gr-qc0211104/red3.ltx
1: 
2: 
3: 
4: 
5: 
6: %\documentstyle[epsfig,prl,aps]{revtex}
7: %\documentstyle[prl,aps,twocolumn]{revtex}
8: %\documentstyle[preprint,eqsecnum,aps]{revtex}\def\btt#1{{\tt$\backslash$#1}}
9: \documentstyle[epsf,prl,aps,twocolumn]{revtex}
10: \begin{document}
11: \draft
12: \title{Gravitational redshifts in electromagnetic bursts occuring
13: near Schwarzschild horizon}
14: \author{ Janusz Karkowski $^1$ and}
15: \author{ Edward Malec $^{1,2}$   }
16: \address{  $^1$ Institute of Physics, Jagiellonian University, 30-064 Krak\'ow,
17: Reymonta 4, Poland   }
18: \address{$^2$ Max Planck Institute for Gravitational Physics AEI, Golm, Germany   }
19: 
20: 
21: \maketitle
22: 
23: \begin{abstract}
24: It was suggested earlier  that the gravitational redshift
25: formula can be invalid  when the effect of the backscattering is strong.
26: It is   demonstrated  here numerically,  for an exemplary electromagnetic pulse that
27: is: i) initially located very close to the horizon of a Schwarzschild black hole
28: and ii)   strongly backscattered, that  a mean frequency does not obey
29: the standard redshift formula.  Redshifts    appear to depend on
30: the frequency and there manifests  a backscatter-induced blueshift in the
31: outgoing radiation.
32: \end{abstract}
33: 
34: 
35: \pacs{ 04.20.-q  04.70.-s    95.30.Sf 98.62.Js  }
36: \date{ }
37: 
38: \section{ Introduction}
39: 
40: 
41: 
42: Standard  derivations of the gravitational redshift base on
43: the approximation of geometric optics (see, for a   discussion,
44: \cite{Sachs} -  \cite{Strauman} ). This problem
45: has been recently reconsidered  in the wave formulation
46: \cite{ME2002}. It was  shown, for a class of
47: compact shocks that  are separated in some sense \cite{sense}
48:  from the horizon
49: of a Schwarzschild black hole, that the energy flux
50: scales accordingly to the redshift formula. Introducing
51:  the concept of photons and assuming that their number is conserved,
52: one again arrives at the standard relation  for   the frequency.
53: As an alternative approach  one can apply  the Fourier
54: analysis,  as outlined below in Sec. II.
55: 
56: The crucial feature behind the above mentioned compactness and
57: separateness conditions is that when they hold true, then the backscatter
58: is negligible \cite{Karkowski}. It was remarked  in \cite{ME2002} that
59: a compact pulse of radiation  that is exposed to a significant
60: backscattering must not obey the familiar redshift relation. This
61: is suggested by the following reasoning. A spatially compact wave
62: packet consists of a superposition of monochromatic waves of different
63: frequencies. The effect of the backscattering is stronger  for low
64: frequency waves than for high frequency ones, that is   spectral
65: amplitudes of the former are stronger damped than those of the latter.
66: Therefore one  expects  that the backscattering induces a shift in a
67: mean frequency, so that it
68: does not satisfy the gravitational redshift formula. The aim of this
69: investigation is to extend results of \cite{ME2002} and  find
70: a numerical example in favour of this conjecture.  This examplary
71: wave pulse {\bf must break } -- and in fact    {\bf does so},
72: in accordance with \cite{ME2002}  -- the compactness condition of \cite{sense}.
73: 
74: In the next section we  write   equations and  review relevant
75: results of the preceding papers \cite{ME2002}, \cite{malec2000}.
76: Sec. III describes an electromagnetic wave that does not
77: comply to the standard redshift formula. Sec. IV presents a short summary.
78: 
79: 
80: \section{Gravitational redshift: classical  derivation}
81: 
82: 
83: 
84: The space-time geometry  is defined  by
85: the Schwarzschild  line element,
86: %
87: \begin{equation} ds^2 =- (1-{2m\over R})dt^2 +
88: {1\over 1-{2m\over R}} dR^2 +
89: R^2 d\Omega^2~,
90: \label{1}
91: \end{equation}
92: %
93: where $t$ is a time coordinate, $R$ is an  areal radius
94: and $d\Omega^2=d\theta^2 +\sin^2 \theta d\phi^2 $
95: (with $0\le \theta \le \pi $ and $0\le \phi <2\pi $)
96: is the line element on the unit sphere.
97: The Newtonian gravitational constant $G$
98: and  the velocity of light $c$ are put equal to 1.
99: We define the tortoise coordinate $r^*=R+2m\ln ({R\over 2m}-1)$
100: and  $\eta_R \equiv 1-{2m\over R}$.
101: We will consider only the dipole electromagnetic term  and,
102: more specifically, choose  the potential one-form $A=\sqrt{3/2}
103: \sin^2\theta \Psi (r^*,t)d\phi $.
104: The unknown function $\Psi $   satisfies the equation  \cite{Wheeler}
105: %
106: \begin{equation}
107: (-\partial_t^2 + \partial_{r^*}^2)\Psi = \eta_R{2\over R^2}
108: \Psi .
109: \label{2}
110: \end{equation}
111: %
112: Let the electromagnetic strength field tensor be $F_{\mu \nu }$.
113: The stress-energy tensor of the electromagnetic field
114: reads   $T_{\mu }^{\nu }=F_{\mu \gamma }
115: F^{\nu \gamma }-(1/4)g_{\mu }^{\nu }F_{\gamma \delta }
116: F^{\gamma \delta }$ and the time-like translational
117: Killing vector is denoted as $\zeta $.  One can
118: define   the energy flux $\hat P_R$,
119: %
120: \begin{eqnarray}
121:  \hat P_R(R,t)\equiv {1\over \sqrt{\eta_r}}\int_{S(R)}dS(R)n_rT^{r\mu }\zeta_{\mu }.
122: \label{2.1}
123: \end{eqnarray}
124: %
125: Here $n$ is the unit normal to the sphere $R$ and $dS$ is the standard
126: area element on $S$.
127: $\hat P_R$ is equal to
128: %
129: \begin{eqnarray}
130:  \hat P_R(R,t)  =
131:  -{4\pi \over   \sqrt{\eta_R}} \partial_t\Psi \partial_{r^*}\Psi .
132: \label{2.5}
133: \end{eqnarray}
134: %
135: Let  $\tilde \Gamma_{R_0}$ be a null geodesic directed outward
136: from the point $R_0$ of the initial hypersurface and let
137: $\tilde \Gamma_{ R_0,  (R,t) }$  be a segment of $\tilde \Gamma_{R_0}$
138: that connects $R_0$ and $(R,t)$. Comparing the   energy fluxes
139: through the spheres $S(R)$ (where $R>>2m$) and the initial   $S(R_0)$,
140: one obtains    \cite{ME2002}
141: %
142: \begin{eqnarray}
143:   \hat P_R(R) \approx
144:     \sqrt{\eta_{R_0}\over \eta_R}    \hat P_R(R_0) ,
145: \label{2.2}
146: \end{eqnarray}
147: %
148: assuming that the initial support of a dipole wave packet is contained
149: in the annulus $(a,b)$ such that   $(b-a)/(a\eta_a^5 )<<1$.
150: Here $a$ and $b$ are the {\bf areal radii}.
151: 
152: There are two ways to derive  the standard   redshift
153: formula from (\ref{2.2}).
154: 
155: i) {\it Eclectic approach}. The  condition
156: %
157: \begin{equation}
158: (b-a)/(a\eta_a^5 )<<1
159: \label{cond}
160: \end{equation}
161: %
162: implies the validity of the geometric optics approximation.
163: Thence  one can write $\hat P_R(R_0)= \hbar \hat \omega_{R_0} N_{R_0}$
164: and $\hat P_R(R)=\hbar \hat \omega_RN_{R}$, where $\hat \omega_{R_0}$
165: and $\hat \omega_R$ are the mean frequencies of the initial and
166: final pulses (as measured by static observers)
167:  and $N_{R_0}$ and $N_R$ are  the respective photon
168: fluxes. If the photon fluxes are  conserved, then from (\ref{2.2})
169: one  infers
170: %
171: \begin{equation}
172: \hat \omega_R=\sqrt{\eta_{R_0}\over \eta_R}\hat
173: \omega_{R_0}.
174: \label{2.4}
175: \end{equation}
176: %
177: ii) {\it Classical approach.}  One can Fourier-analyze  the quantity
178: representing the strength field tensor, $h=(-\partial_t+\partial_{r^*})\Psi $;
179: this is preserved along $\tilde \Gamma_{R_0}$, if  condition (\ref{cond}) is satisfied.
180: Let the spectral strength field  density be $g(\omega_{\tau } )=
181: \int dt e^{-i\omega_{ \tau }t} (-\partial_t+\partial_{r^*})\Psi  $. Let the
182: support of $g(\omega )$ be $\tilde \Omega $. If one assumes the normalization
183: condition $\int_{\tilde \Omega } d\omega_{\tau } |g(\omega_{\tau })|=1$,
184: then  the average frequency can be defined  as $\hat \omega_{\tau } =
185: \int_{\tilde \Omega } d\omega_{\tau } \omega_{\tau } |g(\omega_{\tau })|$.
186: When the analysis is done with respect the asymptotic  time $t$,
187: then  asymptotic  mean frequency $\hat \omega $ is found.
188: The Fourier analysis with respect
189: the proper time $d\tau = \sqrt{\eta_R}dt$ of a static observer located
190: at $R$  gives a corresponding   mean frequency $\hat \omega /
191: \sqrt{\eta_R} $. Thus
192: the two   frequencies satisfy (\ref{2.4}).
193: 
194: 
195: 
196: 
197: 
198: 
199: \section{Counterexample to the standard redshift formula}
200: 
201: 
202: It was pointed out in \cite{ME2002} that (\ref{2.4})
203: is valid for very  compact initial pulses - those satisfying
204: the assumption (\ref{cond}).    This   condition
205: demands not only that a pulse is compact, but also that its relative width
206: is small  in comparison to the relative distance from the black hole horizon.
207: It implies that   the
208: approximation of the geometric optics is valid and that the
209: backscatter is absent. We conjecture, basing on a numerical evidence,  that this
210:   can be relaxed to  $ (r^*(b)-r^*(a))/(2m)<<1$.
211: 
212: In the case of  a radiative pulse that is exposed to a significant
213:   backscattering the relation (\ref{2.4}) would not hold.
214: Let us again review arguments in favour of this conclusion. A  wave
215: pulse is superposed from  monochromatic waves
216: of different frequencies. When resolved spectrally by an observer,
217: the pulse can be seen as a collection   of peaks, each characterized by
218: some mean frequency. The effect of the backscattering is
219: more pronounced for low frequency waves than for high frequency ones
220: - the spectral  amplitudes of the former are stronger damped than of
221: the latter. Therefore one would expect that mean frequencies
222: $\hat \omega_{R_0}, \hat \omega_{R}$ of the
223: initial and final pulses will not conform to (\ref{2.4}). The observed
224: mean frequency  $\hat \omega_{R}$ of an outgoing pulse of radiation
225: can be  blue-shifted  in comparison to
226: the value  $\sqrt{\eta_{R_0}\over \eta_{ R}}\hat \omega_{R_0}$.
227: In these circumstances an attempt to fit observed data
228: of a resolved multi-peak  spectrum to the simple scaling law of
229: (\ref{2.4}) can lead to redshifts depending on the frequency.
230: 
231: Our aim  is to find numerical solutions corresponding to  compact
232: initial data, satisfying the  compactness assumption $b-a<<a$, such that:
233: 
234: i)  the spectral amplitude as a function of $\omega $
235: has  several peaks with well resolved  mean frequencies;
236: 
237: ii) the  evolution exhibits significant backscatter
238: (hence the energy flux is diminished);
239: 
240: iii)   some of  mean frequencies at $R_0$ and $R$ do not comply
241: to (\ref{2.4}).
242: 
243: The condition $b-a<<a$ (we remind the reader that $a$ and $b$ are areal radii)
244: is imposed in order to guarantee that initially
245: the wavelengths are much smaller than the   radius of the wave front
246: and the curvature radius, i.e.,
247: that   the assumption of the geometric optics is satisfied at the emission region.
248: On the other hand we {\bf do not assume}   the validity
249:   of (\ref{cond}); our initial data are such that $r^*(b)-r^*(a) = 4\pi m$,
250: which breaks down this inequality. Let us remark
251: that with such data one can have the geometric optics approximation
252: being valid both at the  emission and detection regions, but being evidently
253: broken around $R\approx 3m$, when the wavelengths and the curvature
254: radius are of the same order.
255: 
256: Below we present data describing one  of many analysed   examples.
257: The mass is normalized to unity and the Schwarzschild radius
258: $R_g$  is equal to 2. Let $I\equiv (r^*(a), r^*(b))$,
259: the  support of initial radiation, be comprised between
260: $a=2+1.97\times 10^{-18}$ and $b= 2+10^{-15}$ or, in terms of the tortoise coordinate,
261: between $r^*(b)\approx -68.46$ and $r^*(a )\approx -68.46- 4\pi $.
262: The purely outgoing initial data have the form $\tilde \Psi = \partial_tf(r^*-t)
263: +f(r^*-t)/R$ \cite{malec2000}, where $f$ is a free datum in $I$ but vanishes
264:  to the left from $a$.
265: With $I$ being so compact in terms of the areal radius it is reasonable
266: to ignore  the $R$ - dependence in this expression and to prescribe
267: initial $\Psi $ as being dependent only on $r^*-t$.
268: Thus the initial data can be  defined  by   prescribing $\tilde \Psi $
269: instead of $f$. We choose $\Psi (r^*)=\sin^2 ((r^*-r^*(a)))$ and
270: $\partial_t\Psi (r^*)=-\sin (2(r^*-r^*(a)))$   for
271: $r^*\in I\equiv (r^*(a), r^*(b))$ and  $\Psi (r^*)=\partial_t\Psi
272: =0$ outside $I$.
273: 
274: Notice that the size of the support of the initial pulse, $i=4\pi $,  satisfies
275: the inequality $i>2$, which is the plausibility condition (according to
276: the foregoing conjecture)  for  a significant backscatter.
277: 
278:    The signal  is measured during the time
279: interval $32 \pi $ at two observation
280: points, $r^*_{0}=-68.44$  and $r^*_1= 199.16$; in terms of the
281: areal radius we have $R_0=2+1.012\times 10^{-15}$ and $R=190.07$.
282: The grid was $120000\times 60000$.
283: Two snapshots of the evolving configuration taken at the
284: above  pair of observation
285: points  are presented  in Fig. 1.
286: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
287: \begin{figure}[1]
288: \epsfxsize=6cm
289: \centerline{\epsffile{j2.eps}}
290: \caption{ Temporal dependence of $h\equiv (-\partial_t+\partial_{r^*})\Psi $
291: (with values of $h$  put on the ordinate)
292: as observed at $R_0=(2+10^{-15})m$ (solid line) and at $R=190.07m$ (broken line).
293: The time (with values put on the abscissa) is measured from the moment
294:  of the arrival of the wave front at each of the
295: observation points. The   part of the radiation   seen at $R$ (broken line)
296: after  $t=4\pi m$ consists solely of the backscattered radiation.
297: The signal is negligible after $t>30m$ - its amplitude is much smaller than that
298: of the oscillatory part.}
299: \end{figure}
300: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
301: 
302:   The main pulse passes through $R$ in less than
303: $4\pi $ (in units of $m$) while the scattered wave appears later and
304: quickly vanishes, becoming negligible after $t=30 m$, when the tail dominates.
305: The decay of the tail asymptotics  should be like $t^{-5}$,
306: according to  \cite{Price}. This asymptotic exponent was in fact obtained
307: in many trial runs of our  numerical code with  initial pulses
308: located outside $a=3m$ \cite{expla}.
309: 
310: 
311: 
312:  Figure 2
313: in turn shows the spectral decomposition of $h=(-\partial_t+
314: \partial_{r^*})\Psi    $, obtained by employing  the FOURCO package
315: (netlib).  Fast Fourier Transform  (IMSL library) was also used,
316: with the purpose of checking the numerical results coming from the
317: application of FOURCO. We depicture
318: {\bf normalized frequencies} in order to see clearly the frequency shift
319: that is caused by the backscatter. The spectrum that is seen at $R_0 $
320: (solid line) is rescaled by $\sqrt{\eta_{R_0}}$  and the
321: spectrum seen at $R$ (broken line) is rescaled by the factor
322: $\sqrt{\eta_R}$.   In the case of negligible
323: backscattering both rescaled spectra would have to coincide.
324: 
325: Figure 2 shows the spectral amplitudes $|g(\omega )|$ in function of the
326: frequency $\omega $. There  is a number of peaks, seen at
327: both observation points. We calculated average frequency for
328: each of the resolved peaks, by choosing as the support $\tilde \Omega $
329: of each pulse the interval between the consecutive minima of
330: $|g(\omega )|$. In the case of the lowest frequency peak one can observe
331: a 25-percent blueshift, from 0.25 to 0.31, with variances 0.1
332: and 0.08, respectively; the broken and solid lines of Fig. 2
333: coincide for $\omega >0.5/m$. In the remaining peaks the effect is
334: seemingly absent. This  can be interpreted as demonstrating that  this
335:  backscatter-induced blueshift is frequency dependent.
336: $1/(4m)$ is approximately the asymptotic frequency of the quasinormal modes.
337: It can be seen from Fig. 2 that the effect of the
338:  backscattering strongly weakens
339:   modes having frequencies $\omega_b$ (as observed by an observer
340: static at $b$), such that $\omega_a\sqrt{\eta_b}$ (which
341: would be the frequency detected by an  asymptotic static observer)
342:  is smaller than  $1/(8m)$ - a half of  the asymptotic frequency of quasinormal modes.
343: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
344: \begin{figure}[2]
345: \epsfxsize=6cm
346: \centerline{\epsffile{j1.eps}}
347: \caption{Frequency amplitude $|g(\omega )|$ (depictured on the ordinate)
348: as observed at $R_0=(2+10^{-15})m$ (solid line) and at $R=190.07m$ (broken line).
349: The {\bf normalized frequencies} (see the text above)
350:  are put on the x-axis  (scaled in units of $1/m$). }
351: \end{figure}
352: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
353: 
354: 
355: Modes with $ \omega_b\sqrt{\eta_b} >1/(2m)$ are    unchanged, while those
356: in the interval $(1/(8m)<\omega_b\sqrt{\eta_b}< 1/(2m)$
357:  undergo a damping of the spectral amplitude,
358: which is  the stronger the smaller  the frequency.   The physical reason
359: for this is following.  Our initial wave pulse has to penetrate through the
360: potential barrier having maximum at $r=3m$; the  modes with frequencies
361: close to those of quasinormal modes are much stronger scattered and their
362: transmission coefficient is small.
363: 
364: Let us point out that the backscatter is negligible -- due to the
365: weakness of the effective potential  --
366: as long as a wave pulse remains in a region
367: characterized by the tortoise variable $r^*<<-1$.
368: Therefore    the standard redshift formula applies in the
369:  zone $r^*<<1$.
370: Define a shifted  radiation pulse as follows:   $\Psi (r^*+D)=\Psi(r^*)$
371: for  $D<0$). Notice that   this translated pulse   will possess
372: a {\bf different}  frequency profile. Let $b_D$ be defined by the equation $r^*(b)+D=
373: b_D+2\ln (b_D/(2m)-1)$ and let $(w_b)$ be a frequency measured by a
374: static observer located at $r^*(b)$.
375: A static observer located at $r^*(b)+D$ will   measure a blueshifted frequency
376:  $(w_b)\sqrt{\eta_b/\eta_{b_D}}$.
377:      On the other hand {\bf a static observer located at infinity
378: would detect the same set  of frequencies irrespective of $D$}.
379:    Moreover, this observer would notice that the
380:  deviation  from the redshift formula (if that can be seen  in the detected
381: frequency range) does not depend on the position of the initial pulse.
382: 
383: If however a {\bf physical} source   of a wave
384: is fixed  and shifted
385:  from $r^*(b) <<-1$ by some $D<0$  then
386:  the two local sets of frequencies $(\omega_{bi}, \omega_{b_Di}$)
387: measured by two (shifted  by a distance $D $) local static observers {\bf do coincide},
388: $\omega_{bi}=\omega_{b_Di}$. (We assume that tidal effects can be ignored.)
389: The    static observer at infinity would then see:
390: i)  two {\bf different sets of redshifted frequencies},  $(\omega_{bi,\infty })$
391: and $(\omega_{b_Di,\infty })$, with   $\omega_{bi,\infty }=\sqrt{\eta_b}\omega_{bi}$
392: and $\omega_{b_Di,\infty }=\sqrt{\eta_{b_D}}\omega_{b_Di}$
393: for all frequencies  which satisfy conditions
394:   $\omega_{bi,\infty }>>1/(4m) $ and $\omega_{b_Di,\infty }>>1/(4m)$;
395:    ii) those frequencies,   for which
396:  $\omega_{bi,\infty }\le 1/(4m)$ and/or  $\omega_{b_Di,\infty }\le 1/(4m)$, would
397:  disobey the   standard redshift formula.  It is clear  that for any fixed wave
398:  source with frequencies $\omega_i$ one can find a location  $b$ close to the
399:  black hole horizon such that some of
400: (the would-be asymptotically detected) frequencies $\eta_b\omega_i$ are of the  order
401: of  $1/(4m)$, so that (according to our numerical example) one could see a failure
402: of the standard redshift formula.
403: 
404: If that numerically discovered  phenomen is generic, then there  would
405: exist a natural cutoff (of the order of  $1/(4m)$) for the asymptotically
406: observed frequency of those wave pulses that are close to an event horizon.
407: A wave  of a  mean frequency $\omega_b$
408: satisfying the condition  $\omega_b\sqrt{\eta_a}<< 1/(4m)$
409: would not be observed   by a distant observer.
410: In the light of the above   the heuristic
411:  analysis that was reported in the
412: beginning of this section, should be probably supplemented by adding
413: the following: {\it  if a wave peak possesses a contribution
414: with frequencies much smaller than   $1/(4m)$ then (and only then) its
415: mean frequency will be influenced by the backscatter}.
416: 
417: 
418: 
419: 
420: 
421: 
422: 
423: 
424: \section{Final remarks}
425: 
426: The backscattering can modify the relation between initial and asymptotic
427: frequencies. A characteristic  frequency  (circa $1/(4m)$ in the hitherto
428: used units) that appears useful in this
429: context is (in standard units) $f_c\approx 0.25\times 31484\times M_0/m$ Hz,
430: where $M_0$   is the solar mass. Let us remark that $f_c$ is the
431: frequency  typical for  the quasinormal modes of the electromagnetic
432: radiation propagating in the Schwarzschild spacetime.
433: 
434: A numerical example of Sec. III reveals a robust difference between
435: the standard redshift prediction    (\ref{2.4}) and the actual observation.
436: It is observed that for a radiation pulse having an asymptotic frequency
437: $f_{\infty }\approx f_c$ the initial frequency $f_R$ is
438: noticeably smaller than $f_c/\sqrt{\eta_R}$, a number following
439: from the redshift formula (\ref{2.4})). The  peaks with frequencies $f_{\infty }>>f_c$
440: are seemingly not influenced by the backscattering. This probably
441: is not always true for all radiation pulses having  mean frequencies
442: much bigger than $f_c$,
443: but we believe that in the latter case the deviation from the redshift
444: formula must be small.  Rephrasing this fact in terms of initial
445: data, it is  probably fair to say that $(r^*(b)-r^*(a)<<2m$ is
446: the sufficient condition  for the validity of the redshift formula (\ref{2.4}).
447: That is a much less stringent condition than that used in
448: \cite{{ME2002}}.
449: Let us stress that in the case of astrophysical black holes of stellar
450: origin   the deviation from the  law (\ref{2.4}) can be seen
451:   by an asymptotic observer in the  radio part  of the electromagnetic
452: spectrum, with wavelengths of the order of  100 km; we do not claim that
453: this phenomenon is of astrophysical interest.
454: 
455: Let us point out that if  the  backscattering is negligible then:
456:  i) all of the energy of an outgoing pulse
457: would get  to infinity; ii) its energy flux would be diminished,
458: according to  (\ref{2.5}). The only mechanism  that can imply
459: the loss of the radiation energy is through the backscatter of the
460: radiation  off the curvature of the spacetime. This is
461: clearly shown in the wave formulation \cite{malec2000}, but a consistent
462: quantum field  theoretic treatment (in the curved Schwarzschild spacetime)
463:  should give the same answer.
464: 
465:  In the present paper the consideration is focused  only on the dipole
466: term, but a similar analysis with the same conclusions can
467: to be done in any multipole order.
468: An analogous phenomenon, leading to the demodulation of  gravitational
469: signals, will also manifest in the spectra of gravitational waves;
470: this in principle may be detected.
471: 
472: Acknowledgments.    This work has been suported
473: in part  by the KBN grant 2 PO3B  006 23.
474: 
475: 
476: 
477: \begin{references}
478: \bibitem{Sachs} J. Kristian and R. K. Sachs, {\it Astrophysics Journal}
479: {\bf 143}, 379(1965).
480: \bibitem{Ehlers} J. Ehlers, "Survey of general relativity theory",
481: in:  Relativity, Astrophysics and Cosmology, W.  Israel (ed.), Reidel
482: Publ. Comp., Dordrecht,
483: p.1   1973.
484: \bibitem{Schneider} P. Schneider, J. Ehlers and E. E. Falco,
485: Gravitational Lenses,
486:   Springer-Verlag, Berlin, Heidelberg, New York, 1992.
487: \bibitem{Strauman} N. Straumann, General Relativity and Relativistic
488: Astrophysics, Springer Verlag 1984.
489: \bibitem{ME2002}   E. Malec, {\it Class. Quantum Grav. } {\bf 19}, 571(2002).
490: \bibitem{sense} It is meant here   that $(b-a)/(a(1-2m/a)^5 )<<1$,
491: where the areal radii $a$ and $b$ are the inner and outer boundaries
492:  of the initial wave pulse.
493: \bibitem{malec2000} E. Malec, {\it Phys. Rev.} {\bf D62}(2000), 084034.
494: \bibitem{Karkowski}  J. Karkowski, E. Malec and Z.  \'Swierczy\'nski,
495:       {\it Class. Quantum Grav. } {\bf 19}, 953(2002).
496: \bibitem{Wheeler} J. A. Wheeler, {\it Phys. Rev.} {\bf 97} 511.
497: \bibitem{Price} R. Price, {\it Phys. Rev. } {\bf D5}, 2419(1972).
498: \bibitem{expla} The integration time, that would be needed in order to obtain the
499: asymptotic tail behaviour in the case of the initial data  of Sec.  III,
500: exceeds by far the integration time required in order to solve our task.
501: 
502: \end{references}
503: 
504: 
505: 
506: 
507: 
508: \end{document}
509: 
510: 
511: 
512: 
513: 
514: 
515: