FINITE GROUPS CAN BE GENERATED BY A p-SUBGROUP

AND A p′ -SUBGROUP

THOMAS BREUER AND ROBERT M. GURALNICK

Abstract. Answering a question of Dan Haran and generalizing some results of Aschbacher-
Guralnick and Suzuki, we prove that given a prime p, any finite group can be generated by
a Sylow p-subgroup and a p′ -subgroup.
arXiv:2103.17216v1 [math.GR] 31 Mar 2021

1. Introduction

Dan Haran recently asked whether, given a prime p, it is true that every finite group can
be generated by a p-group and a p′ -group. In this note, we prove this and a bit more. Recall
that a subgroup H of G is called intravariant if for any automorphism a of G, a(H) and H
are conjugate in G. There are many examples of such subgroups (including Sylow subgroups).
See [8] for many results in that direction for simple groups.
Our main result is the following.
Theorem 1.1. Let p be a prime and G a finite group. Then there exists a Sylow p-subgroup
P of G and an intravariant p′ -subgroup R such that G = hP, Ri.

In the case p = 2, this gives the following:

Corollary 1.2. Let G be a finite group. Then G can be generated by a Sylow 2-subgroup and
an intravariant (solvable) group of odd order.

The solvability of the odd order subgroup depends on the Odd Order Theorem of Feit and
Thompson [5].
In [1, Thm. A], it is shown that every finite group can be generated by two conjugate
solvable groups and so by a cyclic subgroup and a solvable group. This was generalized in
[11] by noting that we can take the solvable group to be intravariant. We note that the proof
given here would give another proof of the Aschbacher-Guralnick and Suzuki results.
A complementary result to the previous corollary is obtained in [7, Cor. 2]. The conclusion
that the subgroup is intravariant was not stated but the proof given in the last section using
the result that any nonabelian finite simple group can be generated by a solvable group and
an involution yields that conclusion.
Corollary 1.3. Let G be a finite group. Then G can be generated by a Sylow 2-subgroup and
an intravariant solvable subgroup containing a Sylow 2-subgroup.

It was also noted in [1] that not every finite group can be generated by two nilpotent
groups (and this was generalized in [4] to show that not every finite group can be generated
by m nilpotent subgroups for any fixed m).
One can translate the result to field extensions.

Date: April 1, 2021.

The first author was funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foun-
dation) – Project-ID 286237555 – TRR 195. The second author was partially supported by the NSF grant
DMS-1901595 and a Simons Foundation Fellowship 609771. We thank Dan Haran for the question and for
useful comments.
1
2 THOMAS BREUER AND ROBERT M. GURALNICK

Corollary 1.4. Let p be a prime. Let K/F be a finite Galois extension. Then there exist
subfields K1 and K2 with F = K1 ∩ K2 such that [K : K1 ] = pa for some a ≥ 0 and [K : K2 ]
is prime to p. Moreover, K is the compositum of K1 and K2 .

There are obvious notions of Sylow subgroups, p-subgroups and p′ -subgroups for profinite
groups. An easy consequence of the result for finite groups using a standard compactness
argument is:
Corollary 1.5. Let p be a prime and let G be a profinite group. If P is a Sylow p-subgroup
of G, then exists a p′ -subgroup H of G such that G = hP, Hi.

The proof requires a somewhat stronger result for simple groups. The classification of
finite simple groups is required for our proof of this.
Theorem 1.6. Let S be a finite nonabelian simple group. There exists a prime r (depending
on S) with r dividing |S| such that given any prime t dividing |S|, S can be generated by a
Sylow r-subgroup and a Sylow t-subgroup.

In the theorem, it seems very likely that one can choose r = 2 in all cases. In order to
simplify the proof, we do not insist on this. Note that the theorem holds for quasisimple
groups as well or indeed for any group G such that G/Φ(G) is simple (here Φ(G) is the
Frattini subgroup of G).
In the next section, we prove Theorem 1.6. In the final section, we deduce Theorem 1.1
from the result on simple groups.

2. Simple Groups

Proposition 2.1. Let S = Altn , n ≥ 5. If t ≤ n is prime, then S can be generated by a

Sylow 2-subgroup and a Sylow t-subgroup.

Proof. For n ≤ 8, this is an easy exercise. Suppose that n > 8 and is odd. Let R be a Sylow
2-subgroup of S. Note that the orbits of R have distinct sizes and√ there are less than log2 n
orbits. Choose x a t-element so that x has a cycle of length ta ≥ n for t odd and at least
n/2 if t = 2. In both cases, the size of the orbit is greater than the number of orbits of R.
Thus, we can choose a conjugate of x so that an orbit of x intersects every orbit of R. Then
J := hR, xi is transitive. Since n is odd and R ≤ J, J is primitive. Since n > 8, the only
primitive subgroup of Altn containing an element moving 4 points is Altn [12, 13.5], whence
the result.
Suppose that n > 9 is even. By the odd case, Altn−1 = hR0 , T i for some t-subgroup T
and Sylow 2-subgroup R0 . Let R be a Sylow 2-subgroup of S properly containing R0 . Then
hR, T i properly contains hR0 , T i = Altn−1 , whence hR, T i = S. 
For the remaining simple groups, we can prove a stronger generation result. The analogous
result for alternating groups fails (for example, consider Alt15 with x a 3-cycle or an involution
moving only four points). Note that the next result also holds for quasisimple groups as long
as x is noncentral.
Proposition 2.2. Let S be a sporadic simple group and let P be a Sylow 2-subgroup of S.
If 1 6= x ∈ S, then S = hP, xg i for some g ∈ S.
Proof. Let S be a sporadic simple group, fix a Sylow 2-subgroup P of S, and let x be a
nonidentity element in S. We use known information about maximal subgroups of S to show
that xS is not a subset of the union of those maximal subgroups in S that contain P .
Let M be a maximal subgroup of S with the property P ≤ M . The number of S-conjugates
of M that contain P is equal to |NS (P )|/|NM (P )| ≤ [NS (P ) : P ], thus these subgroups can
contain at most [NS (P ) : P ]|xS ∩ M | elements from the class xS .
 GENERATION OF FINITE GROUPS 3

Thus the number of elements in xSPthat generate a proper subgroup of S together with P is
bounded from above by [NS (P ) : P ] M |xS ∩M |, where the sum is taken over representatives
M of conjugacy classes of maximal subgroups of odd index in S.
Let 1SM denote the permutation character of S on the cosets of M . We have |xS ∩ M | =
|x |1SM (x)/1SM (1). Hence we are done when we show that
S
X
[NS (P ) : P ] 1SM (x)/1SM (1) < 1 (∗)
M

holds.
The numbers [NS (P ) : P ] can be read off from [13, Table I]. For all sporadic simple
groups S except the Monster group, the permutation characters 1SM can be computed from
the data about maximal subgroups contained in the library of character tables [2] of the GAP
computer algebra system [6]. The Monster group is known to contain exactly five classes of
+
maximal subgroups of odd index, of the structures 21+24 .Co1 , 210+16 .O10 (2), 22+11+22 .(M24 ×
S3 ), 25+10+20 39
.(S3 ×L5 (2)), [2 ].(L3 (2)×3S6 ), and the corresponding permutation characters
are known, see [3].
We are lucky, the crude upper bound from (∗) is smaller than 1 for each S and x in
question. (In fact, the left hand side is much smaller than needed; the maximum of the left
hand side is 3/5, it is attained for x in the class 2B of the group J2 .) 

Finally we turn to the case of finite groups of Lie type. We will typically work with the
simply connected groups and rather than use the prime 2 as the base prime (which almost
certainly would work), we use the prime p that is the characteristic of the group. It is proved
in [7, Thm. A] that any nonabelian finite simple group can be generated by an involution
and a Sylow 2-subgroup.
Proposition 2.3. Let S be a finite simply connected quasisimple group of Lie type in char-
acteristic p. Let P be a Sylow p-subgroup of G. If x ∈ S is a noncentral element of S, then
S = hP, xg i for some g ∈ S.

Proof. By a lemma of Tits [10, 1.6], any maximal subgroup containing P is a parabolic
subgroup and there is precisely one parabolic subgroup containing P in each conjugacy class
of maximal parabolic subgroups.
If S has twisted rank 1, then the Borel subgroup is the unique maximal subgroup contain-
ing T and so clearly no noncentral conjugacy class is contained the Borel, whence the result
holds. So we may assume that the rank is least 2.
First suppose that S is a classical group. If S = SLn (q), then choose a basis for the natural
module v1 , . . . , vn and assume that T stabilizes the flag associated with this ordered basis.
Since S acts 2-transitively on 1-dimensional spaces, given any noncentral element x ∈ S, we
can choose g ∈ S so that g−1 xgv1 is not in the span of v1 , . . . , vn−1 and so xg is not in any
parabolic subgroup containing T .
Now suppose S is any other classical group with natural module V . Since we are assuming
that S has twisted rank at least 2, dim V ≥ 4.
First suppose that S is not an even dimensional spin group in dimension at least 8. Fix T
a Sylow p-subgroup. Then the maximal parabolic subgroups correspond to totally singular
spaces of distinct dimensions. Let v1 be a nonzero element of the 1-dimensional fixed space
of T . If n is the rank of S, then there is an ordered linearly independent set, v1 , . . . , vn so
that any maximal parabolic subgroup containing T is the stabilizer of the totally singular
space Vi = hv1 , . . . , vi i. Let x be a noncentral element of S. Suppose that xv1 = w. If w and
v1 are linearly dependent, then just choose g ∈ S so that xg maps v1 to element not in Vn .
If w is not in Vn , then already x is not in any of the parabolic subgroups containing T . The
last case is if hv1 , wi is a 2-dimensional totally singular subspace of Vn . Then there exists
4 THOMAS BREUER AND ROBERT M. GURALNICK

g ∈ S fixing v1 and mapping w to any singular vector in v1⊥ and since v1⊥ is not contained in
Vn (since dim V ≥ 4), the result holds in this case.
Next suppose that S is an even dimensional spin group of rank n. We work in the orthog-
onal module of dimension 2n. In this case, choose a linearly independent set of basis vectors
v1 , . . . , vn+1 so that the maximal parabolic subgroups containing T are the stabilizer of the
totally singular subspace Vi = hv1 , . . . vi i, i = 1, . . . , n − 2 of dimension i and the two totally
singular subspaces hv1 , . . . , vn−1 , vn i and hv1 , . . . , vn−1 , vn+1 i. Now argue precisely as in the
previous case.
Finally suppose that S is an exceptional group of twisted rank n. Let P1 , . . . , Pn be the
distinct maximal parabolic subgroups containing T . Let C = xS for x a noncentral element
of S. It follows by [9, Thm. 1] that
X
n
|C ∩ Pi |
< 1.
|C|
i=1

Thus, C 6= ∪ni=1 (C ∩ Pi ) and the result follows. 

It remains to consider the special cases where S is the simply connected group but S/Z(S)
is not simple. If S is solvable, then we can ignore the case. This leaves the groups Sp4 (2),
G2 (2), 2 G2 (3) and 2 F4 (2). The first three cases have socle Alt6 , PSU3 (3) and PSL2 (8) and
we have already proved the result for these groups. In the final case, the socle is the derived
subgroup of index 2 but precisely the same argument as above applies (i.e. there are only two
maximal subgroups containing a Sylow 2-subgroup which are parabolic subgroups intersected
with the derived subgroup). This completes the proof of Theorem 1.6.

3. Proof of the main theorem

We now prove Theorem 1.1. Fix a prime p and let G be a counterexample of minimal order.

Step 1. Op (G) = 1.

If not, let 1 6= A be a minimal characteristic subgroup of G with A an elementary abelian

p-group. By minimality, G/A = hP/A, R/Ai where P is a Sylow p-subgroup of G and R is a
subgroup of G containing A with R/A an intravariant p′ -group of G/A.
Since gcd(R|/|A|, |A|) = 1, it follows that H 2 (R/A, A) = 0 and so R = AR1 with R1 a
complement to A in R. In particular, R1 is a p′ -subgroup. Suppose a is an automorphism of
G. Then by conjugating by an element of G, we may assume that a(R) = R. Thus, a(R1 ) is a
complement ot A in R. Since H 1 (R1 , A) = 0, all complements to A in R are conjugate via an
element of A and so a(R1 ) is conjugate to R1 , whence R1 is intravariant. Clearly, G = hP, R1 i.

Step 2. Op′ (G) = 1.

If not, let A be a minimal characteristic subgroup of G with A a p′ -subgroup. By minimal-

ity G/A = hQ/A, R/Ai where Q/A is a Sylow p-subgroup of G/A and R/A is intravariant
p′ -subgroup of G/A. Since A is characteristic, R is intravariant and clearly a p′ -subgroup.
Clearly Q = AP where P is a Sylow p-subgroup of Q (and so of G). Then G = hP, Ri and
the result follows.

Step 3. Completion of the proof.

 GENERATION OF FINITE GROUPS 5

Let A be a minimal characteristic subgroup of G. By Steps 1 and 2, A = L × . . . × L where

L is a nonabelian simple group of order divisible by p. By Theorem 1.6, A = hQ, M i where
Q is a Sylow p-subgroup of A and M is a Sylow r-subgroup of A for some r 6= p.
Let H = NG (M ). Then H is a proper subgroup of G and so H = hD, Ri where D is a
Sylow p-subgroup of H and R is an intravariant p′ -subgroup of H. Since R normalizes M , we
assume that R ≥ M . Note that H is the normalizer of a Sylow subgroup of a characteristic
subgroup of G and so H is intravariant and so R is also intravariant.
We can enlarge D and assume that D contains a Sylow p-subgroup Q1 of A. Then Qa1 = Q
for some a ∈ A. Then D a ≥ Q.
We claim that G = J := hD a , Ri. Note that J ≥ hQ, M i = A. Since a ∈ A, we see that
DA = D a A and so J ≥ hA, D, Ri ≥ hA, Hi. By the Frattini argument, G = HA and so
J = G as claimed. This completes the proof.

References
[1] M. Aschbacher and R. Guralnick, Solvable generation of groups and Sylow subgroups of the lower
central series, J. Algebra 77 (1982), 189–201.
[2] T. Breuer, CTblLib, The GAP Character Table Library, Version 1.3.1 (2020), (GAP package),
http://www.math.rwth-aachen.de/~Thomas.Breuer/ctbllib.
[3] T. Breuer, Four Primitive Permutation Characters of the Monster Group,
http://www.math.rwth-aachen.de/~Thomas.Breuer/ctbllib/doc2/chap0.html.
[4] J. Cossey and T. Hawkes, On generating a finite group by nilpotent subgroups, J. Pure Appl. Algebra
97 (1994), 275–280.
[5] W. Feit and J. G. Thompson, Solvability of groups of odd order, Pacific J. Math. 13 (1963), 775–1029.
[6] GAP – Groups, Algorithms, and Programming, Version 4.11.1, The GAP Group (2021),
https://www.gap-system.org.
[7] R. Guralnick, Generation of simple groups, J. Algebra 103 (1986), 381–401.
[8] R. Guralnick and D. Haran, Frobenius subgroups of free profinite products, Bull. Lond. Math. Soc.
43 (2011), 467–477.
[9] R. Lawther, M. Liebeck and G. Seitz, Fixed point ratios in actions of finite exceptional groups of Lie
type, Pacific J. Math. 205 (2002), 393–464.
[10] G. M. Seitz, Flag-transitive subgroups of Chevalley groups, Ann. of Math. 97 (1973), 27–56.
[11] M. Suzuki, Solvable generation of finite groups, Hokkaido Math. J. 16 (1987), 109–113.
[12] H. Wielandt, Finite permutation groups, Translated from the German by R. Bercov, Academic Press,
New York-London 1964 x+114 pp.
[13] R. A. Wilson, The McKay conjecture is true for the sporadic simple groups, J. Algebra 207(1) (1998),
294–305.

Lehrstuhl für Algebra und Zahlentheorie, RWTH Aachen, Germany

Email address: sam@math.rwth-aachen.de

R.M. Guralnick, Department of Mathematics, University of Southern California, Los An-

geles, CA 90089-2532, USA
Email address: guralnic@usc.edu