Lemma 47.16.1. Let $(A, \mathfrak m, \kappa ) \to (B, \mathfrak m', \kappa ')$ be a finite local map of Noetherian local rings. Let $\omega _ A^\bullet $ be a normalized dualizing complex. Then $\omega _ B^\bullet = R\mathop{\mathrm{Hom}}\nolimits (B, \omega _ A^\bullet )$ is a normalized dualizing complex for $B$.

## 47.16 Dualizing complexes over local rings

In this section $(A, \mathfrak m, \kappa )$ will be a Noetherian local ring endowed with a dualizing complex $\omega _ A^\bullet $ such that the integer $n$ of Lemma 47.15.12 is zero. More precisely, we assume that $R\mathop{\mathrm{Hom}}\nolimits _ A(\kappa , \omega _ A^\bullet ) = \kappa [0]$. In this case we will say that the dualizing complex is *normalized*. Observe that a normalized dualizing complex is unique up to isomorphism and that any other dualizing complex for $A$ is isomorphic to a shift of a normalized one (Lemma 47.15.5).

**Proof.**
By Lemma 47.15.8 the complex $\omega _ B^\bullet $ is dualizing for $B$. We have

by Lemma 47.13.1. Since $\kappa '$ is isomorphic to a finite direct sum of copies of $\kappa $ as an $A$-module and since $\omega _ A^\bullet $ is normalized, we see that this complex only has cohomology placed in degree $0$. Thus $\omega _ B^\bullet $ is a normalized dualizing complex as well. $\square$

Lemma 47.16.2. Let $(A, \mathfrak m, \kappa )$ be a Noetherian local ring with normalized dualizing complex $\omega _ A^\bullet $. Let $A \to B$ be surjective. Then $\omega _ B^\bullet = R\mathop{\mathrm{Hom}}\nolimits _ A(B, \omega _ A^\bullet )$ is a normalized dualizing complex for $B$.

**Proof.**
Special case of Lemma 47.16.1.
$\square$

Lemma 47.16.3. Let $(A, \mathfrak m, \kappa )$ be a Noetherian local ring. Let $F$ be an $A$-linear self-equivalence of the category of finite length $A$-modules. Then $F$ is isomorphic to the identity functor.

**Proof.**
Since $\kappa $ is the unique simple object of the category we have $F(\kappa ) \cong \kappa $. Since our category is abelian, we find that $F$ is exact. Hence $F(E)$ has the same length as $E$ for all finite length modules $E$. Since $\mathop{\mathrm{Hom}}\nolimits (E, \kappa ) = \mathop{\mathrm{Hom}}\nolimits (F(E), F(\kappa )) \cong \mathop{\mathrm{Hom}}\nolimits (F(E), \kappa )$ we conclude from Nakayama's lemma that $E$ and $F(E)$ have the same number of generators. Hence $F(A/\mathfrak m^ n)$ is a cyclic $A$-module. Pick a generator $e \in F(A/\mathfrak m^ n)$. Since $F$ is $A$-linear we conclude that $\mathfrak m^ n e = 0$. The map $A/\mathfrak m^ n \to F(A/\mathfrak m^ n)$ has to be an isomorphism as the lengths are equal. Pick an element

which maps to a generator for all $n$ (small argument omitted). Then we obtain a system of isomorphisms $A/\mathfrak m^ n \to F(A/\mathfrak m^ n)$ compatible with all $A$-module maps $A/\mathfrak m^ n \to A/\mathfrak m^{n'}$ (by $A$-linearity of $F$ again). Since any finite length module is a cokernel of a map between direct sums of cyclic modules, we obtain the isomorphism of the lemma. $\square$

Lemma 47.16.4. Let $(A, \mathfrak m, \kappa )$ be a Noetherian local ring with normalized dualizing complex $\omega _ A^\bullet $. Let $E$ be an injective hull of $\kappa $. Then there exists a functorial isomorphism

for $N$ running through the finite length $A$-modules.

**Proof.**
By induction on the length of $N$ we see that $R\mathop{\mathrm{Hom}}\nolimits _ A(N, \omega _ A^\bullet )$ is a module of finite length sitting in degree $0$. Thus $R\mathop{\mathrm{Hom}}\nolimits _ A(-, \omega _ A^\bullet )$ induces an anti-equivalence on the category of finite length modules. Since the same is true for $\mathop{\mathrm{Hom}}\nolimits _ A(-, E)$ by Proposition 47.7.8 we see that

is an equivalence as in Lemma 47.16.3. Hence it is isomorphic to the identity functor. Since $\mathop{\mathrm{Hom}}\nolimits _ A(-, E)$ applied twice is the identity (Proposition 47.7.8) we obtain the statement of the lemma. $\square$

Lemma 47.16.5. Let $(A, \mathfrak m, \kappa )$ be a Noetherian local ring with normalized dualizing complex $\omega _ A^\bullet $. Let $M$ be a finite $A$-module and let $d = \dim (\text{Supp}(M))$. Then

if $\mathop{\mathrm{Ext}}\nolimits ^ i_ A(M, \omega _ A^\bullet )$ is nonzero, then $i \in \{ -d, \ldots , 0\} $,

the dimension of the support of $\mathop{\mathrm{Ext}}\nolimits ^ i_ A(M, \omega _ A^\bullet )$ is at most $-i$,

$\text{depth}(M)$ is the smallest integer $\delta \geq 0$ such that $\mathop{\mathrm{Ext}}\nolimits ^{-\delta }_ A(M, \omega _ A^\bullet ) \not= 0$.

**Proof.**
We prove this by induction on $d$. If $d = 0$, this follows from Lemma 47.16.4 and Matlis duality (Proposition 47.7.8) which guarantees that $\mathop{\mathrm{Hom}}\nolimits _ A(M, E)$ is nonzero if $M$ is nonzero.

Assume the result holds for modules with support of dimension $< d$ and that $M$ has depth $> 0$. Choose an $f \in \mathfrak m$ which is a nonzerodivisor on $M$ and consider the short exact sequence

Since $\dim (\text{Supp}(M/fM)) = d - 1$ (Algebra, Lemma 10.63.10) we may apply the induction hypothesis. Writing $E^ i = \mathop{\mathrm{Ext}}\nolimits ^ i_ A(M, \omega _ A^\bullet )$ and $F^ i = \mathop{\mathrm{Ext}}\nolimits ^ i_ A(M/fM, \omega _ A^\bullet )$ we obtain a long exact sequence

By induction $E^ i/fE^ i = 0$ for $i + 1 \not\in \{ -\dim (\text{Supp}(M/fM)), \ldots , -\text{depth}(M/fM)\} $. By Nakayama's lemma (Algebra, Lemma 10.20.1) and Algebra, Lemma 10.72.7 we conclude $E^ i = 0$ for $i \not\in \{ -\dim (\text{Supp}(M)), \ldots , -\text{depth}(M)\} $. Moreover, in the boundary case $i = - \text{depth}(M)$ we deduce that $E^ i$ is nonzero as $F^{i + 1}$ is nonzero by induction. Since $E^ i/fE^ i \subset F^{i + 1}$ we get

(see lemma used above) we also obtain the dimension estimate (2).

If $M$ has depth $0$ and $d > 0$ we let $N = M[\mathfrak m^\infty ]$ and set $M' = M/N$ (compare with Lemma 47.11.6). Then $M'$ has depth $> 0$ and $\dim (\text{Supp}(M')) = d$. Thus we know the result for $M'$ and since $R\mathop{\mathrm{Hom}}\nolimits _ A(N, \omega _ A^\bullet ) = \mathop{\mathrm{Hom}}\nolimits _ A(N, E)$ (Lemma 47.16.4) the long exact cohomology sequence of $\mathop{\mathrm{Ext}}\nolimits $'s implies the result for $M$. $\square$

Remark 47.16.6. Let $(A, \mathfrak m)$ and $\omega _ A^\bullet $ be as in Lemma 47.16.5. By More on Algebra, Lemma 15.69.2 we see that $\omega _ A^\bullet $ has injective-amplitude in $[-d, 0]$ because part (3) of that lemma applies. In particular, for any $A$-module $M$ (not necessarily finite) we have $\mathop{\mathrm{Ext}}\nolimits ^ i_ A(M, \omega _ A^\bullet ) = 0$ for $i \not\in \{ -d, \ldots , 0\} $.

Lemma 47.16.7. Let $(A, \mathfrak m, \kappa )$ be a Noetherian local ring with normalized dualizing complex $\omega _ A^\bullet $. Let $M$ be a finite $A$-module. The following are equivalent

$M$ is Cohen-Macaulay,

$\mathop{\mathrm{Ext}}\nolimits ^ i_ A(M, \omega _ A^\bullet )$ is nonzero for at most one $i$,

$\mathop{\mathrm{Ext}}\nolimits ^{-i}_ A(M, \omega _ A^\bullet )$ is zero for $i \not= \dim (\text{Supp}(M))$.

Denote $CM_ d$ the category of finite Cohen-Macaulay $A$-modules of depth $d$. Then $M \mapsto \mathop{\mathrm{Ext}}\nolimits ^{-d}_ A(M, \omega _ A^\bullet )$ defines an anti-auto-equivalence of $CM_ d$.

**Proof.**
We will use the results of Lemma 47.16.5 without further mention. Fix a finite module $M$. If $M$ is Cohen-Macaulay, then only $\mathop{\mathrm{Ext}}\nolimits ^{-d}_ A(M, \omega _ A^\bullet )$ can be nonzero, hence (1) $\Rightarrow $ (3). The implication (3) $\Rightarrow $ (2) is immediate. Assume (2) and let $N = \mathop{\mathrm{Ext}}\nolimits ^{-\delta }_ A(M, \omega _ A^\bullet )$ be the nonzero $\mathop{\mathrm{Ext}}\nolimits $ where $\delta = \text{depth}(M)$. Then, since

(Lemma 47.15.3) we conclude that $M = \mathop{\mathrm{Ext}}\nolimits _ A^{-\delta }(N, \omega _ A^\bullet )$. Thus $\delta \geq \dim (\text{Supp}(M))$. However, since we also know that $\delta \leq \dim (\text{Supp}(M))$ (Algebra, Lemma 10.72.3) we conclude that $M$ is Cohen-Macaulay.

To prove the final statement, it suffices to show that $N = \mathop{\mathrm{Ext}}\nolimits ^{-d}_ A(M, \omega _ A^\bullet )$ is in $CM_ d$ for $M$ in $CM_ d$. Above we have seen that $M[0] = R\mathop{\mathrm{Hom}}\nolimits _ A(N[d], \omega _ A^\bullet )$ and this proves the desired result by the equivalence of (1) and (3). $\square$

Lemma 47.16.8. Let $(A, \mathfrak m, \kappa )$ be a Noetherian local ring with normalized dualizing complex $\omega _ A^\bullet $. If $\dim (A) = 0$, then $\omega _ A^\bullet \cong E[0]$ where $E$ is an injective hull of the residue field.

**Proof.**
Immediate from Lemma 47.16.4.
$\square$

Lemma 47.16.9. Let $(A, \mathfrak m, \kappa )$ be a Noetherian local ring with normalized dualizing complex. Let $I \subset \mathfrak m$ be an ideal of finite length. Set $B = A/I$. Then there is a distinguished triangle

in $D(A)$ where $E$ is an injective hull of $\kappa $ and $\omega _ B^\bullet $ is a normalized dualizing complex for $B$.

**Proof.**
Use the short exact sequence $0 \to I \to A \to B \to 0$ and Lemmas 47.16.4 and 47.16.2.
$\square$

Lemma 47.16.10. Let $(A, \mathfrak m, \kappa )$ be a Noetherian local ring with normalized dualizing complex $\omega _ A^\bullet $. Let $f \in \mathfrak m$ be a nonzerodivisor. Set $B = A/(f)$. Then there is a distinguished triangle

in $D(A)$ where $\omega _ B^\bullet $ is a normalized dualizing complex for $B$.

**Proof.**
Use the short exact sequence $0 \to A \to A \to B \to 0$ and Lemma 47.16.2.
$\square$

Lemma 47.16.11. Let $(A, \mathfrak m, \kappa )$ be a Noetherian local ring with normalized dualizing complex $\omega _ A^\bullet $. Let $\mathfrak p$ be a minimal prime of $A$ with $\dim (A/\mathfrak p) = e$. Then $H^ i(\omega _ A^\bullet )_\mathfrak p$ is nonzero if and only if $i = -e$.

**Proof.**
Since $A_\mathfrak p$ has dimension zero, there exists an integer $n > 0$ such that $\mathfrak p^ nA_\mathfrak p$ is zero. Set $B = A/\mathfrak p^ n$ and $\omega _ B^\bullet = R\mathop{\mathrm{Hom}}\nolimits _ A(B, \omega _ A^\bullet )$. Since $B_\mathfrak p = A_\mathfrak p$ we see that

The second equality holds by More on Algebra, Lemma 15.99.2. By Lemma 47.16.2 we may replace $A$ by $B$. After doing so, we see that $\dim (A) = e$. Then we see that $H^ i(\omega _ A^\bullet )_\mathfrak p$ can only be nonzero if $i = -e$ by Lemma 47.16.5 parts (1) and (2). On the other hand, since $(\omega _ A^\bullet )_\mathfrak p$ is a dualizing complex for the nonzero ring $A_\mathfrak p$ (Lemma 47.15.6) we see that the remaining module has to be nonzero. $\square$

## Post a comment

Your email address will not be published. Required fields are marked.

In your comment you can use Markdown and LaTeX style mathematics (enclose it like `$\pi$`

). A preview option is available if you wish to see how it works out (just click on the eye in the toolbar).

Unfortunately JavaScript is disabled in your browser, so the comment preview function will not work.

All contributions are licensed under the GNU Free Documentation License.

## Comments (0)