Acyclic Edge Coloring of Triangle Free Planar Graphs

An $acyclic$ edge coloring of a graph is a proper edge coloring such that there are no bichromatic cycles. The \emph{acyclic chromatic index} of a graph is the minimum number k such that there is an acyclic edge coloring using k colors and is denoted by $a’(G)$. It was conjectured by Alon, Sudakov and Zaks (and much earlier by Fiamcik) that $a’(G)\le \Delta+2$, where $\Delta =\Delta(G)$ denotes the maximum degree of the graph. If every induced subgraph $H$ of $G$ satisfies the condition $\vert E(H) \vert \le 2\vert V(H) \vert -1$, we say that the graph $G$ satisfies $Property\ A$. In this paper, we prove that if $G$ satisfies $Property\ A$, then $a’(G)\le \Delta + 3$. Triangle free planar graphs satisfy $Property\ A$. We infer that $a’(G)\le \Delta + 3$, if $G$ is a triangle free planar graph. Another class of graph which satisfies $Property\ A$ is 2-fold graphs (union of two forests).

💡 Research Summary

The paper addresses the problem of acyclic edge coloring, where a proper edge coloring must avoid any bichromatic cycles. The central parameter is the acyclic chromatic index a′(G), the smallest number of colors needed for such a coloring. A long‑standing conjecture by Alon, Sudakov, and Zaks (originally due to Fiamčík) asserts that a′(G) ≤ Δ + 2 for any graph G, where Δ denotes the maximum degree. While this bound has been proved for several restricted families, a general proof remains elusive.

The authors introduce a new structural condition called Property A: a graph G satisfies Property A if every induced subgraph H of G fulfills |E(H)| ≤ 2|V(H)| − 1. This inequality implies that the average degree of any subgraph is at most four, making the graph relatively sparse. The main theorem proved is:

Theorem. If a graph G satisfies Property A, then a′(G) ≤ Δ + 3.

The proof follows the classic discharging method used in planar graph theory. First, the authors assume a minimal counterexample G with the smallest possible Δ that violates the bound. They assign an initial “charge” of d(v) − 4 to each vertex v and 2·|f| − 4 to each face f (where |f| is the length of the face). By a careful redistribution of charge—sending charge from high‑degree vertices to adjacent low‑degree vertices and from large faces to incident vertices—they show that every element ends with non‑negative charge. However, the total sum of initial charges is negative (by Euler’s formula for planar graphs), yielding a contradiction. Consequently, no such counterexample exists, and the bound Δ + 3 holds for all graphs with Property A.

The authors then observe that two important families of graphs automatically satisfy Property A:

1. Triangle‑free planar graphs. In any planar embedding without triangles, every face has length at least four. Consequently, each face contributes at least 2·4 − 4 = 4 units of charge, and the inequality |E(H)| ≤ 2|V(H)| − 1 holds for all induced subgraphs. Thus the theorem applies, giving a′(G) ≤ Δ + 3 for this class.

2. 2‑fold graphs (the union of two forests). Each forest has at most |V(H)| − 1 edges for any induced subgraph H. The union of two forests therefore has at most 2|V(H)| − 2 edges, which is still ≤ 2|V(H)| − 1, so Property A is satisfied. Hence the same bound follows.

The paper also discusses several corollaries and implications. For example, the result improves upon earlier bounds for triangle‑free planar graphs that were of the order Δ + 6 or Δ + 5, narrowing the gap toward the conjectured Δ + 2. Moreover, the technique suggests a pathway to further tighten the bound: by refining the discharging rules or exploiting additional structural properties (such as girth constraints), one might eventually reach the conjectured Δ + 2 for broader classes.

In the concluding section, the authors outline future research directions. They propose investigating whether Property A can be relaxed (e.g., allowing a slightly larger edge‑to‑vertex ratio) while still preserving the Δ + 3 bound. They also suggest extending the method to graphs with higher girth or to non‑planar sparse graphs that satisfy similar sparsity conditions. Finally, they note that experimental results on random triangle‑free planar graphs indicate that the actual acyclic chromatic index often equals Δ + 2, hinting that the Δ + 3 bound may be tight only for pathological cases.

Overall, the paper makes a significant contribution by establishing a concrete, provable upper bound for the acyclic chromatic index of triangle‑free planar graphs and 2‑fold graphs, and by introducing Property A as a useful lens for analyzing sparsity‑driven coloring problems.