Dualizing sheaf explained

\omegaX

together with a linear functional

tX:\operatorname{H}n(X,\omegaX)\tok

that induces a natural isomorphism of vector spaces

\operatorname{Hom}X(F,\omegaX)\simeq\operatorname{H}n(X,F)*,\varphi\mapstotX\circ\varphi

for each coherent sheaf F on X (the superscript * refers to a dual vector space). The linear functional

tX

is called a trace morphism.

A pair

(\omegaX,tX)

, if it is exists, is unique up to a natural isomorphism. In fact, in the language of category theory,

\omegaX

is an object representing the contravariant functor

F\mapsto\operatorname{H}n(X,F)*

from the category of coherent sheaves on X to the category of k-vector spaces.

For a normal projective variety X, the dualizing sheaf exists and it is in fact the canonical sheaf:

\omegaX=l{O}X(KX)

where

KX

is a canonical divisor. More generally, the dualizing sheaf exists for any projective scheme.

There is the following variant of Serre's duality theorem: for a projective scheme X of pure dimension n and a Cohen–Macaulay sheaf F on X such that

\operatorname{Supp}(F)

is of pure dimension n, there is a natural isomorphism

\operatorname{H}i(X,F)\simeq\operatorname{H}n-i(X,

*
\operatorname{l{H}om}(F,\omega
X))
.In particular, if X itself is a Cohen–Macaulay scheme, then the above duality holds for any locally free sheaf.

Relative dualizing sheaf

Given a proper finitely presented morphism of schemes

f:X\toY

, defines the relative dualizing sheaf

\omegaf

or

\omegaX/Y

as the sheaf such that for each open subset

U\subsetY

and a quasi-coherent sheaf

F

on

U

, there is a canonical isomorphism
!
(f|
U)

F=\omegafl{OY}F

,which is functorial in

F

and commutes with open restrictions.

Example:If

f

is a local complete intersection morphism between schemes of finite type over a field, then (by definition) each point of

X

has an open neighborhood

U

and a factorization

f|U:U\overset{i}\toZ\overset{\pi}\toY

, a regular embedding of codimension

k

followed by a smooth morphism of relative dimension

r

. Then

\omegaf|U\simeq\wedgeri*

1
\Omega
\pi

\wedgekNU/Z

where
1
\Omega
\pi
is the sheaf of relative Kähler differentials and

NU/Z

is the normal bundle to

i

.

Examples

Dualizing sheaf of a nodal curve

For a smooth curve C, its dualizing sheaf

\omegaC

can be given by the canonical sheaf
1
\Omega
C
.

For a nodal curve C with a node p, we may consider the normalization

\pi:\tildeC\toC

with two points x, y identified. Let

\Omega\tilde(x+y)

be the sheaf of rational 1-forms on

\tildeC

with possible simple poles at x and y, and let

\Omega\tilde(x+y)0

be the subsheaf consisting of rational 1-forms with the sum of residues at x and y equal to zero. Then the direct image

\pi*\Omega\tilde(x+y)0

defines a dualizing sheaf for the nodal curve C. The construction can be easily generalized to nodal curves with multiple nodes.

This is used in the construction of the Hodge bundle on the compactified moduli space of curves: it allows us to extend the relative canonical sheaf over the boundary which parametrizes nodal curves. The Hodge bundle is then defined as the direct image of a relative dualizing sheaf.

Dualizing sheaf of projective schemes

As mentioned above, the dualizing sheaf exists for all projective schemes. For X a closed subscheme of Pn of codimension r, its dualizing sheaf can be given as

r
l{Ext}
Pn

(l{O}X,\omega

Pn

)

. In other words, one uses the dualizing sheaf on the ambient Pn to construct the dualizing sheaf on X.

See also

References

External links