Algebraic surfaces
Divisors
Remark 0.1. Let (A, m) be a Noetherian local ring with residue field k := A/m.
We said that A is a regular local ring if dimA = dimkm/m2. A ring is regular if the localization at every maximal ideal is a regular local ring.
Remark 0.2. Let Y be an affine variety with coordinate ring A(Y ). And P ∈ Y is a point. We say Y is non-singular at P if OP,Y is a regular local ring. We say that Y is non-singular if it’s non-singular at every point.
We say that Y is normal at P if OP,Y is a integrally closed. We say that Y is normal if it’s non-singular at every point.
In general, one has ”non-singular ⇒ normal ⇒ non-singular in codimension 1”.
Remark 0.3. An 1-dimensioanal regular local ring is a DVR. (cf. [Matsumura, p.79] )
Exercise 0.4. Show that dimA2C= 2.
Show that A2k is non-singular.
Let X be a variety (maybe singular), one can define Div(X) to be the divisor group. Let K(X) be the field of rational functions. If X is non-singular in codimen- sion 1, (i.e. localization at height 1 prime gives a regular local ring. An equivalent condition is that the singular locus has codim ≥ 2), then one can define
K(X)∗→ Div(X), by div(f ) =P
Y v(Y ). Where the sum is taking over all codimension 1 subvarieties.
The image is called principal divisors. Two divisor are said to be linearly equivalent if they differ by a principal divisor. We define the divisor class group Cl(X) :=
Div/{principal divisors}.
Exercise 0.5. Prove that Cl(Pn) ∼= Z. Hence we can define the degree of a divisor via Div(Pn) → Cl(Pn) → Z.
However, the divisor class group is in general not that simple. For example, Cl(E) 6∼= Z.
Every divisor on a non-singular variety is locally principal, i.e. in a sufficiently small neighborhood Uα, D|Uα= div(fα) for some rational function fα. We call fα
the local equation of D. Note that on Uα∩ Uβ, fαfβ−1 is regular. On the other hand, if one has an open covering X = ∪Uα and a collection of (fα, Uα) such that fαfβ−1 is regular, then this defines a divisor.
Example 0.6. Consider X = P1= U0∪ U1. Let t, s be local coordinate of U0, U1
respectively. One has s = t−1. Also K(X) = k(t) = k(s). Now consider 1-form dt on U0, it’s clear that dt = −ds/s2. We have {(1, U0), (−1/s2, U1)} which represent the 1-form. The divisor is −2[∞] which is the canonical divisor.
Example 0.7. Similarly, consider X = Pn = U0∪ ... ∪ Un. Conputatation shows that KX= −(n + 1)H for some hyperplane H.
Remark 0.8. It could happen that difference choices of coordinates gives different divisor. Indeed, they might give different divisor but still linear equivalent. One should say that the canonical divisor is the equivalent class of the divisor defined this way. Or sometimes we simply said that a divisor is a canonical divisor if it is in the linear equivalent class.
1
2
For a given a divisor D =P
niDi, we say D is effective, denoted D ≥ 0, if ni≥ 0 for all i. The linear series of D is defined as
|D| := {D0 ∈ Div(X)|D0∼ D, D0≥ 0}.
We say that D is a reduced divisor if ni = 1 for all i, and D is irreducible if ni = 1 for a unique i and nj= 0 otherwise, that is, D is a subvariety.
One of the purpose to realize a divisor as a collection of local equation is that it enable us to define the pull-back of the divisor. Let f : X → Y be a morphism of two non-singular varieties. Given D ∈ Div(Y ), it associate to a collection {(fα, Uα)}
for a small enough open covering {Uα}. This gives an open covering {f−1(Uα)}
on X and also a collection of local equations {(fα◦ f, f−1(Uα))}. We get a divisor on X, denoted f∗D. It’s clear that if D1 ∼ D2 then f∗D1 ∼ f∗D2 because f∗D1− f∗D2= div(g ◦ f ) if D1− D2= div(g).
Example 0.9. Consider the morphism f : P2→ P5via f ([a0, a1, a2]) = [a20, a0a1, a0a2, a21, a1a2, a22].
Let H be the hyperplane Z0+ Z1 = 0, what’s the divisor f∗H? Let H0 be the hy- perplane Z0= 0, what’s the divisor f∗H0?
