The aim of this chapter is to introduce formally two constructions of the theory of groups: semi-direct products and presentations of groups. Later on in the lecture we will relate semi-direct products with a 1st and a 2nd cohomology group. Presentations describe groups by generators and relations in a concise way, they will be useful when considering concrete groups, for instance in examples.
References:
[Hum96] J. F. H��������,A course in group theory, Oxford Science Publications, The Clarendon Press, Oxford University Press, New York, 1996.
[Joh90] D.L. J������,Presentations of groups, London Mathematical Society Student Texts, vol. 15, Cambridge University Press, Cambridge, 1990.
1 Semi-direct Products
The semi-direct product is a construction of the theory of groups, which allows us to build new groups from old ones. It is a natural generalisation of the direct product.
Definition 1.1 (Semi-direct product)
A groupG is said to be the (internal or inner) semi-direct product of a normal subgroup N ú G by a subgroupH §G if the following conditions hold:
(a) G“NH; (b) NXH “ t1u.
Notation: G“N¸H. Example 1
(1) A direct product G1 ˆG2 of two groups is the semi-direct product of N :“ G1 ˆ t1u by H :“ t1u ˆG2.
(2) G“S3 is the semi-direct product of N “C3 “ xp1 2 3qyúS3 und H “C2 “ xp1 2qy§S3. HenceS3–C3¸C2.
Notice that, in particular, a semi-direct product of an abelian subgroup by an abelian subgroup need not be abelian.
7
(3) More generallyG“S�(�•3) is a semi-direct product ofN“A�úS�byH “C2 “ xp1 2qy.
Remark 1.2
(a) If G is a semi-direct product of N byH, then the 2nd Isomorphism Theorem yields G{N “HN{N–H{HXN “H{t1u –H
and this gives rise to a short exact sequence
1›ÑN ›ÑG›ÑH ›Ñ1�
Hence a semi-direct product ofN byH is a special case of an extension ofN by H.
(b) In a semi-direct productG “N¸H of N byH, the subgroup H acts by conjugation on N, namely@�PH,
θ�: N ›Ñ N
� fiÑ ���´1,
is an automorphism of N. In addition θ��1 “θ�˝θ�1 for every �� �1 PH, so that we have a group homomorphism
θ: H ›Ñ AutpNq
� fiÑ θ�.
Proposition 1.3
With the above notation, N�H andθ are sufficient to reconstruct the group law on G.
Proof : Step 1. Each �PGcan be written in a unique way as�“��where�PN,�PH: indeed by (a) and (b) of the Definition, if�“��“�1�1 with�� �1PN,�� �1PH, then
�´1�1“�p�1q´1PNXH“ t1u� hence�“�1 and�“�1.
Step 2. Group law: Let�1 “�1�1� �2 “�2�2 PG with�1� �2PN,�1� �2PH as above. Then
�1�2 “�1�1�2�2“�1�loooomoooon1�2p�´11
θ�1p�2q
�1q�2“ r�1θ�1p�2qs ¨ r�1�2s�
With the construction of the group law in the latter proof in mind, we now consider the problem of constructing an "external" (or outer) semi-direct product of groups.
Proposition 1.4
LetNandH be two arbitrary groups, and letθ:H ›ÑAutpNq� �fiÑθ�be a group homomorphism.
SetG:“NˆH as a set. Then the binary operation
¨: ` GˆG ›Ñ G p�1� �1q�p�2� �2q˘
fiÑ p�1� �1q ¨ p�2� �2q:“ p�1θ�1p�2q� �1�2q
defines a group law onG. The neutral element is1G “ p1N�1Hqand the inverse ofp�� �q PNˆH isp�� �q´1 “ pθ�´1p�´1q� �´1q.
FurthermoreG is an internal semi-direct product of N0 :“Nˆ t1u –N byH0 :“ t1u ˆH–H.
Proof : Exercise 1, Exercice Sheet 1.
Definition 1.5
In the context of Proposition 1.3 we say thatG is theexternal (or outer) semi-direct product ofN byH w.r.t. θ, and we writeG“N¸θH.
Example 2
Here are a few examples of very intuitive semi-direct products of groups, which you have very prob- ably already encountered in other lectures, without knowing that they were semi-direct products:
(1) IfH acts trivially on N (i.e. θ� “IdN @�PH), then N¸θH “NˆH. (2) Let K be a field. Then
GL�pKq “SL�pKq ¸ diagpλ�1� � � � �1q PGL�pKq |λPKˆ(
�
wherediagpλ�1� � � � �1q is the diagonal matrix with (ordered) diagonal entriesλ�1� � � � �1.
(3) Let K be a field and let B:“
#˜˚ ˚
0 ... ˚
¸
PGL�pKq +
p “ upper triangular matricesq�
U:“
#˜1
...˚
0 1
¸
PGL�pKq +
p “ upper unitriangular matricesq�
T :“
#˜λ
1 0
0 ... λ�
¸
PGL�pKq +
p “ diagonal matricesq�
Then B is a semi-direct product ofT byU.
(4) Let C� “ x�yand C� “ x�y (�� �PZ•1) be finite cyclic groups.
Assume moreover that� PZ is such that��”1 pmod �q and set θ: C� ›Ñ AutpC�q
�� fiÑ pθ�q�, whereθ�:C� ›ÑC�� �fiÑ��. Then
pθ�q�p�q “ pθ�q�´1p��q “ pθ�q�´2p��2q “� � �“��� “�
since�p�q “�and��”1 pmod�q. Thus pθ�q�“IdC� and θ is a group homomorphism. It follows that under these hypotheses there exists a semi-direct product ofC� byC� w.r.t. toθ.
Particular case: � •1, �“ 2 and � “ ´1 yield the dihedral group D2� of order 2� with generators� (of order�) and�(of order 2) and the relation θ�p�q “���´1 “�´1.
The details of Examples (1)-(4) will be discussed during the Präsensübung on Wednesday, 11th of April.
2 Presentations of Groups
Idea: describe a group using a set of generators and a set of relations between these generators.
Examples: p1q C� “ x�y “ x�|�� “1y 1generator: � 1relation: �� “1 p2q D2� “C�¸θC2 (see Ex. 2(4)q 2generators: �� �
3relations: �� “1� �2 “1� ���´1 “�´1
p3q Z“ x1Zy 1generator: 1Z
no relation (ù"free group") To begin with we examine free groups and generators.
Definition 2.1 (Free group / Universal property of free groups)
LetX be a set. A free group of basisX (orfree group onX) is a groupF containingX as a subset and satisfying the following universal property: For any group G and for any (set-theoretic) map
� :X ›ÑG, there exists a unique group homomorphism˜� :F ›ÑG such that˜�|X “�, or in other words such that the following diagram commutes:
X G
F
�:“���
� ö
D!˜�s.t.˜�|X“˜�˝�“�
Moreover,|X|is called therank ofF. Proposition 2.2
IfF exists, thenF is the unique free group of basisX up to a unique isomorphism.
Proof : Assume F1 is another free group of basisX.
Let�:X ãÑF be the canonical inclusion ofX inF and let�1:X ãÑF1 be the canonical inclusion of X inF1.
X F1 F
�
�1 D!˜�
D! ˜�1 By the universal property of Definition 2.1, there exists:
- a unique group homomorphism�˜1:F ›ÑF1 s.t. �1“˜�1˝�; and - a unique group homomorphism˜�:F1›ÑF s.t. �“˜�˝�1.
X F
F
�
�
˜�˝�˜1
IdF Thenp˜�˝�˜1q|X “�, but obviously we also haveIdF|X “�. Therefore, by uniqueness, we have˜�˝�˜1“IdF.
A similar argument yields �˜1˝˜� “ IdF1, hence F and F1 are isomorphic, up to a unique isomorphism, namely˜�with inverse�˜1.
Proposition 2.3
IfF is a free group of basisX, then X generates F.
Proof : Let H :“ xXy be the subgroup of F generated by X, and let �H :“ X ãÑ H denote the canonical inclusion ofXinH. By the universal property of Definition 2.1, there exists a unique group homomorphism
�rH such that �rH˝�“�H :
X H
F
�
�H
ö D!�rH
Therefore, lettingκ:HãÑF denote the canonical inclusion ofHinF, we have the following commutative diagram:
X H F
F
�
�H κ
�rH IdF
κ˝�rH
Thus by uniquenessκ˝�rH “IdF, implying that�rH:H›ÑF is injective. Thus F “ImpIdFq “Impκ˝�rHq “Imp�rHqÑH and it follows thatF “H. The claim follows.
Theorem 2.4
For any setX, there exists a free groupF with basis X.
Proof : Set X :“ t�α |α PIuwhereI is a set in bijection with X, set Y :“ t�α |α PIuin bijection withX but disjoint fromX, i.e. XXY “ H, and letZ :“XYY.
Furthermore, setE:“î8
�“0Z�, whereZ0 :“ tp qu(i.e. a singleton), Z1 :“Z,Z2:“ZˆZ,� � � ThenE becomes a monoid for the concatenation of sequences, that is
p�1� � � � � ��q looooomooooon
PZ�
¨ p�looooomooooon11� � � � � ��1q
PZ�
:“ p�loooooooooooomoooooooooooon1� � � � � ��� �11� � � � � ��1q
PZ�`�
�
The law¨is clearly associative by definition, and the neutral element is the empty sequencep q PZ0. Define the followingElementary Operations on the elements ofE:
Type (1): add in a sequencep�1� � � � � ��qtwo consecutive elements�α� �α and obtain p�1� � � � � ��� �α� �α� ��`1� � � � � ��q
Type (1bis): add in a sequencep�1� � � � � ��qtwo consecutive elements�α� �α and obtain p�1� � � � � ��� �α� �α� ��`1� � � � � ��q
Type (2): remove from a sequencep�1� � � � � ��qtwo consecutive elements�α� �α and obtain p�1� � � � � ���ˇ�α��ˇα� ��`1� � � � � ��q
Type (2bis): remove from a sequencep�1� � � � � ��qtwo consecutive elements�α� �α and obtain p�1� � � � � ����ˇα�ˇ�α� ��`1� � � � � ��q
Now define an equivalence relation„onE as follows:
two sequences inE are equivalent :ñ the 2nd sequence can be obtain from the 1st sequence through a succession of Elementary Operations of type (1), (1bis), (2) and (2bis).
It is indeed easily checked that this relation is:
– reflexive: simply use an empty sequence of Elementary Operations;
– symmetric: since each Elementary Operation is invertible;
– transitive: since 2 consecutive sequences of Elementary Operations is again a sequence of Elementary Operations.
Now setF :“E{ „, and writer�1� � � � � ��sfor the equivalence class ofp�1� � � � � ��qinF “E{ „.
Claim 1: The above monoid law onE induces a monoid law on F.
The induced law onF is: r�1� � � � � ��s ¨ r�11� � � � � ��1s “ r�1� � � � � ��� �11� � � � � �1�s.
It is well-defined: ifp�1� � � � � ��q „ p�1� � � � � ��qandp�11� � � � � ��1q „ p�11� � � � � ��1q, then p�1� � � � � ��q ¨ p�11� � � � � ��1q “ p�1� � � � � ��� �11� � � � � ��1q
„ p�1� � � � � ��� �11� � � � � ��1q via Elementary Operations on the 1st part
„ p�1� � � � � ��� �11� � � � � ��1q via Elementary Operations on the 2nd part
“ p�1� � � � � ��q ¨ p�11� � � � � ��1q
The associativity is clear, and the neutral element isrp qs. The claim follows.
Claim 2: F endowed with the monoid law defined in Claim 1 is a group.
Inverses: the inverse of r�1� � � � � ��s P F is the equivalence of the sequence class obtained from p�1� � � � � ��qby reversing the order and replacing each�α with�α and each�α with�α. (Obvious by definition of„.)
Claim 3: F is a free group onX.
LetG be a group and� :X›ÑGbe a map. Define
p�: E ›Ñ G
p�1� � � � � ��q fiÑ �p�1q ¨ ¨ ¨ ¨ ¨�p��q, where� is defined onY by�p�αq:“�p�α´1qfor every �αPY.
Thus, ifp�1� � � � � ��q „ p�1� � � � � ��q, thenp�p�1� � � � � ��q “p�p�1� � � � � ��qby definition of� onY. Hence
� induces a map
rp�: F ›Ñ G
r�1� � � � � ��s fiÑ �p�1q ¨ ¨ ¨ ¨ ¨�p��q,
By constructionp� is a monoid homomorphism, therfore so isrp�, but sinceF and Gare groups,rp� is in fact a group homomorphism. Hence we have a commutative diagram
X G
F
�
� ö
rp�
where�:X ›ÑF� � fiÑ r�sis the canonical inclusion.
Finally, notice that the definition ofrp� is forced if we wantrp� to be a group homorphism, hence we have uniqueness ofrp�, and the universal property of Definition 2.1 is satisfied.
Notation and Terminology
¨ To lighten notation, we identify r�αs P F with �α, hence r�αs with �α´1, and r�1� � � � � ��s with
�1¨ ¨ ¨�� inF.
¨ A sequence p�1� � � � � ��q P E with each letter �� (1 § � § �) equal to an element �α� P X or
�α´1� is called a word in the generators t�α | α P Iu. Each word defines an element of F via:
p�1� � � � � ��qfiÑ�1¨ ¨ ¨��PF. By abuse of language, we then often also call �1¨ ¨ ¨��PF a word.
¨ Two words are calledequivalent :ñ they define the same element of F.
¨ Ifp�1� � � � � ��q PZ�ÑE (�PZ•0), then �is called thelength of the word p�1� � � � � ��q.
¨ A word is said to bereducedif it has minimal length amongst all the words which are equivalent to this word.
Proposition 2.5
Every group G is isomorphic to a factor group of a free group.
Proof : Let S :“ t�α P G | α P Iu be a set of generators for G (in the worst case, take I “ G). Let X :“ t�α |α PIube a set in bijection withS, and let F be the free group onX. Let �:X ãÑF denote the canonical inclusion.
X G
F F{kerp˜�q
�
� D!˜�
���� �����
ö D!p˜�
By the universal property of free groups the map � : X ãÑ G� �α fiÑ �α
induces a unique group homomorphism ˜� : F ›Ñ G such that ˜� ˝�“ �. Clearly ˜� is surjective since the generators ofG are all Imp˜�q. Therefore the 1st Isomorphism Theorem yieldsG–F{kerp˜�q.
We can now consider relations between the generators of groups:
Notation and Terminology
Let S :“ t�α P G | α P Iu be a set of generators for the group G, let X :“ t�α | α P Iu be in bijection with S, and letF be the free group on X.
By the previous proof,G–F{N, whereN :“kerp˜�q (�α Ø�α “�αN via the homomorphismp˜�).
Any wordp�1� � � � � ��qin the�α’s which defines an element ofF inNis mapped inGto an expression of the form
�1¨ ¨ ¨��“1G � where ��:“image of�� inG under the canonical homomorphism.
In this case, the word p�1� � � � � ��q is called a relation in the groupG for the set of generators S.
Now let R:“ t�β|βPJu be a set of generators ofN as normal subgroup of F (this means thatN is generated by the set of all conjugates of R). Such a setR is called aset of defining relations ofG.
Then the ordered pair pX�Rq is called a presentationof G, and we write G “ xX |Ry “ xt�αuαPI | t�βuβPJy�
The group G is said to befinitely presented if it admits a presentation G “ xX |Ry, where both
|X|�|R|†8. In this case, by abuse of notation, we also often write presentations under the form G “ x�1� � � � � �|X||�1“1� � � � � �|R|“1y�
Example 3
The cyclic group C� “ t1� �� � � � � ��´1u of order �PZ•1 generated by S :“ t�u. In this case, we have:
X “ t�u R“ t��u
F “ x�y – pC8�¨q C8 ˜�
›
ÑC�� �fiÑ�has a kernel generated by�� as a normal subgroup. ThenC� “ xt�u | t��uy.
By abuse of notation, we write simplyC�“ x� |��yor alsoC� “ x� |�� “1y.
Proposition 2.6 (Universal property of presentations)
Let G be a group generated by S “ t�α |α P Iu, isomorphic to a quotient of a free group F on X “ t�α |α PIuin bijection with S. Let R :“ t�β |βPJu be a set of relations inG.
Then G admits the presentation G “ xX | Ry if and only if G satisfies the following universal property:
X H
G
�
� ö
�
For every group H, and for every set-theoretic map � : X ›Ñ H such that
˜�p�βq “ 1H @ �β P R, there exists a unique group homomorphism � : G ›Ñ H such that �˝� “�, where � : X ›Ñ G� �α fiÑ �α, and˜� is the unique extension of� to the free group F onX.
Proof : "ñ": Suppose that G “ xX | Ry. Therefore G – F{N, where N is generated by R as normal subgroup. Thus the condition˜�p�βq “1H @ �β PR implies thatNÑkerp˜�q, since
˜�p��β�´1q “˜�p�qloomoon˜�p�βq
“1H
˜�p�q´1 “1H @�β PR�@�PF�
Therefore, by the universal property of the quotient, ˜� induces a unique group homomorphism
� :G –F{N ›Ñ H such that � ˝π “˜�, where π : F ›Ñ F{N is the canonical epimorphism.
Now, if�:X›ÑF denotes the canonical inclusion, then� “π˝�, and as a consequence we have
�˝�“�.
"": Conversely, assume that G satisfies the universal property of the statement (i.e. relatively to X�F�R). SetN:“R for the normal closure ofR. Then we have two group homomorphisms:
�: F{N ›Ñ G
�α fiÑ �α
induced by˜� :F ›ÑG, and
ψ: G ›Ñ F{N
�α fiÑ �α
given by the universal property. Then clearly �˝ψp�αq “ �p�αq “ �α for each α P I, so that
�˝ψ“IdG and similarlyψ˝�“IdF{R. The claim follows.
Example 4
Consider the finite dihedral groupD2� of order 2� with 2§�†8. We can assume that D2� is generated by
� :“ rotation of angle 2π
� and�:“ symmetry through the origin inR2�
Then x�y –C�ÑG,x�y –C2 and we have seen that D2�“ x�y ¸ x�y with three obvious relations
�� “1,�2 “1, and���´1 “�´1.
Claim: D2� admits the presentation x�� �|��“1� �2 “1� ���´1“�´1y.
In order to prove the Claim, we let F be the free group on X :“ t�� �u, R :“ t��� �2� ���´1�u, NúF be the normal subgroup generated byR, andG :“F{N so that
G“ x�� �|��“1� �2 “1� � � �´1� “1y� By the universal property of presentations the map
�: t�� �u ›Ñ D2�
� fiÑ �
� fiÑ �
induces a group homomorphism
�: G ›Ñ D2�
� fiÑ �
� fiÑ �,
which is clearly surjective sinceD2�“ x�� �y. In order to prove that� is injective, we prove that G is a group of order at most 2�. Recall that each element of G is an expression in �� �� �´1� �´1, hence actually an expression in �� �, since �´1 “ ��´1 and �´1 “ �. Moreover, ���´1 “ �´1 implies�� “�´1�, hence we are left with expressions of the form
���� with 0§�§�´1and 0§�§1� Thus we have|G|§2�, and it follows that � is an isomorphism.
Notice that if we remove the relation�� “1, we can also formally define aninfinite dihedral group D8 via the following presentation
D8:“ x�� �|�2 “1� ���´1“�´1y� Theorem 2.7
Let G be a group generated by two distinct elements, � and �, both of order 2. Then G – D2�, where 2§�§8. Moreover,�is the order of �� inG, and
G“ x�� � |�2“1� �2“1�p��q�“1y� (�“ 8simply means "no relation".)
Proof : Set�:“�� and let�be the order of�.
Firstly, note that �• 2, since �“ 1 ñ �� “1 ñ � “ �´1 “ � as �2 “ 1. Secondly, we have the relation���´1 “�´1, since
���´1 “loomoon�p�
“1G
�q�´1“��´1“�´1�´1 “ p��q´1“�´1� ClearlyG can be generated by�and �as�“�� and so� “��.
Now,H:“ x�y –C� andH úG since
���´1“�´1PH and ���´1“�PH (or because|G:H| “2)� SetC:“ x�y –C2.
Claim: �RH.
Indeed, assuming�PH yields�“�� “ p��q� for some0§�§�´1. Hence 1“�2 “�p��q� “ p��q�´1�“ p��looomooon¨ ¨ ¨�q
length�´1
�p��looomooon¨ ¨ ¨�q
length�´1
�
so that conjugating by�, then�, then� � �, then�, we get1“�, contradicting the assumption that�p�q “2.
The claim follows.
Therefore, we have proved that G“HC and HXC “ t1u, so that G “H¸C “D2� as seen in the previous section.
Finally, to prove thatGadmits the presentationx�� �|�2 “1� �2“1�p��q�“1y, we apply the universal property of presentations twice to the maps
�: t��� ��u ›Ñ x�� �|�2 “1� �2 “1�p��q�“1y
�� fiÑ �
�� fiÑ ��
and
�: t��� ��u ›Ñ G“ x�� �|��“1� �2“1� ���´1 “1y
�� fiÑ �
�� fiÑ �� .
This yields the existence of two group homomorphisms
� :G“ x�� �|��“1� �2“1� ���´1 “1y›Ñ x�� �|�2“1� �2 “1�p��q�“1y and �:x�� �|�2“1� �2“1�p��q�“1y›ÑG“ x�� �|��“1� �2“1� ���´1 “1y such that�� “Idand��“Id. (Here you should check the details for yourself!)