3
$\begingroup$

Let $G=\operatorname{GL}_n(\Bbb C)$ act polynomially on some finite-dimensional complex vector space $V$. This means that the action is given by a morphism $\rho\colon G\to\operatorname{GL}(V)$ of algebraic groups which extends to a morphism $\Bbb C^{n\times n}\to\operatorname{GL}(V)$.

Let $v\in V$ be a point with finite stabilizer $H:=G_v$. Assume that there is a sequence $(g_k)_{k\in\Bbb N}$ with $g_k\in G$ such that the sequence $(g_k.v)_{k\in\Bbb N}$ converges to $v$ itself.

Is there a subsequence of $(g_k)$ that converges in $G$ (necessarily to a point of $H$)?

Notably, if there are $h_k\in H$ such that $h_kg_k$ converges, then the answer is affirmative. The latter no longer depends on $H$ being finite. Just an observation, though.

$\endgroup$
0

1 Answer 1

3
$\begingroup$

Yes, there is always a subsequence which converges:

Since $H$ is finite there exists an $\epsilon>0$ such that the sets $h\cdot B(1,\epsilon)$ are disjoint, where $B(1,\epsilon)=\{x|d(1,x)\leq\epsilon\}$.

Consider the continuous map $f:G\to V$ which sends $g$ to $g\cdot v$. This is a polynomial map which is also locally injective (due to the finiteness of $H$). This already implies that this map is a local homeomorphism. It the follows that there is a $\delta>0$ such that $f^{-1}(B(v,\delta))\subseteq \cup_{h\in H} h\cdot B(1,\epsilon)$.

Assume now that $\{g_k\cdot v\}$ converges to $v$. It follows that for almost all $k$ we have that $g_k\in h\cdot B(1,\epsilon)$ for some $h\in H$. Since $H$ is finite, this implies that for some $h\in H$ there are infinitely many $k$'s such that $g_k\in h\cdot B(1,\epsilon)$. These infinitely many $k$'s gives us a subsequence which converges to $h\in H$.

$\endgroup$
2
  • $\begingroup$ Thanks! Could you add more detail for the statement that $f^{-1}(B(v,\delta))\subseteq\bigcup_{h\in H} h\cdot B(1,\varepsilon)$? This is precisely where I failed in an attempt to verify this myself. It's probably just correct application of the inverse function theorem, but I would really appreciate some details. $\endgroup$ Jan 22, 2017 at 16:50
  • $\begingroup$ It is indeed exactly that. The idea is that since $f$ has no singular points $f$ must be open. But then $f(h\cdot B(1,\epsilon)$ is open, and therefore contains a small ball around $v$, say with radius $\delta$. Now take the inverse image of that Ball. $\endgroup$
    – Ehud Meir
    Jan 22, 2017 at 16:52

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.