Table of Contents
Fetching ...

Acyclic dichromatic number of oriented graphs

Jørgen Bang-Jensen, Lucas Picasarri-Arrieta, Anders Yeo

TL;DR

We study the acyclic dichromatic number $\vec{\chi}_{\rm a}(D)$, a directed analogue of the acyclic chromatic number requiring each color class to induce an acyclic subdigraph and all bipartite color-class subdigraphs to be acyclic. We show that $\vec{\chi}_{\rm a}(D)$ can be arbitrarily larger than the ordinary dichromatic number $\vec{\chi}(D)$, even for tournaments with $\vec{\chi}(D)=2$, via a vertex-splitting construction that yields $\vec{\chi}_{\rm a}(D(G))=\lceil\sqrt{\chi(G)}\rceil$. The paper establishes NP-hardness of deciding $\vec{\chi}_{\rm a}(D)\le k$ for fixed $k$, including split digraphs, while showing a polynomial-time algorithm for the case $\vec{\chi}_{\rm a}(T)\le 2$ in tournaments. It introduces acyclic heroes of tournaments and proves several results—most notably that $\Delta(k,1,1)$ is an acyclic hero—along with conjectures aiming to classify all acyclic heroes. Additionally, tight bounds are derived for $\vec{\chi}_{\rm a}$ in tournaments, and the behavior on degenerate graphs is explored, together with open problems and directions for future research.

Abstract

The dichromatic number $\vecχ(D)$ of a digraph $D=(V,A)$ is the minimum number of sets in a partition $V_1,\ldots{},V_k$ of $V$ into $k$ subsets so that the induced subdigraph $D[V_i]$ is acyclic for each $i\in [k]$. This is a generalization of the chromatic number for undirected graphs as a graph has chromatic number at most $k$ if and only if the complete biorientation of $G$ (replace each edge by a directed 2-cycle) has dichromatic number at most $k$. In this paper we introduce the acyclic dichromatic number $\vecχ_{\rm a}(D)$ of a digraph $D$ as the minimum number of sets in a partition $V_1,\ldots{},V_k$ of $V$ so that the induced subdigraph $D[V_i]$ is acyclic for each $i\in [k]$ and each of the bipartite induced subdigraphs $D[V_i,V_j]$ is acyclic for each $1\leq i<j\leq k$. This parameter, which resembles the definition of acyclic chromatic number for undirected graphs, has apparently not been studied before. We derive a number of results which display the difference between the dichromatic number and the acyclic dichromatic number, in particular, there are digraphs $D$ with arbitrarily large $\vecχ_{\rm a}(D)-\vecχ(D)$, even among tournaments with dichromatic number 2 and bipartite tournaments (where the dichromatic number is always 2). We prove several complexity results, including that deciding whether $\vecχ_{\rm a}(D)\leq 2$ is NP-complete already for bipartite digraphs, while it is polynomial for tournaments (contrary to the case for dichromatic number). We also generalize the concept of heroes of a tournament to acyclic heroes of tournaments.

Acyclic dichromatic number of oriented graphs

TL;DR

We study the acyclic dichromatic number , a directed analogue of the acyclic chromatic number requiring each color class to induce an acyclic subdigraph and all bipartite color-class subdigraphs to be acyclic. We show that can be arbitrarily larger than the ordinary dichromatic number , even for tournaments with , via a vertex-splitting construction that yields . The paper establishes NP-hardness of deciding for fixed , including split digraphs, while showing a polynomial-time algorithm for the case in tournaments. It introduces acyclic heroes of tournaments and proves several results—most notably that is an acyclic hero—along with conjectures aiming to classify all acyclic heroes. Additionally, tight bounds are derived for in tournaments, and the behavior on degenerate graphs is explored, together with open problems and directions for future research.

Abstract

The dichromatic number of a digraph is the minimum number of sets in a partition of into subsets so that the induced subdigraph is acyclic for each . This is a generalization of the chromatic number for undirected graphs as a graph has chromatic number at most if and only if the complete biorientation of (replace each edge by a directed 2-cycle) has dichromatic number at most . In this paper we introduce the acyclic dichromatic number of a digraph as the minimum number of sets in a partition of so that the induced subdigraph is acyclic for each and each of the bipartite induced subdigraphs is acyclic for each . This parameter, which resembles the definition of acyclic chromatic number for undirected graphs, has apparently not been studied before. We derive a number of results which display the difference between the dichromatic number and the acyclic dichromatic number, in particular, there are digraphs with arbitrarily large , even among tournaments with dichromatic number 2 and bipartite tournaments (where the dichromatic number is always 2). We prove several complexity results, including that deciding whether is NP-complete already for bipartite digraphs, while it is polynomial for tournaments (contrary to the case for dichromatic number). We also generalize the concept of heroes of a tournament to acyclic heroes of tournaments.

Paper Structure

This paper contains 12 sections, 32 theorems, 46 equations, 5 figures.

Key Result

Proposition 1

For every graph $G$, $\vec{\chi}_{\rm a}(D(G)) = \left\lceil \sqrt{\chi(G)}\right\rceil$.

Figures (5)

  • Figure 1: Figures (b) and (d) illustrate the digraph $D(G)$ for the two graphs shown in Figures (a) and (c), respectively.
  • Figure 2: A digraph $D$ together with one of its $2$-dicolourings $\varphi$ (left), and the corresponding digraph $D^\star$ with one of its acyclic $3$-dicolourings (right).
  • Figure 3: The partition $(V(P),X_0,\dots,X_{\ell+1})$.
  • Figure 4: The only possible structure of any $0$-coherent colouring of $T[Z_0]$. Vertices coloured $1$ are in blue, and vertices coloured $2$ in orange. As indicated by the twisting arrows, the acyclic ordering of each $X_j$, $j\geqslant 1$, goes from right to left.
  • Figure 5: The digraph $W[x,y]$ together with one of its acyclic $2$-dicolourings.

Theorems & Definitions (68)

  • Proposition 1
  • proof
  • Proposition 2
  • proof
  • Theorem 4: Berger et al. bergerJCT103
  • Conjecture 5
  • Theorem 6
  • Theorem 7
  • Theorem 8
  • Theorem 9: Harutyunyan et al. harutyunyanJTCB138
  • ...and 58 more