2
$\begingroup$

I'm trying to prove that every planar graph with no cycles of length $3,4,5$ is $3$-colorable. However, I have no opportunity to receive any validation or correction on it, but it would be very helpful for me to learn about the topic, so any assistance or feedback is appreciated.

Let $G = (V,E)$, $n = |V(G)|$, $e = |E(G)|$, be a planar graph with no cycles of length $3,4,5$.

Claim. $G$ is $3$-colorable.

Proof. We first not that if $G$ has no cycles, then $G$ is a forest and thus $3$-colorable. We may therefore assume $G$ has at leas one cycle. Let $M = \{(e,f) \ | \ e \in E(G), f \in F(G)\}$ where $e$ lies on the boundary of $f$ for every $(e,f) \in M$. Then every edge is on the boundary of exactly two faces, whereas $G$ has at least $6$ edges on the boundary of each face. By double counting thus, $2e = |M| \geq 6f$. By Euler's formula thus

\begin{align*} f = 2 + e - n \iff 2e \geq 6f = 12 + 6e - 6n \iff e \leq \frac{3n}{2} - 3. \end{align*}

We now proceed by induction on $n$. Since $G$ has at least one cycle that cannot be of length $3,4,5$, $n \geq 6$. For $n=6$ we have $G = C_6$, which is $3$-colorable. Assume now $n \geq 7$ and the claim holds for smaller n. Let $v \in V(G)$ such that $d_{G}(v) \leq 2$. Indeed, such vertex exists as otherwise $d_{G}(v) \geq 3 \ \forall v \in V(G)$ and thus by the Handshaking lemma

\begin{align*} e = \frac{1}{2} \sum_{v \in V(G)} d(v) \geq \frac{3n}{2} > \frac{3n}{2} - 3 \end{align*}

which is a contradiction to the result above. Let now $G' = G - \{v\}$ be such coloring. By induction hypothesis, $G'$ is $3$-colorable. Let $c: \ V(G') \mapsto \{1,2,3\}$. We now use $c$ to color $G$ and additionally color $v$ with any color from $\{1,2,3\}$ not already used to color a vertex in $N(v)$. Indeed, this is proper as $d_{G}(v) \leq 2$. Thus $G$ is $3$-colorable and this completes the proof. \qed

Is this correct? Am I doing something wrong?

$\endgroup$

1 Answer 1

1
$\begingroup$

Your proof looks pretty good to me, especially the idea for considering the size of the graph is very brilliant. Good job!

The only flaw to me is the statement "every edge is on the boundary of exactly two faces." This is not always the case unless $G$ is $2$-connected. I was also confused by such illusion at the beginning.

You can consider the case when $G$ is obtained from $C_6$ by joining a new vertex $v$ to anyone on the cycle and put it inside the finite face enclosed by $C_6$. The single edge incident to $v$ only belongs to the finite face.

$\endgroup$
1
  • 1
    $\begingroup$ You're clearly right with mentioning the flaw regarding the set $M$. Perhaps I should say $(e,f) \in M$ if $e$ appears along the boundary of a given face $f$ in a drawing of $G$. Thus every edge is counted twice towards $|M|$. Thank you very much for your insight. :) $\endgroup$
    – ninaPh99
    Commented May 18 at 16:00

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .