## The Weyl Integration Model for KAK decomposition of Reductive Lie Group

### HU Zhi-guang 1 YAN Kui-hua 1 , 2 (1. School of Mathematical Sciences, Peking University, Beijing, 100871; 2. School of Mathematics and Physics, Zhejiang Normal University, Zhejiang Jinhua, 321004) (Email: zhiguang_hu@eyou.com; yankh@zjnu.cn)

Abstract
The Weyl integration model presented by An and Wang can be effectively used to reduce the integration over $G$  -space. In this paper, we construct an especial Weyl integration model for KAK decomposition of Reductive Lie Group and obtain an integration formula which implies that the integration of ${L}^{1}$  -integrable function over reductive Lie group $G$  can be carried out by first integrating over each conjugacy class and then integrating over the set of conjugacy classes.
Key Words: Weyl Integration Model; Reductive Lie Group; KAK Decomposition; Restricted roots Mathematics Subject Classification 2000: 22E15; 58C35 CLC number: O152.5; O186.1

1 Introduction

For a compact connected Lie group $G$  , the classical Weyl integration formula (e.g.
ref. Knapp[1) indicates that an integration of any continuous function over $G$  can be reduced to one over its maximal torus. That is to say the integration can be carried out by first integrating over each conjugacy class and then integrating over the set of conjugacy classes.
Then for a general Lie group, how to reduce the integration of functions over it ?
In Helgason[2, the author presents some integral formulas related to the Cartan, Iwasawa and Bruhat decompositions for semisimple Lie groups. In this paper, we will use the Weyl integration model, which is first introduced by An and Wang in [3and can be used to generalize the reducing integration idea to integrations over G-spaces, to obtain an integral formula for KAK decomposition of reductive Lie groups. It is worthy of indicating that this reducible idea of integrations can be effectively used to calculate the eigenvalue distribution of random matrices in random matrix ensemble theory. But here we do not plan to study those applications in random matrix theory.
Let $G$  be a Lie group, $X$  a $G$  -space, $Y$  a imbedding submanifold of $X$  and $\sigma$  the $G$  -action on $X$  . Suppose that $dx,dy$  and $d\mu$  are proper invariant measures over $X,Y$  and $G/K$  respectively. The Weyl integration model (see section 2 for its details) is a system $\left(G,\sigma ,\left(X,dx\right),\left(Y,dy\right),\left(G/K,d\mu \right)\right)$  satisfying some proper conditions (see  7 - 9 ). Indeed, for a Weyl integration model, the following formula holds for all $f\in {L}^{1}\left(X,dx\right)$  ,
 $\begin{array}{c}{\int }_{X}f\left(x\right)dx=\frac{1}{d}{\int }_{Y}\left({\int }_{G/K}f\left(\sigma \left(g,y\right)\right)d\mu \left(\left[g\right]\right)\right)J\left(y\right)dy,\end{array}$ (1)
where $J\in {C}^{\infty }\left(Y\right)$  and $d$  is the multiplicity of the model. That is to say that the integration over a $G$  -space $X$  can be converted to the integration by first integrating over each orbit and then integrating over the orbits space $Y$  . Moreover, under some orthogonal conditions, $J\left(y\right)$  can be effectively calculated, i.e. there is a constant $C$  such that
 $\begin{array}{c}J\left(y\right)=C|det{\Psi }_{y}|,\text{for all}y\in {Y}^{\prime },\end{array}$ (2)
where ${\Psi }_{y}$  is a mapping induced by G-action $\sigma$  (see  11 ). In this paper, we will construct a Weyl integration model $\left(K×K,\sigma ,\left(G,dg\right),\left(A,da\right),\left(\left(K×K\right)/M,d\mu \right)\right)$  . The meanings of its members are as follow. $G$  is a reductive Lie group with Lie algebra $\mathfrak{g}$  . Let $\theta$  be the Cartan involution on $\mathfrak{g}$  . $\mathfrak{g}=\mathfrak{k}\oplus \mathfrak{p}$  is the Cartan decomposition of $\mathfrak{g}$  , where $\mathfrak{k}$  and $\mathfrak{p}$  are the $+1$  and $-1$  eigenspaces of $\theta$  respectively. Let $K$  be the associated maximal compact subgroup of $G$  with Lie algebra $\mathfrak{k}$  .
Let $\mathfrak{a}$  be a maximal abelian subspace of $\mathfrak{p}$  . Set $A=exp\left(\mathfrak{a}\right)$  , then $A$  is a closed subgroup of $G$  . It is known that $G=KAK$  in the sense that every element in $G$  has a decomposition as ${k}_{{}_{1}}a{k}_{{}_{2}}$  with ${k}_{{}_{1}},{k}_{{}_{2}}\in K$  and $a\in A$  .
To consider the following group $K×K$  -action
 $\begin{array}{c}\sigma :\left(K×K\right)×A\to G,\left(\left({k}_{{}_{1}},{k}_{{}_{2}}\right),a\right)↦{k}_{{}_{1}}a{k}_{{}_{2}}^{-1}.\end{array}$ (3)
Let
 $\begin{array}{c}M=\left\{\left({k}_{{}_{1}},{k}_{{}_{2}}\right)\in K×K|{k}_{{}_{1}}a{k}_{{}_{2}}^{-1}=a,\forall a\in A\right\}.\end{array}$ (4)
Then the $K×K$  -action $\sigma$  can be naturally reduced to a map $\phi$  , i.e.
 $\begin{array}{c}\phi :\left(K×K\right)/M×A\to G,\left(\left[\left({k}_{{}_{1}},{k}_{{}_{2}}\right)\right],a\right)↦{k}_{{}_{1}}a{k}_{{}_{2}}^{-1}.\end{array}$ (5)
Let $dg,da$  be the Haar measures on $G$  and $A$  respectively. There has a $K×K$  -invariant measure $d\mu$  on $\left(K×K\right)/M$  since $K×K$  is compact. Then we can prove that $\left(K×K,\sigma ,\left(G,dg\right),\left(A,da\right),\left(\left(K×K\right)/M,d\mu \right)\right)$  is a Weyl integration model with finite multiplicities.
Thus by  1 , we obtain an integral formula (see theorem  2.6 for details) which reduces the integration over reductive Lie group $G$  to one first over $\left(K×K\right)/M$  and then over its subgroup $A$  . Moreover, by  2 , we can calculate that the according $J\left(y\right)$  has the following formula
 $\begin{array}{c}J\left(a\right)=C{\prod }_{\lambda \in {\Sigma }^{+}}|sinh\left(\lambda \left(H\right)\right){|}^{{\beta }_{{}_{\lambda }}},\end{array}$ (6)
where $a={e}^{H},H\in \mathfrak{a}$  , ${\beta }_{{}_{\lambda }}$  is the dimension of the restricted root space ${\mathfrak{g}}_{{}_{\lambda }}$  with respect to the restricted root $\lambda$  in the positive restricted root system ${\Sigma }^{+}$  and $C$  is a proper constant.
By comparison, it is formally analogous to the integration formula for Cartan decomposition of noncompact semisimple Lie groups (see [2).
The structure of the paper is as follow. In section 2, we briefly introduce the Weyl integration model and the Restricted root system and KAK decomposition for a reductive Lie group. The main results of the paper are presented at the end of this section. In section 3, we prove the main results.

2 Preliminaries and Main Results

2.1 Weyl integration model

In this subsection, we briefly introduce the Weyl integration model. One may refer to [3for details.
Let $G$  be a Lie group which acts on a $n$  -dimensional smooth manifold $X$  . The action is denoted by $\sigma :G×X\to X$  . Let $dx$  be a $G$  -invariant admissible measure on $X$  . $Y$  is an imbedding submanifold of $X$  . Suppose that there is an admissible measure $dy$  on $Y$  , and ${X}_{0}\subset X$  , ${Y}_{0}\subset Y$  are closed zero measure subsets of $X$  and $Y$  respectively. Set ${X}^{\prime }=X\{X}_{0}$  , ${Y}^{\prime }=Y\{Y}_{0}$  and $K=\left\{g\in G|\sigma \left(g,y\right)=y,\forall y\in Y\right\}$  . Let ${K}_{x}=\left\{g\in G|\sigma \left(g,x\right)=x\right\}$  be the isotropic subgroup associated with $x\in X$  and ${O}_{y}=\left\{\sigma \left(g,y\right)|g\in G\right\}$  the orbit of $y\in Y$  . Then $K\subset {K}_{y},\forall y\in Y$  . In the following text we suppose that
 $\begin{array}{c}{X}^{\prime }={\bigcup }_{y\in {Y}^{\prime }}{O}_{y}.\end{array}$ (7)
 $\begin{array}{c}{T}_{y}X={T}_{y}{O}_{y}\oplus {T}_{y}Y,\forall y\in {Y}^{\prime },\end{array}$ (8)
which is called the transversal condition.
 $\begin{array}{c}dim{K}_{y}=dimK,\forall y\in {Y}^{\prime }.\end{array}$ (9)
The G-action $\sigma :G×X\to X$  can be reduced to a map $\phi :G/K×Y\to X$  by $\phi \left(\left[g\right],y\right)=\sigma \left(g,y\right)$  and furthermore to a map which is still denoted by $\phi$  , i.e. $\phi :G/K×{Y}^{\prime }\to {X}^{\prime }$  by restriction. By the above assumption, $\phi$  is surjective. Suppose that there is a $G$  -invariant admissible measure $d\mu$  on $G/K$  .
Proposition 2.1. Suppose the conditions  7 ,  8 and  9 hold, then $\phi :G/K×{Y}^{\prime }\to {X}^{\prime }$  is a local diffeomorphism.
Let $G,X,Y,K,\sigma ,\phi ,dx,dy,d\mu$  be the above objects. A Weyl integration model is a system $\left(G,\sigma ,\left(X,dx\right),\left(Y,dy\right),\left(G/K,d\mu \right)\right)$  , in which we can choose ${X}_{0}$  such that the map $\phi :G/K×{Y}^{\prime }\to {X}^{\prime }$  is a finite-sheeted covering map. The number of sheets of the covering map is called the multiplicity of the model. About a Weyl integration model, we have the following basic theorem.
Theorem 2.2. If $\left(G,\sigma ,\left(X,dx\right),\left(Y,dy\right),\left(G/K,d\mu \right)\right)$  is a Weyl integration model with multiplicity $d$  , then the formula  1 holds for all $f\in {C}^{\infty }\left(X\right)$  with $f\ge 0$  or $f\in {L}^{1}\left(X,dx\right)$  .
Let $\mathfrak{s}$  be a linear subspace of the Lie algebra $\mathfrak{g}$  of $G$  , such that $\mathfrak{g}=\mathfrak{k}\oplus \mathfrak{s}$  , where $\mathfrak{k}$  is the Lie algebra of $K$  . Then $\mathfrak{s}$  can be identified with the tangent space ${T}_{\left[e\right]}\left(G/K\right)$  of $G/K$  at point $\left[e\right]$  in a natural way:
${T}_{\left[e\right]}\left(G/K\right)\sim =\mathfrak{s}.$  By the transversal condition  8 , the tangent map $d\phi :{T}_{\left[e\right]}\left(G/K\right)\oplus {T}_{y}Y\to {T}_{y}X$  of $\phi$  at point $\left(\left[e\right],y\right)$  ( $y\in {Y}^{\prime }$  ) can be regarded as a linear transformation $d\phi :\mathfrak{s}\oplus {T}_{y}Y\to {T}_{y}{O}_{y}\oplus {T}_{y}Y.$  It can be proved that (see [3)
 $\begin{array}{c}d\phi =\left(\begin{array}{cc}{\Psi }_{y}& 0\\ 0& id\end{array}\right),\end{array}$ (10)
where ${\Psi }_{y}:\mathfrak{s}\to {T}_{y}{O}_{y}$  is given by
 $\begin{array}{c}{\Psi }_{y}\left(\xi \right)=\frac{d}{dt}{|}_{t=0}\sigma \left(expt\xi ,y\right),\forall \xi \in \mathfrak{s}.\end{array}$ (11)
Suppose that there is a Riemannian structure on $X$  such that the following orthogonal condition holds
 $\begin{array}{c}{T}_{y}Y\perp {T}_{y}{O}_{y},\text{for all}y\in {Y}^{\prime }.\end{array}$ (12)
Let $dx$  and $dy$  are the Riemannian measures on $X$  and $Y$  respectively. Then there is a constant $C$  such that
 $\begin{array}{c}J\left(y\right)=C|det{\Psi }_{y}|,\text{for all}y\in {Y}^{\prime }.\end{array}$ (13)

2.2 Restricted root system and KAK decomposition for reductive Lie group

In the sense of Knapp ([1, Sec.7.2), a reductive Lie group is a 4-tuple $\left(G,K,\theta ,B\right)$  consisting of a Lie group $G$  , a compact subgroup $K$  of $G$  , a Lie algebra involution $\theta$  of the Lie algebra $\mathfrak{g}$  of $G$  and a nondegenerate $Ad\left(G\right)$  -invariant $\theta$  -invariant bilinear form $B$  on $\mathfrak{g}$  such that (i) $\mathfrak{g}$  is a reductive Lie algebra, (ii) the decomposition of $\mathfrak{g}$  into $+1$  and $-1$  eigenspaces under $\theta$  is $\mathfrak{g}=\mathfrak{k}\oplus \mathfrak{p}$  , where $\mathfrak{k}$  is the Lie algebra of K, (iii) $\mathfrak{k}$  and $\mathfrak{p}$  are orthogonal under $B$  and $B$  is positive definite on $\mathfrak{p}$  and negative definite on $\mathfrak{k}$  , (iv) multiplication, as a map from $K×exp\left(\mathfrak{p}\right)$  into $G$  , is a diffeomorphism onto, and (v) every automorphism $Ad\left(g\right)$  of ${\mathfrak{g}}^{\mathbb{C}}$  is inner for $g\in G$  , i.e., is given by some $x\in Int\left({\mathfrak{g}}^{\mathbb{C}}\right)$  . $K$  is called the associated maximal compact subgroup, $\theta$  the Cartan involution and $B$  the invariant bilinear form. The decomposition of $\mathfrak{g}$  (according $G$  ) in property (iii) (according (iv)) is called (global) Cartan decomposition.
Now let $G$  is a reductive Lie group with Lie algebra $\mathfrak{g}$  . Let $\mathfrak{a}$  be a maximal abelian subspaces of $\mathfrak{p}$  , then $\mathfrak{p}={\bigcup }_{k\in K}Ad\left(k\right)\mathfrak{a}$  . Set ${\mathfrak{g}}_{{}_{\lambda }}=\left\{X\in \mathfrak{g}|\left(adH\right)X=\lambda \left(H\right)X,\text{for all}H\in \mathfrak{a}\right\}$  .
A nonzero $\lambda \in {\mathfrak{a}}^{*}$  is called a restricted root of $\mathfrak{g}$  if ${\mathfrak{g}}_{{}_{\lambda }}$  is nonzero. Accordingly ${\mathfrak{g}}_{{}_{\lambda }}$  is called a restricted root space. The set of restricted roots is denoted by $\Sigma$  and Let ${\Sigma }^{+}$  be the set of positive restricted roots.
Reflections in the restricted roots generate the Weyl group $W\left(\Sigma \right)$  of $\Sigma$  . Denoted by ${N}_{K}\left(\mathfrak{a}\right)$  and ${Z}_{K}\left(\mathfrak{a}\right)$  the normalizer and centralizer of $\mathfrak{a}$  in $K$  respectively, then the Weyl group $W={N}_{K}\left(\mathfrak{a}\right)/{Z}_{K}\left(\mathfrak{a}\right)$  and the Lie algebras of ${N}_{K}\left(\mathfrak{a}\right)$  and ${Z}_{K}\left(\mathfrak{a}\right)$  are $\mathfrak{m}={Z}_{\mathfrak{k}}\left(\mathfrak{a}\right)$  .
The restricted root space ${\mathfrak{g}}_{{}_{\lambda }}$  satisfies the following basic properties
Proposition 2.3. (see [1) (i) $\mathfrak{g}={\mathfrak{g}}_{{}_{0}}\oplus {\oplus }_{\lambda \in \Sigma }{\mathfrak{g}}_{{}_{\lambda }}$  .
(ii) $\left[{\mathfrak{g}}_{{}_{\lambda }},{\mathfrak{g}}_{{}_{\gamma }}\right]\subset {\mathfrak{g}}_{{}_{\lambda +\gamma }}$  and if $\lambda \ne \gamma$  , then ${\mathfrak{g}}_{{}_{\lambda }}\perp {\mathfrak{g}}_{{}_{\gamma }}$  in the sense of $B$  .
(iii) $\theta {\mathfrak{g}}_{{}_{\lambda }}={\mathfrak{g}}_{{}_{-\lambda }}$  , so if $\lambda \in \Sigma$  , then $-\lambda \in \Sigma$  .
(iv) ${\mathfrak{g}}_{{}_{0}}=\mathfrak{a}\oplus \mathfrak{m}$  orthogonally.
Let $A=exp\mathfrak{a}$  . For any reductive Lie group $G$  , it has the following decomposition
Theorem 2.4. (KAK decomposition [1) Every element in G has a decomposition as ${k}_{{}_{1}}a{k}_{{}_{2}}$  with ${k}_{{}_{1}},{k}_{{}_{2}}\in K$  and $a\in A$  . In this decomposition, $a$  is uniquely determined up to conjugation by a member of $W$  . If $a$  is fixed as $expH$  with $H\in \mathfrak{a}$  and $\lambda \left(H\right)\ne 0$  for all $\lambda \in \Sigma$  , then ${k}_{{}_{1}}$  is unique up to right multiplication by a member of ${Z}_{K}\left(\mathfrak{a}\right)$  .

2.3 Main result

Now let us present the main result of the paper. Let $\left(G,K,\theta ,B\right)$  be any reductive Lie group. Let $dg$  and $da$  be the left-invariant measures on $G$  and $A$  respectively corresponding to the Riemannian structure induced by $B$  and $d\mu$  a $K×K$  -invariant measure on $\left(K×K\right)/M$  , where the set $M$  are defined by  4 .
Theorem 2.5. $\left(K×K,\phi ,\left(G,dg\right),\left(A,da\right),\left(\left(K×K\right)/M,d\mu \right)\right)$  is a Weyl integration model with $d$  -sheeted multiplicities, where $d=|W|$  and the map $\phi$  is defined by  5 .
Theorem 2.6. For any $f\in {L}^{1}\left(G,dg\right)$  ,
 $\begin{array}{c}{\int }_{G}f\left(g\right)dg={\int }_{A}\left({\int }_{\left(K×K\right)/M}f\left({k}_{{}_{1}}a{k}_{{}_{2}}^{-1}\right)d\mu \right)J\left(a\right)da,\end{array}$ (14)
where $J\left(a\right)$  has the formula  6 .

3 Proof of Main Results

In this section, we will prove the main results by constructing a Weyl integration model and using the theorem  2.2 .
Now Let $\left(G,K,\theta ,B\right)$  be any reductive Lie group. $\mathfrak{a},A,\mathfrak{m},\Sigma ,{\Sigma }^{+}$  and $W$  are defined in the above section. We come to consider the G-action $\sigma$  defined by  3 and its reduced map $\phi$  defined by  5 . Let ${A}^{\prime }$  be the set of regular elements in $A$  and ${G}^{\prime }=K{A}^{\prime }K$  . It naturally has the following map by restriction of $\phi$  , which is still denoted by $\phi$  .
 $\begin{array}{c}\phi :\left(K×K\right)/M×{A}^{\prime }\to {G}^{\prime }.\end{array}$ (15)
Lemma 3.1. (i) $M$  is isomorphic to ${Z}_{K}\left(\mathfrak{a}\right)$  .
(ii) $\phi$  is a surjective $d$  -sheeted map, where $d=|W|$  .
Proof. (i) Note that the unit element $e\in A$  . If $\left({k}_{{}_{1}},{k}_{{}_{2}}\right)\in M$  , then ${k}_{{}_{1}}e{k}_{{}_{2}}^{-1}=e$  , i.e. ${k}_{{}_{1}}={k}_{{}_{2}}$  .
Thus $M=\left\{\left(k,k\right)\in K×K|ka=ak,\forall a\in A\right\}$  is isomorphic to ${Z}_{K}\left(\mathfrak{a}\right)$  .
(ii) By the conclusion $1$  , it is the direct corollary of theorem  2.4 .
Now for any $x\in G$  , set ${M}_{x}=\left\{\left({k}_{{}_{1}},{k}_{{}_{2}}\right)\in K×K|{k}_{{}_{1}}x{k}_{{}_{2}}^{-1}=x\right\}$  be the isotropic subgroup associated with $x\in G$  and ${O}_{a}=\left\{{k}_{{}_{1}}a{k}_{{}_{2}}^{-1}|\left({k}_{{}_{1}},{k}_{{}_{2}}\right)\in K×K\right\}$  the orbit of $a\in {A}^{\prime }$  . Then $M\subset {M}_{a}$  , for all $a\in {A}^{\prime }$  .
By the definition of reductive group, the invariant bilinear $B$  determines an inner product on $\mathfrak{p}$  , and $-B$  determines an inner product on $\mathfrak{k}$  . We write $\mathfrak{b}={\mathfrak{a}}^{\perp }$  in $\mathfrak{p}$  , and $\mathfrak{l}={\mathfrak{m}}^{\perp }$  in $\mathfrak{k}$  .
By proposition  2.3 , it is obvious that
 $\begin{array}{c}\mathfrak{g}=\mathfrak{m}\oplus \mathfrak{l}\oplus \mathfrak{a}\oplus \mathfrak{b},\mathfrak{b}\oplus \mathfrak{l}={\oplus }_{\lambda \in \Sigma }{\mathfrak{g}}_{{}_{\lambda }},\mathfrak{l}=\mathfrak{k}\bigcap {\oplus }_{\lambda \in \Sigma }{\mathfrak{g}}_{{}_{\lambda }},\mathfrak{b}=\mathfrak{p}\bigcap {\oplus }_{\lambda \in \Sigma }{\mathfrak{g}}_{{}_{\lambda }}.\end{array}$ (16)
For all $\lambda \in {\Sigma }^{+}$  , we choose a normal orthogonal basis $\left\{{\xi }_{{}_{\lambda ,1}},\cdots ,{\xi }_{{}_{\lambda ,{\beta }_{{}_{\lambda }}}}\right\}$  in ${\mathfrak{g}}_{{}_{\lambda }}$  , where ${\beta }_{{}_{\lambda }}=dim{\mathfrak{g}}_{{}_{\lambda }}$  which may be larger than one. Then for all ${\xi }_{{}_{\lambda ,j}}$  , we have
 $\begin{array}{c}\begin{array}{}\text{(17)}& \theta \left({\xi }_{{}_{\lambda ,j}}+\theta {\xi }_{{}_{\lambda ,j}}\right)={\xi }_{{}_{\lambda ,j}}+\theta {\xi }_{{}_{\lambda ,j}}\in \mathfrak{l},\theta \left({\xi }_{{}_{\lambda ,j}}-\theta {\xi }_{{}_{\lambda ,j}}\right)=-\left({\xi }_{{}_{\lambda ,j}}-\theta {\xi }_{{}_{\lambda ,j}}\right)\in \mathfrak{b}.\end{array}\end{array}$ (18)
Indeed, $\left\{\left({\xi }_{{}_{\lambda ,j}}+\theta {\xi }_{{}_{\lambda ,j}}\right)|\lambda \in {\Sigma }^{+},j=1,\cdots ,{\beta }_{{}_{\lambda }}\right\}$  composes of a basis of $\mathfrak{l}$  and $\left\{\left({\xi }_{{}_{\lambda ,j}}-\theta {\xi }_{{}_{\lambda ,j}}\right)|\lambda \in {\Sigma }^{+},j=1,\cdots ,{\beta }_{{}_{\lambda }}\right\}$  a basis of $\mathfrak{b}$  . Then we get
 $\begin{array}{c}dim\mathfrak{l}=dim\mathfrak{b}={\sum }_{\lambda \in {\Sigma }^{+}}{\beta }_{{}_{\lambda }}.\end{array}$ (19)
Therefore,
 $\begin{array}{cc}dim\left(\left(K×K\right)/M×A\right)& =dim\left(\left(\mathfrak{k}\oplus \mathfrak{k}\right)/\mathfrak{m}\oplus \mathfrak{a}\right)=dim\mathfrak{k}+dim\mathfrak{l}+dim\mathfrak{a}\end{array}$
 $\begin{array}{cc}& =dim\mathfrak{k}+dim\mathfrak{b}+dim\mathfrak{a}=dim\mathfrak{k}+dim\mathfrak{p}=dimG.\end{array}$
 $\begin{array}{}\end{array}$
So $\phi$  is a map between the same dimension manifolds.
Lemma 3.2. (i) For $a\in {A}^{\prime }$  , denoted by ${L}_{a}$  the left translation, then
 $\begin{array}{c}{T}_{a}{O}_{a}=\left\{d{L}_{a}\left(Ad\left({a}^{-1}\right){\zeta }_{{}_{1}}-{\zeta }_{{}_{2}}\right)|\left({\zeta }_{{}_{1}},{\zeta }_{{}_{2}}\right)\in \left(\mathfrak{k},\mathfrak{k}\right)\right\}.\end{array}$ (20)
(ii) The following set is composed of a basis of ${T}_{a}{O}_{a}$  ,
 $\begin{array}{c}\begin{array}{cc}F\equiv \left\{d{L}_{a}\left({\eta }_{{}_{i}}\right),d{L}_{a}\left({\xi }_{{}_{\lambda ,j}}^{+}\right),d{L}_{a}\left({\xi }_{{}_{\lambda ,j}}^{-}\right)|i=1,2,\cdots ,m,& \lambda \in {\Sigma }^{+},j=1,2,\cdots ,{\beta }_{{}_{\lambda }}\right\}.\end{array}\end{array}$ (21)
Proof. (i) Note that ${T}_{a}{O}_{a}$  is exactly the set composed of those tangent vectors of the smooth curves $exp\left(t{\zeta }_{{}_{1}}\right)\cdot a\cdot exp\left(-t{\zeta }_{{}_{2}}\right)$  at $t=0$  by the definition of orbit ${O}_{a}$  , where $\left({\zeta }_{{}_{1}},{\zeta }_{{}_{2}}\right)\in \left(\mathfrak{k},\mathfrak{k}\right)$  and $t<|\varepsilon |$  . But
 $\begin{array}{cc}\frac{d}{dt}{|}_{t=0}exp\left(t{\zeta }_{{}_{1}}\right)\cdot a\cdot exp\left(-t{\zeta }_{{}_{2}}\right)& =\frac{d}{dt}{|}_{t=0}a\cdot exp\left(t\left(Ad\left({a}^{-1}\right){\zeta }_{{}_{1}}-{\zeta }_{{}_{2}}\right)+o\left({t}^{2}\right)\right)\end{array}$ (22)
 $\begin{array}{cc}& =d{L}_{a}\left(Ad\left({a}^{-1}\right){\zeta }_{{}_{1}}-{\zeta }_{{}_{2}}\right).\end{array}$ (23)
 $\begin{array}{}\end{array}$
Hence the conclusion is obtained.
(ii) Now suppose that $\left\{{\eta }_{{}_{1}},{\eta }_{{}_{2}},\cdots ,{\eta }_{{}_{m}}\right\}$  is a basis of $\mathfrak{m}$  and denote ${\xi }_{{}_{\lambda ,j}}^{±}\equiv {\xi }_{{}_{\lambda ,j}}±\theta {\xi }_{{}_{\lambda ,j}}$  .
By $\mathfrak{k}=\mathfrak{m}\oplus \mathfrak{l}$  , the following set is composed of a basis of $\left(\mathfrak{k},\mathfrak{k}\right)$  ,
 $\begin{array}{c}\begin{array}{c}\left\{\left({\eta }_{{}_{i}},0\right),\left({\xi }_{{}_{\lambda ,j}}^{+},0\right),\left(0,{\eta }_{{}_{i}}\right),\left(0,{\xi }_{{}_{\lambda ,j}}^{+}\right)|i=1,2,\cdots ,m,\lambda \in {\Sigma }^{+},j=1,2,\cdots ,{\beta }_{{}_{\lambda }}\right\}.\end{array}\end{array}$ (24)
Set $a=expH,H\in \mathfrak{a}$  . Then by $\left[H,{\eta }_{{}_{i}}\right]=0$  , $\left[H,{\xi }_{{}_{\lambda ,j}}^{±}\right]=\lambda \left(H\right){\xi }_{{}_{\lambda ,j}}^{\mp }$  and  22 , it can be calculated that
 $\begin{array}{cc}Ad\left({a}^{-1}\right)\left({\eta }_{{}_{i}}\right)-0& ={\eta }_{{}_{i}},\end{array}$ (25)
 $\begin{array}{cc}Ad\left({a}^{-1}\right)\left({\xi }_{{}_{\lambda ,j}}^{+}\right)-0& =-sinh\left(\lambda \left(H\right)\right){\xi }_{{}_{\lambda ,j}}^{-}+cosh\left(\lambda \left(H\right)\right){\xi }_{{}_{\lambda ,j}}^{+},\end{array}$ (26)
 $\begin{array}{cc}Ad\left({a}^{-1}\right)\left(0\right)-{\eta }_{{}_{i}}& =-{\eta }_{{}_{i}},\end{array}$ (27)
 $\begin{array}{cc}Ad\left({a}^{-1}\right)\left(0\right)-{\xi }_{{}_{\lambda ,j}}^{+}& =-{\xi }_{{}_{\lambda ,j}}^{+}.\end{array}$ (28)
 $\begin{array}{}\end{array}$
Thus by the first conclusion (i), the set $F$  is exactly composed of a basis of ${T}_{a}{O}_{a}$  .
Lemma 3.3. (i) $dim{M}_{a}=dimM$  , for all $a\in {A}^{\prime }$  .
(ii) ${T}_{a}G={T}_{a}A\oplus {T}_{a}{O}_{a}$  , for all $a\in {A}^{\prime }$  .
Proof. (i) Since $M\sim ={Z}_{{}_{K}}\left(\mathfrak{a}\right)$  in lemma  3.1 , the Lie algebra of $M$  is $\mathfrak{m}={Z}_{\mathfrak{k}}\left(\mathfrak{a}\right)$  . By lemma  3.2 , it is obvious that
 $\begin{array}{c}{T}_{a}{O}_{a}=d{L}_{a}\left(\mathfrak{k}\oplus \mathfrak{b}\right).\end{array}$ (29)
Then by  19 ,
 $\begin{array}{cc}dim{M}_{a}& =dim\left(K×K\right)-dim{O}_{a}\end{array}$
 $\begin{array}{cc}& =dim\mathfrak{k}+dim\mathfrak{k}-\left(dim\mathfrak{k}+dim\mathfrak{b}\right)\end{array}$
 $\begin{array}{cc}& =dim\mathfrak{k}-dim\mathfrak{l}\end{array}$
 $\begin{array}{cc}& =dim\mathfrak{m}\end{array}$
 $\begin{array}{cc}& =dimM.\end{array}$
 $\begin{array}{}\end{array}$
(ii) By  29 , it is directly obtained that ${T}_{a}G=d{L}_{a}\left(\mathfrak{g}\right)=d{L}_{a}\left(\mathfrak{k}\oplus \mathfrak{p}\right)=d{L}_{a}\left(\mathfrak{a}\oplus \left(\mathfrak{b}\oplus \mathfrak{k}\right)\right)={T}_{a}A\oplus {T}_{a}{O}_{a}.$  Proof of theorem  2.5 . By Lemma  3.1 - 3.3 , proposition  2.1 and the definition of Weyl integration model, it is obvious.
Proof of theorem  2.6 . By theorem  2.2 and  2.5 , it only needs to calculate $J\left(a\right)$  in  3 . It is obvious that ${T}_{a}A\perp {T}_{a}{O}_{a}$  for any $a\in {A}^{\prime }$  . Then by the second conclusion in lemma  3.3 and  13 , we need to consider the map
 $\begin{array}{c}{\Psi }_{a}:{T}_{\left[\left(e,e\right)\right]}\left(\left(K×K\right)/M\right)\to {T}_{a}{O}_{a},\left({\zeta }_{{}_{1}},{\zeta }_{{}_{2}}\right)↦\frac{d}{dt}{|}_{t=0}exp\left(t{\zeta }_{{}_{1}}\right)\cdot a\cdot exp\left(-t{\zeta }_{{}_{2}}\right).\end{array}$ (30)
Note that the tangent space ${T}_{\left[\left(e,e\right)\right]}\left(\left(K×K\right)/M\right)$  of $\left(K×K\right)/M$  at its unit element $\left[\left(e,e\right)\right]$  is exactly isomorphic to
 $\begin{array}{cc}\left(\mathfrak{k},\mathfrak{l}\right)& \equiv \left\{\left({\zeta }_{{}_{1}},{\zeta }_{{}_{2}}\right)|{\zeta }_{{}_{1}}\in \mathfrak{k},{\zeta }_{{}_{2}}\in \mathfrak{l}\right\}.\end{array}$ (31)
 $\begin{array}{}\end{array}$
Then the set
 $\begin{array}{c}\begin{array}{c}\left\{\left({\eta }_{{}_{i}},0\right),\left({\xi }_{{}_{\lambda ,j}}^{+},0\right),\left(0,{\xi }_{{}_{\lambda ,j}}^{+}\right)|i=1,2,\cdots ,m,\lambda \in {\Sigma }^{+},j=1,2,\cdots ,{\beta }_{{}_{\lambda }}\right\}\end{array}\end{array}$ (32)
is composed of a basis of ${T}_{\left[\left(e,e\right)\right]}\left(\left(K×K\right)/M\right)$  .
It is completely analogous to the calculation in  22 and  25 -(27), then we obtain
 $\begin{array}{cc}{\Psi }_{a}\left(\left({\eta }_{{}_{i}},0\right)\right)& =d{L}_{a}\left({\eta }_{{}_{i}}\right),\end{array}$ (33)
 $\begin{array}{cc}{\Psi }_{a}\left(\left({\xi }_{{}_{\lambda ,j}}^{+},0\right)\right)& =d{L}_{a}\left(-sinh\left(\lambda \left(H\right)\right){\xi }_{{}_{\lambda ,j}}^{-}+cosh\left(\lambda \left(H\right)\right){\xi }_{{}_{\lambda ,j}}^{+}\right),\end{array}$ (34)
 $\begin{array}{cc}{\Psi }_{a}\left(\left(0,{\xi }_{{}_{\lambda ,j}}^{+}\right)\right)& =d{L}_{a}\left(-{\xi }_{{}_{\lambda ,j}}^{+}\right).\end{array}$ (35)
 $\begin{array}{}\end{array}$
Thus by the second conclusion in lemma  3.2 and the invariance of those chosen measures, there is a constant $C$  such that
$J\left(a\right)=C{\prod }_{\lambda \in {\Sigma }^{+}}|sinh\left(\lambda \left(H\right)\right){|}^{{\beta }_{{}_{\lambda }}}.$
It completes the proof of theorem  2.6 .
References

1. KNAPP, A. W., Lie Groups—Beyond an Introduction, 2nd edition, Boston, Birkhäuser,2002.
2. HELGASON, S., Groups and Geometric Analysis—Integral Geometry, Invariant Differetial Operators and Spherical Functions, Mathematical Surveys and Monographs, Vol. 83, New York, AMS, 2000.
3. An, J. P. and Wang, Z. D., A Generalization of Weyl Integration Formula, Institute of Mathematics, Peking University, Research Report No.18, 2004.
4. HELGASON, S., Differential Geometry, Lie Groups, and Symmetric Spaces, New tork, Acadymic Press, 1978.