Published on

Geometric Topology 20

Authors

Geometric Topology 20

Free Groups

The free group on one letter aa is the infinite cyclic group:

F1=⟨a⟩={an:n∈Z}≅(Z,+),F_1 = \langle a \rangle = \{a^n : n \in \mathbb{Z}\} \cong (\mathbb{Z},+),

generated by aa (including a−1a^{-1}). The element aa is not subject to any relation other than the axiom aa−1=e=a−1aaa^{-1} = e = a^{-1}a, and the rule aman=an+ma^ma^n= a^{n+m} (for any n,m∈Zn,m \in \mathbb{Z}) that comes about from the meaning assigned to ama^m for m>0,m=0m>0,m=0 and m<0m <0.

Now let G,HG,H be groups, written multiplicatively. The free product G∗HG {\Large*} H is a group whose elements consist of finite words made up by combining elements from either group (like g1h1g2h2...g_1h_1g_2h_2...) in which identity elements are suppressed (or retained as a combined identity if there are no other letters). The same idea allows us to construct a free group on several letters:

Example: The free group on two letters is:

F2=F1∗F1={am1bn1am2bn2...amkbnk...:mi,nj∈Z,k≥0}.F_2 = F_1 {\Large*} F_1 = \{a^{m_1}b^{n_1}a^{m_2}b^{n_2}...a^{m_k}b^{n_k}...:m_i,n_j \in \mathbb{Z}, k \geq 0\}.

The usual product F1×F1F_1 \times F_1 (pr Z2=Z⊕Z\mathbb{Z}^2 = \mathbb{Z} \oplus \mathbb{Z} in additive notation) is not free; its generators a,ba,b are subject to the relation ab=baab=ba or aba−1b−1=eaba^{-1}b^{-1} = e, making it abelian.

Group Presentations

An element aba−1b−1aba^{-1}b^{-1} of a group is called the commutator of a,ba,b and is sometimes denoted [a,b][a,b]. We can write:

F1×F1=⟨a,b∣aba−1b−1=e⟩F_1 \times F_1 = \langle a,b | aba^{-1}b^{-1} = e \rangle

or more succinctly:

⟨a,b∣aba−1b−1⟩,\langle a,b | aba^{-1}b^{-1} \rangle,

in which a,ba,b are the generators, aba−1b−1=eaba^{-1}b^{-1}=e is a relation and aba−1b−1aba^{-1}b^{-1} is a relator. The latter spawns other relators by taking its inverse and conjugates:

[b,a]=[a,b]−1,[a,b−1]=b−1[b,a]b, etc.[b,a] = [a,b]^{-1},\\ [a,b^{-1}] = b^{-1}[b,a]b, \ \text{etc.}

Example: The dihedral group D3D_3 of symmetries of △\triangle is also the symmetric group S3S_3 of permutations of {1,2,3}\{1,2,3\}. It is generated (for example) by a 33-cycle a=(123)a=(123) and 22-cycle b=(12)b=(12), subject to the equation bab−1=a2bab^{-1}=a^2 or baba=ebaba=e. Thus:

S3≅D3=⟨a,b∣a3,b2,abab⟩.S_3 \cong D_3 = \langle a,b | a^3,b^2,abab \rangle.

Any group GG can be characterised by a number of generators (letters) a1,...,ana_1,...,a_n and a number of relators (words) R1,...,RmR_1,...,R_m giving rise to a so-called presentation:

G=⟨a1,...,an∣R1,...,Rm⟩.G = \langle a_1,...,a_n | R_1,..., R_m \rangle.

Normal Subgroups

The group G=⟨a1,...,an∣R1,...Rm⟩G = \langle a_1,...,a_n | R_1,...R_m \rangle is then a quotient of a free-group FF by a homomorphism whose kernel NN is the smallest normal subgroup containing the RiR_i. (Recall that a subgroup is normal if it contains all its conjugates and that G≅F/NG \cong F/N). Any element of NN can be simplified to the identity, using the relations.

Theorems [Nielsen-Schreier, proofs use covering spaces of graphs]:

  1. Any subgroup of a free group is free, but in general the RiR_i will not generate NN.
  2. If Fn/NF_n/N is finite of size ii then N≅Fi(n−1)+1N \cong F_{i(n-1)+1}.

Examples:

  1. For S3S_3, take j=3j=3 and R1=a3R_1=a^3, R2=b2R_2 = b^2, R3=ababR_3 = abab. Then F2/N≅S3F_2/N \cong S_3, so NN is a subgroup of index i=6i=6 in F2F_2, and N≅F7N \cong F_7. A graph-theoretic algorithm implies that:
    N=⟨a3,b2,abab,b−1a3b,ab−1ab,a−1ba−1b,b−1aba⟩.N = \langle a^3,b^2, abab,b^{-1}a^3b,ab^{-1}ab,a^{-1}ba^{-1}b,b^{-1}aba \rangle.
  2. The kernel of the homomorphism F2→F1×F1F_2 \rightarrow F_1 \times F_1 is the normal subgroup generated by all the commutators xyx−1y−1xyx^{-1}y^{-1} with x,y∈F2x,y \in F_2, and F1×F1≅F2/[F2,F2]F_1 \times F_1 \cong F_2 / [F_2,F_2] is the abelisanisation of F2F_2.

Simplifying Relations

Let KK be a (left or right) trefoil knot in R3\mathbb{R}^3. The so-called Wirtinger presentation of G=π1(R3−K)G = \pi_1(\mathbb{R}^3 - K) is given by:

G=⟨α,β,γ∣α−1βγ−1,β−1γβα−1,γ−1αγβ−1⟩,G = \langle \alpha,\beta,\gamma | \alpha^{-1}\beta \gamma^{-1}, \beta^{-1}\gamma \beta \alpha^{-1}, \gamma^{-1}\alpha \gamma \beta^{-1} \rangle,

with one relation for each crossing. This is explained with some previous work, however we are only concerned with the group theory here.

In GG we have α−1βα=γ=βαβ−1\alpha^{-1}\beta \alpha = \gamma = \beta \alpha \beta^{-1}, which implies the third equation γ−1αγ=β\gamma^{-1}\alpha \gamma = \beta. So we can dispense with γ\gamma, and observe that:

α−1βα=βαβ−1  ⟹  αβα=βαβ.\alpha^{-1}\beta \alpha = \beta \alpha \beta^{-1} \implies \alpha \beta \alpha = \beta \alpha \beta.

Set a=αβa = \alpha \beta and b=αβαb = \alpha \beta \alpha. Then a3=b2a^3 = b^2. Since α=a−1b\alpha = a^{-1}b and β=b−1a2\beta = b^{-1}a^2,

G=⟨a,b∣a3b−2⟩.G = \langle a,b | a^3b^{-2} \rangle.

This presentation reflects the realisation of a trefoil knot wrapping around a torus.