10

Prove that there exists no simple group of order 576.

Suppose $G$ is simple of order 576. It is a straightforward application of the index theorem to determine that the number of Sylow 2-subgroups of $G$ is 9. Let $G$ act on the set of Sylow 2-subgroups by conjugation. The size of the stabilizers on one element is 576/9 = 64 each.

Observation 1. Each element of the stabilizer on one element has order a power of 2; considering that $G$ has an injective image in $A_{9}$, we narrow down the cases to order 8 (8-cycle is odd, eliminated), order 4 (2 4-cycles or 1 4-cycle and 1 2-cycle), or order 2 (2 2-cycles or 4 2-cycles).

Observation 2. The stabilizers are Sylow 2-subgroups themselves, thus each Sylow 2-subgroup is equal to its normalizer. By a counting argument, the intersection of two Sylow 2-subgroups must be greater than 64*64/576, thus is either 8 or 16. 32 is eliminated since then the normalizer of the intersection is of index at most 3 in $G$.

How to proceed from here?

Clearly, combining the two observations, we see that the intersection of two stabilizers has at least 8 elements, thus there exists a 2 2-cycle type element (the power of a 1 4-cycle and 1 2-cycle element is a 2 2-cycle) in the image of $G$, which implies that the intersection of 5 normalizers is non-trivial. I don't see how to lead this argument into a contradiction, though.

In the case that this approach fails, any other solutions to the problem are also appreciated.

Edit: In response to the below comment by Dietrich Burde, I would like to solve the problem using minimal tools; preferably no transfer theory, Burnside's pq theorem, or classification of finite simple groups (obviously).

Mean X
  • 245
  • 1
    If "any other solution" is acceptable, then $576=2^6\cdot 3^2$ implies that every such group is solvable by Burnside, and hence not (nonabelian) simple. For elemenatry proofs see for example here, replacing $5$ by $2$. – Dietrich Burde Feb 12 '20 at 12:06
  • 1
    I should have probably phrased my question better, but Burnside's paqb is a bit overpowered for this, don't you think? Also, if we replace 5 by 2 in the link you gave, we can no longer say $n_{2} = 1$, which was the key-point of the proof given there. – Mean X Feb 12 '20 at 12:57
  • Yes, you are right. On the other hand this site offers more than one elementary proof for several cases of groups of order $p^nq^m$ for smaller $n$ or $m$. Some more general results would be useful, too, given that you also asked about groups of other orders (may be as another user, too). – Dietrich Burde Feb 12 '20 at 13:06
  • I'll search around, of course (I can't seem to find any proofs of specific cases for Burnside's paqb, I would appreciate more links), and try other methods, but the only filters I allow myself at this point are the basic pq/pqr, Sylow 3rd, and index theorem. I don't feel the need to apply further tools for groups of order less than 1000, except for maybe 720. – Mean X Feb 12 '20 at 13:17
  • I haven't searched really, but I just saw this post for general results without using full Burnside. I am sure there are more posts around. Have you looked into Isaacs' book? – Dietrich Burde Feb 12 '20 at 13:23
  • I haven't gone through the full length of the book, but I didn't see any proofs for specific cases, and the assertion before the proof of Burnside remarks that the proofs for the case p, q both odd came before full Burnside, which made me think the proof for p or q = 2 is more difficult. The proof of full Burnside takes 7 steps just to eliminate the case p or q = 2, and it utilizes results of the previous six chapters; not really elementary in my book. – Mean X Feb 12 '20 at 13:34
  • What is your aim? Just a collection of examples of non-simple groups by using only Sylow or only very elementary means? If you want to test all groups of order less than $1000$, it seems worthwhile to read Burnside and other more complicated results. This will pay off quickly. At least I don't see a big motivation to avoid some nice group theory. – Dietrich Burde Feb 12 '20 at 13:37
  • By the way, I found this MO-post for the case $n=720$. This seems to fit into our discussion. – Dietrich Burde Feb 12 '20 at 13:46
  • You're probably right in the regard in that if my aim was to classify simple groups under order 1000, it would be nicer to use stronger tools, but I'm just an undergraduate trying to use undergraduate results. While stronger results are of interest to me, I would like to solve as much as I can using only the tools I would have been given. I did see the MO-post on disproving the existence of simple groups of order 720, but I wanted to hold on to breaking down the proofs until I have actually attempted the problem myself. – Mean X Feb 12 '20 at 13:59
  • May be you can use the following: The order of $A_9$ is $2^6\cdot3^4\cdot5\cdot7$, so the images of Sylow $2$-subgroups of $G$ in $A_9$ also happen to be Sylow $2$-subgroups of $A_9$, hence of $A_8$. The structure of Sylow $2$-subs of $S_8$ is "easy" to describe, see here for a local description. What I'm getting at is that we know the structure of a Sylow-$2$ of $G$ completely. I'm hoping that there might be something there... – Jyrki Lahtonen Feb 12 '20 at 14:35
  • Barring a mistake (likely, as I may be needing a nap shortly) it follows that the center of $P$, a Sylow $2$, is cyclic of order two and contains a product four disjoint $2$-cycles. These must form a conjugacy class of size $9$. Don't know if we run into some kind of difficulties when trying to find a corresponding $3$-subgroup of $A_9$? – Jyrki Lahtonen Feb 12 '20 at 15:05
  • I got to the point that the Sylow 2-subgroup of $A_{8}$ (hence $G$) is isomorphic to $UT(4, 2)$ and thus has center order 2, and the non-identity elements of each Sylow 2 form a conjugacy class of $G$; injection into $A_{9}$. The non-stabilizer elements which are either 3 3-cycles or a 9-cycle are both even. I don't see where we could run into further difficulties here; in the first place, I don't see how this action could yield different results than the original action we had on the 9 Sylow 2's. – Mean X Feb 12 '20 at 16:06
  • MeanX, do you happen to know whether $A_9$ has a subgroup of order $576$? – Jyrki Lahtonen Feb 12 '20 at 17:02
  • Sorry, no. I don't even have any clue on how to prove or disprove the existence of one. – Mean X Feb 12 '20 at 18:04
  • I don't see much coming out of it either. I think it does imply that the pairwise intersections of Sylow-2s all have order eight (isomorphic to the dihedral group). – Jyrki Lahtonen Feb 12 '20 at 19:38
  • Above "it" = "the known structure of Sylow-2s of $A_8$." – Jyrki Lahtonen Feb 12 '20 at 19:47
  • 1
    @JyrkiLahtonen: It turns out that any proper subgroup of $A_9$ containing a $2$-Sylow has a fixed point, i.e. is conjugate to a subgroup of $A_8$. I am not sure if there is an easy way to see this. But at least this has the consequence that if $G$ has $9$ Sylow $2$-subgroups of order $2^6$, then the conjugation action on these subgroups cannot be faithful. (And so $G$ cannot be simple.) – Mikko Korhonen Feb 15 '20 at 15:07
  • Thanks, @Mikko. – Jyrki Lahtonen Feb 15 '20 at 16:40

2 Answers2

9

Let $G$ be simple of order $576 = 64 \times 9$. The number of Sylow $2$-subgroups is 1,3 or 9 by Sylow's Theorem, but $G$ simple implies that it cannot be 1 or 3, so it must be 9. Then the conjugation action of $G$ on the set $\Omega$ of Sylow $2$-subgroups of $G$ induces an embedding $G \to A_9$.

Let $S \in {\rm Syl_2(G)}$. Now a Sylow $2$-sugroup of $A_9$ has order $64 = |S|$ so in its action on $\Omega$, $S$ is a Sylow $2$-subgroup of ${\rm Alt}(\Omega)$ and, in particular $S$ is transitive on $\Omega \setminus \{ S \}$, and hence the action of $G$ on $\Omega$ is 2-transitive.

A 2-point stabilizer in this action is a subgroup $T$ of $S$ of order $8$. Since $T$ fixes more than one point of $\Omega$, it is contained in more than one Sylow $2$-subgroup of $G$, so its normalizer $N_G(T)$ in $G$ has more than one Sylow $2$-subgroup.

Since $N_G(T) \ne G$, $N_G(T)$ cannot have 9 Sylow $2$-subgroups, so it must have 3. Since $T$ is properly contained in $N_S(T)$, we have $|N_S(T) \ge 16$, so $|N_G(T)| = 2^k \times 3$ where $4 \le k \le 6$, and hence $|G:N_G(T)| = 2^{6-k} \times 3 = 3, 6$, or $12$. From its order $G$ is not contained in$A_6$, so we must have $k=4$ and $|N_G(T)|=48$.

To proceed further we can use the fact that $T$ is isomorphic to a Sylow $2$-subgroup of $A_6$. We can identify $T$ with $\langle (4,5)(6,7), (6,7)(8,9), (4,8)(7,9)\rangle$, which is dihedral of order $8$ (the group $D_8$) and has centre $U = \langle (4,5)(8,9) \rangle$ of order $2$.

Now the normalizer of $U$ in $S_8$ contains $D_8 \times D_8$ of order $64$, so its normalizer in $A_8$ and hence in $S$ has order (at least) 32.

Since $N_G(U)$ contains $N_G(T)$ and $|N_S(U)| \ge 32$, we have $|N_G(U)| \ge 32 \times 3$, so $|G: N_G(U)| \le 6$, which is impossible because $G$ does not embed in $A_6$.

Derek Holt
  • 96,726
  • I wonder if it would be possible to find an elementary proof of the fact that a group of order $p^nq^2$ is not simple (elementary in the sense that it does not involve characters/representations and does not follow the Goldschmidt-Bender-Matsuyama argument). – the_fox Mar 02 '20 at 13:39
  • Using the normalizer of $U$ is a nice trick (and something I could follow). I'm probably missing something simple, but why can't $N_G(T)$ have order, say, $16\cdot9$, and then possibly also $9$ Sylow $2$-subgroups? – Jyrki Lahtonen Mar 02 '20 at 20:39
  • Never mind, $G$ would then have a subgroup of index four, and couldn't therefore be simple. – Jyrki Lahtonen Mar 02 '20 at 20:40
3

By Sylow, the number of Sylow $3$-subgroups of $G$ is either $16$ or $64$, as $G$ is simple. Let a Sylow $3$-subgroup $S$ of $G$ act on the set $Syl_3(G)$ of Sylow $3$-subgroups by conjugation. If $g\in S$ fixes a point $T\in Syl_3(G)$, then $\langle g, T\rangle$ is a $3$-group, so $g\in T$. So $S$ has a unique orbit $\{S\}$ of length $1$. The other orbits have length $3$ or $9$.

If all orbits have length $9$, then $S$ intersects all other Sylow $3$-subgroups $T$ trivially ($S\cap T = 1$), and as the Sylow $3$-subgroups are conjugated, any two Sylow $3$-subgroups have trivial intersection. The number of Sylow $3$-subgroups being $1 \bmod 9$ is $64$, and the union of all Sylow $3$-subgroups consists of $8\cdot 64 = 512$ elements. The set of the remaining $64$ elements of $G$ is then the (unique) Sylow $2$-subgroup of $G$, contradicting the simplicity of $G$.

Hence $S$ has an orbit of length $3$, and for $T$ in this orbit $U := S\cap T$ has order $3$. As the Sylow $3$-subgroups are abelian (being of order prime squared), the centralizer $C := C_G(U)$ contains both Sylow $3$-subgroups $S$ and $T$, so by Sylow, the number of Sylow $3$-subgroups of $C$ is a multiple of $4$. As $G$ is simple, the order of $C$ is either $36$ or $72$.

If the order of $C$ is $36$, then $C/U$ is isomorphic to $A_4$ (the only group of order $12$ without normal Sylow $3$-subgroup). A Sylow $2$-subgroup $V$ of $C$ centralizes $U$ and its image $\bmod U$ is normal in $C/U$, so $V$ is normal in $C$. Its normalizer $N:=N_G(V)$ contains $C$, but also elements in $S\setminus V$, as a proper subgroup of a (finite) $p$-group is a proper subgroups of its normalizer. As $G$ is normal, $N$ has order $72$.

In any case, there exists a subgroup $H$ of $G$ of order $72$. As $G$ is simple, its normalizer $N_G(H)$ equals $H$, i.e., $H$ is self-normalizing and $H$ has $8$ conjugates, which are also self-normalizing. Let $H$ act on the set $\Omega := \{H^g\mid g\in G\}$ of its conjugates by conjugation. If $h\in H$ fixes a point $H^g\in\Omega$, then $h\in H\cap H^g$, as $H^g$ is self-normalizing. So $H$ has a unique orbit $\{H\}$ of length $1$, and the other orbits have lengths $2, 3$ or $4$.

If $H^g$ lies in an orbit of length $2$, then $H\cap H^g$ has index $2$ in both $H$ and $H^g$, and therefore has to be normal in both. As $U$ is the maximal normal $3$-subgroup of $H\cap H^g$, it is characteristic in a normal subgroup of $H^g$, and hence normal in $H^g$. So the normalizer of $U$ contains $H$ and $H^g$ contradicting the simplicity of $G$.

As $G$ is simple, its action by conjugation on $\Omega$ is faithful, and $G$ embeds into $A_8$ such that its subgroup $H$ has a fixed point and two orbits of length $3$ and $4$. As $H$ has order $72$, $H = (S_4\times S_3)\cap A_8$, where $S_4$ and $S_3$ are the symmetric groups on $\{1,2,3,4\}$ rsp. $\{5,6,7\}$. So $H$ contains the $3$-cycles $(123), (124), (134), (234)$ and $(567)$.

As the subgroup generated by all $3$-cycles of $G$ is normal in $G$, and as $G$ is simple, $G$ has to contain another $3$-cycle $(abc)$ (not contained in $H$). The subgroup generated by two non-disjoint $3$-cycles acts transitively on the union of their orbits. As $5$ does not divide the order of $G$, $(abc)$ cannot intersect any $3$-cycle of $H$ in exactly one point, so $\{a,b,c\}\subset\{5,6,7,8\}$. But then $(567)$ and $(abc)$ generate $A_4$ on the points $\{5,6,7,8\}$, so $G$ contains $A_4\times A_4$, a subgroup of order $144$ contradicting its simplicity.

j.p.
  • 1,206