1: \documentclass{cernrep}
2: \usepackage{graphicx,here}
3:
4: \def\be{\begin{equation}}
5: \def\ee{\end{equation}}
6:
7: \begin{document}
8:
9: \title{STUDY OF COINCIDENCES BETWEEN RESONANT
10: GRAVITATIONAL WAVE DETECTORS}
11:
12:
13: \author{
14: P. Astone$^1$, M. Bassan$^2$, P. Bonifazi$^3$, P. Carelli$^4$,\\
15: E. Coccia$^2$, C.Cosmelli$^5$, S.D'Antonio$^5$,
16: V. Fafone$^6$, G.Federici$^1$,\\ A. Marini$^6$,
17: Y. Minenkov$^2$, I. Modena$^2$,\\
18: G. Modestino$^6$,
19: A. Moleti$^2$, G. V. Pallottino$^5$,\\ G. Pizzella$^7$,
20: L.Quintieri$^6$, F. Ronga$^6$, R. Terenzi$^3$,
21: M. Visco$^3$, L. Votano$^6$ \\
22: $~$\\
23: $~$
24: }
25:
26: \vskip 0.1 in
27:
28: \institute{
29: {\it ${}^{1)}$ Istituto Nazionale di Fisica Nucleare INFN, Rome}\\
30: {\it ${}^{2)}$ University of Rome "Tor Vergata" and INFN, Rome}\\
31: {\it ${}^{3)}$ IFSI-CNR and INFN, Frascati}\\
32: {\it ${}^{4)}$ University of L'Aquila and INFN, Rome}\\
33: {\it ${}^{5)}$ University of Rome "La Sapienza" and INFN, Rome}\\
34: {\it ${}^{6)}$ Istituto Nazionale di Fisica Nucleare INFN, Frascati}\\
35: {\it ${}^{7)}$ University of Rome "Tor Vergata" and INFN, Frascati}
36: }
37:
38: \maketitle
39:
40: \begin{abstract}
41: Coincidences are searched with the cryogenic
42: resonant gravitational wave detectors EXPLORER and NAUTILUS,
43: during a period of about six months (2 June-14 December 1998)
44: for a total measuring time of 94.5 days, with the purpose to
45: study new algorithms of analysis, based on the
46: physical characteristics of the detectors.
47:
48: PACS:04.80,04.30
49: \end{abstract}
50:
51: %{\bf~~~~June 2000~~~~For circulation only within the ROG group}
52:
53:
54: %\pagestyle{plain}
55: %\setcounter{page}2
56: %\baselineskip=17pt
57: \section{Introduction}
58: After the initial experiments with room temperature resonant detectors, the new
59: generation of cryogenic gravitational wave (GW)
60: antennas entered long term data
61: taking operation in 1990 (EXPLORER~\cite{long}), in 1991
62: (ALLEGRO~\cite{alle}), in 1993 (NIOBE~\cite{NIOBE}),
63: in 1994 (NAUTILUS~\cite{naut}) and in 1997
64: (AURIGA~\cite{auri}).
65:
66: Recently an analysis of the data taken in coincidence among all
67: cryogenic resonant detectors in operation during the years 1997 and 1998
68: has been performed~\cite{5barre}.
69: No coincidence excess was found above background using
70: the event lists produced under the protocol of the
71: International Gravitational Event Collaboration (IGEC), among the
72: groups of ALLEGRO, AURIGA, EXPLORER / NAUTILUS and NIOBE.
73: The coincidence search was done without any particular data
74: selection. However one can consider the possibility to search
75: for coincidences with events selected according to
76: various possible criteria using all available information
77: (we mention criteria based on:
78: the event energy, the event duration, the applied threshold,
79: the shape of the events, the coincidence window, the direction of possible GW,
80: the noise).
81:
82: Here we have used algorithms based on physical
83: characteristics of the detectors, as the event energy (with a new
84: algorithm) and the directionality. In this paper we explore their
85: effect on the coincidence search.
86:
87: For this purpose we shall use IGEC data obtained from 2 June 1998
88: when NAUTILUS, after a stop for instrumental improvements,
89: resumed the operation. We search for coincidences between
90: NAUTILUS and EXPLORER, whose apparatuses differ only in the operating
91: temperatures (respectively 0.15 K and 2.6 K) and in particular
92: have identical readout systems. Extension of the methods we develop
93: here to other detectors in operation during the same period of time
94: is envisaged.
95:
96: We are well aware that any data selection
97: jeopardizes the possibility to express the results
98: by means of a $probability$ that a coincidence excess, if any,
99: had been accidental.
100: With this $proviso$ we shall still use parameters obtained from
101: probability estimations for comparing different situations.
102:
103: \begin{table}
104: \centering
105: \caption{
106: Main characteristics of the two detectors.
107: }
108: \vskip 0.1 in
109: \begin{tabular}{|c|c|c|c|c|c|c|}
110: \hline
111: detector&latitude&longitude&orientation&mass&frequencies&temperature\\
112: &&&&kg&Hz&K\\
113: \hline
114: EXPLORER&46.45 N&6.20 E&$39^o$ E&2270&904.7&2.6\\
115: &&&&&921.3&\\
116: NAUTILUS&41.82 N&12.67 E&$44^o$ E&2270&906.97&0.15\\
117: &&&&&922.46&\\
118: \hline
119: \end{tabular}
120: \label{dire}
121: \end{table}
122:
123: %
124: \section{Events and signals}
125: %
126: We now briefly describe how we obtain $events$ from the measurements.
127: For EXPLORER and NAUTILUS,
128: whose main characteristics are given in table \ref{dire},
129: the data are sampled at intervals of 4.54 ms
130: and are filtered with a filter matched to short bursts~\cite{fast}
131: for the detection of delta-like signals. The filter makes use of power
132: spectra obtained with off-line analysis.
133: After the filtering of the raw-data, $events$ are extracted as follows.
134: Be $x(t)$ the filtered output of the detector.
135: This quantity is normalized, using the detector calibration,
136: such that its square gives the energy innovation E of the oscillation
137: for each sample, expressed in kelvin units.
138: For well behaved noise due only to the thermal
139: motion of the bar and to the electronic noise of the amplifier,
140: the distribution of $x(t)$ is normal with zero mean.
141: The variance (average value of the square of $x(t)$)
142: is called $effective~temperature$ and is indicated with $T_{eff}$. The
143: distribution of $x(t)$ is
144: \be
145: f(x)=\frac{1}{\sqrt{2\pi T_{eff}}}e^{-\frac{x^2}{2T_{eff}}}
146: \label{normal}
147: \ee
148: For extracting $events$ we set a
149: threshold in terms of a critical ratio defined by
150: \be
151: CR=\frac{|x|-\bar{|x|}}{\sigma(|x|)}=
152: \frac {\sqrt{SNR}-\sqrt {\frac {2}{\pi}}}{\sqrt {1-\frac {2}{\pi}}}
153: \label{creq}
154: \ee
155: where $\sigma(|x|)$ is the standard deviation of $|x|$
156: (the moving averages $\bar{|x|}$
157: are made over the preceeding ten minutes) and
158: \be
159: SNR=\frac{E}{T_{eff}}
160: \label{teffeq}
161: \ee
162:
163: The threshold is set at CR=6 in order to obtain,
164: in presence of thermal and electronic noise alone,
165: about one hundred $events$ per day, as agreed among the partners of the IGEC.
166: This threshold corresponds to an energy
167: $E_t=19.5~ T_{eff}$. When $|x|$ goes above the threshold, its
168: time behaviour is considered until it goes below the threshold for
169: more than ten seconds. The maximum amplitude and its occurrence time
170: define the $event$.
171:
172: In general the $event$ is due to a combination of a signal which,
173: in absence of noise, has energy $E_s$ (due to
174: GW or other forces) and the noise.
175: The theoretical probability to detect a signal with a given
176: $SNR_s=\frac{E_s}{T_{eff}}$,
177: in presence of a well behaved Gaussian noise, is calculated as follows.
178: We put $y=(s+x)^2$ where
179: $s\equiv \sqrt{SNR_s}$ is the signal we look for and $x$ is the gaussian
180: noise. We obtain easily \cite{papoulis}
181: \be
182: probability(SNR_s)=\int_{SNR_t}^{\infty} \frac {1}{\sqrt{2 \pi y}}
183: e^{-\frac{(SNR_s+y)}{2}}
184: cosh(\sqrt{y\cdot SNR_s})dy
185: \label{papou}
186: \ee
187: where we put $SNR_t=\frac{E_t}{T_{eff}}=19.5$
188: for the present EXPLORER and NAUTILUS detectors.
189:
190: The behaviour of the integrand is shown in fig. \ref{snrrappo}.
191: \begin{figure}
192: \vspace{9.0cm}
193: \special{psfile=snrrappo.eps hscale=90 vscale=60 angle=0}
194: \caption{
195: Differential probability that the event has the signal-to-noise
196: ratio shown on the abscissa when the signal has $SNR_s=20$.
197: \label{snrrappo} }
198: \end{figure}
199: This figure shows the spread of the event energy due to noise for a given
200: $SNR_s$ of the applied signal. The distinction between the two concepts,
201: $signal$ and $event$, is essential for the analysis we propose in this
202: paper.
203:
204: %
205: \section{Data selection}
206: %
207:
208: All the events which are in coincidence
209: within a time window of $\pm 5~s$ with events produced by a seismometer
210: are eliminated, about $8\%$ of the events.
211:
212: It has been noticed
213: that the experimental data are affected by noise which, in some cases,
214: cannot be observed with any other auxiliary detector.
215: Thus a strategy is needed for deciding
216: when the measurements are considered to be good for the search of
217: coincidences.
218:
219: We are well aware that the selection of the experimental data must
220: be done with great care and the safest strategy is to establish rules
221: before even looking to the data.
222: We have decided to take into consideration $all$ the data recorded by the
223: detectors (except those vetoed by the seismometer)
224: and accept only the events for which the corresponding
225: $T_{eff}$ is below a certain
226: threshold. This threshold must be such that we are confident
227: that no signal is being thrown away.
228: $All$ and $only$ the events which have
229: $T_{eff} \leq 100 ~mK$ (over the preceeding ten minutes) are
230: taken into consideration.
231: The events for which the corresponding $T_{eff}$ is greater than
232: $100 ~mK$ are
233: certainly generated at times the detector is not operating properly.
234:
235: The following information is available on the IGEC Web page for each event:\\
236: Time (UT) of the maximum of the event: YEAR, MONTH, DAY, MINUTE, SECOND.\\
237: $H_o$: Bilateral Fourier amplitude at resonance of the maximum.\\
238: SNR: Signal to noise ratio of amplitude.\\
239: $T_{eff}$: Effective temperature [K] of the previous 10 minutes.\\
240: Duration L of the event, in number of samples (4.54 ms).\\
241: Time in seconds between the beginning and the maximum of the event.
242:
243: The relationship between the Fourier transform $H_o$ of the event amplitude
244: and the energy E of the event is given by \cite{australia}
245: \be
246: H_o=7.97~ 10^{-21} \sqrt{E}
247: \label{acca}
248: \ee
249: with $H_o$ in units of $\frac{1}{Hz}$ and E in kelvin.
250:
251: Looking to these events we have noticed that some events occur during
252: periods of high disturbance.
253: Since we are elaborating here strategies for data analysis, we
254: have thought convenient to select the data
255: to be used in our analysis in various ways according to the noise.
256: Thus another way of choosing the data to
257: be analysed is to select periods with smaller noise. We apply two more
258: data selections, only events with
259: $T_{eff}\leq50~mK$ for both EXPLORER and NAUTILUS
260: and only events with $T_{eff}\leq25~mK$.
261: This data selection has been applied by us in a previously
262: published paper \cite{explniobe}.
263: In fig.\ref{stazpuliti} we show the number of events
264: per day and the hourly averages of $T_{eff}$ for EXPLORER and
265: NAUTILUS for the case $T_{eff}\leq25~mK$.
266:
267: \begin{figure}
268: \vspace{9.0cm}
269: \special{psfile=stazpuliti.eps hscale=120 vscale=90 angle=0}
270: \caption{
271: $T_{eff}\leq25~mK$.
272: The upper two figures show the number of events/day for
273: NAUTILUS (left) and EXPLORER (right).
274: The lower two figures show the noise temperature $T_{eff}$ (kelvin)
275: respectively for NAUTILUS and EXPLORER, daily averaged over the events.
276: \label{stazpuliti} }
277: \end{figure}
278:
279: We notice large fluctuations, in spite of the stringent criteria
280: for the data selection. Information on the
281: various data selections are given in table \ref{2criteri}. We notice
282: that the number of available hours of measurement becomes rather
283: smaller when lowering the threshold for $T_{eff}$, so that any
284: possible result becomes statistically weaker at low $T_{eff}$.
285: \begin{table}
286: \centering
287: \caption{
288: Total number N of events, number of hours of data taking, average
289: noise temperature $<T_{eff}>$ and
290: hours in common, when both detectors were simultaneously operating.
291: }
292: \vskip 0.1 in
293: \begin{tabular}{|c|c|c|c|c||c|c|}
294: \hline
295: &$T_{eff}$&N&hours&$<T_{eff}>$&hours&N\\
296: &mK&&&mK&in common&\\
297: \hline
298: \hline
299: EXPLORER&$\leq100$&55070&3415&40.6&2271&37944\\
300: NAUTILUS&&37734&3450&19.1&&24118\\
301: \hline
302: EXPLORER&$\leq50$&39211&2759&28.9&1816&26481\\
303: NAUTILUS&&34148&3371&14.0&&16677\\
304: \hline
305: EXPLORER&$\leq25$&16172&1498&18.7&931&9765\\
306: NAUTILUS&&27823&3168&9.3&&5999\\
307: \hline
308: \end{tabular}
309: \label{2criteri}
310: \end{table}
311:
312: %
313: \section{Searching for coincidences}
314: %
315:
316: For the search of coincidences it is important to establish
317: the time window. We have decided to adopt the same window used in
318: past analyses, in particular that described in papers \cite{ae1991,5barre},
319: $w=\pm1~s$. This is a reasonable choice considering the present
320: detectors bandwidth (of the order of 1 Hz) and some time inaccuracy.
321:
322: As well known, the analysis in a coincidence search consists essentially
323: in comparing the detected coincidences at zero time delay
324: with the background, that is with coincidences occurring by chance.
325: In order to measure the background due to the accidental
326: coincidences, using a procedure adopted since the beginning of
327: the gravitational wave experiments \cite{weber},
328: we have shifted the time of occurrence of the events of one of
329: the two detectors 1,000 times in steps of 2 s, from -1,000 s to
330: +1,000 s. For each time shift we get a number of coincidences.
331: If the time shift is zero we get the number $n_c$ of $real$ coincidences.
332: The background is calculated from the average number of the $n_{shift}$
333: accidental coincidences obtained from the one thousand time shifts
334: \be
335: \bar{n}=\frac{\sum_1^{1000}n_{shift}}{1000}
336: \label{average}
337: \ee
338: With this experimental procedure for the evaluation of the background
339: we circumvent the problems arising from a non very stationary
340: distribution of the events, provided we test properly the distribution
341: of the shifted coincidences (see fig. \ref{delay} and reference \cite{afp}).
342:
343: The result of our search for coincidences is given in table \ref{combined}.
344: \begin{table}
345: \centering
346: \caption{
347: Number $n_c$ of coincidences and average number $\bar{n}$
348: of accidentals.
349: The total period of time in common when $T_{eff}\leq100~mK$ is 94.5 days.
350: }
351: \vskip 0.1 in
352: \begin{tabular}{|c|c|c|c|}
353: \hline
354: $T_{eff}$&$n_c$&$\bar{n}$&hours\\
355: \hline
356: $\leq100~mK$&223&231.7&2271\\
357: $\leq50~mK$&137&139.8&1816\\
358: $\leq25~mK$&32&36.2&931\\
359: \hline
360: \end{tabular}
361: \label{combined}
362: \end{table}
363: There is no coincidence
364: excess between EXPLORER and NAUTILUS, even for selected periods
365: with smaller noise.
366:
367: %
368: \section{Data selection using the event energy}
369: %
370:
371: We want now to apply data selection algorithms based on the
372: event energy. The most obvious one is to search for pairs of events
373: which have (approximately) the same energy. In the past this
374: energy criterion has been applied, requiring that the responses
375: of the EXPLORER detector at the two resonance modes were within
376: a factor of two one from each other\cite{long}. Later we realized
377: that the effect of the noise on
378: signals near threshold is such that the event energies are
379: only lightly correlated to the signal energies \cite{snr},
380: and this reduces the efficiency of algorithms based on
381: the event energy.
382:
383: Recently an important result was found \cite{blair}. It has been seen
384: that the distribution of the energy ratios of the event
385: energies of two detectors, in the case of non gaussian noise, is different
386: for real coincidences and accidental coincidences.
387: This has pushed us to reconsider the importance to apply selection
388: algorithms based on the event energies.
389:
390: For making use of the event energy, in particular with detectors
391: with different sensitivity, we must consider the result
392: shown in fig. \ref{snrrappo} which indicates the chance to have
393: a certain event-energy for a given signal-energy. In principle,
394: all event-energies are possible, from zero to infinity. Our procedure
395: here is to consider only event-energies within $\pm$ one sigma
396: from the signal energy (that is, we consider events included
397: in 68\% of the area under the line in fig. \ref{snrrappo}).
398:
399: We do not know the signal-energy. The new algorithm we propose
400: is the following. We consider signals in a wide range, say: $E_s$ from 20 mK
401: to 2 K in steps of 20 mK. We find the coincident events, at zero
402: delay (the real coincidences) and at shifted times (for the estimation
403: of the accidentals). For each assumed signal with energy $E_s$ we calculate
404: the $SNR_s$ different for each event, since the noise $T_{eff}$
405: depends on the detected event and it is also different for
406: the two detectors.
407: We then verify if the $SNR_{event}$ falls into $SNR_s \pm1~sigma$,
408: having calculated for each $SNR_s$ the probability curve
409: like that shown in fig. \ref{snrrappo} for $SNR_s=20$.
410: If the two event energies are compatible with the event-energy expected
411: for any of the assumed signals then we accept the coincidence
412: (real or shifted).
413:
414: The result of this analysis is given in table \ref{combi}.
415: \begin{table}
416: \centering
417: \caption{
418: Energy algorithm.
419: Number $n_c$ of coincidences, average number $\bar{n}$
420: of accidentals and the covered time period for the
421: three data selection.
422: }
423: \vskip 0.1 in
424: \begin{tabular}{|c|c|c|c|}
425: \hline
426: $T_{eff}$&$n_c$&$\bar{n}$&hours\\
427: \hline
428: $\leq100~mK$&61&50.5&2271\\
429: $\leq50~mK$&45&37.7&1816\\
430: $\leq25~mK$&11&10.3&931\\
431: \hline
432: \end{tabular}
433: \label{combi}
434: \end{table}
435: We notice that the use of the energy selection algorithm has reduced
436: the number of the accidental coincidences by a factor of three.
437:
438: %
439: \section{Event selection according to the detector orientation
440: with respect to the Galactic Centre}
441: %
442:
443: No extragalactic GW signals should be detected with the present
444: detectors. Therefore we shall focus our attention on possible
445: sources located in the Galaxy. If any of these sources
446: exist we should expect a more favorable condition of detection
447: when the detectors are oriented with their axes perpendicular
448: to the direction toward the Galactic Centre (GC), since,
449: the bar cross-section is proportional to $sin^4(\theta)$,
450: where $\theta$ is the angle between
451: the detector axis and the direction to the GC.
452:
453: After having applied the above energy algorithm,
454: we search for coincidences considering only events
455: obtained when the detectors were oriented with $\theta$ greater
456: than various given values and, according to the previous sections,
457: for the various data selection.
458:
459: The result is given in fig.\ref{versoteta}
460: and shows a larger coincidence excess when the detector
461: axes tend to be perpendicular to the direction towards the GC.
462: \begin{figure}
463: \vspace{9.0cm}
464: \special{psfile=versoteta.eps hscale=120 vscale=90 angle=0}
465: \caption{
466: The upper left figure shows, versus $\theta$ and for $T_{eff}\leq100~mK$,
467: the integral number of coincidences $n_c$ (indicated with asterisks)
468: and the average number of accidentals $\bar{n}$
469: (calculated from
470: the number of coincidences at zero delay and the average
471: background $\bar{n}$ measured with 1000 delays).
472: The right figure shows the Poisson probability that the
473: observed number of coincidences $n_c$ was due to a
474: background fluctuation.
475: Similarly the second line of figures refers to the data selection
476: $T_{eff}\leq50~mK$ and the third line to the data selection
477: $T_{eff}\leq25~mK$.
478: \label{versoteta} }
479: \end{figure}
480: Above $\theta\sim79^o$
481: the number $n_c$ of coincidences drops quickly.
482: If not instrumental, the quick drop
483: could be taken as due to the width of the source. The time spent by
484: the detectors when $\theta\geq79^o$ is 20\% of the total time
485: of 94.5 days.
486:
487: We want now to verify that the evaluation of the background
488: is properly done. We do this
489: in the condition of the greatest coincidence excess, that is for
490: $\theta\geq79^o$. We must consider
491: that by selecting only times when the detectors had certain
492: orientations we have several empty time regions. This makes it
493: possible that in doing the shifting operation for evaluating the
494: background one uses time periods of different duration. We have determined
495: these time periods and found that they vary by a few percent,
496: with a maximum of -10\% for a time shift of +1000 s.
497: In fig. \ref{delay} we show, for the case $\theta\geq79^o$
498: and $T_{eff}\leq100~mK$,
499: the delay histogram with no correction and the delay
500: histogram corrected for the different periods of time for each shift.
501: \begin{figure}
502: \vspace{9.0cm}
503: \special{psfile=delay.eps hscale=120 vscale=90 angle=0}
504: \caption{
505: Data selection with $T_{eff}\leq100~mK$.
506: In the upper figure we show, for $\theta\geq79^o$, the delay histogram.
507: In the lower figure we show the same data
508: normalized for the duration of the time period used for each time
509: shift. In this particular case the normalization turns out to be
510: very small, almost barely visible for delays above 700 s, but it is worth
511: to remark that a possible effect due to different time coverage at
512: various delays has been taken into account.
513: The biggest asterisk indicate the nineteen coincidences at
514: zero time delay.
515: \label{delay} }
516: \end{figure}
517: In this particular case the correction applies for a very small amount
518: only for delays greater than about 700 s.
519: %
520: \section{Conclusions}
521: %
522:
523: In order to make a first step to a complete analysis, we have selected the
524: IGEC events of EXPLORER and NAUTILUS using algorithms based on known physical
525: characteristics of the detectors. In particular a new algorithm
526: which makes use of the event energy has been devised.
527:
528: With event selection based on this algorithm we find an excess of coincidences
529: at zero time delay in the direction of the Galactic Centre. As well known
530: in the scientific community, no g.w. signals are expected to be observed
531: with the present detector sensitivity. Since this result would open new
532: possibilities, a careful Bayesan approach suggests that, given the
533: Poisson probabilities of a few percent, new data with other detectors
534: are required before we can ensure that g.w. from the GC have been
535: indeed observed. Thus, at present, we feel that the coincidence
536: excess is not large enough to establish a claim
537: for detection of true signals, but it is an important information
538: to make available to the scientific community.
539: We believe that the procedures adopted here might be useful for
540: detecting gravitational waves with more or better data.
541:
542: %
543: \section{Acknowledgements}
544: %
545: We thank W.O.Hamilton and W.W.Johnson for discussions and suggestions.
546: We thank the European Center for Nuclear Physics (CERN) for the hospitality
547: and for the supply of the cryogenic liquids. We thank W.O.Hamilton
548: and W.W.Johnson for useful suggestions and
549: F. Campolungo, R. Lenci, G. Martinelli, E. Serrani, R. Simonetti
550: and F. Tabacchioni for precious technical assistance.
551:
552: %
553: %\section{References}
554: %
555:
556: \begin{thebibliography}{99}
557:
558: \bibitem{long} P. Astone et al., Phys. Rev. D. \bf47\rm, 362 (1993).
559: \bibitem{alle} E. Mauceli et al., Phys.Rev. D, \bf54\rm, 1264 (1996)
560: \bibitem{NIOBE}D.G. Blair et al. Phys. Rev. Lett.\bf74\rm, 1908 (1995).
561: \bibitem{naut}P. Astone et al, Astroparticle Physics,\bf7\rm, 231(1997)
562: \bibitem{auri}M.Cerdonio et al.,
563: First Edoardo Amaldi Conference on Gravitational wave
564: Experiments, Frascati, 14-17 June 1994
565: \bibitem{5barre} "Initial operation of the IGEC Collaboration"
566: G.A. Prodi, I.S. Heng, Z.A. Allen ,P. Astone et al.
567: Submitted to 4th Gravitational Wave Data Analysis Workshop (GWDAW 99), Rome,
568: Italy, 2-4 Dec 1999.
569: Submitted to Int.J.Mod.Phys.D
570: e-Print Archive: astro-ph/0003106
571: \bibitem{fast}P.Astone, C.Buttiglione,S.Frasca, G.V.Pallottino, G.Pizzella
572: Il Nuovo Cimento \bf20\rm, 9 (1997)
573: \bibitem{papoulis} A.Papoulis "Probability, Random Variables
574: and Stochastic Processes", McGraw-Hill Book Company (1965), pag 126.
575: \bibitem{australia}P.Astone et al, in "Gravitational Astronomy"
576: Ed. D.E.McClelland and H.A.Bachor, World Scientific (1990).
577: \bibitem{explniobe} P.Astone et al. Astroparticle Phys. \bf10\rm, 83 (1999)
578: \bibitem{ae1991}P.Astone et al. Phys.Rev. D, \bf59\rm, 122001 (1999)
579: \bibitem{weber}J. Weber, Phys. Rev. Lett. \bf22\rm, 1320 (1969).
580: \bibitem{afp} "Background estimation in a gravitational wave experiment"
581: P.Astone, S.Frasca, G.Pizzella ,Dec 1999. 6pp.
582: Submitted to 4th Gravitational Wave Data Analysis Workshop (GWDAW 99), Rome,
583: Italy, 2-4 Dec 1999.
584: Submitted to Int.J.Mod.Phys.D
585: e-Print Archive: gr-qc/0002004
586: \bibitem{snr}P.Astone, G.V.Pallottino, G.Pizzella,
587: Journal of General Relativity and Gravitation, \bf30\rm, 105 (1998)
588: \bibitem{blair}D.Blair, P.Bonifazi et al.,
589: Journal of General Relativity and Gravitation, in press (2000).
590:
591: \end{thebibliography}
592: %
593: \end{document}
594:
595: