<ph f="cmex"> </ph><ph f="cmbx">On the realization of Riemannian symmetric spaces in Lie groups II </ph> <math xmlns="http://www.w3.org/1998/Math/MathML"> <msup> <mrow> <mi> </mi> </mrow> <mrow> <mn>1</mn> </mrow> </msup> </math>

### Zhengdong Wang

School of mathematical science, Peking University, Beijing, 100871, P. R. China E-mail address : anjinpeng@pku.edu.cn School of mathematical science, Peking University, Beijing, 100871, P. R. China E-mail address : zdwang@pku.edu.cn
Abstract. In this paper we generalize a result in [1, showing that an arbitrary Riemannian symmetric space can be realized as a closed submanifold of a covering group of the Lie group defining the symmetric space. Some properties of the subgroups of fixed points of involutions are also proved.
Keywords. Symmetric space, Involution, Embedded submanifold

1 Introduction

Suppose $G$  is a connected Lie group with an involution $\sigma$  . Then the Lie algebra $\mathfrak{g}$  of $G$  has a canonical decomposition $\mathfrak{g}=\mathfrak{k}\oplus \mathfrak{p}$  , where $\mathfrak{k}$  and $\mathfrak{p}$  are the eigenspaces of $d\sigma$  in $\mathfrak{g}$  with eigenvalues $1$  and $-1$  , respectively. Let ${G}^{\sigma }=\left\{g\in G|\sigma \left(g\right)=g\right\}$  , and suppose $K$  is an open subgroup of ${G}^{\sigma }$  . Suppose moreover that $A{d}_{G}\left(K\right){|}_{\mathfrak{p}}$  is compact, that is, $G/K$  has a structure of Riemannian symmetric space. In the particular case that $K={G}^{\sigma }$  , it was proved in [1that $P=exp\left(\mathfrak{p}\right)$  is a closed submanifold of $G$  , and there is a natural isomorphism $G/{G}^{\sigma }\sim =P$  . This gave a realization of the symmetric space $G/{G}^{\sigma }$  in $G$  . In this paper we generalize this result to the case of arbitrary symmetric space $G/K$  , that is, the case that $K$  is an arbitrary open subgroup of ${G}^{\sigma }$  such that $A{d}_{G}\left(K\right){|}_{\mathfrak{p}}$  is compact.
In Section 2 we will make some preparation for this generalized realization. Some properties of the subgroup ${G}^{\sigma }$  will be examined. We will prove the following results.
• $\bullet$  There exists a covering group ${G}^{\prime }$  of $G$  with covering homomorphism $\pi$  such that ${\pi }^{-1}\left(K\right)=\left({G}^{\prime }{\right)}^{{\sigma }^{\prime }}$  , where ${\sigma }^{\prime }$  is the induced involution of $\sigma$  on ${G}^{\prime }$  .
• $\bullet$  The quotient group ${G}^{\sigma }/{G}_{0}^{\sigma }$  is isomorphic to $\left({\mathbb{Z}}_{2}{\right)}^{r}$  for some non-negative integer $r$  , where ${G}_{0}^{\sigma }$  is the identity component of ${G}^{\sigma }$  .
• $\bullet$  ${G}^{\sigma }$  is connected if $G$  is simply connected.
In Section 3 we will give the precise statement of the realization of arbitrary symmetric spaces in Lie groups. Briefly speaking, a symmetric space $G/K$  is diffeomorphic to a closed submanifold ${P}^{\prime }$  of a covering group ${G}^{\prime }$  of $G$  , where ${G}^{\prime }$  is chosen such that ${\pi }^{-1}\left(K\right)=\left({G}^{\prime }{\right)}^{{\sigma }^{\prime }}$  . The idea of the proof is that $G/K\sim ={G}^{\prime }/{\pi }^{-1}\left(K\right)={G}^{\prime }/\left({G}^{\prime }{\right)}^{{\sigma }^{\prime }}\sim ={exp}_{{G}^{\prime }}\left(\mathfrak{p}\right)={P}^{\prime }$  .
We would like to thank the referee of [1, whose suggestion of correction motivated us considering the problems of the paper.

2 The subgroups of fixed points of involutions

For a Lie group $H$  with an automorphism $\theta$  , we always denote ${H}^{\theta }=\left\{h\in H|\theta \left(h\right)=h\right\}$  , and denote the identity component of ${H}^{\theta }$  by ${H}_{0}^{\theta }$  . In this section we prove the following two theorems. The realization of arbitrary symmetric spaces in Lie groups will be based on Theorem  2.1 .
Theorem 2.1. Let $G$  be a connected Lie group with an involution $\sigma$  , K an open subgroup of ${G}^{\sigma }$  . Then there is a covering group ${G}^{\prime }$  of $G$  with covering homomorphism $\pi$  such that ${\pi }^{-1}\left(K\right)=\left({G}^{\prime }{\right)}^{{\sigma }^{\prime }}$  , where ${\sigma }^{\prime }$  is the induced involution of $\sigma$  on ${G}^{\prime }$  .
Theorem 2.2. Let $G$  be a connected Lie group with an involution $\sigma$  . Then the quotient group ${G}^{\sigma }/{G}_{0}^{\sigma }$  is isomorphic to $\left({\mathbb{Z}}_{2}{\right)}^{r}$  for some non-negative integer $r$  , and $r=0$  if $G$  is simply connected.
First we introduce some notations. Let $G$  be a connected Lie group with an involution $\sigma$  . For $g,h\in G$  , denote the set of all continuous paths $\gamma :\left[0,1\right]\to G$  with $\gamma \left(0\right)=g$  and $\gamma \left(1\right)=h$  by $\Omega \left(G,g,h\right)$  , and denote ${\pi }_{1}\left(G,g,h\right)=\left\{\left[\gamma \right]|\gamma \in \Omega \left(G,g,h\right)\right\}$  , where $\left[\gamma \right]$  is the homotopy class relative to endpoints determined by $\gamma$  . Note that ${\pi }_{1}\left(G,g,g\right)$  is the fundamental group ${\pi }_{1}\left(G,g\right)$  of $G$  with basepoint $g$  . For $g\in {G}^{\sigma }$  and ${\gamma }_{g}\in \Omega \left(G,e,g\right)$  , let ${\gamma }_{g}^{\sigma }\in \Omega \left(G,e,e\right)$  be defined as ${\gamma }_{g}^{\sigma }\left(t\right)=\left\{\begin{array}{cc}{\gamma }_{g}\left(2t\right)& t\in \left[0,\frac{1}{2}\right];\\ \sigma \left({\gamma }_{g}\left(2-2t\right)\right)& t\in \left[\frac{1}{2},1\right].\end{array}$  The set ${\pi }_{g}=\left\{\left[{\gamma }_{g}^{\sigma }\right]|{\gamma }_{g}\in \Omega \left(G,e,g\right)\right\}$  is a subset of ${\pi }_{1}\left(G,e\right)$  . For ${g}_{1},{g}_{2}\in {G}^{\sigma }$  , if they belong to the same coset space of ${G}_{0}^{\sigma }$  , it is obvious that ${\pi }_{{g}_{1}}={\pi }_{{g}_{2}}$  . So we can define ${\pi }_{\left[g\right]}={\pi }_{g}$  for $g\in {G}^{\sigma }$  , where $\left[g\right]$  denotes the coset space $g{G}_{0}^{\sigma }$  . We denote the identity element of ${\pi }_{1}\left(G,e\right)$  by $\text{e}$  .
Lemma 2.3. For $g\in {G}^{\sigma }$  , ${\pi }_{\left[g\right]}^{-1}={\pi }_{\left[g\right]}$  , that is, $x\in {\pi }_{\left[g\right]}⇔{x}^{-1}\in {\pi }_{\left[g\right]}$  .
• Proof. It is obvious from the equation $\left[{\gamma }_{g}^{\sigma }{\right]}^{-1}=\left[\left(\sigma \circ {\gamma }_{g}{\right)}^{\sigma }\right]$  .
Lemma 2.4. For $g\in {G}^{\sigma }$  , ${\pi }_{\left[g\right]}={\pi }_{\left[{g}^{-1}\right]}$  .
• Proof. For $\left[{\gamma }_{g}^{\sigma }\right]\in {\pi }_{\left[g\right]}$  , $\left[\left({\gamma }_{g}^{-1}{\right)}^{\sigma }\right]\in {\pi }_{\left[{g}^{-1}\right]}$  , where ${\gamma }_{g}^{-1}\left(t\right)={\gamma }_{g}\left(t{\right)}^{-1}$  . But by Lemma 16.7 in [4, $\left[{\gamma }_{g}^{\sigma }\right]\cdot \left[\left({\gamma }_{g}^{-1}{\right)}^{\sigma }\right]=\left[{\gamma }_{g}^{\sigma }\right]\cdot \left[\left({\gamma }_{g}^{\sigma }{\right)}^{-1}\right]=\left[{\gamma }_{g}^{\sigma }\left({\gamma }_{g}^{\sigma }{\right)}^{-1}\right]=\text{e}$  , where $\left({\gamma }_{g}^{\sigma }\left({\gamma }_{g}^{\sigma }{\right)}^{-1}\right)\left(t\right)={\gamma }_{g}^{\sigma }\left(t\right)\left({\gamma }_{g}^{\sigma }{\right)}^{-1}\left(t\right)$  . So by Lemma  2.3 , $\left[{\gamma }_{g}^{\sigma }\right]=\left[\left({\gamma }_{g}^{-1}{\right)}^{\sigma }{\right]}^{-1}\in {\pi }_{\left[{g}^{-1}\right]}^{-1}={\pi }_{\left[{g}^{-1}\right]}$  . This proves ${\pi }_{\left[g\right]}\subset {\pi }_{\left[{g}^{-1}\right]}$  . By the symmetry, we have the equality.
Lemma 2.5. For ${g}_{1},{g}_{2}\in {G}^{\sigma }$  , $\left[{\gamma }_{{g}_{1}}^{\sigma }\right]\cdot {\pi }_{\left[{g}_{2}\right]}={\pi }_{\left[{g}_{1}{g}_{2}\right]}$  for each $\left[{\gamma }_{{g}_{1}}^{\sigma }\right]\in {\pi }_{\left[{g}_{1}\right]}$  .
Hence ${\pi }_{\left[{g}_{1}\right]}\cdot {\pi }_{\left[{g}_{2}\right]}={\pi }_{\left[{g}_{1}{g}_{2}\right]}$  .
• Proof. Let $\left[{\gamma }_{{g}_{2}}^{\sigma }\right]\in {\pi }_{\left[{g}_{2}\right]}$  . By Lemma 16.7 in [4, $\left[{\gamma }_{{g}_{1}}^{\sigma }\right]\cdot \left[{\gamma }_{{g}_{2}}^{\sigma }\right]=\left[{\gamma }_{{g}_{1}}^{\sigma }{\gamma }_{{g}_{2}}^{\sigma }\right]=\left[\left({\gamma }_{{g}_{1}}{\gamma }_{{g}_{2}}{\right)}^{\sigma }\right]\in {\pi }_{\left[{g}_{1}{g}_{2}\right]}$  . So we have $\left[{\gamma }_{{g}_{1}}^{\sigma }\right]\cdot {\pi }_{\left[{g}_{2}\right]}\subset {\pi }_{\left[{g}_{1}{g}_{2}\right]}$  . To prove the equality, denote $x=\left[{\gamma }_{{g}_{1}}^{\sigma }\right]$  , and consider the map ${l}_{x}:{\pi }_{\left[{g}_{2}\right]}\to {\pi }_{\left[{g}_{1}{g}_{2}\right]}$  defined by ${l}_{x}\left(y\right)=xy$  . Since ${x}^{-1}\in {\pi }_{\left[{g}_{1}\right]}^{-1}={\pi }_{\left[{g}_{1}\right]}={\pi }_{\left[{g}_{1}^{-1}\right]}$  , ${x}^{-1}\cdot {\pi }_{\left[{g}_{1}{g}_{2}\right]}\subset {\pi }_{\left[{g}_{2}\right]}$  . So we can also define the map ${l}_{{x}^{-1}}:{\pi }_{\left[{g}_{1}{g}_{2}\right]}\to {\pi }_{\left[{g}_{2}\right]}$  by ${l}_{{x}^{-1}}\left(y\right)={x}^{-1}y$  . Since ${l}_{x}\circ {l}_{{x}^{-1}}=id$  , ${l}_{x}$  is surjective. This means $x\cdot {\pi }_{\left[{g}_{2}\right]}={\pi }_{\left[{g}_{1}{g}_{2}\right]}$  , hence ${\pi }_{\left[{g}_{1}\right]}\cdot {\pi }_{\left[{g}_{2}\right]}={\pi }_{\left[{g}_{1}{g}_{2}\right]}$  .
Lemma 2.6. Let $K$  be an open subgroup of ${G}^{\sigma }$  . Then ${\pi }_{K}:={\cup }_{k\in K}{\pi }_{\left[k\right]}$  is a subgroup of ${\pi }_{1}\left(G,e\right)$  .
• Proof. By Lemma  2.3 , each ${\pi }_{\left[k\right]}$  is closed under the inverse operation, hence so is ${\pi }_{K}$  . By Lemma  2.5 , ${\pi }_{K}$  is also closed under multiplication.
By Lemma  2.6 , ${\pi }_{\left[e\right]}={\pi }_{\left({G}_{0}^{\sigma }\right)}$  and ${\pi }_{\left({G}^{\sigma }\right)}$  are subgroups of ${\pi }_{1}\left(G,e\right)$  . By Lemma  2.5 , for each $g\in {G}^{\sigma }$  and each $x\in {\pi }_{\left[g\right]}$  , ${\pi }_{\left[g\right]}=x\cdot {\pi }_{\left[e\right]}$  . So ${\pi }_{\left[g\right]}$  is a coset space of ${\pi }_{\left[e\right]}$  in ${\pi }_{\left({G}^{\sigma }\right)}$  , that is, an element in the quotient group ${\Pi }^{\sigma }={\pi }_{\left({G}^{\sigma }\right)}/{\pi }_{\left[e\right]}$  (note that the fundamental group of a Lie group is always abelian). Define the map $f:{G}^{\sigma }/{G}_{0}^{\sigma }\to {\Pi }^{\sigma }$  by $f\left(\left[g\right]\right)={\pi }_{\left[g\right]}$  . By Lemma  2.5 , $f$  is a homomorphism.
Lemma 2.7. Let $G$  be a connected Lie group with an involution $\sigma$  .
Then ${G}^{\sigma }$  has finite many connected components. If moreover $G$  is simply connected, then ${G}^{\sigma }$  is connected.
• Proof. The subgroup $\Sigma =\left\{id,\sigma \right\}$  of the automorphism group $Aut\left(G\right)$  of $G$  has a natural action on $G$  , which we also denote by $\sigma$  . We form the semidirect product $\mathbf{G}=G{×}_{\sigma }\Sigma$  . Denote $x=\left(e,\sigma \right)\in \mathbf{G}$  . Since the identity component ${\mathbf{G}}_{0}$  of $\mathbf{G}$  is naturally isomorphic to $G$  by $i:\left(g,id\right)↦g$  , and $\sigma \left(i\left(y\right)\right)=i\left(xy{x}^{-1}\right),\forall y\in {\mathbf{G}}_{0}$  , to prove the lemma, it is sufficient to show that ${\mathbf{G}}_{0}^{x}=\left\{y\in {\mathbf{G}}_{0}|xy{x}^{-1}=y\right\}$  has finite many connected components, and is connected when ${\mathbf{G}}_{0}$  is simply connected.
Since $x=\left(e,\sigma \right)$  is an element of order two in $\mathbf{G}$  , it lies in some maximal compact subgroup $\mathbf{L}$  of $\mathbf{G}$  . By Theorem 3.1 of Chapter XV in [3, there exist some linear subspaces ${\mathfrak{m}}_{1},\cdots ,{\mathfrak{m}}_{k}$  of the Lie algebra $\mathfrak{g}$  of $\mathbf{G}$  (which is isomorphic to the Lie algebra of $G$  ) such that (1) $\mathfrak{g}=\mathfrak{l}\oplus {\mathfrak{m}}_{1}\oplus \cdots \oplus {\mathfrak{m}}_{k}$  , where $\mathfrak{l}$  is the Lie algebra of $\mathbf{L}$  ; (2) $Ad\left(l\right)\left({\mathfrak{m}}_{i}\right)={\mathfrak{m}}_{i},\forall l\in \mathbf{L},i\in \left\{1,\cdots ,k\right\}$  (in particular, $Ad\left(x\right)\left({\mathfrak{m}}_{i}\right)={\mathfrak{m}}_{i}$  ); (3) the map $\psi :{\mathbf{L}}_{0}×{\mathfrak{m}}_{1}×\cdots ×{\mathfrak{m}}_{k}\to {\mathbf{G}}_{0}$  defined by $\psi \left(l,{X}_{1},\cdots ,{X}_{k}\right)=l{e}^{{X}_{1}}\cdots {e}^{{X}_{k}}$  is a diffeomorphism, where ${\mathbf{L}}_{0}$  is the identity component of $\mathbf{L}$  .
Denote ${\mathbf{L}}_{0}^{x}=\left\{l\in {\mathbf{L}}_{0}|xl{x}^{-1}=l\right\}$  , ${\mathfrak{m}}_{i}^{x}=\left\{X\in {\mathfrak{m}}_{i}|Ad\left(x\right)\left(X\right)=X\right\},i=1,\cdots ,k$  . We claim that ${\mathbf{G}}_{0}^{x}=\psi \left({\mathbf{L}}_{0}^{x}×{\mathfrak{m}}_{1}^{x}×\cdots ×{\mathfrak{m}}_{k}^{x}\right)$  . In fact, for $y\in {\mathbf{G}}_{0}^{x}$  , write $y=l{e}^{{X}_{1}}\cdots {e}^{{X}_{k}}$  , where $l\in {\mathbf{L}}_{0},{X}_{i}\in {\mathfrak{m}}_{i}$  . Then $y=xy{x}^{-1}=\left(xl{x}^{-1}\right){e}^{Ad\left(x\right)\left({X}_{1}\right)}\cdots {e}^{Ad\left(x\right)\left({X}_{k}\right)}$  . By (3), we have $xl{x}^{-1}=l$  , $Ad\left(x\right)\left({X}_{i}\right)={X}_{i},i=1,\cdots ,k$  . That is, $l\in {\mathbf{L}}_{0}^{x}$  , ${X}_{i}\in {\mathfrak{m}}_{i}^{x}$  . Hence we have ${\mathbf{G}}_{0}^{x}\subset \psi \left({\mathbf{L}}_{0}^{x}×{\mathfrak{m}}_{1}^{x}×\cdots ×{\mathfrak{m}}_{k}^{x}\right)$  . The inverse direction is obvious.
Since ${\mathbf{L}}_{0}$  is a connected compact Lie group, ${\mathbf{L}}_{0}^{x}$  has finite many connected components, hence so is ${\mathbf{G}}_{0}^{x}=\psi \left({\mathbf{L}}_{0}^{x}×{\mathfrak{m}}_{1}^{x}×\cdots ×{\mathfrak{m}}_{k}^{x}\right)$  . If ${\mathbf{G}}_{0}$  is simply connected, by (3), ${\mathbf{L}}_{0}$  is also simply connected. By Theorem 8.2 of Chapter VII in [2, ${\mathbf{L}}_{0}^{x}$  is connected, so is ${\mathbf{G}}_{0}^{x}$  . This proves the lemma.
Lemma 2.8. For $g\in {G}^{\sigma }$  , if $\text{e}\in {\pi }_{\left[g\right]}$  , then $\left[g\right]={G}_{0}^{\sigma }$  .
• Proof. We endow the canonical smooth manifold structure and the canonical group structure on $\stackrel{~}{G}={\cup }_{g\in G}{\pi }_{1}\left(G,e,g\right)$  such that $\stackrel{~}{G}$  is the universal covering group of $G$  with covering homomorphism $\pi \left(\left[\gamma \right]\right)=\gamma \left(1\right)$  . The induced involution of $\sigma$  on $\stackrel{~}{G}$  is $\stackrel{~}{\sigma }\left(\left[\gamma \right]\right)=\left[\sigma \circ \gamma \right]$  . Let $g\in {G}^{\sigma }$  . If $\text{e}\in {\pi }_{\left[g\right]}$  , by the definition of ${\pi }_{\left[g\right]}$  , there is a ${\gamma }_{g}\in \Omega \left(G,e,g\right)$  such that $\left[{\gamma }_{g}\right]=\left[\sigma \circ {\gamma }_{g}\right]$  in ${\pi }_{1}\left(G,e,g\right)$  . That is, $\left[{\gamma }_{g}\right]\in \left(\stackrel{~}{G}{\right)}^{\stackrel{~}{\sigma }}$  .
By Lemma  2.7 , $\left(\stackrel{~}{G}{\right)}^{\stackrel{~}{\sigma }}$  is connected. So $g=\pi \left(\left[{\gamma }_{g}\right]\right)\in \pi \left(\left(\stackrel{~}{G}{\right)}^{\stackrel{~}{\sigma }}\right)={G}_{0}^{\sigma }$  , that is, $\left[g\right]={G}_{0}^{\sigma }$  .
Proposition 2.9. Each non-identity element of ${\Pi }^{\sigma }$  has order $2$  , and the homomorphism $f:{G}^{\sigma }/{G}_{0}^{\sigma }\to {\Pi }^{\sigma }$  is an isomorphism.
• Proof. The first assertion follows from Lemma  2.3 , the surjectivity of $f$  follows from the definition of ${\Pi }^{\sigma }$  , and the injectivity of $f$  follows from Lemma  2.8 .
In particular, we have
Lemma 2.10. For ${g}_{1},{g}_{2}\in {G}^{\sigma }$  , if $\left[{g}_{1}\right]\ne \left[{g}_{2}\right]$  , then ${\pi }_{\left[{g}_{1}\right]}\cap {\pi }_{\left[{g}_{2}\right]}=\varnothing$  .
Now we are prepared to prove Theorem  2.1 and Theorem  2.2 .
Proof of Theorem  2.2 . By Lemma  2.7 , ${G}^{\sigma }/{G}_{0}^{\sigma }$  is a finite group. By Proposition  2.9 , ${G}^{\sigma }/{G}_{0}^{\sigma }$  is abelian, and each non-identity element has order $2$  . By the structure theorem of finitely generated abelian groups, ${G}^{\sigma }/{G}_{0}^{\sigma }$  is isomorphic to $\left({\mathbb{Z}}_{2}{\right)}^{r}$  for some non-negative integer $r$  . If $G$  is simply connected, by Lemma  2.7 , ${G}^{\sigma }/{G}_{0}^{\sigma }$  is trivial. Proof of Theorem  2.1 . We define an equivalence relation on the set ${\cup }_{g\in G}\Omega \left(G,e,g\right)$  as follows. For ${\gamma }_{1},{\gamma }_{2}\in {\cup }_{g\in G}\Omega \left(G,e,g\right)$  , ${\gamma }_{1}\sim {\gamma }_{2}$  if and only if ${\gamma }_{1}\left(1\right)={\gamma }_{2}\left(1\right)$  and $\left[{\gamma }_{12}\right]\in {\pi }_{K}$  , where ${\gamma }_{12}\in \Omega \left(G,e,e\right)$  is defined by ${\gamma }_{12}\left(t\right)=\left\{\begin{array}{cc}{\gamma }_{1}\left(2t\right)& t\in \left[0,\frac{1}{2}\right];\\ {\gamma }_{2}\left(2-2t\right)& t\in \left[\frac{1}{2},1\right].\end{array}$  Denote ${G}^{\prime }=\left({\cup }_{g\in G}\Omega \left(G,e,g\right)\right)/\sim$  , and denote the equivalence class of a $\gamma \in {\cup }_{g\in G}\Omega \left(G,e,g\right)$  by $〈\gamma 〉$  . Endow the smooth manifold structure on ${G}^{\prime }$  in the canonical way, and define the group operation on $G$  as follows. For $〈{\gamma }_{1}〉,〈{\gamma }_{2}〉\in G$  , let $\alpha \left(t\right)={\gamma }_{1}\left(t\right){\gamma }_{2}\left(t\right)$  and $\beta \left(t\right)={\gamma }_{1}\left(t{\right)}^{-1}$  , and define $〈{\gamma }_{1}〉〈{\gamma }_{2}〉=〈\alpha 〉,〈{\gamma }_{1}{〉}^{-1}=〈\beta 〉$  .
It is easy to check that these are well defined, and ${G}^{\prime }$  thus assembled becomes a Lie group. Let $\pi :{G}^{\prime }\to G$  be the map $\pi \left(〈\gamma 〉\right)=\gamma \left(1\right)$  , then $\pi$  is a covering homomorphism. Since ${\sigma }_{*}\left({\pi }_{K}\right)={\pi }_{K}$  , the map ${\sigma }^{\prime }:{G}^{\prime }\to {G}^{\prime },{\sigma }^{\prime }\left(〈\gamma 〉\right)=〈\sigma \circ \gamma 〉$  is a well defined involution of ${G}^{\prime }$  . It is obvious that $\pi \circ {\sigma }^{\prime }=\sigma \circ \pi$  , that is, ${\sigma }^{\prime }$  is just the induced involution of $\sigma$  on ${G}^{\prime }$  . Now we have
 $\begin{array}{cc}& 〈\gamma 〉\in \left({G}^{\prime }{\right)}^{{\sigma }^{\prime }}\end{array}$
 $\begin{array}{cc}⇔& 〈\sigma \circ \gamma 〉=〈\gamma 〉\end{array}$
 $\begin{array}{cc}⇔& \sigma \circ \gamma \sim \gamma \end{array}$
 $\begin{array}{cc}⇔& \gamma \left(1\right)\in {G}^{\sigma },\left[{\gamma }_{\gamma \left(1\right)}^{\sigma }\right]\in {\pi }_{K}\end{array}$
 $\begin{array}{cc}⇔& \gamma \left(1\right)\in K\left(\text{by LemmaL:disjoint2.10}\right)\end{array}$
 $\begin{array}{cc}⇔& 〈\gamma 〉\in {\pi }^{-1}\left(K\right).\end{array}$
 $\begin{array}{}\end{array}$
That is, $\left({G}^{\prime }{\right)}^{{\sigma }^{\prime }}={\pi }^{-1}\left(K\right)$  . This complete the proof of Theorem  2.1 .

3 Covering groups and the realization of symmetric spaces

Let $G$  be a connected Lie group with an involution $\sigma$  , and let $K$  be an open subgroup of ${G}^{\sigma }$  . Let $\mathfrak{g}=\mathfrak{k}\oplus \mathfrak{p}$  be the canonical decomposition of the Lie algebra $\mathfrak{g}$  of $G$  . Suppose $A{d}_{G}\left(K\right){|}_{\mathfrak{p}}$  is compact, that is, $G/K$  has a structure of Riemannian symmetric space. Suppose $\pi :{G}^{\prime }\to G$  is a covering homomorphism of Lie groups. Then the twisted conjugate action $\tau$  of $G$  on $G$  , which is defined by ${\tau }_{g}\left(h\right)=gh\sigma \left(g{\right)}^{-1}$  , can be lifted to a well defined action ${\tau }^{\prime }$  of $G$  on ${G}^{\prime }$  , that is, ${\tau }_{g}\left({h}^{\prime }\right)={g}^{\prime }{h}^{\prime }{\sigma }^{\prime }\left({g}^{\prime }{\right)}^{-1},g\in G,{h}^{\prime }\in {G}^{\prime }$  , where ${g}^{\prime }\in {\pi }^{-1}\left(g\right)$  , ${\sigma }^{\prime }$  is the induced involution of $\sigma$  on ${G}^{\prime }$  .
Theorem 3.1. Under the above assumptions, there is a covering group ${G}^{\prime }$  of $G$  with covering homomorphism $\pi$  such that ${P}^{\prime }={exp}_{G}^{\prime }\left(\mathfrak{p}\right)$  is a closed submanifold of ${G}^{\prime }$  , and such that the map $\phi :G/K\to {P}^{\prime }$  defined by $\phi \left(gK\right)={g}^{\prime }{\sigma }^{\prime }\left({g}^{\prime }{\right)}^{-1}$  is a diffeomorphism, where ${g}^{\prime }\in {\pi }^{-1}\left(g\right)$  , ${\sigma }^{\prime }$  is the induced involution of $\sigma$  on ${G}^{\prime }$  . With respect to the actions of $G$  by left multiplication on $G/K$  and by the action ${\tau }^{\prime }$  on ${P}^{\prime }$  , $\phi$  is equivariant.
• Proof. By Theorem  2.1 , there is a covering group ${G}^{\prime }$  of $G$  with covering homomorphism $\pi$  such that ${\pi }^{-1}\left(K\right)=\left({G}^{\prime }{\right)}^{{\sigma }^{\prime }}$  . So $\pi$  induces a $G$  -equivariant diffeomorphism $\overline{\pi }:{G}^{\prime }/\left({G}^{\prime }{\right)}^{{\sigma }^{\prime }}\to G/K$  with respect to left multiplications (note that the left multiplication of $G$  on ${G}^{\prime }/\left({G}^{\prime }{\right)}^{{\sigma }^{\prime }}$  is well defined). Since $A{d}_{{G}^{\prime }}\left(\left({G}^{\prime }{\right)}^{{\sigma }^{\prime }}\right){|}_{\mathfrak{p}}=A{d}_{G}\left(K\right){|}_{\mathfrak{p}}$  is compact, by Corollary 2.6 in [1, ${P}^{\prime }={exp}_{{G}^{\prime }}\left(\mathfrak{p}\right)$  is a closed submanifold of ${G}^{\prime }$  , and the map ${\phi }^{\prime }:{G}^{\prime }/\left({G}^{\prime }{\right)}^{{\sigma }^{\prime }}\to {P}^{\prime },{\phi }^{\prime }\left({g}^{\prime }\left({G}^{\prime }{\right)}^{{\sigma }^{\prime }}\right)={g}^{\prime }{\sigma }^{\prime }\left({g}^{\prime }{\right)}^{-1}$  is a ${G}^{\prime }$  -equivariant diffeomorphism with respect to left multiplication and twisted conjugate action of ${G}^{\prime }$  . Let $\phi ={\phi }^{\prime }\circ \left(\overline{\pi }{\right)}^{-1}$  . Then $\phi \left(gK\right)={\phi }^{\prime }\left({g}^{\prime }\left({G}^{\prime }{\right)}^{{\sigma }^{\prime }}\right)={g}^{\prime }{\sigma }^{\prime }\left({g}^{\prime }{\right)}^{-1}$  . It is obviously a $G$  -equivariant diffeomorphism with respect to left multiplication on $G/K$  and the action ${\tau }^{\prime }$  on ${P}^{\prime }$  .
References

1. An, J., Wang, Z., On the realization of Riemannian symmetric spaces in Lie groups, to appear in Topology Appl., math.GT/0502028.
2. Helgason, S., Differential geometry, Lie groups, and symmetric spaces, Acadymic Press, New York, 1978.
3. Hochschild, G., The structrue of Lie groups, Holden-Day, San Francisco, 1965.
4. Steenrod, N., The topology of fibre bundles, Princeton University Press, Princeton, New Jersey, 1951.

School of mathematical science, Peking University, Beijing, 100871, P. R. China E-mail address : anjinpeng@pku.edu.cn School of mathematical science, Peking University, Beijing, 100871, P. R. China E-mail address : zdwang@pku.edu.cn

1 This work is supported by the 973 Project Foundation of China ( $♯$  TG1999075102). AMS 2000 Mathematics Subject Classifications: Primary 22E15; secondary 57R40, 53C35.