2
$\begingroup$

Let $A$ be an elementary Abelian uncountable $p$-group. Is it known if there is an action of a Prufer $q$-group (here $q$ is a prime not necessarily distinct from $p$) $C_{q^{\infty}}$ onto $A$ such that $A$ does not contain proper uncountable $C_{p^{\infty}}$-invariant subgroups?

$\endgroup$
0

1 Answer 1

5
$\begingroup$

The answer is no:

For all primes $p,q$, for every uncountable elementary abelian $p$-group, for every action of the Prüfer $q$-group $C_{q^\infty}\simeq\mathbf{Z}[1/q]/\mathbf{Z}$ on $A$ by group automorphisms, there exists a proper uncountable $C_{q^\infty}$-invariant subgroup in $A$.

There are 2 distincts proofs: for $q=p$ and $q\neq p$.

When $q=p$: Let $(x_n)_{n\ge 0}$ be generators of $C_{p^\infty}$ with $x_0=1$ and $x_{n+1}^p=x_n$ for all $n$. Define $y_n=x_n-1$ (in the group algebra $\mathbf{F}_p[C_{p^\infty}]$). Then $y_{n+1}^p=y_n$ for all $n$ and $y_0=0$. Define $I_n=y_nA$; then $I_n\subset I_{n+1}$ for all $n$; define $I_\infty=\bigcup I_n$. Then modulo $I_\infty$, every $x_n$ acts as the identity and hence every $\mathbf{F}_p$-subspace of $A$ containing $I_\infty$ is a submodule. Hence if $I_\infty\neq A$, then $A$ contains proper submodules of at most countable index (hence uncountable if $A$ is uncountable).

If $I_\infty=A$ and $A$ is uncountable, then $I_n$ is uncountable for some some $n$. If $I_n\neq A$, we are done. Otherwise, $y_nA=A$ and $y_n$ nilpotent force $A=0$, contradiction.


Now assume that $q\neq p$.

If $A$ is uncountable and splits as a direct sum of two submodules, one of the two is uncountable and we are done.

Assume otherwise ($A$ is "indecomposable"). Again denote by $x_n$ the generators of $C_{q^\infty}$ with $x_0=1$ and $x_{n+1}^q=x_n$. Then $x_n$ is an operator on $A$ vanished by some polynomial (namely $X^{q^n}-1$) with simple roots, hence for each $n$, we have a decomposition $A=\bigoplus_{P\in\mathcal{P}_n}A_P$, where $\mathcal{P}_n$ is the set of monic irreducible divisors of $X^{q^n}-1$ and $A_P=\mathrm{Ker}(P(x_n))$. Since $A$ is indecomposable, there exists a unique $P\in\mathcal{P}_n$ such that $A_P\neq 0$; denote it by $P_n$. So $P_n$ is irreducible and $P_n(x_n)=0$. Thus the subring $R_n$ of $\mathrm{End}_{\mathbf{F}_p}(A)$ generated by $x_n$ is a (finite) field. We have $R_n\subset R_{n+1}$ for all $n$ and hence $K=\bigcup R_n$ is a field as well; it is at most countable. On the other hand, the image of $C_{q^\infty}$ in $\mathrm{End}_{\mathbf{F}_p}(A)$ is contained in $K$. Hence any decomposition of $A$ as a $K$-module is $C_{q^\infty}$-invariant. In particular, we deduce that if $A$ is uncountable, then it is decomposable (as a direct sum of uncountably many countable submodules), contradiction.

$\endgroup$
7
  • $\begingroup$ First of all thanks. When $p\neq q$ nothing is known? $\endgroup$
    – W4cc0
    Commented Apr 20, 2015 at 17:14
  • $\begingroup$ I thought that you don't know about the case $p\neq q$. Anyway still Thanks! $\endgroup$
    – W4cc0
    Commented Apr 21, 2015 at 4:58
  • $\begingroup$ At the time I first posted, I indeed only found a proof for $p=q$... $\endgroup$
    – YCor
    Commented Apr 21, 2015 at 6:44
  • $\begingroup$ I was re-reading the answer and I find that it is unclear to me how to pass from the fact that $x_n$ is an operator on $A$ vanished by... to the decomposition. What you meant by $Ker(P(x_n))$? $\endgroup$
    – W4cc0
    Commented Apr 21, 2015 at 8:43
  • 1
    $\begingroup$ The decomposition lemma I used is the following one (of undergraduate level): let $K$ be a field, $M$ a vector space over $K$ (no dimension restriction). Let $u$ be a self-operator on $M$. Let $P\in K[t]$ be a polynomial such that $P(u)=0$. Assume that $P=\prod P_i$ with the $P_i$ pairwise prime to each other. Then $M=\bigoplus_i\mathrm{Ker}(P_i(u))$. This can usually be used when $P_i$ is a power of an irreducible polynomial; in the precise case here, the $P_i$ can themselves be chosen irreducible because $t^{q^n}-1$ has simple roots in any characteristic $\neq q$. $\endgroup$
    – YCor
    Commented Apr 21, 2015 at 9:56

You must log in to answer this question.

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