2
$\begingroup$

My question will rely on a clarification of a proof, which I simply don't understand.

Let us denote by $X$ a pseudo-riemannian symmetric space and define
$$ Z_{\mathrm{Iso}\left(X\right)}G(X) = \{\, f \in \mathrm{Iso}\!\left(X\right) \mid gf = fg \text{ for all } g \in G(X) \,\} $$ as the centralizer of the transvection group $G(X)$ of the space $X$ and let $\mathrm{Iso}\left(X\right)$ be the isometry group of the regarding space.

In Cahen,Parker,"Pseudo-riemannian symmetric spaces" following assertion has been stated in Proposition 3.9:

Assertion: Any discrete subgroup $A$ of $Z_{\mathrm{Iso}\left(X\right)}G(X)$ acts properly discontinuously on $X$.

Proof: Let $C \subseteq X$ be a compact subset of $X$ with the property that $A_C = \{\, g \in A \mid gC \cap C \neq \emptyset \,\}$ is infinite, $\lvert A_C \rvert = \infty$. There exists a sequence $(a_n) \subseteq A_C$ with distinct elements and a sequence $(c_n) \subseteq C$ such that $a_n c_n \in C$ for all $n \in \mathbb{N}$. By compactness of $C$ and therefore taking sub-sequences, we may assume that $(c_n)$ converges towards $c \in C$ and that $(a_nc_n)$ converges towards $c^\prime \in C$. Let us choose $c$ as a basepoint of $X$. Since the transvection group $G(X)$ acts transitively on $X$ we can find a $g \in G(X)$ such that $gc = c^\prime$ and let $V_1 \subseteq U_1 \subseteq W_1$ be three neighborhoods of the identity in $G(X)$ such that $$ V_1=V_1^{-1}, g^{-1}V_1g\subseteq U_1 \text{ and } U_1^2 \subseteq W_1. $$ Define; $$ W_c = V_1 c, \\ \tilde{W}_c = W_1 c,\\ W_{c^\prime}=gW_c\\\tilde{W}_{c^\prime}=g\tilde{W}_c. $$

Question: Why does this define neighborhoods for the regarding points $c$ and $c^\prime$ and furthermore why does this define a neighborhoodbasis?

There exists an integer $N$ such that for all $n \geq N$ we have $c_n \in W_c$ and $a_nc_n \in W_{c^\prime}$. Indeed, there exists $g_n \in V_1$ such that $c_n = g_nc$ for all $n \geq N$. We want to show that $\lim_{n \to \infty} (a_nc) = c^\prime$. Consider therefore $$ a_n c = a_ng_n^{-1}c_n = g_n^{-1} a_n c_n \in g_n^{-1} g W_c \\ = gg^{-1}g_n^{-1}gW_c \subseteq gU_1V_1c \subseteq gU_1^2c \subseteq gW_1c = \tilde{W}_{c^\prime}, $$ which shows the claimed result.

The sequence $(b_n = a_{n+1}^{-1} a_n)$ in $A$ is clearly such that $\lim_{n \to \infty}b_n c = c$.

Question: Unfortunatley I have to redefine the term "clearly" into "highly unclearly". To me it is not obvious why this series converges towards $c$.

$\endgroup$
1
$\begingroup$

Question 1: This is a standard Baire category argument. Let $U$ be some small neighborhood of $c$. Since $G(X)$ acts continuously on $X$, there is a neighborhood $V$ of the identity in $G(X)$, such that $V^{-1} V c \subseteq U$. Choose a neighbourhood $V_1$ of the identity, such that the closure $\overline{V_1}$ is contained in $V$, and is compact. Since $G(X)$ is separable, it is covered by countably many translates $\{g_i V_1\}$ of $V_1$. Then $X$ is the union of $\{g_i V_1 c\}$, so the Baire Category Theorem implies that the closure of some $g V_1 c$ has nonempty interior. Since $g$ acts by a homeomorphism on $X$, this means that $\overline{V_1c}$ has nonempty interior. Since $\overline{V_1}$ is compact, we know that $\overline{V_1}c$ is closed, so $\overline{V_1c} \subseteq \overline{V_1}c \subseteq Vc$. So $Vc$ contains an open set. Then $V^{-1} V c$ contains a neighborhood of $c$. By assumption, this neighborhood is contained in $U$.


Question 2: Since $a_n c \to c'$, we know that if $n$ is large, then there exist $w,w'$ close to the identity in $G(X)$, such that $a_n c = w' c'$ and $a_{n+1}c = w c'$. Since $a_{n+1}$ commutes with $w$ and $w'$, we have $$ a_{n+1}^{-1} (a_n c) = a_{n+1}^{-1} (w' c') = w' w^{-1} (a_{n+1}^{-1} w c') = w' w^{-1} c .$$ Since $w'$ and $w$ are both close to the identity, this is close to $c$. So $a_{n+1}^{-1} a_n c \to c$.

$\endgroup$

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.