1: \documentclass[prd,aps,showpacs,floats,floatfix,superscriptaddress,twocolumn,nofootinbib]{revtex4}
2:
3: %\renewcommand{\baselinestretch}{2}
4: %\usepackage{showkeys}
5: %Alessandra's style
6:
7: \usepackage{graphicx}
8: \usepackage{color}
9: \usepackage{amssymb}
10: \usepackage{amsmath}
11:
12: \usepackage{longtable}
13: \special{papersize=8.5in,11in}
14:
15: \def\laq{\raise 0.4ex\hbox{$<$}\kern -0.8em\lower 0.62ex\hbox{$\sim$}}
16: \def\gaq{\raise 0.4ex\hbox{$>$}\kern -0.7em\lower 0.62ex\hbox{$\sim$}}
17: \def\tablerule{\noalign {\vskip3truept\hrule\vskip3truept}}
18:
19: \newcommand{\red}{\textcolor{red}}
20: \newcommand{\blue}{\textcolor{blue}}
21:
22: \newcommand{\hLN}{\hat{\mathbf{L}}_{\mathrm{N}}}
23: \newcommand{\hL}{\hat{\mathbf{L}}}
24: \newcommand{\hS}{\hat{\mathbf{S}}}
25: \newcommand{\vh}{\mathbf{h}}
26: \newcommand{\hl}{\hat{\mathbf{\lambda}}}
27: \newcommand{\hn}{\hat{\mathbf{n}}}
28: \newcommand{\hv}{\hat{\mathbf{v}}}
29: \newcommand{\hN}{\hat{\mathbf{N}}}
30: \newcommand{\hJ}{\hat{\mathbf{J}}}
31: \newcommand{\vx}{\mathbf{x}}
32: \newcommand{\vv}{\mathbf{v}}
33: \newcommand{\va}{\mathbf{a}}
34: \newcommand{\ve}{\mathbf{e}}
35: \newcommand{\vT}{\mathbf{T}}
36: \newcommand{\vJ}{\mathbf{J}}
37: \newcommand{\vS}{\mathbf{S}}
38: \newcommand{\vL}{\mathbf{L}}
39:
40:
41: \newcommand{\beq}{\begin{equation}}
42: \newcommand{\eeq}{\end{equation}}
43: \newcommand{\bea}{\begin{eqnarray}}
44: \newcommand{\eea}{\end{eqnarray}}
45: \newcommand{\bse}{\begin{subequations}}
46: \newcommand{\ese}{\end{subequations}}
47: \newcommand{\ba}{\begin{array}}
48: \newcommand{\ea}{\end{array}}
49:
50: \newcommand{\wt}{\hat{t}}
51: \newcommand{\wH}{\widehat{H}}
52: \newcommand{\wF}{\widehat{\cal F}}
53: \newcommand{\ww}{\widehat{\omega}}
54: \newcommand{\pa}{\partial}
55: \newcommand{\vphi}{\varphi}
56:
57: %\newcommand{\eqref}[1]{(\ref{#1})}
58:
59: \newcommand{\ket}[1]{{#1}}
60: \newcommand{\braket}[2]{{\langle #1,#2 \rangle}}
61:
62: \def\np{({\bf n}\cdot{\bf p})}
63: \def\pp{{\bf p}^2}
64: \def\ppp{({\bf p}^2)}
65:
66:
67: \newcommand{\mytextbf}[1]{\textbf{#1}}
68:
69: \newcommand{\mytextrm}[1]{{}}
70: \newcommand{\myremark}[1]{\textbf{#1}}
71:
72: \newcommand{\thispaper}{paper}
73:
74: \newlength{\sizeonefig}
75: \newlength{\sizetwofig}
76: \newlength{\sizeonefigb}
77: \newlength{\sizetwofigb}
78: \setlength{\sizeonefig}{0.45\textwidth}
79: \setlength{\sizetwofig}{0.45\textwidth}
80: \setlength{\sizeonefigb}{0.35\textheight}
81: \setlength{\sizetwofigb}{0.35\textheight}
82:
83: \begin{document}
84:
85: \title{Anatomy of the binary black hole recoil: A multipolar analysis}
86:
87: \author{Jeremy D. Schnittman}
88:
89: \affiliation{Maryland Center for Fundamental Physics,
90: Department of Physics, University of Maryland, College
91: Park, Maryland 20742 }
92:
93: \author{Alessandra Buonanno}
94:
95: \affiliation{Maryland Center for Fundamental Physics,
96: Department of Physics, University of Maryland, College
97: Park, Maryland 20742 }
98:
99: \author{James R. van Meter}
100:
101: \affiliation{Gravitational Astrophysics Laboratory, NASA
102: Goddard Space Flight Center, 8800 Greenbelt Rd., Greenbelt, MD 20771}
103:
104: \affiliation{Center for Space Science \& Technology,
105: University of Maryland Baltimore
106: County, Physics Department, 1000 Hilltop Circle, Baltimore, MD 21250}
107:
108: \author{John G. Baker}
109:
110: \affiliation{Gravitational Astrophysics Laboratory, NASA
111: Goddard Space Flight Center, 8800 Greenbelt Rd., Greenbelt, MD 20771}
112:
113: \author{William D. Boggs}
114:
115: \affiliation{Department of Physics, University of Maryland, College
116: Park, Maryland 20742 }
117:
118: \author{Joan Centrella}
119:
120: \affiliation{Gravitational Astrophysics Laboratory, NASA
121: Goddard Space Flight Center, 8800 Greenbelt Rd., Greenbelt, MD 20771}
122:
123: \author{Bernard J. Kelly}
124:
125: \affiliation{Gravitational Astrophysics Laboratory, NASA
126: Goddard Space Flight Center, 8800 Greenbelt Rd., Greenbelt, MD 20771}
127:
128: \author{Sean T. McWilliams}
129:
130: \affiliation{Department of Physics, University of Maryland, College
131: Park, Maryland 20742 }
132:
133: \begin{abstract}
134: We present a multipolar analysis of the gravitational recoil computed
135: in recent numerical simulations of binary black hole (BH) coalescence, for both unequal
136: masses and non-zero, non-precessing spins. We show that multipole
137: moments up to and including $\ell=4$ are sufficient to accurately
138: reproduce the final
139: recoil velocity (within $\simeq 2\%$) and that only a few dominant modes contribute
140: significantly to it (within $\simeq 5\%$). We describe how the relative
141: amplitudes, and more importantly, the relative phases, of these few modes
142: control the way in which the recoil builds up
143: throughout the inspiral, merger, and ringdown phases.
144: We also find that the numerical results can
145: be reproduced by an ``effective Newtonian'' formula for the multipole moments
146: obtained by replacing the radial separation in the Newtonian formulae
147: with an effective radius computed from the numerical data.
148: Beyond the merger, the numerical results are reproduced by a
149: superposition of three Kerr quasi-normal modes (QNMs). Analytic formulae,
150: obtained by expressing the multipole moments in terms of the
151: fundamental QNMs of a Kerr BH, are able to explain the onset and
152: amount of ``anti-kick'' for each of the simulations. Lastly, we apply
153: this multipolar analysis to help explain the remarkable difference
154: between the amplitudes of planar and non-planar kicks for equal-mass
155: spinning black holes.
156: \end{abstract}
157:
158: \pacs{04.25.Dm, 04.30.Db, 04.70.Bw, 04.25.Nx, 04.30.-w}
159:
160:
161: \date{\today}
162:
163: \maketitle
164:
165: \section{Introduction}
166: \label{intro}
167:
168: After the recent breakthrough in numerical relativity
169: (NR)~\cite{FP,CLMZ,Bakeretal1}, a number of different
170: groups are now able to evolve binary black holes (BHs)
171: through merger~\cite{sperhake,gonzalez,szilagyi}. Recently, a great deal of effort has been directed
172: towards the computation of the recoil velocity of the final
173: BH~\cite{HSL,recoil,recoilJena,recoilPSU,recoilAEI,recoilGoddard,Bigrecoil,recoilFAU,recoilRI}.
174: The fundamental cause of this recoil is a net linear
175: momentum flux in the gravitational radiation, due to some asymmetry in
176: the system~\cite{Bonnor,Peres,Bekenstein,Cooperstock,Fitchett}, typically unequal
177: masses or spins in the case of BH binaries. The recoil has
178: great astrophysical importance because it
179: can affect the growth of supermassive black holes (SMBHs) in the early
180: universe~\cite{HM,Merrittetal,Volonteri07,Schnittman07b}. In those scenarios dark-matter haloes
181: grow through hierarchical mergers. The SMBHs at the centers
182: of such haloes are expected to merge unless they have been kicked out
183: of the gravitational potential well because the recoil velocity
184: gained in a prior merger is larger than the halo's escape
185: velocity.
186:
187: Other astrophysical implications include the displacement of
188: the SMBH, along with its gaseous accretion disk, forming an
189: ``off-center'' quasar~\cite{HDR}. These quasars might also have emission lines
190: highly red- or blue-shifted relative to the host galaxy due to the
191: Doppler shift of the recoil velocity~\cite{Bonning}. Additionally, these displaced
192: SMBHs could in turn displace a significant amount of stellar mass from
193: the galactic nucleus as they sink back to the center via dynamical
194: friction, forming a depleted core of missing mass on the order of
195: twice the SMBH mass~\cite{Merrittetal,Boylan_Kolchin,Lauer}.
196:
197: Numerical simulations have now been used to compute recoil velocities for
198: non-spinning unequal-mass BH binary
199: systems~\cite{HSL,recoil,recoilJena} in the range $m_2/m_1 = (1\cdots 4)$,
200: where $m_1$ and $m_2$ are the individual BH masses;
201: for spinning, non-precessing binary
202: BHs~\cite{recoilAEI,recoilPSU,recoilGoddard}, and also for
203: precessing BHs with both
204: equal~\cite{Bigrecoil,recoilFAU} as well as unequal masses~\cite{recoilRI}.
205: Quite interestingly, there exist initial spin configurations for which
206: the recoil velocity can be quite large, e.g., $\gaq \, 3000$ km/sec~
207: \cite{recoilRI,Bigrecoil,recoilFAU,recoilJena2}. However, it is not yet clear
208: whether those very large recoil velocities are astrophysically
209: likely~\cite{Schnittman04,Volonteri07,bogdanovic,loeb}. So far, due to limited
210: computational resources, the numerical simulations have explored a
211: rather small portion of the total parameter space.
212:
213: Analytic calculations, based on the post-Newtonian (PN) expansion of Einstein's
214: field equations~\cite{LB} and PN-resummation techniques~\cite{DIS98,BD1,BD2,DJS,DJS2,BCD},
215: have made predictions for the recoil velocity~\cite{AW,LK,Favataetal,BQW,DG} before the
216: NR breakthrough. Since the majority of the linear momentum flux
217: is emitted during the merger and ringdown (RD) phases, it is
218: difficult to make definitive predictions for the recoil
219: using {\it only} analytic methods. These methods need to be somehow
220: calibrated to the NR results, so that they can be accurately extended
221: during the transition from inspiral to RD.
222: So far, in the non-spinning case, the PN model~\citep{BQW}
223: has provided results consistent with NR all along the adiabatic
224: inspiral; the effective-one-body (EOB) model~\cite{BD1,DJS,DIS98}
225: can reproduce the total recoil, including the contribution
226: from the RD phase, but with large uncertainties~\cite{DG}.
227: In Ref.~\cite{Sopuerta}, perturbative calculations that make
228: use of the so-called close-limit approximation~\cite{CLA} have been
229: used to predict the recoil for unequal-mass binary BHs
230: moving on circular and eccentric orbits.
231: More recently, Ref.~\cite{SB} provided the first estimates of the
232: distribution of recoil velocities from spinning BH mergers using the
233: EOB model, calibrated to the NR results.
234:
235: In this paper we present a diagnostic of the physics of the recoil,
236: trying to understand how it accumulates during the inspiral,
237: merger, and RD phases. The majority of the analysis is based on
238: several numerical simulations of
239: non-spinning, unequal-mass binary systems, as well as spinning,
240: non-precessing binary systems obtained by the Goddard numerical
241: relativity group. What we learn in this study will be used in a forthcoming paper to
242: improve the PN analytic models~\cite{BQW,DG,SB}, so that they can be used to interpolate
243: between NR results, efficiently and accurately covering the entire
244: parameter space.
245:
246: We frame our understanding using the multipolar formalism originally
247: laid out by Thorne~\cite{KT,BD,BDS,BS,JS}.
248: We work out which multipole moments contribute most significantly
249: to the recoil. We employ analytic, but leading order, formulae
250: for the linear momentum flux during
251: the inspiral phase, and express the multipole moments in terms of
252: a linear superposition of quasi-normal modes (QNMs) during the RD
253: phase~\cite{RD}. These analysis tools help us understand why for some
254: binary mass and spin configurations
255: the so-called ``anti-kick'' is larger than in other cases. By
256: anti-kick, we mean that the recoil velocity reaches a
257: maximum value before decreasing to a final, smaller velocity
258: asymptotically. As shown in Ref.~\cite{recoilGoddard}, even a
259: relatively small range of binary parameters can give rise to a large
260: variety of anti-kick magnitudes (and even the complete lack of an
261: anti-kick in some cases).
262:
263: An example of this multipole analysis is shown in Fig.~\ref{intro_fig},
264: which plots the recoil velocity as a
265: function of time (black curve), along with the separate contributions
266: from the mass-quadrupole--mass-octupole (red), mass-quadrupole--
267: current-quadrupole (blue), and mass-quadrupole--mass-hexadecapole (green)
268: moments. This plot corresponds to a non-spinning system with mass
269: ratio of 1:2. Note in particular how the modes add both
270: constructively and destructively to give the total recoil. For the
271: non-spinning, unequal-mass systems, the kick and anti-kick are
272: dominated by the mass-quadrupole--mass-octupole modes, but also receive
273: significant contributions from the other mode-pairs. For all of the
274: simulations presented in this paper, we scale the time axis around
275: $t_{\rm peak}$, the time at which the mass quadrupole mode reaches a
276: maximum, closely corresponding to the peak in gravitational wave power,
277: as well as the time that a single horizon is formed and the ringdown
278: phase begins.
279:
280: \begin{figure}
281: \includegraphics[width=0.48\textwidth,clip=true]{fig1.ps}
282: \caption{\label{intro_fig} The recoil velocity as a function of time
283: for a binary BH system with mass ratio 1:2 and no spins. The total
284: recoil is plotted in black, along with the contributions from
285: different mode pairs, described below in Sec.~\ref{multipoles}. We denote by $t_{\rm peak}$ the time at
286: which the multipole $I^{22}$ reaches its maximum (see Section
287: \ref{multipoles}).}
288: \end{figure}
289:
290: This paper is organized as follow. In Sec.~\ref{NS}, after introducing our
291: definitions and notations, we review the binary parameters used
292: in the numerical simulations and examine the main features of the
293: numerical runs. In Sec.~\ref{multipoles} we discuss the multipolar expansion
294: of the linear momentum, angular momentum and energy fluxes
295: given in terms of the symmetric trace-free radiative mass
296: and current moments, and show how to
297: compute those fluxes from the multipole decomposition of the
298: Weyl scalar $\Psi_4$.
299: In Sec.~\ref{multipolesNR}, we analyse the multipole content of
300: the numerical waveforms during the inspiral and ringdown phases.
301: In Sec.~\ref{quasi-Newtonian} we show that, by properly normalizing
302: the binary radial separation, the multipole moments computed
303: at leading order in an expansion in $1/c$ can approximate quite well
304: the numerical results. Moreover, a superposition of three QNMs
305: matches the RD phase. In Sec.~\ref{anatomy} we apply the tools
306: developed in the previous sections to understand, using analytic
307: expressions, how the kick builds up during the inspiral,
308: merger, and ringdown phases. We also apply these methods to help explain
309: the large difference between planar and non-planar kicks from
310: equal-mass spinning BHs~\cite{recoilPSU,Bigrecoil,recoilRI}. Finally, Sec.~\ref{discussion}
311: contains a brief discussion of our main results and future
312: research directions. In Appendix ~\ref{app} we discuss recent results
313: for mass ratio $1:4$.
314:
315: \section{Numerical simulations}
316: \label{NS}
317:
318: In this section we introduce our definitions and notation, and review
319: the main features of the numerical simulations. Throughout the paper,
320: we adopt geometrical units with $G=c=1$ (unless otherwise specified)
321: and metric signature $(-1,1,1,1)$.
322:
323: \subsection{Definitions and conventions}
324: Our complex null tetrad is defined using the time-like unit vector normal
325: to a given hypersurface $\hat{\tau}$, the radial unit vector $\hat{r}$, and
326: ingoing ($\vec{\ell}$) and outgoing ($\vec{n}$) null vectors as
327: %
328: \bse
329: \begin{eqnarray}
330: \label{eq:tetrad}
331: \vec{\ell} &\equiv& \frac1{\sqrt{2}}(\hat{\tau} + \hat{r}), \\
332: \vec{n} &\equiv& \frac1{\sqrt{2}}(\hat{\tau} - \hat{r})\,.
333: \end{eqnarray}
334: \ese
335: %
336: We define the complex null vectors $\vec{m}$ and $\vec{m}^*$ by
337: \bse
338: \bea
339: \vec{m} &\equiv& \frac1{\sqrt{2}}(\hat\theta + i\hat\varphi), \\
340: \vec{m}^* &\equiv& \frac1{\sqrt{2}}(\hat\theta - i\hat\varphi),
341: \eea
342: \ese
343: with the standard spherical metric at infinity $ds^2 =
344: -d\tau^2+dr^2+r^2(d\theta^2+\sin^2\theta
345: d\varphi^2)$. The orthogonality relations of this
346: tetrad are then
347: \bse
348: \bea
349: \vec{\ell}\cdot\vec{\ell} &=& \vec{n}\cdot\vec{n} = \vec{m}\cdot\vec{m} =
350: \vec{m}^*\cdot\vec{m}^* = 0 \, , \\
351: \vec{\ell}\cdot\vec{n} &=& -\vec{m}\cdot\vec{m}^* = -1 \, , \\
352: \vec{\ell}\cdot\vec{m} &=& \vec{\ell}\cdot\vec{m}^* =
353: \vec{n}\cdot\vec{m} = \vec{n}\cdot\vec{m}^* = 0 \, .
354: \eea
355: \ese
356:
357: In terms of this tetrad, the complex Weyl scalar $\Psi_4$ is given by
358: %
359: \begin{equation}
360: \label{eq:Psi4_Weyl}
361: \Psi_4 \equiv C_{abcd}\,n^a(m^b)^*n^c(m^d)^*\,,
362: \end{equation}
363: %
364: where $C_{abcd}$ is the Weyl tensor and $*$ denotes complex conjugation.
365:
366: To relate $\Psi_4$ to the gravitational waves (GWs), we note that in
367: the transverse-traceless (TT) gauge (see Chap.~35 in Ref.~\cite{MTW}),
368: %
369: %\begin{widetext}
370: \begin{subequations}\label{eq:Riemann}
371: \begin{eqnarray}
372: \label{eq:Riemann1}
373: \frac14(\ddot{h}^{TT}_{\hat\theta\hat\theta}
374: - \ddot{h}^{TT}_{\hat\varphi\hat\varphi}) &=&
375: -R_{\hat{\tau}\hat\theta\hat{\tau}\hat\theta} =
376: -R_{\hat{\tau}\hat\varphi\hat{r}\hat\varphi} =
377: -R_{\hat{r}\hat\theta\hat{r}\hat\theta} \nonumber\\
378: &=& R_{\hat{\tau}\hat\varphi\hat{\tau}\hat\varphi} =
379: R_{\hat{\tau}\hat\theta\hat{r}\hat\theta} =
380: R_{\hat{r}\hat\varphi\hat{r}\hat\varphi}, \\
381: \label{eq:Riemann2}
382: \frac12\ddot{h}^{TT}_{\hat\theta\hat\varphi} =
383: -R_{\hat{\tau}\hat\theta\hat{\tau}\hat\varphi} &=&
384: -R_{\hat{r}\hat\theta\hat{r}\hat\varphi}=
385: R_{\hat{\tau}\hat\theta\hat{r}\hat\varphi} =
386: R_{\hat{r}\hat\theta\hat{\tau}\hat\varphi}\,.
387: \end{eqnarray}
388: \end{subequations}
389: %\end{widetext}
390: %
391: Following usual convention, we take the $h_+$ and $h_\times$ polarizations
392: of the GW to be given by
393: %
394: \bse
395: \begin{eqnarray}
396: \ddot{h}_+ &=& \frac12(\ddot{h}^{TT}_{\hat\theta\hat\theta}
397: - \ddot{h}^{TT}_{\hat\varphi\hat\varphi})\,, \\
398: \ddot{h}_\times &=& \ddot{h}^{TT}_{\hat\theta\hat\varphi}\,.
399: \end{eqnarray}
400: \ese
401: %
402: Since the Riemann and Weyl tensors coincide in vacuum regions of the
403: spacetime ($R_{abcd} = C_{abcd}$), we find by combining the above
404: equations:
405: %
406: \begin{equation}
407: \label{eq:Psi4_ddh_defn}
408: \Psi_4 = -(\ddot{h}_+ - i\ddot{h}_\times)\,.
409: \end{equation}
410: %
411: Note that this expression for $\Psi_4$ is tetrad-dependent. Here we assume
412: the tetrad given in Ref.~\cite{BCL}, Eqs.~(5.6).
413: It is also common for $\Psi_4$ to be scaled according
414: to an asymptotically Kinnersley tetrad (Ref.~\cite{BCL}, Eqs.~(5.9))
415: which introduces a factor of 2 as in Ref.~\cite{Baker:2006kr}.
416:
417: It is most convenient to deal with $\Psi_4$ in terms of its harmonic
418: decomposition. Given the definition of $\Psi_4$ in
419: Eq.~(\ref{eq:Psi4_Weyl}) and the fact that $\vec{m}^*$ carries a
420: spin-weight of $-1$, it is appropriate to decompose $\Psi_4$ in terms
421: of spin-weight $-2$ spherical harmonics $_{-2}Y_{\ell
422: m}(\theta,\varphi)$ \cite{goldberg}. There is some freedom in the definition of the
423: spin-weighted spherical harmonics. Here, we define them
424: as a linear combination of the scalar spherical
425: harmonics $Y_{\ell m}$ and $Y_{(\ell -1)m}$, as in Ref.~\cite{wiaux}:
426: %
427: \begin{widetext}
428: \begin{equation}
429: {}_{\pm2}Y_{\ell m}\left(\theta,\varphi\right)=\left[\frac{\left(\ell -2\right)!}{\left(\ell +2\right)!}\right]^{1/2}\left[\alpha_{(\ell m)}^{\pm}\left(\theta\right)Y_{\ell m}\left(\theta,\varphi\right)+\beta_{(\ell m)}^{\pm}\left(\theta\right)Y_{(\ell -1)m}\left(\theta,\varphi\right)\right]\,,
430: \end{equation}
431: for $\ell \geq2$ and $\vert m\vert\leq \ell $, and with the functional
432: coefficients
433: \bse
434: \begin{eqnarray}
435: \alpha_{(\ell m)}^{\pm}\left(\theta\right) & = & \frac{2m^{2}-\ell
436: \left(\ell +1\right)}{\sin^{2}\theta}\mp2m\left(\ell
437: -1\right)\frac{\cot\theta}{\sin\theta}+\ell \left(\ell
438: -1\right)\cot^{2}\theta\,, \\
439: \beta_{(\ell m)}^{\pm}\left(\theta\right) & = &
440: 2\left[\frac{2\ell +1}{2\ell -1}\left(\ell
441: ^{2}-m^{2}\right)\right]^{1/2}\left(\pm\frac{m}{\sin^{2}\theta}+\frac{\cot\theta}{\sin\theta}\right)\,.
442: \end{eqnarray}
443: \ese
444: \end{widetext}
445: %
446: Finally, in the far field ($r \gg M$) we decompose the dimensionless
447: Weyl scalar $M r\Psi_4$ as
448: %
449: \begin{equation}\label{eq:psi4Ylmdef}
450: M r \Psi_4(t,\vec{r}) =
451: \sum_{\ell m}{}_{-\!2}C_{\ell m}(t){}_{-\!2}Y_{\ell m}(\theta,\varphi)\,,
452: \end{equation}
453: %
454: where $M$ is the total mass of the binary
455: system (see below for explanations), and $r$ is the radial distance to the
456: binary center of mass. In Eq. (\ref{eq:psi4Ylmdef}), and
457: throughout this paper, the notation $\sum_{\ell m}$ is shorthand for
458: $\sum_{\ell=2}^{\infty}\,\sum_{m=-\ell}^{\ell}$.
459:
460: \subsection{Details of numerical simulations}
461: \label{details_NS}
462:
463: We set up the simulations by placing
464: the BHs on an initial Cauchy surface
465: using the Brandt-Br{\" u}gmann prescription \cite{Brandt97b};
466: the Hamiltonian constraint is solved using the second-order-accurate
467: multigrid solver {\tt AMRMG} \cite{Brown:2004ma}. We use the
468: Bowen-York \cite{Bowen80} framework to incorporate the BH
469: spins and momenta, with the choice
470: of initial tangential momentum informed by the quasi-circular
471: PN approximation of Ref.~\cite{DJS2}, Eq.(5.3). These initial
472: conditions typically result in a small level of orbital eccentricity,
473: which is quickly damped by the radiation reaction losses. The
474: simulations described in Ref.~\cite{recoilGoddard} showed that the
475: final recoil varied by only a few percent over a wider range of
476: initial eccentricities.
477:
478: \begin{center}
479: \begin{table*}
480: \caption{Parameters of the numerical simulations (see
481: Sec.~\ref{details_NS} for explanations). All masses are normalized
482: to an inital total mass of $M=1$.}
483: \label{table:idparams}
484: \begin{tabular}{c| c c c c c c c c c c c c c }
485: \hline \hline
486: Run & $m_{1}$ & $m_{2}$ & $\delta m$ & $q$ & $a_1/m_1$ & $a_2/m_2$
487: & $\Delta^z$ & $\Delta^p$ & $\xi^z$ & $\Sigma_{33}^z$ & $M_{\rm f}$&
488: $a_{\rm f}/M_{\rm f}$ & $v_{\rm f}$(km/s)\\
489: \hline
490: ${\rm EQ}_{+-}$ & 0.503 & 0.503 & 0.0 & 1.0 & 0.198 & -0.198
491: & -0.2 & 0.0 & 0.0 & 0.075 & 0.967 & 0.697 & 90\\
492: ${\rm EQ}_{\rm planar}$ & 0.503 & 0.503 & 0.0 & 1.0 & 0.198 & -0.198
493: & 0.0 & 0.2 & 0.0 & 0.0 & 0.967 & 0.697 & 690\\
494: ${\rm NE}^{2:3}_{00}$ & 0.401 & 0.593 & -0.192 & 0.677 & 0.0 &
495: 0.0 & 0.0 & 0.0 & 0.0 & 0.0 & 0.960 & 0.675 & 100\\
496: ${\rm NE}^{1:2}_{00}$ & 0.333 & 0.667 & -0.333 & 0.500 & 0.0 &
497: 0.0 & 0.0 & 0.0 & 0.0 & 0.0 & 0.966 & 0.633 & 140\\
498: ${\rm NE}^{1:4}_{00}$ & 0.2 & 0.8 & -0.6 & 0.250 & 0.0 &
499: 0.0 & 0.0 & 0.0 & 0.0 & 0.0 & 0.980 & 0.478 & 150\\
500: ${\rm NE}^{2:3}_{+-}$ & 0.399 & 0.610 & -0.210 & 0.655 & 0.201 &
501: -0.194 & -0.2 & 0.0 & 0.002 & 0.072 & 0.971 & 0.640 & 190\\
502: ${\rm NE}^{2:3}_{-+}$ & 0.399 & 0.610 & -0.212 & 0.653 & -0.201 &
503: 0.193 & 0.2 & 0.0 & -0.002 & -0.072 & 0.967 & 0.704 & 70\\
504: \hline \hline
505: \end{tabular}
506: \end{table*}
507: \end{center}
508:
509: The parameters for the runs considered in this paper are shown in
510: Table~\ref{table:idparams}. We use the following notation: ${\rm EQ}$
511: and ${\rm NE}$ indicate equal-mass and unequal-mass runs, respectively.
512: The subscripts $0,+,-$ refer to zero spin, spin aligned, and spin
513: anti-aligned with the orbital angular momentum, respectively (the
514: EQ$_{\rm planar}$ run has spins in the orbital plane and anti-aligned
515: with each other). For the
516: unequal-mass cases we use a superscript to indicate the mass ratio
517: $m_1:m_2$. We denote by $m_1$ the BH horizon mass computed as
518: %
519: \beq
520: m_1 = \sqrt{m_{{\rm irr},1}^2 + \frac{S_1^2}{4 m_{{\rm irr},1}^2}}\,,
521: \eeq
522: %
523: where $\mathbf{S}_1 = a_1 m_1 \mathbf{\hat{S}}_1 = S_1 \mathbf{\hat{S}}_1$ is
524: the spin angular momentum of BH 1,
525: $m_{{\rm irr},1} = \sqrt{A_1/16 \pi}$ is its irreducible mass~\cite{IRM}, and $A_1$
526: is its apparent horizon area. Similar definitions hold for BH 2.
527: The binary's total mass is $M= m_1+m_2$, $\delta m = m_1-m_2$, the mass
528: ratio is $q = m_1/m_2 \le 1$, and the symmetric mass ratio is $\eta =
529: m_1 m_2/M^2$. Following Kidder~\cite{LK}, we further define the spin
530: vectors $\mathbf{S} = \mathbf{S}_1+\mathbf{S}_2$,
531: $\mathbf{\Delta} = M (\mathbf{S}_2/m_2-\mathbf{S}_1/m_1)$, and
532: $\mathbf{\xi}=\mathbf{S}+(\delta m/M)\mathbf{\Delta}$. The spin vector
533: $\Sigma^{z}_{33}$ is defined below in Sec.~\ref{diff_mom}.
534:
535: The mass and spin parameters of the final BH are $M_{\rm f}$ and
536: $a_{\rm f}$. The values of $M_{\rm f}$ and $a_{\rm f}$ listed in
537: Table~\ref{table:idparams} are computed from the loss of energy and
538: angular momentum from the initial time to the end of the RD
539: phase. They are compatible with the values obtained by extracting the
540: fundamental QNMs (see below Sec.~\ref{RD_phase}). All spins are
541: orthogonal to the orbital plane, so
542: $\Delta^x=\Delta^y=0$ (the exception is a single run EQ$_{\rm planar}$
543: with planar spins discussed in Sec.~\ref{planarspins}. In
544: Table~\ref{table:idparams}, the spin components in the orbital plane
545: are represented by $\Delta^p \equiv
546: |\Delta^x+i\Delta^y|$.). Additionally, all runs have
547: $|a_1|/m_1=|a_2|/m_2$ with spins pointing in opposite directions, so
548: $\mathbf{\xi}\approx 0$ within the accuracy of the initial data.
549:
550: The simulations were carried out using the moving puncture
551: method \cite{CLMZ,Bakeretal1} in the finite-differencing
552: code {\tt Hahndol} \cite{Imbiriba:2004tp}, which solves the Einstein equations
553: in a standard 3+1 BSSN conformal formulation. Dissipation \cite{Huebner99}
554: terms (tapered to zero near the punctures) and constraint-damping \cite{Duez:2004uh} terms
555: were added for robust
556: stability. We used the gauge condition recommended in Ref.~\cite{vanMeter:2006vi}
557: for moving punctures, fourth-order-accurate mesh-adapted
558: differencing \cite{Baker:2005xe}
559: for the spatial derivatives, and a fourth-order-accurate
560: Runge-Kutta algorithm for the time-integration. The adaptive mesh
561: refinement and most of the parallelization was handled by the
562: software package {\tt Paramesh} \cite{MacNeice00}, with fifth-order accurate
563: interpolation between mesh refinement regions.
564:
565: The grid spacing in the finest refinement region around each
566: BH is $h_f = 3M/160$. We extract data for the radiation
567: at a radius $r_{\rm ext} = 45M$. The wave
568: extraction was performed by 4th order interpolation to a sphere followed
569: by angular integration with a Newton-Cotes formula. We have found
570: satisfactory convergence of the results. For example, for the 1:2
571: mass ratio run, for which a higher resolution of $h_f=1M/64$ was run in addition
572: to $h_f = 3M/160$, the rates of convergence of the Hamiltonian and momentum
573: constraints are comparable to those found in our equal mass runs reported
574: in \cite{Baker:2006kr}, and the radiated momenta from the two resolutions agree to within $2\%$.
575: This was also true for a 2:3 mass ratio test case with aligned spins
576: (the NE++ run in Ref.~\cite{recoilGoddard}, which is representative
577: of the NE$_{+-}^{2:3}$ and NE$_{-+}^{2:3}$ runs here).
578:
579: \section{Multipolar formalism}
580: \label{multipoles}
581:
582: In this Section we review the most relevant results from
583: Thorne~\cite{KT}, showing how a multipole decomposition of the
584: gravitational radiation field can be used to calculate the energy,
585: angular momentum, and linear momentum fluxes from a BH binary
586: system. When restricting the analysis to leading order terms
587: we shall often express the radiative multipole moments in terms of the
588: source multipole moments~\cite{BD,BDS,BS,JS}, so in much of the
589: discussion below we will use these two descriptions interchangebly.
590:
591: \subsection{Linear momentum flux}
592: \label{linear_flux}
593:
594: In the literature~\cite{HSL,recoil,recoilJena,recoilAEI,recoilRI,recoilPSU}
595: it is common to compute the linear momentum flux, and then the recoil,
596: using the following formula
597: %
598: \begin{equation}
599: \label{dPdt_psi4}
600: \frac{dP_i}{dt} = \frac{r^2}{16\pi}\int d\Omega\, \frac{x_i}{r}\,
601: \left|\int_{-\infty}^t dt \Psi_4 \right|^2\,,
602: \end{equation}
603: %
604: where $r$ is the extraction radius and the antiderivative of
605: $\Psi_4$ is used because the linear momentum flux scales as the square
606: of the first derivative of the wave strain, whereas $\Psi_4$ is
607: proportional to the second derivative of the strain [see
608: Eq.~(\ref{eq:Psi4_ddh_defn}) above]. To study how the different
609: multipole moments contribute to the recoil, we could plug
610: Eq.~(\ref{eq:psi4Ylmdef}) into Eq.~(\ref{dPdt_psi4}), as done, e.g.,
611: in Ref.~\cite{recoilPSU}. Here, we prefer to use the expression of the
612: linear momentum flux given in terms of the symmetric and trace-free
613: (STF) radiative mass and current multipole moments, as done in
614: Refs.~\cite{KT,BD,BDS,BS,JS}.
615:
616: Starting from Eq.~(4.20') in Ref.~\cite{KT}, we write the
617: linear momentum flux as
618: %
619: \begin{widetext}
620: \begin{eqnarray}\label{thorne_420}
621: F_j \equiv \frac{dP_j}{dt} &=& \frac{G}{c^7}\,\sum_{\ell=2}^{\infty}
622: \left[\frac{2(\ell+2)(\ell+3)}{\ell(\ell+1)!(2\ell+3)!!} {}^{(\ell+2)}\mathbf{I}_{jA_\ell}
623: {}^{(\ell+1)}\mathbf{I}_{A_\ell}\,\left (\frac{1}{c}\right)^{2(\ell-2)}
624: + \frac{8(\ell+3)}{(\ell+1)!(2\ell+3)!!}
625: {}^{(\ell+2)}\mathbf{S}_{jA_\ell} {}^{(\ell+1)}\mathbf{S}_{A_\ell}\,\left (\frac{1}{c}\right)^{2(\ell-1)}
626: \right. \nonumber\\
627: & &\left. \hspace{1.4cm} +\frac{8(\ell+2)}{(\ell-1)(\ell+1)!(2\ell+1)!!} \epsilon_{jpq}
628: {}^{(\ell+1)}\mathbf{I}_{pA_{\ell-1}} {}^{(\ell+1)}\mathbf{S}_{qA_{\ell-1}}\,
629: \left (\frac{1}{c}\right)^{2(\ell-2)} \right]\,,
630: \end{eqnarray}
631: \end{widetext}
632: %
633: where $\mathbf{I}_{A_\ell}$ ($\mathbf{S}_{A_\ell}$) are the
634: $\ell$-dimensional STF mass (current) tensors and
635: left-hand superscripts represent time derivatives.
636: From these tensors, we can construct the radiative multipole moments
637: ${\cal I}^{\ell m}$ and ${\cal S}^{\ell m}$ according to the normalization given by
638: Eq.~(4.7) of Ref.~\cite{KT}:
639: %
640: \bse
641: \label{thorne_47}
642: \bea
643: {\cal I}^{\ell m} &=& \frac{16\pi}{(2\ell +1)!!}\nonumber\\
644: && \cdot \left(\frac{(\ell +1)(\ell +2)}{2(\ell -1)\ell }\right)^{1/2} \mathbf{I}_{A_\ell }
645: \mathbf{Y}^{\ell m*}_{A_\ell }\,, \label{thorne_47_I} \\
646: {\cal S}^{\ell m} &=& \frac{-32\pi \ell }{(\ell +1)(2\ell +1)!!}\nonumber\\
647: && \cdot \left(\frac{(\ell +1)(\ell +2)}{2(\ell -1)\ell }\right)^{1/2} \mathbf{S}_{A_\ell }
648: \mathbf{Y}^{\ell m*}_{A_\ell } \,, \label{thorne_47_S}
649: \eea
650: \ese
651: %
652: where $\mathbf{Y}^{\ell m}_{A_\ell }$ are $\ell $-dimensional STF tensors that are
653: closely related to the usual scalar spherical harmonics by
654: %
655: \begin{equation}
656: Y_{\ell m}(\theta,\varphi) = \mathbf{Y}^{\ell m}_{i_1\cdots i_\ell } n^{i_1} \cdots
657: n^{i_\ell }\,,
658: \end{equation}
659: %
660: with $n^i = (\sin\theta \cos\varphi,\sin\theta
661: \sin\varphi,\cos\theta)^i$. Note that the radiative moments ${\cal
662: I}^{\ell m}$ and ${\cal S}^{\ell m}$ are scalar quantities and have no
663: explicit spatial dependence.
664: To simplify the notation below, we incorporate the ($\ell +1$) time
665: derivatives into the radiative multipole moments, and define
666: %
667: \beq
668: \label{new_tensors}
669: I^{\ell m} \equiv {}^{(\ell +1)}{\cal I}^{\ell m}\,,
670: \quad S^{\ell m} \equiv {}^{(\ell +1)}{\cal S}^{\ell m}\,.
671: \eeq
672: %
673:
674: By combining Eqs.~(\ref{thorne_420}), (\ref{thorne_47}), and
675: (\ref{new_tensors}), we find that at leading order (in a $1/c$
676: expansion) the linear momentum flux is given by
677: \begin{widetext}
678: \bea
679: \label{dPdt_radmoments}
680: F^{(0)}_x+iF^{(0)}_y =
681: \frac{1}{336\pi}\hspace{-0.3cm}&&\left[ -14iS^{21}I^{22*}
682: + \sqrt{14}I^{31}I^{22*} -\sqrt{210}I^{22}I^{33*} +
683: 7i\sqrt{6}I^{20}S^{21*}-7i\sqrt{6}S^{20}I^{21*}
684: + \right . \nonumber \\
685: && \left. 14iI^{21}S^{22*} +\sqrt{42}I^{30}I^{21*} - 2\sqrt{21}I^{20}I^{31*} -
686: 2\sqrt{35}I^{21}I^{32*} \right]\,,
687: \eea
688: and
689: \bea
690: \label{dPdtz_radmoments}
691: F^{(0)}_z=
692: \frac{1}{336\pi}\left[4\sqrt{14}\Re(I^{31}I^{21*})-14\Im(I^{21}S^{21*})
693: +2\sqrt{35}\Re(I^{22}I^{32*})-28\Im(I^{22}S^{22*})
694: +3\sqrt{7}I^{20}I^{30}\right]\,.
695: \eea
696: \end{widetext}
697: %
698: Note that Eq.~(\ref{dPdt_radmoments}) coincides with Eq.~(9) in
699: Ref.~\cite{DG} when we equate the radiative multipole moments with
700: the source moments~\cite{BD,BDS,BS,JS} and reduce
701: to a circular, non-spinning orbit in the $x$-$y$ plane. In this case
702: only the first three terms in Eq.~(\ref{dPdt_radmoments}) survive.
703:
704: The next highest order terms ($1/c^2$ with respect to the leading terms) are proportional to the
705: mass octupoles $I^{3m}$, or current quadrupoles $S^{2m}$:
706: \begin{widetext}
707: \bea
708: \label{dPdt_radmoments2}
709: F^{(1)}_x+iF^{(1)}_y =
710: \frac{1}{672\pi} \hspace{-0.3cm}&& \left[ -7i\sqrt{6}S^{32}I^{33*}
711: -14\sqrt{6}I^{33}I^{44*}
712: -4\sqrt{21}S^{20}S^{31*}-4\sqrt{35}S^{21}S^{32*}-2\sqrt{210}S^{22}S^{33*}
713: + \right .\nonumber \\
714: && \left. 2\sqrt{42}S^{30}S^{21*} + 14i\sqrt{3}I^{30}S^{31*}-14i\sqrt{3}S^{30}I^{31*}
715: +7i\sqrt{10}I^{31}S^{32*} -7i\sqrt{10}S^{31}I^{32*} -\right. \nonumber \\
716: && \left . 2\sqrt{105}I^{30}I^{41*}+6\sqrt{7}I^{40}I^{31*}
717: -3\sqrt{70}I^{31}I^{42*} +3\sqrt{14}I^{41}I^{32*} -21\sqrt{2}I^{32}I^{43*} +\right . \nonumber \\
718: && \left . 2\sqrt{14}S^{31}S^{22*}+ \sqrt{42}I^{42}I^{33*} +7i\sqrt{6}I^{32}S^{33*} \right ]\,,
719: \eea
720: and
721: \bea
722: \label{dPdtz_radmoments2}
723: F^{(1)}_z=
724: \frac{1}{336\pi} \hspace{-0.3cm}&& \left[
725: 3\sqrt{7}S^{20}S^{30}
726: +4\sqrt{14}\Re(S^{21}S^{31*})+2\sqrt{35}\Re(S^{22}S^{32*})
727: -7\Im(I^{31}S^{31*})-14\Im(I^{32}S^{32*})-21\Im(I^{33}S^{33*})
728: +\right . \nonumber \\
729: && \left . 2\sqrt{21}I^{30}I^{40}+3\sqrt{35}\Re(I^{31}I^{41*})
730: +6\sqrt{7}\Re(I^{32}I^{42*})+7\sqrt{3}\Re(I^{33}I^{43*}) \right ] \,.
731: \eea
732: \end{widetext}
733: Note that all of the terms in Eqs.~(\ref{dPdt_radmoments}) and
734: (\ref{dPdt_radmoments2}) contain products of multipoles with
735: $m'=m\pm1$, while the terms in Eqs.~(\ref{dPdtz_radmoments}) and
736: (\ref{dPdtz_radmoments2}) have $m'=m$, as with familiar
737: quantum-mechanical operators that involve similar $x_i$-weighted
738: integrations over the sphere. Also note that for mass-mass and
739: current-current terms, $\ell'=\ell\pm 1$, while for mass-current
740: terms, $\ell'=\ell$.
741:
742: The above formulae (\ref{dPdt_radmoments})--(\ref{dPdtz_radmoments2})
743: are valid for completely general orbits, including eccentricity,
744: spin terms and even for binary systems precessing out of the plane. However,
745: we can simplify them significantly by rotating into the frame where
746: the instantaneous orbital angular momentum is along the $z$-axis.
747: Furthermore, by
748: assuming that terms proportional to $\ddot{R}$ ($R$ being the binary radial
749: separation) are negligible,
750: we find $I^{20}=I^{30}=S^{30}=I^{32}=I^{40}=I^{41}=I^{43}=0$. In the
751: approximation of $\ddot{R}=0$, the inclusion of terms linear in
752: $\dot{R}\ne 0$ adds no new multipole modes.
753: In fact, one of the primary reasons the derivations above begin with
754: the mass and current tensors $\mathbf{A}_{A_\ell}$ and
755: $\mathbf{S}_{A_\ell}$ is to facilitate the calculation of the
756: individual radiative moments $I^{\ell m}$ and $S^{\ell m}$ and also
757: identify the contributions from $\dot{R}$ and $\ddot{R}$ terms from a
758: generalized binary orbit~\cite{LK}.
759: In the case of non-spinning BHs, the formulae
760: (\ref{dPdt_radmoments})--(\ref{dPdtz_radmoments2}) can be additionally
761: simplified by setting $S^{20}=I^{21}=S^{22}=S^{31}=S^{33}=0$.
762: Quite interestingly, we obtain that the latter conditions
763: are also valid in the special case of {\it non-precessing} BHs
764: where the spins are aligned or anti-aligned with the orbital
765: angular momentum. Since these are the cases we consider in this paper,
766: we refer often to the following approximate formula for the
767: linear momentum flux:
768: %
769: \begin{widetext}
770: \beq
771: \label{flux_approx}
772: F_x+iF_y \simeq
773: \frac{1}{672\pi} \left[
774: -28iS^{21}I^{22*} -2\sqrt{210}I^{22}I^{33*}
775: -14\sqrt{6}I^{33}I^{44*}+2\sqrt{14}I^{31}I^{22*} -
776: 7i\sqrt{6}S^{32}I^{33*}
777: \right]\,, \quad F_z = 0.
778: \eeq
779: \end{widetext}
780: %
781: As we will see below in Sec.~\ref{inspiral_phase}, the linear momentum
782: flux contributions from $I^{31}I^{22*}$ as well as other higher-$\ell$
783: modes are typically smaller by at least
784: an order of magnitude. When integrating Eq.~(\ref{flux_approx}) to get
785: the recoil velocity, we also find that (due in large part to the
786: relative phases between the modes) the contribution from
787: $S^{32}I^{33*}$ is rather minimal. Thus for most of the analysis that
788: follows, we will focus solely on the first three terms of
789: Eq.~(\ref{flux_approx}).
790:
791: In the following, sometimes we will use
792: %
793: \beq
794: \mathbf{F} = \{F_x,F_y,F_z \}\,, \quad \quad \mathbf{\hat{F}} = \frac{\mathbf{F}}{|\mathbf{F}|}\,.
795: \eeq
796: %
797: All the non-precessing numerical simulations we will analyze have
798: $F_{z}=0$, so we can introduce a complex scalar flux
799: \beq
800: F \equiv F_x + i F_y\,.
801: \eeq
802: %
803:
804: Since what we extract from the numerical simulations are the
805: modes ${}_{-\!2}C_{\ell m}$ computed over the sphere surrounding the
806: binary, we need to relate the ${}_{-\!2}C_{\ell m}$ to the
807: radiative mass and current multipole moments defined above.
808: From Eq.(4.3) of \cite{KT},
809: %
810: \begin{equation}
811: \label{eq:hmm}
812: h=\sum_{\ell m} ({}^{(\ell )}{\cal I}^{\ell m}T_{ab}^{E2,\ell
813: m}m^a m^b + {}^{(\ell )}{\cal S}^{\ell m}T_{ab}^{B2,\ell m}m^a m^b)\, ,
814: \end{equation}
815: %
816: where $h\equiv h_{ab} m^a m^b$ and $h_{ab}$ is the metric perturbation
817: $g_{ab}-\eta_{ab}$ in the transverse traceless gauge, which satisfies
818: Eq.~(\ref{eq:Riemann}), and $T_{ab}^{E2,\ell m}$ and
819: $T_{ab}^{B2,\ell m}$ are the ``pure-spin'' harmonics of Thorne.
820: From Appendix A of \cite{martel05},
821: %
822: \bse
823: \begin{eqnarray}
824: \label{eq:TabE}
825: T_{ab}^{E2,\ell m}&=
826: &\frac{1}{\sqrt2}(\,{}_{-2}Y^{\ell m} m_a m_b +\,{}_{2}Y^{\ell m} m_a^*
827: m_b^*)\\
828: \label{eq:TabB}
829: T_{ab}^{B2,\ell m}&=
830: &\frac{-i}{\sqrt2}(\,{}_{-2}Y^{\ell m} m_a m_b -\,{}_{2}Y^{\ell m} m_a^*
831: m_b^*).
832: \end{eqnarray}
833: \ese
834: Substituting Eqs.~(\ref{eq:TabE})--(\ref{eq:TabB}) into Eq.~(\ref{eq:hmm}) and recalling that $m^a m_a=0$ gives
835: \begin{equation}
836: h=\frac{1}{\sqrt{2}r}\sum_{\ell m}({}^{(\ell )}{\cal I}^{\ell m}+i {}^{(\ell )}{\cal
837: S}^{\ell m})\,{}_{+2}Y^{\ell m}
838: \end{equation}
839: Now taking the complex conjugate and using the fact that
840: $_{+2}Y^{*\ell m}=(-1)^m{}_{-2}Y^{\ell -m}$ [note there is a typo in Eq.~(3.1)
841: of Ref.~\cite{goldberg}]
842: we obtain
843: \begin{eqnarray}
844: h^* &=&\frac{1}{\sqrt{2}r}\sum_{\ell m}(-1)^m({}^{(\ell )}{\cal I}^{\ell m*}-i {}^{(\ell )}{\cal S}^{\ell m*})\,{}_{-2}Y^{\ell -m} \nonumber\\
845: &=&\frac{1}{\sqrt{2}r}\sum_{\ell m}(-1)^m({}^{(\ell )}{\cal
846: I}^{\ell -m*}-i {}^{(\ell )}{\cal S}^{\ell
847: -m*})\,{}_{-2}Y^{\ell m}\,. \nonumber \\
848: \end{eqnarray}
849: %
850: Using the tetrad choice of
851: Eqs.~(\ref{eq:tetrad})--(\ref{eq:Psi4_ddh_defn}), $\partial^2
852: h^*/\partial t^2 =\ddot{h}_+-i\ddot{h}_\times=-\Psi_4$,
853: which decomposed into spin -2 weighted harmonics, gives
854: %
855: \begin{equation}
856: \frac{\partial^2 h^*}{\partial t^2}=-\frac{1}{Mr}\sum_{\ell m}{}_{-2}C_{\ell m} \,{}_{-2}Y^{\ell m},
857: \end{equation}
858: %
859: allowing us to see term-by-term that
860: %
861: \begin{equation}
862: (-1)^m({}^{(\ell +2)}{\cal I}^{\ell -m*}-i {}^{(\ell +2)}{\cal
863: S}^{\ell -m*})=-\sqrt{2} {}_{-2}C_{\ell m} \, .
864: \end{equation}
865: %
866: Recall that $(-1)^m{\cal I}^{\ell -m*} = {\cal I}^{\ell m}$ and $(-1)^m{\cal S}^{\ell -m*} =
867: {\cal S}^{\ell m}$, which allows us to write
868: %
869: \begin{subequations}
870: \begin{eqnarray}
871: \label{Ilm_Clm}
872: {}^{(\ell +2)}{\cal I}^{\ell m} &=& -\frac1{\sqrt2}\left[{}_{-2}C_{\ell m}+\,
873: (-1)^m{}_{-2}C_{\ell-m}^*\right]\,,\\
874: \label{Slm_Clm}
875: {}^{(\ell +2)}{\cal S}^{\ell m} &=& -\frac{i}{\sqrt2}\left[{}_{-2}C_{\ell m}-
876: \,(-1)^m{}_{-2}C_{\ell-m}^*\right]\,.
877: \end{eqnarray}
878: \end{subequations}
879: %
880: Equations (\ref{dPdt_radmoments})--(\ref{flux_approx}) are expressed in terms
881: of $I^{\ell m} \equiv {}^{(\ell +1)}{\cal I}^{\ell m}$ and $S^{\ell m} \equiv
882: {}^{(\ell +1)}{\cal S}^{\ell m}$, which can be computed by integrating
883: Eqs.~(\ref{Ilm_Clm}), (\ref{Slm_Clm}) once in time. To avoid the complication of an
884: undetermined constant of integration, we typically integrate
885: ${}_{-2}C_{\ell m}(t)$ {\it backwards} in time, since in the numerical
886: data (and what we expect happens in reality) all the moments go
887: to zero exponentially after the merger. At early times on the other
888: hand, most of the modes are significantly non-zero and also include
889: a large amount of numerical noise due to the initial conditions.
890:
891: \subsection{Energy and angular momentum flux}
892:
893: Unlike the equations for the linear momentum flux, which all involve
894: ``beating'' between pairs of different modes, the energy and angular-momentum
895: flux expressions involve terms of the form $|I^{\ell m}|^2$,
896: allowing us to isolate the individual contributions
897: from each mode. As
898: we will see below, for the comparable-mass binary systems that
899: we analyse ($m_1$:$m_2$ = 1:1, 2:3, 1:2), the amplitude of the
900: mass quadrupole moment $I^{22}$ is roughly an order of magnitude
901: larger than the next largest mode. Thus it almost completely dominates
902: the energy and angular momentum fluxes, and we can write [see Eq.~(4.16)
903: in Ref.~\cite{KT}]
904: %
905: \begin{equation}\label{dEdt}
906: \frac{dE}{dt} =\frac{1}{32\pi}\sum_{\ell m}
907: \left( |I^{\ell m}|^2 + |S^{\ell m}|^2 \right) \simeq \frac{1}{16\pi}|I^{22}|^2.
908: \end{equation}
909: %
910:
911: The multipole expressions for angular momentum flux are somewhat more
912: complicated, but for the numerical
913: simulations considered in this paper, the only non-zero modes have
914: $\ell +m$ even for $I^{\ell m}$ and $\ell +m$ odd for $S^{\ell m}$, so we can neglect
915: the $(m,m\pm 1)$ cross-terms in Eq.~(4.23) of Ref.~\cite{KT}. These
916: cross-terms are responsible for angular momentum loss in the $x$-$y$
917: plane, so it is reasonable that they must be zero for non-precessing
918: planar orbits. In this case, where the angular momentum is solely
919: along the $\hat{\mathbf{z}}$-axis, we have
920: %
921: \begin{eqnarray}\label{dJzdt}
922: \frac{dJ_z}{dt} &=& \frac{i}{32\pi}\sum_{\ell m}
923: m( {}^{(\ell )}{\cal I}^{\ell m*}\, {}^{(\ell +1)}{\cal I}^{\ell m} +
924: {}^{(\ell )}{\cal S}^{\ell m*}\, {}^{(\ell +1)}{\cal S}^{\ell m} ) \nonumber\\
925: & \simeq & -\frac{1}{8\pi}\, \Im\left[{}^{(2)}{\cal I}^{22*}\,
926: {}^{(3)}{\cal I}^{22} \right]\, ,
927: \end{eqnarray}
928: where we have restored the explicit time
929: derivatives as in Eq.~(\ref{new_tensors}).
930:
931: %\begin{widetext}
932: \begin{center}
933: \begin{table*}
934: \caption{Energy and angular momentum radiated in each of the dominant
935: multipole modes. In parentheses we show the amount radiated only
936: after the peak of GW energy flux. All units are normalized to $M=1$.}
937: \label{table:Elm}
938: \begin{tabular}{c| c c c c c c c c c c }
939: \hline \hline
940: Run & $E_{22}$ & $E_{21}$ & $E_{32}$ & $E_{33}$ & $E_{44}$ &
941: $J_{22}$ & $J_{21}$ & $J_{32}$ & $J_{33}$ & $J_{44}$ \\
942: & $(\times 10^{-2})$ & $(\times 10^{-4})$ & $(\times 10^{-4})$ & $(\times
943: 10^{-4})$ & $(\times 10^{-4})$ & $(\times 10^{-1})$ & $(\times 10^{-4})$ &
944: $(\times 10^{-4})$ & $(\times 10^{-3})$ & $(\times 10^{-3})$ \\
945:
946: \hline
947: ${\rm EQ}_{+-}$ & 3.5 & $0.22$
948: & $1.6$ & $0.04$ & $3.3$ & $2.2$ & $-0.70$ & $7.9$ & $-0.02$ & $1.9$ \\
949: &($1.4$)&($0.17$)&($1.2$)&($0.02$)&($1.5$) &($0.50$)&($-0.46$)&($-2.0$)
950: &($-0.01$)&($0.64$)
951: \smallskip\\
952: ${\rm NE}^{2:3}_{00}$ & 3.1 & $0.61$
953: & $0.90$ & $5.6$ & $2.9$
954: & $2.2$ & $-2.1$ & $3.9$
955: & $-3.1$ & $1.8$ \\
956: &($1.1$) &($0.40$)&($0.66$)
957: &($2.8$)&($1.0$)
958: &($0.45$) &($-0.98$)&($2.5$)
959: &($-1.1$)&($0.46$)
960: \smallskip\\
961: ${\rm NE}^{1:2}_{00}$ & 2.5 & $1.4$
962: & $0.47$ & $12.0$ & $2.7$
963: & $1.8$ & $-4.8$ & $2.4$
964: & $-6.9$ & $1.7$ \\
965: &($0.87$)&($0.94$)&($0.30$)
966: &($5.8$)&($0.73$)
967: &($0.37$) &($-2.4$)&($1.3$)
968: &($-2.3$)&($0.30$)
969: \smallskip\\
970: ${\rm NE}^{1:4}_{00}$ & 1.2 & $2.1$
971: & $0.27$ & $16.0$ & $3.3$
972: & $1.2$ & $-8.0$ & $1.6$
973: & $-11.0$ & $2.4$ \\
974: &($0.35$)&($1.4$)&($0.09$)
975: &($6.6$)&($1.2$)
976: &($0.16$) &($-3.8$)&($0.27$)
977: &($-2.9$)&($0.48$)
978: \smallskip\\
979: ${\rm NE}^{2:3}_{+-}$ & 2.9 & $1.6$
980: & $0.93$ & $5.2$ & $2.6$
981: & $2.0$ & $-5.4$ & $2.1$
982: & $-2.9$ & $1.6$ \\
983: &($1.0$) &($1.0$)&($0.67$)
984: &($2.5$)&($0.82$)
985: &($0.31$) &($-2.9$)&($5.3$)
986: &($-0.98$)&($0.33$)
987: \smallskip\\
988: ${\rm NE}^{2:3}_{-+}$ & 3.3 & $0.14$
989: & $1.1$ & $7.1$ & $2.9$
990: & $2.3$ & $-0.50$ & $4.4$
991: & $-3.9$ & $1.8$ \\
992: &($1.1$) &($0.09$)&($0.78$)
993: &($3.4$)&($0.92$)
994: &($0.44$) &($-0.21$)&($3.1$)
995: &($-1.3$)&($0.37$) \\
996: \hline \hline
997: \end{tabular}
998: \end{table*}
999: \end{center}
1000: %\end{widetext}
1001:
1002: Integrating Eqs.~(\ref{dEdt}) and (\ref{dJzdt}) term-by-term, we can
1003: calculate how much energy and angular momentum are radiated in each of
1004: the dominant modes, similar to the approach of Ref.~\cite{berti07}. We
1005: introduce the quantities $E_{\ell m}$ and $J_{\ell m}$ as the total energy
1006: and angular momentum radiated in each ($\ell$, $m$) mode, computed by
1007: integrating Eqs.~(\ref{dEdt}) and (\ref{dJzdt}) in time, term by
1008: term (for conciseness, we combine both the $m$ and $-m$ terms into
1009: $E_{\ell m}$ and $J_{\ell m}$ and restrict our notation to $m>0$). Note
1010: that while $E_{\ell m}$ is always positive, $J_{\ell m}$ can also be
1011: negative, corresponding to angular momentum in the $-\hat{z}$
1012: direction. These results are shown in Table \ref{table:Elm}, along
1013: with the contributions from just
1014: the RD phase ($t>t_{\rm peak}$, where $t_{\rm peak}$ is the point at which
1015: $|I^{22}|$ reaches its peak, closely corresponding to the peak in GW
1016: energy emission). We will see below in Section
1017: \ref{quasi-Newtonian} that these various energy contributions agree
1018: closely with the Newtonian predictions for the relative
1019: mass-scalings. For example, the energy $E_{22}$ in the inspiral phase
1020: should scale as $\eta$, while the RD contribution should scale like
1021: $\eta^2$. It is important to note that the different moments have
1022: different scalings: $E_{33} \sim \eta^2 \delta m^2$, while the $I^{44}$
1023: contribution has a much weaker dependence on mass ratio: $E_{44} \sim
1024: \eta^2(1-3\eta)^2$.
1025:
1026: In the limit of very large initial
1027: separation (small initial frequency), each of the $E_{\ell m}$ and
1028: $J_{\ell m}$
1029: should converge to a finite value, with the notable exception of
1030: $J_{22}$. It is well-know that the angular momentum of a binary system
1031: scales as $R^{1/2}$, and is thus unbound in the limit of $R\to\infty$,
1032: but it is interesting to see that the higher-order contributions to
1033: the angular momentum all converge at large $R$. This can be understood
1034: directly from Eq.~(\ref{dJzdt}) in the Keplerian limit of
1035: $R=M^{1/3}\omega^{-2/3}$. At leading order, radiation reaction follows the
1036: relation $dt \sim \omega^{-11/3}d\omega$ so the angular momentum in
1037: the inspiral is
1038: %
1039: \bea\label{J22insp}
1040: J_{22} &=& \frac{1}{8\pi}\int_{t=-\infty}^{t_0} dt\,
1041: \Im\left[{}^{(2)}{\cal I}^{22*}\,
1042: {}^{(3)}{\cal I}^{22}\right] \nonumber\\
1043: &\sim& \int_{\omega=0}^{\omega_0} \omega^{2/3} \omega^{5/3}
1044: \omega^{-11/3} d\omega \to \infty.
1045: \eea
1046: %
1047: As we will see below in Section \ref{quasi-Newtonian},
1048: for all the other energy and angular momentum modes, the fluxes from
1049: Eqs.~(\ref{dEdt}),(\ref{dJzdt}) scale as $\omega^{10/3}$ or higher powers,
1050: and thus converge when integrated over $\omega^{-11/3}d\omega$.
1051:
1052: \section{Multipole analysis of the numerical simulations}
1053: \label{multipolesNR}
1054:
1055: In this Section we want to investigate how the different multipole moments
1056: evolve during the inspiral and ringdown phases of BH binary mergers.
1057:
1058: \subsection{Inspiral phase}
1059: \label{inspiral_phase}
1060:
1061: As can be derived in PN theory~\cite{LB} and has been confirmed numerically
1062: in Refs.~\cite{CLMZ, Bakeretal1}, the $\ell=2, m=2$
1063: mode in Eq.~(\ref{eq:psi4Ylmdef}) is circularly polarized to leading order
1064: throughout the coalescence.
1065: Because of this, Ref.~\cite{BCP} defined the
1066: (dominant) orbital angular frequency as
1067: %
1068: \begin{equation}
1069: \label{dom}
1070: \omega_{\rm D}^{\ell m} = -\frac{1}{m} \Im
1071: \left(\frac{{}_{-2}\dot{C}_{\ell m}}{{}_{-2}C_{\ell m}}\right).
1072: \end{equation}
1073: %
1074: Here, we extend Eq.~(\ref{dom}) by defining several
1075: (dominant) orbital angular frequencies, each of them
1076: being related to a specific multipole moment, $I^{\ell m}$ or $S^{\ell m}$,
1077: as
1078: %
1079: \begin{equation}\label{omega_lm}
1080: \omega^{I\ell m}_{\rm D} = -\frac{1}{m} \Im
1081: \left(\frac{\dot{I}^{\ell m}}{I^{\ell m}}\right)\,, \quad
1082: \omega^{S\ell m}_{\rm D} = -\frac{1}{m} \Im
1083: \left(\frac{\dot{S}^{\ell m}}{S^{\ell m}}\right).
1084: \end{equation}
1085: %
1086:
1087: \begin{figure*}
1088: \includegraphics[width=0.48\textwidth,clip=true]{fig2a.ps}
1089: \includegraphics[width=0.48\textwidth,clip=true]{fig2b.ps}
1090: \caption{\label{omega_modes} Dominant orbital angular frequency
1091: obtained from the individual radiative multipole moments, as determined by Eq.~
1092: (\ref{omega_lm}). The different frequencies with $\ell=m$ agree closely
1093: throughout the inspiral and RD phases.
1094: The frequency with $\ell=2,m=1$ decouples from the others at earlier time
1095: and reaches a much higher plateau. The left panel refers to the NE$_{00}^{2:3}$ run
1096: and the right panel to the NE$_{00}^{1:2}$ run. We denote with $t_{\rm peak}$ the
1097: time at which $I^{22}$ reaches its maximum.}
1098: \end{figure*}
1099:
1100: \begin{figure*}
1101: \includegraphics[width=0.48\textwidth,clip=true]{fig3a.ps}
1102: \includegraphics[width=0.48\textwidth,clip=true]{fig3b.ps}
1103: \caption{\label{abs_lm} Amplitudes of the dominant radiative
1104: multipole moments.
1105: On the left panel we show the modes for the NE$^{2:3}_{00}$ run,
1106: while on the right panel the modes for the NE$^{1:2}_{00}$ run.
1107: The leading-order mass quadrupole $I^{22}$ is about an
1108: order of magnitude stronger than any other mode. The oscillating
1109: behavior of the $S^{32}$ moment during RD is likely due to mode
1110: mixing with $I^{22}$. We denote with $t_{\rm peak}$ the
1111: time at which $I^{22}$ reaches its maximum.}
1112: \end{figure*}
1113:
1114: We plot these frequencies in Fig.~\ref{omega_modes} for the dominant
1115: multipole moments $I^{22}$, $S^{21}$, $I^{33}$, $I^{44}$, and
1116: $S^{32}$, for the NE$_{00}^{2:3}$ (left panel) and NE$_{00}^{1:2}$
1117: (right panel) runs. The amplitudes of the $I^{31}$ and $I^{42}$ modes are too
1118: weak and dominated by noise to extract a dominant frequency. In this
1119: figure, as well as most shown in the rest of the paper, we plot the
1120: time variable with respect to $t_{\rm peak}$. We notice
1121: that the frequencies corresponding to the modes with $\ell=m$ agree quite
1122: well throughout the inspiral and ringdown, but the frequency of the $S^{21}$ mode
1123: decouples from the others approximately $50 M$ before the peak in the
1124: $I^{22}$ mode. As we shall see in Sec.~\ref{anatomy}, this is due to
1125: the fact that, during the ringdown phase, the dominant angular frequency
1126: associated to the $S^{21}$ mode is almost twice as large as those of
1127: the other leading modes~\cite{L85,E89,BCW}. This decoupling plays a
1128: major role in determining the shape of the kick and anti-kick (see
1129: Sec.~\ref{anatomy} below), and also suggests that the
1130: transition to RD may begin long before the peak of the GW
1131: flux. Similarly, the $S^{32}$ mode should converge to a higher RD
1132: frequency ($\omega_{320}/2 \simeq 0.37/M_{\rm f}$ for these runs), but
1133: may be limited by numerical noise here, as well as possible mode
1134: mixing with the dominant $I^{22}$ moment.
1135:
1136: \begin{figure*}
1137: \includegraphics[width=0.48\textwidth,clip=true]{fig4a.ps}
1138: \includegraphics[width=0.48\textwidth,clip=true]{fig4b.ps}
1139: \caption{\label{flux_lm} Linear momentum flux of the strongest radiative
1140: multipole moments, i.e., the ones in Eq.~(\ref{flux_approx}).
1141: On the left panel we show the modes for the NE$^{2:3}_{00}$ run,
1142: while on the right panel the modes for the NE$^{1:2}_{00}$ run.
1143: We denote with $t_{\rm peak}$ the time at which $I^{22}$ reaches its
1144: maximum.}
1145: \end{figure*}
1146:
1147: In Fig.~\ref{abs_lm} we show the amplitudes of the multipole moments
1148: in Eq.~(\ref{flux_approx}). Again, the left panel refers
1149: to the NE$^{2:3}_{00}$ run, while the right panel to the
1150: NE$^{1:2}_{00}$ run. The mass-quadrupole moment $I^{22}$ clearly
1151: dominates in both cases, while the $I^{31}$ and $I^{42}$ modes are so
1152: weak as to be almost completely overwhelmed by numerical noise.
1153: In addition to having dissimilar amplitudes, the different moments
1154: also peak at slightly different times, which
1155: may be related to the fact that RD modes are excited at different times.
1156: In particular, the modes mentioned above with $\ell \ne m$ tend to peak later
1157: in time, perhaps due to a longer transition to the higher QNM frequency.
1158: As we shall see in Sec.~\ref{quasi-Newtonian}, as the mass ratio becomes more
1159: extreme (i.e., decreasing $\eta$), the higher-order modes increase in relative amplitude, with
1160: $I^{33}$ and $S^{21}$ both proportional to $\eta\, \delta m$. $I^{44}$ and
1161: $S^{32}$, however, scale as $\eta(1-3\eta)$, so they increase
1162: only slightly in the range of masses considered here.
1163:
1164: Next, in Fig.~\ref{flux_lm}, we show the amplitude of the linear
1165: momentum flux from the mode-pairs included in
1166: Eq.~(\ref{flux_approx}). Here we define the complex flux
1167: $F^{21,22}=(-14i/336\pi)S^{21}I^{22*}$ and other $F^{\ell m,\ell' m'}$
1168: analogously from Eq.~(\ref{flux_approx}). As in Fig.~\ref{abs_lm}, the
1169: mass-quadrupole terms dominate,
1170: with significantly smaller contributions from the $S^{32}$ and
1171: $I^{31}$ modes. However, note the appreciable flux amplitude from the
1172: $F^{33,44}\sim I^{33}I^{44*}$
1173: term, which is formally a higher-order correction in a ($1/c$) expansion~\cite{BQW,DG}.
1174: From Fig.~\ref{flux_lm}, we expect that the first three pairs of modes in
1175: Eq.~(\ref{flux_approx}) should contribute most significantly to the
1176: recoil. Including the complex phase relations between the different
1177: modes, we find this result will be supported further by the analysis in
1178: Sec.~\ref{diff_mom}.
1179:
1180: \subsection{Ringdown phase}
1181: \label{RD_phase}
1182:
1183: We now extract the QNMs, notably the fundamental
1184: and the first two overtones,
1185: present in the most significant multipole moments during the RD phase.
1186: We follow the procedure outlined in Ref.~\cite{BCP}. To avoid possible
1187: constant offsets introduced by integrating
1188: Eqs.~(\ref{Ilm_Clm}), (\ref{Slm_Clm}), we prefer to extract the QNMs
1189: directly from the ${}_{-2}C_{\ell m}$ instead of using $I_{\ell m}$ or
1190: $S_{\ell m}$. Additionally, from
1191: Eqs.~(\ref{Ilm_Clm}), (\ref{Slm_Clm}), we see that ${}^{(1)}I^{\ell m}$ and
1192: ${}^{(1)}S^{\ell m}$ are made up of both ${}_{-2}C_{\ell m}$ {\it and}
1193: ${}_{-2}C_{\ell -m}$, which in general do not have the same QNM
1194: frequencies, so it is more reliable to extract the RD modes from just
1195: ${}_{-2}C_{\ell m}$ (however, in practice we find that the RD phase is
1196: dominated by modes with positive $m$). Following the approach of
1197: Ref.~\cite{BCW}, we define the
1198: complex frequencies $\sigma_{\ell mn}$:
1199: \beq
1200: \sigma_{\ell mn} \equiv \omega_{\ell mn} - i/\tau_{\ell mn},
1201: \eeq
1202: and each RD mode is proportional to $\exp(-i\sigma_{\ell mn}t)$. In this
1203: notation, $\omega_{\ell mn}$ are the QNM oscillation frequencies [not to be
1204: confused with the dominant frequencies of Eq.~(\ref{omega_lm})] and
1205: $\tau_{\ell mn}$ are the mode decay times, all functions of the final black
1206: hole mass and spin. The subscripts $\ell $ and $m$ are the same
1207: spherical wavenumbers used above, and $n=0$ denotes the fundamental
1208: mode, with $n=1,2,\cdots$, corresponding to the higher overtones. The
1209: fundamental QNM frequencies $\sigma_{\ell m0}$ are listed in
1210: Table~\ref{table:QNM_freqs} for the NR runs listed above. All
1211: frequencies and decay times are measured in units of the final mass
1212: $M_{\rm f}$.
1213:
1214: \begin{center}
1215: \begin{table*}
1216: \caption{Frequencies and decay times for the fundamental QNMs for
1217: each of the numerical simulations. $\omega_{\ell m 0}$ is in units of
1218: $M_{\rm f}^{-1}$ and $\tau_{\ell m 0}$ is in units of $M_{\rm f}$.}
1219: \label{table:QNM_freqs}
1220: \begin{tabular}{c| c r r r r r r r r r r }
1221: \hline \hline
1222: Run & $a_{\rm f}/M_{\rm f}$
1223: & \hspace{0.4cm} $\omega_{210}$ & $\tau_{210}$
1224: & \hspace{0.4cm} $\omega_{220}$ & $\tau_{220}$
1225: & \hspace{0.4cm} $\omega_{320}$ & $\tau_{320}$
1226: & \hspace{0.4cm} $\omega_{330}$ & $\tau_{330}$
1227: & \hspace{0.3cm} $\omega_{440}$ & $\tau_{440}$ \\
1228: \hline
1229: ${\rm EQ}_{+-}$ & 0.697 & 0.454 & 12.2 & 0.531 & 12.4
1230: & 0.758 & 11.9 & 0.841 & 12.0 & 1.14 & 11.8 \\
1231: ${\rm NE}^{2:3}_{00}$ & 0.675 & 0.450 & 12.1 & 0.521 & 12.2
1232: & 0.749 & 11.7 & 0.827 & 11.9 & 1.12 & 11.7 \\
1233: ${\rm NE}^{1:2}_{00}$ & 0.633 & 0.442 & 11.9 & 0.505 & 12.1
1234: & 0.734 & 11.6 & 0.803 & 11.7 & 1.09 & 11.5 \\
1235: ${\rm NE}^{1:4}_{00}$ & 0.423 & 0.411 & 11.5 & 0.445 & 11.5
1236: & 0.674 & 11.1 & 0.711 & 11.1 & 0.963 & 10.9 \\
1237: ${\rm NE}^{2:3}_{+-}$ & 0.640 & 0.443 & 11.9 & 0.507 & 12.1
1238: & 0.736 & 11.6 & 0.806 & 11.7 & 1.09 & 11.5 \\
1239: ${\rm NE}^{2:3}_{-+}$ & 0.704 & 0.456 & 12.2 & 0.533 & 12.4
1240: & 0.760 & 11.9 & 0.845 & 12.1 & 1.14 & 11.9 \\
1241: \hline \hline
1242: \end{tabular}
1243: \end{table*}
1244: \end{center}
1245:
1246: We present the RD analysis only for the NE$_{00}^{2:3}$ run, but the
1247: others are qualitatively very similar.
1248: We have extracted the various QNM contributions to the
1249: $_{-\!2}C_{\ell m}$ RD signal in the following way (see also Ref.~\cite{BCP}):
1250: We expect that at late times the $n=0$ QNM dominates.
1251: We fit the signal after time $t_{\rm peak}+t_r$ to this single mode
1252: using non-linear regression and choose $t_r$ to minimize the error in
1253: the fit. We have four dimensionless parameters in this non-linear
1254: fit: the QNM amplitude and phase, ${\cal C}_{\ell m 0}$ and $\phi_{\ell m 0}$,
1255: and the QNM frequency and decay time $M\omega_{\ell m 0}$ and
1256: $\tau_{\ell m 0}/M$. However, instead of fitting directly for these four
1257: parameters, we treat $M\omega_{\ell m 0}$ and $\tau_{\ell m 0}/M$ as
1258: functions of $a_{\rm f}/M_{\rm f}$ and $M_{\rm f}/M$ (which can be obtained via
1259: interpolation from tabulated values given in Ref.~\cite{BCW}).
1260: The advantage of using $(a_{\rm f}/M_{\rm f},M_{\rm f}/M,{\cal C}_{\ell m 0},\phi_{\ell m 0})$ for
1261: the set of fitting parameters comes when we fit to higher overtones.
1262: As done in Ref.~\cite{BCP}, we extract the QNMs treating
1263: the real and imaginary parts of $_{-\!2}C_{\ell m}$ as independent.
1264: Below we shall list results obtained from ${\rm Re}[_{-\!2}C_{\ell m}]$.
1265:
1266: By applying this procedure to the dominant mode, ${}_{-2}C_{22}$,
1267: we obtain $a_{\rm f}/M_{\rm f} = 0.669$ and $M/M_{\rm f} = 0.965$
1268: together with the amplitude and phase of the fundamental
1269: QNM. We include additional overtones ($n>0$) successively.
1270: For each value of $n$, we refit the entire function, so for $n=0$ there are 4 parameters in the
1271: fit, for $n=1$ there are 6, for $n=2$ there are 8, and so forth.
1272: Thus, applying a 6-parameter fit we successfully extract also the
1273: first overtone simultaneously, obtaining
1274: slightly different values for $a_{\rm f}/M_{\rm f} = 0.661$
1275: and $M/M_{\rm f} = 0.958$. We find it impossible to extract, with a
1276: single 8-parameter fit, also the second
1277: overtone. By contrast if we keep $a_{\rm f}/M_{\rm f}$ and $M/M_{\rm f}$ fixed
1278: and equal to the values obtained when extracting the fundamental
1279: QNM, we find that we can fit up to the second overtone. Moreover,
1280: quite interestingly, the fit provides waveforms that compare
1281: very well with the NR waveforms up to the peak of $I_{22}$,
1282: as can be seen in the upper left panel of Fig.~\ref{fig:rd22}.
1283:
1284: \begin{figure*}
1285: \includegraphics[width=0.48\textwidth,clip=true]{fig5a.ps}
1286: \includegraphics[width=0.48\textwidth,clip=true]{fig5b.ps}\\
1287: \includegraphics[width=0.48\textwidth,clip=true]{fig5c.ps}
1288: \includegraphics[width=0.48\textwidth,clip=true]{fig5d.ps}
1289: \caption{\label{fig:rd22} Comparison of numerical and QNM
1290: waveforms for the NE$_{00}^{2:3}$ run. The dominant modes analyzed are
1291: ${}_{-2}C_{22}$ ({\it upper left}), ${}_{-2}C_{33}$
1292: ({\it upper right}), ${}_{-2}C_{32}$ ({\it lower left}), and
1293: ${}_{-2}C_{44}$ ({\it lower right}). Note that the ${}_{-2}C_{32}$
1294: waveform includes contributions from the $\ell=2,m=2$ modes as well.
1295: We denote with $t_{\rm peak}$ the time of the peak of $I^{22}$.}
1296: \end{figure*}
1297:
1298: The remaining panels in Fig.~\ref{fig:rd22} show
1299: results for the other relevant modes ${}_{-2}C_{33}$, ${}_{-2}C_{44}$ and
1300: ${}_{-2}C_{32}$. As obtained in Ref.~\cite{BCP},
1301: we find a ``mode-mixing'' in ${}_{-2}C_{32}$, i.e., the RD waveform is a combination
1302: of $\ell=2,m=2$ and $\ell=3,m=2$ QNMs. This effect appears to be most
1303: important between modes with the same $m$ value, and may possibly be
1304: explained by the fact that the QNMs should really be expressed as {\it
1305: spheroidal}, not {\it spherical} harmonics \cite{BCW,BCP}. Including
1306: both sets of modes means that the ${}_{-2}C_{32}$ is actually fit
1307: using 14 parameters: the final mass and spin, and the amplitude and
1308: phase of 6 QNMs.
1309:
1310: By fitting the fundamental QNM for each ringdown waveform,
1311: we obtain $a_{\rm f}/M_{\rm f} = 0.671$ and $M/M_{\rm f} = 0.972$;
1312: $a_{\rm f}/M_{\rm f} = 0.527$ and $M/M_{\rm f} = 0.884$; $a_{\rm
1313: f}/M_{\rm f} = 0.686$ and $M/M_{\rm f} = 0.981$,
1314: for ${}_{-2}C_{33}$, ${}_{-2}C_{44}$ and ${}_{-2}C_{32}$,
1315: respectively. We also are able to extract the fundamental QNM for the
1316: ${}_{-2}C_{21}$ mode (not shown in Fig.~\ref{fig:rd22}) and find
1317: $a_{\rm f}/M_{\rm f}=0.678$ and $M/M_{\rm f}=0.960$.
1318: All of these values for the inferred final BH spin and mass are
1319: rather consistent, except for ${}_{-2}C_{44}$. This discrepancy
1320: might be due to numerical resolution effects, and will be the object
1321: of future investigations.
1322:
1323: Thus we find that although we cannot simultaneously extract three QNMs
1324: (the fundamental and two overtones) and we are not able to
1325: clearly determine the onset of the RD phase,
1326: we {\it do} obtain that for $t > t_{\rm peak}$
1327: the numerical waveforms can be well fitted by a superposition of
1328: three QNMs. This result explains why the simple matching procedure
1329: from inspiral to RD adopted in the EOB model~\cite{BD2,DG,BCP} can
1330: almost always work succesfully (see Ref.~\cite{EOB4PN} for some caveats).
1331: In Sec.~\ref{matching_RD} we shall adopt the same matching procedure of the EOB
1332: model when building the full waveform using the pseudo-analytic
1333: model of Sec.~\ref{quasi-Newtonian}.
1334:
1335: \section{Effective Newtonian model}\label{quasi-Newtonian}
1336:
1337: In an attempt to better understand the amplitudes and frequencies of
1338: the various modes during the inspiral and merger phases, we present
1339: here what we call the ``effective Newtonian'' (eN) model. It begins with
1340: calculating the leading-order Newtonian formulae for each multipole
1341: moment of the source, as a function of the BH masses, binary
1342: separation $R$, and orbital phase $\phi$. To extend these formulae through
1343: the end of the inspiral and into the merger phase, we introduce an
1344: effective radial separation to absorb PN effects into
1345: the leading-order multipole expressions. Each multipole moment is then
1346: individually matched to a linear superposition of ringdown modes, as
1347: is done in the effective-one-body model~\cite{BD2,DG,BCP}. Taken
1348: together with the match to Kerr QNMs, this
1349: eN model provides an excellent framework within which
1350: we can understand the details of the linear momentum flux and
1351: net recoil velocity.
1352:
1353: \subsection{Newtonian Multipole Moments}\label{effective_radius}
1354:
1355: Working at leading Newtonian order for each mode, we equate
1356: the radiative multipole moments to the source multipole
1357: moments. Restricting ourselves to circular, planar orbits, we find
1358: that for non-spinning systems, the dominant modes are
1359: \cite{BD,BDS,BS,JS}
1360: %
1361: \begin{subequations}\label{I_nospin}
1362: \bea
1363: \label{S21}
1364: S_{\rm nospin}^{21} &=& -\frac{8}{3}i\sqrt{\frac{2\pi}{5}}\,\frac{\delta m}{M} \, \mu
1365: \,R^3\, \omega^4 \,e^{-i\phi}, \\
1366: \label{I22}
1367: I_{\rm nospin}^{22} &=& 16i\sqrt{\frac{2\pi}{5}}\, \mu\, R^2\, \omega^3\, e^{-2i\phi}, \\
1368: \label{I31}
1369: I_{\rm nospin}^{31} &=& -\frac{2}{3}\sqrt{\frac{\pi}{35}}\, \frac{\delta m}{M} \, \mu\,
1370: R^3\, \omega^4\, e^{-i\phi}, \\
1371: \label{S32}
1372: S_{\rm nospin}^{32} &=& -\frac{16}{3}\sqrt{\frac{2\pi}{7}}\, \mu\,(1-3\eta)\, R^4\, \omega^5
1373: \,e^{-2i\phi}, \\
1374: \label{I33}
1375: I_{\rm nospin}^{33} &=& 54\sqrt{\frac{\pi}{21}}\, \frac{\delta m}{M} \, \mu\,
1376: R^3\, \omega^4\, e^{-3i\phi}, \\
1377: \label{I42}
1378: I_{\rm nospin}^{42} &=& \frac{16}{63}i\sqrt{2\pi}\,\mu\,(1-3\eta)\,R^4\,\omega^5\,
1379: e^{-2i\phi}, \\
1380: \label{I44}
1381: I_{\rm nospin}^{44} &=& -\frac{256}{9}i\sqrt{\frac{2\pi}{7}}\,\mu\,(1-3\eta)\,
1382: R^4\,\omega^5\, e^{-4i\phi},
1383: \eea
1384: \end{subequations}
1385: %
1386: where $R$ is the radial separation
1387: and $\omega = \dot{\phi}$ is the binary orbital frequency.
1388: Considering only the mass quadrupole terms in the linear momentum flux
1389: (i.e., the terms proportional to $S^{21}I^{22*}$, $I^{31}I^{22*}$,
1390: and $I^{22}I^{33*}$),
1391: we obtain the well-known result valid at Newtonian order~\cite{DG}:
1392: %
1393: \begin{equation}
1394: \label{flux_Newt0}
1395: F^{(0)} = -i \frac{464}{105}\frac{\delta m}{M} \, \mu^2\, R^5\, \omega^7\, e^{i\phi}.
1396: \end{equation}
1397: %
1398: Including the next-highest order moments in
1399: Eq.~(\ref{dPdt_radmoments2}), we get
1400: \begin{equation}
1401: \label{flux_Newt1}
1402: F^{(1)} = -i \frac{11120}{1323}\frac{\delta m}{M} \, \mu^2\, (1-3\eta) R^7\, \omega^9\, e^{i\phi}.
1403: \end{equation}
1404: While there may also be next-to-leading order contributions from a PN
1405: expansion of the multipole moments included in
1406: Eq.~(\ref{dPdt_radmoments}) that would show up in
1407: Eq.~(\ref{flux_Newt1}), we can effectively absorb those corrections
1408: into the $R$ variable, as will be described below.
1409:
1410: \begin{figure*}
1411: \includegraphics[width=0.48\textwidth,clip=true]{fig6a.ps}
1412: \includegraphics[width=0.48\textwidth,clip=true]{fig6b.ps}
1413: \caption{\label{r_eff} Effective radius for different modes, derived
1414: from Eqs.~(\ref{omega_lm}), (\ref{S21})--(\ref{I44}).
1415: The close agreement for the $R_{\rm eff}^{lm}$ suggests we can use a single
1416: effective radius $R_{\rm eff}(t)$ for the Newtonian expressions. We believe
1417: that the large oscillations in $R_{\rm eff}^{21}$ are due to initial
1418: eccentricity at early times. Also plotted is the ADM radius (dashed
1419: curves) derived from the orbital frequency via Eq.~(\ref{eqn:R_ADM}),
1420: the coordinate separation of the BH punctures (dot-dashed curves), and
1421: the empirical fit $R_{\rm fit}$ (dotted curves) obtained by shifting
1422: $R_{\rm ADM}$ by 0.65. The results correspond to the
1423: NE$_{00}^{2:3}$ (left panel) and NE$_{00}^{1:2}$ (right panel)
1424: runs. We denote with $t_{\rm peak}$ the time at which $I^{22}$ reaches its maximum.}
1425: \end{figure*}
1426:
1427: Combining Eqs.~(\ref{flux_Newt0}) and (\ref{flux_Newt1}) we find the
1428: linear momentum flux scales like
1429: \bea\label{jena_formula}
1430: |F^{(0)}+F^{(1)}| &\propto & \frac{\delta m}{M}\, \mu^2
1431: \left[1+\frac{3475}{1827}(1-3\eta)R^2 \omega^2\right] \nonumber\\
1432: &\approx& \frac{3}{2}\frac{\delta m}{M}\, \mu^2 (1-0.9\eta),
1433: \eea
1434: which is remarkably similar to the result found in
1435: Ref.~\cite{recoilJena}. Here we have used $R^2\omega^2\approx 0.23-0.25$ at
1436: the peak of the energy flux, which seems to be quite robust for a
1437: range of mass ratios. However, the extremely close agreement with
1438: Ref.~\cite{recoilJena} is probably to some degree a coincidence, since
1439: this simple Newtonian formula does not include any details of the
1440: phase relations between different modes, which become especially
1441: important during the transition from inspiral to ringdown (see
1442: Sec.~\ref{transition} below). Since Eq.~(\ref{jena_formula}) really
1443: only applies to the inspiral portion, if anything, it should be a
1444: predictor of how the {\it peak} recoil velocity scales. This is not
1445: necessarily the same as the {\it final} recoil, since
1446: we find that more extreme-mass-ratio BH binaries have a
1447: relatively smaller anti-kick, which should also play an important role
1448: in the scaling relation of Ref.~\cite{recoilJena}.
1449:
1450: If we compute the above multipole moments (\ref{S21})--(\ref{I44})
1451: using $\omega$ as given by Eq.~(\ref{omega_lm}) and $R$ as obtained from the puncture trajectories,
1452: we do not find a very good agreement with the numerical results. This is not
1453: surprising since there is no reason to believe that the Newtonian
1454: approximation should work well all along the inspiral phase. We
1455: should expect that higher-order PN corrections become important as
1456: we approach the merger. Furthermore, $R$ is a
1457: coordinate-dependent quantity, and thus does not necessarily have the
1458: same meaning in a PN expression as in NR. Since our scope is limited to a diagnostic
1459: of the NR results, and not to a precise comparison with PN calculations,
1460: instead of including PN corrections in
1461: Eqs.~(\ref{I_nospin})-(\ref{flux_Newt1}), we investigate whether by
1462: properly scaling the
1463: Newtonian expressions we can get a better agreement until the merger. We can also think
1464: of this normalization as a way of resumming the PN expansion.
1465:
1466: Quite interestingly, if we compute the amplitudes
1467: $|I^{\ell m}|$ or $|S^{\ell m}|$ from the numerical data,
1468: and the angular frequency $\omega$ from Eq.~(\ref{omega_lm}),
1469: we find that the radii $R^{\ell m}$ which appear in the RHS of
1470: Eqs.~(\ref{S21})--(\ref{I44}) are rather
1471: independent of the multipole moments $\ell $ and $m$, as Fig.~\ref{r_eff}
1472: shows. We denote the radii $R^{\ell m}$ computed numerically as {\it effective}
1473: radii $R_{\rm eff}^{\ell m}$. The close agreement between the frequencies (see
1474: Fig.~\ref{omega_modes}) and effective radii for each mode suggests we
1475: can use the Newtonian expressions and a single $R_{\rm eff}(t)$ and
1476: orbital frequency $\omega(t)$,
1477: e.g., $R_{\rm eff}^{22}(t)$ and $\omega_{\rm D}^{I22}$for all modes with a high degree of accuracy for
1478: the entire inspiral phase and even during the transition to merger.
1479:
1480: For comparison we also show in Fig.~\ref{r_eff} the radius
1481: from the puncture trajectory (dot-dashed curves) and the radius
1482: computed using the Arnowitt-Deser-Misner transverse-traceless gauge
1483: (dashed curves), given as a function of frequency through 3PN order
1484: by~\cite{BI}
1485: %
1486: \begin{widetext}
1487: \beq\label{eqn:R_ADM}
1488: R_{\rm ADM} = M^{1/3}\, \omega^{-2/3}\,\left [ 1 + \omega^{2/3}\,\left (-1 + \frac{\eta}{3} \right )
1489: + \omega^{4/3}\,\left ( -\frac{1}{4} + \frac{9}{8}\,\eta + \frac{\eta^2}{9} \right )
1490: + \omega^2\,\left ( -\frac{1}{4} - \frac{1625}{144}\,\eta + \frac{167}{192}\,\eta\, \pi^2
1491: - \frac{3}{2}\,\eta^2 + \frac{2}{81}\,\eta^3\right ) \right ]\,.
1492: \eeq
1493: \end{widetext}
1494: %
1495: Here we use the orbital frequency $\omega$ derived from the $I^{22}$
1496: mode via Eqn.~(\ref{omega_lm}), giving a constant value during
1497: the RD phase when the orbital frequency is meaningless.
1498: Fig.~\ref{r_eff} shows interesting agreement between $R_{\rm ADM}$ and
1499: the radius from the puncture trajectory, and a constant offset
1500: between $R_{\rm ADM}$ and $R_{\rm eff}$. The latter is due to the fact
1501: that the amplitude of the multipole moments computed at
1502: leading Newtonian order does not reproduce the numerical
1503: relativity amplitude~\cite{BCP,Baker:2006kr}, and higher order
1504: PN corrections need to be included. Motivated by this similarity
1505: between $R_{\rm ADM}$ and $R_{\rm eff}$, we attempt to fit empirically
1506: the $R_{\rm eff}$ curves in Fig.~\ref{r_eff} by simply shifting
1507: $R_{\rm ADM}$ by $0.65$. The fit curve is included as
1508: a dotted curve in Fig.~\ref{r_eff}. As we accumulate longer and
1509: more accurate NR data for a wider range of $\eta$ values, and
1510: study possible analytic resummation of higher-order PN
1511: amplitude corrections, we should be able to work out
1512: a widely applicable amplitude-scaling factor to be included
1513: in leading-order analytic waveforms~\cite{EOB4PN}.
1514:
1515: In the next section, we shall investigate how
1516: this simple eN model can be combined
1517: with a superposition of QNMs, as described in Sec.~\ref{RD_phase},
1518: giving a good representation of the NR results.
1519:
1520: \subsection{Matching to ringdown}
1521: \label{matching_RD}
1522:
1523: We now match the inspiral and RD
1524: waveforms in a mode-by-mode fashion following the philosophy of the
1525: EOB approach~\cite{BD2}. Note this is not the same
1526: analysis of Section \ref{RD_phase}, where we {\it fit} the numerical
1527: data throughout the RD phase with a superposition of QNMs. Here we
1528: {\it match} the data at a single point at the transition from inspiral
1529: to RD and see how well it agrees with the rest of the RD phase. A
1530: similar attempt was followed in Ref.~\cite{DG},
1531: where for simplicity the authors performed the matching to the
1532: Schwarzschild QNM frequencies, while we use the Kerr QNM frequencies
1533: and match to the fundamental QNM frequency and the first
1534: two overtones, as done in Ref.~\cite{BCP}.
1535: We obtain the QNM frequencies and decay times from Ref.~\cite{BCW} as
1536: a function of $a_{\rm f}/M_{\rm f}$ (taken from
1537: Table~\ref{table:idparams} above).
1538: For the fundamental and two overtone QNMs, we can match a given multipole
1539: mode by equating it and two time derivatives to a linear combination
1540: of QNMs.
1541:
1542: We write
1543: %
1544: \beq
1545: \label{RD}
1546: I^{\ell m}(t) = A(t)\,e^{-i\phi(t)} = \sum_{n=0}^{\infty} A_{\ell mn}\,
1547: e^{-i\sigma_{\ell mn} (t-t_{\rm match})},
1548: \eeq
1549: %
1550: where the complex QNM frequencies are known functions of the final BH
1551: mass and spin, and we must solve for the complex amplitudes
1552: $A_{\ell mn}$. Matching three QNMs we get
1553: %
1554: \bse
1555: \begin{eqnarray}
1556: I^{\ell m}(t_{\rm match}) &=& \sum_{n=0}^{2} A_{\ell mn}, \\
1557: \frac{d}{dt}I^{\ell m}(t_{\rm match}) &=& -i \,\sum_{n=0}^{2} \sigma_{\ell mn}
1558: A_{\ell mn}, \\
1559: \frac{d^2}{dt^2}I^{\ell m}(t_{\rm match}) &=& - \sum_{n=0}^{2} \sigma_{\ell mn}^2 \,A_{\ell mn},
1560: \end{eqnarray}
1561: \ese
1562: %
1563: or as a simple matrix equation
1564: %
1565: \begin{equation}
1566: \left(\begin{array}{ccc}
1567: 1 & 1 & 1 \\
1568: -i\sigma_{\ell m0} & -i\sigma_{\ell m1} & -i\sigma_{\ell m2} \\
1569: -\sigma_{\ell m0}^2 & -\sigma_{\ell m1}^2 & -\sigma_{\ell m2}^2
1570: \end{array}\right) \left(\begin{array}{c} A_{\ell m0} \\
1571: A_{\ell m1} \\ A_{\ell m2} \end{array}\right) =
1572: \left(\begin{array}{c} I^{\ell m} \\
1573: \dot{I}^{\ell m} \\ \ddot{I}^{\ell m} \end{array}\right).
1574: \end{equation}
1575: %
1576:
1577: \begin{figure*}
1578: \includegraphics[width=0.48\textwidth,clip=true]{fig7a.ps}
1579: \includegraphics[width=0.48\textwidth,clip=true]{fig7b.ps}\\
1580: \includegraphics[width=0.48\textwidth,clip=true]{fig7c.ps}
1581: \includegraphics[width=0.48\textwidth,clip=true]{fig7d.ps}
1582: \caption{\label{compare_ring} Comparison of the effective Newtonian
1583: and NR radiative modes during inspiral, merger and RD phases. The
1584: data refer to the NE$_{00}^{1:2}$ run. We denote with $t_{\rm peak}$
1585: the time at which $I^{22}$ reaches its maximum.}
1586: \end{figure*}
1587:
1588: \begin{figure}
1589: \includegraphics[width=0.5\textwidth,clip=true]{fig8.ps}
1590: \caption{\label{kick_match} Comparison of the effective Newtonian model and NR predictions for the recoil velocity
1591: for a range of inspiral-RD matching points. We denote with $t_{\rm peak}$ the
1592: time at which $I^{22}$ reaches its maximum. The data refer to the NE$_{00}^{2:3}$ run.}
1593: \end{figure}
1594:
1595: In Fig.~\ref{compare_ring}, we compare the NR modes to
1596: the modes obtained by the effective Newtonian model described in
1597: Sec.~\ref{effective_radius} until $t_{\rm match}$ and by the superposition
1598: of three QNMs for $t > t_{\rm match}$. During the
1599: inspiral, the different moments are calculated according to
1600: Eqs.~(\ref{S21})-(\ref{I44}), using a single $R_{\rm eff}$ and
1601: $\omega_{\rm D}$ determined from the $I^{22}$ mode, with the
1602: exception of the $S^{21}$ mode, where we instead use the higher
1603: frequency $\omega_{\rm D}^{S21}$ (but same $R_{\rm eff}$).
1604: We treat $t_{\rm match}$ as a free parameter: if we stop the inspiral
1605: too early, the eN mode amplitudes are still growing,
1606: so the sudden transition to decaying RD modes prematurely reduces
1607: them. On the other hand, if the inspiral is continued too long, we
1608: tend to lose the important phase shifts
1609: between the modes that only begin during the transition to
1610: RD. This is particularly evident in the $I^{44}$
1611: mode, which undergoes an unexplained phase-shift around
1612: the transition to RD, and also decays at somewhat different rate
1613: than is predicted from QNM theory (see above, Sec.~\ref{RD_phase}).
1614: Motivated by the results of Sec.~\ref{RD_phase}, notably by the fact
1615: that a superposition of three QNMs can fit very well the NR waveforms
1616: starting from the peak of the energy flux, we choose as best-matching
1617: point the peak of the energy flux.
1618:
1619: Having shown a reasonably close match for each of the
1620: radiative multipoles between the effective Newtonian model and the
1621: numerical data, it stands to reason that the total recoil calculated
1622: with this model should agree as well. This is shown in Fig.~\ref{kick_match},
1623: where we have also varied the matching point around $t_{\rm
1624: peak}$. We first note the close agreement between the eN
1625: models with varying $t_{\rm match}$, suggesting the
1626: inspiral-to-ringdown matching method described above is relatively
1627: robust. Not surprisingly, since the individual modes agree, we also
1628: find reasonable agreement between the NR data and the eN predictions
1629: for the recoil.
1630:
1631: However, this agreement may be partially fortuitous, since the eN model
1632: cannot predict the mode phase shifts
1633: around $t=t_{\rm peak}$, most notably that of the $I^{44}$ mode
1634: described above.
1635: In Section \ref{transition} below, we will examine
1636: this phasing in greater detail and show how it affects the overall
1637: kick. At this point, we unfortunately do not have a clear
1638: understanding of the underlying cause of the phase shift, but it may
1639: well be related to the slightly different times of transition from
1640: inspiral to ringdown for the different modes. Preliminary results
1641: also suggest that this de-phasing effect is reduced in more
1642: extreme-mass-ratio systems, as we shall see in Appendix~\ref{app}.
1643:
1644: \section{Anatomy of the kick}
1645: \label{anatomy}
1646:
1647: In the above Sections, we have laid the groundwork for a multipolar
1648: analysis of the gravitational recoil, describing the momentum flux as
1649: a combination of radiative multipole modes. Along with the
1650: psuedo-analytic models for the inspiral and ringdown phases, we can
1651: now give a detailed description of the ``anatomy'' of the kick, namely
1652: the way the different modes combine to produce a peak recoil velocity,
1653: followed by a characteristic anti-kick and then asymptotic approach to
1654: the final value of the BH recoil.
1655:
1656: \subsection{Contribution from different moments}
1657: \label{diff_mom}
1658:
1659: In Sec.~\ref{linear_flux}, we showed how the radiative multipole
1660: moments contribute to the linear momentum flux through the integral
1661: of the $\Psi_4$ scalar
1662: [Eqs.~(\ref{eq:psi4Ylmdef}),(\ref{dPdt_psi4})]. Here, we want to
1663: determine exactly which modes we need to include in the multipole expansion
1664: Eq.~(\ref{thorne_420}) to get a good representation of the full recoil, and
1665: which are the pairs of modes in Eq.~(\ref{flux_approx}) that
1666: contribute most.
1667:
1668: By including only a select choice of terms in the $\psi_4$ expansion
1669: Eq.~(\ref{eq:psi4Ylmdef}), we can calculate the linear momentum flux
1670: by direct integration of Eq.~(\ref{dPdt_psi4}) and compare it with the
1671: predictions of Eqs.~(\ref{dPdt_radmoments})-(\ref{flux_approx}), in
1672: each case including only the appropriate moments. This is a good
1673: way of double-checking those lengthy equations term-by-term, and in
1674: practice we find excellent agreement, limited only by the numerical
1675: accuracy of the simulations. Similarly, we can use this method of
1676: truncated expansion to determine which modes are necessary for
1677: calculating the recoil up to a given accuracy.
1678: The results of using higher and higher order multipolar moments are
1679: shown in Figs.~\ref{v_tot1} and \ref{v_tot2} for the NE$_{00}^{2:3}$ and
1680: NE$_{00}^{1:2}$ runs, respectively.
1681:
1682: \begin{figure*}
1683: \includegraphics[width=0.48\textwidth,clip=true]{fig9a.ps}
1684: \includegraphics[width=0.48\textwidth,clip=true]{fig9b.ps}
1685: \caption{\label{v_tot1} In the left panel we show the net recoil kick, integrated from the
1686: linear momentum flux via Eq.~(\ref{dPdt_psi4}) (solid curve),
1687: from all modes with $\ell\le 4$ (dashed curve) and also limiting the
1688: modal composition of $\Psi_4$ to just the three dominant mode pairs
1689: in Eq.~(\ref{flux_approx}) (dotted curve). In the right panel
1690: we show the difference between the exact result and the $\Psi_4$
1691: expansion Eq.~(\ref{eq:psi4Ylmdef}), limited to $\ell\leq 3,4,5,6$.
1692: The data refer to the NE$_{00}^{2:3}$ run.
1693: We denote with $t_{\rm peak}$ the time at which $I^{22}$ reaches its maximum.}
1694: \end{figure*}
1695:
1696: \begin{figure*}
1697: \includegraphics[width=0.48\textwidth,clip=true]{fig10a.ps}
1698: \includegraphics[width=0.48\textwidth,clip=true]{fig10b.ps}
1699: \caption{\label{v_tot2} Same as Fig.~\ref{v_tot1}, but for the
1700: NE$_{00}^{1:2}$ run.}
1701: \end{figure*}
1702:
1703: In the left panels of Figs.~\ref{v_tot1} and \ref{v_tot2} we show with a
1704: solid curve the exact recoil velocity from Eq.~(\ref{dPdt_psi4}),
1705: with a dashed curve the contribution from terms up to $\ell=4$, i.e., those
1706: obtained from Eq.~(\ref{dPdt_radmoments}) and
1707: (\ref{dPdt_radmoments2}), and with a dotted curve the contribution
1708: from just the three leading terms in Eq.~(\ref{flux_approx}), valid
1709: for non-precessing BHs with kicks in the orbital plane.
1710: We conclude that the linear momentum flux is dominated by the
1711: $I^{33}I^{22*}$, $I^{33}I^{44*}$, and $S^{21}I^{22*}$ terms, which
1712: combine to produce
1713: the primary kick and anti-kick agreeing with the exact result
1714: within $\lesssim 10\%$ throughout the entire merger. Note that the
1715: flux from the $S^{32}I^{33*}$ term, while not insignificant in
1716: Fig.~\ref{flux_lm}, contributes almost nothing to the net recoil
1717: velocity. This is largely due to phase relations between the various
1718: modes during the transition from inspiral to ringdown, described below
1719: in Sec.~\ref{transition}.
1720:
1721: In the right panels of Figs.~\ref{v_tot1} and \ref{v_tot2} we
1722: show the difference between the calculation obtained including terms up
1723: to $\ell=3,4,5,6,$ and the exact result. It seems clear that we need modes
1724: up to and including $\ell=4$ to get an accurate
1725: estimate of the recoil velocity. For more extreme mass ratios,
1726: higher-order moments become relatively more important, but remain
1727: strongly sub-dominant to the $\ell\le 4$ modes \cite{recoilAEI,berti07}.
1728:
1729: \begin{figure}
1730: \includegraphics[width=0.48\textwidth,clip=true]{fig11.ps}
1731: \caption{\label{r_eff_mp} $R_{\rm eff}$ derived from different
1732: multipole modes, as in Fig.~\ref{r_eff}, for the NE$_{-+}^{2:3}$
1733: run. The $S^{21}$ mode for this run has comparable contributions
1734: from $\delta m$ and $\Delta^z$, making it difficult to derive a
1735: reasonable $R_{\rm eff}(S^{21})$.}
1736: \end{figure}
1737:
1738: To understand more clearly the relative contributions of the different modes
1739: to the total recoil, we will include analysis of a few
1740: more simulations including non-precessing spins. As mentioned above in
1741: Sec.~\ref{linear_flux}, non-precessing spins do not introduce any
1742: additional moments compared to the non-spinning simulations, but
1743: simply modify the relative amplitudes of the different modes in
1744: Eq.~(\ref{flux_approx}) by adding the spin terms. Thus, once we
1745: determine how the spins modify the individual modes, we
1746: can use the same analysis for the spinning and non-spinning cases.
1747:
1748: Again equating the radiative multipole moments with the source moments,
1749: we get the leading order spin-orbit modifications
1750: to Eqs.~(\ref{S21})--(\ref{I44}) [see Eqs.~(3.14),(3.20) in
1751: Ref.~\cite{LK} and Eq.~(5.5) in Ref.~\cite{BBF}]:
1752: \begin{subequations}\label{spinorbit_moments}
1753: \begin{eqnarray}
1754: S^{21}_{\rm SO} &=& -4i\sqrt{\frac{2\pi}{5}}\,\eta\, R\, \omega^3 \,e^{-i\phi}
1755: \Delta^z, \\
1756: I^{22}_{\rm SO} &=& \frac{64}{3}i\sqrt{\frac{2\pi}{5}}\,\eta\, R^2\,
1757: \omega^4\, e^{-2i\phi}\, \xi^z\, \\
1758: S^{32}_{\rm SO} &=& -\frac{32}{3}\sqrt{\frac{2\pi}{7}}\,\eta\, R^2\,
1759: \omega^4\, e^{-2i\phi}\, \xi^z, \\
1760: I^{31}_{\rm SO} &=& -\frac{2}{3}\sqrt{\frac{\pi}{35}}\, \eta\,
1761: R^3\, \omega^5\, e^{-i\phi}\, \Sigma_{31}^z, \\
1762: I^{33}_{\rm SO} &=& 54\sqrt{\frac{\pi}{21}}\, \eta\,
1763: R^3\, \omega^5\, e^{-3i\phi}\, \Sigma_{33}^z,
1764: \end{eqnarray}
1765: \end{subequations}
1766: where we have introduced the spin vectors
1767: \begin{subequations}\label{sigma_z}
1768: \bea
1769: \Sigma_{31} &\equiv& \frac{11}{2}\frac{\delta m}{M}\, \mathbf{S} +
1770: \frac{1}{2}(11-39\eta)\mathbf{\Delta}, \\
1771: \Sigma_{33} &\equiv& \frac{3}{2}\frac{\delta m}{M}\, \mathbf{S} +
1772: \frac{3}{2}(1-5\eta)\mathbf{\Delta}.
1773: \eea
1774: \end{subequations}
1775:
1776: \begin{figure*}
1777: \includegraphics[width=0.48\textwidth,clip=true]{fig12a.ps}
1778: \includegraphics[width=0.48\textwidth,clip=true]{fig12b.ps}\\
1779: \includegraphics[width=0.48\textwidth,clip=true]{fig12c.ps}
1780: \includegraphics[width=0.48\textwidth,clip=true]{fig12d.ps}
1781: \caption{\label{ne_modeamps} Relative amplitudes of the dominant
1782: multipole mode-pairs in the linear momentum flux. Also shown in the dashed
1783: curves are the eN model predictions for the flux amplitudes.
1784: We denote with $t_{\rm peak}$ the time at which $I^{22}$ reaches its maximum.}
1785: \end{figure*}
1786:
1787: In all of the simulations considered here, the dimensionless spins are equal
1788: ($|a_1|/m_1=|a_2|/m_2$) and point in opposite directions, $\xi^z=0$,
1789: so for the leading-order terms in Eqn.~(\ref{flux_approx}) we are left
1790: only with the
1791: modifications of $S^{21}$ and $I^{33}$, due to $\Delta^z$ and
1792: $\Sigma_{33}^z$, respectively. Then Eqs.~(\ref{I_nospin}) and
1793: (\ref{spinorbit_moments}) give the linear momentum flux
1794: during the inspiral for each of the first three dominant terms in
1795: Eq.~(\ref{flux_approx}):
1796: %
1797: \begin{widetext}
1798: \begin{subequations}\label{newtonian_flux}
1799: \bea
1800: F_{\rm insp}^{21,22} &=& \frac{16}{45}i \frac{\mu^2}{M}\, R^3\, \omega^6\,
1801: (2\delta m \, R^2 \omega +3\Delta^z)\, e^{i\phi},
1802: \label{newtonian_flux1} \\
1803: F_{\rm insp}^{22,33} &=& -\frac{36}{7}i \frac{\mu^2}{M}\,
1804: R^5\,\omega^7\, (\delta m + \omega\, \Sigma_{33}^z)
1805: \,e^{i\phi},
1806: \label{newtonian_flux2} \\
1807: F_{\rm insp}^{33,44} &=& -\frac{64}{7}i \frac{\mu^2}{M}\, (1-3\eta)\,
1808: R^7\,\omega^9\, (\delta m + \omega\, \Sigma_{33}^z)\, e^{i\phi}
1809: \label{newtonian_flux3}.
1810: \eea
1811: \end{subequations}
1812: \end{widetext}
1813: %
1814: While these flux formulae contain terms of various orders in
1815: $\omega$, we expect that the effective Newtonian scaling of $R$
1816: ensures that we are including all relevant PN terms, at least in the
1817: cases where the $\delta m$ terms dominate over the spin
1818: corrections. When the spin terms begin to dominate, we find that it
1819: becomes more difficult to use a single effective $R$ for all
1820: modes. This can be seen in Fig.~\ref{r_eff_mp}, which plots $R_{\rm
1821: eff}$ as in Fig.~\ref{r_eff}, but for the NE$_{-+}^{2:3}$ run, where
1822: the $\Delta^z$ and $\delta m$ terms in Eq.~(\ref{newtonian_flux1})
1823: are comparable, making it difficult to derive a reasonable $R_{\rm
1824: eff}(S^{21})$.
1825:
1826: Even for non-spinning runs, in order to get reasonable agreement with
1827: the NR data, we find that one
1828: must be careful towards the end of the inspiral to distinguish between
1829: $\omega_{\rm D}^{I22}$ and
1830: $\omega_{\rm D}^{S21}$ in Eq.~(\ref{newtonian_flux1}):
1831: %
1832: \begin{equation}
1833: F_{\rm insp}^{21,22} \propto (\mu^2/M) R^3\, (\omega_{\rm D}^{I22})^3\,
1834: (\omega_{\rm D}^{S21})^3\, (2\delta m \, R^2 \omega_{\rm D}^{S21}
1835: +3\Delta^z)\, .
1836: \end{equation}
1837:
1838: The amplitudes of these fluxes are plotted in Fig.~\ref{ne_modeamps}
1839: for the four runs NE$_{-+}^{2:3}$, NE$_{+-}^{2:3}$, NE$_{00}^{2:3}$,
1840: and EQ$_{+-}$. As seen in Table~\ref{table:idparams}, the
1841: NE$_{-+}^{2:3}$ run has $\Delta^z = 0.2M^2$, while the
1842: NE$_{+-}^{2:3}$ run has $\Delta^z = -0.2M^2$, respectively adding
1843: destructively and constructively with the $\delta m$ term in
1844: Eq.~(\ref{newtonian_flux1}). This difference is clearly seen in the
1845: blue curves in the top two panels of Fig.~\ref{ne_modeamps}. Also notable
1846: in these plots is the somewhat smaller difference in the amplitudes of
1847: $F^{22,33}$, due to a similar
1848: effect from the constructive/destructive additions of $\delta m$ and
1849: $\Sigma_{33}^z$ in Eq.~(\ref{newtonian_flux2}).
1850: As we see in Fig.~\ref{ne_modeamps}, NE$_{00}^{2:3}$ appears to be the
1851: average of NE$_{+-}^{2:3}$ and NE$_{-+}^{2:3}$, while the flux from
1852: EQ$_{+-}$ is strongly suppressed due to the $\delta m=0$ terms in
1853: Eq.~(\ref{newtonian_flux}), leaving only the flux from the terms
1854: proportional to $\Delta^z=-0.2M^2$ and $\Sigma_{33}^z=0.075$. However,
1855: as noted above, when the spin terms dominate the flux, as in the
1856: case of equal-mass BHs, the eN model with a single $R_{\rm eff}$
1857: begins to break down. Yet even in this situation,
1858: Eqs.~(\ref{newtonian_flux1})-(\ref{newtonian_flux3}) still have
1859: qualitative (if not quantitative) predictive value, including the
1860: relative phases between the different mode-pair fluxes during the
1861: inspiral.
1862:
1863: In each panel of Fig.~\ref{ne_modeamps}, we also plot
1864: with dashed lines the eN prediction for the various flux
1865: amplitudes. In almost all cases, the eN flux is quite close to the
1866: NR results up to about $10M$ before $t_{\rm peak}$, when the
1867: eN model begins to break down, especially for the spinning runs.
1868: The amplitude differences near the peaks are comparable to those seen in
1869: Fig.~\ref{compare_ring} for the NE$_{00}^{2:3}$ run. The notable
1870: exception is the $F^{21,22}$ flux from the NE$_{-+}^{2:3}$ and
1871: EQ$_{+-}$ runs, where the spin terms dominate over the $\delta m$
1872: terms.
1873:
1874: \subsection{Transition to ringdown and the de-phasing of the multipole
1875: modes}\label{transition}
1876:
1877: Since the flux vectors defined by Eq.~(\ref{newtonian_flux})
1878: will not generally be co-linear, to understand the time evolution of the recoil velocity,
1879: we must first understand the phase relations between the different
1880: modes. From Eqs.~(\ref{I_nospin}), (\ref{spinorbit_moments}), and
1881: (\ref{newtonian_flux}),
1882: we see that during the inspiral phase, the individual moments and the resulting
1883: flux vectors evolve according to a single orbital phase $\phi$, with
1884: $F_{\rm insp}^{21,22}$ pointing
1885: in the opposite direction to $F_{\rm insp}^{22,33}$ and
1886: $F_{\rm insp}^{33,44}$. However, as we can see from
1887: Fig.~\ref{omega_modes}, as the binary evolves from inspiral to
1888: RD, the frequency (and thus phase) of the $S^{21}$ mode decouples
1889: from the other dominant modes. Upon closer inspection, we find that
1890: even the $I^{22}, I^{33}$ and $I^{44}$ modes deviate from each other
1891: enough to undergo a significant phase shift at the inspiral-RD
1892: transition.
1893:
1894: To quantify these effects, we define the following phase differences:
1895: %
1896: \begin{subequations}
1897: \bea
1898: \label{phase_shifts1}
1899: \cos\psi^{2-3} &=& \mathbf{\hat{F}}_{\rm insp}^{21,22} \cdot
1900: \mathbf{\hat{F}}_{\rm insp}^{22,33}, \\
1901: \label{phase_shifts2}
1902: \cos\psi^{2-4} &=& \hat{\mathbf{F}}_{\rm insp}^{21,22}\cdot
1903: \mathbf{\hat{F}}_{\rm insp}^{33,44}, \\
1904: \label{phase_shifts3}
1905: \cos\psi^{3-4} &=& \hat{\mathbf{F}}_{\rm insp}^{22,33}\cdot
1906: \mathbf{\hat{F}}_{\rm insp}^{33,44}.
1907: \eea
1908: \end{subequations}
1909: %
1910: Here we use the notation $\psi^{m-m'}$ to describe
1911: the phase difference between two complex flux vectors, where $m$ and
1912: $m'$ correspond to the {\it larger} $m$-values of each mode pair
1913: that makes up the flux. These definitions are valid throughout the
1914: inspiral, merger, and ringdown phases.
1915: In the inspiral phase, we can see that for the unequal-mass runs where
1916: $\delta m$ dominates with respect to the spin terms in
1917: Eqs.~(\ref{newtonian_flux1})--(\ref{newtonian_flux3}), we have
1918: %
1919: \beq\label{phase_shifts_insp}
1920: \cos\psi_{\rm insp}^{2-3}=\cos\psi_{\rm insp}^{2-4}=-1\,, \quad \cos\psi_{\rm insp}^{3-4}=1\,.
1921: \eeq
1922: %
1923: For the EQ$_{+-}$ run with $\delta m=0$, Eq.~(\ref{newtonian_flux})
1924: predicts that all phases have $\cos\psi_{\rm insp}=1$ during
1925: the inspiral (as shown in Table~\ref{table:idparams}, $\Delta^z$ and
1926: $\Sigma^z_{33}$ have opposite signs, so all the flux vectors in
1927: Eq.~(\ref{newtonian_flux}) are parallel).
1928: During the RD phase, using Eq.~(\ref{RD}), we can approximate the flux
1929: vectors and phase evolution in terms of the fundamental QNM
1930: frequencies $\sigma_{\ell m0}$:
1931: %
1932: \begin{widetext}
1933: \begin{subequations}\label{flux_RD}
1934: \bea
1935: \label{flux_RD1}
1936: F_{\rm RD}^{21,22} &\simeq& F_{\rm match}^{21,22}
1937: \exp[-i(\sigma_{210}-\sigma_{220}^*)(t-t_{\rm match})]\, ,\\
1938: \label{flux_RD2}
1939: F_{\rm RD}^{22,33} &\simeq& F_{\rm match}^{22,33}
1940: \exp[-i(\sigma_{220}-\sigma_{330}^*)(t-t_{\rm match})]\, ,\\
1941: \label{flux_RD3}
1942: F_{\rm RD}^{33,44} &\simeq& F_{\rm match}^{33,44}
1943: \exp[-i(\sigma_{330}-\sigma_{440}^*)(t-t_{\rm match})]\, ,
1944: \eea
1945: \end{subequations}
1946: where the $F^{\ell m,\ell' m'}_{\rm match}$ fluxes include complex
1947: phase information at the matching point. Taking the phase differences
1948: between these RD modes gives
1949: \begin{subequations}\label{phase_shifts_RD}
1950: \bea
1951: \label{phase_shifts_RD1}
1952: \cos\psi_{\rm RD}^{2-3} &\simeq&
1953: \cos[(\omega_{210}-2\omega_{220}+\omega_{330})(t-t_{\rm
1954: match})+\Phi_{\rm match}^{2-3}]\,, \\
1955: \label{phase_shifts_RD2}
1956: \cos\psi_{\rm RD}^{2-4} &\simeq&
1957: \cos[(\omega_{210}-\omega_{220}-\omega_{330}+\omega_{440})(t-t_{\rm
1958: match})+\Phi_{\rm match}^{2-4}]\,,\\
1959: \label{phase_shifts_RD3}
1960: \cos\psi_{\rm RD}^{3-4} &\simeq&
1961: \cos[(\omega_{220}-2\omega_{330}+\omega_{440})(t-t_{\rm
1962: match})+\Phi_{\rm match}^{3-4}]\,.
1963: \eea
1964: \end{subequations}
1965: \end{widetext}
1966: %
1967: Here $\Phi_{\rm match}$ is a phase offset determined at the transition
1968: from inspiral to ringdown.
1969: Quite interestingly, we find that for the range of final BH spin parameters
1970: $0.5\le a_{\rm f}/M_{\rm f} \le 0.8$, the linear combinations of frequencies
1971: in Eqs.~(\ref{phase_shifts_RD1})--(\ref{phase_shifts_RD3}) vary by
1972: less than $\sim 5\%$. Thus, if we compute the above expressions
1973: for the $\omega_{lm0}$ corresponding to $a_{\rm f}/M_{\rm f}=0.7$, we
1974: have \cite{BCW}
1975: %
1976: \begin{subequations}\label{phase_shifts_RDe}
1977: \bea
1978: \label{phase_shifts_RD1e}
1979: \cos\psi_{\rm RD}^{2-3} &\simeq & \cos \left [
1980: \frac{0.23}{M_{\rm f}} (t- t_{\rm match}) +
1981: \Phi_{\rm match}^{2-3} \right ]\,,\\
1982: \label{phase_shifts_RD2e}
1983: \cos\psi_{\rm RD}^{2-4} &\simeq &\cos\left [
1984: \frac{0.22}{M_{\rm f}}(t-t_{\rm match}) +\Phi_{\rm match}^{2-4}
1985: \right ]\,,\\
1986: \label{phase_shifts_RD3e}
1987: \cos\psi_{\rm RD}^{3-4} &\simeq &\cos \left [
1988: \frac{0.012}{M_{\rm f}}(t-t_{\rm match})+\Phi_{\rm match}^{3-4}
1989: \right ].
1990: \eea
1991: \end{subequations}
1992: %
1993: \begin{figure*}
1994: \includegraphics[width=0.48\textwidth,clip=true]{fig13a.ps}
1995: \includegraphics[width=0.48\textwidth,clip=true]{fig13b.ps}\\
1996: \includegraphics[width=0.48\textwidth,clip=true]{fig13c.ps}
1997: \includegraphics[width=0.48\textwidth,clip=true]{fig13d.ps}
1998: \caption{\label{ne_modephase1} Phase differences between different
1999: mode-pair flux vectors, as defined by Eqs.~(\ref{phase_shifts1})--(\ref{phase_shifts3}).
2000: The data refer to the NE$_{+-}^{2:3}$ (upper left panel),
2001: NE$_{-+}^{2:3}$ (upper right panel), NE$_{00}^{2:3}$ (lower left
2002: panel), and EQ$_{+-}$ (lower right panel) runs. The
2003: dashed curves are the eN model predictions of
2004: Eqs.~(\ref{phase_shifts_insp}),(\ref{phase_shifts_RDe}).
2005: We denote with $t_{\rm peak}$ the time at which $I^{22}$ reaches its maximum.}
2006: \end{figure*}
2007:
2008: Even more intriguing, we find that for the unequal-mass simulations
2009: described above, the phase relations during the inspiral and RD
2010: are almost identical, regardless of spin orientations. This can be
2011: seen clearly in Fig.~\ref{ne_modephase1}, which plots $\cos\psi$
2012: during inspiral, merger and RD for the different runs.
2013: The colinearity of the flux vectors is clear during the inspiral
2014: phase, and the sinusoidal oscillations of the phases during RD agree
2015: well with the analytic predictions (plotted in dashed
2016: curves in Fig.~\ref{ne_modephase1}). Since the analytic models are most reliable
2017: during the inspiral and RD phases (but have more difficulty tracing
2018: the merger portion), we omit in Fig.~\ref{ne_modephase1} the
2019: transition region of $-10M\le (t-t_{\rm peak})\le 10 M$. The analytic phase
2020: relations during inspiral are determined by
2021: Eq.~(\ref{phase_shifts_insp}) and during ringdown by
2022: Eqs.~(\ref{phase_shifts_RD1e})--(\ref{phase_shifts_RD3e}). Here we use a
2023: $t_{\rm match}$ (and corresponding $\Phi_{\rm match}$) about $20M$ after $t_{\rm peak}$ to ensure that the
2024: multipole moments are truly dominated by the fundamental QNMs, and thus
2025: Eqs.~(\ref{phase_shifts_RD1e})--(\ref{phase_shifts_RD3e}) are valid.
2026: Note that the phase differences for EQ$_{+-}$ are particularly noisy
2027: since the amplitude of the $I^{33}$ moment is zero to leading order, and
2028: thus it is more difficult to extract a clear phase for that mode.
2029:
2030: The feature that is most difficult to explain from an analytic model
2031: alone (and is thus omitted from the eN curves in Fig.~\ref{ne_modephase1})
2032: is the roughly 180-degree jump in phase between $F_{\rm insp}^{22,33}$ and
2033: $F_{\rm insp}^{33,44}$, beginning around $20M$ before the
2034: peak. This appears to be a feature in all the
2035: unequal-mass runs examined, but preliminary results suggest that is
2036: less significant (i.e., a smaller phase shift) for more extreme-mass
2037: ratio systems, as we shall discuss in Appendix~\ref{app}.
2038: We are not able to explain it with the additional RD
2039: overtone modes described in Sec.~\ref{matching_RD}, but using
2040: slightly different RD matching points for the different multipoles
2041: may help explain the issue.
2042:
2043: \subsection{The anti-kick}\label{antikick}
2044:
2045: These flux amplitudes and phase relations can now be used to
2046: understand the amplitude of the kick and anti-kick, by which we mean
2047: the difference between the peak and the final recoil velocities (see
2048: Fig.~\ref{intro_fig} for an example). Throughout the
2049: inspiral phase, the amplitude and rotational frequency of the flux
2050: vectors in Eq.~(\ref{newtonian_flux}) are monotonically increasing,
2051: giving the familiar outward-spiraling trajectory for the velocity
2052: vector. Then, in the RD phase, the dominant frequencies are nearly
2053: constant while the amplitudes decay exponentially for each mode,
2054: giving an inward-spiral that decays like a damped harmonic oscillator around
2055: the final asymptotic recoil velocity.
2056:
2057: These trajectories in velocity
2058: space can be seen in Fig.~\ref{ne_spiralmodes1}, along with the
2059: instantaneous flux vectors from the competing mode-pairs. Clearly,
2060: even small changes in the mass ratios and spins orientations of the
2061: BHs can give a rather diverse selection of velocity
2062: trajectories. Note in particular the difference between the
2063: NE$_{-+}^{2:3}$ run, dominated by the $F^{22,33}$ flux and a large
2064: anti-kick, and the EQ$_{+-}$ run, which in contrast is dominated by
2065: the $F^{21,22}$ flux. We find that the EQ$_{+-}$ run has {\it no}
2066: anti-kick, which can be explained by the slowly rotating flux vector
2067: that does not spiral back inwards, but rather drifts off slowly
2068: towards infinity during the ringdown. The difference between these
2069: two runs can be explained entirely by examining the real part of
2070: their fundamental QNM frequencies $\sigma_{\ell m 0}$, which in
2071: turn determine the rotation rates of the flux vectors in
2072: Eq.~(\ref{flux_RD}): EQ$_{+-}$ is dominated by
2073: $\omega_{220}-\omega_{210}=0.08/M_{\rm f}$, a much
2074: slower frequency than $\omega_{330}-\omega_{220}=0.31/M_{\rm
2075: f}$, which causes the rapid inward-spiral of the NE$_{-+}^{2:3}$
2076: run.
2077:
2078: \begin{figure*}
2079: \includegraphics[width=0.48\textwidth,clip=true]{fig14a.ps}
2080: \includegraphics[width=0.48\textwidth,clip=true]{fig14b.ps}\\
2081: \includegraphics[width=0.48\textwidth,clip=true]{fig14c.ps}
2082: \includegraphics[width=0.48\textwidth,clip=true]{fig14d.ps}
2083: \caption{\label{ne_spiralmodes1} The recoil velocity vector evolving in
2084: the $v_x$--$v_y$ plane ({\it black solid curve}), along with the flux
2085: vectors due to the three mode pairs at each time interval along the
2086: velocity trajectory (same color
2087: scheme as Fig.~\ref{ne_modeamps}).
2088: The data refer to the NE$_{+-}^{2:3}$ ({\it upper left panel}),
2089: NE$_{-+}^{2:3}$ ({\it upper right panel}), NE$_{00}^{2:3}$ ({\it lower left
2090: panel}), and EQ$_{+-}$ ({\it lower right panel}) runs.
2091: We denote with the label {\it peak} the time at which $I^{22}$ reaches
2092: its maximum.}
2093: \end{figure*}
2094:
2095: To calculate the recoil velocity, we must integrate the linear
2096: momentum flux vectors in time. (For the initial velocity vector, we
2097: integrate the post-Newtonian approximation for the momentum flux from
2098: $t=-\infty$ to the beginning of the numerical simulation \cite{BQW}. This
2099: effectively sets the centers of the spiral curves in
2100: Fig.~\ref{ne_spiralmodes1} to correspond to the origin in velocity
2101: space.) We can get a reasonable analytic
2102: approximation by using Eqs.~(\ref{newtonian_flux}) and (\ref{flux_RD})
2103: for the inspiral and RD phases, respectively. In the adiabatic
2104: inspiral, the complex recoil velocity $v = v_x + iv_y$ can be written as
2105: %
2106: \begin{widetext}
2107: \beq\label{v_insp}
2108: v_{\rm insp}= \int_{-\infty}^{t_{\rm match}} F(t') \, dt' \simeq
2109: \frac{1}{i\omega_{\rm match}}\,F_{\rm match},
2110: \eeq
2111: %
2112: while for the RD portion we have
2113: %
2114:
2115: \bse\label{v_RD}
2116: \bea
2117: v_{\rm RD}(t) = \int_{t_{\rm match}}^{t} F(t')\,dt' &\simeq&
2118: \sum_{\ell m,\ell'm'}\,
2119: \frac{i\, F^{\ell m,\ell'm'}_{\rm match}}{\sigma_{\ell
2120: m0}-\sigma_{\ell'm'0}^{*}}\,
2121: \left[e^{-i(\sigma_{\ell m 0}-\sigma_{\ell' m' 0}^*)(t-t_{\rm
2122: match})}-1\right], \\
2123: v_{\rm f} \equiv v_{\rm RD}(t\to \infty) &\simeq &
2124: \sum_{\ell m,\ell'm'}\,
2125: \frac{-i}{\sigma_{\ell m0}-\sigma_{\ell'm'0}^{*}}\,F^{\ell
2126: m,\ell'm'}_{\rm match},
2127: \eea
2128: \ese
2129: summing the contributions from each pair of modes $(\ell m,\ell'm')$. Then the total velocity
2130: in each of the dominant mode pairs is given by
2131: %
2132: \begin{subequations}\label{vtot_modes}
2133: \bea
2134: \label{vtot_modes1}
2135: \int F^{21,22}(t')\, dt' &=& \frac{16}{45}\, \frac{\mu^2}{M} \,R_{\rm match}^3\, \omega_{\rm match}^5
2136: (2\delta m \, R_{\rm match}^2 \omega_{\rm match}+3\Delta^z)
2137: \left[1-\frac{i\omega_{\rm match}\, e^{i\phi^{21,22}_{\rm match}}}{\sigma_{210}-\sigma_{220}^{*}}
2138: \right], \\
2139: \label{vtot_modes2}
2140: \int F^{22,33}(t')\, dt' &=& -\frac{36}{7}\, \frac{\mu^2}{M}\, R_{\rm
2141: match}^5\,\omega_{\rm match}^6 (\delta m + \omega_{\rm match} \Sigma_{33}^z)
2142: \left[1-\frac{i\omega_{\rm match}\, e^{i\phi^{22,33}_{\rm match}}}{\sigma_{220}-\sigma_{330}^{*}}
2143: \right], \\
2144: \label{vtot_modes3}
2145: \int F^{33,44}(t')\, dt' &=& -\frac{64}{7}\, \frac{\mu^2}{M}\, (1-3\eta)\,
2146: R_{\rm match}^7\,\omega_{\rm match}^8\, (\delta m + \omega_{\rm match}
2147: \Sigma_{33}^z) \left[1-\frac{i\omega_{\rm match}\,
2148: e^{i\phi^{33,44}_{\rm match}}}
2149: {\sigma_{330}-\sigma_{440}^{*}}\right]\,.
2150: \eea
2151: \end{subequations}
2152: \end{widetext}
2153: %
2154: The phase $\phi^{21,22}_{\rm match}$ is defined as the angle made between the flux
2155: vector $F^{21,22}$ and the velocity vector $\mathbf{v}$ at
2156: the beginning of the ringdown (with other phases $\phi^{22,33}_{\rm match}$,
2157: $\phi^{33,44}_{\rm match}$ defined analogously). Because of the anomalous phase
2158: shifts and departure from adiabaticity at the transition from inspiral
2159: to ringdown,
2160: these angles are difficult to predict with an
2161: independent analytic model, but can be calculated easily from plots
2162: like Fig.~\ref{ne_spiralmodes1}. However, the accuracy of
2163: Eq.~(\ref{vtot_modes}) is limited both by the adiabaticity condition of
2164: Eq.~(\ref{v_insp}) as well as the accuracy of the spin-orbit
2165: corrections to the eN model (see Fig.~\ref{r_eff_mp}). Therefore, in analyzing the anti-kick in
2166: terms of RD modes, we find it more useful simply to integrate
2167: Eq.~(\ref{v_insp}) directly from the numerical data during the
2168: inspiral, and then attach the fundamental QNM terms from
2169: Eq.~(\ref{v_RD}) at the matching point $t_{\rm match} = t_{\rm peak}$.
2170:
2171: Given $v_{\rm match}$ at the end of the inspiral, we can use this
2172: quasi-analytic approach to predict the maximum and final recoil
2173: velocities ($v_{\rm max}$ and $v_{\rm f}$, respectively). These
2174: predictions are plotted as black dashed curves in
2175: Fig.~\ref{ne_kickmodes1}, to be compared with the solid black curves
2176: of the exact NR results. Within this
2177: context, we define the anti-kick magnitude as
2178: %
2179: \beq\label{f_ak}
2180: f_{\rm ak}\equiv \frac{v_{\rm f}-v_{\rm max}}{v_{\rm max}}
2181: \eeq
2182: %
2183: and the net ringdown contribution as
2184: %
2185: \beq\label{f_RD}
2186: f_{\rm RD}\equiv \frac{v_{\rm f}-v_{\rm match}}{v_{\rm match}}\, ,
2187: \eeq
2188: %
2189: where $v_{\rm max}$ and $v_{\rm f}$ are the (real-valued) velocity
2190: magnitudes calculated analytically from Eq.~(\ref{v_RD}).
2191:
2192: In the case of the NE$_{-+}^{2:3}$ run, where the recoil is almost entirely
2193: dominated by the $F^{22,33}$ flux, we find a large anti-kick
2194: with $f_{\rm ak}=-0.53$ and $f_{\rm RD}=-0.5$. On the other hand, for
2195: the NE$_{+-}^{2:3}$ run, as can be seen in Fig.~\ref{ne_kickmodes1},
2196: the net recoil velocity continues to increase after $t_{\rm
2197: match}=t_{\rm peak}$ before turning around for a small anti-kick of
2198: $f_{\rm ak}=-0.11$. The total effect of the ringdown phase is actually
2199: to increase the recoil with $f_{\rm RD}=0.68$. An intermediate effect
2200: is seen for the NE$_{00}^{2:3}$ run, with $f_{\rm ak}=-0.28$ and
2201: $f_{\rm RD}=-0.04$. However, as seen above in
2202: Fig.~\ref{ne_spiralmodes1}, for the EQ$_{+-}$ run, we see no
2203: anti-kick, with $f_{\rm ak}=-0.01$ and $f_{\rm RD}=0.58$.
2204:
2205: In general, we find the magnitude
2206: of the anti-kick is primarily dependent on the relative magnitudes of the
2207: $S^{21}$ and $I^{33}$ moments. When $S^{21}$ dominates (e.g.,
2208: when $\delta m$ and $\Delta^z$ add constructively), the ringdown
2209: rotation is slow and there is a small anti-kick, whereas a dominant
2210: $I^{33}$ mode (e.g., large $\delta m$ or no spins) gives a rapidly rotating ringdown flux and thus a large
2211: anti-kick. Furthermore, from Eq.~(\ref{newtonian_flux}), we see that for
2212: non-spinning BHs, both the $S^{21}$ and $I^{33}$ modes share the same
2213: mass and frequency scaling, so the relative size of the anti-kick should
2214: be roughly independent of mass ratio (see Appendix \ref{app} for a caveat).
2215:
2216: We would like a more quantitative picture of how these flux vectors add
2217: constructively and destructively to give the total recoil velocity to
2218: support the analytic estimates presented above.
2219: Using $\mathbf{v}=\int \mathbf{F} dt$, we can write
2220: %
2221: \begin{equation}
2222: \label{dvhat_dt}
2223: \frac{d}{dt}|\mathbf{v}| =
2224: \frac{d}{dt}(\hat{\mathbf{v}}\cdot\mathbf{v}) = \hat{\mathbf{v}}\cdot
2225: \mathbf{F}\,,
2226: \end{equation}
2227: %
2228: where $\hat{\mathbf{v}} \cdot \hat{\mathbf{v}} = 1$.
2229: Breaking $\mathbf{F}$ up into the contributions of the
2230: dominant modes as above, and then integrating in time gives
2231: %
2232: \bse
2233: \begin{eqnarray}
2234: \label{vdotf}
2235: v^{21,22}&=&\int \hat{\mathbf{v}}\cdot\mathbf{F}^{21,22}\,dt \, ,\\
2236: v^{22,33}&=&\int \hat{\mathbf{v}}\cdot\mathbf{F}^{22,33}\,dt \, ,\\
2237: v^{33,44}&=&\int \hat{\mathbf{v}}\cdot\mathbf{F}^{33,44}\,dt \, ,
2238: \end{eqnarray}
2239: \ese
2240: %
2241: which add linearly to give to total recoil velocity:
2242: %
2243: \begin{equation}
2244: |\mathbf{v}|=v^{21,22}+v^{22,33}+v^{33,44}.
2245: \end{equation}
2246: %
2247: Note that with these definitions, the $v^{\ell
2248: m,\ell' m'}$ are all real, but can be positive or negative.
2249: These different velocities are plotted in Fig.~\ref{ne_kickmodes1},
2250: with the same color scheme as in Figs.~\ref{ne_modeamps} and \ref{ne_spiralmodes1},
2251: along with the total recoil velocity in
2252: solid black curves. Also shown in Fig.~\ref{ne_kickmodes1} is the
2253: velocity $v^{32,33}$ (dashed blue curves), defined analogously to
2254: Eq.~(\ref{vdotf}) for the $S^{32}I^{33*}$ flux terms. The
2255: small contribution from this mode pair further justifies our focus
2256: on the more dominant pairs of Eq.~(\ref{flux_approx}) and
2257: Fig.~\ref{flux_lm}.
2258:
2259: \begin{figure*}
2260: \includegraphics[width=0.48\textwidth,clip=true]{fig15a.ps}
2261: \includegraphics[width=0.48\textwidth,clip=true]{fig15b.ps} \\
2262: \includegraphics[width=0.48\textwidth,clip=true]{fig15c.ps}
2263: \includegraphics[width=0.48\textwidth,clip=true]{fig15d.ps}
2264: \caption{\label{ne_kickmodes1} Relative contributions to the total
2265: recoil velocity from the different multipole
2266: mode-pairs. $I^{22}I^{33*}$ (red curve) is the dominant mode for
2267: unequal-mass
2268: binary systems, while $S^{21}I^{22*}$ (blue curve) dominates for spinning,
2269: equal-mass binary systems. Also plotted is the
2270: contribution from the $S^{32}I^{33*}$ flux terms (blue dashed
2271: curve), demonstrating its very small contribution to the total
2272: recoil velocity. For $t>t_{\rm match}=t_{\rm peak}$, we include the
2273: quasi-analytic prediction for $v_{\rm RD}(t)$ (black dashed
2274: curves), based on the fundamental RD modes from Eq.~(\ref{v_RD}).
2275: The data refer to the NE$_{+-}^{2:3}$ (upper left panel),
2276: NE$_{-+}^{2:3}$ (upper right panel), NE$_{00}^{2:3}$ (lower left
2277: panel), and EQ$_{+-}$ (lower right panel) runs.
2278: We denote with $t_{\rm peak}$ the time at which $I^{22}$ reaches its maximum.}
2279: \end{figure*}
2280:
2281: In the NE$_{-+}^{2:3}$ run, where the modal analysis shows the
2282: $F^{21,22}$ and $F^{33,44}$ flux terms canceling out, we see that the
2283: total recoil velocity (black curves in
2284: Fig.~\ref{ne_kickmodes1}) is almost entirely dominated by the
2285: $F^{22,33}$ flux (red curves). On the other hand, for the
2286: NE$_{+-}^{2:3}$ run, the $F^{21,22}$ flux is much stronger, adding
2287: destructively with the $F^{22,33}$ flux during the RD. This has the effect of both
2288: increasing the peak velocity and also decreasing the relative strength
2289: of the anti-kick, due to the slow rotation frequency of the $F^{21,22}$
2290: flux during ringdown, as described above. As expected, the
2291: NE$_{00}^{2:3}$ run displays behavior intermediate between these two
2292: extremes. The EQ$_{+-}$ run, however, is entirely dominated by the
2293: $F^{21,22}$ flux, and thus experiences {\it no} anti-kick, but
2294: rather drifts off slowly in a nearly constant direction, as seen in the
2295: bottom-right panel of Fig.~\ref{ne_spiralmodes1}.
2296:
2297: \subsection{Application to non-planar kicks}\label{planarspins}
2298:
2299: One of the most remarkable results from the recent renaissance in
2300: numerical relativity was the prediction of extremely large kicks from
2301: equal-mass BHs with spins pointing opposite to each other and normal
2302: to the orbital angular momentum, producing a recoil out of the orbital
2303: plane \cite{recoilRI,Bigrecoil,recoilFAU,recoilJena2}. While this configuration can
2304: produce recoils of nearly $4000$ km/sec, the analogous non-precessing
2305: configuation (EQ$_{+-}$ run in this paper) gives a kick of only $\sim
2306: 500$ km/sec in the case of maximal spin
2307: \cite{recoilAEI,recoilPSU,recoilGoddard}. The
2308: multipole analysis tools developed above can be used for understanding
2309: and explaining this remarkable difference.
2310:
2311: First, we should note that leading-order PN
2312: estimates of the linear momentum flux during inspiral suggest that the
2313: discrepancy should be less than a factor of two. For example, Eq.~(3.31b)
2314: of Kidder \cite{LK} gives the spin-orbit contribution to the momentum
2315: flux for circular, Keplerian orbits as
2316: %\begin{widetext}
2317: \beq
2318: \mathbf{F}_{\rm SO} = \frac{16}{15}\mu^2\, M\,\frac{\omega^2}{R^3}
2319: [\hat{\mathbf{n}}\times \mathbf{\Delta}+
2320: (\hat{\mathbf{n}}\times\hat{\mathbf{v}})
2321: (\hat{\mathbf{v}}\cdot\mathbf{\Delta})],
2322: \eeq
2323: %\end{widetext}
2324: with $\hat{\mathbf{n}}$ and $\hat{\mathbf{v}}$ being the normalized
2325: separation and velocity vectors, respectively. For spins parallel to
2326: the orbital angular momentum, the term in square brackets has magnitude
2327: $\Delta^z$, while for planar spins, it is
2328: $2\Delta^p\sin\phi_{\Delta}$, where
2329: $\phi_{\Delta}$ is the angle between $\mathbf{\Delta}$ and
2330: $\mathbf{n}$, and $\Delta^p$ is the
2331: magnitude of $\mathbf{\Delta}$ in the orbital plane.
2332:
2333: Not surprisingly, we get the exact same results from the multipole
2334: analysis of Eqs.~(\ref{dPdt_radmoments}), (\ref{dPdtz_radmoments}),
2335: (\ref{spinorbit_moments}), and one new multipole moment:
2336: \beq\label{s22_SO}
2337: S^{22}_{\rm SO} = 4i\sqrt{\frac{2\pi}{5}}\,\eta\, R\, \omega^3 \,e^{-i\phi}
2338: (\Delta^x-i\Delta^y),
2339: \eeq
2340: while on the other hand, the $S^{21}$ mode is zero for the planar spin
2341: configuration. Combining these equations, we get
2342: \begin{widetext}
2343: \beq\label{parallel_flux}
2344: F_x+iF_y \approx \frac{1}{336\pi}(-14iS^{21}I^{22*})=
2345: \frac{16}{15}i \frac{\mu^2}{M}\, R^3\, \omega^6\, \Delta^z\, e^{i\phi}
2346: \eeq
2347: and using Eq.~(\ref{dPdtz_radmoments}) we obtain
2348: \bea\label{planar_flux}
2349: F_z\approx \frac{1}{336\pi}[-28\Im(I^{22}S^{22*})] &=&
2350: \frac{32}{15} \frac{\mu^2}{M}\, R^3\, \omega^6\, (\Delta^x\sin\phi
2351: -\Delta^y\cos\phi) \nonumber\\
2352: &=& \frac{32}{15} \frac{\mu^2}{M}\, R^3\, \omega^6\, \Delta^p\sin\phi_\Delta\, ,
2353: \eea
2354: \end{widetext}
2355: where $\phi$ is the orbital phase of the binary.
2356: So in both paradigms, we see that, when maximizing over
2357: $\sin\phi_{\Delta}$, the planar-spin orientation should result in a
2358: recoil twice as large as the parallel-spin case, leaving a factor
2359: of roughly 4 difference unexplained.
2360:
2361: From Eqs.~(\ref{parallel_flux}),(\ref{planar_flux}) we see that the
2362: only relevant modes involved should be $I^{22}$, $S^{21}$, and
2363: $S^{22}$ (for these equal-mass systems the momentum flux is
2364: dominated by a single mode pair, responsible for $\gtrsim 95$\% of the
2365: final recoil value). In the left panel of Fig.~\ref{I22_planar} we plot the
2366: amplitude of $I^{22}$
2367: from the EQ$_{+-}$ simulation, along with that of a planar-spin
2368: simulation EQ$_{\rm planar}$. All other binary parameters and the
2369: initial conditions are the same. Remarkably, the mass-quadrupole
2370: moments $I^{22}$ are nearly identical (and dominant) in both runs, and
2371: this suggests that the energy and angular momentum fluxes are the same
2372: [see Eqs.~(\ref{dEdt}),(\ref{dJzdt})].
2373: This is in fact quite reasonable since the total spin of the system is zero in both
2374: cases. However, we see in the right-hand panel of
2375: Fig.~\ref{I22_planar} that the peak amplitude of the $S^{22}$ mode is a
2376: factor of $\sim 2.5$ greater than that of the $S^{21}$ mode from the
2377: EQ$_{\rm planar}$ and EQ$_{+-}$ runs, respectively.
2378:
2379: Yet Eqs.~(\ref{spinorbit_moments}),(\ref{s22_SO}) suggest that these
2380: two modes should have exactly the same magnitudes, at least during the
2381: inspiral phase, and presumably during the RD as well, since the RD
2382: amplitudes are completely determined by the mode amplitudes at the
2383: matching point. It appears from Fig.~\ref{I22_planar} that $S^{22}$
2384: and $S^{21}$ do in fact have the same amplitude at early times, but
2385: the relatively noisy data and short duration of the simulations make
2386: it impossible to say for certain. If this is the case, one possible
2387: explanation for the sudden remarkable increase in the amplitude of
2388: $S^{22}$ might be mode-coupling with the dominant $I^{22}$ mode, as the
2389: inspiral phase begins to transition to the RD phase. This coupling is
2390: analogous to that of $S^{32}$ and $I^{22}$ described above in
2391: Sec.~\ref{RD_phase}, an effect that is apparently only
2392: important between modes with the same $m$-number\cite{BCW,BCP}. We
2393: hope to address this question in the future with longer
2394: simulations to confirm the agreement at early times, as well as other
2395: spin configurations that should enhance specific multipole modes and
2396: may help identify other similar cases of mode amplification.
2397:
2398: \begin{figure*}
2399: \includegraphics[width=0.48\textwidth,clip=true]{fig16a.ps}
2400: \includegraphics[width=0.48\textwidth,clip=true]{fig16b.ps}
2401: \caption{\label{I22_planar} {\it left panel:} Comparison of the
2402: multipole amplitudes
2403: $I^{22}$ for the two different equal-mass simulations: EQ$_{\rm
2404: planar}$ (solid line) and EQ$_{+-}$ (dashed line). {\it right
2405: panel:} The $S^{22}$ amplitude from the planar-spins run (EQ$_{\rm
2406: planar}$, solid line) and the $S^{21}$ amplitude from the
2407: parallel-spins run (EQ$_{+-}$, dashed line). We denote with $t_{\rm peak}$
2408: the time at which $I^{22}$ reaches its maximum.}
2409: \end{figure*}
2410:
2411: \begin{figure*}
2412: \includegraphics[width=0.48\textwidth,clip=true]{fig17a.ps}
2413: \includegraphics[width=0.48\textwidth,clip=true]{fig17b.ps}
2414: \caption{\label{planar_recoil} {\it left panel:} Comparison of the
2415: linear momentum flux for the two different equal-mass simulations: EQ$_{\rm
2416: planar}$ (solid line) and EQ$_{+-}$ (dashed line). {\it right
2417: panel:} The total recoil velocity from the planar-spins run (EQ$_{\rm
2418: planar}$, solid line) and the parallel-spins run (EQ$_{+-}$,
2419: dashed line). We denote with $t_{\rm peak}$
2420: the time at which $I^{22}$ reaches its maximum.}
2421: \end{figure*}
2422:
2423: Lastly, from the ringdown contribution to the velocity
2424: [Eqs.~(\ref{v_RD}),(\ref{vtot_modes})], we can understand another
2425: difference between the planar- and parallel-spin
2426: orientations.
2427: Instead of having two different RD frequencies
2428: $\sigma_{210}$ and $\sigma_{220}$ combine to give a slowly rotating
2429: flux vector, for the planar-spin case, we have two {\it identical} RD
2430: frequencies for $I^{22}$ and $S^{22}$ in Eq.~(\ref{planar_flux}),
2431: giving precisely zero rotation to the RD
2432: flux. Furthermore, as the spin vector $\mathbf{\Delta}$ is precessing
2433: faster and faster in a positive direction around the orbital angular
2434: momentum vector, even during the inspiral the two modes $I^{22}$ and
2435: $S^{22}$ become nearly locked in phase, producing a relatively
2436: long-duration burst of linear momentum flux in a single direction
2437: during the merger phase. Combined, these effects essentially
2438: straighten out the spiral curve in the lower-right panel of
2439: Fig.~\ref{ne_spiralmodes1}, providing another factor of $\sim 1.6$
2440: of increased recoil velocity for planar spins.
2441:
2442: In Fig.~\ref{planar_recoil} we show the combination of the above
2443: effects. In the left panel, we plot the linear momentum flux from
2444: Eqs.~(\ref{parallel_flux}),(\ref{planar_flux}), showing the factor of
2445: two increase predicted by the Kidder formula and our
2446: Eqs.~(\ref{dPdt_radmoments}), (\ref{dPdtz_radmoments}), along with the factor of
2447: $2.5$ increase in the amplitude of $S^{22}$ relative to $S^{21}$. In the
2448: right panel, we plot the recoil velocity for both runs,
2449: which includes the effect of flux rotation during the merger and
2450: inspiral phases, accounting for another factor of $\sim 1.6$, giving a
2451: total discrepancy of $v({\rm EQ}_{\rm planar})/v({\rm EQ}_{+-})\approx
2452: 2.5\times 2\times 1.6 = 8$.
2453:
2454: \section{Discussion}
2455: \label{discussion}
2456:
2457: In this paper we analysed several numerical simulations of binary BH
2458: coalescence, focusing on the physics of the recoil. We developed tools,
2459: based on the multipolar expansion~\cite{KT,BD,BDS,BS,JS}, that can be
2460: used as a diagnostic of the numerical results, and
2461: understand how the recoil velocity evolves during the inspiral,
2462: merger, and ringdown phases of the coalescence.
2463:
2464: We wrote explicit expressions for the linear momentum flux expressed
2465: in terms of radiative multipole moments through $\ell=4$, valid for
2466: generic spinning, precessing BH binary systems. We found that
2467: these formulae are sufficient to obtain the total recoil velocity with
2468: high accuracy. By comparing the amplitudes of the different
2469: multipole moments, we found that in the case of non-precessing
2470: spins--and thus a recoil in the orbital plane--only three pairs
2471: of modes contribute to most of the linear momentum flux,
2472: notably ${S^{21}I^{22*}}$, ${I^{22}I^{33*}}$ and ${I^{33}I^{44*}}$.
2473: Those modes account for the total recoil with an accuracy on the order
2474: of $\sim 5-10\%$ throughout the simulations. (see Figs.~\ref{v_tot1},
2475: \ref{v_tot2}).
2476:
2477: The way in which the contribution from these three pairs of modes
2478: builds up is not trivial, since not only the relative amplitudes, but
2479: especially, the relative phases are also quite important. We found that the
2480: relative phases between the three mode-pairs are nearly constant during
2481: the inspiral phase, but start diverging at the onset of the
2482: transition from inspiral to RD (see Fig.~\ref{ne_modephase1}).
2483: The late-time evolution can be described reasonably well with
2484: analytic formula obtained expressing the mode-pairs in terms
2485: of fundamental QNMs of a Kerr BH. We showed that it is the relative
2486: magnitude of the current-quadrupole mode $S^{21}$ and the mass-octupole
2487: mode $I^{33}$, together with the differences of the
2488: QNM fundamental frequencies for each of the dominant modes,
2489: that determine the difference between the recoil at the
2490: peak of the linear momentum flux, and the final recoil velocity, i.e.,
2491: the magnitude of the anti-kick.
2492:
2493: With the final goal of improving analytic PN models, we also explored
2494: whether simple modifications of the Newtonian formula for the
2495: linear momentum flux allow us to match the numerical
2496: results all along the binary evolution. We found that,
2497: if we treat the binary radial separation in the Newtonian multipole
2498: modes (\ref{S21})--(\ref{I33}) with an effective radius, which is
2499: computed from the numerical simulations assuming that each multipole
2500: mode is described by a dominant frequency (see
2501: Fig.~\ref{omega_modes}), the leading Newtonian modes reproduce quite
2502: well the numerical ones (see Figs.~\ref{compare_ring}, \ref{kick_match})
2503: up to the end of the inspiral phase. We also found, confirming the results
2504: in Ref.~\cite{BCP}, that a superposition of three
2505: QNMs can fit the numerical waveforms very well from the peak
2506: of the radiation through the RD phase.
2507:
2508: The tools developed in this paper will be employed to
2509: improve current analytic predictions for the recoil
2510: velocity~\cite{BQW,DG} using PN analytic models
2511: ~\cite{LB} and the EOB approach~\cite{BD1,DJS,DIS98,BD2,BCD,
2512: EOB4PN}. An accurate, fully analytic description of the recoil velocity
2513: can be adopted in fast Monte Carlo
2514: simulations to predict recoil distributions from
2515: BH mergers with uncertainties smaller than in Ref.~\cite{SB}.
2516: Those recoil distributions can in turn be included in simulations
2517: of hierarchical merger models of supermassive BHs providing
2518: more robust predictions for LISA.
2519:
2520: \acknowledgments
2521: We thank Emanuele
2522: Berti for providing us with tabulated data for the Kerr QNM
2523: frequencies. We would like to thank the anonymous referee for their
2524: careful and constructive comments.
2525:
2526: A.B.\ and J.D.S.\ acknowledge support from NSF grant PHYS-0603762, and
2527: A.B.\ was also supported by the Alfred P. Sloan Foundation. The work at
2528: Goddard was supported in part by NASA grant 05-BEFS-05-0044 and
2529: 06-BEFS06-19. The simulations were carried out using
2530: Project Columbia at the NASA Advanced Supercomputing Division (Ames
2531: Research Center) and at
2532: the NASA Center for Computational Sciences (Goddard
2533: Space Flight Center). B.J.K.\ was supported by the NASA Postdoctoral Program at
2534: the Oak Ridge Associated Universities. S.T.M.\ was supported in part
2535: by the Leon A.\ Herreid Graduate Fellowship.
2536:
2537:
2538: \appendix
2539:
2540: \section{Results from 1:4 mass ratio}
2541: \label{app}
2542:
2543: In addition to the simulations presented in the main body of this
2544: paper, we have also recently analyzed a non-spinning
2545: system with mass ratio 1:4 ($\eta = 0.16$). The
2546: results of this analysis are
2547: presented briefly in this appendix, as well as in
2548: Tables~\ref{table:idparams}--\ref{table:QNM_freqs} (labeled
2549: appropriately as NE$_{00}^{1:4}$). More details can be
2550: found in Ref.~\cite{EOB4PN}.
2551:
2552: \begin{figure}
2553: \includegraphics[width=0.48\textwidth,clip=true]{fig18.ps}
2554: \caption{\label{app_fig1} Flux amplitudes from the NE$_{00}^{1:4}$
2555: run, as in Fig.~\ref{flux_lm}.}
2556: \end{figure}
2557:
2558: \begin{figure}
2559: \includegraphics[width=0.48\textwidth,clip=true]{fig19.ps}
2560: \caption{\label{app_fig2} Phase differences from the NE$_{00}^{1:4}$
2561: run, as in Fig.~\ref{ne_modephase1}.}
2562: \end{figure}
2563:
2564: \begin{figure}
2565: \includegraphics[width=0.48\textwidth,clip=true]{fig20.ps}
2566: \caption{\label{app_fig3} Relative contributions to the total
2567: recoil velocity from the different multipole mode-pairs for the NE$_{00}^{1:4}$
2568: run, as in Fig.~\ref{ne_kickmodes1}.}
2569: \end{figure}
2570:
2571: In Fig.~\ref{app_fig1} we show the flux amplitudes from the different
2572: modes, as in Fig.~\ref{flux_lm} above. We find the relative amplitudes
2573: almost identical to those of the NE$_{00}^{1:2}$ run, with a slightly
2574: stronger contribution from the $I^{44}$ mode, as expected from
2575: Eq.~(\ref{I44}), which predicts a maximum in the $I^{44}$ amplitude for
2576: $\eta=0.167$.
2577:
2578: In Fig.~\ref{app_fig2} we plot the phase relations between the
2579: different flux vectors, defined in
2580: Eqs.~(\ref{phase_shifts1})-(\ref{phase_shifts3}). As anticipated in
2581: Sec.~\ref{transition} above, we find a smaller phase shift in
2582: $\psi^{3-4}$ at
2583: the transition from inspiral to ringdown for this more extreme
2584: mass-ratio system. The other phases appear to behave as expected.
2585:
2586: Lastly, in Fig.~\ref{app_fig3}, we show the total recoil velocity
2587: along with the relative contributions from the dominant modes for the
2588: NE$_{00}^{1:4}$ run. Again, the qualitative behavior is quite similar
2589: to the NE$_{00}^{2:3}$ and NE$_{00}^{1:2}$ runs, but we can now
2590: identify a clear trend of a smaller anti-kick for smaller values of
2591: $\eta$. As mentioned above in Section \ref{antikick}, the
2592: amplitude of the anti-kick is most strongly dependent on the relative
2593: amplitudes of the $S^{21}$ and $I^{33}$ modes, but for non-spinning BH
2594: binaries, these modes both scale the same with mass ratio. However,
2595: the amplitude of the $I^{22}$ mode decreases with decreasing
2596: $\eta$, while the amplitude of $I^{44}$ increases with
2597: decreasing $\eta$, at least over the range considered here. Thus the
2598: amplitude of the $F^{33,44}$ flux increases relative to the
2599: $F^{22,33}$ flux for more extreme mass ratios. From
2600: Figs.~\ref{ne_kickmodes1} and \ref{app_fig3}, we see that the
2601: $F^{22,33}$ flux dominates the anti-kick, while the
2602: $F^{33,44}$ flux contributes almost nothing to it, so by
2603: increasing the relative amplitude of $F^{33,44}$, we have effectively
2604: decreased the size of the anti-kick.
2605:
2606: \renewcommand{\prd}{\emph{Phys. Rev. D}}
2607:
2608: \begin{thebibliography}{999}
2609: \frenchspacing
2610: \bibitem{FP} F. Pretorius, Phys. Rev. Lett. {\bf 95}, 121101 (2005).
2611: %
2612: \bibitem{CLMZ} M. Campanelli, C.O. Lousto, P. Marronetti, and Y. Zlochower, Phys. Rev. Lett. {\bf 96}, 111101 (2006).
2613: %
2614: \bibitem{Bakeretal1} J. Baker, J. Centrella, D. Choi, M. Koppitz, and J. van Meter, Phys. Rev. Lett. {\bf 96}, 111102 (2006).
2615: %
2616: \bibitem{sperhake} U. Sperhake, Phys. Rev. D \textbf{76}, 104015 (2007).
2617: %
2618: \bibitem{gonzalez} J. Gonz\'{a}lez, U. Sperhake, B. Br\"{u}gmann,
2619: M. Hannam, and S. Husa, Phys. Rev. Lett. \textbf{98}, 091101
2620: (2007).
2621: %
2622: \bibitem{szilagyi} B. Szilagyi, D. Pollney, L. Rezzolla, J. Thornburg
2623: and J. Winicour, Class. Quantum Grav. {\bf 24}, S275 (2007).
2624: %
2625: \bibitem{HSL} F. Herrmann, I. Hinder, D. Shoemaker, and P. Laguna,
2626: Class. Quantum Grav. {\bf 24}, S33 (2007).
2627: %
2628: \bibitem{recoil} J.G. Baker, J. Centrella, D. Choi, M. Koppitz, J.R. van Meter, and M.C Miller, Astrophys. J {\bf 653}, L93 (2006).
2629: %
2630: \bibitem{recoilJena} J.A. Gonzalez, U. Sperhake, B. Br\"{u}gmann, M. Hannam, and S. Husa, Phys.Rev.Lett.{\bf 98},091101 (2007).
2631: %
2632: \bibitem{recoilPSU} F. Herrmann, I. Hinder, D. Shoemaker, P. Laguna,
2633: and R.A. Matzner, Astrophys. J {\bf 661}, 430 (2007).
2634: %
2635: \bibitem{recoilAEI}
2636: M. Koppitz, D. Pollney, C. Reisswig, L. Rezzolla, J. Thornburg,
2637: P. Diener, and E. Schnetter, Phys. Rev. Lett. {\bf 99},041102 (2007).
2638: %
2639: \bibitem{recoilGoddard} J.G. Baker, W.D. Boggs, J.Centrella,
2640: B.J. Kelly, S.T. McWilliams, M.C. Miller, and J.R. van Meter,
2641: Astrophys. J {\bf 668}, 1140 (2007).
2642: %
2643: \bibitem{Bigrecoil} J.A. Gonzalez, M.D. Hannam, U. Sperhake,
2644: B. Brugmann, and S. Husa, Phys. Rev. Lett. {\bf 98}, 231101 (2007).
2645: %
2646: \bibitem{recoilFAU} W. Tichy and P. Marronetti, Phys. Rev. D {\bf 76},
2647: 061502 (2007).
2648: %
2649: \bibitem{recoilJena2} B. Br\"{u}gmann, J.A. Gonzalez, M. Hannam,
2650: S. Husa, and U. Sperhake, \url{arXiv:0707.0135}.
2651: %
2652: \bibitem{recoilRI} M. Campanelli, C.O. Lousto, Y. Zlochower, and D. Merritt, Astrophys. J. Lett. {\bf 659}, L5 (2007).
2653: %
2654: \bibitem{MTW} C.W. Misner, K.S. Thorne, and J.A. Wheeler, {\it Gravitation}, W.H. Freeman, San Francisco, 1973.
2655: %
2656: \bibitem{BCL} J. Baker, M. Campanelli, C. Lousto, Phys. Rev. D {\bf 65}, 044001 (2002).
2657: %
2658: \bibitem{Bonnor} W.B. Bonnor and M.A. Rotenberg, Proc. R. Soc. London
2659: A {\bf 265}, 109 (1961).
2660: %
2661: \bibitem{Peres} A. Peres, Phys. Rev. {\bf 128}, 2471 (1962).
2662: %
2663: \bibitem{Bekenstein} J.D. Bekenstein, Astrophys. J. {\bf 183}, 657 (1973).
2664: %
2665: \bibitem{Cooperstock} F.I. Cooperstock, Astrophys. J. {\bf 213}, 250 (1977).
2666: %
2667: \bibitem{Fitchett} M.J. Fitchett, Mon. Not. R. Astr. Soc. {\bf 203}, 1049 (1983);
2668: M.J. Fitchett and S. Detweiler, Mon. Not. R. Astr. Soc. {\bf 211}, 933 (1984).
2669: %
2670: \bibitem{HM} M.G. Haehnelt, Mon. Not. R. Astr. Soc. {\bf 269}, 199 (1994);
2671: K. Menou, Z. Haiman, and V.K. Narayanan, Astrophys.J. {\bf 558}, 535 (2001);
2672: M. Volonteri, F. Haardt, and P. Madau, Astrophys.J. {\bf 582}, 559 (2003).
2673: %
2674: \bibitem{Merrittetal} D. Merritt, M. Milosavljevic, M. Favata, and S.A. Hughes, Astrophys. J. {\bf 607}, L9 (2004).
2675: %
2676: \bibitem{Volonteri07} M. Volonteri, Astrophys. J {\bf 663}, L5 (2007).
2677: %
2678: \bibitem{Schnittman07b} J.D. Schnittman, Astrophys. J {\bf 667}, L133 (2007).
2679: %
2680: \bibitem{HDR} M.G. Haehnelt, M.B. Davies, and M.J. Rees,
2681: Mon. Not. R. Astr. Soc. {\bf 366}, L22 (2006).
2682: %
2683: \bibitem{Bonning} E.W. Bonning and G.A. Shields, Astrophys. J {\bf
2684: 666}, L13 (2007).
2685: %
2686: \bibitem{Boylan_Kolchin} M. Boylan-Kolchin, C.-P. Ma, and E. Quataert,
2687: Astrophys. J. Lett. {\bf 613}, L37 (2004).
2688: %
2689: \bibitem{Lauer} T. R. Lauer, et al., Astrophys. J {\bf 662}, 808 (2007).
2690: %
2691: \bibitem{SB} J.D. Schnittman and A. Buonanno, Astrophys.J. {\bf 662},
2692: L63 (2007).
2693: %
2694: \bibitem{Schnittman04} J.D. Schnittman, Phys. Rev. D {\bf 70}, 124020
2695: (2004).
2696: %
2697: \bibitem{bogdanovic} T. Bogdanovic, C.S. Reynolds, and M.C. Miller,
2698: Astroph. J {\bf 661}, L147 (2007).
2699: %
2700: \bibitem{loeb} A. Loeb, Phys. Rev. Lett. {\bf 99}, 041103 (2007).
2701: %
2702: \bibitem{LB} See, e.g., L. Blanchet, Living Rev. Rel. {\bf 5}, 3 (2002).
2703: %
2704: \bibitem{DIS98} T. Damour, B.R. Iyer and B.S. Sathyaprakash, Phys. Rev. D {\bf 57}, 885 (1998).
2705: %
2706: \bibitem{BD1} A. Buonanno and T. Damour, Phys. Rev. D {\bf 59}, 084006 (1999).
2707: %
2708: \bibitem{BD2} A. Buonanno and T. Damour, Phys. Rev. D {\bf 62}, 064015 (2000).
2709: %
2710: \bibitem{DJS} T. Damour, P. Jaranowski, and G. Sch\"{a}fer, Phys. Rev. D {\bf 62}, 084011 (2000).
2711: %
2712: \bibitem{DJS2} T. Damour, P. Jaranowski, and G. Sch\"{a}fer, Phys. Rev. D {\bf 62}, 044024 (2000).
2713: %
2714: \bibitem{BCD} A. Buonanno, Y. Chen, and T. Damour, Phys. Rev. D {\bf 74}, 104005 (2006).
2715: %
2716: \bibitem{AW} A. Wiseman, Phys. Rev. D {\bf 46}, 1517 (1992).
2717: %
2718: \bibitem{LK} L. Kidder, Phys. Rev. D {\bf 52}, 821 (1995).
2719: %
2720: \bibitem{Favataetal} M. Favata et al., Astrophys. J. {\bf 607}, L5 (2004).
2721: %
2722: \bibitem{BQW} L. Blanchet, M.S.S. Qusailah, and C.M. Will, Astrophys. J. {\bf 635}, 508 (2006).
2723: %
2724: \bibitem{DG} T.\ Damour and A.\ Gopakumar, Phys.\ Rev.\ D {\bf 73}, 124006 (2006).
2725: %
2726: \bibitem{Sopuerta} C.F. Sopuerta, N. Yunes, and P. Laguna, Astrophys. J. Lett. {\bf 656}, L9 (2007).
2727: %
2728: \bibitem{CLA} R.H. Price and J. Pullin, Phys. Rev. Lett. {\bf 72}, 3297 (1994).
2729: %
2730: \bibitem{KT} K.S. Thorne, Rev. Mod. Phys. {\bf 52}, 299 (1980).
2731: %
2732: \bibitem{BD} L. Blanchet and T. Damour, Ann. Inst. H. Poincar\'e {\bf 50}, 377 (1989).
2733: %
2734: \bibitem{BS} L. Blanchet and G. Sch\"{a}fer, Mon. Not. R. Astr. Soc. {\bf 239}, 845 (1989).
2735: %
2736: \bibitem{BDS} L. Blanchet, T. Damour, and G. Sch\"{a}fer, Mon. Not. R. Astr. Soc. {\bf 242}, 289 (1990).
2737: %
2738: \bibitem{JS} W. Junker and G. Sch\"{a}fer, Mon. Not. R. Astr. Soc. {\bf 254}, 146 (1992).
2739: %
2740: \bibitem{RD} C.V. Vishveshwara, Nature {\bf 227}, 936 (1970);
2741: M. Davis, R. Ruffini, W.H. Press and R.H. Price, Phys. Rev. Lett. {\bf 27}, 1466 (1971);
2742: W. Press, Astrophys J. Letters {\bf 170}, L105 (1971);
2743: M. Davis, R. Ruffini and J. Tiomno, Phys. Rev. D {\bf 5}, 2932 (1972);
2744: S. Chandrasekhar and S. Detweiler, Proc. R. Soc. Lond. A {\bf 344},
2745: 441 (1975).
2746: %
2747: \bibitem{goldberg} J.N. Goldberg, A.J. Macfarlane, E.T. Newman,
2748: F. Rohrlich, and E.C.G. Sundarshan, J. Math.\ Phys.\ {\bf 8}, 2155
2749: (1967).
2750: %
2751: \bibitem{wiaux} Y. Wiaux, L. Jacques, P. Vandergheynst, \url{astro-ph/0508514}.
2752: %
2753: \bibitem{Brandt97b} S. Brandt and B. Br\"{u}gmann, Phys. Rev. Lett. {\bf 78}, 3606 (1997).
2754: %
2755: \bibitem{Brown:2004ma} J.D. Brown and L.L. Lowe, J. Comput. Phys. {\bf 209}, 582 (2005).
2756: %
2757: \bibitem{Bowen80} J. Bowen and J.W. York, Phys. Rev. D {\bf 21}, 2047 (1980).
2758: %
2759: \bibitem{IRM} D. Christodoulou, Phys. Rev. Lett. {\bf 25}, 1596 (1970); D. Christodoulou
2760: and R. Ruffini, Phys. Rev. D {\bf 4}, 3552 (1971).
2761: %
2762: \bibitem{Imbiriba:2004tp} B. Imbiriba, J.G. Baker, D.-I. Choi, J. Centrella, D.R. Fiske, J.D. Brown, J.R. van Meter,
2763: and K. Olson, Phys. Rev. D {\bf 70}, 124025 (2004).
2764: %
2765: \bibitem{Huebner99} P. H\"{u}bner, Class. Quantum Grav. {\bf 16}, 2823 (1999).
2766: %
2767: \bibitem{Duez:2004uh} M.D. Duez, S.L. Shapiro, and H.-J. Yo, Phys. Rev. D {\bf 69}, 104016 (2004).
2768: %
2769: \bibitem{vanMeter:2006vi} J. van Meter, J.G. Baker, M. Koppitz, and D.-I. Choi, Phys. Rev D {\bf 73}, 124011 (2006).
2770: %
2771: \bibitem{Baker:2005xe} J.G. Baker and J. van Meter, Phys. Rev. D {\bf 72}, 104010 (2005).
2772: %
2773: \bibitem{MacNeice00} P. MacNeice, K. Olson, C. Mobarry, R. de Fainchtein, and C. Packer, Computer Physics Comm. {\bf 126}, 330 (2000).
2774: %
2775: \bibitem{Baker:2006kr} J.G. Baker, S.T. McWilliams, J.R. van Meter,
2776: J. Centrella, D.I. Choi, B.J. Kelly, and M. Koppitz, Phys. Rev. D {\bf 75}, 124024 (2007).
2777: %
2778: \bibitem{martel05} K. Martel and E. Poisson, Phys. Rev. D {\bf 71},
2779: 104003 (2005).
2780: %
2781: \bibitem{L85} E.W. Leaver, Proc. R. Soc. Lond. A {\bf 402}, 285 (1985).
2782: %
2783: \bibitem{E89} F. Echeverria, Phys. Rev. D {\bf 40}, 3194 (1997).
2784: %
2785: \bibitem{BCW} E. Berti, V. Cardoso and C. Will, Phys. Rev. D {\bf 73}, 064030 (2006).
2786: %
2787: \bibitem{BCP} A. Buonanno, G. Cook and F. Pretorius, Phys. Rev. D {\bf 75}, 124018 (2007).
2788: %
2789: \bibitem{BI} L. Blanchet and B. Iyer, Class. Quant. Grav. {\bf 20}, 755 (2003).
2790: %
2791: \bibitem{berti07} E. Berti et al., Phys. Rev. D {\bf 76}, 064034 (2007).
2792: %
2793: \bibitem{BBF} L. Blanchet, A. Buonanno, and G. Faye, Phys.\ Rev.\ D
2794: {\bf 74}, 104034 (2006); Erratum-ibid. D {\bf 75}, 049903 (2007).
2795: %
2796: \bibitem{EOB4PN} A. Buonanno, Y. Pan, J.G. Baker, J. Centrella,
2797: B.J. Kelly, S.T. McWilliams, and J.R. van Meter, \url{arXiv:0706.3732}.
2798: %
2799: \end{thebibliography}
2800: \end{document}
2801:
2802: