8
$\begingroup$

Let $M$ be a finitely generated graded module over a graded ring $R$. Let $\mathcal{F}$ be the corresponding coherent sheaf on $\operatorname{Proj} R$. There is a natural map of graded $R$-modules $$\phi \colon M \to \Gamma^*(\mathcal{F}) := \bigoplus_{n} \Gamma(\operatorname{Proj} R, \mathcal{F}(n)).$$ If I recall Ravi Vakil's notes correctly, $M$ is called saturated if $\phi$ is an isomorphism.

Is there a term (perhaps semi-saturated, or some such) for modules $M$ such that $\phi$ is injective?

This concept is appealing for several reasons. For one thing, it is easier to test "semi-saturatedness" than saturatedness; e.g., unless I am mistaken, $\phi$ is automatically injective if $M$ admits any positive-degree homogeneous nonzerodivisor. For another, at least if $R$ is a polynomial ring, $M = R/I$ is "semi-saturated" iff $I$ is a saturated ideal of $R$. (Note that the definition of "saturated ideal" is different from the definition given above for "saturated module", and I do not think the two are equivalent for ideals.)

$\endgroup$
2
  • 2
    $\begingroup$ I don't know a name for this, but I have a comment. This is quite close to being torsion free (or more precisely, $R_+$-torsion free) if $\mathcal{F}$ is already S2. Ok, for example, lets suppose that $R$ is normal and $\mathcal{F}$ is reflexive on $X$ (this is equivalent to S2, depending on your choice of terms). Then $\Gamma^*(\mathcal{F})$ is the S2-ification = reflexification of $M$. There's a paper of Hartshorne called something like "Generalized divisors on Gorenstein schemes" where he points out that in the context I described, the injectivity is equivalent to being torsion free. $\endgroup$ Jul 18, 2011 at 20:11
  • $\begingroup$ You can make sense of this even without the reflexification statements too. $\endgroup$ Jul 18, 2011 at 20:12

2 Answers 2

6
$\begingroup$

To elaborate on Karl's comment:

Let $m$ be the irrelevant ideal of $R$, then there is a short exact sequence:

$$0 \to H_m^0(M) \to M \to \Gamma^*(\mathcal{F}) \to H_m^1(M) \to 0$$

(see Eisenbud's book, Theorem A4.1, p. 693). Here $H_m^i(M)$ denote the local cohomology modules. So the map is injective precisely when $H_m^0(M):= \cup_n (0:_M m^n) = 0$. I believe such module is called $m$-torsion-free (don't know a reference off hand, may be Brodmann-Sharp's book on local cohomology?).

Also, it is equivalent to $m$ contains a non-zerodivisor on $M$ (may be that's what you meant in the last paragraph?)

$\endgroup$
3
$\begingroup$

As Hailong wrote, the injectivity means the vanishing of $H^0_m(M)$. But $\operatorname{depth}(m,M)$ is the least non-vanishing local cohomology, thus the map $M \to H^0_*(\tilde{M})$ is injective iff $M$ has depth $\ge$ 1. If $M$ has finite projective dimension, e.g. if R is regular, then this happens iff $M$ has projective dimension at most depth $R-1$ by the Auslander-Buchsbaum formula.

$\endgroup$

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.