Tuesday, March 1, 2016

The canonical bundle of projective space and hypersurfaces

Let Pn be projective n-space with coordinates [x0::xn]. Cover Pn with affine pieces Ui={xi0}, each of which are An, in coordinates (y1,,yn), where yj=xj/xi. Recall that the canonical bundle of Pn is the n-fold wedge of the cotangent bundle of Pn, or ωPn=nTPn. The canonical bundle for an arbitrary variety is defined analogously.

Definition: Let X be a projective n-dimensional variety. The sheaf of regular functions on X is OX, with OX(U)={f/g : f,gk[x1,,xn]/I(X),g0}, and the restriction maps are function restriction.

There is a natural grading on OX, given by deg(f)deg(g). A shift in the grading may be applied, called a {\it Serre twist}, to get a differently graded (but isomorphic) module: for φOX with deg(φ)=k, set φOX() to have deg(φ)=k.

Let α=dy1dynωPn, which is well-defined on all of Ui. We claim this is well-defined on all of Pn. We check this on the overlap U0Un (for nicer notation), but the approach is analogous for UiUj.
U0={(y1,,yn) : yi=xi/x0}yi=zi+1zidyi=z1dzi+1zi+1dz1z21Un={(z1,,zn) : zi=xi1/xn}yn=1z1dyn=dz1z21

Therefore
α=dy1dyn=z1dz2z2dz1z21z1dznzndz1z21dz1z21=dz2z1dznz1dz1z21=(1)nzn+11dz1dzn.

Since the transition function has a pole of order n+1 when z1=0, which happens when x0=0, we have that α has a pole of order n+1 at . Therefore ωPnOPn(n1).

Let XPn be a smooth hypersurface defined by a degree d equation F(x0,,xn)=0. On the affine piece U0 this becomes f(y1,,yn)=F(1,x1x0,,xnx0) with yi=xi/x0. The total derivative is
fy1dy1++fyndyn=ni=1fyidyi=0,

and since X is smooth, the terms never all vanish at the same time. Let Vi={fyi0}, and set
βi=(1)i1f/yidy1^dyidynωX,

which is well-defined on all of ViU0. We claim that the choice of Vi does not matter, and indeed, assuming i<j,
βj=(1)j1f/yjdy1^dyjdyn=(1)j1+i1dyif/yjdy1^dyi^dyjdyn=(1)j1+i11f/yi(fy1dy1++^fyidyi++fyndyn)f/yjdy1^dyi^dyjdyn=(1)j1+i1+11f/yifyjdyjf/yjdy1^dyi^dyjdyn=(1)j1+i1+1+j2f/yidy1^dyidyn=(1)i1f/yidy1^dyidyn=βi.

Hence βi is well-defined on all of U0, and we call it simply β. Next we claim it is well-defined on all of X. Again we only check on the overlap of U0Un. On the affine piece Un this becomes g(z1,,zn)=F(x0xn,,xn1xn,1)=f(z2z1,,znz1,1z1) with zi=xi1/xn. We employ the chain rule fyi=fzjzjyi and the results above to find that
β=(1)i1f/yidy1^dyidyn=(1)i1f/zjzj/yiz1dz2z2dz1z21^dyiz1dznzndz1z21dz1z21=(1)i1f/zjzj/yi(1)n1zn1dz1^dzidzn=(1)i+n(1z1)d1(c+)zn1dz1^dzidzn=(1)i+nznd+11(c+)dz1^dzidzn,

for some constant c. This comes from expressing f in terms of the zis and factoring. Since the transition function has a pole of order nd+1 when z1=0, which happens when x0=0, we have that β has a pole of order nd+1 at . Therefore ωXOX(n+d1).

References: Griffiths and Harris (Principles of Algebraic Geometry, Chapter 1.2)

No comments:

Post a Comment