Let $\mathbb P_1$ be the one dimensional complex projective space. What is the connected component of the full automorphism of $\mathbb C^*\times \mathbb P_1$. Is it a complex Lie group? I mean is it finite dimensional?

We know that ${\rm Aut}(\mathbb C)^\ast$ is ${\mathbb C}^\ast$, and $Aut^\circ(\mathbb P_1)$ is $PSL(2,\mathbb C)$.


It is not a finite dimensional Lie group. For example, all of the maps $$ \Phi\bigl(z,[a,b]\bigr) = \bigl(z, [a+p(z)b,\ b]\bigr), $$ where $p:\mathbb{C}^\ast\to \mathbb{C}$ is holomorphic, belong to this group. More generally, the automorphism group is essentially the semi-direct product of $\mathbb{C}^\ast$ with the group of holomorphic mappings $p:\mathbb{C}^\ast\to PSL(2,\mathbb{C})$.

To see this, note that the foliation of $\mathbb{C}^\ast\times \mathbb{P}^1$ by the fibers $\{z\}\times \mathbb{P}^1$ for $z\in\mathbb{C}^\ast$ must be preserved by any automorphism, since these are the only compact complex curves that represent a generator of the second homology group (with the orientation given by the complex structure). Thus, every automorphism $\Phi$ has a corresponding automorphism $\phi:\mathbb{C}^\ast\to\mathbb{C}^\ast$ that intertwines $\Phi$ and $\phi$ with the projection $\mathbb{C}^\ast\times\mathbb{P}^1\to\mathbb{C}^\ast$ onto the first factor. Of course $\phi$ is represented by scalar multiplication by a nonzero complex number $\lambda$, and the map $\Phi\mapsto\lambda$ is a homomorphism from this automorphism group to $\mathbb{C}^\ast$. The kernel of this homomorphism is clearly the group of holomorphic mappings $p:\mathbb{C}^\ast\to PSL(2,\mathbb{C})$, since $PSL(2,\mathbb{C})$ is the full group of automorphisms of $\mathbb{P}^1$.

An historical note: Of course, Lie (and Cartan) would certainly have regarded this automorphism group as a Lie group (albeit not one of finite dimension), since it can be defined as the set of solutions of a system of differential equations. By contrast, this is not so for the group $\text{Aut}(\mathbb{C}^2)$, which cannot be defined as the set of solutions of a system of differential equations.


I have troublesome understanding Robert's answer, but since I can not give comments, I write my comments here. If the maps $\Phi : (z,[a,b])\rightarrow (z,[a+p(z),b])$ were an isomorphism, then the maps $[a,b]\mapsto [a+p(z),b]$ must be an isomorphism of $\mathbb{P}^1$. But I think the latter can not be a map from $\mathbb{P}^1$ to itself, except when $p(z)$ is the zero function. For example, the map is not homogeneous in a and b.

  • $\begingroup$ Oh, you are right. That's a typo that I will fix. It should be $[a,b]\mapsto [a + p(z)b,b]$. $\endgroup$ – Robert Bryant Oct 16 '12 at 17:27
  • $\begingroup$ Why can't you give comments? $\endgroup$ – Robert Bryant Oct 16 '12 at 17:28
  • 3
    $\begingroup$ That requires sufficient reputation. But answering doesn't. $\endgroup$ – Allen Knutson Oct 16 '12 at 17:42
  • 2
    $\begingroup$ @Allen: Ah! I didn't know that. I really must read the MO directions some time; they probably explain why we have this rule, which, apparently, doesn't allow questions from beginners about someone's answer, except as another answer (which seems weird). $\endgroup$ – Robert Bryant Oct 16 '12 at 20:37

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.