On $\mathbb{CP}^1$ with standard complex structure, how to prove that there are only two types of antiholomorphic involution, given by $$ \tau :[z:w]\mapsto [\bar w, \bar z] \qquad \eta :[z:w]\mapsto [\bar w, -\bar z]\ ?$$

For $\mathbb {CP}^n$, there should be one involution for even $n$, two for odd $n$.

As suggested: in homogeneous coordinates, let us write our involution as $z \to \overline {Az}$ for some $A \in GL_{n+1}(C)$. Then we have $\overline{A} A = \lambda id$ for some real $\lambda$. By rescaling $\lambda$ and taking determinant, we see that for even $n$ only $\overline{A}A=1$ is possible (class $\tau$ above), while for odd $n$ also $\overline{A}A=-1$ is possible (so also $\eta$ above).

  • 5
    $\begingroup$ In even dimension you have only one, but in odd dimension you have two such. I suggest you to write your antiholomorphic involution as the product of the standard one with a biholomorphic map, and then to write down what is such a map. $\endgroup$ Jun 1, 2014 at 16:24
  • $\begingroup$ Cross posted on MSE. $\endgroup$
    – MvG
    Jun 2, 2014 at 6:28
  • $\begingroup$ It's a question about involutive elements in $PGL\setminus PSL$, hence just something about eigenvalues (simple linear algebra). I presume that you are speaking about the standard complex structure; it has nothing to do with the metric. $\endgroup$ Jun 2, 2014 at 6:58
  • $\begingroup$ @JérémyBlanc Do you know a reference to the result? Just to avoid repeatition of this well-known linear algebra computation in a research paper. $\endgroup$ Nov 13, 2019 at 5:37
  • 1
    $\begingroup$ @AlexDegtyarev: no, $PGL_{n+1}(C)=PSL_{n+1}(C)$. It's a question about an overgroup of index 2 of $PGL_{n+1}(C)$. $\endgroup$
    – YCor
    Nov 26, 2019 at 9:54

1 Answer 1


Coda (Lemma 11 from arXiv:2002.01355):

Projective transformations of $\mathbb{C}P^{n}$ act on the solutions of the equation $L\bar L=I$, respectively $L\bar L=-I$, by transformations $L\mapsto U^{-1}L\bar U$, where $U$ is an invertible complex $(n+1)\times (n+1)$ matrix. It remains to show that each solution can be transformed to $I$ and $J=\left(\begin{smallmatrix} 0 & -I\\ I & 0\end{smallmatrix}\right)$ respectively, i.e., each matrix $L$ satisfying $L\bar L=I$ (respectively, $L\bar L=-I$) has form $L=U^{-1}\bar U$ (respectively, $L=U^{-1}J\bar U$ ) for some invertible matrix $U$.

If $L\bar L=I$, then take $a\in\mathbb{R}$ such that $-e^{2ia}$ is not an eigenvalue of $\bar L$. Then $U:=e^{ia}I+e^{-ia}\bar L$ is the required invertible matrix because $UL=(e^{ia}I+e^{-ia}\bar L)L=e^{ia}L+e^{-ia}I=\bar U$ by the equation $\bar LL=L\bar L=I$.

If $L\bar L=-I$, then take $a\in\mathbb{R}$ such that $e^{2ia}$ is not an eigenvalue of $J\bar L$. Then $U:=e^{ia}J+e^{-ia}\bar L$ is the required matrix: $UL=(e^{ia}J+e^{-ia}\bar L)L=e^{ia}JL+e^{-ia}J\bar J=J\bar U$ because $\bar LL=L\bar L=-I=J\bar J$.


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.