$\newcommand{\Cof}{\operatorname{cof}}$ $\newcommand{\id}{\operatorname{Id}}$ $\newcommand{\End}{\operatorname{End}}$ $\newcommand{\GL}{\operatorname{GL}}$

Let $V$ be a real $d$-dimensional vector space ($d \ge 4$). Let $2 \le k \le d-2$ be odd. Define $H_{>k}=\{ A \in \text{End}(V) \mid \operatorname{rank}(A) > k \}$. $H_{>k}$ is an open submanifold of $ \text{End}(V)$.

We also define, for a given number $s$, the open submanifold

$$\tilde H_{>s}=\{ B \in \End(\bigwedge^k V) \mid \operatorname{rank}(B) > s \} \subseteq \End(\bigwedge^k V),$$

where $\bigwedge^{k} V$ is the $k$-th exterior power of $V$.

Consider the map $$ \psi:H_{>k} \to \tilde H_{>k} \, \,, \, \, \psi(A)=\bigwedge^{k}A, %\psi:H_r \to \text{End}(\bigwedge^{k}V) \, \,, \, \, \psi(A)=\bigwedge^{k}A, $$

$\psi$ is a smooth injective immersion but not an embedding. (The injectivity uses the fact $k$ is odd, since otherwise $\psi(A)=\psi(-A)$).

Question: Is $\text{Image}(\psi)=\psi(H_{>k})$ closed in $\tilde H_{>k}$?

Here are some thoughts:

Let's try prove the image is closed. Let $A_n \in H_{>k}$, and suppose $\psi(A_n)=\bigwedge^k A_n$ converges to some $D \in \tilde H_{>k}$. We can assume all the $A_n$'s have the same rank $r$. If $\text{rank}(D)=\binom {r}{k}$ then we are done, since the restriction of $\psi$ to the space of matrices of rank $r$ is proper, as I explain below.

The problem is that the rank of the limit $D$ can fall below the shared rank of the $\psi(A_n)$. We can ask whether or not $D \in \tilde H_{\binom {i}{k}}$ for some $i>k$? That is, if the rank of $D$ "falls", must it fall to another "legal" value, which is one of the values that are obtainable from endomorphisms of $V$?

Explanation regarding the stratified/filtered structure:

Note that $H_{>k}=\cup_{r=k+1}^dH_r$, where $H_{r}=\{ A \in \text{End}(V) \mid \operatorname{rank}(A) = r \}$.

Denote $\tilde H_s=\{ B \in \text{End}(\bigwedge^kV) \mid \operatorname{rank}(B) = s \}$. If $\operatorname{rank}(A) = r $, then $\operatorname{rank}(\bigwedge^kA) = \binom {r}{k} $, which means $\psi(H_r) \subseteq \tilde H_{\binom {r}{k}}$, so we have

$$ \psi(H_{>k})= \cup_{r=k+1}^d \psi(H_r) \subseteq \cup_{r=k+1}^d \tilde H_{\binom {r}{k}}. $$

I know that each restriction $\psi|_{H_r}:H_r \to \tilde H_{\binom {r}{k}}$ is proper, hence $\psi(H_r)$ is an embedded closed submanifold of $\tilde H_{\binom {r}{k}}$.

  • $\begingroup$ I suggest you try the case of $d=3, k=2$ to see that $\psi(H_k)$ is not closed. $\endgroup$ – Mohan Jun 4 '18 at 15:42
  • $\begingroup$ Thanks, you are right. I meant to suggest $\psi(H_k)$ is closed in a certain subspace of $End(\bigwedge^k V)$. I have edited the question to reflect this. $\endgroup$ – Asaf Shachar Jun 5 '18 at 5:55

The answer is negative. The image of $\psi$ is not closed:

Let $d=5,k=3$:

Set $A_n=\text{diag}(n,\frac{1}{\sqrt{n}},\frac{1}{\sqrt{n}},\frac{1}{\sqrt{n}},\frac{1}{\sqrt{n}}) \in H_{>3}$. Then $\psi(A_n)=\bigwedge^3 A_n=\text{diag}(1,1,1,1,1,1,(\frac{1}{\sqrt{n}})^3,(\frac{1}{\sqrt{n}})^3,(\frac{1}{\sqrt{n}})^3,(\frac{1}{\sqrt{n}})^3) $ converges to $D=\text{diag}(1,1,1,1,1,1,0,0,0,0) \in \tilde H_{>3}$.

However, $D \notin \psi(H_{>3})$ since it does not have the right rank:

$\text{rank}(D)=6$, and $\psi(H_{>3})= \psi(H_4)\cup \psi(H_5) \subseteq \tilde H_{\binom {4}{3}} \cup H_{\binom {5}{3}}=\tilde H_4 \cup \tilde H_{10}$.

Similar examples exist when $k$ is even.


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.