math0208215/ko.tex
1: \documentclass{amsart}
2: 
3: \input epsf.tex
4: 
5: \newtheorem{theorem}{Theorem}[section]
6: \newtheorem{lemma}[theorem]{Lemma}
7: \newtheorem{prop}[theorem]{Proposition}
8: 
9: 
10: 
11: \theoremstyle{definition}
12: \newtheorem{definition}[theorem]{Definition}
13: \newtheorem{example}[theorem]{Example}
14: \newtheorem{xca}[theorem]{Exercise}
15: \newtheorem{conjecture}[theorem]{Conjecture}
16: 
17: 
18: \theoremstyle{remark}
19: \newtheorem{remark}[theorem]{Remark}
20: 
21: \numberwithin{equation}{section}
22: 
23: %    Absolute value notation
24: \newcommand{\abs}[1]{\lvert#1\rvert}
25: 
26: \begin{document}
27: \author[B. Kr\"otz]{Bernhard Kr\"otz}
28: \author[M. Otto]{Michael Otto}
29: \address{The Ohio State University, Department of Mathematics, 231 West 18th Avenue, 
30: Columbus, OH 43210-1174}
31: \email{kroetz@math.ohio-state.edu, otto@math.ohio-state.edu}
32: \thanks {The first author was supported in part by NSF-grant DMS-0097314}
33: \title[Double coset decomposition]
34: {A Convexity property for the $SO(2,{\mathbb C})$-double coset 
35: decomposition of $SL(2,{\mathbb C})$ and applications to spherical functions}
36: 
37: 
38: \begin{abstract} 
39: We make a fine study of the $SO(2,{\mathbb C})$-double coset decomposition in $SL(2,{\mathbb C})$
40: and give a full description of the intersection of the various cells with the complex 
41: crown $\Xi$ of $SL(2,{\mathbb R})/ SO(2)$. A non-linear convexity theorem is proved 
42: and applications to analytically continued spherical functions are given. 
43: \end {abstract}
44: \maketitle
45: \section{Introduction}
46: 
47: With a Riemannian symmetric space $X=G/K$ of the non-compact type
48: comes a natural complexification $\Xi$, the so-called 
49: {\it complex crown} of $X$. 
50: One can define $\Xi$ in various ways. To begin 
51: with let $G=NAK$ be an Iwasawa decomposition and let 
52: $\mathfrak{a}=\mbox{Lie}(A)$ be the Lie algebra of $A$. 
53: Studying proper $G$-actions on $X_{\mathbb C}=G_{\mathbb C}/ K_{\mathbb C}$
54: Akhiezer and Gindikin were led to the following definition 
55: (cf.\ \cite{AG})  
56: 
57: \begin{equation}\label{a} \Xi=G\exp(i\Omega)K_{\mathbb C}/ K_{\mathbb C}\ ,\end{equation}
58: where $\Omega\subseteq\mathfrak{a}$ is a certain bounded 
59: convex set (cf.\ Section 2). Observe that  $\Xi$ is a $G$-invariant 
60: domain in $X_{\mathbb C}$ containing $X$. Moreover,
61: the definition of $\Xi$ is independent from the choice of $\mathfrak{a}$
62: and hence $\Xi$ is generically defined through $X$. 
63: \par Equivalently, one can define $\Xi$ by 
64: \begin{equation}\label{b} 
65: \Xi=\left(\bigcap_{g\in G} g N_{\mathbb C} A_{\mathbb C} K_{\mathbb C}/ K_{\mathbb C} \right)_0\ , 
66: \end{equation}
67: where $(\cdot)_0$ refers to the connected component of $(\cdot)$
68: containing $X$ (cf.\ \cite{FK, GM1, KS1}, \cite{BA, HU},  \cite{FH, GM2, HW}
69: and \cite{KS3, MA2}).  
70: Using the {\it complex convexity theorem} from \cite{GK2} one can 
71: significantly improve on \ref{b}, namely 
72: \begin{equation}\label{c} 
73: \Xi=\left(\bigcap_{g\in G} g N_{\mathbb C} A\exp(i\Omega)  K_{\mathbb C}/ K_{\mathbb C} \right)_0 \ . 
74: \end{equation}
75: \par The various definitions of $\Xi$ are useful. In \cite{KS2}, where the foundations 
76: for the $G$-invariant complex geometry of $\Xi$ were laid out 
77: (construction of natural  $G$-invariant psh exhaustion functions, K\"ahler metrics etc.), one mainly used the 
78: description (\ref{a}). For other purposes such as analytic continuation of eigenfunctions or the
79: heat kernel on $X$ to holomorphic  functions on $\Xi$ (cf.\ \cite{KS1}, \cite {KS2})
80: the characterizations in (\ref{b}) and in particular (\ref{c}) are more appropriate. 
81: 
82: \smallskip One of the main objectives in the study of complex crowns is to achieve 
83: a better understanding of harmonic analyis on symmetric spaces. For example 
84: it was shown in \cite{GK1} that all non-compactly causal symmetric 
85: spaces $G/H$ associated to $G$ appear in the so-called {\it distinguished boundary} of $\Xi$. 
86: Subsequently this was used in \cite{GKO, GKO2} to construct a Hardy-space for 
87: non-compactly causal symmetric spaces $G/H$ yielding first progress towards a geometric realization of some of the continuous series in $L^2(G/H)$ (Gelfand-Gindikin program).  
88: 
89: \medskip  The motivation for this article stems from our interest to obtain 
90: a first understanding of the various kinds of weighted Bergman-spaces which 
91: can be associated with $\Xi$, most notably Fock-spaces which play 
92: a prominent role in the context of the heat kernel transform on $X$ (cf.\ \cite{KS2}). 
93: Intimately related to this circle of problems is the growth behavior of 
94: analytically continued spherical functions. 
95: \par A spherical function $\phi_\lambda$
96: on $X$ is left $K$-invariant, hence its analytic continuation $\tilde\phi_\lambda$  to $\Xi$ 
97: is necessarily (locally) $K_{\mathbb C}$-invariant. 
98: This suggests that one should study the relation of $\Xi$ with the complexified polar decomposition 
99: $\Xi_P\:=K_{\mathbb C} A_{\mathbb C} K_{\mathbb C}/ K_{\mathbb C}$ in $X_{\mathbb C}$. 
100: 
101: \par Whereas the complex crown $\Xi$ is 
102: contained in the complexified Iwasawa decomposition $N_{\mathbb C} A_{\mathbb C} K_{\mathbb C}/ K_{\mathbb C}$ 
103: (cf.\ (\ref{b})), the same does not hold for the complexified polar decomposition, i.e., 
104: $\Xi\not\subseteq \Xi_P$. Furthermore, the 
105: polar domain $\Xi_P$ is not even open in $X_{\mathbb C}$. These are disappointing 
106: facts. But before giving up, we at least wanted to understand 
107: what actually goes ``wrong'' in the most important case of $G=SL(2,{\mathbb R})$. 
108: It turned out that the above mentioned difficulties  can be overcome, and, moreover, 
109: by settling the technical problems we could make new 
110: observations regarding the structure theory of complex crowns. 
111: \par This paper focuses mostly on $G=SL(2,{\mathbb R})$, but  
112: we made an effort to present  results in a fashion which allows 
113: immediate generalization 
114: to all semisimple Lie groups. An approach in full generality will be given elsewhere. 
115: 
116: \par Let us now describe some of our results in more detail. First observe that 
117: the polar domain $\Xi_P$ contains a Zariski open subset 
118: of $X_{\mathbb C}$;  in particular it is dense in $X_{\mathbb C}$. Thus almost all elements in $\Xi$ are 
119: contained in $\Xi_P$. Our main observation is the following 
120: non-linear convexity theorem:
121: 
122: \medskip\noindent{\bf Theorem 4.1.} {\it Let $G=SL(2, {\mathbb R})$ and $x=g\exp(iY)K_{\mathbb C} \in \Xi\cap \Xi_P$ with $g\in G$ and 
123: $Y\in \Omega$ (cf.\ (\ref{a})). Then 
124: $$x\in K_{\mathbb C} A\exp(i{\rm conv}({\mathcal W}Y))K_{\mathbb C}$$
125: with ${\rm conv}({\mathcal W}Y)$ the convex hull of the Weyl group orbit of $Y$.}
126: 
127: \medskip We conjecture that Theorem 4.1 holds true for all semisimple Lie groups 
128: and we provide additional  evidence with a discussion of the Lorentz groups $G=SO_e(1,n)$. 
129: \par Using general  results of \cite{KS1} one obtains from Theorem 4.1 the following estimate:
130: 
131: \medskip\noindent{\bf Theorem 5.2.} {\it  Let $G=SL(2, {\mathbb R})$. Let $\phi_\lambda$ be a positive definite 
132: spherical function on $X$ and $\tilde\phi_\lambda$ its analytic continuation 
133: to $\Xi$. Then $\tilde\phi_\lambda$ is bounded.}
134: 
135: \bigskip Finally,  as an application of our methods we give an 
136: estimate for the $A_{\mathbb C}$-projection in $\Xi_P$ for 
137: elements in $\Xi$. 
138: \par It is our pleasure to thank the referee for his very careful
139: reading and his useful suggestions. In particular the easy proof
140: of Theorem \ref{union} is due to him.
141: 
142: 
143: \section{Notation and general facts} 
144: 
145: Let $\mathfrak{g} $ be a real semisimple Lie algebra and 
146: $\ \mathfrak{g}=\mathfrak{k}+\mathfrak{p} \ $ a Cartan 
147: decomposition. For a maximal abelian subspace $\mathfrak{a}$
148: of $\mathfrak{p}$ let $\ \Sigma = \Sigma (\mathfrak{g},
149: \mathfrak{a} ) \subseteq \mathfrak{a}^{*} \ $ be the 
150: corresponding restricted root system. 
151: Then $\mathfrak{g}$ admits a root space decomposition
152: $$ \mathfrak{g} = \mathfrak{a} \oplus \mathfrak{m} \oplus
153: \bigoplus_{\alpha \in \Sigma} \mathfrak{g}^{\alpha} ,$$
154: where $\ \mathfrak{m}=\mathfrak{z}_{\mathfrak{k}} (\mathfrak{a})
155: \ $ and $\ \mathfrak{g}^{\alpha}=\{ X \in \mathfrak{g} :
156: (\forall H\in \mathfrak{a}) \  \left[ H,X \right] = \alpha(H)X\} \ $ . \\
157: For a fixed positive system $\Sigma^{+}$ define $\
158: \mathfrak{n} := \bigoplus_{\alpha \in \Sigma^{+}}
159: \mathfrak{g}^{\alpha} \ $ . Then we have the Iwasawa 
160: decomposition on the Lie algebra level:
161: \[ \mathfrak{g} = \mathfrak{k} \oplus \mathfrak{a} \oplus
162: \mathfrak{n} . \]
163: We write $\ \mathcal{ W}=N_{K}(\mathfrak{a})/Z_{K}(\mathfrak{a})
164: \ $ for the corresponding {\it Weyl group}. 
165: For an element $\ X \in \mathfrak{a} \ $ we denote by 
166: $ {\rm conv}(\mathcal{W}X) $ the convex hull of the Weyl group
167: orbit of $X$. \\
168: \\
169: For any real Lie algebra $\mathfrak{l}$ we write 
170: $\mathfrak{l}_{\mathbb C}$ for its complexification. \\
171: \par In the sequel $G_{\mathbb C}$ will denote  a simply connected Lie group with Lie
172: algebra $\mathfrak{g}_{\mathbb C}$. We write  $\ G, K, K_{\mathbb C}, 
173: A, A_{\mathbb C}, N \ $ and $N_{\mathbb C}$ the analytic 
174: subgroups of $G_{\mathbb C}$ corresponding to subalgebras $\ \mathfrak{g}, 
175: \mathfrak{k}, \mathfrak{k}_{\mathbb C}, \mathfrak{a}, 
176: \mathfrak{a}_{\mathbb C}, \mathfrak{n} \ $ and 
177: $\mathfrak{n}_{\mathbb C}$ , respectively. \\ 
178: \\
179: The following bounded and convex subset of $\mathfrak{a}$ plays a central role:
180: \[ \Omega := \{ X\in \mathfrak{a} : \abs{\alpha(X)}<
181: \frac{\pi}{2} \quad \forall \ \alpha \in \Sigma \} . \]
182: \\
183: With $\Omega$ we define a left $G$ and right $K_{\mathbb C}$-invariant domain in 
184: $G_{\mathbb C}$ by 
185: $$ \tilde \Xi=G\exp(i\Omega)K_{\mathbb C}\ .$$
186: Also we write 
187: $$\Xi=\tilde \Xi/ K_{\mathbb C}$$
188: for the union of right $K_{\mathbb C}$-cosets of $\tilde \Xi$ in the complex symmetric space 
189: $G_{\mathbb C}/ K_{\mathbb C}$. We refer to $\Xi$ as the {\it complex crown}
190: of the symmetric space $G/K$. Notice that $\Xi$ is independent of the 
191: choice of $\mathfrak a$ and hence generically defined through $G/K$. 
192: 
193: 
194: \par As every root $\alpha \in \Sigma $ 
195: is analytically integral, it gives rise to a character 
196: of $A_{\mathbb C}$ by 
197: $$\ \xi_{\alpha} : A_{\mathbb C} \longrightarrow
198: \mathbb{C}^* , \ \xi_{\alpha}(\exp(X))=e^{\alpha(X)} \ .$$ \\
199: Let us define the set of {\it regular elements} in $A_{\mathbb C}$ by 
200: $$A_{\mathbb{C},\mbox{reg}} = \{ z \in 
201: A_{\mathbb C} : \xi_{\alpha}^2(z) \neq 1 \quad \forall \ \alpha
202: \in \Sigma \} .$$
203: Notice that $A_{\mathbb{C},\mbox{reg}}$ is an algebraic variety
204: ( principal open set in $A_{\mathbb C}$ ). \\
205: 
206: We also define 
207: $$A_{{\mathbb C}, \mbox{sing}}=A_{\mathbb C}\backslash
208: A_{\mathbb{C},\mbox{reg}}$$
209: and call it the {\it singular set} in $A_{\mathbb C}$. 
210: 
211: 
212: \begin{lemma}
213: The following assertions hold:
214: \begin{enumerate}
215: \item The multiplication mapping
216: \[ m:K_{\mathbb C} \times A_{\mathbb{C},\mbox{reg}} \times
217: K_{\mathbb C} \longrightarrow G_{\mathbb C} , \ 
218: (k_{1},a,k_{2}) \mapsto k_{1}ak_{2} , \]
219: is submersive. In particular, $\ K_{\mathbb C} 
220: A_{\mathbb{C},\mbox{reg}} K_{\mathbb C} \subseteq G_{\mathbb C}
221: \ $ is open.
222: \item $K_{\mathbb C} A_{\mathbb{C},\mbox{reg}} K_{\mathbb C}$
223: is dense in $G_{\mathbb C}$.
224: \end{enumerate}
225: \end{lemma}
226: \begin{proof}
227: (1) is a standard computation which will not be repeated here. 
228: \par\noindent (2) Notice that $\ K_{\mathbb C} A_{{\mathbb C},\mbox{reg}} K_{\mathbb C}
229: \subseteq G_{\mathbb C} \ $ is a constructible set as the 
230: image under the regular mapping 
231: $\ m:K_{\mathbb C} \times A_{{\mathbb C},\mbox{reg}} \times K_{\mathbb C} 
232: \longrightarrow G_{\mathbb C} \ $ . Thus
233: it follows from (1) that $ K_{\mathbb C} A_{{\mathbb C},\mbox{reg}} 
234: K_{\mathbb C} $ contains a Zariski-open subset of 
235: $G_{\mathbb C}$. This proves (2). 
236: \end{proof} 
237: 
238: 
239: \begin{remark} Notice that the $K_{\mathbb C}$-bi-invariant  domain 
240: $K_{\mathbb C}A_{\mathbb C} K_{\mathbb C}$ is not open in 
241: $G_{\mathbb C}$ (cf. our discussion of $G=SL(2,{\mathbb R})$ in Section 3).
242: \end{remark}
243: 
244: 
245: 
246: 
247: \section{The structure of the double $K_{\mathbb C}$-cosets}
248: In this section we will give a detailed analysis of the double $K_{\mathbb C}$-coset decomposition 
249: for $G_{\mathbb C}=SL(2,\mathbb C)$. Further we will start the investigation of the intersections
250: of $\tilde \Xi$ with the various cells in the double $K_{\mathbb C}$-coset 
251: decomposition. 
252: 
253: 
254: 
255: \par Let us introduce the necessary notation. Our choice of a maximal compact subgroup of $G$ is $K=SO(2,{\mathbb R})$ and our choice of 
256: $\mathfrak a$ will be 
257: $${\mathfrak a}=\left\{ \left( \begin{array}{cc} t & 0 \\ 0 & -t \end{array} \right):\ t\in {\mathbb R}\right\} \ .$$
258: We choose $\Sigma^+$ such that 
259: $${\mathfrak n}=\left\{ \left( \begin{array}{cc} 0 & x \\ 0 & 0\end{array} \right):\ x\in {\mathbb R}\right\} \ .$$
260: Notice that $G_{\mathbb C}=SL(2,{\mathbb C})$ is simply connected and $K_{\mathbb C}=SO(2,{\mathbb C})$. 
261: 
262: 
263: \par In order to study $K_{\mathbb C}$-double cosets the use of spherical functions is 
264: useful. 
265: 
266: \begin{definition} On $G_{\mathbb C}=SL(2,{\mathbb C})$ we define the {\it elementary spherical function} $\Phi$
267: by 
268: \[ \Phi :G_{\mathbb{C}} \longrightarrow \mathbb{C}, \quad 
269: \Phi (g) := \mbox{tr}(gg^{t}) \quad \forall \ g\in G_{\mathbb{C}} . \]
270: ( For $\ g = \left( \begin{array}{cc} a & b \\ c & d \end{array} \right)\ $ we have $\ \Phi(g)=a^2 +b^2 +c^2+d^2$. )
271: \par Note that the function $\Phi $ is holomorphic and 
272: $K_{\mathbb C}$-bi-invariant.
273: \end{definition}
274: \par The set of singular elements in $A_{\mathbb C}$ is given by 
275: 
276: $$A_{{\mathbb C}, \mbox{sing}}=\left\{ \left( \begin{array}{cc} z & 0 \\ 0 & z^{-1} \end{array} \right) \
277: :\ z\in\{-i,i,-1,1\}\right\} \ .$$
278: 
279: 
280: \par Let $\ g \in G_{\mathbb C} \setminus K_{\mathbb C} 
281: A_{\mathbb C} K_{\mathbb C} . \ $ According to Lemma 2.1 
282: we can find sequences \[
283: (k_{n}) \subset K_{\mathbb C}, \ (a_{n}) \subset A_{\mathbb C}
284: , \ (\tilde{k}_{n}) \subset K_{\mathbb C} \quad
285: \mbox{with} \quad k_{n}a_{n}\tilde{k}_{n} \longrightarrow g . 
286: \]
287: 
288: \begin{prop}\label{limits}
289: Let $\ g \in G_{\mathbb C} \setminus K_{\mathbb C} A_{\mathbb C} K_{\mathbb C} \ $ be the limit of the sequence $\ 
290: (k_{n}a_{n}\tilde{k}_{n})
291: \subset K_{\mathbb C} A_{\mathbb C} K_{\mathbb C} . $ 
292: Then  $(a_{n})\subset A_{\mathbb C}$ is bounded and for every 
293: convergent subsequence $(a_{n_k})$ we have  
294: $$\lim_{k \to \infty} a_{n_k} \in A_{{\mathbb C}, \mbox{sing}}\ . $$
295: \end{prop}
296: \begin{proof}
297: Write $\ a_{n}=\left( \begin{array}{cc} z_{n} & 0 \\ 0 & z_{n}^{-1} \end{array} \right) \ $. 
298: By continuity of the map $\Phi $ we get
299: \[ \Phi (g) = \mbox{tr}(gg^{t}) = \lim_{n \to \infty } \mbox{tr}(a_{n}a_{n}^{t})
300: = \lim_{n \to \infty} (z_{n}^{2}+z_{n}^{-2}) . \]
301: It follows that there exist $\ m, M > 0 \ $ such that
302: \[ 0 < m < \abs{z_{n}} < M < \infty \quad \forall n . \]
303: Passing to an appropriate subsequence we can assume that $\ 
304: z_{n} \longrightarrow z_{0} \neq 0 . $ \\ 
305: \\
306: Suppose $(k_{n})$ has a convergent subsequence. W.l.o.g. we may assume that  \[
307: k_{n} \longrightarrow k_{0} 
308: \in K_{\mathbb C}, \quad a_{n} \longrightarrow a_{0} \in A_{\mathbb C} . \]
309: But then $\ \tilde{k}_{n} = a_{n}^{-1} k_{n}^{-1} (k_{n}a_{n}\tilde{k}_{n}) \
310: $ also converges. \\
311: We have now
312: \[ k_{n} \longrightarrow k_{0} 
313: \in K_{\mathbb C}, \quad a_{n} \longrightarrow a_{0} \in A_{\mathbb C}, 
314: \quad \tilde{k}_{n} \longrightarrow \tilde{k}_{0} \in K_{\mathbb C} . \]
315: But this would imply $\ g = \lim_{n\to\infty} k_{n} a_{n} \tilde{k}_{n} \in 
316: K_{\mathbb C} A_{\mathbb C} K_{\mathbb C} , \ $ contradicting the assumption.
317: \\
318: By the same argument we see that $(\tilde{k}_{n})$ cannot have a 
319: convergent subsequence. \\ 
320: \\
321: We want to know more about the limits $z_{0}$. \\
322: Notice that $k_n a_n^2 k_n^t$ converges in $G_{\mathbb C}$. 
323: For simplicity we omit the indices writing $k=k_{n}$ and
324: $a=a_{n}$. We look at the elements $\ ka^{2}k^{t} \in G_{\mathbb C}/
325: K_{\mathbb C} . $  \\ 
326: Due to the $K_{\mathbb C}$-bi-invariance of $ K_{\mathbb C} A_{\mathbb C} 
327: K_{\mathbb C} $ and the compactness of $K$ we can assume that 
328: \[ k=\left( \begin{array}{cc}
329: \cosh t & i\sinh t \\ -i\sinh t & \cosh t \end{array} \right) 
330: = \left( \begin{array}{cc}
331: \cos it & \sin it \\ -\sin it & \cos it \end{array} \right),  \]
332: where $t\in {\mathbb R}$ and $|t|\to \infty$. 
333: Then with $a=\left( \begin{array}{cc}
334: z& 0 \\ 0 & z^{-1} \end{array} \right)$ we have 
335: \begin{eqnarray}\label{matrix}
336: ka^{2}k^{t} & = & \left( \begin{array}{cc}
337: \cosh t & i\sinh t \\ -i\sinh t & \cosh t \end{array} \right) 
338: \left( \begin{array}{cc}
339: z^{2} & 0 \\ 0 & z^{-2} \end{array} \right)
340: \left( \begin{array}{cc}
341: \cosh t & -i\sinh t \\ i\sinh t & \cosh t \end{array} \right) 
342: \nonumber \\
343: & = &
344: \left( \begin{array}{cc}
345: z^{2} \cosh^{2}t-z^{-2}\sinh^{2}t  & (z^{-2}-z^{2})i\sinh t \cosh t \\
346: (z^{-2}-z^{2})i\sinh t \cosh t & z^{-2} \cosh^{2}t -z^{2} \sinh^{2} t
347: \end{array} \right) . 
348: \end{eqnarray}
349: We know that $\ z \to z_{0} \ $ and $\ \abs{t} \to \infty . $ \\
350: The upper left entry in (\ref{matrix}) must converge, therefore
351: \[ \frac{z^{2} \cosh^{2}t-z^{-2}\sinh^{2}t}{\sinh^{2}t} = 
352: z^{2} \frac{\cosh^{2}t}{\sinh^{2}t}-z^{-2} \longrightarrow 0 . \]
353: But this implies $\ z_{0}^{2}=z_{0}^{-2} \ $ , hence $\ z_{0}^{2}=\pm 1 . 
354: $ 
355: \end{proof}
356: 
357: 
358: \par The vector space $\mathfrak{p}_{\mathbb C}$ admits a decomposition 
359: $\mathfrak{p}_{\mathbb C}=\mathfrak{p}^+\oplus \mathfrak{p}^-$
360: into irreducible $\mathfrak{k}_{\mathbb C}$-modules with 
361: 
362: \[ \mathfrak{p}^{-} = \mathbb{C} \left( \begin{array}{cc}
363: 1 & -i \\ -i & -1 \end{array} \right) \quad \mbox{and} \quad
364: \mathfrak{p}^{+} = \mathbb{C} \left( \begin{array}{cc}
365: 1 & i \\ i & -1 \end{array} \right) . \]
366: The corresponding analytic subgroups of $G_{\mathbb C}$ are given by 
367: \[ P^{-} = \left\{ \left( \begin{array}{cc} 1-u & iu \\ iu & 1+u
368: \end{array} \right) : u \in \mathbb C \right\} \quad \mbox{and} \quad
369: P^{+} = \left\{ \left( \begin{array}{cc} 1+u & iu \\ iu & 1-u
370: \end{array} \right) : u \in \mathbb C \right\} . \]
371: We introduce punctured discs in $P^-$ and $P^+$ by 
372: $$P_{1\over 2}^-=\left\{ \left( \begin{array}{cc} 1-v & iv \\ iv & 1+v \end{array} 
373: \right) : v \in \mathbb{C}, 0<\abs{v}<\frac{1}{2} \right\} $$
374: and similarily we define $P_{1\over 2}^+$. \\
375: Finally, we set
376: \[ A_{\mathbb C}':= A_{\mathbb C} \setminus \{ \pm {\bf 1} \} \quad
377: \mbox{and} \quad P^{\mp '}:= P^{\mp} \setminus \{ {\bf 1} \} , \]
378: \\
379: and define an element 
380: $$y_0=\left( \begin{array}{cc} i & 0 \\ 0 & -i
381: \end{array} \right) \in A_{{\mathbb C},\mbox{sing}}\ .$$
382: We can now describe how $G_{\mathbb C}$ decomposes into a disjoint
383: union of $K_{\mathbb C}$-bi-invariant subsets. 
384: Furthermore, we will give a complete description of the 
385: intersection of $\tilde\Xi=G\exp(i\Omega)K_{\mathbb C}$ with the lower 
386: dimensional cells. The 
387: intersection of $\tilde\Xi$ with the big cell $K_{\mathbb C} A_{\mathbb C} 
388: K_{\mathbb C} $ will be subject of the next section.
389: \\
390: \\
391: \begin{theorem}\label{union}
392: \begin{enumerate}
393: \item The $K_{\mathbb C}$-double coset decomposition of $G_{\mathbb C}$ is given  by:
394: \[ G_{\mathbb C}=K_{\mathbb C} A_{\mathbb C}' K_{\mathbb C} \
395: \amalg \ P^{-'} K_{\mathbb{C}} \ \amalg \ 
396: P^{+'} K_{\mathbb{C}} \ \amalg \ P^{-'} y_0 K_{\mathbb{C}} \ 
397: \amalg \ P^{+'} y_0 K_{\mathbb{C}} \ \amalg \ K_{\mathbb C}
398: . \]
399: \item The following equality holds:
400: $$\tilde\Xi  \setminus K_{\mathbb C} A_{\mathbb C} 
401: K_{\mathbb C} = P_{1\over 2}^-K_{\mathbb C} \amalg P_{1\over 2}^+K_{\mathbb C}\ .$$
402: \end{enumerate}
403: \end{theorem}
404: \medskip
405: In order to prove Theorem \ref{union} it is useful to adopt a more geometric point of view. 
406: We closely follow the suggestions of the referee. 
407: \par In the sequel we realize $X=SL(2,\mathbb{R})/SO(2, \mathbb{R})$ as 
408: the upper sheet of the two-sheeted hyperboloid
409: \[ x_{0}^{2}-x_{1}^{2}-x_{2}^{2}=1, \ x_{0}>0 \quad (x=(x_0, x_1, x_2)\in \mathbb{R}^{3}) \]
410: and $X_{\mathbb{C}}=SL(2,\mathbb{C})/SO(2,\mathbb{C})$
411: as the complex quadric 
412: \[ z_{0}^{2}-z_{1}^{2}-z_{2}^{2}=1 \quad (z=(z_0,z_1,z_2)\in \mathbb{C}^{3}). \]
413: Let $e_0=(1,0,0)$. Then the polar decomposition in $X$ is given by $X=KA. e_{0}$,
414: \[ x_{0}=\cosh t, \ x_{1}=\sinh t \cos \phi , \ x_{2}=\sinh t \sin \phi  \]
415: with $t,\varphi\in \mathbb{R}$. 
416: \medskip
417: 
418: \begin{proof}
419: $(Thm.\ref{union})$
420: Complexifying the real polar decomposition of $X$ it follows that a point 
421: $z\in X_{\mathbb{C}}$ belongs to $K_{\mathbb{C}}A_{\mathbb{C}}. e_{0}$
422: if and only if there exist $a,b,u,v \in \mathbb{C}$ with
423: \[ a^{2}+b^{2}=1, \quad u^{2}-v^{2}=1 , \]
424: such that
425: \[ z_{0}=u, \ z_{1}=va, \ z_{2}=vb . \]
426: Therefore $X_{\mathbb{C}}\setminus K_{\mathbb{C}}A_{\mathbb{C}}. e_{0} $
427: decomposes into four $K_{\mathbb{C}}$-orbits
428: \begin{eqnarray*}
429: (1) & & z_{0}=1, z_{1}=\tau, z_{2}=i\tau, \\
430: (2) & & z_{0}=1, z_{1}=\tau, z_{2}=-i\tau, \\
431: (3) & & z_{0}=-1, z_{1}=\tau, z_{2}=i\tau, \\
432: (4) & & z_{0}=-1, z_{1}=\tau, z_{2}=-i\tau, 
433: \end{eqnarray*}
434: where $\tau \in \mathbb{C}^{*}$. \\
435: The orbits $(1),(2),(3),(4)$ correspond to the $K_{\mathbb{C}}$-bi-invariant 
436: sets $P^{-'} K_{\mathbb{C}}$, $P^{+'} K_{\mathbb{C}}$, $P^{-'} y_0 K_{\mathbb{C}}$,
437: $P^{+'} y_0 K_{\mathbb{C}} $, respectively. \\
438: This proves part $(1)$ of the theorem. \\
439: \\
440: The points $z=x+iy \in \Xi \subset X_{\mathbb{C}}$ are characterized by
441: the property
442: \[ x_{0}>0, \quad x_{0}^{2}-x_{1}^{2}-x_{2}^{2}>0 . \]
443: For an element $z=(1,\tau,i\tau)$ in the orbit $(1)$ this translates to
444: $\abs{\tau}<1$. Therefore,
445: \[ \{ (1,\tau,i\tau)\in X_{\mathbb{C}} : \tau \in \mathbb{C}^{*} \}
446: \cap \Xi = \{ (1,\tau,i\tau)\in X_{\mathbb{C}} : \tau \in \mathbb{C}^{*},
447: \ \abs{\tau}<1 \} \ , \]
448: and this subset of $\Xi$ corresponds to $P_{1\over 2}^-K_{\mathbb C} \subset
449: \tilde{\Xi} $. \\
450: Similarly one relates orbit $(2)$ with $P_{1\over 2}^{+}K_{\mathbb C} \subset
451: \tilde{\Xi} $. \\
452: The intersections of both orbits $(3)$ and $(4)$ with $\tilde \Xi$ are empty.
453: \end{proof}
454: 
455: 
456: \begin{remark} (a) Theorem \ref{union}(1) can be deduced 
457: from  general results of Matsuki (cf.\ \cite{MA1}).
458: \par\noindent (b) The cells in the decomposition of $G_{\mathbb C}$ in Theorem \ref{union}
459: are characterized by the values of $\Phi $. 
460: The function $\Phi $ attains the value $2$ on $K_{\mathbb C}$ and on $\ P^{-'} 
461: K_{\mathbb{C}} \amalg P^{+'} K_{\mathbb{C}} \ $, the value $-2$ on $ \
462: P^{-'} y_0 K_{\mathbb{C}} \amalg P^{+'} y_0 K_{\mathbb{C}} \ $, and $\Phi $
463: can take any value except $2$ on $K_{\mathbb C}A_{\mathbb C}'K_{\mathbb C}$.
464: \end{remark}
465: 
466: 
467: 
468: 
469: 
470: 
471: \section{A non-linear convexity theorem}
472: In this section we will investigate the intersection of $\tilde\Xi$ with the big cell 
473: $K_{\mathbb C} A_{\mathbb C} K_{\mathbb C}$. 
474: An element $\ g \in
475:  G \exp (i\Omega) K_{\mathbb{C}} \cap K_{\mathbb C} 
476: A_{\mathbb C} K_{\mathbb C} \ $ can be written as 
477: 
478: \begin{equation}\label{g}
479: g = h \left( \begin{array}{cc} e^{i\theta} & 0 \\ 0 & 
480: e^{-i\theta} \end{array} \right) k = \tilde{k_{1}} \tilde{a} 
481: \tilde{k_{2}} , 
482: \end{equation}
483: where $\tilde k_1, \tilde k_2\in K_{\mathbb C}$ and 
484: \[ h = \left( \begin{array}{cc} a & b \\ c & d
485: \end{array} \right) \in G \quad \mbox{and} \quad \tilde{a}=\left( 
486: \begin{array}{cc} z & 0 \\ 0 & z^{-1} \end{array} \right)
487: \in A_{\mathbb C} . \]
488: 
489: 
490: We want to describe the dependence of $\tilde{a}$ from $\theta$ and $h$.
491: \\
492: \\
493: Recall that for $SL(2,\mathbb{R})$ the Weyl group consists of only two
494: elements and for $X\in \mathfrak{a}$ we have $\ {\rm conv}
495: (\mathcal{W}.X) = [-1,1] \cdot X \ $.
496: 
497: 
498: \begin{theorem}\label{conv}
499: Let $ X \in \Omega $ and $ h \in G $. Suppose that $\ 
500: g = h \exp(iX) = \tilde{k_{1}} \tilde{a} \tilde{k_{2}} \in 
501: K_{\mathbb C} A_{\mathbb C} K_{\mathbb C} \ $. Then
502: \[ g \in K_{\mathbb C} A \exp(i{\rm conv}(\mathcal{W}X)) 
503: K_{\mathbb C}
504: . \]
505: \end{theorem}
506: \begin{proof}
507: Using the Iwasawa decomposition $G=KAN$ we may assume that
508: \[
509: \begin{array}{cccccccc}
510: g & = & \underbrace{\left( \begin{array}{cc} 1 & r \\ 0 & 1 \end{array} 
511: \right) } & \underbrace{a\exp iX } & = & \underbrace{\left( \begin{array}{cc}
512: \cos \omega & \sin \omega \\ -\sin \omega & \cos \omega \end{array} 
513: \right) } & \underbrace{\left( \begin{array}{cc} z & 0 \\ 0 & z^{-1} 
514: \end{array} \right) } & \underbrace{\tilde{k}_{2} } \\
515: & & \in N & \in A\exp (i\Omega) & & \in K_{\mathbb C} & 
516: =\tilde{a} \in A_{\mathbb C} &
517: \in K_{\mathbb C}  
518: \end{array} \]
519: If we set $\ \left( \begin{array}{cc} u & 0 \\ 0 & u^{-1} 
520: \end{array} \right) := a\exp iX \ $ and $\ \theta := \mbox{Arg}(u) \ $,
521: then
522: \[ gg^{t}= \left( \begin{array}{cc}
523: u^{2}+r^{2}u^{-2} & ru^{-2} \\ ru^{-2} & u^{-2} \end{array} \right) =
524: \left( \begin{array}{cc} z^{2}\cos^{2}\omega + z^{-2}\sin^{2} \omega & 
525: (z^{-2}-z^{2})\sin \omega \cos \omega \\ (z^{-2}-z^{2})\sin \omega \cos 
526: \omega & z^{2}\sin^{2} \omega + z^{-2}\cos^{2} \omega \end{array} \right) ,
527: \] yielding 
528: \begin{equation}\label{form1}
529: \Phi (g)=\mbox{tr}(gg^{t})=u^{2}+(1+r^{2})u^{-2} = z^{2}+z^{-2} . 
530: \end{equation}
531: In Figure 1 the set of points of the hyperbola is
532: $\ \{ l e^{2i\theta} + l^{-1} e^{-2i\theta} : l \in \mathbb{R}, l>0
533: \} \ $. The complex number $z^{2}+z^{-2}$ lies within the shaded region. 
534: Therefore, there exists $w=l e^{i\theta}$ such that 
535: $\ z^{2}+z^{-2} = \lambda (w^{2}+w^{-2}) \ $ for some real number $\lambda >1$. 
536: 
537: 
538: \medskip
539: \epsfbox{febfig1.eps}
540: \medskip 
541: \par 
542: We now obtain the assertion of the theorem with the following Lemma \ref{convlem}.  
543: \end{proof}
544: 
545: \begin{lemma}\label{convlem} Let $r_1, r_2>0$ and $-{\pi\over 2}<\varphi_1,\varphi_2<{\pi\over 2}$.
546: Define complex numbers by $u=r_{1} e^{i\varphi_{1}}$ and $v=r_{2} e^{i\varphi_{2}}$. 
547: Assume that $\ v+v^{-1} = \lambda (u+u^{-1}) \ $, with $\lambda \in \mathbb{R}, \lambda
548: >1 $. Then, $\quad \abs{\varphi_{2}} \leq \abs{\varphi_{1}} \ $.
549: \end{lemma}
550: \begin{proof}
551: Clearly we can assume $\varphi_{2} \neq 0 $. 
552: Comparing real and imaginary parts of $\ v+v^{-1} = \lambda 
553: (u+u^{-1}) \ $ , followed by squaring, yields
554: \[  \quad \lambda^{2}(r_{1}+r_{1}^{-1})^{2}\cos^{2}{\varphi_{1}} = (r_{2}+r_{2}^{-1})^{2}\cos^{2}{\varphi_{2}} \ , \] and
555: \begin{equation}\label{eq1}
556: \lambda^{2}(r_{1}-r_{1}^{-1})^{2}\sin^{2}{\varphi_{1}} = (r_{2}-r_{2}^{-1})^{2}\sin^{2}{\varphi_{2}} \ . 
557: \end{equation} 
558: Adding these two equations we get 
559: \begin{equation}\label{eq2} 
560: \lambda^{2}(r_{1}-r_{1}^{-1})^{2}+4\lambda^{2}\cos^{2}{\varphi_{1}} = 
561: (r_{2}-r_{2}^{-1})^{2} + 4 \cos^{2}{\varphi_{2}} \ . 
562: \end{equation} 
563: Combining (\ref{eq1}) and (\ref{eq2}) gives
564: 
565: \begin{equation}\label{eq3} 
566: \lambda^{2}(r_{1}-r_{1}^{-1})^{2}+4\lambda^{2}\cos^{2}{\varphi_{1}} = \lambda^{2}(r_{1}-r_{1}^{-1})^{2}
567: \frac{\sin^{2}{\varphi_{1}}}{\sin^{2}{\varphi_{2}}} + 4\cos^{2}{\varphi_{2}} \ .
568: \end{equation}
569: If we assume $\ \abs{\varphi_{2}} > \abs{\varphi_{1}} \ $, then 
570: \[ \lambda^{2}(r_{1}-r_{1}^{-1})^{2} > \lambda^{2}(r_{1}-r_{1}^{-1})^{2} \frac{\sin^{2}{\varphi_{1}}}{\sin^{2}{\varphi_{2}}}
571: , \] 
572: and \[ 4\lambda^{2}\cos^{2}{\varphi_{1}} > 4\cos^{2}{\varphi_{2}} \ , \] 
573: a contradiction to (\ref{eq3}). 
574: \end{proof}
575: 
576: 
577: \begin{conjecture} Theorem \ref{conv} holds for any real connected semisimple
578: Lie group $G$. \end{conjecture}
579: 
580: 
581: \begin{example} To give more evidence for our conjecture we will now discuss the case 
582: of $G=SO_{e}(1,n)$, $n\geq 2$. 
583: \par On the Lie algebra level we have the Cartan decomposition $ \mathfrak{g}=
584: \mathfrak{k} + \mathfrak{p} $, where
585: \begin{eqnarray*}
586: \mathfrak{k} & = & \left\{ \left( \begin{array}{cc} 0 & 0 \\ 0 & B \end{array} 
587: \right) \in \mbox{Mat}(n+1,\mathbb{R}) : B \in \mathfrak{so}(n) \right\} \ ,  \\ 
588: \mathfrak{p} & = &\left\{ \left( \begin{array}{cc} 0 & v \\ v^{t} & 0 \end{array} \right)
589: \in \mbox{Mat}(n+1,\mathbb{R}) : v \in \mathbb{R}^{n} \right\} . 
590: \end{eqnarray*}
591: Choosing 
592: \[ \mathfrak{a} = \left\{ \left( \begin{array}{ccc} 0 & 0 & t \\ 0 & 0 & 0 \\
593: t & 0 & 0 \end{array} \right) \in \mathfrak{p} : t \in \mathbb {R} \right\}  \]
594: as maximal subalgebra of $\mathfrak{p}$ yields the root system $\ \Sigma =
595: \{ \alpha , -\alpha \} \ $ with 
596: \[ \alpha \left( \begin{array}{ccc} 0 & 0 & 1 \\ 0 & 0 & 0 \\
597: 1 & 0 & 0 \end{array} \right) = 1. \]
598: Therefore,
599: \[ \Omega = \{ X\in \mathfrak{a} : \abs{\alpha(X)}<
600: \frac{\pi}{2} \quad \forall \ \alpha \in \Sigma \} = \left\{ \left( 
601: \begin{array}{ccc} 0 & 0 & t \\ 0 & 0 & 0 \\ t & 0 & 0 \end{array} \right) :
602: t\in \mathbb{R}, \abs{t}<\frac{\pi}{2} \right\} . \]
603: We choose $\alpha $ as the positive root. Then the subgroups in the 
604: Iwasawa decomposition $G=KAN$ are 
605: \begin{eqnarray*}
606: K & = & \left\{ \left( \begin{array}{cc} 1 & 0 \\ 0 & B \end{array} 
607: \right) : B \in SO(n) \right\} \ \simeq \ SO(n) , \\
608: N & = & \left\{ \left( \begin{array}{ccc} 1+\frac{1}{2}\abs{\abs{v}}^{2} & v & 
609: -{1\over 2}\abs{\abs{v}}^{2} \\ v^{t} & \mbox{Id}_{n-1} & -v^{t} \\
610: \frac{1}{2}\abs{\abs{v}}^{2} & v & 1-\frac{1}{2}\abs{\abs{v}}^{2} 
611: \end{array} \right) : v \in \mathbb{R}^{n-1} \right\} , \\
612: A & = & \left\{ \left( \begin{array}{ccc} \cosh t & 0 & \sinh t \\ 0 & 
613: \mbox{Id}_{n-1} & 0 \\ \sinh t & 0 & \cosh t \end{array} \right) : t \in
614: \mathbb{R} \right\} .
615: \end{eqnarray*}
616: We consider the standard representation $(\pi ,V)$ of $G$ on $V=\mathbb{C}^{
617: n+1}$. \\
618: Clearly, $\ v_{0}=e_{1}=(1,0,\cdots ,0)^{t} \ $ is a $K$-fixed vector. The 
619: elementary spherical function $\Phi $ on $\ G_{\mathbb C}\simeq 
620: SO(n+1,\mathbb{C}) \ $ is given by
621: \[ \Phi :G_{\mathbb C} \longrightarrow \mathbb{C}, \quad \Phi (g)= \langle \pi(g)
622: v_{0},v_{0}\rangle  . \]
623: Note that $\Phi(g)$ is just the upper left entry $g_{11}$ in the matrix
624: $(g_{ij})_{1\leq i,j \leq n+1} $. \\
625: Now let $\ x \in G\exp(i\Omega) \cap K_{\mathbb C}A_{\mathbb C}K_{\mathbb C}
626: \ $. W.l.o.g. we can write 
627: \[ x=na\exp iX = \tilde{k}_{1} \tilde{a} \tilde{k}_{2} , \]
628: for some
629: \begin{eqnarray*}
630: n  & = & \left( \begin{array}{ccc} 1+\frac{1}{2}\abs{\abs{v}}^{2} & v & 
631: -{1\over 2}\abs{\abs{v}}^{2} \\ v^{t} & \mbox{Id}_{n-1} & -v^{t} \\
632: \frac{1}{2}\abs{\abs{v}}^{2} & v & 1-\frac{1}{2}\abs{\abs{v}}^{2} 
633: \end{array} \right) \in N ,  \\
634: a\exp iX   & = &   \left( \begin{array}{ccc} \cosh z & 0 & \sinh z \\ 0 & 
635: \mbox{Id}_{n-1} & 0 \\ \sinh z & 0 & \cosh z \end{array} \right) \in A
636: \exp (i\Omega) \\
637: \end{eqnarray*}
638: and 
639: $$\tilde{a}= \left( \begin{array}{ccc} \cosh \omega & 0 & \sinh \omega \\ 0 & 
640: \mbox{Id}_{n-1} & 0 \\ \sinh \omega & 0 & \cosh \omega \end{array} \right)  
641: \in A_{\mathbb C} , \quad
642: \tilde{k}_{1,2}=\left( \begin{array}{cc} 1 & 0 \\ 0 & B_{1,2} \end{array} 
643: \right) \ ( \ B_{1,2} \in SO(n,\mathbb{C}) \ ) \ .$$
644: Applying $\Phi $ yields
645: \[ \Phi(x)=(1+\frac{1}{2}\abs{\abs{v}}^{2}) \cosh z - 
646: \frac{1}{2}\abs{\abs{v}}^{2} \sinh z = \cosh \omega , \]
647: or
648: \begin{equation}\label{SO(1,n)}
649: e^{z}+ (1+\abs{\abs{v}}^{2})e^{-z} = e^{\omega}+e^{-\omega} . 
650: \end{equation}
651: Equation (\ref{SO(1,n)}), Figure 1 (with a slight modification in the 
652: notation) and Lemma \ref{convlem}  show again that $\ \tilde{a} \in A\exp(i{\rm conv}(\mathcal{W}.X))
653: K_{\mathbb C} \ $, as claimed. 
654: \end{example}
655: 
656: \section{Applications to spherical functions}
657: 
658: Let us briefly introduce the necessary notation. Write $G=NAK$ for the Iwasawa
659: decomposition of $G=SL(2,{\mathbb R})$ with $N$, $A$ and $K$ as in Section 3. 
660: For $g\in G$ we define $a(g)\in A$ by $g\in Na(g)K$. For $\lambda\in {\mathfrak a}_{\mathbb C}^*$
661: and $a\in A$ we set $a^\lambda=e^{\lambda(\log a)}$. 
662: \par Following Harish-Chandra we define the {\it spherical function on $G/K$ with parameter 
663: $\lambda\in {\mathfrak a}_{\mathbb C}^*$} by 
664: 
665: $$\phi_\lambda(gK)=\int_K a(kg)^{\rho-\lambda}\ dk \qquad (g\in G), $$
666: where $dk$ denotes the normalized Haar-measure on $K$ and $\rho={1\over 2}\alpha$
667: with $\Sigma^+=\{\alpha\}$. 
668: \par Define 
669: $$\Pi=\{\lambda\in \mathfrak{a}_{\mathbb C}^*:\  \phi_\lambda \ \mbox{is positive definite}\}$$
670: and recall that $i{\mathfrak a}^*\subseteq \Pi$. 
671: \par We have to recall some facts from \cite{KS1} \S4  on 
672: the analytic continuation of the spherical functions (actually 
673: valid for all semisimple Lie groups $G$).  
674: 
675: 
676: \begin{prop}\label{KS}
677:  Let $\lambda\in {\mathfrak a}_{\mathbb C}^*$. Then the following assertions hold:
678: \begin{enumerate}
679: \item  The spherical function $\phi_\lambda$ has a unique holomorphic extension 
680: to $\Xi$. 
681: \item The spherical function $\phi_\lambda$ has a unique continuation 
682: to a $K_{\mathbb C}$-bi-invariant function on $K_{\mathbb C} A\exp(2i\Omega)K_{\mathbb C}/ K_{\mathbb C}$
683: such that the restriction to $A\exp(2i\Omega)$ is holomorphic. 
684: Moreover, if $\lambda\in \Pi $ and $\Omega_c\subseteq\Omega$ is 
685: a compact subset, then the restriction of $\phi_\lambda$ to 
686: $K_{\mathbb C} A\exp(2i\Omega_c)K_{\mathbb C}/ K_{\mathbb C}$ is bounded. 
687: \end{enumerate}
688: \end{prop}
689: 
690: In the sequel we also denote by $\phi_\lambda$ the holomorphic extension of $\phi_\lambda$
691: to $\Xi$. We denote by $Z=\{\pm{\bf 1}\}$ the center of $G$. 
692: Our result in this section then is:
693: 
694: \begin{theorem} \label{sph} 
695: Let $\lambda\in \Pi$. Then the spherical function $\phi_\lambda$ is 
696: bounded on $\Xi$, i.e., 
697: $$\|\phi_\lambda\|_{\infty, \Xi}<\infty\ .$$
698: \end{theorem}
699: 
700: \begin{proof} By Theorem \ref{conv} we have 
701: $$\Xi\bigcap K_{\mathbb C} A_{\mathbb C} K_{\mathbb C}/ K_{\mathbb C} \subseteq K_{\mathbb C} A\exp(i\Omega)K_{\mathbb C}/ K_{\mathbb C}\ .$$
702: Thus Proposition \ref{KS} implies that $\phi_\lambda$ is bounded 
703: on $\Xi\bigcap K_{\mathbb C} A_{\mathbb C} K_{\mathbb C}/ K_{\mathbb C}$. 
704: \par Let now $z\in \Xi$ such that $z\not\in K_{\mathbb C} A_{\mathbb C} K_{\mathbb C}/ K_{\mathbb C}$. 
705: By the density of $K_{\mathbb C} A_{\mathbb C} K_{\mathbb C}$ in $G_{\mathbb C}$
706: we can find sequences $(k_n)\subseteq K_{\mathbb C}$, 
707: $(a_n)\subseteq A_{\mathbb C}$ such that $k_n a_n K_{\mathbb C} \in \Xi$ for all
708: $n\in {\mathbb N}$ and $z=\lim_{n\to\infty} k_n a_n K_{\mathbb C}$. By Theorem \ref{conv} 
709: we have $a_n\in ZA\exp(i\Omega)$ and thus 
710: $$\lim_{n\to \infty} a_n=\pm {\bf 1}$$
711: by Proposition \ref{limits}.  By the continuity of $\phi_\lambda$ we hence obtain 
712: that 
713: $$\phi_\lambda(z)=\lim_{n\to \infty}\phi_\lambda(k_n a_n K_{\mathbb C})
714: =\lim_{n\to \infty}\phi_\lambda(a_n K_{\mathbb C})=\phi_\lambda(\pm {\bf 1})=1, $$
715: concluding the proof of the theorem.
716: \end{proof}
717: 
718: \begin{remark} 
719: (a) Note that Proposition \ref{KS} holds for all 
720: semisimple Lie groups. Thus in order to generalize Theorem \ref{sph} 
721: to all semisimple Lie groups one needs Proposition 
722: \ref{limits} and Theorem \ref{conv} in full generality (cf. our conjecture 
723: in Section 4). 
724: 
725: \par\noindent  (b) Fix a $\lambda\in {\mathfrak a}_{\mathbb C}^*$  
726: and consider the following function on $G/K\times K$
727: $${\mathcal P}_\lambda(gK, k)\mapsto a(kg)^{\rho-\lambda}\ .$$
728: Identifying $G/K$ with the unit disc  $D$ and $K$ with its boundary, this 
729: function is easily seen to be a power of the Poisson kernel on $D$. According 
730: to \cite{KS1} the function ${\mathcal P}_\lambda$ admits an analytic continuation to 
731: $\Xi\times K$ which is holomorphic in the first variable. 
732: Also by \cite{KS1}  we know that this function is unbounded for $\lambda\in \Pi\setminus\{\rho, -\rho\}$. 
733: Observe 
734: that the spherical function on $\Xi$ is given by 
735: $$\phi_\lambda(z)=\int_K {\mathcal P}_\lambda(z, k) \ dk \qquad (z\in \Xi)\ .$$
736: Now notice that Theorem \ref{sph} says that $\phi_\lambda$ stays 
737: bounded despite the fact that the integrands become singular 
738: towards the boundary of $\Xi$.
739: \end{remark}
740: 
741: 
742: 
743: 
744: \section{A norm estimate for the middle projection}
745:  
746: 
747: In this section we will investigate the growth of the 
748: middle projection in the $K_{\mathbb C} A_{\mathbb C} K_{\mathbb C}$-decomposition 
749: for elements in $\Xi$. 
750: 
751: \par To start with let us introduce a {\it norm} on $G=SL(2,{\mathbb R})$ by 
752: setting 
753: $$\|g\|=\mbox{tr}(gg^t)\qquad (g\in G)\ . $$
754: 
755: \par Define a compact (here actually finite)
756: group 
757: $$L=N_{K_{\mathbb C}}(A_{\mathbb C}) \times Z= N_K(A)\times Z\ .$$
758: We let $L$ act on $A_{\mathbb C}$ via
759: $$(k,z)a=kak^{-1}z\qquad (k,z)\in L,\  a\in A_{\mathbb C}\ .$$
760: Then for every $x\in K_{\mathbb C} A_{\mathbb C} K_{\mathbb C}$ we 
761: have $x\in K_{\mathbb C} a_{K_{\mathbb C}}(x) K_{\mathbb C}$ with a unique continuous function 
762: 
763: $$a_{K_{\mathbb C}}: K_{\mathbb C} A_{\mathbb C} K_{\mathbb C} \to A_{\mathbb C}/ L, \ \  x\mapsto a_{K_{\mathbb C}}(x)\ . $$
764: In view of Theorem \ref{sph} and its proof we also obtain 
765: a holomorphic function 
766: 
767: \[ a_{K_{\mathbb C}}:\Xi \to A_{\mathbb C}/ L, \ \  x\mapsto 
768: \left\{ \begin{array}{rl}
769: a_{K_{\mathbb C}}(x) & \mbox{if}\  x\in K_{\mathbb C} A_{\mathbb C} K_{\mathbb C} / K_{\mathbb C},  \\
770: {\bf 1} & \mbox{if}\  x\not\in  K_{\mathbb C} A_{\mathbb C} K_{\mathbb C} / K_{\mathbb C}\ .\\
771: \end{array}\right. \]
772: 
773: Finally we define a {\it norm} on $A_{\mathbb C}$ by setting 
774: $$|\cdot|: A_{\mathbb C} \mapsto {\mathbb R}^+, \ \ a=\left(\begin{array}{cc} z & 0 \\ 0 & z^{-1} \end{array} \right)
775: \mapsto |z|^2 +|z|^{-2}\ . $$
776: Notice that $|\cdot |$ is $L$-invariant hence factors to 
777: a continuous positive function on $A_{\mathbb C}/ L$ also denoted by $|\cdot|$. 
778: 
779: 
780: \begin{prop}\label{norm}
781: For all $g=h\exp(iX)K_{\mathbb C} \in \Xi$, ($h\in G, \ X\in \Omega$),  we have 
782: $$|a_{K_{\mathbb C}}(g)|\leq \|h\|\ .$$
783: \end{prop}
784: 
785: \begin{proof}
786: Using $G=KAK$ we can write 
787: \[\begin{array}{cccc}
788: g& = & \underbrace{\left( \begin{array}{cc} \cos r & \sin r \\ 
789: -\sin r & \cos r \end{array} \right)\left( \begin{array}{cc} 
790: s & 0 \\ 0 & s^{-1} \end{array} 
791: \right) \left( \begin{array}{cc} \cos t & \sin t \\ 
792: -\sin t & \cos t \end{array} \right) } & \underbrace{\left( 
793: \begin{array}{cc} 
794: e^{i\theta} & 0 \\ 0 & e^{-i\theta} \end{array} \right) } K_{\mathbb C} \\
795: & & =h \in KAK & \in \exp (i\Omega)
796: \end{array} \]
797: We may assume that $g\in K_{\mathbb C} A_{\mathbb C} K_{\mathbb C}/ K_{\mathbb C}$ and write 
798: $a_{K_{\mathbb C}}(g)=\left(\begin{array}{cc} z & 0 \\ 0 & z^{-1} \end{array} \right)$. 
799: Then 
800: \begin{eqnarray}
801: \Phi (g) & = & (s^{2}\cos^{2}t+s^{-2}\sin^{2}t)e^{2i\theta} + 
802: (s^{2}\sin^{2}t+s^{-2}\cos^{2}t)e^{-2i\theta} \nonumber \\
803: & = & \cos^{2}t(s^{2}e^{2i\theta}+s^{-2}e^{-2i\theta}) +
804: \sin^{2}t(s^{-2}e^{2i\theta}+s^{2}e^{-2i\theta}) \nonumber \\
805: & = & z^{2}+z^{-2} . 
806: \end{eqnarray}
807: Hence $\ z^{2}+z^{-2} \ $ is a convex combination of 
808: $\ s^{2}e^{2i\theta}+s^{-2}e^{-2i\theta} \ $ and
809: $\ s^{-2}e^{2i\theta}+s^{2}e^{-2i\theta} \ $  
810: and therefore lies on the vertical segment shown in Figure 2.
811: \\
812: \\
813: 
814: \epsfbox{febfig2.eps}
815: 
816: 
817: The following Lemma \ref{normlem} applied to $u=s^2e^{2i\theta}$ and $v=z^2$ completes the proof
818: of the theorem. 
819: \end{proof}
820: 
821: 
822: \begin{lemma}\label{normlem} Let $r_1, r_2>0$ and $-{\pi\over 2}<\varphi_1, \varphi_2<{\pi\over 2}$. 
823: Define complex numbers by $u=r_1 e^{i\varphi_{1}}$ and $v=r_2 e^{i\varphi_{2}}$. 
824: Assume that 
825: $\ \Re(v+v^{-1}) = \Re(u+u^{-1}) \ $ and $\ \Im(v+v^{-1}) = \mu \Im(u+u^{-1})
826:  \ $ for some $\mu \in \mathbb{R}, \abs{\mu} \leq 1 $. 
827: Then $\quad \abs{\varphi_{2}} \leq \abs{\varphi_{1}} \ $ and $r_2+r_2^{-1}\leq r_1+r_1^{-1}$. 
828: \end{lemma}
829: \begin{proof}
830: The assumption written out in polar coordinates followed by squaring yields 
831: \begin{equation} \label{hirsch}
832: (r_{1}+r_{1}^{-1})^{2}\cos^{2}{\varphi_{1}}  =  (r_{2}+r_{2}^{-1})^{2}\cos^{2}{\varphi_{2}} \ ,
833: \end{equation}
834: \begin{equation*}
835: \mu^{2} (r_{1}-r_{1}^{-1})^{2}\sin^{2}{\varphi_{1}}  =  (r_{2}-r_{2}^{-1})^{2}\sin^{2}{\varphi_{2}} \ .
836: \end{equation*}
837: 
838: If $\abs{\varphi_{2}} > \abs{\varphi_{1}}$, then (\ref{hirsch}) implies
839: \[ (r_{1}+r_{1}^{-1})^{2} < (r_{2}+r_{2}^{-1})^{2} \ . \]
840: Since $\ (r_{1}+r_{1}^{-1})^{2} = (r_{1}-r_{1}^{-1})^{2}+4 \ $ this means
841: \[ (r_{1}-r_{1}^{-1})^{2} < (r_{2}-r_{2}^{-1})^{2} \ . \]
842: But now we end up with
843: \[ \mu^{2}(r_{1}-r_{1}^{-1})^{2}\sin^{2}{\varphi_{1}} < (r_{2}-r_{2}^{-1})^{2}\sin^{2}{\varphi_{2}} \ , \]
844: which contradicts the assumption. Thus $ \abs{\varphi_{2}} \leq \abs{\varphi_{1}}$ proving 
845: our first assertion. The second assertion is now immediate from (\ref{hirsch}).
846: 
847: \end{proof}
848: 
849: 
850: 
851: 
852: 
853: \bibliographystyle{amsplain}
854: \begin{thebibliography}{10}
855: 
856: \bibitem {AG} Akhiezer, D. and S. Gindikin, 
857: \textit{ On Stein extensions of
858: real symmetric spaces}, Math.\ Ann.\ {\bf 286}, 1--12, 1990
859: 
860: \bibitem {BA} Barchini, L., {\it Stein extensions of real symmetric spaces and the 
861: geometry of the flag manifold}, Math. Ann., to appear 
862: 
863: \bibitem{FK} Faraut, J. and T. Kobayashi, uncirculated notes and private communication 
864: 
865: \bibitem{FH} Fels, G. and A. Huckleberry, {\it Characterization of  cycle domains via
866: Kobayashi hyperbolicity}, arXiv:math.AG/0204341
867: 
868: \bibitem {GK1} Gindikin, S., and B.\ Kr\"otz, {\it Complex crowns of Riemannian
869: symmetric spaces and non-compactly causal symmetric spaces}, Trans. Amer. 
870: Math. Soc. {\bf 354(8)}, 3299--3327, 2002
871: 
872: \bibitem {GK2} Gindikin, S., and B.\ Kr\"otz, {\it Invariant Stein domains in Stein symmetric spaces and a
873: non-linear complex convexity theorem}, IMRN {\bf 18} (2002), 959-971
874: 
875: \bibitem {GKO} Gindikin, S., B.\ Kr\"otz, and G.\ \'Olafsson, {\it Hardy spaces for
876: non-compactly causal symmetric spaces and the most continuous spectrum}, MSRI
877: preprint 2001-043, Math. Ann., to appear 
878: 
879: \bibitem {GKO2} Gindikin, S., B.\ Kr\"otz, and G.\ \'Olafsson, {\it Hardy spaces for
880: NCC symmetric spaces}, preprint
881: 
882: 
883: \bibitem {GM1} Gindikin, S. and T. Matsuki, {\it Stein Extensions of Riemann symmetric 
884: spaces and dualities of orbits on flag manifolds}, Transformation groups, to appear  
885: 
886: \bibitem{GM2} Gindikin, S. and T. Matsuki, {\it A remark on Schubert cells and duality 
887: of orbits on flag manifolds}, arXiv:math.RT/0208071
888: 
889: \bibitem{HU} Huckleberry, A., {\it On certain domains in cycle spaces of flag manifolds}, 
890: Math. Ann., to appear 
891: 
892: \bibitem{HW} Huckleberry, A. and J.A. Wolf, {\it Schubert Varieties and Cycle Spaces}, 
893: arXiv:math.AG/0204033 
894: 
895: \bibitem {KS1} Kr\"otz, B., and R.J.  Stanton, {\it Holomorphic extensions of
896: representations:  (I) automorphic functions}, Annals of Math., to appear 
897: 
898: \bibitem {KS2} Kr\"otz, B., and R.J.  Stanton, {\it Holomorphic extensions of
899: representations:  (II) geometry and harmonic analysis}, preprint
900: 
901: \bibitem {KS3} Kr\"otz, B., and R.J.  Stanton, {\it Complex crowns of Riemannian 
902: symmetric spaces -- exercises for the exceptional cases}, in preparation
903: 
904: \bibitem{MA1} Matsuki, T., {\it Double coset decompositions of algebraic groups arising from 
905: two involutions. I},  J. Algebra {\bf 175} (1995), no. {\bf 3}, 865--925
906: 
907: \bibitem{MA2} Matsuki, T., {\it Stein extensions of Riemann symmetric spaces and its generalization}, arXiv:math.RT/0208175,
908: J. Lie Theory, to appear 
909: 
910: \end{thebibliography}
911:  
912: 
913: 
914: 
915: 
916: 
917: 
918: \end{document}
919: