Exact Sequences with a flat cokernel and a sketch of properties of Tor

Exact Sequences with a flat cokernel and a sketch of properties of Tor We give two arguments to prove the following fact: Lemma. Let 0 → N → M → F → 0 be an exact sequence of R-modules such that F is R-flat. Then for every R-module A, 0 → A ⊗R N → A ⊗R M → A ⊗R F → 0 is exact. The first proof is completely elementary. The second uses properties of Tor, and is quite straightforward if one is familiar with Tor. Prior to giving the argument, we give a brief treatment of properties of Tor. Elementary proof. Map a free R-module G onto A, and let B denote the kernel, so that 0 → B → G → A → 0 is exact. Think of this exact sequence as written as a column. We may tensor the row and column together to obtain a double array (all tensor products are over R) in which the squares commute: 0   y B ⊗ N −−−−→ B ⊗ M −−−−→ B ⊗ F −−−−→ 0      α y y y g 0 −−−−→ G ⊗ N −−−−→ G ⊗ M −−−−→ G ⊗ F −−−−→ 0       y y y f A ⊗ N −−−−→ A ⊗ M −−−−→ A ⊗ F −−−−→ 0       y y y 0 0 0 By the right exactness of tensor product, the rows are exact except at the left and the columns are exact except at the top. However, because G is free, the middle row is exact, and because F is flat, the rightmost column is exact. We want to show that f is injective. Suppose that u ∈ Ker (f ). Pick v ∈ G ⊗ N that maps to u. Let w be ithe image of v in G ⊗ M . Then w maps to 0 in A ⊗ M , since the image of w is the image of v, and if we use the maps forming the other two sides of the square, we get v 7→ u 7→ f (u) = 0. Hence, w is the image of some x ∈ B ⊗ M . Let y be the image of w in B ⊗ F . We claim that y = 0. Since α is injective, it suffices to show that the image z of y in G ⊗ F is 0. If we use the other two sides of the square, we see that this is the same as the image of w. Since w is the image of v, this is 0. Since y = 0, x is the mage of an element x0 in B ⊗ N . Let v 0 be the image of x in G ⊗ N . Then v and v 0 both map to w in G ⊗ M , so that v − v 0 7→ 0. Since g is injective, we have that v = v 0 . But then u is the image of v 0 , and so it is the image of x0 . Hence, u = 0.  1 2 For those familiar with Tor, we give a different argument. First note that when C, D are R-modules the R-modules TorR i (C, D) are defined for all integers i. The superscript R is frequently omitted when the base ring is understood from context. For negative i, they vanish, while Tor0 (C, D) = C ⊗ D. There is a canonical isomorphism Tori (C, D) ∼ = Tori (D, C) that generalizes the symmetry of the tensor product. When D is flat and, hence, when C is flat, Tori (C, D) = 0 for all i ≥ 1. Moreover, if 0 → C 0 → C → C 00 → 0 is exact, there is a (functorial) long exact sequence: · · · → Tori (C 0 , D) → Tori (C, D) → Tori (C 00 , D) → Tori−1 (C 0 , D) → · · · → Tor1 (C 0 , D) → Tor1 (C, D) → Tor1 (C 00 , D) → C 0 ⊗ D → C ⊗ D → C 00 ⊗ D → 0. We only need the rightmost four terms of the long exact sequence, which we may utilize in the case where C 00 = F , C 0 = N , C = M , and D = A. The conclusion we want is immediate.  We make the following further comments about Tor. If C is held fixed, Tori (C, ) is a covariant functor form R-modules to R-modules. The same holds for Tori ( , D) when D is held fixed. If R is Noetherian and C, D are Noetherian, so are all the modules Tori (C, D). If r ∈ R, the map given by multiplication by r on C induces the map given by multiplication by r on Tori (C, D), and the same applies for the map given by multiplciation by r on D. It follows that Tori (C, D) is killed by the sum of the annihilators of the modules C and D. One way of defining Tor is as follows. If A is an R-module, one can construct a left free resolution of A as follows. Map a free R-module onto A, call it P0 . Then map a free Rmodule P1 onto Ker (P0  A). This yields a sequence P1 → P0 → A → 0 exact except at P1 . Recursively, if one has an exact sequence Pi → Pi−1 → · · · P1 → P0 → A → 0, one may extend the resolution further by mapping a free module Pi+1 onto Ker (Pi → Pi−1 ). This yields a (very possibly infinite) free resolution · · · → Pn → Pn−1 · · · → P1 → P0 → A → 0 which is exact. Let P• be the sequence obtained by replacing A by 0. This sequence is exact except at P0 . It is still a complex, i.e., the composition of any two consecutive maps is 0. One gets a new complex P• ⊗R D by applying ⊗R D, namely · · · → Pn ⊗R D → Pn−1 ⊗R · · · → P1 ⊗R → P0 ⊗R → 0. Given any complex G• , say · · · Gn → Gn−1 · · · we may define its homology Hn (G• ) at the n th spot as Zn /Bn , where Zn = ker(Gn → Gn−1 ) and Bn = Im (Gn+1 → Gn . Then Torn (C, D) = Hn (P• ⊗R D). This turns out to be independent of the free resolution up to canonical isomorphism. In fact, one may use any projective resolution or even a resolution by flat modules to calculuate Tor.