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

Interval Posets and Polygon Dissections

Eli Bagno Jerusalem College of Technology, 21 HaVaad HaLeumi St., Jerusalem, Israel    Estrella Eisenberg Jerusalem College of Technology, 21 HaVaad HaLeumi St., Jerusalem, Israel    Shulamit Reches and Moriah Sigron Jerusalem College of Technology, 21 HaVaad HaLeumi St., Jerusalem, Israel bagnoe@g.jct.ac.il
Abstract

The Interval poset of a permutation is an effective way of capturing all the intervals of the permutation and the inclusions between them and was introduced recently by Tenner. Thi paper explores the geometric interpretation of interval posets of permutations. We present a bijection between tree interval posets and convex polygons with non-crossing diagonals, offering a novel geometric perspective on this purely combinatorial concept. Additionally, we provide an enumeration of interval posets using this bijection and demonstrate its application to block-wise simple permutations.

1 Introduction

In [5], Tenner defined the concept of an interval poset of a permutation. This is an effective way of capturing all the intervals of a permutation and the set of inclusions between them in one glance. Tenner dealt with structural aspects of the interval poset and characterized the posets PP that can be seen as interval posets of some permutations.

An interval poset might correspond to more than one permutation. For instance, all simple permutations of a given order nn share the same interval poset. Tener, in the aforementioned paper, enumerated binary interval posets and binary tree interval posets but left open the following question:

Question 1.1.

How many tree interval posets have nn minimal elements?

This question was answered by Bouvel, Cioni and Izart in [3]. They also noted that the number of tree interval posets is equal to the number of ways to place non-crossing diagonals in a convex (n+1)(n+1)-gon such that no quadrilaterals are created.

In this work we suggest a simple bijection between the set of tree interval posets and the set of (n+2)(n+2)-gons, satisfying the conditions listed above. We use this bijection also for enumerating the whole set of interval posets by using a broader set of polygons. In [3], the enumeration of the entire set of interval posets was done in an algebraic way, using generating functions, while our bijection grants a geometric view to the interval posets.

Another set of interval posets that can be enumerated by polygons is the one corresponding to block-wise simple permutations, a term that was introduced in a recent paper by the current authors [2].

2 Background

Definition 2.1.

Let 𝒮n\mathcal{S}_{n} the symmetric group on nn elements. Let π=a1an𝒮n\pi=a_{1}\cdots a_{n}\in\mathcal{S}_{n}. An interval (or block) of π\pi is a non-empty contiguous sequence of entries aiai+1ai+ka_{i}a_{i+1}\cdots a_{i+k} whose values also form a contiguous sequence of integers. For a<ba<b, [a,b][a,b] denotes the interval of values that range from aa to bb. Clearly, [n]:=[1,n][n]:=[1,n] is an interval, as well as {i}\{i\} for each i[n]i\in[n]. These are called trivial intervals. The other intervals are called proper.

For example, the permutation π=314297856\pi=314297856 has [5,9]=97856[5,9]=97856 as a proper interval as well as the following proper intervals: [1,4],[5,6],[7,8],[7,9][1,4],[5,6],[7,8],[7,9], [5,8][5,8].

A permutation π𝒮n\pi\in\mathcal{S}_{n} is called simple if it does not have proper intervals. For example, the permutation 35172463517246 is simple.

Following Tenner [5], we define an interval poset for each permutation as follows:

Definition 2.2.

The interval poset of a permutation π𝒮n\pi\in\mathcal{S}_{n} is the poset P(π)P(\pi) whose elements are the non-empty intervals of π\pi; the order is defined by set inclusion (see for example Figures 1 and 2). The minimal elements are the intervals of size 11.

In [5], the interval poset is embedded in the plane so that each node’s direct descendants are increasingly ordered according to the minimum of each interval from left to right. We note that in [3] another embedding of the same poset was presented.

[1,7][1,7][1,6][1,6]{7}\{7\}[1,3][1,3][1,2][1,2][2,3][2,3]{4}\{4\}{5}\{5\}{6}\{6\}{1}\{1\}{2}\{2\}{3}\{3\}
Figure 1: Interval poset of the permutations: 5123647, 5321647, 4612357, 4632157, 7463215, 7461235, 7532164, 7512364

If π\pi is a simple permutation, the interval poset of π\pi comprises the entire interval [1,,n][1,\dots,n] with minimal elements {1},,{n}\{1\},\dots,\{n\} as its only descendants. Hence, all simple permutations of a given order nn share the same interval poset (see for example Figure 2).

[1,4][1,4]{1}\{1\}{2}\{2\}{3}\{3\}{4}\{4\}
Figure 2: Interval poset of permutations 3142 and 2413.

3 Geometrical view of interval posets

3.1 General interval posets

Bouvel, Cioni and Izart[3], provided a formula for the number of interval posets with nn minimal elements and added it to OEIS as sequence A348479 [4].

Here we provide a geometrical view to the interval posets by providing a bijection from the set of interval posets with nn minimal elements to a distinguished set of dissections of the convex (n+1)(n+1)- gon, which we define below.

We identify a polygon with its set of vertices and denote a diagonal or an outer edge of the polygon from vertex ii to vertex jj by {i,j}\{i,j\}.

Definition 3.1.

A dissection of an (n+1)(n+1)- gon will be called diagonally framed if for each two crossing diagonals, their vertices are connected to each other. Explicitly, if {a,b}\{a,b\} and {c,d}\{c,d\} are two crossing diagonals, then the diagonals or outer edges {a,d},{b,d},{c,b},{c,a}\{a,d\},\{b,d\},\{c,b\},\{c,a\} must also exist. See Figure 5 for an example.

Before we proceed, we have to present two observations which provide some details on the structure of interval posets and will be used in the sequel.

Observation 3.2.

Let π𝒮n\pi\in\mathcal{S}_{n}. If II and JJ are intervals of π\pi such that IJI\nsubseteq J and JIJ\nsubseteq I and IJI\cap J\neq\emptyset, then IJI\cap J, IJI\cup J, IJI-J and JIJ-I are intervals of π\pi.

For example, take π=3124576\pi=3124576, then I=[1,5]I=[1,5] and J=[4,7]J=[4,7] are intersecting intervals of π\pi and thus IJ=[1,7],IJ=[4,5],IJ=[1,3],JI=[6,7]I\cup J=[1,7],I\cap J=[4,5],I-J=[1,3],J-I=[6,7] are also intervals of π\pi, as can be seen in Figure 3 which depicts the permutation π\pi in the common graphical way.

Refer to caption
Figure 3: The permutation π=3124576\pi=3124576 and its blocks in a graphical way.
Observation 3.3.

If P(π)P(\pi) is the interval poset of π𝒮n\pi\in\mathcal{S}_{n}, then no element of P(π)P(\pi) has exactly 33 direct descendants, since every permutation of order 33 must contain a block of order 22.

Refer to caption
Figure 4: Right: the interval poset P. Left: the polygon Φ(P)\Phi(P)

We are ready now to present the main result of this subsection.

Theorem 3.4.

The number of interval posets with nn minimal elements is equal to the number of diagonally framed dissections of the convex (n+1n+1)-gon such that no quadrilaterals are present (see Figure 6 in the appendix for some examples of the bijection in small values of nn).

Proof.

We define a bijection between the set of interval posets with nn minimal elements and the set of diagonally framed dissections of convex (n+1)(n+1)- gons without quadrilaterals as follows:

Let PP be the interval poset of some π𝒮n\pi\in\mathcal{S}_{n}. We set Φ(P)\Phi(P) to be the convex (n+1)(n+1)-gon whose set of diagonals is

{{a,b+1}|[a,b] is an internal node of P},\{\{a,b+1\}|[a,b]\text{ is an internal node of }P\},

i.e. to each interval of the form [a,b][a,b] corresponds a diagonal {a,b+1}\{a,b+1\} in Φ(P)\Phi(P); note that singletones intervals correspond to outer edges in the polygon (see Figure 4 for an example).

We claim now that Φ(P)\Phi(P) must be a diagonally framed (n+1)(n+1) - gon. Indeed, if {a,c+1}\{a,c+1\} and {b,d+1}\{b,d+1\} are two crossing diagonals in Φ(P)\Phi(P), where abcda\leq b\leq c\leq d, then I=[a,c]I=[a,c] and J=[b,d]J=[b,d] are intersecting intervals in PP and by Observation 3.2 we have that IJ=[a,d]I\cup J=[a,d], IJ=[b,c]I\cap J=[b,c], IJ=[a,b1]I-J=[a,b-1] and JI=[c+1,d]J-I=[c+1,d] are intervals in PP corresponding respectively to the diagonals {a,d+1},{b,c+1},{a,b}\{a,d+1\},\{b,c+1\},\{a,b\} and {c+1,d+1}\{c+1,d+1\}. (See Figure 5 for an illustration).

Moreover, Φ(P)\Phi(P) must not contain any quadrilateral. Otherwise, if a<b<c<da<b<c<d are such that {a.b,c,d}\{a.b,c,d\} is a quadrilateral (without any subdivision) then PP must contain the intervals [a,b1],[b,c1],[c,d1][a,b-1],[b,c-1],[c,d-1] and [a,d1][a,d-1] so we must have that the first three intervals are direct descendants of the fourth one and they are the only ones. By Observation 3.3, this is impossible. ∎

Refer to caption
Figure 5:

3.2 Tree interval posets

A tree poset is a poset whose Hasse diagram is a tree.

In [3], the authors calculated the generating function of the number of tree interval posets using generating functions and mentioned that this is equal to the number of ways to place non-crossing diagonals in a convex (n+2)(n+2)-gon such that no quadrilaterals are created (sequence A054515 from OEIS [4]).

Using the function Φ\Phi defined above, one can easily produce a combinatorial proof of the following result.

Theorem 3.5.

The number of tree interval posets with nn minimal elements is equal to the number of non crossing dissections of the convex (n+1)(n+1)- gon such that no quadrilaterals are present (see Figure 7 in the appendix for some examples of the bijection).

Proof.

We use the same mapping Φ\Phi which was applied in the proof of Theorem 3.4. It is now sufficient to prove that no crossing diagonals are obtained. This is implied by the fact that intersecting diagonals stem from intersecting intervals which can not exist in a tree since they cause a circle. (See Figure 1).

3.3 Interval posets of block-wise simple permutations

In [2], the current authors introduced the notion of block-wise simple permutations. We cite here the definition:

Definition 3.6.

A permutation π𝒮n\pi\in\mathcal{S}_{n} is called block-wise simple if it has no interval of the form p1p2p_{1}\oplus p_{2} or p1p2p_{1}\ominus p_{2}, where \oplus and \ominus stand for direct and skew sums of permutations respectively.

There are no block-wise simple permutations of orders 22 and 33. For n{4,5,6}n\in\{4,5,6\}, a permutation is block-wise simple, if and only if it is simple. One of the first nontrivial examples of block-wise simple permutations is 42537164253716.

In [2], the current authors enumerated the interval posets of block-wise permutations.

The first few values of the sequence of these numbers are 1,1,1,5,10,16,45,109,222,5401,1,1,5,10,16,45,109,222,540. This is sequence A054514 from OEIS [4] which also counts the number of ways to place non-crossing diagonals in a convex (n+4)(n+4)-gon such that there are no triangles or quadrilaterals.

The geometrical interpretation of interval posets of block-wise permutations is as follows:

Theorem 3.7.

The number of interval posets that represent a block-wise simple permutation of order nn is equal to the number of ways to place non-crossing diagonals in a convex (n+1n+1)-gon such that no triangles or quadrilaterals are present (see Figure 8 in the appendix for some examples of the bijection).

Proof.

We use again the mapping Φ\Phi, defined earlier. In [5] (Theorem 6.1), the author claimed that P(σ)P(\sigma) is a tree interval poset if and only if σ\sigma contains no interval of the form p1p2p3p_{1}\oplus p_{2}\oplus p_{3} or p1p2p3p_{1}\ominus p_{2}\ominus p_{3}. From here, and by Definition 3.6, it is obvious that an interval poset of a block-wise simple permutation is a tree. Hence it is sufficient to prove that for an interval poset PP of a block-wise permutation, Φ(P)\Phi(P) has no triangles. This holds due to the fact that if Φ(P)\Phi(P) contains a triangle with edges {a,b},{b,c},{a,c}\{a,b\},\{b,c\},\{a,c\} with a<b<ca<b<c then PP must contain the intervals [a,b1],[b,c1][a,b-1],[b,c-1] and [a,c1][a,c-1] and thus [a,c1][a,c-1] is the direct parent of [a,b1][a,b-1] and [b,c1][b,c-1] which contradicts the definition of block-wise simple permutations. ∎

References

4 Appendix

Refer to caption
Refer to caption
Figure 6: The bijection for small values of nn
Refer to caption
Refer to caption
Refer to caption
Refer to caption
Figure 7: Examples for the bijection of tree intervals for small values of nn
Refer to caption
Refer to caption
Refer to caption
Figure 8: Examples for the bijection of block-wise simple intervals for small values of nn