-2
$\begingroup$

Let $F$ be some function graph to graph which preserve graph isomorphism.

Example of such $F$ are the line graph, the $k$-subdivision of $G$ and many others.

$F$ need not preserve the order, the degree sequence, the girth, the size of the maximum independent set and others.

Q1 Must all $F$ preserve the order of the automorphism group?

According to my limited experiments, the answer is positive.


Added: In comments Brendan McKay suggested counterexample for the subdivision transformation and it is the cycle graph.

According to my experiments (modulo errors), subdividing a non-rigid graphs preserves the order of the automorphism group.

Q2 How to explain this observation?


Some tested subdivision examples from sagemath and the "nauty and traces" database:

PetersenGrap, PaleyGraph(73), SRG_100_44_18_20, "had-20" of order 80, "sts-21" of order 70.

$\endgroup$
11
  • 1
    $\begingroup$ In addition to Martti's example, subdivision does not preserve automorphism group. Consider subdividing a cycle. $\endgroup$ Commented Jul 1, 2023 at 14:23
  • $\begingroup$ @BrendanMcKay Thanks, you are right. $\endgroup$
    – joro
    Commented Jul 1, 2023 at 14:40
  • $\begingroup$ @BrendanMcKay Is it coincidence that subdividing random/selected non-rigid graphs preserves the automorphism group? $\endgroup$
    – joro
    Commented Jul 1, 2023 at 15:44
  • 1
    $\begingroup$ The line graph functor also does not preserve the automorphism group. For example, $L(K_4)$ is the complement of $K_2 \sqcup K_2 \sqcup K_2$, which has automorphism group $C_2^3 \rtimes S_3$, of order $48$. $\endgroup$ Commented Jul 2, 2023 at 18:57
  • 1
    $\begingroup$ Although Q1 itself was too naive, it seems to be a source of somewhat interesting questions to take a given injective functor $F$ from graphs to graphs and ask for which graphs $F:\operatorname{Aut}(G) \to \operatorname{Aut}(F(G))$ fails to be an isomorphism. $\endgroup$ Commented Jul 2, 2023 at 19:13

2 Answers 2

5
$\begingroup$

This is an answer to the follow-up question about automorphisms of a subdivision.

Suppose $G$ is a connected graph which is not $2$-regular. Let $G^{(k)}$ be the $k$-subdivision of $G$, i.e., the graph obtained by replacing each edge with a path of length $k$. Then $\operatorname{Aut}(G^{(k)}) \cong \mathrm{Aut}(G)$.

Proof: Clearly every automorphism of $G$ uniquely extends to $G^{(k)}$. Conversely, let $\alpha \in \mathrm{Aut}(G^{(k)})$ and let $v_0 \in V(G) \subseteq V(G^{(k)})$ be a vertex of degree $d(v_0) \ne 2$. Since the vertices introduced by subdivision all have degree $2$, $\alpha(v_0)$ must again be a vertex of $G$. Since the vertices of $G$ are precisely those such that $d_{G^{(k)}}(v,v_0)$ is divisible by $k$, $\alpha$ preserves $V(G)$ and induces an automorphism of $G$ (whose unique extension is $\alpha$).

So subdivision can introduce automorphisms only if $G$ has a connected component which is $2$-regular, i.e., either a finite cycle or a bi-infinite path.

$\endgroup$
2
  • $\begingroup$ Thanks. I am not good at group theory, but doesn't this construction gives infinitely many groups of given order? (The orders of $G$ and $G^k$ are different). $\endgroup$
    – joro
    Commented Jul 2, 2023 at 13:58
  • 2
    $\begingroup$ @joro They are only different as permutation groups. As abstract groups they are the same. For any abstract group there are infinitely many permutation representations, that's all. $\endgroup$ Commented Jul 2, 2023 at 14:05
3
$\begingroup$

If you send a graph to two disjoint copies of the same graph, you get a transformation sending isomorphic graphs to isomorphic graphs (in fact, you get an endofunctor on the category of graphs), but this does not preserve the order of the automorphism group.

$\endgroup$
1
  • $\begingroup$ Thanks, this works. $\endgroup$
    – joro
    Commented Jul 1, 2023 at 13:19

You must log in to answer this question.

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