Localizing a sequence of modules at a multiplicative set preserves exactness.
Exactness of localization
Localization is not just a way to invert elements in a ring (localization of rings)
; it is also a well-behaved operation on modules (localization of modules)
. One of its most important formal properties is that it preserves exact sequences.
S−1(imh)=im(S−1h),
so localization commutes with kernels and images.
One convenient conceptual reformulation is that localization is extension of scalars
along the ring map R→S−1R:
S−1M≅M⊗RS−1R,
and this tensor description is what makes the exactness behave so cleanly.
Examples
Localizing a short exact sequence over Z. Consider
0→Z×nZ→Z/nZ→0.
Localize at S=Z∖(p) (so S−1Z=Z(p)). Exactness says
0→Z(p)×nZ(p)→(Z/nZ)(p)→0
is exact. Concretely:
if p∤n, then multiplication by n becomes an isomorphism on Z(p), so (Z/nZ)(p)=0;
if n=pkm with p∤m, then m becomes invertible in Z(p), and the cokernel is “the p-primary part,” isomorphic to Z/pkZ as a Z(p)-module.
Localization can kill torsion. Let R=k[x] and consider the exact sequence
0→R×xR→R/(x)→0.
Localize at S={1,x,x2,…}, so S−1R=Rx. Then x becomes a unit in Rx, hence multiplication by x on Rx is an isomorphism. Exactness forces
(R/(x))x=0,
which reflects that R/(x) is “x-torsion.”
Exactness on kernels and images. Let R=Z, h:Z→Z be multiplication by 6, and localize at S={1,2,22,…}. Then kerh=0 and exactness implies ker(S−1h)=0 as well. Meanwhile imh=6Z localizes to 6Z[1/2], which equals the image of the localized map Z[1/2]×6Z[1/2].