This paper was converted on www.awesomepapers.org from LaTeX by an anonymous user.
Want to know more? Visit the Converter page.

Triangular faces of the order and chain polytope of
a maximal ranked poset

Aki Mori Aki Mori, Learning center, Institute for general education, Setsunan University, Neyagawa, Osaka, 572-8508, Japan [email protected]
Abstract.

Let π’ͺ​(P){\mathcal{O}}(P) and π’žβ€‹(P){\mathcal{C}}(P) denote the order polytope and chain polytope, respectively, associated with a finite poset PP. We prove the following result: if PP is a maximal ranked poset, then the number of triangular 22-faces of π’ͺ​(P){\mathcal{O}}(P) is less than or equal to that of π’žβ€‹(P){\mathcal{C}}(P), with equality holding if and only if PP does not contain an XX-poset as a subposet.

Key words and phrases:
order polytope, chain polytope, partially ordered set, ff-vector
2020 Mathematics Subject Classification:
52B05, 06A07

1. Introduction

The order polytope π’ͺ​(P){\mathcal{O}}(P) and chain polytope π’žβ€‹(P){\mathcal{C}}(P) associated with a finite partially ordered set PP, were introduced by Stanley [7]. These are significant classes of nn-dimensional polytopes that have been widely studied in the fields of combinatorics and commutative algebra. We are interested in the face structure and the number of faces of π’ͺ​(P){\mathcal{O}}(P) and π’žβ€‹(P){\mathcal{C}}(P). For an nn-dimensional polytope, (f0,f1,…,fnβˆ’1)(f_{0},f_{1},\ldots,f_{n-1}) is called the f-vector which denotes the number of ii-dimensional faces, where 0≀i≀nβˆ’10\leq i\leq n-1. Hibi and Li proposed the following conjecture regarding the ff-vectors of π’ͺ​(P){\mathcal{O}}(P) and π’žβ€‹(P){\mathcal{C}}(P):

Conjecture 1.1 ([4, Conjecture 2.4]).

Let PP be a finite poset with |P|=d>1|P|=d>1. Then
(a) fi​(π’ͺ​(P))≀fi​(π’žβ€‹(P))f_{i}({\mathcal{O}}(P))\leq f_{i}({\mathcal{C}}(P)) for all 1≀i≀dβˆ’11\leq i\leq d-1.
(b) If fi​(π’ͺ​(P))=fi​(π’žβ€‹(P))f_{i}({\mathcal{O}}(P))=f_{i}({\mathcal{C}}(P)) for some 1≀i≀dβˆ’11\leq i\leq d-1, then π’ͺ​(P){\mathcal{O}}(P) and π’žβ€‹(P){\mathcal{C}}(P) are unimodularly equivalent.

In the same paper [4], it is proven that π’ͺ​(P){\mathcal{O}}(P) and π’žβ€‹(P){\mathcal{C}}(P) are unimodularly equivalent if and only if PP does not contain an XX-poset (Figure 1) as a subposet. As supporting evidence for this conjecture, we present the results that have already been shown. It is shown that the number of vertices of π’ͺ​(P){\mathcal{O}}(P) is equal to that of π’žβ€‹(P){\mathcal{C}}(P) in [7], and similarly, the number of edges of π’ͺ​(P){\mathcal{O}}(P) is equal to that of π’žβ€‹(P){\mathcal{C}}(P) in [5]. In this study, it should be noted that Conjecture 1.1 (b) is interpreted as 2≀i≀dβˆ’12\leq i\leq d-1 based on the results in [5]. In [4], it is shown that the number of facets of π’ͺ​(P){\mathcal{O}}(P) is less than or equal to that of π’ͺ​(P){\mathcal{O}}(P), with equality holding if and only if PP does not contain an XX-poset as a subposet. The recent result described below proves Conjecture 1.1 (a) when applied to specific classes of posets. In [1], it is shown that if PP is a poset called a maximal ranked poset, then fi​(π’ͺ​(P))≀fi​(π’žβ€‹(P))f_{i}({\mathcal{O}}(P))\leq f_{i}({\mathcal{C}}(P)) holds for all 1≀i≀dβˆ’11\leq i\leq d-1. Furthermore, in [2], it is shown that if PP is a poset built inductively by taking disjoint and connected unions of posets, then fi​(π’ͺ​(P))≀fi​(π’žβ€‹(P))f_{i}({\mathcal{O}}(P))\leq f_{i}({\mathcal{C}}(P)) holds for all 1≀i≀dβˆ’11\leq i\leq d-1. The result by [2] encompasses the result by [1]. Building upon the previous studies mentioned above, this paper focuses on the triangular 22-faces of π’ͺ​(P){\mathcal{O}}(P) and π’žβ€‹(P){\mathcal{C}}(P). The main theorem of this paper states that if PP is a maximal ranked poset, then the number of triangular 22-faces of π’ͺ​(P){\mathcal{O}}(P) is less than or equal to that of π’žβ€‹(P){\mathcal{C}}(P), with equality holding if and only if PP does not contain an XX-poset as a subposet. In other words, when PP is a maximal ranked poset containing an XX-poset, the number of triangles in the 11-skeleton of π’žβ€‹(P){\mathcal{C}}(P) is strictly greater than that in the 11-skeleton of π’ͺ​(P){\mathcal{O}}(P). This result contributes to the advancement of the Hibi and Li’s Conjecture.

Refer to caption
Figure 1. XX-poset

2. Basics on the faces of order and chain polytope

Let P={p1,…,pd}P=\{p_{1},\ldots,p_{d}\} be a finite poset equipped with a partial order ≀\leq. To each subset WβŠ‚PW\subset P, we associate ρ​(W)=βˆ‘i∈W𝐞iβˆˆβ„d\rho(W)=\sum_{i\in W}{\bf e}_{i}\in{\mathbb{R}}^{d}, where 𝐞1,…,𝐞d{\bf e}_{1},\ldots,{\bf e}_{d} are the canonical unit coordinate vectors of ℝd{\mathbb{R}}^{d}. In particular ρ​(βˆ…)\rho(\emptyset) is the origin of ℝd{\mathbb{R}}^{d}. A poset ideal of PP is a subset IβŠ‚PI\subset P such that if pi∈Ip_{i}\in I and pj≀pip_{j}\leq p_{i}, then pj∈Ip_{j}\in I. An antichain of PP is a subset AβŠ‚PA\subset P such that pip_{i} and pjp_{j} belonging to AA with iβ‰ ji\neq j are incomparable. Note that the empty set βˆ…\emptyset is considered both a poset ideal and an antichain of PP. We say that pjp_{j} covers pip_{i} if pi<pjp_{i}<p_{j} and pi<pk<pjp_{i}<p_{k}<p_{j} for no pk∈Pp_{k}\in P.

The two polytopes associated with a poset, namely the order polytope

π’ͺ​(P)={(x1,…,xd)βˆˆβ„d:0≀xi≀1​for all​ 1≀i≀d,xiβ‰₯xj​if​pi≀pj​in​P}\mathcal{O}(P)=\{(x_{1},\dots,x_{d})\in{\mathbb{R}}^{d}:0\leq x_{i}\leq 1\;\mbox{for all}\;1\leq i\leq d,\;x_{i}\geq x_{j}\;\mbox{if}\;p_{i}\leq p_{j}\;\mbox{in}\;P\}

and the chain polytope

π’žβ€‹(P)={(x1,…,xd)βˆˆβ„d:xiβ‰₯0​for all​ 1≀i≀d,xi1+β‹―+xik≀1​if​pi1<β‹―<pik​in​P}\mathcal{C}(P)=\{(x_{1},\dots,x_{d})\in{\mathbb{R}}^{d}:x_{i}\geq 0\;\mbox{for all}\;1\leq i\leq d,\;\;x_{i_{1}}+\cdots+x_{i_{k}}\leq 1\;\mbox{if}\;p_{i_{1}}<\cdots<p_{i_{k}}\;\mbox{in}\;P\}

were introduced by [7].

We briefly review the properties of these two polytopes as established by [7]. One has dim​π’ͺ​(P)\mathrm{dim}{\mathcal{O}}(P) == dimβ€‹π’žβ€‹(P)=d\mathrm{dim}{\mathcal{C}}(P)=d. Each vertex of π’ͺ​(P){\mathcal{O}}(P) corresponds to ρ​(I)\rho(I), where II is a poset ideal of PP and each vertex of π’žβ€‹(P){\mathcal{C}}(P) corresponds to ρ​(A)\rho(A), where AA is an antichain of PP. Notice that the poset ideals one-to-one correspond to the antichains. It then follows that the number of vertices of π’ͺ​(P){\mathcal{O}}(P) is equal to that of π’žβ€‹(P){\mathcal{C}}(P). Then the facets of π’ͺ​(P){\mathcal{O}}(P) are obtained by the the hyperplane whose defining equation is as follows:

  • β€’

    xi=0x_{i}=0, where pi∈Pp_{i}\in P is maximal;

  • β€’

    xj=1x_{j}=1, where pj∈Pp_{j}\in P is minimal;

  • β€’

    xi=xjx_{i}=x_{j}, where pjp_{j} covers pip_{i},

the facets of π’žβ€‹(P){\mathcal{C}}(P) are the following:

  • β€’

    xi=0x_{i}=0, for all pi∈Pp_{i}\in P;

  • β€’

    xi1+β‹―+xik=1x_{i_{1}}+\cdots+x_{i_{k}}=1, where pi1<β‹―<pikp_{i_{1}}<\cdots<p_{i_{k}} is a maximal chain of PP.

Recall that the comparability graph Com​(P){\rm Com}(P) of PP is the finite simple graph on the vertex set {p1,…,pd}\{p_{1},\ldots,p_{d}\}, where the edges are defined by {pi,pj}\{p_{i},p_{j}\} with iβ‰ ji\neq j, such that pip_{i} and pjp_{j} are comparable in PP. In general, we say that a nonempty subset Q={pi1,…,pik}Q=\{p_{i_{1}},\ldots,p_{i_{k}}\} of PP is connected in PP if the induced subgraph of Com​(P){\rm Com}(P) on {pi1,…,pik}\{p_{i_{1}},\ldots,p_{i_{k}}\} is connected.

The descriptions of edges and triangular 22-faces of π’ͺ​(P){\mathcal{O}}(P), as well as those of π’žβ€‹(P){\mathcal{C}}(P), respectively, are obtained as follows. Note that A​△​BA\triangle B denote symmetric difference of AA and BB, that is A​△​B=(Aβˆ–B)βˆͺ(Bβˆ–A)A\triangle B=(A\setminus B)\cup(B\setminus A).

Lemma 2.1 ([3, Lemma 4, Lemma 5]).

Let PP be a finite poset.

  1. (a)

    Let II and JJ be poset ideals of PP with Iβ‰ JI\neq J. Then the conv​({ρ​(I),ρ​(J)}){\rm conv}(\{\rho(I),\rho(J)\}) is an edge of π’ͺ​(P){\mathcal{O}}(P) if and only if IβŠ‚JI\subset J and Jβˆ–IJ\setminus I is connected in PP.

  2. (b)

    Let AA and BB be antichains of PP with Aβ‰ BA\neq B. Then the conv​({ρ​(A),ρ​(B)}){\rm conv}(\{\rho(A),\rho(B)\}) is an edge of π’žβ€‹(P){\mathcal{C}}(P) if and only if A​△​BA\triangle B is connected in PP.

Lemma 2.2 ([6, Theorem 1, Theorem 2]).

Let PP be a finite poset.

  1. (a)

    Let II, JJ, and KK be pairwise distinct poset ideals of PP. Then the conv​({ρ​(I),ρ​(J),ρ​(K)}){\rm conv}(\{\rho(I),\rho(J),\rho(K)\}) is a 2-face of π’ͺ​(P)\mathcal{O}(P) if and only if IβŠ‚JβŠ‚KI\subset J\subset K and Jβˆ–IJ\setminus I, Kβˆ–JK\setminus J, Kβˆ–IK\setminus I are connected in PP.

  2. (b)

    Let AA, BB, and CC be pairwise distinct antichains of PP. Then the conv​({ρ​(A),ρ​(B),ρ​(C)}){\rm conv}(\{\rho(A),\rho(B),\rho(C)\}) is a 2-face of π’žβ€‹(P)\mathcal{C}(P) if and only if A​△​BA\triangle B, B​△​CB\triangle C and C​△​AC\triangle A are connected in PP.

Remark 2.3.

The statement of Lemma 2.2 (a) in [5] is β€œIβŠ‚JβŠ‚KI\subset J\subset K and Kβˆ–IK\setminus I is connected in PP”, which is incorrect. Lemma 2.2 implies that the triangles in the 11-skeleton of π’ͺ​(P)\mathcal{O}(P) or π’žβ€‹(P)\mathcal{C}(P) correspond to the triangular 22-faces of each polytope.

3. The number of triangular 22-faces

The rank of a poset PP is the maximal length of a chain in PP, where the length of a chain CC is |C|βˆ’1|C|-1. A poset PP is graded of rank n if every maximal chain in PP has the same length nn. Let PiP_{i} denote the set of elements of rank ii. Namely, if PP is graded of rank nn, then it can be written P=βˆͺi=0nPiP=\cup_{i=0}^{n}P_{i} such that every maximal chain has the form p0<p1<β‹―<pnp_{0}<p_{1}<\cdots<p_{n}, where pi∈Pip_{i}\in P_{i}. If any two elements in distinct rank of a graded poset PP are comparable, then PP is called a maximal ranked poset [1]. Given a poset ideal IβŠ‚PI\subset P, we denote max​(I){\rm max}(I) for the set of maximal element of II. Similarly, given an antichain AβŠ‚PA\subset P, we denote ⟨A⟩\langle A\rangle as the poset ideal of PP generated by AA, that is ⟨A⟩={x:x≀p​for​some​p∈A}\langle A\rangle=\{x:x\leq p\;{\rm for\;some}\;p\in A\}. This establishes a bijection between poset ideal II and antichain AA of PP.

Let Eπ’ͺ​(P)βˆ—E^{*}_{\mathcal{O}(P)} denote the set of pairs {I,J}\{I,J\}, where II and JJ are poset ideals of PP with Iβ‰ JI\neq J, and conv​({ρ​(I),ρ​(J)}){\rm conv}(\{\rho(I),\rho(J)\}) forms an edge of π’ͺ​(P)\mathcal{O}(P), while conv​({ρ​(max​(I)),ρ​(max​(J))}){\rm conv}(\{\rho({\rm max}(I)),\rho({\rm max}(J))\}) does not form an edge of π’žβ€‹(P)\mathcal{C}(P). Let Eπ’žβ€‹(P)βˆ—E^{*}_{\mathcal{C}(P)} denote the set of pairs {A,B}\{A,B\}, where AA and BB are antichains of PP with Aβ‰ BA\neq B, and conv​({ρ​(A),ρ​(B)}){\rm conv}(\{\rho(A),\rho(B)\}) forms an edge of π’žβ€‹(P)\mathcal{C}(P), while conv​({ρ​(⟨A⟩),ρ​(⟨B⟩)}){\rm conv}(\{\rho(\langle A\rangle),\rho(\langle B\rangle)\}) does not form an edge of π’ͺ​(P)\mathcal{O}(P).

Lemma 3.1.

Let PP be a maximal ranked poset of rank nn.

  1. (a)

    Let II and JJ be poset ideals of PP with Iβ‰ JI\neq J. Then {I,J}∈Eπ’ͺ​(P)βˆ—\{I,J\}\in E^{*}_{\mathcal{O}(P)} if and only if I=βˆ…I=\emptyset, JJ is connected in PP, and |max​(J)|β‰₯2|{\rm max}(J)|\geq 2.

  2. (b)

    Let AA and BB be antichains of PP with Aβ‰ BA\neq B. Then {A,B}∈Eπ’žβ€‹(P)βˆ—\{A,B\}\in E^{*}_{\mathcal{C}(P)} if and only if there exists some β„“\ell such that A=Pβ„“βˆ’1A=P_{\ell-1}, BβŠ‚Pβ„“B\subset P_{\ell}, and |B|β‰₯2|B|\geq 2, where 1≀ℓ≀n1\leq\ell\leq n.

Proof.

(a) (β€œIf”) From Lemma 2.1(a), if I=βˆ…I=\emptyset and JJ is connected, then conv​({ρ​(I),ρ​(J)}){\rm conv}(\{\rho(I),\rho(J)\}) forms an edge of π’ͺ​(P)\mathcal{O}(P). Since |max​(J)|β‰₯2|{\rm max}(J)|\geq 2, max​(I)​△​max​(J)=max​(J){\rm max}(I)\,\triangle\,{\rm max}(J)={\rm max}(J) is disconnected. Thus, by Lemma 2.1(b), conv​({ρ​(max​(I)),ρ​(max​(J))}){\rm conv}(\{\rho({\rm max}(I)),\rho({\rm max}(J))\}) does not form an edge of π’žβ€‹(P)\mathcal{C}(P).
(β€œOnly if”) According to Lemma 2.1, {I,J}∈Eπ’ͺ​(P)βˆ—\{I,J\}\in E^{*}_{\mathcal{O}(P)} if and only if IβŠ‚JI\subset J, Jβˆ–IJ\setminus I is connected, and max​(I)​△​max​(J){\rm max}(I)\,\triangle\,{\rm max}(J) is disconnected in PP. If max​(I)​△​max​(J){\rm max}(I)\triangle{\rm max}(J) is disconnected, then either max​(I)=βˆ…{\rm max}(I)=\emptyset and |max​(J)|β‰₯2|{\rm max}(J)|\geq 2, or max​(I),max​(J){\rm max}(I),{\rm max}(J) are subsets of PkP_{k}, where 0≀k≀n0\leq k\leq n, since PP is a maximal ranked poset. If max​(I),max​(J)βŠ‚Pk{\rm max}(I),{\rm max}(J)\subset P_{k}, then max​(I)​△​max​(J)=max​(J)βˆ–max​(I)=Jβˆ–I{\rm max}(I)\;\triangle\;{\rm max}(J)={\rm max}(J)\setminus{\rm max}(I)=J\setminus I, since IβŠ‚JI\subset J, namely max​(I)βŠ‚max​(J){\rm max}(I)\subset{\rm max}(J). Since Jβˆ–IJ\setminus I is connected and max​(I)​△​max​(J){\rm max}(I)\;\triangle\,{\rm max}(J) is disconnected, this leads to a contradiction. Hence, we have max​(I)=βˆ…{\rm max}(I)=\emptyset and |max​(J)|β‰₯2|{\rm max}(J)|\geq 2, as desired.

(b) (β€œIf”) If there exists some β„“\ell such that A=Pβ„“βˆ’1A=P_{\ell-1}, BβŠ‚Pβ„“B\subset P_{\ell}, and |B|β‰₯2|B|\geq 2, where 1≀ℓ≀n1\leq\ell\leq n, then A​△​BA\triangle B is connected. Then, from Lemma 2.1(b), conv​({ρ​(A),ρ​(B)}){\rm conv}(\{\rho(A),\rho(B)\}) forms an edge of π’žβ€‹(P)\mathcal{C}(P). Furthermore, ⟨AβŸ©βŠ‚βŸ¨B⟩\langle A\rangle\subset\langle B\rangle and ⟨BβŸ©βˆ–βŸ¨A⟩\langle B\rangle\setminus\langle A\rangle is disconnected, because ⟨BβŸ©βˆ–βŸ¨A⟩=B\langle B\rangle\setminus\langle A\rangle=B and |B|β‰₯2|B|\geq 2. Hence, by Lemma 2.1(a), conv({ρ(⟨A⟩),ρ(⟨B⟩)){\rm conv}(\{\rho(\langle A\rangle),\rho(\langle B\rangle)) does not form an edge of π’ͺ​(P)\mathcal{O}(P).
(β€œOnly if”) According to Lemma 2.1, {A,B}∈Eπ’žβ€‹(P)βˆ—\{A,B\}\in E^{*}_{\mathcal{C}(P)} if and only if the following conditions hold:

  • β€’

    A​△​BA\triangle B is connected in PP ;

  • β€’

    ⟨AβŸ©βˆ–βŸ¨BβŸ©β‰ βˆ…\langle A\rangle\setminus\langle B\rangle\neq\emptyset and ⟨BβŸ©βˆ–βŸ¨AβŸ©β‰ βˆ…\langle B\rangle\setminus\langle A\rangle\neq\emptyset or ⟨AβŸ©βŠ‚βŸ¨B⟩\langle A\rangle\subset\langle B\rangle and ⟨BβŸ©βˆ–βŸ¨A⟩\langle B\rangle\setminus\langle A\rangle is disconnected in PP.

If ⟨AβŸ©βˆ–βŸ¨BβŸ©β‰ βˆ…\langle A\rangle\setminus\langle B\rangle\neq\emptyset and ⟨BβŸ©βˆ–βŸ¨AβŸ©β‰ βˆ…\langle B\rangle\setminus\langle A\rangle\neq\emptyset, then there exists some kk such that A,BβŠ‚PkA,B\subset P_{k} and Aβˆ–Bβ‰ βˆ…A\setminus B\neq\emptyset and Bβˆ–Aβ‰ βˆ…B\setminus A\neq\emptyset, where 0≀k≀n0\leq k\leq n, because PP is a maximal ranked poset. This contradicts A​△​BA\triangle B is connected. Thus, assuming ⟨AβŸ©βŠ‚βŸ¨B⟩\langle A\rangle\subset\langle B\rangle and ⟨BβŸ©βˆ–βŸ¨A⟩\langle B\rangle\setminus\langle A\rangle is disconnected, there exists some β„“\ell such that ⟨BβŸ©βˆ–βŸ¨AβŸ©βŠ‚Pβ„“\langle B\rangle\setminus\langle A\rangle\subset P_{\ell} and |⟨BβŸ©βˆ–βŸ¨A⟩|β‰₯2|\langle B\rangle\setminus\langle A\rangle|\geq 2. In this case, either A,BβŠ‚Pβ„“A,B\subset P_{\ell} or A=Pβ„“βˆ’1,BβŠ‚Pβ„“A=P_{\ell-1},B\subset P_{\ell} holds, but A,BβŠ‚Pβ„“A,B\subset P_{\ell} contradicts A​△​BA\triangle B is connected. Hence, there exists some β„“\ell such that A=Pβ„“βˆ’1,BβŠ‚Pβ„“A=P_{\ell-1},B\subset P_{\ell} and |⟨BβŸ©βˆ–βŸ¨A⟩|=|B|β‰₯2|\langle B\rangle\setminus\langle A\rangle|=|B|\geq 2, where 1≀ℓ≀n1\leq\ell\leq n. ∎

Definition 3.2.

Let PP be a finite poset. Let II, JJ, and KK be pairwise distinct poset ideals of PP, and let AA, BB, and CC be pairwise distinct antichains of PP. We define the sets of triples corresponding to triangular 22-faces of π’ͺ​(P)\mathcal{O}(P) and π’žβ€‹(P)\mathcal{C}(P) as follows:

Ξ”π’ͺ​(P)\displaystyle\Delta_{{\mathcal{O}}(P)} :=\displaystyle:= {{I,J,K}:conv​({ρ​(I),ρ​(J),ρ​(K)})​is a 2-face of​π’ͺ​(P)};\displaystyle\bigl{\{}\{I,J,K\}\;:\;{\rm conv}(\{\rho(I),\rho(J),\rho(K)\})\;\mbox{is a 2-face of}\;\mathcal{O}(P)\bigr{\}};
Ξ”π’žβ€‹(P)\displaystyle\Delta_{{\mathcal{C}}(P)} :=\displaystyle:= {{A,B,C}:conv​({ρ​(A),ρ​(B),ρ​(C)})​is a 2-face ofβ€‹π’žβ€‹(P)};\displaystyle\bigl{\{}\{A,B,C\}\;:\;{\rm conv}(\{\rho(A),\rho(B),\rho(C)\})\;\mbox{is a 2-face of}\;\mathcal{C}(P)\bigr{\}};
Ξ”π’ͺ​(P)βˆ—\displaystyle\Delta^{*}_{{\mathcal{O}}(P)} :=\displaystyle:= {{I,J,K}βˆˆΞ”π’ͺ​(P):at least one of​{I,J},{J,K},or​{I,K}​belongs to​Eπ’ͺ​(P)βˆ—};\displaystyle\bigl{\{}\{I,J,K\}\in\;\Delta_{\mathcal{O}(P)}:\mbox{at least one of}\;\{I,J\},\{J,K\},\;\mbox{or}\;\{I,K\}\;\mbox{belongs to}\;E^{*}_{\mathcal{O}(P)}\bigr{\}};
Ξ”π’žβ€‹(P)βˆ—\displaystyle\Delta^{*}_{{\mathcal{C}}(P)} :=\displaystyle:= {{A,B,C}βˆˆΞ”π’žβ€‹(P):at least one of​{A,B},{B,C},or​{A,C}​belongs to​Eπ’žβ€‹(P)βˆ—}.\displaystyle\bigl{\{}\{A,B,C\}\in\;\Delta_{\mathcal{C}(P)}:\mbox{at least one of}\;\{A,B\},\{B,C\},\;\mbox{or}\;\{A,C\}\;\mbox{belongs to}\;E^{*}_{\mathcal{C}(P)}\bigr{\}}.

The first claim we wish to present is that the number of triangular 22-faces of π’ͺ​(P){\mathcal{O}}(P) is less than or equal to that of π’žβ€‹(P){\mathcal{C}}(P), namely |Ξ”π’ͺ​(P)|≀|Ξ”π’žβ€‹(P)||\Delta_{{\mathcal{O}}(P)}|\leq|\Delta_{{\mathcal{C}}(P)}|. This assertion is guided by the following two lemmas.

Lemma 3.3.

Let PP be a finite poset. Then the number of triples in Ξ”π’ͺ​(P)βˆ–Ξ”π’ͺ​(P)βˆ—\Delta_{\mathcal{O}(P)}\setminus\Delta^{*}_{\mathcal{O}(P)} equals that in Ξ”π’žβ€‹(P)βˆ–Ξ”π’žβ€‹(P)βˆ—\Delta_{\mathcal{C}(P)}\setminus\Delta^{*}_{\mathcal{C}(P)}.

Proof.

From Lemma 2.2, the set of triples {ρ​(I),ρ​(J),ρ​(K)}\{\rho(I),\rho(J),\rho(K)\} where {I,J,K}βˆˆΞ”π’ͺ​(P)βˆ–Ξ”π’ͺ​(P)βˆ—\{I,J,K\}\in\Delta_{\mathcal{O}(P)}\setminus\Delta^{*}_{\mathcal{O}(P)}, corresponds to the set of triangles in the graph obtained by removing the all edges {ρ​(I1),ρ​(I2)}\{\rho(I_{1}),\rho(I_{2})\}, where {I1,I2}∈Eπ’ͺ​(P)βˆ—\{I_{1},I_{2}\}\in E^{*}_{\mathcal{O}(P)} from the 11-skeleton of π’ͺ​(P){\mathcal{O}}(P). Similarly, the set of triples {ρ​(A),ρ​(B),ρ​(C)}\{\rho(A),\rho(B),\rho(C)\} where {A,B,C}βˆˆΞ”π’žβ€‹(P)βˆ–Ξ”π’žβ€‹(P)βˆ—\{A,B,C\}\in\Delta_{\mathcal{C}(P)}\setminus\Delta^{*}_{\mathcal{C}(P)}, corresponds to the set of triangles in the graph obtained by removing the all edges {ρ​(A1),ρ​(A2)}\{\rho(A_{1}),\rho(A_{2})\}, where {A1,A2}∈Eπ’žβ€‹(P)βˆ—\{A_{1},A_{2}\}\in E^{*}_{\mathcal{C}(P)} from the 11-skeleton of π’žβ€‹(P){\mathcal{C}}(P). Since these graphs are clearly isomorphic, the number of triangles in them coincides. ∎

Lemma 3.4.

Let PP be a maximal ranked poset. Then the number of triples in Ξ”π’ͺ​(P)βˆ—\Delta^{*}_{\mathcal{O}(P)} is less than or equal to that in Ξ”π’žβ€‹(P)βˆ—\Delta^{*}_{\mathcal{C}(P)}.

Proof.

From Lemma 2.2(a) and Lemma 3.1(a), the following conditions hold for {I,J,K}βˆˆΞ”π’ͺ​(P)βˆ—\{I,J,K\}\in\Delta^{*}_{\mathcal{O}(P)}, where II, JJ, and KK be pairwise distinct poset ideals of PP :

  • β€’

    I=βˆ…I=\emptyset, JβŠ‚KJ\subset K and JJ, KK, and Kβˆ–JK\setminus J are connected in PP;

  • β€’

    |max​(J)|β‰₯2|{\rm max}(J)|\geq 2 or |max​(K)|β‰₯2|{\rm max}(K)|\geq 2.

Let PP has rank nn. Since PP is a maximal ranked poset, we may assume that max​(J)βŠ‚Pj{\rm max}(J)\subset P_{j} and max​(K)βŠ‚Pk{\rm max}(K)\subset P_{k}, where 0≀j≀k≀n0\leq j\leq k\leq n. We define the map Ο†:Ξ”π’ͺ​(P)βˆ—β†’Ξ”π’žβ€‹(P)βˆ—\varphi\;:\;\Delta^{*}_{\mathcal{O}(P)}\rightarrow\Delta^{*}_{\mathcal{C}(P)} by

{I,J,K}↦{{Pjβˆ’1,max​(J),max​(K)}if​|max​(J)|β‰₯2,{Pkβˆ’1,max​(J),max​(K)}if​|max​(J)|=1​and​kβˆ’jβ‰ 1,{Pkβˆ’1,min​(Kβˆ–J),max​(K)}if​|max​(J)|=1​and​kβˆ’j=1.\{I,J,K\}\mapsto\begin{cases}\{P_{j-1},\;{\rm max}(J),\;{\rm max}(K)\}&\mbox{if}\;\;|{\rm max}(J)|\geq 2,\\ \{P_{k-1},\;{\rm max}(J),\;{\rm max}(K)\}&\mbox{if}\;\;|{\rm max}(J)|=1\;\mbox{and}\;k-j\neq 1,\\ \{P_{k-1},\;{\rm min}(K\setminus J),\;{\rm max}(K)\}&\mbox{if}\;\;|{\rm max}(J)|=1\;\mbox{and}\;k-j=1.\end{cases}

We will show that the map is an injection.

Case 1. |max​(J)|β‰₯2.|{\rm max}(J)|\geq 2.
Since JJ is connected, we may assume that jβ‰₯1j\geq 1. Now, Pjβˆ’1P_{j-1}, max​(J){\rm max}(J) and max​(K){\rm max}(K) are pairwise distinct antichains of PP. As PP is a maximal ranked poset, Pjβˆ’1​△​max​(J)P_{j-1}\triangle{\rm max}(J) and Pjβˆ’1​△​max​(K)P_{j-1}\triangle{\rm max}(K) are connected. If j<kj<k, then max​(J)​△​max​(K){\rm max}(J)\triangle{\rm max}(K) is connected. If j=kj=k, then max​(J)​△​max​(K)=max​(K)βˆ–max​(J)=Kβˆ–J,{\rm max}(J)\;\triangle\;{\rm max}(K)={\rm max}(K)\setminus{\rm max}(J)=K\setminus J, because JβŠ‚KJ\subset K, namely max​(J)βŠ‚max​(K){\rm max}(J)\subset{\rm max}(K). Since Kβˆ–JK\setminus J is connected, max​(J)​△​max​(K){\rm max}(J)\triangle{\rm max}(K) is also connected. Furthermore, since |max​(J)|β‰₯2|{\rm max}(J)|\geq 2, from Lemma 3.1(b), we have {Pjβˆ’1,max​(J)}∈Eπ’žβ€‹(P)βˆ—\{P_{j-1},{\rm max}(J)\}\in E^{*}_{\mathcal{C}(P)}, implying {Pjβˆ’1,max​(J),max​(K)}βˆˆΞ”π’žβ€‹(P)βˆ—\{P_{j-1},{\rm max}(J),{\rm max}(K)\}\in\Delta^{*}_{\mathcal{C}(P)}.

Case 2. |max​(J)|=1|{\rm max}(J)|=1 and kβˆ’jβ‰ 1k-j\neq 1.
Now, |max​(K)|β‰₯2|{\rm max}(K)|\geq 2. Since PP is a maximal ranked poset and kβˆ’jβ‰ 1k-j\neq 1, max​(J)βŠ„Pkβˆ’1{\rm max}(J)\not\subset P_{k-1}. Thus Pkβˆ’1P_{k-1}, max​(J){\rm max}(J) and max​(K){\rm max}(K) are pairwise distinct antichains of PP and Pkβˆ’1​△​max​(J)P_{k-1}\triangle{\rm max}(J) and Pkβˆ’1​△​max​(K)P_{k-1}\triangle{\rm max}(K) are connected. If kβˆ’jβ‰₯2k-j\geq 2, then max​(J)​△​max​(K){\rm max}(J)\triangle{\rm max}(K) are connected. If kβˆ’j=0k-j=0, then we can show max​(J)​△​max​(K){\rm max}(J)\triangle{\rm max}(K) is connected by the same argument in Case 1. Furthermore, since |max​(K)|β‰₯2|{\rm max}(K)|\geq 2, from Lemma 3.1(b), we have {Pkβˆ’1,max​(K)}∈Eπ’žβ€‹(P)βˆ—\{P_{k-1},{\rm max}(K)\}\in E^{*}_{\mathcal{C}(P)}, implying {Pkβˆ’1,max​(J),max​(K)}βˆˆΞ”π’žβ€‹(P)βˆ—\{P_{k-1},{\rm max}(J),{\rm max}(K)\}\in\Delta^{*}_{\mathcal{C}(P)}.

Case 3. |max​(J)|=1|{\rm max}(J)|=1 and kβˆ’j=1k-j=1.
Now, |max​(K)|β‰₯2|{\rm max}(K)|\geq 2. Since PP is a maximal ranked poset and kβˆ’j=1k-j=1, min​(Kβˆ–J)=Pjβˆ–max​(J)=Pkβˆ’1βˆ–max​(J)βŠ‚Pkβˆ’1{\rm min}(K\setminus J)=P_{j}\setminus{\rm max}(J)=P_{k-1}\setminus{\rm max}(J)\subset P_{k-1}. Thus Pkβˆ’1P_{k-1}, min​(Kβˆ–J){\rm min}(K\setminus J) and max​(K){\rm max}(K) are pairwise distinct antichains of PP and Pkβˆ’1​△​max​(K)P_{k-1}\triangle{\rm max}(K) and min​(Kβˆ–J)​△​max​(K){\rm min}(K\setminus J)\triangle{\rm max}(K) are connected. Since min​(Kβˆ–J)=Pkβˆ’1βˆ–max​(J){\rm min}(K\setminus J)=P_{k-1}\setminus{\rm max}(J), then Pkβˆ’1​△​min​(Kβˆ–J)=max​(J)P_{k-1}\triangle{\rm min}(K\setminus J)={\rm max}(J). Hence, Pkβˆ’1​△​min​(Kβˆ–J)P_{k-1}\triangle{\rm min}(K\setminus J) is connected because |max​(J)|=1|{\rm max}(J)|=1. Furthermore, since |max​(K)|β‰₯2|{\rm max}(K)|\geq 2, from Lemma 3.1(b), we have {Pkβˆ’1,max​(K)}∈Eπ’žβ€‹(P)βˆ—\{P_{k-1},{\rm max}(K)\}\in E^{*}_{\mathcal{C}(P)}, implying {Pkβˆ’1,max​(J),max​(K)}βˆˆΞ”π’žβ€‹(P)βˆ—\{P_{k-1},{\rm max}(J),{\rm max}(K)\}\in\Delta^{*}_{\mathcal{C}(P)}.

In Case 1 or Case 2, the map is clearly injective. In Case 3, let {βˆ…,J,K}\{\emptyset,J,K\} and {βˆ…,Jβ€²,Kβ€²}\{\emptyset,J^{\prime},K^{\prime}\} belong to Ξ”π’ͺ​(P)βˆ—\Delta^{*}_{\mathcal{O}(P)}. We note the following holds:

  • β€’

    max​(Jβ€²)βŠ‚Pjβ€²{\rm max}(J^{\prime})\subset P_{j^{\prime}} and max​(Kβ€²)βŠ‚Pkβ€²{\rm max}(K^{\prime})\subset P_{k^{\prime}}, where 0≀j′≀k′≀n0\leq j^{\prime}\leq k^{\prime}\leq n;

  • β€’

    |max​(Jβ€²)|=1|{\rm max}(J^{\prime})|=1 and kβ€²βˆ’jβ€²=1k^{\prime}-j^{\prime}=1.

Suppose that Pkβˆ’1=Pkβ€²βˆ’1P_{k-1}=P_{k^{\prime}-1}, min​(Kβˆ–J)=min​(Kβ€²βˆ–Jβ€²){\rm min}(K\setminus J)={\rm min}(K^{\prime}\setminus J^{\prime}), and max​(K)=max​(Kβ€²){\rm max}(K)={\rm max}(K^{\prime}). Then K=Kβ€²K=K^{\prime}. Since kβˆ’j=kβ€²βˆ’jβ€²=1k-j=k^{\prime}-j^{\prime}=1, max​(Kβˆ–J)=max​(K)=max​(Kβ€²)=max​(Kβ€²βˆ–Jβ€²){\rm max}(K\setminus J)={\rm max}(K)={\rm max}(K^{\prime})={\rm max}(K^{\prime}\setminus J^{\prime}). Thus max​(Kβˆ–J)=max​(Kβ€²βˆ–Jβ€²){\rm max}(K\setminus J)={\rm max}(K^{\prime}\setminus J^{\prime}). Now, we have min​(Kβˆ–J)=min​(Kβˆ–Jβ€²){\rm min}(K\setminus J)={\rm min}(K\setminus J^{\prime}) and max​(Kβˆ–J)=max​(Kβˆ–Jβ€²){\rm max}(K\setminus J)={\rm max}(K\setminus J^{\prime}) by K=Kβ€²K=K^{\prime}. Since Kβˆ–J=min​(Kβˆ–J)βˆͺmax​(Kβˆ–J)K\setminus J={\rm min}(K\setminus J)\cup{\rm max}(K\setminus J), it follows that Kβˆ–J=Kβˆ–Jβ€²K\setminus J=K\setminus J^{\prime}, namely J=Jβ€²J=J^{\prime}, as desired. ∎

From Lemma 3.3 and Lemma 3.4, it follows that |Ξ”π’ͺ​(P)|≀|Ξ”π’žβ€‹(P)||\Delta_{\mathcal{O}(P)}|\leq|\Delta_{\mathcal{C}(P)}|. Consequently, we obtain the following.

Corollary 3.5.

Let PP be a maximal ranked poset. Then the number of triangular 22-faces of π’ͺ​(P)\mathcal{O}(P) is less than or equal to that of π’žβ€‹(P)\mathcal{C}(P).

Furthermore, we can characterize whether the number of triangular 22-faces of π’ͺ​(P){\mathcal{O}}(P) is equal to that of π’žβ€‹(P){\mathcal{C}}(P), namely |Ξ”π’ͺ​(P)|=|Ξ”π’žβ€‹(P)||\Delta_{\mathcal{O}(P)}|=|\Delta_{\mathcal{C}(P)}|, by the XX-poset from Figure 1.

Theorem 3.6.

Let PP be a maximal ranked poset. Then the number of triangular 22-faces of π’ͺ​(P)\mathcal{O}(P) is equal to that of π’žβ€‹(P)\mathcal{C}(P) if and only if the poset PP does not contain an XX-poset as a subposet.

Proof.

(β€œIf”) The poset PP does not contain an XX-poset from as a subposet if and only if π’ͺ​(P)\mathcal{O}(P) and π’žβ€‹(P)\mathcal{C}(P) are unimodularly equivalent by [4, Theorem 2.1]. Thus, the 11-skeleton of π’ͺ​(P)\mathcal{O}(P) and that of π’žβ€‹(P)\mathcal{C}(P) are isomorphic as finite graphs. In particlar, the number of triagles in those 11-skeletons coincides. Therefore |Ξ”π’ͺ​(P)|=|Ξ”π’žβ€‹(P)||\Delta_{\mathcal{O}(P)}|=|\Delta_{\mathcal{C}(P)}|.
(β€œOnly if”) Let PP has rank nn. Suppose that the poset PP contains an XX-poset from as a subposet. Let the XX-poset be {a,b,c,d,e}\{a,b,c,d,e\}, where each element is pairwise distinct such that (i) aa and bb are incomparable, (ii) dd and ee are incomparable, and (iii) a<c<da<c<d, b<c<eb<c<e. We may assume that d,e∈Psd,e\in P_{s}, c∈Psβˆ’1c\in P_{s-1} and a,b∈Psβˆ’ta,b\in P_{s-t}, where 2≀s≀n2\leq s\leq n and 2≀t≀s2\leq t\leq s. Now, Psβˆ’tP_{s-t}, Psβˆ’1P_{s-1}, and PsP_{s} are pairwise distinct antichains. As PP is a maximal ranked poset, Psβˆ’t​△​Psβˆ’1P_{s-t}\triangle P_{s-1}, Psβˆ’1​△​PsP_{s-1}\triangle P_{s}, and Psβˆ’t​△​PsP_{s-t}\triangle P_{s} are connected in PP. Since |Ps|β‰₯2|P_{s}|\geq 2, from Lemma 3.1(b), {Psβˆ’1,Ps}∈Eπ’žβ€‹(P)βˆ—\{P_{s-1},P_{s}\}\in E^{*}_{\mathcal{C}(P)}, implying {Psβˆ’t,Psβˆ’1,Ps}βˆˆΞ”π’žβ€‹(P)βˆ—\{P_{s-t},P_{s-1},P_{s}\}\in\Delta^{*}_{\mathcal{C}(P)}. We will show that there does not exist any {I,J,K}βˆˆΞ”π’ͺ​(P)βˆ—\{I,J,K\}\in\Delta^{*}_{\mathcal{O}(P)} such that φ​({I,J,K})={Psβˆ’t,Psβˆ’1,Ps}\varphi(\{I,J,K\})=\{P_{s-t},P_{s-1},P_{s}\}, where Ο†\varphi is the injection obtained in the proof of Lemma 3.4.

Case 1. Suppose that there exists some {I,J,K}βˆˆΞ”π’ͺ​(P)βˆ—\{I,J,K\}\in\Delta^{*}_{\mathcal{O}(P)} such that |max​(J)|β‰₯2|{\rm max}(J)|\geq 2 and φ​({I,J,K})\varphi(\{I,J,K\}) coincides {Psβˆ’t,Psβˆ’1,Ps}\{P_{s-t},P_{s-1},P_{s}\}. Since |max​(J)|β‰₯2|{\rm max}(J)|\geq 2, then φ​({I,J,K})={Pjβˆ’1,max​(J),max​(K)}\varphi(\{I,J,K\})=\{P_{j-1},\;{\rm max}(J),\;{\rm max}(K)\}. We may assume that Ps=max​(K)P_{s}={\rm max}(K), Psβˆ’1=max​(J)P_{s-1}={\rm max}(J), and Psβˆ’t=Pjβˆ’1P_{s-t}=P_{j-1}. Then K=⟨Ps⟩K=\langle P_{s}\rangle and J=⟨Psβˆ’1⟩J=\langle P_{s-1}\rangle. However, since ⟨PsβŸ©βˆ–βŸ¨Psβˆ’1⟩\langle P_{s}\rangle\setminus\langle P_{s-1}\rangle is disconnected, this contradicts Kβˆ–JK\setminus J is connected, because {I,J,K}βˆˆΞ”π’ͺ​(P)βˆ—\{I,J,K\}\in\Delta^{*}_{\mathcal{O}(P)}.

Case 2. Suppose that there exists some {I,J,K}βˆˆΞ”π’ͺ​(P)βˆ—\{I,J,K\}\in\Delta^{*}_{\mathcal{O}(P)} such that |max​(J)|=1|{\rm max}(J)|=1, kβˆ’jβ‰ 1k-j\neq 1, and φ​({I,J,K})\varphi(\{I,J,K\}) coincides {Psβˆ’t,Psβˆ’1,Ps}\{P_{s-t},P_{s-1},P_{s}\}. Since |max​(J)|=1|{\rm max}(J)|=1 and kβˆ’jβ‰ 1k-j\neq 1, then φ​({I,J,K})={Pkβˆ’1,max​(J),max​(K)}\varphi(\{I,J,K\})=\{P_{k-1},\;{\rm max}(J),\;{\rm max}(K)\}. We may assume that Ps=max​(K)P_{s}={\rm max}(K), Psβˆ’1=Pkβˆ’1P_{s-1}=P_{k-1}, and Psβˆ’t=max​(J)P_{s-t}={\rm max}(J). However, Since |Psβˆ’t|β‰₯2|P_{s-t}|\geq 2, this contradicts |max​(J)|=1|{\rm max}(J)|=1.

Case 3. Suppose that there exists some {I,J,K}βˆˆΞ”π’ͺ​(P)βˆ—\{I,J,K\}\in\Delta^{*}_{\mathcal{O}(P)} such that |max​(J)|=1|{\rm max}(J)|=1, kβˆ’j=1k-j=1, and φ​({I,J,K})\varphi(\{I,J,K\}) coincides {Psβˆ’t,Psβˆ’1,Ps}\{P_{s-t},P_{s-1},P_{s}\}. Since |max​(J)|=1|{\rm max}(J)|=1 and kβˆ’j=1k-j=1, then φ​({I,J,K})={Pkβˆ’1,min​(Kβˆ–J),max​(K)}\varphi(\{I,J,K\})=\{P_{k-1},\;{\rm min}(K\setminus J),\;{\rm max}(K)\}. We may assume that Ps=max​(K)P_{s}={\rm max}(K), Psβˆ’1=Pkβˆ’1P_{s-1}=P_{k-1}, and Psβˆ’t=min​(Kβˆ–J)P_{s-t}={\rm min}(K\setminus J). Then K=⟨Ps⟩K=\langle P_{s}\rangle and J=⟨Psβˆ’tβˆ’1⟩J=\langle P_{s-t-1}\rangle. However, Since sβˆ’(sβˆ’tβˆ’1)=t+1β‰₯3s-(s-t-1)=t+1\geq 3, this contradicts kβˆ’j=1k-j=1.

By the above arguments, the map Ο†\varphi turns out to be not surjective. Consequently, it follows that |Ξ”π’ͺ​(P)βˆ—|<|Ξ”π’žβ€‹(P)βˆ—||\Delta^{*}_{\mathcal{O}(P)}|<|\Delta^{*}_{\mathcal{C}(P)}|. From Lemma 3.3, we establish |Ξ”π’ͺ​(P)|<|Ξ”π’žβ€‹(P)||\Delta_{\mathcal{O}(P)}|<|\Delta_{\mathcal{C}(P)}|, as desired. ∎

In the following statement, we set the XX-poset to be {a,b,c,d,e}\{a,b,c,d,e\}, where d,e∈Psd,e\in P_{s}, c∈Psβˆ’1c\in P_{s-1}, and a,b∈Psβˆ’ta,b\in P_{s-t}, with 2≀s≀n2\leq s\leq n and 2≀t≀s2\leq t\leq s, following the same notation as in the proof of Theorem 3.6.

Corollary 3.7.

Let PP be a maximal ranked poset of rank nn which contains an XX-poset as a subposet. The number of triangular 22-faces of π’žβ€‹(P)\mathcal{C}(P) is equal to that of π’ͺ​(P)\mathcal{O}(P) plus

βˆ‘s=2nβˆ‘t=2sβˆ‘m=2|Ps|βˆ‘β„“=2|Psβˆ’t|(|Ps|m)​(|Psβˆ’t|β„“).\sum_{s=2}^{n}\;\sum_{t=2}^{s}\;\sum_{m=2}^{|P_{s}|}\;\sum_{\ell=2}^{|P_{s-t}|}\binom{|P_{s}|}{m}\binom{|P_{s-t}|}{\ell}.
Proof.

Let {d,e}βŠ‚Psβ€²βŠ‚Ps\{d,e\}\subset P^{\prime}_{s}\subset P_{s} and {a,b}βŠ‚Psβˆ’tβ€²βŠ‚Psβˆ’t\{a,b\}\subset P^{\prime}_{s-t}\subset P_{s-t}. Then {Psβˆ’tβ€²,Psβˆ’1,Psβ€²}βˆˆΞ”π’žβ€‹(P)βˆ—\{P^{\prime}_{s-t},P_{s-1},P^{\prime}_{s}\}\in\Delta^{*}_{\mathcal{C}(P)}, and there does not exist any {I,J,K}βˆˆΞ”π’ͺ​(P)βˆ—\{I,J,K\}\in\Delta^{*}_{\mathcal{O}(P)} such that φ​({I,J,K})={Psβˆ’tβ€²,Psβˆ’1,Psβ€²}\varphi(\{I,J,K\})=\{P^{\prime}_{s-t},P_{s-1},P^{\prime}_{s}\}, by the same argument as in the proof of Theorem 3.6. Fixing ss, there are βˆ‘m=2|Ps|(|Ps|m)\sum_{m=2}^{|P_{s}|}\binom{|P_{s}|}{m} ways to choose Psβ€²P^{\prime}_{s} and βˆ‘t=2sβˆ‘β„“=2|Psβˆ’t|(|Psβˆ’t|β„“)\sum_{t=2}^{s}\sum_{\ell=2}^{|P_{s-t}|}\binom{|P_{s-t}|}{\ell} ways to choose Psβˆ’tβ€²P^{\prime}_{s-t}. Therefore,

βˆ‘s=2n{βˆ‘m=2|Ps|(|Ps|m)β€‹βˆ‘t=2sβˆ‘β„“=2|Psβˆ’t|(|Psβˆ’t|β„“)}=βˆ‘s=2nβˆ‘t=2sβˆ‘m=2|Ps|βˆ‘β„“=2|Psβˆ’t|(|Ps|m)​(|Psβˆ’t|β„“).\sum_{s=2}^{n}\biggl{\{}\sum_{m=2}^{|P_{s}|}\binom{|P_{s}|}{m}\;\sum_{t=2}^{s}\sum_{\ell=2}^{|P_{s-t}|}\binom{|P_{s-t}|}{\ell}\biggr{\}}=\sum_{s=2}^{n}\;\sum_{t=2}^{s}\;\sum_{m=2}^{|P_{s}|}\;\sum_{\ell=2}^{|P_{s-t}|}\binom{|P_{s}|}{m}\binom{|P_{s-t}|}{\ell}.

∎

Acknowledgement

The author would like to thank Takayuki Hibi for numerous helpful discussion.

References

  • [1] I. Ahmad, G. Fourier, and M. Joswig, Order and chain polytopes of maximal ranked posets, arXiv : 2309.01626.
  • [2] R. Freij-Hollanti and T. LundstrΓΆm, ff-vector Inequalities for Order and Chain Polytopes, arXiv : 2312.13890.
  • [3] T. Hibi, N. Li, Cutting convex polytopes by hyperplanes, Mathematics 7 (2019), #381.
  • [4] T. Hibi, N. Li, Unimodular equivalence of order and chain polytopes, Math. Scand. 118 (2016), 5–12.
  • [5] T. Hibi, N. Li, Y. Sahara, A. Shikama, The numbers of edges of the order polytope and the chain polytope of a finite partially ordered set, Discrete Math. 340 (2017), 991–994.
  • [6] A. Mori, Faces of 2-dimensional simplex of order and chain polytopes, Mathematics 7 (2019), #851.
  • [7] R. Stanley, Two poset polytopes, Discrete Comput. Geom. 1 (1986), 9–23.