1: \documentclass{article}
2:
3: \usepackage{amssymb,eufrak,epsfig,latexsym}
4: \parindent0mm
5: \setlength{\parskip}{1ex}
6:
7: %\input /afs/aei-potsdam.mpg.de/u/jurke/latex/kwg/prom/Doktorarbeit/Latex/STYLE-FI/tcilatex
8: \input tcilatex
9: \newenvironment{beweis}{{\bfseries\upshape\selectfont Proof:}}
10: {\ifmmode\rlap{\hskip\displaywidth\rule{2mm}{2mm}}\else\hfill $\rule{2mm}{2mm}$\fi}
11: \def\ABB#1#2#3{\vskip6mm\def\Test{#2}\ifx\Test\empty\centerline{%
12: \epsfig{figure=#1}}\else\centerline{\epsfig{figure=#1, width=#2}}\fi%
13: \def\Test{#3}\ifx\Test\empty\else\vskip0mm\global\advance\c@Abbildung %
14: by 1\def\@currentlabel{(\thechapter-\theAbbildung)}%
15: \centerline{\mbox{\rmfamily\mdseries\upshape\fontsize{9pt}{13pt}%
16: \selectfont \begin{tabular}{rl} Abbildung (\thechapter-\theAbbildung):%
17: \kern -3mm & #3\end{tabular}}}\fi\vskip5mm}
18:
19: \begin{document}
20:
21: \title{On future asymptotics of polarized Gowdy
22: $\mathbb T^3$-models}
23: \author{Thomas Jurke\\
24: {\em Max-Planck-Institut f\"ur Gravitationsphysik}\\
25: {\em Am M\"uhlenberg 1}\\
26: {\em 14476 Golm}\footnote{current address:
27: Weierstra{\ss}-Institut
28: f\"ur Angewandte Analysis und
29: Stochastik, Mohrenstra{\ss}e 39, 10117 Berlin}}
30: \maketitle
31:
32: \begin{abstract}
33: Gowdy's model of cosmological spacetimes is a much investigated
34: subject in classical and quantum gravity. Depending on spatial
35: topology re\-collapsing as well as expanding models are known.
36: Several analytic tools were used in order to clarify singular
37: behaviour in this class of spacetimes. Here we investigate the
38: structure of a certain subclass, the polarized Gowdy models with
39: spatial $\mathbb T^3$-topology, in the large.
40: The a\-sym\-pto\-tics for general solutions of the dynamical equation for one
41: of the gravitational degrees of freedom plays a key role
42: while the asymptotic behaviour of the remaining metric function
43: is a result of solving the Hamiltonian constraint equation.
44: Using both we are able to prove (future) geodesic
45: completeness in all spacetimes of this type.
46: \end{abstract}
47:
48:
49: \section{Introduction}
50: %\begin{eqnarray}
51: %\frac{\partial^2}{\partial t^2}W\kern1mm+\kern1mm
52: %\frac1t\frac{\partial}{\partial t}W\kern1mm-\kern1mm
53: %\frac{\partial^2}{\partial x^2}W&=&\sinh W\cosh W
54: %\left(\left(\frac{\partial Q}{\partial t}\right)^2-
55: %\left(\frac{\partial Q}{\partial x}\right)^2\right)
56: % \label{GowdyW}\\
57: %\frac{\partial^2}{\partial t^2}Q\kern1mm+\kern1mm
58: %\frac1t\frac{\partial}{\partial t}Q\kern1mm-\kern1mm
59: %\frac{\partial^2}{\partial x^2}Q&=&-2\coth W
60: %\left(\frac{\partial Q}{\partial t}\frac{\partial W}{\partial t}-
61: %\frac{\partial Q}{\partial x}\frac{\partial W}{\partial x}\right)
62: % \label{GowdyQ}\\
63: %\frac{\partial^2}{\partial t^2}a\kern1mm+\kern1mm
64: %\frac1t\frac{\partial}{\partial t}a\kern1mm-\kern1mm
65: %\frac{\partial^2}{\partial x^2}a&=&\frac12
66: %\left(\left(\frac{\partial W}{\partial x}\right)^2+\sinh^2 W
67: %\left(\frac{\partial Q}{\partial x}\right)^2\right)
68: % \label{Gowdya}
69: %\end{eqnarray}
70: %\begin{eqnarray}
71: %\frac{\partial}{\partial t}a&=&
72: %-\frac1{4t}+\frac{t}4\left[
73: %\left(\frac{\partial W}{\partial t}\right)^2+
74: %\left(\frac{\partial W}{\partial x}\right)^2+
75: %\right.\nonumber\\
76: %&&\left.
77: %\sinh^2 W\left(
78: %\left(\frac{\partial Q}{\partial t}\right)^2+
79: %\left(\frac{\partial Q}{\partial x}\right)^2
80: %\right)
81: %\right]
82: % \label{Gowdyat}\\
83: %\frac{\partial}{\partial x}a&=&\frac{t}2\left(
84: %\frac{\partial W}{\partial t}
85: %\frac{\partial W}{\partial x}+
86: %\sinh^2 W\frac{\partial Q}{\partial t}
87: %\frac{\partial Q}{\partial x}
88: %\right)
89: % \label{Gowdyax}
90: %\end{eqnarray}
91: The cosmological spacetime model by Gowdy (1974) has experienced a great
92: deal of attention over the last years. These models are especially
93: interesting because they allow rigorous analytical investigations in
94: inhomogeneous spacetimes. Various spatial topologies are allowed but~--~as
95: is mostly done~--~we will concentrate on vacuum spacetimes with $\mathbb
96: R_+\times\mathbb T^3$-topology. A basic analytical tool for this class of
97: spacetimes was provided by Moncrief~(1981) who proved existence of global
98: smooth solutions for the corresponding Einstein equations. Investigations
99: concerning the singular asymptotic behaviour were subjects of interest
100: almost exclusively. Now, we have a quite clear picture what happens
101: in the polarized as well as the full Gowdy $\mathbb T^3$-model (see
102: Isenberg, Moncrief~(1990), Rendall~(2000), Ringstr\"om~(2002) and references
103: therein). On the other hand, due to Chru\'sciel, Isenberg and Moncrief~(1990)
104: there exist some statements about the asymptotic behaviour for large times in
105: the case of polarized spacetimes. In this paper we prove such theorems,
106: describing the long time asymptotics for general polarized Gowdy models
107: and using this results to prove future completeness of any causal geodesic.
108:
109:
110: \section{The polarized model}
111: By definition a smooth, orientable spacetime is a cosmological model of Gowdy
112: type if it is a maximally extended, globally hyperbolic solution of the
113: vacuum Einstein equations having compact Cauchy surfaces and a
114: pseudo-Riemannian metric, invariant under an effective $U(1)\times U(1)$
115: group action on these Cauchy surface. This action is generated by two
116: spacelike Killing vector fields which are assumed to have vanishing twist
117: constants. A Gowdy model is called polarized if the defining Killing vector
118: fields are everywhere mutually orthogonal. By virtue of this definition
119: we may describe a polarized Gowdy metric by the line element
120: \begin{eqnarray}
121: {\rm d}s^2&=&{\rm e}^{2a(t,x)}\left(-{\rm d}t^2+{\rm d}x^2\right)\kern1mm+
122: \kern1mmt\kern.5mm\left({\rm e}^{W(t,x)}{\rm d}y^2+{\rm e}^{-W(t,x)}
123: {\rm d}z^2\right).
124: \end{eqnarray}
125: The corresponding vacuum Einstein equations are
126: \begin{eqnarray}
127: 0&=&\frac{\partial^2}{\partial t^2}W\kern1mm+\kern1mm
128: \frac1t\frac{\partial}{\partial t}W\kern1mm-\kern1mm
129: \frac{\partial^2}{\partial x^2}W\label{Darboux}\\[2mm]
130: 0&=&\frac{\partial^2}{\partial t^2}a\kern1mm-\kern1mm
131: \frac{\partial^2}{\partial x^2}a
132: \kern1mm-\kern1mm\frac1{4t^2}\kern1mm+\kern1mm\frac14
133: \left[\left(\frac\partial{\partial t}W\right)^2\kern1mm-\kern1mm
134: \left(\frac\partial{\partial x}W\right)^2
135: \right]\label{Wellena}\\[2mm]
136: 0&=&
137: \frac2t\frac{\partial}{\partial x}a\kern1mm-\kern1mm
138: \frac{\partial}{\partial t}W\kern1mm\frac{\partial}{\partial x}W
139: \label{Impuls1}\\[2mm]
140: 0&=&
141: \frac2t\frac{\partial}{\partial t}a\kern1mm+\kern1mm
142: \frac1{2t^2}\kern1mm-\kern1mm\frac12\left[\left(
143: \frac{\partial}{\partial t}W\right)^2\kern1mm+\kern1mm
144: \left(\frac{\partial}{\partial x}W\right)^2\right].
145: \label{Hamilton1}
146: \end{eqnarray}
147: As in the general non-polarized case equation~(\ref{Wellena}) is always
148: satisfied as
149: a consequence of the others. Thus, we have essentially to deal with a single
150: linear equation of second order while
151: function~$a$ is the integral of the constraint
152: equation~(\ref{Hamilton1}).
153:
154: \section{On the asymptotics for solutions of the
155: Euler-Poisson-Darboux equation}
156:
157: Now, we start by investigating equation~(\ref{Darboux}). As in
158: literature we will sometimes call this equation
159: ``Euler-Poisson-Darboux equation''.
160: Following Moncrief~(1981) it will be sufficient to have only
161: functions $W=W(t,x)$ of a certain regularity in mind
162: which are defined for all
163: $(t,x)\in\mathbb R_+\times\mathbb T^1$.
164:
165: As a first result we get asymptotic homogenity for all
166: solutions of~(\ref{Darboux}). Namely, we will show
167: that for large~$t$ the mean value,
168: \begin{eqnarray}
169: \bar W(t) &\mathop{=}\limits_{\rm Df}& \frac1{2\pi}
170: \intop\limits_0^{2\pi} W(t, x)\kern1mm {\rm d} x,
171: \label{W_mittel}
172: \end{eqnarray}
173: will dominate while the $x$-dependent terms show a
174: $t^{-\frac12}$ fall-off behaviour.
175:
176: \begin{theorem}
177: Let $W\in\mathcal C^2(\mathbb R_+\times\mathbb T^1;\mathbb R)$
178: be a solution of Euler-Poisson-Darboux equation~(\ref{Darboux})
179: and $\bar W$ as decleared in~(\ref{W_mittel}). Then $\bar W$
180: solves~(\ref{Darboux}), too.
181: \end{theorem}
182:
183: \begin{beweis}
184: For every $t\in\mathbb R_+$ function $W$ as well as its first and
185: second derivative are continuous, hence
186: $\mathbb T^1$-integrable and $\bar W$ is well-defined.
187: Since~$W$ is sufficiently smooth $\bar W$ depends on its
188: parameter~$t\in\mathbb R_+$ smoothly, hence
189: $\bar W\in\mathcal C^2(\mathbb R_+;\mathbb R)$ and
190: \begin{eqnarray*}
191: \frac{\rm d^2}{{\rm d}t^2}\bar W(t)&=&\frac1{2\pi}
192: \frac{\rm d}{{\rm d}t}\int_0^{2\pi}\frac\partial{\partial t}
193: W(t,x)\kern1mm{\rm d}x\kern2mm=\kern2mm
194: \frac1{2\pi}\int_0^{2\pi}\frac{\partial^2}{\partial t^2}
195: W(t,x)\kern1mm{\rm d}x.
196: \end{eqnarray*}
197: Now it is easy to integrate~(\ref{Darboux})
198: \begin{eqnarray*}
199: \frac1{2\pi}
200: \intop\limits_0^{2\pi}
201: \frac{\partial^2}{\partial t^2}W \kern1mm {\rm d} x
202: \kern2mm+\kern2mm
203: \frac1{2\pi t}
204: \intop\limits_0^{2\pi}
205: \frac{\partial }{\partial t}W\kern1mm {\rm d} x &=&
206: \frac1{2\pi}
207: \intop\limits_0^{2\pi}
208: \frac{\partial ^2}{\partial x^2}W\kern1mm {\rm d} x\\[2mm]
209: %\end{eqnarray*}
210: %\begin{eqnarray*}
211: \frac{\rm d^2}{{\rm d} t^2}\bar W \kern2mm+\kern2mm
212: \frac1t\frac{\rm d }{{\rm d} t}\bar W&=&
213: \frac1{2\pi}\left.
214: \frac{\partial W}{\partial x}\right|^{2\pi}_0
215: \end{eqnarray*}
216: and since the derivative of smooth periodic functions have same
217: periods we get
218: \begin{eqnarray*}
219: \frac{\partial^2}{\partial t^2}\bar W \kern2mm+\kern2mm
220: \frac1t\frac{\partial }{\partial t}\bar W&=&0\kern2mm
221: \mathop{=}\limits^! \kern2mm
222: \frac{\partial^2}{\partial x^2}\bar W.
223: \end{eqnarray*}
224: \end{beweis}
225:
226: The general solution of~(\ref{Darboux}) which does not depend on $x$
227: is
228: \begin{eqnarray}
229: \bar W(t)&=&\gamma\kern1mm+\kern1mm
230: \beta\cdot\ln t, \label{Whom}
231: \end{eqnarray}
232: where $\beta$ and $\gamma$ may be fixed by some initial values
233: of~$\bar W$ and~${\bar W}_t$ at $t_0>0$. Due to the
234: linearity of~(\ref{Darboux}) $\psi$ in
235: \begin{eqnarray*}
236: t^{-\frac12}\psi(t,x)&\mathop{=}
237: \limits_{\rm Df}& W(t,x)-\bar W(t)
238: \end{eqnarray*}
239: is unique and solves (\ref{Darboux}), too. Obviously this
240: newly defined function~$\psi$,
241: $\psi\in\mathcal C^2(\mathbb R_+\times\mathbb T^1;\mathbb R)$,
242: has zero mean value for every~$t\in\mathbb R_+$
243: \begin{eqnarray}
244: \bar\psi(t)&\mathop{=}\limits_{\rm Df}&
245: \frac1{2\pi}\intop\limits_0^{2\pi}\psi(t,x)\kern1mm{\rm d}x
246: \kern2mm=\kern2mm0\label{mittellos}
247: \end{eqnarray}
248: and satisfies
249: \begin{eqnarray}
250: \frac{\partial^2}{\partial t^2}\psi \kern1mm-\kern1mm
251: \frac{\partial ^2}{\partial x^2}\psi&=&
252: -\frac1{4t^2}\psi.
253: \label{Wellenpsi}
254: \end{eqnarray}
255:
256: Now, in order to prove that any solution of this equation has
257: to be bounded we need the following estimate as prerequisite.
258: Here we have formulated the lemma in dependence of a
259: parameters~$t\in\mathbb R_+$, since this is the form we will need.
260: \begin{lemma}
261: Let \label{Hilfssatz2}
262: $f\colon{\mathbb R}_+\times {\mathbb T}^1\to {\mathbb R}$ be
263: continuously differentiable and
264: \begin{eqnarray*}
265: \frac1{2\pi}\int_0^{2\pi} f(t,x)\kern1mm {\rm d} x&=&0
266: \end{eqnarray*}
267: for every $t$. Then we have for each $(t,x)\in\mathbb R_+
268: \times\mathbb T^1:$
269: \begin{eqnarray*}
270: |f(t,x)|^2 &\le& 2\pi\intop\limits_0^{2\pi}\left|
271: \frac\partial{\partial x} f(t,x)
272: \right|^2\kern1mm {\rm d} x\kern2mm\equiv\kern2mm 2\pi\kern1mm
273: \left|\left| f_x(t,.)\right|\right|_{L^2}^2
274: \end{eqnarray*}
275: \end{lemma}
276:
277: \begin{beweis}
278: The mean value theorem of integral calculus for continuous
279: functions gives us to any~$t$ some
280: $x_0 \in {\mathbb T}^1$ with $f(t, x_0)=0$. Furthermore
281: $|x-x_0|\le 2\pi$ is true for each $x\in\mathbb T^1$. Using
282: Schwartz' inequality we get
283: \begin{eqnarray*}
284: |f(t,x)| &=&
285: \left|\intop\limits_{\phantom{.}x_0}^{x}
286: \frac\partial{\partial\xi} f(t,\xi)
287: \kern1mm {\rm d} \xi \right|
288: \kern2mm\le\kern2mm \intop\limits_0^{2\pi}\left|
289: \frac\partial{\partial x} f(t,x)
290: \right|\kern1mm {\rm d} x \\
291: &\le& \left(\intop\limits_0^{2\pi}1^2
292: \kern1mm {\rm d} x\right)^{\frac12}\cdot
293: \left(
294: \intop\limits_0^{2\pi}\left|
295: \frac\partial{\partial x} f(t,x)
296: \right|^2\kern1mm {\rm d} x
297: \right)^{\frac12}
298: \end{eqnarray*}
299: that is
300: \begin{eqnarray*}
301: |f(t,x)|^2 &\le& 2\pi\intop\limits_0^{2\pi}\left|
302: \frac\partial{\partial x} f(t,x)
303: \right|^2\kern1mm {\rm d} x.
304: \end{eqnarray*}
305: \end{beweis}
306:
307: Now we are able to show boundedness of~$\psi$,
308: uniformly on every strictly positive intervall. The key element
309: in the proof is the investigation of a certain energy
310: functional corresponding to~(\ref{Wellenpsi}).
311:
312: \begin{theorem}
313: To every function~$\psi\in\mathcal C^2(\mathbb R_+
314: \times\mathbb T^1;\mathbb R)$, satisfying both~(\ref{mittellos})
315: and~(\ref{Wellenpsi}), there is a positive constant~$C_{t_0}$
316: depending only on~$t_0>0$, such that for every~$t\ge t_0$ and
317: all $x\in\mathbb T^1$ the inequality
318: \label{Satz4}
319: \begin{eqnarray}
320: |\psi(t, x)|<C_{t_0} \label{Schaetz}
321: \end{eqnarray}
322: holds.
323: \end{theorem}
324:
325: \begin{remark}
326: One verifies that equation (\ref{Wellenpsi}) has unbounded
327: solutions, too. But none of them satisfies~(\ref{mittellos})
328: which underlines the importance of this assumption.
329: \end{remark}
330:
331: \begin{beweis}
332: Let us define the energy
333: \begin{eqnarray*}
334: \varepsilon(t)&\mathop{=}\limits_{\rm Df}&\frac12
335: \int_0^{2\pi}\left({\psi_t}^2+{\psi_x}^2\right){\rm d}x.
336: \end{eqnarray*}
337: Our smoothness assumptions in~$\psi$ guarantees
338: differentiability of~$\varepsilon$ in $\mathbb R_+$.
339: Integrating by parts and using equation~(\ref{Wellenpsi}) yields
340: \begin{eqnarray*}
341: \frac{\rm d}{{\rm d} t}\varepsilon&=&
342: -\frac1{4 t^2}\int \limits_0^{2\pi}
343: \psi_t \psi\kern1mm{\rm d}x.
344: \end{eqnarray*}
345: Since $\frac12({\psi_t}^2+\psi^2)+\psi_t\psi\ge0$ we have
346: \begin{eqnarray*}
347: \frac{\rm d}{{\rm d} t}\varepsilon
348: &\le&
349: \frac1{4 t^2}\cdot\frac12\int \limits_0^{2\pi}
350: \left({\psi_t}^2+\psi^2\right)\kern1mm{\rm d}x.
351: \end{eqnarray*}
352: Now, lemma~\ref{Hilfssatz2} permits an estimate of the left
353: hand side of the inequality in terms of $\varepsilon$ itself:
354: \begin{eqnarray*}
355: \frac{\rm d}{{\rm d} t}\varepsilon&\le&\frac{\pi^2}{t^2}\varepsilon
356: \end{eqnarray*}
357: Since $t\ge t_0>0$ we find
358: \begin{eqnarray*}
359: \varepsilon (t)&\le&
360: \varepsilon (t_0)\kern1mm+\kern1mm
361: \pi^2\intop\limits_{t_0}^t s^{-2}
362: \varepsilon(s)\kern1mm {\rm d}s
363: \end{eqnarray*}
364: and Gronwall's inequality gives the uniform bound
365: \begin{eqnarray*}
366: \varepsilon (t)&\le&
367: \varepsilon (t_0)\kern1mm+\kern1mm
368: \varepsilon (t_0)\left({\rm e}^{\frac{\pi^2}{t_0}
369: -\frac{\pi^2}{t\phantom{_0}}}
370: -1\right)
371: \kern2mm<\kern2mm
372: \varepsilon (t_0)\kern1mm
373: {\rm e}^{\frac{\pi^2}{t_0}}.
374: \end{eqnarray*}
375: Hence, the energy of an arbitrary solution is for every
376: $t>t_0$ bounded and by definition non-negative.
377: Now, up to a factor we can control the
378: $L^2(\mathbb T^1)$-norm
379: of~$\psi_x$ for all $t\ge t_0$ by the square root of the same
380: constant and since~$\psi$ satisfies all suppositions of
381: lemma~\ref{Hilfssatz2} a short calculation yield finally
382: \begin{eqnarray*}
383: |\psi(t,x)|&<&\sqrt{4\pi\kern.5mm\varepsilon(t_0)}\kern1mm
384: {\rm e}^{\frac{\pi^2}{2t_0}}\kern2mm\equiv\kern2mm C_{t_0}
385: \end{eqnarray*}
386: for all $t\ge t_0>0$ and all $x\in\mathbb T^1$, hence $C_{t_0}$
387: fulfills all needs.
388: \end{beweis}
389:
390: We have shown that every spatially periodic solution of the
391: Euler-Poisson-Darboux equation can be cast into the form
392: \begin{eqnarray*}
393: W(t,x)&=&\gamma\kern1mm+\kern1mm\beta\cdot\ln t
394: \kern1mm+\kern1mmt^{-\frac12}\psi(t,x)
395: \end{eqnarray*}
396: where $\psi$ is bounded in $t\in[t_0,\infty)$. On the other
397: hand for applications in the following sections we need deeper
398: insights into the asymptotic behaviour of~$\psi$.
399: Using the boundedness of~$\psi$ we can show that solutions
400: to
401: \begin{eqnarray*}
402: \frac{\partial^2}{\partial t^2}\psi\kern1mm-\kern1mm
403: \frac{\partial^2}{\partial x^2}\psi&=&
404: -\frac1{4t^2}\psi
405: \end{eqnarray*}
406: behave like solutions of the free wave equation in two
407: dimensions with a remainder falling off in time.
408:
409: \begin{theorem}
410: Let $\psi\in\mathcal C^2(\mathbb R_+\times\mathbb T^1;\mathbb R)$
411: be a bounded solution of \label{Satz5}
412: \begin{eqnarray}
413: \frac{\partial^2}{\partial t^2}\psi\kern1mm-\kern1mm
414: \frac{\partial^2}{\partial x^2}\psi&=&
415: -\frac1{4t^2}\psi,\label{Psi2}
416: \end{eqnarray}
417: with the side condition
418: \begin{eqnarray*}
419: \frac1{2\pi}\int_0^{2\pi}\psi(t,x)\kern1mm{\rm d}x&=&0
420: \end{eqnarray*}
421: for every $t\in\mathbb R_+$. Then there exist uniquely
422: defined functions
423: $\nu\in\mathcal C^2(\mathbb R_+\times\mathbb T^1;\mathbb R)$ and
424: $\omega\in\mathcal C^2(\mathbb R_+\times\mathbb T^1;\mathbb R)$
425: as well as a constant~$C$ depending only on the choise of $t_0$
426: such that
427: \begin{eqnarray*}
428: \psi(t,x)&=&\nu(t,x)\kern1mm+\kern1mm\omega(t,x)\\[2mm]
429: \frac{\partial^2}{\partial t^2}\nu&=&
430: \frac{\partial^2}{\partial x^2}\nu \\[2mm]
431: |\omega(t,x)|&\le& C\cdot t^{-1}\\[2mm]
432: |\omega_t (t,x)|&\le& C\cdot t^{-1}
433: \end{eqnarray*}
434: for all $t\in[t_0,\infty)$, $t_0>0$ and all $x\in\mathbb T^1$.
435: \end{theorem}
436:
437: \begin{beweis}
438: Let
439: \begin{eqnarray*}
440: u&\mathop{=}\limits_{\rm Df}&t\kern1mm+\kern1mm x\\[2mm]
441: v&\mathop{=}\limits_{\rm Df}&t\kern1mm-\kern1mm x.
442: \end{eqnarray*}
443: In this coordinates~(\ref{Psi2}) reads
444: \begin{eqnarray}
445: \frac{\partial^2}{\partial u\kern1mm\partial v}\psi&=&
446: -\frac14\frac{\psi}{(u+v)^2}.\label{Psi3}
447: \end{eqnarray}
448: As a consequence of this transformation some parallelogram,
449: say~$\overline{1234}$
450: \ABB{x-t.eps}{10cm}
451: {%Integrationsgebiet in\\ \phantom{eine kleine Verschiebung}
452: %charakteristischen Koordinaten
453: }
454: \mbox{}\vskip-16mm
455: \begin{eqnarray*}
456: &&\kern45mm
457: \begin{array}{ll}
458: \left(t^{(1)}\equiv t_0,\right.&\left.x^{(1)}\equiv x_0=u_0-t_0
459: \right)\\[2mm]
460: \left(t^{(2)} = t_0,\right.&\left.x^{(2)}=x_0+2\pi
461: \right)\\[2mm]
462: \left(t^{(3)}= t_0+U,\right.&\left.x^{(3)}=x_0-U
463: \right)\\[2mm]
464: \left(t^{(4)}= t_0+U,\right.&\left.x^{(4)}=x_0+2\pi-U
465: \right)
466: \end{array}
467: \end{eqnarray*}
468: \vskip5mm
469:
470: has new vertices as follows:
471: \ABB{u-v.eps}{10cm}
472: {%Integrationsgebiet in\\ \phantom{eine kleine Verschiebung}
473: %charakteristischen Koordinaten
474: }
475: \mbox{}\vskip-16mm
476: \begin{eqnarray*}
477: &&\kern45mm
478: \begin{array}{ll}
479: \left(u^{(1)}\equiv u_0,\right.&\left.v^{(1)}\equiv v_0=2t_0-u_0
480: \right)\\[2mm]
481: \left(u^{(2)} = u_0+2\pi,\right.&\left.v^{(2)}=v_0-2\pi
482: \right)\\[2mm]
483: \left(u^{(3)}= u_0,\right.&\left.v^{(3)}=v_0+2U
484: \right)\\[2mm]
485: \left(u^{(4)}= u_0+2\pi,\right.&\left.v^{(4)}=v_0-2\pi+2U
486: \right)
487: \end{array}
488: \end{eqnarray*}
489: \vskip7mm
490:
491: The periodicity in $x$-space is the identity
492: \begin{eqnarray}
493: (u,v)&\equiv&(u+2\pi n,v-2\pi n)\label{Periode}
494: \end{eqnarray}
495: for all $n\in\mathbb Z$ and the over-all condition
496: $t>0$ converts to $u+v>0$. Both are generally assumed in what
497: follows.~--~The estimate
498: \begin{eqnarray*}
499: |\psi_u(u,v_2)\kern1mm-\kern1mm
500: \psi_u(u,v_1)|&\le&\intop_{v_1}^{v_2}\frac{|\psi|\kern1mm{\rm d}v}
501: {(u+v)^2}\kern2mm\mathop{\le}\limits^{(\ref{Schaetz})}\kern2mm
502: C_{t_0}\left(\frac1{u+v_1}\kern1mm-\kern1mm\frac1{u+v_2}\right)
503: \end{eqnarray*}
504: yields boundedness of $|\psi_u(u,v)|$ for any $u$ while
505: $v\to\infty$. But the same inequality also proves with
506: $v_2=v_1+\Delta$, $\Delta>0$, the existence of a finite
507: limit $F^\prime(u)$ for any $u$, hence
508: \begin{eqnarray}
509: F^\prime(u)&\mathop{=}\limits_{\rm Df}&
510: \psi_u(u,2t_0-u)+\lim_{U\to\infty}
511: \intop_{2t_0-u}^{2t_0-u+2U} \psi_{uv}(u,v)
512: \kern1mm {\rm d}v \label{FStr}
513: \end{eqnarray}
514: is well-defined. Periodicity of $F^\prime(u)$ is due
515: to~(\ref{Periode}) along with
516: \begin{eqnarray*}
517: F^\prime(u+2\pi)&=&
518: \psi_u(u+2\pi,2t_0-u-2\pi)-\frac14
519: \intop_{2t_0-u-2\pi}^\infty \frac{\psi(u+2\pi,v)}
520: {(u+2\pi+v)^2}
521: \kern1mm {\rm d}v\\[2mm]
522: &=&
523: \psi_u(u,2t_0-u)-\frac14
524: \intop_{2t_0-u}^\infty \frac{\psi(u+2\pi,\tilde v-2\pi)}
525: {(u+\tilde v)^2}
526: \kern1mm {\rm d}\tilde v\\
527: &=& F^\prime(u).
528: \end{eqnarray*}
529: Another property of $F^\prime(u)$ is its zero mean value.
530: To show this we transform the defining integral back to
531: $t$-$x$-coordinates according to the scetches from the beginning
532: of this proof. Let $(u_0,v_0)$, that is $(t_0,x_0)$,
533: $t_0>0$, an arbitrary point in characteristic or space-time
534: coordinates, respectively. Then
535: \begin{eqnarray}
536: \intop_{u_0}^{u_0+2\pi} F^\prime(u)\kern1mm {\rm d}u
537: &=& -\intop_{u_0}^{u_0+2\pi}
538: \lim\limits_{U\to\infty}\intop\limits_{2t_0-u}^{2t_0-u+2U}
539: \frac {\psi(u,v)}{4(u+v)^2}
540: \kern1mm {\rm d}v\kern1mm {\rm d}u \\
541: &=&\lim\limits_{U\to\infty}\intop_{t_0}^{t_0+U}\frac1{2t^2}
542: \intop\limits_{x_0+t_0-t}^{x_0+t_0-t+2\pi}
543: \psi(t,x)
544: \kern1mm {\rm d}x\kern1mm {\rm d}t\kern2mm=\kern2mm0,
545: \label{F-Periode}
546: \end{eqnarray}
547: since the term under the integral is zero for every~$t$.
548: A direct consequence is the periodicity of primitives. Let
549: \begin{eqnarray*}
550: \tilde F(u)&\mathop{=}\limits_{\rm Df}&\int F^\prime(u)\kern1mm
551: {\rm d}u
552: \end{eqnarray*}
553: such an (arbitrary but fixed) primitive. Then the function
554: \begin{eqnarray*}
555: F(u)&\mathop{=}\limits_{\rm Df}&\tilde F(u)
556: \kern1mm-\kern1mm\frac1{2\pi}\intop_{u_0}^{u_0+2\pi}
557: \tilde F(u)\kern1mm{\rm d}u
558: \end{eqnarray*}
559: is defined uniquely and especially
560: \begin{eqnarray}
561: \intop_{u_0}^{u_0+2\pi} F(u)\kern1mm
562: {\rm d}u&=&0 \label{FS-Periode}
563: \end{eqnarray}
564: is always satisfied.
565: Now, starting with the inequality
566: \begin{eqnarray*}
567: |\psi_v(u_2,v)\kern1mm-\kern1mm
568: \psi_v(u_1,v)|&\le&\intop_{u_1}^{u_2}\frac{|\psi|\kern1mm{\rm d}u}
569: {(u+v)^2}\kern2mm\le\kern2mm
570: C_{t_0}\left(\frac1{u_1+v}\kern1mm-\kern1mm\frac1{u_2+v}\right)
571: \end{eqnarray*}
572: we construct a likewise uniquely defined function
573: $G(v)$, where all steps follow as above
574: endowing $G$ with corresponding properties. Let us finally
575: introduce the functions:
576: \begin{eqnarray}
577: \nu(u,v)&\mathop{=}\limits_{\rm Df}&
578: F(u)\kern1mm+\kern1mm G(v)\label{Nue}\\[2mm]
579: \omega(u,v)&\mathop{=}\limits_{\rm Df}&
580: \psi(u,v)\kern1mm-\kern1mm \nu(u,v)
581: \end{eqnarray}
582: Obviously $\nu$ and $\omega$ are uniquely defined for any fixed
583: $\psi$. $\nu$ solves the free wave equation $\nu_{uv}=0$
584: and regarding the remainder~$\omega$ we estimate for every $u$, $v$
585: subject to $u+v>0$
586: \begin{eqnarray}
587: |\omega_u(u,v)|&=&|\psi_u(u,v)-\nu_u(u,v)|
588: \nonumber\\ [2mm]
589: &=&\left|\psi_u(u,v) - \psi_u(u,2t_0-u) -
590: \intop_{2t_0-u}^\infty\psi_{uv}(u,v)\kern1mm{\rm d}v\right|
591: \nonumber\\[2mm]
592: &=&\left|-\intop^{2t_0-u}_v
593: \psi_{u\tilde v}(u,\tilde v)\kern1mm
594: {\rm d}\tilde v\kern1mm-\kern1mm
595: \intop_{2t_0-u}^\infty\psi_{uv}(u,v)\kern1mm{\rm d}v\right|
596: \nonumber\\ [2mm]
597: &=&
598: \left|\frac14\intop_v^\infty
599: \frac{\psi(u,\tilde v)}{(u+\tilde v)^2}\kern1mm{\rm d}
600: \tilde v\right|\kern2mm\le\kern2mm\frac{C_{t_0}}{4(u+v)},
601: \label{Itom}
602: \end{eqnarray}
603: where the estimate follows from~(\ref{Schaetz}). In exactly the same
604: way one proves
605: \begin{eqnarray*}
606: |\omega_v(u,v)|&\le&\frac{C_{t_0}}{4(u+v)}.
607: \end{eqnarray*}
608: Now, since
609: \begin{eqnarray}
610: |\omega_t(t,x)|&=&
611: |\omega_u(u,v)+\omega_v(u,v)|
612: \kern2mm\le\kern2mm\frac {C_{t_0}}{2(u+v)}\kern2mm=\kern2mm
613: \frac{C_{t_0}}{4t}\label{Itom2}\\[2mm]
614: |\omega_x(t,x)|&=&
615: |\omega_u(u,v)-\omega_v(u,v)|
616: \kern2mm\le\kern2mm\frac {C_{t_0}}{2(u+v)}\kern2mm=\kern2mm
617: \frac{C_{t_0}}{4t}\label{Itom1}
618: \end{eqnarray}
619: for all $t\in[t_0,\infty)$, $x\in\mathbb T^1$ we have
620: \begin{eqnarray*}
621: ||\omega_x(t,.)||_{L^2}^2&=&
622: \intop_0^{2\pi}
623: {\omega_x}^2\kern1mm{\rm d}x\kern2mm\le\kern2mm
624: \frac{\pi {C_{t_0}}^2}{8t^2}.
625: \end{eqnarray*}
626: Corollary~\ref{om-Eigen} below proves the properties
627: of~$\omega$
628: \begin{eqnarray*}
629: \omega(t,x+2\pi)&=&\omega(t,x) \\
630: \intop_0^{2\pi}\omega(t,x)\kern1mm{\rm d}x&=&0
631: \end{eqnarray*}
632: for every $(t,x)\in\mathbb R_+\times\mathbb T^1$. Hence all
633: assumptions of lemma~\ref{Hilfssatz2} are satisfied which
634: ensures the existence of some constant
635: $C\mathop{=}\limits_{\rm Df}\frac\pi2 C_{t_0}>0$ with
636: \begin{eqnarray}
637: |\omega(t,x)|&\le& \frac Ct\label{A6}
638: \end{eqnarray}
639: for all $t\in[t_0,\infty)$, $t_0>0$ and all $x\in\mathbb T^1$.
640: \end{beweis}
641:
642: The existence and uniqueness in the definition of function $\nu$
643: allow us to introduce the following term.
644:
645: \begin{definition}
646: The function $\nu$ is called the $\psi$-associated solution
647: of the free wave equation, if~$\psi\in\mathcal C^2
648: (\mathbb R_+\times\mathbb T^1;\mathbb R)$ solves
649: \begin{eqnarray*}
650: \frac{\partial^2}{\partial t^2}\psi\kern1mm-\kern1mm
651: \frac{\partial^2}{\partial x^2}\psi&=&
652: -\frac1{4 t^2}\psi
653: \end{eqnarray*}
654: with the side condition
655: \begin{eqnarray*}
656: \intop_0^{2\pi}\psi(t,x)\kern1mm{\rm d}x&=&0
657: \end{eqnarray*}
658: for every $t\in[t_0,\infty)$ and $\nu\in\mathcal C^2
659: (\mathbb R_+\times\mathbb T^1;\mathbb R)$ is defined as in
660: equation~(\ref{Nue}).
661: \end{definition}
662:
663: Here follows a series of corollaries establishing properties
664: of~$\nu$ as well as~$\omega$ which were used in the previous
665: proof or are of frequent use in the sections below.
666:
667: \begin{corollary}
668: Let $\nu$ be the $\psi$-associated solution of the free
669: wave equation and $t_0>0$. Then there exists a positive constant
670: $C$, such that for all $t\in[t_0,\infty)$ and $x\in\mathbb T^1$
671: the functions
672: $\nu$, $\nu_t$ and $\nu_x$ are bounded:\label{nu-Eigen}
673: \begin{eqnarray}
674: \mathop{\mathop{\rm max}\limits_{t\ge t_0>0}}\limits_{x\in\mathbb T^1}
675: \kern1mm
676: \left\{|\nu(t,x)|,|\nu_t(t,x)|,|\nu_x(t,x)|\right\}&<&C \label{nuE1}
677: \end{eqnarray}
678: Further, for all $n\in\mathbb Z$ are:
679: \begin{eqnarray}
680: \nu(t,x+2\pi n)&=&\nu(t,x)
681: \kern2mm=\kern2mm\nu(t+2\pi n,x) \label{nuE2} \\[2mm]
682: \nu_t(t+2\pi n,x)&=&\nu_t(t,x) \label{nuE4} \\[2mm]
683: \nu_x(t+2\pi n,x)&=&\nu_x(t,x) \label{nuE5} \\[2mm]
684: \intop_0^{2\pi}\nu(t,x)\kern1mm{\rm d}x&=&0 \label{nuE3}
685: \end{eqnarray}
686: \end{corollary}
687:
688: \begin{beweis}
689:
690: (\ref{nuE1})\kern3mm
691: By definition it is $\nu(u,v)=F(u)+G(v)$. $F$ is periodic with vanishing
692: mean value. W.l.o.g.\ is $F(u_0)=0$,
693: $\psi_u(u,v_\infty)\kern1mm\mathop{=}\limits_{\rm Df}\kern1mm
694: \mathop{\rm lim}\limits_{v\to\infty}\psi_u(u,v)$. Then
695: \begin{eqnarray*}
696: F(u)&=&\intop_{u_0}^{u}F^\prime(\tilde u)\kern1mm{\rm d}\tilde u
697: \kern2mm=\kern2mm
698: \intop_{u_0}^{u}\psi_{\tilde u}(\tilde u,v_\infty)
699: \kern1mm{\rm d}\tilde u
700: \kern2mm=\kern2mm\psi(u,v_\infty)\kern1mm-\kern1mm\psi(u_0,v_\infty).
701: \end{eqnarray*}
702: Theorem~\ref{Satz4} shows boundedness of $\psi$
703: (uniformly in the considered domain).
704: The same is true for $G$. Finally, boundedness of $|\nu_t(t,x)|$
705: and $|\nu_x(t,x)|$ is a consequence of
706: $|F^\prime(u)|$ and $|G^\prime(v)|$ boundedness established
707: during the proof of theorem~\ref{Satz5}.
708:
709: (\ref{nuE2})\kern3mm
710: With~(\ref{FS-Periode}) we have
711: \begin{eqnarray*}
712: \nu(t+2\pi n,x)&=&F(t+2\pi n+x)\kern1mm+\kern1mm G(t+2\pi n-x)\\
713: &=&F(t+x)\kern1mm+\kern1mmG(t-x)
714: \kern2mm=\kern2mm\nu(t,x)
715: \end{eqnarray*}
716: and analogously
717: \begin{eqnarray*}
718: \nu(t,x+2\pi n)&=&F(t+x+2\pi n)\kern1mm+\kern1mm G(t-x-2\pi n)\\
719: &=&\nu(t,x).
720: \end{eqnarray*}
721:
722: (\ref{nuE4})\kern3mm
723: Is a direct consequence of (\ref{nuE2}).
724:
725: (\ref{nuE5})\kern3mm
726: Follows directly from~(\ref{nuE2}).
727:
728: (\ref{nuE3})\kern3mm
729: Property~(\ref{FS-Periode}) and correspondingly for~$G$ yield:
730: \begin{eqnarray*}
731: \intop_0^{2\pi}
732: \nu(t,x)\kern1mm{\rm d}x&=&
733: \intop_t^{t+2\pi}
734: \nu(u,2t-u)\kern1mm{\rm d}u\\
735: &=&
736: \intop_t^{t+2\pi}\left(
737: F(u)\kern1mm+\kern1mm G(2t-u)\right)
738: \kern1mm{\rm d}u\\
739: &=&
740: \intop_t^{t+2\pi}
741: F(u)\kern1mm{\rm d}u
742: \kern1mm+\kern1mm
743: \intop_{t-2\pi}^{t} G(v)
744: \kern1mm{\rm d}v\kern2mm=\kern2mm0
745: \end{eqnarray*}
746: \end{beweis}
747:
748: \begin{corollary}
749: Let $\nu$ be the $\psi$-associated solution of the free wave
750: equation. Then the following are equivalent:\label{nu-Aequ}
751: \begin{eqnarray}
752: \psi(t,x)&\equiv&0 \label{nuA1} \\[2mm]
753: F(t+x)&\equiv&0 \kern5mm\text{and}\kern5mm
754: G(t-x)\kern2mm\equiv\kern2mm0 \label{nuA2} \\[2mm]
755: \nu(t,x)&\equiv&0 \label{nuA3}
756: \end{eqnarray}
757: \end{corollary}
758:
759: \begin{beweis}
760:
761: (\ref{nuA1}) $\Rightarrow$ (\ref{nuA2}):\kern3mm
762: $F$ is constant due to~(\ref{FStr}),
763: from~(\ref{FS-Periode}) follows the statement with respect to~$F$;
764: analogously for $G$.
765:
766: (\ref{nuA2}) $\Leftrightarrow$ (\ref{nuA3}):\kern3mm
767: ($\Rightarrow$) is due
768: to~(\ref{Nue}) and ($\Leftarrow$)~due to~(\ref{Nue})
769: and~(\ref{FS-Periode}).
770:
771: (\ref{nuA3}) $\Rightarrow$ (\ref{nuA1}):\kern3mm
772: It is $\psi(t,x)=\omega(t,x)$. The estimate~(\ref{Itom1})
773: reads here in a first step
774: \begin{eqnarray*}
775: |\psi_x(t,x)|_1&\le&\frac{C_{t_0}}{4t}
776: \end{eqnarray*}
777: and according to lemma~\ref{Hilfssatz2}
778: \begin{eqnarray*}
779: |\psi(t,x)|_1&\le&\frac{\pi C_{t_0}}{2t}.
780: \end{eqnarray*}
781: Now, this improved estimate can be used iteratively in~(\ref{Itom}),
782: leading to an $n$-th order estimate
783: \begin{eqnarray*}
784: |\psi(t,x)|_n&\le&\left(\frac\pi 2\right)^n \frac{C_{t_0}}{n!}
785: t^{-n}
786: \end{eqnarray*}
787: for every $x\in\mathbb T^1$ and every $t\in[t_0,\infty)$.
788: Due to~(\ref{Itom2}) the same estimate is valid for $\psi_t$.
789: Consequently exists to any
790: $\varepsilon>0$ an $n_0$, such that
791: \begin{eqnarray*}
792: |\psi(t_0,x)|_n&<&\varepsilon\\
793: |\psi_t(t_0,x)|_n&<&\varepsilon
794: \end{eqnarray*}
795: for all $x\in\mathbb T^1$ and all $n\ge n_0$, that is
796: $\psi(t_0,x)\kern1mm=\kern1mm\psi_t(t_0,x)\kern1mm=\kern1mm0$.
797: Due to the uniqueness of its solutions
798: $\psi$ is the trivial solution of~(\ref{Psi2}).
799: \end{beweis}
800:
801: \begin{corollary}
802: Let $\nu$ be the $\psi$-associated solution of the free wave
803: equation, $\omega\mathop{=}\limits_{\rm Df}\psi-\nu$
804: and $t_0>0$. Then there exists a positive constant
805: $C$, such that for all $t\in[t_0,\infty)$, $x\in\mathbb T^1$
806: and $n\in\mathbb Z$:\label{om-Eigen}
807: \begin{eqnarray}
808: |\omega(t,x)|&<&C \label{omE1} \\[2mm]
809: \omega(t,x+2\pi n)&=&\omega(t,x) \label{omE2} \\[2mm]
810: \intop_0^{2\pi}\omega(t,x)\kern1mm{\rm d}x
811: &=&0 \label{omE3}
812: \end{eqnarray}
813: \end{corollary}
814:
815: \begin{beweis}
816: All properties are direct consequences from the definition
817: of~$\omega$, the corresponding properties of~$\psi$ along
818: with corollary~\ref{nu-Eigen} above.
819: \end{beweis}
820:
821: \begin{corollary}
822: Let $\nu$ be the $\psi$-associated solution of the free wave
823: equation and $\omega\mathop{=}\limits_{\rm Df}\psi-\nu$.
824: Then the following is equivalent:\label{om-Aequ}
825: \begin{eqnarray}
826: \psi(t,x)&\equiv&0 \label{omA1} \\[2mm]
827: \omega(t,x)&\equiv&0 \label{omA2}
828: \end{eqnarray}
829: \end{corollary}
830:
831: \begin{beweis}
832:
833: (\ref{omA1}) $\Rightarrow$ (\ref{omA2}):\kern3mm
834: With corollary~\ref{nu-Aequ} is $\nu\equiv 0$.
835:
836: (\ref{omA2}) $\Rightarrow$ (\ref{omA1}):\kern3mm
837: It is $\psi\equiv\nu$. Since
838: \begin{eqnarray*}
839: \frac{\partial^2}{\partial t^2}\psi\kern1mm-\kern1mm
840: \frac{\partial^2}{\partial x^2}\psi&=&-\frac1{4t^2}\psi
841: \kern3mm\mathop{\equiv}\limits^{!}\kern3mm0
842: \end{eqnarray*}
843: the statement follows.
844: \end{beweis}
845:
846: Now we summarize the most important features in the
847: asymptotic behaviour of the metric function~$W$.
848: Here we use as remainder
849: $\varkappa\mathop{=}\limits_{\rm Df}t^{-\frac12}\omega$.
850:
851: \begin{corollary}\label{Folgerung6}%
852: Let $W\in\mathcal C^2(\mathbb R_+\times \mathbb T^1)$
853: be a real-valued solution of the Euler-Poisson-Darboux
854: equation
855: \begin{eqnarray*}
856: \frac{\partial^2}{\partial t^2}W(t,x)\kern1mm+\kern1mm
857: \frac1t
858: \frac{\partial}{\partial t}W(t,x)
859: \kern1mm-\kern1mm
860: \frac{\partial^2}{\partial x^2}W(t,x)&=&0.
861: \end{eqnarray*}
862: Then there exists uniquely defined, bounded, real-valued
863: functions $\nu\in\mathcal C^2(\mathbb R_+\times\mathbb T^1)$
864: and $\varkappa\in\mathcal C^2(\mathbb R_+\times\mathbb T^1)$,
865: constants $\beta$, $\gamma$, and a positive constant $C_{t_0}$
866: such that
867: \begin{eqnarray}
868: W(t,x)&=&\gamma\kern1mm+\kern1mm\beta\cdot\ln t
869: \kern1mm+\kern1mm t^{-\frac12}\nu(t,x)\kern1mm+\kern1mm
870: \varkappa(t,x)\label{A1}\\
871: \frac{\partial^2}{\partial t^2}\nu(t,x)&=&
872: \frac{\partial^2}{\partial x^2}\nu(t,x)\label{A2}
873: \end{eqnarray}
874: and
875: \begin{eqnarray}
876: |\varkappa(t,x)|&\le&C_{t_0}\cdot t^{-\frac32}\label{A3}\\
877: |\varkappa_t(t,x)|&\le&C_{t_0}\cdot t^{-\frac32}\label{A4}\\
878: |\varkappa_x(t,x)|&\le&C_{t_0}\cdot t^{-\frac32}\label{A5}
879: \end{eqnarray}
880: for all $t\ge t_0>0$ and all $x\in\mathbb T^1$.
881: \end{corollary}
882: \begin{proof}
883: We have already shown (\ref{A1}) and (\ref{A2}), explicitly. But on the
884: other hand estimates (\ref{A3}), (\ref{A4}) and (\ref{A5}) follow directly
885: from the above definition of~$\varkappa$ together with (\ref{A6}),
886: (\ref{Itom2}) and (\ref{Itom1}), respectively.
887: \end{proof}
888:
889: \section{On the integration of constraints}
890:
891: As mentioned in the beginning part of this paper a special
892: feature of Gowdy's model is the decoupling of the dynamical
893: quantity~$a$. Furthermore, it is well-known that the
894: momentum constraint
895: \begin{eqnarray}
896: \frac\partial{\partial x}a&=&\frac12\kern.5mm t \kern.5mm
897: \frac\partial{\partial t}W\kern.5mm
898: \frac\partial{\partial x}W \label{Impuls}
899: \end{eqnarray}
900: is conserved along $t$ developments; that means it is satisfied
901: always if it is satisfied initially.
902:
903: With this remark in mind we will investigate now the
904: Hamiltonian constraint
905: \begin{eqnarray}
906: \frac\partial{\partial t}a&=&-\frac1{4t}\kern1mm+
907: \kern1mm\frac14\kern.5mmt
908: \left[\left(\frac\partial{\partial t}W\right)^2
909: \kern1mm+\kern1mm
910: \left(\frac\partial{\partial x}W\right)^2\right]
911: \label{Hamilton}
912: \end{eqnarray}
913: exclusively.
914:
915: \subsection*{Spatially homogeneous spacetimes}
916: According to (\ref{Whom}) we find as the general solution
917: of the Euler-Poisson-Darboux equation in spatially homogeneous
918: spacetimes
919: \begin{eqnarray*}
920: W(t)&=&\beta\cdot\ln t\kern1mm+\kern1mm\gamma.
921: \end{eqnarray*}
922: The constraint (\ref{Hamilton}) reads
923: \begin{eqnarray*}
924: a_t(t)&=&\frac14(\beta^2-1)\cdot t^{-1},
925: \end{eqnarray*}
926: leading to
927: \begin{eqnarray}
928: a(t)&=&\frac14(\beta^2-1)\cdot\ln t\kern1mm+\kern1mm
929: \zeta \label{ahom}
930: \end{eqnarray}
931: as solution where the arbitrary constant~$\zeta$ has to be
932: specialized by an initial condition for $a$.
933:
934: \subsection*{Not spatially homogeneous spacetimes}
935: The exceptional position of spatially homogeneous among
936: the polarized Gowdy spacetimes is a consequence of the
937: following consideration.
938:
939: \begin{theorem}
940: Let $W\in\mathcal C^2(\mathbb R_+\times\mathbb T^1;\mathbb R)$
941: be a solution of the Euler-Poisson-Darboux equation.
942: Then there is a positive constant~$C$, such that
943: \begin{eqnarray*}
944: |a_t(t,x)|&\le&C
945: \end{eqnarray*}
946: for all $t\in[t_0,\infty)$ and all $x\in\mathbb T^1$.
947: \end{theorem}
948:
949: \begin{beweis}
950: With corollary~\ref{Folgerung6} we have
951: \begin{eqnarray*}
952: W(t,x)&=&\gamma\kern1mm+\kern1mm\beta\cdot\ln t\kern1mm+\kern1mm
953: t^{-\frac12}\nu(t,x)\kern1mm+\kern1mm\varkappa(t,x)
954: \end{eqnarray*}
955: with corresponding properties of $\nu$ and estimates for the
956: remainder~$\varkappa$; therefore
957: \begin{eqnarray}
958: a_t&=&\frac14({\nu_t}^2+{\nu_x}^2)\kern1mm+\kern1mm\frac12
959: \kern.5mm t^{-\frac12}\kern.5mm\beta\kern.5mm\nu_t \nonumber\\
960: &&+\kern1mm\frac14 \kern.5mmt^{-1}\kern.5mm
961: \left(\beta^2-1+2\nu_t\left(t^\frac32\varkappa_t-\frac12\nu\right)
962: +2\nu_x t^\frac32\varkappa_x\right) \nonumber\\
963: &&+\kern1mm\mathcal O(t^{-\frac32}). \label{at}
964: \end{eqnarray}
965: The statement follows from corollaries~\ref{nu-Eigen}
966: and~\ref{Folgerung6}.
967: \end{beweis}
968:
969: \begin{remark}
970: The above equation~(\ref{at}) already reflects the qualitatively
971: different asymptotic behaviour in the metric function~$a$ depending on the
972: spatial homogenity or not spatial homogenity of the underlying
973: model. This is so because corollary~\ref{nu-Aequ} along with
974: corollary~\ref{om-Aequ}
975: states that a polarized Gowdy spacetime is spatially
976: homogeneous if and only if
977: \begin{eqnarray*}
978: \psi\kern1mm\equiv\kern1mm0&\leftrightarrow&
979: \nu\kern1mm\equiv\kern1mm0\kern2mm\leftrightarrow\kern2mm
980: \varkappa\kern1mm\equiv\kern1mm0.
981: \end{eqnarray*}
982: \end{remark}
983:
984: Since the asymptotics in spatially homogeneous models has been
985: completely investigated we will in what follows w.l.o.g.\
986: assume that
987: $\nu_t\kern1mm=\kern1mm\nu_x\kern1mm=\kern1mm0$ is valid
988: not everywhere.
989:
990: \begin{theorem}
991: Let $a\in\mathcal C^2(\mathbb R_+\times\mathbb T^1;\mathbb R)$
992: be a solution of equation~(\ref{Hamilton}) and
993: $\partial_x a\kern1mm=\kern1mm0$ not everywhere. Then there
994: exist positive constants $\bar\nu$ and $C$ along with a uniquely
995: determined function $\delta\in\mathcal C^2(\mathbb R_+\times
996: \mathbb T^1;\mathbb R)$ such that
997: \begin{eqnarray*}
998: a(t,x)&=&\bar\nu\cdot t\kern1mm+\kern1mm\delta(t,x)
999: \end{eqnarray*}
1000: with
1001: \begin{eqnarray*}
1002: |\delta(t,x)|&\le&C\cdot (1+t^\frac12)
1003: \end{eqnarray*}
1004: for every $t\in[t_0,\infty)$, $t_0>0$ and every $x\in\mathbb T^1$.
1005: \end{theorem}
1006:
1007: \begin{beweis}
1008: Since $a$ is continously differentiable, there is in
1009: $\mathbb R_+\times\mathbb T^1$ an open domain with
1010: $a_x\kern1mm\neq\kern1mm0$ hence $\nu_x\kern1mm\neq\kern1mm0$.
1011: Consequently
1012: \begin{eqnarray}
1013: \bar\nu&\mathop{=}\limits_{\rm Df}&\frac1{8\pi}
1014: \intop_{t_0}^{t_0+2\pi}({\nu_t}^2+{\nu_x}^2)\kern1mm{\rm d}t
1015: \label{nuq}
1016: \end{eqnarray}
1017: is a positive constant since it is also $x$-independent due to
1018: $\nu_{tt}\kern1mm=\kern1mm\nu_{xx}$ and
1019: \begin{eqnarray*}
1020: \frac{\rm d}{{\rm d}x}\bar\nu&=&\frac1{4\pi}
1021: \int_{t_0}^{t_0+2\pi}(\nu_t\nu_{tx}+\nu_x\nu_{xx})\kern1mm
1022: {\rm d}t\\
1023: &=&\frac1{4\pi}\int_{t_0}^{t_0+2\pi}\nu_x(-\nu_{tt}+\nu_{xx})
1024: \kern1mm{\rm d}t\kern1mm+\kern1mm\left.\frac1{4\pi}\kern.5mm\nu_t
1025: \kern.5mm\nu_x\kern.5mm\right|_{t_0}^{t_0+2\pi}\kern2mm=\kern2mm0.
1026: \end{eqnarray*}
1027: Let be further
1028: \begin{eqnarray*}
1029: \delta_t(t,x)&\mathop{=}\limits_{\rm Df}&
1030: -\bar\nu\kern1mm+\kern1mm a_t(t,x)\\
1031: &=&-\bar\nu\kern1mm+\kern1mm\frac14({\nu_t}^2+{\nu_x}^2)\\
1032: &&+\kern1mm\frac12
1033: \kern.5mm t^{-\frac12}\kern.5mm\beta\kern.5mm\nu_t \\
1034: &&+\kern1mm\frac14 \kern.5mmt^{-1}\kern.5mm
1035: \left(\beta^2-1\kern.5mm+\kern.5mm
1036: 2\nu_t\left(t^\frac32\varkappa_t-\frac12\nu\right)
1037: \kern.5mm+\kern.5mm2\nu_x t^\frac32\varkappa_x\right) \\
1038: &&+\kern1mm\frac12\kern.5mmt^{-\frac32}\kern.5mm\beta\kern.5mm
1039: \left(t^\frac32\varkappa_t\kern.5mm-\kern.5mm\frac12\nu\right) \\
1040: &&+\kern1mm\frac14\kern.5mmt^{-2}\left[\left(t^\frac32\varkappa_t-
1041: \frac12\nu\right)^2\kern.5mm+\kern.5mm\left(t^\frac32\varkappa_x
1042: \right)^2\right]
1043: \end{eqnarray*}
1044: which is uniquely determined for every fixed solution $W$
1045: of the Euler-Poisson-Darboux equation. So we have
1046: \begin{eqnarray*}
1047: \delta(t,x)&=&\delta(t_0,x)\kern1mm+\kern1mm\frac14\int_{t_0}^t
1048: \left({\nu_\tau}^2\kern.5mm+\kern.5mm{\nu_x}^2\right)
1049: \kern1mm{\rm d}\tau\kern1mm-\kern1mm\bar\nu\cdot(t-t_0)
1050: \kern1mm+\kern1mm\mathcal O(t^\frac12).
1051: \end{eqnarray*}
1052: Now we consider the second and third term of the
1053: right hand side using~(\ref{nuq}) in detail:
1054: \begin{eqnarray*}
1055: \lefteqn{\mbox{}\kern-12mm\frac14\int_{t_0}^t\left({\nu_\tau}^2
1056: \kern.5mm+\kern.5mm
1057: {\nu_x}^2\right)\kern1mm{\rm d}\tau\kern1mm-\kern1mm
1058: \bar\nu\cdot(t-t_0)} \\[5mm]
1059: \mbox{}\kern12mm&=&
1060: \int_{t_0}^t\left[\frac14({\nu_\tau}^2+{\nu_x}^2)-\frac1{8\pi}
1061: \intop_{t_0}^{t_0+2\pi}({\nu_t}^2+{\nu_x}^2)\kern1mm{\rm d}t
1062: \right]\kern1mm{\rm d}\tau \\[5mm]
1063: \mbox{}\kern12mm&=&
1064: \frac14\int_{t_0}^{t_0+2\pi\left[\frac{t-t_0}{2\pi}\right]}
1065: ({\nu_t}^2+{\nu_x}^2)\kern1mm
1066: {\rm d}t\kern1mm-\kern1mm\frac{\left[\frac{t-t_0}{2\pi}\right]}{4}
1067: \intop_{t_0}^{t_0+2\pi}({\nu_t}^2+{\nu_x}^2)\kern1mm{\rm d}t\\[2mm]
1068: \mbox{}\kern12mm&&+
1069: \int_{t_0+2\pi\left[\frac{t-t_0}{2\pi}\right]}^t
1070: \left[\frac14({\nu_\tau}^2+{\nu_x}^2)
1071: \kern1mm-\kern1mm\frac1{8\pi}
1072: \intop_{t_0}^{t_0+2\pi}({\nu_t}^2+{\nu_x}^2)\kern1mm{\rm d}t
1073: \right]\kern1mm{\rm d}\tau
1074: \end{eqnarray*}
1075: Here as usual $[n]$ is the largest integer not
1076: greater then $n$. Now, due to periodicity the first two terms cancel
1077: each other. The integrand of the third is bounded, the domain
1078: of integration less then $2\pi$ so it is
1079: \begin{eqnarray*}
1080: \left|\intop_{t_0}^t\left[\frac14({\nu_\tau}^2+{\nu_x}^2)
1081: \kern1mm-\kern1mm\bar\nu
1082: \right]\kern1mm{\rm d}\tau\right|&\le&2\pi C
1083: \end{eqnarray*}
1084: and finally
1085: \begin{eqnarray*}
1086: |\delta(t,x)|&\le&C\cdot(1+t^\frac12)
1087: \end{eqnarray*}
1088: for a positive constant $C$, all $x\in\mathbb T^1$
1089: and every~$t\in[t_0,\infty)$, $t_0>0$.
1090: \end{beweis}
1091:
1092: All results concerning the integration of constrains are
1093: summarized by the following theorem.
1094:
1095: \begin{theorem}\label{Satz7}%
1096: Let $W\in\mathcal C^2(\mathbb R_+\times\mathbb T^1;\mathbb R)$
1097: be a solution of $W_{tt}+t^{-1}W_t-W_{xx}=0$.
1098: Then any solution
1099: $a\in\mathcal C^2(\mathbb R_+\times\mathbb T^1;\mathbb R)$
1100: of
1101: \begin{eqnarray*}
1102: a_t&=&-\frac14t^{-1}\kern1mm+\kern1mm\frac14t\kern.5mm
1103: ({W_t}^2+{W_x}^2)
1104: \end{eqnarray*}
1105: may be cast for all $t\ge t_0>0$ and all $x\in\mathbb T^1$
1106: into the form
1107: \begin{eqnarray}
1108: a(t,x)&=&\left\{
1109: \begin{array}{ll}
1110: \frac14(\beta^2-1)\cdot\ln t\kern1mm+\kern1mm\zeta,&
1111: \text{if }\partial_x a(t,x)\equiv0\\[2mm]
1112: \bar\nu\cdot t\kern1mm+\kern1mm\delta(t,x),&\text{otherwise}
1113: \end{array}
1114: \right. \label{aasym}
1115: \end{eqnarray}
1116: where $\bar\nu$ is a positive, $\zeta$ an arbitrary constant
1117: and $\delta$ satisfies the inequalities
1118: \begin{eqnarray*}
1119: |\delta(t,x)|&\le&C\cdot (1+t^\frac12)\\[2mm]
1120: |\delta_t(t,x)|&\le&C
1121: \end{eqnarray*}
1122: with another positive constante~$C$.
1123: \end{theorem}
1124:
1125: \section{On geodesic completeness in polarized Gowdy spacetimes}
1126:
1127: In contrast to the analytical treatments in the first part
1128: of this paper we will now investigate some open
1129: geometrical questions in polarized Gowdy spacetimes.
1130:
1131: Isenberg and Moncrief~(1990) established important results
1132: with respect to the singular behaviour in this class of
1133: spacetimes. However, the question concerning possible
1134: singularities in its future development remained open
1135: (at least for the case with spatial $\mathbb T^3$-topology) and
1136: shall be the subject of this section. As usual we will
1137: identify the term ``singularity'' with the term
1138: ``existence of an uncomplete non-spacelike geodesic''
1139: and use the basic concepts and relations of the theory
1140: of Lorentzian manifolds and causality without repeating
1141: all its definitions. However, for convenience we will summarize
1142: some facts for short.
1143:
1144: Future will always mean the expanding direction
1145: of the universe.
1146:
1147: In local coordinates
1148: $(x^0,x^1,x^2,x^3)$ the geodesic equation reads:
1149: \begin{eqnarray}
1150: \frac{{\rm d}^2 x^i}{{\rm d}\tau^2}\kern1mm+\kern1mm
1151: \sum_{j,k}\kern.5mm\Gamma^i_{jk}\kern.5mm
1152: \frac{{\rm d} x^j}{{\rm d}\tau}\frac{{\rm d} x^k}{{\rm d}\tau}
1153: &=&0,\kern10mmi,j,k=0,\dots,3\label{Geo}
1154: \end{eqnarray}
1155: Among the properties of affinly parametrized geodesics $\mathfrak x$
1156: the following are of special importance:
1157:
1158: 1.) The tangent vector is normalized
1159: \begin{eqnarray}
1160: \sum_{i,j}\kern.5mmg_{ij}\kern.5mm\frac{{\rm d}x^i}{{\rm d}\tau}
1161: \kern.5mm\frac{{\rm d}x^j}{{\rm d}\tau}&=&-\varepsilon,
1162: \label{Norm}
1163: \end{eqnarray}
1164: where $\varepsilon=0$, if the geodesic is null and
1165: $\varepsilon=1$, if it is timelike.
1166:
1167: 2.) The scalar product of any Killing vectorfield $X$ and the
1168: geodesic tangent is constant along the curve; in coordinates:
1169: \begin{eqnarray}
1170: \frac{{\rm d}}{{\rm d}\tau}\kern.5mm\sum_{i,j}\kern.5mmg_{ij}
1171: \kern.5mm{X_\mathfrak x}^i\kern.5mm\frac{{\rm d}x^j}{{\rm d}\tau}
1172: &=&0
1173: \label{KVF}
1174: \end{eqnarray}
1175:
1176: Gowdy spacetimes are known to be maximally extended, globally
1177: hyperbolic regions which permit foliations by
1178: $(t={\rm const})$-hypersurfaces. Consequently, any nonspacelike
1179: curve intersects every such hypersurface exactly once.
1180: This is important in view of the possibility to take
1181: $t\to\infty$ limits along an arbitrary nonspacelike curve.
1182: Furthermore, according to our choise of orientation it is
1183: \begin{eqnarray}
1184: \frac{{\rm d}t}{{\rm d}\tau}&>&0 \label{Orient}
1185: \end{eqnarray}
1186: if $\tau$ is as before the parameter of the future directed curve.
1187:
1188: Now we can start with the investigations. Since the asymptotic
1189: expansion of polarized Gowdy spacetimes is quite different depending
1190: on the spatial homogenity/inhomogenity of the model it is
1191: necessary to split the proofs accordingly.
1192:
1193: \subsection*{Geodesics in spatially homogeneous, polarized
1194: spacetimes}
1195: Spatially homogeneous, polarized spacetimes correspond to
1196: exactly solvable Einstein equations. We had found~(\ref{Whom})
1197: and~(\ref{ahom}),
1198: \begin{eqnarray}
1199: W(t)&=&\gamma\kern1mm+\kern1mm\beta\cdot\ln t\label{Whom1}\\
1200: a(t)&=&\zeta\kern1mm+\kern1mm\frac14(\beta^2-1)\cdot\ln t,
1201: \label{ahom1}
1202: \end{eqnarray}
1203: with constants $\beta$, $\gamma$, $\zeta$. We need the
1204: following estimate to prove geodesic completeness in such
1205: models.
1206:
1207: \begin{lemma}
1208: Let $\left(\mathbb R_+\times\mathbb T^3,g={\rm diag}(g_{00},g_{11},g_{22},g_{33})\right)$
1209: be a spatially homogeneous solution of Einstein's field equation
1210: for the polarized Gowdy model. Then there exist a constant~$C_\beta$,
1211: such that the Christoffel symbols satisfy for every
1212: $t\in[t_0,\infty)$:\label{Voll1}
1213: \begin{eqnarray*}
1214: \Gamma^0_{11}(t)&>&-\frac12 \kern1mm g^{00}\kern1mm
1215: C_\beta\kern1mm t^{-1}\cdot g_{11} \\ [2mm]
1216: \Gamma^0_{22}(t)&>&-\frac12 \kern1mm g^{00}\kern1mm
1217: C_\beta\kern1mm t^{-1}\cdot g_{22} \\ [2mm]
1218: \Gamma^0_{33}(t)&>& -\frac12\kern1mm g^{00}\kern1mm
1219: C_\beta\kern1mm t^{-1}\cdot g_{33}
1220: \end{eqnarray*}
1221: \end{lemma}
1222:
1223: \begin{beweis}
1224: The following section~\ref{Symbole} contains the relevant
1225: Christoffel symbols.
1226: Using~(\ref{Whom1}) and~(\ref{ahom1}) we find:
1227: \begin{eqnarray*}
1228: \Gamma^0_{11}(t)&=&-\frac12g^{00}\partial_t g_{11}
1229: \kern2mm=\kern2mm-\frac12g^{00}\cdot\frac12(\beta^2-1)\kern1mm
1230: t^{-1}\cdot g_{11}\\[2mm]
1231: \Gamma^0_{22}(t)&=&-\frac12g^{00}\partial_t g_{22}
1232: \kern2mm=\kern2mm-\frac12g^{00}\cdot\phantom{\frac12^2}
1233: (\beta+1)\kern1mm
1234: t^{-1}\cdot g_{22}\\ [2mm]
1235: \Gamma^0_{33}(t)&=&-\frac12g^{00}\partial_t g_{33}
1236: \kern2mm=\kern2mm-\frac12g^{00}\cdot\phantom{\frac12}
1237: (-\beta+1)\kern1mm
1238: t^{-1}\cdot g_{33}
1239: \end{eqnarray*}
1240: The constant $C_\beta$,
1241: \begin{eqnarray}
1242: C_\beta&\mathop{=}\limits_{\rm Df}&-\frac12(\beta^2+3),\label{Cbeta}
1243: \end{eqnarray}
1244: proves the statement. Observe that~$C_\beta<0$.
1245: \end{beweis}
1246:
1247: \begin{theorem}
1248: Every inextendable, future directed, nonspacelike geodesic in an
1249: arbitrary but spatially homogeneous, polarized Gowdy manifold
1250: is in its future direction complete.
1251: \end{theorem}
1252:
1253: \begin{beweis}
1254: It is sufficient to regard the $t$-component of an arbitrary
1255: nonspacelike geodesic with affin parameter~$\tau$. Due to the
1256: assumed homogenity not all of the symbols in section~\ref{Symbole}
1257: are different from zero. More precisely, the $0$-component
1258: of~(\ref{Geo}) reads
1259: \begin{eqnarray*}
1260: \frac{{\rm d}^2t}{{\rm d}\tau^2}\kern1mm+\kern1mm
1261: \Gamma^0_{00}\left(\frac{{\rm d}t}{{\rm d}\tau}\right)^2\kern1mm+\kern1mm
1262: \Gamma^0_{11}\left(\frac{{\rm d}x}{{\rm d}\tau}\right)^2\kern1mm+\kern1mm
1263: \Gamma^0_{22}\left(\frac{{\rm d}y}{{\rm d}\tau}\right)^2\kern1mm+\kern1mm
1264: \Gamma^0_{33}\left(\frac{{\rm d}z}{{\rm d}\tau}\right)^2&=&0.
1265: \end{eqnarray*}
1266: The causal character of the curve and the choise of time orientation
1267: imply~(\ref{Orient}) and further
1268: \begin{eqnarray}
1269: \lefteqn{%\mbox{}\kern-4mm
1270: \frac{\rm d}{{\rm d}t}\ln\left(\frac{{\rm d}t}{{\rm d}\tau}\right)^{-1}
1271: \kern2mm=}\nonumber\\[2mm]
1272: &&\kern-12mm\left(\frac{{\rm d}t}{{\rm d}\tau}\right)^{-2}\left[
1273: \Gamma^0_{00}\left(\frac{{\rm d}t}{{\rm d}\tau}\right)^2\kern1mm+\kern1mm
1274: \Gamma^0_{11}\left(\frac{{\rm d}x}{{\rm d}\tau}\right)^2\kern1mm+\kern1mm
1275: \Gamma^0_{22}\left(\frac{{\rm d}y}{{\rm d}\tau}\right)^2\kern1mm+\kern1mm
1276: \Gamma^0_{33}\left(\frac{{\rm d}z}{{\rm d}\tau}\right)^2
1277: \right].\label{kern}
1278: \end{eqnarray}
1279: Now, using lemma~\ref{Voll1} we can estimate the symbols and get
1280: \begin{eqnarray*}
1281: \lefteqn{%
1282: \Gamma^0_{11}\left(\frac{{\rm d}x}{{\rm d}\tau}\right)^2\kern1mm+\kern1mm
1283: \Gamma^0_{22}\left(\frac{{\rm d}y}{{\rm d}\tau}\right)^2\kern1mm+\kern1mm
1284: \Gamma^0_{33}\left(\frac{{\rm d}z}{{\rm d}\tau}\right)^2\kern2mm>\kern2mm}\\
1285: &&\kern14mm-\frac12 g^{00}
1286: C_\beta \kern1mmt^{-1}\left[
1287: g_{11}\left(\frac{{\rm d}x}{{\rm d}\tau}\right)^2\kern1mm+\kern1mm
1288: g_{22}\left(\frac{{\rm d}y}{{\rm d}\tau}\right)^2\kern1mm+\kern1mm
1289: g_{33}\left(\frac{{\rm d}z}{{\rm d}\tau}\right)^2
1290: \right].
1291: \end{eqnarray*}
1292: On the other hand we have~(\ref{Norm})
1293: \begin{eqnarray*}
1294: g_{11}\left(\frac{{\rm d}x}{{\rm d}\tau}\right)^2\kern1mm+\kern1mm
1295: g_{22}\left(\frac{{\rm d}y}{{\rm d}\tau}\right)^2\kern1mm+\kern1mm
1296: g_{33}\left(\frac{{\rm d}z}{{\rm d}\tau}\right)^2&=&
1297: -\varepsilon-g_{00}\left(\frac{{\rm d}t}{{\rm d}\tau}\right)^2 \\
1298: &<&-g_{00}\left(\frac{{\rm d}t}{{\rm d}\tau}\right)^2
1299: \end{eqnarray*}
1300: where $\varepsilon\in\{0,1\}$ so we get finally
1301: ($C_\beta<0$)
1302: \begin{eqnarray*}
1303: \Gamma^0_{11}\left(\frac{{\rm d}x}{{\rm d}\tau}\right)^2\kern1mm+\kern1mm
1304: \Gamma^0_{22}\left(\frac{{\rm d}y}{{\rm d}\tau}\right)^2\kern1mm+\kern1mm
1305: \Gamma^0_{33}\left(\frac{{\rm d}z}{{\rm d}\tau}\right)^2
1306: &>&\frac12 C_\beta \kern1mmt^{-1}
1307: \left(\frac{{\rm d}t}{{\rm d}\tau}\right)^2.
1308: \end{eqnarray*}
1309: Using~(\ref{kern}) as well as~$\Gamma^0_{00}=
1310: \frac14(\beta-1)\cdot t^{-1}$ and~(\ref{Cbeta})
1311: \begin{eqnarray*}
1312: \frac{\rm d}{{\rm d}t}\ln\left(\frac{{\rm d}t}{{\rm d}\tau}\right)^{-1}
1313: &>&\frac12
1314: \left[
1315: \frac12(\beta^2-1)+C_\beta
1316: \right]\kern.5mm t^{-1}\kern2mm=\kern2mm-t^{-1}
1317: \end{eqnarray*}
1318: we get an suitable estimate. Integrating twice the result is
1319: \begin{eqnarray*}
1320: \tau(t)&>&\tau(t_0)\kern1mm+\kern1mmC\cdot\ln\frac{t}{t_0}
1321: \end{eqnarray*}
1322: for some positive $C$. Obviously, by taking $t\to\infty$ we have
1323: proved that $\tau\to\infty$ holds along any nonspacelike
1324: geodesic.
1325: \end{beweis}
1326:
1327: \subsection*{Geodesics in not spatially homogeneous,
1328: polarized spacetimes}
1329:
1330: In the case under consideration the components of Gowdy metric
1331: are not everywhere $x$-independent and consequently, the
1332: solutions of Einstein equations are not known explicitly.
1333: Due to limitation on asymptotic behaviour as well as the
1334: more complex form of geodesic equation proofs are more technical here.
1335:
1336: We begin by poving some estimates:
1337:
1338: \begin{lemma}
1339: Let $t_0>0$ and $W\colon\mathbb R_+\times\mathbb T^1\to\mathbb R$
1340: some smooth solution of $W_{tt}+t^{-1}W_t-W_{xx}=0$,
1341: $W_x\equiv\kern-3.0mm/ \kern1.5mm0$. Then there is a positive
1342: constant $C$, such that for all $t>t_0$
1343: \label{Voll2}
1344: \begin{eqnarray*}
1345: \left|-\frac1{2t^2}+\frac12({W_t}^2-{W_x}^2)\right|&\le&C\cdot t^{-1}
1346: \end{eqnarray*}
1347: holds.
1348: \end{lemma}
1349:
1350: \begin{beweis}
1351: For $t$ sufficiently large we know according to
1352: corollary~\ref{Folgerung6}
1353: $W$ is of the form
1354: \begin{eqnarray}
1355: W(t,x)&=&\gamma\kern1mm+\kern1mm\beta\cdot\ln t\kern1mm+\kern1mm
1356: t^{-\frac12}\left(\nu(t,x)\kern1mm+\kern1mm\omega(t,x)\right).
1357: \label{W3}
1358: \end{eqnarray}
1359: Theorem~\ref{Satz4} proved boundedness of $\nu+\omega$
1360: while its partial derivatives are bounded by theorem~\ref{Satz5},
1361: hence the lemma.
1362: \end{beweis}
1363:
1364: \begin{lemma}
1365: For any polarized Gowdy spacetime model there is a constant~$C_1$
1366: and a positive constant~$C$, such that
1367: \label{Voll3}
1368: \begin{eqnarray*}
1369: |g_{ab}|\kern1mm+\kern1mm|\partial_t g_{ab}|
1370: \kern1mm+\kern1mm|g^{ab}|\kern1mm+\kern1mm
1371: |\partial_t g^{ab}|&\le&C\cdot t^{C_1}
1372: \end{eqnarray*}
1373: where $a,b\in\{2,3\}$ and $t_0<t<\infty$.
1374: \end{lemma}
1375:
1376: \begin{beweis}
1377: (\ref{W3}) shows logarithmic growth of $W$ if $\beta\neq0$,
1378: consequently the metric coefficients $g_{ab}$, its inverses
1379: and derivatives can increase at most with power~$\beta+1$.
1380: So a constant~$C_1\mathop{=}\limits_{\rm Df}|\beta|+1$
1381: will be sufficient for our needs.
1382: \end{beweis}
1383:
1384: \begin{theorem}
1385: Every inextendable, future directed, nonspacelike geodesic in an
1386: arbitrary but not spatially homogeneous, polarized Gowdy manifold
1387: is in its future direction complete.
1388: \end{theorem}
1389:
1390:
1391: \begin{beweis}
1392: Let $\mathfrak x=(t,x,y,z)$ be the components of some
1393: arbitrary nonspacelike geodesic in local coordinates,
1394: having affin parameter~$\tau$ and taking values in an
1395: likewise arbitrary, polarized (not spatially homogeneous)
1396: Gowdy manifold. As before it will sufficient to restrict
1397: considerations on the $t$-component. Here it holds:
1398: \begin{eqnarray}
1399: 0&=&\frac{{\rm d}^2t}{{\rm d}\tau^2}\kern1mm+\kern1mm\Gamma^0_{00}
1400: \left(\frac{{\rm d}t}{{\rm d}\tau}\right)^2\kern1mm+\kern1mm
1401: 2\kern.5mm\Gamma^0_{01}\kern.5mm\frac{{\rm d}t}{{\rm d}\tau}
1402: \frac{{\rm d}x}{{\rm d}\tau}\kern1mm+\kern1mm
1403: \Gamma^0_{11}\left(\frac{{\rm d}x}{{\rm d}\tau}\right)^2\nonumber\\
1404: &&+\kern1mm
1405: \Gamma^0_{22}\left(\frac{{\rm d}y}{{\rm d}\tau}\right)^2\kern1mm+\kern1mm
1406: \Gamma^0_{33}\left(\frac{{\rm d}z}{{\rm d}\tau}\right)^2
1407: \label{kern2}
1408: \end{eqnarray}
1409: Again, Christoffel symbols are taken from section~\ref{Symbole}.
1410: Using constraints~(\ref{Hamilton}) and~(\ref{Impuls})
1411: we get:
1412: \begin{eqnarray*}
1413: \lefteqn{%
1414: \mbox{}\kern-25mm
1415: \Gamma^0_{00}\left(\frac{{\rm d}t}{{\rm d}\tau}\right)^2\kern1mm+\kern1mm
1416: 2\kern.5mm\Gamma^0_{01}\kern.5mm\frac{{\rm d}t}{{\rm d}\tau}
1417: \frac{{\rm d}x}{{\rm d}\tau}\kern1mm+\kern1mm
1418: \Gamma^0_{11}\left(\frac{{\rm d}x}{{\rm d}\tau}\right)^2}\\[2mm]
1419: \mbox{}\kern25mm&=&
1420: \frac12 t\left(W_t\frac{{\rm d}t}{{\rm d}\tau}+
1421: W_x\frac{{\rm d}x}{{\rm d}\tau}\right)^2
1422: \kern1mm-\kern1mm\frac1{4t}
1423: \left[\left(\frac{{\rm d}t}{{\rm d}\tau}\right)^2+
1424: \left(\frac{{\rm d}x}{{\rm d}\tau}\right)^2\right]\\[2mm]
1425: && -\kern1mm
1426: \frac t4({W_t}^2-{W_x}^2)
1427: \left[\left(\frac{{\rm d}t}{{\rm d}\tau}\right)^2-
1428: \left(\frac{{\rm d}x}{{\rm d}\tau}
1429: \right)^2\right]
1430: \end{eqnarray*}
1431: The first term on the right hand side is nonnegative and
1432: will be neglected. Using condition~(\ref{Norm})
1433: in the form
1434: \begin{eqnarray*}
1435: \left(\frac{{\rm d}x}{{\rm d}\tau}\right)^2&=&
1436: \left(\frac{{\rm d}t}{{\rm d}\tau}\right)^2
1437: \kern1mm+\kern1mm g^{00}
1438: \left[\varepsilon\kern.5mm+\kern.5mm
1439: g_{22}\left(\frac{{\rm d}y}{{\rm d}\tau}\right)^2
1440: \kern.5mm+\kern.5mm
1441: g_{33}\left(\frac{{\rm d}z}{{\rm d}\tau}\right)^2\right]
1442: \end{eqnarray*}
1443: where as before $\varepsilon\in\{0,1\}$ depending on causal
1444: character of geodesic, we get
1445: \begin{eqnarray*}
1446: \Gamma^0_{00} \left(\frac{{\rm d}t}{{\rm d}\tau}\right)^2\kern1mm+\kern1mm
1447: 2\kern.5mm\Gamma^0_{01}\kern.5mm\frac{{\rm d}t}{{\rm d}\tau}
1448: \frac{{\rm d}x}{{\rm d}\tau}\kern1mm+\kern1mm
1449: \Gamma^0_{11}\left(\frac{{\rm d}x}{{\rm d}\tau}\right)^2
1450: &\ge&-\frac1{2t}\kern.5mm
1451: \left(\frac{{\rm d}t}{{\rm d}\tau}\right)^2 \\
1452: &&\mbox{}\kern-65mm+\left[-\frac1{4t}\kern.5mm+\kern.5mm
1453: \frac t4({W_t}^2-{W_x}^2)\right]
1454: \kern.5mm g^{00} \kern.5mm
1455: \left[\varepsilon\kern.5mm+\kern.5mm
1456: g_{22}\left(\frac{{\rm d}y}{{\rm d}\tau}\right)^2
1457: \kern.5mm+\kern.5mm
1458: g_{33}\left(\frac{{\rm d}z}{{\rm d}\tau}
1459: \right)^2\right].
1460: \end{eqnarray*}
1461: This estimate yield for the factor
1462: ($2t^{-1}$) of geodesic equation~(\ref{kern2})
1463: \begin{eqnarray}
1464: \lefteqn{%
1465: \frac{\rm d}{{\rm d}t}\left[t^{-1}\left(
1466: \frac{{\rm d}t}{{\rm d}\tau}\right)^2\right]
1467: \kern2mm\le}\nonumber\\[5mm]
1468: &&\mbox{}\kern10mm
1469: -g^{00}\left[-\frac1{2t^2}+\frac 12({W_t}^2-{W_x}^2)\right]
1470: \left[\varepsilon\kern.5mm+\kern.5mm
1471: g_{22}\left(\frac{{\rm d}y}{{\rm d}\tau}\right)^2\kern.5mm+\kern.5mm
1472: g_{33}\left(\frac{{\rm d}z}{{\rm d}\tau}\right)^2
1473: \right]\nonumber\\[2mm]
1474: &&\mbox{}\kern10mm
1475: +g^{00}\left[\partial_t
1476: g_{22}\cdot t^{-1}\kern.5mm
1477: \left(\frac{{\rm d}y}{{\rm d}\tau}\right)^2\kern.5mm+\kern.5mm
1478: \partial_t
1479: g_{33}
1480: \cdot t^{-1}\kern.5mm\left(\frac{{\rm d}z}{{\rm d}\tau}\right)^2
1481: \right],\label{Voll4}
1482: \end{eqnarray}
1483: where we have used~(\ref{Orient}).
1484:
1485: Due to~(\ref{KVF}) the scalar products~$g(\dot\mathfrak x,\cdot)$
1486: in
1487: \begin{eqnarray*}
1488: \left(\frac{{\rm d}y}{{\rm d}\tau}\right)^2
1489: &=&
1490: \left[g^{22}\cdot g\left(
1491: \dot\mathfrak x,\frac\partial{\partial y}\right)\right]^2\\[2mm]
1492: \left(\frac{{\rm d}z}{{\rm d}\tau}\right)^2
1493: &=&
1494: \left[g^{33}\cdot g\left(
1495: \dot\mathfrak x,\frac\partial{\partial z}\right)\right]^2
1496: \end{eqnarray*}
1497: are constant along the geodesic.
1498: Thanks to lemmas~\ref{Voll2} and~\ref{Voll3} we can estimate
1499: all brackets on the right hand side of the inequality~(\ref{Voll4})
1500: by an $\mathcal O(t^C)$-term, where the constant~$C$ only depends
1501: on the model parameter $\beta$ under consideration.
1502: Furthermore, due to~(\ref{aasym}) it holds
1503: $g^{00}\sim\mathcal O ({\rm e}^{-2\bar\nu t})$, $\bar\nu>0$,
1504: $g^{00}$ will dominate the development on the right hand side.
1505: Consequently there are positive constants
1506: $C_1$, $C_2$ with
1507: \begin{eqnarray*}
1508: \frac{\rm d}{{\rm d}t}\left[t^{-1}\left(
1509: \frac{{\rm d}t}{{\rm d}\tau}\right)^2\right]
1510: &\le&
1511: C_1\kern.5mm{\rm e}^{-C_2t}
1512: \end{eqnarray*}
1513: and furthermore some positive constant
1514: ${C_3}^{-2}$, such that for sufficiently large $t$ it holds
1515: \begin{eqnarray*}
1516: t^{-1}\left(
1517: \frac{{\rm d}t}{{\rm d}\tau}
1518: \right)^2&\le&{C_3}^{-2}
1519: \end{eqnarray*}
1520: and a final integration gives
1521: \begin{eqnarray*}
1522: \tau(t)&\ge&\tau(t_1)\kern1mm+\kern1mm2{C_3}\left(\sqrt{t}-
1523: \sqrt{t_1}\right),
1524: \end{eqnarray*}
1525: which proves $\tau\to\infty$ for $t\to\infty$.
1526: \end{beweis}
1527:
1528: We summarize as follows:
1529:
1530: \begin{corollary}
1531: Every inextendable, future directed, nonspacelike geodesic in an
1532: arbitrary polarized Gowdy manifold
1533: is in its future direction complete.
1534: \end{corollary}
1535:
1536:
1537: \section{Nonvanishing Christoffel symbols in polarized Gowdy spacetimes}
1538: \label{Symbole}
1539:
1540: In polarized Gowdy models the nonvanishing components of the metric
1541: tensor are
1542: \begin{eqnarray*}
1543: g_{00}&=&-{\rm e}^{2a}\kern2cm g_{22}\kern2mm=\kern2mmt\kern.5mm{\rm e}^{W}\\
1544: g_{11}&=&\phantom{-}{\rm e}^{2a}\kern2cm g_{33}\kern2mm=\kern2mmt\kern.5mm{\rm e}^{-W}
1545: \end{eqnarray*}
1546: where $a=a(t,x)$, $W=W(t,x)$, so up to symmetry the only nonvanishing
1547: symbols are
1548: \begin{eqnarray*}
1549: \begin{array}{lcrclclcrcl}
1550: \Gamma^0_{00}&=&\frac12g^{00}\partial_t g_{00}&=&a_t&\phantom{mmm}&
1551: \Gamma^0_{11}&=&-\frac12g^{00}\partial_t g_{11}&=&a_t\\[2mm]
1552: \Gamma^0_{01}&=&\frac12g^{00}\partial_x g_{00}&=&a_x&\phantom{mmm}&
1553: \Gamma^1_{01}&=&\frac12g^{11}\partial_t g_{11}&=&a_t\\[2mm]
1554: \Gamma^1_{00}&=&-\frac12g^{11}\partial_x g_{00}&=&a_x&\phantom{mmm}&
1555: \Gamma^1_{11}&=&\frac12g^{11}\partial_x g_{11}&=&a_x
1556: \end{array}
1557: \end{eqnarray*}
1558: \begin{eqnarray*}
1559: \begin{array}{lcrcl}
1560: \Gamma^0_{22}&=&-\frac12g^{00}\partial_t g_{22}&=&-\frac12g^{00}
1561: (t^{-1}+W_t)g_{22}\\[2mm]
1562: \Gamma^2_{02}&=&\frac12g^{22}\partial_t g_{22}&=&\phantom{-}\frac12(t^{-1}+W_t)\\[2mm]
1563: \Gamma^1_{22}&=&-\frac12g^{11}\partial_x g_{22}&=&-\frac12g^{11}
1564: W_x g_{22}\\[2mm]
1565: \Gamma^2_{12}&=&\frac12g^{22}\partial_x g_{22}&=&\phantom{-}\frac12W_x\\[2mm]
1566: \Gamma^0_{33}&=&-\frac12g^{00}\partial_t g_{33}&=&-\frac12g^{00}
1567: (t^{-1}-W_t)g_{33}\\[2mm]
1568: \Gamma^3_{03}&=&\frac12g^{33}\partial_t g_{33}&=&\phantom{-}\frac12(t^{-1}-W_t)\\[2mm]
1569: \Gamma^1_{33}&=&-\frac12g^{11}\partial_x g_{33}&=&\phantom{-}\frac12g^{11}
1570: W_x g_{33}\\[2mm]
1571: \Gamma^3_{13}&=&\frac12g^{33}\partial_x g_{33}&=&-\frac12W_x
1572: \end{array}
1573: \end{eqnarray*}
1574:
1575: %\begin{itemize}
1576: %\item Interpretation: L"osungen der Einsteingleichungen
1577: %korrespondieren zu linear polarisierten Graitationswellen mit
1578: %$\mathbb T^3$-Topologie
1579: %\item Isenberg und Moncrief (1990) untersuchten singul"ares
1580: %Verhalten f"ur alle Topologien; $\mathbb S^3$ und
1581: %$\mathbb S^2\times\mathbb S^1$ sind auch zukunftssingul"ar
1582: %$\Rightarrow$ f"ur polarisierte Modelle blieb also nur offen,
1583: %was in diesem Abschnitt abgehandelt wurde
1584: %\item CIM machen Aussagen "uber nichtgenerische F"alle;
1585: %Verbindung zu flachen Kasner-/Taubraumzeiten erw"ahnen; ggf.\
1586: %kurze Rechnung wie zu Misnerschen Raumzeiten; i.d.S.\ sind
1587: %auch Bemerkungen im Schlu"sfolgerungsabschnitt von Moncrief (1981)
1588: %sehr wertvoll $\Rightarrow$ ggf.\ m"ussen diese im obigen
1589: %ausgeschlossen werden, was jedoch {\em sehr} unwahrscheinlich ist
1590: %\end{itemize}
1591:
1592: %Das vorliegende Kapitel befa"ste sich mit der Struktur polarisierter
1593: %Gowdymannigfaltigkeiten im Gro"sen. Der Schl"ussel dazu war
1594: %die Absch"atzung von L"osungen
1595: %der Euler-Poisson-Darboux-Gleichung. Die topologische
1596: %Natur des Problems ausnutzend, h"atte man ebensogut mit einem
1597: %Ansatz $W(t,x)=Z(t)R(x)$ beginnen und $R$ nach einem vollst"andigen
1598: %System von Eigenfunktionen entwickeln k"onnen. Das Ergebnis
1599: %ist eine formale L"osung
1600: %\begin{eqnarray*}
1601: %W(t,x)&=&\gamma+\beta\cdot\ln t+\sum_n^{1,\infty}
1602: %b_n(J_0(nt)+N_0(nt))\cos nx +\\[-5mm]
1603: %&&\phantom{\gamma+\beta\cdot\ln t+\sum_n^{1,\infty}}
1604: %a_n(J_0(nt)+N_0(nt))\sin nx,
1605: %\end{eqnarray*}
1606: %in der $J_0$ und $N_0$ die Besselsche bzw.\ Neumannsche Funktion
1607: %nullter Ordnung bezeichnen. Diese Reihe wurde bereits von
1608: %in der Arbeit von Berger~(1974) angegeben.
1609:
1610: %F"ur unsere Zwecke bietet eine
1611: %solch kurze Darstellung aber nur auf den ersten Blick
1612: %Vorteile, da wir "uber die Anfangsbedingungen, die schlie"slich
1613: %$a_n$ und $b_n$ bestimmen, damit aber auch "uber Konvergenzverhalten
1614: %und Summe der Reihe zun"achst nichts wissen.
1615:
1616: %Besselsche Funktionen wurden zum ersten Mal von Schl"omilch~(1857)
1617: %um ihrer selbst willen untersucht. Er betrachtete dabei auch
1618: %Reihen des angegebenen Typs, die seither nach ihm benannt werden.
1619: %Im Zusammenhang mit eindimensionalen Luftschwingungen gab
1620: %Rayleigh~(1911) sp"ater eine physikalische Interpretation
1621: %und von~Ignatowsky (1914, 1915) wurde etwa zur gleichen Zeit
1622: %bei seinen Untersuchungen "uber Beugungsgitter auf dieselbe Reihe
1623: %gef"uhrt, die er unabh"angig von den Vorgenannten
1624: %untersuchte.~--~Jedenfalls ist bekannt
1625: %(Twer\-sky (1961), Cooke (1928)),
1626: %da"s sich nahezu
1627: %beliebige Funktionen in Schl"omilchsche Reihen entwickeln lassen,
1628: %was Aussagen "uber das asymptotische Verhalten von~$W$
1629: %vermittels der angegebenen Reihe entsprechend erschwert.
1630:
1631: %\begin{itemize}
1632: %\item Rayleigh- und Ignatkowskiverweise weglassen; letzten
1633: %Satz umformulieren $\Rightarrow$ Reihe ist gliedweise L\"osung;
1634: %Asymptotik der Besselschen Funktionen Richtlinie des Beweises
1635: %\item Diskussion der Arbeit von Chru\'sciel, Isenberg und Moncrief
1636: %\end{itemize}
1637:
1638: %Es sollte die Verbindung des vollen Modells / des polarisierten
1639: %Modells zu den Kasnerl"osungen (insbesondere nat"urlich der
1640: %r"aumlich homogenen L"osungen) ausgearbeitet werden.
1641: %Die Rendall-Kichenassamy-Arbeit ist sicher kein ungeeigneter
1642: %Ansatz dazu.
1643:
1644: %$\Rightarrow$ Asymptotisch homogen. In Abh\"angigkeit von gew\"ahlten
1645: %Anfangsdaten werden alle Punkte mit Ausnahme eines Taubuniversums
1646: %approximiert. Kurze Berechnungsskizze. Erw\"ahnung der Analogie von
1647: %Isenberg/Moncrief f\"ur Kontraktion.
1648:
1649:
1650: \section*{Acknowledgement}
1651:
1652: I am indebted to A.D.\ Rendall for suggesting me this problem and
1653: for useful discussions.
1654:
1655: \section*{Literature}
1656:
1657: %{\bf Berger, B.K.} (1974) Quantum graviton creation in a model
1658: %universe,
1659: %{\em Ann. Phys.}, {\em 83}: 458-490.
1660:
1661: {\bf Chru\'sciel, P.T., J.~Isenberg and V.~Moncrief} (1990)
1662: Strong cosmic censorship in polarised Gowdy spacetimes,
1663: {\em Class. Quant. Grav.}, {\em 7}: 1671-1680.
1664:
1665: {\bf Gowdy R.H.} (1974)
1666: Vacuum spacetimes with two-parameter spacelike isometry groups
1667: and compact invariant hypersurfaces: topology and boundary
1668: conditions,
1669: {\em Ann. Phys.}, {\em 83}: 203-241.
1670:
1671: %{\bf Hawking, S.W.} (1968) The existence of cosmic time
1672: %functions,
1673: %{\em Proc. Roy. Soc. Lond.}, {\em A 308}: 433-435.
1674:
1675: %{\bf Hawking, S.W. and R. Penrose} (1970) The singularities of
1676: %gravitaional collapse and cosmology,
1677: %{\em Proc. Roy. Soc. Lond.}, {\em A 314}: 529-548.
1678:
1679: %{\bf Hopf, H. and W. Rinow} (1931) "Uber den Begriff der
1680: %vollst"andigen differentialgeometrischen Fl"ache,
1681: %{\em Comm. Math. Helv.}, {\em 3}: 209-225.
1682:
1683: %{\bf Isenberg, J. and V.~Moncrief} (1982)
1684: %The existence of constant mean curvature foliations of Gowdy
1685: %3-torus spacetimes,
1686: %{\em Comm. Math. Phys.}, {\em 86}: 485-493.
1687:
1688: {\bf Isenberg, J. and V.~Moncrief} (1990)
1689: Asymptotic behavior of the gravitational field and the nature
1690: of singularities in Gowdy spacetimes,
1691: {\em Ann. Phys.}, {\em 199}: 84-122.
1692:
1693: {\bf Moncrief, V.} (1981)
1694: Global properties of Gowdy spacetimes with $\mathbb T^3\times\mathbb R$
1695: topology, {\em Ann. Phys.}, {\em 132}: 87-107.
1696:
1697: {\bf Rendall, A.D.} (2000)
1698: Fuchsian analysis of singularities in Gowdy spacetimes beyond analyticity,
1699: {\em Class. Quant. Grav.}, {\em 17}: 3305-3316.
1700:
1701: {\bf Ringstr\"om, H.} (2002)
1702: On Gowdy vacuum spacetimes,
1703: {\em preprint}, {\tt gr-qc/0204044}.
1704:
1705: %{\bf Seifert, H.-J.} (1977)
1706: %Smoothing and extending cosmic time functions,
1707: %{\em Gen. Rel. Grav.}, {\em 8}: 815-831.
1708:
1709: \end{document}
1710:
1711: