2
$\begingroup$

The definition of block is

Block of $G$ is a maximal subgraph $G'$ of $G$ with no cut vertex of $G'$ itself.

Of course, there can exist many blocks in $G$.
In particular, isolated vertices, edges in bridge, and maximal $2$-connected subgraph of $G$ are blocks of $G$.

The definition of leaf block is

Leaf block of $G$ is a block of $G$ containing only one cut-vertex of $G$.

First, I found a lemma about $k$-regular tree, whose vertices except leaves have degree $k$.

Lemma: If a tree $T$ of order $n$ has $m$ vertices of degree $k$ and $n-m$ leaves, $k−1$ must divide $n−2$.

Proof: By handshaking lemma, $$ km + (n-m) = \sum_{v \in V(T)} \deg(v) = 2\vert E(T) \vert = 2(n-1)$$ Then $$ (k-1)m=n-2 $$ Since all terms $\in \mathbb{Z}$, $k-1$ divides $n-2$.

I will use $(k-1)m=n-2$ part here.
And this is my guess:

To make the cubic graph with maximum # of leaf blocks, make a $3$-regular tree, and change its leaves into the smallest $3$-regular leaf blocks.

I think the smallest $3$-regular leaf block is as following (but not certain):

enter image description here

So my conclusion is

Suppose $G$ of order $n$ has $\ell$ number of leaf blocks. First, we construct a $3$-regular tree with $n-4\ell$ vertices and $\ell$ leaves. By the lemma, $2(n-5\ell)=n-4\ell-2$. Then $\ell=\frac{n+2}{6}$. After constructing a tree with $n-4\ell$ vertices, change all leaves into the smallest $3$-regular leaf blocks shown above. It adds $4$ vertices to each leaf. Then we can make a $3$-regular graph with $n$ vertices and $\ell$ leaf blocks.

But there are so many assumptions, and such tree exists only when $n \equiv 4$ (mod 6).
So I cannot conclude $\ell = \lfloor \frac{n+2}{6} \rfloor$ is the maximum. Would you help me?

$\endgroup$
3
  • $\begingroup$ 2 mod 6 is not correct. Draw the smallest examples. $\endgroup$ Commented Oct 10, 2021 at 4:29
  • $\begingroup$ @BrendanMcKay Sorry, it's a mistake. Now I revised that. $\endgroup$
    – okw1124
    Commented Oct 10, 2021 at 5:38
  • $\begingroup$ @PeterTaylor Thanks. Now I revised it. $\endgroup$
    – okw1124
    Commented Oct 13, 2021 at 2:04

1 Answer 1

0
+50
$\begingroup$

Yes, your conjecture is true. In fact, we can prove something stronger. All the extremal examples actually come from your construction. Say that a graph is special if it can be obtained from a $3$-regular tree $T$ (a tree where every vertex has degree $1$ or $3$) by replacing each leaf with the gadget in your picture.

Lemma. Let $\ell \geq 2$, and let $G$ be a smallest connected cubic graph with exactly $\ell$ leaf blocks. Then $G$ is special.

Proof. Let $B_1, \dots, B_\ell$ be the leaf blocks of $G$ and $x_1, \dots, x_\ell$ be the cut-vertices of $G$ such that $x_i \in V(B_i)$. Fix $i \in [\ell]$. Since $B_i$ is $2$-connected, $\deg_{B_i}(x)=2$. Therefore, $B_i$ has $|V(B_i)|-1$ vertices of degree $3$ and exactly one vertex of degree $2$. The smallest such graph with this property is the gadget, and it is the unique smallest graph. Therefore, by the minimality of $|V(G)|$, each $B_i$ is isomorphic to the gadget. Since $G$ is cubic, $x_1, \dots, x_\ell$ are distinct vertices. Let $Y= \bigcup_{i \in [\ell]} V(B_i) \setminus \{x_i\}$ and $T=G-Y$. Since $G$ is connected, $T$ is a connected graph with exactly $\ell$ vertices of degree $1$ and $|V(T)|-\ell$ vertices of degree $3$. The smallest such graph with this property is a $3$-regular tree. By the minimality of $|V(G)|$, it follows that $T$ is a $3$-regular tree. In other words, $G$ is special. $\square$

The smallest cubic graph with exactly one leaf block is $K_4$, which technically is not covered by the above lemma.

Corollary. Every connected $n$-vertex cubic graph has at most $\lfloor \frac{n+2}{6} \rfloor$ leaf blocks.

Proof. If $G$ is a connected $n$-vertex cubic graph with exactly $\ell$ leaf blocks, then $G$ has at least as many vertices as a special graph with $\ell$ leaf blocks by the previous lemma. Thus, $n \geq 6\ell-2$. $\square$

$\endgroup$
5
  • $\begingroup$ Thanks a lot! But I have one more question: If $n \equiv 2$ (mod $6$) or $n \equiv 0$ (mod $6$), there are $4$ or $2$ remaining vertices. We should replace one gadget to bigger block, or add them to $T$. I think the former is reasonable way, since the latter makes $T$ no more tree. Then how can we assure that $G$ is the smallest? $\endgroup$
    – okw1124
    Commented Oct 13, 2021 at 9:15
  • $\begingroup$ The way I wrote the proof, we do not really care about $n$. That upperbound holds for all values of $n$. However, you can get examples that match the upper bound for all values of $n$ by doing the following operation. Start with a special graph, choose two edges $e$ and $f$ and subdivide them. Then add an edge between the two new subdivided vertices. This creates a new cubic graph with two more vertices. You can apply this operation by choosing two edges in the same leaf block or two edges in the tree. It doesn't matter. You'll need to do it twice if $n \equiv 0 \pmod{6}$. $\endgroup$
    – Tony Huynh
    Commented Oct 13, 2021 at 9:31
  • $\begingroup$ It may seem obvious, but can I ask one more thing? Here we fixed $\ell$ and found that $n=6\ell-2$ is the smallest $n$ with the given condition. But can we say $\ell=\frac{n+2}{6}$ is the largest $\ell$ with fixed $n$? $\endgroup$
    – okw1124
    Commented Oct 13, 2021 at 11:05
  • $\begingroup$ If you take the floor, then yes. The upperbound on $\ell$ follows from the corollary. The lowerbound follows from special graphs (and their modifications as I described in my previous comment). $\endgroup$
    – Tony Huynh
    Commented Oct 13, 2021 at 11:58
  • $\begingroup$ Now I fully understood the proof. Thanks a lot! $\endgroup$
    – okw1124
    Commented Oct 15, 2021 at 12:57

You must log in to answer this question.

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