Convex expansion for finite distributive lattices with applications
Extracted source text · arXiv 1810.06762
\documentclass[UTF8]{article}
\usepackage{graphicx}
\usepackage{hyperref}
\usepackage{amssymb}
\usepackage[thmmarks]{ntheorem}
\usepackage[tbtags]{amsmath}
\usepackage{caption}
\usepackage{subcaption}
\usepackage{tikz} %画个平行四边形符号
\usetikzlibrary{calc,arrows}
\usepackage[bottom=25mm,left=25mm,right=25mm,top=25mm]{geometry}
{ % define difinition environment
\theoremstyle{nonumberplain}
\theoremheaderfont{\bfseries}
\theorembodyfont{\normalfont}
\theoremsymbol{$\square$}
\newtheorem{proof}{Proof}
}
%\hyphenpenalty=1000
\tolerance=1000
\renewcommand{\baselinestretch}{1.25}
\allowdisplaybreaks[3]
\newtheorem{definition}{Definition}[section]
\newtheorem{theorem}{Theorem}[section]
\newtheorem{corollary}[theorem]{Corollary}
\newtheorem{lemma}[theorem]{Lemma}
\newtheorem{proposition}[theorem]{Proposition}
\newtheorem{remark}[theorem]{Remark}
\begin{document}
\title{Convex expansion for finite distributive lattices with applications\thanks{This work was supported by NSFC (Grant No. 11761064).}
}
\author{Xu Wang, Xuxu Zhao and Haiyuan Yao\footnote{Corresponding author.}%
\\ {\footnotesize College of Mathematics and Statistics, Northwest Normal University, Lanzhou 730070, PR China}}
\date{}
\maketitle
\begin{abstract}
The concept of cutting is first explicitly introduced. By the concept, a convex expansion for finite distributive lattices is considered. Thus, a more general method for drawing the Hasse diagram is given, and the rank generating function of a finite distributive lattice is obtained.
In addition, we have several enumerative properties on finite distributive lattices and verify the generalized Euler formula for polyhedrons.
\textbf{Key words:} finite distributive lattice; cutting (sublattice); convex expansion; filter lattice; enumerative property
\textbf{2010 AMS Subj. Class.:} 06D05; 06A07; 06B05 %%MSCclass
\end{abstract}
\section{Introduction}
Day introduced a doubling construction in a proof in \cite{aDay1970}. Using the construction, J\'onsson \cite{aJonss1971} considered the free products of lattices, Day \cite{aDay1977} provided an affirmative answer to a McKenzie's problem \cite{aMcKen1972}, Davey et al.\ \cite{aDaveyDQR1978} studied the exponents of lattices for finite distributive lattices, and Siv\'ak \cite{aSivak1978} characterized a class of finite lattices put by Schein \cite{aSchei1972}. Day et al.\ \cite{aDayGP1979} obtained a result of the construction on distributive lattice. In addition, Day et al.\ \cite{aDayNT1989} found various applications of doubling construction in the study of finite lattices etc., and Day \cite{aDay1992} obtained some results for the construction. Geyer \cite{aGeyer1993} extended the construction to obtain a lattice, and Nation \cite{aNatio1995} surveyed the doubling construction. Recently, Ern\'{e} et al.\ \cite{aErneHR2002} obtained a convex expansion for finite distributive lattices by the construction. Bertet et al.\ \cite{aBerteC2002} characterized the doubling convex sets in lattices and got a recognition algorithms.
In the paper a cutting of a finite distributive lattice is first explicitly introduced; in fact, Day \cite{aDay1992,aDayGP1979} already had the thought on cutting and got some results, we give two equivalent characterizations by the fundamental theorem for finite distributive lattices \cite{aBirkh1937}.
Using the concept of cutting, we obtain a convex expansion for finite distributive lattices; moreover, the expansion extends a result for finite distributive lattices in \cite{aErneHR2002} and the method for drawing the Hasse diagram of a filter lattice in \cite[P293]{bStanl2011}.
Using the convex expansion for finite distributive lattices, we have a series of consequences, including the rank generating function \cite{bStanl2011}, the number of convex Boolean sublattices of finite distributive lattice, and the number of particular elements; especially, the conclusions of Fibonacci and Lucas cubes \cite{aHsuPL1993,aMunarCS2001}. Finally, we verify the generalized Euler formula for polyhedrons by the relation of convex Boolean sublattices and antichains of posets.
\section{Preliminaries}
A set $P$ equipped with a partial order relation $\le$ is said to be a \emph{partially ordered set} (poset for short).
We write $x\parallel y$ if $x\nleq y$ and $y\ngeq x$.
A subposet $S$ of $P$ is a \emph{chain} if any two elements of $S$ are comparable, and denoted by $\mathbf{n}$ if $|S|=n$ \cite{bDaveyP2002}. The chain $S$ of $P$ is called \emph{saturated} if there does not exist $u \in P \setminus S$ such that $s < u < t$ for some $s, t \in S$ and such that $S\cup \{u\}$ is a chain \cite{bStanl2011}.
The set consisting of all minimal (resp.\ maximal) elements of $P$ is denoted by $\operatorname{Min}P$ (resp.\ $\operatorname{Max}P$).
Let $x\prec y$ denote $y$ \emph{cover}s $x$ in $P$, i.e.\ $x<y$ and $x\le z < y$ implies $z=x$.
The subset $Q$ of the poset $P$ is called \emph{convex} if $a,b \in Q$, $c \in P$, and $a \le c \le b$ imply that $c \in Q$.
The set of all filters of a poset $P$ is denoted by $\mathcal{F}(P)$, and carries the usual anti-inclusion order, forms a finite distributive lattice called \emph{filter lattice}.
Note that if $P=\emptyset$, then $\mathcal{F}(P) = \mathbf{1}$.
For a finite lattice $L$, we denote by $\hat0_L$ (resp. $\hat1_L$) the minimum (resp. maximum) element in $L$.
For a finite distributive lattice $L$, let $\mathop{\mathrm{Mi}}(L)$ denote the set of all meet-irreducible elements, regarded as a poset under the ordering of $L$ \cite{bGraet2011}.
For some concepts and notations not explained in the paper, refer to \cite{bDaveyP2002,bGraet2011,bStanl2011}.
\begin{theorem}[\cite{aBirkh1937,bStanl2011}]\label{th:ftfdl}
Let $L$ be a finite distributive lattice. Then there is a unique (up to isomorphism) poset $P$ for which $L\cong \mathcal{F}(P)$. In fact $P\cong \mathop{\mathrm{Mi}}(L)$.
\end{theorem}
\section{Convex expansion}
%\begin{definition}
% An interval is cutting if each maximal chain of the lattice has a nonempty intersection with it.
%\end{definition}
\begin{definition}\label{def:cutsub}
Let $L$ be a finite distributive lattice and let $K$ be a convex sublattice, i.e.\ interval, of $L$. We say the lattice $K$ is a \emph{cutting} of $L$ if any maximal chain of $L$ contains at least one element of $K$.
\end{definition}
Especially, if $K=\{a\}$ ($a \ne \hat0_L$ or $\hat1_L$) is a cutting of a finite distributive lattice $L$, then $a$ is called a \emph{cutting element} of $L$ \cite{aYaoZ2015}.
Let $L=\mathcal{F}(P)$ be a finite distributive lattice and let $K$ be a interval of $L$, then $S = \hat{0}_K\setminus \hat1_K$ is a convex subposet of $P$.
In addition, let $S_0:= \mathop{\mathrm{Max}}(P\setminus\hat{0}_K) = \{\,z \in P\setminus\hat{0}_K \mid \hat{0}_K\cup\{z\} \in\mathcal{F}(P)\,\}$, and $S_1:= \mathop{\mathrm{Min}}\hat{1}_K = \{\,y \in \hat{1}_K \mid \hat{1}_K\setminus\{y\} \in\mathcal{F}(P)\,\}$.
Clearly $P={\downarrow S_0} \mathbin{\dot\cup} S \mathbin{\dot\cup} {\uparrow S_1}$ and $z \not> y$ for all $z \in S_0,\,y \in S_1$, where $P\mathbin{\dot\cup}S$ is the disjoint union of two posets $P$ and $S$ \cite{bDaveyP2002}.
Let $K$ be a cutting of a finite distributive lattice $L = \mathcal{F}(P)$.
The poset $P_K:=(P\cup \{x_K\}, \le)$ is defined as follows: $x\le y$ if $x\le y$ in $P$ for all $x,y\in P$; and $x_K\succ z$ for all $z\in S_0$, $x_K\prec y$ for all $y\in S_1$ and $x_K\parallel w$ for all $w\in S$.
This partially order is well-defined by Theorem~\ref{th:cslt}.
Let $P$ be a poset and $x$ is an element not necessarily in $P$. We denote by $P-x$ and $P*x$ the induced subposet obtained from $P\setminus\{x\}$ and $P\setminus({\mathop{\uparrow} x}\cup{\mathop{\downarrow} x})$, respectively. Obviously $x\parallel w$ for every $w\in P*x$, and $P*x \subseteq (P-x)$ for all $x\in P$, moreover $P*x \cong P-x$ if and only if $P \cong (P-x) \mathbin{\dot{\cup}} \{x\}$.
\begin{theorem}\label{th:cslt}
Let $K$ be a interval of a finite distributive lattice $L=\mathcal{F}(P)$. The posets $S$, $S_0$ and $S_1$ are defined as above, then
the following are equivalent:
\begin{enumerate}
\itemsep=0em \parskip=0em
\item[(1)] $L={\uparrow \hat{0}_K}\cup{\downarrow \hat{1}_K}$;
\item[(2)] $K$ is a cutting of $L$;
\item[(3)] $z<y$ for all $z\in S_0$ and $y\in S_1$;
\item[(4)] there exist $x_K$ such that $S = P_K*x_K$.
\end{enumerate}
\end{theorem}
\begin{proof}
It is trivial that (3) $\iff$ (4) and it is not hard to verify (1) $\iff$ (2), we prove (2) $\iff$ (3) in what follows.
Firstly, by contradiction, suppose that $z\in S_0$ (thus $z\notin \hat0_k$) and $y \in S_1 \subseteq \hat{1}_K$ such that $z \parallel y$, thus there exists $M\in K(\cong \mathcal{F}(S))$ such that $M\cup\{z\} \prec M \prec M\setminus\{y\}$, therefore $M\cup \{z\}\setminus \{y\}, \hat{0}_K\cup\{z\}, \hat{1}_K\setminus\{y\} \in \mathcal{F}(P)\setminus K$, it follows that there exists a saturated chain $\{\hat{0}_K\cup\{z\}, \dots, \hat{0}_K\cup\{z\}, M\cup \{z\}\setminus \{y\}, \hat{1}_K\setminus\{y\}, \dots, \hat{1}_K\setminus\{y\}\}$ containing no any element in $K$ (see Figure~\ref{fig:csl}), a contradiction.
Therefore, the partial order on $P_K$ is well-defined.
Conversely, let $P = \hat0_L \prec \cdots \prec M_0 \prec M \prec \cdots \prec \hat1_L=\emptyset$ be a saturated chain in $L$ such that $M_0\setminus M \subseteq S_0$ and $M \cap S_0 = \emptyset$. By the assumption, $\hat1_K = {\uparrow S_1} \subseteq M_0$, thus $\hat1_K \subseteq M$ and $\hat1_K \subseteq M \subseteq \hat0_K$, i.e.\ $M \in K$, thus $K$ is a cutting of $L$.
% Next by the definitions of $S_0$, $S_1$ and the partial order on $P_K$, either $x_K>x$ or $x_K<x$ for every $x\in P\setminus S$ (In fact ${\downarrow x_K} \setminus \{x_K\} = {\downarrow S_0}$ and ${\uparrow x_K} \setminus \{x_K\} = {\uparrow S_1}$). In addition $x_K\parallel w$ for all $w\in S$, thus $S = P*x_K$.
%
% Finally, if $S\cong P*x_K$, then every saturated chain from $P$ to $\emptyset$ in $\mathcal{F}(P)$ do contain at least one element in $K(\cong\mathcal{F}(S))$, that is $K$ is a cutting of $L$.
\end{proof}
\begin{figure}[!htb] %%nocsl.tex
\centering
\begin{tikzpicture}[scale=0.8]
\fill[lightgray] (-1,1) -- (0,2) -- (-1,3) -- (-2,2) -- cycle;
\foreach \i in {0,1,2}
{
\foreach \k in {0,1}
{
\draw (\k-\i,\i+\k) -- (\k-\i+1,\i+\k+1)
(\i-\k,\i+\k) -- (\i-\k-1,\i+1+\k);
}
}
\foreach \i in {0,1,2}
{
\foreach \j in {0,1,2}
{
\filldraw[fill=white] (\i-\j,\i+\j) circle (1.5pt);
}
}
\node at (-1,2) {$K$};
\node[right] at (0,2) {$M$};
\node[right] at (1,1) {$M\cup\{z\}$};
\node[right] at (1,3) {$M\setminus\{y\}$};
\node[right] at (2,2) {$M\cup\{z\}\setminus\{y\}$};
\end{tikzpicture}
\caption{Illustrating proof of Theorem~\ref{th:cslt}}\label{fig:csl}
\end{figure}
Observe that $\mathcal{F}(P_K)$ is a convex expansion of $L$, let $L\boxplus K$ denote the expansion, namely $L\boxplus K \cong \mathcal{F}(P_K)$ (see Figure~\ref{fig:lplk}, where $L\cong \mathcal{F}(P)$, $L\boxplus K \cong \mathcal{F}(P_K)$ and $K \cong \mathcal{F}(P*x_K)$); in general, if $L\boxplus K$ exists, then $K\boxplus L$ does not exist unless $L\cong K$.
Day \cite{aDay1992,aDayGP1979} had obtained a result on the distributivity of $L \boxplus K$.
\begin{lemma}[\cite{aDay1992,aDayGP1979}]
Let $L$ be a finite distributive lattice. If $K \subseteq L$ is a interval, then $L \boxplus K$ ($L[K]$ in \cite{aDay1992,aDayGP1979}) is again distributive if and only if $L={\uparrow \hat{0}_K}\cup{\downarrow \hat{1}_K}$.
\end{lemma}
Moreover, it is easy to prove the following convex expansion for finite distributive lattices.
\begin{theorem}\label{th:cetfdl}
If $K$ is a cutting of a finite distributive lattice $L = \mathcal{F}(P)$, then there exists $x_K$ such that
\[
\mathcal{F}(P_K) \cong \mathcal{F}(P) \boxplus \mathcal{F}(P_K*x_K).
\]
On the other hand, $\mathcal{F}(P)$ is also considered as a convex expansion, that is for every $x\in P$,
\[
\mathcal{F}(P)\cong \mathcal{F}(P-x)\boxplus \mathcal{F}(P*x).
\]
\end{theorem}
\begin{figure}[!ht] %%l&k.tex and lbpk.tex
\centering
\begin{subfigure}[b]{0.45\linewidth}
\centering
\begin{tikzpicture}
\filldraw[fill=lightgray] (0,0) -- (2,1) -- (2,2) -- (0,1) -- cycle;
\draw (0,1) -- (-1*0.306,1.732*0.306+1) -- (-1*0.306+2,1.732*0.306+2) -- (2,2)
(0,0) -- (1*0.5,-1.732*0.5) -- (1*0.5+2,-1.732*0.5+1) -- (2,1);
\foreach \i in {0,2}
{
\filldraw[fill=white] (\i,\i) circle (1.5pt);
}
\node at (1,1) {$K$};
\node at (-0.2,0) {$\hat{0}_K$};
\node at (2.25,2) {$\hat1_K$};
\end{tikzpicture}
\caption{the finite distributive lattice $L$ with a cutting $K$}
\end{subfigure}
\begin{subfigure}[b]{0.35\linewidth}
\centering
\begin{tikzpicture}
\filldraw[fill=lightgray] (0,0) -- (2,1) -- (2,2) -- (0,1) -- cycle
(0.707,-0.707) -- (2.707,1-0.707) -- (2.707,2-0.707) -- (0.707,1-0.707) -- cycle;
\draw (0,1) -- (-1*0.306,1.732*0.306+1) -- (-1*0.306+2,1.732*0.306+2) -- (2,2);
\draw[xshift=0.707cm,yshift=-0.707cm] (0,0) -- (1*0.5,-1.732*0.5) -- (1*0.5+2,-1.732*0.5+1) -- (2,1);
\foreach \i in {0,1}
{
\foreach \j in {0,1}
{
\draw (2*\i,\i+\j)++(0.707,-0.707) -- (2*\i,\i+\j);
}
}
\foreach \i in {0,2}
{
\foreach \j in {0,1}
{
\filldraw[fill=white] (\i+\j*0.707,\i-\j*0.707) circle (1.5pt);
}
}
\node at (0.2,-0.507) {$x_K$};
\node at (1,1) {$K$};
\node at (1+0.707,1-0.707) {$K'(\cong K)$};
\end{tikzpicture}
\caption{the expansion $L \boxplus K$}
\end{subfigure}
\caption{The finite distributive lattices $L$ and $L\boxplus K$}\label{fig:lplk}
\end{figure}
In particular, if $P-x \cong P*x$, then $\mathcal{F}(P) \cong \mathcal{F}(P-x) \mathbin{\square} \mathbf{2}$. Note that Theorem~\ref{th:cetfdl} holds for all $x\in P$, thus we get a more general method for drawing the Hasse diagram of $\mathcal{F}(P)$ than Stanley's in \cite[P293]{bStanl2011}. On the other hand, we have a consequence on Day's doubling construction.
\begin{remark}
Let $L$ be a finite distributive lattice and let $K$ be a cutting of $L$. If $\hat{1}_K=\hat{1}_L$, then $L\uparrow \hat{0}_K$ in \cite{aErneHR2002} is the same as $L\boxplus K$. Thus, the operation $L\boxplus K$ is a special case of doubling construction on lattices in \cite{aDay1970}, and a generalization of the case in \cite[Section~2]{aErneHR2002}.
\end{remark}
In addition, we have two more general results than those in \cite[Section~2]{aErneHR2002}, namely Corollaries~\ref{th:ce} and \ref{th:mce}.
\begin{corollary}\label{th:ce}
Let $P = P_0 \cup P_m$ be a finite poset, where $P_0 \cap P_m = \emptyset$ and $P_i=\{x_1,x_2,\dots,x_i\}$ for $i\in \{1,2,\dots,m\}$, then
\begin{multline*}
\mathcal{F}(P) \cong (\cdots((\mathcal{F}(P_0) \boxplus \mathcal{F}((P_0\cup P_1)*x_1)) \boxplus \mathcal{F}((P_0\cup P_2)*x_2)) \boxplus \cdots \\
\boxplus \mathcal{F}((P_0\cup P_{m-1})*x_{m-1})) \boxplus \mathcal{F}((P_0\cup P_m)*x_m).
\end{multline*}
\end{corollary}
Specially, if $P_0 = \emptyset$ in Corollary~\ref{th:ce}, a more general result is obtained by Theorem~\ref{th:ftfdl}.
\begin{corollary}\label{th:mce}
Every finite distributive lattice with more than one element can be generated from one element lattice $\mathbf{1}$ by finite expansions $\boxplus$.
\end{corollary}
The consequences for cut element is also obtained.
\begin{corollary}\label{coro:cut}
Let $P$ be a poset and $x\in P$.
Then $\mathcal{F}(P-x)$ has a cut element, if and only if there exist two cut elements $M_1,M_2\in \mathcal{F}(P)$ such that $M_1 \cong M_2\cup \{x\}$, if and only if $P*x = \emptyset$.
% If $P*x$ is empty, then the filter lattice $\mathcal{F}(P-x)$ has a cut element; and there exist two cut elements $M_1,M_2\in \mathcal{F}(P)$ such that %$M_1\prec M_2$ %, where $M_1\in \mathop{\mathrm{Mi}}(\mathcal{F}(P))$ and $M_2\in \mathop{\mathrm{Ji}}(\mathcal{F}(P))$ and
% $M_1 \cong M_2\cup \{x\}$.
% In addition, if there exists a unique element $x$ such that $P*x=\emptyset$, then $\mathcal{F}(P-x)$ has a unique cut element.
\end{corollary}
Let $P$ and $S$ be posets, by %$P\mathbin{\dot\cup}S$,
$P+S$ and $P\dotplus S$ denote %the disjoint union,
ordinal sum and vertical sum \cite{bDaveyP2002} of $P$ and $S$, respectively. As a consequence of Theorem \ref{th:cetfdl}, we have the following result.
\begin{corollary}[\cite{bDaveyP2002}]
Let $P$ and $S$ be posets, then
$\mathcal{F}(P+\mathbf{1})\cong\mathcal{F}(P)+\mathbf{1}, \mathcal{F}(\mathbf{1}+P)\cong\mathbf{1}+\mathcal{F}(P)$.
Moreover,
$\mathcal{F}(P\mathbin{\dot{\cup}} S) \cong \mathcal{F}(P)\mathbin{\square} \mathcal{F}(S)$ and $\mathcal{F}(P+S)\cong \mathcal{F}(P)\dotplus\mathcal{F}(S)$.
\end{corollary}
\section{Enumerative properties}
It is easy to show that the enumeration result on $L \boxplus K$ by counting.
Let $R_L(K,x)$ denote the rank generating function of $K$ in $L$, and $R(L,x) := R_L(L,x)$; and let $h_L(x)$ denote the height of $x$ in $L$ for $x \in L$.
The rank generating function of $L \boxplus K$ is obtained by Theorem~\ref{th:cetfdl}.
\begin{proposition}
Let $L$ be a finite distributive lattice and $K$ is a cutting of $L$. The rank generating function of $L\boxplus K$ is
% \[
% R(L\boxplus K,x) = R_L({\downarrow K},x) + xR_L({\uparrow K},x) = R_{\downarrow K}({\downarrow K},x) + x^{\phi_L(\hat0_K)+1}R_{\uparrow K}({\uparrow K},x)
% \]
\[
R(L\boxplus K,x) = R_L({\downarrow K},x) + xR_L({\uparrow K},x) = R({\downarrow K},x) + x^{h_L(\hat0_K)+1}R({\uparrow K},x)
\]
\end{proposition}
In particular, we have the follow corollary.
\begin{corollary}\label{coro:rec-rgf}
Let $L$ be a finite distributive lattice and $K$ is a cutting of $L$. The rank generating function of $L\boxplus K$ is
\[
R(L\boxplus K,x) =
\begin{cases}
R(L,x) + x^{h_L(\hat0_K)+1}R(K,x), & \text{if } \hat1_K = \hat1_L; \\
R(K,x) + xR(L,x), & \text{if } \hat0_K = \hat0_L.
\end{cases}
\]
\end{corollary}
Let $q_k(L)$ denote the number of convex Boolean lattice $\mathbf{B}_k$ in $L$, and let $q_k(L)=0$ if no $\mathbf{B}_k$ exists.
In fact, $q_k(L)$ is equal to the number of antichains of $k$ elements in $\operatorname{Mi}(L)$.
\begin{theorem}\label{th:enum}
Let $L$ be a finite distributive lattice and let $K$ be a cutting of $L$. For $k\ge 0$, we have
\[
q_k(L\boxplus K) = q_k(L) + q_k(K) + q_{k-1}(K).
\]
% \[
% p_d(L\boxplus K\boxplus K) = p_d(L\boxplus K) + p_{d-2}(K)
% \]
Especially, $|L \boxplus K| = |L| + |K|.$
\end{theorem}
For some filter lattices of special posets, not only the number of elements is calculated, but also the structure is obtained.
%For example, the undirected Hasse diagram of filter lattice of the fences \cite{bStanl2011} is Fibonacci cubes \cite{pYao2007}.
For example, the undirected Hasse diagrams of filter lattices of the fences \cite{bStanl2011} and crown \cite{bBakerFR1970,aKellyR1974} are Fibonacci cubes and Lucas cubes, respectively \cite{pYao2007}.
\begin{figure}[!htb] %%fences.tex, fences-n.tex, fencesxn.tex
\centering
\begin{subfigure}[b]{0.32\linewidth}
\centering
\begin{tikzpicture}[scale=0.8, transform shape]
\foreach \i in {0,...,4}
{
\draw (\i,1) -- (\i,0) -- (\i+1,1);
\filldraw[fill=white] (\i,1) circle (1.5pt)
(\i,0) circle (1.5pt)
(\i+1,1) circle (1.5pt);
}
\node[above] at (0,1) {$1$};
\node[above] at (2,1) {$n-1$};
\node[above] at (3,1) {$n+1$};
\node[below] at (2,0) {$n$};
\node[above] at (5.25,1) {$m+n-2$};
\end{tikzpicture}
\caption{$Z_{m+n-2}$}
\end{subfigure}
\begin{subfigure}[b]{0.32\linewidth}
\centering
\begin{tikzpicture}[scale=0.8, transform shape]
\draw[dashed] (2,1) -- (2,0) -- (3,1);
\foreach \i in {0,1,3,4}
{
\draw (\i,1) -- (\i,0) -- (\i+1,1);
\filldraw[fill=white] (\i,1) circle (1.5pt)
(\i,0) circle (1.5pt);
}
\filldraw[fill=white,dashed] (2,0) circle (1.5pt);
\filldraw[fill=white] (2,1) circle (1.5pt) (5,1) circle (1.5pt);
\node[above] at (0,1) {$1$};
\node[above] at (2,1) {$n-1$};
\node[above] at (3,1) {$n+1$};
\node[below] at (2,0) {$n$};
\node[above] at (5.25,1) {$m+n-2$};
\end{tikzpicture}
\caption{$Z_{m+n-2}-n$}
\end{subfigure}
\begin{subfigure}[b]{0.32\linewidth}
\centering
\begin{tikzpicture}[scale=0.8, transform shape]
\draw (1,1) -- (1,0) (3,0) -- (4,1);
\foreach \i in {0,4}
{
\draw (\i,1) -- (\i,0) -- (\i+1,1);
}
\foreach \i in {2,3}
{
\draw[dashed] (\i-1,0) -- (\i,1) -- (\i,0);
}
\foreach \i in {0,1}
{
\foreach \j in {0,1}
{
\filldraw[fill=white] (\i,\j) circle (1.5pt)
(\i+\j+3,\j) circle (1.5pt);
}
}
\filldraw[fill=white,dashed] (2,0) circle (1.5pt) (2,1) circle (1.5pt) (3,1) circle (1.5pt);
\node[above] at (0,1) {$1$};
\node[above] at (2,1) {$n-1$};
\node[above] at (3,1) {$n+1$};
\node[below] at (2,0) {$n$};
\node[above] at (5.25,1) {$m+n-2$};
\end{tikzpicture}
\caption{$Z_{m+n-2}*n$}
\end{subfigure}
\caption{An example for applications of Theorem~\ref{th:enum}}\label{fig:fibnum}
\end{figure}
By Theorems~\ref{th:cetfdl} and \ref{th:enum}, we have some results on Fibonacci cubes and Lucas cubes, that is Corollaries~\ref{coro:fib} and \ref{coro:lucas}.
\begin{corollary}\label{coro:fib}
For the fences as shown in Figure~\ref{fig:fibnum}, we have
\[
\Gamma_{m+n-2} \cong (\Gamma_{m-2} \mathbin{\square} \Gamma_{n-1}) \boxplus (\Gamma_{m-3} \mathbin{\square} \Gamma_{n-2}),
\]
and
\begin{equation}\label{eq:fc}
q_k(\Gamma_{m+n-2}) = q_k(\Gamma_{m-2} \mathbin{\square} \Gamma_{n-1})
+ q_k(\Gamma_{m-3} \mathbin{\square} \Gamma_{n-2}) + q_{k-1}(\Gamma_{m-3} \mathbin{\square} \Gamma_{n-2}),
\end{equation}
where $\Gamma_n$ is the $n$-th Fibonacci cube.
\end{corollary}
Hence some formulas for Fibonacci cubes are shown easily. For instance, let $k=0$ in \eqref{eq:fc}, we have
\[
F_{m+n} = F_mF_{n+1}+F_{m-1}F_n.
\]
In particular, let $m=2$, the equation \eqref{eq:fc} implies
%\[
% \Gamma_n \cong \Gamma_{n-1} \boxplus \Gamma_{n-2}
%\]
%and
\[
F_{n+2} = F_{n+1}+F_n.
\]
%In addition, let $F_n^{(m)}$ denote the number of $m$-dimensional hypercubes in $\Gamma_{n-2}$, we get
%\[
%F_n^{(m)} = F_{n-1}^{(m)} + F_{n-2}^{(m)} + F_{n-2}^{(m-1)},
%\]
%for $m \in \{1,2,\dots,{\lceil \frac n2 \rceil-1}\}$. Obviously $F_n^{(0)} = F_n$.
\begin{figure}[!htb] %%crown2n.tex, crown2n-m.tex, crown2nxm.tex
\centering
\begin{subfigure}[b]{0.32\linewidth}
\centering
\begin{tikzpicture}[scale=0.8, transform shape]
\foreach \i in {0,...,4}
{
\draw (\i,1) -- (\i,0) -- (\i+1,1);
\filldraw[fill=white] (\i,1) circle (1.5pt)
(\i,0) circle (1.5pt)
(\i+1,1) circle (1.5pt);
}
\draw (5,1) -- (5,0) -- (0,1);
\foreach \i in {0,...,5}
{
\filldraw[fill=white] (\i,1) circle (1.5pt)
(\i,0) circle (1.5pt);
}
\node[above] at (0,1) {$1$};
\node[below] at (5,0) {$2n$};
\end{tikzpicture}
\caption{$\bowtie_{2n}$}
\end{subfigure}
\begin{subfigure}[b]{0.32\linewidth}
\centering
\begin{tikzpicture}[scale=0.8, transform shape]
\draw[dashed] (2,1) -- (2,0) -- (3,1);
\draw (5,1) -- (5,0) -- (0,1);
\foreach \i in {0,1,3,4}
{
\draw (\i,1) -- (\i,0) -- (\i+1,1);
\filldraw[fill=white] (\i,1) circle (1.5pt)
(\i,0) circle (1.5pt);
}
\filldraw[fill=white] (5,0) circle (1.5pt);
\filldraw[fill=white,dashed] (2,0) circle (1.5pt);
\filldraw[fill=white] (2,1) circle (1.5pt) (5,1) circle (1.5pt);
\node[above] at (0,1) {$1$};
\node[below] at (5,0) {$2n$};
\node[below] at (2,0) {$m$};
\end{tikzpicture}
\caption{$\bowtie_{2n}-m$}
\end{subfigure}
\begin{subfigure}[b]{0.32\linewidth}
\centering
\begin{tikzpicture}[scale=0.8, transform shape]
\draw (3,0) -- (4,1) -- (4,0) -- (5,1) -- (5,0) -- (0,1) -- (0,0) -- (1,1) -- (1,0);
\foreach \i in {2,3}
{
\draw[dashed] (\i-1,0) -- (\i,1) -- (\i,0);
}
\foreach \i in {0,1}
{
\foreach \j in {0,1}
{
\filldraw[fill=white] (\i,\j) circle (1.5pt)
(\i+4,\j) circle (1.5pt);
}
}
\filldraw[fill=white] (3,0) circle (1.5pt);
\filldraw[fill=white,dashed] (2,0) circle (1.5pt) (2,1) circle (1.5pt) (3,1) circle (1.5pt);
\node[above] at (0,1) {$1$};
\node[below] at (5,0) {$2n$};
\node[below] at (2,0) {$m$};
\end{tikzpicture}
\caption{$\bowtie_{2n}*m$}
\end{subfigure}
\caption{Another example for applications of Theorem~\ref{th:enum}}\label{fig:nmlc}
\end{figure}
\begin{corollary}\label{coro:lucas}
For the crown as shown in Figure~\ref{fig:nmlc}, we have
\[
\Lambda_{2n} \cong \Gamma_{2n-1} \boxplus \Gamma_{2n-3}^*,
\]
and
\begin{equation}\label{eq:nmlc}
q_k(\Lambda_{2n}) = q_k(\Gamma_{2n-1}) + q_k(\Gamma_{2n-3}) + q_{k-1}(\Gamma_{2n-3}),
\end{equation}
where $\Lambda_n$ is the $n$-th Lucas cube.
\end{corollary}
Hence some formulas for Lucas cubes are shown easily too. For instance, using the same method in \eqref{eq:fc} for \eqref{eq:nmlc}, we have
\[
L_{2n} = F_{2n+1}+F_{2n-1}.
\]
Moreover, as similar to results in \cite{aMunarS2002b}, we can obtain the recurrence relations of rank generating functions of $\Gamma_n$ and $\Lambda_{2n}$, by Corollary~\ref{coro:rec-rgf}, respectively, but we shall omit them.
Let $d_k^-(L) = |\{\,x \in L \mid \operatorname{deg}_L^-(x) = k\,\}|$,
$d_k^+(L) = |\{\,x \in L \mid \operatorname{deg}_L^+(x) = k\,\}|$,
and
$d_k(L) = |\{\,x \in L \mid \operatorname{deg}_L(x) = k\,\}|$, where $\operatorname{deg}_L^-(x) = |\{\,y \in L \mid x \prec y\,\}|$, $\operatorname{deg}_L^+(x) = |\{\,y \in L \mid y \prec x\,\}|$ and $\operatorname{deg}_L(x) = \operatorname{deg}_L^-(x) + \operatorname{deg}_L^+(x)$ for $x \in L$.
It is not difficult to see that $d_0^-(L) =1$ and $d_1^-(L) = |\operatorname{Mi}(L)|$ for all finite distributive lattice $L$.
\begin{remark}
If $L$ is a finite distributive lattice, thus both $d_k^-(L)$ and $d_k^+(L)$ is equal to the number of maximal antichains in $\operatorname{Mi}(L)$ with only $k$ elements.
Therefore, it suffices to consider $d_k^-(L)$.
\end{remark}
In addition, the recurrence relations on $d_k^-(L)$ and $d_k(L)$ are obtained easily (see Figures~\ref{fig:deg} and \ref{fig:indeg}).
\begin{theorem}
Let $L$ be a finite distributive lattice. If $K$ is a cutting of $L$, then
\[
d_k((L\boxplus K) \boxplus K) = d_k(L \boxplus K) + d_{k-2}(K).
\]
and
\[
d_k^-(L\boxplus K) = d_k^-(L) + d_{k-1}^-(K),
\]
\end{theorem}
\begin{figure}[!ht] %%deg-lkk0.tex and deg-lkk.tex
\centering
% \begin{subfigure}[b]{0.28\linewidth}
% \centering
\begin{tikzpicture}[scale=1]
\filldraw[fill=lightgray] (0,0) -- (2,1) -- (2,2) -- (0,1) -- cycle;
\filldraw[fill=lightgray, xshift=0.707cm, yshift=-0.707cm] (0,0) -- (2,1) -- (2,2) -- (0,1) -- cycle;
\filldraw[fill=lightgray, xshift=1.414cm, yshift=-1.414cm] (0,0) -- (2,1) -- (2,2) -- (0,1) -- cycle;
\draw (0,1) -- (-1*0.306,1.732*0.306+1) -- (-1*0.306+2,1.732*0.306+2) -- (2,2);
% (0,0) -- (1*0.5,-1.732*0.5) -- (1*0.5+2,-1.732*0.5+1) -- (2,1);
\draw[xshift=1.414cm,yshift=-1.414cm] (0,0) -- (1*0.5,-1.732*0.5) -- (1*0.5+2,-1.732*0.5+1) -- (2,1);
% (0.707,-0.707) -- (1*0.75+0.707,-1.732*0.75-0.707) -- (1*0.75+2.707,-1.732*0.75+1-0.707) -- (2.707,1-0.707)
% (0,0) -- (0.707,-0.707) (2,1) -- (2.707,0.293) (2,2) -- (2.707,1.293) (0,1) -- (0.707,0.293);
\foreach \i in {0,1}
{
\foreach \j in {0,1}
{
\draw (2*\i,\i+\j)++(1.414,-1.414) -- (2*\i,\i+\j);
}
}
\foreach \i in {0,2}
{
\foreach \j in {0,1,2}
{
\filldraw[fill=white] (\i+\j*0.707,\i-\j*0.707) circle (1.5pt);
}
}
\node at (0.2,-0.507) {$x_K$};
\node at (0.2+0.707,-0.507-0.707) {$x'_K$};
\node at (1,1) {$K$};
\node at (1+0.707,1-0.707) {$K'(\cong K)$};
\node at (1+1.414,1-1.414) {$K''(\cong K)$};
\end{tikzpicture}
% \caption{$L$ with a cutting $K$}
% \end{subfigure}
\raisebox{15ex}{${}\longrightarrow{}$}
% \begin{subfigure}[b]{0.45\linewidth}
% \centering
\begin{tikzpicture}[scale=1]
\filldraw[fill=lightgray] (0,0) -- (2,1) -- (2,2) -- (0,1) -- cycle;
\filldraw[fill=lightgray, xshift=4.707cm, yshift=-0.707cm] (0,0) -- (2,1) -- (2,2) -- (0,1) -- cycle;
\filldraw[fill=lightgray, xshift=1.414cm, yshift=-1.414cm] (0,0) -- (2,1) -- (2,2) -- (0,1) -- cycle;
\draw (0,1) -- (-1*0.306,1.732*0.306+1) -- (-1*0.306+2,1.732*0.306+2) -- (2,2);
% (0,0) -- (1*0.5,-1.732*0.5) -- (1*0.5+2,-1.732*0.5+1) -- (2,1);
\draw[xshift=1.414cm,yshift=-1.414cm] (0,0) -- (1*0.5,-1.732*0.5) -- (1*0.5+2,-1.732*0.5+1) -- (2,1);
% (0.707,-0.707) -- (1*0.75+0.707,-1.732*0.75-0.707) -- (1*0.75+2.707,-1.732*0.75+1-0.707) -- (2.707,1-0.707)
% (0,0) -- (0.707,-0.707) (2,1) -- (2.707,0.293) (2,2) -- (2.707,1.293) (0,1) -- (0.707,0.293);
\foreach \i in {0,1}
{
\foreach \j in {0,1}
{
\draw (2*\i,\i+\j)++(1.414,-1.414) -- (2*\i,\i+\j);
}
}
\foreach \i in {0,1}
{
\foreach \j in {0,1}
{
\draw[xshift=4cm] (2*\i,\i+\j)++(1.414,-1.414) -- (2*\i,\i+\j);
}
}
\foreach \i in {0,2}
{
\foreach \j in {0,2} %%{0,1,2}
{
\filldraw[fill=white] (\i+\j*0.707,\i-\j*0.707) circle (1.5pt);
}
\filldraw[fill=white] (\i+4.707,\i-0.707) circle (1.5pt);
}
% \node at (0.2,-0.507) {$x_K$};
% \node at (0.2+0.707,-0.507-0.707) {$x'_K$};
\node at (0.2+4,-0.507) {$x_K$};
\node at (0.2+4+0.707,-0.507-0.707) {$x'_K$};
\node at (1,1) {$K$};
\node at (1+4+0.707,1-0.707) {$K'(\cong K)$};
\node at (1+1.414,1-1.414) {$K''(\cong K)$};
\end{tikzpicture}
% \caption{$L$ with a cutting $K$}
% \end{subfigure}
\caption{Illustrating the recurrence relation of $d_k(L \boxplus K \boxplus K)$}\label{fig:deg}
\end{figure}
\begin{figure}[!ht] %%l&k.tex and lbpk.tex
\centering
% \begin{subfigure}[b]{0.27\linewidth}
% \centering
\begin{tikzpicture}[scale=1]
\filldraw[fill=lightgray] (0,0) -- (2,1) -- (2,2) -- (0,1) -- cycle
(0.707,-0.707) -- (2.707,1-0.707) -- (2.707,2-0.707) -- (0.707,1-0.707) -- cycle;
\draw (0,1) -- (-1*0.306,1.732*0.306+1) -- (-1*0.306+2,1.732*0.306+2) -- (2,2);
\draw[xshift=0.707cm,yshift=-0.707cm] (0,0) -- (1*0.5,-1.732*0.5) -- (1*0.5+2,-1.732*0.5+1) -- (2,1);
\foreach \i in {0,1}
{
\foreach \j in {0,1}
{
\draw (2*\i,\i+\j)++(0.707,-0.707) -- (2*\i,\i+\j);
}
}
\foreach \i in {0,2}
{
\foreach \j in {0,1}
{
\filldraw[fill=white] (\i+\j*0.707,\i-\j*0.707) circle (1.5pt);
}
}
\node at (0.2,-0.507) {$x_K$};
\node at (1,1) {$K$};
\node at (1+0.707,1-0.707) {$K'(\cong K)$};
\end{tikzpicture}
% \caption{$L$ with a cutting $K$}
% \end{subfigure}
\raisebox{12ex}{${}\longrightarrow{}$}
% \begin{subfigure}[b]{0.45\linewidth}
% \centering
\begin{tikzpicture}[scale=1]
\filldraw[fill=lightgray] (0,0) -- (2,1) -- (2,2) -- (0,1) -- cycle;
\filldraw[fill=lightgray, xshift=3.5cm, yshift=-0.707cm] (0,0) -- (2,1) -- (2,2) -- (0,1) -- cycle;
\draw (0,1) -- (-1*0.306,1.732*0.306+1) -- (-1*0.306+2,1.732*0.306+2) -- (2,2);
\draw (0,0) -- (1*0.5,-1.732*0.5) -- (1*0.5+2,-1.732*0.5+1) -- (2,1);
\foreach \i in {0,1}
{
\foreach \j in {0,1}
{
\draw[xshift=3.5cm, yshift=-0.707cm, -stealth] (2*\i,\i+\j)++(-0.707,0.707) -- (2*\i,\i+\j);
}
}
\foreach \i in {0,2}
{
\foreach \j in {0,1}
{
\filldraw[fill=white] (\i+\j*3.5,\i-\j*0.707) circle (1.5pt);
}
}
\node at (3,-0.507) {$x_K$};
\node at (1,1) {$K$};
\node at (1+3.5,1-0.707) {$K'(\cong K)$};
\end{tikzpicture}
% \caption{$L$ with a cutting $K$}
% \end{subfigure}
\caption{Illustrating the recurrence relation of $d_k(L \boxplus K)$}\label{fig:indeg}
\end{figure}
By definitions of $q_k(L)$ and $d_k^-(L)$, we have the results on them.
\begin{proposition}\label{prop:rel-qd-}
Let $L$ be a finite distributive lattice and $m$ is the size of set of all maximal antichains in $\operatorname{Mi}(L)$. Then
\[
q_k(L) = \sum_{j=k}^m \binom jk d_j^-(L);
\]
moreover, by binomial transform,
\[
d_k^-(L) = \sum_{j=k}^m (-1)^{j-k}\binom jk q_j(L).
\]
\end{proposition}
Let $Q_L(x) = \sum_{k=0}^m q_k(L) x^k$ and $D_L^-(x) = \sum_{k=0}^m d_k^-(L) x^k$, from Proposition~\ref{prop:rel-qd-}, a simple calculation leads to the following two corollaries.
\begin{corollary}
Let $L$ be a finite distributive lattice and let $m$ denote the size of set of all maximal antichains in $\operatorname{Mi}(L)$. We have
\[
Q_L(x) = D_L^-(1+x).
\]
\end{corollary}
In addition, for a finite distributive lattice $L$, because of the fact that there is only one element covered by no element in $L$ and the element having exactly one cover is meet-irreducible, the following corollary is immediate.
\begin{corollary}
Let $L$ be a finite distributive lattice. Then
\[
Q_L(-1) = q_0(L) - q_1(L) + q_2(L) - q_3(L) +\cdots = D_L^-(0) = 1,
\]
and
\[
\left. \frac{\mathrm{d}\, Q_L(x)}{\mathrm{d}\, x} \right|_{x=-1} = \left. \frac{\mathrm{d}\, D_L^-(x)}{\mathrm{d}\, x} \right|_{x=0} = d_1^{-}(L) = |\operatorname{Mi}(L)|.
\]
\end{corollary}
This means that we verify the generalized Euler formula for polyhedrons with Euler characteristic $\chi = 1$ \cite{bDodsoP1997} and a special case of \cite{aSkrek2001}.
We finally remark that all results on filter can be written dually the conclusions on ideal and remain correct.
\section*{Acknowledgments}
\addcontentsline{toc}{section}{Acknowledgments}
The authors are grateful to the referees for their careful reading and many valuable suggestions.
%\bibliographystyle{acmke}
%\bibliography{../../../JabRef/paper,../../../JabRef/book}
\begin{thebibliography}{10}
\bibitem{bBakerFR1970}
Baker K.~A., Fishburn P.~C., and Roberts F.~S.
\newblock {\em Partial orders of dimension 2, interval orders, and interval
graphs}.
\newblock Rand Corporation, 1970.
\bibitem{aBerteC2002}
Bertet K., and Caspard N.
\newblock Doubling convex sets in lattices: Characterizations and recognition
algorithms.
\newblock {\em Order 19}, 2 (Jun 2002), 181--207.
\bibitem{aBirkh1937}
Birkhoff G.
\newblock Rings of sets.
\newblock {\em Duke Math J 3}, 3 (09 1937), 443--454.
\bibitem{aDaveyDQR1978}
Davey B.~A., Duffus D., Quackenbush R.~W., and Rival I.
\newblock Exponents of finite simple lattices.
\newblock {\em J London Math Soc 17}, 2 (1978), 203--211.
\bibitem{bDaveyP2002}
Davey B.~A., and Priestley H.~A.
\newblock {\em Introduction to Lattices and Order}, 2nd~ed.
\newblock Cambridge University Press, Cambridge, 2002.
\bibitem{aDay1970}
Day A.
\newblock A simple solution to the word problem for lattices.
\newblock {\em Canad Math Bull 13}, 2 (1970), 253--254.
\bibitem{aDay1977}
Day A.
\newblock Splitting lattices generate all lattices.
\newblock {\em Algebra Univ 7}, 1 (Dec 1977), 163--169.
\bibitem{aDay1992}
Day A.
\newblock Doubling constructions in lattice theory.
\newblock {\em Canad. J. Math. 44}, 2 (1992), 252--269.
\bibitem{aDayGP1979}
Day A., Gaskill H., and Poguntke W.
\newblock Distributive lattices with finite projective covers.
\newblock {\em Pacific Journal of Mathematics 81}, 1 (1979), 45--59.
\bibitem{aDayNT1989}
Day A., Nation J.~B., and Tschantz S.
\newblock Doubling convex sets in lattices and a generalized semidistributivity
condition.
\newblock {\em Order 6}, 2 (Jun 1989), 175--180.
\bibitem{bDodsoP1997}
{Dodson} C. T.~J., and {Parker} P.~E.
\newblock {\em A User’s Guide to Algebraic Topology}, vol.~387 of {\em
Mathematics and Its Applications}.
\newblock Kluwer Academic Publishers,, Dordrecht, 1997.
\bibitem{aErneHR2002}
Ern\'e M., Heitzig J., and Reinhold J.
\newblock On the number of distributive lattices.
\newblock {\em Electron J Combin 9}, 1 (2002), \#R24.
\bibitem{aGeyer1993}
Geyer W.
\newblock Generalizing semidistributivity.
\newblock {\em Order 10}, 1 (Mar 1993), 77--92.
\bibitem{bGraet2011}
Gr\"atzer G.
\newblock {\em Lattice Theory: Foundation}.
\newblock Birkh\"auser, Basel, 2011.
\bibitem{aHsuPL1993}
Hsu W.-J., Page C.~V., and Liu J.-S.
\newblock {Fibonacci} cubes -- a class of self-similar graphs.
\newblock {\em {Fibonacci} Quarterly 31\/} (1993), 65--72.
\bibitem{aJonss1971}
J\'onsson B.
\newblock Relatively free products of lattices.
\newblock {\em Algebra Univ 1}, 1 (Dec 1971), 362--373.
\bibitem{aKellyR1974}
Kelly D., and Rival I.
\newblock Crowns, fences, and dismantlable lattices.
\newblock {\em Canad. J. Math 26}, 5 (1974), 1257--1271.
\bibitem{aMcKen1972}
McKenzie R.
\newblock Equational bases and non-modular lattice varieties.
\newblock {\em Trans. Amer. Math. Soc. 174\/} (1972), 1--43.
\bibitem{aMunarCS2001}
Munarini E., Cippo C.~P., and Salvi N.~Z.
\newblock On the {Lucas} cubes.
\newblock {\em {Fibonacci} Quarterly 39}, 1 (2001), 12--21.
\bibitem{aMunarS2002b}
Munarini E., and Salvi N.~Z.
\newblock On the rank polynomial of the lattice of order ideals of fences and
crowns.
\newblock {\em Discrete Math 259}, 1 (2002), 163--177.
\bibitem{aNatio1995}
Nation J.~B.
\newblock {A}lan {D}ay's doubling construction.
\newblock {\em Algebra Univ 34}, 1 (Mar 1995), 24--34.
\bibitem{aSchei1972}
Schein B.~M.
\newblock A representation theorem for lattices.
\newblock {\em Algebra Univ 2}, 1 (1972), 177--178.
\bibitem{aSivak1978}
Siv\'ak B.
\newblock Representation of finite lattices by orders on finite sets.
\newblock {\em Math. Slovaca\/} (1978), 203--215.
\bibitem{aSkrek2001}
{\v{S}krekovski} R.
\newblock Two relations for median graphs.
\newblock {\em Discrete Mathematics 226}, 1-3 (2001), 351--353.
\bibitem{bStanl2011}
Stanley R.~P.
\newblock {\em Enumerative Combinatorics: Volume 1}, 2nd~ed., vol.~49 of {\em
Cambridge studies in advanced mathematics}.
\newblock Cambridge University Press, Cambridge, 2011.
\bibitem{pYao2007}
Yao H.
\newblock {\em Perfect Matchings of Plane Bipartite Graphs and Distributive
Lattice Structure}.
\newblock {PhD}~thesis, Lanzhou University, Lanzhou, May 2007.
\newblock in Chinese.
\bibitem{aYaoZ2015}
Yao H., and Zhang H.
\newblock Non-matchable distributive lattices.
\newblock {\em Discrete Math 338}, 3 (2015), 122--132.
\end{thebibliography}
\end{document}