MATH 6121 lecture notes

Transcribed and formatted by John M. Campbell

[email protected]

7 September 29 lecture
7.1 Errata
Recall that the normalizer of a subset A of the underlying set of a group G is usually defined as
NG (A) = {g ∈ G : gAg −1 = A}.
However, the normalizer of a subset A ⊆ G was incorrectly defined in class as {g ∈ G : ∀a ∈ A gag −1 ∈
A}. Let sets of this form be denoted as follows:

MG (A) := g ∈ G : ∀a ∈ A gag −1 ∈ A .


As indicated in the lecture notes for the previous lecture, MG (A) may not form a subgroup, as may be
verified by letting G denote the group of permutations of Z, and letting A = {σ ∈ G : ∀i < 0 σ(i) = i}
and g(x) = x + 1, with g −1 (x) = x − 1. It is thus easily seen that MG (A) is not closed under inverses.
Consider the classification of infinite non-abelian groups G with a subset A such that MG (A) is not a
subgroup. How can groups of this form be classified?

7.2 Sylow theory

Theorem 7.1. Letting G be a group of prime power order, with |G| = pa , we have that:
1. Z(G) 6= {1};

2. For N E G, N ∩ Z(G) 6= {1}; and

3. For H ≤ G, where H 6= G, NG (H) 6= H.

Recall that we proved that Z(G) 6= {1} if |G| = pa using a clever counting argument. Theorem 7.1.2.
may be proven using a similar enumerative technique.
Sketch of a proof of Theorem 7.1.3.: Find a proper normal subgroup K E G and K / H, such that K is
maximal and that G/K is not trivial. That is, we take K to be the largest normal subgroup of G which
is also contained in H. For example, it is possible that we could take {1}. Observe that H/K ≤ G/K.
The quotient group G/K is also a p-group, so there exists a non-identity element zK in the center
Z(G/K) of G/K, with z ∈ / K since zK 6= eK. For any h ∈ H, we have that hK ∈ H/K, so

zhK = zKhK = hKzK = hzK,

and we thus have that hK = z −1 hzK. Therefore, zhz −1 ∈ hK ⊆ hH = H. So zhz −1 ∈ H for all h ∈ H.
Since MG (A) = NG (A) if G is finite, we have that z ∈ NG (H), but z ∈
/ H.
Exercise 7.2. Show that z ∈
/ H with respect to the above argument.
Corollary 7.3. If H ≤ G, |H| = pα−1 , and |G| = pα , then H E G.

1
 
Sketch of a proof: If NG (H) 6= H, then |H| ≤ |NG (H)| and |NG (H)||G| = pα , so |NG (H)| = pα , and
NG (H) = G.
Corollary 7.4. If |G| = pn , then G is solvable.
Sketch of a proof: Our strategy is to use induction. Recall that a subnormal series of a given group
is a sequence of subgroups such that each such subgroup is a (proper) normal subgroup of the next.
Consider the following subnormal series:

{1} E Z(G) E G.

This subnormal series will have a composition series such that the composition factors are all abelian.
Recall that a finite group is said to be solvable if it has a subnormal series whose factor groups are all
abelian. In particular, abelian groups are solvable. If G is of order pn , and Z(G) is nontrivial, and is of
order pk , then we have that G/Z(G) is of order pn−k .

7.2.1 An illustration of a 2-group of order 8

Let the dihedral group D4 be denoted as follows:

D4 = 1, a, a2 , a3 , b, ba, ba2 , ba3 .



Question 7.5. What is Z(D4 )?

We claim that Z(D4 ) = {1, a2 }. It is useful to think about this equality from something of a geometric
perspective: one may prove geometrically that the only isometries in D4 which commute with all the
functions in D4 are the identity isometry, together with the mapping in D4 given by a half-turn rotation.
Example 7.6. The mapping ba is not in Z(D4 ), since ba = a3 b 6= ab.
Example 7.7. The mapping ba2 is not in Z(D4 ), because a(ba2 ) = ba3 a2 = ba, and since (ba2 )a =
ba3 6= ba.
We thus arrive at the following subnormal series:

{1} E {1, a2 } E D4 .

What is D4 /{1, a2 }? It is easily seen that:

D4 /{1, a2 } = {{1, a2 }, {a, a3 }, {b, ba2 }, {ba, ba3 }}.

Question 7.8. What is the center of D4 /Z(D4 )?

It is clear that Z(D4 /Z(D4 )) = D4 /Z(D4 ), since D4 /Z(D4 ) is of order 4, and therefore must be abelian.
Recall that up to isomorphism, there are only two groups of order 4, namely: the cyclic group Z/4Z,
and the Klein four-group (Z/2Z) × (Z/2Z).
We thus arrive at the following subnormal series:

{{1, a2 }} E {{1, a2 }, {a, a3 }} / D4 /{1, a2 }.

Also consider the following subnormal series:

{1} E {1, a2 } E {1, a, a2 , a3 } E D4 .

2
7.3 Sylow p-subgroups
Recall that a finite group is a p-group iff its order is a power of p.
Similarly, a p-subgroup is a subgroup which is also a p-group.
A Sylow p-subgroup of a group |G| = pα m is a p-subgroup of order pα with (p, m) = 1.
There are many equivalent ways of defining Sylow p-subgroups. In particular, Sylow p-subgroups are
defined in the following equivalent ways in John Fraleigh’s A First Course in Abstract Algebra and
Joseph Gallian’s Contemporary Abstract Algebra, respectively.
Definition 7.9. A Sylow p-subgroup P of a group G is a maximal p-subgroup of G, that is, a
p-subgroup contained in no larger p-subgroup.
Definition 7.10. Let G be a finite group and let p be a prime. If pk divides |G| and pk+1 does not
divide |G|, then any subgroup of G of order pk is called a Sylow p-subgroup of G.
Let np denote the # of Sylow p-subgroups, and let Sylp (G) denote the set of all Sylow p-subgroups.
Theorem 7.11. Letting G be a finite group, and letting p ∈ N be a prime number, we have that:
1. Sylow p-subgroups always exist, i.e., np ≥ 1;

2. # of Sylow p-subgroups divides |G|, i.e., np |G|;
3. # of Sylow p-subgroups ≡ 1(mod p), i.e., np = kp + 1 for some k ≥ 0; and
4. All Sylow p-subgroups are conjugate, i.e., if K, H ∈ Sylp (G), then ∃g ∈ G gHg −1 = K.
Exercise 7.12. Illustrate the above theorem using the Sylow p-subgroups of S3 ∼ = D3 . Recall that S3
denotes the group consisting of permutations on {1, 2, 3}.
Remark 7.13. This theorem is particularly useful for finding normal subgroups of a group (and thereby
showing that G is not simple). List all values of np and see if any are forced to be 1 by this theorem.
Proof sketch for Theorem 7.1.1: We inductively assume
 that the theorem holds for all n < |G|. Infor-
mally, our strategy is to “get at the center”. If p |Z(G)|, then we can find an element c ∈ Z(G) such

that order(c) = p, and |G/hci| = pα−1 m . So it must have a group of order pα−1 . Let
π : G → G/hci
denote the canonical morphism.
A Sylow p-subgroup of G/hci will have order pα .
If p - |Z(G)|, then:
X |G|
|G| = |Z(G)| = .
i=1
|NG ({xi })|
|c(xi )|>1

We know that p|G|, but p - |Z(G)|, so we know that there exists an index i such that
|G|
p- .
|NG ({xi })|
This implies that |NG ({xi })| = pα m0 for some m0 6= m. So there exists a nontrivial subgroup of smaller
order than G, so NG ({ai }) contains a subgroup of order pα by induction.

3
Question 7.14. Why is it true that if p divides the order of Z(G), then there exists an element c ∈ Z(G)
of order p?

Observe that since Z(G) is abelian, this may be justified using the Fundamental Theorem of Finitely-
Generated Abelian Groups. However, we haven’t covered this in class.
As indicated on the course webpage, rather than cover the proof of Sylow’s theorem in full detail, there
is an outline available1 on the course webpage that includes the proofs with spartan explanations and
some details left as exercises.

7.4 Cauchy’s Theorem

In class, we tried to prove that if p divides the order of an abelian group, then there exists an element
in G of order p.
However, it is actually true in general that if p divides the order of a group G, which may or may not
be abelian, then G must have an element of order p.
This important result in group theory is known as Cauchy’s theorem.
Cauchy’s theorem is not especially difficult to prove.
The following proof of Cauchy’s theorem is based on a proof of this result given in Fraleigh’s A First
Course in Abstract Algebra.

Theorem 7.15. Cauchy’s Theorem: Let p be a prime, and let G be a finite group such that p divides
|G|. Then there exists an element in G of order p.

Proof. Let X denote the following subset of the direct product Gp :

X = {(g1 , g2 , . . . , gp ) ∈ Gp | g1 g2 · · · gp = e} .

It is natural to consider the cardinality of X. Given a choice of a first entry g1 of an element in X,

and a choice of a second entry g2 , and a choice of a third entry, and so forth, including a choice for a
(p − 1)th entry, we observe that the value of gp must be uniquely determined: in particular, we have that
gp = (g1 g2 · · · gp−1 )−1 . We thus have that |X| = |G|p−1 . Consequently, p divides the order of X.
Let σ denote the cycle (1, 2, 3, . . . , p) in the symmetric group Sp . Write:


σ(g1 , g2 , . . . , gp ) = gσ(1) , gσ(2) , . . . , gσ(p) = (g2 , g3 , . . . , gp , g1 ) ,

letting (g1 , g2 , . . . , gp ) ∈ X, and observe that since

g1 g2 · · · gp = e,

we have that
g2 g3 · · · gp = g1−1 ,
and we thus have that
g2 g3 · · · gp g1 = e,
1
http://garsia.math.yorku.ca/~zabrocki/math6121f16/documents/100616sylows.pdf.

4
thus proving that
σ(g1 , g2 , . . . , gp ) ∈ X.
More generally, we have that

gρ(1) , gρ(2) , . . . , gρ(p) ∈ X
for a cyclic permutation ρ in the cyclic subgroup hσi of Sp . We thus have that the cyclic group hσi ≤ Sp
acts on the set X in a natural way, letting

∗ : hσi × X → X

denote the group action whereby

ρ ∗ (g1 , g2 , . . . , gp ) = gρ1 , gρ2 , . . . , gρp

for ρ ∈ hσi and (g1 , g2 , . . . , gp ) ∈ G.

But the hσi-set X must be a disjoint union of orbits, say

(Orbit(x1 ) ] Orbit(x2 ) ] · · · ] Orbit(xn1 )) ] (Orbit(y1 ) ] Orbit(y2 ) ] · · · ] Orbit(yn2 )) ,

where Orbit(xi ) is a singleton set for all indices i and Orbit(yi ) is not a singleton set for all indices i.
But since | hσi | = p, and since orders of the orbits of the hσi-set X divide | hσi | by the orbit-stabilizer
theorem, we may thus deduce that either

X = Orbit(y)

for some element y or

X = Orbit(x1 ) ] Orbit(x2 ) ] · · · ] Orbit(xp ),
with n1 = p. But it is clear that

Orbit(e, e, . . . , e) = {(e, e, . . . , e)},

so X must be of the form

X = Orbit(x1 ) ] Orbit(x2 ) ] · · · ] Orbit(xp ),

which shows that there must be an element a 6= e such that ap = e. But then the order of a must be
p since p is a prime: it cannot be the case that a` for some natural number ` < p, because otherwise `
and p would be relatively prime, and this would contradict that a 6= e.