2
$\begingroup$

Let $G$ be an abelian group. If all proper subgroups of $G$ are finite, what can we say about $G$. Are the any properties, $G$ is guaranteed to have.

$\endgroup$
3
  • $\begingroup$ Maybe check this out math.stackexchange.com/questions/155521/… $\endgroup$
    – H_B
    Sep 13, 2015 at 1:44
  • $\begingroup$ If $G$ is finitely generated it must be finite under your assumption. $\endgroup$
    – Sempliner
    Sep 14, 2015 at 2:02
  • $\begingroup$ Not exactly the same thing, but a group whose finitely generated subgroups are all finite is called locally finite. It's a wider class than the one you're considering (which only contains finite groups and the Prüfer groups). $\endgroup$ Sep 15, 2015 at 12:16

1 Answer 1

4
$\begingroup$

$\newcommand{\bZ}{\mathbb Z} \newcommand{\soc}{\operatorname{soc}}$Obviously, finite abelian groups as well as the $p$-Prüfer groups $\bZ(p^\infty)$ for primes $p$ have this property. These are in fact all the possibilities.

Proposition. Let $G$ be an abelian group such that all proper subgroups of $G$ are finite. Then either $G$ is finite or there exists a prime $p$ such that $G \cong \bZ(p^\infty)$.

Proof: Suppose that $G$ is not finite. First note that $G$ is indecomposable: If $G \cong H \oplus K$ with nonzero $H$ and $K$, then at least one of $H$ and $K$ must be infinite. Hence $G$ has a proper infinite subgroup.

Since all cyclic subgroups of $G$ are finite, $G$ is a torsion group. Hence it decomposes as $G \cong \bigoplus_{p \in \mathbb P} G_p$, where the $G_p$s are (uniquely determined) $p$-groups. But since $G$ is indecomposable, $G \cong G_p$ for some $p$. Therefore $G$ is an indecomposable infinite abelian $p$-group.

Recall that the height, $H(x)$, of an element $x \in G$ is defined as the maximal $n \in \mathbb N_0 \cup \{\infty\}$ such that there exists $x_n \in G$ with $x=p^nx_n$. Let $G^\infty \subset G$ denote the subset of all elements of infinite height. It is easily checked that $G^\infty$ is a subgroup of $G$.

Claim: If $G$ does not have finite exponent, then $G^\infty$ is infinite.

For $k \ge 0$, let $\soc^k(G)$ denote the $k$-th socle of $G$, that is, the subgroup consisting of all elements of order at most $p^k$. By assumption, $\soc^k(G)$ is finite for all $k$. Since we also assume that $G$ does not have finite exponent, $\soc^{k-1}(G) \subsetneq \soc^{k}(G)$ for all $k \ge 1$. We will show: For all $k$, there exists an $x \in \soc^k(G) \setminus \soc^{k-1}(G)$ (i.e., $\operatorname{ord}(x) = p^k$) with $H(x) = \infty$. Note that $$ \bigcup_{l \ge k} p^{l-k}(\soc^l(G) \setminus \soc^{l-1}(G)) \subset \soc^k(G) \setminus \soc^{k-1}(G). $$ Since the set on the right hand side is finite, while each of the sets in the infinite union on the left is non-empty, it follows that there must exist at least one $x \in \soc^k(G) \setminus \soc^{k-1}(G)$ which is divisible by arbitrary high powers of $p$. But then $H(x) = \infty$. Thus $G^\infty$ is infinite and the claim is shown.

Now we have two cases: If $G$ has finite exponent, then $G=\soc^n(G)$ where $p^n = \operatorname{exp}(G)$. However, then $\soc^{n-1}(G)$ is finite as a proper subgroup, and $\soc^n(G)/\soc^{n-1}(G) \cong (\bZ/p\bZ)^{(I)}$ for some index set $I$. If $I$ is infinite, one easily produces an infinite proper subgroup of $G$. Therefore $I$, and hence also $G$, is finite.

If $G$ does not have finite exponent, then we must have $G = G^\infty$. But then every element is divisible by arbitrary high powers of $p$. Since $G$ is a $p$-group, this implies that $G$ is divisible. It follows that $G$ is a direct sum of $p$-Prüfer groups. Since $G$ is also indecomposable, in fact $G \cong \bZ(p^\infty)$.

$\endgroup$
1
  • $\begingroup$ Maybe I should also add that the nonabelian picture is of course completely different: For instance, there are the Tarski monsters. $\endgroup$
    – moonlight
    Sep 15, 2015 at 12:42

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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