Polytopes and posets associated to preorders
Extracted source text · arXiv 2605.26916
\documentclass[11pt]{amsart}
\usepackage[utf8]{inputenc}
\usepackage{amsmath,amssymb,amsthm}
\usepackage[colorlinks,citecolor=blue]{hyperref}
\usepackage{cleveref}
% \usepackage{ntheorem}
\usepackage{enumitem}
\usepackage{booktabs}
\usepackage{mathrsfs}
\usepackage{graphicx}
\usepackage{tikz}
%
\usetikzlibrary{patterns, shapes, arrows}
\usetikzlibrary{trees}
\usetikzlibrary{cd, graphs.standard}
%
\textwidth=16cm
\textheight=21cm
\hoffset=-55pt
\footskip=25pt
%
\newcommand{\ZZ}{\mathbb{Z}}
\newcommand{\NN}{\mathbb{N}}
\newcommand{\CC}{\mathbb{C}}
\newcommand{\QQ}{\mathbb{Q}}
\newcommand{\RR}{\mathbb{R}}
\newcommand{\TA}{\mathbb{A}}
\newcommand{\TB}{\mathbb{B}}
\newcommand{\TC}{\mathbb{C}}
\newcommand{\TD}{\mathbb{D}}
\newcommand{\TE}{\mathbb{E}}
\newcommand{\TF}{\mathbb{F}}
\newcommand{\TyH}{\mathbb{H}}
\newcommand{\TI}{\mathbb{I}}
\newcommand{\TG}{\mathbb{G}}
\newcommand{\aA}{{\mathcal A}}
\newcommand{\eE}{{\mathcal E}}
\newcommand{\gG}{{\mathcal G}}
\newcommand{\iI}{{\mathcal I}}
\newcommand{\jJ}{{\mathcal J}}
\newcommand{\pP}{{\mathcal P}}
\newcommand{\qQ}{{\mathcal Q}}
\newcommand{\rR}{{\mathcal R}}
\newcommand{\uU}{{\mathcal U}}
\newcommand{\zZ}{{\mathcal Z}}
\newcommand{\wW}{{\mathcal W}}
\newcommand{\asc}{{\rm asc}}
\newcommand{\des}{{\rm des}}
\newcommand{\Ehr}{{\rm Ehr}}
\newcommand{\rev}{{\rm rev}}
\newcommand{\tw}{\tau_{\mathsf{w}}}
\newcommand{\ov}[1]{\overline{#1}}
\newcommand{\bu}{\overline{u}}
\newcommand{\bx}{\overline{x}}
\newcommand{\bDelta}{\boldsymbol\Delta}
\newcommand{\mX}{\mathsf{X}}
\newcommand{\mY}{\mathsf{Y}}
\newtheorem{theorem}{Theorem}[section]
\newtheorem{proposition}[theorem]{Proposition}
\newtheorem{conjecture}[theorem]{Conjecture}
\newtheorem{corollary}[theorem]{Corollary}
\newtheorem{lemma}[theorem]{Lemma}
\newtheorem{definition}{Definition}
\newtheorem{question}[theorem]{Question}
\newtheorem{remark}[theorem]{Remark}
\newtheorem{example}[theorem]{Example}
\newtheorem{problem}[theorem]{Problem}
\newtheorem{algorithm}[theorem]{Algorithm}
\newcommand{\po}{\mathscr{P}}
\newcommand{\op}{\operatorname{op}}
\newcommand{\eps}{\varepsilon}
\newcommand{\oeis}[1]{\href{https://oeis.org/#1}{{#1}}}
\newcommand*\pFq[4]{{}_{#1}F_{#2}\biggl(\genfrac..{0pt}{}{#3}{#4};1\biggr)}
\newcommand{\defi}[1]{\textit{\color{red}{#1}}}
\newcommand{\tp}{\operatorname{top}}
\newcommand{\ztop}{Z^{\operatorname{top}}}
\newcommand{\down}{\mathscr{D}}
\newcommand{\up}{\mathscr{U}}
\newcommand{\Zeta}{\mathsf{Zeta}}
\newcommand{\tM}{\mathbb{M}}
\newcommand{\haut}{\operatorname{ht}}
\newcommand{\trunc}{\mathsf{T}}
\newcommand{\lap}{\mathcal{L}}
\newcommand{\rt}{r}
\newcommand{\cov}{\operatorname{cov}}
\newcommand{\nz}{\operatorname{nz}}
\newcommand{\NC}{\mathsf{NC}}
\newcommand{\posi}{\operatorname{posi}}
\newcommand{\nega}{\operatorname{nega}}
\newcommand{\Max}{\operatorname{Max}}
\newcommand{\pell}{\mathcal{P}}
\newcommand{\re}{\operatorname{Re}}
\newcommand{\nara}{\operatorname{Nar}}
\newcommand{\biE}{\mathbb{E}}
% \definecolor{racines}{gray}{0.8}
% \definecolor{racines}{RGB}{187,222,251}
\definecolor{racines}{RGB}{197,225,165}
%
\begin{document}
\title[Polytopes and posets associated to preorders]
{Polytopes and posets associated to preorders}
\author{Christos~A.~Athanasiadis}
\address{Department of Mathematics\\
National and Kapodistrian University of Athens\\
Panepistimioupolis\\ 15784 Athens, Greece}
\email{caath@math.uoa.gr}
\author{Frédéric~Chapoton}
\address{Institut de Recherche Mathématique
Avancée, UMR 7501
\\ Université de Strasbourg et CNRS\\
7 rue René-Descartes\\
67000 Strasbourg, France}
\email{chapoton@unistra.fr}
\date{\today}
\thanks{ \textit{Mathematics Subject
Classifications}: 52B20, 05A15, 06A07}
\thanks{ \textit{Key words and phrases}.
Preorder, lattice polytope, generalized
permutohedron, Ehrhart polynomial,
$h^\ast$-polynomial, zeta polynomial, reflexive polytope.
}
\begin{abstract}
Preorder polytopes, defined from preorders on
finite sets, are introduced and studied from a
lattice point enumeration point of view. They
naturally generalize arbor polytopes, recently
introduced and studied by the second named
author. Preorder
polytopes are shown to be lattice polytopes
which satisfy a certain duality relating their
Ehrhart polynomials with the zeta polynomials
of their posets of lattice points. A combinatorial
interpretation of the normalized volume of a
preorder polytope is proven, together with
formulas for the Ehrhart polynomial and
the $h^\ast$-polynomial, and a combinatorial
interpretation of the latter is conjectured.
Several conjectures and results on the lattice
point enumeration of arbor polytopes are
generalized to preorder polytopes, new
conjectures are proposed and new interesting
examples of preorder polytopes are studied.
\end{abstract}
\maketitle
\section{Introduction}
\label{sec:intro}
Arbor polytopes are lattice polytopes
introduced and studied in~\cite{Cha25b}. This
paper studies a natural generalization. Let us
first recall the definition and main properties
of arbor polytopes. Given a finite set $E$, an
\emph{arbor} on $E$ is a rooted tree with
vertices the blocks of a partition of $E$. The
\emph{arbor polytope} associated to such an arbor
$\sigma$ is the polytope $\qQ_\sigma$ in $\RR^E$
defined by the linear inequalities $x_e \ge 0$
for $e \in E$ and
%
\begin{equation} \label{eq:arbor}
\sum_{e \in \down(v)} x_e \le |\down(v)|
\end{equation}
%
for every vertex $v$ of $\sigma$, where
$\down(v)$ is the union of all vertices which
are descendants of $v$ in $\sigma$ (including
$v$ itself). The set of lattice points of
$\qQ_\sigma$, denoted by $P_\sigma$ and
partially ordered as a subposet of the product
order $\NN^E$, also plays a central role in
\cite{Cha25b}. It was shown there that
$\qQ_\sigma$ is a simple lattice polytope with
very interesting properties from a lattice
point enumeration point of view.
Let us present two conjectures and one result
about arbor polytopes which partly motivated
the present work (see Section~\ref{sec:pre}
for relevant background and any undefined
terminology). Let us denote by $h_i(\sigma)$
the number of lattice points of $\qQ_\sigma$
having $i$ nonzero coordinates for $0 \le i
\le n$ and set $h(\sigma,t) = \sum_{i=0}^n h_i
(\sigma)t^i$, where $n$ is the cardinality of
the ground set $E$. Following \cite{Cha25b},
we call $h(\sigma,t)$ the \emph{$h$-polynomial}
of $\sigma$.
%
\begin{conjecture} \label{conj:arbor-h}
{\rm (\cite[Conjecture~1.1]{Cha25b})}
The polynomial $h(\sigma,t)$ is equal to the
$h$-polynomial of an $n$-dimensional simplicial
polytope for every arbor $\sigma$ on an
$n$-element set $E$. In particular, $h(\sigma,t)$
is palindromic and unimodal.
\end{conjecture}
%
\begin{conjecture}
{\rm (\cite[Conjecture~3.2]{Cha25b})}
\label{conj:arbor-roots}
For every arbor $\sigma$, all roots of the
Ehrhart polynomial of $\qQ_\sigma$ are real and
lie in the interval $[-1,0]$. In particular, the
$h^\ast$-polynomial of $\qQ_\sigma$ has only real
roots.
\end{conjecture}
An arbor $\sigma$ is said to be \emph{linear} if
the associated rooted tree is a path. The reverse
arbor $\rev(\sigma)$ is then defined by reversing
the order of the vertices of this path. The
following statement was conjectured in
\cite[Section~4]{Cha25b}.
%
\begin{theorem} \label{thm:Ath26}
{\rm (\cite[Theorem~1.3]{Ath26})}
The Ehrhart polynomial of $\qQ_\sigma$ and the
zeta polynomial of $P_{\rev(\sigma)}$ are related
by the identity
%
\[ \Ehr(\qQ_\sigma, t) =
\zZ(P_{\rev(\sigma)}, t+1) \]
%
for every linear arbor $\sigma$.
\end{theorem}
\begin{figure}[!h]
\centering
\begin{tikzpicture}
\tikzstyle{every node}=[draw,shape=circle, thick]
\node at (0,0) (ac) {$a,c$};
\node at (2,0) (b) {$b$};
\node at (0,2) (e) {$e$};
\node at (2,2) (d) {$d$};
\path [draw] (ac) -- (e);
\path [draw] (b) -- (e);
\path [draw] (b) -- (d);
\end{tikzpicture}
\caption{The Hasse diagram of a preorder on the
set $\{a,b,c,d,e\}$.}
\label{fig1}
\end{figure}
This paper generalizes arbor polytopes by
replacing arbors on finite sets by arbitrary
preorders. A \emph{preorder} on a finite set
$E$ can be viewed as a partial order on the
set of blocks of a partition of $E$, called
vertices (see Section~\ref{sec:preorders}).
The example shown in Figure~\ref{fig1} will
serve as a running example. Given a preorder
$\tau$ on a finite set $E$, the \emph{preorder
polytope} associated to $\tau$ is the polytope
$\qQ_\tau$ in $\RR^E$ defined by the linear
inequalities $x_e \ge 0$ for $e \in E$ and
%
\begin{equation} \label{eq:Q-tau-ineq-b}
\sum_{e \in \iI} x_e \le |\iI|
\end{equation}
%
for all order ideals $\iI$ of $\tau$. This
definition recovers and extends the definition
of arbor polytopes. Indeed, every arbor $\sigma$
can be considered as a preorder $\tau$ having
the root of $\sigma$ as the maximum vertex.
Then, for every vertex $v$ of $\sigma$ the set
$\down(v)$ is a principal order ideal of $\tau$
and inequalities of the form
\eqref{eq:Q-tau-ineq-b} associated to
nonprincipal order ideals of $\tau$ are
consequences (namely sums) of inequalities
associated to principal order ideals.
It is not obvious that preorder polytopes are
lattice polytopes and no longer true that they
are always simple. Our first main result shows
that preorder polytopes are indeed lattice
polytopes which form a subclass of the class of
generalized permutohedra \cite{post09}.
%
\begin{theorem} \label{thm:main-a}
Let $\tau$ be a preorder on an $n$-element set.
Then, $\qQ_\tau$ is an $n$-dimensional lattice
polytope which is unimodularly equivalent to a
generalized permutohedron.
\end{theorem}
An advantage of considering preorder polytopes,
compared to the more restrictive class of arbor
polytopes, is that there is a natural notion of
duality for preorders. This notion allows for a
generalization of Theorem~\ref{thm:Ath26} which
we call \emph{Ehrhart-Zeta duality}. Another
obvious advantage is that the new setting allows
for more interesting examples. We denote by
$\tau^\ast$ the dual of the preorder $\tau$ and
define the poset $P_\tau$ of lattice points of
$\qQ_\tau$ as we did for arbors. The following
statement is the second main result of this
paper.
%
\begin{theorem} \label{thm:main-b}
Let $\tau$ be a preorder on an $n$-element set
$E$.
%
\begin{itemize}
\itemsep=0pt
\item[{\rm (a)}]
The Ehrhart polynomial of $\qQ_\tau$ is given by
the formula
%
\[ \Ehr(\qQ_\tau, t) = \sum_{(a_e) \in
P_{\tau^\ast}}
\prod_{e \in E} \binom{t+a_e-1}{a_e}, \]
%
where the sum ranges over all lattice points
$(a_e)$ of $\qQ_{\tau^\ast}$. In particular, the
number of lattice points of $\qQ_\tau$ is equal to
that of $\qQ_{\tau^\ast}$.
The normalized volume of $\qQ_\tau$ is equal to the
number of words $w \in E^n$ such that for every order
ideal $\iI$ of $\tau$, the elements of $\iI$ appear a
total of at least $|\iI|$ times in $w$.
\item[{\rm (b)}]
The Ehrhart polynomial of $\qQ_\tau$ and the zeta
polynomial of $P_{\tau^\ast}$ are related by the identity
%
\[ \Ehr(\qQ_\tau, t) = \zZ(P_{\tau^\ast}, t+1). \]
%
\end{itemize}
\end{theorem}
The proofs of both results rely on the theory
of generalized permutohedra \cite{post09}. An
interesting feature of our study of preorder
polytopes is the variety of conjectures on their
lattice point enumeration that can be stated,
some of which extend
Conjecture~\ref{conj:arbor-h} and (partly)
Conjecture~\ref{conj:arbor-roots} on arbor
polytopes.
The content and structure of this paper can be
summarized as follows. Section~\ref{sec:pre}
discusses notation and reviews basic
definitions and background on poset and preposet
combinatorics, Ehrhart theory of lattice
polytopes, generalized permutohedra and related
topics. Section~\ref{sec:basic} establishes
basic properties of preorder polytopes and
their posets of lattice points.
Theorems~\ref{thm:main-a} and~\ref{thm:main-b}
are proven in Sections~\ref{sec:basic}
and~\ref{sec:Ehrhart}, respectively, by
applying results from the theory of generalized
permutohedra \cite{post09}.
Section~\ref{sec:Ehrhart} also conjectures
a stronger form of Ehrhart-Zeta duality, as well
as a combinatorial interpretation of the
$h^\ast$-polynomial of any preorder polytope
which refines that of the normalized volume
given by Theorem~\ref{thm:main-b} (see
Conjectures~\ref{conj:matrixEZ}
and~\ref{conj:h-ast}) and provides partial
results and a formula for the
$h^\ast$-polynomial. Ehrhart polynomials of
preorder polytopes may have nonreal complex
roots. However, a weaker property (called
magic positivity) which implies the
real-rootedness of the $h^\ast$-polynomial
is conjectured in Section~\ref{sec:Ehrhart}
(see Conjecture~\ref{conj:magic}).
Section~\ref{sec:h-vectors} discusses the
natural extension of
Conjecture~\ref{conj:arbor-h} for preorder
polytopes and conjectures the
$\gamma$-positivity and real-rootedness of
$h$-polynomials of preorders, as well as
their invariance under duality.
Section~\ref{sec:reflexive} studies reflexive
polytopes which can be obtained as
Minkowski sums of preorder polytopes and
standard simplices. Section~\ref{sec:double}
provides a formula for the double Ehrhart
polynomial of a preorder polytope and the
standard simplex and conjectures an identity
(see Conjecture~\ref{conj:double}) which is
stronger than Conjecture~\ref{conj:h-ast}.
Section~\ref{sec:M-triangles} briefly
discusses an important enumerative invariant
of the poset of lattice points of a preorder
polytope, namely the $M$-triangle.
Section~\ref{sec:exa} discusses some
interesting examples and confirms some of
the conjectures of this paper for them.
\section{Preliminaries}
\label{sec:pre}
This section establishes notation to be followed
throughout the paper and reviews definitions and
background on the combinatorics of partial orders
and preorders, lattice polytopes and generalized
permutohedra. More information about these topics
can be found in \cite{HiAC,StaCCA,StaEC1,post09}.
\subsection{Notation}
\label{sec:notation}
Let $E$ be a nonempty finite set. We will denote
by ${({\rm \mathbf{e}}_e)}_{e \in E}$ the standard
basis of the space $\RR^E$. We will use boldface
letters to denote vectors $\mathbf{x} =
(x_e)_{e \in E}$ of $\RR^E$, or collections of
real numbers. We will denote by $|\mathbf{x}|$
the sum of the coordinates of a point $\mathbf{x}
\in \RR^E$ and by $|S|$ or $\# S$ the cardinality
of a finite set $S$.
\subsection{Preorders}
\label{sec:preorders}
Let $E$ be a nonempty finite set. A \emph{preorder}
on $E$ is defined as a reflexive and transitive
relation $\le$ on $E$. Declaring two elements $a, b
\in E$ to be equivalent if $a \le b$ and $b \le a$,
one gets an equivalence relation on $E$. The induced
order on the set of equivalence classes, which we
call \emph{vertices}, is a partial order denoted by
$\preceq$. Thus, a preorder $\le$ on the ground set
$E$ can also be defined as a partial order $\preceq$
on the set of blocks of a partition of $E$. We will
depict $\le$ by the Hasse diagram
\cite[Section~3.1]{StaEC1} of the associated partial
order $\preceq$ (where, as usual, the order increases
from bottom to top). In the language of
species~\cite{BLL98}, we are working with the
composition of the species of partial orders with
that of nonempty finite sets. Preorders on a finite
set $E$ are in one-to-one correspondence with
topologies on $E$ \cite[Exercise~3.3]{StaEC1}. We
will usually denote preordered finite sets by small
greek letters. We will refer to the cardinality of
the ground set $E$ as the \emph{size} of a preorder
on $E$ and to the cardinality of a vertex as the
\emph{size} of this vertex. Figure~\ref{fig1} shows
the Hasse diagram of a preorder of size 5; there is
one vertex of size two and three single element
vertices.
Basic constructions for posets
\cite[Sections~3.1-3.2]{StaEC1}, such as
disjoint unions, ordinal sums, duality, order
ideals and filters and principal order ideals and
filters, extend naturally to preorders. For example,
an \emph{order ideal} of a preorder on the ground
set $E$ is defined as any subset $\iI \subseteq E$
such that for every $b \in \iI$ we have $a \in \iI$
for every $a \in E$ with $a \le b$. Thus, order
ideals are unions of vertices which are ideals of
the associated partial order $\preceq$. Similarly,
an \emph{order filter} (or \emph{dual order ideal})
is defined as any subset $\jJ \subseteq E$ such
that for every $a \in \jJ$ we have $b \in \jJ$ for
every $b \in E$ with $a \le b$. We will denote by
$\down(e)$ (respectively, $\up(e)$) the principal
order ideal (respectively, principal order filter)
generated by $e \in E$. We will sometimes abuse notation
and write $\down(v)$ and $\up(v)$ for a vertex $v$.
A preorder $\le$ on $E$ is called
a \emph{chain} if $a \le b$ or $b \le a$ for all
$a, b \in E$ (equivalently, the induced partial
order $\preceq$ is a total order). Chains are
preorders associated to linear arbors
\cite[Section~4]{Cha25b}; the corresponding
arbor polytopes were studied in~\cite{Ath26}
and called \emph{composition polytopes}.
\subsection{Poset combinatorics}
\label{sec:poset-comb}
Let $(P, \preceq)$ be a finite poset and let
$d$ be the length (one less than the cardinality)
of the longest chain in $P$. The \emph{zeta
polynomial} \cite[Section~3.12]{StaEC1} of $P$
is the unique polynomial $\zZ(P, t)$
with the property that $\zZ(P, m+1)$ is equal
to the number of $m$-element multichains in $P$,
meaning sequences $(p_1, p_2,\dots,p_m)$ of
elements of $P$ such that $p_1 \preceq p_2
\preceq \cdots \preceq p_m$, for every $m \in
\NN$. The degree of $\zZ(P, t)$ is equal to $d$
and the leading coefficient is equal to the
number of (necessarily maximal) chains of $P$
of length $d$, divided by $d!$. We say that
$P$ is \emph{graded} if all maximal chains in
$P$ have length $d$.
\subsection{Ehrhart theory}
\label{sec:Ehr-hstar}
Let $\qQ$ be a $d$-dimensional lattice polytope
in $\RR^n$ (thus, every vertex of $\qQ$ belongs
to $\ZZ^n$). The \emph{Ehrhart polynomial} of
$\qQ$ is the unique polynomial $\Ehr(\qQ, t)$
with the property that $\Ehr(\qQ, m) = \#
(m\qQ \cap \NN^n)$ for every $m \in \NN$. The
Ehrhart polynomial $\Ehr(\qQ, t)$ has degree
$d$ and thus
%
\begin{equation} \label{eq:def-hstar}
\sum_{m \ge 0} \Ehr(\qQ, m) t^m =
\frac{h^\ast(\qQ, t)}{(1-t)^{d+1}}
\end{equation}
%
for some polynomial $h^\ast(\qQ, t)$ of degree at
most $d$, called the \emph{$h^\ast$-polynomial}.
The coefficients of $h^\ast(\qQ, t)$ are in fact
nonnegative integers \cite[Theorem~33.9]{HiAC}
\cite[Section~I.5]{StaCCA}. The sequence
$h^\ast(\qQ) = (h^\ast_0(\qQ),
h^\ast_1(\qQ),\dots,h^\ast_d(\qQ))$ of
coefficients of $h^\ast(\qQ, t) = \sum_{i=0}^d
h^\ast_i(\qQ) t^i$ is the \emph{$h^\ast$-vector}
of $\qQ$. The leading coefficient of $\Ehr (\qQ,
t)$ is equal to the $d$-dimensional volume of
$\qQ$. Multiplied by $d!$, this number is a
positive integer called the \emph{normalized
volume}; it is equal to the sum of the entries
of the $h^\ast$-vector. Thus, the normalized
volume is refined by the $h^\ast$-vector.
The polytope $\qQ$ is said to be \emph{Ehrhart
positive} if all coefficients of $\Ehr(\qQ, t)$
are nonnegative. Finding classes of Ehrhart
positive lattice polytopes is a very
interesting problem \cite{FH24, Liu19}.
Following~\cite{FH24}, we say that $\Ehr(\qQ,
t)$ is \emph{magic positive} if
%
\[ \Ehr(\qQ, t) = \sum_{i=0}^d c_i
t^i(1+t)^{d-i} \]
%
for some nonnegative real numbers $c_0,
c_1,\dots,c_d$. By a theorem of Br\"and\'en
\cite[Theorem~4.2]{Bra06}, this property (which
is stronger than Ehrhart positivity) implies
that $h^\ast(\qQ, t)$ has only real roots.
\subsection{Generalized permutohedra}
\label{sec:genperm}
We briefly recall, with slightly adapted notation,
the definitions and some basic results from
Postnikov's theory of generalized permutohedra, as
introduced in~\cite{post09}. This involves two
distinct notions, namely generalized permutohedra
and the subclass of $Y$-type generalized
permutohedra.
Let $E$ be a nonempty finite set. For a nonempty
subset $I$ of $E$, the standard simplex $\Delta_I$
is the convex hull of the set of standard basis
vectors
$\{ {\rm \mathbf{e}}_e : e \in I\}$. $Y$-type
generalized permutohedra are defined as Minkowski
sums of scaled standard simplices. More precisely,
given a collection $\mathbf{y} = {\{y_I\}}_I$ of
nonnegative real numbers indexed by nonempty
subsets of $E$,
%
\begin{equation} \label{eq:Py}
\pP^y(\mathbf{y}) = \sum_{\varnothing \not=
I \subseteq E} y_I \Delta_I
\end{equation}
%
is the \emph{$Y$-type generalized permutohedron}
associated to $\mathbf{y}$.
On the other hand, generalized permutohedra are
defined by linear equations and inequalities as
follows. Given a collection $\mathbf{z} =
{\{z_I\}}_I$ of nonnegative real numbers indexed
by nonempty subsets of $E$,
%
\begin{equation} \label{eq:Pz}
\pP^z(\mathbf{z}) = \bigg\{(x_e) \in \RR^E
\biggm| \sum_{e \in E} x_e = z_{E}, \quad
\sum_{i \in I} x_i \ge z_I \quad \forall I \not=
\varnothing \bigg\}
\end{equation}
%
is the \emph{generalized permutohedron} associated
to $\mathbf{z}$. By~\cite[Proposition~6.3]{post09},
every $Y$-type generalized permutohedron is a
generalized permutohedron. More precisely,
defining the collection $\mathbf{z}$ from the
collection $\mathbf{y}$ by setting
%
\begin{equation} \label{eq:zy}
z_J = \sum_{\varnothing \not= I \subseteq J} y_I
\end{equation}
%
for nonempty $J \subseteq E$, there is an equality
of polytopes: $\pP^y(\mathbf{y}) = \pP^z
(\mathbf{z})$.
We now present a basic formula from~\cite{post09}
for the Ehrhart polynomial of an integral $Y$-type
generalized permutohedron. Let us consider nonempty
subsets $I_0, I_1,\dots,I_m$ of $E$ and numbers
$y_0, y_1,\dots,y_m \in \NN$. It is convenient to
consider the bipartite graph $G$ with $m+1$
vertices $\ell_0, \ell_1,\dots,\ell_m$ on the left
side and vertices $r_e$ for $e \in E$ on the right
side for which $\ell_j$ is adjacent to $r_e$ if
and only if $e \in I_j$. A sequence
$(a_0, a_1,\dots,a_m) \in \NN^{m+1}$ is called
\emph{$G$-draconian} \cite[Definition~9.2]{post09}
if $a_0 + a_1 + \cdots + a_m = |E| - 1$ and
%
\begin{equation} \label{eq:draco2}
1 + \sum_{\alpha \in \aA} a_\alpha \le \#
\bigcup_{\alpha \in \aA} I_\alpha
\end{equation}
%
for every nonempty $\aA \subseteq \{0, 1,\dots,m\}$.
The following statement is a consequence of
\cite[Theorem~11.3]{post09}.
%
\begin{theorem} [\cite{post09}] \label{thm:post}
Assume that the bipartite graph $G$ just
described has no isolated vertices and that $I_0 =
E$. Then, for all nonnegative integers $y_j$, the
Ehrhart polynomial of the $Y$-type generalized
permutohedron
%
\[ \qQ = \sum_{j=0}^m y_j \Delta_{I_j} \]
%
in $\RR_{\ge 0}^E$ is given, by the formula
%
\[ \Ehr(\qQ, t) = \sum_{(a_0, a_1,\dots,a_m) \in
\gG_G} \binom{y_0 t + a_0}{a_0}
\prod_{i=1}^m \binom{y_i t + a_i - 1}{a_i},
\]
%
where $\gG_G$ stands for the set of $G$-draconian
sequences. In particular, $\qQ$ is Ehrhart
positive.
\end{theorem}
\section{Basic properties}
\label{sec:basic}
This section establishes basic properties of
preorder polytopes and their posets of lattice
points and proves Theorem~\ref{thm:main-a}.
Throughout this and the following sections
(unless stated otherwise), $\tau$ will be a
preorder of size $n$ on the ground set $E$; we
assume that $E$ does not contain 0.
\subsection{Elementary properties}
\label{sec:elementary}
We recall that $\qQ_\tau$ is defined by $x_e
\ge 0$ for every $e \in E$ and the linear
inequalities~(\ref{eq:Q-tau-ineq-b}) for all
order ideals $\iI$ of $\tau$, where $(x_e)$ are
the standard coordinates of $\RR^E$. For our
running example from Figure~\ref{fig1}, these
inequalities are
%
\begin{align*}
x_a + x_c & \le 2 \\
x_b & \le 1 \\
x_a + x_b + x_c + x_e & \le 4 \\
x_b + x_d & \le 2 \\
x_a + x_b + x_c + x_d + x_e & \le 5.
\end{align*}
%
Since $E$ is an order ideal of $\tau$, one of
the linear inequalities defining $\qQ_\tau$ is
%
\[ \sum_{e \in E} x_e \le n \]
%
and hence $\qQ_\tau$ is indeed a bounded
polyhedron in $\RR^E$. Clearly, one may disregard
the inequalities~(\ref{eq:Q-tau-ineq-b})
associated to order ideals which are disconnected
as subgraphs of the Hasse diagram of $\tau$,
since these are implied by the ones associated
to connected order ideals.
\begin{remark} \label{rem:not-simple} \rm
(a) Contrary to the situation with arbor
polytopes~\cite{Cha25b}, preorder polytopes are
not necessarily simple. This property fails
already for the three-element poset $\{a,b,c\}$
with relations $a \prec b$ and $a \prec c$. The
preorder polytope for Figure~\ref{fig1} is not
simple either; its $f$-vector is $(1, 27, 69, 72,
38, 10, 1)$.
(b) The polytope $Q_\tau$ is an anti-blocking polytope,
in the sense of~\cite{fulkerson}; see also
\cite[Section~9.3]{schrijver}. This means that
$Q_\tau \subset \RR^E_{\ge 0}$ and $\mathbf{y}
\in \qQ_\tau \Rightarrow \mathbf{x} \in
\qQ_\tau$ whenever $\mathbf{x}, \mathbf{y} \in
\RR^E_{\ge 0}$ with $\mathbf{x} \le \mathbf{y}$
coordinatewise.
\qed
\end{remark}
\subsection{Preorder polytopes as generalized
permutohedra}
\label{sec:Minkowski}
This section shows that every preorder polytope
is unimodularly equivalent to a generalized
permutohedron and deduces Theorem~\ref{thm:main-a}.
This construction, which is explicit, will be
essential for other results of this paper as well.
We will work with a larger family of polytopes
than that of preorder polytopes, since this will
be convenient in Sections~\ref{sec:reflexive} and~\ref{sec:double}.
For $I \subseteq E$, let us denote by $\bDelta_I$
the lattice simplex in $\RR^E$ with vertices the
basis vectors ${\rm \mathbf{e}}_e$
for $e \in I$ and the zero vector of $\RR^E$. For
a vertex $v$ of $\tau$, let us denote by
$\boldsymbol\uU_v$ the $|v|$th Minkowski power
of the simplex $\bDelta_{\up(v)}$, where $\up(v)$
is the principal order filter of $\tau$ generated
by $v$ (thus, $\up(v)$ is a union of vertices).
Given $r, s \in \NN$, we consider the Minkowski
sum
%
\begin{equation} \label{eq:def-Qrs}
\qQ_\tau(r,s) = r \sum_{v} \boldsymbol\uU_v + s
\bDelta_E
\end{equation}
%
in $\RR^E$, where the sum ranges over all vertices
$v$ of $\tau$.
\begin{proposition} \label{prop:y-gen}
The polytope $\qQ_\tau(r,s)$ is unimodularly
equivalent to the $Y$-type generalized permutohedron
%
\[ r \sum_{v} \uU_v + s \Delta_{E \sqcup \{0\}} \]
%
in $\RR^{E \sqcup \{0\}}$, where the sum ranges
over all vertices $v$ of $\tau$ and $\uU_v$ is the
$|v|$th Minkowski power of the simplex
$\Delta_{\up(v) \sqcup \{0\}}$ in
$\RR^{E \sqcup \{0\}}$.
\end{proposition}
\begin{proof}
The projection $\rho: \RR^{E \sqcup \{0\}} \to
\RR^E$ which forgets the zeroth coordinate induces
a bijection from the hyperplane consisting of all
points in $\RR^{E \sqcup \{0\}}$ with sum of
coordinates equal to 1 onto $\RR^E$ and a
unimodular equivalence between the lattice simplices
$\Delta_{I \sqcup \{0\}} \subset \RR^{E \sqcup \{0\}}$
and $\bDelta_I \subset \RR^E$ for every $I \subseteq
E$. Thus, $\rho$ induces the claimed unimodular
equivalence as well.
\end{proof}
Let us define a collection $\mathbf{y} = \mathbf{y}
(r,s)$ of nonnegative integers, indexed by nonempty
subsets of $E \sqcup \{0\}$, by $y_I = 0$ for every
nonempty $I \subseteq E$ and
%
\begin{equation} \label{eq:def-y(r,s)}
y_{I \sqcup \{0\}} = \begin{cases}
r|v| + s \delta_{I, E},
\text{if} \ I = \up(v) \ \text{for some vertex} \
v \ \text{of} \ \tau, \\
s \delta_{I, E}, \text{otherwise}
\end{cases}
\end{equation}
%
for every $I \subseteq E$, where
%
\[ \delta_{I, E} = \begin{cases}
1, \text{if} \ I = E, \\
0, \text{ otherwise.}
\end{cases} \]
%
We have just shown that $\qQ_\tau(r,s)$ can be
identified with the $Y$-type generalized
permutohedron $\pP^y(\mathbf{y})$. The following
statement describes explicitly $\qQ_\tau(r,s)$
in terms of linear inequalities.
\begin{proposition} \label{prop:Q(r,s)-ineq}
The polytope $\qQ_\tau(r,s)$ is defined by the
inequalities $x_e \ge 0$
%%
%%\begin{equation} \label{eq:Q(r,s)-ineq-a}
%%x_e \ge 0
%%\end{equation}
%%
for every $e \in E$ and
%
\begin{equation} \label{eq:Q(r,s)-ineq-b}
\sum_{e \in \iI} x_e \le r |\iI| + s
\end{equation}
%
for every order ideal $\iI$ of $\tau$.
In particular, $\qQ_\tau(r,s)$ coincides with the
preorder polytope $\qQ_\tau$ for $r=1$ and $s=0$.
\end{proposition}
\begin{proof}
Let $\mathbf{z} = \mathbf{z}(r,s)$ be the
collection of nonnegative integers, indexed
by nonempty subsets of $E \sqcup \{0\}$, defined
by~(\ref{eq:zy}). Then, $z_I = 0$ for every
nonempty $I \subseteq E$ and
%
\[ z_{I \sqcup \{0\}} = r \left( \,
\sum_{v \, : \, \up(v) \subseteq I} |v|
\right) + s \delta_{I,E} \]
%
for every $I \subseteq E$. In particular,
%
\begin{equation} \label{eq:zn}
z_{E \sqcup \{0\}} = r \left( \, \sum_v |v|
\right) + s = rn+s,
\end{equation}
%
where the sum ranges over all vertices $v$ of
$\tau$. By~(\ref{eq:Pz}), the unique linear
equation defining the generalized permutohedron
$\pP^y(\mathbf{y})$ becomes
%
\begin{equation} \label{eq:eliminate}
x_0 + \sum_{e \in E} x_e = rn + s,
\end{equation}
%
which we will use to eliminate $x_0$ in the
linear inequalities associated with nonempty
subsets of $E \sqcup \{0\}$. Clearly, the
linear inequalities of~(\ref{eq:Pz}) for nonempty
$I \subseteq E$ amount to $x_e \ge 0$ for every
$e \in E$, since $z_I = 0$ for all such $I$.
%%
%%Similarly, for $I = \{0\}$ one gets $x_0 \ge 0$.
We now consider a subset $J \sqcup \{0\}$, where
$J \subseteq E$; the associated inequality is
%
\begin{equation} \label{eq:associate}
x_0 + \sum_{e \in J} x_e \ge
z_{J \sqcup \{0\}}.
\end{equation}
%
Let $\jJ$ be the maximal order filter of $\tau$
contained in $J$; in other words, $\jJ$ is the
union of all vertices $v$ of $\tau$ such that
$\up(v) \subseteq J$. Then,
%
\[ z_{J \sqcup \{0\}} = r \left( \,
\sum_{v \, : \, \up(v) \subseteq J} |v|
\right) + s \delta_{J,E} = r |\jJ| + s
\delta_{J,E} \]
%
and hence, in view of
Equation~(\ref{eq:eliminate}), the inequality
(\ref{eq:associate}) is equivalent to
%
\[ \sum_{e \in I} x_e \le r |\iI| +
s(1-\delta_{J,E}), \]
%
where $I$ and $\iI$ are the complements of $J$
and $\jJ$ in $E$, respectively. The special case
when $J=E$ gives a trivial inequality and can be
ignored. Otherwise, this inequality
yields~(\ref{eq:Q(r,s)-ineq-b}) when $J = \jJ$
is an order filter of $\tau$ itself, so that $I
= \iI$ is an order ideal. It is also implied
by~(\ref{eq:Q(r,s)-ineq-b}) since $I \subseteq
\iI$ and hence
%
\[ \sum_{e \in I} x_e \le \sum_{e \in \iI} x_e.
\]
%
Since complementation induces a one-to-one
correspendence between order ideals and order
filters of $\tau$, the proof follows from our
considerations and Proposition~\ref{prop:y-gen}.
\end{proof}
\begin{proof}[Proof of Theorem~\ref{thm:main-a}]
Combining Propositions~\ref{prop:y-gen}
and~\ref{prop:Q(r,s)-ineq} shows that $\qQ_\tau
= \qQ_\tau(1,0)$ is a lattice polytope which is
unimodularly equivalent to a $Y$-type generalized
permutohedron. The dimension of $\qQ_\tau$ is
equal to $n$ since it contains the $n$-dimensional
cube $[0,1]^E$.
\end{proof}
\subsection{Lattice points}
\label{sec:lattice}
As mentioned in the introduction, we denote by
$P_\tau$ the set of lattice points of $\qQ_\tau$
and consider it a subposet of the product poset
$\NN^E$. Thus, for elements $\mathbf{a} = (a_e)$
and $\mathbf{b} = (b_e)$ of $P_\tau$ we have
$\mathbf{a} \preceq \mathbf{b}$ if and only if
$a_e \le b_e$ for every $e \in E$. In our
running example from Figure~\ref{fig1}, $P_\tau$
has cardinality 92.
The poset $P_\tau$ is an order ideal of $\NN^E$.
In particular, the zero vector $\mathbf{0}$ is
its minimum element and every interval
$[\mathbf{0},x]$ is a product of chains.
Moreover, $P_\tau$ is graded with rank function
given by the sum of coordinates. The following
multiplicative property is a direct consequence
of the relevant definitions.
\begin{proposition} \label{prop:prod}
Let $\tau$ be the disjoint union of two preorders
$\tau_1$ and $\tau_2$. Then:
%
\begin{itemize}
\itemsep=0pt
\item[{\rm (a)}]
$\qQ_\tau = \qQ_{\tau_1} \times \qQ_{\tau_2}$.
\item[{\rm (b)}]
$P_\tau = P_{\tau_1} \times P_{\tau_2}$.
\end{itemize}
%
\end{proposition}
The polytope $\qQ_\tau$ has no interior lattice
points, but it does have some once scaled by 2,
such as the point $(1, 1,\dots,1)$. Let us call
\emph{upper boundary} of $\qQ_\tau$ the set of
its points which satisfy at least one of the
defining inequalities~\eqref{eq:Q-tau-ineq-b}
as an equality and denote this set by
$\hat{\partial}(\qQ_\tau)$.
\begin{proposition} \label{prop:upper}
The number of interior lattice points of $2
\qQ_{\tau}$ is equal to the number of lattice
points of $\qQ_\tau$ not in $\hat{\partial}
(\qQ_\tau)$.
\end{proposition}
\begin{proof}
There is a bijection between the two sets of
lattice points, given by translation by the
vector $(1, 1,\dots,1)$.
\end{proof}
\section{Ehrhart theory}
\label{sec:Ehrhart}
This section studies the Ehrhart polynomial of
the preorder polytope $\qQ_\tau$ and proves
Theorem~\ref{thm:main-b}. An explicit
combinatorial interpretation of the
$h^\ast$-polynomial of $\qQ_\tau$ is conjectured,
partial results are given in this direction and
a conjectural generalization of part (b) of
Theorem~\ref{thm:main-b} (Ehrhart-Zeta duality)
is included.
\subsection{Ehrhart-Zeta duality}
\label{sec:EZ}
\begin{figure}[!h]
\centering
\begin{tikzpicture}[scale=0.5]
\tikzstyle{every node}=[draw,shape=circle, thick]
\node [draw, color=green] at (0,0) (e) {$\ell_e$};
\node [draw, color=magenta] at (0,2) (d) {$\ell_d$};
\node at (0,4) (c) {$\ell_c$};
\node [draw, color=blue] at (0,6) (b) {$\ell_b$};
\node [draw, color=red] at (0,8) (a) {$\ell_a$};
\node at (0,10) (l0) {$\ell_0$};
\node at (4,0) (re) {$r_e$};
\node at (4,2) (rd) {$r_d$};
\node at (4,4) (rc) {$r_c$};
\node at (4,6) (rb) {$r_b$};
\node at (4,8) (ra) {$r_a$};
\node at (4,10) (r0) {$r_0$};
\path [draw, dotted] (l0) -- (ra);
\path [draw, dotted] (l0) -- (rb);
\path [draw, dotted] (l0) -- (rc);
\path [draw, dotted] (l0) -- (rd);
\path [draw, dotted] (l0) -- (re);
\path [draw, dotted] (r0) -- (a);
\path [draw, dotted] (r0) -- (b);
\path [draw, dotted] (r0) -- (c);
\path [draw, dotted] (r0) -- (d);
\path [draw, dotted] (r0) -- (e);
\path [draw, color=red] (a) -- (ra);
\path [draw, color=red] (a) -- (rc);
\path [draw, color=red] (a) -- (re);
\path [draw, color=blue] (b) -- (rb);
\path [draw, color=blue] (b) -- (rd);
\path [draw, color=blue] (b) -- (re);
\path [draw] (c) -- (ra);
\path [draw] (c) -- (rc);
\path [draw] (c) -- (re);
\path [draw, color=magenta] (d) -- (rd);
\path [draw, color=green] (e) -- (re);
\end{tikzpicture}
\caption{Example of bipartite graph $G_\tau$.}
\label{fig2}
\end{figure}
Let us define a bipartite graph $G_\tau$ with
$n+1$ left and $n+1$ right vertices as follows.
There are vertices $\ell_e$ on the left side
and $r_e$ on the right side, for $e \in E
\sqcup \{0\}$. Vertex $\ell_0$ is connected by
an edge to all vertices on the right and $r_0$
is connected by an edge to all vertices on the
left. For every $e \in E$, there are edges from
$\ell_e$ to all vertices $r_f$ such that $f$
belongs to the principal order filter $\up(e)$
of $\tau$. The graph $G_\tau$ has no isolated
vertex; it is shown in Figure~\ref{fig2} in the
case of our running example from
Figure~\ref{fig1}. We set
%
\[ I_e = \{f \in E \sqcup \{0\} \mid \{\ell_e,
r_f\} \ \text{is an edge of} \ G_\tau \} \]
%
for $e \in E \sqcup \{0\}$. Equivalently, $I_e
= \up(e) \sqcup \{0\}$ for every $e \in E$ and
$I_0 = E \sqcup \{0\}$. According to
Propositions~\ref{prop:y-gen}
and~\ref{prop:Q(r,s)-ineq}, the preorder polytope
$\qQ_\tau$ is unimodularly equivalent to the
$Y$-type generalized permutohedron
%
\[ \sum_{e \in E \sqcup \{0\}} y_e \Delta_{I_e}
\]
%
in $\RR_{\ge 0}^{E \sqcup \{0\}}$ with parameters
$y_0 = 0$ and $y_e = 1$ for every $e \in E$. To
apply Theorem~\ref{thm:post}, we need to
understand the $G_\tau$-draconian sequences.
%
\begin{lemma} \label{lem:tau-drac}
A sequence $(a_e)_{e \in E \sqcup \{0\}} \in
\NN^{E \sqcup \{0\}}$ is $G_\tau$-draconian if
and only if
%
\begin{equation} \label{eq:draco1}
a_0 + \sum_{e \in E} a_e = n
\end{equation}
%
and $(a_e)_{e \in E}$ is a lattice point of
$\qQ_{\tau^\ast}$.
\end{lemma}
\begin{proof}
According to the definition (see
Section~\ref{sec:genperm}), $(a_e)_{e \in E
\sqcup \{0\}} \in \NN^{E \sqcup \{0\}}$ is
$G_\tau$-draconian if and only if
Equation~\eqref{eq:draco1} is valid
and~\eqref{eq:draco2} holds for every nonempty
$\aA \subseteq E \sqcup \{0\}$. Let us analyze
these inequalities and show that some of them
are superflous.
Since $I_0 = E \sqcup \{0\}$, if $0 \in \aA$
then \eqref{eq:draco2} becomes $\sum_{e \in \aA}
a_e \le n$, which is a consequence
of~\eqref{eq:draco1}. Thus, we may assume that
$\aA \subseteq E$. Then, we get
%
\begin{equation} \label{eq:minidraco1}
\sum_{e \in \aA} a_e \leq |\up(\aA)|,
\end{equation}
%
where $\up(\aA)$ is the order filter of $\tau$
generated by $\aA$. In particular, if $\aA$ is
an order filter $\jJ$ itself, so that $\up(\aA)
= \jJ$, we get
%
\begin{equation} \label{eq:minidraco2}
\sum_{e \in \jJ} a_e \le |\jJ|.
\end{equation}
%
In general,
%
$ \sum_{e \in \aA} a_e \le \sum_{e \in \up(\aA)}
a_e$,
%
so \eqref{eq:minidraco1} follows from
\eqref{eq:minidraco2} for $\jJ = \up(\aA)$. As
a conclusion, the $G_\tau$-draconian conditions
are equivalent to Equation~\eqref{eq:draco1}
and the validity of~\eqref{eq:minidraco2} for
all order filters $\jJ$ of $\tau$. The latter
means that $(a_e)_{e \in E}$ is a lattice point
of $\qQ_{\tau^\ast}$ and the proof follows.
\end{proof}
We are now in a position to prove
Theorem~\ref{thm:main-b}.
%
\begin{proof}[Proof of Theorem~\ref{thm:main-b}]
The formula of part (a) for the Ehrhart
polynomial follows directly by combining
Theorem~\ref{thm:post}, applied to the graph
$G_\tau$ for the parameters $y_0 = 0$ and $y_e =
1$ for every $e \in E$, with
Lemma~\ref{lem:tau-drac}.
To deduce part (b), let us count the
$m$-multichains in the poset $P_{\tau^\ast}$ for
$m \ge 1$. One can can condition on the top element
$\mathbf{a} = (a_e)_{e \in E} \in P_{\tau^\ast}$ of
such a multichain. Since the interval
$[0,\mathbf{a}]$ is isomorphic to a product of
chains of lengths $a_e$, the number of
$m$-multichains in $P_{\tau^\ast}$ with top
element $\mathbf{a}$ is equal to the product
$\prod_{e \in E} \binom{m+a_e-1}{a_e}$. As a
result,
%
\[ \zZ(P_{\tau^\ast},m+1) =
\sum_{(a_e) \in P_{\tau^\ast}} \prod_{e \in E}
\binom{m+a_e-1}{a_e} \]
%
for every $m \ge 1$ and part (b) follows by
comparison with the formula of part (a).
We may now deduce the last statement of part (a)
as follows. By part (b), the normalized volume
of $\qQ_\tau$ is equal to $n!$ times the
coefficient of $t^n$ in $\zZ(P_{\tau^\ast},t+1)$
and hence to the number of chains of
$P_{\tau^\ast}$ of length $n$ (since
$P_{\tau^\ast}$ is graded of rank $n$, these are
exactly the maximal chains of $P_{\tau^\ast}$).
To every such chain, one can associate the word
of length $n$ in the alphabet $E$ which records
at each step of the chain the unique coordinate
which increases. The resulting map is one-to-one
and, by definition of $\qQ_{\tau^\ast}$, the
words thus obtained are those words $w \in E^n$
such that the elements of any order ideal $\jJ$
of $\tau^\ast$ appear at most $|\jJ|$ times in
$w$. Equivalently, the elements of any order
ideal $\iI$ of $\tau$ should appear at least
$|\iI|$ times in $w$.
\end{proof}
\subsection{Matrix Ehrhart-Zeta duality}
\label{sec:matrixEZ}
One can extend Ehrhart-Zeta duality in terms of
infinite matrices as follows. For $k, \ell \in
\NN$ we denote by $\nabla_\tau(k,\ell)$ the
number of $k$-multichains in the poset of
lattice points of $\ell \qQ_\tau$ (where the
order is induced from $\NN^n$, as usual). This
function
can be viewed as an infinite matrix with rows
and columns indexed by nonnegative integers,
every column of which is the sequence of values
of a zeta polynomial. We have $\nabla_\tau(k,0)
= \nabla_\tau(0,\ell) = 1$, $\nabla_\tau(k,1) =
\zZ(P_\tau,k+1)$ and $\nabla_\tau(1,\ell) =
\Ehr(\qQ_\tau,\ell)$ for all $k, \ell \in \NN$.
The following conjecture generalizes part (b)
of Theorem~\ref{thm:main-b}; it has been verified
by computer on certain finite pieces of
$\nabla_\tau$ for preorders of size at most 5.
%
\begin{conjecture} \label{conj:matrixEZ}
The matrix $\nabla_\tau$ is the transpose of
$\nabla_{\tau^\ast}$ for every preorder $\tau$.
\end{conjecture}
A bijective proof of Ehrhart-Zeta duality
was given for chain preorders in~\cite{Ath26}.
This proof can be easily extended in the setting
of the matrix Ehrhart-Zeta duality. We sketch the
proof of the following proposition, which
generalizes the first proof of
Theorem~\ref{thm:Ath26} given in
\cite[Section~4]{Ath26}.
%
\begin{proposition} \label{prop:matrixEZ}
Conjecture~\ref{conj:matrixEZ} holds for all
chain preorders.
\end{proposition}
\begin{proof}
Let $\tau$ be a chain $v_1 \prec v_2 \prec
\cdots \prec v_p$ with vertices of sizes $r_1,
r_2,\dots,r_p$, respectively. We set $s_i = r_1
+ r_2 + \cdots + r_i$ for $0 \le i \le p$, so
that where $s_0 := 0$ and $s_p = n$. We may
assume that $E = \{1, 2,\dots,n\}$ with elements
indexed so that $\qQ_\tau$ is described by the
inequalities $x_i \ge 0$ for $1 \le i \le n$ and
%
\[ x_1 + x_2 + \cdots + x_{s_i} \le s_i \]
%
for $1 \le i \le p$.
By \cite[Lemma~4.1]{Ath26}, the $k$-multichains
in the poset of lattice points of $\ell \qQ_\tau$
are in bijection with the points
$(a_1, a_2,\dots,a_{nk}) \in \NN^{nk}$ such that
%
\begin{equation} \label{eq:ineq-ai}
a_1 + a_2 + \cdots + a_{s_i k} \le s_i \ell
\end{equation}
%
for $1 \le i \le p$. Given such a point, for
$0 \le j \le n\ell$ we define $b_j$ as the
number of $i \in \{1, 2,\dots,nk\}$ such
that $a_1 + a_2 + \cdots + a_i = j$. Then,
%
\begin{equation} \label{eq:sum-bj}
b_0 + b_1 + \cdots + b_{n \ell} = nk
\end{equation}
%
and
%
\begin{equation} \label{eq:ineq-bj-1}
b_0 + b_1 + \cdots + b_{s_i \ell} \ge s_i k
\end{equation}
%
for $0 \le i \le p$. Using
Equation~(\ref{eq:sum-bj}) to eliminate $b_0$
we find that the solutions $(a_1, a_2,\dots,a_{nk})
\in \NN^{nk}$ of the system of linear
inequalities (\ref{eq:ineq-ai}) are in bijection
with the solutions $(b_1, b_2,\dots,b_{n \ell})
\in \NN^{n\ell}$ of the system of linear
inequalities
%
\begin{equation} \label{eq:ineq-bj-2}
b_{s_i \ell + 1} + b_{s_i \ell + 2} + \cdots +
b_{n \ell} \le (n - s_i) k
\end{equation}
%
for $0 \le i \le p-1$. By \cite[Lemma~4.1]{Ath26},
and since $\qQ_{\tau^\ast}$ is described by the
inequalities $x_i \ge 0$ for $1 \le i \le n$ and
%
\[ x_{s_i + 1} + x_{s_i + 2} + \cdots + x_n
\le n - s_i \]
%
for $0 \le i \le p-1$, these solutions are in
bijection with the $\ell$-multichains in the
poset of lattice points of $k \qQ_{\tau^\ast}$
and the proof follows.
\end{proof}
\subsection{On the $h^\ast$-polynomial}
\label{sec:h-ast}
We recall from Section~\ref{sec:Ehr-hstar} that
the $h^\ast$-polynomial $h^\ast(\qQ_\tau, t)$
has nonnegative coefficients which sum to the
normalized volume of $\qQ_\tau$. Thus, the
following conjecture is a natural refinement of
the second statement of Theorem~\ref{thm:main-b}
(a). We fix an arbitrary total order $\le$ on the ground
set $E$ and, given a word $w = (w_1,
w_2,\dots,w_n) \in E^n$, we denote by $\des(w)$
the number of \emph{descents} of $w$, meaning
indices $i \in \{1, 2,\dots,n-1\}$ such that
$w_i > w_{i+1}$. The following conjecture extends
\cite[Conjecture~5.2]{AXY26} from arbor polytopes
to all preorder polytopes.
%
\begin{conjecture} \label{conj:h-ast}
Let $\tau$ be a preorder on a totally ordered
$n$-element set $E$. Then
%
\[ h^\ast(\qQ_\tau, t) = \sum_{w \in \wW_\tau}
t^{n-1-\des(w)}, \]
%
where $\wW_\tau$ is the set of words $w \in E^n$
such that for every order ideal $\iI$ of $\tau$,
the elements of $\iI$ appear a total of at least
$|\iI|$ times in $w$.
\end{conjecture}
This conjecture has been verified by computer
for all preorders of size at most 7. The motivation
behind it comes from \cite[Theorem~1.5]{AXY26},
which verifies it in the special case in which
$\tau$ is the preorder associated to an arbor
which consists of a root of any size and several
single element leaves.
A combinatorial interpretation of
$h^\ast(\qQ_\tau, t)$ which seems different from
that of Conjecture~\ref{conj:h-ast} was given in
\cite[Section~4]{Ath26} in the special case that
the preorder $\tau$ is a chain. The proof is
based on the theory of pure shellability of posets
and generalizes to all preorders with a minimum
vertex. To state the result, it is convenient to
assume that $E = \{1, 2,\dots,n\}$ is totally
ordered as usual. Given a word $w =
(w_1, w_2,\dots,w_n) \in \{1, 2,\dots,n\}^n$, we
denote by $\asc^\ast(w)$ the number of indices
(called \emph{modified ascents}) $i \in \{0,
1,\dots,n-1\}$ such that $w_i < w_{i+1}$, where
$w_0 := 1$. The following statement generalizes
\cite[Proposition~4.2]{Ath26}. To avoid making
too long of a digression we assume familiarity
with the relevant parts of
\cite[Section~4]{Ath26} and only sketch the
proof, which directly extends that of
\cite[Proposition~4.2]{Ath26}.
%
\begin{proposition} \label{prop:h-ast-min}
Let $\tau$ be a preorder of size $n$ on the
ground set $E = \{1, 2,\dots,n\}$ which has a
minimum vertex. Assuming that 1 belongs to this
vertex we have
%
\[ h^\ast(\qQ_\tau, t) = \sum_{w \in \wW_\tau}
t^{\asc^\ast(w)}, \]
%
where $\wW_\tau$ is as in
Conjecture~\ref{conj:h-ast}.
\end{proposition}
\begin{proof}
As explained in the proof of
\cite[Proposition~4.2]{Ath26} we have
%
\[ h^\ast(\qQ_\tau, t) = h(\Delta(P_{\tau^\ast}),
t) \]
%
by Ehrhart-Zeta duality, where $\Delta
(P_{\tau^\ast})$ stands for the order complex of
$P_{\tau^\ast}$.
We consider the poset
$\widehat{P}_{\tau^\ast} = P_{\tau^\ast} \cup
\{ \hat{1} \}$, obtained from $P_{\tau^\ast}$
by adding a maximum element $\hat{1}$, and an
integer valued edge labeling $\lambda$ of this
poset, defined as follows. For a cover
relation $(u,v)$ of $P_{\tau^\ast}$ we set
$\lambda(u,v) = i$, if $u$ and $v$ differ in
the $i$th coordinate. We also set
$\lambda(u, \hat{1}) = 1$ for every maximal
element $u \in P_\tau$ and claim that $\lambda$
is an EL-labeling. This means that, given any
nonsingleton closed interval $[u, v]$ of
$\widehat{P}_{\tau^\ast}$, there is a unique
rising (with respect to $\lambda$) maximal
chain in $[u, v]$ and that this chain is
lexicographically least among all maximal
chains of $[u, v]$. This is clear if $v \ne
\hat{1}$, since then the labels of the maximal
chains of $[u, v]$ are the permutations of a
multiset. Suppose that $v = \hat{1}$. By
assumption, $\tau^\ast$ has a maximum vertex
and this vertex contains 1. This implies that
from the linear inequalities
(\ref{eq:Q-tau-ineq-b}) which define
$\qQ_{\tau^\ast}$, only the one associated
to the order ideal $E$ of $\tau^\ast$ involves
the variable $x_1$. As a result, there is a
unique maximal chain in $[u, \hat{1}]$ with
label $(1, 1,\dots,1)$ and this chain is the
only rising maximal chain of $[u, \hat{1}]$.
Since $\lambda$ is EL,
$h(\Delta(P_{\tau^\ast}), t)$ can be
interpreted as the descent enumerator (with
respect to $\lambda$) of the maximal chains
of $\widehat{P}_{\tau^\ast}$. As already
observed in the proof of
Theorem~\ref{thm:main-b}, these chains are in
bijection with the elements of $\wW_\tau$ and
the proof follows.
\end{proof}
The proof of Proposition~\ref{prop:h-ast-min}
shows that the poset $P_\tau$ is shellable, and
hence Cohen-Macaulay (over $\ZZ$ and any field),
if $\tau$ has a maximum vertex. This raises the
following question.
\begin{question} \label{que:shellability}
Is the poset $P_\tau$ shellable for every
preorder $\tau$? Is it Cohen--Macaulay for
every preorder $\tau$?
\end{question}
A formula for $h^\ast(\qQ_\tau, t)$ may be
derived from part (a) of Theorem~\ref{thm:main-b}
as follows.
%
\begin{proposition} \label{prop:h-ast}
Let $\tau$ be a preorder of size $n$ on a
totally ordered ground set $E$. Then
%
\[ h^\ast(\qQ_\tau, t) = (1-t)^n +
\sum_{w \in \widehat{\wW}_\tau \smallsetminus
\{\varnothing\}} (1-t)^{n-\ell(w)}
t^{\des(w)}, \]
%
where $\widehat{\wW}_\tau$ is the set of words
$w$ of length (number of letters) $\ell(w) \le n$
with letters from $E$ such that for every order
filter $\jJ$ of $\tau$, the elements of $\jJ$
appear a total of at most $|\jJ|$ times in $w$.
\end{proposition}
\begin{proof}
Given $\mathbf{a} = (a_e) \in \NN^E$, let us
denote by $\mathfrak{S}_{\mathbf{a}}$ the set
of permutations of the multiset which consists
of $a_e$ copies of $e$, for $e \in E$. By part
(a) of Theorem~\ref{thm:main-b},
%
\[ \begin{aligned}
\sum_{m \ge 0} \Ehr(\qQ_\tau,m) t^m &= 1 +
\sum_{m \ge 0} \Ehr(\qQ_\tau,m+1) t^{m+1} \\
&= 1 + t \sum_{m \ge 0} \left( \,
\sum_{(a_e) \in P_{\tau^\ast}} \prod_{e \in E}
\binom{m+a_e}{a_e} \right) t^m \\
&= 1 + t \sum_{\mathbf{a} \in P_{\tau^\ast}}
\left( \, \sum_{w \in \mathfrak{S}_{\mathbf{a}}}
t^{\des(w)} \right) / (1-t)^{1 + |\mathbf{a}|},
\end{aligned} \]
%
where the third equality applies MacMahon's
formula \cite[p.~211]{Mac04}
%
\[ \sum_{m \ge 0} \left( \, \prod_{e \in E}
\binom{m+a_e}{a_e} \right) t^m
= \frac{\displaystyle\sum_{w \in
\mathfrak{S}_{\mathbf{a}}}
t^{\des(w)}}{(1-t)^{1 + |\mathbf{a}|}} \]
%
for $\mathbf{a} = (a_e) \in \NN^E$. This
computation shows that
%
\[ h^\ast(\qQ_\tau, t) = (1-t)^{n+1} + t
\sum_{\mathbf{a} \in P_{\tau^\ast}}
(1-t)^{n-|\mathbf{a}|}
\sum_{w \in \mathfrak{S}_{\mathbf{a}}}
t^{\des(w)}. \]
%
We now observe that the summand of the outer
sum on the right-hand side for $\mathbf{a} =
\mathbf{0}$ is equal to $(1-t)^n$ and that
$\widehat{\wW}_\tau$ is the disjoint union of
the sets $\mathfrak{S}_{\mathbf{a}}$ for
$\mathbf{a} \in P_{\tau^\ast}$ and the proof
follows.
\end{proof}
As mentioned in the introduction, Ehrhart
polynomials of preorder polytopes are not always
real-rooted. For example, if $\tau$ is a partial
order consisting of a minimum element and two
other elements incomparable to each other which
cover it, then $\Ehr(\qQ_\tau, t) = 1 + 23t/6 +
5t^2 + 13t^3/6$ has two nonreal complex roots.
We conjecture that $\Ehr(\qQ_\tau, t)$ has the
weaker magic positivity property and hence that
$h^\ast(\qQ_\tau,t)$ is real-rooted.
%
\begin{conjecture} \label{conj:magic}
The Ehrhart polynomial $\Ehr(\qQ_\tau, t)$ is
magic positive for every preorder $\tau$. In
particular, $h^\ast(\qQ_\tau,t)$ is real-rooted.
\end{conjecture}
This conjecture has been proven for the special
class of the arbor polytopes studied in~\cite{AXY26}
and also follows from \cite[Theorem~13]{SP02}
when $\tau$ is a chain with vertices of size one
(in which case $\qQ_\tau$ is the standard
Pitman--Stanley polytope). The coefficients of
the Ehrhart polynomial in the magic basis
$t^i(1+t)^{n-i}$, multiplied by $n!$, have
interesting combinatorial interpretations in both
cases; see \cite[Section~1]{AFM26}
\cite[Section~3]{AXY26} \cite[Theorem~10.1]{GS06}.
Conjecture~\ref{conj:magic} has been verified by
computer for all preorders of size at most 7.
\subsection{On the zeta polynomial}
\label{sec:zeta-conj}
Computational evidence suggests that $(-1)^n
\zZ(P_\tau,-1)$ is equal to the number of maximal
elements of $P_\tau$. A refined conjecture may be
formulated using the notion of $q$-zeta polynomial
introduced in~\cite{Cha25a}. Given a finite graded
poset $P$ with rank function $\rho: P \to \NN$, it
was shown in~\cite{Cha25a} that there exists a
polynomial $\zZ_q(P,t) \in \QQ(q)[t]$ such that
%
\[ \zZ_q(P,[m]_q) = \sum_{p_1 \preceq p_2 \preceq
\cdots \preceq p_{m-1}} q^{\rho(p_1) + \rho(p_2)
+ \cdots + \rho(p_{m-1})} \]
%
for every integer $m \ge 2$, where the sum ranges
over all multichains $p_1 \preceq p_2 \preceq
\cdots \preceq p_{m-1}$ in $P$ and $[m]_q =
(q^m-1)/(q-1)$ is the standard $q$-analog of $m$.
Let us denote by $\Max(P)$ the set of maximal
elements of a finite poset $P$.
%
\begin{conjecture} \label{conj:zeta}
For every preorder $\tau$ of size $n$
%
\begin{equation} \label{eq:conj-qzeta}
\zZ_q(P_\tau, [-1]_q) =
(-1)^n \sum_{\mathbf{a} \in \Max(P_\tau)}
q^{-\cov(\mathbf{a})},
\end{equation}
%
where $\cov(\mathbf{a})$ is the number of elements
of $P_\tau$ covered by $\mathbf{a}$ and $[-1]_q =
-1/q$. In particular,
%
\begin{equation} \label{eq:conj-zeta}
\zZ(P_\tau,-1) = (-1)^n \, \# \Max(P_\tau)
\end{equation}
%
for every preorder $\tau$ of size $n$.
\end{conjecture}
This conjecture has been verified by
computer for all preorders of size at most 6.
For the example of Figure~\ref{fig1} we have
$\zZ(P_\tau,-1) = -18$ and
%
\[ \zZ_q(P_\tau, [-1]_q) = - (2q + 1)
(q^2 + 4q + 1) / q^5 \]
%
and $P_\tau$ has 18 maximal elements.
\begin{remark} \label{rem:conj-zeta} \rm
(a) Equation~(\ref{eq:conj-zeta}) is a consequence
of Conjecture~\ref{conj:h-ast}. Indeed, we have
$\zZ(P_\tau,-1) = \Ehr(\qQ_{\tau^\ast},-2)$ by
Ehrhart-Zeta duality and
%
\[ \Ehr(\qQ_{\tau^\ast}, t) = \sum_{i=0}^n
h^\ast_i(\qQ_{\tau^\ast}) \binom{t+n-i}{n} \]
%
by Equation~(\ref{eq:def-hstar}). Moreover,
$h^\ast_n(\qQ_{\tau^\ast}) = 0$ since
$\qQ_{\tau^\ast}$ has no interior lattice points
and hence
%
\[ \Ehr(\qQ_{\tau^\ast},-2) = (-1)^n \,
h^\ast_{n-1}(\qQ_{\tau^\ast}). \]
%
According to Conjecture~\ref{conj:h-ast},
$h^\ast_{n-1}(\qQ_{\tau^\ast})$ is equal to the
number of words $w \in \wW_{\tau^\ast}$ with no
descent. These are the (weakly) increasing words
$w \in E^n$ such that the elements of any order
filter $\jJ$ of $\tau$ appear a total of at least
$|\jJ|$ times in $w$. Equivalently, the elements
of any order ideal $\iI$ of $P_\tau$ should appear
at most $|\iI|$ times in $w$. These words are
obviously in bijection with the maximal elements
of $P_\tau$.
(b) By Ehrhart reciprocity, $(-1)^n \,
\Ehr(\qQ_{\tau^\ast},-2)$ is equal to the number
of lattice points in the interior of
$2\qQ_{\tau^\ast}$. Thus, in view of
Proposition~\ref{prop:upper},
Equation~(\ref{eq:conj-zeta}) is equivalent to
the statement that the number of maximal elements
of $P_\tau$ is equal to the number of lattice
points of $\qQ_{\tau^\ast}$ not in the upper
boundary $\hat{\partial} (\qQ_{\tau^\ast})$.
\qed
\end{remark}
\begin{remark} \label{rem:conj-zeta-chains} \rm
Equation~(\ref{eq:conj-zeta}) holds for every
chain preorder $\tau$. To explain why, let us
adopt the notation of the proof of
Proposition~\ref{prop:matrixEZ}. By part (b)
of Remark~\ref{rem:conj-zeta}, it suffices to
show that the number of maximal elements
of $P_{\tau^\ast}$ is equal to the number of
lattice points of $\qQ_\tau \smallsetminus
\hat{\partial} (\qQ_\tau)$. The maximal elements
of $P_{\tau^\ast}$ are the tuples $(a_1,
a_2,\dots,a_n) \in \NN^n$ such that
%
$a_1 + a_2 + \cdots + a_n = n$
%
and
%
\[ a_{s_i + 1} + a_{s_i + 2} + \cdots +
a_n \le n - s_i \]
%
for $1 \le i \le p-1$ or, equivalently,
%
\[ a_1 + a_2 + \cdots + a_{s_i} \ge s_i \]
%
for $1 \le i \le p-1$. The lattice points
of $\qQ_\tau \smallsetminus \hat{\partial}
(\qQ_\tau)$ are the tuples $(b_1, b_2,\dots,b_n)
\in \NN^n$ such that
%
\[ b_1 + b_2 + \cdots + b_{s_i} < s_i \]
%
for $1 \le i \le p$. Given a maximal element
$(a_1, a_2,\dots,a_n)$ of $P_{\tau^\ast}$, for
$1 \le j \le n$ we define $b_j$ to be the
number of indices $i \in \{1, 2,\dots,n\}$
such that $a_1 + a_2 + \cdots + a_i = j-1$.
We leave it to the reader to verify that
$(b_1, b_2,\dots,b_n)$ is a lattice point of
$\qQ_\tau \smallsetminus \hat{\partial}
(\qQ_\tau)$ and that the resulting map is a
bijection from the set of maximal elements of
$P_{\tau^\ast}$ to the set of lattice points
of $\qQ_\tau \smallsetminus \hat{\partial}
(\qQ_\tau)$.
\end{remark}
% Could it be due to the cubical nature of the posets? not really..
% This equality fails for the Dexter semlattice D(4), not ranked
% echoue aussi pour un poset ranked mais pas cubique
% P=Poset({0:[1,2],1:[3],2:[4],3:[5],4:[5,6]})
% echoue pour un poset assez cubique:
% P=Poset({0:[1,2],1:[3,4],2:[4,5],3:[6],4:[6,7],5:[7]})
% echoue en type B pour toute autre tranche que h=0
\section{Lattice point enumeration}
\label{sec:h-vectors}
This section extends to preorder polytopes some
conjectures of~\cite{Cha25b} on the lattice point
enumeration of arbor polytopes. Throughout this
and the following two sections, $\tau$ continues
to be a preorder of size $n$ on the ground set $E$.
For $0 \le i \le n$, we let $h_i(\tau)$ be
the number of lattice points of $\qQ_\tau$ which
have exactly $i$ nonzero coordinates. The vector
$h(\tau) = (h_0(\tau), h_1(\tau),\dots,h_n(\tau))$
is called the \emph{$h$-vector} of $\tau$ and the
polynomial $h(\tau, t) = \sum_{i=0}^n h_i(\tau)t^i$
is called the \emph{$h$-polynomial}. These concepts
directly generalize the notion of $h$-vector of an
arbor, introduced in \cite{Cha25b}. For the example
of Figure~\ref{fig1} we have $h(\tau) = (1, 11, 34,
34, 11, 1)$. We note that $h_0(\tau) = h_1(\tau) =
1$. The following conjecture, the parts of
which appear in the order of increasing strength,
generalizes Conjecture~1.1 in~\cite{Cha25b}. For
undefined terminology about simplicial polytopes,
see \cite{StaCCA}.
%
\begin{conjecture} \label{conj:hvector}
For every preorder $\tau$ of size $n$:
%
\begin{itemize}
\itemsep=0pt
\item[{\rm (a)}]
$h(\tau)$ is palindromic, meaning that $h_i(\tau)
= h_{n-i}(\tau)$ for $0 \le i \le n$.
\item[{\rm (b)}]
$h(\tau)$ is palindromic and unimodal, meaning
that $h_i(\tau) = h_{n-i}(\tau)$ for $0 \le i \le
n$ and $h_0(\tau) \le h_1(\tau) \le \cdots \le
h_{\lfloor n/2 \rfloor}(\tau)$.
\item[{\rm (c)}]
$h(\tau)$ is equal to the $h$-vector of an
$n$-dimensional simplicial polytope.
\item[{\rm (d)}]
$h(\tau)$ is equal to the $h$-vector of an
$n$-dimensional flag simplicial polytope.
\end{itemize}
\end{conjecture}
For example, if $\tau$ is the chain preorder
of size $n$ having $n$ vertices of cardinality
1 or one vertex of cardinality $n$, then the
role of the flag simplicial polytope of part (d)
can be played by the simplicial associahedron of
of type $\TA_n$ and $\TB_n$, respectively; see
Sections~7 and~8 of \cite{Cha25b}. The following
statement was implicitly conjectured for arbor
polytopes in \cite[Section~1.1]{Cha25b}.
%
\begin{conjecture} \label{conj:gamma}
The polynomial $h(\tau,t)$ is $\gamma$-positive,
meaning that
%
\[ h(\tau,t) = \sum_{i=0}^{\lfloor n/2 \rfloor}
\gamma_i(\tau) t^i (1+t)^{n-2i} \]
%
for some nonnegative integers $\gamma_0(\tau),
\gamma_1(\tau),\dots,\gamma_{\lfloor n/2 \rfloor}
(\tau)$, for every preorder $\tau$ of size $n$.
In particular, the vector $h(\tau)$ is palindromic
and unimodal.
\end{conjecture}
A more optimistic conjecture is the following.
%
\begin{conjecture} \label{conj:roots}
The polynomial $h(\tau,t)$ has only real roots
for every preorder $\tau$.
\end{conjecture}
The following statement substantially
generalizes Conjecture~4.2 in~\cite{Cha25b}.
%
\begin{conjecture} \label{conj:hvector-dual}
We have $h(\tau) = h(\tau^\ast)$ for every
preorder $\tau$.
\end{conjecture}
All previous conjectures, other than
Conjecture~\ref{conj:roots}, were proven for
chain preorders in \cite{Ath26}. They have been
verified (together with
Conjecture~\ref{conj:roots}) in several other
special cases in \cite{AXY26}
\cite[Sections~7-10]{Cha25b}; see also
Section~\ref{sec:exa}. The authors have verified
by computer Conjecture~\ref{conj:hvector} (c) for
all preorders of size at most 7 and arbors of
size at most 8, Conjectures~\ref{conj:gamma}
and~\ref{conj:hvector-dual} for all preorders
of size at most 7 and
Conjecture~\ref{conj:roots} for all preorders
of size at most 7 and arbors of size at most 10.
%%\begin{figure}[!h]
%%\centering
%%\begin{tikzpicture}
%%\tikzstyle{every node}=[draw,shape=circle, thick]
%%\node at (0,0) (a) {$a$};
%%\node at (2,0) (b) {$b$};
%%\node at (4,0) (c) {$c$};
%%\node at (1,2) (d) {$d$};
%%\node at (3,2) (e) {$e$};
%%\path [draw] (a) -- (d);
%%\path [draw] (b) -- (d);
%%\path [draw] (b) -- (e);
%%\path [draw] (c) -- (e);
%%\end{tikzpicture}
%%\end{figure}
%%The common $h$-vector for $\tau$ and $\tau'$
%%is $(1, 9, 25, 25, 9, 1)$.
\begin{remark} \label{rem:christos-example} \rm
We do not know of a class of generalized
permutohedra larger than that of preorder
polytopes for which the $h$-vector may be
palindromic. For example, $\qQ =
\bDelta_{\{1,2\}} + \bDelta_{\{2,3\}} +
\bDelta_{\{3\}}$ is a $Y$-type generalized
permutohedron in $\RR^3$ described by the
inequalities $x_1, x_2, x_3 \ge 0$ and
%
\begin{align*}
x_1 & \le 1 \\
x_1 + x_2 & \le 2 \\
x_3 & \le 2 \\
x_1 + x_2 + x_3 & \le 3.
\end{align*}
%
Counting its lattice points by the number of
nonzero coordinates gives the vector $(1,5,6,1)$,
which is not palindromic.
\end{remark}
\begin{remark} \label{rem:series} \rm
One may naturally consider the formal power
series
%
\[ F(\tau, t) =
\frac{h(\tau,t)}{{(1-t)}^{n+1}}. \]
%
The coefficients of $F(\tau, t)$ are the values
of a polynomial at nonnegative integers. Curiously,
this polynomial has all its complex roots on the
line $\re(z) = -1/2$ for all preorders of size at
most 7 and all arbors of size at most 8. The arbor
\medskip
\begin{center}
\begin{tikzpicture}
[grow=down]
\tikzstyle{every node}=[draw,shape=circle]
\node [fill=racines,thick] {$a$}
child {node {$b,b'$}}
child {node {$c,c'$}}
child {node {$d,d'$}}
child {node {$e,e'$}}
;
\end{tikzpicture}
\end{center}
\medskip
\noindent
is the unique arbor of size 9 for which this
property fails.
%
%%There are 5 similar counter-examples among
%%arbors of size $10$.
\end{remark}
\section{Preorder polytopes and reflexive polytopes}
\label{sec:reflexive}
This section shows that, after suitable
translation, the lattice polytope
%
\[ \qQ_\tau(1,1) = \qQ_\tau + \bDelta_E \]
%
(see Section~\ref{sec:Minkowski}) is reflexive
and investigates its properties. We recall that
an $n$-dimensional lattice polytope $\qQ$ in
$\RR^n$ having $\mathbf{0}$ in its
interior is said to be \emph{reflexive} if the
polar dual of $\qQ$ is also a lattice polytope.
An equivalent condition \cite[Theorem~35.8]{HiAC}
is that the $h^\ast$-polynomial of $\qQ$ is
palindromic (of degree $n$).
Let $\rR_\tau$ denote the translation of
$\qQ_\tau(1,1)$ by the negative of the vector
$\mathbf{1} = \sum_{e \in E} {\rm \mathbf{e}}_e
\in \RR^E$. By Proposition~\ref{prop:Q(r,s)-ineq},
$\qQ_\tau(1,1)$ is defined by the inequalities
$x_e \ge 0$ for every $e \in E$ and
%
\[ \sum_{e \in \iI} x_e \le |\iI| + 1 \]
%
for every order ideal $\iI$ of $\tau$. As a result,
$\rR_\tau$ is defined by the inequalities
%
\begin{equation} \label{eq:Rtau-a}
x_e \ge -1
\end{equation}
%
for every $e \in E$ and
%
\begin{equation} \label{eq:Rtau-b}
\sum_{e \in \iI} x_e \le 1
\end{equation}
%
for every order ideal $\iI$ of $\tau$.
%
\begin{proposition}
The lattice polytope $\rR_\tau$ is reflexive for
every preorder $\tau$.
\end{proposition}
\begin{proof}
The inequalities~(\ref{eq:Rtau-a})
and~(\ref{eq:Rtau-b}) show that $\mathbf{0}$ is
the unique interior lattice point of $\rR_\tau$
and that its polar dual has vertices with integer
coordinates.
\end{proof}
Let $\rR^\vee_\tau$ be the reflexive polytope
which is polar dual to $\rR_\tau$. By definition,
$\rR^\vee_\tau$ is the convex hull of the vectors
$-{\rm \mathbf{e}}_e$ for all $e \in E$ and
$\sum_{e \in \iI} {\rm \mathbf{e}}_e$ for all
nonempty order ideals $\iI$ of $\tau$. Let us
denote this set of vectors by $V_\tau$. Since
$V_\tau$ is a subset of the set of vertices of
the convex hull of the unit cube $[0,1]^n$ and
the vectors $-{\rm \mathbf{e}}_e$ for all
$e \in E$, every element of $V_\tau$ is a vertex
of $\rR^\vee_\tau$ and $\rR^\vee_\tau$ has no
lattice points other than $\mathbf{0}$ and its
vertices.
We now state two conjectures about the
$h^*$-vector of $\rR^\vee_\tau$. For the
example of Figure~\ref{fig1} we have
$h^\ast(\rR_\tau) = (1,228,1614,1614,218,1)$
and $h^\ast(\rR^\vee_\tau) = (1,7,16,16,7,1)$;
in particular, $\rR_\tau$ has 234 lattice
points and $\rR^\vee_\tau$ has 13 lattice
points.
\begin{conjecture} \label{conj:Rtau-a}
We have $h^\ast(\rR^\vee_\tau) = h^\ast
(\rR^\vee_{\tau^\ast})$ (equivalently, $\Ehr
(\rR^\vee_\tau, t) = \Ehr(\rR^\vee_{\tau^\ast},
t)$) for every preorder $\tau$.
\end{conjecture}
\begin{conjecture} \label{conj:Rtau-b}
We have $h^\ast(\rR^\vee_\tau,t) = h^\ast
(\rR^\vee_{\tau_1},t) \, h^\ast(\rR^\vee_{\tau_2},t)$
if the preorder $\tau$ is the ordinal sum of
$\tau_1$ and $\tau_2$.
\end{conjecture}
We are unable to explain the latter conjecture
in terms of the structure of $\rR^\vee_\tau$
since, in general, this polytope cannot be
expressed as a free product.
\begin{remark} \rm
Curiously, the roots of the Ehrhart polynomial
of $\rR^\vee_\tau$ lie on the line
$\operatorname{Re}(z)=-1/2$ for every preorder
$\tau$ of size at most 6. Assuming the validity
of Conjecture~\ref{conj:Rtau-b}, this property
fails when $\tau$ is the ordinal sum of two
5-element antichains.
\end{remark}
%%It should also fail for the antichain poset
%%of size $9$, where the $h^*$-vector is expected
%%to be
%
%%\[ (1, 511, 16867, 131275, 344551, 344551,
%% 131275, 16867, 511, 1). \]
%
\section{Double Ehrhart theory}
\label{sec:double}
This section is concerned with the lattice point
enumeration of the polytope $\qQ_\tau(r,s)$,
introduced in Section~\ref{sec:Minkowski}.
\begin{proposition}
There exists a polynomial $\biE_\tau(u,v)$ of
total degree $n$ such that $\biE_\tau(r,s)$ is
equal to the number of lattice points of
$\qQ_\tau(r,s)$ for all $r,s \in \NN$.
\end{proposition}
\begin{proof}
Since $\qQ_\tau(r,s)$ can be expressed as the
Minkowski sum of two $n$-dimensional lattice
polytopes with dilation factors $r$ and $s$ by
Equation~(\ref{eq:def-Qrs}), the result follows
from a general theorem of Bernstein--McMullen
which defines multivariate Ehrhart polynomials;
see, for instance, \cite[Theorem 19.4]{gruber}.
Alternatively, the explicit formula
%
\begin{equation} \label{eq:Ehr-double}
\biE_\tau(r,s) = \sum_{(a_e) \in \gG_{G_\tau}}
\binom{s + a_0}{a_0} \prod_{e \in E}
\binom{r + a_e - 1}{a_e}
\end{equation}
%
for the number of lattice points of $\qQ_\tau
(r,s)$ follows from Theorem~\ref{thm:post} for
$t=1$ and the parameters $y_0 = s$ and $y_e = r$
for $e \in E$, where the sum ranges over all
$G_\tau$-draconian sequences
$(a_e) \in \NN^{E \sqcup \{0\}}$ characterized
in Lemma~\ref{lem:tau-drac}.
\end{proof}
For our running example from Figure~\ref{fig1}
we have
%
\begin{multline*}
\biE_\tau(u,v) = (760 u^5 + 1680 u^4 v + 990
u^3 v^2 + 230 u^2 v^3 + 25 uv^4 + v^5 + 2740
u^4 + 4350 u^3 v \\ + 1830 u^2 v^2 + 280 uv^3
+ 15 v^4 + 3860 u^3 + 4180 u^2 v + 1115 uv^2
\\ + 85 v^3 + 2660 u^2 + 1760 uv + 225 v^2 +
900 u + 274 v + 120) / 120
\end{multline*}
%
and, in particular,
%
\[ \begin{aligned}
\Ehr(\qQ_\tau, t) &= (760 t^5 + 2740 t^4 +
3860 t^3 + 2660 t^2 + 900 t + 120)/120 \\
\Ehr(\qQ_\tau(1,1), t) &= (3686 t^5 + 9215
t^4 + 9240 t^3 + 4645 t^2 + 1174 t + 120)/
120. \end{aligned} \]
%
The terms of total degree $n=5$ of
$\biE_\tau(u,v)$ yield the decomposition $1 +
25 + 230 + 990 + 1680 + 760 = 3686$ of the
normalized volume of the reflexive polytope
$Q_\tau(1,1)$.
Since $\biE_\tau(t,t)$ is the Ehrhart polynomial
of the reflexive polytope $\qQ_{\tau}(1,1)$, we
have
%
\[ \biE_\tau(-u,-u) = (-1)^n \,
\biE_\tau(u-1,u-1) \]
%
for every preorder $\tau$ of size $n$ by Ehrhart
reciprocity. The following stronger statement has
been verified by computer for all preorders of
size at most 5.
%
\begin{conjecture} \label{conj:double}
The double Ehrhart polynomial $\biE_\tau(u,v)$
satisfies the identity
%
\begin{equation} \label{eq:conj-double}
\biE_\tau(-u,-v) = (-1)^n \, \biE_\tau(u-1,v-1)
\end{equation}
%
for every preorder $\tau$ of size $n$.
\end{conjecture}
\begin{remark} \label{rem:equivalence} \rm
The specialization $v=0$ of
Conjecture~\ref{conj:double} is in fact
equivalent to Conjecture~\ref{conj:h-ast}.
Indeed, let us use the notation of
Section~\ref{sec:Ehrhart}. Setting $u=-m$ and
$v=0$ in~(\ref{eq:conj-double}) and applying
Equation~(\ref{eq:Ehr-double}) we get
\[ \begin{aligned}
\Ehr(\qQ_\tau, m) &= \biE_\tau(-m-1,-1) \\
&= (-1)^n \sum_{(a_e) \in \gG_{G_\tau}}
\binom{a_0-1}{a_0} \prod_{e \in E}
\binom{-m + a_e - 2}{a_e},
\end{aligned} \]
%
where the sum ranges over all $G_\tau$-draconian
sequences $(a_e) \in \NN^{E \sqcup \{0\}}$
characterized in Lemma~\ref{lem:tau-drac}. Since
the summand is nonzero only for $a_0 = 0$, we
reach the formula
%
\[ \Ehr(\qQ_\tau, m) = \sum_{(a_e)} \prod_{e \in E}
\binom{m + 1}{a_e}, \]
%
where the sum ranges over all points $\mathbf{a} =
(a_e) \in P_{\tau^\ast}$ such that $|\mathbf{a}|=n$
(these are the maximal elements of $P_{\tau^\ast}$).
Since (see, for instance, \cite[Section~4]{AXY26})
%
\[ \sum_{m \ge 0} \left( \, \prod_{e \in E}
\binom{m+1}{a_e} \right) t^m =
\frac{\displaystyle\sum_{w \in
\mathfrak{S}_{\mathbf{a}}}
t^{n-1-\des(w)}}{(1-t)^{n+1}} \]
%
for every such $\mathbf{a} \in P_{\tau^\ast}$,
our formula for $\Ehr(\qQ_\tau, m)$ is equivalent
to Conjecture~\ref{conj:h-ast}.
\qed
\end{remark}
\begin{remark} \rm
There are nonisomorphic preorders with equal
double Ehrhart polynomials. The smallest pair
for which this coincidence occurs is
%
\begin{figure}[!h]
\centering
\begin{tikzpicture}
\tikzstyle{every node}=[draw,shape=circle, thick]
\node at (0,0) (a) {$a$};
\node at (1,0) (bc) {$b,c$};
\end{tikzpicture}
\quad\text{and}\quad
\begin{tikzpicture}
\tikzstyle{every node}=[draw,shape=circle, thick]
\node at (0,0) (a) {$a$};
\node at (2,0) (b) {$b$};
\node at (1,1) (c) {$c$};
\path [draw] (a) -- (c);
\path [draw] (b) -- (c);
\end{tikzpicture}.
\end{figure}
\end{remark}
\section{The $M$-triangle}
\label{sec:M-triangles}
This section discusses another enumerative
invariant of the poset of lattice points of a
preorder polytope, namely the $M$-triangle, and
extends in this setting a conjecture of
\cite{Cha25b} on arbor polytopes. This
conjecture provided much of the motivation
behind the study of arbor polytopes; see
\cite[Section~1.1]{Cha25b} for the global
picture, which should extend verbatim to all
preorder polytopes.
Let $P$ be a finite graded poset with rank
function $\rho: P \to \NN$. The
\emph{$M$-triangle} of $P$ is defined by the
formula
%
\[ M_P(x,y) = \sum_{a, b \, \in P : \,
a \preceq b}
\mu_P (a,b) x^{\rho(a)} y^{\rho(b)}, \]
%
where $\mu_P$ is the Möbius function of $P$.
The \emph{transmutation} of a rational
function $f(x,y) \in \QQ(x,y)$ is defined as
%
\[ \overline{f}(x,y) = f \left( \frac{1-y}{1-xy},
1-xy \right). \]
\smallskip
\noindent
We note that transmutation is an involution on
$\QQ(x,y)$ and that $\overline{M}_P(x,y) \in
\ZZ[x,y]$ for every graded poset $P$. The following
conjecture is stated implicitly for arbors in
\cite[Section~1.1]{Cha25b}.
%
\begin{conjecture} \label{conj:lattice}
For every preorder $\tau$ of size $n$ there
exists a finite graded lattice $L_\tau$ of
rank $n$ such that $\zZ(P_\tau,t) = \zZ(L_\tau,
t)$ and
%
\begin{equation} \label{eq:M-transmute}
\overline{M}_{P_\tau}(x,y) = M_{L_\tau}(x,y).
\end{equation}
%
In particular, the rank-generating polynomial
of $L_\tau$ is equal to $h(\tau,t)$.
\end{conjecture}
%
For example, if $\tau$ is the chain preorder
of size $n$ having $n$ vertices of cardinality
1 or one vertex of cardinality $n$, then the
role of $L_\tau$ can be played by the
noncrossing partition lattice of type $\TA_n$
and $\TB_n$, respectively; see Sections~7
and~8 of \cite{Cha25b}. Thus, the poset
$P_\tau$ relates the combinatorics of the
noncrossing partition lattice to that of the
associahedron of corresponding type, the
$h$-polynomial of which is equal to $h(\tau,
t)$. Conjecture~\ref{conj:lattice} predicts
a similar relationship in a much more general
context; we refer the reader to
\cite[Section~1.1]{Cha25b} for an extensive
discussion of this conjectural relationship
in the setting of arbors.
The fact that Conjecture~(\ref{conj:lattice})
implies that the rank-generating polynomial
of $L_\tau$ is equal to $h(\tau,t)$ follows
as in the special case of arbors; see
\cite[Proposition~5.1]{Cha25b}.
Our second conjecture states that the
transmuted $M$-triangle of $P_\tau$ satisfies
a certain duality.
%
\begin{conjecture} \label{conj:M-duality}
We have
%
\[ (xy)^n \, \overline{M}_{P_\tau}(1/x,1/y)
= \overline{M}_{P_\tau}(y,x) \]
%
for every preorder $\tau$ of size $n$.
Equivalently, $M_{L_\tau}(x,y) =
M_{(L_\tau)^\ast}(x,y)$, where $(L_\tau)^\ast$
is the lattice dual to $L_\tau$.
\end{conjecture}
\begin{example} \rm
Let us illustrate these concepts and
conjectures with our running example.
The $M$-triangle can be displayed as a matrix
of coefficients, with the constant term on the
bottom left. For the running example from
Figure~\ref{fig1} in dimension 5 with 92
elements and 18 maximal elements, this matrix
is
\begin{equation*}
\begin{array}{rrrrrr}
-1 & 11 & -43 & 75 & -60 & 18 \\
5 & -35 & 84 & -84 & 30 & \\
-10 & 43 & -57 & 24 & & \\
10 & -24 & 14 & & & \\
-5 & 5 & & & & \\
1 & & & & &
\end{array}
\end{equation*}
where the diagonal counts elements according
to their sum of coordinates. The leftmost column
comes from the unit cube contained in $\qQ_\tau$.
The transmutation of this $M$-triangle is
displayed by the matrix
\begin{equation*}
\begin{array}{rrrrrr}
-18 & 60 & -75 & 43 & -11 & 1 \\
60 & -174 & 180 & -77 & 11 & \\
-75 & 180 & -139 & 34 & & \\
43 & -77 & 34 & & & \\
-11 & 11 & & & & \\
1 & & & & &
\end{array}
\end{equation*}
where the diagonal is now the $h$-vector. The
identity claimed in Conjecture~\ref{conj:M-duality}
is represented by the symmetry of the transmuted
matrix in the main diagonal. Note that the
number of maximal elements becomes, up to sign, the
corner entry of the transmuted matrix.
\end{example}
\section{Examples}
\label{sec:exa}
The context of preorders allows for several
interesting new examples of preorder polytopes,
compared to the more restrictive setting of arbors.
This section studies such examples in some detail.
We let
%
\[ \nara^\TA_n(t) = \sum_{i=0}^n \frac{1}{i+1}
\binom{n}{i}\binom{n+1}{i} t^i \]
%
and
%
\[ \nara^\TB_n(t) = \sum_{i=0}^n \binom{n}{i}^2
t^i \]
%
be the Narayana polynomials of type $\TA_n$ and
$\TB_n$, respectively.
\subsection{Zig-zag posets}
\label{sec:zig-zag}
Given a positive integer $n$, the
\emph{zig-zag poset} $Z_n$ is defined as the
partial order on the set $\{1, 2,\dots,n\}$
with cover relations $i \prec i+1$ for all odd
$i \in \{1, 2,\dots,n-1\}$ and $i \succ i+1$
for all even $i \in \{1, 2,\dots,n-1\}$. Let
us denote by $\qQ_n$ the preorder polytope
and by $h_n(t)$ the $h$-polynomial associated to
$\tau = Z_n$. The $n$th \emph{Delannoy
polynomial} \oeis{A008288}, denoted by $d_n(t)$,
is defined by the recurrence
%
\begin{equation} \label{eq:Delannoy}
d_n(t) = (1+t) d_{n-1}(t) + t d_{n-2}(t)
\end{equation}
%
for $n \ge 2$, with initial conditions $d_0(t)
= 1$ and $d_1(t) = 1+t$; we refer the reader
to \cite{WZC19} and the references given there
for combinatorial interpretations and explicit
formulas for these polynomials. The sum of the
coefficients of $d_n(t)$ satisfies the
recurrence $d_n(1) = 2 d_{n-1}(1) + d_{n-2}(1)$
for $n \ge 2$, where $d_0(1) = 1$ and $d_1(1) =
2$, which yields the Pell numbers \oeis{A000129}.
%
\begin{proposition} \label{prop:zig-zag}
We have $h_n(t) = d_n(t)$ for every $n \ge 1$.
In particular, the number of lattice points of
$\qQ_n$ is equal to the Pell number $d_n(1)$.
\end{proposition}
\begin{proof}
Let $n \ge 4$ be even. The set of lattice
points $(a_1, a_2,\dots,a_n)$ of $\qQ_n$ can
be expressed as the disjoint union of nonempty
subsets by the following conditions:
%
\begin{itemize}
\item $a_n=0$,
\item $a_n=1$,
\item $a_n=2$, which implies $a_{n-1}=0$.
\end{itemize}
%
As one can easily verify, these sets are in
one-to-one correspondence with the set of
lattice points of $\qQ_{n-1}$, $\qQ_{n-1}$
and $\qQ_{n-2}$, respectively, so that
%
\[ h_n(t) = (1+t) h_{n-1}(t) + th_{n-2}(t). \]
Similarly, the set of lattice points $(a_1,
a_2,\dots,a_{n+1})$ of $\qQ_{n+1}$ can be
expressed as the disjoint union of nonempty
subsets by the following conditions:
%
\begin{itemize}
\item $a_{n+1} = 1$,
\item $a_{n+1} = 0$ and $a_n = 0$,
\item $a_{n+1} = 0$ and $a_n = 1$,
\item $a_{n+1} = 0$ and $a_n = 2$,
\item $a_{n+1} = 0$ and $a_n = 3$, which
implies $a_{n-1} = 0$.
\end{itemize}
%
These sets are in one-to-one correspondence
with the set of lattice points of $\qQ_n$,
$\qQ_{n-1}$, $\qQ_{n-1}$, $\qQ_{n-1}$ and
$\qQ_{n-2}$, respectively, so
%
\[ h_{n+1}(t) = t h_n(t) + (1+2t) h_{n-1}(t)
+ t h_{n-2}(t) = (1+t) h_n(t) + t
h_{n-1}(t). \]
%
Setting $h_0(t) = 1$, and since $h_1(t) =
1+t$, $h_2(t) = 1+3t+t^2$ and $h_3(t) =
1+5t+5t^2+t^3$, we have shown that $h_n(t)$
satisfies the recurrence (\ref{eq:Delannoy})
for $n \ge 2$ and the proof follows.
\end{proof}
The Delannoy polynomial $d_n(t)$ is known to
have all properties stated in
Conjectures~\ref{conj:hvector} -
\ref{conj:roots}. The palindromicity of $d_n
(t)$ follows directly from~(\ref{eq:Delannoy});
its $\gamma$-positivity and real-rootedness
was shown in \cite{WZC19}. The polynomial
$d_n(t)$ was shown to be equal to the
$h$-polynomial of a flag simplicial polytope
(whose boundary complex is isomorphic to the
independence complex of a planar ternary
graph) in \cite{BDHKY26}. We suspect that
$d_n(t)$ is also equal to the $h$-polynomial
of the $n$-dimensional
pellytope~\cite{pellytopes} \cite[\S 4]{HLRZ20};
see \cite[Conjecture~6.1.36]{almeter} for a
related conjecture.
This example is also related to the lattices
of sashes~\cite{Law13,Law14}.
%% For Almeter, see also https://emis.de/ft/26209
%% In this Pell case, it often happens that the
%% $M$-triangle factorizes in
%% non-trivial ways. The factors could be
%% interpreted as the
%% $M$-triangles of preorders modulo a symmetry
%% (someting like type $\TB$-associahedra as
%% middle slice of type $\TA$-associahedra).
\subsection{Ordinal sums of antichains}
\label{sec:antichains}
For positive integers $m,n$ we will denote by
$\tau_{m,n}$ the set $\{1, 2,\dots,m\} \sqcup
\{1^\prime, 2^\prime,\dots,n^\prime\}$ partially
ordered so that the cover relations are $i \prec
j^\prime$ for all $i,j$. Thus, $\tau_{m,n}$ is
the ordinal sum of an $m$-element antichain and
an $n$-element antichain. Let us denote by
$\qQ_{m,n}$ the preorder polytope and by
$h_{m,n}(t)$ the $h$-polynomial associated to
$\tau_{m,n}$.
%
\begin{proposition} \label{prop:antichains}
The number of lattice points of $\qQ_{m,n}$ is
equal to
%
\[ \sum_{k=0}^m \sum_{\ell=0}^n
\binom{k+\ell}{k} \binom{m}{k} \binom{n}{\ell}
\]
%
and
%
\[ \begin{aligned}
h_{m,n}(t) &= \sum_{k=0}^m \sum_{\ell=0}^n
\binom{k+\ell}{k} \binom{m}{k} \binom{n}{\ell}
t^{m-k+\ell} \\
&= \sum_{i \ge 0} \binom{m}{i} \binom{n}{i}
t^i (1+t)^{m+n-2i}
\end{aligned} \]
%
for all $m, n$. In particular, $h_{m,n}(t)$
is palindromic and $\gamma$-positive. Moreover,
all roots of $h_{m,n}(t)$ are real.
\end{proposition}
\begin{proof}
Clearly, it suffices to prove the proposed
formulas for $h_{m,n}(t)$. By definition, the
lattice points of $\qQ_{m,n}$ are the tuples
$(a_1, a_2,\dots,a_m, b_1, b_2,\dots,b_n) \in
\NN^{m+n}$ such that $a_i \in \{0, 1\}$ for all
$1 \le i \le n$ and
%
\begin{equation} \label{eq:ab}
b_{j_1} + b_{j_2} + \cdots + b_{j_r} + a_1 +
a_2 + \cdots + a_m \le m+r
\end{equation}
%
for all $1 \le j_1 < j_2 < \cdots < j_r \le n$.
Suppose that we have $a_i = 1$ for $k$ indices
$i \in \{1, 2,\dots,m\}$ and $b_j \ge 1$ for
$\ell$ indices $j \in \{1, 2,\dots,n\}$ and
set $c_j = b_j - 1$ for such indices $j$. By
symmetry, we may assume that $b_1,
b_2,\dots,b_\ell \ge 1$. Then, the inequalities
(\ref{eq:ab}) become
%
\[ c_{j_1} + c_{j_2} + \cdots + c_{j_r} \le
m-k \]
%
for $1 \le j_1 < j_2 < \cdots < j_r \le \ell$
and are thus equivalent to the single
inequality
%
\[ c_1 + c_2 + \cdots + c_\ell \le m-k. \]
%
As a result, there are $\binom{m}{k}$ ways
to choose $a_1, a_2,\dots,a_m \in \{0,1\}$ and
for every such choice there are
$\binom{n}{\ell}\binom{m-k+\ell}{\ell}$ ways
to choose $b_1, b_2,\dots,b_n$. Therefore,
%
\[ h_{m,n}(t) = \sum_{k=0}^m \sum_{\ell=0}^n
\binom{m-k+\ell}{\ell} \binom{m}{k}
\binom{n}{\ell} t^{k+\ell}. \]
%
This formula is equivalent to the first formula
proposed for $h_{m,n}(t)$. Since the right-hand
side is equal to the rank-generating polynomial
of the poset of shuffles of $(1, 2,\dots,m)$
and $(1^\prime, 2^\prime,\dots,n^\prime)$
\cite{Gr88}, the second formula proposed for
$h_{m,n}(t)$ follows from
\cite[Corollary~4.8]{Gr88}.
Finally, we observe that $\sum_{i \ge 0}
\binom{m}{i} \binom{n}{i} t^i$ is real-rooted as
a Hadamard product of real-rooted polynomials
\cite[V~155]{PS72}. By \cite[Remark~3.1.1]{Ga05},
so is $h_{m,n}(t)$.
\end{proof}
The enumerative combinatorics of the shuffle
poset \cite{Gr88} is closely related to that
of the bubble lattice for the same parameters,
introduced and studied in~\cite{MM22,MM24,MM25}.
A shellable triangulation of a sphere with
$h$-polynomial equal to $h_{m,n}(t)$ is the
noncrossing bipartite complex of
\cite[Section~4]{MM25}; see also
\cite[Remark~4.16]{MM25}.
% See also ~\cite{shuffle_bubble}, cf Remark 5.3.
\subsection{Grid posets}
\label{sec:grid}
Let us consider the direct product $C_n \times
C_2$ of an $n$-element chain and a two-element
chain, for positive integers $n$. For small
values of $n$, the number of lattice points of
the associated preorder polytope and the
$h$-vector are as follows:
\bigskip
\begin{center}
\begin{tabular}{c c c}
\toprule
$n$ & number of lattice points & $h$-vector\\
\midrule
1 & 5 & $(1, 3, 1)$ \\
2 & 38 & $(1, 9, 18, 9, 1)$ \\
3 & 352 & $(1, 18, 86, 142, 86, 18, 1)$ \\
4 & 3659 & $(1, 30, 260, 882, 1313, 882, 260,
30, 1)$ \\
\bottomrule
\end{tabular}
\end{center}
\bigskip
\noindent
This suggests that the number of lattice points
is given by sequence \oeis{A158266}. For the
poset obtained from $C_n \times C_2$ by removing
the minimum and maximum elements we have
\bigskip
\begin{center}
\begin{tabular}{c c c}
\toprule
$n$ & number of lattice points & $h$-vector\\
\midrule
1 & 4 & $(1, 2, 1)$ \\
2 & 29 & $(1, 7, 13, 7, 1)$ \\
3 & 265 & $(1, 15, 65, 103, 65, 15, 1)$ \\
4 & 2745 &
$(1, 26, 206, 659, 961, 659, 206, 26, 1)$ \\
\bottomrule
\end{tabular}
\end{center}
\bigskip
\noindent
which suggests that the number of lattice points
is given by sequence \oeis{A370955}. Curiously,
the generating functions
%
\[ 1 + x + 5 x^2 + 38 x^3 + 352 x^4 + 3659 x^5
+ \cdots \]
%
and
%
\[ 1 - x - 4 x^2 - 29 x^3 - 265 x^4 - 2745 x^5
- \cdots \]
%
of these sequences are inverses of each other.
This seems to extend to the generating
functions of the $h$-polynomials as follows.
Setting
%
\[ \begin{aligned}
F(t,x) &= 1 + x + \left(t + 3 + t^{-1} \right)
x^2 + \left(t^2 + 9t + 18 + 9t^{-1} +
t^{-2} \right) x^3 + \cdots \\
G(t,x) &= 1 - x - \left(t + 2 + t^{-1} \right)
x^2 - \left(t^2 + 7t + 13 + 7t^{-1}
+ t^{-2} \right) x^3 - \cdots
\end{aligned} \]
%
we suspect that $F(t,x) = 1/G(t,x) =
\exp(S(t,x))$, where
%
\begin{equation*}
s_n(t) = \sum_{i=0}^{n-1} \binom{n}{i}
\binom{n - 1}{i} t^i \quad \text{and} \quad
S(t,x) = \sum_{n \geq 1} s_n(t) s_n(t^{-1}) x^n/n.
\end{equation*}
For the poset obtained from $C_n \times C_2$ by
removing the maximum element we have
\bigskip
\begin{center}
\begin{tabular}{c c c}
\toprule
$n$ & number of lattice points & $h$-vector\\
\midrule
1 & 2 & $(1, 1)$ \\
2 & 12 & $(1, 5, 5, 1)$ \\
3 & 100 & $(1, 12, 37, 37, 12, 1)$ \\
4 & 980 &
$(1, 22, 138, 329, 329, 138, 22, 1)$ \\
\bottomrule
\end{tabular}
\end{center}
\bigskip
\noindent
which suggests that the number of lattice
points of the associated preorder polytope is
given by sequence \oeis{A000888} and that the
$h$-polynomial is equal to $\nara^{\TA}_{n-1}
(t) \nara^{\TB}_n(t)$.
\subsection{Double chains}
\label{sec:double-chains}
Let us consider the chain preorder having $n$
vertices, each of cardinality 2. For small
values of $n$, we have:
\bigskip
\begin{center}
\begin{tabular}{c c c}
\toprule
$n$ & number of lattice points & $h$-vector\\
\midrule
1 & 6 & $(1, 4, 1)$ \\
2 & 53 & $(1, 12, 27, 12, 1)$ \\
3 & 554 & $(1, 24, 134, 236, 134, 24, 1)$ \\
4 & 6362 & $(1, 40, 410, 1540, 2380, 1540,
410, 40, 1)$ \\
\bottomrule
\end{tabular}
\end{center}
\bigskip
\noindent
The lattice path interpretation of the lattice
points described (for all chain preorders) in
\cite[Section~2]{Ath26} shows that the number of
lattice points is given by sequence
\oeis{A066357}. The generating function
%
\[ 1 + x + 6 x^2 + 53 x^3 + 554 x^4 + 6362 x^5 +
\cdots \]
%
of the number of lattice points is the inverse of
%
\[ 1 - \sum_{n \ge 1} \frac{1}{2n} \binom{4n-2}{2n-1}
x^n = 1 - x - 5 x^2 - 42 x^3 - 429 x^4 - 4862 x^5
- \cdots \]
%
and the generating function
%
\[ 1 + x + \left(t + 4 + t^{-1} \right) x^2 +
\left(t^2 + 12t + 27 + 12t^{-1} + t^{-2}
\right) x^3 + \cdots \]
%
of the $h$-polynomial is the inverse of $1 -
\sum_{n \ge 1} \nara^\TA_{2n-2}(t) \, t^{-n+1}
x^n$.
\subsection{Multiple chains}
\label{sec:multiple}
Let us consider the chain preorder $\tau_{n,k}$
having $n$ vertices, each of cardinality $k$.
Experimental evidence suggests that
%
\[ 1 + \sum_{n \ge 1} h(\tau_{n,k},t^2) \, t^{-nk}
x^n = \exp \left( \, \sum_{n \ge 1}
\nara^\TB_{kn}(t^2) \, t^{-kn} \, \frac{x^n}{n}
\right). \]
%
When $k$ is even, one can replace $t$ by $t^{1/2}$
and still get Laurent polynomials in $t$. For
$k=2$ this gives a different interpretation of
the $h$-polynomial from that of
Section~\ref{sec:double-chains}; see \oeis{A181145}.
% exp(sum(binomial(4*n,2*n)/n*x**n))
\subsection{Comb posets}
\label{sec:comb}
Let $\tau_n$ be the poset defined recursively
for $n \in \NN$ as follows. The poset $\tau_0$
is empty and $\tau_{n+1}$ is obtained from $\tau_n$
by first taking the disjoint union with a poset
with one element and then adding a new minimum
element. For small values of $n$, we have:
\bigskip
\begin{center}
\begin{tabular}{c c c}
\toprule
$n$ & number of lattice points & $h$-vector\\
\midrule
1 & 5 & $(1, 3, 1)$ \\
2 & 33 & $(1, 8, 15, 8, 1)$ \\
3 & 249 & $(1, 15, 61, 95, 61, 15, 1)$ \\
4 & 2033 & $(1, 24, 166, 484, 683, 484, 166,
24, 1)$ \\
\bottomrule
\end{tabular}
\end{center}
\bigskip
\noindent
which suggests that the number of lattice
points of the associated preorder polytope
is given by sequence \oeis{A034015}. The
generating function
%
\[ x + 5 x^2 + 33 x^3 + 249 x^4 + 2033 x^5 +
\cdots \]
%
of this sequence is the compositional inverse
of
%
\[ \sum_{n \ge 1} (-1)^{n-1} \left( 1 + 2^n
(n-1) \right) x^n = x - 5 x^2 + 17 x^3 -
49 x^4 + 129 x^5 - \cdots, \]
%
see \oeis{A000337}. This leads us to suspect
that the generating function of the
$h$-polynomial
%
\[ x + \left(t + 3 + t^{-1} \right) x^2 +
\left(t^2 + 8t + 15 + 8t^{-1} + t^{-2}
\right) x^3 + \cdots \]
%
is the compositional inverse of
%
\[ \sum_{n \ge 1} (-1)^{n-1} \left( 1 +
{(1+t)}^n (1+t+\cdots+t^{n-2}) \, t^{-n+1}
\right) x^n. \]
%%\bibliographystyle{alpha} \bibliography{preorder}
\begin{thebibliography}{ALOGV24}
%
\bibitem[Alm22]{almeter}
J.~Almeter,
\textit{$P$-graph associahedra and hypercube
graph associahedra},
Ph.D thesis, North Carolina State University, 2022,
{\tt arXiv:2211.02113}.
%
%%\bibitem[Ath18]{Ath18}
%%C.A.~Athanasiadis,
%%\textit{Gamma-positivity in combinatorics and geometry},
%%S\'em. Lothar. Combin. {\bf~77} (2018), Article B77i,
%%64pp (electronic).
%
\bibitem[Ath25]{Ath26}
C.A.~Athanasiadis,
\textit{Lattice point enumeration of polytopes
associated to integer compositions},
Ann. Comb. (to appear),
{\tt arXiv:2510.23903}.
%
\bibitem[AFM26]{AFM26}
N.~Avila, L.~Ferroni and A.H.~Morales,
\textit{Luck and magic for Pitman--Stanley polytopes
and parking functions},
{\tt arXiv:2603.19194}.
%
\bibitem[AXY26]{AXY26}
C.A.~Athanasiadis, Q.~Xiao and X.~Yan,
\textit{Lattice point enumeration of some arbor
polytopes},
{\tt arXiv: 2603.11654}.
%
\bibitem[BDHKY26]{BDHKY26}
M.~Bayer, R.~Danner, T.~Holleben, M.~Kramer and
Y.~Yang,
\textit{Planar ternary graphs, flag spheres, and
Delannoy polynomials},
{\tt arXiv:2509.21705}.
%
\bibitem[BLL98]{BLL98}
F.~Bergeron, G.~Labelle and P.~Leroux,
Combinatorial species and tree-like structures,
Transl. from the French by Margaret Readdy,
Encyclopedia of Mathematics and its
Applications {\bf~67},
Cambridge University Press, 1998.
%
\bibitem[BTT25]{pellytopes}
L.~Bossinger, M.L.~Telek and H.~Tillmann-Morris,
\textit{Binary geometries from pellytopes},
Mathematische {\bf~80} (2025), 189--209.
%
\bibitem[Bra06]{Bra06}
P.~Br\"and\'en,
\textit{On linear transformations preserving the
P\'olya frequency property},
Trans. Amer. Math. Soc. {\bf~358} (2006),
3697--3716.
%
\bibitem[Cha25a]{Cha25a}
F.~Chapoton,
\textit{On a $q$-analogue of the Zeta polynomial
of posets},
{\tt arXiv:2402.11979v2}.
%
\bibitem[Cha25b]{Cha25b}
F.~Chapoton,
\textit{On posets and polytopes attached to arbors},
Math. Scand. {\bf~131} (2025), 401--449.
%
\bibitem[FH24]{FH24}
L.~Ferroni and A.~Higashitani,
\textit{Examples and counterexamples in Ehrhart
theory},
EMS Surv. Math. Sci. (to appear),
{\tt arXiv:2307.10852}.
%
\bibitem[Fu71]{fulkerson}
D.R.~Fulkerson,
\textit{Blocking and anti-blocking pairs of
polyhedra},
Math. Program. {\bf~1} (1971), 168--194.
%
\bibitem[Gal05]{Ga05}
S.R.~Gal,
\textit{Real root conjecture fails for five-
and higher-dimensional spheres},
Discrete Comput. Geom. {\bf~34} (2005), 269--284.
%
\bibitem[GS06]{GS06}
I.M.~Gessel and S.~Seo,
\textit{A refinement of Cayley's formula for trees},
Electron. J. Combin. {\bf~11} (2006),
Paper no. R27, 23pp (electronic).
%
\bibitem[Gre88]{Gr88}
C.~Greene,
\textit{Posets of shuffles},
J. Combin. Theory Series A {\bf~47} (1988), 191--206.
%
\bibitem[Gru07]{gruber}
P.M.~Gr\"uber,
Convex and Discrete Geometry,
Grundlehren der Mathematischen Wissenschaften
{\bf~336}, Springer, 2007.
%
\bibitem[HLRZ20]{HLRZ20}
S.~He, Z.~Li, P.~Raman and C.~Zhang,
\textit{Stringy canonical forms and binary
geometries from associahedra, cyclohedra and
generalized permutohedra},
J. High Energy Phys. {\bf~2020} (2020),
Paper No. 54, 35pp.
%
\bibitem[Hi92]{HiAC}
T.~Hibi,
Algebraic Combinatorics on Convex Polytopes,
Carslaw Publications, Australia, 1992.
%
\bibitem[Law13]{Law13}
S.~Law,
\textit{Combinatorial realization of certain
Hopf algebras of pattern-avoiding permutations},
Ph.D thesis, University of Michigan, 2013.
%
\bibitem[Law14]{Law14}
S.~Law,
\textit{Combinatorial realization of the Hopf
algebra of shashes},
in \emph{Proceedings of the 26th International
Conference on Formal Power Series and Algebraic
Combinatorics (FPSAC 2014), Chicago, IL, USA,
June 29 -- July 3, 2014}, pp.~621--632,
Discrete Mathematics and Theoretical Computer
Science, Nancy, 2014.
%
\bibitem[Liu19]{Liu19}
F.~Liu,
\textit{On positivity of Ehrhart polynomials},
in \emph{Recent Trends in Algebraic Combinatorics},
pp.~189--237,
Association for Women in Mathematics Series
{\bf~16}, Springer, 2019.
%
\bibitem[Mac04]{Mac04}
P.A.~MacMahon,
Combinatorial analysis, Vol.~II,
Dover Publications, 2004.
%
\bibitem[MM22]{MM22}
T.~McConville and H.~M\"uhle,
\textit{Shuffle lattices and bubble lattices},
S\'em. Lothar. Combin. {\bf~86B} (2022), Article 63,
12pp (electronic).
%
\bibitem[MM24]{MM24}
T.~McConville and H.~M\"uhle,
\textit{Bubble lattices I: Structure},
Algebra Univers. {\bf~85} (2024),
Paper No. 12, 31pp.
%
\bibitem[MM25]{MM25}
T.~McConville and H.~M\"uhle,
\textit{Bubble lattices II: Combinatorics},
Ann. Comb. {\bf~29} (2025), 657--690.
%
\bibitem[PS72]{PS72}
G.~P\'olya and G.~Szego,
Problems and Theorems in Analysis, vol.~2,
Springer--Verlag, 1972.
%
\bibitem[Pos09]{post09}
A.~Postnikov,
\textit{Permutohedra, associahedra, and beyond},
Int. Math. Res. Not. {\bf~2009} (2009),
1026--1106.
%
\bibitem[Sch86]{schrijver}
A.~Schrijver,
Theory of Integer and Linear Programming,
Wiley-Intersc. Ser. Discrete Math.,
Wiley, 1986.
%
\bibitem[SP02]{SP02}
R.P.~Stanley and J.~Pitman,
\textit{A polytope related to empirical distributions,
plane trees, parking functions and the associahedron},
Discrete Comput. Geom. {\bf~27} (2002), 603--634.
%
\bibitem[Sta96]{StaCCA}
R.P.~Stanley,
Combinatorics and Commutative Algebra,
Progress in Mathematics {\bf~41},
Birkh\"auser, second edition, 1996.
%
\bibitem[Sta11]{StaEC1}
R.P.~Stanley,
Enumerative Combinatorics, vol.~1,
Cambridge Studies in Advanced Mathematics {\bf~49},
Cambridge University Press, second edition,
Cambridge, 2011.
%
\bibitem[WZC19]{WZC19}
Y.~Wang, S-N.~Zheng and X.~Chen,
\textit{Analytic aspects of Delannoy numbers},
Discrete Math. {\bf~342} (2019), 2270--2277.
\end{thebibliography}
\end{document}
%%
%%Let us consider the finest Minkowski decomposition
%%of $Q_{\tau}^{m, h}$, where each summand $U_v$ is
%%further decomposed as a Minkowski sum of $|v|$
%%equal terms. In this decomposition, each summand
%%is the convex hull in $\ZZ^E$ of $0$ and some
%%vectors ${\{b_i\}}_{i \in I}$ where $I$ is a
%%non-empty subset of $E$. This summand is
%%therefore the polytope $\bDelta_I$ in $\ZZ^E$, to
%%which one associates a standard simplex $\Delta_I$
%%in $\ZZ^{\{0\} \sqcup E}$ by replacing the vertex
%%$0$ by the basis vector $b_0$. Let us consider the
%%linear map $\rho$ from $\ZZ^{\{0\} \sqcup E}$ to
%%$\ZZ^E$ that forgets the first coordinate. Then
%%$\rho$ maps $\Delta_I$ to $\bDelta_I$ and is
%%moreover a unimodular equivalence from $\Delta_I$
%%to $\bDelta_I$.
%%
%%Let us write $Q_{\tau}^{m, h} = \sum_{s \in S}
%%\bDelta_{I_s}$ and $\overline{Q}_{\tau}^{m, h} =
%%\sum_{s \in S} \Delta_{I_s}$ over the appropriate
%%indexing set $S$. Then $\rho$ induces an
%%invertible integral affine map from the hyperplane
%%$H$ in $\ZZ^{\{0\}\sqcup E}$ where coordinates sum
%%to $|S|$ to $\ZZ^E$, that maps $Q_{\tau}^{m, h}$
%%to $\overline{Q}_{\tau}^{m, h}$. This is the
%%expected unimodular equivalence.
%%\begin{proposition}
%%Every element of $V_\tau$ is a vertex of
%%$\rR^\vee_\tau$. The lattice polytope
%%$\rR^\vee_\tau$ has no lattice points other than
%%$\mathbf{0}$ and its vertices.
%%\end{proposition}
%%\begin{proof}
%%We will show that every lattice point of
%%$\rR^\vee_\tau$ which is a convex combination
%%of elements of $V_\tau$ is either an element
%%of $V_\tau$, as a one-term convex combination, or
%%equal to $\mathbf{0}$.
%%Let us denote by $J(\tau)$ the set of order
%%ideals of $\tau$ and let $\mathbf{a} = (a_e) \in
%%\NN^E$ be a lattice point of $\rR^\vee_\tau$ such
%%that
%
%%\[ \mathbf{a} = \sum_{e \in E} c_e
%% (-{\rm \mathbf{e}}_e) + \sum_{\iI \in J(\tau)}
%% d_\iI \sum_{e \in \iI} {\rm \mathbf{e}}_e \]
%
%%for some nonnegative real numbers $c_e$ and
%%$d_\iI$ satisfying
%
%%\begin{equation} \label{eq:bary}
%%\sum_{e \in E} c_e + \sum_{\iI \in J(\tau)}
%%d_\iI = 1.
\end{equation}
%
%%Then,
%
%%\begin{equation} \label{eq:coordi}
%%a_e = - c_e + \sum_{\iI \in J(\tau) : \, e \in \iI}
%%d_\iI
%%\end{equation}
%
%%for every $e \in E$. Since $0 \le c_e \le 1$
%%and $0 \le \sum_{\iI \in J(\tau) : \, e \in \iI}
%%d_\iI \le 1$, we must have $a_e \in \{-1,0,1\}$
%%for every $e \in E$.
%%Let us first assume that there exists $e \in E$
%%such that $a_e = -1$. Then, $c_e = 1$ by
%%Equation~(\ref{eq:coordi}) and all other
%%coefficients must vanish by~(\ref{eq:bary}),
%%so that $\mathbf{a} = -{\rm \mathbf{e}}_e$.
Finally, we assume that $a_e \in \{0, 1\}$ for
every $e \in E$. We may assume further that $a_f
= 1$ for some $f \in E$ (otherwise, $\mathbf{a}
= \mathbf{0}$).
Equations~(\ref{eq:bary}) and~(\ref{eq:coordi})
then imply that $c_e = 0$ for every $e \in E$
and $d_\iI = 0$ for every $\iI \in J(\tau)$
which does not contain $f$. In particular,
$\mathbf{a}$ can be expressed as a convex
combination
%
\[ \mathbf{a} = \sum_{\iI \in J(\tau)}
d_\iI \sum_{e \in \iI} {\rm \mathbf{e}}_e \]
%
with
%
\[ \sum_{\iI \in J(\tau)} d_\iI = 1. \]
%
As a result, the set of $e \in E$ such that
$a_e \ne 0$ is equal to the union of all order
ideals $\iI$ of $\tau$ such that $d_\iI > 0$
and thus is an order ideal of $\tau$, say $\jJ$,
itself. In particular, $\mathbf{a} =
\sum_{e \in \jJ} {\rm \mathbf{e}}_e$ and we
claim that $d_\iI = 0$ for every order ideal
$\iI$ of $\tau$ other than $\jJ$. Indeed,
suppose that $d_{\iI_0} > 0$ for some order ideal
$\iI_0$ of $\tau$ other than $\jJ$. We then
have $\iI_0 \subset \jJ$ by definition of $\jJ$
and for every $e \in \jJ \smallsetminus \iI_0$
%
\[ a_e = \sum_{\iI \in J(\tau) : \, e \in \iI}
d_\iI \le 1 - d_{\iI_0} < 1, \]
%
contrary to the fact that $a_e = 1$. This
contradiction verifies our claim and the proof
follows.
\end{proof}
%%cf Rietsch + Williams for this terminology :
%%https://arxiv.org/pdf/2406.15803