35
$\begingroup$

It seems that often in using counting arguments to show that a group of a given order cannot be simple, it is shown that the group must have at least $n_p(p^n-1)$ elements, where $n_p$ is the number of Sylow $p$-subgroups.

It is explained that the reason this is the case is because distinct Sylow $p$-subgroups intersect only at the identity, which somehow follows from Lagrange's Theorem.

I cannot see why this is true.

Can anyone quicker than I tell me why? I know it's probably very obvious.

Note: This isn't a homework question, so if the answer is obvious I'd really just appreciate knowing why.

Thanks!

$\endgroup$
2
  • 1
    $\begingroup$ Yeah, this is definitely not true in general, and now that I look back at what I've been studying, this trick is only used in the case for a subgroup of order p. Thanks everybody! $\endgroup$ Commented Mar 2, 2011 at 23:28
  • 3
    $\begingroup$ There are some groups where the Sylow p-subgroups intersect trivially even though they have order greater than p. These groups are very special and somewhat important. For p=2, Suzuki discovered exotic new simple groups by looking at such a configuration. $\endgroup$ Commented Mar 3, 2011 at 1:27

4 Answers 4

20
$\begingroup$

Suppose $P$ and $Q$ are Sylow p-subgroups of prime order p (so not just any power of p; as others remarked, then it is not true in general). Note that $P\cap Q$ is a subgroup of $P$ (and of $Q$). So by Lagrange, the order $|P\cap Q|$ divides p. As p is prime, it is 1 or p. But it cannot be p, as $P$ and $Q$ are distinct. So $|P\cap Q|=1$ and consequently the intersection is trivial.

$\endgroup$
4
  • 1
    $\begingroup$ But the order of $P$ and $Q$ may be $p^n$ for some $n$, right? $\endgroup$ Commented Mar 2, 2011 at 23:22
  • 4
    $\begingroup$ but if the p group has order $p^n, n\geq2$ this doesnt work and in fact the claim in the question isnt true. $\endgroup$
    – yoyo
    Commented Mar 2, 2011 at 23:23
  • $\begingroup$ Okay thankyou, that clears things up then! $\endgroup$ Commented Mar 2, 2011 at 23:25
  • 1
    $\begingroup$ Interesting that a false statement got more than 20 votes. $\endgroup$ Commented Apr 8 at 11:41
16
$\begingroup$

That's because it is not true in general. Look at $2$-Sylows in $S_5$: they have nontrivial intersection.

$\endgroup$
8
$\begingroup$

In some situations, to prove that groups of order $n$ cannot be simple, you can use the counting argument if all Sylow subgroups have trivial intersection, and a different argument otherwise.

For example let $G$ be a simple group of order $n=144 = 16 \times 9$. The number $n_3$ of Sylow 3-subgroups is 1, 4 or 16. If $n_3 = 1$ then there is normal Sylow subgroup and if $n_3= 4$ then $G$ maps nontrivially to $S_4$, so we must have $n_3 = 16$.

If all pairs of Sylow 3-subgroups have trivial intersection, then they contain in total $16 \times 8$ non-identity elements, so the remaining 16 elements must form a unique and hence normal Sylow 2-subgroup of $G$.

Otherwise two Sylow 3-subgroups intersect in a subgroup $T$ of order 3. Then the normalizer $N_G(T)$ of $T$ in $G$ contains both of these Sylow 3-subgroups, so by Sylow's theorem it has at least 4 Sylow 3-subgroups, and hence has order at least 36, so $|G:N_G(T)| \le 4$ and $G$ cannot be simple.

$\endgroup$
2
$\begingroup$

It simply isn't true, Sylow p-subgroups can very well intersect non-trivially, Plop gave an example thereof.

Well, it seems like you actually cannot say the following, see comments. I'm just leaving it here as a mistake one shouldn't make, so I won't mind if a moderator deletes it since it's not an actual answer.

[wrong]You could say that the number of elements of order $p$ is at least $n_p(p^n-p^{n-1})+p^{n-1}$, which is the case when all $p$-groups intersect maximally. [/wrong] Note that in this case however the intersection of all Sylow $p$-subgroups is a normal subgroup (even a characteristic subgroup, it is called $\mathbf O_p(G)$), so this cannot occur in a simple group.

$\endgroup$
8
  • 1
    $\begingroup$ S5's Sylow 2-subgroups actually overlap a lot. The union of the Sylow 2-subgroups of S5 has 1+10+15+30=56 elements, but there are 15 Sylow 2-subgroups of order 8. 15*(8-4)+4=64 is actually larger than 56, the number of 2-elements. $\endgroup$ Commented Mar 3, 2011 at 1:19
  • $\begingroup$ This being a comment to my answer, I'm presuming you mean the non-trivial intersection of those sylow 2-subgroups is in contradiction with myself saying this could indeed be the case? $\endgroup$
    – Myself
    Commented Mar 3, 2011 at 1:22
  • $\begingroup$ Hehe, I'm not sure I can parse that. I just mean in the first sentence of the second paragraph: a group need not have as many elements of order (a power of) p as you said, because Sylow p-subgroups can overlap like crazy. Your number is 64, but the real number is only 56. $\endgroup$ Commented Mar 3, 2011 at 1:29
  • 1
    $\begingroup$ Indeed, theorem 1.38 there states that if $D = S\cap T$ is the minimal such intersection of two Sylow p-subgroups, then then $\mathbf O_p(G)$ is the unique largest subgroup of $D$ that is normal in both $S$ and $T$. And there are other fascinating theorems in that wonderful book about the world of Sylow-intersections that turns out to be even more interesting than I believed. $\endgroup$
    – Myself
    Commented Mar 3, 2011 at 1:55
  • 1
    $\begingroup$ I found a family of finite groups with q³ Sylow 2-subgroups but only 3q²+1 elements of order a power of 2, so it looks like the ratio of p-elements to Sylow p-subgroups can be arbitrarily small ((p²−1)/q + 1/q³). $\endgroup$ Commented Mar 3, 2011 at 19:17

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .