2
$\begingroup$

I'm trying to show all the non-projectively equivalent configurations of linear subspaces.
Two subspaces $\Lambda_1,\Lambda_2$ are projectively equivalent if

$\exists\omega:\mathbb{P}^n\longrightarrow \mathbb{P}^n$ homography such that $\omega(\Lambda_1)=\Lambda_2$.

For example, in $\mathbb{P}^4$ let $l_1,l_2$ be two distinct lines and $P$ a point such that such that $P \notin l_1\cup l_2$.

What's the number of non-projectively equivalent configurations of the two lines and the point $P$?

My idea is to study the dimension of the intersection space of the two lines.
Infact, for Grassmann's formula we have: $$\mathrm{dim}(l_1 \cup l_2)=\mathrm{dim}(l_1)+\mathrm{dim}(l_2)-\mathrm{dim}(l_1 \cap l_2) \le4$$
We know that $\mathrm{dim}(l_1 \cap l_2)= \begin{cases} 0 \\ -1 \end{cases}$

If $\mathrm{dim}(l_1 \cap l_2)=0$ it means that $l_1 \cap l_2=\{Q\}$ (case $1$), while if $\mathrm{dim}(l_1 \cap l_2)=-1$ it means that $l_1 \cap l_2 = \emptyset$ (case $2$). We can now study all these cases.

$1$a) $<l_1,l_2>=\pi$, where $\pi$ is a plane, $P \in \pi$ and $\mathrm{dim}(<l_1,l_2,P>)=2$

$1$b) $<l_1,l_2>=\pi$, where $\pi$ is a plane, $P \notin \pi$ and $\mathrm{dim}(<l_1,l_2,P>)=3$

$2$a) $<l_1,l_2>=\Theta$, where $\Theta$ is a hyperplane, $P \in \Theta$ and $\mathrm{dim}(<l_1,l_2,P>)=3$

$2$b) $<l_1,l_2>=\Theta$, where $\Theta$ is a hyperplane, $P \notin \Theta$ and $\mathrm{dim}(<l_1,l_2,P>)=4$

All these are non-projectively equivalent because a homography $\omega$ doesn't change the dimension of linear subspaces in $\mathbb{P}^n$. Is that correct? Thank you in advance.

$\endgroup$
2
  • $\begingroup$ A projective line means the image by $k^{5}\to\Bbb{P}^4$ of a plane passing through the origin. Thus your equivalent configurations classes are elements of $GL_5(k)\setminus P^3 \ / {\scriptstyle\pmatrix{GL_2(k)&&\\&GL_2(k)&\\&&GL_2(k)}}$ where $P^3$ is the subset of $k^{5\times 6}$ of matrices where the $2j-1$-th and $2j$-th column span a plane. Are you searching for a simpler description of the configurations ? $\endgroup$
    – reuns
    Commented Jun 14, 2020 at 1:12
  • $\begingroup$ Yes... We've never seen this way to solve it. Anyway, I edited my question to explain how we should proceed. $\endgroup$
    – Vajra
    Commented Jun 14, 2020 at 6:39

1 Answer 1

1
$\begingroup$

Yep, you are correct about why your four cases are all distinct. In fact, they are the only distinct cases: any two configurations in the same case are equivalent. To prove this, I find it convenient to pull everything back to $k^5$ (where $k$ is the field) and just think about linear subspaces of $k^5$ using linear algebra.

So, for instance, suppose we are in your case (2a) (similar arguments can be made in your other three cases). Then $\ell_1$ and $\ell_2$ correspond to two-dimensional subspaces $V_1,V_2\subseteq k^5$ with trivial intersection, and $P$ corresponds to a 1-dimensional subspace $W$ which is contained in $V_1+V_2$ but not contained in $V_1$ or $V_2$. Pick a nonzero vector $w\in W$, and write $w=w_1+w_2$ for $w_1\in V_1$ and $w_2\in V_2$ (since $w\in W\subseteq V_1+V_2$). Since $W$ is not contained in $V_1$ or $V_2$, both $w_1$ and $w_2$ are nonzero. So, we can pick $v_1\in V_1$ and $v_2\in V_2$ such that $\{v_1,w_1\}$ and $\{v_2,w_2\}$ are bases for $V_1$ and $V_2$. Finally, pick a nonzero vector $u\in k^5\setminus (V_1+V_2)$, so $\{v_1,w_1,v_2,w_2,u\}$ is a basis for $k^5$.

Now, given any other configuration $(\ell_1',\ell_2',P')$ which is also in case (2a), we can similarly pick a basis $\{v_1',w_1',v_2',w_2',u'\}$. There is then a linear isomorphism $T:k^5\to k^5$ mapping $v_1$ to $v_1'$, $w_1$ to $w_1'$, and so on. This $T$ will then map $V_1$ to $V_1'$, $V_2$ to $V_2'$, and $W$ to $W'$, so it induces a homography that maps $\ell_1$ to $\ell_1'$, $\ell_2$ to $\ell_2'$, and $P$ to $P'$.

$\endgroup$
5
  • $\begingroup$ Maybe I know its meaning but I use a different notation... what is $k^5$? Thank you for your answer! $\endgroup$
    – Vajra
    Commented Jul 3, 2020 at 20:46
  • $\begingroup$ I just mean the $5$-dimensional vector space whose projectivization is $\mathbb{P}^4$. $\endgroup$ Commented Jul 3, 2020 at 20:52
  • $\begingroup$ Ok perfect, thank you for your clear answer. $\endgroup$
    – Vajra
    Commented Jul 3, 2020 at 21:04
  • $\begingroup$ Maybe you didn't write this fact because the 6th point is arbitrary(?) $\endgroup$
    – Vajra
    Commented Jul 3, 2020 at 21:14
  • 1
    $\begingroup$ I'm picking vectors in $k^5$, not points in $\mathbb{P}^4$. $\endgroup$ Commented Jul 3, 2020 at 21:17

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .