Take the 2-minute tour ×
Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

I know that if we got two finite index subgroups $U,H$ of a group $G$. Then the index of the intersection $H\cap U$ of both is smaller than the product of the indices of $H,U$. We proof this with an injective map from the cosets of $H\cap U$ in H to the cosets of $U$ in $G$. But is there a chance to show that this map is surjective in some cases? Especially in the case:

$U\neq H< G$ infinite groups, $p$ prime number. $|G:U|=p=|G:H|$. Then $|G:H\cap U|= p^2$?

Or when the indices are coprime?

Or can we show that the index of the intersection $H\cap U$ in $G$ divides the product of the indices of $H$ and $U$ in $G$.

If we can't do this for infinite groups. Do anyone have an example where $|G:H\cap U|$ does not divide $|G:U|\cdot|G:H| $?

Thanks for help.

share|improve this question
    
Since $H$ and $U$ are assumed to have finite index, $H\cap U$ has finite index, and so contains a normal subgroup of $G$ of finite index. Moding out by this subgroup you can reduce to the case of $G$ finite. –  Arturo Magidin May 28 '12 at 13:04

2 Answers 2

up vote 2 down vote accepted

Finiteness or infiniteness is irrelevant; a finite example yields an infinite one by taking a direct product with an infinite group. And every infinite example can be turned into a finite example by moding out by a normal subgroup of $G$ of finite index contained in $H\cap U$.

In general, if $G$ is a group, $H$ and $K$ are subgroups, then $$[H\colon H\cap K] \leq [G:K]$$ in the sense of cardinalities; but you may or may not get equality. In particular, if $[G:H]=[G:K]=p$, then $[G:H\cap K]\leq p^2$, and is a multiple of $p$, but you may or may not have equality to $p^2$.

Proposition. Let $G$ be a group, $H$ and $K$ subgroups. Then $$[H:H\cap K] \leq [G:K]$$ in the sense of cardinalities, with equality if and only if $HK=G$.

Proof. We define a map from the left cosets of $H\cap K$ in $H$ to the left cosets of $K$ in $G$ by mapping $h(H\cap K)$ to $hK$. I claim the map is well-defined and one-to-one: $$\begin{align*} h(H\cap K) = h'(H\cap K) &\iff h'^{-1}h\in H\cap K\\ &\iff h'^{-1}h\in K\quad\text{(it is always in }H\text{)}\\ &\iff hK = h'K. \end{align*}$$ The map is onto if and only if for every $g\in G$ there exists $h\in H$ such that $gK=hK$, if and only if for every $g\in G$ there exists $h\in H$, $k\in K$ such that $g=hk$, if and only if $G\subseteq HK$, if and only if $G=HK$. $\Box$

In particular, for you situation, if $HU$ is a subgroup of $G$ (it need not be!) then $HU=G$, so you will have $[G:H\cap U] = [G:H][H:H\cap U] = [G:H][G:U] = p^2$. But if $HU$ is not a subgroup, then the best you have is that $p$ divides $[G:H\cap U]$, and $p\lt [G:H\cap U] \leq p^2$; so it could be any of $2p$, $3p,\ldots,(p-1)p,p^2$.

In the example Gerry Myerson gives, for instance, the intersection has index $6=2p$ ($p=3$).

share|improve this answer
    
Thank you very much. For my field of interest, I wanted to have that the index is a power of $p$ not a multiple of $p$. But I thoughted that it won't work, too. But didn't know how to proof it. –  Peter May 28 '12 at 14:28

$S_4$ has two (in fact, three) subgroups of index 3, but their intersection is not of index 9. Yes, I know, $S_4$ is a finite group, but just do $S_4\times G$ for some infinite group, $G$.

share|improve this answer
    
Okay. But the index of the interesection divides the product of the indices of the subgroups, so it divides 9. right? –  Peter May 28 '12 at 13:20
1  
Does it? The only numbers dividing 9 are 1, 3, and 9. The index of the intersection can't be 9 cuz 9 doesn't divide the order of the group $S_4$. And it can't be 3 or less since the intersection is smaller than either subgroup. But you know it's not that hard to find those two subgroups of $S_4$ and see for yourself what the intersection is and what its index is. Try it! –  Gerry Myerson May 28 '12 at 13:24

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

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