1: \documentclass[a4,12pt,fleqn]{article}
2: \usepackage{epsfig,times,latexsym,enumerate,lscape,float,flafter,amsmath,amssymb,graphicx}
3:
4: \mathindent = 0.0cm
5:
6: \newcommand{\prd}{Phys. Rev. D }
7: \newcommand{\mnras}{MNRAS }
8: \newcommand{\cqg}{Class. Quantum Grav. }
9: \newcommand{\aaa}{Astron. and Astrophys. }
10: \newcommand{\pp}{Preprint gr-qc/ }
11:
12: \hoffset = 0.71in
13: \textwidth=6.5in
14: \textheight=8.50in
15: \oddsidemargin = -0.5in
16: \evensidemargin = 0.0in
17:
18: \begin{document}
19:
20: \title{On the independent points in the sky for the search of periodic gravitational wave}
21: \author{S.K. Sahay\thanks{email: ssahay@iucaa.ernet.in, ssahay@bits-goa.ac.in}\\
22: {\small Birla Institute of Technology and Science-Pilani, Goa Campus, Physics Group, Zuarinagar, Goa - 403726, India.}}
23:
24: \date{}
25: \maketitle
26:
27: \begin{abstract} We investigate the independent points in the sky require
28: to search the periodic gravitational wave, assuming the noise power spectral density to be flat. We have made an analysis with different initial azimuth of
29: the Earth for a week data set. The analysis shows
30: significant difference in the independent points in the sky for the search. We numerically obtain an approximate relation to make trade-off between computational cost and sensitivities. We also discuss the feasibility of the coherent search in small frequency band in reference to advanced LIGO.
31: \end{abstract}
32: \indent\indent\indent{\bf Keywords:} {\it gravitational wave - data analysis - periodic sources}
33:
34:
35:
36: \section{Introduction}
37: The first generation of laser interferometric gravitational wave observatory (LIGO) [Abramovici (1992)] is in operation.
38: However, its has been planned that by the end of the initial LIGO science run it would undergo upgrades to significantly improve the sensitivity. Now referred to as advanced LIGO (formerly LIGO II) [Weinstein, 2002]. The detector will sweep their broad
39: quadrupolar beam pattern across the sky as the earth moves. Hence the data
40: analysis system will have to carry
41: out all sky searches for its sources. In this, search of the PGW without a priori knowledge appears to
42: computationally quite demanding even by the standard computers
43: expected to available in the near future. It appears that due to limited computational resource it will be not feasible to perform all sky all frequency
44: search
45: in the months/year data set. However, if advanced LIGO achieve its design sensitivity $\sim 10^{-23}$ or better (Weinstein, 2002), then it may be feasible to perform all sky search for a day to week data set in small frequency
46: band for the sources emitting signal of amplitude $\gtrsim 10^{-26}$. The
47: search of the potential sources may be more significant, if done in
48: the frequency
49: band where most of the Pulsars are detected by other means. Also, the choice of sophisticated, optimal data analysis methods and a clever programming is also
50: integral part to search the signal buried in the noise with the available computation power.
51:
52: \par The current status of the
53: search indicates that its important to detect the gravitational waves (GW) rather finding the source
54: location more accurately. Hence, one would like to do minimum Doppler
55: correction or/and to make the search
56: templates. In reference to the all sky search, Schutz (1991), has introduced the
57: concept of patch in the
58: sky as the region of space throughout which the required Doppler correction
59: remains the same. He roughly estimated the number of patches required and shown that
60: for $10^7$ sec. observation data set of one KHz signal would be about $1.3
61: \times 10^{13}$. This also indicates the bank of templates require
62: for the coherent all sky search. The size of the patch may be increased, hence reducing the number of points require for the Doppler correction,
63: by manipulating the output of the detector. Which in turn demand the detail investigation
64: of the parameters affecting the phase of the modulated signal. In this, the initial azimuth of the Earth play a vital role in the modulation of the signal, particularly for the analysis of day to week data set. Hence, in the next section, we incorporate the
65: initial azimuth of the Earth in the Fourier transform (FT) obtained by
66: Srivastava and Sahay (2002a,b) for arbitrary observation time.
67:
68: \par The limited computation power make to search the signal in the short observation data set, may be a day/week, so in section 3, employing the concept of fitting factor (Apostolatos, 1995), we investigate the independent points in the sky ($N_{sky}$) require for an all sky
69: search for a week data set with different initial azimuth of the Earth,
70: assuming the noise power spectral density to be flat. Also, we numerically obtain an approximate relation to make trade-off between computational cost and sensitivities. In section 4, we discuss the feasibility of the coherent search in reference to advanced LIGO. Section 5 contains the conclusions of the paper.
71:
72: \section{Fourier transform}
73:
74: The FT analysis of the frequency modulated PGW signal has been done by
75: Srivastava and Sahay (2002a,b) by taking account the effects arising due to
76: the rotational as well
77: as orbital motion of the Earth. However, they have neglected an important
78: parameter, the initial azimuth of the Earth, which affects significantly in the spacing of the parameters space for an all sky search. To obtain the FT with taking account the Earth initial azimuth, we rewrite the phase of the received PGW signal of frequency $f_\circ$ at
79: time $t$ given by them and may be written as
80:
81: \begin{eqnarray}
82: \Phi (t) & = & 2\pi f_\circ t + {\cal Z}\cos (a\xi_{rot} - \sigma ) + {\cal N}\cos (\xi_{rot} - \delta ) - {\cal M}
83: \end{eqnarray}
84: \newpage
85: \noindent where \\
86: \begin{equation}
87: \left.\begin{array}{lcl}
88: \vspace{0.2cm}
89: {\cal M}& = & \frac{2\pi f_\circ}{c}\left(R_{se}\sin\theta\cos\sigma \sqrt{{\cal P}^2 + {\cal Q}^2 }\cos\delta\right) , \\
90: \vspace{0.2cm}
91: {\cal Z}& = & \frac{2\pi f_\circ}{c} R_{se}\sin\theta\, ,\\
92: \vspace{0.2cm}
93: {\cal N}&= & \frac{2\pi f_\circ}{c} \sqrt{{\cal P}^2 + {\cal Q}^2 } \, ,\\
94: \vspace{0.2cm}
95: {\cal P}& = & R_e\sin\alpha (\sin\theta\sin\phi\cos\epsilon + \cos\theta \sin\epsilon )\, ,\\
96: \vspace{0.2cm}
97: {\cal Q}& = & R_e \sin\alpha\sin\theta\cos\phi \, ,\\
98: \vspace{0.2cm}
99: \sigma &=& \phi - \beta_{orb}\, , \quad \delta = \tan^{- 1}\frac{{\cal P}}{{\cal Q}} - \beta_{rot}\, , \\
100: \vspace{0.2cm}
101: a &= & w_{orb}/w_{rot}\; \approx \; 1/365.26, \; w_{orb}t \; = \; a\xi_{rot} , \\
102: \vspace{0.2cm}
103: {\bf\textstyle n} & = & \left(\sin\theta\cos\phi , \; \sin\theta\sin\phi\, , \;\cos\theta\right), \; \xi_{rot} \; =\; w_{rot}t \\
104: \end{array} \right\}
105: \end{equation}
106:
107: \noindent where $\theta$, $\phi$, $R_{e}$, $R_{se}$, $w_{orb}$ $w_{rot}$, $\alpha$ and
108: $\epsilon$ represent
109: respectively the celestial co-latitude, longitude, Earth radius, average distance between Earth centre from
110: the origin of SSB frame, orbital and rotational angular velocity of the
111: Earth, co-latitude of the detector and obliquity of the ecliptic. Here $\beta_{orb}$ and $\beta_{rot}$ are the initial azimuth of
112: the Earth and detector respectively.\\
113:
114: \noindent To estimate $N_{sky}$, it is sufficient to consider either of the two polarisation of the signal given as,
115: \begin{equation}
116: h_+(t) = h_{\circ_+}\cos [\Phi (t)]
117: \label{eq:hpt}
118: \end{equation}
119: \begin{equation}
120: h_\times (t) = h_{\circ_\times}\sin [\Phi (t)]
121: \label{eq:hct}
122: \end{equation}
123: \noindent hence, we consider the `$+$' polarisation of amplitude unity, given as
124: \begin{equation}
125: h(t) = \cos[\Phi (t)]
126: \label{eq:cosphit}
127: \end{equation}
128:
129: \noindent where $h_{\circ_+}$, $h_{\circ_\times}$ are constant amplitude of the two
130: polarizations.\\
131:
132: Now let us assume $h(t)$ to be given on finite time interval $[0,T_{obs}]$ assumed to be the observation period. Now its straight forward to obtain FT in the similar way as obtain by Srivastava and Sahay (2002b), and may be given as
133:
134: \begin{eqnarray}
135: \tilde{h}(f)& =& \int_0^{T_{obs}} \cos[\Phi (t)]e^{-i2\pi ft}dt \nonumber \\
136: && \simeq \frac{\nu}{2 w_{rot}} \sum_{k = -
137: \infty}^{k = \infty} \sum_{m = - \infty}^{m = \infty} e^{ i {\cal A}}{\cal
138: B}[ \tilde{{\cal C}} - i\tilde{{\cal D}} ] \; ; \;
139: \label{eq:hf}
140: \end{eqnarray}
141:
142: \noindent where
143: \begin{equation}
144: \left.\begin{array}{lcl}
145: \vspace{0.2cm}
146: \nu & = & \frac{f_\circ - f}{f_{rot}} \\
147: \vspace{0.2cm}
148: {\cal A}& = &{(k + m)\pi\over 2} - {\cal M} \\
149: \vspace{0.2cm}
150: {\cal B} & = & {J_k({\cal Z}) J_m({\cal N})\over {\nu^2 - (a k + m)^2}}\\
151: \vspace{0.2cm}
152: \tilde{{\cal C}} &=& \sin \nu\xi_\circ \cos ( a k \xi_\circ + m\xi_\circ - k \sigma - m \delta )\\
153: \vspace{0.2cm}
154: && - { a k + m \over \nu}\{\cos\nu\xi_\circ \sin ( a k \xi_\circ + m\xi_\circ - k \sigma - m \delta )+ \sin ( k \sigma + m \delta )\}\\
155: \vspace{0.2cm}
156: \tilde{{\cal D}} & = & \cos \nu\xi_\circ \cos ( a k \xi_\circ + m\xi_\circ - k \sigma - m \delta ) \\
157: \vspace{0.2cm}
158: && + {k a + m \over \nu}\sin \nu \xi_\circ \sin ( a k \xi_\circ + m\xi_\circ - k \sigma - m \delta ) - \cos ( k \sigma + m \delta )\\
159: \vspace{0.2cm}
160: \xi_o & = & w_{rot}T_{obs}
161: \end{array} \right\}
162: \end{equation}
163:
164: \noindent where J stands for the Bessel function of first kind. Using the symmetrical property of Bessel function we reduce the computation time appreciably by rewriting $\tilde{h}(f)$ as
165: \begin{eqnarray}
166: \tilde{h}(f)&\simeq & { \nu \over w_{rot}}\left[ {J_\circ({\cal Z}) J_\circ({\cal N}) \over 2\nu^2}\left[ \{ \sin{\cal M} - \sin
167: ({\cal M} - \nu\xi_\circ )\}\; \right.+ i\{ \cos{\cal M} - \cos ({\cal M} - \nu\xi_\circ
168: )\} \right]\;+ \nonumber \\
169: && J_\circ ({\cal Z})\sum_{m = 1}^{m = \infty} {J_m({\cal N})\over
170: \nu^2 - m^2} \left[ ( {\cal Y} {\cal U} - {\cal X} {\cal V} ) - i (
171: {\cal X} {\cal U} + {\cal Y} {\cal V} ) \right]\; + \nonumber \\
172: && \left.\sum_{k = 1 }^{k = \infty}\sum_{m = -
173: \infty}^{m = \infty} e^{ i {\cal A}}{\cal B}\left(
174: \tilde{{\cal C}} - i\tilde{{\cal D}} \right)\right]\; ;
175: \label{eq:fm_code}
176: \end{eqnarray}
177:
178: \begin{equation}
179: \left.\begin{array}{ccl}
180: {\cal X}& =& \sin ({\cal M} - m \pi/2 )\\
181: {\cal Y}& =& \cos ({\cal M} - m \pi/2 )\\
182: {\cal U}& =& \sin \nu\xi_\circ \cos m ( \xi_\circ - \delta ) - {m\over \nu}\left\{\cos
183: \nu\xi_\circ \sin m ( \xi_\circ - \delta ) - \sin m\delta\right\}\\
184: {\cal V}& =& \cos \nu\xi_\circ \cos m ( \xi_\circ - \delta ) + {m\over \nu}\sin
185: \nu\xi_\circ \sin m ( \xi_\circ - \delta ) - \cos m\delta\\
186: \end{array}\right\}
187: \end{equation}
188:
189: The transform contains double infinite series of Bessel function.
190: However, for analysis the order of Bessel function required to compute
191: $\tilde{h}(f)$ in the infinite series are given as (Srivastava and Sahay, 2002b).
192: \begin{equation}
193: k\approx 3133.22 \times 10^3 \sin\theta \left(\frac{f_\circ}{1 kHz}\right)\, ,
194: \end{equation}
195: \begin{equation}
196: m\approx 134 \left(\frac{f_\circ}{1 kHz}\right).
197: \end{equation}
198:
199: The accuracy and range of validity for large order and/or argument has been discussed by Chishtie et.al (2005).
200:
201: \section{Independent points in the sky}
202: The study of the independent points for an all sky search
203: has been made by many research workers [Schutz
204: (1991), Brady et. al. (1998), Brady and Creighton (2000), Jaranowski and Kr\'olak (1999, 2001), Astone et.al (2002)] for the coherent and/or incoherent search. The coherent search means
205: cross correlating the data with the bank of search templates. While incoherent search implies adding of the power spectra by dividing the data into $N$ subsets, performing a full search for each subset, and adds up the power spectra of the resulting
206: searches. In this case, there is loss in S/N ratio by a factor of $\sqrt{N}$ in relation to coherent search as power spectra are added incoherently.
207: However, irrespective of the method of search, the optimal spacing in the $(\theta,\phi)$ parameters for an all sky search is a problem of interest.
208:
209: \par For the coherent search, one have to make bank of search templates to detect the
210: signal. The bank of search
211: templates are discrete set of signals from
212: among the continuum of possible signals. Consequently all the
213: signals will not get detected with equal efficiency. However, it is possible to
214: choose judiciously the set of templates so that all the signals of a given
215: amplitude are
216: detected with a given minimum detection loss. Fitting factor ($FF$) is one of the standard
217: measure for deciding what class of wave form is good enough and
218: quantitatively describes the
219: closeness of the true signals to the
220: template manifold in terms of the reduction of S/N arising due to the cross
221: correlation of
222: a signal outside the manifold with the best matching templates lying inside the
223: manifold, given as
224:
225: \begin{eqnarray}
226: {FF} & = & \frac{\langle h(f)|
227: h_T(f;\theta_T , \phi_T)\rangle}{\sqrt{\langle h_T(f;\theta_T , \phi_T )|h_T(f;
228: \theta_T , \phi_T )\rangle\langle h(f)|h(f)\rangle}}
229: \label{eq:ff1}
230: \end{eqnarray}
231:
232: \noindent where $h(f)$ and $h_T(f; \theta_T , \phi_T)$ represent the
233: FTs of the actual signal and the templates respectively. The inner product of two
234: signal $h_1$ and $h_2$ is defined as
235: \begin{eqnarray}
236: \langle h_1|h_2\rangle & =& 2\int_0^\infty \frac{\tilde{h}_1^*(f)\tilde{h}_2(f)
237: + \tilde{h}_1(f)\tilde{h}_2^*(f)}{S_n(f)}df \nonumber \\
238: & = &
239: 4\int_0^\infty \frac{\tilde{h}_1^*(f)\tilde{h}_2(f)}{S_n(f)}df
240: \label{eq:ip}
241: \end{eqnarray}
242:
243: \noindent where $^*$ denotes complex conjugation and $S_n(f)$ is the
244: spectral noise density of the detector.
245:
246: \par To estimate optimal spacing in the parameters space one have to careful investigate the parameters contain in the
247: phase of the modulated signal. Hence we check
248: the
249: effect for different $\beta_{orb}$ in $ \tilde{h}(f)$ for a week data set for the LIGO detector at Livingston [the position and orientation of the detector
250: can be found in Allen (1995)] of unit amplitude signal for $f_\circ = 50$ Hz and ($\theta,\phi$) $=$($\pi /18, \pi /4$). We take the ranges of $k$ and $m$ as 1 to 27300 and -15
251: to 15 respectively. We found that Earth azimuth affects the FM spectrum severely. Hence, for the coherent search, we investigate its effect for the number of independent of points in the sky.
252:
253: \par To estimate $N_{sky}$, we consider the LIGO detector at Livingston, receive a PGW signal of
254: frequency $f_\circ =50$ Hz from a source
255: located at $(\theta , \phi) = (1^\circ,45^\circ)$. First, we chosen the data set such that $\beta_{orb} =0$ at $t=0$. In this case we take
256: the ranges of $k$ and
257: $m$ as $1$ to $2800$ and $-15$ to $15$ respectively and bandwidth equal to
258: $50 \pm 3.28 \times 10^{-4}$ Hz for the integration. Now, we select
259: the spacing $\bigtriangleup\theta = 0.45 \times 10^{-4}$, thereafter we maximize over $\phi$ by introducing
260: spacing $\bigtriangleup\phi$ in the so obtained $N_{sky}$ and determine
261: the resulting $FF$. In similar manner we obtain $N_{sky}$ for $\beta_{orb}=\pi/6$ and $\pi/2$. The results obtained are shown in the Fig.~(\ref{fig:Nsky}). We also plot the templates spacing $\bigtriangleup\phi$ in the $\phi-$parameter with FF shown in the Fig.~(\ref{fig:deltaphi}). Interestingly, the nature
262: of these curves are similar. we have obtained a best fit of the graphs, given as
263: \begin{equation}
264: N_{sky} = 10^{18} [a_\circ +a_1x -a_2x^2 +a_3x^3 -a_4x^4 +a_5x^5 -a_6x^6 +a_7x^7]; \quad 0.85 \le x \le 0.995;\end{equation}
265: \begin{equation}
266: FF = b_\circ + b_1y - b_2y^2 + b_3y^3;\quad 0.037 \le y \le 0.69;
267: \label{eq:ff}
268: \end{equation}
269: \noindent where $a_\circ..a_7$ and $b_\circ..b_3$ are constants as given in
270: Table 1 and 2 respectively.\\
271:
272: \begin{table}[h]
273: %\begin{table}[h]
274: \label{coefficients1}
275: \centering
276: {\begin{tabular}{ccccccccc}\\
277: \hline
278: %&&&&&&\\
279: $\beta_{orb}$ & $a_\circ$ & $a_1$ & $a_2$ & $a_3$ & $a_4$ & $a_5$ & $a_6$ & $a_7$ \\
280: %&&&&&&\\
281: \hline
282: &&&&&&\\
283: $0$&$-1.71829$& $13.1670 $ & $43.2234$ & $78.7956 $&$86.1505$&$56.4921$&$20.5716 $&$3.20919$\\
284: &&&&&&\\
285: $\pi/6$&$-35.7699$& $273.770$ & $897.636$ & $1634.42$&$1784.84 $&$1168.99$&$425.178$&$66.2488$\\
286: &&&&&&\\
287: $\pi/2$ &$-40.7053$& $311.849$ & $1023.49$ & $1865.42 $&$2039.12 $&$1336.86$&$486.720$&$75.9142$\\
288: &&&&&&\\\hline
289: \end{tabular}}
290: \caption{Coefficients of the best fit graphs obtained for the $N_{sky}$ with $FF$.}
291: \end{table}
292:
293: \begin{table}[hbt]
294: \label{coefficients2}
295: \centering
296: {\begin{tabular}{ccccc}\\
297: \hline
298: %&&&&\\
299: $\beta_{orb}$ & $b_\circ$ & $b_1$ & $b_2$ & $b_3$\\
300: \hline
301: %&&&&\\
302: 0 & $0.994123$ & $0.149451$ & $3.94494$ & $2.64819$\\
303: %&&&&\\
304: &&&&\\
305: $\pi/6$ & $0.993544$ & $2.00186$ & $366.375$ &$1651.88$\\
306: %&&&&\\
307: &&&&\\
308: $\pi/2$ &$0.996774$ & $2.11099$ &$2914.21$ &$37817.3$\\
309: %&&&&\\
310: \hline
311: \end{tabular}}
312: \caption{Coefficients of the best fit graphs obtained for the $FF$ with $\bigtriangleup\phi$.}
313: \end{table}
314:
315: In view of the above investigation, the
316: spacing $\bigtriangleup\phi $ in the $\phi-$parameter may be
317: expressed as
318: \begin{equation}
319: \bigtriangleup\phi = {\cal G}(FF, f_\circ, \theta ,\phi ,T_{obs}, \beta_{orb})
320: \label{eq:phifn}
321: \end{equation}
322:
323: Equation~(\ref{eq:ff}) is a third order polynomial, hence complicates the solution. However, one would like to do data analysis for $FF > 0.90$. Therefore from the figure~(\ref{fig:deltaphi}) we obtain a very good dependence of $FF$ for minimum $N_{sky}$ and may be given as
324: \begin{equation}
325: FF = - 2.85657 \bigtriangleup\phi^2 + 0.00230059 \bigtriangleup\phi + 1.00018
326: \label{eq:phiff}
327: \end{equation}
328:
329: \noindent From Eqs.~(\ref{eq:phifn}), and~(\ref{eq:phiff}), we may write
330: \begin{eqnarray}
331: &{\cal F}(FF,50,1^\circ,45^\circ,1w,\beta_{orb}) \approx 4.02684 \times 10^{-4} \pm 0.591667 \sqrt{1.00018 - FF}
332: \label{eq:tradeoff}
333: \end{eqnarray}
334:
335: Above equation can be relevant to make trade-off between computational costs and sensitivities i.e. for the selected $FF$ one can estimate $N_{sky}$. However, there is no unique choice for it.
336: Here we are interested in the estimation of $N_{sky}$ such that the
337: spacing is maximum resulting into the least number of points require for the search. As mentioned earlier,
338: there is stringent requirement on reducing computational costs. Accordingly,
339: there is serious
340: need of adopting some procedure/formalism to achieve this. For example, one
341: may adopt the method of
342: hierarchical search.
343:
344: \begin{figure}[h]
345: \centering
346: \epsfig{file=N_sky.ps,height=16.0cm,angle=-90}
347: \vspace{1.0cm}
348: \caption{$N_{sky}$ with FF at different $\beta_{orb}$.}
349: \label{fig:Nsky}
350: \end{figure}
351:
352: \begin{figure}[h]
353: \centering
354: \epsfig{file=deltaphi_FF.ps,height=16.0cm,angle=-90}
355: \caption{FF with $\bigtriangleup\phi$ at different $\beta_{orb}$.}
356: \label{fig:deltaphi}
357: \end{figure}
358:
359:
360: \section{Computational costs}
361: In view of the above analysis it will be interesting to know the feasibility
362: of the all sky search with the target sensitivity of the advanced LIGO. The computational costs of the data analysis basically depends on the floating point operations (flops) require to perform the Fast Fourier transform (FFT). Hence in terms of FFT, the flops for the data reduction upto frequency $f$ for $T_{obs}$ of the interferometer output may be given as (Press et al., 1986)
363: \begin{equation}
364: N_{flops}= 2fT_{obs}log_2(2f_{max}T_{obs})
365: \end{equation}
366:
367: \noindent Now for the given mismatch ($FF$), if $N_p$ is the number of
368: independent points to perform all sky search, then the flops will be
369: \begin{equation}
370: N_{flops}(FF,N_p)= 2fT_{obs}N_plog_2(2f_{max}T_{obs})
371: \end{equation}
372:
373: Hence, for the mismatch of 3\% and without manipulating the data in reference to
374: $\beta_{orb}$, the flops for the search of PGW signal upto 50 HZ in the week data set
375: will be $2.38 \times 10^{19}, 2.58 \times 10^{20}$ and $6.23 \times 10^{20}$
376: for $\beta_{orb} = 0,\pi/6$ and $\pi/2$ respectively, assuming other operation need negligible flops compare to FFT. However, the lower cut off frequency of the LIGO I/II is 10/40 Hz. Hence the analysis shall be done above the lower cut off. Also, the search will be more significant if one perform in the most sensitive band of the detector. In this, if one would like to perform all sky search in a small band say 5 Hz then the minimum flops for the on-line analysis
377: (a week data gets analyzed in a $\sim$ week time) will be
378: $3.44 \times 10^{12}$. The flops require may be further reduce, if one perform hierarchal search. Hence, it may be feasible to perform limited frequency all sky search of signal amplitude $\gtrsim 10^{-26}$ in the output of such a sensitive detector with a $\sim$Tflops computer.
379:
380: \section{Conclusions}
381:
382: In this paper we have incorporated the initial azimuth of the Earth in the
383: FT of the frequency modulated PGW signal and investigated its effect in the independent points in the sky require for the search of PGW. We found that the $N_{sky}$ for the search in the output of one week data set varies significantly with $\beta_{orb}$. For the case investigated here, we observe that for $FF=0.97$ approximately $1.53 \times 10^{10}$, $1.66 \times 10^{11}$ and $4.0 \times 10^{11}$
384: $N_{sky}$ will be require when
385: $\beta_{orb} = 0, \pi/6$ and $\pi/2$ respectively.
386: Hence, the analysis may be useful to reduce the computational cost for a coherent all sky search. However, also the inspection of the phase of the modulated signal reveals that reduction in $N_{sky}$ depends on time scale of integration,
387: shorter the $T_{obs}$ more the difference in $N_{sky}$.
388:
389: \par The reduction in $N_{sky}$ is large, so we studied the feasibility of all sky search in reference to advanced LIGO and found that in the band of 5 Hz one may perform on-line all sky search of the PGW signal of amplitude $\gtrsim 10^{-26}$ with a $\sim$ Tflops computer. The relation given by equation~(\ref{eq:tradeoff}) may be useful to make trade-off between computational costs and sensitivities for the search of periodic gravitational waves. The issue to reduce the flops for the all sky search is a problem of interests and hence need more studies/investigations.\\
390:
391: \section*{Acknowledgements}
392: I am thankful to Prof. D.C. Srivastava, Department of Physics, DDU Gorakhpur University, Gorakhpur for useful discussions. I thankfully acknowledge the computation facilities provided by IUCAA.
393:
394: \begin{thebibliography}{}
395: \bibitem{} Abramovici A., Althouse W.E., Drever R.W.P.,
396: G$\ddot{u}$rsel Y., Kanwamura S., Raab F.J. Shoemaker D., Sievers L.,
397: Spero R.E., Thorne K.S., Vogt R.E., Weiss R., Whitcomb S.E., and
398: Zucker Z.E., Sciences {\bf 256}, 325, (1992); http://www.ligo.caltech.edu/
399: \bibitem{} Allen, B., 1996, arxiv:gr-qc/9607075
400: \bibitem{} Apostolatos T.A., 1995, \prd {\bf 52}, 605
401:
402: \bibitem{} Astone P., Borkowski K.M., Jaranowski P. and Kr\'olak A.,
403: 2002, \prd {\bf 65}, 042003
404:
405: \bibitem{} Brady P.R., Creighton T., Cutler C., Schutz B.F., 1998, \prd , 57, 2101
406: \bibitem{} Brady, P.R. and Creighton T, 2000, \prd , 61, 082001
407:
408: \bibitem{} Chishtie F.A, Valluri S.R., Rao K.M., Sikorski D., Williams, 2005 arxiv: math-ph/0503037; High Performance Computing Symposium, May 15-18 (HPCS 2005) Guelph, Ontario, Canada
409:
410: \bibitem{} Jaranowski P. and Kr\'{o}lak A., 2000, \prd , {\bf 61}, 062001
411:
412: \bibitem{} Jaranowski P. and Kr\'{o}lak A., 1999, \prd , {\bf 59}, 063003
413: \bibitem{} Press W.H., Flannery B.P., Teukoslsky S.A., Vetterling W.T./, 1986, Numerical Recipes: The Art of Scientific Computing, Cambridge Univ. Press, Cambridge
414: \bibitem{} Schutz, B.F., 1991, in Blair D.G. ed., The Detection of Gravitational Waves, Cambridge Universtiy Press, Cambridge
415: \bibitem{} Srivastava D.C. and Sahay S.K., 2002a, \mnras, 337, 305
416: \bibitem{} Srivastava D.C. and Sahay S.K., 2002b, \mnras, 337, 315
417: \bibitem{14} Sahay S.K., 2003, Int. J. Mod. Phys. D., 2003, Vol. 12, No. 7, 1227
418: \bibitem{} Srivastava D.C. and Sahay S.K., 2002c, \mnras, 337, 322
419: \bibitem{} Tagoshi H., Kanda N., Tanaka T., Tatsumi D., Telada S., Ando M., Arai K., Araya A., Asada H. et al. 2001, \prd {\bf 63}, 062001
420:
421: \bibitem{} Weinstein. A, 2002, \cqg {\bf 19}, 1575-1584; http://www.ligo.caltech.edu/$\sim$ligo2/
422:
423: \end{thebibliography}
424: \end{document}
425:
426:
427: