I'm studying an introduction to line bundles and I'm struggling with a particular proof. I'm following some notes that present this theorem:

Let $ L \rightarrow X $ be a holomorphic line bundle on a compact manifold $X$. Suppose $ s \in Γ(X, L) $ is not the zero section. Then either one of the following holds:

  1. $L \cong X × C $ is the trivial bundle and $s$ has no zeros.
  2. $Γ(X, L^{−1}) = 0$ and the section $s$ admits at least one zero.

(Note: here $L^{-1}$ denotes the dual bundle of $L$.)

To prove this the author considers a global section $t$ of the dual bundle and, considering local descriptions $L ↔ \{U_α, g_{α\beta}\}$, $L^{-1} ↔ \{U_α, g_{α\beta}^{-1}\}$, $s ↔ \{U_α, s_α\}$ and $t ↔ \{U_α, t_α\}$ he observes that on every $U_\alpha \cap U_\beta$ we have $s_\alpha t_\alpha = s_\beta t_\beta$, so we can glue these $s_\alpha t_\alpha$ in a "global" function $F: X \rightarrow \mathbb{C}$. By the maximum principle, since $X$ is compact, $F \equiv c \in \mathbb{C}$.

Now if $c\neq0$ we have the first result; otherwise if $c=0$ we have that $t$ must be the zero section, thus implying $\Gamma(X,L^{-1}) = 0$. I'm not sure about this last implication: since $F$ depends on the choice of $t$ also $c$ does; what if we have another section $t'\in \Gamma(X,L^{-1})$ s.t. $c_{t'} \neq 0$? Would this lead to a contradiction I'm not able to see?


Suppose that $L$ is not the trivial bundle, for every section $t$, you can construct $F_t=c_t$, then $c_t$ has to be zero for every chosen $t$, (since $t$ vanishes in an open subset (and $M$ is connected)) we deduce that $t=0$.

  • $\begingroup$ I'm sorry, I just edited my question (in the last sentences I wanted to know what happens if I find $t'$ s.t. $c_{t'}\neq 0$. By the way, why can I assume that $L$ is not the trivial bundle? I only know that a trivial bundle has always a nowhere vanishing global section: if $s$ has this property and $t$ is s.t. $c_t \neq 0$ I can see no contradiction. $\endgroup$ – M. Rinetti Oct 19 '18 at 13:25
  • $\begingroup$ You want to show that there are two cases: 1. $L$ is trivial, 2. $\Gamma(X,L^{-1})=0$. To achieve this, you can show that if $L$ is not trivial, the second case holds. $\endgroup$ – Tsemo Aristide Oct 19 '18 at 13:28
  • $\begingroup$ If you assume that $L$ is not trivial, you cannot have $c_{t}\neq 0$ for any $t$ since it implies that $L$ is trivial, so $c_t=0$ for every $t$. $\endgroup$ – Tsemo Aristide Oct 19 '18 at 13:31
  • $\begingroup$ What if I have $c = 0$ and $L$ trivial? I also have to show that when $L$ is trivial $s$ has not zeros, but supposing $L$ is not trivial whenever I find $c = 0$ I'm discarding a possible case, am I not? $\endgroup$ – M. Rinetti Oct 19 '18 at 13:43

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.