COMPLEX ALGEBRAIC SURFACES CLASS 7
RAVI VAKIL
CONTENTS
1. How basic aspects of surfaces change under blow-up 1 2. Rational maps of surfaces, linear systems, and elimination of indeterminacy 3 3. The universal property of blowing up 4 3.1. Applications of the universal property of blowing up 4
Recap of last time. Last time we began discussing blow-ups:
¼ ¼
Ô ¾ Ë Ë = BÐ Ë : Ë ! Ë
Given , there is a surface Ô and a morphism , unique up to
½
´Ë fÔgµ
isomorphism, such that (i) the restriction of to is an isomorphism onto
½ ½ ½
fÔg ´Ôµ È ´Ôµ Ô Ë , and (ii) is isomorphic to . is called the exceptional divisor , and is called the exceptional divisor.
A key example, and indeed the analytic-, formal-, or etale-local situation, was given by
¾
= A blowing up Ë at the origin, which I’ll describe again soon when it comes up in a proof.
For the definition, complex analytically, you can take the same construction. Then you
need to think a little bit about uniqueness. There is a more intrinsic definition that works
¼ d
Á Ë = ÈÖÓ j ¨ Á
¼ algebraically, let be the ideal sheaf of the point. Then d .
1. HOW BASIC ASPECTS OF SURFACES CHANGE UNDER BLOW-UP
×ØÖicØ
Ë C C
Definition. If C is a curve on , define the strict transform of to the the closure of
ÔÖÓÔ eÖ
£
´C \ Ë Ôµ Ë Ô j C
the pullback on , i.e. ¼ . The proper transform is given by the Ë E
£ ¼
¼
Ç ´C µ=Ç ´C µ Ë
pullback of the defining equation, so for example Ë .
×ØÖicØ ÔÖÓÔ eÖ ×ØÖicØ £
C · ÑE C Ñ Ô C = C · ÑE C
Lemma. If has multiplicity at , then , i.e. = . Ñ
Proof. The multiplicity of C being means that in local coordinates, the defining equation
Ñ Ñ
has terms of degree Ñ, but not lower. (Better: the defining equation lies in but not
Ñ·½
Ü Ý Ñ .) Analytically, this means that the leading term in and has degree m.
Date: Wednesday, October 23.
1
Í =
We do this by local calculation which will be useful in general. (Draw picture.) ¼
f´´Ü ;Ý µ; [½; Ú ]µ : Ý = Ü Ú g = ËÔ ec k [Ü ;Ý ;Ú]=Ý = Ü Ú = ËÔ ec k [Ü ;Ú]
¼ ¼ ¼ ¼ ¼ ¼ ¼ ¼
¼ . The exceptional
E Ý =¼
divisor is given by ¼ (after morphism).
Í = f´´Ü ;Ý µ; [Ù; ½]µ : Ü = Ý Ùg = ËÔ ec k [Ü ;Ý ;Ù]=Ü = Ý Ù = ËÔ ec k [Ý ;Ù]
½ ½ ½ ½ ½ ½ ½ ½ ½
½ . The
E Ý =¼
exceptional divisor is given by ½ .
´Ü; Ý µ = ËÔ ec k [Ü; Ý ] ´Ü ;Ý ;Úµ 7! ´Ü ;Ý µ ´Ü ;Ý ;Úµ 7! ´Ü ;Ý µ
¼ ¼ ¼ ½ ½ ½ ½
Map down to . ¼ , .
f ´Ü; Ý µ=¼ Í f ´Ü; Ý µ= f ´Ü; Ý µ · higheÖ = f ´Ü ; Ü Úµ ·
Ñ Ñ ¼ ¼
Given a function . Pull it back to ¼ : ¡¡¡ higheÖ · .
Exercise. To see if you understood that, do the same calculation on patch 2.
¼ ¼
: Ë ! Ë Ë Ô E Ë
Theorem. Suppose is a blow up of at , with exceptional curve . Let
¼ £ £ ¼ ¼ £ ¾
D Ë D ¡ D = D ¡ D E ¡ D =¼ E = ½
D and be divisors on . Then , , .
½
½
Remark. A curve on a smooth surface that is isomorphic to È and has self-intersection
½µ is called a ´ -curve.
Proof. The first we did yesterday. The second: by Serre’s moving lemma, we can move
Ô C Ô
D away from , then pull back. For the third: choose a curve passing through with
×ØÖicØ
C ¡ E = ½
multiplicity 1. (How to do this: hyperplane section of Ë .) Then . Also
ÔÖÓÔ eÖ ×ØÖicØ ÔÖÓÔ eÖ
¡ E =¼ C · E = C
C .as , we’re done.
¼ £
Ë ¨ Z ! Èic Ë ´D; Òµ 7! D · ÒE
Theorem. (a) There is an isomorphism Èic defined by . ÆË (b) The same with Èic replaced by .
Proof. The arguments are the same for both parts, so I’ll do (a). It is surjective: the divisors E
upstairs are either E or strict transforms (which are proper transforms plus ’s). It is
£
D · ÒE = ¼ E Ò = ¼
injective: if , then intersect with to see that ; then apply £ to see =¼
that D .
£
¼
= Ã · E Ã Ë
Theorem. Ë .
£
¼
Ã Ñ E à = = à · ÑE
Ë E
Proof. Clearly Ë for some . By the adjunction formula for ,
¼
´E µj à E
Ë . Taking degrees:
£
¾= ´ Ã · ÑE · E µ ¡ E = Ñ ½: Ë
Exercise/Remark. If you want practice with the canonical bundle in local coordinates,
à Ô
take a meromorphic section of Ë that has neither zero nor pole at (possible by Serre’s
f ´Ü; Ý µdÜ ^ dÝ Í
moving lemma), write it as , and pull it back to the open set ½ to see that
f ´Ü ;Ü Ú µdÜ ^ d´Ü Ú µ= f ´Ü ;Ü Ú µÜ dÜ ^ dÚ
¼ ¼ ¼ ¼ ¼ ¼ ¼ you get ¼ .
2 2. RATIONAL MAPS OF SURFACES, LINEAR SYSTEMS, AND ELIMINATION OF
INDETERMINACY
99 ÃX X
A rational map Ë , where is a variety, means a morphism from an dense open C set of Ë . Recall that a rational map from a curve to a projective variety can always be
extended to a morphism. Similarly, a rational map from a surface Ë to a projective variety
can be extended over most points; the set of indeterminacy is a finite set of points. More
Ò
: Ë 99 ÃÈ Ò ·½ precisely, given a map . This is given by sections of some line bundle. It
makes sense except where the sections are all zero. This will be in codimension 2.
´Ë F µ Ë ´Ë µ
Let F be this finite set. We’ll denote the image of , and denote it . (I’m
Ë ´C F µ
not sure we need to take the closure.) If C is a curve on , then we’ll denote the
´C µ
image of C , and denote it . Here we definitely need to take the closure.
Ë Î Ò
Now suppose you have a divisor D on . Given a subspace of dimension of
¼ £
´Ë; Ç ´D µµ ÈÎ À , we might hope to get a map to projective space . (This is called a linear
system of dimension Ò; I should have introduced this notation earlier.) If it is base point free, we do.
If it has base points, the locus could have components of dimension 1. Such a compo- Î
nent is called a fixed component of the linear system Î . The fixed part of is the biggest
F D F divisor contained in every element of Î . So if this fixed part is , then has no fixed components.
(I’m not happy with how I explained the previous paragraphs in class. I hope this is clearer.)
Lemma. If the linear systems has no fixed part, then it has only a finite number of fixed points.
Proof. Take two general sections, and look at their two zero-sets. Where do they intersect? ¾
At a bunch of points. Hence we get at most D ?
We’ve basically shown that there is a bijection between:
Ò
: Ë 99 ÃÈ ´Ë µ g
(i) f rational maps such that is contained in no hyperplane
Ë Ò g (ii) f linear systems on without fixed part and of dimension
(Explain the correspondence.)
: Ë 99 Ã X
Theorem (Elimination of indeterminacy). Let be a rational map from a
¼ ¼
: Ë ! Ë
surface to a projective variety. Then there exists a surface Ë , a morphism
¼
: Ë ! X which is the composite of a finite number of blow-ups, and a morphism f such
3
that the diagram
¼
Ë
f
º ²
Ë 99 Ã X
is commutative. ¾
Proof. Idea: blow up fixed points, show that D decreases.
Ñ
È ´Ë µ
We immediately reduce to the case where X is , and isn’t contained in any
Ñ
Î jD j Ò Ë hyperplane of È . Then corresponds to a linear system of dimension on ,
with no fixed component. If Î has no base point, then we’re done.
Ü Ë ! Ë Ü
Otherwise, we blow up a base point , and consider ½ at (and hence a rational
Ë 99 ÃË
map ½ ). The exceptional curve is now in the fixed part of the linear system, with
½ kE
some multiplicity k . So we subtract to get rid of the fixed part, i.e. get a new linear
£
Î j D kE j : Ë 99 ÃË D = D kE
½ ½ ½ system ½ , to get the same rational map , given by . If this is a morphism, we win, otherwise we keep going.
At some point, this process must stop (and hence we win in the long run). We prove
¾
= i i i = ¼
this is the case when D , by induction on . Base case, : the number of fixed
¾
= ¼ i > ¼
points is bounded by D , so there aren’t any. Inductive step: Now . Then
¾
D =
we blow-up once, and we get a new surface with divisor class. On this surface, ½
¾ ¾ ¾
D kE µ´D kE µ = D k < D ´ . So by the inductive hypothesis, the process will terminate on this new surface, completing the induction.
3. THE UNIVERSAL PROPERTY OF BLOWING UP
: X ! Ë
Theorem (Universal property of blowing up). Let f be a birational morphism
½
Ô Ë f of surfaces, and suppose that the rational map f is undefined at a point of . Then
factorizes as
g
~
f : X ! Ë := BÐ Ë ! Ë
Ô
Ô where g is a birational morphism and is the blow-up at .
Proof: next day.
3.1. Applications of the universal property of blowing up. Two theorems.
f : Ë ! Ë
Theorem (all birational morphisms factor into blow-ups). Let ¼ be a birational
: Ë ! Ë k ½;:::;Ò
k k ½
morphism of surfaces. Then there is a sequence of blow-ups k ( )
Ù : Ë ! Ë f = Æ ¡¡¡ Æ Æ Ù
½ Ò
and an isomorphism Ò such that .
½
f Ô Ë f
Proof. If is an isomorphism, we’re done. Otherwise, there is a point of ¼ such that
Ô Ë ! Ë =BÐ Ë
Ô ¼
is undefined At , and we can factor through ½ . We can repeat this.
Ò´f µ Ò´f µ < Ò´f µ E
k k ½
If k is the number of contracted curves of :if is the exceptional
: Ë ! Ë E Ë
k k ½ divisor of k , then the preimage of in contains a curve which is contracted
4
f f
½ k by k but not . As the number of contracted curves can’t be negative, the process must
terminate.
¼
: Ë 99 Ã Ë
Theorem (all birational maps can be factored into blow-ups). Let be a ¼¼
birational map of surfaces. Then there is a surface Ë and a commutative diagram
¼¼
Ë
f g
º ²
¼
Ë 99 Ã Ë g where the morphisms f and are composites of blow-ups.
Proof. By the theorem of elimination of indeterminacy, we can find such a diagram such g that f is a composition of blow-ups. By the Theorem above, must then be a composition of blow-ups too.
We’ve now proved some powerful stuff, so let’s take a step back and see what we now know, and how it relates to classification.
Two surfaces are birational iff they can be be related by sequences of blow-ups. We’ll
be interested in birational classification, but biregular classification is very close.
¼
f : Ë ! Ë Ò ÆË´Ë µ
If is birational which is the composition of blow-ups, then =
¼ Ò
ÆË´Ë µ ¨ Z ,so Ò is independent of the choice of blow-ups. Exercise: Use this to show
that every birational morphism from Ë to itself is an isomorphism.
i ¼
: Ë ! Ë
Fact. In a blow-up, À of the structure sheaf is preserved, i.e. if is a blow-up,
£ i i
¼
: À ´Ç µ ! À ´Ç µ Ë then Ë is an isomorphism.
The algebraic way of proving this fact comes from the Leray spectral sequence, and the
i
¼ ¼
Ç Ê Ç i > ¼ = Ç = ¼
£ Ë £ Ë fact that Ë and for . This in turn requires some infinitesimal analysis, in the form of “formal function theorems”. I suspect that there should be a relatively straightforward analytic proof.
In particular, by these numbers are birational invariants.
So look at what this means for the Hodge diamond. When you blow up, you add 1 to the central entry (the rank of the Neron-Severi group). Everything else is constant.
Next day: More consequences of these powerful theorems. Proof of the universal prop- erty of blowing up. Castelnuovo’s criterion for blowing down curves.
5