Understanding the claim that holomorphic divisors contained in a divisor class are in bijection with the projectivisation of the space of sections

81 Views Asked by At

Let $V$ be a complex manifold. I am trying to approach the subject of divisors through complex geometry since I am not familiar enough with algebraic geometry to come that way.

A divisor $D$ on $V$ is defined as a system of meromorphic place functions $\{ f_i \}$ on an open cover $\{ U_i \}$ such that on $U_i \cap U_j$, $f_i/f_j$ has neither zeros nor poles.

My confusion might stem from not properly understanding when two divisors are equivalent. Provisionally, I am assuming this definition:

Two divisors $D, D'$ are defined as equivalent if there is an open cover $\{ U_i \}$ for both and a global meromorphic function $f$ such that $f_i/f_i' = f$ on each $U_i$.

(Auxiliary) Question 0. Is this definition correct?

The source I have considers the sheaf $\mathcal M$ of local meromorphic not identically $0$ functions and defines divisors up to equivalence as global sections in the sheaf $\mathcal M/ \mathcal H^*$ where $\mathcal H^*$ is the sheaf of never $0$ local holomorphic functions. I suspect this definition is not the same as the one I gave above, but it is cumbersome and I have no idea how to work with it, as you will see below. It is also said that $H^0(V, \mathcal M)$ is the set of all meromorphic functions on $V$ which are not identically $0$ on any connected component of $V$, which I don't truly understand why.

Moving to what I am trying to understand, two divisors $D, D'$ are called linearly equivalent if $D - D'$ is the divisor $(g)$ of $g \in H^0(V, \mathcal M)$. For an arbitrary divisor $D$ given by place functions $\{ f_i \}$, $L(D)$ is defined as the vector space of all global meromorphic functions $g$ (not necessarily satisfying $g \in H^0(V, \mathcal M)$) such that $gf_i$ are holomorphic on each $U_i$.

The claim I am trying to understand is that if $V$ is compact and connected then the set of holomorphic divisors which are linearly equivalent to $D$ is in bijection with the projectivisation of the vector space $L(D)$.

I get that the gist of this claim is that if $D' = \{ f_i' \} $ is a holomorphic linearly equivalent divisor to $D = \{ f_i \}$, then there exists $g \in H^0(V, \mathcal M)$ such that $f_i (1/g) = f_i'$, and since $D'$ is holomorphic, it follows that $1/g \in L(D)$.

This gives me a tentative map from the set of holomorphic linearly equivalent divisors to $D$ into $L(D)$. I am not sure this map is well defined. For example, according to my definition of when divisors are equivalent, if $c \in \mathbb C^*$, then the divisor given by $f_i' = f_i (c/g)$ would be equivalent to the one given by $f_i' = f_i (1/g)$, thus this map only has a chance of being well defined if we take indeed the projectivisation of $L(D)$ i.e. compress all multiples of a meromorphic function in $L(D)$ to a point. But even then, what guarantees that it is well defined? In other words:

Question 1 If the divisor given by $f_i (1/g)$ is equivalent to $f_i (1/h)$ where $g, h \in H^0(V, \mathcal M)$ (whatever "divisor equivalence" ends up meaning), can we deduce that $g = c h$ with $c \in \mathbb C$ and, if so, how can we deduce this?

Then regarding surjectivity, I suspect what happens if $V$ is not compact or not connected is that we can find some meromorphic function $g$ such that $1/g \not\in H^0(V, \mathcal M)$, but nevertheless $gf_i$ is holomorphic; in other words $gf_i$ is holomorphic, but the holomorphic divisor $D+(g)$ is not linearly equivalent to $D$. I suspect in the end this point has nothing to do with holomorphicity also, but just with the existence of meromorphic functions which are not in $H^0(V, \mathcal M)$, right?

Question 2 Why if $V$ is compact connected then for any meromorphic $g$, $D+(g)$ is linearly equivalent to $D$?

Examples to work through in illustration of my points of confusion would certainly help. Also some references which explain these things more clearly would be beneficial.

1

There are 1 best solutions below

6
On

The source I have considers the sheaf $\mathcal M$ of local meromorphic not identically 0 functions and defines divisors up to equivalence as global sections in the sheaf $\mathcal M/\mathcal H^∗$ where $\mathcal H^∗$ is the sheaf of never 0 local holomorphic functions.

That is the formally correct definition. Quotient sheaves might be a bit scary at first (they are defined by sheafification), but essentially this means what you write before: Given a global section $D \in \Gamma(V, \mathcal{M/H}^*)$, there exists an open cover $\{U_i\}_i$, and for each $i$ a preimage $f_i \in \Gamma(V, \mathcal M)$ of the section $D|_{U_i} \in \Gamma(U_i, \mathcal M/\mathcal H^*)$. Of course on $U_{ij} = U_j \cap U_j$, $f_i$ and $f_j$ should give the same section $D|_{U_{ij}}$, which translates in the condition that $f_i / f_j \in \Gamma(U_{ij}, \mathcal H^*)$. So the tuple $(U_i, f_i)_i$ represents the divisor $D$. Naturally, because we are taking the quotient by $\mathcal H^*$, two tuples $(U_i, f_i)$ and $(U_i', f_i')$ define the same divisor $D$, if $f_i / f_i'$ is holomorphic and nonzero on $U_i \cap U_i'$.

Two divisors $D,D′$ are defined as equivalent if there is an open cover $\{U_i\}$ for both and a global meromorphic function $f$ such that $f_i/f_i'=f$ on each $U_i$.

If you allow $f$ to be meromorphic, you get the notion of linear equivalence, which you also write later. However if $f$ is in fact holomorphic, we say the two divisors are equal (because they define the same global section of $\mathcal M / \mathcal H^*$.

Now to the map $$\{ \text{holomorphic divisors linearly equivalent to } D\} \to \mathbb P(L(D)), D' \mapsto [1/g]$$ where $(g) = D - D'$.

Asking if this map is well-defined boils is the following problem: If $g, h \in \Gamma(V, \mathcal M)$ are meromorphic with $(g) = (h)$, then $g = c \cdot h$ for some $c \in \mathbb C$. Note that $(g) = (h)$ means that their quotient $\frac g h$ is holomorphic and nonzero on $V$ (see the above discussion about equality of divisors). As $V$ is compact, this quotient has to be a nonzero constant.

I suspect in the end this point has nothing to do with holomorphicity also, but just with the existence of meromorphic functions which are not in $H^0(V,\mathcal M)$, right?

I'm not sure what you are talking about here. Isn't $H^0(V, \mathcal M)$ the set of meromorphic functions by definition? Also, a meromorphic function is locally the quotient of two holomorphic functions. So for each $g \in \Gamma(V, \mathcal M)$, the function $1/g$ is also contained in $\Gamma(V, \mathcal M)$. This also has to be the case so that linear equivalence is symmetric, otherwise it wouldn't be an equivalence relation. Then it is self-evident that $D$ is linearly equivalent to $D + (g)$.