THE SYLOW THEOREMS

KEITH CONRAD

1. Introduction
The converse of Lagrange’s theorem is false: if G is a finite group and d | |G|, then there
may not be a subgroup of G with order d. The simplest example of this is the group A4 , of
order 12, which has no subgroup of order 6. The Norwegian mathematician Peter Ludwig
Sylow [1] discovered that a converse result is true when d is a prime power: if p is a prime
number and pk | |G| then G must contain a subgroup of order pk . Sylow also discovered
important relations among the subgroups with order the largest power of p dividing |G|,
such as the fact that all subgroups of that order are conjugate to each other.
For example, a group of order 100 = 22 · 52 must contain subgroups of order 1, 2, 4, 5,
and 25, the subgroups of order 4 are conjugate to each other, and the subgroups of order
25 are conjugate to each other. It is not necessarily the case that the subgroups of order 2
are conjugate or that the subgroups of order 5 are conjugate.
Definition 1.1. Let G be a finite group and p be a prime. Any subgroup of G whose order
is the highest power of p dividing |G| is called a p-Sylow subgroup of G. A p-Sylow subgroup
for some p is called a Sylow subgroup.
In a group of order 100, a 2-Sylow subgroup has order 4, a 5-Sylow subgroup has order
25, and a p-Sylow subgroup is trivial if p 6= 2 or 5.
In a group of order 12, a 2-Sylow subgroup has order 4, a 3-Sylow subgroup has order 3,
and a p-Sylow subgroup is trivial if p > 3. Let’s look at a few examples of Sylow subgroups
in groups of order 12.
Example 1.2. In Z/(12), the only 2-Sylow subgroup is {0, 3, 6, 9} = h3i and the only
3-Sylow subgroup is {0, 4, 8} = h4i.
Example 1.3. In A4 there is one subgroup of order 4, so the only 2-Sylow subgroup is
{(1), (12)(34), (13)(24), (14)(23)} = h(12)(34), (14)(23)i.
There are four 3-Sylow subgroups:
{(1), (123), (132)} = h(123)i, {(1), (124), (142)} = h(124)i,
{(1), (134), (143)} = h(134)i, {(1), (234), (243)} = h(234)i.
Example 1.4. In D6 there are three 2-Sylow subgroups:
{1, r3 , s, r3 s} = hr3 , si, {1, r3 , rs, r4 s} = hr3 , rsi, {1, r3 , r2 s, r5 s} = hr3 , r2 si.
The only 3-Sylow subgroup of D6 is {1, r2 , r4 } = hr2 i.
In a group of order 24, a 2-Sylow subgroup has order 8 and a 3-Sylow subgroup has order
3. Let’s look at two examples.
1
2 KEITH CONRAD

Example 1.5. In S4 , the 3-Sylow subgroups are the 3-Sylow subgroups of A4 (an element
of 3-power order in S4 must be a 3-cycle, and they all lie in A4 ). We determined the 3-Sylow
subgroups of A4 in Example 1.3; there are four of them.
There are three 2-Sylow subgroups of S4 , and they are interesting to work out since they
can be understood as copies of D4 inside S4 . The number of ways to label the four vertices
of a square as 1, 2, 3, and 4 is 4! = 24, but up to rotations and reflections of the square
there are really just three different ways of carrying out the labeling, as follows.
2 3 2 4 3 2

1 4 1 3 1 4

Any other labeling of the square is a rotated or reflected version of one of these three squares.
For example, the square below is obtained from the middle square above by reflecting across
a horizontal line through the middle of the square.
1 3

2 4

When D4 acts on a square with labeled vertices, each motion of D4 creates a permutation
of the four vertices, and this permutation is an element of S4 . For example, a 90 degree
rotation of the square is a 4-cycle on the vertices. In this way we obtain a copy of D4 inside
S4 . The three essentially different labelings of the vertices of the square above embed D4
into S4 as three different subgroups of order 8:
{1, (1234), (1432), (12)(34), (13)(24), (14)(23), (13), (24)} = h(1234), (13)i,
{1, (1243), (1342), (12)(34), (13)(24), (14)(23), (14), (23)} = h(1243), (14)i,
{1, (1324), (1423), (12)(34), (13)(24), (14)(23), (12), (34)} = h(1324), (12)i.
These are the 2-Sylow subgroups of S4 .
Example 1.6. The group SL2 (Z/(3)) has order 24. An explicit tabulation of the elements
of this group reveals that there are only 8 elements in the group with 2-power order:
       
1 0 0 −1 1 1 −1 1
, , , ,
0 1 1 0 1 −1 1 1
       
−1 0 0 1 −1 −1 1 −1
, , , .
0 −1 −1 0 −1 1 −1 −1
These form the only 2-Sylow subgroup, which is isomorphic to Q8 by labeling the matrices
in the first row as 1, i, j, k and the matrices in the second row as −1, −i, −j, −k.
There are four 3-Sylow subgroups: h( 10 11 )i, h( 11 01 )i, h( 02 12 )i, and h( 01 22 )i.
 THE SYLOW THEOREMS 3

Here are the Sylow theorems. They are often given in three parts. The result we call
Sylow III* is not always stated explicitly as part of the Sylow theorems.
Theorem 1.7 (Sylow I). A finite group G has a p-Sylow subgroup for every prime p and
each p-subgroup of G lies in some p-Sylow subgroup of G.
Theorem 1.8 (Sylow II). For each prime p, the p-Sylow subgroups of G are conjugate.
Theorem 1.9 (Sylow III). For each prime p, let np be the number of p-Sylow subgroups of
G. Write |G| = pk m, where p doesn’t divide m. Then
np ≡ 1 mod p and np | m.
Theorem 1.10 (Sylow III*). For each prime p, let np be the number of p-Sylow subgroups
of G. Then np = [G : N(P )], where P is a p-Sylow subgroup and N(P ) is its normalizer.
The existence part of Sylow I has been illustrated in all the previous examples.
Sylow II says for two p-Sylow subgroups H and K of G that there is some g ∈ G such
that gHg −1 = K. This is illustrated in the table below, where Example 1.2 is skipped since
Z/(12) is abelian.
Example Group Size p H K g
1.3 A4 12 3 h(123)i h(124)i (243)
1.4 D6 12 2 hr3 , si hr3 , rsi r2
1.5 S4 24 2 h(1234), (13)i h(1243), (14)i (34)
1.6 SL2 (Z/(3)) 24 3 h( 10 11 )i h( 11 01 )i ( 02 11 )
When trying to conjugate one cyclic subgroup to another cyclic subgroup, be careful: not
all generators of the two groups have to be conjugate. For example, in A4 the subgroups
h(123)i = {(1), (123), (132)} and h(124)i = {(1), (124), (142)} are conjugate, but the
conjugacy class of (123) in A4 is {(123), (142), (134), (243)}, so there’s no way to conjugate
(123) to (124) by an element of A4 ; we must conjugate (123) to (142). The 3-cycles (123)
and (124) are conjugate in S4 , but not in A4 . Similarly, ( 10 11 ) and ( 11 01 ) are conjugate in
GL2 (Z/(3)) but not in SL2 (Z/(3)), so when Sylow II says the subgroups h( 10 11 )i and h( 11 01 )i
2
are conjugate in SL2 (Z/(3)) a conjugating matrix must send ( 10 11 ) to ( 11 01 ) = ( 12 01 ).
Let’s see what Sylow III tells us about the number of 2-Sylow and 3-Sylow subgroups
of a group of order 12. For p = 2 and p = 3 in Sylow III, the divisibility conditions are
n2 | 3 and n3 | 4 and the congruence conditions are n2 ≡ 1 mod 2 and n3 ≡ 1 mod 3. The
divisibility conditions imply n2 is 1 or 3 and n3 is 1, 2, or 4. The congruence n2 ≡ 1 mod 2
tells us nothing new (1 and 3 are both odd), but the congruence n3 ≡ 1 mod 3 rules out
the option n3 = 2. Therefore n2 is 1 or 3 and n3 is 1 or 4 when |G| = 12.
If |G| = 24 we again find n2 is 1 or 3 while n3 is 1 or 4. (For instance, from n3 | 8 and
n3 ≡ 1 mod 3 the only choices are n3 = 1 and n3 = 4.) Therefore as soon as we find more
than one 2-Sylow subgroup there must be three of them, and as soon as we find more than
one 3-Sylow subgroup there must be four of them. The table below shows the values of n2
and n3 in the examples above.
Example Group Size n2 n3
1.2 Z/(12) 12 1 1
1.3 A4 12 1 4
1.4 D6 12 3 1
1.5 S4 24 3 4
1.6 SL2 (Z/(3)) 24 1 4
4 KEITH CONRAD

2. Proof of the Sylow Theorems

Our proof of the Sylow theorems will use group actions, which we assume the reader
knows. The table below is a summary. For each theorem the table lists a group, a set it
acts on, and the action. Let Sylp (G) be the set of p-Sylow subgroups of G, so np = | Sylp (G)|.
Theorem Group Set Action
Sylow I p-subgroup H G/H left mult.
Sylow II p-Sylow subgroup Q G/P left mult.
Sylow III (np ≡ 1 mod p) P ∈ Sylp (G) Sylp (G) conjugation
Sylow III (np | m) G Sylp (G) conjugation
∗
Sylow III G Sylp (G) conjugation
The two conclusions of Sylow III are listed separately in the table since they are proved
using different group actions.
Our proofs will usually involve the action of a p-group on a set and use the fixed-point
congruence for such actions: |X| ≡ | FixΓ (X)| mod p, where X is a finite set being acted on
by a finite p-group Γ and FixΓ (X) is the fixed points of Γ in X.
Proof of Sylow I: Let pk be the highest power of p in |G|. The result is obvious if k = 0,
since the trivial subgroup is a p-Sylow subgroup, so we can take k ≥ 1, hence p | |G|.
Our strategy for proving Sylow I is to prove a stronger result: there is a subgroup of
order pi for 0 ≤ i ≤ k. More specifically, if |H| = pi and i < k, we will show there is a
p-subgroup H 0 ⊃ H with [H 0 : H] = p, so |H 0 | = pi+1 . Then, starting with H as the trivial
subgroup, we can repeat this process with H 0 in place of H to create increasingly larger
subgroups
{e} = H0 ⊂ H1 ⊂ H2 ⊂ · · ·
i
where |Hi | = p , and after k steps we reach Hk , which is a p-Sylow subgroup of G. And if
we start with H as a p-subgroup, we will have shown H is contained in a p-Sylow subgroup.
Consider the left multiplication action of H on the left cosets G/H (this need not be
a group). This is an action of a finite p-group H on the set G/H, so by the fixed-point
congruence for actions of nontrivial p-groups,
(2.1) |G/H| ≡ | FixH (G/H)| mod p.
Here is what it means for gH in G/H to be fixed by the group H acting by left multiplication:
hgH = gH for all h ∈ H ⇐⇒ hg ∈ gH for all h ∈ H
⇐⇒ g −1 hg ∈ H for all h ∈ H
⇐⇒ g −1 Hg ⊂ H
⇐⇒ g −1 Hg = H because |g −1 Hg| = |H|
⇐⇒ g ∈ N(H).
Thus FixH (G/H) = {gH : g ∈ N(H)} = N(H)/H, so (2.1) becomes
(2.2) [G : H] ≡ [N(H) : H] mod p.
Because H C N(H), N(H)/H is a group.
When |H| = pi and i < k, the index [G : H] is divisible by p, so the congruence
(2.2) implies [N(H) : H] is divisible by p, so N(H)/H is a group with order divisible by
p. Thus N(H)/H has a subgroup of order p by Cauchy’s theorem. All subgroups of the
quotient group N(H)/H have the form H 0 /H, where H 0 is a subgroup between H and
 THE SYLOW THEOREMS 5

N(H). Therefore a subgroup of order p in N(H)/H is H 0 /H such that [H 0 : H] = p, so

|H 0 | = p|H| = pi+1 . This can be repeated until we reach a subgroup of order pk , and we’re
done.
Proof of Sylow II: For p-Sylow subgroups P and Q, we want to show they are conjugate.
Consider the action of Q on G/P by left multiplication. Since Q is a finite p-group,
|G/P | ≡ | FixQ (G/P )| mod p.
The left side is [G : P ], which is nonzero modulo p since P is a p-Sylow subgroup. Thus
| FixQ (G/P )| can’t be 0, so there is a fixed point in G/P . Call it gP . That is, qgP = gP
for all q ∈ Q. Equivalently, qg ∈ gP for all q ∈ Q, so Q ⊂ gP g −1 . Therefore Q = gP g −1 ,
since Q and gP g −1 have the same size and we’re done.
Proof of Sylow III: We will prove np ≡ 1 mod p and then np | m.
To show np ≡ 1 mod p, consider the action of P on the set Sylp (G) by conjugation. The
size of Sylp (G) is np . Since P is a finite p-group, by the fixed-point congruence we have
np ≡ |{fixed points}| mod p.
Fixed points for P acting by conjugation on Sylp (G) are Q ∈ Sylp (G) such that gQg −1 = Q
for all g ∈ P . One choice for Q is P . For all such Q, P ⊂ N(Q). Also Q ⊂ N(Q), so P
and Q are p-Sylow subgroups in N(Q). Applying Sylow II to the group N(Q), P and Q are
conjugate in N(Q). Since Q C N(Q), the only subgroup of N(Q) conjugate to Q is Q, so
P = Q. Thus P is the only fixed point when P acts on Sylp (G), so np ≡ 1 mod p.
To show np | m, consider the action of G by conjugation on Sylp (G). Since the p-Sylow
subgroups are conjugate to each other (Sylow II), there is one orbit. A set on which a group
acts with one orbit has size dividing the size of the group, so np | |G|. From np ≡ 1 mod p,
the number np is relatively prime to p, so np | m and we’re done.
Proof of Sylow III∗ : Let P be a p-Sylow subgroup of G and let G act on Sylp (G) by
conjugation. By the orbit-stabilizer formula,
np = | Sylp (G)| = [G : Stab{P } ].
The stabilizer Stab{P } of the “point” P in Sylp (G) (viewing P as a point is why we write
{P }) is
Stab{P } = {g : gP g −1 = P } = N(P ).
Thus np = [G : N(P )] and we’re done.
In the proof of Sylow I, we saw that if H is a p-subgroup of G that is not a p-Sylow
subgroup then N(H) is strictly larger than H. What can be said about N(P ) when P is a
p-Sylow subgroup? It may or may not be larger than P , but we will show that taking the
normalizer a second time will not give anything new.
Theorem 2.1. Let P be a p-Sylow subgroup of a finite group G. Then N(N(P )) = N(P ).
More generally, if H is a subgroup of G that contains N(P ) then N(H) = H.
Proof. We will prove H ⊂ N(H) and N(H) ⊂ H. The containment H ⊂ N(H) is easy,
To prove N(H) ⊂ H let x ∈ N(H), so xHx−1 = H. Since P ⊂ N(P ) ⊂ H we have
xP x−1 ⊂ xHx−1 = H, so P and xP x−1 are both p-Sylow subgroups of H. By Sylow II
for the group H, there is y ∈ H such that xP x−1 = yP y −1 . Thus y −1 xP (y −1 x)−1 = P , so
y −1 x ∈ N(P ) ⊂ H, so x ∈ yH = H. 
6 KEITH CONRAD

3. Historical Remarks
Sylow’s proof of his theorems appeared in [1]. Here is what he showed (of course, without
using the label “Sylow subgroup”).
(1) There exist p-Sylow subgroups. Moreover, [G : N(P )] ≡ 1 mod p for each p-Sylow
subgroup P .
(2) Let P be a p-Sylow subgroup. The number of p-Sylow subgroups is [G : N(P )]. All
p-Sylow subgroups are conjugate.
(3) Any finite p-group G with size pk contains an increasing chain of subgroups
{e} = G0 ⊂ G1 ⊂ G2 ⊂ · · · ⊂ Gk ⊂ G,
where each subgroup has index p in the next one. In particular, |Gi | = pi for all i.
Here is how Sylow phrased his first theorem (the first item on the above list):1
Si pα désigne la plus grande puissance du nombre premier p qui divise l’ordre
du groupe G, ce groupe contient un autre H de l’ordre pα ; si de plus pα ν
désigne l’ordre du plus grand groupe contenu dans G dont les substitutions
sont permutables à H, l’ordre de G sera de la forme pα ν(pm + 1).
In English, using current terminology, this says
If pα is the largest power of the prime p which divides the size of the group
G, this group contains a subgroup H of order pα ; if moreover pα ν is the size
of the largest subgroup of G that normalizes H, the size of G is of the form
pα ν(pm + 1).
Sylow did not have the abstract concept of a group: all groups for him arose as subgroups
of symmetric groups, so groups were always “groupes de substitutions.” The condition that
an element x ∈ G is “permutable” with a subgroup H means xH = Hx, or in other words
x ∈ N(H). The end of the first part of his theorem says the normalizer of a Sylow subgroup
has index pm + 1 for some m, which means the index is ≡ 1 mod p.

4. Analogues of the Sylow Theorems

There are analogues of the first two Sylow theorems and Theorem 2.1 for other types of
subgroups.
(1) A Hall subgroup of a finite group G is a subgroup H whose order and index are
relatively prime. For example, in a group of order 60 a subgroup of order 12 has
index 5 and thus is a Hall subgroup. A p-subgroup is a Hall subgroup if and only if
it is a Sylow subgroup. In 1928 Philip Hall proved in every solvable group of order n
that there is a Hall subgroup of each order d dividing n where (d, n/d) = 1 and two
Hall subgroups with the same order are conjugate. Also the normalizer of a Hall
subgroup of a solvable group is its own normalizer. Conversely, Hall proved that a
finite group of order n containing a Hall subgroup of order d for each d dividing n
such that (d, n/d) = 1 has to be a solvable group.
(2) In a compact connected Lie group G, maximal tori (maximal connected abelian
subgroups of G) satisfy properties analogous to Sylow subgroups: they exist, every
torus is in a maximal torus, and all maximal tori are conjugate. The proof of conju-
gacy uses the Lefschetz fixed point theorem. Like normalizers of Sylow subgroups,
1We modify some of his notation: he wrote the subgroup as g, not H, and the prime as n, not p.
 THE SYLOW THEOREMS 7

the normalizer of a maximal torus is its own normalizer. Unlike Sylow subgroups,
maximal tori are always abelian and every element of G is in some maximal torus.
(3) In a connected linear algebraic group, maximal connected unipotent subgroups are
like Sylow subgroups: they exist, every connected unipotent subgroup is in a maxi-
mal connected unipotent subgroup, and all maximal connected unipotent subgroups
are conjugate. The proof of conjugacy uses the Borel fixed point theorem. The nor-
malizer of a maximal connected unipotent subgroup is called a Borel subgroup, and
like normalizers of Sylow subgroups each Borel subgroup is its own normalizer.

References
[1] L. Sylow, Théorèmes sur les groupes de substitutions, Mathematische Annalen 5 (1872), 584–594. Trans-
lation into English by Robert Wilson at http://www.maths.qmul.ac.uk/~raw/pubs_files/Sylow.pdf.