Let $G$ be a finitely generated abelian group and $a$ be a nontrival element of $G$ contained in all nontrivial subgroups of $G$. Is $G$ necessarily cyclic?
-
5$Q_{8}$ or something... – Gina Jul 27 '14 at 02:07
-
@Gina: Is $-1$ common in all subgroups of $Q_8$? – Minimus Heximus Jul 27 '14 at 02:10
-
2@QuangHoang: Putting your mouse over the down arrow shows the text "This question does not show any research effort; it is unclear or not useful." I completely understand the downvote. – Michael Albanese Jul 27 '14 at 02:12
-
So probably it is true in abelian case? – Minimus Heximus Jul 27 '14 at 02:15
-
2@user138171:It's the only element with order $2$, and by Cauchy's Lagrange's, every nontrivial subgroup must have element of order $2$. – Gina Jul 27 '14 at 02:16
-
It's not even true in the abelian case. – Quang Hoang Jul 27 '14 at 02:36
-
@Gina: what if $G$ has an odd order? – Jack D'Aurizio Jul 27 '14 at 02:37
-
1@QuangHoang: I think it should be true in abelian case. Since if the cyclic decomposition have at least 2, then there are trivially intersected subgroup. – Gina Jul 27 '14 at 02:38
-
I'm thinking the "centered product" of two copies of $\mathbb Z/p^2$ with identical $\mathbb Z/p$. Am I missing something? – Quang Hoang Jul 27 '14 at 02:41
-
For sure, a group with a "persistent element" must be a p-group, since if two different primes $p$,$q$ divide the order of $G$, the subgroups generated by an element of order $p$ and an element of order $q$ intersect trivially. – Jack D'Aurizio Jul 27 '14 at 02:43
-
@QuangHoang: wouldn't that be just $\mathbb{Z}{p}\times\mathbb{Z}{p}$? Or am I missing something? (never worked with central product before) – Gina Jul 27 '14 at 02:46
-
Possible duplicate of http://math.stackexchange.com/questions/863390/subgroup-contained-in-all-other-subgroups/863600#863600 – mesel Jul 27 '14 at 02:47
-
1@mesel: that's for abelian case only. The way it is currently phrased, this question is about the nonabelian case, even though the asker's comment indicate it might have been originally meant to be just abelian. – Gina Jul 27 '14 at 02:49
-
@mesel, Jack A'Aurizio: that question is on finite groups. I did not restrict the question to finite groups. btw, I do not disagree with restricting to abelian groups. – Minimus Heximus Jul 27 '14 at 02:51
-
1OK. So it's true for abelian case for obvious reason (which I failed to see). @Gina: that group is of order $p^3$ and is not abelian :-(. – Quang Hoang Jul 27 '14 at 02:52
-
@user13871: It doesn't need to be finite in the other question. By the way, I vote to close this for duplication. – Quang Hoang Jul 27 '14 at 02:53
-
@user138171: where did you see this question ? – mesel Jul 27 '14 at 02:55
-
@mesel: I encountered it while proving something. – Minimus Heximus Jul 27 '14 at 02:56
-
@user138171: Is $G$ abelian ? – mesel Jul 27 '14 at 02:59
-
@mesel: yes. But I wondered if it is really needed. – Minimus Heximus Jul 27 '14 at 03:00
-
@QuangHoang: It seems the proof there is based on finiteness. – Minimus Heximus Jul 27 '14 at 03:05
-
The answer uses the same idea with @Gina's comment: Cyclic decomposition of finitely generated abelian group. – Quang Hoang Jul 27 '14 at 03:10
-
@QuangHoang: Gina's comment is complete proof in abelian case. That question is completely for finite case. – Minimus Heximus Jul 27 '14 at 03:13
-
@bof: yes but it may contain a power of it (equal to $a$) – Minimus Heximus Jul 27 '14 at 03:25
-
@JackD'Aurizio: nope, there are no counterexample with odd order. See my answer. In particular, $Q_{8}$ is indeed rather representative of the rest of all the possible counterexample. – Gina Jul 27 '14 at 03:40
3 Answers
Let's consider what is the condition such that $G$ would have such element $a$. For simplicity, only finite $G$ are examined.
First, $|G|$ must be a prime power, otherwise it is false by Cauchy's. Let $|G|=p^{n}$ ($p$ prime, $n\geq 1$). Then any abelian subgroup of $G$ must be cyclic, by the proof using cyclic decomposition as outline in the comment: for any nontrivial abelian subgroup $H$ of $G$, we can always do the cyclic decomposition $H=\mathbb{Z}_{k_{1}}\times \mathbb{Z}_{k_{2}}\times\ldots\mathbb{Z}_{k_{t}}$. If $t\geq 2$ then $H$ have 2 subgroups that intersect trivially: $\mathbb{Z}_{k_{1}}$ and $\mathbb{Z}_{k_{2}}$ (or more correctly, their embedding in $H$), which in turn mean that they are both subgroup of $G$ that not both containing $a$. Hence $t=1$ so $H$ is cyclic.
By a theorem, http://books.google.com/books?id=0268b52ghcsC&pg=262#v=onepage&q&f=false all subgroup of a such group must be either cyclic or a generalized quarternion group. In particular, $G$ is either cyclic or a generalized quarternion group. Hence the only counterexample would be the generalized quarternion group:
$<x,y|x^{2n}=y^{4}=1,x^{n}=y^{2},y^{-1}xy=x^{-1}>$
Because $|G|=p^{n}$ this means $p=2$. Generalized quarternion group have order at least $8$. Hence $G$ is a generalized quarternion group with order $2^{n}$ for $n\geq 3$ are all the exception to the claim. Note that the only possible persistent element in those counterexample are the lone element with order $2$.
As for infinite $G$, I have no ideas. Note that every cyclic group must have finite order, as infinite cyclic group do not have persistant element. Also, just like above, we still have $a$ have order $p$ for some prime $p$, and so each element still have order a power of $p$. Hence $G$ is now an infinite $p$-group that is finitely generated. The existence of such a group is the famous nontrivial Burnside's problem. There is a theorem that prove such a group exist http://en.wikipedia.org/wiki/Golod-Shafarevich_theorem but there is no guarantee (and in fact rather unlikely) that the group constructed by that theorem will have a persistent element. Hence I will leave this to the expert.
Now if we remove the requirement that the group have to be finitely generated, then an example can be found. Notice that you can embedded $Q_{4n}$ inside $Q_{4(2n)}$ (the generalized quarternion group). Hence by keep embedding like that, you get a chain of group (chain ordered by subset relation). Simply use take the union of the whole chain to get the infinite group you want.
- 5,408
-
that's interesting. But do you have any counterexample for infinite non-abelian case? – Minimus Heximus Jul 27 '14 at 03:52
-
1@user138171: nope, that's even tougher than the infamous Burnside's problem. However, I edited in the explanation why it is hard, and a counterexample when finitely generated requirement is removed. – Gina Jul 27 '14 at 05:25
-
-
@user138171: well an early sign that this is going to be difficult is that there are no infinite abelian group that satisfy the requirement, so you can't even get a cylic group to fulfill the requirement. In fact, I suspect such a group cannot exist, cyclic or not. – Gina Jul 27 '14 at 06:07
-
-
I accept this answer because of your comments above answering original and current question. – Minimus Heximus Jul 27 '14 at 08:22
-
Another counter-example to the infinite-generated case is Prufer Quasi-cyclic groups (same idea, but with cyclic groups). With respect to finitely generated, infinite $p$-groups, Tarski monsters do not have persistent elements. – user1729 Jul 27 '14 at 14:32
The group $G$ is supposed to be abelian. The subgroup generated by the persistent element must be simple. The group can't have any other simple subgroups, because the sum of simple subgroups is a direct sum of simple groups. Also the group must be indecomposable, because the persistent element can't belong to both a direct summand and its complement.
If $H$ is the simple subgroup generated by the persistent element, it intersects non trivially every non zero subgroup, so it is essential. In particular, the injective envelope of $G$ is the same as the injective envelope of $H$, so it is the Prüfer $p$-group, where $p=|H|$, because the Prüfer groups are the only indecomposable divisible torsion groups.
In particular $G$ is isomorphic to a subgroup of a Prüfer group; if it is finitely generated, it is cyclic; otherwise it coincides with the Prüfer group.
- 244,946
-
-
@user138171 Yes, of course. Otherwise almost none of the argument would work. – egreg Jul 27 '14 at 17:59
-
"The subgroup generated by the persistent element must be simple" how can Prüfer group be simple?! – Minimus Heximus Sep 12 '14 at 07:32
-
@user138171 The Prüfer group is not simple, but it has persistent elements, precisely those of order $p$. Where's the problem? The subgroup generated by a persistent element is simple. – egreg Sep 12 '14 at 09:34
-
Since $G$ is finitly generatet abelian group;
$$G\cong \mathbb Z^n\oplus \mathbb Z_{n_1}\oplus...\mathbb Z_{n_k}$$
From the question; we have $G\cong \mathbb Z^n\oplus \mathbb Z_{n_i}$ where $n_i=p^n$.
since intersection of components trivial in the direct sum, $G\cong \mathbb Z$ or $G\cong \mathbb Z_{p^n}$. But in $\mathbb Z$, there is no integer which is divisible by all primes so f0r all $n\in Z$ there exist $pZ$ which does not contain it. Thus the answer is same $G\cong \mathbb Z_{p^n}$.