3
$\begingroup$

In one the the answers to this thread " Can one embedd the projectivezed tangent space of CP^2 in a projective space? " it was mentioned that " $\mathbb{P}(T\mathbb{P}^2)$ isomorphic to the variety of complete flags in the vector space $\mathbb{C^3}$ ".

I'm having a hard time understanding why this is true, and can't seem to find any references.

$\endgroup$
2
  • $\begingroup$ Someone answered this and I was reading it, but then my page refreshed and it was gone... does that just mean they deleted their answer? $\endgroup$
    – Lauren
    May 22, 2014 at 20:58
  • $\begingroup$ One way to look at it is that if $z\in \mathbb{P}^2$, corresponding to a line $L_z\subset \mathbb{C}^3$, then a line in $T_z$ corresponds uniquely to a direction orthogonal to $L_z$, which corresponds uniquely to a $2$-dimensional subspace of $\mathbb{C}^3$ containing $L_z$. $\endgroup$ May 22, 2014 at 21:01

2 Answers 2

2
$\begingroup$

Let $\pi:\mathbb{P}(T_{\mathbb{P}^2})\rightarrow\mathbb{P}^2$ be the projectivized tangent bundle. The point $x:=(p,[L])\in\mathbb{P}(T_{\mathbb{P}^2})$ corresponds to the point $p = \pi((p,[L]))\in\mathbb{P}^2$ and to the class of the line $L\subset\mathbb{P}^2$ passing through $p$. Now, the point $p$ is a line through the origin $V_p\subset\mathbb{C}^3$, the line $L_p$ corresponds to a plane $\Pi_p\subset\mathbb{C}^3$. Clearly, we have $V_p\subset\Pi_p\subset\mathbb{C}^3$. Now, the morphism $$\phi:\mathbb{P}(T_{\mathbb{P}^2})\rightarrow F(1,2,3),\: x:=(p,[L])\mapsto (V_p,\Pi_p),$$ is an isomorphism because $\phi$ is injective, and $\mathbb{P}(T_{\mathbb{P}^2})$, $F(1,2,3)$ are both smooth.

$\endgroup$
7
  • $\begingroup$ Thanks! I think the main problem I'm having is that I'm not understanding $T\mathbb{P}^2$ correctly. Given a point $z$ in $\mathbb{P}^2$, how to you describe the fiber over the point $z$; i.e. what is $T_z\mathbb{P}^2$? And then my understanding is that you projectivize each point in $T_z\mathbb{P}^2$ to obtain the fiber over $z$ in $\mathbb{P}(T\mathbb{P}^2)$. $\endgroup$
    – Lauren
    May 22, 2014 at 23:21
  • $\begingroup$ Exactly, a point in the fiber of $\mathbb{P}(T_\mathbb{P}^2)$ over $z\in\mathbb{P}^2$ corresponds to a line in $\mathbb{P}^2$ through $z$. $\endgroup$
    – Puzzled
    May 23, 2014 at 9:01
  • $\begingroup$ And a point in the fiber of $T\mathbb{P}^2$ over $z\in\mathbb{P}^2$, would that be a 2-dimensional subspace of $\mathbb{C}^3$ containing the line <z>? $\endgroup$
    – Lauren
    May 24, 2014 at 15:00
  • $\begingroup$ Exactly. That's it. $\endgroup$
    – Puzzled
    May 24, 2014 at 15:22
  • $\begingroup$ @Lauren Another way of thinking about it is as follows: Given a point $\ell\in \Bbb P^2$, the space $T_\ell \Bbb P^2$ is simply $\operatorname{Hom}(\ell,\ell^\perp)$, where $\ell^\perp$ denotes the orthogonal complement with respect to the standard inner product on $\Bbb C^3$. Now pick an element $[v]$ of $\Bbb P(T_\ell \Bbb P^2)$. Then any representative of $[v]$ corresponds to a linear map $\ell\to \ell^\perp$. Thus, picking any $0\neq z\in \ell$, it is sent to $w\in \ell^\perp$. [continued in the next comment] $\endgroup$
    – Danu
    Dec 5, 2018 at 10:59
3
$\begingroup$

I will take $P^2$ to mean the space of lines in $C^3$. The tangent space at a particular point in $P^2$ (say represented by a line $L$) is a linear map from $L$ to $C^3/L$. Let $\mathcal{L}$ be the line bundle whose total space is $\{(L,x) \in P^2 \times C^3 \mid x \in L\}$ (this is $O(-1)$ but it doesn't matter). Then from the first fact, the tangent bundle is $Hom(\mathcal{L}, C^3/\mathcal{L}) \cong \mathcal{L}^* \otimes (C^3/\mathcal{L})$ where I use $C^3$ to be the trivial bundle $C^3 \times P^2$.

The projectivization of a vector bundle ignores tensoring with line bundles, so $P(TP^2)$ is the same as the space of projectivization of $C^3/\mathcal{L})$. So its points correspond to a choice of line $L$ and a choice of line in $C^3/L$. The latter is equivalent to a 2-dimensional subspace in $C^3$ containing $L$.

In general, $P(TP^n)$ is just the space of partial flags of type $(1,2)$ in $C^{n+1}$.

EDIT. A derivation of the above fact for tangent spaces of projective spaces can be found here: http://concretenonsense.wordpress.com/2009/08/17/tangent-bundle-of-the-grassmannian/

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.