10
$\begingroup$

Here are the exact definitions of the terms:

Let $G$ be a topological group. Then $G$ has the small index property if every subgroup of countable (including finite) index is open in $G$. Furthermore, $G$ has automatic continuity if every group homomorphism from $G$ to any separable topological group is continuous.

It is well-known that $G$ having the small index property is equivalent to the fact that every homomorphism from $G$ to the symmetric group $S_{\infty}$ on a countably infinite set $X$ is continuous, where $S_{\infty}$ is given the usual pointwise topology with sub-basis $\{f\in S_{\infty}: f(x)=y\}$ over all points $x,y \in X$.

Hence automatic continuity of $G$ implies the small index property. Is an example known that shows the converse is false?

$\endgroup$
3
  • $\begingroup$ I wonder if this would be somehow connected to the results about automatic continuity in Banach spaces (and other related spaces) under ZF+DC+"Every set of reals has the Baire measurable". In other words, I wonder if any counterexample would have to be somehow non-constructive. $\endgroup$
    – Asaf Karagila
    Commented Dec 12, 2019 at 18:45
  • $\begingroup$ It would certainly be better to assume $G$ Hausdorff, and probably $G$ Polish would be more interesting. $\endgroup$
    – YCor
    Commented Dec 13, 2019 at 9:04
  • 3
    $\begingroup$ A Polish example would be extra interesting. $\endgroup$ Commented Dec 13, 2019 at 10:09

1 Answer 1

11
$\begingroup$

Here is a counterexample. Consider $\mathbb{R}$ with addition. We define a topology on this group by giving the cosets of all countable index subgroups as a sub-basis.

This subbasis is actually a basis as the intersection of finitely many subgroups of countable index is another subgroup of countable index. Moreover as $\mathbb{R}$ is abelian we can see this is a group topology, as if G is a countable index subgroup then $(G+x) + (G+y) \subseteq (G+x+y)$ and $-(G+x)= (G-x)$ for all $x,y \in \mathbb{R}$. This topolgical group has the small index property by the definition of the topology.

To see it doesn't have automatic continuity, consider the idenity map from the reals with this topology the the reals with the standard topology. This is a homomorphism (even an isomorphism), but the preimage of $(-1, 1)$ is itself. This is not open in our new topology as all open sets in our new topology contain a translation of a non-trivial group, and are thus unbounded.

$\endgroup$
2
  • 2
    $\begingroup$ I think this topology is Hausdorff but not Polish $\endgroup$ Commented Dec 13, 2019 at 9:43
  • 1
    $\begingroup$ Very nice. I'd still be interested in a Polish example. $\endgroup$ Commented Dec 14, 2019 at 7:39

You must log in to answer this question.

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