Let $L^*$ be the total space of the line bundle $\mathcal{O}_{\mathbb{P}^n}(k)$ minus its zero section.

How can one compute the fundamental group of $L^*$?

For k = 0 the space $L^*$ is $\mathbb{P}^n \times \mathbb{C}^*$ hence $\pi_1(L^*) = \mathbb{Z}$.

For k=-1 the $L^*$ is $\mathbb{C}^{n+1} \setminus \{0\}$, therefore $\pi_1(L^*) = 0$.

What about the other $k$ ?

The long exact sequence of homotopy of a Serre fibration $\mathbb{C}^* \rightarrow L^* \rightarrow \mathbb{P}^n$ gives

$\pi_2(\mathbb{C}^*) = 0 \rightarrow \pi_2(L^*) \rightarrow \pi_2(\mathbb{P}^n) \simeq \mathbb{Z} \rightarrow \pi_1(\mathbb{C}^*)\simeq \mathbb{Z} \rightarrow \pi_1(L^*) \rightarrow \pi_1(\mathbb{P}^n) = 0$.

So one needs to understand the map $\pi_2(\mathbb{P}^n)\rightarrow \pi_1(\mathbb{C}^*)$.

  • 3
    $\begingroup$ You can use van Kampen's theorem to compute this, where the two opens $U$ and $V$ are the inverse images of the standard opens $\mathbb{C}$ in $\mathbb{CP}^1$ whose intersection is $\mathbb{C}^\times$. $\endgroup$ – Jason Starr Sep 28 '17 at 12:47
  • 1
    $\begingroup$ Anyway, the answer is that the fundamental group is isomorphic to $\mathbb{Z}/k\mathbb{Z}$. $\endgroup$ – Jason Starr Sep 28 '17 at 12:50
  • 3
    $\begingroup$ Now I see that $n$ is arbitrary. I thought the OP specified that $n$ should equal $2$. So you need more than two "standard opens" to cover $\mathbb{CP}^n$. However, it does not change the answer. You can remove a codimension $2$ linear subvariety from $\mathbb{CP}^n$, and its inverse image in $L^\times$, without changing the fundamental group. After doing that, there is a covering by two opens $\mathbb{C}^n$ whose intersection is $\mathbb{C}^\times \times \mathbb{C}^{n-1}$. So the same computation works. $\endgroup$ – Jason Starr Sep 28 '17 at 12:55

The fibration $\mathbb{C}^\times\to L^\times\to\mathbb{P}^n$ can be "delooped" to a fibration $L^\ast \to\mathbb{P}^n\to{\rm B}\mathbb{C}^\times$ where the last map is the classifying map for the line bundle. Now we have ${\rm B}\mathbb{C}^\times\cong\mathbb{P}^\infty$, and we want to identify the map $\pi_2(\mathbb{P}^n)\to\pi_2(\mathbb{P}^\infty)\cong\pi_1(\mathbb{C}^\times)$ induced by the classifying map for $\mathcal{O}(k)$. Note that the generator for $\pi_2(\mathbb{P}^\infty)$ is given by the natural inclusion $\mathbb{P}^1\to\mathbb{P}^\infty$ classifying $\mathcal{O}(1)$. In particular, the map $\pi_2(\mathbb{P}^n)\to\pi_2(\mathbb{P}^\infty)$ classifying $\mathcal{O}(k)$ must be multiplication by $k$. Of course, the result is the same as in Jason Starr's comments. Just another funny way to see it.


Another way to see that the fundamental group is $\mathbb{Z}/k$ is using geometry of cyclic quotient singularities and it goes as follows.

We consider the negative twists first. In this case the total space $\mathcal{O}_{\mathbb{P}^n}(-k)$ is the resolution of singularities of the quotient space $\mathbb{C}^{n+1} / G$ with $G=\mathbb{Z}/k$ acting diagonally by $k$-th roots of unity. To see this one computes the coordinate algebra of the quotient which is precisely the affine cone over the degree $k$ Veronese embedding of $\mathbb{P}^n$. Now the latter cone is resolved by a single blow up of its vertex and the resulting space is the total space $\mathcal{O}_{\mathbb{P}^n}(-k)$.

In particular the complement to the zero section of $\mathcal{O}(-k)$ is isomorphic to the punctured cone $(\mathbb{C}^{n+1} - 0) / G$, and so has fundamental group isomorphic to $G$ as $\mathbb{C}^{n+1} - 0$ is simply-connected.

Now if the twist was positive, contracting zero section is impossible, but replacing $k$ by $-k$ makes a homeomorphic space, so it is $\mathbb{Z}/k$ in both cases; and the answer is $\mathbb{Z}$ for $k=0$ as the bundle is trivial then.


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.