1: \documentclass[12pt]{amsart}
2: \newif\ifpdf\ifx\pdfoutput\undefined\pdffalse\else\pdfoutput=1\pdftrue\fi
3: \newcommand{\pdfgraphics}{\ifpdf\DeclareGraphicsExtensions{.pdf,.jpg}\else\fi}
4: \usepackage{graphicx,amsmath,amsfonts,latexsym,amssymb,amsthm,amscd}
5: \usepackage{latexsym}
6:
7: \pagestyle{plain}
8: \makeatletter
9: \makeatother
10:
11: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
12:
13:
14:
15: \newcommand{\emb}{\hookrightarrow}
16:
17: \newcommand{\ca}[1]{\mathcal{#1}}
18: \newcommand{\frk}[1]{\mathfrak{#1}}
19:
20: \newcommand{\field}[1]{\mathbb{#1}}
21: \newcommand{\rz}{\field{R}}
22: \newcommand{\cz}{\field{C}}
23: \newcommand{\zz}{\field{Z}}
24: \newcommand{\nz}{\field{N}}
25:
26: \newcommand{\I}{\mathrm{i}}
27:
28: \newcommand{\fr}{\frac}
29: \newcommand{\pax}{\partial_x}
30:
31: \newcommand{\la}{ \lambda }
32: \newcommand{\ep}{ \epsilon }
33: \newcommand{\om}{ \omega }
34: \newcommand{\et}{ \eta }
35: \newcommand{\vp}{\varphi }
36:
37: \newcommand{\ti}{\tilde }
38: \newcommand{\eq}{\Leftrightarrow}
39: \newcommand{\imp}{\Rightarrow}
40: \newcommand{\lr}{\Leftrightarrow}
41: \newcommand{\spc}{ $ \textrm{Spin}^\cz $ }
42: \newcommand{\cp}{\field{C}\field{P}^2}
43: \newcommand{\U}{\mathcal{U}}
44: \newcommand{\D}{\mathcal{D}}
45: \newcommand{\Hi}{\mathcal{H}}
46: \newcommand{\dom}{\mathrm{dom}}
47: \newcommand{\ind}{\mathrm{ind}}
48: \newcommand{\supp}{\mathrm{supp}}
49: \newcommand{\spec}{\mathrm{spec}}
50: \newcommand{\res}{\mathrm{Res}}
51: \newcommand{\Rg}{\mathrm{Rg}}
52:
53: \newcommand{\lieg}{\mathfrak{g}}
54: \newcommand{\liek}{\mathfrak{k}}
55: \newcommand{\liea}{\mathfrak{a}}
56: \newcommand{\lieh}{\mathfrak{h}}
57: \newcommand{\liep}{\mathfrak{p}}
58: \newcommand{\liem}{\mathfrak{m}}
59: \newcommand{\lien}{\mathfrak{n}}
60: \newcommand{\liegc}{\mathfrak{g}_\cz}
61: \newcommand{\liekc}{\mathfrak{k}_\cz}
62: \newcommand{\lieac}{\mathfrak{a}_\cz}
63: \newcommand{\Ad}{\mathrm{Ad}}
64: \newcommand{\killing}{\kappa}
65: \newcommand{\hc}{\mathbf{c}}
66:
67:
68: \newcommand{\id}{1 \hspace{-1mm} \mbox{\sf I}}
69:
70: %%%%%%%%%%%%%%%%%%%% Environments %%%%%%%%%%%%%%%%%%%%%%%
71:
72: \newtheorem{theorem}{Theorem}[section]
73: \newtheorem{definition}[theorem]{Definition}
74: \newtheorem{exa}[theorem]{Example}
75: \newtheorem{lem}[theorem]{Lemma}
76: \newtheorem{cor}[theorem]{Corollary}
77: \newtheorem{pro}[theorem]{Proposition}
78: \newtheorem{hyp}[theorem]{Hypothesis}
79: \newtheorem{rem}[theorem]{Remark}
80:
81:
82: %\setlength{\parindent}{0pt} \setlength{\parskip}{5pt plus 2pt minus 1pt}
83:
84: % paper formats
85:
86: \linespread{1.3}
87: \addtolength{\oddsidemargin}{-15mm}
88: \addtolength{\evensidemargin}{-15mm}
89: \addtolength{\textwidth}{30mm}
90: \setlength{\topmargin}{0mm}
91: \setlength{\headsep}{0pt}
92: \addtolength{\textheight}{33mm}
93: \addtolength{\footskip}{5mm}
94:
95: %\numberwithin?{equation}{section}
96:
97: \DeclareMathOperator{\tr}{tr}
98:
99:
100: %\allowdisplaybreaks
101:
102: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
103:
104: \title{Analytic Continuation of Resolvent Kernels on noncompact Symmetric Spaces}
105:
106: \author{Alexander Strohmaier}
107:
108: %\date{\today}
109:
110:
111: \begin{document}
112:
113: \maketitle
114:
115:
116: \begin{abstract}
117: \noindent
118: Let $X=G/K$ be a symmetric space of noncompact type and let $\Delta$ be the
119: Laplacian associated with a $G$-invariant metric on $X$. We show that the
120: resolvent kernel of $\Delta$ admits a holomorphic extension to a Riemann
121: surface depending on the rank of the symmetric space. This Riemann
122: surface is a branched cover of the complex plane with a certain part of the
123: real axis removed. It has a branching point at the bottom of the spectrum
124: of $\Delta$. It is further shown that this branching point is quadratic if
125: the rank of $X$ is odd, and is logarithmic otherwise. In case $G$ has only
126: one conjugacy class of Cartan subalgebras the resolvent kernel extends to a
127: holomorphic function on a branched cover of $\cz$ with the only branching
128: point being the bottom of the spectrum.
129: \end{abstract}
130:
131: %{\em Keywords:} \\
132: {\small \bf Mathematics Subject Classification (2000):} 58J50 (11F72) \\
133:
134: %\newpage
135:
136: \section{Introduction}
137:
138: If $(M,g)$ is a complete Riemannian manifold the metric Laplace operator
139: $\Delta_g$ is a selfadjoint operator in the Hilbert space $L^2(M)$. Its
140: resolvent
141: \begin{gather}
142: R_z(\Delta_g)=(\Delta_g -z)^{-1}
143: \end{gather}
144: is a holomorphic function on $\cz \backslash \textrm{spec}(\Delta_g)$ with
145: values in the bounded operators in $L^2(M)$. Whereas $\cz \backslash
146: \textrm{spec}(\Delta_g)$ is the largest domain where all matrix elements of
147: the resolvent
148: \begin{gather*}
149: \langle \phi, R_z(\Delta_g) \psi \rangle \quad \phi,\psi \in L^2(M)
150: \end{gather*}
151: are holomorphic functions it may happen that for all $\psi$ and $\phi$ in some
152: dense subset of $L^2(M)$ those matrix elements can be continued to holomorphic
153: functions on a Riemann surface extending the resolvent set $\cz \backslash
154: \textrm{spec}(\Delta_g)$.
155:
156: A lot of examples where such a situation occurs are known. If $(M,g)$ is the
157: Euclidean space of dimension $n$, then the functions
158: \begin{gather*}
159: \langle \phi, R_z(\Delta_g) \psi \rangle \quad \phi,\psi \in
160: L^2(\rz^n,e^{+\epsilon x^2} dx), \quad \epsilon >0
161: \end{gather*}
162: extend to holomorphic functions on the concrete Riemann surface associated
163: with either the function $\sqrt{z}$ or $\log(z)$, depending on whether the
164: dimension $n$ is odd or even (see e.g. \cite{MR96k:35129}).
165: If a compactly supported potential is added,
166: the same holds, except that in this case the continuation is meromorphic.
167: Poles of these matrix elements are commonly referred to as scattering poles or
168: resonances. The theory that provides analytic continuation in the case of the
169: Laplacian in $\rz^n$ with potential is well developed and plays a crucial role
170: for example in proving the absence of singular continuous spectrum (see e.g.
171: \cite{CFKS} and references therein). Also in
172: order to make sense of the notion of resonances one needs to proof the
173: existence of a meromorphic continuation across the spectrum. From the results
174: known for the Euclidean space one may obtain further examples employing
175: perturbation techniques (e.g. \cite{MR96j:35183}).
176:
177: Here we will be mainly interested in a continuation of the resolvent kernel,
178: i.e. we view $R_z(\Delta_g)$ as a function with values in $\mathcal{D}'(M
179: \times M)$. Hence, we look at the matrix elements $\langle \phi, R_z(\Delta_g)
180: \psi \rangle$ where $\phi,\psi \in C^\infty_0(M)$.
181:
182: Examples of manifolds where a meromorphic continuation of the resolvent kernel
183: is known to exist include manifolds with cylindrical ends
184: (\cite{MR96k:35129,MR98f:11053,MR96g:58180}),
185: certain strictly pseudoconvex domains (\cite{MR92i:32016}),
186: manifolds with asymptotically constant negative curvature
187: (\cite{MR89c:58133,MR96h:58172}), and noncompact symmetric spaces of rank $1$
188: (\cite{MR93i:58160,CarrPed:03}). Concerning this question little was known
189: however about higher rank symmetric spaces of noncompact type.
190: Only recently the method of complex scaling
191: was used by Mazzeo and Vasy (\cite{MazVas:02,MazVas:03}) to show that
192: the resolvent kernel has a meromorphic extension across the spectrum to a
193: region which is obtained by rotating the spectrum up to the angle $\pi/2$ and
194: by adding further branching points.
195: It was conjectured however in their papers that this continuation has no poles
196: and that the surface can be chosen without branching points.
197: A special case which can be considered as well understood is that
198: of a symmetric space $X=G/K$, where $G$ is a complex semisimple Lie group.
199: In this case there is an explicit
200: formula for the heat kernel (see \cite{MR39:360}). By Laplace transform
201: one can obtain an explicit expression for the resolvent kernel in terms of
202: the modified Bessel function of the second kind.
203: In this way one finds that the resolvent kernel admits a holomorphic extension
204: to a branched cover of the complex plane.
205:
206: We will show that in the case of a noncompact Riemannian symmetric space
207: $X=G/K$ of arbitrary rank the resolvent kernel admits a holomorphic extension
208: to a Riemann surface depending on the rank of the symmetric space. This
209: Riemann surface is a branched cover of the complex plane with a part of the
210: real axis removed. A branching point occurs at the bottom of the spectrum.
211: This branching point turns out to be quadratic if the rank of the space is odd
212: and logarithmic otherwise. This Riemann surface contains the region
213: obtained by rotating the spectrum up to the angle $\pi$, and our result
214: implies the absence of poles in that region.
215: In case the rank of the space is one we reproduce the known
216: results, i.e. the resolvent kernel extends to a meromorphic function on the
217: branched cover of $\cz$ associated with the function $\sqrt{z-\mu}$,
218: where $\mu$ is the bottom of the spectrum. We get even
219: stronger results in case the group $G$ has only one conjugacy class of Cartan
220: subalgebras. In this case the resolvent kernel behaves almost like in
221: Euclidean space, i.e. there exists a holomorphic continuation to a Riemann
222: surface which is a branched cover of $\cz$ with the only branching point
223: being the bottom of the spectrum.
224: The branching point is quadratic or logarithmic depending on
225: whether the rank of the space is odd or even. This special case includes the
226: real hyperbolic spaces of odd dimension, the spaces $SU^*(2n)/Sp(n)$,
227: $E_{6(-26)}/F_4$ as well as the case $G/K$ when $G$ is a complex group.
228:
229: Our method relies on the Fourier transform on symmetric spaces of noncompact
230: type and on the meromorphic continuation of the Harish-Chandra $\hc$-function.
231:
232: \section{Notations and Background material}
233:
234: \subsection{Symmetric spaces of noncompact type}
235: Suppose that $X=G/K$ is a symmetric space of noncompact type, that is $G$ is a
236: real connected noncompact semisimple Lie group with finite center and $K
237: \subset G$ is a maximal compact subgroup. Let $\lieg$ be the Lie algebra of $G$ and let
238: $\liek \subset \lieg$ be the Lie algebra of $K$. We denote by $\killing(\cdot,\cdot)$
239: the Killing form on $\lieg$, i.e.
240: \begin{gather}
241: \killing(X,Y)=\mathrm{Tr}(\Ad(X) \circ \Ad(Y)).
242: \end{gather}
243: Then there exists a Cartan involution $\theta: \lieg \to \lieg$ with fixed
244: point algebra $\liek$, i.e. $\theta$ is an involutive automorphism such that
245: the bilinear form $\langle X,Y \rangle:=-\killing(X,\theta Y)$ is positive definite
246: and such that the $+1$ eigenspace of $\theta$ coincides with $\liek$. Hence,
247: the decomposition of $\lieg$ into $+1$ and $-1$ eigenspaces of $\theta$ reads
248: \begin{gather}
249: \lieg = \liek \oplus \liep,
250: \end{gather}
251: where $\liep \subset \lieg$ is a linear subspace. If $\liegc$ is the
252: complexification of $\lieg$ then
253: \begin{gather}
254: \lieg_c=\liek + \I \liep \subset \liegc
255: \end{gather}
256: is a compact real form of $\liegc$. Now let $\liea$ be a maximal abelian
257: subspace of $\liep$ and let $\liem$ be the centralizer of $\liea$ in $\liek$.
258: If $\mathfrak{b}$ is a maximal abelian subalgebra of $\liem$ then $\lieh =
259: \mathfrak{b} + \liea$ is a $\theta$-stable Cartan subalgebra of $\lieg$. The
260: set of restricted roots $\Delta(\lieg,\liea)$ w.r.t. $\liea$ is the set of
261: nonzero linear functionals $\alpha \in \liea^*$ such that
262: \begin{gather}
263: [x,y]=\alpha(x) y, \quad \forall x \in \liea
264: \end{gather}
265: for some nonzero element $y \in \lieg$. The multiplicity $m_\alpha$ of a
266: restricted root is the dimension of the vector space $\{y \in \lieg;\;
267: [x,y]=\alpha(x) y, \quad \forall x \in \liea\}$. This gives rise to a root
268: space decomposition of $\lieg$:
269: \begin{gather}
270: \lieg = \liea \oplus \liem \oplus \sum_{\alpha \in \Delta(\lieg,\liea)}
271: \lieg_\alpha,
272: \end{gather}
273: where $\lieg_\alpha$ are the root subspaces, i.e.
274: \begin{gather}
275: \lieg_\alpha=\{y \in \lieg;\; [x,y]= \alpha(x) y, \quad \forall x \in
276: \liea\}.
277: \end{gather}
278: Note that each restricted root $\Delta(\lieg,\liea)$ coincides with the restriction
279: of a root $\alpha \in \Delta(\lieg,\lieh)$ to $\liea$. We may choose now a subsystem of positive roots
280: $\Delta^+(\lieg,\lieh)$ and accordingly a subsystem of positive restricted
281: roots $\Delta^+(\lieg,\liea)$. As usual we use the notation
282: \begin{gather}
283: \rho=\frac{1}{2} \sum_{\alpha \in \Delta^+(\lieg,\liea)} m_\alpha \alpha
284: \end{gather}
285: for half of the sum of positive roots counting multiplicities. Let $\lien
286: \subset \lieg$ be the nilpotent Lie-subalgebra
287: \begin{gather}
288: \lien=\sum_{\alpha \in \Delta^+(\lieg,\liea)} g_\alpha.
289: \end{gather}
290: Then the Iwasawa decomposition of $\lieg$ reads as follows:
291: \begin{gather}
292: \lieg = \liek \oplus \liea \oplus \lien.
293: \end{gather}
294:
295: Now let $A$ and $N$ be the analytic subgroups of $G$ with Lie algebras $\liea$
296: and $\lien$ respectively. Then
297: \begin{gather}
298: N \times A \times K \to G, \; (n,a,k) \to nak
299: \end{gather}
300: is a diffeomorphism from $N \times A \times K$ to $G$. This is the
301: classical Iwasawa decomposition of $G$. For $g \in G$ we denote by $A(g) \in
302: \liea$ the unique element such that $g$ can be expressed as
303: \begin{gather}
304: g = n \exp(A(g)) k
305: \end{gather}
306: with $n \in N$ and $k \in K$.
307: Let $M$ be the centralizer of $A$ in $K$ and denote by $B$ be the compact homogeneous
308: space $K/M$. Then $A(gK,kM):=A(k^{-1} g)$ defines a smooth function on $X
309: \times B$ with values in $\liea$.
310: If $M'$ is the normalizer of $A$ in $K$ then the restricted Weyl group is
311: defined by $W:=M'/M$. The restricted Weyl group acts on $\liea$ by
312: $kM a = \mathrm{Ad}(k) a$ and by duality it also acts on the dual $\liea^*$ of
313: $\liea$. This representation of $W$ on $\liea$ is injective and $W$ can be
314: identified in this way with a group of reflections in $\liea$.
315: As usual let $\liea_+^*$ be the positive Weyl chamber
316: $\{\lambda \in \liea^*;\; \lambda(\alpha)>0 \quad \forall \alpha \in \Delta^+(\lieg,\liea)\}$.
317: Then $\liea_+^*$ is a fundamental domain for the action of $W$ on $\liea^*$.
318:
319: \subsection{The Helgason transform and the Paley-Wiener theorem}
320:
321: Let $X=G/K$ be a symmetric space of noncompact type. We use the notations and
322: conventions from above. The generalized Fourier transform (Helgason
323: transform) of a function $f \in C^\infty_0(X)$ is the function $\hat f :
324: \liea^* \times B \to \cz$ defined by
325: \begin{gather}
326: \hat f(\lambda,b)=\int_X f(x) e^{(-\I \lambda + \rho)(A(x,b))} dx,
327: \end{gather}
328: where we integrate with respect to some invariant measure on $X$ which we
329: choose to be normalized as in \cite{MR41:8587}
330: \footnote{the normalization of the measure is not important for our
331: considerations}.
332: For suitable normalizations of the Euclidean measure on $\liea^*$ and the invariant
333: measure $db$ on $B$ the inverse Fourier transform (see \cite{MR41:8587}) is given by
334: \begin{gather}
335: f(x)= w^{-1} \int_{\liea^* \times B}
336: \hat f(\lambda,b) e^{(+\I \lambda + \rho)(A(x,b))} \frac{d
337: \lambda}{|\hc(\lambda)|^2} db,
338: \end{gather}
339: where $\hc(\lambda)$ is the Harish-Chandra $\hc$-function, and $w$ is the order of
340: the restricted Weyl group. Moreover, for $f,g \in C^\infty_0(X)$ we have
341: \begin{gather}
342: \int_X \overline f(x) g(x) dx = w^{-1} \int_{\liea^* \times B}
343: \overline{\hat f}(\lambda,b) \hat
344: g(\lambda,b) \frac{d \lambda}{|\hc(\lambda)|^2} db .
345: \end{gather}
346: The Fourier transform $\mathcal{F}$ extends to an
347: isometry
348: \begin{gather*}
349: L^2(X) \to L^2(\liea^*_+ \times B,\frac{d
350: \lambda}{|\hc(\lambda)|^2} db).
351: \end{gather*}
352: For $f \in C^\infty_0(X)$ the Fourier transform $\hat f$ extends to a function on
353: $\liea^*_\cz \times B$ which is entire in the first variable. Let
354: $\Hi(a^*_\cz)$ be the space of holomorphic functions of uniform exponential
355: type on $a^*_\cz \times B$,
356: i.e. the space functions $\phi$ on $a^*_\cz \times B$, entire in the
357: first variable, such that
358: there exists a constant $R>0$ with
359: \begin{gather}
360: |\phi(\lambda,b)| \leq C_N (1+|\lambda|_\killing)^{-N} e^{R |\Im(\lambda)|_\killing}
361: \end{gather}
362: for all $N \in \nz_0$ with some constants $C_N>0$. Here $| \cdot|_\killing$ denotes
363: the norm induced by the Killing form.
364: The Paley-Wiener theorem for symmetric spaces of noncompact type (see
365: \cite{MR51:3804}) is precisely the statement that the image of $C^\infty_0(X)$
366: under the Fourier transform coincides with the space of functions
367: $f \in \Hi(a^*_\cz)$ which satisfy
368: \begin{gather}
369: \int_B f(s\lambda,b) e^{(+\I s\lambda + \rho)(A(x,b))} db=
370: \int_B f(\lambda,b) e^{(+\I \lambda + \rho)(A(x,b))} db \quad \forall s
371: \in W, x \in X, \lambda \in \liea^*_\cz.
372: \end{gather}
373:
374: The $\hc$-Function satisfies $|\hc(\lambda)|^2=\hc(\lambda) \hc(-\lambda)$ and is
375: known to extend to a a meromorphic function on $\liea^*_\cz$. This follows
376: from the product formula of Gindikin and Karpelevic (\cite{MR27:240}) which
377: leads to the following explicit expression for the $\hc$-function.
378: \begin{gather} \label{GKFormula}
379: \hc(\lambda)=\hc_0 \prod_{\alpha \in \Delta(\lieg,\liea)_0^+} \frac{2^{-\langle
380: \I \lambda, \alpha_0 \rangle} \Gamma(\langle \I \lambda, \alpha_0
381: \rangle)} {\Gamma(\frac{1}{2}(\frac{1}{2} m_\alpha + 1 + \langle \I
382: \lambda, \alpha_0 \rangle)) \Gamma(\frac{1}{2}(\frac{1}{2} m_\alpha +
383: m_{2 \alpha} + \langle \I \lambda, \alpha_0 \rangle))},
384: \end{gather}
385: where $\alpha_0 = \frac{\alpha}{\langle \alpha,\alpha \rangle}$ and the
386: constant $\hc_0$ is determined by $\hc(-\I \rho)=1$. The product is over the set
387: $\Delta(\lieg,\liea)_0^+$ of indecomposable positive restricted roots, i.e.
388: \begin{gather}
389: \Delta(\lieg,\liea)_0:=\{\alpha \in \Delta(\lieg,\liea)^+;\; \frac{1}{2}
390: \alpha \notin \Delta(\lieg,\liea)\}.
391: \end{gather}
392: The scalar product $\langle \cdot,\cdot\rangle$ is the positive definite
393: scalar product on $\liea^*$ induced by the Killing form.
394:
395: \subsection{The metric Laplacian}
396:
397: As above let $X=G/K$ be a symmetric space of noncompact type and let
398: $\lieg=\liek \oplus \liep$ the corresponding Cartan decomposition. The
399: tangent space $T_{eK} X$ at the point $eK$ in $X$ is canonically identified
400: with $\liep$ and invariant metrics on $X$ are therefore in $1-1$
401: correspondence with $Ad(K)$-invariant positive definite scalar products on
402: $\liep$. Suppose now that $g$ is an invariant metric on $X$ and denote by
403: $(\cdot,\cdot)$ the corresponding scalar product on $\liep$. This induces a
404: scalar product on $\liea$ and on its dual $\liea^*$. We denote the dual
405: scalar product on $\liea^*$ again by $(\cdot,\cdot)$ \footnote{in contrast to
406: the scalar product induced by the Killing form which we denoted by $\langle
407: \cdot,\cdot\rangle$}. The Laplace operator $\Delta_g: C^\infty_0(X) \to
408: L^2(X)$ associated with the metric is a $G$-invariant differential operator
409: and hence, its spectral decomposition is strongly related to the Fourier
410: transform. For $\lambda \in \liea^*$ and $b \in B$ we define the function
411: $\phi_{\lambda,b}$ on $X$ by $\phi_{\lambda,b}(x)=e^{(-\I \lambda +
412: \rho)(A(x,b))}$ and obtain (see \cite{MR20:925} and \cite{MR51:3804} p. 458, 462)
413: \footnote{this formula appears in the literature for the case that the metric is induced
414: by the Killing form. On each irreducible factor of the symmetric space any
415: invariant metric is proportional to the one induced by the Killing form. Taking this into
416: account it is easy to see that the formula holds in the general situation.}
417: \begin{gather}
418: \Delta_g \phi_{\lambda,b} = ( |\lambda|^2 + |\rho|^2 ) \phi_{\lambda,b},
419: \end{gather}
420: where the norm $|\lambda|=( \lambda ,\lambda )^{1/2}$ is with respect to the
421: scalar product induced by the metric $g$. As a consequence if $\mathcal{F}:L^2(X) \to
422: L^2(\liea^*_+ \times B,\frac{d \lambda}{|\hc(\lambda)|^2} db)$ is the
423: Fourier transform the operator $\mathcal{F} \Delta_g \mathcal{F}^{-1}$ is a
424: multiplication operator. More precisely, if $m_{|\lambda|^2 + |\rho|^2}$ is
425: the selfadjoint operator in $L^2(\liea^*_+ \times B, \frac{d
426: \lambda}{|\hc(\lambda)|^2} db)$ defined by
427: \begin{gather} \label{multi}
428: (m_{|\lambda|^2 + |\rho|^2}f)(\lambda,k)= ( |\lambda|^2 + |\rho|^2 )
429: f(\lambda,k).
430: \end{gather}
431: with maximal domain, the operator $\mathcal{F} \Delta_g \mathcal{F}^{-1}$
432: coincides with its restriction to the set $\mathcal{F}(C^\infty_0(X))$. Therefore,
433: $m_{|\lambda|^2 + |\rho|^2}$ coincides with $\mathcal{F} \overline{\Delta_g}
434: \mathcal{F}^{-1}$, where $\overline{\Delta_g}$ denotes the closure of
435: $\Delta_g$ in $L^2(X)$. Hence, the
436: Fourier transform yields the spectral decomposition of the selfadjoint
437: operator $\overline{\Delta_g}$.
438:
439: \subsection{The resolvent of the Laplacian}
440:
441: Let $\Hi$ be a Hilbert space and let $A$ be a closed operator in $\Hi$. The
442: resolvent set of $A$ is defined to be the set of all $z \in \cz$ such
443: that the operator $A-z$ has a bounded inverse. The resolvent
444: $R_z(A):=(A-z)^{-1}$ is a holomorphic function with values in
445: $\mathcal{B}(\Hi)$ on the resolvent set $\rho(A)$ of $A$. The complement of
446: the resolvent set is called spectrum of $A$. If $A$ is selfadjoint the
447: spectrum is contained in the real line.
448:
449: We want to investigate the resolvent of the closure of the Laplace operator
450: $\Delta_g$ on a symmetric space $X=G/K$ of noncompact type with invariant
451: Riemannian metric $g$. It follows from (\ref{multi}) and the above that
452: the spectrum of
453: $\overline \Delta_g$ is exactly the set
454: \begin{gather}
455: \spec(\overline \Delta_g)=[|\rho|^2,\infty).
456: \end{gather}
457: For $f,g \in L^2(X)$ we have by the spectral theorem
458: \begin{gather}
459: \langle f, (\Delta_g-z)^{-1} g \rangle=
460: \int_{\liea^*_+ \times B} \frac{\overline{\hat
461: f(\lambda,b)}\hat g(\lambda,b)} { |\lambda|^2 + |\rho|^2 -z}\frac{d
462: \lambda}{|\hc(\lambda)|^2} db= \nonumber\\
463: =w^{-1} \int_{\liea^* \times B} \frac{\overline{\hat
464: f(\lambda,b)}\hat g(\lambda,b)} { |\lambda|^2 + |\rho|^2 -z} \frac{d
465: \lambda}{|\hc(\lambda)|^2} db\;.
466: \end{gather}
467:
468: \section{Analytic continuation of the resolvent kernel}
469:
470: Let $X=G/K$ by a symmetric space of noncompact type with invariant Riemannian
471: metric $g$. Then the resolvent of the Laplacian
472: \begin{gather}
473: R_z(\overline \Delta_g)=(\overline \Delta_g - z)^{-1}
474: \end{gather}
475: can be regarded as a map $C^\infty_0(X) \to \mathcal{D}'(X)$ and hence has a
476: distributional kernel. This resolvent kernel is a holomorphic function on the
477: set $\cz \backslash [|\rho|^2,\infty)$ with values in $\mathcal{D}'(X \times
478: X)$. Throughout this section $r$ will denote the positive real number
479: \begin{gather} \label{defr}
480: r:= \mathrm{min} \{|\alpha| j(m_\alpha) \}_{\alpha \in \Delta(\lieg,\liea)_0^+},
481: \end{gather}
482: where $j(x)=\frac{x}{2}$ if $x$ is odd and $j(x)=\frac{x}{2}+1$ if $x$ is even.
483: Note that here $|\alpha|$ is the length of $\alpha$ with respect to the metric
484: $g$ which is not necessarily induced by the Killing form. Our main result is
485: that the resolvent kernel admits an analytic continuation to a larger
486: Riemann surface.
487:
488: \begin{theorem} \label{th1}
489: Let $X=G/K$ be a symmetric space of noncompact type with invariant
490: Riemannian metric $g$. Suppose that the rank of $X$ is odd. Let $H^-:=\{z
491: \in \cz;\; \Im(z) < 0\}$ be the lower half plane. For each $f,g \in
492: C^\infty_0(X)$ define the holomorphic function
493: \begin{gather*}
494: F: H^- \to \cz,\quad z \to \langle f,\; (\Delta_g
495: -|\rho|^2 - z^2)^{-1} \;g \rangle.
496: \end{gather*}
497: Then the function $F$ admits a holomorphic continuation
498: to the open set $\cz \backslash \I [r,\infty)$.
499: If $\mathrm{rk}(X)=1$ the function $F$ admits a meromorphic
500: continuation to $\cz$ with all poles contained in the set
501: $\I [r,\infty)$.
502: \end{theorem}
503:
504: Denote by $\Lambda^{o}$ the concrete Riemann surface associated with the function
505: $\sqrt{z-|\rho|^2}$ on the domain $\cz \backslash{[|\rho|^2,\infty)}$. This
506: means that $\Lambda^{o}$ is a branched double cover of the complex plane
507: with branching point $|\rho|^2$.
508: The original domain $\cz \backslash{[|\rho|^2,\infty)}$ is commonly referred
509: to as the physical sheet, whereas its complement with the spectrum $[|\rho|^2,\infty)$
510: removed is called the unphysical sheet.
511: Let
512: $\Lambda^o$ be the surface with the half line $(-\infty,|\rho|^2-r^2]$
513: removed on the unphysical sheet. Then the above means that
514: the resolvent kernel regarded as a distribution extends to a function
515: which is holomorphic on $\Lambda_r^o$.
516: In case the rank is $1$ the resolvent kernel extends to a meromorphic
517: function on $\Lambda^o$ with all poles contained in the half
518: line $(-\infty,|\rho|^2-r^2]$ on the unphysical sheet.
519: Note that in the literature sometimes another parameterization is chosen.
520: By a change of variables $z=\I(s-|\rho|)$ one can see that the functions
521: \begin{gather}
522: \langle f,\left (\Delta_g - s(4 |\rho|^2-s)\right)^{-1} g \rangle,
523: \end{gather}
524: which are defined in the half plane $\Re(s) < |\rho|$ have an analytic
525: continuation across the line $\Re(s) = |\rho|$.
526:
527: In case the symmetric space has even rank there is an analogous result.
528:
529: \begin{theorem} \label{th2}
530: Let $X=G/K$ by a symmetric space of noncompact type with invariant
531: Riemannian metric $g$. Suppose that the rank of $X$ is even. Denote by
532: $S_{a,b}$ the strip $\{z \in \cz;\; a < \Im(z) < b\}$. Then for each $f,g
533: \in C^\infty_0(X)$ the holomorphic function
534: \begin{gather*}
535: F: S_{-\pi,0} \to \cz,\quad z \to \langle f,\; (\Delta_g
536: -|\rho|^2 - e^{2z})^{-1} \;g \rangle
537: \end{gather*}
538: admits a holomorphic continuation to the open set
539: \begin{gather*}
540: \mathcal{U}:=\{z \in \cz;\; z \notin \I \pi (n+\frac{1}{2}) + [\log(r),\infty)
541: \quad \forall n \in (\zz \backslash \{-1\})\}.
542: \end{gather*}
543: \end{theorem}
544:
545: Let $\Lambda^e$ be the concrete Riemann surface associated with the function
546: $\log(z-|\rho|^2)$ on the domain $\cz \backslash{[|\rho|^2,\infty)}$.
547: This means, $\Lambda^e$ is the logarithmic covering of $\cz \backslash {|\rho|^2}$.
548: Denote by $\Lambda_r^e$ the Riemann
549: surface $\Lambda^e$ with the half line $(-\infty,|\rho|^2-r^2]$
550: removed on all unphysical sheets. Then our result means that the resolvent
551: kernel regarded as a distribution has a holomorphic extension to $\Lambda_r^e$.
552:
553: If all Cartan subalgebras of $G$ are conjugate there are even stronger
554: statements. This condition is known to be equivalent to each of the
555: following (see \cite{HelgI}, Ch. IX, Th. 6.1)
556: \begin{enumerate}
557: \item all restricted roots have even multiplicity, i.e. $m_\alpha$ is even
558: for all $\alpha \in \Delta(\lieg,\liea)$.
559: \item $\mathrm{rk}(G)=\mathrm{rk}(X)+\mathrm{rk}(K)$.
560: \end{enumerate}
561: In this case it follows that $m_{2 \alpha}=0$ and that the function
562: $\hc(\lambda)^{-1}$ is a polynomial in $\lambda$ (see \cite{HelgII}, Ch. IV,
563: Cor. 6.15). The irreducible symmetric spaces of noncompact type, where this
564: happens are
565: \begin{itemize}
566: \item the real hyperbolic spaces of odd dimension, i.e. $SO_0(2n+1,1)/SO(2n)$,
567: \item the spaces $SU^*(2n)/Sp(n)$,
568: \item the spaces $G/K$ where $G$ complex,
569: \item the exceptional space $E_{6(-26)}/F_4$.
570: \end{itemize}
571:
572: \begin{theorem} \label{th3}
573: Let $X=G/K$ by a symmetric space of noncompact type with invariant
574: Riemannian metric $g$. Suppose that $G$ has only one conjugacy class
575: of Cartan subalgebras. Let $f,g \in C^\infty_0(X)$ and
576: $$
577: F(z):=\langle f, (\Delta_g
578: -|\rho|^2 - z)^{-1} g \rangle
579: $$
580: Then the following holds.
581: \begin{enumerate}
582: \item if $\mathrm{rk}(X)$ is odd then $F(z^2)$
583: has an analytic continuation to the whole complex plane.
584: \item if $\mathrm{rk}(X)$ is even then $F(e^{2z})$ has an analytic
585: continuation to the whole complex plane.
586: \end{enumerate}
587: \end{theorem}
588:
589: Hence, in this special situation the resolvent kernel has a holomorphic extension
590: to $\Lambda^e$ of $\Lambda^o$, depending on whether the rank of $X$ is even or
591: odd.
592:
593: \section{Proof of the main results}
594:
595: We will split the proof of this theorem into several propositions.
596: By the spectral theorem we have for $f,g \in C^\infty_0(X)$ and $z \notin
597: [|\rho|^2,\infty)$
598: \begin{gather}
599: \langle f, (\Delta_g-z)^{-1} g \rangle= \nonumber \\
600: =w^{-1} \int_{\liea^*} \frac{1}{|\hc(\lambda)|^2(|\lambda|^2 +
601: |\rho|^2 -z)} \int_B
602: \overline{\hat f(\lambda,b)}\hat g(\lambda,b) db \;d\lambda=\\
603: = \int_{\liea^*} \frac{1}{|\hc(\lambda)|^2(|\lambda|^2 + |\rho|^2 -z)}
604: V(\lambda) d \lambda, \nonumber
605: \end{gather}
606: where $V(\lambda)=w^{-1} \int_B \hat {\bar{f}} (-\lambda,b)\hat
607: g(\lambda,b) db$ is rapidly decaying and admits a continuation to an entire
608: function on $\cz$. This follows from the analytic properties of the Fourier
609: transforms and the fact that $B$ is compact. Now denote by $S$ the unit
610: sphere in $\liea^*$, i.e. $S=\{\lambda \in \liea^*;\; |\lambda|=1\}$. Then using
611: polar coordinates we obtain
612: \begin{gather} \label{specform}
613: \langle f, (\Delta_g-|\rho|^2-z)^{-1} g \rangle = \int_{\rz^+}
614: \frac{F(x)}{(x^2 -z )} dx,
615: \end{gather}
616: where
617: \begin{gather}\label{deff}
618: F(x)=C x^{\mathrm{dim}(\liea-1)} \int_S
619: \frac{V(x\lambda)}{\hc(x \lambda)\hc(- x \lambda)} d \mu_S(\lambda),
620: \end{gather}
621: $\mu_S$ is the usual measure on the sphere and $C$ is a constant not depending
622: on $x$.
623:
624: \begin{pro} \label{holf}
625: The function $F: \rz^+ \to \cz$ defined by (\ref{deff}) is bounded and admits a
626: holomorphic continuation to the set $\cz \backslash (\I [r,\infty) \cup -\I
627: [r,\infty))$ (see Fig. \ref{domain}). Moreover,
628: $F(-z)=(-1)^{\mathrm{rk}(X)-1} F(z)$ and
629: $\lim_{z \to 0} z^{1-\mathrm{rk}(X)} F(z)=0$. In case $m_\alpha$ is even for all
630: $\alpha \in \Delta(\lieg,\liea)_0^+$ the function $F$ admits a continuation
631: to an entire function. In case the rank of $X$ is one then $F$ has a
632: meromorphic extension to the whole complex plane.
633: \end{pro}
634:
635: \begin{proof}
636: The inverse of the $\hc$-function is polynomially bounded (see
637: \cite{HelgII}, Ch. IV, Prop. 7.2) whereas $V$ is
638: rapidly decreasing. This immediately implies that $F$ is rapidly decreasing
639: as well and hence, bounded.
640: For each $\alpha \in \Delta(\lieg,\liea)_0^+$ we define the meromorphic function
641: \begin{gather}
642: h_\alpha(z):=
643: \frac{\Gamma(\frac{1}{2}(\frac{1}{2} m_\alpha + 1 + z)) \Gamma(\frac{1}{2}(\frac{1}{2} m_\alpha +
644: m_{2 \alpha} + z))}{\Gamma(z)}.
645: \end{gather}
646: Note that all poles of $h_\alpha$ are on the negative real axis.
647: In case $m_\alpha$ is even and $m_{2 \alpha}=0$ the function $h_\alpha$
648: is entire in $z$. If $m_\alpha$ is odd it is known that $m_{2
649: \alpha}=0$ (see \cite{MR27:3743}, or \cite{HelgI} Chapter X, Exercise F).
650: Hence, in this case the set of poles of $h_\alpha$ is
651: $-(\frac{m_\alpha}{2}+n) \quad n \in \nz_0$. As in (\ref{defr}) let
652: $j(x)=\frac{x}{2}$ if $x$ is odd and $j(x)=\frac{x}{2}+1$ if $x$ is even.
653: Hence, if $|z|< j(m_\alpha)$ then $z$ is not a pole of $h_\alpha$.
654: Now suppose that $|\lambda| < |\alpha| j(m_\alpha)$.
655: Then we have
656: \begin{gather}
657: |\langle \lambda , \frac{\alpha}{\langle \alpha,\alpha
658: \rangle} \rangle| \leq |\lambda|_\killing |\alpha|_\killing^{-1}=|\lambda|
659: |\alpha|^{-1} < j(m_\alpha),
660: \end{gather}
661: and therefore the function $h_\alpha(\langle \I \lambda,\alpha_0 \rangle)$
662: is analytic in the ball $|\lambda| < r$, where $r$ is defined by (\ref{defr}).
663: By the product formula (\ref{GKFormula}) the function
664: $\frac{1}{\hc(-\lambda) \hc(\lambda)}$ is analytic in the ball
665: $|\lambda|<r$.
666: Now suppose that $z \in \cz \backslash \I \rz$ and $\lambda \in S$. Then we have either
667: $\langle \I z \lambda ,\alpha_0 \rangle=0$ or
668: $\Im (\langle \I z \lambda ,\alpha_0 \rangle) \not= 0$. In this case $z
669: \lambda$ is not a pole of $h_\alpha(\langle \I \lambda,\alpha_0 \rangle)$.
670: We conclude that the function $\cz \backslash (\I [r,\infty) \cup -\I
671: [r,\infty)) \times S \to \cz$
672: \begin{gather*}
673: (z,\lambda) \to \frac{V(z\lambda)}{\hc(z \lambda)\hc(- z \lambda)}
674: \end{gather*}
675: is holomorphic in the first variable.
676: Hence, the defining formula (\ref{deff}) for
677: the function $F$ yields the desired analytic continuation. The formula for
678: $F(-z)$ is immediate from (\ref{deff}). Since the $\Gamma$-function has a
679: pole at $0$ the function $\hc(\lambda)^{-1}$ vanishes at $0$. Hence, the
680: integral on the right hand side of (\ref{deff}) vanishes at
681: $\lambda=0$. This implies $\lim_{z \to 0} z^{1-\mathrm{rk}(X)} F(z)=0$.
682: If all
683: $m_\alpha$ are even the function $\frac{1}{\hc(-\lambda) \hc(\lambda)}$ is a
684: polynomial and hence entire. In this case $F$ is entire as well by the same
685: argument. In case the rank of $X$ is one the function $F$ is meromorphic
686: in the whole complex plane, since in this case the function
687: $\frac{1}{\hc(-\lambda) \hc(\lambda)}$ is a meromorphic function of one
688: variable and the integral in (\ref{deff}) is a sum.
689: \end{proof}
690:
691: \begin{figure}
692: \centerline{\includegraphics*[width=7cm]{domain}}
693: \caption{Domain of analyticity of the function F(z)}\label{domain}
694: \end{figure}
695:
696: By the above observation we are interested in evaluating integrals of the
697: form
698: $$
699: \int_{\rz^+} \frac{f(x)}{x^2-z} dx,
700: $$
701: where $f$ is a meromorphic function on $\cz \backslash (\I [r,\infty)\cup
702: -\I [r,\infty))$ with no poles on the real axis. This integral clearly
703: defines a holomorphic function on $\cz \backslash \rz^+$, and we may ask,
704: whether it has an extension to a meromorphic function on a larger Riemann
705: surface. We have the following results.
706: \begin{pro} \label{pro1}
707: Let $r$ be a positive real number and suppose that $f$ is a meromorphic
708: function on the set $\cz \backslash (\I [r,\infty)\cup -\I [r,\infty))$ such
709: that
710: \begin{gather*}
711: |f(x)| \leq C, \quad \forall x \in \rz \quad \textrm{and}\\
712: f(x)=f(-x),\quad \forall x \in \rz.
713: \end{gather*}
714: Define the holomorphic function $G$ on $H^-:=\{z \in \cz;\; \Im(z)<0\}$ by
715: $$
716: G(z):= \int_{\rz} \frac{f(x)}{x^2-z^2} dx.
717: $$
718: Then $G(z)$ has a meromorphic continuation to $\cz \backslash \I
719: [r,\infty)$. Except for the point $0$ all poles of $G(z)$ are contained in
720: the set of poles of $f$. The singular behaviour of $G(z)$ at $0$ is like
721: $\frac{\I \pi f(z)}{z}$. Hence, if $f(0)=0$ then $0$ is not a pole.
722: \end{pro}
723: \begin{proof}
724: The function $G$ is clearly holomorphic in the lower half plane, since the
725: integral converges absolutely. Let $\gamma$ be a path in $\cz$ chosen
726: like in Fig. \ref{patha}, such that $\gamma$ does not meet any poles.
727: \begin{figure}
728: \centerline{\includegraphics*[width=7cm]{path1}}
729: \caption{The path $\gamma$, poles of $f(z)$ are indicated by crosses}\label{patha}
730: \end{figure}
731: Denote the domain $\cz \backslash (\I [r,\infty)\cup -\I [r,\infty))$ by
732: $\mathcal{U}$. The path divides $\mathcal{U}$ into two components. Let
733: $\mathcal{U}^-_\gamma$ be the interiour of the lower component. Then
734: \begin{gather}
735: G_\gamma(z):= \int_{\gamma} \frac{f(x)}{x^2-z^2} dx
736: \end{gather}
737: defines a holomorphic function on the open set
738: $O_\gamma=\mathcal{U}^-_\gamma \cap -\mathcal{U}^-_\gamma$. Let $H^+$ and
739: $H^-$ be the upper and lower half plane respectively, i.e. $H^\pm=\{z \in
740: \cz;\; \Im(z)\gtrless 0\}$.
741:
742: By Cauchies theorem we have for $z \in O_\gamma \cap H^-$
743: \begin{gather}
744: G(z)-G_\gamma(z)=2 \pi \I \sum_a \mathrm{Res}_{x=a}(\frac{f(x)}{x^2-z^2}),
745: \end{gather}
746: where the sum is taken over all poles $a$ contained in $O_\gamma \cap H^+$
747: of the function $x \to \frac{f(x)}{x^2-z^2}$. Now choose $z \in O_\gamma \cap H^-$
748: such that $-z$ is not a pole of $f$. Then the function
749: $\frac{f(x)}{x^2-z^2}$ has a simple pole at $x=-z$ with residuum
750: $\frac{f(z)}{2z}$. If $a$ is a pole of $f$ of order $k$ we have
751: \begin{gather}
752: \mathrm{Res}_{x=a}(\frac{f(x)}{x^2-z^2})= \frac{1}{(k-1)!}
753: (\frac{\partial}{\partial x})^{k-1} \frac{f(x)}{x^2-z^2}|_{x=a}
754: \end{gather}
755: This expression is a finite sum of terms of the form $b_n (z^2-a^2)^{-n}$
756: where the constants $b_n$ do not depend on $z$. Hence, if $a$ is a pole of
757: $f$ then $R_a(z)=\mathrm{Res}_{x=a}(\frac{f(x)}{x^2-z^2})$ is a meromorphic
758: function of $z$ in the whole complex plane with poles only at $z=\pm a$. Let
759: $\mathcal{P}$ be the set of poles of $f$ in $O_\gamma \cap H^+$. We conclude
760: that
761: \begin{gather}
762: R(z)=\frac{f(z)}{2 z} + \sum_{a \in \mathcal{P}} R_a(z)
763: \end{gather}
764: extends to a meromorphic function on the complex plane with poles only at
765: $0$ and at the poles of $f$. The residuum of $R(z)$ at $0$ is $f(0)/2$.
766: Moreover, if $z \in O_\gamma \cap H^-$ and $-z$ is not a pole of $f$ we have
767: the equation
768: \begin{gather}
769: G(z)-G_\gamma(z)=2 \pi \I R(z).
770: \end{gather}
771: Since the set of such points is open this equation holds everywhere in
772: $O_\gamma \cap H^-$. Since $G_\gamma(z)$ is holomorphic in $O_\gamma$ the
773: equation
774: \begin{gather}
775: G(z)=2 \pi \I R(z) + G_\gamma(z)
776: \end{gather}
777: defines a meromorphic continuation of $G(z)$ to $O_\gamma \cup H^-$.
778: Since $\gamma$ can be chosen such that an arbitrary point in $\cz \backslash
779: \I [r,\infty)$ is contained in $O_\gamma$ this completes the proof.
780: \end{proof}
781:
782: This allows us to analytically continue integrals of the form
783: $$
784: \int_{\rz^+} \frac{f(x)}{x^2-z^2} dx
785: $$
786: in case $f(x)=f(-x)$. The case $f(x)=-f(-x)$ is covered by the following
787: proposition.
788:
789: \begin{pro} \label{pro2}
790: Suppose that $f: \cz \backslash (\I
791: [r,\infty)\cup -\I [r,\infty)) \to \cz$ is a holomorphic function such that
792: \begin{gather*}
793: |f(x)| \leq C, \quad \forall x \in \rz \quad \textrm{and},\\
794: f(x)=-f(-x), \quad \forall x \in \rz.
795: \end{gather*}
796: Let $S_{a,b}$ be the strip $\{z \in \cz;\; a < \Im(z) < b\}$ and define the
797: holomorphic function $G$ on $S_{-\pi,0}$ by
798: $$
799: G(z):= \int_{\rz^+} \frac{f(x)}{x^2-e^{2z}} dx.
800: $$
801: Denote by $\mathcal{U}$ the open set
802: \begin{gather*}
803: \mathcal{U}:=\{z \in \cz;\; z \notin \I \pi (n+\frac{1}{2}) + [\log(r),\infty)
804: \quad \forall n \in (\zz \backslash \{-1\})\}.
805: \end{gather*}
806: Then $G$ has an analytic continuation to $\mathcal{U}$. Moreover, denoting
807: the analytic continuation again by $G$ we have
808: \begin{gather*}
809: G(z+\I \pi)-G(z)= \pi \I f(e^z) e^{-z}.
810: \end{gather*}
811: \end{pro}
812: \begin{proof}
813: A simple change of variables gives
814: \begin{gather}
815: G(z)=\int_{\rz} \frac{f(e^{x})}{e^{2x}-e^{2z}} e^{x} dx.
816: \end{gather}
817: We define $\mathcal{U}_0$ by
818: \begin{gather*}
819: \mathcal{U}_0:=\{z \in \cz;\; z \notin \I \pi (n+\frac{1}{2}) + [\log(r),\infty)
820: \quad \forall n \in \zz\}.
821: \end{gather*}
822: By assumption the function $g(x):=f(e^{x}) e^{x}$ is holomorphic in
823: $\mathcal{U}_0$ and is periodic in the sense that $g(x + \pi \I)=g(x)$. Now
824: choose a path $\gamma$ like in Fig. \ref{pathb}.
825: \begin{figure}
826: \centerline{\includegraphics*[width=7cm]{path2}}
827: \caption{The Path $\gamma$}\label{pathb}
828: \end{figure}
829: Denote by $O_\gamma$ the connected component of $-\infty-\I \frac{\pi}{2}$ in
830: $\mathcal{U}_0 \backslash (\gamma \cup (\gamma-\I \pi \zz))$. Then
831: \begin{gather}
832: G_\gamma(z)=\int_{\gamma} \frac{g(x)}{e^{2x}-e^{2z}} dx.
833: \end{gather}
834: defines a function holomorphic in $O_\gamma$. Now let $z$ be a point in
835: $O_\gamma \cap S_{-\pi,0}$. Then by Cauchies integral theorem we have
836: \begin{gather}\label{equ40}
837: G(z)=G_\gamma(z).
838: \end{gather}
839: For each point $x
840: \in \mathcal{U}_0 \cap H^+$ there is a curve, such that $x \in
841: O_\gamma$. Hence, Equ. (\ref{equ40}) defines
842: an analytic continuation of $G$ to $\mathcal{U} \cap H^+$.
843: Suppose now that $\gamma_1$ and $\gamma_2$ are two different paths like in
844: Fig. \ref{pathb} such that $z \in O_{\gamma_1}$ and $z+\I \pi \in
845: O_{\gamma_2}$. Then Cauchies integral theorem gives
846: \begin{gather}
847: G_{\gamma_2}(z+\I \pi)-G_{\gamma_1}(z)=2 \pi \I \frac{g(z)}{2} e^{-2z}= \pi
848: \I f(e^z) e^{-z}.
849: \end{gather}
850: Hence, the analytically continued function satisfies
851: \begin{gather} \label{period}
852: G(z+\I \pi)-G(z)= \pi \I f(e^z) e^{-z}.
853: \end{gather}
854: This formula provides an extension of $G$ to all of $\mathcal{U}$ satisfying
855: (\ref{period}).
856: \end{proof}
857:
858: In view of the formula (\ref{specform}) the combination of
859: Prop. \ref{holf} with Prop. \ref{pro1} and Prop. \ref{pro2} proves the
860: Theorems \ref{th1}, \ref{th2} and \ref{th3}.
861:
862: \section{Concluding Remarks}
863:
864: By elliptic regularity the resolvent kernel (and its analytic continuation)
865: is smooth off the diagonal. Hence, we automatically obtain an analytic
866: continuation of the functions $R_z(x,y)$ for $x \not= y$ to the same
867: Riemann surfaces.
868:
869: Our result may be used to obtain meromorphic continuations of resolvent
870: kernels associated to operators of the form $\Delta_g + V$, where
871: $V$ is a compactly supported potential. This can be obtained using standard
872: perturbation arguments based on the analytic Fredholm theorem.
873:
874: We would also like to mention that our method
875: does not rely on the fact that the Fourier transforms of the
876: test functions are in the space of rapidly decaying functions of
877: uniform exponential type $\Hi(a^*_\cz)$.
878: We only use the fact that the Fourier transforms
879: extend to entire function on $a^*_\cz$ and that the function $F$ defined
880: by $(\ref{deff})$ is integrable. The latter happens to be the case for
881: all $f,g \in L^2(X)$. The former requires certain growth conditions
882: for $f$ and $g$ at infinity.
883: Suppose for example that $\alpha$ is a smooth function on $X$
884: such that $\alpha>1$ and such that for all $b \in B$
885: the functions $\alpha(x)^{-\frac{1}{2}} e^{\lambda(A(x,b))}$
886: are entire functions of $\lambda$ with values in $L^2(X)$.
887: Then our results remain true for $f,g$ in the weighted Hilbert space
888: $L^2(X,\alpha(x) dx)$.
889: Denoting $\Hi_+=L^2(X,\alpha(x) dx)$ and $\Hi_-=L^2(X,\alpha^{-1}(x) dx)$
890: we therefore see that the resolvent regarded as function with values
891: in $\mathcal{L}(\Hi_+,\Hi_-)$ has an analytic (meromorphic) continuation
892: to the considered Riemann surfaces. This may be useful when considering
893: perturbations of the Laplace operator by potentials that are not compactly
894: supported.
895:
896: %\break
897: ${}$\\[10pt]
898: {\bf Acknowledgments}\\[6pt]
899:
900: I would like to thank W. M\"uller for raising my interest in this
901: topic and for many discussions and hints. For the latter
902: I would also like to thank D. Sch\"uth and H. Thaler.
903:
904: \begin{thebibliography}{10}
905:
906: \bibitem{MR27:3743}
907: S.~I. Araki, \emph{On root systems and an infinitesimal classification of
908: irreducible symmetric spaces}, J. Math. Osaka City Univ. \textbf{13} (1962),
909: 1--34.
910:
911: \bibitem{MR96j:35183}
912: A.~S{\'a} Barreto and M.~Zworski, \emph{Existence of resonances in three
913: dimensions}, Comm. Math. Phys. \textbf{173} (1995), no.~2, 401--415.
914:
915: \bibitem{CarrPed:03}
916: G.~Carron and E.~Pedon, \emph{On the differential form spectrum of hyperbolic
917: manifolds}, (2003), math.DG/0303348.
918:
919: \bibitem{CFKS}
920: H.~L. Cycon, R.~G. Froese, W.~Kirsch, and B.~Simon, \emph{{Schr\"odinger
921: Operators}}, Springer Verlag, Berlin Heidelberg New York, 1987.
922:
923: \bibitem{MR92i:32016}
924: C.~L. Epstein, R.~B. Melrose, and G.~A. Mendoza, \emph{Resolvent of the
925: {L}aplacian on strictly pseudoconvex domains}, Acta Math. \textbf{167}
926: (1991), no.~1-2, 1--106.
927:
928: \bibitem{MR39:360}
929: R.~Gangolli, \emph{Asymptotic behavior of spectra of compact quotients of
930: certain symmetric spaces}, Acta Math. \textbf{121} (1968), 151--192.
931:
932: \bibitem{MR27:240}
933: S.~G. Gindikin and F.~I. Karpelevi{\v{c}}, \emph{Plancherel measure for
934: symmetric {R}iemannian spaces of non-positive curvature}, Dokl. Akad. Nauk
935: SSSR \textbf{145} (1962), 252--255.
936:
937: \bibitem{MR96h:58172}
938: L.~Guillop{\'e} and M.~Zworski, \emph{Polynomial bounds on the number of
939: resonances for some complete spaces of constant negative curvature near
940: infinity}, Asymptotic Anal. \textbf{11} (1995), no.~1, 1--22.
941:
942: \bibitem{MR20:925}
943: Harish-Chandra, \emph{Spherical functions on a semisimple {L}ie group. {I}},
944: Amer. J. Math. \textbf{80} (1958), 241--310.
945:
946: \bibitem{MR41:8587}
947: S.~Helgason, \emph{A duality for symmetric spaces with applications to group
948: representations}, Adv. in Math. \textbf{5} (1970), 1--154.
949:
950: \bibitem{MR51:3804}
951: S.~Helgason, \emph{The surjectivity of invariant differential operators on
952: symmetric spaces. {I}}, Ann. of Math. (2) \textbf{98} (1973), 451--479.
953:
954: \bibitem{HelgI}
955: S.~Helgason, \emph{{Differential Geometry, Lie Groups, and Symmetric Spaces}},
956: Academic Press, New York, 1978.
957:
958: \bibitem{HelgII}
959: \bysame, \emph{{Groups and Geometric Analysis}}, Academic Press, New York,
960: 1984.
961:
962: \bibitem{MazVas:02}
963: R.~Mazzeo and A.~Vasy, \emph{Analytic continuation of the resolvent of the
964: {L}aplacian on {$SL(3)/SO(3)$}}, (2002), math.AP/0208116v2.
965:
966: \bibitem{MazVas:03}
967: \bysame, \emph{Analytic continuation of the resolvent of the {L}aplacian on
968: symmetric spaces of noncompact type}, (2003), math.AP/0308043v2.
969:
970: \bibitem{MR89c:58133}
971: R.~B. Mazzeo, R.~Melrose, \emph{Meromorphic extension of the resolvent on
972: complete spaces with asymptotically constant negative curvature}, J. Funct.
973: Anal. \textbf{75} (1987), no.~2, 260--310.
974:
975: \bibitem{MR96g:58180}
976: R.~B. Melrose, \emph{The {A}tiyah-{P}atodi-{S}inger index theorem}, Research
977: Notes in Mathematics, vol.~4, A K Peters Ltd., Wellesley, MA, 1993.
978:
979: \bibitem{MR96k:35129}
980: \bysame, \emph{Geometric scattering theory}, Stanford Lectures, Cambridge
981: University Press, Cambridge, 1995.
982:
983: \bibitem{MR93i:58160}
984: R.~Miatello and N.~R. Wallach, \emph{The resolvent of the {L}aplacian on
985: locally symmetric spaces}, J. Differential Geom. \textbf{36} (1992), no.~3,
986: 663--698.
987:
988: \bibitem{MR98f:11053}
989: W.~M{\"u}ller, \emph{On the analytic continuation of rank one {E}isenstein
990: series}, Geom. Funct. Anal. \textbf{6} (1996), no.~3, 572--586.
991:
992: \end{thebibliography}
993:
994:
995: \vspace{0.5cm}
996:
997: \parskip0ex
998:
999: Mathematisches Institut
1000:
1001: Universit\"at Bonn
1002:
1003: Beringstr. 1
1004:
1005: 53115 Bonn
1006:
1007: Germany
1008:
1009: \vspace{0.4cm}
1010:
1011: E-Mail: {\tt strohmai@math.uni-bonn.de}
1012:
1013:
1014: \end{document}
1015:
1016:
1017:
1018:
1019:
1020:
1021:
1022:
1023:
1024:
1025:
1026:
1027:
1028:
1029:
1030: