2
$\begingroup$

Let $G$ be a locally compact group with the group of topological group automorphisms $Aut(G)$ furnished with the compact-open topology. Let $B$ be a subgroup of $Aut(G)$. We call $G$ an [IN$]_B$ if there is a $B$-invariant relatively compact neighbourhood of the identity element of the group $G$.

It is known that (e.g. look at [1, Theorem 2.5]) for an [IN$]_B$ group $G$, the intersection of all $B$-invariant relatively compact neighbourhoods of the identity forms a compact subgroup of $G$, denoted here by $K_B$.

I am looking for a group $G$ with a subgroup $B$ in $Aut(G)$ so that $K_B$ is not a normal subgroup.

Remarks:

  1. One can easily note that the group $B$ cannot include all the inner automorphisms.
  2. By [1, Theorem 0.1], we can conclude that $B$ cannot be relatively compact in $Aut(G)$. Otherwise, $G$ would be [SIN$]_B$ and in this case, $K_B$ is the trivial group of the identity.

[1] Grosser, Siegfried; Moskowitz, Martin; Compactness conditions in topological groups. J. Reine Angew. Math. 246 1971 1–40, DOI: 10.1515/crll.1971.246.1.

$\endgroup$

2 Answers 2

1
$\begingroup$

Take $G=\text{GL}_2(\mathbb{Z})\ltimes (\mathbb{R}/\mathbb{Z})^2$ and take $B<G$ to be the $\begin{bmatrix} 1 & * \\ 0 & 1 \end{bmatrix} \simeq \mathbb{Z}$. Then $K_B$ is one of the factors $\mathbb{R}/\mathbb{Z}$.

$\endgroup$
2
  • $\begingroup$ Would $K_B$ be non-normal? It seems central to me. $\endgroup$
    – Mahmood Al
    Jun 15, 2017 at 8:16
  • $\begingroup$ it will not commute with the opposite unipotent. $\endgroup$
    – Uri Bader
    Jun 15, 2017 at 8:32
0
$\begingroup$

Let $F$ be a non commutative finite group and let $\Bbb{Z}$ be the group of integers. By $\prod_{n \in \Bbb{Z}} F$ we mean the compact group of all $(a_n)_{n \in \Bbb{Z}}$, $a_n \in F$ $\forall n \in \Bbb{Z}$, equipped with the product topology. It is easy to observe that $$ \psi: \Bbb{Z} \rightarrow Aut\left( \prod_{n \in \Bbb{Z}} F\right) $$ where $$ \psi(k)\left((a_n)_n\right)=(a_{n+k})_n, \quad (a_n)_{n \in \Bbb{Z}}\in \prod_{n \in \Bbb{Z}} F. $$ Let $N = \prod_{n \in \Bbb{Z}} F \rtimes_\psi \Bbb{Z}$ be the semi direct product of these two groups.

Let $\bigoplus_{n \in \Bbb{Z}} F$ be the discrete group of the direct product of the groups $F$ index by $\Bbb{Z}$ which is the set of $(b_n)_{n\in \Bbb{Z}}$ where $b_n$ is the identity of $F$ except for finitely many $n \in \Bbb{Z}$.

Note that $\phi$ is a group homomorphism given by $$ \phi: N \rightarrow Aut\big(\bigoplus_{n \in \Bbb{Z}} F\big) $$ where $$ \phi\big( (a_n)_n, k \big) (b_n)_n = (a_n b_{n+k} a_n^{-1}). $$

Let $$ G = \bigoplus_{n \in \Bbb{Z}} F \rtimes_\phi \big( \prod_{n \in \Bbb{Z}} F \rtimes_\psi \Bbb{Z}\big). $$ and let $B$ be the group of inner automorphisms of the elements of subgroup $\{0\} \times \{0\} \times \Bbb{Z}$. Indeed, for each $\beta_\ell \in B$ ($\ell \in \Bbb{Z}$),
$$ \beta_\ell \big( (a_n)_n, (b_n)_n, k\big):= \big(0,0,\ell\big)\big( (a_n)_n, (b_n)_n, k\big)\big(0,0,-\ell\big) = ( (a_{n+\ell}), (b_{n+\ell}), k\big). $$ One can show that $K_B$ (defined in the question) is nothing but the compact group $$ \{0\} \times \prod_{n \in \Bbb{Z}} F \times \{0\}. $$ But $K_B$ cannot be normal because, \begin{eqnarray*} & & \big((a_n)_n, (c_n)_n, k) ( 0 , (b_n)_n, 0\big) \big((a_n)_n, (c_n)_n, k\big)^{-1}\\ &=& \big((a_n)_n, (c_n b_{n+k})_n, k\big) \big( (c_{n-k}^{-1} a_{n-k}^{-1} c_{n-k})_n, (c_{n-k}^{-1}), -k \big)\\ &=& \big( (a_n c_n b_{n+k} c_{n}^{-1} a_{n}^{-1} c_n b_{n+k}^{-1} c_n^{-1})_n, ( c_n b_{n+k}c_n^{-1})_n, 0 \big). \end{eqnarray*} So if $F$ is non-commutative enough (and I have not made any mistakes in my claims), this would be an answer to my own question.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.