5
$\begingroup$

[This was first posted on MSE but did not get any answer. I apologize if it is not suited for MO.]

Let $C$ be a curve, i.e. a purely one-dimensional scheme, embedded in a smooth projective threefold $X$. For a coherent sheaf $E$ of codimension $c$ on $X$, let $E^D=\mathscr Ext_X^c(E,\omega_X)$ be the Grothendieck dual of $E$. It is a reflexive sheaf of codimension $c$ on $X$. Reflexive means that the natural map to the double dual is an isomorphism. Let $\mathscr O_C$ be the structure sheaf of $C$, viewed as a torsion sheaf on $X$.

Question. For which curves $C\subset X$ is $\mathscr O_C$ reflexive?

Certainly, when $C$ is smooth. Let us assume $C$ is singular. I was thinking that maybe if $C$ is a local complete intersection then the first dual $\mathscr O_C^D$ could already be isomorphic to $\mathscr O_C$. Is this true? If so, then in particular dualizing twice does not do anything. But probably someone knows a more convincing (or even more true) statement. Thank you for any help!

$\endgroup$

2 Answers 2

4
$\begingroup$

Indeed, this condition is equivalent to $C$ being CM (Regardless of $X$ as long as it is also CM). This "dual" is called the $\omega$-dual in Kol13. You might be interested in reading section 2.5, or just the part directly dealing with this on pp.80-83. In particular, the statement you need is Cor. 2.70. Furthermore, if $X$ is not necessarily CM, then you get the same result using the dualizing complex instead of the dualizing sheaf.

As Mohan points out, the first dual will only be isomorphic to $\mathscr O$ if its own dualizing sheaf is trivial. Then again, why would you want that? That's a freak accident and doesn't give you too much mileage. Reflexivity is a (or "the") natural condition.

$\endgroup$
2
$\begingroup$

If $C$ is Cohen-Macaulay (in particular, if it is local complete intersection), then $\mathcal{O}_C$ is reflexive with your definition. Even when $C$ is smooth, $\mathcal{O}_C^D$ is not necessarily isomorphic to $\mathcal{O}_C$, since the dual is just $K_C$, which may or may not be isomorphic to $\mathcal{O}_C$.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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