I'm studying algebraic geometry and I need some help to understand the Riemann-Roch theorem. Let us consider a holomorphic line bundle $\xi$ over a Riemann surface $X$. The unique invariant of a holomorphic bundle is the equivalence class $[D]$ of the divisor $D$ of an arbitrary section, but it looks like this class is "complicated to calculate".

The most simplest invariant of line bundles is the degree $\text{deg}\,D$ of divisor $D$. In fact, it is just the Euler characteristic. This number is enough to classify all line bundles over $\mathbb{CP}^{1}$.

The second clear invariant is the count of linearly independent holomorphic sections. This number is precisely $l(-D)$.

According to the Riemann-Roch theorem we have the formula $$ l(D) - l'(D) = 1 - g - \text{deg}\,D, $$ that we can use to calculate the $l(D)$. Am I understanding this theorem correctly?

Ok. Could you show me some example of a non-trivial holomorphic line bundle over torus that has degree zero.


  • $\begingroup$ If $p$ and $q$ are distinct points on the torus, then $O(p) \otimes O(-q)$ is a nontrivial line bundle of degree zero. $\endgroup$ – user64687 Jul 9 '14 at 11:00
  • $\begingroup$ What is $O(p)$? $\endgroup$ – Gleb Jul 9 '14 at 11:03
  • $\begingroup$ $O(p)$ is the line bundle associated to the divisor $p$. $\endgroup$ – user64687 Jul 9 '14 at 11:11
  • $\begingroup$ Thank you. Now I understand that it is very easy to construct a line bundle for arbitrary divisor. $\endgroup$ – Gleb Jul 11 '14 at 9:43

Let $\mbox{Pic}^0(X)$ denote the group of all line bundles of degree 0 on $X$. Then it is a classical fact that $\mbox{Pic}^0(X)$ is an abelian variety (that is, a projective complex torus) of dimension $g$ called the Jacobian of $X$ (I'm assuming that $X$ is compact). So basically, for $g\geq1$ there are "plenty" of line bundles of degree 0 that are non-trivial. Actually, for $g\geq1$, $X$ is contained in $\mbox{Pic}^0(X)$; this can be seen by using the line bundle that Asal Beag Dubh described in the comments. Let $q\in X$ be fixed, and define the map $$\phi_q:X\to\mbox{Pic}^0(X)$$ where $p\mapsto\mathcal{O}_X(p)\otimes\mathcal{O}_X(-q)$. This is a holomorphic immersion.

Since we have the exact sequence $$0\to\mbox{Pic}^0(X)\to\mbox{Pic}(X)\stackrel{\deg}{\to}\mathbb{Z}\to0,$$ we see that the degree classifies line bundles on $\mathbb{P}^1$, but for higher genus this is no longer true. For example, for $g=1$, a divisor $n_1p_1+\cdots+n_rp_r$ is linearly equivalent to zero (that is, defines the trivial line bundle) if and only if it is of degree 0 and if the sum $n_1p_1+\cdots+n_rp_r$ in $X$ (remember that $X$ is a torus) is 0. This then generalizes to higher genus by using what I said above, and so a divisor $D$ on $X$ is linearly equivalent to 0 if and only if it is of degree 0 and $\alpha_p(D)=0$ in $\mbox{Pic}^0(X)$ (the map $\alpha_p$ extends linearly to $\mbox{Div}(X)$).

Your Riemann-Roch formula isn't quite right. The Riemann-Roch formula actually says that $$\chi(D):=h^0(D)-h^1(D)=\deg D+1-g.$$ Using Serre duality, $h^1(D)=h^0(K-D)$, where $K$ is a canonical divisor of $X$ (that is, the zero set of a differential form). Here $h^0(D)$ denotes the dimension of the global sections of $D$ and $h^1(D)$ denotes the dimension of the first cohomology group of $D$ (actually of $\mathcal{O}_X(D)$). Another thing you mentioned is $l(-D)$; if $D$ is effective, this is actually the dimension of all holomorphic functions that vanish on $D$, and so equals 0.

  • $\begingroup$ Many thanks! Your answer has been very useful for me. I denoted by $l'(D)$ the space of differential forms which have divisors greater than $-D$. $\endgroup$ – Gleb Jul 11 '14 at 9:50

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.