manuscripta Math. 36(2),
147–162 (1981/82)
20. Przhiyalkovskiı̆, V.V., Cheltsov, I.A., Shramov, K.A.: Fano threefolds with infinite automor-
phism groups. Izv. Ross. Akad. Nauk Ser. Mat. 83(4), 226–280 (2019)
21. Süss, H.: Fano threefolds with 2-torus action: a picture book. Doc. Math. 19, 905–940 (2014)
Chapter 17
The Reflexive Dimension
of (0, 1)-Polytopes

Akiyoshi Tsuchiya

Abstract Haase and Melnikov showed that every lattice polytope is unimodularly
equivalent to a face of some reflexive polytope. The reflexive dimension of a lattice
polytope P is the minimal d so that P is unimodularly equivalent to a face of some d-
dimensional reflexive polytope. Computing the reflexive dimension of a lattice poly-
tope is a hard problem in general. In this survey, we discuss the reflexive dimension
of a (0, 1)-polytope. In particular, virtue of the algebraic technique on Gröbner bases
and a linear algebraic technique, many families of reflexive polytopes arising from
several classes of (0, 1)-polytopes are presented, and we see that the (0, 1)-polytopes
are unimodularly equivalent to facets of some reflexive polytopes.

Keywords Reflexive polytope · Reflexive dimension · Order polytope · Stable set

polytope · Edge polytope · Gröbner basis

17.1 Introduction

The reflexive polytope is one of the keywords belonging to the current trends in the
research of convex polytopes. In fact, many authors have studied reflexive polytopes
from the viewpoints of combinatorics, commutative algebra and algebraic geometry.
Hence, finding new classes of reflexive polytopes is an important problem.
A lattice polytope is a convex polytope all of whose vertices have integer coor-

dinates. Two lattice polytopes P ⊆ Rd and P ⊆ Rd are said to be unimodularly
equivalent if there exists an affine map from the affine span aff(P) of P to the

affine span aff(P ) of P that maps Zd ∩ aff(P) bijectively onto Zd ∩ aff(P ) and
that maps P to P . Note that every lattice polytope is unimodularly equivalent to a
full-dimensional one. A lattice polytope P ⊂ Rd of dimension d is called reflexive
if the origin of Rd is the unique lattice point belonging to the interior of P and its

A. Tsuchiya (B)
Graduate School of Mathematical Sciences, University of Tokyo, 3-8-1 Komaba, Meguro-ku,
Tokyo 153-8914, Japan
e-mail: [email protected]

© Springer Nature Switzerland AG 2022 351

A. M. Kasprzyk and B. Nill (eds.), Interactions with Lattice Polytopes,
Springer Proceedings in Mathematics & Statistics 386,
https://doi.org/10.1007/978-3-030-98327-7_17
352 A. Tsuchiya

dual polytope
P∨ := {y ∈ Rd | x, y ≤ 1 for all x ∈ P}

is also a lattice polytope, where x, y is the usual inner product of Rd . It is known
that reflexive polytopes correspond to Gorenstein toric Fano varieties, and they are
related to mirror symmetry, e.g., [1, 3]. In each dimension, there exist only finitely
many reflexive polytopes up to unimodular equivalence [15]. Moreover, Haase and
Melnikov [4] showed that every lattice polytope is unimodularly equivalent to a face
of some reflexive polytope. From this result they defined the reflexive dimension of
a lattice polytope.

Definition 1 Let P be a lattice polytope. The reflexive dimension of P, denoted

by rdim(P), is the minimal integer d such that P is unimodularly equivalent to a face
of some d-dimensional reflexive polytope.

We immediately know the following proposition from the fact that there are only
finitely many reflexive polytopes up to unimodular equivalence.

Proposition 2 Given a positive integer d, there exist up to unimodular equivalence

only finitely many lattice polytopes whose reflexive dimensions are equal to d.

Our interest is to classify lattice polytopes whose reflexive dimensions are a given
integer. We remark that classifying lattice polytopes whose reflexive dimensions
are equal to their dimensions is equivalent to classifying reflexive polytopes. In
particular, it is known that there is one reflexive polytope in dimension one, there
are 16 in dimension two, 4,319 in dimension three and 473,800,776 in dimension
four according to computations by Kreuzer and Skarke [14]. As a next step, we focus
on lattice polytopes whose reflexive dimensions are equal to their dimensions plus
one. Namely, we consider the following question.

Question 3 For which lattice polytope P, does it follow that rdim(P) = dim(P) +
1?

We note that for a lattice polytope P, rdim(P) = dim(P) + 1 if and only if P is

not reflexive and P is a facet of some reflexive polytope. For example, the reflexive
dimensions of a d-dimensional unit cube and a d-dimensional unit simplex are d + 1.
A stronger question is the following.

Question 4 For every (0, 1)-polytope P, does it follow that rdim(P) = dim(P) + 1?
Equivalently, is every (0, 1)-polytope a facet of some reflexive polytope?

In order to show that this second Question 4 has a positive answer for some
class of (0, 1)-polytopes, we give higher-dimensional construction of lattice poly-
topes. Given two lattice polytopes P ⊂ Rd and Q ⊂ Rd , we set the lattice polytope
(P, Q) ⊂ Rd+1 with

(P, Q) := conv{(P × {1}) ∪ (−Q × {−1})}.

17 The Reflexive Dimension of (0, 1)-Polytopes 353

If P = Q, then we will write (P) := (P, P). We remark that the origin of Rd+1
is always a relative interior lattice point of (P). Assume that P is full-dimensional.
Then (P) is also full-dimensional. In particular, P × {1} is a facet of (P). Hence
for a lattice polytope P ⊂ Rd of dimension d, if (P) is reflexive, then we know
that P is unimodularly equivalent to a facet of some reflexive polytope.
In this survey, by using this construction, we will show that for several classes
of (0, 1)-polytopes, Question 4 has a positive answer. In particular, we will present
many large families of reflexive polytopes arising from well-known classes of (0, 1)-
polytopes.
This survey is organized as follows: In Sect. 17.2, we will introduce an algebraic
technique to show that a given lattice polytope is reflexive. In particular, we will recall
basic materials and notation on toric ideals. In Sect. 17.3, we will give two families
of reflexive polytopes arising from the order polytopes and the chain polytopes of
finite partially ordered sets. In Sect. 17.4, we will give a family of reflexive polytopes
arising from the stable set polytopes of perfect graphs. Finally, in Sect. 17.5, we will
introduce a linear algebraic technique to show that a lattice polytope is reflexive, and
we will give a family of reflexive polytopes arising from the edge polytopes of finite
simple graphs.

17.2 Toric Ideals and Reflexive Polytopes

In this section, we introduce an algebraic technique to show that a lattice poly-

tope is reflexive. First, we recall basic materials and notation on toric ideals.
Let K [t±1 , s] := K [t1±1 , . . . , td±1 , s] be the Laurent polynomial ring in d + 1 vari-
ables over a field K . If a = (a1 , . . . , ad ) ∈ Zd , then ta s is the Laurent mono-
mial t1a1 · · · tdad s ∈ K [t±1 , s]. Let P ⊂ Rd be a lattice polytope of dimension d
and P ∩ Zd = {a1 , . . . , an }. Then, the toric ring of P is the subalgebra K [P]
of K [t±1 , s] generated by {ta1 s, . . . , tan s} over K . We regard K [P] as a homo-
geneous algebra by setting each deg tai s = 1. Let K [x] := K [x1 , . . . , xn ] denote
the polynomial ring in n variables over K . The toric ideal IP of P is the ker-
nel of the surjective homomorphism π : K [x] → K [P] defined by π(xi ) = tai s
for 1 ≤ i ≤ n. It is known that IP is generated by homogeneous binomials, see,
e.g., [27]. Let < be a monomial order on K [x] and in< (IP ) the initial ideal of IP
with respect to <. The initial ideal in< (IP ) is called squarefree if in< (IP ) is gener-
ated by squarefree monomials. The reverse lexicographic order on K [x] induced
by the ordering xn <rev · · · <rev x1 is the total order <rev on the set of mono-
n variablesxn1 , x2 , . . . , xn 
mials in the x1a1 x2a2 · · · xnan <rev x1b1 x2b2 · · · xnbn if
by setting 
n n
either (i) i=1 ai < i=1 bi , or (ii) i=1 ai = i=1 bi and the rightmost nonzero
component of the vector (b1 − a1 , b2 − a2 , . . . , bn − an ) is negative. A reverse lex-
icographic order is also called a graded reverse lexicographic order. Please refer
to [5, Chaps. 1 and 5] for more details on Gröbner bases and toric ideals.
Next, we recall important classes of lattice polytopes. We say that a lattice
polytope P ⊂ Rd possesses the integer decomposition property if, for each inte-
354 A. Tsuchiya

ger k ≥ 1 and for each a ∈ kP ∩ Z N , there exist a1 , . . . , ak belonging to P ∩ Z N

with a = a1 + · · · + ak , where kP := {kx : x ∈ P} is the kth dilated polytope of P.
The integer decomposition property is particularly important in the theory and appli-
cation of integer programming [25, Sect. 22.10]. Moreover, a lattice polytope which
possesses the integer decomposition property is normal and very ample. Hence
these properties play important roles in algebraic geometry. A lattice polytope P
is called compressed [28] if all its “pulling triangulations” are unimodular. In partic-
ular, if any lattice point in Zd+1 is a linear integer combination of the lattice points
in P × {1}, then P is compressed if and only if every reverse lexicographic initial
ideal of IP is squarefree [27]. We note that a lattice polytope with a unimodular
triangulation possesses the integer decomposition property, hence, all compressed
polytopes possess the integer decomposition property.
Now, we introduce an algebraic technique to show that a lattice polytope is reflex-
ive.

Lemma 5 ([8, Lemma 1.1]) Let P ⊂ Rd be a lattice polytope of dimension d such

that the origin 0 of Rd is contained in its interior and P ∩ Zd = {a1 , . . . , an }. Suppose
that any lattice point in Zd+1 is a linear integer combination of the lattice points
in P × {1} and there exists an ordering of the variables xi1 <rev · · · <rev xin for
which ai1 = 0 such that the initial ideal in<rev (IP ) of the toric ideal IP with respect
to the reverse lexicographic order <rev on the polynomial ring K [x] induced by the
ordering is squarefree. Then P is a reflexive polytope with a regular unimodular
triangulation.

Example 6 Set ai = ei for i = 1, . . . , d, ad+1 = −e1 − · · · − ed and ad+2 = 0,

where e1 , . . . , ed are the standard coordinate unit vectors of Rd . Let P ⊂ Rd be
the lattice polytope with P ∩ Zd = {a1 , . . . , ad+2 }. Then one has

IP = (x1 · · · xd+1 − xd+2

d+1
).

Hence the initial ideal in<rev (IP ) of the toric ideal IP with respect to the reverse lexi-
cographic order <rev on the polynomial ring K [x] induced by the ordering xd+2 <rev
xd+1 <rev · · · <rev x1 is squarefree. Therefore, by Lemma 5, P is a reflexive polytope
with a regular unimodular triangulation.

Recently, several families of reflexive polytopes with regular unimodular triangula-

tions have been given by using this technique [7–13, 21–24].

17.3 Reflexive Polytopes Arising from Order Polytopes and

Chain Polytopes

In this section, we give two families of reflexive polytopes with regular unimodular
triangulations arising from order polytopes and chain polytopes of finite partially
ordered sets. In particular, we show that Question 4 has a positive answer for order
and for chain polytopes.
17 The Reflexive Dimension of (0, 1)-Polytopes 355

First, we recall some terminologies of finite partially ordered sets and introduce
two lattice polytopes arising from finite partially ordered sets. Let P denote a finite
partially ordered set (poset, for short) on the ground set [d] := {1, . . . , d}. A subset I
of [d] is called a poset ideal of P if i ∈ I and j ∈ P together with j ≤ i in P,
then j ∈ I . Note that the empty set ∅ and [d] are poset ideals of P. Let J(P)
denote the set of poset ideals of P. A subset A of [d] is called an antichain of P if
i and j belonging to A with i = j are incomparable. In particular, the empty set ∅
and each 1-element subsets { j} are antichains of P. Let A(P) denote the set of
antichains of P. For a poset ideal I of P, we write max(I ) for the set of maximal
elements of I . In particular, max(I ) is an antichain. A linear extension of P is a
permutation σ = i 1 i 2 · · · i d of [d] which satisfies a < b if i a < i b in P.
Stanley [26] introduced two classes of lattice polytopes arising from finite posets,
which are called order polytopes and chain polytopes. The order polytope O P of P
is defined to be the convex polytope consisting of those (x1 , . . . , xd ) ∈ Rd such that:
1. 0 ≤ xi ≤ 1 for 1 ≤ i ≤ d;
2. xi ≥ x j if i ≤ j in P.
The chain polytope C P is defined to be the convex polytope consisting of those
(x1 , . . . , xd ) ∈ Rd such that:
1. xi ≥ 0 for 1 ≤ i ≤ d;
2. xi1 + · · · + xik ≤ 1 for every maximal chain i 1 < · · · < i k of P.

For each subset I ⊂ [d], we define the (0, 1)-vectors ρ(I ) := i∈I ei . In particu-
lar ρ(∅) is the origin 0 of Rd . Both order polytopes and chain polytopes are (0, 1)-
polytopes of dimension d. In fact, in [26, Corollary 1.3 and Theorem 2.2], it is shown
that

{the set of vertices of O P } = {ρ(I ) : I ∈ J(P)},

{the set of vertices of C P } = {ρ(A) : A ∈ A(P)}.

Moreover, both order polytopes and chain polytopes are compressed, hence, they
possess the integer decomposition property. However, the class of order polytopes is
different from the class of chain polytopes [6, Example 3.5 and Corollary 3.9].
Now, we consider the two lattice polytopes (O P ) and (C P ) for a finite poset P
on [d]. In particular, we will see that the toric ideals of these lattice polytopes are
squarefree with respect to some monomial orders on K [x]. Let

K [O] := K [{x I , y I } I ∈J(P) ∪ {z}]

K [C] := K [{xmax(I ) , ymax(I ) } I ∈J(P) ∪ {z}]

denote the polynomial rings over K , and define the surjective ring homomorphisms
πO and πC by the following:
356 A. Tsuchiya

1. πO : K [O] → K [(O P )] by setting

πO (x I ) = tρ(I ∪{d+1}) s,
πO (y J ) = t−ρ(J ∪{d+1}) s, and
πO (z) = s;

2. πC : K [C] → K [(C P )] by setting

πC (xmax(I ) ) = tρ(max(I )∪{d+1}) s,

πC (ymax(J ) ) = t−ρ(max(J )∪{d+1}) s, and
πC (z) = s.

Here I, J ∈ J(P). Then the toric ideal I(O P ) (resp. I(C P ) is the kernel of πO
(resp. πC ).
Next, we introduce monomial orders <O and <C , and GO and GC which are the
set of binomials. Let <O denote a reverse lexicographic order on K [O] satisfying:
1. z <O y J <O x I ;
2. x I  <O x I if I  ⊂ I ;
3. y J  <O y J if J  ⊂ J ;
and GO ⊂ K [O] the set of binomials:

x I x I  − x I ∪I  x I ∩I  ,
y J y J  − y J ∪J  y J ∩J ,
x I y J − x I \{i} y J \{i} ,
x∅ y∅ − z 2 .

Let <C denote a reverse lexicographic order on K [C] satisfying:

1. z <C ymax(J ) <C xmax(I ) ;
2. xmax(I  ) <C xmax(I ) if I  ⊂ I ;
3. ymax(J  ) <C ymax(J ) if J  ⊂ J ;
and GC ⊂ K [C] the set of binomials:

xmax(I ) xmax(I  ) − ymax(I ∪I  ) ymax(I ∗I  ) ,

ymax(J ) ymax(J  ) − ymax(J ∪J  ) ymax(J ∗J  ) ,
xmax(I ) ymax(J ) − xmax(I )\{i} ymax(J )\{i} ,
x∅ y∅ − z 2 .

Here
1. I and I  (resp. J and J  ) are poset ideals of P which are incomparable in J(P);
17 The Reflexive Dimension of (0, 1)-Polytopes 357

2. I ∗ I  (resp. J ∗ J  ) is the poset ideal of P generated by max(I ∩ I  ) ∩ (max(I ) ∪

max(I  )) (resp. max(J ∩ J  ) ∩ (max(J ) ∪ max(J  )));
3. i is a maximal element of both I and J .

Proposition 7 ([11, Propositions 2.2 and 2.4]) Let the notation be as above. Then GO
(resp. GC ) is a Gröbner basis of I(O P ) (resp. I(C P ) ) with respect to <O (resp. <C ).

By combining Lemma 5 and Proposition 7 we obtain the following theorem.

Theorem 8 ([11, Theorem 1.3]) Let P be a finite poset on [d]. Then each of (O P )
and (C P ) is a reflexive polytope with a regular unimodular triangulation.

Hence we have:
Corollary 9 Let P be a finite poset on [d]. Then one has

rdim(O P ) = rdim(C P ) = d + 1.

Remark 10 In [11], larger families of reflexive polytopes are given. In fact, for two
finite posets P and Q with |P| = |Q| = d, the three lattice polytopes (O P , O Q ),
(O P , C Q ) and (C P , C Q ) are studied. By the same technique, we know that (O P ,
O Q ) is a reflexive polytope (with a regular unimodular triangulation) if and only if P
and Q have a common linear extension, and (O P , C Q ) and (C P , C Q ) are always
reflexive polytopes with regular unimodular triangulations [11, Theorem 1.3].

Remark 11 In [7, 9, 10], by using other constructions of lattice polytopes, large

families of reflexive polytopes are presented. Given two lattice polytopes P ⊂ Rd
and Q ⊂ Rd , we set the lattice polytope (P, Q) ⊂ Rd with

(P, Q) := conv{P ∪ (−Q)}.

If P = Q, then we will write (P) := (P, P). The two lattice polytopes (P, Q)
and (P, Q) often have the same properties. In fact, (O P , O Q ) is a reflexive polytope
(with a regular unimodular triangulation) if and only if P and Q have a common
linear extension [7, Corollary 2.2]. Moreover, (O P , C Q ) and (C P , C Q ) are always
reflexive polytopes with regular unimodular triangulations; see [10, Corollary 1.2]
and [9, Corollary 1.3]. In [9, 29], combinatorial properties of these polytopes, for
example, their volumes, are studied.

17.4 Reflexive Polytopes Arising from the Stable Set

Polytopes of Perfect Graphs

In this section, we give a family of reflexive polytopes with regular unimodular

triangulations arising from the stable set polytopes of perfect graphs. In particular,
we show that Question 4 has a positive answer for the stable set polytopes of perfect
graphs.
358 A. Tsuchiya

First, we recall what perfect graphs are and introduce the stable set polytopes of
finite simple graphs. Let G be a finite simple graph on the vertex set [d] and E(G)
the set of edges of G. (A finite graph G is called simple if G possesses no loop and
no multiple edge.) A subset W ⊂ [d] is called stable if, for all i and j belonging
to W with i = j, one has {i, j} ∈ / E(G). We remark that a stable set is often called
an independent set. A clique of G is a subset W ⊂ [d] which is a stable set of
the complementary graph G of G. The clique number ω(G) of G is the maximal
cardinality of a clique of G. The chromatic number χ (G) of G is the smallest
integer t ≥ 1 for which there exist stable set W1 , . . . , Wt of G with [d] = W1 ∪
· · · ∪ Wt . In general, it follows that ω(G) ≤ χ (G). A finite simple graph G is said to
be perfect [2] if, for any induced subgraph H of G including G itself, one has ω(H ) =
χ (G). Perfect graphs include many important classes of graphs, for example, chordal
graphs and comparability graphs. Moreover, it is known that the complementary
graph of a perfect graph is perfect. This characterization of perfect graphs is called
the perfect graph theorem. Recently, a stronger characterization of perfect graphs,
which is called the strong perfect graph theorem, is known. An odd hole is an induced
odd cycle of length ≥ 5 and an odd antihole is the complementary graph of an odd
hole.

Proposition 12 ([2, Strong Perfect Graph Theorem]) A finite simple graph G is

perfect if and only if G has no odd hole and no odd antihole as induced subgraph.

Next, we introduce the stable set polytopes of finite simple graphs. Let S(G)
denote the set of stable sets of G. One has ∅ ∈ S(G) and {i} ∈ S(G) for each i ∈ [d].
The stable set polytope QG ⊂ Rd of G is the (0, 1)-polytope which is the convex hull
of {ρ(W ) : W ∈ S(G)} in Rd . Then the dimension QG is equal to d. It is known that
every chain polytope is a stable set polytope. In fact, let P be a finite poset on [d].
Its comparability graph G P is the finite simple graph on [d] such that {i, j} ∈ E(G P )
if and only if i < j or j < i in P. Then a stable set of G P corresponds to an antichain
of P. Moreover, one has C P = QG P . Since every comparability graph is perfect, the
class of the chain polytopes is contained in the class of the stable set polytopes of
perfect graphs. We see a characterization of perfect graphs in terms of the stable set
polytopes.

Proposition 13 ([19, Example 1.3 (c)]) Let G be a finite simple graph on [d].
Then G is perfect if and only if QG is compressed.

Now, we consider the lattice polytope (QG ) for a perfect graph G on [d]. In
particular, we see that the toric ideal of this lattice polytope is squarefree with respect
to some monomial order on K [x]. Let

K [Q] := K [{x S , yS } S∈S(G) ∪ {z}]

denote the polynomial ring over K and define the surjective ring homomorphism
πQ : K [Q] → K [(QG )] by:
17 The Reflexive Dimension of (0, 1)-Polytopes 359

πQ (x S ) = tρ(S∪{d+1}) s,
πQ (yT ) = t−ρ(T ∪{d+1}) s, and
πQ (z) = s.

Here S, T ∈ S(G). Then the toric ideal I(QG ) is the kernel of πQ .

Next, we introduce monomial order <Q and MQ which is the set of monomials.
Let <Q denote a reverse lexicographic order on K [Q] satisfying:
1. z <Q yT <O x S ;
2. x S  <Q x S if S  ⊂ S;
3. yT  <Q yT if T  ⊂ T ;
and set

M := {x∅ y∅ } ∪ {x S yT : S, T ∈ S(G), S ∩ T = ∅} ∪ MG ∪ MG ,

where MG (resp. MG ) is the minimal set of squarefree monomial generators

of in<Q (IQG ) (resp. in<Q (I(−QG ) )). Let GQ be a finite set of binomials belonging
to I(QG ) with M = {in<Q ( f ) : f ∈ GQ }.

Proposition 14 ([12, Theorem 2.1]) Let the notation be as above. Then GQ is a

Gröbner basis of I(QG ) with respect to <Q .

By combining Lemma 5 and Proposition 14, we can show that (QG ) is a reflexive
polytope with a regular unimodular triangulation. Moreover, we can give a polyhedral
characterization of perfect graphs.

Theorem 15 ([12, Theorem 1.1 (b)]) Let G be a finite simple graph on [d]. Then
the following arguments are equivalent:
1. G is perfect;
2. (QG ) is a reflexive polytope with a regular unimodular triangulation;
3. (QG ) has a regular unimodular triangulation.

Hence we can obtain the following corollary.

Corollary 16 Let G be a perfect graph on [d]. Then one has rdim(QG ) = d + 1.

Remark 17 In [12], a lager family of reflexive polytopes is given. In fact, for two
finite simple graphs G 1 and G 2 on [d], the lattice polytope (QG 1 , QG 2 ) is studied.
By the same technique, it follows from [12, Theorem 1.1 (b)] that the following
arguments are equivalent:
1. G 1 and G 2 are perfect;
2. (QG 1 , QG 2 ) is a reflexive polytope with a regular unimodular triangulation;
3. (QG 1 , QG 2 ) has a regular unimodular triangulation.
360 A. Tsuchiya

Remark 18 In [21], the lattice polytope (QG 1 , QG 2 ) for two finite simple graphs G 1 ,
G 2 on [d] is studied. The two lattice polytopes (QG 1 , QG 2 ) and (QG 1 , QG 2 ) have
a very similar but different property. In fact, it follows from [21, Theorem 2.8] that
the following arguments are equivalent:
1. G 1 and G 2 are perfect;
2. (QG 1 , QG 2 ) is a reflexive polytope with a regular unimodular triangulation;
3. (QG 1 , QG 2 ) is a reflexive polytope.

Remark 19 In [13], the lattice polytopes (QG , O P ) and (QG , O P ) for a finite
simple graph G on [d] and a finite poset P on [d] are studied. In fact, it follows
from [13, Theorem 1.2] that the following arguments are equivalent:
1. G is perfect;
2. (QG , O P ) has a regular unimodular triangulation;
3. (QG , O P ) is a reflexive polytope.
In particular, if G is perfect, then each of (QG , O P ) and (QG , O P ) is a reflexive
polytope with a regular unimodular triangulation.

17.5 Reflexive Polytopes Arising from Edge Polytopes

When we show that (P) is reflexive for a lattice polytope P ⊂ Zd of dimension d

by using Lemma 5, we need to assume that P possesses the integer decomposition
property. In fact, order polytopes, chain polytopes and the stable set polytopes of per-
fect graphs are compressed, hence they possess the integer decomposition property.
In order to show that Question 4 has a positive answer for a class of (0, 1)-polytopes
which do not necessarily possess the integer decomposition property, we should con-
sider other approaches. In this section, we introduce a linear algebraic technique to
show that a lattice polytope which does not necessarily possess the integer decom-
position property is reflexive, and we give one family of reflexive polytopes arising
from the edge polytopes of finite simple graphs.
For two d × d integer matrices A, B, we write A ∼ B if B can be obtained from A
by some row and column operations over Z. Let X = {x1 , . . . , xn } be a set of lattice
points in Rd . Given a subset I = {i 1 , . . . , i d+1 } ⊂ [n], let X I be the (d + 1) × (d +
1) matrix whose ith row vector is (xi , 1).

Lemma 20 ([16, Proof of Theorem 2.1]) Let X = {x1 , . . . , xn } be a set of lattice

points in Rd and P ⊂ Rd a lattice polytope of dimension d all of whose vertices
belong to X . Assume that for any subset I = {i 1 , . . . , i d+1 } ⊂ [n] with det(X I ) = 0,
it follows that for some integer 0 ≤ s ≤ d,
17 The Reflexive Dimension of (0, 1)-Polytopes 361
⎛ ⎞⎫
1 ⎬
⎜ .. ⎟ s
⎜ .
⎜
⎜ 1
0 ⎟⎭
⎟
⎟
XI ∼ ⎜
⎜
⎟ .
⎟
⎜ 2 ⎟
⎜ ⎟
⎝ 0 ..
.
2
⎠

Then (P) is reflexive.

In order to give a family of reflexive polytopes, to find X which satisfies the

condition of Lemma 20 is an interesting problem. In fact, by using this lemma, we
can get a large family of reflexive polytopes.

Proposition 21 ([16, Proposition 2.3]) Set

X = {0} ∪ {ei : 1 ≤ i ≤ d} ∪ {ei + e j : 1 ≤ i ≤ j ≤ d} ⊂ Rd .

Then for any subset I = {i 1 , . . . , i d+1 } ⊂ [n] with det(X I ) = 0, it follows that for
some integer 0 ≤ s ≤ d,
⎛ ⎞⎫
1 ⎬
⎜ .. ⎟ s
⎜ .
⎜
⎜ 1
0 ⎟⎭
⎟
⎟
XI ∼ ⎜
⎜
⎟ .
⎟
⎜ 2 ⎟
⎜ ⎟
⎝ 0 ..
.
2
⎠

By combining Lemma 20 and Proposition 21, we can obtain the following theo-
rem.

Theorem 22 ([16, Theorem 2.1]) Let P ⊂ Rd be a full-dimensional lattice polytope

all of whose vertices belong to

{0} ∪ {ei : 1 ≤ i ≤ d} ∪ {ei + e j : 1 ≤ i ≤ j ≤ d}.

Then (P) is reflexive.

The class of lattice polytopes which satisfy the condition of Theorem 22 contains
a well-known large family of (0, 1)-polytopes, namely the edge polytopes of finite
simple graphs. Let G be a simple graph on [d]. The edge polytope PG ⊂ Rd of G
is the convex hull of all vectors ei + e j such that {i, j} ∈ E(G). This means that
the edge polytope of PG of G is the convex hull of all row vectors of the incidence
matrix A G of G, where A G is the matrix in {0, 1} E(G)×[d] with
362 A. Tsuchiya

1 if v ∈ e,
ae,v =
0 otherwise.

Moreover, the dimension of PG equals rank(A G ) − 1. Hence edge polytopes are

not full-dimensional. However, given an edge polytope PG , one can easily get a
full-dimensional unimodularly equivalent copy P of PG by considering the lattice
polytope defined as the convex hull of the row vectors of A G with some columns
deleted. This implies that every edge polytope PG is unimodularly equivalent to a
lattice polytope which satisfies the condition of Theorem 22. In particular, (PG ) is
unimodularly equivalent to (P). Hence we can get the following corollary.

Corollary 23 Let G be a finite simple graph on [d]. Then one has rdim(PG ) =
dim(PG ) + 1.

In general, an edge polytope does not possess the integer decomposition property.
In fact, it is known when the edge polytope of a connected finite simple graph
possesses the integer decomposition property.

Proposition 24 ([18, Corollary 2.3]) Let G be a connected finite simple graph

on [d]. Then PG possesses the integer decomposition property if and only if for any
two odd cycles C and C  of G having no common vertex, there exists an edge of G
joining a vertex of C with a vertex of C  .

However, even if the edge polytope PG of a connected finite simple graph G

possesses the integer decomposition property, the reflexive polytope (PG ) does
not always possess the integer decomposition property.

Example 25 Let G be the connected simple graph as follows:

G:

Then G satisfies the condition of Proposition 24. Hence PG possesses the integer
decomposition property. However, (PG ) does not possess the integer decomposi-
tion property.

Finally, we characterize when (PG ) possesses the integer decomposition prop-

erty for a connected finite simple graph G.

Theorem 26 ([16, Theorem 3.2]) Let G be a connected finite simple graph on [d].
Then (PG ) possesses the integer decomposition property if and only if G does not
contain two disjoint odd cycles.
17 The Reflexive Dimension of (0, 1)-Polytopes 363

Remark 27 We remark that the condition of Theorem 26 characterizes some class

of edge polytopes. In fact, a connected finite simple graph G does not contain two
disjoint odd cycles if and only if PG is unimodular, that is, all its triangulations are
unimodular [17, Example 3.6 b].

Remark 28 In [20], the lattice polytope (PG ) for a connected finite simple graph G
is studied. In fact, if G does not contain two disjoint odd cycles, then (PG ) is a
reflexive polytope which possesses the integer decomposition property. However,
in general, (PG ) is not always reflexive. For instance, if G is the connected finite
simple graph which appears in Example 25, then (PG ) is not reflexive and does not
possess the integer decomposition property.

Acknowledgements The author is very grateful to the anonymous referee for his or her insightful
report that led to significant improvements of the form of the paper. This manuscript was pre-
pared as a contribution to the conference proceedings of the Interactions with Lattice Polytopes at
Otto-von-Guericke-Universität Magdeburg on September 14th – 16th, 2017. The author thanks the
organizers, Christopher Borger, Alexander Kasprzyk, Benjamin Nill and Johannes Hofscheier, for
their support during this conference. The author was partially supported by Grant-in-Aid for JSPS
Fellows 16J01549.

References

1. Batyrev, V.V.: Dual polyhedra and mirror symmetry for Calabi-Yau hypersurfaces in toric
varieties. J. Algebraic Geom. 3(3), 493–535 (1994)
2. Chudnovsky, M., Robertson, N., Seymour, P., Thomas, R.: The strong perfect graph theorem.
Ann. of Math. (2) 164(1), 51–229 (2006)
3. Cox, D.A., Little, J.B., Schenck, H.K.: Toric varieties, Graduate Studies in Mathematics, vol.
124. American Mathematical Society, Providence (2011)
4. Haase, C., Melnikov, I.V.: The reflexive dimension of a lattice polytope. Ann. Comb. 10(2),
211–217 (2006)
5. Hibi, T.: A quick introduction to Gröbner bases. In: Gröbner bases, pp. 1–54. Springer, Tokyo
(2013)
6. Hibi, T., Li, N., Li, T.X., Mu, L.L., Tsuchiya, A.: Order-chain polytopes. Ars Math. Contemp.
16(2), 299–317 (2019)
7. Hibi, T., Matsuda, K.: Quadratic Gröbner bases of twinned order polytopes. Eur. J. Combin.
54, 187–192 (2016)
8. Hibi, T., Matsuda, K., Ohsugi, H., Shibata, K.: Centrally symmetric configurations of order
polytopes. J. Algebra 443, 469–478 (2015)
9. Hibi, T., Matsuda, K., Tsuchiya, A.: Gorenstein Fano polytopes arising from order polytopes
and chain polytopes (2015). arXiv:1507.03221 [math.CO]
10. Hibi, T., Matsuda, K., Tsuchiya, A.: Quadratic Gröbner bases arising from partially ordered
sets. Math. Scand. 121(1), 19–25 (2017)
11. Hibi, T., Tsuchiya, A.: Facets and volume of Gorenstein Fano polytopes. Math. Nachr. 290(16),
2619–2628 (2017)
12. Hibi, T., Tsuchiya, A.: Reflexive polytopes arising from perfect graphs. J. Combin. Theory Ser.
A 157, 233–246 (2018)
13. Hibi, T., Tsuchiya, A.: Reflexive polytopes arising from partially ordered sets and perfect
graphs. J. Algebraic Combin. 49(1), 69–81 (2019)
364 A. Tsuchiya

14. Kreuzer, M., Skarke, H.: Complete classification of reflexive polyhedra in four dimensions.
Adv. Theor. Math. Phys. 4(6), 1209–1230 (2000)
15. Lagarias, J.C., Ziegler, G.M.: Bounds for lattice polytopes containing a fixed number of interior
points in a sublattice. Canad. J. Math. 43(5), 1022–1035 (1991)
16. Nagaoka, T., Tsuchiya, A.: Reflexive polytopes arising from edge polytopes. Linear Algebra
Appl. 557, 438–454 (2018)
17. Ohsugi, H., Herzog, J., Hibi, T.: Combinatorial pure subrings. Osaka J. Math. 37(3), 745–757
(2000)
18. Ohsugi, H., Hibi, T.: Normal polytopes arising from finite graphs. J. Algebra 207(2), 409–426
(1998)
19. Ohsugi, H., Hibi, T.: Convex polytopes all of whose reverse lexicographic initial ideals are
squarefree. Proc. Amer. Math. Soc. 129(9), 2541–2546 (2001)
20. Ohsugi, H., Hibi, T.: Centrally symmetric configurations of integer matrices. Nagoya Math. J.
216, 153–170 (2014)
21. Ohsugi, H., Hibi, T.: Reverse lexicographic squarefree initial ideals and Gorenstein Fano poly-
topes. J. Commut. Algebra 10(2), 171–186 (2018)
22. Ohsugi, H., Tsuchiya, A.: Enriched chain polytopes. Israel J. Math. 237(1), 485–500 (2020)
23. Ohsugi, H., Tsuchiya, A.: Reflexive polytopes arising from bipartite graphs with γ -positivity
associated to interior polynomials. Selecta Math. (N.S.) 26(4), Paper No. 59, 22 (2020)
24. Ohsugi, H., Tsuchiya, A.: Enriched order polytopes and enriched Hibi rings. Eur. J. Math. 7(1),
48–68 (2021)
25. Schrijver, A.: Theory of linear and integer programming. Wiley-Interscience Series in Discrete
Mathematics. Wiley, Chichester (1986). A Wiley-Interscience Publication
26. Stanley, R.P.: Two poset polytopes. Discrete Comput. Geom. 1(1), 9–23 (1986)
27. Sturmfels, B.: Gröbner bases and convex polytopes. University Lecture Series, vol. 8. American
Mathematical Society, Providence (1996)
28. Sullivant, S.: Compressed polytopes and statistical disclosure limitation. Tohoku Math. J. (2)
58(3), 433–445 (2006)
29. Tsuchiya, A.: Volume, facets and dual polytopes of twinned chain polytopes. Ann. Comb.
22(4), 875–884 (2018)