Bourbaki

Theorem 1

Tag 0DC6, permanent. Cite it as [Bourbaki, Tag 0DC6].

Commutative Algebra, chapter X, § 9. Modules dualisants

Let MM be a finitely generated AA-module. Then αM\alpha_M is a homologism: we have Hi(D(D(M)))=0H_i(D(D(M))) = 0 for i≠0i \neq 0 and the homomorphism αM\alpha_M is bijective.

Let us first take M=AM = A. The mapping e:Ω→D(A)e : \Omega \to D(A) is a homomorphism (Example 1), hence so also is the mapping D(e):D(D(A))→D(Ω)D(e) : D(D(A)) \to D(\Omega). The mapping e~:A→D(Ω)\tilde{e} : A \to D(\Omega) is a homomorphism (Example 2), and we have D(e)∘αA=e~D(e) \circ \alpha_A = \tilde{e}; thus αA\alpha_A is a homomorphism, which proves the theorem in this case. It follows that αM\alpha_M is a homomorphism when the AA-module MM is a finitely generated free module.

Let us pass to the general case; we shall prove by induction on the integer nn the following assertion:

(An)(A_n) for every finitely generated AA-module MM, the homomorphism Hi(αM)H_i(\alpha_M) is bijective for i≤ni \leq n.

This also means that Hi(D(D(M)))H_i(D(D(M))) is zero for i≠0i \neq 0 and i≤ni \leq n, and that αM\alpha_M is bijective if n≥0n \geq 0. Observe that (An)(A_n) holds for n<−dn < -d, where dd is the length of the complex II; in fact the AA-module D(D(M))iD(D(M))_i is equal to ⨁pHomA(HomA(M,Ip),Ip−i)\bigoplus_p \mathrm{Hom}_A(\mathrm{Hom}_A(M, I^p), I^{p-i}), and is therefore zero for i<−di < -d and i>di > d.

Let us prove the implication (An)⇒(An+1)(A_n) \Rightarrow (A_{n+1}). Let MM be a finitely generated AA-module. There exists a finitely generated free AA-module LL and an exact sequence 0→N→uL→vM→00 \to N \xrightarrow{u} L \xrightarrow{v} M \to 0.

The sequence 0→D(M)→D(v)D(L)→D(u)D(N)→00 \to D(M) \xrightarrow{D(v)} D(L) \xrightarrow{D(u)} D(N) \to 0 is exact; analogously, if we put u′=D(D(u))u' = D(D(u)) and v′=D(D(v))v' = D(D(v)), the sequence 0→D(D(N))→u′D(D(L))→v′D(D(M))→00 \to D(D(N)) \xrightarrow{u'} D(D(L)) \xrightarrow{v'} D(D(M)) \to 0 is exact.

Since Hi(D(D(L)))H_i(D(D(L))) is zero for i≠0i \neq 0, we have isomorphisms

Hi(D(D(M))))⟶Hi−1(D(D(N)))for i≠0,1; H_i(D(D(M)))) \longrightarrow H_{i-1}(D(D(N))) \quad \text{for } i \neq 0, 1 ;

this yields the implication (An)⇒(An+1)(A_n) \Rightarrow (A_{n+1}) for n≠−1n \neq -1 and n≠0n \neq 0. Consider the commutative diagram with exact rows

0→N→uL→vM→0↓αN↓αL↓αM0→H1(D(D(M)))⟶H0(D(D(N)))→H0(u′)H0(D(D(L)))→H0(v′)H0(D(D(M)))⟶H−1(D(D(N))) \begin{array}{ccccccccc} 0 & \to & N & \xrightarrow{u} & L & \xrightarrow{v} & M & \to & 0 \\ & & \downarrow{\alpha_N} & & \downarrow{\alpha_L} & & \downarrow{\alpha_M} & & \\ 0 & \to & H_1(D(D(M))) & \longrightarrow & H_0(D(D(N))) & \xrightarrow{H_0(u')} & H_0(D(D(L))) & \xrightarrow{H_0(v')} & H_0(D(D(M))) \longrightarrow H_{-1}(D(D(N))) \end{array}

where αL\alpha_L is bijective. If (A0)(A_0) holds, the homomorphism αN\alpha_N is likewise bijective, hence H0(u′)H_0(u') is injective and one obtains H1(D(D(M)))=0H_1(D(D(M))) = 0, whence (A1)(A_1). If (A−1)(A_{-1})

Thus (An)(A_n) is true for every nn, which proves the theorem.

Let MM be a finitely generated AA-module; put c=dim⁡(A)−dim⁡A(M)c = \dim(A) - \dim_A(M). Let D′(M)D'(M) denote the subcomplex of D(M)D(M) equal to ⨁i<cD(M)i⨁Zc(D(M))\bigoplus_{i < c} D(M)^i \bigoplus Z^c(D(M)), and

j:D′(M)→D(M) j : D'(M) \to D(M)

the canonical injection. From the canonical surjection Zc(D(M))→Hc(D(M))Z^c(D(M)) \to H^c(D(M)) and the isomorphism φ(M,I)\varphi(M, I) one deduces a morphism of complexes

p:D′(M)(−c)→ExtAc(M,Ω) p : D'(M)(-c) \to \mathrm{Ext}_A^c(M, \Omega)

since Hi(D(M))H^i(D(M)) is zero for i<ci < c (No. 1, prop. 3 a)), (D′(M)(−c),p)(D'(M)(-c), p) is a left resolution of ExtAc(M,Ω)\mathrm{Ext}_A^c(M, \Omega). By A, X, p. 100, th. 1, one has a canonical isomorphism

φ0(D′(M)(−c),I):H0(D(D′(M)))→ExtAc(ExtAc(M,Ω),Ω) \varphi^0(D'(M)(-c), I) : H_0(D(D'(M))) \to \mathrm{Ext}_A^c(\mathrm{Ext}_A^c(M, \Omega), \Omega)

by composing it with the homomorphism H0(D(j)):H0(D(D(M)))→H0(D(D′(M)))H_0(D(j)) : H_0(D(D(M))) \to H_0(D(D'(M)))

one therefore obtains a homomorphism

H0(D(D(M)))→ExtAc(ExtAc(M,Ω),Ω), H_0(D(D(M))) \to \mathrm{Ext}_A^c(\mathrm{Ext}_A^c(M, \Omega), \Omega),

whence finally, by composition with αM\alpha_M, a canonical homomorphism

βM:M→ExtAc(ExtAc(M,Ω),Ω). \beta_M : M \to \mathrm{Ext}_A^c(\mathrm{Ext}_A^c(M, \Omega), \Omega).

Bourbaki puts the proof directly after the statement, so what is above is the statement and its proof together.

Cited by

Nothing.

Cites

Nothing.

Translated with machine help, not yet reviewed by a person.