8

Can there be an infinite group $G$ and a nontrivial proper subgroup $1<H<G$ such that if $1<K<G$ and $|K|=|H|$, then $K=H$? Although I'm particularly interested in when $|H|$ is infinite, some examples for finite $H$ have been given in the comment:

  1. $(G,H)=(\mathbb{Z}\times\mathbb{Z}/2\mathbb{Z},\{0\}\times\mathbb{Z}/2\mathbb{Z})$
  2. $(G,H)=(\mathbb{R}^*,\{1,-1\})$
  3. Prüfer groups (I do not know about this)

Note: Originally I allowed $K=G$. I wasn't aware of this distinction but Qiaochu Yuan's answer shows that if $K=G$ is allowed, it is impossible to find such $(G,H)$ where $H$ is infinite.

durianice
  • 2,624
  • 5
    Maybe I'm missing something, but wouldn't the group $\mathbb{Z}\times\mathbb{Z}/2\mathbb{Z}$ work (for your finite case)? It has only one subgroup of order $2$. – Christian E. Ramirez Oct 01 '22 at 07:21
  • 1
    Same idea with $G=(\mathbb R^*,\cdot)$. – Anne Bauval Oct 01 '22 at 07:24
  • 5
    Pruefer groups (also known as quasicyclic groups) are infinite $p$-groups with a single subgroup of order $p^n$ for any fixed $n$. – David A. Craven Oct 01 '22 at 07:30
  • 1
    Thanks for the examples for finite $H$. I changed the question to find infinite $H$. – durianice Oct 01 '22 at 07:53
  • can you explain what is "infinite non-trivial proper subgroup that has a certain order"? (It is confused at first glance to me) – Maths Rahul Oct 03 '22 at 07:34
  • 1
    @MathsRahul You need to interpret the sentence together with "has exactly one". $G$ having exactly one something that has a certain order means that if you fix an order (a cardinality or cardinal), say $\kappa$, then there is such a something with order $\kappa$ in $G$, and such something with order $\kappa$ is unique (but not necessarily for any other something with order not equal to $\kappa$). In this case, the something is an infinite (nontrivial) proper subgroup. – durianice Oct 03 '22 at 13:10

3 Answers3

8

There does indeed exist a group $G$ with precisely one proper subgroup $H$ of infinite order.

Preamble:

Before we proceed, let's discuss the Prüfer $p$-group. Let $p$ be a prime.

We know that for every $n\geq 1$, the cyclic group $C_{p^n}$ has precisely one subgroup isomorphic to $C_{p^{n-1}}$. Therefore, we may form an infinite chain of groups $S_0<S_1<S_2<S_3<\cdots$ where each $S_i\cong C_{p^i}$. We define the Prüfer $p$-group as: \begin{equation} C_{p^\infty}\equiv\bigcup_{i\in\mathbb{N}}S_i \end{equation} The group is also isomorphic to the group given by $\{e^{2ki\pi/p^n}:n\in\mathbb{N},0\leq k<p^n\}$ under standard complex multiplication.

It should be somewhat clear from these definitions that $C_{p^\infty}$ has no infinite proper subgroups, since any infinite subgroup $K$ must contain elements of arbitrarily high order $p^i$, and thus contain $S_i$ as a subgroup for arbitrarily high index $i$, meaning that $K=\bigcup_{i\in\mathbb{N}}S_i=C_{p^\infty}$.

Explicit construction:

Now, to solve your problem, we can use the groups $G=C_q\times C_{p^\infty}$ and $H=\{e\}\times C_{p^\infty}$ where $p,q$ are distinct primes. To see this, suppose that $K<G$ is a subgroup of infinite order. There are two cases:

Case $1$: $K$ has no element of the form $(g,h)$ where $g$ is a generator of $C_q$, then it is clear that $K\leq H$, but since $H$ cannot have infinite proper subgroups, then $K=H$.

Case $2$: $K$ contains an element $(g,h)$ where $g$ is a generator of $C_q$. Suppose that $h$ has order $m=p^n$, then $(g,h)^m=(g^m,e)\in K$, (note that $g^m$ is a generator of $C_q$, since $q,m$ are coprime) and thus $K$ contains as a subgroup $C_q\times \{e\}$. Now, Since $K$ is infinite, Abelian, has no elements of infinite order, and is not the product of infinitely many cyclic groups, then $K$ must have elements of arbitrarily high order; this means that for $N\geq 1$, $K$ must contain an element $(g',h')$ of order $o(g',h')\geq qN$. In fact $N\leq\frac{1}{q}o(g',h')=\frac{1}{q}o(g')o(h')\leq o(h')=o({h'}^q)$(since $p,q$ are coprime). We have that $(g',h')^q=(e,{h'}^q)\in K$. In other words, $K$ contains an element $(e,{h'}^q)\in\{e\}\times C_{p^\infty}$ of order $o(e,{h'}^q)=o({h'}^q)\geq N$ for every $N\geq 1$. This means that $K$ contains a subgroup of $\{e\}\times C_{p^\infty}$ of infinite order, but we know that the only subgroup of $\{e\}\times C_{p^\infty}$ of infinite order is itself, so $K$ must contain $\{e\}\times C_{p^\infty}$. Finally, since $K$ contains $\{e\}\times C_{p^\infty}$ and $C_q\times \{e\}$ as subgroups, then it must contain $G=C_q\times C_{p^\infty}$ as a subgroup, which means that $K=G$.

We have thus proven that $G$ has only one proper subgroup of infinite order, which is $H$, as desired.

  • 1
    For Case 2, note that if $H$ is an infinite subgroup of $A\times B$ then its projection along either $A$ or $B$ is infinite. In this case, this means that $H$ contains some $(g,h)$ for any $h$ in the second factor. Now raise to the $q$th power to obtain a generator for $\langle h\rangle$. Thus the Pruefer factor lies inside $H$ and done. – David A. Craven Oct 01 '22 at 21:11
  • 1
    That $H$ is the only proper infinite subgroup follows easily from Goursat's Lemma; if $M$ is any infinite subgroup, the projection onto $C_{p^{\infty}}$ must be infinite, hence all of $C_{p^{\infty}}$. If the projection onto $C_q$ is trivial, you have $H$; if the projection is not trivial, then it is all of $C_q$, so $M$ is a subdirect product. By Goursat's Lemma, it is the graph of an isomorphism between a quotient of $C_q$ and a quotient of $C_{p^{\infty}}$. But the only possible such isomorphism is between $C_q/C_q$ and $C_{p^{\infty}}/C_{p^{\infty}}$, so $M=G$. – Arturo Magidin Oct 01 '22 at 21:12
  • 2
    For a non-abelian example, let $H$ be a Tarski monster for some prime $p$, and let $G=C_2\times H$. The same proof as above works. – David A. Craven Oct 01 '22 at 21:14
  • 1
    @ArturoMagidin and David A. Craven. Of course you're right but I wanted to keep things elementary for the asker, who was already unaware of what a Prüfer $p$-group was. – Christian E. Ramirez Oct 01 '22 at 21:50
  • 1
    @C-RAM That applies to my invocation of Goursat's Lemma, but David's argument is elementary. – Arturo Magidin Oct 01 '22 at 21:56
  • @ArturoMagidin Fair; that is indeed simpler. I didn't quite see that while I was building the argument. – Christian E. Ramirez Oct 01 '22 at 22:01
5

Let $\kappa$ be an infinite cardinal. The existence of a group with just one proper subgroup of cardinality $\kappa$ is equivalent to the existence of a Jónsson group of cardinality $\kappa$, that is, a group of cardinality $\kappa$ with no proper subgroup of cardinality $\kappa$. If a group $G$ has just one proper subgroup of cardinality $\kappa$, then that subgroup is a Jónsson group; conversely, if $H$ is a Jónsson group of cardinality $\kappa$, then the group $G=H\times(\mathbb Z/2\mathbb Z)$ has just one proper subgroup of cardinality $\kappa$.

The first example of a Jónsson group of cardinality $\aleph_1$ was constructed by Saharon Shelah. Jónsson groups of cardinality $\aleph_n$ are known to exist for all finite $n$, and it is consistent with ZFC that there are Jónsson groups of all infinite cardinalities; on the other hand, there is no Jónsson group of cardinality $\kappa$ if $\kappa$ is a Rowbottom cardinal. See Samuel M. Corson, Jónsson groups of various cardinalities, Proc. Amer. Math. Soc. 150 (2022), 2771-2775, arXiv:2101.11035, for a discussion of the possible cardinalities of Jónsson groups.

bof
  • 78,265
4

(Edit: this addresses the question as stated in the body, where $K = G$ is allowed.)

For infinite $H$ this is impossible. If $H$ is infinite then its cardinality is unaffected by adjoining an additional element, so $H$ always has the same cardinality as the subgroup generated by $H$ and some $g \in G \setminus H$. Since $H$ is by hypothesis a proper subgroup such a $g$ exists and gives another subgroup $K$ of the same cardinality as $H$ but by hypothesis not equal to it.

Qiaochu Yuan
  • 419,620
  • 3
    The body of the post allows $K=G$, but the subject line suggests $K$ should be proper, and here your subgroup could be all of $G$. – Arturo Magidin Oct 01 '22 at 17:23
  • I was wondering why $\langle H,g \rangle$ necessarily has the same cardinality as $H$, but https://math.stackexchange.com/q/2191267/806703 seems to answer it (though I do not understsand it). I modified my question to include both possibilities of $K$. – durianice Oct 02 '22 at 04:22