Romyar SharifiLECTURE NOTES
READING EDITIONPDF

LECTURE NOTES / Chapter 7

Abstract Algebra

Romyar Sharifi

Chapter 7 Topics in group theory

Book contents

Part 2 A Second Course

Chapter 7
Topics in group theory

7.1. Semidirect products

Proposition 7.1.1.

Let N and H be groups and φ : H Aut(N) be a homomorphism. Then there exists a group G with underlying set N ×H and group operation

(n,h)(n,h) = (𝑛𝜑(h)(n),hh)

for all n,n N and h,h H. Moreover, H = {e}×H G and N = N ×{e} G. In fact, in G we have φ(h)(n) = h𝑛h1 for all h H and n N.

Proof.

We note that (e,e) G is an identity, that (φ(h1)(n1),h1) is inverse to (n,h), and we leave it to the reader to check associativity. Clearly H,N G by definition of the multiplication, and we check that for h H and n N, we have

h𝑛h1 = (e,h)(n,e)(e,h1) = (e,h)(n,h1) = (φ(h)(n),e) = φ(h)(n) N.

Definition 7.1.2.

For groups N and H and a homomorphism φ : H Aut(N), the group defined by Proposition 7.1.1 is known as the semidirect product of N and H relative to φ and is denoted by N φH.

Example 7.1.3.

If H and N are groups and φ : H Aut(N) satisfies φ(h) = idN for all h H, then H φN is the direct product H ×N.

Example 7.1.4.

Let φ : (𝑛ℤ)× Aut(𝑛ℤ) be the isomorphism taking a (𝑛ℤ)× to multiplication by a. Set G = 𝑛ℤφ(𝑛ℤ)×. Then G Aff(𝑛ℤ) via (b,a) ( a b 0 1 ), so G is also isomorphic to Aut(Dn) by Proposition 4.3.5.

Proposition 7.1.5.

Let G be a group with normal subgroup N and subgroup H such that N H = {e} and 𝑁𝐻 = G. Define a homomorphism φ : H Aut(N) by φ(h)(n) = h𝑛h1. Then we may define an isomorphism of groups by

ψ : N φH G,ψ(n,h) = 𝑛h

for all n N and h H.

Proof.

Any g G can be written as 𝑛h for some n N and h H by assumption, so f is onto. For n,n N and h,h H, we have

ψ((n,h)(n,h)) = ψ(𝑛𝜑(h)(n),hh) = 𝑛𝜑(h)(n)hh = 𝑛hnh = ψ((n,h))f((n,h)).

If ψ(n,h) = 𝑛h = e, then n = h1 N H, so n = h = e.

The proposition we have just proven has Proposition 4.11.4 as a corollary.

Proof

Alternate proof of Proposition 4.11.4. By Proposition 7.1.5, we need only that φ : H Aut(N) given by φ(h)(n) = h𝑛h1 for h H and n N is the trivial map. That is, we need that h𝑛h1 = n, or [h,n] = e, for all such h and n. This follows as H and N are normal, so [h,n] H N = {e}.

Definition 7.1.6.

If G is a group with subgroups N and H such that 𝐺≅𝑁 φH for φ : H Aut(N) given by φ(h)(n) = h𝑛h1, then we say that G is the internal semidirect product of N and H and write G = N H to denote this.

Definition 7.1.7.

Let G be a group with normal subgroup N. A complement to N in G is a subgroup H such that G is the internal semidirect product N H of N and H.

Remark 7.1.8.

There are often many complements to a normal subgroup. In particular, if G = N H and n N, then 𝑛𝐻n1 is also a complement to N. If N is abelian, then we have the equality γ𝑛hn1 = γh of conjugation maps, but if N is nonabelian, then these may not be equal. The map φ : H Aut(N) given by φ(h) = γ𝑛hn1 would then satisfy 𝐺≅𝑁 ×φH, though not necessarily internally. Rather, this is simply an expression of the fact that G = N 𝑛𝐻n1.

The following rather general result can be used to show that two semidirect products are isomorphic.

Proposition 7.1.9.

Let H, H, N, and N be groups and φ : H Aut(N) and φ: H Aut(N) be homomorphisms. Suppose that there exist isomorphisms ψ : H H and 𝜃 : N N, and define

Θ: Aut(N) Aut(N)

by Θ(α) = 𝜃 α 𝜃1 for any α Aut(N). If Θφ = φψ, then the map

f : N φH N φH

defined by f(n,h) = (𝜃(n),ψ(h)) for all n N and h H is an isomorphism.

Proof.

Note that f has an inverse given by f1(n,h) = (𝜃1(n),ψ1(h)) for all n N and h H, so we need only show that f is a homomorphism. Letting n1,n2 N and h1,h2 H, we calculate:

f((n1,h1)(n2,h2)) = f(n1φ(h1)(n2),h1h2) = (𝜃(n1)𝜃(φ(h1)(n2)),ψ(h1)ψ(h2)), f(n1,h1)f(n2,h2) = (𝜃(n1),ψ(h1))(𝜃(n2),ψ(h2)) = (𝜃(n1)φ(ψ(h1))(𝜃(n2)),ψ(h1)ψ(h2)).

To see that the first coordinates of these expressions are equal, we check that

φ(ψ(h1))(𝜃(n2)) = Θ(ϕ(h1))(𝜃(n2)) = (𝜃 ϕ(h1)𝜃1)(𝜃(n2)) = 𝜃(ϕ(h1)(n2)).

Thus, f is a homomorphism.

We can use this to completely classify groups of order a product of two distinct primes, completing the study begun in Theorem 4.11.5.

Theorem 7.1.10.

Let p and q be distinct primes with q 1modp. Then there exists a unique isomorphism class of nonabelian groups of order 𝑝𝑞.

Proof.

Let G be a nonabelian group. By Theorem 4.11.5, we have that it has a unique normal subgroup Q of order q, and let P be a subgroup of order p. By Proposition 7.1.5, we have that G = QP. We have 𝑃≅ℤ𝑝ℤ and 𝑄≅ℤ𝑞ℤ. Fixing such isomorphisms and recalling the canonical isomorphism Aut(𝑞ℤ)(𝑞ℤ)×, we are reduced to showing that there is a unique isomorphism class of semi-direct product (𝑞ℤ)φ(𝑝ℤ), where φ : 𝑝ℤ (𝑞ℤ)× is a nontrivial homomorphism. The group (𝑞ℤ)× is cyclic by Corollary 6.5.5. Let a (𝑞ℤ)× be a generator.

Any nontrivial homomorphism φ : 𝑝ℤ (𝑞ℤ)× must send 1 to an element of order p in (𝑞ℤ)×. If we set b = a(q1)p, then φ(1) = bi for some i with i0modp. Let us denote this particular homomorphism by φi, and define ψi: 𝑝ℤ 𝑝ℤ to be multiplication by i. Then φi = φ1 ψi since both maps send 1 to bi. Proposition 7.1.9 then tells us that the semidirect products (𝑞ℤ)(𝑝ℤ) defined by φ1 and φi are isomorphic. That is, there is a unique isomorphism class of nonabelian semidirect product of order 𝑝𝑞.

7.2. Composition series

First, we explain how simple groups may be used in building arbitrary finite groups, starting with the following definition.

Definition 7.2.1.

Any collection (Hi)i of subgroups of a group G with Hi1 Hi for i is called a series of subgroups of G.

Definition 7.2.2.

Let 𝒞 = (Hi)i be a series of subgroups of a group G.

a.

We say that 𝒞 is an ascending series if Hi = 1 for i sufficiently small.

b.

We say that 𝒞 is a descending series if Hi = G for i sufficiently large.

c.

We say that 𝒞 is a finite series if it is both ascending and descending.

d.

The length of a finite series 𝒞 is difference ji of the smallest integer j such that Hj = G and largest integer i such that Hi = 1.

Notation 7.2.3.

We use the notation

1 = H0 H1 Ht1 Ht = G

to denote a finite series of subgroups Hi of a group G with H0 = 1, Ht = G. It has length t if H0H1 and Ht1Ht.

Remark 7.2.4.

To say that a series (Hi)i of subgroups of G is finite is stronger than simply saying it has only finitely many terms. For instance, if G is nontrivial, then Hi = 1 for all i provides a series with only one distinct subgroup, but it is not finite as no Hi equals G.

Remark 7.2.5.

A descending series in G is often taken to be a list (Hi)i of subgroups of G with Hi Hi1 for all i and Hi = G for i sufficiently small. This agrees with the usual notion in the sense that letting Ki = Hi will provide a descending series (Ki)i in the sense of the original definition.

Definition 7.2.6.

A finite series

1 = H0 H1 Ht1 Ht = G

of subgroups of G is said to be a subnormal series if Hi1 Hi for all 1 i t. It is called a normal series if Hi G for all 0 i t 1.

Definition 7.2.7.

Two subnormal series (Hi)i=0t and (Ki)i=0t are equivalent if there exists σ St such that HiHi1Kσ(i)Kσ(i)1 for all 1 i t.

Definition 7.2.8.

A refinement of a subnormal series (Hi)i=0t in a group G is a subnormal series (Ki)i=0s such that there exists an increasing function f : {0,,t}{0,,s} such that Hi = Kf(i) for 0 i t.

Theorem 7.2.9 (Schreier refinement theorem).

Any two subnormal series in a group G have refinements that are equivalent.

Proof.

Let (Hi)i=0t and (Ki)i=0s be subnormal series in G. For 0 i < t and 0 j < s, let

M𝑠𝑖+j = Hi(Hi+1 Kj) and N𝑡𝑗+i = Kj(Kj+1 Hi).

Set M𝑚𝑛 = N𝑚𝑛 = G as well. Then M𝑠𝑖+j M𝑠𝑖+j+1 for 0 i t 1 and 0 j s2 as Kj Kj+1, and M𝑠𝑖+s1 Hi+1 = Ms(i+1) for 0 i t 1 as Hi Hi+1 and Hs1 G. Thus (Mi)i=0𝑚𝑛 is a subnormal series, as is (Ni)i=0𝑚𝑛. In fact, we see from this that (Mi)i refines (Hi)i and (Ni)i refines (Ki)i.

It remains to see that (Mi)i and (Ni)i are equivalent. For 0 i t 1 and 0 j s2, note that

M𝑠𝑖+j+1 M𝑠𝑖+j Hi(Hi+1 Kj+1) Hi(Hi+1 Kj) Kj(Kj+1 Hi+1) Kj(Kj+1 Hi) N𝑡𝑗+i+1 N𝑡𝑗+i

by the butterfly lemma, and

Ms(i+1) Ms(i+1)1 Hi+1 Hi(Hi+1 Ks1)Ks1Hi+1 Ks1Hi Ks1(KsHi+1) Ks1(KsHi) K(s1)t+i+1 K(s1)t+i

for 0 i t 1 since Ks = G. Thus, the two refinements are equivalent.

Definition 7.2.10.

A subnormal series of subgroups

1 = H0 H1 Ht1 Ht = G

of a group G is called a composition series for G if HiHi1 is simple for each 1 i t. The simple groups HiHi1 are referred to as the composition factors of the series.

Lemma 7.2.11.

Let

1 = H0 H1 Ht1 Ht = G

be a composition series for G, and let N be a proper normal subgroup of G.

a.

There exists s t and an increasing function f : {0,,s}{0,,t} with f(0) = 0 such that

1Hf(1) N Hf(2) N Hf(s) N = N

is a composition series for N with composition factors

Hf(i) N Hf(i1) N Hf(i) Hf(i)1.
b.

Set H¯i = Hi(HiN) for 0 i t. There exists an r t and an increasing function f: {0,1,,r}{0,1,,t} with f(0) = 0 such that

1 = H¯0 H¯f(1) H¯f(r1) H¯f(r) = GN

is a composition series for GN with composition factors

H¯f(i)H¯f(i1)Hf(i)Hf(i)1.
c.

In the notation of parts a and b, the images of f and f to be complementary away from 0,x and r+s to equal t.

Proof.

Let 0 i t 1. The quotient (HiN)(Hi1 N) is a subgroup of the simple group HiHi1 and therefore necessarily trivial or improper. Let s be the number of simple quotients. Let f(0) = 0, and for 1 j s, let f(j) be the smallest positive integer greater than f(j1) and such that (Hf(j) N)(Hf(j)1 N) is simple. Then Hf(j)1 N = Hf(j1) N, and the result follows.

Similarly, by the third isomorphism theorem, we have

H¯i H¯i1 Hi Hi1(HiN),

which is a quotient of HiHi1 by the image of Hi+1 N in it. Since Hi+1Hi is simple, this image is either trivial or Hi+1Hi. That is, H¯i+1H¯i is either trivial or simple. Let s be the number of simple terms. Set f(0) = 0, and for 1 j r, take f(j) to be the smallest integer greater than f(j1) such that H¯f(j)H¯f(j)1 is simple. Then H¯f(j)1 = H¯f(j1).

Note that H¯iH¯i1 if and only if HiN = Hi1 N. Then r+s = t and the images of f and f are complementary by construction.

We leave the straightforward proof of the following lemma to the reader.

Lemma 7.2.12.

Let G be a group, and let N be a normal subgroup. Suppose that N has a composition series

1 = H0 H1 Ht1 Hs = N

and GN has a composition series

1 = Q0 Q1 Qr1 Qr = GN.

For 1 i r, let Hs+i denote the unique subgroup of G containing N and such that Hs+iN = Qi, which exists by Proposition 2.13.10. Then the series

1 = H0 H1 Ht1 Ht = G

is a composition series of G with composition factors satisfying Hs+iHs+i1 = QiQi1 for 1 i r.

Corollary 7.2.13.

Let G be a group and N a normal subgroup. If N and GN have composition series, then G has a composition series. Moreover, its list of composition factors consists of the concatenation of the list of composition factors of N by the list of composition factors of G.

Theorem 7.2.14 (Jordan-Hölder theorem).

a.

Every finite group has a composition series.

b.

Let G be a nontrivial group with composition series

1 = N0 N1 Ns1 Ns = G

and

1 = H0 H1 Ht1 Ht = G.

Then s = t and there exists a permutation σ St such that

Hσ(i)Hσ(i)1NiNi1

for all 1 i t.

Proof.

To show part a, we work by induction on the order n of the group G. It is clear in the case that G is trivial, with t = 0. Now, if G is nontrivial of order n, then either it is simple, and the composition series is 1 G, or it is not, and there exists a nontrivial normal subgroup K G, and then K and GK have composition series by induction. The result is then immediate from Lemma 7.2.13.

To see that the composition series is unique in the stated sense of part b, start with two composition series as in the statement of the theorem. We work by induction on the minimal length s of a composition series for G. If s = 0, then G is trivial. If s = 1, then G is simple, so it cannot have a nontrivial normal subgroup, and all composition series must have length 1. Consider N = Ns1, which has the composition series

1N1 Ns2 Ns1 = N,

as well as a composition series

1Hf(1) N Hf(r1) N Hf(r) N = N

for some r t and increasing f : Xr Xt by Lemma 7.2.11a. Since the minimal length of a composition series of N is less than s, we have by induction that r = s1 and there exists σ Ss1 such that

NiNi1(Hf(σ(i)) N)(Hf(σ(i)1) N)Hf(σ(i))Hf(σ(i))1

for all i, again by Lemma 7.2.11a.

Let k < t be maximal such that Hk1 N. Then

Hk1 N = Hk1 HkN < Hk.

Since HkHk1 is simple, this forces Hk1 = HkN. In particular, k is not in the image of f. Moreover, we have

HkHk1Hk(HkN)HkN𝑁≅𝐺N,

the latter step as N is a maximal normal subgroup of G and Hk≰N. As we have found the final composition factor in the series (Ni)i among those of the series (Hi)i, it remains only to show that s = t. If (HiN)(Hi1 N) nontrivial for any for ik, then by Lemma 7.2.11c, the group GN has a composition series of length at least 2, but GN is simple, so this is impossible. Thus, r = t 1 as well, as needed.

Definition 7.2.15.

The Jordan-Hölder factors of a group G are the terms in a list of the isomorphism classes of the composition factors in a composition series for G.

Examples 7.2.16.

a.

The group pn for a prime p has n copies of 𝑝ℤ as its Jordan-Hölder factors, which arise from its unique composition series

0pn1pn2ppn.
b.

The group 6 has two composition series

026 and 036,

both of which have Jordan-Hölder factors 2 and 3.

c.

Let G be a nonabelian group order 𝑝𝑞 with p and q distinct primes and q 1modp. Then G has a unique composition series 1QG, where Q has order q, and it has Jordan-Hölder factors 𝑝ℤ and 𝑞ℤ.

d.

For n 6, the group Sn has a unique composition series 1AnG with Jordan-Hölder factors An and 2.

Remark 7.2.17.

The set of Jordan-Hölder factors of a group tell us a great deal about the structure of a group, but they do not tell us the group. For instance, n2 and (𝑛ℤ)2 have the same Jordan-Hölder factors for any n 2.

7.3. Solvable groups

Definition 7.3.1.

Let G be a group. The derived series of G is the unique descending series (G(i))i0 of subgroups of G with G(0) = G and G(i) = [G(i1),G(i1)] for all i 1.

Notation 7.3.2.

Often, one writes G for G(1) = [G,G] and G for G(2) = [[G,G],[G,G]].

Definition 7.3.3.

A group G is solvable if its derived series is finite.

Examples 7.3.4.

a.

The derived series of an abelian group satisfies G(i) = 1 for all i 1. Hence, abelian groups are solvable.

b.

The derived series of a nonabelian simple group G satisfies G(i) = G for i 0. Hence, nonabelian simple groups are not solvable.

Example 7.3.5.

Let R be a commutative ring. Consider the group

T = Heis(R) = {( 1 a c 1 b 1 )a,b,c R}GL3(R).

The reader should verify that this group has commutator subgroup equal to its center, which is

Z(T ) = [T,T ] = {( 1 0 c 1 0 1 )c R}GL3(R).

In particular, T (2) = [Z(T ),Z(T )] = 0. In fact, the reader might find a rather canonical isomorphism from Heis() to the group presented by x,y,z[x,y] = z,[x,z] = [y,z] = e.

Lemma 7.3.6.

The groups G(i) for i 1 are characteristic subgroups of a group G.

Proof.

First, Lemma 4.3.15c tells us that G(i) is characteristic in G(i1) for each i 1, and then the result follows recursively from Lemma 4.3.16.

We can now prove the following equivalence of definitions of solvability.

Proposition 7.3.7.

The following statements regarding a group G are equivalent:

i.

G is solvable,

ii.

G has a normal series with abelian composition factors, and

iii.

G has a subnormal series with abelian composition factors.

Proof.

That (i) implies (ii) is a consequence of the facts that the group G(i) are characteristic, hence normal, and that G(i1)G(i) is the quotient of G(i1) by its commutator subgroup, hence abelian. That (ii) implies (iii) is obvious. So, suppose (iii) and let

G = N0 N1 Nt1 Nt = 1

be a subnormal series of length t. (Note the reversed indexing, as in Remark 7.2.5.) We claim that G(i) Ni for each i 0. For i = 0, we have G = G(0) = N0. In general, suppose inductively that G(i1) Ni1. Then G(i) [Ni1,Ni1] by definition, and we have [Ni1,Ni1] Ni as Ni1Ni is abelian. Therefore, we have that Gt = Nt = 1, and G is solvable.

We also have the following.

Proposition 7.3.8.

a.

Every subgroup of a solvable group is solvable.

b.

Every quotient group of a solvable group is solvable.

c.

If G is a group and N is a normal subgroup of G such that N and GN are both solvable, then G is solvable as well.

Proof.

Let G be a group and N a normal subgroup. If G is solvable, then it has a composition series with abelian factors, so N and GN are solvable by Lemma 7.2.11. Part (iii) is a corollary of Corollary 7.2.13, since the derived series of N and GN have abelian composition factors.

Proposition 7.3.9.

A group G with a composition series is solvable if and only if it is finite and its Jordan-Hölder factors are all cyclic of prime order.

Proof.

If G has cyclic Jordan-Hölder factors, then G is solvable by Proposition 7.3.7. If G is solvable and has a composition series, then the composition factors are abelian by Proposition 7.3.7 and the uniqueness in Theorem 7.2.14. As composition factors, they are also simple, hence cyclic of prime order, from which it follows that G is finite.

Example 7.3.10.

All groups of order 𝑝𝑞 for distinct primes p and q are solvable, as their Jordan-Hölder factors are 𝑝ℤ and 𝑞ℤ.

Definition 7.3.11.

A Hall subgroup of a finite group G is a subgroup H such that |H| and [G : H] are relatively prime.

7.4. Nilpotent groups

Definition 7.4.1.

Let G be a group. The lower central series of G is the unique descending series (Gi)i1 of G with G1 = G and

Gi+1 = [G,Gi] = {[a,b]a G,b Gi}

for each i 1.

Remark 7.4.2.

By convention, Gi starts with G1 = G, while G(i) starts with G(0) = G.

Remark 7.4.3.

For a group G, we have G = G2, but G = [G,G] can be smaller than G3 = [G,G]. In fact, we clearly have G(n+1) Gn for all n 1.

The reader will easily verify the following by induction.

Lemma 7.4.4.

The groups Gi in the lower central series of a group G are characteristic subgroups of G.

Definition 7.4.5.

A group G is nilpotent if its lower central series is finite.

Definition 7.4.6.

The nilpotency class of a nilpotent group is the length of its lower central series, which is to say the smallest n 0 such that Gn+1 = 1.

Lemma 7.4.7.

Let G be a group. Then G(i) Gi for all i.

Proof.

This is almost trivial by induction, as

G(i) = [G(i1),G(i1)] [G,G(i1)] [G,G i1] = Gi.

Corollary 7.4.8.

Nilpotent groups are solvable.

Examples 7.4.9.

a.

The lower central series of an abelian group satisfies Gi = 1 for all i 1.

b.

Let T be as in Example 7.3.5. Then T 1 = Z(T ) and T 2 = 1, so T is nilpotent.

c.

Let G be the group Aff() of upper-triangular matrices in GL2() with lower-right entry 1, as in Example 2.12.13). We have

[( a 0 0 1 ), ( 1 b 0 1 )] = ( 1 b(1a) 0 1 )

for all a × and b . It follows easily from this that

G2 = G = {( 1 b 0 1 )|b }

and then Gi = G2 for all i 1. On the other hand, G is abelian, so G = 1. Thus, Aff() is solvable but not nilpotent. (Note that can be replaced by any nonzero commutative ring R in which R× and 1R× have nontrivial intersection.)

We can give an alternative characterization of nilpotent groups through the ascending series of the following definition.

Definition 7.4.10.

The upper central series of a group G is the unique ascending series (Zi(G))i0 with Z0(G) = 1 and Zi+1(G) equal to the inverse image of Z(GZi(G)) under the projection map G GZi(G) for all i 0.

Remark 7.4.11.

For any group G, we have Z1(G) = Z(G), and Zi(G) Zi+1(G) for all i 0.

Lemma 7.4.12.

If G is a nontrivial nilpotent group, then Z(G)1.

Proof.

Let n be the nilpotency class of G. Then Gn is nontrivial but central in G since [G,Gn] = 1.

Proposition 7.4.13.

A group G is nilpotent if and only if Zi(G) = G for i sufficiently large. In this case, the nilpotency class of G equals the smallest n such that Zn(G) = G, and we have Gn+1i Zi(G) for all 1 i n.

Proof.

The result is clear for abelian groupts, which are the nilpotent groups of nilpotency class 1. Let G¯ = GZ(G) for brevity of notation.

Suppose that G is nilpotent of nilpotency class n 2. As Gn+1 = [G,Gn] = 1, we have 1Gn Z(G). Since Z(G) is central in G, it follows that

G¯i = GiZ(G)Z(G)

for all i. Since Gn1 is not central in G by definition, we have that G¯n11. By induction on n, we then have

Zi1(G¯)G¯ ni Zi(G¯)

for 1 i n1, and Zn1(G¯) = G¯. Taking the inverse images of these groups under the quotient map from G, we obtain

Zi(G)G ni Zi+1(G)

for 1 i n1 as well, and in particular Zn1(G) < Zn(G) = G.

Conversely, if Zn(G) = G for some minimial n 2, then Zn1(G¯) = G¯, so G¯ is nilpotent of nilpotency class n1 by induction on n. This means that Gn Z(G), and therefore Gn+1 = 1, so G is nilpotent.

The following corollary can also be seen directly, using Proposition 4.9.6.

Corollary 7.4.14.

Finite p-groups are nilpotent.

Proof.

This follows by induction on the order of a nontrivial p-group P, since Z(P)1, and PZ(P) is a p-group which we suppose by induction to be nilpotent. Then Zi+1(P) is the inverse image of Zi(PZ(P)) in P, so P is nilpotent as well.

Theorem 7.4.15 (Frattini’s argument).

Let G be a finite group and N a normal subgroup. Let P be a Sylow p-subgroup of N. Then G = N NG(P).

Proof.

For any g G, we have 𝑔𝑃g1 N, as N is normal, so 𝑔𝑃g1 is also a Sylow p-subgroup of N. As such, it is conjugate to P in N, which is to say there exists a N such that 𝑎𝑔𝑃(𝑔𝑎)1 = P, or 𝑎𝑔 NG(G). In other words, g N NG(P).

We are now ready to prove the following equivalent conditions for nilpotency.

Theorem 7.4.16.

Let G be a finite group. Then the following are equivalent:

i.

the group G is nilpotent,

ii.

every proper subgroup of G is a proper subgroup of its normalizer in G,

iii.

every Sylow p-subgroup of G is normal,

iv.

G is the direct product of its Sylow p-subgroups,

v.

every maximal proper subgroup of G is normal.

Proof.

Suppose that G is nilpotent of nilpotence class n, and let H be a proper subgroup of G. If 𝐻𝑍(G) = G, then H is a proper subgroup of NG(H) = G. Thus, we may suppose that 𝐻𝑍(G)G. As we always have that NG(𝐻𝑍(G)) = NG(H), we may further assume that Z(G) H in proving (ii). In this case, HZ(G) is a proper subgroup of GZ(G), which has nilpotence class less than n as Gn Z(G), so (GZ(G))n = 1. Thus, HZ(G) is a proper subgroup of NG(HZ(G)) by induction, but the latter group is NG(H)Z(G) since Z(G) NG(H), and therefore HZ(G) is a proper subgroup of NG(H)Z(G). Thus, (i) implies (ii).

Next, suppose (ii). If G is a p-group, (iii) obviously holds, so suppose this is not the case. Let P be a Sylow p-subgroup of G for some p|G|, and note that P < G. Let N = NG(P). By part (ii), we have that P < N. Note also that P is a normal subgroup of NG(N) in that it is characteristic in N, which forces NG(N) = N. Since (ii) holds, N cannot be proper in G, and thus P is normal in G. Hence, (ii) implies (iii).

Suppose (iii). Let s be the number of primes dividing |G|, and let P1,P2,,Ps be the distinct Sylow subgroups of G. If s = 1, then we are done. In general, we set H = P1Ps1 G, and by induction we have that 𝐻≅P1 ×P2 ××Ps1. We then note that H Ps = 1 and HPs = G, so 𝐺≅𝐻 ×Ps. Hence, (iii) implies (iv).

Suppose (iv), and let M be a maximal proper subgroup of G. Let P1,P2,Ps be the distinct Sylow subgroups of G. If M PiPi for some i, then M Pj = Pj for all ji, since otherwise M < MPi < G. Thus, M is the direct product of M Pi and the Pj for ji. By the first Sylow theorem, M Pi is normal in Pi, so M is normal in G. Thus, (iv) implies (v).

Suppose (v). Let P be a Sylow p-subgroup of G, and suppose it is not normal. Let M be a maximal proper subgroup of G containing NG(P). Then M is normal in G, and Frattini’s argument implies that G = MNG(P) = M, a contradiction. So, P is normal and (iii) holds, and so (iv) holds. It then suffices to note that finite p-groups are nilpotent by Corollary 7.4.14, as this tells us that (v) implies (i).

The following is a useful fact regarding nilpotent groups.

Proposition 7.4.17.

Let G be a nilpotent group, and let S be a subset of G with image in Gab a generating set. Then S generates G.

Proof.

We prove this by induction on the nilpotence class n of nilpotent groups G. It is clear if G is abelian, or n = 1. For n 2, consider GGn, and note that its abelianization is Gab, so by induction GGn is generated by the image of S. Thus, if we let H = S, we have G = GnH. This implies that H is normal in G since Gn Z(G) and thus elements of Gn and of H normalize H. We have that

Gn = [GnH,Gn1] = [H,Gn1] H,

the first equality as GnH = G, the second as Gn Z(G), and the third as H G. It follows that

G = GnH = H = S,

as claimed.

7.5. Groups of order p3

We note the following useful fact.

Lemma 7.5.1.

Let G be a group such that GZ(G) is cyclic. Then G is abelian, so G = Z(G).

Proof.

Any b GZ(G) has image generating GZ(G), so G = Z(G)b. As b commutes with itself and every element of Z(G), it is in the center of G, a contradiction.

Let us classify the groups of order p3 for a prime number p.

Theorem 7.5.2.

Let p be a prime number. There are exactly two isomorphism classes of nonabelian groups of order p3. These are represented by:

a.

if p = 2, the dihedral group D4 and the quaternion group Q8, and

b.

if p is odd, the Heisenberg group Heis(𝑝ℤ) and the group

K = {( a b 0 1 ) Aff(p2)|a 1modp}.
Proof.

Let G be a nonabelian group of order p3. By Lemma 7.5.1, the quotient GZ(G) cannot be cyclic. This eliminates the possibility that |Z(G)| = p2, since then GZ(G) would be cyclic of order p. Also, |Z(G)|1 as G is a p-group. Thus, we have |Z(G)| = p, and GZ(G) is a direct product of two cyclic groups of order p. Note that [G,G] Z(G) since GZ(G) is abelian, which forces [G,G] = Z(G) since G is nonabelian. Let a,b G with images together generating GZ(G). Then G = a,b by Proposition 7.4.17, since finite p-groups are nilpotent. Moreover, z = [b,a] generates Z(G), and note that this means 𝑎𝑏 = 𝑏𝑎𝑧.

Now, suppose that G has an element of order p2. Without loss of generality, we may suppose that it is b. Then bp generates Z(G), so we have b𝑝𝑖 = z for some i with 0 < i < p. We then have 𝑎𝑏 = b1+𝑝𝑖a, which in particular tells us that b is a normal subgroup of G. Suppose that we can also choose a to have order p. Then G = ba, and in fact G has a presentation

𝐺≅a,bap = bp2 = e,𝑎𝑏 = b1+𝑝𝑖a.

If p is odd, then we have an isomorphism

f : GK,f(a) = ( 1+𝑝𝑖 0 0 1 ),f(b) = ( 1 1 0 1 ),

and if p = 2, then f : G D4 defined by f(a) = s, f(b) = r works as well.

Suppose now that we cannot choose a and b with G = a,b and either a or b of order p. If p = 2, then this implies that a2 = b2 = (𝑎𝑏)2 = z, and from this one checks that G is a quotient of the group with presentation

x,yx2 = y2 = (𝑥𝑦)2,x4 = e,

and the latter group is isomorphic to Q8 under the map which takes x to i and y to j. If p is odd, then one sees that

(𝑏𝑎)p = bpapzp(p1)2 = bpap = z𝑝𝑖z𝑝𝑗 = zp(i+j)

for some i,j prime to p. Note that we have used p is odd here, since otherwise p(p1)2 = 1, which is not a multiple of p = 2. If we replace a by ak where k with 𝑖𝑘 jmodp, then (𝑏𝑎)p = e, which yields a contradiction.

Finally, suppose that G has no element of order p2. If p = 2, then z = [b,a] = (𝑏𝑎)2 = e, which is a contradiction. If p is odd, then G is a quotient of the group that has presentation

x,yxp,yp,[x,y]p,[x,[x,y]],[y,[x,y]],

but Heis(𝑝ℤ) has this presentation, so it is isomorphic to G in that G has order p3.

Find in the notes