2
$\begingroup$

Let $ X $ be a smooth (projective) variety and $ \mathcal{F} $ a torsion-free coherent sheaf of rank $ r $ on $ X $. The determinant $ \det \mathcal{F} $ can be defined by

(1) $ \det \mathcal F := ( \wedge^r \mathcal{F} )^{\vee \vee} $. Here the double dual denotes the reflexive hull of the rank $ 1 $ sheaf $ \wedge^r \mathcal{F} $ and therefore is a line bundle (as $ X $ is smooth).

(2) $ \det \mathcal F := \bigotimes_j (-1)^j \det E^j $ for a finite length resolution $ E^{\bullet} $ of $ \mathcal{F} $ by locally free sheaves. Here the notation $ (-1)^j \det E^j $ just means $ \det E^j $ for even $ j $ and $ (\det E^j)^{\vee} $ for odd $ j $. The determinant computed this way can be shown to be independent of the chosen complex $ E^{\bullet} $.

Are these two definitions the same? And if yes, why? I have been going crazy over this point!

$\endgroup$
3
  • 3
    $\begingroup$ Restrict over the maximal open on which the coherent sheaf is locally free. $\endgroup$ Commented Aug 31, 2023 at 1:02
  • 3
    $\begingroup$ @JasonStarr Oh wow. This maximal open set U will have complement codimension 2 or larger - because the local ring at the generic point of a codimension 1 subvariety is of dimension 1 hence a DVR so torsion-free stalk of F is the same as free stalk of F there. Now on that maximal open set, the definitions of det agree and the restriction Pic X to Pic U is an isomorphism because of codimension >=2 again. Correct? $\endgroup$ Commented Aug 31, 2023 at 1:09
  • $\begingroup$ Yes that is correct. $\endgroup$ Commented Aug 31, 2023 at 1:57

0

You must log in to answer this question.