2024-05-06
Algebra-I
00

Contents

§ 6 Free Groups
§ 6.1 Words and Letters
§ 6.2 Reduction of Words
§ 6.3 Definition of the Free Group
§ 6.4 Equivalence Relations
§ 6.5 Existence of a Unique Reduction
§ 6.6 Applications of Free Groups

§ 6 Free Groups

One way to define a group is to specify a collection of generators together with a collection of relations satisfied by those generators.

Question: What does a group with a set of generators but no relations look like? If the set of generators is SS, such a group is called the free group on SS.

Example 6.1 g=image(ZG) ngn\begin{aligned} \textbf{Example 6.1}~\langle g\rangle=\mathrm{image}(\mathbb{Z}&\to G)\ n&\mapsto g^{n} \end{aligned}

We are going to generalize Z\mathbb{Z}. An analogous description of Z\mathbb{Z} is

 g GZG.\begin{array}{rcl} *&\xrightarrow{~g~}&G\\ &\Downarrow&\\ \mathbb{Z}&\longrightarrow&G. \end{array}

For any map of sets

SG,S\to G,

we will obtain a group homomorphism

F(S)G,F(S)\to G,

where F(S)F(S) is the free group on the set SS.

§ 6.1 Words and Letters

Definition 6.1 Let XX be a set. A word on XX is a finite ordered collection of elements of XX. Given a word ww, an element of ww is called a letter of ww.

Equivalently, a word is:

  • an element of
Xnfor some n0X^{n}\tag*{for some $n\geqslant 0$}
  • a sequence
X1X2Xn1XnX_{1}X_{2}\cdots X_{n-1}X_{n}

where each XiXX_i\in X and n0n\geqslant0.

Remark The empty word is the word containing no letters.

We write Word(X)\mathrm{Word}(X) for the set of all words on XX.

Example 6.2 If X=aX={a},

Word(X)={,a,aa,aaa,}.\mathrm{Word}(X)=\{\varnothing,a,aa,aaa,\ldots\}.

If X=a,bX={a,b},

Word(X)={,a,b,ab,ba,aa,bb,}.\mathrm{Word}(X)=\{\varnothing,a,b,ab,ba,aa,bb,\ldots\}.

Given two words in XX, we may concatenate them to produce a new word:

Word(X)×Word(X)Word(X)(w1,w2)w1w2.\begin{aligned} \mathrm{Word}(X)\times\mathrm{Word}(X)&\to \mathrm{Word}(X)\\ (w_{1},w_{2})&\mapsto w_{1}w_{2}. \end{aligned}

Example 6.3

(ba,ab)baab(ab,ba)abba.\begin{aligned} (ba,ab)&\mapsto baab\\ (ab,ba)&\mapsto abba. \end{aligned}

This is clearly associative, and the empty word looks like an identity element. But there are no inverses! Let us fix this.

Given a set

S={a,b,c},S=\{a,b,c\},

let SS^{\prime} be the set of symbols

S={a1,b1,c1,}.S^{\prime}=\{a^{-1},b^{-1},c^{-1},\ldots\}.

Thus SS and SS^{\prime} are in bijection.

We define

S=SS={a,a1,b,b1,}.\begin{aligned} \overline{S} &=S\cup S^{\prime}\\ &=\{a,a^{-1},b,b^{-1},\ldots\}. \end{aligned}

Example 6.4 A word in S\overline{S} might look like

w=babb1a1c1ca.w=babb^{-1}a^{-1}c^{-1}ca.

§ 6.2 Reduction of Words

Definition 6.2 A word in S\overline{S} is called unreduced if, for some aSa\in S, one of the strings

aa1 or a1aaa^{-1}~\text{or}~a^{-1}a

occurs in the word. If a word is not unreduced, it is called reduced.

Example 6.5

aaab1a1b is reducedacbcb1bc1acbcbb1c1}are both unreduced.\begin{aligned} aaab^{-1}a^{-1}b~&\text{is reduced}\\ \left. \begin{aligned} acbcb^{-1}bc^{-1}\\ acbcbb^{-1}c^{-1} \end{aligned} \right\} &\text{are both unreduced}. \end{aligned}

Definition 6.3 If ww^{\prime} is obtained from ww by removing, or cancelling, one occurrence of aa1aa^{-1} or a1aa^{-1}a, then we say that ww^{\prime} is obtained from ww by cancellation. We write

ww.w\rightsquigarrow w^{\prime}.

Example 6.6

  • The empty word is obtained by cancellation from b1bb^{-1}b and from bb1bb^{-1}.

  • abab is obtained by cancellation from

abb1bc1cabacc1b.\begin{align*} abb^{-1}b \tag*{we may remove $a\underline{bb^{-1}}b$ or $ab\underline{b^{-1}b}$}\\ c^{-1}cab\\ acc^{-1}b. \end{align*}

Definition 6.4 If ww^{\prime} is obtained from ww through a sequence of cancellations,

ww,w\rightsquigarrow\cdots\rightsquigarrow w^{\prime},

and ww^{\prime} is reduced, then ww^{\prime} is called a reduction of ww.

Proposition 6.1 If ww^{\prime} and ww^{\prime\prime} are reductions of ww, then

w=w.w^{\prime}=w^{\prime\prime}.

Proof: We use induction on the length ll of the word ww. Notice that

wulength(u)<length(w).w\rightsquigarrow u \quad\Rightarrow\quad \mathrm{length}(u)<\mathrm{length}(w).

When l=0l=0: the empty word is reduced.

When l=1l=1: the word contains only one element, so it is impossible for aa and a1a^{-1} to occur next to each other. Thus every word of length 11 is reduced.

Assume that every word of length l1l-1 has a unique reduction. We prove the result for words of length ll.

If ww has length ll and is already reduced, we are done.

Otherwise, somewhere in ww there is an occurrence of aa1aa^{-1} or a1aa^{-1}a. There may be several such occurrences.

For example, consider

w=a1aa1aa1a,w=a^{-1}aa^{-1}aa^{-1}a,

which has length 66. Choose one occurrence

a1a.\cdots\underline{a^{-1}a}\cdots.

A reduction of ww may be obtained in one of the following ways:

(i) At some stage, cancel the chosen a1a\underline{a^{-1}a}.

(ii) Never cancel the chosen a1a\underline{a^{-1}a}.

Case (ii) can occur only in the following situations:

In (6.1), cancelling a1aa1\underline{a^{-1}a}a^{-1} using either or produces the same word.

The same is true for (6.2). Therefore, we may assume that the chosen a1a\underline{a^{-1}a} is cancelled at some stage. Any reduction obtained by (ii) can also be obtained by (i).

Thus, we have a reduction

Thus, whether we cancel aa1aa^{-1} at Step 1 or at some later stage, we obtain the same reduction.

 ~\tag*{$\square$}

§ 6.3 Definition of the Free Group

Definition 6.5 Let SS be a set. The free group F(S)F(S) on SS is

(F(S),m)(F(S),m)

where

  • F(S)F(S) is the set of reduced words in
S=SS,\overline{S}=S\cup S^{\prime},

where

S:={w1wS};S^{\prime}:=\{w^{-1}\mid w\in S\};
  • the map
m:F(S)×F(S)F(S)m:F(S)\times F(S)\to F(S)

sends (w1,w2)(w_1,w_2) to the reduction of w1w2w_1w_2.

Example 6.7 Consider a set with one element,

S={a}.S=\{a\}.

A word in S\overline{S} might look like

aaaa1aa1aaa1a.aaaa^{-1}aa^{-1}aaa^{-1}a.

If a word is reduced, it looks like

aaaaaan times\underbrace{aaaaa\cdots a}_{n~\text{times}}

or

a1a1a1n times.\underbrace{a^{-1}a^{-1}\cdots a^{-1}}_{n~\text{times}}.

Therefore,

ZF(S)n{aan timesn>0a1a1n timesn<0n=0\begin{aligned} \mathbb{Z}&\to F(S)\\ n&\mapsto \begin{cases} \underbrace{a\cdots a}_{n~\text{times}}&n>0\\ \underbrace{a^{-1}\cdots a^{-1}}_{n~\text{times}}&n<0\\ \varnothing&n=0 \end{cases} \end{aligned}

is a bijection. It is an isomorphism.

Example 6.8 Consider a set with two elements,

S={a,b}.S=\{a,b\}.

Then F(S)F(S) is the free group on two generators. Its elements look like

l=0=:1l=1a,b,a1,b1l=2aa,bb,ba,ab,a1a1,b1b1,a1b1,b1a1.\begin{array}{ll} l=0 & \varnothing=:1 \\ l=1 & a,b,a^{-1},b^{-1}\\ l=2 & aa,bb,ba,ab,a^{-1}a^{-1},b^{-1}b^{-1},a^{-1}b^{-1},b^{-1}a^{-1}. \end{array}

Proposition 6.2 The inverse of S1SnS_1\cdots S_n is Sn1S11S_n^{-1}\cdots S_1^{-1}.

Proof: The inverse of S1SnS_1\cdots S_n is Sn1S11S_n^{-1}\cdots S_1^{-1} because

 ~\tag*{$\square$}

§ 6.4 Equivalence Relations

Definition 6.6 Let XX be a set. An equivalence relation on XX is a subset

RX×XR\subset X\times X

satisfying:

(1) For every xXx\in X,

(x,x)R.(x,x)\in R.

(2) If

(x,y)R,(x,y)\in R,

then

(y,x)R.(y,x)\in R.

(3) If

(x,y)R(x,y)\in R

and

(y,z)R,(y,z)\in R,

then

(x,z)R.(x,z)\in R.

We write

xyx\sim y

if

(x,y)R.(x,y)\in R.

Example 6.9 Let GG act on a set XX. Define

xyx\sim y

if and only if

y=gxy=gx

for some gGg\in G.

This is an equivalence relation because

(1)

x=1Gx,x=1_Gx,

so

xx.x\sim x.

(2)

xyy=gxx=g1yyxx\sim y \Rightarrow y=gx \Rightarrow x=g^{-1}y \Rightarrow y\sim x

for some gg.

(3)

yzz=gy,y\sim z \Rightarrow z=g^{\prime}y,

so

z=ggy,z=g^{\prime}gy,

and hence

zx.z\sim x.

Thus, for every xx, Ox\mathcal{O}_x is the equivalence class of xx:

Ox={yyx}.\mathcal{O}_x = \{y\mid y\sim x\}.

The orbit space X/GX/G is the set of equivalence classes.

Here is another example.

Example 6.10 Let SS be a set and define

S={x1}xSS=SS.\begin{aligned} S^{\prime}&=\{x^{-1}\}_{x\in S}\\ \overline{S}&=S\cup S^{\prime}. \end{aligned}

§ 6.5 Existence of a Unique Reduction

Theorem 6.3 Every

wWord(S)w\in\mathrm{Word}(\overline{S})

has a unique reduction.

Proof: We use induction on the length ll.

The case l=0l=0 is clear.

The case l=1l=1 is clear.

Assume that for every word ww^{\prime} of length l1\leqslant l-1, the set

{reductions of w}\{\text{reductions of }w^{\prime}\}

has exactly one element.

We must prove the same for every word ww of length ll.

  • If ww is already reduced, then no other word can be obtained from ww by cancellation. Therefore
{reductions of w} has exactly one element—w itself.\{\text{reductions of }w\} ~\text{has exactly one element—}w~\text{itself}.
  • Otherwise, somewhere in ww there is an occurrence of
aa1ora1a.aa^{-1}\quad\text{or}\quad a^{-1}a.

Fix one such occurrence:

w=aa1.w=\cdots\underline{aa^{-1}}\cdots.

We have underlined it.

Consider the following:

1\textcircled{\small 1} is an equality because if the chosen aa1\underline{aa^{-1}} is cancelled at Step NN, then we may instead cancel it first and then perform Steps 11 through N1N-1, obtaining the same reduction.

2\textcircled{\small 2} is either an equality or the set is empty, because not cancelling aa1\underline{aa^{-1}} means that at some stage one must perform a cancellation like or .

Together, 1\textcircled{\small 1} and 2\textcircled{\small 2} tell us that every reduction of ww can be obtained by first cancelling the chosen aa1\underline{aa^{-1}}.

But

w=aa1w^{\prime} = \cdots\underline{aa^{-1}}\cdots

is a word of length less than ll!

Moreover, every reduction of ww obtained by first cancelling aa1\underline{aa^{-1}} is a reduction of ww^{\prime}.

Therefore,

{reductions of wobtained by first cancelling  aa1}={reductions of w}=a set containing exactly one element.\left\{ \begin{aligned} &\text{reductions of }w\\ &\text{obtained by first cancelling }~\underline{aa^{-1}} \end{aligned} \right\} = \{\text{reductions of }w^{\prime}\} = \text{a set containing exactly one element}.
 ~\tag*{$\square$}

Example 6.11 If

S=,S=\varnothing,

then Word(S)\mathrm{Word}(S) is a set containing one element—the empty word, which has length zero.

Example 6.12 If

S=,S=\varnothing,

then

Free(S)={reduced words in  S=},\mathrm{Free}(S) = \{\text{reduced words in }~\overline{S}=\varnothing\},

which is the set containing the empty word.

Therefore, when S=S=\varnothing, Free(S)\mathrm{Free}(S) is a group containing exactly one element.

§ 6.6 Applications of Free Groups

Proposition 6.4 Let GG be a group. Suppose

j:SGj:S\to G

is a map of sets. Then it extends to a group homomorphism

F(S)G.F(S)\to G.

Proof: Let sSs\in S be an element of the set, and let j(s)Gj(s)\in G be its image in GG.

Let

jˉ:SG\bar{j}:\overline{S}\to G

be the function defined by

sj(s),s1j(s)1.s\mapsto j(s), \qquad s^{-1}\mapsto j(s)^{-1}.

We then define a function

ϕj:Word(S)G\phi_j:\operatorname{Word}(\overline{S})\to G

by sending any word

W=s1sl,W=s_1\ldots s_l,

where siSs_i\in\overline{S}, to

ϕj(s1)ϕj(s2)ϕj(sl).\phi_j(s_1)\cdot\phi_j(s_2)\cdots\phi_j(s_l).

We must prove that this gives a well-defined map on F(S)F(S) and that it is a homomorphism.

Indeed, if ww is a reduced word obtained from WW, then it is obtained by cancelling adjacent inverse pairs. On the other hand, whenever a letter ss occurs next to its inverse s1s^{-1} in WW, the corresponding product in GG contains

ϕj(s)\phi_j(s)

next to

ϕj(s1)=ϕj(s)1.\phi_j(s^{-1})=\phi_j(s)^{-1}.

Thus, if we cancel two inverse letters in the word WW to obtain a new word ww^{\prime}, then

ϕj(W)=ϕj(w).\phi_j(W)=\phi_j(w^{\prime}).

More explicitly, in a product of several elements of GG, removing an occurrence of

ϕj(s)ϕj(s)1\phi_j(s)\phi_j(s)^{-1}

or

ϕj(s)1ϕj(s)\phi_j(s)^{-1}\phi_j(s)

does not change the value of the product:

ϕj(s1)ϕj(s)ϕj(s)1ϕj(sl)=ϕj(s1)1Gϕj(sl)=ϕj(s1)ϕj(sl).\begin{aligned} \phi_j(s_1)\cdots\phi_j(s)\phi_j(s)^{-1}\cdots\phi_j(s_l) &= \phi_j(s_1)\cdots1_G\cdots\phi_j(s_l)\\ &= \phi_j(s_1)\cdots\phi_j(s_l). \end{aligned}

The same applies when cancelling an occurrence of ϕj(s)1ϕj(s)\phi_j(s)^{-1}\phi_j(s).

Thus, if the words wiw_i^{\prime}, i=1,,Ii=1,\ldots,I, are the intermediate words appearing while reducing WW to its reduced word ww, then

ϕj(W)=ϕj(w1)==ϕj(wI)=ϕj(w).\phi_j(W) = \phi_j(w_1^{\prime}) = \cdots = \phi_j(w_I^{\prime}) = \phi_j(w).

This shows that ϕj\phi_j is well defined on F(S)F(S).

To prove that ϕj\phi_j defines a homomorphism, let

W=wwW=w^{\prime}\cdot w^{\prime\prime}

be the concatenation of two words, and let ww be its reduction. We must prove

ϕj(w)=ϕj(w)ϕj(w).\phi_j(w) = \phi_j(w^{\prime})\cdot\phi_j(w^{\prime\prime}).

By well-definedness, it is enough to prove

ϕj(W)=ϕj(w)ϕj(w).\phi_j(W) = \phi_j(w^{\prime})\cdot\phi_j(w^{\prime\prime}).

This is immediate because if

w=s1sl,w=s1sl,w^{\prime} = s_1^{\prime}\cdots s_{l^{\prime}}^{\prime}, \qquad w^{\prime\prime} = s_1^{\prime\prime}\cdots s_{l^{\prime\prime}}^{\prime\prime},

then

ϕj(W)=ϕj(s1)ϕj(sl)ϕj(s1)ϕj(sl)=(ϕj(s1)ϕj(sl))(ϕj(s1)ϕj(sl))=ϕj(w)ϕj(w).\begin{aligned} \phi_j(W) &= \phi_j(s_1^{\prime})\cdots \phi_j(s_{l^{\prime}}^{\prime}) \phi_j(s_1^{\prime\prime})\cdots \phi_j(s_{l^{\prime\prime}}^{\prime\prime})\\ &= \left( \phi_j(s_1^{\prime})\cdots \phi_j(s_{l^{\prime}}^{\prime}) \right) \left( \phi_j(s_1^{\prime\prime})\cdots \phi_j(s_{l^{\prime\prime}}^{\prime\prime}) \right)\\ &= \phi_j(w^{\prime})\cdot \phi_j(w^{\prime\prime}). \end{aligned}
 ~\tag*{$\square$}

Proposition 6.5 There is a bijection of sets

{group homomorphisms F(S)G}{maps of sets SG}.\{\text{group homomorphisms}~F(S)\rightarrow G\} \cong \{\text{maps of sets}~S\rightarrow G\}.

Proof: Above, for every function

j:SG,j:S\to G,

we defined a homomorphism

ϕj:F(S)G.\phi_j:F(S)\to G.

This defines a function

Φ:{maps of sets SG}{group homomorphisms F(S)G}\Phi: \{\text{maps of sets}~S\to G\} \to \{\text{group homomorphisms}~F(S)\to G\}

given by

jϕj.j\mapsto\phi_j.

We define an inverse map Ψ\Psi as follows.

If ϕ\phi is a group homomorphism, it assigns a value to every one-letter reduced word sSs\in S. Thus we define

Ψ(ϕ):=ψϕ\Psi(\phi):=\psi_\phi

to be the function sending

sϕ(s).s\mapsto\phi(s).

We must show that

ΨΦ\Psi\circ\Phi

and

ΦΨ\Phi\circ\Psi

are identity maps.

Indeed, given a homomorphism

ϕ:F(S)G,\phi:F(S)\to G,

let

w=s1sl,w=s_1\ldots s_l,

where each sis_i is a letter of ww lying in S\overline{S}.

By the group homomorphism property, together with the fact that every word is a product of one-letter words,

ϕ(w)=ϕ(s1sl)=ϕ(s1)ϕ(sl).\phi(w) = \phi(s_1\cdots s_l) = \phi(s_1)\cdots\phi(s_l).

Therefore, the values of ϕ\phi on one-letter words—that is, its values on SS—determine its value on every element of F(S)F(S). This shows that

ΦΨ\Phi\circ\Psi

is the identity.

On the other hand, Φ\Phi was defined so that

Φ(j)=ϕj\Phi(j)=\phi_j

simply sends a one-letter word ss to j(s)j(s). Therefore,

ΨΦ\Psi\circ\Phi

is also the identity.

 ~\tag*{$\square$}

Example 6.13 Let SS be a set and GG a group. For every function

SGS\to G

there exists a group homomorphism

F(S)G.F(S)\to G.

When

S=,S=\varnothing,

there is a unique function

ϕG.\phi\to G.

What is the corresponding group homomorphism

F(ϕ)G?F(\phi)\to G?

It sends the empty word to

1G.1_G.

Proposition 6.6 If w1w_1 and w2w_2 have the same reduced word, then

w1w2w_1\sim w_2

defines an equivalence relation.

Proof: w1w2w_1\sim w_2 is an equivalence relation because

(1)

www\sim w

is immediate:

reduction(w)=reduction(w).\mathrm{reduction}(w)=\mathrm{reduction}(w).

(2)

w1w2w2w1w_1\sim w_2 \Rightarrow w_2\sim w_1

is immediate because

reduction(w1)=reduction(w2)reduction(w2)=reduction(w1).\begin{aligned} &\mathrm{reduction}(w_1) = \mathrm{reduction}(w_2)\\ \Rightarrow\quad& \mathrm{reduction}(w_2) = \mathrm{reduction}(w_1). \end{aligned}

(3)

w1w2,w2w3w1w3w_1\sim w_2, \qquad w_2\sim w_3 \Rightarrow w_1\sim w_3

because

reduction(w1)=reduction(w2)reduction(w2)=reduction(w3)reduction(w1)=reduction(w3).\begin{aligned} \mathrm{reduction}(w_1)&=\mathrm{reduction}(w_2)\\ \mathrm{reduction}(w_2)&=\mathrm{reduction}(w_3) \end{aligned} \Rightarrow \mathrm{reduction}(w_1) = \mathrm{reduction}(w_3).

All of these follow from the uniqueness of reduction and from the fact that equality is an equivalence relation.

 ~\tag*{$\square$}

Remark Therefore, if

w1w2w_1\rightsquigarrow w_2

by cancellation, then

w1w2.w_1\sim w_2.

The converse need not hold.

Proposition 6.7 Let F(S)F(S) be the set of reduced words in S\overline{S}. There is a bijection

F(S){equivalence classes of words in  Word(S)}.F(S) \to \{\text{equivalence classes of words in }~ \mathrm{Word}(\overline{S})\}.

Proof: Send

w[w],w\mapsto[w],

i.e. send ww to its equivalence class.

Every equivalence class has a unique element of shortest length—namely the common reduced word of any

w[w].w^{\prime}\in[w].

This defines the inverse map.

 ~\tag*{$\square$}

Proposition 6.8 The operation

{equivalence classes of words}×{equivalence classes of words}concatenation{equivalence classes of words}([w1],[w2])[w1w2]\begin{aligned} \{\text{equivalence classes of words}\} \times \{\text{equivalence classes of words}\} &\xrightarrow{\text{concatenation}} \{\text{equivalence classes of words}\}\\ ([w_1],[w_2]) &\longmapsto[w_1w_2] \end{aligned}

is well defined.

Proof: Let r1r_1 and r2r_2 be the reduced words of w1w_1^{\prime} and w2w_2^{\prime}, respectively.

Notice that

r1r2 can be obtained by cancellation from w1w2.r_1r_2 ~\text{can be obtained by cancellation from}~ w_1^{\prime}w_2^{\prime}.

Simply perform all cancellations in the w1w_1^{\prime} part of the word, and then all cancellations in the w2w_2^{\prime} part.

Thus, for any

w1[w1],w2[w2],w_1^{\prime}\in[w_1], \qquad w_2^{\prime}\in[w_2],

we have

w1w2r1r2,w_1^{\prime}w_2^{\prime}\sim r_1r_2,

and hence

[w1w2]=[r1r2].[w_1^{\prime}w_2^{\prime}] = [r_1r_2].

Therefore, regardless of which representatives

w1[w1],w2[w2]w_1^{\prime}\in[w_1], \qquad w_2^{\prime}\in[w_2]

we choose, the equivalence class of w1w2w_1^{\prime}w_2^{\prime} does not change.

 ~\tag*{$\square$}

Corollary 6.9 The free-group operation

F(S)×F(S)F(S)(r1,r2)reduction(r1r2)\begin{aligned} F(S)\times F(S)&\to F(S)\\ (r_1,r_2)&\mapsto\mathrm{reduction}(r_1r_2) \end{aligned}

is associative.

Proof:

Therefore, it is enough to prove that the operation

([r1],[r2])[r1r2]([r_1],[r_2]) \mapsto [r_1r_2]

is associative.

Indeed,

([r1][r2])[r3]=[r1r2][r3]=[(r1r2)r3]=[r1(r2r3)]=[r1][r2r3]=[r1]([r2][r3]),([r_1][r_2])[r_3] = [r_1r_2][r_3] = [(r_1r_2)r_3] = [r_1(r_2r_3)] = [r_1][r_2r_3] = [r_1]([r_2][r_3]),

where the third equality follows from associativity of ordinary word concatenation.

 ~\tag*{$\square$}