Let G be a group and f a mapping of X into G. There exists one and only one homomorphism fˉ of F(X) into G which extends f.
The uniqueness of fˉ follows from the fact that the group F(X) is generated by X. For all x in X, let fx be the homomorphism n↦f(x)n of Z into G. By no. 3, Proposition 4, there exists a homomorphism fˉ of F(X) into G such that fˉ(xn)=fx(n) for x∈X and n∈Z; in particular, fˉ(x)=fx(1)=f(x) for all x∈X and hence fˉ extends f.
Let u:X→Y be a mapping. By Proposition 8 there exists one and only one homomorphism of F(X) into F(Y) which coincides with u on X; it is denoted by F(u). As in the case of magmas (no. 1) the formula
F(v∘u)=F(v)∘F(u)
is established for every mapping v:Y→Z and it is shown that F(u) is injective (resp. surjective, bijective) if u is. For every subset S of X, F(S) will be identified with the subgroup of F(X) generated by S.
Let I be a set. In certain cases it is of interest not to identify i in I with its canonical image ϕi(1) in the free group F(I); the latter will be denoted by Ti (or Ti′,Xi,… as the case may be) and called the indeterminate of index i. The free group F(I) is then denoted by F((Ti)i∈I) or F(T1,…,Tn) if I={1,2,…,n}.
Let G be a group and t=(ti)i∈I a family of elements of G. By Proposition 8 there exists a homomorphism ft of F((Ti)i∈I) into G characterized by ft(Ti)=ti for all i∈I. The image of an element w of F((Ti)i∈I) under ft will be denoted by w(t) or w(t1,…,tn) if I={1,2,…,n}; w(t) is said to result from the substitution Ti↦ti in w. In particular, if we take G=F((Ti)i∈I) and (ti)=(Ti)=T,fT is the identity homomorphism of G, whence w(T)=w; for I={1,2,…,n}, then w(T1,…,Tn)=w.
Let G and G′ be two groups, u a homomorphism of G into G′ and t=(t1,…,tn) a finite sequence of elements in G. Let t′=(u(t1),…,u(tn)); the homomorphism u∘ft of F(T1,…,Tn) into G′ maps Ti to u(ti) for 1≤i≤n and hence is equal to ft′; for w in F(T1,…,Tn), then
u(w(t1,…,tn))=w(u(t1),…,u(tn)).
Let w be given in F(T1,…,Tn) and elements v1,…,vn in the free group F(T1′,…,Tm′). The substitution Ti↦vi defines an element w′=w(v1,…,vn) of F(T1′,…,Tm′). Let G be a group, t1,…,tm elements of G and u the homomorphism of F(T1′,…,Tm′) into G characterized by u(Tj′)=tj for 1≤j≤m. Then u(vi)=vi(t1,…,tm) and u(w′)=w(t1,…,tm); formula (8) thus implies
w′(t1,…,tm)=w(v1(t1,…,tm),…,vn(t1,…,tm)).
This justifies the "functional notation" w(t1,…,tn). The reader is left to extend formulae (8) and (9) to the case of arbitrary indexing sets.
Bourbaki puts the proof directly after the statement, so what is above is the statement and its proof together.