Tournament transitivity of graphs

Abstract

Let $G=(V,E)$ be a graph where $V$ and $E$ are the vertex and edge sets, respectively. For two disjoint subsets $A$ and $B$ of $V$, we say $A$ dominates $B$ if every vertex of $B$ is adjacent to at least one vertex of $A$ in $G$. A vertex partition $\pi=\{V_{1},V_{2},\ldots,V_{k}\}$ of $G$ is called a transitive partition of size $k$ if $V_{i}$ dominates $V_{j}$ for all $1\leq i<j\leq k$. A vertex partition $\pi=\{V_{1},V_{2},\ldots,V_{k}\}$ of $G$ is called a tournament transitive partition of size $k$ if $V_{i}$ dominates $V_{j}$ for all $1\leq i<j\leq k$ and $V_{j}$ does not dominate $V_{i}$ for $i<j$. The maximum integer $k$ for which the above partition exists is called tournament transitivity of $G$, and it is denoted by $TTr(G)$. The Maximum Tournament Transitivity Problem is to find a tournament transitive partition of a given graph with the maximum number of parts. In this article, we study this variation of transitive partition from a structure and algorithmic point of view. We show that the decision version of this problem is NP-complete for chordal graphs (connected), perfect elimination bipartite graphs (disconnected) and doubly chordal graphs (disconnected). On the positive side, we prove that this problem can be solved in polynomial time for trees. Furthermore, we characterize Type-I BCG with equal transitivity and tournament transitivity and find some sufficient conditions under which the above two parameters are equal for a Type-II BCG. Finally, we show that for Type-III BCG, these two parameters are never equal.

Keywords. Tournament transitivity, NP-completeness, chordal graphs, Polynomial-time algorithm, trees, bipartite chain graphs.

1 Introduction

Partitioning a graph is one of the fundamental problems in graph theory. In the partitioning problem, the objective is to partition the vertex set (or edge set) into some parts with desired properties, such as independence, minimal edges across partite sets, etc. In literature, partitioning the vertex set into certain parts so that the partite sets follow particular domination relations among themselves has been studied [2, 7, 9, 10, 11, 15, 16, 18]. Let $G$ be a graph with $V(G)$ as its vertex set and $E(G)$ as its edge set. When the context is clear, $V$ and $E$ are used instead of $V(G)$ and $E(G)$. The neighbourhood of a vertex $v\in V$ in a graph $G=(V,E)$ is the set of all adjacent vertices of $v$ and is denoted by $N_{G}(v)$. The degree of a vertex $v$ in $G$, denoted as $\deg_{G}(v)$, is the number of edges incident to $v$. A vertex $v$ is said to dominate itself and all its neighbouring vertices. A dominating set of $G=(V,E)$ is a subset of vertices $D$ such that every vertex $x\in V\setminus D$ has a neighbour $y\in D$, that is, $x$ is dominated by some vertex $y$ of $D$. For two disjoint subsets $A$ and $B$ of $V$, we say $A$ dominates $B$ if every vertex of $B$ is adjacent to at least one vertex of $A$.

There has been a lot of research on graph partitioning problems based on a domination relationship between the different sets. Cockayne and Hedetniemi introduced the concept of domatic partition of a graph $G=(V,E)$ in 1977, in which the vertex set is partitioned into $k$ parts, say $\pi=\{V_{1},V_{2},\ldots,V_{k}\}$, such that each $V_{i}$ is a dominating set of $G$ [4]. The number representing the highest possible order of a domatic partition is called the domatic number of G, denoted by $d(G)$. Another similar type of partitioning problem is the Grundy partition. Christen and Selkow introduced a Grundy partition of a graph $G=(V,E)$ in 1979 [3]. In the Grundy partitioning problem, the vertex set is partitioned into $k$ parts, say $\pi=\{V_{1},V_{2},\ldots,V_{k}\}$, such that each $V_{i}$ is an independent set and for all $1\leq i<j\leq k$, $V_{i}$ dominates $V_{j}$. The maximum order of such a partition is called the Grundy number of $G$, denoted by $\Gamma(G)$. In 2018, J. T. Hedetniemi and S. T. Hedetniemi [11] introduced a transitive partition as a generalization of the Grundy partition. A transitive partition of size $k$ is defined as a partition of the vertex set into $k$ parts, say $\pi=\{V_{1},V_{2},\ldots,V_{k}\}$, such that for all $1\leq i<j\leq k$, $V_{i}$ dominates $V_{j}$. The maximum order of such a transitive partition is called the transitivity of $G$ and is denoted by $Tr(G)$. Recently, in 2020, Haynes et al. generalized the idea of domatic partition as well as transitive partition and introduced the concept of upper domatic partition of a graph $G$, where the vertex set is partitioned into $k$ parts, say $\pi=\{V_{1},V_{2},\ldots,V_{k}\}$, such that for each $i,j$, with $1\leq i<j\leq k$, either $V_{i}$ dominates $V_{j}$ or $V_{j}$ dominates $V_{i}$ or both [10]. The maximum order of such an upper domatic partition is called the upper domatic number of $G$, denoted by $D(G)$. All these problems, domatic number [2, 20, 21], Grundy number [5, 7, 12, 18, 19], transitivity [9, 11, 15, 16], upper domatic number [10, 17] have been extensively studied both from an algorithmic and structural point of view.

The concept of tournament transitive partition was introduced by Haynes et al. in 2019 [9] as an open problem. So far, no research has been done on this variation of transitivity. In this article, we study this variation of the transitivity problem from a structural and algorithmic point of view. For two disjoint subsets $A$ and $B$, we say $A$ dominates $B$ if every vertex of $B$ is adjacent to at least one vertex of $A$. A vertex partition $\pi=\{V_{1},V_{2},\ldots,V_{k}\}$ of $G$ is called a tournament transitive partition of size $k$ if $V_{i}$ dominates $V_{j}$ and $V_{j}$ does not dominate $V_{i}$ for all $1\leq i<j\leq k$. The maximum integer $k$ for which the above partition exists is called tournament transitivity of $G$, and it is denoted by $TTr(G)$. A tournament transitive partition of order $TTr(G)$ is called a $TTr(G)$-partition. The Maximum Tournament Transitivity Problem is to find a tournament transitive partition of a given graph with the maximum number of parts. Note that every tournament transitive partition is also a transitive partition. Therefore, for any graph $G$, $1\leq TTr(G)\leq Tr(G)\leq n$. The Maximum Tournament Transitivity Problem and its corresponding decision version are defined as follows.

Maximum Tournament Transitivity Problem(MTTP) Instance: A graph $G=(V,E)$ Solution: A tournament transitive partition of $G$ with maximum size

Maximum Tournament Transitivity Decision Problem(MTTDP) Instance: A graph $G=(V,E)$, integer $k$ Question: Does $G$ have a tournament transitive partition of order at least $k$?

Note that the only tournament transitive partition for the graph $K_{n}$ is $\{V\{K_{n}\}\}$. But on the other hand, according to [11], we know that $Tr(K_{n})=n$. This distinction motivates us to investigate this new parameter. The other similarities and differences can be found in the properties of tournament transitivity section. The rest of the paper is organized as follows: Section 2 contains basic definitions and notations that are followed throughout the article. This section also discusses the properties of tournament transitivity of graphs. Section 3 shows that the MTTDP is NP-complete in chordal graphs (Connected), perfect elimination bipartite graphs (disconnected) and doubly chordal graphs (disconnected). In Section 4, we design a polynomial-time algorithm for solving MTTP in trees. Finally, Section 5 concludes the article.

2 Preliminaries

2.1 Notation and definition

Let $G=(V,E)$ be a graph with $V$ and $E$ as its vertex and edge sets, respectively. A graph $H=(V^{\prime},E^{\prime})$ is said to be a subgraph of a graph $G=(V,E)$ if and only if $V^{\prime}\subseteq V$ and $E^{\prime}\subseteq E$. For a subset $S\subseteq V$, the induced subgraph on $S$ of $G$ is defined as the subgraph of $G$ whose vertex set is $S$ and edge set consists of all of the edges in $E$ that have both endpoints in $S$, and it is denoted by $G[S]$. The complement of a graph $G=(V,E)$ is the graph $\overline{G}=(\overline{V},\overline{E})$, such that $\overline{V}=V$ and $\overline{E}=\{uv|uv\notin E\text{ and }u\neq v\}$. The open neighbourhood of a vertex $x\in V$ is the set of vertices $y$ adjacent to $x$, denoted by $N_{G}(x)$. The closed neighborhood of a vertex $x\in V$, denoted as $N_{G}[x]$, is defined by $N_{G}[x]=N_{G}(x)\cup\{x\}$. Let $S\subseteq V$, then we define $N_{G}(S)=\displaystyle\bigcup_{x\in S}N_{G}(x)$.

A subset of $S\subseteq V$ is said to be an independent set of $G$ if every pair of vertices in $S$ are non-adjacent. A subset of $K\subseteq V$ is said to be a clique of $G$ if every pair of vertices in $K$ are adjacent. The cardinality of a clique of maximum size is called clique number of $G$, and it is denoted by $\omega(G)$. A path of length $k-1$ in a graph $G=(V,E)$, denoted by $P_{k}$, is a sequence of distinct vertices $v_{1},\ldots,v_{k}$ such that $v_{i-1}v_{i}\in E$ for $2\leq i\leq k$. Vertices $v_{1}$ and $v_{k}$ are called the end vertices of $P_{k}$. For any two vertices $y,z$, we denoted a path starting with $y$ and ending at $z$ by $yPz$.

A graph is called bipartite if its vertex set can be partitioned into two independent sets. A star $S_{t}$ is the complete bipartite graph $K_{1,t}$. An edge $uv$ in a bipartite graph $G$ is called bisimplicial if $N(u)\cup N(v)$ induces a biclique in $G$. For an edge ordering $(e_{1},e_{2},\ldots,e_{k})$, let $S_{i}$ be the set of endpoints of $\{e_{1},e_{2},\ldots,e_{i}\}$ and $S_{0}=\emptyset$. An ordering $(e_{1},e_{2},\ldots,e_{k})$ is a perfect edge elimination ordering for a bipartite graph $G=(V,E)$ if $G[V\setminus S_{k}]$ has no edges and each edge $e_{i}$ is a bisimplicial edge in $G[V\setminus S_{i-1}]$. A graph $G$ is a perfect elimination bipartite if and only if it admits a perfect edge elimination ordering [8].

A bipartite graph $G=(X\cup Y,E)$ is called a bipartite chain graph if there exists an ordering of vertices of $X$ and $Y$, say $\sigma_{X}=(x_{1},x_{2},\ldots,x_{n_{1}})$ and $\sigma_{Y}=(y_{1},y_{2},\ldots,y_{n_{2}})$, such that $N(x_{n_{1}})\subseteq N(x_{n_{1}-1})\subseteq\ldots\subseteq N(x_{2})\subseteq N(x_{1})$ and $N(y_{n_{2}})\subseteq N(y_{n_{2}-1})\subseteq\ldots\subseteq N(y_{2})\subseteq N(y_{1})$. Such ordering of $X$ and $Y$ is called a chain ordering, and it can be computed in linear time [14]. Let $G$ be a bipartite chain graph but not a complete bipartite graph. Also, let $\sigma_{X}=(x_{1},x_{2},\ldots,x_{n_{1}})$ and $\sigma_{Y}=(y_{1},y_{2},\ldots,y_{n_{2}})$ be the chain ordering of $G$ and $t$ be the maximum integer such that $G$ contains a $K_{t,t}$ as an induced subgraph. A bipartite chain graph $G$ is called $(i)$ Type-I BCG if $x_{t+1}y_{t}\notin E(G)$ and $x_{t}y_{t+1}\notin E(G)$, $(ii)$ Type-II BCG if either $x_{t+1}y_{t}\in E(G)$ or $x_{t}y_{t+1}\in E(G)$ not both, $(iii)$ Type-III BCG if $x_{t+1}y_{t}\in E(G)$ and $x_{t}y_{t+1}\in E(G)$.

An edge between two non-consecutive vertices of a cycle is called a chord. If every cycle in $G$ of length at least four has a chord, then $G$ is called a chordal graph. A vertex $v\in V$ is called a simplicial vertex of $G$ if $N_{G}[v]$ induces a clique in $G$. A perfect elimination ordering (PEO) of $G$ is an ordering of the vertices, say $\sigma=(v_{1},v_{2},\ldots,v_{n})$, such that $v_{i}$ is a simplicial vertex of $G_{i}=G[\{v_{i},v_{i+1},\ldots,v_{n}\}]$ for all $1\leq i\leq n$. Chordal graphs can be characterized by the existence of PEO; that is, a graph $G$ is chordal if and only if $G$ has a PEO [6]. A vertex $u\in N_{G}[v]$ is called a maximum neighbour of $v$ in $G$ if $N_{G}[w]\subseteq N_{G}[u]$ for every vertex $w\in N_{G}[v]$. A vertex $v$ in $G$ is called doubly simplicial if it is simplicial and has a maximum neighbour in $G$. An ordering $\delta=(v_{1},v_{2},\ldots,v_{n})$ of vertices of $G$ is called a doubly perfect elimination ordering (DPEO) if $v_{i}$ is a doubly simplicial vertex in $G_{i}=G[\{v_{i},v_{i+1},\ldots,v_{n}\}]$ for all $1\leq i\leq n$. A graph is doubly chordal if it admits a doubly perfect elimination ordering [1].

2.2 Properties of tournament transitivity

In this subsection, we present some properties of tournament transitivity that motivate us to study this variation of transitivity. First, we show the following bounds for tournament transitivity.

Proposition 1.

For any graph $G$, $TTr(G)\leq min\{Tr(G),Tr(\overline{G})\}$.

Proof.

Let $\pi=\{V_{1},V_{2},\ldots,V_{k}\}$ be a $TTr$-partition of $G$. By the definition of tournament transitive partition, $\pi$ is also a transitive partition of $G$. Thus, we can say that $TTr(G)\leq Tr(G)$. Now we show that $\pi$ is also a transitive partition of $\overline{G}$. Since $\pi$ is a tournament transitive partition, for $i<j$, $V_{i}$ dominates $V_{j}$ and $V_{j}$ does not dominate $V_{i}$. Therefore, there exists a vertex $x\in V_{i}$ such that $xy\notin E(G)$ for all $y\in V_{j}$, and hence $V_{i}$ dominates $V_{j}$ in $\overline{G}$. Thus, $\pi$ is also a transitive partition of $\overline{G}$. Therefore, $TTr(G)\leq Tr(\overline{G})$. Hence, for any graph $G$, $TTr(G)\leq\min\{Tr(G),Tr(\overline{G})\}$. ∎

Next, we present two upper bounds: one in terms of the maximum and minimum degrees of a graph and the other in terms of the number of vertices.

Proposition 2.

For any graph $G$, $TTr(G)\leq min\{\Delta(G)+1,n-\delta(G)\}$, where $n$ is the number of vertex and $\Delta(G),\delta(G)$ are maximum and minimum degree of $G$, respectively.

Proof.

From [9], we know that $Tr(G)\leq\Delta(G)+1$, for any graph $G$. Also, from the Proposition 1, we have $TTr(G)\leq min\{Tr(G),Tr(\overline{G})\}$. Therefore, $TTr(G)\leq min\{\Delta(G)+1,\Delta(\overline{G})+1\}=min\{\Delta(G)+1,n-\delta(G)\}$. ∎

Proposition 3.

For any graph $G$, $1\leq TTr(G)\leq\left\lfloor\frac{n+1}{2}\right\rfloor$, where $n$ is the number of vertex of $G$.

Proof.

Taking all vertices in one set will produce a tournament transitive partition of size $1$. So, $TTr(G)\geq 1$. Now, let $\pi=\{V_{1},V_{2},\ldots,V_{k}\}$ be a $TTr$-partition of $G$. Consider $x\in V_{k}$. As $\pi$ $V_{i}$ dominates $V_{k}$ in $G$ for all $1\leq i<k$, $deg_{G}(x)\geq k-1$. Also, we know that $V_{k}$ does not dominate $V_{j}$ for all $1\leq j<k$. Therefore, $deg_{\overline{G}}(x)\geq k-1$. So, $deg_{G}(x)+deg_{\overline{G}}(x)\geq 2k-1$, which implies $n-1\geq deg_{G}(x)+deg_{\overline{G}}(x)\geq 2k-1$. Hence, $k=TTr(G)\leq\left\lfloor\frac{n+1}{2}\right\rfloor$. ∎

Similarly, as transitivity, if we union any two sets of a tournament transitive partition of order $k$, we create a tournament transitive partition of order $k-1$. Hence, we have a similar proposition as transitivity.

Proposition 4.

If $\pi=\{V_{1},V_{2},\ldots,V_{k}\}$ is a tournament transitive partition of a graph $G$, then for any two sets $V_{i}$ and $V_{j}$ in $\pi$, where $i<j$, the partition $\pi=\{V_{1},V_{2},\ldots,V_{i-1},V_{i}\cup V_{j},V_{i+1},\ldots,V_{j-1},V_{j+1},\ldots,V_{k}\}$ is a tournament transitive partition.

Given this, we have the following interpolation result.

Proposition 5.

Let $G$ be a graph and $k$ be the order of a tournament transitive partition of $G$, then for every $j$, $1\leq j\leq k$, $G$ has a tournament transitive partition of order $j$.

Proposition 6.

Let $G$ be a connected graph and $TTr(G)=k$, $k\geq 3$. Then there exists a tournament transitive partition of $G$ of size $k$, say $\pi=\{V_{1},V_{2},\ldots,V_{k}\}$ of $G$, such that $|V_{k}|=1$, $|V_{k-1}|=2$. Moreover, $V_{k-1}=\{x,y\}$ and $V_{k}=\{z\}$ such that $xz\in E(G)$ and $yz\notin E(G)$.

Next, we find the tournament transitivity of complete graphs, paths, cycles and complete bipartite graphs.

Proposition 7.

For the complete graph $K_{n}$ of $n$ vertices $TTr(K_{n})=1$.

Proof.

The partition $\{V(K_{n})\}$ is a tournament transitive partition of $K_{n}$ of size $1$. Moreover, from the Proposition 2, we have $TTr(G)\leq min\{\Delta(K_{n})+1,n-\delta(K_{n})\}=min\{n,1\}=1$. Therefore, $TTr(K_{n})=1$. ∎

Proposition 8.

Let $P_{n}$ be a path of $n$ vertices, and then the tournament transitivity of $P_{n}$ is given as follows:

$TTr(P_{n})=\begin{cases}1&n=1,2\\ 2&n=3,4\\ 3&n\geq 5\end{cases}$

Proof.

For $n=1,2$, as $P_{1}$ and $P_{2}$ are $K_{1}$ and $K_{2}$, respectively, form the Proposition 7, we have $TTr(P_{n})=1$ for $n=1,2$. Let us consider $n=3,4$. According to Proposition 3, we have $TTr(P_{n})\leq 2$. Also, from the Figure1, it is clear that $TTr(P_{n})\geq 2$. Hence, $TTr(P_{n})=2$ for $n=2,3$.

Figure 1: Tournament transitive partition of $P_{n}$, $n\geq 3$

Assume $n=5$. We show that $TTr(P_{5})=3$. From the Proposition1, we have $TTr(P_{5})\leq 3$. Let $\{v_{1},v_{2},v_{3},v_{4},v_{5}\}$ be the vertices of $P_{5}$, where $v_{1},v_{5}$ are its end vertices. Consider a vertex partition $\pi=\{V_{1},V_{2},V_{3}\}$ as follows: $V_{1}=\{v_{2},v_{5}\}$, $V_{2}=\{v_{1},v_{3}\}$, and $V_{3}=\{v_{4}\}$ (see Figure 1). We show that $\pi$ is a tournament transitive partition. Clearly, $V_{i}$ dominates $V_{j}$, and $V_{j}$ does not dominate $V_{i}$, for all $1\leq i<j\leq j$. Therefore, $TTr(P_{5})=3$. Finally, consider $n>5$. Also, let $\{v_{1},v_{2},v_{3},v_{4},v_{5},V_{6},\ldots,v_{n}\}$ be the path $P_{n}$. Consider a vertex partition $\pi=\{V_{1},V_{2},V_{3}\}$ as follows: $\{v_{2},v_{5}\}\subseteq V_{1}$, $V_{2}=\{v_{1},v_{3}\}$, and $V_{3}=\{v_{4}\}$ and all the other vertices in $V_{1}$ (see Figure 1). Similarly, we can show that $TTr(P_{n})=3$ for all $n>5$. Hence, we have $TTr(P_{n})=3$, $n\geq 5$. ∎

Proposition 9.

Let $C_{n}$ be a path of $n$ vertices, and then the tournament transitivity of $P_{n}$ is given as follows:

$TTr(C_{n})=\begin{cases}1&n=3\\ 2&n=4,5\\ 3&n\geq 6\end{cases}$

Proof.

Note that $TTr(C_{n})\leq 3$, for all $n\geq 3$. We show that $TTr(C_{5})=2$; for the other cases, we can easily prove the statements. If possible, assume $TTr(C_{5})=3$ and $\{V_{1},V_{2},V_{3}\}$ be a tournament transitive partition of $C_{5}$. Also, assume $C_{5}=\{v_{1},v_{2},v_{3},v_{4},v_{5}\}$. Without loss of generality, assume $v_{1}\in V_{3}$. As $V_{2},V_{2}$ dominate $V_{3}$, each of $V_{1},V_{2}$ contains exactly one vertex from $\{v_{2},v_{5}\}$. Again, $V_{3}$ does not dominate $V_{1},V_{2}$; there exists a vertex from $\{v_{3},v_{4}\}$ that must be in $V_{2}$, and a vertex from $\{v_{3},v_{4}\}$ must be in $V_{1}$. In this situation, $V_{1}$ dominates $V_{2}$, and $V_{2}$ dominates $V_{1}$, a contradiction as $\pi$ is a tournament transitive partition. Hence, $TTr(C_{5})=2$. ∎

Proposition 10.

For a complete bipartite graph $K_{m,n}$ with $m+n$ vertices $TTr(K_{m,n})=1$ if and only if $m=n=1$.

Proof.

First, we show that $TTr(K_{m,n})=2$ if either $m\neq 1$ or $n\neq 1$. Let the vertex set of $K_{m,n}$ be $V=X\cup Y$ such that $|X|=m,|Y|=n$ and $m\geq 2$. Consider a vertex partition $\pi=\{V_{1},V_{2}\}$, where $V_{1}=(X\setminus\{x\})\cup Y$, $V_{2}=\{x\}$ for some $x\in X$. Since $m\geq 2$, $\pi$ is a tournament transitive partition of $K_{m,n}$ and hence $TTr(K_{m,n})\geq 2$. To prove $TTr(K_{m,n})=2$, we show that $TTr(K_{m,n})$ cannot be more than $2$. If possible, let $TTr(K_{m,n})=k\geq 3$ and $\pi=\{V_{1},V_{2},\ldots,V_{k}\}$ be a tournament transitive partition of $K_{m,n}$ of size $k$. We show that $V_{i}$, for $3\leq i\leq k$, cannot contain any vertex from $X$ or $Y$. If $V_{i}$ contains a vertex, say $x$, from $X$, there must exists $y_{1}\in V_{1}$ and $y_{2}\in V_{2}$ such that $xy_{1},xy_{2}\in E(G)$. As $G$ is a complete bipartite graph and $V_{i}$ does not dominate $V_{1}$ and $V_{2}$, there must be $x_{1}\in V_{1}$ and $x_{2}\in V_{2}$. So, $V_{1}$ and $V_{2}$ contains vertices from $X$ and $Y$ and hence $V_{1}$ dominates $V_{2}$ and $V_{2}$ dominates $V_{1}$, which is a contradiction as $\pi$ is a tournament transitive partition of $K_{m,n}$. On the other hand, assume $V_{i}$ contains a vertex, say $y$ from $Y$. Similarly, we can show a contradiction. Therefore, $TTr(K_{m,n})=2$, when $m\geq 2$. With similar arguments, we can show that $TTr(K_{m,n})=2$, when $n\geq 2$. Hence, $TTr(K_{m,n})=1$, if $m=1$ and $n=1$.

For the converse part, assume $m=1$ and $n=1$. Clearly, $K_{m,n}$ is a $P_{2}$ and from Proposition 7, we have $TTr(K_{m,n})=1$. ∎

For the transitivity, if $H$ is a subgraph of $G$, then $Tr(H)\leq Tr(G)$ [11]. But for tournament transitivity, this is not true. For example, let us consider $G=C_{5}$ and $H=P_{5}$. Clearly, $H$ is a subgraph of $G$, and $TTr(C_{5})=2$ (by Proposition 9), $TTr(P_{5})=3$ (by Proposition 8). Another example is $G=K_{4}$ and $H=C_{4}$, where $TTr(K_{4})=1$ and $TTr(C_{4})=2$.

In the above examples, neither of the subgraphs is an induced one. However, the subgraph containment relation holds for tournament transitive partition if we consider an induced subgraph instead of a subgraph.

Proposition 11.

If H is an induced subgraph of a graph G, then $TTr(H)\leq TTr(G)$.

Proof.

Let $\pi=\{V_{1},V_{2},\ldots,V_{k}\}$ be a $TTr$-transitive partition of $H$. Consider $\pi^{\prime}=\{V_{1}^{\prime},V_{2}^{\prime},\ldots,V_{k}^{\prime}\}$ a vertex partition of $G$ such that $V_{1}^{\prime}=(V(G)\setminus V(H))\cup V_{1}$ and $V_{p}^{\prime}=V_{p}$ for all $2\leq p\leq k$. As $H$ is an induced subgraph, $V_{i}^{\prime}$ dominates $V_{j}^{\prime}$, but $V_{j}^{\prime}$ does not dominate $V_{i}^{\prime}$ in $\pi^{\prime}$. So, $\pi^{\prime}$ is a tournament transitive partition of size $k$. Therefore, $k=TTr(H)\leq TTr(G)$. ∎

For the case of transitivity, if $G$ is a disconnected graph with connected component $C_{1},C_{2},\ldots,C_{t}$, then $Tr(G)=\max\{Tr(C_{i})|1\leq i\leq k\}$ [9]. But we show that for tournament transitivity, $TTr(G)\geq\max\{TTr(C_{i})|1\leq i\leq k\}$ and the difference $TTr(G)-\max\{TTr(C_{i})|1\leq i\leq k\}$ can be arbitrarily large. First, we prove that if $G$ is a disjoint union of $t(\leq n)$ number of complete graphs of $n$ vertices, then $TTr(G)=t$.

Lemma 12.

Let $G$ be the disjoint union of $t(\leq n)$ number of a complete graph of $n$ vertices. Then $TTr(G)=t$.

Proof.

Let $G=K_{n}\cup K_{n}\cup\ldots\cup K_{n}$ ($t$ times). Let $\pi^{j}=\{V_{1}^{j},V_{2}^{j},\ldots,V_{t}^{j},V_{t+1}^{j},\ldots,V_{n}^{j}\}$ be a transitive partition of size $n$ of the $j$-th component of $G$, and also we denote the $j$-th component of $G$ as $K_{n}^{j}$. Let us consider $\pi=\{V_{1},V_{2},\ldots,V_{t}\}$ be a vertex partition of $G$ of size $t$ as follows:

$$V_{1}=(\bigcup_{j=1}^{t}(V_{t+1}^{j}\cup V_{t+2}^{j}\cup\ldots\cup V_{n}^{j}))\cup(V_{1}^{1}\cup V_{1}^{2}\cup\ldots\cup V_{1}^{t-1}\cup(V_{1}^{t}\cup V_{2}^{t}\cup\ldots\cup V_{t}^{t})),$$

$$V_{j}=(V_{j}^{1}\cup V_{j}^{2}\cup\ldots\cup V_{j}^{t-j}\cup(V_{j}^{t-j+1}\cup V_{j+1}^{t-j+1}\cup\ldots\cup V_{t}^{t-j+1}))$$

for all $2\leq j\leq t$ (See Figure 2 for reference).

Figure 2: $(a)$ represents partitions $\pi^{j}$ and $(b)$ represents the partition $\pi$

Now we show that $\pi$ is a tournament transitive partition of size $t$. Let us consider $V_{i}$ and $V_{j}$, for $i<j$. Clearly, $V_{i}\supseteq(V_{i}^{1}\cup V_{i}^{2}\cup\ldots\cup V_{i}^{t-i}\cup(V_{i}^{t-i+1}\cup V_{i+1}^{t-i+1}\cup\ldots\cup V_{t}^{t-i+1}))$ and $V_{j}=(V_{j}^{1}\cup V_{j}^{2}\cup\ldots\cup V_{j}^{t-j}\cup(V_{j}^{t-j+1}\cup V_{j+1}^{t-j+1}\cup\ldots\cup V_{t}^{t-j+1}))$. Since $\pi^{p}$ is a transitive partition of $K_{n}^{p}$, $V_{i}^{p}$ dominates $V_{j}^{p}$ for all $1\leq p\leq t-j$. Again, as $i<j$, which implies $t-i\geq t-j+1$ and $V_{i}^{t-j+1}\subseteq V_{i}$. So, $V_{i}^{t-j+1}$ dominates $(V_{j}^{t-j+1}\cup V_{j+1}^{t-j+1}\cup\ldots\cup V_{t}^{t-j+1})$ as $K_{n}^{t-j+1}$ is a complete graph. Therefore, $V_{i}$ dominates $V_{j}$ for all $1\leq i<j\leq t$. However, $V_{j}$ does not dominate $V_{i}$, as $V_{i}$ contains vertices from $K_{n}^{t-i}$, whereas $V_{j}$ does not contain any vertex from $K_{n}^{t-i}$. So, $\pi$ is a tournament transitive partition of $G$, and hence $TTr(G)\geq t$.

To prove $TTr(G)=t$, we show that $TTr(G)$ cannot be more than $t$. If possible, let $TTr(G)=k\geq t+1$. By the Proposition 5, we have a tournament transitive partition of $G$ of size $t+1$, say $\pi=\{V_{1},V_{2},\ldots,V_{t+1}\}$. Without loss of generality, assume $V_{t+1}$ contains a vertex from $K_{n}^{1}$. As $\pi$ is a tournament transitive partition, $V_{t+1}$ does not dominate $V_{t}$, which implies there exists a vertex $x\in V_{t}$ and $x$ does not belong to $K_{n}^{1}$. Assume $x\in K_{n}^{2}$. Again, as $V_{t}$ does not dominate $V_{t-1}$ and $K_{n}^{1}\cap V_{t-1},K_{n}^{2}\cap V_{t-1}$ are non-empty, there exists a vertex $y\in V_{t-1}$, and $y$ must be a vertex that does not belong to $K_{n}^{1}\cup K_{n}^{2}$. Assume $y\in K_{n}^{3}$. If we go like this, we end up with a situation where we need a vertex in $V_{1}$ that does not belong to $K_{n}^{1}\cup K_{n}^{2}\cup\ldots\cup K_{n}^{t}$, a contradiction. As a result, $\pi$ cannot be a tournament transitive partition of $G$, and hence $TTr(G)\leq t$. Therefore, $TTr(G)=t$. ∎

Proposition 13.

Let $G$ is a disconnected graph with connected component $C_{1},C_{2},\ldots,C_{t}$, then $TTr(G)\geq\max\{Tr(C_{i})|1\leq i\leq k\}$ and the difference $Tr(G)-\max\{Tr(C_{i})|1\leq i\leq k\}$ can be arbitrarily large.

Proof.

Let $C_{i}$ be the component such that $TTr(C_{i})=\max\{TTr(C_{i})|1\leq i\leq k\}$. Also, let $\pi=\{V_{1},V_{2},\ldots,V_{k}\}$ be a $TTr$-partition of $C_{i}$. Now if we consider a vertex partition $\pi^{\prime}=\{V_{1}^{\prime},V_{2}^{\prime},\ldots,V_{k}^{\prime}\}$, where $V_{1}^{\prime}=V_{1}\displaystyle\bigcup_{p=1,p\neq i}^{t}V(C_{p})$ and $V_{i}^{\prime}=V_{i}$, for all $2\leq i\leq k$. Clearly, $\pi^{\prime}$ is a tournament transitive partition of $G$ of size $k$. Therefore, $TTr(G)\geq\max\{Tr(C_{i})|1\leq i\leq k\}$.

Now let us consider $G$ to be a graph, which is the disjoint union of $t(\leq n)$ number of complete graphs of $n$ vertices. Then by the Lemma 12 and Proposition 7, we have $TTr(G)=t$ and $TTr(C_{i})=1$, for all $i$. Therefore, the difference $Tr(G)-\max\{Tr(C_{i})|1\leq i\leq k\}=t-1$ arbitrarily large for $t$. ∎

Next, we characterize connected graphs $G$ with $TTr(G)=1$, and we give a necessary and sufficient condition for $TTr(G)\geq 2$.

Proposition 14.

For any connected graph $G$, $TTr(G)=1$ if and only if $G=K_{n}$.

Proof.

From the Proposition7, it is clear that $TTr(G)=1$ if $G=K_{n}$. For the converse part, assume $TTr(G)=1$ and $G$ connected graph with $n\geq 3$. If $G$ is not a complete graph, there exists $x,y\in V(G)$ such that $xy\notin E(G)$. Since $G$ is connected, there exists a path connecting $x$ and $y$ with a length of at least $3$. Taking $y\in V_{2}$, $x\in V_{1}$, and other vertices are in $V_{1}$ form a tournament transitive partition of size $2$. As a result, $G$ must be a complete graph $K_{n}$. ∎

Proposition 15.

For any graph $G$, $TTr(G)\geq 2$ if and only if $G$ contains $P_{3}$ as an induced subgraph.

Proof.

Let us assume $G$ contains $P_{3}$ as an induced subgraph. Then from Proposition8 and Proposition11, we have $TTr(G)=2$. Conversely, assume $TTr(G)\geq 2$. As $V_{1}$ dominates $V_{2}$ and $V_{2}$ does not dominate $V_{1}$, also $G$ is a connected graph, there must be a induce $P_{3}$ in $G$. ∎

Remark 1.

From the definition of a tournament transitive partition, it is clear that it is also a transitive partition of $G$. But the difference $Tr(G)-TTr(G)$ can be arbitrarily large. From the Proposition 7, we have $TTr(K_{n})=1$ and from [9], we know that $Tr(K_{n})=n$, hence the difference $Tr(K_{n})-TTr(K_{n})=n-1$ is arbitrarily large for $n$. Moreover, we know that $Tr(K_{m,n})=n+1$, when $n<m$. But from Proposition 10, we have $TTr(K_{m,n})=2$, when either $m\neq 1$ or $n\neq 1$. So, the difference $Tr(K_{n})-TTr(K_{n})=n-1$ is arbitrarily large for $n$.

3 NP-Completeness of MTTDP

In this section, we present NP-completeness results for Maximum Tournament Transitivity Decision Problem. Given a vertex partition $\pi=\{V_{1},V_{2},\ldots,V_{k}\}$ of a graph, we can verify in polynomial time whether $\pi$ is a tournament transitive partition of that graph or not. Hence, the Maximum Tournament Transitivity Decision Problem (MTTDP) is in NP. To prove that this problem is NP-hard, we show a polynomial time reduction from the Maximum Transitivity Decision Problem in general graphs, which is known to be NP-complete [13]. The reduction is as follows: given an instance of the Maximum Transitivity Decision Problem, that is, a graph $G=(V,E)$ and an integer $k$, let us consider $3(\Delta(G)+1)$ copies of $G$. Furthermore, we take another three vertices, say $x,x^{\prime}$ and $x^{\prime\prime}$. Now we connect every vertices of first $\Delta(G)+1$ copies of $G$ to $x$, second $\Delta(G)+1$ copies of $G$ to $x^{\prime}$ and final $\Delta(G)+1$ copies of $G$ to $x^{\prime\prime}$. Finally, add two edges $xx^{\prime}$ and $x^{\prime}x^{\prime\prime}$. Clearly, the resultant graph $G^{\prime}=(V^{\prime},E^{\prime})$ is a chordal graph if $G$ is a chordal graph and having $(3n(\Delta(G)+1)+3)$ vertices and $(3m(\Delta(G)+1)+2)$ edges. The construction of $G^{\prime}$ is illustrated in Figure 3.

Figure 3: Construction of $G^{\prime}$ from $G$

Let $\mathcal{G}$ be the disjoint union of $\Delta(G)+1$ copies of $G$, and also we denote the $i$-th component of $\mathcal{G}$ as $G^{i}$. Next, we prove that $Tr(G)=k$ if and only if $TTr(G)=k$, as shown in the following lemma.

Lemma 16.

Let $\mathcal{G}$ be the disjoint union of ($\Delta(G)+1$)-copies of $G$, then $Tr(G)=k$ if and only if $TTr(\mathcal{G})=k$.

Proof.

Let $G$ be a graph with $Tr(G)=k$. Also, let $\mathcal{G}=G^{1}\cup G^{2}\cup\ldots\cup G^{t}$, where each $G^{i}$ is a graph $G$ and $t=\Delta(G)+1$. Let $\pi^{j}=\{V_{1}^{j},V_{2}^{j},\ldots,V_{k}^{j}\}$ be a transitive partition of size $k$ of the $j$-th component of $\mathcal{G}$. Let us consider $\pi=\{V_{1},V_{2},\ldots,V_{k}\}$ be a vertex partition of $\mathcal{G}$ of size $k$ as follows:

$$V_{1}=(\displaystyle\bigcup_{j=k+1}^{t}V(G^{j}))\cup(V_{1}^{1}\cup V_{1}^{2}\cup\ldots\cup V_{1}^{k-1}\cup(V_{1}^{k}\cup V_{2}^{k}\cup\ldots\cup V_{k}^{k})),$$

$$V_{j}=(V_{j}^{1}\cup V_{j}^{2}\cup\ldots\cup V_{j}^{k-j}\cup(V_{j}^{k-j+1}\cup V_{j+1}^{k-j+1}\cup\ldots\cup V_{k}^{k-j+1}))$$

for all $2\leq j\leq k$.

Now we show that $\pi$ is a tournament transitive partition of size $k$. Let us consider $V_{i}$ and $V_{j}$ for $i<j$. Clearly,  $V_{i}\supseteq(V_{i}^{1}\cup V_{i}^{2}\cup\ldots\cup V_{i}^{k-i}\cup(V_{i}^{k-i+1}\cup V_{i+1}^{k-i+1}\cup\ldots\cup V_{k}^{k-i+1}))$ and $V_{j}=(V_{j}^{1}\cup V_{j}^{2}\cup\ldots\cup V_{j}^{k-j}\cup(V_{j}^{k-j+1}\cup V_{j+1}^{k-j+1}\cup\ldots\cup V_{k}^{k-j+1}))$. Since $\pi^{p}$ is a transitive partition of $G^{p}$, $V_{i}^{p}$ dominates $V_{j}^{p}$ for all $1\leq p\leq k-j$. Again, as $i<j$, which implies $k-i\geq k-j+1$ and $V_{i}^{k-j+1}\subseteq V_{i}$. So, $V_{i}^{t-j+1}$ dominates $(V_{j}^{k-j+1}\cup V_{j+1}^{t-j+1}\cup\ldots\cup V_{t}^{t-j+1})$ as $\pi^{k-j+1}$ is a transitive partition of $G^{k-j+1}$. Therefore, $V_{i}$ dominates $V_{j}$ for all $1\leq i<j\leq t$. However, $V_{j}$ does not dominate $V_{i}$ because $V_{i}$ contains vertices from $G^{k-i}$, but $V_{j}$ does not contain any vertex from $G^{k-i}$. So, $\pi$ is a tournament transitive partition of $\mathcal{G}$, and hence $TTr(\mathcal{G})\geq k$. To prove $TTr(\mathcal{G})=k$, we show that $TTr(\mathcal{G})$ cannot be more than $k$. By the Proposition 1, we know that $TTr(\mathcal{G})\leq Tr(\mathcal{G})$, and also from [9], we have $Tr(\mathcal{G})=\max\{Tr(G^{i}),1\leq i\leq t\}=k$. Therefore, $TTr(\mathcal{G})=k$.

Conversely, assume $TTr(\mathcal{G})=k$. Then from [9] and by the Proposition 1, we have $k=TTr(\mathcal{G})\leq Tr(\mathcal{G})=\max\{Tr(G^{i}),1\leq i\leq t\}=Tr(G)$. Clearly, $Tr(G)$ cannot be more than $k$; otherwise, $TTr(\mathcal{G})$ is more than $k$. Therefore, $Tr(G)=k$ if and only if $TTr(\mathcal{G})=k$. ∎

In the next claim, we show that $G$ has a transitive partition of size $k$ if and only if $G^{\prime}$ has a tournament transitive partition of size $k+2$.

Claim 17.

The graph $G$ has a transitive partition of size $k$ if and only if $G^{\prime}$ has a tournament transitive partition of size $k+2$.

Proof.

Let $G$ be a graph with a transitive partition of size $k$. Also, let $\mathcal{G}_{i}$ be the disjoint union of $i$-th set of $(\Delta(G)+1)$ copies of $G$, for all $1\leq i\leq 3$. By the Lemma 16, we know that $TTr(\mathcal{G}_{i})\geq k$, for all $i$. Now, if $\pi^{i}=\{V_{1}^{i},V_{2}^{i},\ldots,V_{k}^{i}\}$ is a transitive partition of $\mathcal{G}_{i}$ of size $k$, we can construct a vertex partition, say $\pi^{\prime}=\{V^{\prime}_{1},V^{\prime}_{2},\ldots,V^{\prime}_{k+2}\}$ of $G^{\prime}$, as follows: $V^{\prime}_{i}=V_{i}^{1}\cup V_{i}^{2}\cup V_{i}^{3}$ for all $1\leq i\leq k$ and $V^{\prime}_{k+1}=\{x^{\prime},x^{\prime\prime}\},V^{\prime}_{k+2}=\{x\}$. We show that $\pi^{\prime}$ is a tournament transitive partition of $G^{\prime}$. For any pair of sets $V^{\prime}_{i}$ and $V^{\prime}_{j}$ with $1\leq i<j\leq k$, $V^{\prime}_{i}$ dominates $V^{\prime}_{j}$ in $G^{\prime}$ as $V_{i}^{p}$ dominates $V_{j}^{p}$ in $\mathcal{G}_{p}$ and $V^{\prime}_{j}$ does not dominate $V^{\prime}_{i}$ in $G^{\prime}$ as $V_{j}^{p}$ does not dominate $V_{i}^{p}$ in $\mathcal{G}_{p}$. Also, from the construction of $G^{\prime}$, it is clear that $V_{i}$ dominates $V_{j}$ but $V_{j}$ does not dominate $V_{i}$, for all $1\leq i<j\leq k+2$. Therefore, $\pi^{\prime}$ is a tournament transitive partition of $G^{\prime}$ of size $k+2$.

Conversely, let $\pi=\{V_{1},V_{2},\ldots,V_{k+2}\}$ be a tournament transitive partition of $G^{\prime}$ of size $k+2$. So, $Tr(G^{\prime})\geq k+2$. Now consider the graph $\mathcal{G}_{1}^{\prime}$, where $V(\mathcal{G}_{1}^{\prime})=V(\mathcal{G}_{1})\cup\{x\}$ and $E(\mathcal{G}_{1}^{\prime})=E(\mathcal{G}_{1})\cup\{xy|y\in V(\mathcal{G}_{1})\}$. From [11], we know that for a graph $H$ and a vertex $v$, $Tr(H)-1\leq Tr(H-v)\leq Tr(H)$, and if $H$ is a disjoint union of $C_{1},C_{2},\ldots,C_{r}$, then $Tr(H)=\max\{Tr(C_{i}),1\leq i\leq r\}$. So, $Tr(G^{\prime})-1\leq Tr(G^{\prime}-x^{\prime})=Tr(\mathcal{G}_{1}^{\prime})$, which implies $k+1\leq Tr(G^{\prime})-1\leq Tr(\mathcal{G}_{1}^{\prime})$. Furthermore, $Tr(\mathcal{G}_{1}^{\prime})-1\leq Tr(\mathcal{G}_{1}^{\prime}-x)=Tr(G)$. Hence, $Tr(G)\geq k$, and $G$ has a transitive partition of size $k$. ∎

Therefore, we have the following main theorem for this section.

Theorem 18.

The MTTDP is NP-complete for chordal graphs (connected).

Remark 2.

The Maximum Transitivity Decision Problem is known to be NP-complete for the perfect elimination bipartite graphs [15] and doubly chordal graphs [16]. Therefore, from Lemma16, we can say that MTTDP is NP-complete for perfect elimination bipartite graphs (disconnected) and doubly chordal graphs (disconnected).

4 Tournament transitivity in trees

In this section, we design a polynomial-time algorithm for finding the tournament transitivity of a given tree $T$. To design the algorithm, we first prove that the tournament transitivity of a tree $T$ can be either $Tr(T)-1$ or $Tr(T)$, where $Tr(T)$ is the transitivity of $T$.

Lemma 19.

For a tree $T$, $Tr(T)-1\leq TTr(T)\leq Tr(T)$.

Proof.

According to the tournament transitive partition definition, for any graph G, $TTr(G)\leq Tr(G)$. Hence, for a tree $T$, we have $TTr(T)\leq Tr(T)$. Note that if $Tr(T)=1$, by [11] $T$ is a single vertex graph, and in that case, $TTr(T)=1$. Again, when $Tr(T)=2$, $T$ is a star graph, according to [11]. Let $T$ be a star $S_{t}$. According to Proposition 10, we know that $TTr(S_{t})=1$ if $t=1$, otherwise $TTr(S_{t})=2$. Therefore, we assume that $T$ is a tree such that $Tr(T)\geq 3$.

To prove that $TTr(G)\geq Tr(T)-1$ for tree, let us consider $\pi=\{V_{1},V_{2},\ldots,V_{k}\}$ be a transitive partition of $T$ with size $Tr(T)\geq 3$. As $Tr(T)\geq 3$, from [9], we can modify the above transitive partition into another transitive partition, say, $\pi^{\prime}=\{V_{1}^{\prime},V_{2}^{\prime},\ldots,V_{k}^{\prime}\}$, such that $|V_{k}^{\prime}|=|V_{k-1}^{\prime}|=1$ and $|V_{k-i}^{\prime}|\leq 2^{i-1}$ for all $2\leq i\leq k-2$. Since $T$ is a tree, $|V_{k}^{\prime}|=|V_{k-1}^{\prime}|=1$, $|V_{k-i}^{\prime}|=2^{i-1}$ for all $2\leq i\leq k-2$ and $|V_{1}^{\prime}|\geq 2^{k-2}$. Let us consider $V_{i}^{\prime}$ and $V_{j}^{\prime}$ for $i<j\leq k-1$. Let $x\in V_{i}^{\prime}$ and $z\in V_{k}^{\prime}$ such that $xz\in E(T)$. We show that $x$ has no neighbour in $V_{j}^{\prime}$. If possible, assume $y_{j}\in V_{j}^{\prime}$ and $xy_{j}\in E(T)$. As $|V_{j}^{\prime}|=2^{k-j-1}=|(V_{j+1}^{\prime}\cup V_{j+2}^{\prime}\cup\ldots\cup V_{k}^{\prime})|$ and $T$ is a tree for $y_{j}$, there exists $y_{j+1}\in V_{j+1}^{\prime}$ such that $y_{j}y_{j+1}\in E(T)$. Similarly, for $y_{j+1}$ we have $y_{j+2}\in V_{j+2}^{\prime}$ such that $y_{j+1}y_{j+2}\in E(T)$. In this way, we end up with a cycle with the vertices from the set $\{x,y_{j},y_{j+1},\ldots,y_{k-1},z\}$, which is a contradiction as $T$ is a tree. So, $x$ has no neighbour in $V_{j}^{\prime}$. Therefore, in $\pi^{\prime}$, $V_{j}^{\prime}$ does not dominate $V_{i}^{\prime}$ for all $1\leq i<j\leq k-1$. Now consider $\pi^{\prime\prime}=\{V_{1}^{\prime}\cup V_{k}^{\prime},V_{2}^{\prime},\ldots,V_{k-1}^{\prime}\}$. Based on the above discussion, we can say that $\pi^{\prime\prime}$ is a tournament transitive partition of $T$ with size $k-1$. Therefore, $TTr(G)\geq Tr(T)-1$. Hence, for a $T$, $Tr(T)-1\leq TTr(T)\leq Tr(T)$. ∎

Next, we characterize the trees with tournament transitivity equal to $Tr(T)$. We use a function, TransitiveNumber$(x,T)$, which takes a vertex $x$ and a tree $T$ and returns the transitive number of $x$ in $T$. The transitive number of a vertex $v$ in $T$ is the maximum integer $p$ such that $v\in V_{p}$ in a transitive partition $\pi=\{V_{1},V_{2},\ldots,V_{k}\}$, where the maximum is taken over all transitive partitions of $T$. The transitive number of a vertex $v$ in $T$ is denoted by $t(v,T)$. The function TransitiveNumber$(x,T)$ is presented in [12], where the authors design an algorithm to find the Grundy number of a given tree. As we know that transitivity and Grundy number are the same for a tree [11], the algorithm in [12] also finds the transitivity of a tree. From the description in [12], it follows that TransitiveNumber$(x,T)$ correctly calculates the transitive number of $x$ in $T$.

Let us consider two vertices $u,v\in V(T)$ and the path $uPv$ as $(u,v_{a},\ldots,v_{b},v)$. Additionally, we assume that the transitive number of $u$ is $t(u,T)=t(u)$, and the transitive number of $v$ is $t(v,T)=t(v)$. Let $T^{c}$ denote the tree $T$ rooted at $c$ and $T^{[c,c^{\prime}]}$ denote the subtree rooted at $c$, which is obtained by deleting $c^{\prime}$ from $T^{c}$. Let us define the set $X$ as the set of vertices required from the path $\{u,v_{a},\ldots,v_{b},v\}$ such that $u\in V_{t(u)}$ for some transitive partition $\pi$ of $T$. Clearly, $u\in X$ and the vertices of $X$ are consecutive vertices from the path $(u,v_{a},\ldots,v_{b},v)$. We calculate $X$ iteratively by checking whether the transitive number of $u$ is $t(u)-1$ in the tree $T^{[u,v_{\alpha}]}$ or not for all $v_{\alpha}\in\{v_{a},v_{a+1},\ldots,v_{b},v\}$. Also, we define $Y$ as the set of vertices required from $\{u,v_{a},\ldots,v_{b},v\}$ to achieve the transitive number of $v$ as $t(v)$ and calculate $Y$ similarly. For a vertex $w\in X\cap Y$, we say that $w$ agrees for $u$ and $v$ if there exist two transitive partitions of $T$ (not necessarily distinct), say $\pi_{p}$ and $\pi_{q}$, such that $u\in V_{t(u)}$ and $w\in V_{t}$ in $\pi_{p}$ and $v\in V_{t(v)}$ and $w\in V_{t}$ in $\pi_{q}$, respectively. In other words, the index of the sets in which $w$ belongs is the same in both $\pi_{p}$ and $\pi_{q}$. In the following lemma, we characterize trees with equal transitivity and tournament transitivity.

Lemma 20.

Let $T$ be a tree and $Tr(T)=k(\geq 3)$. Then $TTr(G)=Tr(T)$ if and only if there exist two vertices $y,z$ of $T$ such that the following conditions hold.

1. Transitive number of $z$ in $T$ is $k$, that is, $t(z,T)=k$.

2. $y\notin N_{T}(z)$.

3. Transitive number of $y$ in $T$ is at least $k-1$, that is, $t(y,T)\geq k-1$.

4. Either $X\cap Y=\phi$ or $X\cap Y\neq\phi$ and every vertex of $X\cap Y$ agrees for $y$ and $z$.

   [Where $X$ is the set of vertices required from the path $yPz$ such that $y\in V_{k-1}$ for some transitive partition $\pi$ of $T$ and $Y$ is the set of vertices required from the path $yPz$ such that $z\in V_{k}$ for some transitive partition $\pi$ of $T$. For a vertex $w\in X\cap Y$, we say that $w$ agrees for $y$ and $z$ if there exist two transitive partitions of $T$ (not necessarily distinct), say $\pi_{p}$ and $\pi_{q}$, such that $y\in V_{k-1}$ and $w\in V_{t}$ in $\pi_{p}$ and $z\in V_{k}$ and $w\in V_{t}$ in $\pi_{q}$, respectively. In other words, the index of the sets in which $w$ belongs is the same in both $\pi_{p}$ and $\pi_{q}$.]

Proof.

First, assume that $TTr(T)=Tr(T)=k(\geq 3)$. By the Proposition 6, we have a tournament transitive partition of $T$ of size $k$, say $\pi=\{V_{1},V_{2},\ldots,V_{k}\}$, such that $|V_{k}|=1$ and $|V_{k-1}|=2$. Moreover, $V_{k-1}=\{x,y\}$ and $V_{k}=\{z\}$ such that $xz\in E(T)$ and $yz\notin E(T)$. Since $\pi$ is also a transitive partition of $T$, by the definition of transitive number, we have $t(z,T)=k$ and $t(y,T)\geq k-1$. Also, by our choice, $y\notin N_{T}(z)$. Furthermore, $\pi$ is a transitive partition such that $y\in V_{k-1}$ and $z\in V_{k}$. Now, from the definition of $X$ and $Y$, we have either $X\cap Y=\phi$ or $X\cap Y\neq\phi$, and every vertex of $X\cap Y$ agrees for $y$ and $z$. Therefore, we have two vertices, namely $y$ and $z$, that satisfy the given conditions.

Conversely, assume there exist two vertices $y,z$ of $T$ such that the given conditions hold. From the conditions $(a),(b),(c)$ and $(d)$, we have a transitive partition of $T$, say, $\pi=\{V_{1},V_{2},\ldots,V_{k-1},V_{k}\}$ such that $V_{k}=\{z\}$ and $y\in V_{k-1}$. Now we show that $TTr(T)=Tr(T)=k$. Let $V_{i}$ and $V_{j}$ from $\pi$ and $i<j\leq k-1$. Since $V_{i}$ dominates $V_{k}$, there exists $x\in V_{i}$ such that $xz\in E(T)$. Now if $x$ has no neighbour in $V_{j}$, then $V_{j}$ does not dominate $V_{i}$. On the other hand, assume $x$ has a neighbour in $V_{j}$, say $y_{j}$. If $y_{j}$ is not used to dominate any sets from $\{V_{j+1},V_{j+2},\ldots,V_{k}\}$, we modify the partition $\pi$ into another partition by moving $y_{j}$ in $V_{1}$. In this partition, $V_{j}$ does not dominate $V_{i}$. Now consider the case when $y_{j}$ is used to dominate a vertex from $\{V_{j+1},V_{j+2},\ldots,V_{k}\}$, and let $y_{s}\in V_{s}$ such that $y_{j}y_{s}\in E(T)$. If $V_{s}$ is $V_{k}$, then we have a cycle $\{x,y_{j},y_{s}=z\}$, which is a contradiction as $T$ is a tree. Therefore, we can assume $V_{s}$ is not $V_{k}$.

Similarly, if $y_{s}$ is not used to dominate sets from $\{V_{s+1},V_{s+2},\ldots,V_{k}\}$, we can modify the partition $\pi$ into another partition by moving $y_{j},y_{s}$ in $V_{1}$. Note that, in the new partition, $V_{j}$ does not dominate $V_{i}$. If we continue this process, we have a path $\{x,y_{j},y_{s},\ldots,y_{p}\}$, and if $p\leq k-2$, similarly as above, we have our required modified partition. Now, consider $y_{p}\in V_{q}$, where $q\leq k-1$. If $q=k$, then this path $\{x,y_{j},y_{s},\ldots,y_{p}\}$ becomes a cycle, which is a contradiction as $T$ is a tree. Assume, $p=k-1$. The vertex $y_{p}$ is either $y$ or not. If it is not $y$, then we have a modified partition where $V_{j}$ does not dominate $V_{i}$ for all $1\leq i<j\leq k$, and hence $Tr(T)=TTr(T)=k$. Assume $y_{p}$ is $y$, as $V_{k-1}$ dominates $V_{k}$, for $z\in V_{k}$, we have $x^{\prime}\in V_{k-1}$ such that $x^{\prime}z\in E(T)$. Similarly for this $x^{\prime}$ we have $x^{\prime\prime}\in V_{i}$ and a path $\{x^{\prime\prime},y_{j}^{\prime},y_{s}^{\prime},\ldots,y_{p}^{\prime}\}$ such that $y_{j}^{\prime}\in V_{j}$. Now, either we have a desired partition or $y_{p}^{\prime}$ is any vertex from $V_{q}$, $q\geq k-1$. If $q=k$ we have a cycle $\{x^{\prime},x^{\prime\prime},y_{j}^{\prime},\ldots,z\}$. Assume $q=k-1$. If $y_{p}^{\prime}$ is $x^{\prime}$, again we have a cycle; if $y_{p}^{\prime}$ is $y$, we have a cycle $\{z,x,y_{j},\ldots,y,\ldots,y_{j}^{\prime},x^{\prime\prime},x^{\prime}\}$. In other cases, we have a desired partition such that $V_{j}$ does not dominate $V_{i}$ for all $1\leq i<j\leq k$, and hence $TTr(T)=Tr(T)=k$. ∎

4.1 Finding the set $X$ and $Y$

In this subsection, we calculate the sets $X$ and $Y$, which are defined as above. Now, we investigate those transitive partitions of $T$, where $u\in V_{t(u)}$, and show that for those transitive partitions, the vertices of $X$ can be in some specific sets. To show this result, we need to understand how to find the transitive number of $x$ in a rooted tree $T^{x}$, and when $x$ achieves that transitive number, then to which sets the children of $x$ belong. The following lemma shows that.

Lemma 21.

Let $v_{1},v_{2},\ldots,v_{k}$ be the children of $x$ in a rooted tree $T^{x}$, and for each $1\leq i\leq k$, $l_{i}$ denotes the rooted transitive number of $v_{i}$ in $T^{x}$ with $l_{1}\leq l_{2}\leq\ldots\leq l_{k}$. Let $z$ be the largest integer such that there exists a subsequence of $\{l_{i}:1\leq i\leq k\}$, say $(l_{i_{1}}\leq l_{i_{2}}\leq\ldots\leq l_{i_{z}})$ such that $l_{i_{j}}\geq j$, for all $1\leq j\leq z$.

1. In this case, the transitive number of $x$ in $T^{x}$ is $1+z$.

2. (i) For all $1\leq j\leq z$, there exists a transitive partition $\pi=\{V_{1},V_{2},\ldots,V_{1+z}\}$ such that $x\in V_{1+z}$ and $v_{i_{j}}$ belongs to any of the sets from $V_{j},V_{j+1},\ldots,V_{\min\{l_{i_{j}},z\}}$.

   (ii) If there exists a neighbour of $x$, say $v$, other than $\{v_{i_{1}},v_{i_{2}},\ldots,v_{i_{z}}\}$, such that the rooted transitive number of $v$ in $T^{x}$ is at least $j$, then there exists a transitive partition $\pi=\{V_{1},V_{2},\ldots,V_{1+z}\}$ such that $x\in V_{1+z}$ and $v_{i_{j}}$ belongs to any of the sets from $V_{1},V_{2},\ldots,V_{j-1}$. Otherwise, let $r$ be the minimum index such that $l_{i_{t}}\geq t+1$ for all $r\leq t\leq j-1$. Then there exists a transitive partition $\pi=\{V_{1},V_{2},\ldots,V_{1+z}\}$ such that $x\in V_{1+z}$ and $v_{i_{j}}$ belongs to any of the sets from $V_{r},V_{r+1},\ldots,V_{j-1}$ if and only if such $r$ exists.

Proof.

$(a)$ The proof is similar to the result used in [12], where authors designed an algorithm to calculate the Grundy number of a tree. We rephrase the statement in terms of transitivity.

$(b)(i)$ Let us denote the subtree of $T^{x}$, rooted at $y$, as $T^{x}_{y}$. Since $l_{i_{j}}\geq j$, for all $1\leq j\leq z$, there exists a transitive partition of $T^{x}_{v_{i_{j}}}$, say $\{V_{1}^{j},V_{2}^{j},\ldots,V_{j}^{j}\}$ such that $v_{i_{j}}\in V_{j}^{j}$. Now let us consider a vertex partition of $T^{x}$, say $\{V_{1},V_{2},\ldots,V_{z},V_{1+z}\}$ as follows: $V_{r}=\bigcup_{s=r}^{z}V^{s}_{r}$ for all $1\leq r\leq z$ and $V_{1+z}=\{x\}$. Clearly, the above partition is a transitive partition of $T^{x}$, such that $x\in V_{1+z}$ and $v_{i_{j}}$ belongs to the set $V_{j}$. Now consider the vertex $v_{i_{j}}$ and let $s\in\{j+1,j+2,\ldots,\min\{l_{i_{j}},z\}\}$. Since $l_{i_{j}}\geq s$, there exists a transitive partition of $T^{x}_{v_{i_{j}}}$, say $\{V_{1}^{j},V_{2}^{j},\ldots,V_{s}^{j}\}$ such that $v_{i_{j}}\in V_{s}^{j}$. Also, since $l_{i_{s}}\geq s>j$, there exists a transitive partition of $T^{x}_{v_{i_{s}}}$, say $\{V_{1}^{s},V_{2}^{s},\ldots,V_{j}^{s}\}$ such that $v_{i_{s}}\in V_{j}^{s}$. Similarly, as before, we can show that there exists a transitive partition $\pi=\{V_{1},V_{2},\ldots,V_{1+z}\}$ such that $x\in V_{1+z}$ and $v_{i_{j}}$ belongs to the set $V_{s}$. Therefore, we always have a transitive partition of $T^{x}$, say $\pi=\{V_{1},V_{2},\ldots,V_{1+z}\}$ such that $x\in V_{1+z}$ and $v_{i_{j}}$ belongs to any of the sets from $V_{j},V_{j+1},\ldots,V_{\min\{l_{i_{j}},z\}}$.

$(b)(ii)$ Let $v$ be a neighbour of $x$ other than $\{v_{i_{1}},v_{i_{2}},\ldots,v_{i_{z}}\}$, such that the rooted transitive number of $v$ in $T^{x}$ is at least $j$. In this case, let $s\in\{1,2,\ldots,j-1\}$. Since $l_{i_{j}}\geq s$, there exists a transitive partition of $T^{x}_{v_{i_{j}}}$, say $\{V_{1}^{j},V_{2}^{j},\ldots,V_{s}^{j}\}$ such that $v_{i_{j}}\in V_{s}^{j}$. Also, since the rooted transitive number of $v$ in $T^{x}$ is at least $j$, there exists a transitive partition of $T^{x}_{v}$, say $\{V_{1}^{s},V_{2}^{s},\ldots,V_{j}^{s}\}$ such that $v\in V_{j}^{s}$. By taking the unions of the sets of these partitions as in $(b)(i)$, we can construct a transitive partition of $T^{x}$, say $\pi=\{V_{1},V_{2},\ldots,V_{1+z}\}$ such that $x\in V_{1+z}$ and $v_{i_{j}}$ belongs to the set $V_{s}$.

Now assume there does not exist any neighbour of $x$ other than $\{v_{i_{1}},v_{i_{2}},\ldots,v_{i_{z}}\}$, having at least $j$ as its rooted transitive number in $T^{x}$. In this case, let $r$ be the minimum index such that $l_{i_{t}}\geq t+1$ for all $r\leq t\leq j-1$. For $1\leq s\leq r-1$ and $j+1\leq s\leq z$ as $l_{i_{t}}\geq t$, we know that there exists a transitive partition of $T^{x}_{v_{i_{s}}}$, say $\{V_{1}^{s},V_{2}^{s},\ldots,V_{s}^{s}\}$ such that $v_{i_{s}}\in V_{s}^{s}$. Again, since $l_{i_{t}}\geq t+1$ for all $r\leq t\leq j-1$, we have a transitive partition of $T^{x}_{v_{i_{t}}}$, say $\{V_{1}^{t},V_{2}^{t},\ldots,V_{t+1}^{t}\}$ such that $v_{i_{t}}\in V_{t+1}^{t}$, for all $r\leq t\leq j-1$. As $l_{i_{j}}\geq j$, we have a transitive partition of $T^{x}_{v_{i_{j}}}$, say $\{V_{1}^{j},V_{2}^{j},\ldots,V_{r}^{j}\}$ such that $v_{i_{j}}\in V_{r}^{j}$. By taking the unions of the sets of these partitions as in $(b)(i)$, we can construct a transitive partition of $T^{x}$, say $\pi=\{V_{1},V_{2},\ldots,V_{1+z}\}$ such that $x\in V_{1+z}$ and $v_{i_{j}}$ belongs to the set $V_{r}$. Further, using similar arguments as in $(b)(i)$, we can show that there exists a transitive partition $\pi=\{V_{1},V_{2},\ldots,V_{1+z}\}$ such that $x\in V_{1+z}$ and $v_{i_{j}}$ belongs to any of the sets from $V_{r+1},V_{r+2},\ldots,V_{j-1}$.

Conversely, let no such $r$ exist. In this case, we show that there does not exist any transitive partition $\pi=\{V_{1},V_{2},\ldots,V_{1+z}\}$ such that $x\in V_{1+z}$ and $v_{i_{j}}$ belongs to one of the sets from $V_{1},V_{2},\ldots,V_{j-1}$. Let $\pi=\{V_{1},V_{2},\ldots,V_{1+z}\}$ be a transitive such that $x\in V_{1+z}$ and $v_{i_{j}}$ belongs to $V_{s}$, for some $1\leq s\leq j-1$. We know that $l_{i_{1}}\leq l_{i_{2}}\leq\ldots\leq l_{i_{z}}$ and $r$ do not exist. This implies that $l_{i_{1}}\leq l_{i_{2}}\leq\ldots\leq l_{i_{j-1}}<j$. So, vertices of $\{v_{i_{1}},v_{i_{2}},\ldots v_{i_{j-1}}\}$ cannot belong to any of the sets $\{V_{j},V_{j+1},\ldots,V_{z}\}$. As we assume that $v_{i_{j}}\in V_{s}$ for some $1\leq s\leq j-1$ and all the neighbours of $x$, other than $\{v_{i_{1}},v_{i_{2}},\ldots,v_{i_{z}}\}$, have rooted transitive number less than $j$, the vertices $\{v_{i_{j+1}},v_{i_{j+2}},\ldots v_{i_{z}}\}$ are the only neighbours of $x$ that can belong to the sets $\{V_{j},V_{j+1},\ldots,V_{z}\}$. Clearly, there exists a set from $\{V_{j},V_{j+1},\ldots,V_{z}\}$, which does not contain any children of $x$, which contradicts the fact that $\pi$ is a transitive partition. Therefore, there does not exit any transitive partition $\pi=\{V_{1},V_{2},\ldots,V_{1+z}\}$ such that $x\in V_{1+z}$ and $v_{i_{j}}$ belongs to any of the sets from $V_{1},V_{2},\ldots,V_{j-1}$. Hence, we have the lemma. ∎

Based on the above lemma, we now show that in a transitive partition of $T$, if $u\in V_{t(u)}$, then the vertices of $X$ can be in some specific sets. A similar result is true for $Y$ as well.

Lemma 22.

Let $u$ and $v$ be two vertices of $T$ and they are connected by the path $(u,v_{a},\ldots,v_{b},v)$ in $T$. Let $t(u,T)=t(u)$ and $t(v,T)=t(v)$. Further, let $X=\{x_{1},x_{2},\ldots x_{\alpha}\}$ be the set of consecutive vertices, starting with $x_{1}=u$, from the path that is required to achieve the transitive number of $u$ is $t(u)$. Similarly, let $Y=\{y_{1},y_{2},\ldots y_{\beta}\}$ be the set of consecutive vertices, starting with $y_{1}=v$, from the path that is required to achieve the transitive number of $v$ as $k$. Also, assume that $\pi_{p}$ and $\pi_{q}$ are two transitive partitions of $T$ such that $u\in V_{t(u)}$ in $\pi_{p}$ and $v\in V_{t(v)}$ in $\pi_{q}$, respectively. Then for all $2\leq j\leq\alpha-1$, $x_{j}$ belongs to a unique set $V_{p_{j}}$ in $\pi_{p}$, where $p_{j}=t(x_{j},T^{[x_{j},x_{j-1}]})$ and $x_{\alpha}$ belongs to a set of sets in $\pi_{p}$ as described in Lemma 21(b). Similarly, for all $2\leq j\leq\beta-1$, $y_{j}$ belongs to a unique set $V_{q_{j}}$ in $\pi_{q}$, where $q_{j}=t(y_{j},T^{[y_{j},y_{j-1}]})$ and $y_{\beta}$ belongs to a set of sets in $\pi_{q}$ as described in Lemma 21(b).

Proof.

Let $x_{j}\in X$ for $2\leq j\leq\alpha-1$. The transitive number of $x_{j}$ in $T^{[x_{j},x_{j-1}]}$ is given by $t(x_{j},T^{[x_{j},x_{j-1}]})$. Note that, in $\pi_{p}$, to obtain the transitive number of $u$ as $t(u)$, the vertex $x_{j}$ must belong to $V_{p_{j}}$ in $\pi_{p}$, where $p_{j}=t(x_{j},T^{[x_{j},x_{j-1}]})$ (see Figure 4). Because if $x_{j}$ belongs to some $V_{r}$, where $r<{p_{j}}$, then $x_{j+1}$ does not belong to $X$ as we have transitive partition of $T^{[x_{j},x_{j-1}]}\setminus\{x_{j+1}\}$ such that $x_{j}\in V_{r}$. For the vertex, $x_{\alpha}$ can be assigned from a set of sets in $\pi_{p}$, as described in Lemma 21(b). The proof for $Y$ is similar. ∎

Figure 4: Calculation of X and Y in the tree $T$

Based on the above two lemmas, we have Algorithm1 that describes the process of calculating the sets $X$ and $Y$ for two specific vertices $y$ and $z$ with $t(y,T)\geq k-1$ and $t(z,T)=k$.

Algorithm 1 Calculation_X_and_Y

Input: The tree $T$ and two vertices $y$ and $z$ with $t(y,T)\geq k-1$ and $t(z,T)=k$

Output: Set $X$ and $Y$ corresponding $y$ and $z$, respectively .

1:Let the path $P=(v_{a_{0}}=y,v_{a},\ldots,v_{b},z=v_{b_{0}})$;

2:Initially, $X=\{y\}$ and $Y=\{z\}$;

3:if $t(y)=t(y,T)=k$ then

4:     $t(y,T^{[y,v_{a}]})\geq k-1$;

5:     $X=\{y\}$;

6:else

7:     for all $r$ in $\{v_{a},v_{a+1},\ldots,v_{b},z\}$ do

8:         if $t(y,T^{[y,r]})=k-2$ then

9:              $X=X\cup\{r\}$;

10:         else

11:              Stop;

12:         end if

13:     end for

14:end if

15:for all $s$ in $\{v_{b},v_{b-1},\ldots,v_{a},y\}$ do

16:     if $t(z,T^{[z,s]})=k-1$ then

17:         $Y=Y\cup\{s\}$;

18:     else

19:         Stop;

20:     end if

21:end for

22:Calculate the set(s) in $\pi_{p}$ to which the vertices of $X$ belong using to the Lemma 22;

23:Calculate the set(s) in $\pi_{q}$ to which the vertices of $Y$ belong using to the Lemma 22;

Based on Lemmas 19 and 20, we can design an algorithm for finding tournament transitivity of a given tree.

Algorithm 2 Tournament_Transitivity(T)

Input: A tree $T$

Output: Tournament transitivity of $T$, that is, $TTr(T)$

1:Let $V=\{u_{1},u_{2},\ldots,u_{n}\}$ be the set of all vertices of $T$;

2:for all $u\in V$ do

3:     TransitiveNumber$(u,T)$ ;

4:end for

5:$S=\{(y,z)|y,z\in V\text{and }y\notin N_{T}(z)\text{ and }t(y,T)\geq k-1,t(z,T)=k\}$

6:if $S\neq\phi$ then

7:     for all $(y,z)\in S$ do

8:         Calculate the sets $X$ and $Y$ according to the Algorithm 1;

9:         if $X\cap Y=\phi$ then

10:              $TTr(T)=Tr(T)$;

11:         else if $X\cap Y\neq\phi$ and every vertex of $X\cap Y$ agrees for $y$ and $z$ then

12:              $TTr(T)=Tr(T)$;

13:         else

14:              $TTr(T)=Tr(T)-1$;

15:         end if

16:     end for

17:else

18:     $TTr(T)=Tr(T)-1$;

19:end if

20:return ($TTr(G)$);

4.2 Running time of the algorithm

In this subsection, we analyse the running time of Algorithm 2 (Tournament_Transitivity(T)). In lines $2-4$ of the algorithm, we have computed the transitive number of every vertex. According to the running time analysis of [12], it will take $O(n)$ to calculate the transitive number of every vertex. Moreover, for a fixed pair of vertices $\{y,z\}$, we can check whether it is in $S$ or not in a constant time.

Now we analyse the running time of Algorithm 1 (Calculation_X_and_Y) for a pair of vertices $\{y,z\}$ such that $y\notin N_{T}(z)$ and $t(y,T)\geq k-1,t(z,T)=k$. While computing the set $X$, the checking in line $8$ takes $\displaystyle\sum_{v\in T}(O(deg(v))$ time. This implies that we can compute the set $X$ in $k(\displaystyle\sum_{v\in T}O(deg(v)))$ time, where $k$ is the length of the path $yPz$. In the worst case, $X$ in lines $3-14$ can be computed in $O(n^{2})$ time. Similarly, $Y$ can be computed in $O(n^{2})$ time. The process of computing $X$ and $Y$ also indicates the set(s) to which the vertices of $X$ and $Y$ belong. Therefore, lines $22$ and $23$ take constant time. So for all vertices from $S$, to calculate the sets $X$ and $Y$, we need at most $n^{2}*O(n^{2})=O(n^{4})$ time. As a result, we get that the running time of Algorithm 2 is $O(n)+O(n^{4})=O(n^{4})$. Hence, we have the following main theorem.

Theorem 23.

The Maximum Tournament Transitivity Problem can be solved in polynomial time for trees.

5 Tournament transitivity in bipartite chain graphs

In this section, we characterize some bipartite chain graphs with equal tournament transitivity and transitivity. Let $G=(X\cup Y,E)$ be a bipartite chain graph also let $\sigma_{X}=(x_{1},x_{2},\ldots,x_{n_{1}})$ and $\sigma_{Y}=(y_{1},y_{2},\ldots,y_{n_{2}})$ be the chain ordering. We know from the Proposition 10, for complete bipartite graph $TTr(K_{m,n})=2$, if and only if either $m\neq 1$ or $n\neq 1$. Let us assume only bipartite chain graphs, which are not complete bipartite graphs. Now we find some bipartite chain graphs which are not a complete bipartite graph with equal tournament transitivity and transitivity. Assume $t$ be the maximum integer such that $G$ contains $K_{t,t}$ as an induced subgraph. It is known form [15], that $K_{t,t}=G[X_{t}\cup Y_{t}]$, where $X_{t}=\{x_{1},x_{2},\ldots,x_{t}\}$ and $Y_{t}=\{y_{1},y_{2},\ldots,y_{t}\}$. Now in $G$, the edges $x_{t+1}y_{t},x_{t}y_{t+1}$ may or may not present. Based on this, a bipartite chain graph, which is not a complete bipartite graph, can be partitioned into three subclasses.

Definition 24.

Let $G=(X\cup Y,E)$ be a bipartite chain graph which is not a complete bipartite graph. Also, let $\sigma_{X}=(x_{1},x_{2},\ldots,x_{n_{1}})$ and $\sigma_{Y}=(y_{1},y_{2},\ldots,y_{n_{2}})$ be the chain ordering of $G$ and $t$ be the maximum integer such that $G$ contains a $K_{t,t}$ as an induced subgraph. A bipartite chain graph $G$ is called $(i)$ Type-I BCG if $x_{t+1}y_{t}\notin E(G)$ and $x_{t}y_{t+1}\notin E(G)$, $(ii)$ Type-II BCG if either $x_{t+1}y_{t}\in E(G)$ or $x_{t}y_{t+1}\in E(G)$ not both, $(iii)$ Type-III BCG if $x_{t+1}y_{t}\in E(G)$ and $x_{t}y_{t+1}\in E(G)$.

It is known form [15], that for a bipartite chain graph $G$, $Tr(G)=t+1$, where $t$ is the maximum integer such that $G$ contains either $K_{t,t}$ or $K_{t,t}-\{e\}$ as an induced subgraph. Also, from the definition of tournament transitivity, we know that $TTr(G)\leq Tr(G)$ for a graph $G$. Next, we find the conditions for which the transitivity and tournament transitivity are the same for Type-I BCG and for some Type-II BCG graphs. Also, we show that for Type-III BCG, $TTr(G)<Tr(G)$ always.

5.1 Tournament transitivity of Type-I BCG

In this subsection, we find the condition under which $TTr(G)=Tr(G)$ for a Type-I BCG. For that, we have the following theorem.

Theorem 25.

Let $G$ be a Type-I BCG with $\sigma_{X}=(x_{1},x_{2},\ldots,x_{n_{1}})$ and $\sigma_{Y}=(y_{1},y_{2},\ldots,y_{n_{2}})$ be the chain ordering of $G$. Then $TTr(G)=Tr(G)=t+1$ if and only if there exist vertices $\{z_{1},z_{2},\ldots,z_{t}\}$ from $\{x_{t+1},x_{t+2},\ldots,x_{n_{1}},y_{t+1},y_{t+2},\ldots,y_{n_{2}}\}$ such that $|N(z_{1})|=t-1$ and $t-j+1\geq|N(z_{j})|\geq t-j$ for all $2\leq j\leq t$.

Proof.

First, assume that there exist vertices $\{z_{1},z_{2},\ldots,z_{t}\}$ from $\{x_{t+1},x_{t+2},\ldots x_{n_{1}},y_{t+1},y_{t+2},\ldots,y_{n_{2}}\}$ such that $|N(z_{1})|=t-1$ and $t-j+1\geq|N(z_{j})|\geq t-j$ for all $2\leq j\leq t$. Consider a vertex partition $\pi=\{V_{1},V_{2},\ldots,V_{t},V_{t+1}\}$ of $T$ as follows: $\{x_{1},y_{1},z_{t}\}\subseteq V_{1}$; for all $2\leq i\leq t-1$, we set $V_{i}=\{x_{i},y_{i},z_{t-i+1}\}$, $V_{t}=\{x_{t},z_{1}\}$, and $V_{t}=\{y_{t}\}$. We put other vertices of $G$ in $V_{1}$. This partition $\pi$ is illustrated in Figure 5.

Figure 5: The partition $\pi$ of Type-I BCG $G$, where dotted edges are not present in $G$

We show that $\pi$ is a tournament transitive partition of $G$ with size $t+1$. Since $G$ is a Type-I BCG, the set $\{x_{1},x_{2},\ldots x_{t},y_{1},\ldots,y_{t}\}$ induces a complete bipartite graph $K_{t,t}$ in $G$. Let us consider $V_{i}$ and $V_{j}$ for $2\leq i<j\leq t-1$. From $\pi$, we have $V_{i}=\{x_{i},y_{i},z_{t-i+1}\}$ and $V_{j}=\{x_{j},y_{j},z_{t-j+1}\}$. As $j\geq|N(z_{t-j+1})|\geq j-1$, which implies either $\{x_{1},x_{2},\ldots,x_{j-1}\}\subseteq N(z_{t-j+1})$ or $\{y_{1},y_{2},\ldots,y_{j-1}\}\subseteq N(z_{t-j+1})$. So, either $x_{i}$ is a neighbour of $z_{t-j+1}$ or $y_{i}$ is a neighbour of $z_{t-j+1}$. Furthermore, because $G$ is a Type-I BCG, $x_{j}$ is adjacent with $y_{i}$, and $y_{j}$ is adjacent with $x_{i}$. Therefore, $V_{i}$ dominates $V_{j}$. Moreover, as $i\geq|N(z_{t-i+1})|\geq i-1$, which implies either $N(z_{t-i+1})\subseteq\{x_{1},x_{2},\ldots,x_{i}\}$ or $N(z_{t-i+1})\subseteq\{y_{1},y_{2},\ldots,y_{i}\}$. Since Type-I BCG and $i<j$, $V_{j}$ does not dominate $V_{i}$.

Now for $V_{t-1}$ and $V_{t}$, clearly $V_{t-1}$ dominates $V_{t}$ but $V_{t}$ does not dominate $V_{t-1}$ as $t\geq|N(z_{t-j+1})|\geq t-1$ and $G$ is a Type-I BCG. Similarly, we can show that $V_{i}$ dominates $V_{j}$ but $V_{j}$ does not dominate $V_{j}$ for others $1\leq i<j\leq k$. Hence, $\pi$ is a tournament transitive partition of $G$ of size $t+1$. Therefore, $TTr(G)\geq t+1$. As we already know from [15], for a bipartite chain graph $G$, $Tr(G)=t+1$, where $t$ is the maximum integer such that $G$ contains either $K_{t,t}$ or $K_{t,t}-\{e\}$ as an induced subgraph. Since $G$ is a Type-I BCG, $G$ contains $K_{t,t}$ as an induced subgraph for maximum $t$. Hence, $Tr(G)=t+1$. According to the definition of tournament transitivity, we know that $TTr(G)\leq Tr(G)$. Therefore, $TTr(G)=Tr(G)=t+1$.

Conversely, assume $TTr(G)=Tr(G)=t+1$. To show the existence of $\{z_{1},z_{2},\ldots,z_{t}\}$ from $\{x_{t+1},x_{t+2},x_{n_{1}},y_{t+1},y_{t+2},\ldots,y_{n_{2}}\}$ such that $|N(z_{1})|=t-1$ and $t-j+1\geq|N(z_{j})|\geq t-j$ for all $2\leq j\leq t$, first we prove the following claim.

Claim 26.

Let $\pi=\{V_{1},V_{2},\ldots,V_{t},V_{t+1}\}$ be a tournament transitive partition of $G$ of size $t+1$ such that $|V_{t}|=2$ and $|V_{t+1}|=1$. Then one of the following conditions holds.

1. Each $V_{1},V_{2},\ldots,V_{t}$ contains exactly one vertex from $\{x_{1},x_{2},\ldots,x_{t}\}$ and each $V_{1},V_{2},\ldots,V_{t-1},V_{t+1}$ contains exactly one vertex from $\{y_{1},y_{2},\ldots,y_{t}\}$.

2. Each $V_{1},V_{2},\ldots,V_{t}$ contains exactly one vertex from $\{y_{1},y_{2},\ldots,y_{t}\}$, and each $V_{1},V_{2},\ldots,V_{t-1},V_{t+1}$ contains exactly one vertex from $\{x_{1},x_{2},\ldots,x_{t}\}$.

Proof.

According to Proposition 6, we always have a $TTr(G)$-partition $\pi=\{V_{1},V_{2},\ldots,V_{t},V_{t+1}\}$ of $G$ such that $|V_{t}|=2$ and $|V_{t+1}|=1$. Since $G$ is a Type-I BCG, the degree of each vertices from $(X\setminus X_{t})\cup(Y\setminus Y_{t})$ is at most $t-1$. Now, for any vertex $x\in V_{t+1}$, the degree of $x$ must be at least $t$. Therefore, only vertices from $X_{t}\cup Y_{t}$ can be in $V_{t+1}$. Now we prove that if $V_{t+1}$ contains a vertex from $Y_{t}$, $\pi$ satisfy the condition $(a)$, otherwise $\pi$ satisfy the condition $(b)$.

Let us assume $V_{t+1}=\{y_{j}\}$ for some $1\leq j\leq t$. Since $\pi$ is a tournament transitive partition of $G$, $V_{t}$ must contain a vertex adjacent to $y_{j}$, and the degree of that vertex is at least $t$. So, for some $1\leq i\leq t$, assume $x_{i}\in V_{t}$. Moreover, $V_{t+1}$ does not dominate $V_{t}$, which implies that there exists a vertex in $V_{t}$ that is not adjacent to $y_{j}$, and the degree of that vertex is at least $t-1$. Let $v\in V_{t}$ other than $x_{i}$, then $v\in(X\setminus X_{t})\cup Y$. If $v\in(X\setminus X_{t})$, to dominate $v$, each $V_{1},V_{2},\ldots,V_{t-1}$ contains exactly one vertex from $\{y_{1},y_{2},\ldots,y_{t-1}\}$, and the vertex $y_{j}$ must be the vertex $y_{t}$. As $y_{t}\in V_{t+1}$ and $N(y_{t})=X_{t}$, each $V_{1},V_{2},\ldots,V_{t}$ contains exactly one vertex from $\{x_{1},x_{2},\ldots,x_{t}\}$. Therefore, in this case, the condition $(a)$ holds. On the other hand, if $v\in(Y\setminus Y_{t})$, we can similarly show the condition $(a)$.

Finally, assume $v\in Y_{t}$. As $V_{t-1}$ dominates $V_{t}$, for $v\in V_{t}$ there must exist a vertex from $V_{t-1}$, say $u$, such that $v$ is adjacent to $u$ and $|N(u)\setminus\{v,y_{j}\}|\geq t-2$. If $u\in(X\setminus X_{t})$, then $|N(u)\setminus\{v,y_{j}\}|=|N(u)|-2\leq t-1-2=t-3$, which implies $u$ must be a vertex from $X_{t}$. Let us assume $u=x_{r}$ for some $1\leq r\leq t$ and $r\neq i$. As we know, $V_{t}$ does not dominate $V_{t-1}$, so there exists a vertex from $V_{t-1}$, say $v^{\prime}$, such that $v^{\prime}$ is not adjacent to the vertices from $\{x_{i},v\}$. Therefore, $v^{\prime}$ must be a vertex from $(X\setminus X_{t})\cup(Y\setminus Y_{t})$. Consider the case when $v^{\prime}\in(X\setminus X_{t})$. Since $v^{\prime}\in V_{t-1}$, $|N(v^{\prime})\setminus\{v,y_{j}\}|\geq t-2$. But $|N(v^{\prime})\setminus\{v,y_{j}\}|=|N(v^{\prime})|-2\leq t-1-2=t-3$. So, $v^{\prime}$ cannot be a vertex from $(X\setminus X_{t})$. Furthermore, for $v^{\prime}\in(Y\setminus Y_{t})$, $|N(v^{\prime})\setminus\{x_{r},x_{i}\}|\geq t-2$. But $|N(v^{\prime})\setminus\{x_{r},x_{i}\}|=|N(v^{\prime})|-2\leq t-1-2=t-3$. So, $v^{\prime}$ cannot be a vertex from $(X\setminus X_{t})$. We have a contradiction, which implies that $v$ cannot be in $V_{t}$. Hence, we have the claim. ∎

From the Claim 26, we have either each $V_{1},V_{2},\ldots,V_{t}$ contains exactly one vertex from $\{x_{1},x_{2},\ldots,x_{t}\}$ and each $V_{1},V_{2},\ldots,V_{t-1},V_{t+1}$ contains exactly one vertex from $\{y_{1},y_{2},\ldots,y_{t}\}$ or each $V_{1},V_{2},\ldots,V_{t}$ contains exactly one vertex from $\{y_{1},y_{2},\ldots,y_{t}\}$ and each $V_{1},V_{2},\ldots,V_{t-1},V_{t+1}$ contains exactly one vertex from $\{x_{1},x_{2},\ldots,x_{t}\}$. Since $G$ is a bipartite chain graph, without loss of generality assume $x_{r},y_{r}\in V_{r}$ for all $1\leq r\leq t-1$ and either $x_{t}\in V_{t}$ and $y_{t}\in V_{t+1}$ or $y_{t}\in V_{t}$ and $x_{t}\in V_{t+1}$.

Now we are ready to show the existence of $\{z_{1},z_{2},\ldots,z_{t}\}$ vertices from $(X\setminus X_{t})\cup(Y\setminus Y_{t})$ such that $|N(z_{1})|=t-1$ and $t-j+1\geq|N(z_{j})|\geq t-j$ for all $2\leq j\leq t$. Since $\pi$ is a tournament transitive partition, $V_{t+1}$ does not dominate $V_{t}$. So, there must exist a vertex from $\{x_{t+1},x_{t+2},x_{n_{1}},y_{t+1},y_{t+2},\ldots,y_{n_{2}}\}$ that belongs to $V_{t}$ and is not adjacent to either $y_{t}$ or $x_{t}$, depending on whether $V_{t+1}=\{y_{t}\}$ or $V_{t+1}=\{x_{t}\}$. Let $z_{1}$ be such a vertex from $(X\setminus X_{t})\cup(Y\setminus Y_{t})$. As $z_{1}\in V_{t}$, $|N(z_{1})|\geq t-1$ and we know that $|N(z_{1})|\leq t-1$, which implies $|N(z_{1})|=t-1$. Let us consider $V_{j}$ and $V_{j+1}$ for some $1\leq j\leq t-1$. Since $V_{j+1}$ does not dominate $V_{j}$, there exists a vertex form $(X\setminus X_{t})\cup(Y\setminus Y_{t})$, say $z_{t-j+1}\in V_{j}$ such that $z_{t-j+1}\notin N(V_{j+1})$. As $z_{t-j+1}\in V_{j}$ and $\pi$ is a tournament transitive partition of $G$, $|N(z_{t-j+1})|\geq j-1$. Moreover, $z_{t-j+1}\notin N(V_{j+1})$ and $G$ is a bipartite chain graph, implies that $z_{t-j+1}\notin N(x_{p})$ and $z_{t-j+1}\notin N(y_{p})$, for all $j+1\leq p\leq t$. So, $|N(z_{t-j+1})|\leq j$. Therefore, $j\geq|N(z_{t-j+1})|\geq j-1=t-s+1\geq|N(z_{s})|\geq t-s$. Hence, we have $\{z_{1},z_{2},\ldots,z_{t}\}$ from $\{x_{t+1},x_{t+2},x_{n_{1}},y_{t+1},y_{t+2},\ldots,y_{n_{2}}\}$ such that $|N(z_{1})|=t-1$ and $t-s+1\geq|N(z_{s})|\geq t-s$ for all $2\leq s\leq t$. ∎

5.2 Tournament transitivity of Type-II BCG

In this subsection, we find a sufficient condition under which $TTr(G)=Tr(G)$ for Type-II BCG. Let $G$ be a Type-II BCG with $\sigma_{X}=(x_{1},x_{2},\ldots,x_{n_{1}})$ and $\sigma_{Y}=(y_{1},y_{2},\ldots,y_{n_{2}})$ be the chain ordering of $G$ and $x_{t}y_{t+1}\in E(G)$. Further, we divide a Type-II BCG into two subclasses, based on whether $x_{t+1}y_{t-1}\notin E(G)$ or $x_{t+1}y_{t-1}\in E(G)$. A Type-II BCG $G$ is called $(i)$ Type-II(a) BCG if $x_{t+1}y_{t-1}\notin E(G)$, $(ii)$ Type-II(b) BCG if $x_{t+1}y_{t-1}\in E(G)$. The following theorems find a sufficient condition under which $TTr(G)=Tr(G)$ for a Type-IIBCG.

Theorem 27.

Let $G$ be a Type-II(a) BCG with $\sigma_{X}=(x_{1},x_{2},\ldots,x_{n_{1}})$ and $\sigma_{Y}=(y_{1},y_{2},\ldots,y_{n_{2}})$ be the chain ordering of $G$. Then $TTr(G)=Tr(G)=t+1$ if there exist vertices $\{z_{1},z_{2},\ldots,z_{t-1}\}$ from $\{x_{t+1},x_{t+2},\ldots x_{n_{1}},y_{t+2},y_{t+3},\ldots,y_{n_{2}}\}$ such that $t-j\geq|N(z_{j})|\geq t-j-1$ for all $1\leq j\leq t-1$.

Proof.

Let us assume there exist vertices $\{z_{1},z_{2},\ldots,z_{t-1}\}$ from $\{x_{t+1},x_{t+2},\ldots x_{n_{1}},y_{t+2},y_{t+3},\ldots,y_{n_{2}}\}$ such that $t-j\geq|N(z_{j})|\geq t-j-1$ for all $1\leq j\leq t-1$. As $G$ is a Type-II(a) BCG, $x_{t}y_{t+1}\in E(G)$ and $x_{t+1}y_{t-1}\notin E(G)$. Consider a vertex partition $\pi=\{V_{1},V_{2},\ldots,V_{t},V_{t+1}\}$ of $G$ as follows: $\{x_{1},y_{1},z_{t-1}\}\subseteq V_{1}$; for all $2\leq i\leq t-1$, we set $V_{i}=\{x_{i},y_{i},z_{t-i}\}$, $V_{t}=\{x_{t},y_{t}\}$, and $V_{t+1}=\{y_{t+1}\}$. We put the other vertices of $G$ in $V_{1}$. This partition $\pi$ is illustrated in Figure 6.

Figure 6: The partition $\pi$ of Type-II(a) BCG $G$, where dotted edges are not present in $G$

We show that $\pi$ is a tournament transitive partition of $G$ with size $t+1$. Since $G$ is a Type-II BCG, the set $\{x_{1},x_{2},\ldots x_{t},y_{1},\ldots,y_{t}\}$ induces a complete bipartite graph $K_{t,t}$ in $G$. Let us consider $V_{i}$ and $V_{j}$ for $2\leq i<j\leq t-1$. From $\pi$, we have $V_{i}=\{x_{i},y_{i},z_{t-i}\}$ and $V_{j}=\{x_{j},y_{j},z_{t-j}\}$. As $j\geq|N(z_{t-j})|\geq j-1$, which implies either $\{x_{1},x_{2},\ldots,x_{j-1}\}\subseteq N(z_{t-j})$ or $\{y_{1},y_{2},\ldots,y_{j-1}\}\subseteq N(z_{t-j})$. Since $i<j$, either $x_{i}$ is a neighbour of $z_{t-j}$ or $y_{i}$ is a neighbour of $z_{t-j}$. Furthermore, because $G$ is a Type-II BCG, $x_{j}$ is adjacent with $y_{i}$, and $y_{j}$ is adjacent with $x_{i}$. Therefore, $V_{i}$ dominates $V_{j}$ in $\pi$. Moreover, as $i\geq|N(z_{t-i})|\geq i-1$, which implies either $N(z_{t-i})\subseteq\{x_{1},x_{2},\ldots,x_{i}\}$ or $N(z_{t-i})\subseteq\{y_{1},y_{2},\ldots,y_{i}\}$. Since Type-II BCG and $i<j$, $V_{j}$ does not dominate $V_{i}$. Now for $V_{t}$ and $V_{t+1}$, clearly $V_{t}$ dominates $V_{t+1}$ but $V_{t+1}$ does not dominate $V_{t}$. Similarly, we can show that $V_{i}$ dominates $V_{j}$ but $V_{j}$ does not dominate $V_{j}$ for others $1\leq i<j\leq k$. Hence, $\pi$ is a tournament transitive partition of $G$ of size $t+1$. Therefore, $TTr(G)\geq t+1$. As we already know from [15], for a bipartite chain graph $G$, $Tr(G)=t+1$, where $t$ is the maximum integer such that $G$ contains either $K_{t,t}$ or $K_{t,t}-\{e\}$ as an induced subgraph. Since $G$ is a Type-II BCG, $G$ contains $K_{t,t}$ as an induced subgraph for maximum $t$. Hence, $Tr(G)=t+1$. According to the definition of tournament transitivity, we know that $TTr(G)\leq Tr(G)$. Therefore, $TTr(G)=Tr(G)=t+1$. ∎

For Type-II(b) BCG, we have the following theorem.

Theorem 28.

Let $G$ be a Type-II(b) BCG with $\sigma_{X}=(x_{1},x_{2},\ldots,x_{n_{1}})$ and $\sigma_{Y}=(y_{1},y_{2},\ldots,y_{n_{2}})$ be the chain ordering of $G$. Then $TTr(G)=Tr(G)=t+1$ if there exist vertices $\{z_{1},z_{2},\ldots,z_{t-2}\}$ from $\{x_{t+2},x_{t+3},\ldots,x_{n_{1}},y_{t+2},y_{t+3},\ldots,y_{n_{2}}\}$ such that $t-j-1\geq|N(z_{j})|\geq t-j-2$ for all $1\leq j\leq t-2$.

Proof.

Let us assume there exist vertices $\{z_{1},z_{2},\ldots,z_{t-2}\}$ from $\{x_{t+2},x_{t+3},\ldots,x_{n_{1}},y_{t+2},y_{t+3},\ldots,y_{n_{2}}\}$ such that $t-j-1\geq|N(z_{j})|\geq t-j-2$ for all $1\leq j\leq t-2$. As $G$ is a Type-II(b) BCG, $x_{t}y_{t+1}\in E(G)$ and $x_{t+1}y_{t-1}\in E(G)$. Consider a vertex partition $\pi=\{V_{1},V_{2},\ldots,V_{t},V_{t+1}\}$ of $G$ as follows: $\{x_{1},y_{1},z_{t-2}\}\subseteq V_{1}$; for all $2\leq i\leq t-2$, we set $V_{i}=\{x_{i},y_{i},z_{t-i-1}\}$, $V_{t-1}=\{x_{t-1},y_{t-1}\}$, $V_{t}=\{x_{t},x_{t+1}\}$, and $V_{t+1}=\{y_{t}\}$. We put the other vertices of $G$ in $V_{1}$. This partition $\pi$ is illustrated in Figure LABEL:fig:BCG_type-II(b)_partition.

Figure 7: The partition $\pi$ of Type-II(b) BCG $G$, where dotted edges are not present in $G$

We show that $\pi$ is a tournament transitive partition of $G$ with size $t+1$. Since $G$ is a Type-II BCG, the set $\{x_{1},x_{2},\ldots x_{t},y_{1},\ldots,y_{t}\}$ induces a complete bipartite graph $K_{t,t}$ in $G$. Let us consider $V_{i}$ and $V_{j}$ for $2\leq i<j\leq t-2$. From $\pi$, we have $V_{i}=\{x_{i},y_{i},z_{t-i-1}\}$ and $V_{j}=\{x_{j},y_{j},z_{t-j-1}\}$. As $j\geq|N(z_{t-j-1})|\geq j-1$, which implies either $\{x_{1},x_{2},\ldots,x_{j-1}\}\subseteq N(z_{t-j-1})$ or $\{y_{1},y_{2},\ldots,y_{j-1}\}\subseteq N(z_{t-j-1})$. Since $i<j$, either $x_{i}$ is a neighbour of $z_{t-j-1}$ or $y_{i}$ is a neighbour of $z_{t-j-1}$. Furthermore, because $G$ is a Type-II BCG, $x_{j}$ is adjacent with $y_{i}$, and $y_{j}$ is adjacent with $x_{i}$. Therefore, $V_{i}$ dominates $V_{j}$ in $\pi$. Moreover, as $i\geq|N(z_{t-i-1})|\geq i-1$, which implies either $N(z_{t-i-1})\subseteq\{x_{1},x_{2},\ldots,x_{i}\}$ or $N(z_{t-i-1})\subseteq\{y_{1},y_{2},\ldots,y_{i}\}$. Since Type-II BCG and $i<j$, $V_{j}$ does not dominate $V_{i}$. Now for $t-1\leq p<q\leq t+1$, as $x_{t+1}y_{t-1}\in E(G)$, clearly, $V_{p}$ dominates $V_{q}$ and $V_{q}$ does not dominate $V_{p}$. Similarly, we can show that $V_{i}$ dominates $V_{j}$ but $V_{j}$ does not dominate $V_{j}$ for others $1\leq i<j\leq k$. Hence, $\pi$ is a tournament transitive partition of $G$ of size $t+1$. Therefore, $TTr(G)\geq t+1$. As we already know from [15], for a bipartite chain graph $G$, $Tr(G)=t+1$, where $t$ is the maximum integer such that $G$ contains either $K_{t,t}$ or $K_{t,t}-\{e\}$ as an induced subgraph. Since $G$ is a Type-II BCG, $G$ contains $K_{t,t}$ as an induced subgraph for maximum $t$. Hence, $Tr(G)=t+1$. According to the definition of tournament transitivity, we know that $TTr(G)\leq Tr(G)$. Therefore, $TTr(G)=Tr(G)=t+1$. ∎

5.3 Tournament transitivity of Type-III BCG

In this subsection, we show that $TTr(G)<Tr(G)$ for a Type-III BCG. For that, we have the following theorem.

Theorem 29.

Let $G$ be a Type-III BCG. Then $TTr(G)<Tr(G)$.

Proof.

By the definition of Type-III BCG, $G$ contains $K_{t,t}$ for maximum $t$ and $x_{t+1}y_{t},x_{t}y_{t+1}\in E(G)$. From [15], as $G$ contains $K_{t+1,t+1}\setminus\{e\}$ for maximum $t$, we have $Tr(G)=t+2$. Now we show that $TTr(G)$ cannot be $t+2$. If possible assume $TTr(G)=t+2$ and let $\pi=\{V_{1},V_{2},\ldots,V_{t},V_{t+1},V_{t+2}\}$ be a tournament transitive partition of $G$. By the Proposition 6 we can assume that $|V_{t+1}|=2$ and $|V_{t+2}|=1$.

Since $G$ is a Type-III BCG, the degree of each vertices from $(X\setminus X_{t})\cup(Y\setminus Y_{t})$ is at most $t$. Now, for any vertex $u\in V_{t+2}$, $u$ must have a degree of at least $t+1$. Therefore, only vertices from $X_{t}\cup Y_{t}$ can be in $V_{t+2}$ (See Figure 8).

Figure 8: The partition $\pi$ of Type-III BCG $G$, where dotted edges are not present in $G$

Let us assume $V_{t+2}=\{y_{j}\}$ for some $1\leq j\leq t$. Since $\pi$ is a tournament transitive partition of $G$, $V_{t+1}$ must contain a vertex adjacent to $y_{j}$, and the degree of that vertex is at least $t+1$. So, for some $1\leq i\leq t$, assume $x_{i}\in V_{t+1}$. Moreover, $V_{t+2}$ does not dominate $V_{t+1}$, which implies that there exists a vertex in $V_{t+1}$ that is not adjacent to $y_{j}$, and the degree of that vertex is at least $t$. Let $z\in V_{t+1}$ other than $x_{i}$, then $z\in(X\setminus X_{t})\cup Y$.

If $z\in(X\setminus X_{t})$, to dominate $z$, each $V_{1},V_{2},\ldots,V_{t}$ contains exactly one vertex from $\{y_{1},y_{2},\ldots,y_{t}\}$, and the vertex $y_{j}$ must be the vertex $y_{t+1}$. As $y_{t+1}\in V_{t+2}$ and $N(y_{t+1})=X_{t}$, each $V_{1},V_{2},\ldots,V_{t+1}$ contains exactly one vertex from $X_{t}$. We have a contradiction as $|X_{t}|=t$, and we require $t+1$ vertices. Assume $z\in(Y\setminus Y_{t})$. To dominate $z$, each $V_{1},V_{2},\ldots,V_{t}$ contains exactly one vertex from $\{x_{1},x_{2},\ldots,x_{t}\}\setminus\{x_{i}\}$. Again, we have contradiction as $|X_{t}|\setminus\{x_{i}\}=t-1$ and we required $t$ vertices to dominate $z$.

Finally, assume $z\in Y_{t}$. As $V_{t}$ dominates $V_{t+1}$, for $z\in V_{t+1}$ there must exist a vertex from $V_{t}$, say $z^{\prime}$, such that $z$ is adjacent to $z^{\prime}$ and $|N(z^{\prime})\setminus\{z,y_{j}\}|\geq t-1$. If $z^{\prime}\in(X\setminus X_{t})$, then $|N(z^{\prime})\setminus\{z,y_{j}\}|=|N(z^{\prime})|-2\leq t-2=t-2$, which implies $z^{\prime}$ must be a vertex from $X_{t}$. Let us assume $z^{\prime}=x_{r}$ for some $1\leq r\leq t$ and $r\neq i$. As we know, $V_{t+1}$ does not dominate $V_{t}$, so there exists a vertex from $V_{t}$, say $z^{\prime\prime}$, such that $z^{\prime\prime}$ is not adjacent to the vertices from $\{x_{i},z\}$. Therefore, $z^{\prime\prime}$ must be a vertex from $(X\setminus(X_{t}\cup\{x_{t+1}\}))\cup(Y\setminus(Y_{t}\cup\{y_{t+1}\}))$. Consider the case when $z^{\prime\prime}\in(X\setminus(X_{t}\cup\{x_{t+1}\}))$. Since $z^{\prime\prime}\in(X\setminus X_{t}\cup\{x_{t+1}\}))$ and $z^{\prime\prime}\in V_{t}$, we must have $|N(z^{\prime\prime})\setminus\{z,y_{j}\}|\geq t-1$. But $|N(z^{\prime\prime})\setminus\{z,y_{j}\}|=|N(z^{\prime})|-2\leq t-2$. So, $z^{\prime\prime}$ cannot be a vertex from $(X\setminus(X_{t}\cup\{x_{t+1}\}))$. Furthermore, for $z^{\prime\prime}\in(Y\setminus(Y_{t}\cup\{y_{t+1}\}))$, $|N(z^{\prime\prime})\setminus\{x_{r},x_{i}\}|\geq t-1$. But $|N(z^{\prime\prime})\setminus\{x_{r},x_{i}\}|=|N(z^{\prime\prime})|-2\leq t-2$. So, $z^{\prime\prime}$ cannot be a vertex from $(Y\setminus(Y_{t}\cup\{y_{t+1}\}))$. We have a contradiction, which implies that $z$ cannot be in $V_{t+1}$. From the above discussion, we have if $TTr(G)=t+2$, there does not exist any $z\in V_{t+1}$ such that $z$ is not adjacent with $y\in V_{t+1}\cap Y_{t}$. Similarly, we can show that contradiction when $x\in V_{t+1}\cap X_{t}$. Hence, $TTr(G)<t+2=Tr(G)$. ∎

6 Conclusion and future works

In this paper, we have studied the notion of tournament transitivity in graphs, which is a variation of transitivity. We have shown that the decision version of this problem is NP-complete for chordal graphs (connected), perfect elimination bipartite graphs (disconnected), and doubly chordal graphs (disconnected). On the positive side, we prove that this problem can be solved in polynomial time for trees. Furthermore, we have characterized Type-I BCG with equal transitivity and tournament transitivity and find some sufficient conditions under which the above two parameters are equal for a Type-II BCG. Finally, we have shown that for Type-III BCG, these two parameters are never equal.

This paper concludes by addressing some of the several unresolved problems in the study of tournament transitivity of a graph.

1. What is the necessary condition for a Type-II BCG with $TTr(G)=Tr(G)$?

2. We know form [15], in linear time we can solved the transitivity problem in bipartite chain graphs. Can we design an algorithm for tournament transitivity in a bipartite chain graph?

3. Characterize connected graphs with $TTr(G)=2$ or $TTr(G)=3$.

4. What is the necessary and sufficient condition for $TTr(G)\geq t$, for an integer $t$?

It would be interesting to investigate the complexity status of this problem in other graph classes. Designing an approximation algorithm for this problem would be another challenging open problem.