8
$\begingroup$

Let $R$ be a noetherian commutative ring of dimension $n$, and let $M$ be a faithful finite $R$-module. Let $I$ be a proper ideal of $R$, and let $x\in I$ be a non-zerodivisor on $M$.

When does multiplication by $x$ induce an injection $H^n_I(M)\hookrightarrow H^n_I(M)$?

$\endgroup$
5
  • 3
    $\begingroup$ The local cohomology modules are $I$ torsion, so multiplication by $x$ will actually be the $0$ map. $\endgroup$ Apr 4, 2011 at 1:44
  • $\begingroup$ Uh-oh! Really? $\endgroup$ Apr 4, 2011 at 2:01
  • 5
    $\begingroup$ Not necessarily the zero map, but definitely not injective. For example, when $M = R$ is a Gorenstein local ring and $I = m$ is the maximal ideal, the top local cohomology is the injective hull of $R/m$, which is $m$-torsion. $\endgroup$ Apr 4, 2011 at 2:38
  • 1
    $\begingroup$ @Harry: The $I$-torsionness follows from the fact that there is an isomorphism ${\rm H}^i_I(M) = \varinjlim_d {\rm Ext}^i_R(R/I^d, M)$ $\endgroup$
    – Steven Sam
    Apr 4, 2011 at 2:41
  • $\begingroup$ Graham was right, let me give a careful answer then. $\endgroup$ Apr 4, 2011 at 4:37

1 Answer 1

6
$\begingroup$

Graham was right, the map is not necessarily $0$ as I wrote in the first comment. However, it is true that $H_I^n(M)$ is $I$-torsion, so it will be injective if and only if $H_I^n(M)=0$.

Amusingly, I will observe that the map is actually surjective.

Apply $\Gamma_I(-)$ to the sequence:

$$ M \stackrel{x}{\to} M \to M/xM$$

to get $$ \to H_I^n(M) \stackrel{x}{\to} H_I^n(M) \to H_I^n(M/xM) \to $$

Now note that $\dim M/xM < \dim M = n$ because $x$ is $M$-regular, so $H_I^n(M/xM) = 0$. (In general, $H_I^n(N) =0 $ for $n>\dim N$). So the multiplication by $x$ map is surjective, as claimed (may be this is what you had in mind anyway).

For completeness, the question of when $H_I^n(M) =0$ is rather subtle. It will be true, for example, if $I$ can be generated up to radical by at most $n-1$ elements, because you can calculate local cohomology using the Cech complex on those elements.

Another instance is when $R$ is a complete local domain, and $\dim R/I>0$ (this is known as the Hartshorne-Lichtenbaum vanishing theorem). I do not know an easy equivalent condition off the top of my head.

$\endgroup$
2
  • $\begingroup$ Heh, for what it's worth, I already knew it was surjective by that exact argument =). $\endgroup$ Apr 4, 2011 at 4:57
  • 1
    $\begingroup$ @Harry: I am glad you agree! $\endgroup$ Apr 4, 2011 at 5:02

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.