Suppose that in the commutative diagram (10) the sequences (u,v) and (u′,u′) are exact. Then:
(i) v1∘u1=0; if u′ is injective, the sequence (u1,v1) is exact. (ii) v2∘u2=0; if v is surjective, the sequence (u2,v2) is exact. (iii) Suppose that u′ is injective and v is surjective. Then there exists one and only one homomorphism d:Ker(c)→Coker(a) with the following property: if x∈Ker(c), y∈B and t′∈A′ satisfy the relations v(y)=k(x) and u′(t′)=b(y), then d(x)=p(t′). Moreover the sequence
Let us prove (i). As u1 and v1 have the same graphs as the restrictions of u and v to Ker(a) and Ker(b) respectively, we have v1∘u1=0. Then
Ker(v1)=Ker(b)∩Ker(v)=Ker(b)∩Im(u)=Im(j)∩Im(u).
But, by Proposition 1, (i), Ker(v1)=Im(j∘u1)=Im(u1) if u′ is injective.
Let us prove (ii). As u2 and v2 are obtained from u and v by taking quotients, it is clear that v2∘u2=0. Suppose that v is surjective; as q and p are surjective, it follows from the hypotheses and Proposition 1, (ii) that
Finally let us prove (iii). For x∈Ker(c) there exists y∈B such that v(y)=k(x) since v is surjective; moreover v′(b(y))=c(k(x))=0 and consequently there exists a uniquet′∈A′ such that u′(t′)=b(y), since u′ is injective. We now show that the element p(t′)∈Coker(a) is independent of the element y∈B such that v(y)=k(x). For if y′∈B is another element such that v(y′)=k(x), then y′=y+u(z), where z∈A; we show that if t′′∈A′ is such that u′(t′′)=b(y′) then t′′=t′+a(z); for
Finally, it follows that p(t′′)=p(t′)+p(a(z))=p(t′). Then it is possible to set d(x)=p(t′) and the mapping d:Ker(c)→Coker(a) has thus been defined.
If now x1,x2 are elements of Ker(c) and x=x1+x2, we take elements y1,y2 of B such that v(y1)=k(x1) and v(y2)=k(x2) and choose for y∈B the element y1+y2; it is then immediate that d(x)=d(x1)+d(x2) and hence d is a homomorphism.
Suppose that x=v1(x′) for some x′∈Ker(b); then take y∈B to be the element j(x′). As b(j(x′))=0, it follows that d(x)=0, hence d∘v1=0. Conversely, suppose that d(x)=0. In the above notation we then have t′=a(s), where s∈A. In this case we have b(y)=u′(t′)=u′(a(s))=b(u(s)), or b(y−u(s))=0. The element y−u(s) is therefore of the form j(n), where n∈Ker(b), and we have k(x)=v(y)=v(u(s)+j(n))=v(j(n))=k(v1(n)); as k is injective, x=v1(n), which proves that the sequence (*) is exact at Ker(c).
Finally, we have (always in the same notation)
u2(d(x))=u2(p(t′))=q(u′(t′))=q(b(y))=0
and hence u2∘d=0. Conversely, suppose that an element w=p(t′) of Coker(a) is such that u2(w)=u2(p(t′))=0 (where t′∈A). Then q(u′(t′))=0 and consequently u′(t′)=b(y) for some y∈B; as v′(u′(t′))=0, we have v′(b(y))=0, hence c(v(y))=0, in other words v(y)=k(x) for some x∈Ker(c), and by definition w=d(x), which shows that the sequence (*) is exact at
Coker(a). It has been seen in (i) that it is exact at Ker(b) and in (ii) it is exact at Coker(b), which completes the proof of (iii).
Bourbaki puts the proof directly after the statement, so what is above is the statement and its proof together.