0710.1897/ms.tex
1: %\documentclass[12pt,preprint]{aastex}  %Needed for submission
2: \documentclass[apjl]{emulateapj}
3: \usepackage{apjfonts}  %For emulateapj
4: 
5: %%%%%%%%%%%%%%%%%%%%%%%%%%
6: 
7: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
8: %
9: % some definitions
10: %
11: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
12: %
13: \newcommand{\be}{\begin{equation}}
14: \newcommand{\ee}{\end{equation}}
15: \newcommand{\ba}{\begin{eqnarray}}
16: \newcommand{\ea}{\end{eqnarray}}
17: %
18: \newcommand{\Mc}{{\cal M}}
19: \newcommand{\Ms}{M_{\odot}}
20: %
21: \newcommand{\m}{\langle}
22: \newcommand{\M}{\rangle}
23: %
24: \def\ltsima{$\; \buildrel < \over \sim \;$}
25: \def\simlt{\lower.5ex\hbox{\ltsima}}
26: \def\gtsima{$\; \buildrel > \over \sim \;$}
27: \def\simgt{\lower.5ex\hbox{\gtsima}}
28: %
29: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
30: %
31: % end definitions
32: %
33: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
34: 
35: \shorttitle{Spinning Inspiral Binaries and Gravitational-Wave Astronomy}
36: \shortauthors{M.V.\ van der Sluys et al.}
37: 
38: \begin{document}
39: 
40: \title{Gravitational-Wave Astronomy with Inspiral Signals of Spinning Compact-Object Binaries}
41: 
42: \author{M.V.\ van der Sluys$^{1}$, C.\ R\"{o}ver$^{2,3}$, A.\ Stroeer$^{1,4}$, V. Raymond$^{1}$, I.\ Mandel$^{1}$, \\
43:   N.\ Christensen$^{5}$, V.\ Kalogera$^{1}$, R.\ Meyer$^{2}$, A.\ Vecchio$^{1,4}$}
44: \altaffiltext{1}{Physics \& Astronomy, Northwestern U., Evanston IL, USA} 
45: \altaffiltext{2}{Statistics, U.\ of Auckland, Auckland, New Zealand}
46: \altaffiltext{3}{Max-Planck-Institut f\"{u}r Gravitationsphysik, Hannover, Germany}
47: \altaffiltext{4}{Physics \& Astronomy, U. of Birmingham, Edgbaston, Birmingham, UK}
48: \altaffiltext{5}{Physics \& Astronomy, Carleton College, Northfield  MN, USA}
49: 
50: \begin{abstract}
51:   Inspiral signals from binary compact objects (black holes and neutron stars) are primary 
52:   targets of the ongoing searches by ground-based gravitational-wave interferometers 
53:   (LIGO, Virgo, GEO-600 and TAMA-300). 
54:   We present parameter-estimation simulations for inspirals of black-hole--neutron-star 
55:   binaries using Markov-chain Monte-Carlo methods. 
56:   For the first time, we have both estimated the parameters of a binary inspiral source 
57:   with a spinning component and determined the accuracy of the parameter estimation,
58:   for simulated observations with ground-based gravitational-wave detectors.
59:   We demonstrate that we can obtain the distance, sky position, and binary orientation
60:   at a higher accuracy than previously suggested in the literature.
61:   For an observation of an inspiral with sufficient spin and two or three detectors
62:   we find an accuracy in the determination of the sky position of typically a few tens
63:   of square degrees.
64: \end{abstract}
65: 
66: \keywords{Binaries: close, Gamma rays: bursts, Gravitational waves, Relativity}
67: 
68: 
69: \section{Introduction}
70: \label{sec:intro}
71: 
72: Binary systems with compact objects --- neutron stars (NS) and black holes (BH) --- in the mass 
73: range $\sim 1\,\Ms - 100\,\Ms$ are among the most likely sources of gravitational waves (GWs) 
74: for ground-based laser interferometers currently in operation (Cutler \& Thorne, 2002): LIGO 
75: (Barish \& Weiss 1999), Virgo (Arcese et al.\ 2004), GEO-600 (Willke et al.\ 2004) and TAMA-300 
76: (Takahashi et al.\ 2004). 
77: Merger-rate estimates are quite uncertain and for BH-NS binaries current detection-rate 
78: estimates reach as high as 0.1\,yr$^{-1}$ (\emph{e.g.} O'Shaughnessy et al.\ 2008) for first-generation 
79: instruments. 
80: Upgrades to Enhanced LIGO/Virgo (2008--2009) and Advanced LIGO/Virgo (2011--2014)
81: are expected to increase detection rates by factors of about $\sim 8$ and $10^3$, respectively. 
82: 
83: The measurement of astrophysical source properties holds major promise for 
84: improving our physical understanding and requires reliable methods for parameter estimation. 
85: This is a challenging problem because of the large 
86: number of parameters ($> 10$) and the presence of strong correlations among them, leading 
87: to a highly-structured parameter space. In the case of high mass ratio binaries (\emph{e.g.} BH-NS),
88: these issues are amplified 
89: for significant spin magnitudes and large spin misalignments 
90: (Apostolatos et al.\ 1994; Grandcl{\'e}ment et al.\ 2003; Buonanno et al.\ 2003). 
91: However, the presence of  
92: spins benefits parameter estimation through the signal modulations, although still  
93: presenting us with a considerable computational challenge. This has been highlighted  
94: in the context of LISA observations (see Vecchio 2004; Lang \& Hughes 2006) but no study has  
95: been devoted so far to ground-based observations.
96: 
97: 
98: In this \emph{Letter} we examine for the first time the potential for parameter estimation of  
99: spinning binary inspirals with ground-based interferometers, including twelve
100: physical parameters. Earlier studies (\emph{e.g.} Cutler \& Flanagan 1994, Poisson \& Will 1995, 
101: Van den Broeck \& Sengupta 2007) have estimated the theoretical accuracy with which some
102: of these parameters should be measured, without determining 
103: the parameters themselves. Also, (R\"{o}ver et al.\ 2006, 2007) have explored parameter 
104: estimation for non-spinning binaries. We focus on BH-NS binaries  
105: where spin effects are strongest 
106: (Apostolatos et al.\ 1994), while at the same time we are justified to ignore the NS spin.
107: We employ a newly developed Markov-chain Monte-Carlo (MCMC) algorithm (van der Sluys et al.\ 2008)
108: applied on spinning inspiral signals injected into synthetic ground-based noise and we derive posterior  
109: probability-density functions (PDFs) of all twelve signal parameters. 
110: We show that although sky position is degenerate 
111: when using two detectors, we can still determine the mass and 
112: spin parameters to reasonable accuracy.  With three detectors, the sky position 
113: and binary orientation can be fully resolved.
114: We show that our accuracies are good enough to associate an inspiral event with an electromagnetic
115: detection, such as a short gamma-ray burst (\emph{e.g.} Nakar 2007).
116: 
117: 
118: 
119: \section{Signal and observables}
120: \label{sec:GW}
121: 
122: In this \emph{Letter} we concentrate on the signal produced during the inspiral phase of two  
123: compact objects of masses $M_{1,2}$ in circular orbit. We  
124: focus on a fiducial BH-NS binary system with $M_1 = 10\,\Ms$ and $M_2 = 1.4\,\Ms$, so  
125: that we can ignore the NS spin. The BH spin $\mathbf{S}$
126: couples to the orbital angular momentum, leading to amplitude and phase modulation  
127: of the observed radiation due to the precession of the orbital plane during the observation. 
128: Here we model GWs by post-Newtonian (pN) waveforms at $^{1.5}$-pN order in phase and Newtonian 
129: amplitude.
130: We adopt the \emph{simple-precession} 
131: limit (Eqs.\,51, 52, 59 \& 63 in Apostolatos et al.\ 1994), appropriate  
132: for the single-spin system considered here. For simplicity (to speed up the  
133: waveform calculation), we ignore the \emph{Thomas-precession phase} (Apostolatos et al.\ 1994).
134: In this simple-precession approximation, the orbital angular momentum $\mathbf{L}$ and
135: spin $\mathbf{S}$ precess with the \emph{same} angular frequency 
136: around a fixed direction $\hat{\mathbf{J}}_0 \approx \hat{\mathbf{J}}$, where $\mathbf{J} = \mathbf{L} + \mathbf{S}$.
137: During the inspiral phase the spin misalignment
138:  $\theta_\mathrm{SL} \equiv \mathrm{arccos}({\bf \hat S} \cdot {\bf \hat L})$ and $S = |\mathbf{S}|$ 
139: are constant. These approximated waveforms retain (at the leading  
140: order) all the salient qualitative features introduced by the spins, while allowing us to  
141: compute the waveforms analytically, at great computational  
142: speed. While this approach is justified for exploration of GW astronomy and development of  
143: parameter-estimation algorithms, more accurate waveforms (\emph{e.g.} Kidder 1995;  
144: Faye et al.\ 2006; Blanchet et al.\ 2006) will be necessary for the analysis of real signals. 
145: 
146: A circular binary inspiral with one spinning compact object is described by a 12-dimensional parameter  
147: vector $\vec{\lambda}$. With respect to a fixed geocentric coordinate system our choice of  
148: independent parameters is:
149: \be
150: \vec{\lambda} = \{\Mc,\eta,\mathrm{R.A.},\mathrm{Dec},\cos\theta_{J_0},\phi_{J_0},\log{d_L},a_\mathrm{spin},\cos\theta_\mathrm{SL}, \phi_c,\alpha_c, t_c\},
151: \label{e:lambda}
152: \ee
153: where $\Mc = \frac{(M_1 M_2)^{3/5}}{(M_1 + M_2)^{1/5}}$ and $\eta = \frac{M_1 M_2}{(M_1 + M_2)^2}$ are  
154: the chirp mass and symmetric mass ratio, respectively; $\mathrm{R.A.}$ (right ascension)  
155: and $\mathrm{Dec}$ (declination) identify the source position in the sky;  
156: the angles $\theta_{J_0}$ and $\phi_{J_0}$ (defined in the range $\theta_{J_0} \in \left[-\frac{\pi}{2},\frac{\pi}{2} \right]$  
157: and $\phi_{J_0}\in \left[0, 2\pi \right[$) identify the unit vector $\hat{\mathbf{J}}_0$; $d_L$ is the  
158: luminosity distance to the source and $0 \le a_\mathrm{spin} \equiv S/M_1^2 \le 1$ is the  
159: dimensionless spin magnitude; $\phi_c$ and $\alpha_c$ are integration constants that specify  
160: the GW phase and the location of $\mathbf{S}$ on the precession cone, respectively, at 
161: the time of coalescence $t_c$. 
162: 
163: Given a network comprising $n_\mathrm{det}$ detectors, the data collected at the $a-$th  
164: instrument ($a = 1,\dots, n_\mathrm{det}$) is given by $x_a(t) = n_a(t) + h_a(t;\vec{\lambda})$,  
165: where $h_a(t;\vec{\lambda}) = F_{a,+}(t)\,h_{a,+}(t;\vec{\lambda}) + F_{a,\times}(t)\,h_{a,\times}(t;\vec{\lambda})$  
166: is the GW strain at the detector (see Eqs.\,2--5 in Apostolatos et al.\ 1994)
167: and $n_a(t)$ is the detector noise. The astrophysical signal  
168: is given by the linear combination of the two independent polarisations $h_{a,+}(t;\vec{\lambda})$  
169: and $h_{a,\times}(t;\vec{\lambda})$ weighted by the \emph{time-dependent} antenna beam  
170: patterns $F_{a,+}(t)$ and $F_{a,\times}(t)$. An example of $h_a$ for $\theta_\mathrm{SL}=20^\circ$  
171: and $a_\mathrm{spin}=0.1$ and $0.8$ is shown in panels a--b of Fig.\,\ref{fig:pdf}. In our  
172: analysis we model the noise in each detector as a zero-mean Gaussian,  stationary random  
173: process, with one-sided noise spectral density $S_a(f)$ at the initial-LIGO design sensitivity,  
174: where $f$ is the frequency. 
175: 
176: \section{Parameter estimation: Methods and results}
177: \label{sec:results}
178: 
179: The goal of our analysis is to determine the \emph{posterior} PDF of the unknown parameter  
180: vector $\vec{\lambda}$ in Eq.~(\ref{e:lambda}), given the data sets $x_a$ collected by a  
181: network of $n_\mathrm{det}$ detectors and the \emph{prior} $p(\vec{\lambda})$ on the  
182: parameters.  We use wide, flat priors (see Van der Sluys et al.\ 2008 for details).
183: Bayes' theorem provides a rigorous mathematical rule to assign such a probability: 
184: \be
185: p(\vec{\lambda}|x_a ) = \frac{p(\vec{\lambda}) \, {\cal L} (x_a|\vec{\lambda})}{p(x_a)}\,;
186: \label{e:jointPDF}
187: \ee
188: in the previous Equation
189: \be
190: {\cal L}(x_a|\vec{\lambda}) \propto 
191: \exp\left\{-2
192: \int_{f_l}^{f_h} \frac{\left| \tilde{x}_a(f) - \tilde{h}_a(f;\vec{\lambda})\right|^2}{S_a(f)}\,\mathrm{d}f
193: \right\}
194: \label{e:La}
195: \ee
196: is the \emph{likelihood function} of the data given the model, 
197: which measures the fit of the data to the model,
198: and $p(x_a)$ is the  
199: \emph{marginal likelihood} or \emph{evidence}; $\tilde x(f)$ stands for the Fourier  
200: component of $x(t)$. For multi-detector observations involving a network of detectors with  
201: uncorrelated noise --- this is the case of this paper, where we do not use the pair of  
202: co-located LIGO instruments 
203: --- we have $p(\vec{\lambda}|\{x_a; a  
204: = 1,\dots,n_\mathrm{det}\}) = \prod_{a=1}^{n_\mathrm{det}}\, p(\vec{\lambda}|x_a)\,.$
205: 
206: The numerical computation of the joint and \emph{marginalised} PDFs involves the  
207: evaluation of integrals over a large number of dimensions. Markov-chain Monte-Carlo  
208: (MCMC) methods (\emph{e.g.} Gilks et al.\ 1996; Gelman et al.\ 1997, and references therein)  
209: have proved to be particularly effective in tackling these numerical problems.
210: We have developed an adaptive (see Figueiredo \& Jain, 2002; Atchade \& Rosenthal 2003)  
211: MCMC algorithm, intended to explore the parameter  
212: space efficiently while requiring the least amount of tuning for the specific signal at  
213: hand; the code is an extension of the one developed by some of the authors to explore  
214: MCMC methods for non-spinning binaries (R\"{o}ver et al.\ 2006, 2007) and takes advantage  
215: of techniques explored by some of us in the context of LISA data analysis (Stroeer et al.\  
216: 2007).  A summary of the methods used in our MCMC code has been published (Van der Sluys 
217: et al.\ 2008); more technical details will be provided elsewhere.
218: 
219: Here we present results obtained by adding a signal in simulated initial-LIGO noise 
220: and computing the posterior PDFs with MCMC techniques for a fiducial  
221: source consisting of a $10\,M_\odot$ spinning BH and a $1.4\,M_\odot$ non-spinning NS  
222: in a binary system with a signal-to-noise ratio (SNR) of 17.0 for the network of 2 or 3 detectors
223: (obtained by scaling the distance). We consider a number of cases for which we  
224: change the BH spin magnitude ($a_\mathrm{spin} = 0.0, 0.1, 0.5, 0.8$) and the angle between  
225: the spin and the orbital angular momentum ($\theta_\mathrm{SL} = 20^\circ, 55^\circ$);   
226: the remaining ten parameters, including source position and orientation of the total  
227: angular momentum, are kept constant ($R.A. = 14.3h$, Dec. $= 12^\circ$, $\theta_{J_0} = 4^\circ$
228: and $\phi_{J_0} = 289^\circ$ for this study). For each of the seven ($a_\mathrm{spin}$, $\theta_\mathrm{SL}$)  
229: combinations (six for finite spin, one for zero spin), we run the analysis using the 
230: data from (i) the 4-km LIGO detector at Hanford (H1) and the Virgo detector near Pisa
231: ($n_\mathrm{det}=2$), and (ii) the two LIGO 4-km interferometers (H1 and L1) and the Virgo 
232: detector ($n_\mathrm{det}=3$). This results in a total of 14 signal cases explored in this study. 
233: The MCMC analysis that we carry out on each data set consists of 5 separate serial chains, 
234: each with a length of $3.5\times10^6$ iterations ($n_\mathrm{det}=2$) or $2.5\times10^6$ 
235: iterations ($n_\mathrm{det}=3$),
236: sampled after a \emph{burn-in} period (see \emph{e.g.} Gilks et al.\ 1996) that is determined
237: automatically as follows: we determine the absolute maximum likelihood $L_\mathrm{max}$ 
238: that is obtained in any of the five chains, and for each chain include all the iterations 
239: {\it after} the chain has reached a likelihood value of $L_\mathrm{max}-2$.
240: Each chain starts at offset ({\it i.e.}, non-true) parameter values. The starting values
241: for chirp mass and the time of coalescence are drawn from a Gaussian distribution 
242: centred on the true parameter value, with a standard deviation of about $0.1\,M_\odot$ and 
243: 30\,ms respectively.  The other ten parameters are drawn randomly from the allowed ranges.
244: Multiple chains starting from offset parameters and locking on to the same values for the
245: parameters and likelihood provide convincing evidence of convergence in a blind analysis.
246: Our MCMC code needs to run for typically one week to show the first results and 10--14 days
247: to accumulate a sufficient number of iterations for good statistics, each serial chain using
248: a single 2.8\,GHz CPU.
249: An example of the PDFs obtained for a signal characterised by $a_\mathrm{spin}=0.1$ and  
250: $\theta_\mathrm{SL}=20^\circ$ is shown in panels c--f of Fig.\,\ref{fig:pdf}, for the cases  
251: of 2 and 3 detectors; the PDFs for $M_1$ and $M_2$ in Fig.\,\ref{fig:pdf}d are constructed  
252: from those obtained for $\Mc$ and $\eta$.
253: 
254: %\begin{figure}  %For aastex
255: \begin{figure*} %For emulateapj
256:   %\figurenum{}
257:   \epsscale{1.0}
258:   \plotone{f1.eps} %For submission
259:   %\plotone{figures/waveform_pdfs.eps} %Local
260:   \figcaption{
261:     {\bf (a)} Part of the waveform from a source with $a_\mathrm{spin}=0.1$ and  
262:     $\theta_\mathrm{SL}=20^\circ$.   
263:     {\bf (b)} the same waveform, but for $a_\mathrm{spin}=0.8$. 
264:     {\bf (c)} Posterior PDF of the luminosity distance for a signal with $a_\mathrm{spin}=0.5$  
265:     and $\theta_\mathrm{SL}=20^\circ$, as determined with the signal of two (left PDF) and  
266:     three (right PDF) detectors. The dashed lines show the true distance, which is higher 
267:     for the three-detector case to obtain the same SNR.   
268:     {\bf (d--f)} Two-dimensional posterior PDF showing the 99\%-probability areas for 
269:     the same runs as (c), for the individual masses, 
270:     where the ellipses are aligned with the line of constant $\cal{M}$ 
271:     (d), the spin parameters (e) and the position in the sky (f). 
272:     The dashed lines display the true parameter values.
273:     Upward and downward hashes show the result for two and three detectors respectively 
274:     in panels {\bf (c--f)}. 
275:     \label{fig:pdf}
276:   }
277: %\end{figure}  %For aastex
278: \end{figure*} %For emulateapj
279: 
280: In order to evaluate the parameter-estimation accuracy we compute probability intervals; in  
281: Table\,\ref{tab:deltas} we report the 90\%-probability interval for each of the parameters,  
282: defined as the smallest range for which the posterior probability of a given parameter to be  
283: in that range is 0.9.  For the two-dimensional cases (position and orientation)
284: we quote the smallest \emph{area} that contains 90\% of the probability. 
285: From the 140 marginalised PDFs considered here (ignoring the 
286: derived parameters $M_1$, $M_2$ and combining R.A., Dec as position and $\theta_\mathrm{J0}$, 
287: $\phi_\mathrm{J0}$ as orientation), the true parameter values lie outside the $90\%$ probability 
288: range in 27 cases, marked with superscripts in Table\,\ref{tab:deltas}. For  
289: the 21 cases marked with $^a$ the true parameter is within the 99\%-probability range,
290: the 6 cases marked with $^b$ lie outside the 99\% but inside the 100\% range.
291: 
292: 
293: %\begin{deluxetable}{lllllllllllllllllll} %For aastex
294: \begin{deluxetable*}{lllllllllllllllllll} %For emulateapj
295:   %\tabletypesize{\scriptsize} %For aastex
296:   %\tablewidth{0pt}
297:   \tablecolumns{19}
298:   \tablecaption{
299:     Injection details and widths of the 90\%-probability intervals of the MCMC runs described in the text
300:     \label{tab:deltas}
301:   }
302:   \tablehead{
303:     \colhead{$n_\mathrm{det}$$\!\!\!$} & \colhead{$a_\mathrm{spin}$$\!\!\!$} & \colhead{$\theta_\mathrm{SL}$$\!\!\!$} & \colhead{$d_\mathrm{L}$$\!\!\!$} & 
304:     \colhead{~} &
305:     \colhead{$M_1$$\!\!\!$} & \colhead{$M_2$$\!\!\!$} &
306:     \colhead{~} &
307:     \colhead{$\cal{M}$$\!\!\!$} & \colhead{$\eta$$\!\!\!$} & \colhead{$t_\mathrm{c}$$\!\!\!$} & \colhead{$d_\mathrm{L}$$\!\!\!$} & 
308:     \colhead{$a_\mathrm{spin}$$\!\!\!$} & \colhead{$\theta_\mathrm{SL}$$\!\!\!$} & 
309:     \colhead{$\phi_\mathrm{c}$$\!\!\!$} & 
310:     \colhead{$\alpha_\mathrm{c}$$\!\!\!$} & 
311:     \colhead{~} &
312:     \colhead{Pos.$\!\!\!$} & \colhead{Ori.$\!\!\!$} \\
313:     %
314:     & & ($^\circ$) & (Mpc) & 
315:     &
316:     (\%) & (\%) & 
317:     &
318:     (\%) & (\%) & (ms) & (\%) & 
319:     & ($^\circ$) & 
320:     ($^\circ$) &  ($^\circ$)  &  
321:     &
322:     ($^{\circ^2}$) &  ($^{\circ^2}$) 
323:   }
324:   \startdata
325:   2  &     0.0  &       0  &    16.0  &  &    95       &    83       &  &   2.6   &   138   &    18   &    86   &  0.63   &---   &   323   &---   & &   537   & 19095  \\
326:   2  &     0.1  &      20  &    16.4  &  &   102       &    85       &  &   1.2   &    90   &    10   &    91   &  0.91   &   169   &   324   &   326$^a$   & &   406   & 16653  \\
327:   2  &     0.1  &      55  &    16.7  &  &    51       &    38       &  &  0.88   &    59   &   7.9   &    58   &  0.32   &   115   &   322   &   326   & &   212   &  3749  \\
328:   2  &     0.5  &      20  &    17.4  &  &    53$^b$   &    42$^a$   &  &  0.90   &    50$^b$   &   5.4   &    46$^a$   &  0.26   &    56   &   330   &   301$^b$   & &   111$^a$   &  3467$^a$  \\
329:   2  &     0.5  &      55  &    17.3  &  &    31       &    24       &  &  0.62   &    41   &   4.9   &    21   &  0.12   &    24   &   323   &   269$^a$   & & 19.8   &   178$^a$  \\
330:   2  &     0.8  &      20  &    17.9  &  &    54$^a$   &    42$^a$   &  &  0.86$^a$   &    54$^a$   &   6.0   &    56   &  0.16   &    25$^a$   &   325   &   319   & &  104$^a$   &  1540  \\
331:   2  &     0.8  &      55  &    17.9  &  &    21       &    16       &  &  0.66   &    29   &   4.7   &    22   &  0.15   &    15   &   320   &   323   & & 22.8   &   182$^a$  \\
332:   & & & &  &  & &  &   & & & & & & & &  & & \\
333:   3  &     0.0  &       0  &    20.5  &  &   114       &    90       &  &   2.6   &   119   &    15   &    69   &  0.98$^b$   &---   &   325   &---   & &  116   &  4827  \\
334:   3  &     0.1  &      20  &    21.1  &  &    70       &    57       &  &  0.92   &    72   &   7.0   &    60   &  0.49   &   160   &   321   &   322$^a$   & &  64.7   &  3917  \\
335:   3  &     0.1  &      55  &    21.4  &  &    62       &    48       &  &  0.93   &    68   &   6.2   &    51   &  0.52   &   123   &   325   &   308$^a$   & & 48.7   &   976  \\
336:   3  &     0.5  &      20  &    22.3  &  &    54$^b$   &    44$^a$   &  &  0.89$^a$   &    48$^b$   &   3.3   &    52   &  0.28$^a$   &    69   &   318   &   229$^b$ &  &  28.8   &   849  \\
337:   3  &     0.5  &      55  &    22.0  &  &    33       &    25       &  &  0.62   &    43   &   4.6   &    23$^a$   &  0.14   &    27   &   322   &   324   & &  20.7   &   234$^a$  \\
338:   3  &     0.8  &      20  &    23.0  &  &    53$^b$   &    41$^a$   &  &  0.85$^a$   &    52$^b$   &   3.8   &    55   &  0.17   &    23$^a$   &   320   &   327$^a$  & &  36.4$^a$   &   645  \\
339:   3  &     0.8  &      55  &    22.4  &  &    30       &    22       &  &  0.86   &    40   &   5.0   &    26   &  0.21   &    21   &   322   &   323   & &  27.2   &   288 
340:   \enddata
341:   \tablenotetext{~}{$^a$ the true value lies outside the 90\%-probability range; $^b$ idem, outside the 99\%-probability range, but inside the 100\% range}
342: %\end{deluxetable} %For aastex
343: \end{deluxetable*} %For emulateapj
344: 
345: 
346: 
347: 
348: We find that most of these outliers are caused by a degeneracy between the mass and spin 
349: parameters.  A parameter set with different values for $\cal{M}$, $\eta$, $a_\mathrm{spin}$ 
350: and $\theta_\mathrm{SL}$ can produce a waveform that is almost identical to the signal we
351: injected.  For the chirp mass and spin parameters, the distance between the two degenerate
352: regions is relatively small. 
353: However, for the mass ratio $\eta$, these two regions ($\eta\approx0.11$, the injected value
354: and $\eta\approx0.2$) are far apart and seem disconnected. 
355: A comparison of waveforms from the two degenerate regions demonstrates that their overlap is so
356: high (match $>99.5\%$) that it would be impossible to tell which is the true signal even at 
357: high SNR.  This degeneracy could be physical or could be caused by the simplified waveform model;
358: further investigation is warranted.
359: 
360: 
361: For a detection with two interferometers, the sky position and binary orientation are 
362: degenerate; for low spin, our PDFs show an incomplete ring in the sky where the source might 
363: be.  When the BH spin increases, the allowed sky location shrinks appreciably until mere arcs 
364: are left (Fig.\,\ref{fig:pdf}f).
365: For intermediate and high spin, and $\theta_\mathrm{SL}=55^\circ$, we typically find only one
366: such arc, reducing the sky position to several degrees (Table\,\ref{tab:deltas}).
367: 
368: Even with two detectors the source parameters can be measured at 
369: astrophysically interesting levels when sufficient spin is present, including distance, individual  
370: masses, spin magnitude and tilt angle; 
371: for $a_\mathrm{spin}=0.5$ or more, the typical uncertainty in the sky position is
372: a few tens of square degrees, the distance is determined with 20--60\% accuracy and the timing 
373: accuracy is 6\,ms or better. 
374: 
375: The accuracy of the parameter determination is affected by the number of 
376: detectors used, a result well established in studies of inspirals of non-spinning 
377: compact objects (\emph{e.g.} Jaranowski, P., \& Krolak 1994; Pai et al.\ 2001; Cavalier et al.\ 2006; 
378: R\"{o}ver et al.\ 2007).
379: Unlike some other studies, we keep the SNR of the detector network constant; when a third detector is added, the
380: distance to the source is increased (Fig.\,\ref{fig:pdf}c).  This way, we see the effect of the additional information
381: that is provided by the extra interferometer and eliminate the effect of the higher SNR.
382: Table\,\ref{tab:deltas} shows that the effect on the uncertainty in the mass and spin 
383: parameters is marginal when adding a third interferometer to the network.
384: The uncertainty in the distance and time of coalescence 
385: decreases typically by 20--25\% when using three detectors, but the largest effect is on the
386: accuracy for sky position and binary orientation; Table\,\ref{tab:deltas}
387: shows that the (two-dimensional) uncertainties in these parameters decrease by 50\% and
388: 40\% respectively on average.
389: %See ~/work/GW/programs/MCMC/output/papers/ApJL_01/draft_0806/table
390: 
391: 
392: The parameter-estimation accuracy also depends strongly on the actual spin 
393: parameters of the system: as a general trend, the larger $a_\mathrm{spin}$ and $\theta_\mathrm{SL}$,  
394: the stronger the modulations in the waveform induced by precession, and the more information
395: is coded up in the waveform.  
396: When we divide our simulations into low spin ($a_\mathrm{spin}=0.0,0.1$) and high spin 
397: ($a_\mathrm{spin}=0.5,0.8$) cases, we find that the uncertainties in the high-spin case are smaller
398: by 40--60\% for the masses, time of coalescence and distance, by 65--70\% for the 
399: spin parameters and by 80--90\% for the sky position and binary orientation.
400: However, the width of the 90\%-probability interval is in fact not strictly monotonic as a function 
401: of $a_\mathrm{spin}$ and $\theta_\mathrm{SL}$ (Table\,\ref{tab:deltas}).
402: The increasingly complex structure of the likelihood 
403: function and stronger correlations amongst different parameters for higher spin have an important 
404: effect on the sampling efficiency of the MCMC. 
405: 
406: 
407: 
408: Earlier studies
409: (\emph{e.g.} Cutler \& Flanagan (1994, their Tables\,II \& III and Fig.\,7);
410: Poisson \& Will (1995, their Table\,II); Van den Broeck \& Sengupta (2007, their Table\,III))
411: have reported on the theoretical accuracy of parameter estimation. 
412: These explorations are based on the Fisher 
413: matrix, which yields the expected uncertainty (for unimodal 
414: distributions), without actually estimating the parameter values themselves.  The quoted accuracies 
415: for masses and the time and phase of coalescence are typically better than or similar to the 
416: values in our Table\,\ref{tab:deltas}.  We have been able to estimate distance, sky position and 
417: binary orientation to better accuracy than suggested in these studies.
418: 
419: 
420: \section{Conclusions}
421: \label{sec:concl}
422: 
423: We have explored for the first time the 
424: parameter estimation
425: of all physical parameters --- including masses, spin, 
426: distance, sky location and binary orientation --- on ground-based gravitational-wave
427: observations of binary inspirals with spinning compact objects. 
428: We show that for two detectors and sufficient spin ($a_\mathrm{spin}\geq0.5$) or
429: for three detectors, the obtained accuracy in sky position, distance and time of 
430: coalescence is good enough to allow the identification of electromagnetic 
431: counterparts of compact-binary mergers, \emph{e.g.} short gamma-ray bursts (Nakar 2007).
432: A direct measurement of mass, spin, distance and orientation can be obtained from inspiral GWs, which
433: is notoriously difficult for electromagnetic observations. 
434: 
435: The analysis presented here is the first step of a more detailed  
436: study that we are currently carrying out, exploring a much larger parameter space,  
437: developing techniques to reduce the computational cost of these simulations, and  
438: testing the methods with actual LIGO data. The waveform model used here, though adequate for  
439: exploratory studies, is not sufficiently accurate for the analysis of real detections, 
440: and we are finalising the implementation of a more realistic waveform. Simulations
441: with this improved waveform may also shed light on the degeneracy between mass
442: and spin parameters discussed in Sect.\,\ref{sec:results}, and may improve the 
443: accuracy of our parameter estimation appreciably (\emph{e.g.} Van den Broeck \& 
444: Sengupta 2007).
445: Finally, we intend to further develop our Bayesian approach into one of the standard  
446: tools that can be included in the analysis pipeline used for the processing of the  
447: `science data' collected by ground-based laser interferometers.
448: 
449: \acknowledgments 
450: 
451: This work is partially supported by a Packard Foundation Fellowship, a NASA BEFS grant  
452: (NNG06GH87G), and a NSF Gravitational Physics grant (PHY-0353111) to VK; NSF Gravitational  
453: Physics grant PHY-0553422 to NC; Royal Society of New Zealand Marsden Fund grant UOA-204  
454: to RM and CR; UK Science and Technology Facilities Council grant to AV; Computations were  
455: performed on the Fugu computer cluster funded by NSF MRI grant PHY-0619274 to VK. 
456: 
457: \begin{thebibliography}{}
458: 
459: \bibitem[Apostolatos et al.(1994)]{1994PhRvD..49.6274A} 
460: Apostolatos, T.~A., Cutler, C., Sussman, G.~J., \& Thorne, K.~S.\ 1994, \prd, 49, 6274 
461: 
462: \bibitem{virgo}
463: Arcese, F., et al.\ 2004, Class.\ Quant.\ Grav., 21,  385
464: 
465: \bibitem{AtchadeRosenthal:2003}
466: Atchade, Y.~F., \& Rosenthal, J.~S.\ 2005,  Bernoulli 11(5) 815
467: 
468: \bibitem[Barish \& Weiss(1999)]{1999PhT....52j..44B} 
469: Barish, B.~C., \& Weiss, R.\ 1999, Physics Today, 52, 44
470: 
471: %\bibitem{BelczynskiEtAl:2008}
472: %Belczynski, K., Taam, R.~E., Rantsiou, E. \& van der Sluys, M.\ 2008, ApJ, in press, arXiv:astro-ph/0703131
473: 
474: \bibitem[Blanchet et al.(1995)]{1995PRL..74L3515B} 
475: Blanchet, L., Damour, T., Iyer, B.~R., Will, C.~M., \& Wiseman, A.~G.\ 1995, \prl, 74, 3515
476: 
477: \bibitem[Blanchet et al.(2006)]{2006PhRvD..74j4034B} 
478: Blanchet, L., Buonanno, A., \& Faye, G.\ 2006, \prd, 74, 104034
479: 
480: \bibitem[Buonanno et al.(2003)]{2003PhRvD..67j4025B} 
481: Buonanno, A., Chen, Y., \& Vallisneri, M.\ 2003, \prd, 67, 104025. Erratum-ibid.\ 2006, 74, 029904(E)
482: 
483: \bibitem[Cavalier et al.(2006)]{2006PhRvD..74h2004C}
484: Cavalier, F., et al.\  2006, \prd, 74, 082004
485: 
486: \bibitem[Cutler \& Flanagan(1994)]{1994PhRvD..49.2658C}
487: Cutler, C. \& Flanagan, E.~E.\ 1994, \prd, 49, 2658
488: 
489: \bibitem{CutlerThorne:2002}
490: Cutler, C. and Thorne, K.~S., \ 2002,  gr-qc/0204090
491: 
492: \bibitem[Faye et al.(2006)]{2006PhRvD..74j4033F} 
493: Faye, G., Blanchet, L., \& Buonanno, A.\ 2006, \prd, 74, 104033
494: 
495: \bibitem{FigueiredoJain:2002}
496: Figueiredo, M.~A.~T., \& Jain, A.~K.\ 2002, IEEE Transactions on Pattern Analysis and Machine Intelligence, 24, 281
497: 
498: \bibitem{GelmanEtAl:1997}
499: Gelman, A., Carlin, J.~B., Stern, H., \& Rubin, D.~B.\ 1997,  \emph{Bayesian data analysis}, Chapman \& Hall, Boca Raton
500: 
501: \bibitem{GilksEtAl:1996}
502: Gilks, W.~R., Richardson, S., \& Spiegelhalter, D.~J.\ 1996, \emph{Markov chain Monte Carlo in practice}, Chapman \& Hall
503: 
504: \bibitem[Grandcl{\'e}ment et al.(2003)]{2003PhRvD..67d2003G} 
505: Grandcl{\'e}ment, P., Kalogera, V., \& Vecchio, A.\ 2003, \prd, 67, 042003
506: 
507: \bibitem{JK94}
508: Jaranowski, P., \& Krolak, A.\ 1994, \prd, 49, 1723
509: 
510: \bibitem[Kidder(1995)]{1995PhRvD..52..821K}
511: Kidder, L.~E.\ 1995, \prd, 52, 821
512: 
513: \bibitem[Lang \& Hughes(2006)]{2006PhRvD..74l2001L}
514: Lang, R.~N., \& Hughes, S.~A.\ 2006, \prd, 74, 122001. Erratum-ibid.\ 2007, 75, 089902(E)
515: 
516: \bibitem[Nakar(2007)]{2007PhR...442..166N}
517: Nakar, E.\ 2007, \physrep, 442, 166
518: 
519: \bibitem[O'Shaughnessy et al.(2008)]{2008ApJ...672.479O}
520: O'Shaughnessy, R., Kalogera, V., \& Belczynski, K.\ 2008, \apj, 672, 479
521: 
522: \bibitem[Pai et al.(2001)]{2001PhRvD..64d2004P}
523: Pai, A., Dhurandhar, S., \& Bose, S.\ 2001, \prd, 64, 042004
524: 
525: \bibitem[Poisson \& Will(1995)]{1995PhRvD..52.848P}
526: Poisson, E. \& Will, C.~M.\ 1995, \prd, 52, 848
527: 
528: \bibitem{RoeverEtAl:2006a}
529: R\"{o}ver, C., Meyer, R., \& Christensen, N.\ 2006, Class.\ Quant.\ Grav., 23, 4895
530: 
531: \bibitem{RoeverEtAl:2007}
532: R\"{o}ver, C., Meyer, R., \& Christensen, N.\ 2007, \prd,  75, 062004
533: 
534: \bibitem[Stroeer et al.(2007)]{2007CQGra..24S.541S}
535: Stroeer, A., et al.\  2007, Class.\ Quant.\ Grav., 24, 541
536: 
537: \bibitem[Takahashi \& the TAMA Collaboration(2004)]{2004CQGra..21S.403T}
538: Takahashi, R., \& the TAMA Collaboration 2004, Class.\ Quant.\ Grav.,  21, 403
539: 
540: \bibitem[Van den Broeck \& Sengupta(2007)]{2007CQGra..24.1089V}
541: Van den Broeck, C. \& Sengupta, A.~S.\ 2007, Class.\ Quant.\ Grav.\ 24, 1089
542: 
543: \bibitem[Van der Sluys et al.(2008)]{arXiv:0805.1689}
544: Van der Sluys, M., et al.\  2008,  Class.\ Quant.\ Grav., in press, arXiv:0805.1689
545: 
546: \bibitem[Vecchio(2004)]{2004PhRvD..70d2001V}
547: Vecchio, A.\ 2004, \prd, 70, 042001
548: 
549: \bibitem[Willke et al.(2004)]{2004CQGra..21S.417W}
550: Willke, B., et al.\ 2004, Class.\ Quant.\ Grav., 21, 417
551: 
552: 
553: \end{thebibliography}
554: 
555: 
556: \end{document}
557: 
558: 
559: