5
$\begingroup$

If $G$ is a finite abelian group, then we have a decomposition $$G\cong \prod_{p} G(p)$$ where $G(p)$ is the $p$-Sylow subgroup of $G$. This product makes sense as for all but finitely many primes $p$, we have $G_p=\{0\}$. This is proven by showing that the cardinality of $G$ and $\prod_{p} G(p)$ agree. If we now assume that $P$ is a profinite abelian group, there still exists the notion of a $p$-Sylow subgroup $P(p)$ which is now a pro-$p$-group. I'm curious if there exists an isomorphism $$P\cong \prod_{p} P(p).$$

$\endgroup$
2
  • 4
    $\begingroup$ One can verify this by computing indices as 'supernatural numbers', in the sense of Serre's "Galois cohomology"; but probably easier is to consider the maps $P \to P(p)$ given by $g \mapsto \lim_{N \to \infty} g^{p^N - 1}$, where $N \to \infty$ in the divisibility order. (I might also consider calling your profinite group something other than $P$, which looks like a pro-$p$ group ….) $\endgroup$
    – LSpice
    Jul 31, 2020 at 16:40
  • $\begingroup$ Oh, by the way there is a very natural proof for finite groups, which consists in proving that the natural map $\prod_p G(p)\to G$ is a group isomorphism (rather than computing cardinalities). $\endgroup$
    – YCor
    Jul 31, 2020 at 22:28

1 Answer 1

10
$\begingroup$

This is Proposition 2.3.8 of Ribes and Zaleskii - Profinite groups (second edition). (I originally gave references specifically for the finer structure of profinite Abelian groups, but assuming finite generation, in Section 4.3 of the same book.)

$\endgroup$
2
  • $\begingroup$ I'm particularly interested in the non-finitely generated case. Does anyone know more about that case? $\endgroup$ Jul 31, 2020 at 17:01
  • 2
    $\begingroup$ It turns out those Section 4.3 references were mainly about the finer structure of pro-$p$ groups. I modified now to cite Proposition 2.3.8, which is about pronilpotent groups in general (no finite generation hypotheses). Sorry for the confusion! $\endgroup$
    – LSpice
    Jul 31, 2020 at 17:04

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.