# Special holomorphic triples on Riemann surfaces and branch divisors

Let $X$ be a compact connected Riemann surface and $E \rightarrow X$ a holomorphic vector bundle of rank 2. Then for any holomorphic sub-line bundle $L \subset E$, there exists a holomorphic section $s$ of $\mathbb{P}(E)$ associated to it.The inverse is also true, that is, for any holomorphic section $s \in \Gamma(X,\mathbb{P}(E))$, we can construct a holomorphic sub-line bundle $L \subset E$.

Now, suppose $E$ is projectively flat, which means that $\mathbb{P}(E)$ is a flat $\mathbb{P}^1$ bundle. We fix a projectively flat structure on $E$, i.e. we choose a family of trivializations $\phi_i: \mathbb{P}(E)|_{U_i} \rightarrow U_i \times \mathbb{P}^1$ such that the transition function $g_{ij}: U_i \cap U_j \rightarrow PGL(2,\mathbb{C})$ is a constant map, where $g_{ij}(x) = (\phi_i \circ \phi_j^{-1})|_{\{x\} \times \mathbb{P}^1}$.

Under this projectively flat structure $(U_i, \phi_i)$ of $\mathbb{P}(E)$, the section $s \in \Gamma(X,\mathbb{P}(E))$ can be re-written as $(U_i, s_i)$ such that

1. $s_i: U_i \rightarrow \mathbb{P}^1$ is holomorphic;
2. $\phi_i \circ s|_{U_i}(x) = (x, s_i(x))$ for $x \in U_i$.

If $s_i$ is not a constant map, we could assume $U_i$ is sufficiently small such that $s_i$ has at most one branch point $p$ in $U_i$ and $p \not\in U_i\cap U_j$ for any $j\neq i$. Then we define a positive divisor $B_s = \sum_{x\in X} \nu_{s_i}(x)\cdot x$ for some $i$ such that $x \in U_i$, where $\nu_{s_i}(x)$ is the branching order.

In summary, for any given holomorphic triple $L \subset E$ on $X$ with a fixed projectively flat structure $(U_i,\phi_i)$ on $E$ such that

• rank($L$)=1, rank($E$)=2.
• The section $s$ corresponding to $L$ is not constant, i.e. $s_i$ is not constant for some $i$.

My questions are

1. Can we determine the divisor $B_s$ directly by $L$ and $E$?
2. What can we say about the relation between $\deg L$, $\deg E$ and $\deg B_s$?

Any comments, suggestion or references are welcome. Thank you!

• It was worked out by Mandelbaum in 70s. I will look for references when I have time. – Misha Feb 9 '17 at 2:01

See the formula in the proof of Corollary 11.3.1 in The Monodromy Groups of Schwarzian Equations on Closed Riemann Surfaces: $$deg(L)= g-1 +(deg(E) -deg(B_s))/2.$$ As for your first question (can we?), the answer is yes, we can: $$g-1$$ comes from the half-canonical bundle of your Riemann surface.