astro-ph0204487/ms.tex
1: %\documentstyle[12pt,aaspp4,flushrt,psfig]{article}
2: \documentclass[12pt,preprint]{aastex}
3: %\documentstyle[emulateapj,psfig]{article}
4: \newcommand{\beq}{\begin{equation}}
5: \newcommand{\beqa}{\begin{eqnarray}}
6: \newcommand{\eeq}{\end{equation}}
7: \newcommand{\eeqa}{\end{eqnarray}}
8: \newcommand{\sol}{M_{\sun}}   
9: \newcommand{\etal}{{\it et al. }}
10: \newcommand{\siml}{\la}
11: \newcommand{\simg}{\ga}
12: \newcommand{\lsim}{\la}
13: \newcommand{\gsim}{\ga}
14: \newcommand{\psim}{\mbox{\raisebox{-1.0ex}{$~\stackrel{\textstyle propto}
15: {\textstyle \sim}~$ }}}
16:   
17: \newcommand{\lmk}{\left(}
18: \newcommand{\rmk}{\right)}
19: \newcommand{\lnk}{\left\{ }
20: \newcommand{\nn}{\nonumber}
21: \newcommand{\rnk}{\right\} }
22: \newcommand{\lkk}{\left[}
23: \newcommand{\rkk}{\right]}
24: \newcommand{\lla}{\left\langle}
25: \newcommand{\p}{\partial}
26: \newcommand{\rra}{\right\rangle}
27: %\slugcomment{KUNS-1722}
28: %
29: \shorttitle{}
30: \shortauthors{Takahashi \& Seto}
31: 
32: \begin{document}
33: 
34: %\if0
35: \title{
36: Parameter estimation for  Galactic binaries
37:  by LISA
38: }
39: \author{Ryuichi Takahashi}
40: \affil{
41: Department of Physics, Kyoto University,
42: Kyoto 606-8502, Japan 
43: }
44: \author{Naoki Seto}
45: \affil{
46: Department of Earth and Space Science, Osaka University,
47: Toyonaka 560-0043, Japan
48: }
49: 
50: %\date{April 18, 2001}
51: %\fi
52: 
53: \begin{abstract}
54: We calculate how accurately parameters of the short-period binaries
55:  $(10^{-4}~{\mbox Hz} \lsim f\lsim 10^{-2}~{\mbox Hz})$
56:  will be determined from the gravitational waves   by LISA.
57: In our analysis   the chirp signal ${\dot f}$ is newly included as a
58:  fitting parameters and dependence on observational period  or wave
59:  frequency is studied in detail. Implications for gravitational
60: wave astronomy are also discussed quantitatively.
61: \end{abstract}
62: 
63: \keywords{gravitational waves -- gravitation -- binaries}
64: %\pacs{PACS numbers:  98.80.Es; 98.80.Cq}
65: %\vfill\eject
66: 
67: %\baselineskip 8mm
68: 
69: \section{Introduction}
70: 
71: The Laser Interferometer space Antenna (LISA), a joint project of 
72: NASA
73: and ESA (European Space Agency) would establish
74: gravitational wave astronomy at low frequency band $(10^{-4}~{\mbox 
75: Hz}
76: \lsim f\lsim 10^{-1}~{\mbox Hz})$.   It would bring us essentially 
77: new
78: information of the  Universe (Bender et al. 1998). For example
79: gravitational waves from 
80: merging super massive black holes (SMBHs) would be detected with
81: significant 
82: signal-to-noise ratio ($SNR$) $\gsim 10^3$,  though event rate of 
83:  such merging is highly unknown  ({\it e.g.}
84: Haehnelt 1994,  Vecchio 1997). Galactic binaries are promising sources 
85: of
86: LISA (Mironowskii 1965, 
87: Evans, Iben \& Smarr 1987, Hils, Bender \& Webbink 1991, Webbink \& Han
88: 1998).  
89: Gravitational
90: waves from some known binaries ({\it e.g.} X-ray 
91: binary 4U1820-30) would  be detected with $SNR>5$ by one year
92: integration (Bender et al. 1998). In addition more than thousands of 
93: close
94: white dwarf binaries (CWDBs) are expected to exist in LISA band (for
95: recent studies see Yungelson et al 2001, Nelemans et al. 2001,
96: Napiwotzki et al. 2002). 
97: Our  target in this article is these Galactic binaries. We examine 
98: how
99: accurately information of binaries can be extracted from 
100: gravitational
101: waves observed  by LISA.
102: 
103: Cutler (1998) studied the estimation errors   for binary parameters 
104: with
105: special attention to  angular variables  such as direction and
106: orientation of binaries (see also Peterseim et al. 1997; Vecchio \&
107: Cutler 1998; Cutler \& Vecchio 1998;  Hughes 2002; Moore \&  
108: Hellings 2002).
109: He used approximation that emitted 
110: gravitational waves would be  monochromatic, namely
111: neglected the effects of the chirp signal $\dot f$.
112: But the wave frequency or the chirp signal are fundamental quantities
113: for gravitational wave astronomy. From the measured chirp signal
114:  $\dot f$ we can obtain the so called chirp mass
115: $M_c=M_1^{3/5}M_2^{3/5}(M_1+M_2)^{-1/5}$  ($M_1, M_2$: masses of two
116: stars) for a binary whose orbital evolution is determined by
117: gravitational radiation reaction. Furthermore the distance to the 
118: binary
119: could be estimated from the chirp signal $\dot f$ and the amplitude of
120: the wave signal (Schutz 1986). 
121: The frequency $f$ itself  also contains important information. One of 
122: the
123: authors  (Seto 2001) pointed out that signature of the periastron 
124: advance
125: could be detected in gravitational waves  from an eccentric binary by
126: measuring its wave frequencies preciously. If this method works well,
127:  we can estimate
128: the total 
129: mass $M_{total}=M_1+M_2$ of the binary beside the chirp mass, and
130: consequently 
131: each mass of the binary is obtained separately.
132: 
133: 
134: Estimation errors for fitting parameters
135: correlate  complicatedly to each other and
136:  depend largely on observational situations. For example longer
137: observational periods  would improve not only signal to noise ratio but
138: also
139: resolution 
140: of the frequency space. Note that the latter is crucial for reducing 
141: Galactic
142: binary confusion noise, as the number of resolved
143: binaries increases with decrease of the frequency bin $\propto T_{obs}^
144: {-1}$ 
145: (see Seto 2002 for details). 
146: At LISA age it would become an interesting observational
147: challenge 
148: to  
149: optically  identify the binaries whose gravitational waves are 
150: detected
151: by LISA. We also discuss impacts of these observational efforts on
152:  estimation of binary parameters.  
153: 
154: 
155: 
156: The magnitude of
157: the chirp 
158: signal $\dot f$ has strong dependence on   the wave frequency $f$ as
159: ${\dot f}\propto M_c^{5/3}f^{11/3}$.  At lower frequencies the
160: estimation error 
161: $\Delta {\dot f}$ would be larger than the signal $\dot f$ itself.  
162: Then
163: it would be better to remove  $\dot f$ from  fitting parameters 
164: and
165: simply 
166: put ${\dot f}=0$ from the beginning. By evaluating the estimation 
167: error
168: $\Delta {\dot f}$ we can quantitatively  discuss these prescriptions 
169: for
170:  signal  analysis.  
171: 
172: 
173: In this article we only study parameter estimation errors on the
174: assumption that (i) signal has been detected with high SNR, and (ii)
175: noises are Gaussian distributed. We evaluate the estimation errors 
176: using
177: the Fisher 
178: information matrix for maximum likelihood method. For low SNR our 
179: simple
180: analysis would not be valid,  as the probability distribution function 
181: of the
182: fitting parameters could become highly complicated ({\it e.g.}
183: multimodal) (Balasubramanian \&
184: Dhurandhar 1998). Thus our result should
185: be regarded as a lower bound of the estimation errors. Non-Gaussian
186: nature of noises would farther increases 
187: errors. This is a very important problem, but detailed quantitative
188: analysis would   be a formidable task. 
189: 
190: This article is organized as follows. In \S 2 we  briefly discuss 
191: the gravitational waveforms of chirping binaries  and the data stream
192: obtained by LISA. Then we mention the parameter estimation  based
193: on the Matched filtering analysis.  In \S 3 we numerically evaluate 
194: the 
195: parameter estimation errors, and discuss its dependence on 
196: observational
197: period or wave frequency in detail. 
198: \S 4 is devoted to  summary and discussions.
199: 
200: 
201: 
202: 
203: 
204: \section{Gravitational Waveforms and Parameter Extraction}
205: 
206: \subsection{Gravitational Wave Measurement with LISA}
207: LISA consists of three  spacecrafts forming an equilateral
208:  triangle, and orbits around the Sun, trailing $20^{\circ}$ behind the
209: Earth. 
210: The sides of the triangle are $L=5 \times 10^6$ km in length, and the
211: plane of 
212:  triangle is inclined at $60^{\circ}$ with respect to the ecliptic.
213: The triangle rotates annually. 
214: The gravitational wave signal is reconstructed from the three data
215: streams  that effectively correspond to three
216: time-varying  
217:  armlength data, ($\delta L_1, \delta L_2, \delta L_3$) for 
218: gravitational
219:  waves with $\lambda \lsim L$.
220: We basically analyze two data streams given by 
221:  $s_{I}(t)=(\delta L_1(t)-\delta L_2(t))/L$ and $s_{II}(t)=
222:  (\delta L_1(t)+\delta L_2(t)-2\delta L_3(t))/\sqrt{3}L$.
223: These data can be  regarded as the response of two
224:  $90^{\circ}$-interferometers  rotated by $45^{\circ}$ to one
225: another (Cutler 1998).   
226: The data  $s_{I,II}(t)$ contain both  gravitational waves signal
227: $h_{I,II}(t)$  to be fitted  by matched filtering and   the  noise
228: $n_{I,II}(t)$. 
229: The latter is constituted by the detectors noise
230: and the binary confusion noise.
231: As in  Cutler (1998) we assume that the noises are stationary, Gaussian
232: and uncorrelated with 
233: each  other. 
234: 
235: 
236: The gravitational wave 
237: signals  $h_{I,II}(t)$ from a binary  are written as 
238: \beq
239:   h_{I,II}(t)=\frac{\sqrt{3}}{2} \left[ F^{+}_{I,II}(t)h_+(t)
240:  +F^{\times}_{I,II}(t)h_{\times}(t) \right],
241: \eeq
242: where $F^{+,\times}_{I,II}(t)$ are the pattern functions which depend 
243: on
244: the 
245:  source's angular position of the binary ($\bar{\theta}_S,\bar{\phi}_S$)
246: , its
247:  orientation ($\bar{\theta}_L,\bar{\phi}_L$) and detector's 
248: configuration. 
249: The angular variables with bars  are defined  in a fixed  barycenter
250: frame of 
251:  the solar system, and the  direction and orientation of the 
252: binary  are assumed to
253:  be constant  during the observation in this frame. The quantities
254: $h_{+,\times}(t)$ are the two polarization modes of 
255:  gravitational radiation  from  the  binary.  
256:  We can estimate both
257: ($\bar{\theta}_S,\bar{\phi}_S$) and ($\bar{\theta}_L,\bar{\phi}_L$) 
258: from
259: the time profiles  of the two  signals due to LISA's annual  rotation
260: and revolution. 
261:  Further discussion and details about the pattern functions are seen in
262:  Cutler (1998).  We basically use his formulation but newly
263:  include the effects of
264:  the chirp signal $\dot f$.
265: 
266: \subsection{Gravitational Waveforms}
267: We study   short-period
268:  ($10^{-4}~{\rm Hz}\lsim f \lsim 10^{-2}~{\rm Hz}$) 
269: binaries such as the close white dwarf 
270: binaries 
271:  (CWDBs) or the neutron star binaries (NBs) in our  Galaxy   
272:  (Hils, Bender \& Webbink 1991). 
273: We only discuss binaries with circular orbits. This is an excellent
274: approximation for CWDBs, as their orbits are circularized by strong
275:  tidal interaction in their earlier evolutionary stages.
276:    At LISA band some NBs or neutron star-white dwarf
277: binaries would 
278: have non-negligible eccentricities $e\sim 0.1$ induced at their
279:   supernova
280:  explosions (see {\it e.g.} Brown, Lee, Zwart, \& Bethe  2001).   But
281: extension 
282: to these eccentric binaries is straightforward.
283:  
284: 
285:  The chirping 
286: gravitational waveform  is given  by the quadrupole
287:  approximation \cite{peters64} as 
288: \beqa
289:   h_{+}(t) &=& A \cos \left[ 2 \pi \left( f + \frac{1}{2} {\dot f} t 
290:  \right) t + \phi_D(t) + \phi_0 \right] \times \left[ 1+ \left( {\hat L
291: }
292:  \cdot {\hat n} \right)^2 \right],   \nonumber \\ 
293:   h_{\times}(t) &=& -2 A \sin \left[ 2 \pi \left( f + \frac{1}{2} {\dot
294:  f}
295:  t \right) t + \phi_D(t) + \phi_0 \right] \times \left( {\hat L} \cdot
296:  {\hat n} \right),
297: \label{wf}
298: \eeqa   
299: where ${\hat L}$ (given by $\bar{\theta}_L,\bar{\phi}_L$) is the unit
300: vector in the direction of 
301:  the binary's orbital angular
302:  momentum, ${\hat n}$ (given by $\bar{\theta}_S,\bar{\phi}_S$) is the
303: unit  vector toward the binary 
304:  and $\phi_0$ is an integral  constant. 
305: We regard the frequency $f$ and its time variation
306: ${\dot f}$ as constants in the above equations.
307: A purely  monochromatic waveform has ${\dot f}=0$. This is the case 
308:  studied by 
309:  Cutler (1998).
310: When gravitational radiation reaction
311:  dominates evolution of the binary as in the
312: case of CWDBs or NBs, the  chirp signal ${\dot f}$ is given as
313:  ${\dot f}=( 96 \pi^{8/3} /5 )
314: f^{11/3} M_c^{5/3}$ with  the chirp mass $M_c$.
315: The perturbative expansion for the intrinsic 
316: evolution of the wave frequency 
317:  in equation (\ref{wf}) are valid under the condition 
318:  ${\dot f} T_{obs} \ll f$, where $T_{obs}$ is the observational period.
319: This condition is expressed as
320: \beq
321:  f \ll 0.14  \left( \frac{T_{obs}}{10 \mbox{yr}} 
322:  \right)^{-3/8} \left( \frac{M_c}{1 M_{\odot}} \right)^{-5/8}~\mbox{Hz}.
323: \eeq  
324: The amplitude $A$ in equation (\ref{wf}) is given in terms of the wave
325: frequency  $f$, chirp signal  $\dot f$ and the distance $D$ as
326: \beq
327:  A=\frac{5}{96 \pi^2} \frac{\dot f}{f^3 D}.
328: \label{amp}
329: \eeq
330: Thus we could determine the distance $D$, if we could measure three
331: observables $f$, $\dot f$ and $A$ (Schutz 1986). This is an important
332: aspect of gravitational wave astronomy.
333: 
334: The term  $\phi_D(t)$ in equation (\ref{wf}) is caused by  revolution 
335: of
336: LISA around the Sun and called the Doppler phase.
337: Its explicit form  is given by
338: \beq
339:   \phi_D(t)=2 \pi f R \sin \bar{\theta}_S \cos \left[ \bar{\phi}(t)-
340:  \bar{\phi}_S \right],
341: \label{dop}
342: \eeq
343: where $R=1$ AU and $\bar{\phi}(t)=2\pi t/T~(T=1 \mbox{yr})$ is the
344: direction of LISA in the fixed barycenter frame.
345: 
346: \subsection{Parameter Extraction}
347: Let us briefly discuss the matched filtering analysis and the parameter
348: estimation errors \cite{finn92,cf94}. We assume that
349:  the signal $h_\alpha(t)$ is characterized by some
350: unknown parameters $\gamma_i$ (eight  parameters in the present case:
351: $(A,f,{\dot f},\phi_0,{\bar \theta_S},
352:  {\bar \phi_S},{\bar \theta_L},{\bar \phi_L})$).
353: In the matched filtering analysis the variance-covariance matrix of the
354: parameter estimation error $\Delta 
355: \gamma_i$ is given by   inverse of the Fisher 
356: information matrix $\Gamma_{ij}$ as 
357: $\langle \Delta  
358:  \gamma_i \Delta \gamma_j \rangle = \left( \Gamma^{-1} \right)_{ij}$.
359: For a quasi-monochromatic binary (${\dot f}T_{obs}\ll f$) the noise
360: spectrum $S_n(f)$ is nearly constant in the frequency region swept by
361: the binary and the Fisher matrix simply becomes \cite{cutler98}
362: \beq 
363:   \Gamma _{ij} 
364:  = \frac{2}{S_n(f)} \sum_{\alpha=I,II} \int_0^{T_{obs}} dt~
365:  \frac{\partial h_{\alpha}(t)}{\partial \gamma_i}
366:  \frac{\partial h_{\alpha}(t)}{\partial \gamma_j}.
367: \label{fis}
368: \eeq
369: 
370: The  error boxes for the angular parameters ${\hat L}$ and ${\hat n}$
371: become
372: ellipses  in the celestial sphere  due to the correlation of 
373: two parameters $\theta$ and $\phi$. 
374: In this article we represent the estimation errors for direction and
375: orientation of binaries in the form defined in Cutler (1998) as
376: follows 
377: \beq
378:  \Delta \Omega = 2 \pi \sqrt{  \langle \Delta \mu^2\rangle
379: \langle \Delta \phi^2\rangle   
380:   -   \langle \Delta \mu \Delta \phi \rangle^2 },
381: \eeq    
382: where we have defined $\mu=\cos \theta$.
383: In the same manner the signal to noise ratio ($SNR$) is given by
384: \beq
385:   (SNR)^{~2} =  \frac{2}{S_n(f)}\sum_{\alpha=I,II}  \int_0^{T_{obs}} dt
386: ~
387: h_{\alpha}(t)  h_{\alpha}(t). 
388: \label{snr}
389: \eeq
390: 
391:   From equations (\ref{fis}) and (\ref{snr}) it is apparent that the
392: expressions for the 
393:  estimation errors  $\lla \Delta \gamma_i\Delta \gamma_j\rra$
394:  do not depend on the noise spectrum $S_n(f)$ when they are normalized
395: by the signal-to-noise ratio \citep{cutler98}.  In this article we
396: extensively use this  normalization method. For example the parameter
397: estimation 
398: errors for a simple wave form $h(t)=A \sin[2\pi(f+{\dot f}t/2)t+\phi_0]
399:  $ 
400: with four  fitting parameters $(A,f,{\dot f},\phi_0)$ is easily
401: evaluated as in Seto (2002). The explicit forms for $\Delta A, \Delta f
402: $ and
403: $\Delta \dot f$ are given by 
404: \beqa
405: \frac{\Delta A}{A} &=& \frac{1}{SNR} =
406: 0.1 \lmk \frac{SNR}{10} \rmk^{-1}, \label{aa} \\ 
407: \Delta f&=&\frac{4\sqrt3}{\pi}\frac{T^{-1}_{obs}}{SNR}=0.22 \lmk
408: \frac{SNR}{10}\rmk^{-1} T_{obs}^{-1},\label{af}\\
409: \Delta {\dot f}&=&\frac{6\sqrt5}{\pi}\frac{T^{-2}_{obs}}{SNR}=0.43 \lmk
410: \frac{SNR}{10}\rmk^{-1} T_{obs}^{-2}.\label{ac}
411: \eeqa
412: This simple analysis does not include information of the angular
413: parameters, but 
414: would be helpful to understand
415:  more detailed numerical analysis in the following
416: section. 
417: \section{Results}
418: 
419: \subsection{General Behavior}
420: 
421: We have numerically evaluated the uncertainties of the estimated
422:  parameters for  various
423: quasi-monochromatic binaries. In this subsection
424:  we show results for  a typical example with a fixed set of  
425: angular parameters  at $\cos \bar{\theta}_S=0.3,\bar{\phi}_S=5.0,
426:  \cos \bar{\theta}_L=-0.2 $ and $ \bar{\phi}_L=4.0$.
427: In Table \ref{t1} we show LISA's measurement accuracy for parameters 
428:   $(A,f,{\dot f},\Omega_S,\Omega_L)$  at  
429: frequencies $f=10^{-4}, 10^{-3}$ and $10^{-2}$Hz. We present our 
430: results
431: for two   observational
432: periods,  $T_{obs}=1$ and 10yr. 
433: All results are normalized by $SNR=10$ after integration over each
434:  observational period $T_{obs}$. 
435: These results simply scale as  $(SNR/10)^{-1}$ for errors $\Delta A, 
436:  \Delta f$ and $\Delta {\dot f}$ and $(SNR/10)^{-2}$ for error
437:  ellipses $\Delta \Omega_{S,L}$.  
438: The first row for each observational period 
439:  $T_{obs}$ represents  the case when all the eight
440: parameters $(A,f,{\dot f}, \phi_0,\bar{\theta}_S,\bar{\phi}_S,
441:  \bar{\theta}_L,\bar{\phi}_L)$ are 
442: included in the matched filtering analysis. The second row corresponds
443: to the case without the chirp signal $\dot f$.  Results
444: for $T_{obs}=1$yr are obtained under the same condition with
445:  Table 1 (case A) in Cutler (1998). Our numerical values  completely
446:  coincide with his ones.  The third row is the case 
447:  when direction  
448:  of the binary ($\bar{\theta}_S,\bar{\phi}_S$) 
449: is given exactly by other
450: method ({\it e.g.} optical identification of the binary) and 
451: removed from  the fitting parameters.
452:  
453: 
454: Figs 1 and 2 show clearly  dependence of the orbital parameters
455:  on wave frequency and observational period $T_{obs}$.
456: In Fig.\ref{f1} we show LISA's measurement accuracy as a function of
457:  observational period $T_{obs}$ at given frequencies
458:  $f=10^{-4},10^{-3}$ and $10^{-2}$ Hz.
459: All results are normalized by  $SNR=10$ at  integration period 
460: $T_{obs}=1$ yr. The solid lines are results for  fitting all the eight
461: parameters.
462: The dotted lines represent the case when the angular position
463:  $(\bar{\theta}_S,\bar{\phi}_S)$ are removed from the fitting
464:  parameters (see also Hughes 2002).
465: The dashed lines are results with fitting  only the angular position
466:  $(\bar{\theta}_S,\bar{\phi}_S)$.
467: For observational period  $T_{obs} \gsim 2$ yr, the
468:  difference between the solid and the dotted lines is very small
469:  especially for $\Delta A$, $\Delta f$ and $\Delta {\dot f}$
470:  irrespective of the frequency. Thus optical determination of the 
471: source direction $\Omega_S$ would only
472: slightly 
473: reduce the estimation errors of other parameters.
474: Asymptotic behaviors of errors $\Delta A$, $\Delta f$ and $\Delta {\dot
475: f}$ are given by
476: \beqa
477: \frac{\Delta A}{A} &=& 0.20 \lmk \frac{SNR}{10} \rmk^{-1} \label{ga},\\ 
478: \Delta f&=& 0.22 \lmk
479: \frac{SNR}{10}\rmk^{-1} T_{obs}^{-1} \label{gf}, \\
480: \Delta {\dot f}&=&0.43 \lmk
481: \frac{SNR}{10}\rmk^{-1} T_{obs}^{-2} \label{gc}, 
482: \eeqa 
483: (see also Table.1), and
484:  the asymptotic time dependence is given as
485:  $\Delta A \propto T_{obs}^{-1/2}, 
486:  \Delta f \propto T_{obs}^{-3/2},$ and $ \Delta {\dot f} \propto
487:  T_{obs}^{-5/2}$, since we have $SNR \propto T_{obs}^{1/2}$ from 
488: equation
489:  (\ref{snr}).
490: We have also
491:  $\Delta \Omega_{S,L} \propto SNR^{-2}\propto T_{obs}^{-1}$ for
492: directions and orientations of binaries. Numerical
493:  results for
494: $\Delta f$ and $\Delta {\dot f}$ in equations (\ref{gf}) and (\ref{gc})
495:  are almost identical to analytical ones in equations (\ref{af}) and
496:  (\ref{ac}). 
497: Hence we can expect that the information of the angular parameters  
498:  (the direction and orientation) does not affect the accuracy of
499:  the estimation for the frequency $f$ and the chirp signal ${\dot f}$
500:  for observational period $T_{obs} \gsim 2$ yr. 
501: But for the amplitude $A$, the above result (in [\ref{ga}]) is
502:  two times as large as  equation (\ref{aa}).
503: This is due to the fact that the amplitude $A$ is tied with the
504:  angular parameters (the inclination ${\hat L} \cdot {\hat n}$) in
505:  equation (\ref{wf}) and the estimation error $\Delta A$
506:  strongly depends on the error in ${\hat L} \cdot {\hat n}$
507:  (see the detailed discussion in the next subsection).
508: 
509: 
510: Fig.\ref{f2} is same as Fig.\ref{f1} but given as a function of
511: frequency  $f$ with fixed observational period
512:  $T_{obs}=1$ and 10 yr.
513: We find that for $T_{obs}=10$ yr the errors 
514: $\Delta A,\Delta f$ and $\Delta \dot{f}$
515:  do not depend on the frequency,  as given in equations
516:  (\ref{ga})-(\ref{gc}).
517: But the angular resolution $\Delta \Omega_S$ depends on the frequency
518:  and is given  by
519: \beqa
520: \Delta \Omega_S &=& 4.8 \times 10^{-4}
521:  \lmk \frac{f}{10^{-2} \mbox{Hz}} \rmk^{-2}
522:  \lmk \frac{SNR}{10} \rmk^{-2} \mbox{sr},
523: \label{fdos}
524: \eeqa
525: for higher frequencies $f \gsim 10^{-3}$ Hz, and nearly constant for
526:  lower frequency $f \lsim 10^{-3}$ Hz.  
527: 
528: Now we discuss correlation between Fisher matrix elements for $T\gsim 2
529: $yr. As expected
530:  from Figs 1 and 2,  the source direction  $\Omega_S$ has almost no
531: correlation with other parameters at higher frequencies $f>10^{-3}$Hz
532: \footnote{In other words, Fisher information matrix $\Gamma_{ij}$ in 
533: equation
534:  (\ref{fis}) becomes diagonalized
535:  to the two parts; 
536: $
537: \Gamma \simeq \left( 
538:            \begin{array}{@{\,}cc@{\,}}
539:            \Gamma_A &     0        \\
540:                0    &  \Gamma_S    \\ 
541:            \end{array} \right),  
542: $
543: where $\Gamma_A$ represents a $6 \times 6$ matrix which corresponds
544:  to $(A,f,{\dot f},\phi_0,{\bar \theta_L},{\bar \phi_L})$ components 
545: and
546:  $\Gamma_S$ represent a $2 \times 2$ matrix which corresponds to
547:  $(\bar{\theta}_S,\bar{\phi}_S)$ components.}
548: and weakly correlates with source orientation $\Omega_L$ at lower  
549: frequencies $f<10^{-3}$Hz.  This dependence  seems reasonable 
550: considering 
551: the information used to determine the source direction  $\Omega_S$,
552: namely, the Doppler phase at $f>10^{-3}$Hz and amplitude modulation at
553: $f<10^{-3}$Hz. The orientation $\Omega_L$ is  strongly related to
554: the latter. As a result angular variables $\Omega_S$ and $\Omega_L$
555: correlate at lower frequencies. 
556: %%%%%%%revvv
557: The detected wave form $h_\alpha$ is modulated by annual revolution
558: and rotation of the detectors, and its phase has oscillating part due
559: to the modulation that is more prominent at higher
560: frequencies. Amplitudes of the derivatives $\p h/\p \Omega_s$ also show
561: this oscillation.  In  the Fisher matrix their cross terms with
562:  other derivatives are significantly canceled with long time
563:  integration. As a result the estimation errors for the angular 
564: direction  $\Omega_s$  show very weak correlation with other errors.  
565: 
566: In the same manner we can also understand the asymptotic frequency
567: dependence of the angular resolution $\Omega_S$. In the derivatives $\p
568: h_{\alpha}/\p \gamma_i$ for the  Fisher matrix
569: elements in equation (6) the frequency $f$ appears only through the 
570: Doppler phase
571: $\phi_D(t)$ in equation (\ref{dop}). At lower frequencies $f<10^{-3}$Hz
572:  terms from the Doppler
573: phase ($\propto f^1$) is smaller than the terms from the amplitude
574: modulation ($\propto f^0$), and consequently  the
575: estimation error $\Delta \Omega_S$ does not depend on the frequency $f$.
576: At higher frequencies  $f>10^{-3}$Hz the Doppler phase term  becomes
577: dominant  and the Fisher matrix are diagonalized. Thus the error
578:  $\Delta \Omega_S$ depends on the frequency $f$ as 
579: $\Delta \Omega_S\propto f^{-2}$ (Cutler \& Vecchio 1998, Moore \& 
580: Hellings 2002).
581: 
582: 
583: 
584: Finally we analytically investigate how  the estimation errors  improve
585:  with the observational period $T_{obs}$.  We reanalyze the simple toy 
586: model
587:  $h(t)=A \sin[2\pi(f+{\dot f}t/2)t+\phi_0]$ (see the sentences before
588:  equation (\ref{aa})) adding the information of the source direction
589:  $(\bar{\theta}_S,\bar{\phi}_S)$ by the Doppler phase  $\phi_D(t)$.
590: This waveform is given as
591: \beq
592:   h(t)=A \sin[~2\pi(f+{\dot f}t/2)t+\phi_D(t)+\phi_0],
593: \label{swf}
594: \eeq
595: with six fitting parameters $\gamma_i=(A,f,{\dot f},\phi_0,
596:  \bar{\theta}_S,\bar{\phi}_S)$. 
597: We evaluate the magnitude of the variance-covariance matrix $\langle
598:  \Delta \gamma_i \Delta 
599:  \gamma_j \rangle$
600:  by its   determinant as
601:   $\det \langle \Delta \gamma_i \Delta \gamma_j \rangle 
602:  = (\det \Gamma)^{-1}$.
603: After some algebra the time dependence of $\det \Gamma$ is given
604: analytically  from
605: equations 
606:  (\ref{fis}) and (\ref{swf}) as,      
607: \beq
608:  \det \Gamma \propto T_{obs}^{12} \lmk 1- \frac{6}{\pi^2
609:  \lmk T_{obs}/ 1 \mbox{yr} \rmk^{2}} \rmk 
610:   \lmk 1- \frac{90}{\pi^4 \lmk T_{obs}/1 \mbox{yr}\rmk^{4}} \rmk,
611: \label{det}
612: \eeq       
613: where we assume that the observational period  $T_{obs}(\gg f^{-1})$ is
614:  a integer in
615: units of year.
616: In  the above expression the quantity  $(\det \Gamma)^{-1}$ formally
617:   diverges at
618:  $T_{obs}=\sqrt[4]{90}/\pi=0.98$ [yr] which is very close to 1, thus 
619: the
620:  estimation errors have large value at $T_{obs}=1$ yr.  
621: Fig.\ref{f3} shows  the inverse of the determinant of the Fisher matrix
622: in the form  $T_{obs}^{12}(\det \Gamma)^{-1}$ as a function of the 
623: observational
624:  period $T_{obs}$.
625:   From Fig.\ref{f3}, $T_{obs}^{12} (\det \Gamma)^{-1}$ rapidly 
626: converges at
627:  $T_{obs} \sim 2$ yr.
628: In this case we also obtain the estimation errors as follows;
629: \beqa
630: \frac{\Delta A}{A} &=& \frac{1}{SNR} ,\\ 
631: \Delta f&=&\frac{4\sqrt3}{\pi}\frac{T^{-1}_{obs}}{SNR} 
632:  \times \sqrt{\frac{1-(45/8)~x^2 (1+x^2)}{(1-6 x^2)(1-90 x^4)}}, \\
633: \Delta {\dot f}&=&\frac{6\sqrt5}{\pi}\frac{T^{-2}_{obs}}{SNR}
634:  \times \frac{1}{\sqrt{1-90 x^4}} , \\
635: \Delta \Omega_S &=& \frac{1}{\pi f^2 R^2 \cos{\bar{\theta}_S} SNR^2}
636:  \times \frac{1}{\sqrt{(1-6 x^2)(1-90 x^4)}}  \\
637:  &\simeq& 4.3 \times 10^{-4} \lmk \frac{\cos{\bar{\theta}_S}}{0.3} \rmk
638: ^{-1}
639:  \lmk \frac{f}{10^{-2} \mbox{Hz}} \rmk^{-2}
640:  \lmk \frac{SNR}{10} \rmk^{-2} \mbox{sr}, \label{doms}
641: \eeqa
642: where $x=\pi^{-1} (T_{obs}/1 \mbox{yr})^{-1}$ and equation
643:  (\ref{doms}) is valid for $T_{obs} \gsim 2$ yr and is similar to
644:  equation (\ref{fdos}).
645: The above results for $\Delta f$, $\Delta \dot{f}$ and
646:  $\Delta \Omega_S$ are about $40 \%$ smaller than the the results
647:  in Fig.\ref{f1} at $T_{obs}=1$ yr, because we do not include the
648:  information of the source orientation.  
649: $\Delta f$ and $\Delta \dot{f}$ are asymptotically same as
650:  equations (\ref{af}) and (\ref{ac}) for $T_{obs} \gsim 2$ yr.  
651: 
652: 
653: \subsection{Statistical Analysis}
654: So far we have studied a specific set of the angular parameters $\hat n
655: $
656: and $\hat L$. In this subsection we present statistical results for
657: their various combinations  at the asymptotic region $T_{obs}\gsim2$yr.
658:  
659: We have made 100 realizations of  $\hat n$
660: and $\hat L$ that  are distributed randomly on  celestial
661: spheres. Then we 
662: calculate the
663: estimation errors $\Delta f$ and $\Delta {\dot f}$ for each binary
664: normalized by $SNR=10$ with $T_{obs}=10$yr.  We find that theses errors
665:  depend very  weakly (less than 10\% scatter)  on  
666: the directions  $\hat n$
667: and $\hat L$ in contrast  to the results for  $T_{obs}=1$yr
668:  (scattering typically factor $2^{\pm1}$ around the average).
669: We calculate their mean values and
670: obtain  results 
671: given in the following forms
672: \beqa
673: \Delta f &=&0.22 \lmk \frac{SNR}{10} \rmk^{-1} T_{obs}^{-1},
674: \label{nf} \\
675: \Delta {\dot f} &=&0.43 \lmk \frac{SNR}{10} \rmk^{-1} T_{obs}^{-2}.
676: \label{nc}  
677: \eeqa
678: Note that these   results do not depend on the  frequency $f$ and
679: would be useful for quantitative analysis of quasi-monochromatic
680: binaries with $T_{obs}\gsim 2$yr. As expected from  the previous
681: subsection the simple analytical
682:  estimations given in  equations (\ref{af})
683:  and (\ref{ac})
684: are  very close to equations (\ref{nf}) and (\ref{nc}). 
685: 
686: 
687: 
688: We have also studied the estimation  errors $\Delta \Omega_S$, 
689: $\Delta \Omega_L$ and $\Delta A/A$.   Distributions of  the latter two
690: have very large scatters.  This is because (i) the Fisher matrix 
691: elements
692: relating to the orbital orientation $\hat L$ become singular at the  
693: highly symmetric face-on configuration ${\hat n}=\pm {\hat L}$ and (ii)
694:  the
695: estimation of the amplitude $A$
696: is closely related to the inclination.  For majority of realizations
697: with  $|{\hat n}\cdot {\hat L}| \lsim 0.8$  we have typically
698:  $\Delta A/A\sim 0.2 (SNR/10)^{-1}$. To determine the source
699: direction $\hat n$ we can use the information of the Doppler phase in
700: addition to the annual amplitude modulation caused by LISA's rotation. 
701: Thus the error $\Delta \Omega_S$ does not show such a bad behavior. For
702: the above 100 realization at $f=0.01$Hz  with $SNR=10$ and $T_{obs}=10$
703: yr
704:  the errors $\Delta
705: \Omega_S$ are distributed as  $1.3\times 10^{-4}{\rm sr}\le \Delta
706: \Omega_S \le 3.7 \times 10^{-3}$sr with the mean value $\Delta
707: \Omega_S =7.1 \times 10^{-4}$sr. Therefore the following relation 
708: roughly
709: gives  the estimation error for the
710:  source direction with $T_{obs}\ge 2$yr  
711: \beq
712: \Delta
713: \Omega_S \sim 7.1 \times 10^{-4} \lmk \frac{SNR}{10} \rmk^{-2}\lmk
714: \frac{f}{10^{-2}\rm Hz} \rmk^{-2} {\rm sr},
715: \eeq
716: at $f\gsim 2\times 10^{-3}$Hz where the Doppler phase becomes more 
717: important
718: than the annual amplitude due to the rotation of LISA.
719: 
720:  
721: Now let us discuss  issues concerned with the chirp signal $\dot f$
722: based on 
723:  equation (\ref{nc}). We  assume that the chirp signal is dominated by
724: gravitational radiation reaction  and introduce a  parameter
725:  $R\equiv \Delta {\dot
726: f}/ {\dot f}$ that represents relative accuracy  of the
727: measured signal $\dot f$.  For a given threshold $R$ we can solve the
728: corresponding frequency $f_R$ as
729: \beq
730: f_R=9.2\times 10^{-4} \lmk \frac{M_c}{1M_\odot}\rmk^{-5/11}
731: \lmk\frac{SNR}{10} \rmk^{-3/11}  \lmk \frac{T_{obs}}{10 {\rm yr}}
732: \rmk^{-6/11} 
733: \lmk \frac{R}{1.0}\rmk^{-3/11}  ~~{\rm Hz} 
734: \eeq
735: for observational period  $T_{obs}\gsim 2$yr.
736: The frequency  $f_{R=1}$ can be regarded as the critical frequency for 
737: treatment of the
738: chirp signal $\dot f$. At higher frequencies $f\gsim f_{R=1}$ the
739: parameter 
740: $\dot f$ should be included in the matched filtering,  but the simple
741: prescription 
742: $\dot f=0$  
743: would be better  at  $f\lsim f_{R=1}$ since the expected signal would 
744: be
745: completely buried  in error.  If  the chirp signal is measured  with
746: accuracy $R(\ll 1)$,  the chirp mass can be estimated with relative
747: accuracy $\Delta M_c/M_c\simeq 3R/5$ from relation ${\dot f}\propto
748: M_c^{5/3}$. 
749: The chirp signal $\dot f$  is also  essential to
750: determine the distance $D$ to the binary. From the
751: simple relation $A=5{\dot f}/96\pi^2f^3D$,  
752: the estimation error for distance $D$ is roughly evaluated as
753: \beqa
754:   \frac{\Delta D}{D} &\simeq& \frac{\Delta A}{A}+\frac{\Delta {\dot f}}
755:  {\dot f},  \\
756: &\simeq& {\rm Max}\lnk \frac{\Delta A}{A}, ~~~\frac{\Delta {\dot f}}
757:  {\dot f}\rnk,  \\
758:   &\simeq& 0.2 \left( \frac{SNR} {10} \right)^{-1} {\rm Max} \lnk 1,
759: ~~~\lmk\frac{f}{1.4\times 10^{-3}{\rm Hz}} \rmk^{-11/3}
760: \lmk\frac{M_c}{1 M_\odot} \rmk^{-5/3}
761: \lmk\frac{T_{obs}}{10{\rm yr}} \rmk^{-2}   \rnk,  
762: \label{dis}
763: \eeqa
764: where we have used the typical error for the amplitude estimation
765: $\Delta A/A\sim 0.2(SNR/10)^{-1}$. For  compact binaries with chirp 
766: mass $M_c\sim
767: 1M_\odot$ such as NBs or CWDBs  the chirp signal $\dot f$ is the 
768: dominant
769: source of the error    at  lower frequencies $f\lsim
770: 10^{-3}$Hz and observational period $T_{obs}\sim 10$yr.   The amplitude
771: $A$ becomes dominant one  at  $f\gsim 
772: 10^{-3}$Hz. 
773: 
774: 
775: 
776: Finally, we calculate the estimation error for the three dimensional
777:  position of the
778:  Galactic binaries. 
779: The signal to noise ratio is calculated from equation (\ref{snr})
780:  as $SNR \sim 380 
781:  ~( f/5 \times 10^{-3} \mbox{Hz} )^{2/3}$
782:  $( D/10 \mbox{kpc} )^{-1} 
783:  ~( T_{obs}/10 \mbox{yr} )^{1/2}
784:  ~( M_c/1 M_{\odot} )^{5/3}
785:  ~( \sqrt{S_n}/ 8 \times 10^{-21} \mbox{Hz}^{-1/2} )^{-1/2}$.
786:  The noise spectrum $S_n(f)$ (in units of ${\rm Hz}^{-1}$) is nearly
787:  constant for $3 \times 10^{-3}~\mbox{Hz} \lsim f \lsim 10^{-2}$ Hz
788: % \cite{cutler98}.
789:  (see Fig.5 in Cutler 1998).
790: Hence one could determine both the angular position and the distance
791:  to the binary with the accuracy of
792: \beqa
793:  \Delta {\bar \theta_S} &\sim& 2~
794:  \left( \frac{f}{5 \times 10^{-3} \mbox{Hz}} \right)^{-5/3}
795:  \left( \frac{D}{10 \mbox{kpc}} \right)
796:  \left( \frac{T_{obs}}{10 \mbox{yr}} \right)^{-1/2} \nonumber \\
797:  && ~~~~~~~\times \left( \frac{M_c}{1 M_{\odot}} \right)^{-5/3}
798:  \left( \frac{\sqrt{S_n}}{8 \times 10^{-21} \mbox{Hz}^{-1/2}}
799:  \right)^{1/2} \mbox{arcmin},\\
800:  \frac{\Delta D}{D} &\sim& 5 \times 10^{-3}
801:  \left( \frac{f}{5 \times 10^{-3} \mbox{Hz}} \right)^{-2/3}
802:  \left( \frac{D}{10 \mbox{kpc}} \right)
803:  \left( \frac{T_{obs}}{10 \mbox{yr}} \right)^{-1/2} \nonumber \\
804:  && ~~~~~~~\times \left( \frac{M_c}{1 M_{\odot}} \right)^{-5/3}
805:  \left( \frac{\sqrt{S_n}}{8 \times 10^{-21} \mbox{Hz}^{-1/2}} \right)^{
806: 1/2}.
807: \eeqa  
808: 
809: 
810: 
811: \section{Conclusion}
812: We have calculated LISA's measurement accuracy for  short-period 
813: binaries
814:  $(10^{-4}~{\mbox Hz} \lsim$ $f$ $\lsim 10^{-1}~{\mbox Hz})$
815:  in our  Galaxy, including the effects of chirp signal ${\dot f}$
816:  and dependence  on observational period $T_{obs}$.
817: We find that the measurement accuracy  rapidly improves
818:  for $T_{obs} \gsim 2$yr comparing with $T_{obs}\sim 1$yr. This might 
819: be
820: an important element  for discussing  operation period of LISA.
821: 
822: At observational period 
823: $T_{obs} \gsim 2$ the errors for quantities  $f$ and ${\dot f}$
824:  is independent on the information of the positions and orientations of
825: binaries  in contrast to 
826:  $\Delta A$ and $\Delta \Omega_{S,L}$.
827: It is also found that the estimation errors $\Delta A$, $\Delta f$ and
828: $\Delta{\dot f}$ 
829:  are almost independent on the frequency $f$. 
830: The fitting formulae of the estimation errors for the source
831:  parameters (such as frequency $f$, chirp signal ${\dot f}$,
832:  amplitude $A$, angular position $\Omega_S$ and distance $D$)
833:  are given as functions of $f$ and $T_{obs}$. We expect
834: these would be  powerful tools for quantitatively studying possibility
835:  gravitational wave astronomy.
836: 
837: 
838: The relative motion between the Sun and a binary could affect the
839:  frequency $f$ by the Doppler factor, $(1+v_R)$ where 
840:  $v_R$ is the relative radial velocity.
841: The tangential velocity $v_T$ or accelerating motion $\dot v$  would
842: affect  
843:  the time variation of the frequency, $(\dot{f}/f)_{gal}  \sim
844: v_{T}^2/cD$ or $\sim {\dot v}/c\sim
845: v_{rot}^2/cR_g$ respectively 
846:  where $v_{rot}$ is the rotation velocity ($\sim 200$km/s) of the
847: Galaxy, $R_g$ is the  Galactic radius $\sim 10$kpc and $D$  
848:  is the distance to the binary (see {\it e.g.} Damour \& Taylor 1991). 
849: For most Galactic binaries in LISA band  this should be negligibly 
850: small
851: compared with  
852:  $(\dot{f}/f)_{GW}$ due to gravitational radiation reaction.
853: This condition is expressed as $f \gg 3\times10^{-5}
854:  (M_c/1M_{\odot})^{-5/8} (v_{rot}/200 \mbox{km})^{3/4}
855:  (D/10\mbox{kpc})^{-3/8}$ Hz.
856: For  binaries very close to us  $D \lsim 100$ pc   
857:  these effects would be important.  But note that we might measure the
858: proper 
859: motion of such binaries by optical observation.
860: 
861: 
862: 
863:    
864: 
865: \acknowledgments
866: We would like to thank an anonymous referee for helpful comments to
867: improve the manuscript,  and Takeshi Chiba, Hideyuki Tagoshi and 
868: Hirotaka
869: Takahashi  for useful comments and discussions.
870: This work was
871: supported in part by Grant-in-Aid of Scientific Research
872: of the Ministry of Education, Culture, Sports, Science and
873: Technology No. 0001416.
874: %%%%%%%%% references %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
875: \begin{thebibliography}{}
876: 
877: %\bibitem[Bender \& Hils 1997]{bh97}
878: %Bender, P.L. \& Hils, D., Classical Quantum Gravity, {\bf 14}, 1439, 
879: %(1997).
880: 
881: \bibitem[Bender et al 1998]{lisa}
882: Bender, P.L. \etal, {{\it LISA Pre-Phase A Report}}, Second Edition,
883:  July 1998. 
884: 
885: \bibitem[Brown, Lee, Zwart, \& Bethe  2001]{2001ApJ...547..345B}
886:  Brown, G.~E., Lee, C.-H., Zwart, S.~F.~P., \& Bethe, H.~A.,
887:  ApJ, {\bf 547}, 345, (2001)
888: 
889: \bibitem[Cutler \& Flanagan 1994]{cf94}
890: Cutler, C. \& Flanagan, E.F., Phys. Rev. D {\bf 49}, 2658, (1994). 
891: 
892: \bibitem[Cutler 1998]{cutler98}
893: Cutler, C., Phys. Rev. D {\bf 57}, 7089, (1998). 
894: 
895: \bibitem[Cutler \& Vecchio 1998]{cv98}
896: Cutler, C. \& Vecchio, A., in Folkner, W.M., editor, {{\it Laser
897:  Interferometer Space Antenna, Second International LISA Symposium on 
898: the
899:  Detection and Observation of Gravitational Waves in Space}}, AIP 
900: Conference
901:  Proceedings, Vol.456, p95, (1998).    
902: 
903: 
904: \bibitem[Damour \& Taylor 1991]{dt91}
905:  Damour, T. \& Taylor, J.H., ApJ., {\bf 366}, 501, (1991).
906: 
907: \bibitem[Balasubramanian \& Dhurandhar(1998)]{1998PhRvD..57.3408B} 
908: Balasubramanian, R.~\& Dhurandhar, S.~V.\  \prd, 57, 3408, (1998) 
909: 
910: \bibitem[Evans, Iben, \& Smarr 1987]{1987ApJ...323..129E}
911:  Evans, C.\ R., Iben, I.\ J., \& Smarr, L., ApJ, {\bf 323}, 129, (1987)
912: 
913: \bibitem[Finn 1992]{finn92}
914: Finn, L.S., Phys. Rev. D {\bf 46}, 5236, (1992).  
915: 
916: \bibitem[Haehnelt 1994]{haehnelt94}
917: Haehnelt, M.G., MNRAS, {\bf 269}, 199, (1994).
918: 
919: \bibitem[Hils, Bender \& Webbink 1991]{hbw91}
920: Hils, D., Bender, P.L. \& Webbink, R.F., ApJ, {\bf 369}, 271, (1991).
921: 
922: 
923: \bibitem[Hughes 2002]{hughes01}
924: Hughes, S. A.
925: %``Untangling the merger history of massive black holes with LISA,''
926: MNRAS, {\bf 331}, 805, (2002)
927: %%CITATION = ASTRO-PH 0108483;%%
928: 
929: 
930: \bibitem[Mironovskii 1965]{1965SvA.....9..752M}
931:  Mironovskii, V.\ N., Soviet Astronomy, {\bf 9}, 752, (1965) 
932: 
933: \bibitem[Moore \& Hellings 2002]{mh99}
934: Moore, T.M. \& Hellings, R.W., Phys. Rev. D {\bf 65}, 062001, (2002).
935: 
936: \bibitem[n02]{n02}
937: Napiwotzki,  R.  et al. arXiv:astro-ph/0203145
938: 
939: \bibitem[n01]{n01}
940:  Nelemans, G., Yungelson, L. R.,  \& Portegies
941:  Zwart, S. F. A\&A, {\bf 365}, 49 (2001)
942: 
943: 
944: \bibitem[Peters 1964]{peters64}
945: Peters, P.C., Phys. Rev. {\bf 136}, B1224, (1964).
946:   
947: 
948: \bibitem[Peterseim \etal 1997]{pjds97}
949: Peterseim, M., Jennrich, O., Danzmann, K. \& Schutz, B.F.,
950:  Classical Quantum Gravity, {\bf 14}, 1507, (1997).
951: 
952: \bibitem[Schutz 1986]{schutz86}
953: Schutz, B.F., Nature {\bf 323}, 310, (1986).
954: 
955: 
956: %\cite{Seto:2002dz}
957: \bibitem{Seto:2002dz}
958: Seto, N.
959: %``Long term operation of LISA and galactic close white dwarf binaries,
960: ''
961: arXiv:astro-ph/0202364.
962: %%CITATION = ASTRO-PH 0202364;%%
963: 
964: 
965: \bibitem[Seto 2001]{seto01b}
966: Seto, N., Phys. Rev. Lett., {\bf 87}, 251101, (2001). 
967: 
968: 
969: \bibitem[Vecchio]{v97}
970: Vecchio, A., Classical Quantum Gravity, {\bf 14}, 1431, (1997).
971: 
972: 
973: \bibitem[Vecchio \& Cutler 1998]{vc98}
974: Vecchio, A. \& Cutler, C., in Folkner, W.M., editor, {{\it Laser
975:  Interferometer Space Antenna, Second International LISA Symposium on 
976: the
977:  Detection and Observation of Gravitational Waves in Space}}, AIP 
978: Conference
979:  Proceedings, Vol.456, p101, (1998). 
980: 
981: 
982: \bibitem[Webbink \& Han(1998)]{1998lain.conf...61W} Webbink, R.~F.~\&
983:  Han,  Z.\ in Folkner, W.M., editor, {{\it Laser
984:  Interferometer Space Antenna, Second International LISA Symposium on 
985: the
986:  Detection and Observation of Gravitational Waves in Space}}, AIP 
987: Conference
988:  Proceedings, Vol.456, p61, (1998). 
989: 
990:  \bibitem[y01]{y01}
991: Yungelson, L. R., Nelemans, G.,  Portegies
992:  Zwart, S. F. \&  Verbunt, F.  in ed. Vanbeveren, D., editor,
993: The Influence of Binaries on 
994:  Stellar Population Studies,  Kluwer, p339,  (2001)
995:   
996: 
997: \end{thebibliography}
998: 
999: \newpage
1000: 
1001: \begin{table}
1002:   \begin{center}
1003: %  \tabcolsep=2mm
1004:   \setlength{\tabcolsep}{10pt}
1005:   \renewcommand{\arraystretch}{1.1}
1006: %  \small
1007:   \begin{tabular}{cccccc} \hline\hline
1008:    $T_{obs}$ (yr) & ~~$\Delta A/A$~~ & ~$T_{obs}\Delta f$~ & ~$T_{obs}^
1009: 2\Delta {\dot f}$~ & ~$\Delta \Omega_S$ (sr)~ & ~$\Delta \Omega_L$ (sr)
1010:  \\ \hline
1011:   \multicolumn{6}{l}{$f=10^{-4}$ Hz} \\ 
1012:    1yr & 0.205 & 0.33 & 0.59 & 8.47 $\times 10^{-2}$ &
1013:    0.201 \\ 
1014: & 0.204 & 0.076 &   & 8.27 $\times 10^{-2}$ &
1015:    0.199 \\ 
1016:        & 0.204 &  0.31 & 0.58 & & 0.154 \\
1017:    10yr & 0.204 & 0.22 & 0.43 & 6.78 $\times 10^{-2}$ &
1018:    0.185 \\ 
1019: & 0.204 & 0.055 &   & 6.78 $\times 10^{-2}$ &
1020:    0.185 \\ 
1021:         & 0.204 & 0.22 & 0.43 & & 0.152 \\  \hline
1022:   \multicolumn{6}{l}{$f=10^{-3}$ Hz} \\ 
1023:    1yr & 0.205 & 0.62 & 1.1 & 7.59 $\times 10^{-2}$ &
1024:    0.185 \\
1025: & 0.204 & 0.092 &  & 3.97 $\times 10^{-2}$ &
1026:    0.169 \\     
1027:         & 0.204 &0.31 & 0.58 & & 0.154 \\
1028:    10yr & 0.204 & 0.22 & 0.43 & 2.70 $\times 10^{-2}$ &
1029:    0.161 \\
1030:         & 0.204 &0.055 &  & 2.70 $\times 10^{-2}$ & 0.161\\
1031:  & 0.204 & 0.22 & 0.43 &  &
1032:    0.161 \\ \hline
1033:   \multicolumn{6}{l}{$f=10^{-2}$ Hz} \\
1034:    1yr & 0.205 &1.1 &2.2 &  
1035:     4.10 $\times 10^{-3}$ & 0.155 \\
1036: & 0.204 &0.14 &0.20 &  
1037:     1.08 $\times 10^{-3}$ & 0.153 \\
1038:        & 0.204 & 0.31 &0.58 & & 0.153 \\
1039:    10yr & 0.204 &0.22 &0.43 &
1040:    4.77 $\times 10^{-4}$ & 0.110 \\
1041:  & 0.204 &0.055 & &
1042:    4.77 $\times 10^{-4}$ & 0.153 \\
1043:        & 0.204 &0.22 &0.43 & &
1044:    0.109 \\ \hline\hline
1045:   \end{tabular}
1046:   \end{center}
1047: \caption{
1048: LISA's measurement accuracy for  binaries with angular parameters
1049:  $(\cos \bar{\theta}_S=0.3, \bar{\phi}_S=5.0, \cos \bar{\theta}_L=-0.2,
1050:  \bar{\phi}_L=4.0)$.   
1051: Results are normalized by $SNR=10$. Errors scale as $(SNR/10)^{-1}$ for 
1052:  $\Delta A,\Delta f$ and $\Delta {\dot f}$, and $(SNR/10)^{-2}$ for
1053:  $\Delta \Omega_{S,L}$.
1054: %for $f=10^{-4}, 10^{-3},10^{-2}$ Hz and $T_{obs}=1,10$ yr.
1055: The second lines in each observational period  $T_{obs}$ represent the 
1056: case with removing the
1057:  chirp signal $\dot f$ from fitting parameters and the third lines with
1058:  removing the direction of the source $(\theta_S, \phi_S)$.  }
1059: \label{t1}
1060: \end{table}
1061: 
1062: 
1063: \begin{figure}
1064:   \vspace{0.1cm}
1065:   \hspace{3.75cm}
1066: %  \includegraphics[height=7.5cm,clip]{f1.eps}
1067:   \includegraphics[height=7.5cm,clip]{262.ps}
1068:   \vspace{0.1cm}
1069:     \\
1070:   \begin{minipage}[t]{7.5cm}
1071:     \vspace{0.1cm}
1072: %    \includegraphics[height=7.5cm,clip]{f2.eps} 
1073:      \includegraphics[height=7.5cm,clip]{242.ps} 
1074:     \vspace{0.1cm}
1075:   \end{minipage}
1076:   \begin{minipage}[t]{7.5cm}
1077:     \vspace{0.1cm}
1078: %    \includegraphics[height=7.5cm,clip]{f3.eps} 
1079:      \includegraphics[height=7.5cm,clip]{252.ps}  
1080:     \vspace{0.1cm}
1081:   \end{minipage}
1082:     \\
1083:   \caption{LISA's measurement accuracy for the binaries
1084:  as a function of the observational period $T_{obs}$ with angular
1085:  parameters 
1086:  $(\cos \bar{\theta}_S=0.3, \bar{\phi}_S=5.0, \cos \bar{\theta}_L=-0.2,
1087:  \bar{\phi}_L=4.0)$. 
1088: The solid lines correspond to $\Delta {\dot f},
1089:  \Delta f, \Delta A,
1090:  \Delta \Omega_L$ and $\Delta \Omega_S$ from top to bottom.
1091:  The dotted lines represent the case source positions $(\bar{\theta}_S,
1092:  \bar{\phi}_S)$ are removed from the fitting parameters,
1093:  and the dashed lines represent the
1094:  case the only source positions are included in the fitting parameters. 
1095: The accuracies are normalized by $SNR=10$ at 1yr
1096:  observation. 
1097: } 
1098: \label{f1}
1099: \end{figure}
1100: 
1101: 
1102: 
1103: \begin{figure}
1104: %  \plottwo{f4.eps}{f5.eps}
1105:   \plottwo{316.ps}{317.ps}
1106:   \caption{Same as Fig.\ref{f1}, but as a function of the frequency $f$
1107:  with the observational period $T_{obs}=1,10$ yr. The accuracies are
1108:  normalized by $SNR=10$.} 
1109: \label{f2}
1110: \end{figure}
1111: 
1112: %\newpage
1113: %\clearpage
1114: 
1115: 
1116: 
1117: \begin{figure}
1118:   \hspace{3.75cm}
1119: %  \includegraphics[height=7.5cm,clip]{f6.eps}
1120:    \includegraphics[height=7.5cm,clip]{263.ps} 
1121: %  \plotone{263.ps}
1122: 
1123: \caption{The inverse of the determinant of the Fisher matrix for the
1124:  simple waveform given in equation (16).
1125: The filled circles represent with integer  $T_{obs}$ in units of year.
1126: We normalize the overall scale by unity  at $T_{obs}=+\infty$.
1127: }
1128: 
1129: \label{f3}
1130: \end{figure}
1131: 
1132: 
1133: \end{document}
1134: