1: % what format do we use
2: \documentclass[aps,prd,twocolumn,amssymb,eqsecnum,showpacs,nofootinbib]{revtex4}
3:
4: % packages
5: \usepackage{bm}
6: \usepackage{amsmath}
7: \usepackage{euscript}
8: \usepackage{graphicx}
9:
10: % make your own operator
11: \DeclareMathOperator{\sgn}{sgn}
12:
13: % enter the main document
14: \begin{document}
15:
16: \title{Gravitational wave snapshots of generic extreme mass ratio inspirals}
17:
18: \author{Steve Drasco}
19: %\email{sdrasco@jpl.nasa.gov}
20: \affiliation{Jet Propulsion Laboratory, California Institute of
21: Technology, Pasadena, CA 91109}
22: \affiliation{Center for Radiophysics and Space Research, Cornell
23: University, Ithaca, NY 14853}
24:
25: \author{Scott A.\ Hughes}
26: %\email{sahughes@mit.edu}
27: \affiliation{Department of Physics and MIT Kavli Institute, MIT, 77
28: Massachusetts Ave., Cambridge, MA 02139}
29:
30: \begin{abstract}
31: Using black hole perturbation theory, we calculate the gravitational
32: waves produced by test particles moving on bound geodesic orbits about
33: rotating black holes. The orbits we consider are generic ---
34: simultaneously eccentric and inclined. The waves can be described as
35: having radial, polar, and azimuthal ``voices'', each of which can be
36: made to dominate by varying eccentricity and inclination. Although
37: each voice is generally apparent in the waveform, the radial voice is
38: prone to overpowering the others. We also compute the radiative
39: fluxes of energy and axial angular momentum at infinity and through
40: the event horizon. These fluxes, coupled to a prescription for the
41: radiative evolution of the Carter constant, will be used in future
42: work to adiabatically evolve through a sequence of generic orbits.
43: This will enable the calculation of inspiral waveforms that, while
44: lacking certain important features, will approximate those expected
45: from astrophysical extreme mass ratio captures sufficiently well to
46: aid development of measurement algorithms on a relatively short
47: timescale.
48: \end{abstract}
49: \pacs{04.70.-s, 97.60.Lf}
50: \maketitle
51:
52: \section{Introduction}
53: \label{s:intro}
54:
55: One of the leading problems in modern general relativity research is
56: the analysis of binary systems. This problem is quite nasty in
57: general, especially when the members of the binary are strongly
58: interacting and gravitational-wave (GW) emission is significant. It
59: simplifies considerably in the extreme mass ratio limit, in which one
60: member of the binary is much smaller than the other (taken to be a
61: black hole). This is largely because the small mass ratio makes it
62: possible to bring perturbative techniques to bear.
63:
64: This paper reports the development of a key piece of infrastructure
65: needed for an approximate approach to this problem: the ability to
66: accurately calculate the instantaneous GW emission from arbitary bound
67: geodesic orbits of rotating Kerr black holes. Our work generalizes
68: previous calculations which specialized the orbit to be either
69: circular (of constant Boyer-Lindquist coordinate radius
70: {\cite{circular}}) or equatorial (confined to the black hole's plane of
71: reflection symmetry {\cite{equatorial}}).
72:
73: By knowing the GWs, we also calculate the flux of energy $E$ and axial
74: angular momentum $L_z$ that the radiation carries to infinity and down
75: the black hole's event horizon. Global conservation allows us to
76: evolve the energy and angular momentum associated with the orbit,
77: fixing the evolution of two of the three ``constants'' which
78: characterize black hole orbits (up to initial conditions). The
79: remaining constant, $Q$, is known as the Carter constant\footnote{The
80: constant $K = Q + (L_z - aE)^2$ is often used as well.}
81: {\cite{Carter,MTW}}. In the limit of zero black hole spin, the hole
82: is spherically symmetric, geodesic orbits conserve total angular
83: momentum $L$, and Carter's constant is given by $Q = L^2 - L_z^2$;
84: though not strictly accurate, the intuitive interpretation ``$Q = $
85: the rest of the angular momentum'' is useful even for non-zero black
86: hole spin. It has recently become clear that, when the system evolves
87: ``slowly'' (quantified below), it is possible to fix the evolution of
88: $Q$ using techniques not too different from those used to evolve $E$
89: and $L_z$ {\cite{Mino, Hughes et al, Drasco Flanagan
90: Hughes,sthn2005}}.
91:
92: For such slowly evolving systems, we can then evolve {\it all} of
93: the orbital constants; by doing so, we can build reasonably realistic
94: inspirals corresponding to generic initial conditions, as well as the
95: associated waveforms. The qualifier ``reasonably realistic'' refers
96: to the fact that this procedure neglects, by construction, the
97: influence of ``conservative'' pieces of the small body's self
98: interaction; only the ``dissipative'' influence can be analyzed in
99: this way. As we discuss further below, neglect of the conservative
100: influence will limit the reliability of our inspirals; for the
101: applications we have in mind, this limitation is a
102: reasonable trade off in order to build waveforms on a rapid timescale.
103:
104: We now briefly summarize our major motivations for analyzing this
105: problem, sketch the techniques that we have developed and our major
106: results, and discuss the utility and limitations of our results.
107:
108: \subsection{Astrophysical context}
109: \label{s:context}
110:
111: Though an idealization of a much more difficult general problem, the
112: extreme mass ratio limit in fact corresponds precisely to an important
113: subset of astrophysical binaries: stellar mass compact objects
114: captured onto highly relativistic orbits of massive black holes in
115: galaxy cores. Such binaries are expected to be created by multibody
116: scattering processes in the stellar cluster that surrounds these
117: central black holes.
118:
119: It has now become clear that the central bulges of essentially all
120: galaxies contain a massive black hole with properties strongly
121: correlated to that of the bulge {\cite{fm2000,getal2000}}. It also
122: appears likely that {\it mass segregation} (the sinking of massive
123: stellar objects to the bottom of a gravitational potential due to
124: equipartition of kinetic energy {\cite{spitzer69}}) is likely to drive
125: the largest stellar mass compact objects into the vicinity of the
126: massive black hole, increasing the probability of forming a capture
127: binary --- see, for example, Ref.\ {\cite{meg2000}} for discussion
128: pointing to the possibility of a ``minicluster'' of stellar mass black
129: holes in the center of our galaxy, and Ref.\ {\cite{munoetal2005}} for
130: evidence of an overabundance of black holes in binaries in the
131: galaxy's inner parsec. Mass segregation is also seen numerically in
132: galactic center N-body simulations by Freitag {\cite{freitag2001}}.
133: The abundance of galactic black holes plus mass segregation suggest
134: that the capture and GW driven inspiral of stellar mass compact
135: objects into massive black holes may be a relatively common phenomenon
136: in the universe.
137:
138: The GWs produced by these {\it extreme mass ratio inspirals} (EMRIs)
139: are thus ideal targets for low-frequency GW detectors, particularly
140: the space-based {\it LISA} antenna, currently under development as a
141: joint NASA-ESA mission {\cite{LISA,lisaesa}}. By combining
142: estimates for the rate at which such captures are likely to occur (per
143: galaxy) with the projected sensitivity of {\it LISA} (which determines
144: the distance to which sources can be measured, and hence sets the
145: number of galaxies that are relevant), one finds that dozens to
146: thousands of EMRI events may be measured over {\it LISA}'s mission
147: lifetime {\cite{Gair et al}}.
148:
149: Because the small body only slightly perturbs the spacetime of the
150: large black hole, the GWs are largely set by --- and therefore encode
151: --- the nature of the black hole's spacetime. A typical measured EMRI
152: event is likely to last for $\sim 10^4 - 10^5$ orbits; by coherently
153: tracking the waveform's phase over these orbits, {\it LISA} should
154: determine parameters characterizing the binary with great precision.
155: Black hole masses and spins should be measured with fraction of a
156: percent accuracy or better {\cite{Barack Cutler}}; it should even be possible
157: to check whether the spacetime has the ``shape'' (higher multipolar
158: structure) required by the no-hair theorems of general relativity
159: {\cite{ryan95,ryan97,ch2004}}. EMRI measurements are expected to
160: provide a cornucopia of data of interest to astrophysicists and
161: general relativists.
162:
163: This promising astrophysical scenario is a major motivator for much of
164: the effort in this problem (certainly for the present authors).
165: Reliable theoretical models of the inspiral waves will be needed both
166: in order to maximize the science return from {\it LISA} data, but also
167: to assure detection of these events. Ideally, the final
168: science measurement of a detected signal will be performed with a
169: ``measurement template'' --- a model waveform computed accurately
170: enough that it remains in phase with a fractional accuracy of roughly
171: $1/\rho$ (where $\rho$ is the measured signal-to-noise ratio) over the
172: entire inspiral. The techniques which we describe here cannot
173: construct waveforms accurate enough for this task; as we describe in
174: Sec.\ {\ref{ss:a_and_l}}, measurement templates require a more
175: accurate and complete analysis of the small body's self-interaction
176: than our techniques encompass.
177:
178: Our goal instead is to develop waveform models that are sufficiently
179: reliable that they may be used to develop data analysis techniques for
180: EMRI {\it detection}. EMRI waveforms are characterized by 14
181: parameters\footnote{A useful counting of these parameters is as
182: follows: the 2 masses, the black hole spin ${\bf S}$, the initial
183: relative position ${\bf r}_0$, the initial relative velocity ${\bf
184: v}_0$, and the binary's position relative to the observer ${\bf R}$.
185: Each vector has 3 components, for a total of 14 parameters. Including
186: the smaller body's spin ${\bf s}$ raises the count to 17; fortunately,
187: ${\bf s}$ can be neglected for our purposes {\cite{Barack Cutler}}.}.
188: The number of measurement templates that would be needed to fully
189: cover the 14 dimensional manifold of waveforms would render any
190: search for these waves by this method infeasible {\cite{Gair
191: et al}}. Detection will instead be done hierarchically using
192: (relatively) short segments of the EMRI signal {\cite{Gair et al}}.
193: Each short segment need only match coherently to a model for
194: about $10^3 - 10^4$ orbits. The accuracy requirements on such
195: ``detection templates'' are less stringent than those needed for
196: measurement templates.
197:
198: \subsection{Sketch of our calculation}
199: \label{ss:sketch}
200:
201: The key pieces of our formalism have been summarized in depth in
202: previous papers, particularly {\cite{circular}}; here, we provide a very
203: brief sketch largely to set the context for the following discussion.
204:
205: We use the Teukolsky equation {\cite{Teukolsky}} to calculate the
206: influence of a perturbation to a black hole spacetime. This equation
207: describes the evolution of a complex scalar $\psi_4$ which is
208: constructed from the Weyl (vacuum) curvature of the spacetime; it is
209: basically a wave equation for the Weyl curvature, linearized around
210: the Kerr background\footnote{Indeed, Michael Ryan has shown that the
211: Teukolsky equation can be derived from the ``Penrose equation''
212: {\cite{mpryan74}}, a nonlinear wave equation for the Riemann curvature
213: tensor that is constructed from the Bianchi identity; see
214: {\cite{penrose60}}.}, with a source. Schematically, the Teukolsky
215: equation is of the form
216: \begin{equation}
217: {\cal D}^2\psi_4 = {\cal T}[\vec z(t)]\;.
218: \label{eq:teukschematic}
219: \end{equation}
220: The general form of the operator ${\cal D}^2$ and the source function
221: ${\cal T}$ can be found in {\cite{Teukolsky}}; $\vec z(t)$ represents
222: the worldline of the orbiting body. Coordinate time $t$ (which
223: amounts to time as measured by distant observers) is a particularly
224: convenient parameterization for our purposes. Derived forms of Eq.\
225: (\ref{eq:teukschematic}) relevant to our analysis are given in Sec.\
226: {\ref{ss:tsn}}.
227:
228: A key thing to note at this point is that the source depends on the
229: orbiting body's worldline ${\vec z}(t)$. In order to construct ${\cal
230: T}$, we use the ``zeroth'' order worldline, neglecting radiation.
231: This worldline is built from geodesics of the background black hole
232: spacetime; as we discuss in Sec.\ {\ref{ss:a_and_l}}, we
233: incur an important cost due to this setup. From this worldline, we
234: build the source function, and then compute $\psi_4$. This complex
235: scalar completely encodes the GW flux to distant observers, and down
236: the black hole's event horizon {\cite{HH}} (equivalently, the
237: tidal interaction of the hole with the orbiting body {\cite{hartle}}).
238: All of the quantities which we use to describe GW emission and GW
239: induced orbit evolution are encoded in $\psi_4$.
240: %As discussed above,
241: %from $\psi_4$ we construct all quantities of interest to this
242: %analysis.
243:
244: We take advantage of the fact that the Teukolsky equation is separable
245: --- $\psi_4$ can be expanded into Fourier and spheroidal harmonics,
246: allowing it to be computed mode-by-mode. Each mode is characterized
247: by a spherical-harmonic-like integer index $l$, and by three integer
248: indices ($m$, $k$, $n$) that label harmonics of the three fundamental
249: frequencies that describe the orbits. Thanks to the linear nature of
250: the Teukolsky equation, the modes are independent of one another.
251: Codes that solve for $\psi_4$ in this manner are thus easily
252: parallelized --- many mode contributions can be computed separately
253: and independently. Indeed, we have found that the code developed for
254: this work exhibits almost perfect $1/N$ computation-time scaling (where $N$ is
255: the number of processors) {\cite{Hughes et al}}.
256:
257: We incur an important cost by expanding $\psi_4$ in modes: expanding
258: formally requires that we understand our source's behavior for all
259: time, $-\infty < t < \infty$, in order that the Fourier integral
260: exist. In practice, this means that the orbit cannot evolve
261: ``quickly'': we require the radiation to backreact {\it
262: adiabatically}, so that over a typical ``orbital time'' $T_{\rm orb}$
263: the change in any parameter that should be constant is much smaller
264: than the parameter itself (e.g., $T_{\rm orb}\dot E \ll E$). In
265: principle, the adiabaticity requirement could be circumvented by
266: working in the time domain --- solving the wave equation for $\psi_4$
267: directly rather than expanding in modes. Indeed, time domain codes
268: have proven superior to frequency domain codes for analyzing problems
269: without sources (i.e. black hole ringdown), and are a more natural approach to evolving radiation
270: propagating in a black hole spacetime {\cite{td1,td2,td3,td4}}.
271: Although new techniques may eventually change this story \cite{Sopuerta et al},
272: at present, frequency domain codes appear better suited to
273: handle the point-like sources appropriate to the EMRI problem.
274:
275: The next natural step is to approximate the inspiral and associated
276: waveform from a sequence of geodesic orbits. Beginning with some
277: starting orbit, we compute $\psi_4$. From it, we extract a snapshot
278: waveform, which instantaneously approximates the true waveform, as
279: well as the rates of change of the orbital constants. These rates of
280: change tell us how the trajectory evolves from the present orbit to
281: the next orbit in the sequence. By repeating this process many times,
282: we can build an adiabatic waveform that usefully approximates the true
283: waveform. We have not yet taken this step for generic orbits, though
284: most of the tools needed to do so are now in hand
285: {\cite{Mino, Hughes et al, Drasco Flanagan Hughes,sthn2005}}.
286:
287: \begin{table}
288: \begin{ruledtabular}
289: \begin{tabular}{c|c|c|c|c}
290: Ref. & $a \ne 0$ & eccentric orbits &
291: inclined orbits & evolve orbits \\ \hline
292: {\cite{ckp}} & & \checkmark &
293: $\cdots$ & \checkmark \\ \hline
294: {\cite{Finn Thorne}} & \checkmark & &
295: & \checkmark \\ \hline
296: {\cite{Shibata circular, circular}} & \checkmark & &
297: \checkmark & \\ \hline
298: {\cite{Shibata equatorial, equatorial}} & \checkmark & \checkmark &
299: & \\ \hline
300: {\cite{circularII}} & \checkmark & &
301: \checkmark & \checkmark \\ \hline
302: %{\cite{equatorial}} & \checkmark & \checkmark &
303: % & \\ \hline
304: Here & \checkmark & \checkmark &
305: \checkmark &
306: \end{tabular}
307: \end{ruledtabular}
308: \caption{\label{t:history} A sketch of the history of this technique.
309: The magnitude of the massive black hole's spin angular momentum is
310: given by $aM$. Due to spherical symmetry all geodesics of
311: Schwarzschild ($a=0$) black holes are planar, and can be considered
312: equatorial.}
313: \end{table}
314:
315: Table {\ref{t:history}} summarizes the history of this approach. To
316: date, the program has been completed only for non-spinning black holes
317: {\cite{ckp}}, circular-equatorial orbits \cite{Finn Thorne} circular orbits {\cite{circularII}}, and equatorial
318: orbits\footnote{Although Refs.~\cite{equatorial} and \cite{Shibata
319: equatorial} do not use their snapshot data to compute model inspirals,
320: it is a straightforward extension to do so.} \cite{Shibata
321: equatorial, equatorial}. For the astrophysical EMRI problem, these
322: are unrealistic restrictions. Since extreme mass ratio binaries are
323: created through capture processes, we expect inclination to be
324: randomly distributed. We also expect the eccentricity to be
325: substantial --- although radiative backreaction strongly reduces
326: eccentricity, EMRI events have such large initial eccentricities that
327: an estimated 50\% of all observable EMRIs will have eccentricities $e
328: > 0.2$ as they approach their last stable orbit {\cite{Barack
329: Cutler}}. We must understand generic EMRIs in order to realize the
330: event rates predicted in Ref.\ {\cite{Gair et al}}.
331:
332: \subsection{Applicability and limitations of our approach}
333: \label{ss:a_and_l}
334:
335: The ``flux balancing-adiabatic evolution'' approach we advocate here is, as
336: emphasized above, an approximation to the evolution of extreme mass
337: ratio binaries. A more rigorous approach, upon which most workers in
338: this field are focused, is based on computation of the {\it self
339: force} --- the small body's self interaction with its own
340: gravitational perturbation. The gravitational self force is analogous
341: to the Abraham-Lorentz-Dirac electromagnetic self force
342: {\cite{ALD,Dirac}}. Its complete general relativistic formulation was
343: worked out in detail by Mino, Sasaki, and Tanaka {\cite{MST}} and by
344: Quinn and Wald {\cite{QW}}; Poisson {\cite{ericlivrev}} provides a
345: very readable summary of this formalism. Developing this approach to
346: the point that one can build a code around it to study the evolution
347: of generic orbits of Kerr black holes remains some time in the future;
348: however, efforts are intense and progress is rapid
349: {\cite{cqg_special}}.
350:
351: At least heuristically, the self force can be broken into two pieces:
352: a ``dissipative'' force which carries energy and angular momentum away
353: from the binary (and thus drives the smaller body's inspiral), and a
354: ``conservative'' force which does not. These forces are computed
355: relative to geodesics of the {\it background} spacetime --- the
356: spacetime of the large black hole. The conservative force tells us
357: that, even in the absence of radiation emission, the trajectory is
358: modified relative to the background geodesics --- the spacetime is
359: deformed from that of a black hole, so that black hole geodesics do
360: not precisely describe the motion of the small body.
361:
362: Our approach totally neglects the conservative self force --- by
363: construction, we can only analyze dissipative effects. This can be
364: seen in the schematic description of our calculation near Eq.\
365: (\ref{eq:teukschematic}) --- the worldline used in the source function
366: ${\cal T}$ is built from geodesics of the {\it background} spacetime,
367: without incorporating the influence of the conservative force. We
368: have argued previously {\cite{Hughes et al}} that this should be
369: adequate for scoping out EMRI waveforms and exploring issues related
370: to {\it LISA} data analysis. Our argument was based in part on how
371: the phase error resulting from our approach scales with the mass ratio
372: $\mu/M$ [Eq.\ (1.2) of Ref.\ {\cite{Drasco Flanagan Hughes}}]. That
373: scaling was in turn based on the idea that (quoting from Ref.\
374: {\cite{Hughes et al}}) ``{\it Dissipative terms accumulate secularly;
375: conservative terms do not}''. In other words, dissipative aspects of
376: the self force would accumulate phase effects over an EMRI event; we
377: expected that conservative terms would oscillate, and thus not
378: contribute as strongly over an inspiral.
379:
380: To our chagrin, it has now been clearly demonstrated that our claim
381: that conservative effects do not accumulate secularly is wrong
382: (although the scaling of phase errors with mass ratio which we derived
383: from this is correct). Using a compelling toy model to describe the
384: influence of the conservative self force, Pound, Poisson, and Nickel
385: {\cite{ppn05}} show that there is a component to the phase evolution
386: missed by a ``dissipative only'' evolution. A simple way to
387: understand this additional component is as follows: The background
388: geodesics are characterized by oscillatory motion with three
389: frequencies, $\Omega_\phi$, $\Omega_\theta$, and $\Omega_r$, where
390: $\Omega_x$ describes oscillations (or orbits) associated with the
391: coordinate $x$. A conservative self force changes the ``potentials''
392: that determine orbital motion, and thus modifies these three
393: frequencies:
394: \begin{eqnarray}
395: \Omega_\phi &\to& \Omega_\phi + \delta\Omega_\phi
396: \nonumber\\
397: \Omega_\theta &\to& \Omega_\theta + \delta\Omega_\theta
398: \nonumber\\
399: \Omega_r &\to& \Omega_r + \delta\Omega_r\;.
400: \label{eq:freqshift}
401: \end{eqnarray}
402: The shifts are smaller than their frequencies by (roughly) the mass
403: ratio: $\delta\Omega_x \sim (\mu/M)\Omega_x$, where $\mu$ is the mass
404: of the smaller body and $M$ is the mass of the black hole.
405:
406: Some of the most interesting strong field features are due to beating
407: between these frequencies. For example, periastron precession (well
408: known in the solar system due to planetary perihelion precession)
409: occurs at
410: \begin{equation}
411: \Omega_{\rm PP} = \Omega_\phi - \Omega_r\;.
412: \label{eq:peri_prec}
413: \end{equation}
414: Another effect is Lense-Thirring precession, the rotation of
415: the orbital plane due to frame dragging; it occurs at
416: \begin{equation}
417: \Omega_{\rm LT} = \Omega_\phi - \Omega_\theta\;.
418: \label{eq:lt_prec}
419: \end{equation}
420: The conservative self force shifts these precessions:
421: \begin{eqnarray}
422: \delta\Omega_{\rm PP} &=& \delta\Omega_\phi - \delta\Omega_r\;,
423: \nonumber\\
424: \delta\Omega_{\rm LT} &=& \delta\Omega_\phi - \delta\Omega_\theta\;.
425: \label{eq:precshift}
426: \end{eqnarray}
427: It is likely that $\delta\Omega_\phi \simeq \delta\Omega_\theta$; they
428: are presumably exactly equal for Schwarzschild black holes (spherical
429: symmetry). However, $\delta\Omega_r$ is likely to differ quite a bit
430: from the other two frequency shifts, importantly modifying periastron
431: precession. We thus expect that eccentric orbits in particular will
432: be impacted by the conservative self force. This, indeed, is what is
433: found by Pound, Poisson, and Nickel.
434:
435: Our inability to incorporate this conservative effect into our
436: analysis limits its utility. We advocate our approach primarily to
437: begin exploring the space of EMRI waveforms in preparation for {\it
438: LISA}'s EMRI data analysis. The best understanding at the moment,
439: after Ref.~\cite{ppn05}, is that the phase errors scale like
440: $(\mu/M)^0$ and $(v/c)^{-3}$; consequently, in the regime $v\sim c$
441: relevant to {\it LISA} the errors are formally of order unity. It is
442: thus not yet clear if these errors are large enough to prevent
443: adiabatic waveforms being useful for search templates. At any rate,
444: though adiabatic waveforms may miss an important contribution to the
445: phase evolution of extreme mass ratio binaries, they accurately
446: represent the spectral spread that can be expected. Experience from
447: circular {\cite{circular,circularII}} and equatorial
448: {\cite{ckp,equatorial}} studies shows that waves from strong field
449: orbits radiate significant power into high harmonics of the orbital
450: frequencies. The {\it LISA} datastream is expected to contain many
451: simultaneous confused sources --- certainly millions of white dwarf
452: binaries (whose radiation is essentially monochromatic), perhaps $\sim
453: 10^3$ simultaneous EMRI sources, all lying under the signals from
454: massive coalescing cosmological black hole binaries. Learning to
455: detangle these many signals will require models for the different
456: waves which accurately describe their time and frequency overlap. By
457: modeling the frequency spread of EMRI waves with good accuracy (though
458: not necessarily the frequency evolution), adiabatic waveforms should
459: be a very useful tool.
460:
461: {\it If} it turns out to be possible to separate conservative and
462: dissipative self force contributions, it should be easy to modify our
463: approach to include the leading conservative effects: we would simply
464: replace the geodesic worldlines presently used with worldlines that
465: are augmented by these conservative effects. We would then use this augmented worldline
466: to construct the Teukolsky source function. Doing so would allow us
467: to calculate an inspiral that accurately accounted for the leading
468: dissipative {\it and} conservative effects\footnote{One might label a
469: self force that allows this separation ``compassionately
470: conservative''.}.
471:
472: It should be strongly emphasized that this is a rather big ``if'':
473: Because the gravitational self force is a gauge dependent quantity
474: (though its impact on binary orbits must be gauge independent), it is
475: far from clear that one can cleanly separate the ``conservative'' from
476: the ``dissipative'' influence. It may be that the only way to
477: discriminate the two influences is to run the equations of motion
478: which the self force implies through a self-consistent wave-generation
479: calculation {\cite{ppn05}}, in which case there is no need to separate
480: these influences in the first place.
481:
482: \subsection{Organization of this paper}
483: \label{ss:org}
484:
485: We begin this paper by summarizing, in Sec.\ {\ref{s:Bound black hole
486: orbits}}, relevant properties of bound Kerr black hole orbits. These
487: orbits determine, in turn, the properites of the Teukolsky equation's
488: source term. Our goal is to present enough detail to make it clear
489: how we build this term. Of particular importance is a frequency
490: domain description of functions that are built from these orbits
491: {\cite{Schmidt,Drasco Hughes}}.
492:
493: We next (Sec.\ \ref{s:Perturbing a black hole with an orbiting test
494: mass}) summarize the relevant details of the black hole perturbation
495: formalism that we use. We go into some detail in this section. This
496: is in part to make the analysis self-contained, but also to correct
497: some small errors that have appeared in previous papers, notably Ref.\
498: {\cite{circular}} (hopefully without introducing too many new errors).
499: We first review in Sec.\ {\ref{ss:tsn}} how Teukolsky equation
500: solutions are built. This is a somewhat subtle numerical problem ---
501: due to the long-rangedness of the separated Teukolsky equation's
502: radial potential, the amplitudes of ingoing ($\propto e^{-i\omega
503: r^*}$) and outgoing ($\propto e^{i\omega r^*}$) radiative solutions
504: grow at different rates. [The quantity $r^*$ is the Kerr ``tortoise
505: coordinate''; cf.\ Eq.\ (\ref{eq:rstarofr}) below.] One component of
506: the solution can easily swamp the other, causing numerical resolution
507: to be lost. Sasaki and Nakamura {\cite{SN}} found an equation with
508: short-ranged potential whose solutions are related to those of the
509: Teukolsky equation by a simple differential transformation; the
510: combined Sasaki-Nakamura-Teukolsky formalism makes for a very
511: well-behaved numerical problem. We next (Sec.\ \ref{ss:Bound test
512: mass sources}) discuss explicitly how the source function is built,
513: taking advantage of the results that we presented in Sec.\
514: {\ref{s:Bound black hole orbits}}, and then discuss some pratical
515: issues related to the numerics (Sec.\ \ref{ss:Numerical
516: considerations}). We wrap up this section with a brief discussion of
517: how the gravitational waveforms and fluxes of ``conserved'' quantities
518: are extracted from these solutions.
519:
520: Section \ref{s:Numerical algorithm} discusses several practical issues
521: related to numerics. We first describe the algorithms used to
522: numerically represent the geodesics, a key element for the Teukolsky
523: source term, before describing issues related to computing each of the
524: modes from which we build $\psi_4$. We then discuss in great depth
525: the algorithms we have developed to truncate our modal expansion.
526: Strictly speaking, the number of modes that should be used to build
527: $\psi_4$ is infinite; picking a finite value to truncate this
528: expansion that is ``large enough'' is somewhat subtle. We describe a
529: scheme in Sec.\ \ref{ss:Truncation} that has proven to be robust
530: enough to work well for our present purposes. There is clearly room
531: for improvement, however; we describe some ways in which this
532: procedure could be made better. Finally, we conclude in Sec.\
533: \ref{ss:Validation} with detailed discussion of validation tests that
534: we made against previous results in the appropriate limits. We find
535: that the code agrees perfectly with results from Ref.\
536: {\cite{circular}} when the eccentricity is zero\footnote{The
537: ``generic'' code is a direct descendent of the ``circular'' code used
538: in Ref.\ {\cite{circular}}, so it is perhaps not too surprising that
539: there is agreement in the circular limit. However, {\it many} details
540: of the generic code's inner workings are significantly different from
541: the circular incarnation, so this agreement is far from trivial.}, and
542: agrees quite well with the ``equatorial'' code of Glampedakis and
543: Kennefick in {\cite{equatorial}} in the limit of zero inclination.
544: Except for some (very small) down-horizon modes that do not
545: significantly impact the overall flux, we typically find agreement at
546: the level of $10^{-3}$ or better with Glampedakis and Kennefick. This
547: is the level of accuracy that they cite for their code.
548:
549: Our main results are presented in Sec.\ \ref{s:Results}. The modal
550: decomposition allows us to break the waveform into ``voices'':
551: \begin{equation}
552: H \equiv h_+ - ih_\times = \sum_{kn} H_{kn} e^{-i\omega_{kn}(t - r^*)}\;,
553: \label{eq:voice_def}
554: \end{equation}
555: where $\omega_{kn} \equiv k\Omega_\theta + n\Omega_r$. [A sum over
556: $l$ and $m$, including oscillations with frequency $m\Omega_\phi$, is
557: hidden in the definition of $H_{kn}$; see Eq.\ (\ref{eq:Hkn_def})
558: below.] The ``polar voice'' is composed of the terms with $n = 0$ and
559: $k \ne 0$; The ``radial voice'' is composed of the terms with $k = 0$
560: and $n \ne 0$; The ``azimuthal voice'' is the term with $k = n = 0$,
561: and the ``mixed voice'' is composed of the terms with $k \ne 0$ and $n
562: \ne 0$. A similar voice-by-voice labeling can be applied to the
563: fluxes. We find that the importance of the various voices is fairly
564: simply controlled by the eccentricity and the inclination angle: Polar
565: voices become progressively more important as the purely polar orbit
566: (inclination $90^\circ$) is approached from either direction; radial
567: voices rapidly become important as eccentricity, $e$, approaches unity,
568: typically dominating the waves' spectrum by $e \sim 0.3 - 0.5$. We
569: speculate that this multivoice character may facilitate approximations
570: in the design of GW detection schemes, making it possible to detect
571: the most important voices of signals, rather than needing to detect
572: the (rather ornate) chorus of all voices together.
573:
574: Section \ref{s:Summary and future work} presents our final
575: conclusions, and outlines future work to which we plan to apply this
576: formalism. Chief among these future tasks will be augmenting this
577: approach with an adiabatic scheme to evolve the Carter constant
578: {\cite{Drasco Flanagan Hughes,sthn2005}}, and then using the complete
579: flux data to build model adiabatic inspirals and their associated
580: waveforms (as discussed in Sec.\ {\ref{ss:sketch}}). We are
581: optimistic that this can be completed relatively quickly since there
582: appear to be no major hurdles or issues of principle that must be
583: overcome first. Producing adiabatic waveforms for initial data
584: analysis algorithm development will then ``just'' be a matter of
585: finding sufficient CPU power.
586:
587: \section{Bound black hole orbits}
588: \label{s:Bound black hole orbits}
589:
590: In this section we review the geodesic motion of a non-spinning test
591: mass on a bound orbit of a Kerr black hole. Kerr orbits are not a new
592: subject of investigation \cite{Carter, Wilkins}, but interest has been
593: renewed recently because of their relevance to EMRI GWs \cite{Schmidt,
594: Mino, Drasco Hughes}. The main new development has been Mino's
595: exploitation of the fact that these orbits are fundamentally periodic
596: entities \cite{Mino}. The utility of exploiting this property is
597: discussed in detail in Ref.~\cite{Drasco Hughes}, and details of its
598: application to EMRIs can be found in Sec.\ 3 of Ref.~\cite{Drasco
599: Flanagan Hughes}. Here we will discuss the relation between an
600: orbit's geometry and its natural frequencies. Using a special choice
601: of initial conditions (the fiducial geodesics from Ref.~\cite{Drasco
602: Flanagan Hughes}) we also derive a new system of equations for
603: efficient numerical evaluation of these orbits.
604:
605: Kerr black holes are characterized by two parameters: the mass $M$ of
606: the black hole, and the magnitude $aM$ of its spin angular momentum,
607: with $0\le a \le M$. Throughout this paper, we will use
608: Boyer-Lindquist \cite{Boyer Lindquist} coordinates
609: $(t,r,\theta,\phi)$, with units chosen so that $G=c=1$. The line
610: element for the Kerr geometry is then given by \cite{MTW}
611: \begin{eqnarray}
612: \label{Kerr}
613: ds^2_\text{Kerr} &=&
614: - \left( 1-\frac{2Mr}{\Sigma} \right) ~dt^2
615: + \frac{\Sigma}{\Delta}~dr^2
616: + \Sigma~d\theta^2 \nonumber \\
617: &&+ \left( r^2+a^2 + \frac{2Ma^2r}{\Sigma}\sin^2\theta \right)\sin^2\theta~d\phi^2 \nonumber \\
618: &&- \frac{4Mar}{\Sigma}\sin^2\theta~dt~d\phi,
619: \end{eqnarray}
620: where
621: \begin{align}
622: & \Sigma = r^2 + a^2\cos^2\theta, & \Delta = r^2 - 2Mr + a^2. &
623: \end{align}
624:
625: Bound black hole orbits admit four constants of the motion which allow
626: us to rewrite the geodesic equations as a system of first order
627: differential equations. Three of these constants are fairly
628: straightforward --- if the test particle has 4-momentum $\vec{p}$,
629: these constants are
630: \begin{align}
631: &\vec{p}\cdot\vec{p} = -\mu^2,&
632: &\vec{ \partial_t } \cdot \vec{p} = -E,&
633: &\vec{ \partial_\phi } \cdot \vec{p} = L_z,&
634: \end{align}
635: where $\mu$ is its rest mass, $E$ is the orbit's energy, and $L_z$ is
636: its axial angular momentum. Carter discovered a fourth constant $Q$
637: which allows the motion to be completely described by a system of
638: first order equations \cite{Carter}. As discussed in the
639: Introduction, it is often useful to think of the Carter constant as
640: representing the geodesic's non-axial angular momentum (a
641: correspondance which is exact for non-rotating black holes).
642:
643: Carter's first order equations can be written in the following form
644: \begin{align}
645: &\left(\frac{dr}{d\lambda}\right)^2 = V_r(r), &
646: &\frac{dt}{d\lambda} = V_t(r,\theta),&
647: \nonumber \\
648: &\left(\frac{d\theta}{d\lambda}\right)^2 = V_\theta(\theta), &
649: &\frac{d\phi}{d\lambda} = V_\phi(r,\theta).&
650: \label{geodesics}
651: \end{align}
652: The functions $V_t$, $V_r$, $V_\theta$, and $V_\phi$ are shown in
653: Appendix \ref{s:Geodesic details}. The Mino time parameter $\lambda$
654: is related to the test mass' proper time $\tau$ by
655: \begin{equation}
656: \frac{d\tau}{d\lambda} = \Sigma.
657: \end{equation}
658: Mino time decouples the radial and polar equations of motion so that
659: $V_r=V_r(r)$ and $V_\theta = V_\theta(\theta)$. This property
660: appears to have been recognized by Carter; however, Mino appears to be
661: the first to use $\lambda$ to fully exploit the periodic nature of Kerr
662: orbits \cite{Mino}. Since the first order equations for $r$ and
663: $\theta$ are purely quadratic in their derivatives, $r(\lambda)$ and
664: $\theta(\lambda)$ are periodic functions for bound orbits. The functions
665: $dt/d\lambda$ and $d\phi/d\lambda$ are then biperiodic functions in the
666: sense defined by Ref.~\cite{Drasco Hughes}. As a result,
667: after subtracting a term proportional to $\lambda$, the coordinates
668: $t(\lambda)$ and $\phi(\lambda)$ can be represented with a two dimensional
669: Fourier series. These properties, discussed in more detail below, are extremely powerful.
670:
671: Solutions of the geodesic equations (\ref{geodesics}) are uniquely
672: determined if we specify $E$, $L_z$, $Q$ and the initial position
673: $\vec{z}(\lambda = 0)$ (or some other equivalent set of constants).
674: We now focus on the orbit's geometry or
675: shape, which is determined by $E$, $L_z$, and $Q$. This fact can be
676: roughly understood in that the orbit must be bounded by two radii
677: $r_{\min} \leq r \leq r_{\max}$, and [because the geometry
678: (\ref{Kerr}) is symmetric across the equatorial plane] one polar angle
679: $ \theta_{\min} \leq \theta \leq (\pi - \theta_{\min})$. The orbit is
680: then confined to a toroidal region sketched in
681: Fig.~\ref{f:torus-pretzle}.
682: \begin{figure*}
683: \includegraphics[width = 0.95\textwidth]{merged_BW}
684: \caption{The orbital torus and the evolution of $r$ and $\cos\theta$
685: for a generic geodesic orbit. The magnitude of the black hole's spin
686: is $a = 0.998 M$. The orbit shown here has
687: eccentricity $e = 1/3$, inclination $\theta_\text{inc} = 30^\circ$,
688: and semilatus rectum $p = 7$. We start the orbit at $(r,\cos\theta) =
689: (r_{\min},\cos\theta_{\min})$, and end it around $(r,\cos\theta) =
690: (10.5M,-0.1)$, after several complete radial and polar
691: oscillations. The orbit is not closed: over time, it would eventually
692: fill the orbital torus.}
693: \label{f:torus-pretzle}
694: \end{figure*}
695: The boundaries of the orbital torus could have equivalently been
696: determined by an eccentricity $e$, a semilatus rectum $p$, and an
697: inclination $\theta_\text{inc}$ defined by
698: \begin{equation}
699: r_\text{min} = \frac{pM}{1+e},\;\;\;
700: r_\text{max} = \frac{pM}{1-e},\;\;\;
701: \theta_\text{inc} +(\sgn L_z) \theta_{\min} = \frac{\pi}{2}\;.
702: \end{equation}
703: Here we have included a factor of $\sgn L_z$ so as to make
704: $\theta_\text{inc}$ most closely resemble another common definition
705: for the orbital inclination angle $\iota$ \cite{circular}:
706: \begin{equation} \label{iota}
707: \cos \iota = \frac{L_z}{\sqrt{L_z^2 + Q}}.
708: \end{equation}
709: A nice property of the angles $\theta_\text{inc}$ and $\iota$ (not
710: shared by the angle $\theta_{\min}$) is that
711: they automatically encode a notion of prograde and retrograde ---
712: prograde orbits ($\phi$ motion parallel to the hole's rotation) have
713: $\theta_\text{inc},\iota < 90^\circ$; retrograde orbits ($\phi$ motion
714: antiparallel to the hole) have $\theta_\text{inc},\iota > 90^\circ$.
715: We have
716: found that in general $\iota \approx \theta_\text{inc}$. Even in the
717: strong field the two quantities typically differ by less than 10\%.
718: For the orbits which we discuss in detail here (Sec.~\ref{s:Results})
719: $\iota$ and $\theta_\text{inc}$ differ by no more than $0.5\%$.
720:
721: Explicit algebraic relationships between the geometric orbital
722: parameters $(e,p,\theta_\text{inc})$ and the physical constants
723: $(E,L_z,Q)$ were first computed by Schmidt \cite{Schmidt}, and are
724: shown below in Appendix \ref{s:Geodesic details}. We find that when
725: exploring the orbital parameters space, it is best to first think in
726: terms of $(e,p,\theta_\text{inc})$, and then to convert these
727: (following Appendix \ref{s:Geodesic details}) to $(E,L_z,Q)$ which are
728: used in all further calculations.
729:
730: Using a special choice of initial conditions, we now derive a set of
731: equations for efficient numerical evaluation of bound orbits. Since
732: the solutions to the radial and polar geodesic equations
733: (\ref{geodesics}) are periodic in Mino time, each can be expressed as
734: a Fourier series,
735: \begin{align} \label{Fseries}
736: &\theta(\lambda) = \sum_{k=-\infty}^\infty \theta_k e^{-ik\Upsilon_\theta \lambda}, &
737: &r(\lambda) = \sum_{n=-\infty}^\infty r_n e^{-in\Upsilon_r\lambda},&
738: \end{align}
739: where $\theta_k$ and $r_n$ are constants, and where
740: $\Upsilon_{r,\theta}$ are the orbital frequencies in Mino time. In
741: Appendix \ref{s:Geodesic details} we list explicit expressions for
742: these frequencies as functions of the orbital parameters described in
743: the previous section. It is convenient to write these series in terms
744: of angle-variables,
745: \begin{align}
746: &w_r = \Upsilon_r \lambda,&
747: &w_\theta = \Upsilon_\theta \lambda.&
748: \end{align}
749: The result is
750: \begin{align} \label{Fseries w}
751: &\theta(w_\theta) = \sum_{k=-\infty}^\infty \theta_k e^{-ik w_\theta}, &
752: &r(w_r) = \sum_{n=-\infty}^\infty r_n e^{-in w_r}.&
753: \end{align}
754:
755: We now specialize to a specific choice of initial conditions for the
756: radial and polar motion. We require that the orbit begins at a
757: turning point of both the $r$ and $\theta$ coordinates,
758: \begin{align} \label{initial rt}
759: &r(w_r=0) = r_{\min},&
760: &\theta(w_\theta=0) = \theta_{\min}.&
761: \end{align}
762: This choice of initial conditions will result in a simplified
763: numerical evaluation of the geodesics. For example, the coordinates
764: $r$ and $\theta$ are now even functions of $w_{r,\theta}$,
765: \begin{align} \label{even coords}
766: &r(-w_r) = r(w_r),&
767: &\theta(-w_\theta) = r(w_\theta),&
768: \end{align}
769: and the Fourier series (\ref{Fseries}) become cosine series,
770: \begin{eqnarray}
771: \theta(w_\theta) &=& \theta_0 + 2\sum_{k = 1}^\infty \theta_k \cos(k w_\theta),
772: \\
773: r(w_r) &=& r_0 + 2\sum_{n = 1}^\infty r_n \cos(n w_r).
774: \end{eqnarray}
775: In this paper we will not evaluate these series
776: explicitly\footnote{Current investigations suggest that doing so may
777: lead to greater computational efficiency.}, though we will use the
778: fact that they are even functions.
779:
780: We now derive a similarly simplified series expansion for $t$ and
781: $\phi$. Both the $t$ and $\phi$ coordinates will be treated in a
782: similar way, so to save space we define
783: \begin{align}
784: &x = t,\phi,&
785: &\dot x = V_t,V_\phi,&
786: \end{align}
787: such that an overdot represents a derivative with respect to Mino time
788: $\lambda$. Following the analysis preceeding Eqs.~(3.18) and (3.19)
789: in Ref.~\cite{Drasco Hughes}, we write the derivatives $\dot x$ in
790: the form
791: \begin{subequations}\label{dot series}
792: \begin{eqnarray}
793: \dot x(\lambda) &=& \dot x_{00} + \sum_{k=1}^\infty (\dot x_k^\theta e^{-ik\Upsilon_\theta\lambda} + \text{c.c.})
794: \nonumber \\
795: && + \sum_{n=1}^\infty (\dot x_n^r e^{-in \Upsilon_r\lambda} + \text{c.c.}),
796: \\
797: \dot x_k^\theta &=& \frac{1}{2\pi} \int_0^{2\pi} dw_\theta~ \dot x^\theta[\theta(w_\theta)] e^{i k w_\theta},
798: \\
799: \dot x_n^r &=& \frac{1}{2\pi} \int_0^{2\pi} dw_r~ \dot x^r[r(w_r)] e^{i n w_r}.
800: \end{eqnarray}
801: \end{subequations}
802: Here ``c.c.'' means the complex conjugate of the preceeding term. The
803: leading constants are defined by
804: \begin{equation}
805: \dot x_{00} = \dot x_0^\theta + \dot x_0^r\;.
806: \end{equation}
807: We have made use of the fact that the derivatives of both $t$ and
808: $\phi$ separate into a sum of two functions, each depending only on
809: one of the coordinates\footnote{It's interesting to note that
810: Teukolsky has shown that the master equation (\ref{master}) separates
811: in any coordinates where this property (\ref{split potentials}) is
812: preserved \cite{Teukolsky}.},
813: \begin{equation} \label{split potentials}
814: {\dot x} = {\dot x}^\theta(\theta)+ {\dot x}^r(r).
815: \end{equation}
816: %From Eqs.~(\ref{tdot}) and (\ref{phidot}), we have
817: %\begin{subequations}
818: %\begin{eqnarray}
819: %{\dot t}^r(r) &=& E\frac{\varpi^4}{\Delta} + aL_z \left(1 - \frac{\varpi^2}{\Delta} \right),
820: %\\
821: %{\dot t}^\theta(\theta)&=& - Ea^2\sin^2\theta,
822: %\\
823: %{\dot \phi}^r(r) &=&aE\left(\frac{\varpi^2}{\Delta} - 1 \right) - \frac{a^2L_z}{\Delta},
824: %\\
825: %{\dot \phi}^\theta(\theta)&=& L_z \csc^2\theta.
826: %\end{eqnarray}
827: %\end{subequations}
828: Following the notation of Ref.~\cite{Drasco Hughes}, we use the
829: following symbols for ${\dot x}_{00}$ in the cases $x=t$ and $x=\phi$:
830: \begin{align}
831: &\Gamma = (V_t)_{00},&
832: &\Upsilon_{\phi} = (V_\phi)_{00}.&
833: \end{align}
834: The constant $\Gamma$ is the analog of the Lorentz factor of special
835: relativity. For example, in the case of an orbit which is both
836: circular and equatorial, we have $\Gamma = dt/d\lambda = (dt/d\tau)
837: \Sigma$ ($\Sigma$ is constant for circular-equatorial orbits).
838: $\Gamma$ also relates the Mino time frequencies
839: $\Upsilon_{\phi,\theta,r}$ to coordinate time frequencies $\Omega_r$,
840: $\Omega_\theta$, and $\Omega_\phi$:
841: \begin{align} \label{t frequencies}
842: &\Omega_\phi = \frac{\Upsilon_\phi}{\Gamma},&
843: &\Omega_\theta = \frac{\Upsilon_\theta}{\Gamma},&
844: &\Omega_r = \frac{\Upsilon_r}{\Gamma}.&
845: \end{align}
846: Schmidt \cite{Schmidt} provides an elegant derivation of closed form
847: expressions for the orbital frequencies. His results were converted
848: into the Mino time frequencies in Ref.~\cite{Drasco Hughes} (see
849: Appendix \ref{s:Geodesic details} for details).
850:
851: As a consequence of their initial values (\ref{initial rt}), the
852: functions $\theta(w_\theta)$ and $r(w_r)$ are even. The functions
853: $\dot x^\theta(w_\theta)$ and $\dot x^r(w_r)$ are then also even, and
854: the series (\ref{dot series}) above simplifies to
855: \begin{subequations}\label{simple dot series}
856: \begin{eqnarray}
857: \dot x(\lambda) &=& \dot x_{00} + 2\sum_{k=1}^\infty \dot x_k^\theta \cos(k\Upsilon_\theta\lambda)
858: + 2\sum_{n=1}^\infty \dot x_n^r \cos(n \Upsilon_r\lambda). \label{expanded xdot}
859: \nonumber \\ && \\
860: \dot x_k^\theta &=& \frac{1}{\pi} \int_0^{\pi} dw_\theta~ \dot x^\theta[\theta(w_\theta)] \cos(k w_\theta),
861: \\
862: \dot x_n^r &=& \frac{1}{\pi} \int_0^{\pi} dw_r~ \dot x^r[r(w_r)] \cos(n w_r).
863: \end{eqnarray}
864: \end{subequations}
865:
866: If we now assume initial values for $t$ and $\phi$,
867: \begin{align}
868: &t(\lambda = 0) = 0\;,&
869: &\phi(\lambda = 0) = 0\;,&
870: \end{align}
871: so that we are now discussing the fiducial geodesics from
872: Ref.~\cite{Drasco Flanagan Hughes}, a similarly simplified expression
873: for the coordinates $x$ can be found by integrating the series
874: (\ref{expanded xdot}) for their derivatives $\dot x$:
875: \begin{subequations}
876: \begin{align} \label{expanded x}
877: x(\lambda) &= {\dot x}_{00} \lambda + \Delta x^\theta(\lambda) + \Delta x^r(\lambda)&
878: \\
879: \Delta x^\theta(\lambda) &= \sum_{k=1}^{\infty} \Delta x_k^\theta \sin(k \Upsilon_\theta \lambda), &
880: \label{Dxtheta}
881: \\
882: \Delta x^r(\lambda) &= \sum_{n=1}^{\infty} \Delta x_n^r \sin(n \Upsilon_r \lambda), &
883: \label{Dxr}
884: \\
885: \Delta x_k^\theta &= \frac{2}{k\pi\Upsilon_k} \int_0^\pi dw_\theta~ {\dot x}^\theta(w_\theta) \cos(k w_\theta),&
886: \label{Dxk}
887: \\
888: \Delta x_n^r &= \frac{2}{n\pi\Upsilon_r} \int_0^\pi dw_r~ {\dot x}^r(w_r) \cos(n w_r).&
889: \label{Dxn}
890: \end{align}
891: \end{subequations}
892: Note that in the cases of orbits with special geometries, the
893: coordinates $x$ simplify accordingly:
894: \begin{subequations}\label{geodesic sym}
895: \begin{align}
896: \Delta x^r(\lambda) &= 0 = \Delta x^r_n & &\text{(for circular orbits)}\;, & \\
897: \Delta x^\theta(\lambda) &= 0 = \Delta x^\theta_k & &\text{(for equatorial orbits)}\;.&
898: \end{align}
899: \end{subequations}
900:
901: %Although in general none of the coordinates are periodic in coordinate time,
902: %[e.g. $r(t + 2\pi/\Omega_r) \ne r(t)$] the frequencies (\ref{t frequencies}) reduce to the
903: %usual orbital frequencies in the appropriate limits. For example,
904: %one can readily see that for a circular-equatorial orbit $\Omega_\phi = d\phi/dt$.
905: %Similarly, in the Newtonian limit all three frequencies are identical,
906: %while in the strong field regime they are generally distinct.
907: %Figure \ref{f:frequencies} demonstrates that a typical EMRI's orbital frequencies
908: %are easily distinguished in the LISA frequency band, indicating that EMRIs are
909: %strongly relativistic (as expected).
910: %\begin{figure}
911: %\includegraphics[width = .45\textwidth]{frequencies}
912: %\caption{Here we show the coordinate time frequencies as a function
913: %of the minimum orbital radius. The massive black hole has mass
914: %$M = 10^6 M_{\odot}\approx 1.5 \times 10^6$ km, and spin $a = 0.998M$.
915: %The orbits have have eccentricity $e = 1/2$, and are inclined
916: %by $\theta_\text{inc} = 45^\circ$.
917: %The dash-dot curve is the axial frequency $\Omega_\phi / (2\pi)$, the solid curve is
918: %the polar frequency $\Omega_\theta/(2\pi)$, and the dashed curve is the radial
919: %frequency $\Omega_r/(2\pi)$.}
920: %\label{f:frequencies}
921: %\end{figure}
922: %Note also that there is a critical radius (at $r = 2.3M$ or $3.5\times 10^6$ km
923: %in this case) below
924: %which the radial frequency actually decreases as the hole is approached.
925:
926: Evaluating the integrands in Eqs.\ (\ref{Dxk}) and (\ref{Dxn}) as
927: functions of $r(w_r)$ and $\theta(w_\theta)$, would first require a
928: direct integration of the radial and polar geodesic equations.
929: Unfortunately this is somewhat difficult because the derivatives $V_r$
930: and $V_\theta$ vanish (by definition) at the orbital turning points.
931: This means, for example, that the integrand in the direct expression
932: for $r(\lambda)$,
933: \begin{equation}
934: \lambda = \int_{r_{\min}}^{r(\lambda)} \frac{dr'}{\pm \sqrt{V_r(r')}},
935: \end{equation}
936: contains singularities which, although integrable, complicate the
937: numerics. This is a well known problem (see the Appendix of
938: Ref.~\cite{Drasco Hughes} for more details) which can be avoided if we
939: change the integration variables to more well behaved coordinates
940: $\chi(\theta)$ and $\psi(r)$, where
941: \begin{align} \label{new coords}
942: &\cos\chi = \frac{\cos\theta}{\cos\theta_{\min}},&
943: &r = \frac{pM}{1+e\cos\psi}.&
944: \end{align}
945: Recall that $p$ is the semilatus rectum of the orbit, and $e$ is the
946: orbital eccentricity. So in practice we replace Eqs.(\ref{Dxk}) and
947: (\ref{Dxn}) with
948: \begin{eqnarray} \label{final delta}
949: \Delta x_k^\theta &=& \frac{2}{k\pi\Upsilon_\theta} \int_0^\pi d\chi~\frac{dw_\theta}{d\chi} {\dot x}^\theta(w_\theta) \cos(k w_\theta),
950: \label{Dxk2}
951: \\
952: \Delta x_n^r &=& \frac{2}{n\pi\Upsilon_r} \int_0^\pi d\psi~ \frac{dw_r}{d\psi}{\dot x}^r(w_r) \cos(n w_r).
953: \label{Dxn2}
954: \end{eqnarray}
955: In order to use this result, we of course need explicit expressions
956: for $w_\theta(\chi)$, $w_r(\psi)$, and their derivatives. These
957: functions were first derived in Ref.~\cite{Drasco Hughes}, and the
958: results are shown below in Appendix \ref{s:Geodesic details}. It
959: turns out that $w_\theta(\chi)$ and $dw_\theta/d\chi$ can be written
960: in terms of elliptic integrals. The functions $w_r(\psi)$ and
961: $dw_r/d\psi$ are slightly more involved.
962:
963: We have found that this technique is an extremely efficient way to
964: evaluate bound orbits and functions of them (see Ref.~\cite{Drasco
965: Hughes} for a simple example). The two expansions given by
966: Eqs.~(\ref{Dxtheta}) and (\ref{Dxr}) converge very rapidly. Typically
967: only about $50$ terms are needed in order to obtain a fractional
968: accuracy of $10^{-12}$.
969:
970: \section{Perturbing a black hole with an orbiting test mass}
971: \label{s:Perturbing a black hole with an orbiting test mass}
972:
973: In this section we will discuss how a non-spinning test mass $\mu$ on
974: a bound geodesic perturbs the Kerr geometry (\ref{Kerr}). The
975: radiative information describing the linear order perturbation can be
976: extracted from the Weyl curvature scalar \cite{Wald}
977: \begin{equation} \label{Weyl}
978: \psi_4 = -C_{\alpha\beta\gamma\delta} n^\alpha \bar m^\beta n^\gamma
979: \bar m^\delta\;.
980: \end{equation}
981: Here, overbar denotes complex conjugation, $C_{abcd}$ is the Weyl
982: curvature tensor (the Riemann tensor in vacuum), and the vectors are
983: elements of a Newman-Penrose null-basis \cite{NP, Chandra} \{$
984: l^\alpha,m^\alpha,\bar m^\alpha,n^\alpha$\}. For distant observers,
985: the curvature scalar $\psi_4$ is simply related to the metric
986: perturbation. For these same observers, and also for observers at the
987: event horizon, $\psi_4$ is simply related to the fluxes of energy and
988: angular momentum (see Sec.~\ref{ss:Waveforms and fluxes} for details).
989: Since we are only analyzing perturbations to first order in the mass
990: ratio, $\psi_4$ always represents the leading $O(\mu/M)$ contribution
991: to the curvature perturbation.
992:
993: \subsection{Teukolsky-Sasaki-Nakamura formalism}
994: \label{ss:tsn}
995:
996: Teukolsky showed that $\psi_4$ is a solution to an equation of the
997: form \cite{Teukolsky}
998: \begin{align} \label{master}
999: \left[ \widehat{U}_{t\phi r}(r) + \widehat{V}_{t\phi \theta}(\theta) \right]
1000: \varphi = -\EuScript{T},
1001: \end{align}
1002: where $\EuScript{T}$ is the source term (described below
1003: in Sec.~\ref{ss:Bound test mass sources}),
1004: \begin{align}
1005: &\varphi = \rho^{-4}\psi_4, &
1006: &\rho = -(r - ia\cos \theta)^{-1},&
1007: \end{align}
1008: and where $\widehat{U}_{t\phi r}(r)$ and $\widehat{V}_{t\phi
1009: \theta}(\theta)$ are second order differential operators containing
1010: derivatives with respect to the variables shown as subscripts [see
1011: Eqs.~(\ref{Utphir}) and (\ref{Vtphitheta})]. Equation (\ref{master})
1012: is known as the master equation. In the next two sections, we will
1013: summarize Teukolsky's technique for solving the master equation by
1014: separation of variables \cite{Teukolsky}, converting
1015: Eq.~(\ref{master}) into a pair of ordinary differential equations.
1016: This technique is referred to as solving the master equation ``in the
1017: frequency domain''. As briefly discussed in the Introduction, one
1018: could instead solve the master equation numerically as a partial
1019: differential equation ``in the time domain'' (see Refs.~\cite{Poisson
1020: PDE,Khanna} and references therein). Such an approach can be
1021: advantageous for problems where there is no source, or where the
1022: source is not pointlike. To date at least, time domain treatments are
1023: accurate to about 10\% \cite{Khanna} for EMRI sources (at least for
1024: quantities like orbit averaged fluxes of energy and angular momentum).
1025:
1026: In the source free case $\EuScript{T} = 0$, the master equation
1027: (\ref{master}) is satisfied by functions of the form
1028: \begin{equation}
1029: \varphi_m(\omega,C) =
1030: \widetilde{R}_m(r;\omega,C)S_m(\theta;\omega,C)e^{-i\omega t + im\phi},
1031: \end{equation}
1032: where $C$ is some constant, $m$ is an integer (since $\psi$ must be
1033: periodic in $\phi$), and $\omega$ is any real number. The functions
1034: $\widetilde{R}$ and $S$ are solutions to ordinary differential
1035: equations of the form
1036: \begin{subequations}
1037: \begin{align}
1038: \left[ \widehat{U}_r(r,m,\omega) - C\right]\widetilde{R}_m(r,\omega,C)
1039: &= 0,& \label{radial}
1040: \\
1041: \left[ \widehat{V}_\theta(\theta,m,\omega) + C\right]
1042: S_m(\theta,\omega,C) &= 0.& \label{angular}
1043: \end{align}
1044: \end{subequations}
1045: Requiring that the solutions to the angular equation (\ref{angular})
1046: (known as spin-weighted spheroidal harmonics with spin weight $-2$)
1047: be regular results in a discrete spectrum of eigenvalues
1048: $C = C_{lm}(\omega)$,
1049: \begin{equation}
1050: S_m[\theta,\omega,C_{lm}(\omega)] \equiv S_{lm}(\theta,\omega)\;,
1051: \end{equation}
1052: where $l \ge \max(|m|,2)$ is an integer.
1053: The functions $S_{lm}(\theta,\omega)$ are uniquely defined only after we specify
1054: boundary conditions, or equivalently after we chose a normalization convention.
1055: We use the convention\footnote{Although not stated there, this is the normalization
1056: convention used in
1057: Ref.~\cite{circular}.}
1058: \begin{equation} \label{S norm}
1059: \int_0^{\pi} d\theta~ \left[S_{lm}(\theta,\omega)\right]^2 \sin\theta
1060: = \frac{1}{2\pi}\;.
1061: \end{equation}
1062: The choice of normalization is arbitrary in the sense that it does
1063: not change any physical predictions. However, in order to
1064: compare equations from other papers in which a different convention was
1065: used, we must state our choice explicitly.
1066: The functions $S_{lm}(\theta,\omega)$ will be used as a basis to
1067: express functions $f(t,\theta,\phi)$ as follows:
1068: \begin{equation}
1069: f(t,\theta,\phi) = \sum_{l=2}^\infty \sum_{m=-l}^l
1070: \int_{-\infty}^\infty\!\! d\omega\,\langle lm\omega|f\rangle
1071: S_{lm}(\theta,\omega) e^{-i\omega t + i m \phi},
1072: \end{equation}
1073: where
1074: \begin{equation}
1075: \left<lm\omega|f\right> = \frac{1}{2\pi}\int dt~ \int d\Omega~
1076: f(t,\theta,\phi) S_{lm}(\theta,\omega) e^{i\omega t - i m\phi},
1077: \end{equation}
1078: so that we have $\left<l'm'\omega'|lm\omega\right> =
1079: \delta_{ll'}\delta_{mm'}\delta(\omega - \omega')$.
1080:
1081: We now return to the case where the source $\EuScript{T}$ in
1082: the master equation (\ref{master}) is nonvanishing. Assuming a
1083: solution of the form
1084: \begin{equation} \label{psi_4}
1085: \varphi = \rho^{-4}\psi_4 = \sum_{l=2}^\infty \sum_{m=-l}^l \int d\omega~ \varphi_{lm}(\omega)\;,
1086: \end{equation}
1087: where
1088: \begin{equation} \label{psilmomega}
1089: \varphi_{lm}(\omega) =
1090: R_{lm}(r,\omega)S_{lm}(\theta,\omega)e^{-i\omega t + im\phi},
1091: \end{equation}
1092: implies that the radial function $R_{lm}(r,\omega)$ must satisfy
1093: \begin{equation} \label{Teukolsky}
1094: \left[\Delta^2 \frac{d}{dr}\left( \frac{1}{\Delta} \frac{d}{dr}\right)
1095: - \mathcal{V}_{lm}(r,\omega)\right] R_{lm}(r,\omega) = -
1096: \EuScript{T}_{lm}(r,\omega),
1097: \end{equation}
1098: where $\EuScript{T}_{lm}(r,\omega) =
1099: \left<lm\omega|\EuScript{T}\right>$, and the potential is given by
1100: \begin{equation}
1101: \mathcal{V}_{lm}(r,\omega) = -\frac{K^2+4i(r-M)K}{\Delta} + 8i\omega r + \lambda_{lm}(\omega).
1102: \end{equation}
1103: Here $K = (r^2 + a^2)\omega - ma$, and
1104: \begin{equation} \label{lambda eig}
1105: \lambda_{lm}(\omega) = -C_{lm}(\omega) - 2am\omega.
1106: \end{equation}
1107: The radial equation (\ref{radial}) is just the source free
1108: (homogeneous) version of the previous radial equation (\ref{Teukolsky}). The
1109: two independent solutions to the homogeneous radial equation will be
1110: used to construct the solution to the inhomogeneous equation. We use
1111: the so-called ``in-up basis'' $\{ R^\text{H}_{lm}(r,\omega),
1112: R^\infty_{lm}(r,\omega)\}$, which is defined by the following limiting
1113: behavior \cite{Drasco Flanagan Hughes}:
1114: \begin{subequations}\label{inup}
1115: \begin{align}
1116: R_{lm}^\text{H}(r\to r_+,\omega) &= B_{lm}^{\rm hole}(\omega)\Delta^2
1117: e^{-i P r^*},&
1118: \label{RH H} \\
1119: R_{lm}^\text{H}(r\to \infty,\omega) &= B_{lm}^{\rm out}(\omega)r^3
1120: e^{i\omega r^*} + \frac{B_{lm}^{\rm in}(\omega)}{r} e^{-i\omega
1121: r^*},&
1122: \label{RH inf}\\
1123: R_{lm}^\infty(r\to r_+,\omega) &= D_{lm}^{\rm out}(\omega)e^{i P r^*}
1124: + D_{lm}^{\rm in}(\omega) \Delta^2 e^{-i P r^*},&
1125: \label{Rinf H} \\
1126: R_{lm}^\infty(r\to \infty,\omega) &= D_{lm}^{\infty}(\omega)r^3
1127: e^{i\omega r^*}.&
1128: \label{Rinf inf}
1129: \end{align}
1130: \end{subequations}
1131: Here $P = \omega - ma/(2Mr_+)$, and the ``tortoise coordinate'' $r^*$
1132: satisfies $dr^*/dr = (r^2 + a^2)/\Delta$:
1133: \begin{equation}
1134: r^*(r) = r + \frac{2M r_+}{r_+ - r_-}\ln\frac{r-r_+}{2 M}
1135: - \frac{2M r_-}{r_+ - r_-}\ln\frac{r-r_-}{2 M} ,
1136: \label{eq:rstarofr}
1137: \end{equation}
1138: where $r_\pm = M \pm \sqrt{M^2 - a^2}$ are the roots of $\Delta$
1139: ($r=r_+$ is the location of the event horizon).
1140:
1141: The numerical calculation of the homogeneous solutions, say
1142: $R_{lm}^\text{H}(r)$, would in principle work as follows. First set
1143: $B_{lm}^\text{in}(\omega) = 1$ (a normalization convention). Then
1144: note that
1145: \begin{align}
1146: \frac{R_{lm}^\text{H}(r\to r_+)}{B_{lm}^{\rm hole}(\omega)} &=
1147: \Delta^2 e^{-i P r^*},&
1148: \\
1149: \frac{R_{lm}^\text{H}(r\to \infty)}{B_{lm}^{\rm hole}(\omega)}
1150: &= \frac{B_{lm}^{\rm out}(\omega)}{B_{lm}^{\rm hole}(\omega)}r^3
1151: e^{i\omega r^*} + \frac{1}{B_{lm}^{\rm hole}(\omega)r} e^{-i\omega
1152: r^*}.&
1153: \label{inf ratio}
1154: \end{align}
1155: Next starting near the horizon at $r = r_+$, integrate forward. When
1156: we reach $r \approx \infty$, we can read off
1157: $B^\text{out}/B^\text{hole}$ and $1/B^{\rm hole}$. (When clarity
1158: permits, we will often drop the somewhat cumbersome subscripts $l,m$
1159: and the dependence on $r,\omega$.) Since the first term in Eq.\
1160: (\ref{inf ratio}) grows rapidly with $r$, the two terms cannot be
1161: extracted with equal accuracy. We work around this by instead solving
1162: the homogeneous Sasaki-Nakamura equation \cite{SN} (which was designed
1163: to avoid this problem) and converting the result to $R^{H,\infty}$;
1164: see Ref.~\cite{circular} for details. Recently Fujita and Tagoshi
1165: have worked out a sophisticated numerical scheme for computing the
1166: functions $R^{H,\infty}$ numerically with great accuracy (with
1167: fractional accuracies $\sim10^{-14}$ in the corresponding energy
1168: fluxes for orbits which are both circular and equatorial) \cite{Fujita
1169: Tagoshi}.
1170:
1171: The general solution to the radial equation (\ref{Teukolsky}),
1172: corresponding to the retarded solution of the master equation
1173: (\ref{master}), is
1174: \begin{align}
1175: R_{lm}(r,\omega) =
1176: Z^{H}_{lm}(r,\omega)R^{\infty}_{lm}(r,\omega)
1177: + Z^{\infty}_{lm}(r,\omega)R^{H}_{lm}(r,\omega),
1178: \end{align}
1179: where the functions $Z$ are radial integrals over the source
1180: term:
1181: \begin{subequations}\label{Z}
1182: \begin{align}
1183: Z^\text{H}_{lm}(r,\omega) &= -\frac{1}{\mathcal{A}_{lm}(\omega)} \int_{r_+}^{r} dr'~
1184: \frac{R^\text{H}_{lm}(r',\omega)}{\Delta'^2} \EuScript{T}_{lm}(r',\omega),& \label{ZH} \\
1185: Z^\infty_{lm}(r,\omega) &= -\frac{1}{\mathcal{A}_{lm}(\omega)} \int_{r}^\infty dr'~
1186: \frac{R^\infty_{lm}(r',\omega)}{\Delta'^2} \EuScript{T}_{lm}(r',\omega)\;.& \label{ZI}
1187: \end{align}
1188: \end{subequations}
1189: The function $\mathcal{A}_{lm}(\omega)$ is given by\footnote{The value
1190: of $\mathcal{A}$ reported in Ref.~\cite{circular} has the wrong sign.}
1191: \begin{equation}
1192: \mathcal{A}_{lm}(\omega) = 2i\omega B^\text{in}_{lm}(\omega)D^\infty_{lm}(\omega).
1193: \end{equation}
1194: This construction of $Z^{\text{H},\infty}$ and ${\cal A}$ follows from
1195: the theory of Green's functions {\cite{Arfken}}.
1196:
1197: In this paper, we only evaluate the perturbed field $\psi_4$ [and
1198: therefore the functions $Z$ (\ref{Z})] outside the orbital torus ---
1199: $r < r_{\min}$ and $r > r_{\max}$. Evaluating $\psi_4$ at arbitrary
1200: locations within the orbital torus $r_{\min} < r < r_{\max}$ would
1201: require a more complicated numerical apparatus than the one developed
1202: here.
1203:
1204: \subsection{Bound test mass sources}
1205: \label{ss:Bound test mass sources}
1206:
1207: Beginning from Teukolsky's original expression for the source term
1208: $\EuScript{T}$ \cite{Teukolsky}, Breuer \cite{Breuer} has computed the
1209: explicit form of the projected source (see Ref.~\cite{Sasaki Tagoshi}
1210: or Sec.~IV C of Ref.~\cite{circular}):
1211: \begin{equation} \label{source}
1212: \EuScript{T}_{lm}(r,\omega) = \int d\Omega~dt~\EuScript{B}_m(t,r,\theta,\phi,\omega)S_{lm}(\theta,\omega)
1213: e^{i\omega t - im\phi},
1214: \end{equation}
1215: where
1216: \begin{eqnarray}
1217: \EuScript{B}_m &=&
1218: \sqrt{2} \Delta^2 \rho^3 L_{-1}[\rho^{-4}\bar\rho^2J_+(\rho^{-2}\bar\rho^{-2}\Delta^{-1}T_{n\bar m})] \nonumber \\
1219: &&-2 \rho^3 L_{-1}[\rho^{-4}L_0(\rho^{-2}\bar\rho^{-1}T_{nn})] \nonumber \\
1220: &&+\sqrt{2} \Delta^2 \rho^3 J_+[\rho^{-4}\bar\rho^2\Delta^{-1}L_{-1}(\rho^{-2}\bar\rho^{-2}T_{n\bar m})] \nonumber \\
1221: &&-\Delta^2 \rho^3 J_+[\rho^{-4}J_+(\rho^{-2}\bar\rho T_{\bar m\bar m})]\;.
1222: \end{eqnarray}
1223: This function is given in terms of the tetrad components of the
1224: orbiting particle's energy-momentum tensor,
1225: \begin{equation}
1226: T_{ab}= T^{\alpha\beta}a_\alpha b_\beta\;,
1227: \end{equation}
1228: where $T^{\alpha\beta}$ is the energy-momentum tensor of the particle,
1229: and $a$ and $b$ are either $n$ or $\bar m$. We define the derivative
1230: operators
1231: \begin{eqnarray}
1232: J_+&=& \partial_r + i K/\Delta\\
1233: L_s&=& \partial_\theta + m\csc\theta - a\omega \sin\theta + s\cot\theta.
1234: \end{eqnarray}
1235: In Kerr spacetime, a point particle of mass $\mu$ has an
1236: energy-momentum tensor given by
1237: \begin{equation} \label{energy momentum}
1238: T^{\alpha\beta}(t,r,\theta,\phi) = \frac{\mu u^\alpha u^\beta}{\Sigma \sin\theta} \frac{d\tau}{dt}
1239: \delta[r-r(t)]\delta[\theta-\theta(t)]\delta[\phi-\phi(t)],
1240: \end{equation}
1241: where $u^\alpha$ is the particle's 4-velocity. The quantities $r(t)$,
1242: $\theta(t)$, and $\phi(t)$ appearing in Eq.~(\ref{energy momentum})
1243: are the coordinates of the geodesic at coordinate time $t$, and should
1244: not be confused with $r$, $\theta$, and $\phi$ in the definition of
1245: the source term (\ref{source}).
1246:
1247: We use the Newman-Penrose basis given by Kinnersly
1248: \cite{Kinnersly} [see Eqs.~(\ref{K vectors}) and (\ref{K 1forms})].
1249: This means the tetrad components of $T^{\alpha\beta}$ are given by
1250: \begin{equation} \label{tradition}
1251: T_{ab} = \frac{C_{ab}}{\sin\theta}\delta[r-r(t)]\delta[\theta-\theta(t)]\delta[\phi-\phi(t)],
1252: \end{equation}
1253: where\footnote{In Ref.~\cite{Minoetalreview}, the $d\theta/d\tau$
1254: terms are missing from the expressions for $C_{n\bar m}$ and $C_{\bar
1255: m \bar m}$. Also, in Ref.~\cite{equatorial} $d\theta/d\lambda$ is
1256: erroneously replaced with $(d\theta/d\lambda)^2$.} \cite{Sasaki
1257: Tagoshi}
1258: \begin{subequations} \label{C's}
1259: \begin{align}
1260: C_{nn}&= \frac{d\lambda}{dt}\frac{\mu}{4\Sigma^2}
1261: \left[ E\left(r^2+a^2\right) - aL_z + \frac{dr}{d\lambda} \right]^2,&
1262: \label{Cnn}\\
1263: C_{\bar m\bar m}&= \frac{d\lambda}{dt}\frac{\mu\rho^2}{2}
1264: \left[i\left(aE-\frac{L_z}{\sin^2\theta}\right)\sin\theta + \frac{d\theta}{d\lambda}\right]^2,&
1265: \label{Cnm}\\
1266: C_{n \bar m}&= \frac{d\lambda}{dt} \frac{\mu\rho}{2\sqrt{2}\Sigma}
1267: \left[ E\left(r^2+a^2\right) - aL_z + \frac{dr}{d\lambda}\right]&
1268: \nonumber \\
1269: & \times \left[ i\left(aE-\frac{L_z}{\sin^2\theta}\right)\sin\theta + \frac{d\theta}{d\lambda}\right].&
1270: \label{Cmm}
1271: \end{align}
1272: \end{subequations}
1273: These functions (\ref{C's}) depend explicitly on $dr/d\lambda$ and
1274: $d\theta/d\lambda$, rather than on their squares $V_{r,\theta}$.
1275: Extra bookkeeping is thus needed to keep track of the signs of
1276: $dr/d\lambda$ and $d\theta/d\lambda$ when evaluating them numerically.
1277:
1278: Following the details shown in (for example) Ref.~\cite{circular}, we
1279: can now write the projected source term (\ref{source}) in the
1280: following form
1281: \begin{eqnarray} \label{TandA}
1282: \EuScript{T}_{lm}(r,\omega)&=&\int dt~\Delta^2
1283: \{[A_{nn0} + A_{n\bar m 0} + A_{\bar m \bar m 0}] \delta[r-r(t)] \nonumber\\
1284: &&+ \partial_r[ (A_{n\bar m 1} + A_{\bar m \bar m 1}) \delta(r-r\{t\})] \nonumber\\
1285: &&+ \partial^2_r[A_{\bar m \bar m 2}\delta(r-r\{t\})] \}e^{i\omega t - im\phi(t)}\;.
1286: \end{eqnarray}
1287: The quantities $A_{abc}$ are shown explicitly in Appendix
1288: \ref{s:Perturbation details}.
1289:
1290: Substituting Eq.~(\ref{TandA}) into Eq.~(\ref{Z}) and eliminating the
1291: radial delta functions gives
1292: \begin{eqnarray} \label{ZA}
1293: Z_{lm}^\star(r,\omega) &=& -\frac{1}{A} \int dt~ e^{i\omega t -
1294: im\phi(t)} \Theta^{\star}[r,r(t)]\nonumber\\ && \left[
1295: (A_{nn0} + A_{n\bar m 0} + A_{\bar m \bar m 0}) \right. \nonumber\\ &&
1296: - (A_{n\bar m 1} + A_{\bar m \bar m 1}) \frac{d}{dr} + \left. A_{\bar
1297: m \bar m 2} \frac{d^2}{dr^2}\right] R_{lm}^\star, \nonumber\\
1298: \end{eqnarray}
1299: where $\star = \text{H},\infty$. We define the step functions
1300: \begin{equation}
1301: \Theta^\infty(x_1,x_2) = \Theta^\text{H}(x_2,x_1) = \Theta(x_2 -x_1),
1302: \end{equation}
1303: in terms of the Heaviside step function $\Theta(x)$. All other
1304: functions of $r$ and $\theta$ under the integral in Eq.~(\ref{ZA}) are
1305: to be evaluated at $r(t)$ and $\theta(t)$ respectively. Because $r$
1306: only appears inside the step functions, the quantities
1307: $Z^{\text{H},\infty}$ are independent of $r$ for all $r > r_{\max}$
1308: and $r < r_{\min}$.
1309:
1310: To clean up this expression, we now absorb most of the integrand into
1311: a single function:
1312: \begin{align} \label{short}
1313: Z^\star_{lm}(r,\omega) =\int dt~e^{i\omega t - im\phi(t)}
1314: I_{lm}^\star(t,r,\omega)\;.
1315: \end{align}
1316: Following the arguments in Sec.\ V B of Ref.\ {\cite{Drasco Hughes}},
1317: we exploit the harmonic structure of the geodesics (Sec.~\ref{s:Bound
1318: black hole orbits}) to simplify Eq.~(\ref{short}). In the end, we
1319: will arrive at a general expression for $\psi_4$ (\ref{psi_4}) as a
1320: discrete sum over frequencies, rather than an integral. First we
1321: insert
1322: \begin{align}
1323: &t = \Gamma \lambda + \Delta t,&
1324: &\phi = \Upsilon_\phi \lambda + \Delta \phi,&
1325: \end{align}
1326: into Eq.~(\ref{short}), and we change the integration variable from
1327: $t$ to $\lambda$. The result is
1328: \begin{eqnarray} \label{new short}
1329: Z^\star_{lm}(r,\omega) = \int d\lambda~ e^{i(\omega\Gamma - m\Upsilon_\phi)\lambda}
1330: J_{lm}^\star(\lambda,r,\omega) ,
1331: \end{eqnarray}
1332: where
1333: \begin{align} \label{J}
1334: J_{lm}^\star(\lambda,r,\omega) = \frac{dt}{d\lambda}
1335: e^{i\omega \Delta t - im\Delta\phi}
1336: I_{lm}^\star(\lambda,r,\omega)\;.
1337: \end{align}
1338: The function $J_{lm}^\star(\lambda,r,\omega)$ is biperiodic, so we can
1339: write it as \cite{Drasco Hughes}
1340: \begin{equation} \label{J exp}
1341: J_{lm}^\star(\lambda,r,\omega) = \sum_{kn} J^{\star}_{lmkn} e^{-i(k \Upsilon_\theta + n \Upsilon_r)\lambda}\;;
1342: \end{equation}
1343: the constants $J^\star_{lmkn}$ are given by
1344: \begin{eqnarray}
1345: J^\star_{lmkn} &=& \frac{1}{(2\pi)^2}\int_0^{2\pi}dw_\theta \int_0^{2\pi}dw_r~e^{i(k w_\theta + n w_r)}
1346: \nonumber \\
1347: &&\times J^\star_{lm}[r(w_r),\theta(w_\theta),r,\omega] .
1348: \end{eqnarray}
1349: By inserting the expansion (\ref{J exp}) into Eq.~(\ref{new short}),
1350: we find that the integral (\ref{new short}) is just a sum of delta
1351: functions:
1352: \begin{equation} \label{zdelta}
1353: Z^\star_{lm}(r,\omega) = \sum_{kn} Z^\star_{lmkn}(r) \delta(\omega -
1354: \omega_{mkn})\;.
1355: \end{equation}
1356: We have used the coordinate time frequencies (\ref{t frequencies}) to
1357: define
1358: \begin{equation}
1359: \omega_{mkn} = m\Omega_\phi + k\Omega_\theta + n\Omega_r.
1360: \end{equation}
1361: The expansion coefficients for $Z^\star_{lm}(r,\omega)$ are given by
1362: \begin{align} \label{long Zlmnk}
1363: Z^\star_{lmkn}(r) =
1364: &\frac{1}{2\pi\Gamma}\int_0^{2\pi}dw_\theta~\int_0^{2\pi} dw_r~e^{ikw_\theta + inw_r}&
1365: \nonumber\\
1366: &\times J^\star_{lm}(w_r,w_\theta,r,\omega_{mkn}),&
1367: \end{align}
1368: By substituting the expanded form (\ref{zdelta}) of $Z$ into the
1369: general expression (\ref{psi_4}) for $\psi_4$, we eliminate the
1370: frequency integral to obtain
1371: \begin{equation}
1372: \psi_4 = \rho^4 \sum_{lmkn} R_{lmkn}(r)S_{lmkn}(\theta)
1373: e^{-i\omega_{mkn}t+im\phi} \label{psi exp}\;.
1374: \end{equation}
1375: The radial and angular functions are now discrete functions of
1376: frequency, inheriting the angular harmonic index $k$ and the radial
1377: harmonic index $n$:
1378: \begin{align}
1379: S_{lmkn}(\theta) = &S_{lm}(\theta,\omega_{mkn}),& \\
1380: R^\star_{lmkn}(r) = &R^\star_{lm}(r,\omega_{mkn}),& \\
1381: R_{lmkn}(r) = &Z^\text{H}_{lmkn}(r)R^{\infty}_{lmkn}(r)
1382: + Z^{\infty}_{lmkn}(r)R^\text{H}_{lmkn}(r).
1383: \label{eq:greens}&
1384: \end{align}
1385:
1386: \subsection{Numerical considerations}
1387: \label{ss:Numerical considerations}
1388:
1389: Because the functions $r(w_r)$ and $\theta(w_\theta)$ are periodic and
1390: even [cf.\ Eq.\ (\ref{even coords})], the range of the integrals
1391: (\ref{long Zlmnk}) can be reduced by a factor of two. Writing the
1392: integrand in (\ref{long Zlmnk}) as $z(w_r,w_\theta)$, we can expand
1393: the integrals into the form
1394: \begin{align}
1395: Z =
1396: &\left[ \int_0^{\pi}\!dw_\theta \int_0^{\pi} \!dw_r
1397: + \int_0^{\pi}\!dw_\theta \int_\pi^{2\pi} \!dw_r \right.&
1398: \nonumber \\
1399: &\left.\!\!+ \int_\pi^{2\pi}\!dw_\theta \int_0^{\pi} \!dw_r
1400: + \int_\pi^{2\pi}\!dw_\theta \int_\pi^{2\pi} \!dw_r \right] z(w_r,w_\theta).&
1401: \end{align}
1402: Since the integrand has a period of $2\pi$ in both variables, we can
1403: shift all of the $\{\pi,2\pi\}$ branches to $\{-\pi,0\}$:
1404: \begin{align}
1405: Z =
1406: &\left[ \int_0^{\pi}dw_\theta \int_0^{\pi} dw_r
1407: + \int_0^{\pi}dw_\theta \int_{-\pi}^{0} dw_r \right.&
1408: \nonumber \\
1409: &\left.\!\!+ \int_{-\pi}^0dw_\theta \int_0^{\pi} dw_r
1410: + \int_{-\pi}^0 dw_\theta \int_{-\pi}^0 dw_r \right] z(w_r,w_\theta).&
1411: \end{align}
1412: We can now reflect all of the $\{-\pi,0\}$ branches across the origin to get
1413: \begin{align} \label{short z}
1414: Z = \int_0^{\pi}dw_\theta \int_0^{\pi} dw_r \sum_{D_r = \pm} \sum_{D_\theta = \pm}
1415: z(D_r w_r,D_\theta w_\theta)\;.
1416: \end{align}
1417:
1418: Direct evaluation of Eq.~(\ref{short z}) is complicated by the fact
1419: that it is difficult to evaluate $r(w_r)$ and $\theta(w_\theta)$ [see
1420: discussion preceeding Eq.~(\ref{final delta})]. We avoid this problem
1421: by changing integration variables from $w_r$ and $w_\theta$ to $\psi$
1422: and $\chi$ [see Eq.~(\ref{new coords})].
1423:
1424: Eccentric orbits introduce another numerical nuance. The integrand in
1425: Eq.~(\ref{short z}) is moderately divergent if the orbit covers a
1426: large range in $r$ (i.e., if the orbit is highly eccentric). This
1427: problem can be traced to the scaling behavior (\ref{inup}) of
1428: solutions to the homogeneous radial equation. In particular, the
1429: typical size of the integrand at $r_{\min}$ can be different from
1430: typical values at $r_{\max}$ by orders of magnitude. This causes an
1431: artificial focusing of numerical accuracy. Our choice of boundary
1432: conditions (\ref{initial rt}) is such that the turning points
1433: $r_{\min}$ and $r_{\max}$ occur at $\psi = 0$ and $\psi = \pi$
1434: respectively. To avoid this problem, we introduce a change of
1435: integration variables from $\psi$ to $\zeta$ where $d\psi/d\zeta =
1436: e^{-\psi/\Psi^\star}$, with $\star = \text{H},\infty$, and
1437: \begin{align}
1438: \zeta(\psi) = \Psi^\star \left( e^{\psi/\Psi^\star} - 1\right).
1439: \end{align}
1440: Here $\Psi^\star$ is a constant (which can be positive or negative).
1441: Before integrating, we tune the value of $\Psi^\star$ so that the
1442: typical magnitude of the integrand at the outer turning point is
1443: comparable to that at the inner turning point. In general, the value
1444: of $\Psi^\star$ will decrease if either $l$ or eccentricity $e$ are
1445: increased.
1446:
1447: Applying these various changes of variable and massaging of the
1448: integration ranges to Eq.~(\ref{long Zlmnk}), our final expression for
1449: the $Z^\star_{lmkn}(r)$ coefficients is given by
1450: \begin{align} \label{final Z}
1451: Z^\star_{lmkn}(r) =
1452: &\frac{1}{2\pi\Gamma}\int_0^{\pi}d\chi \int_0^{\zeta(\pi)}\!\! d\zeta
1453: \sum_{D_r = \pm} \sum_{D_\theta = \pm}\!\!
1454: \frac{dw_\theta}{d\chi}\frac{dw_r}{d\psi}\frac{dt}{d\lambda}&
1455: \nonumber \\
1456: &\times e^{-\psi /\Psi^\star} I^\star_{lm}\left(w_r, w_\theta, D_r, D_\theta, r, \omega_{mkn} \right)&
1457: \nonumber \\
1458: &\times \exp \left[iD_\theta \left( kw_\theta + \omega_{mkn} \Delta t^\theta - m \Delta \phi^\theta \right)\right]&
1459: \nonumber \\
1460: &\times \exp \left[iD_r \left( nw_r + \omega_{mkn} \Delta t^r - m \Delta \phi^r \right)\right],&
1461: \end{align}
1462: where $I^\star_{lm}$ is defined by Eq.~(\ref{short}). The functions
1463: $r(t)$, $\theta(t)$ and their derivatives are evaluated as
1464: \begin{align}
1465: &r(t) \to r(\zeta),&
1466: &\frac{dr}{d\lambda}[r(t)] \to D_r\left| \frac{dr}{d\lambda}[r(\zeta)] \right|,&
1467: \nonumber \\
1468: &\theta(t) \to \theta(\chi),&
1469: &\frac{d\theta}{d\lambda}[\theta(t)] \to D_\theta\left| \frac{d\theta}{d\lambda}[\theta(\chi)] \right|.&
1470: \end{align}
1471:
1472: Lastly, note that applying the simultaneous transformations $m\to -m$,
1473: $\omega\to -\omega$ to the radial (\ref{Teukolsky}) and the angular
1474: equations (\ref{angular}), one can derive a simple symmetry rule for
1475: expansion coefficients \cite{circular,equatorial}
1476: \begin{equation} \label{Zsym}
1477: Z^\star_{l(-m)(-k)(-n)}(r) = (-1)^{l+k} \bar Z^\star_{lmkn}(r).
1478: \end{equation}
1479: This symmetry is used to reduce the computation time for various
1480: summations (see Sec.~\ref{ss:Truncation}).
1481:
1482: \subsection{Waveforms and fluxes}
1483: \label{ss:Waveforms and fluxes}
1484:
1485: In this section, we use the expanded form of the curvature
1486: perturbation (\ref{psi exp}) to derive expressions for the leading
1487: order metric perturbation at infinity, and for the radiated fluxes of
1488: energy and angular momentum to infinity and down the black hole's
1489: event horizon. Since we will be interested in quantities at both
1490: infinity and the horizon, it will be useful to first define the
1491: limiting values of the coefficients $Z_{lmkn}^\star(r)$ as follows:
1492: \begin{subequations}\label{lim Zs}
1493: \begin{align}
1494: &Z_{lmkn}^\text{H} = D^\infty_{lmkn} Z_{lmkn}^\text{H}(r > r_{\max}), &
1495: \label{lim ZH}
1496: \\
1497: &Z_{lmkn}^\infty = B^\text{hole}_{lmkn} Z_{lmkn}^\infty(r < r_{\min}).&
1498: \label{lim ZI}
1499: \end{align}
1500: \end{subequations}
1501: Recall that the functions $Z_{lmkn}^\star(r)$ are independent of $r$ for all
1502: $r > r_{\max}$ and $r < r_{\min}$.
1503:
1504: At infinity, the leading order curvature and metric perturbations are
1505: related by
1506: \begin{equation} \label{curvature perturbation}
1507: \psi_4(r \to \infty) = \frac{1}{2}\frac{\partial^2}{\partial t^2}
1508: \left( h_+ - ih_\times \right).
1509: \end{equation}
1510: Here $h_+$ and $h_\times$ are the two independent components of the metric
1511: perturbation, defined by
1512: \begin{equation} \label{metric perturbation}
1513: h_{\alpha\beta}(r\to\infty) = h_+ e^+_{\alpha\beta} + h_\times e^\times_{\alpha\beta} + O(1/r^2)\;,
1514: \end{equation}
1515: where $e^{+}_{\alpha\beta}$ and $e^{\times}_{\alpha\beta}$ are
1516: polarization tensors \cite{Blandford Thorne}. If we now substitute
1517: the expanded form of the curvature perturbation (\ref{psi exp}) into
1518: the left hand side of Eq.~(\ref{curvature perturbation}), evaluate the
1519: limit, and integrate twice in coordinate time (setting the arbitrary
1520: integration constants to zero), we find
1521: \begin{align} \label{waveform}
1522: h_+ - i h_\times = -\frac{2}{r}\sum_{lmkn} \frac{Z_{lmkn}^\text{H}}{\omega_{mkn}^{2}}
1523: S_{lmkn}(\theta) e^{-i\omega_{mkn}(t-r)+im\phi}\;.
1524: \end{align}
1525: We have used the definition (\ref{lim ZH}) for the limiting value of
1526: $Z_{lmkn}^\text{H}(r)$.
1527:
1528: It is important to note that the expression for $Z^\text{H}_{lmkn}$
1529: derived above [e.g. Eq.~(\ref{final Z})] assumed geodesics with a
1530: specific choice of initial position: $r(0) = r_{\min}$, $\theta(0) =
1531: \theta_{\min}$, $t(0) = 0$, and $\phi(0) = 0$. These are the fiducial
1532: geodesics from Ref.~\cite{Drasco Flanagan Hughes}. Making a different
1533: choice of initial position will result in an overall phase change for
1534: $Z^\text{H}_{lmkn}$. The details of this phase can be found in
1535: Ref.~\cite{Drasco Flanagan Hughes}. As we will see below, the
1536: expression for the waveform (\ref{waveform}) is unique in that, unlike
1537: the formulas for evolving $E$, $L_z$, it depends explicitly on this
1538: overall phase, and therefore on the initial position of the geodesic.
1539:
1540: At infinity, the effective energy-momentum tensor
1541: $T^\text{GW}_{\alpha\beta}$ is easily built from the GWs
1542: \cite{Isaacson}. The result can be expressed as \cite{Blandford
1543: Thorne}
1544: \begin{equation}
1545: T^\text{GW}_{\alpha\beta} = \frac{1}{16\pi}
1546: \left<
1547: \frac{\partial h_+}{\partial x^\alpha} \frac{\partial h_+}{\partial x^\beta}
1548: +
1549: \frac{\partial h_\times}{\partial x^\alpha} \frac{\partial h_\times}{\partial x^\beta}
1550: \right>\;.
1551: \end{equation}
1552: Brackets indicate an average over several wavelengths of the
1553: gravitational waves. Using this result and the expansion
1554: (\ref{waveform}), it is straightforward to show that waves carry an
1555: averaged flux of energy and angular momentum given by
1556: \begin{subequations}
1557: \begin{align}
1558: &\left< \frac{dE}{dt} \right>^\infty = \sum_{lmkn} \frac{1}{4\pi\omega_{mkn}^2} \left| Z_{lmkn}^\text{H}\right|^2,& \\
1559: &\left< \frac{dL_z}{dt} \right>^\infty = \sum_{lmkn}\frac{m}{4\pi\omega_{mkn}^3} \left| Z_{lmkn}^\text{H}\right|^2.&
1560: \end{align}
1561: \end{subequations}
1562: Similar expressions can be found for the flux of energy and angular
1563: momentum through the horizon; we refer the reader to
1564: Refs.~\cite{Poisson PDE,teukpress,circular} for a detailed derivation.
1565: The resulting fluxes are
1566: \begin{subequations} \label{Horizon fluxes}
1567: \begin{align}
1568: &\left< \frac{dE}{dt} \right>^\text{H} = \sum_{lmkn} \frac{1}{4\pi\omega_{mkn}^2} \alpha_{lmkn}\left| Z_{lmkn}^\infty\right|^2,& \\
1569: &\left< \frac{dL_z}{dt} \right>^\text{H} = \sum_{lmkn}\frac{m}{4\pi\omega_{mkn}^3}\alpha_{lmkn} \left| Z_{lmkn}^\infty\right|^2.&
1570: \end{align}
1571: \end{subequations}
1572: The superscripts we use here are somewhat confusing: the fluxes to
1573: infinity depend upon the {\it horizon} coefficients,
1574: $Z^\text{H}_{lmkn}$, and the fluxes down the horizon depend upon the
1575: {\it infinity} coefficients $Z^\infty_{lmkn}$. This seemingly obtuse
1576: convention comes from the Green's function solution we constructed,
1577: Eq.\ (\ref{eq:greens}). The coefficients $Z^\text{H}$ set the
1578: amplitude of the radial behavior towards infinity; the coefficients
1579: $Z^\infty$ set the amplitude towards the horizon.
1580:
1581: The coefficient $\alpha_{lmkn}$ appearing in Eq.\ (\ref{Horizon
1582: fluxes}) is given by \cite{circular}
1583: \begin{equation}
1584: \alpha_{lmkn} = \frac{256 (2Mr_+)^5 P(P^2 + 4\epsilon^2)(P^2 +
1585: 16\epsilon^2)\omega_{mkn}^3}{|C_{lmkn}|^2}\;,
1586: \label{eq:horizonalphadef}
1587: \end{equation}
1588: with $\epsilon = \sqrt{M^2 - a^2}/4Mr_+$, where $r_+$ is the location of the event horizon, and
1589: \begin{align}
1590: |C_{lmkn}|^2 &= \left[(\lambda_{lmkn}+2)^2 + 4 a \omega_{mkn} - 4 a^2 \omega_{mkn}^2\right]& \nonumber \\
1591: &\times\left(\lambda_{lmkn}^2 + 36 m a \omega_{mkn} - 36 a^2\omega_{mkn}^2\right)& \nonumber\\
1592: &+ \left(2\lambda_{lmkn}+3\right) \left(96 a^2\omega_{mkn}^2 - 48 m a \omega_{mkn}\right) & \nonumber\\
1593: &+ 144 \omega_{mkn}^2(M^2 - a^2).&
1594: \end{align}
1595: Recall that $P = \omega_{mkn} - ma/(2Mr_+)$, and that $\lambda_{lmkn}$
1596: is related to angular equation's eigenvalue [cf.\ Eq.\
1597: (\ref{angular})] via Eq.~(\ref{lambda eig}) with $\omega =
1598: \omega_{mkn}$.
1599:
1600: To go from snapshot waveforms to adiabatic inspiral waveforms, we must
1601: evolve not just energy and angular momentum, but also the Carter
1602: constant $Q$. This will eventually be done with the formalism
1603: recently derived by Mino \cite{Mino}. However, if we assume that
1604: radiation does not influence an orbit's inclination [$\iota$ or
1605: equivalently $\theta_\text{inc}$], we can can infer an expression for
1606: $\left< dQ/dt \right>$. This scheme\ was first suggested by Curt
1607: Cutler \cite{Cutler private}. From the definition (\ref{iota}) of
1608: $\iota$, we see that setting $\left< d\iota/dt \right> = 0$ gives
1609: \begin{equation}\label{Qdot}
1610: \left< \frac{dQ}{dt} \right> = \frac{2Q}{L_z} \left< \frac{dL_z}{dt} \right>,
1611: \end{equation}
1612: where $\left< dL_z/dt \right> = \left< dL_z/dt \right>^\text{H} +
1613: \left< dL_z/dt \right>^\infty$. Hughes showed that this approximation
1614: is reasonable in the case of circular orbits \cite{circularII}. It is
1615: currently unknown how accurate it may be in the case of generic orbits
1616: (though it is almost certainly {\it pathological} for orbits with
1617: nearly vanishing $L_z$ {\cite{gair_private}}). In
1618: Sec.~\ref{s:Results} below, we report values of $\left< dQ/dt \right>$
1619: given by Eq.~(\ref{Qdot}). We should emphasize however that, since
1620: Eq.~(\ref{Qdot}) is itself a speculation, those results will very
1621: likely be less accurate than the associated fluxes of energy and
1622: angular momentum.
1623:
1624:
1625: \section{Numerical algorithm}
1626: \label{s:Numerical algorithm}
1627:
1628: For a given orbit, the calculation of the snapshot waveform is
1629: essentially identical to the calculation of each of the radiative
1630: fluxes. Given an engine which can compute the $Z_{lmkn}^\star$
1631: coefficients, which we will refer to simply as modes, each calculation
1632: is essentially just a long four dimensional summation. In the
1633: following three subsections, we describe the details of our algorithms
1634: for (i) computing the frequency domain representations of the
1635: geodesics, (ii) computing the individual modes $Z_{lmkn}^\star$, and
1636: (iii) truncating the sums representing the waveforms and fluxes.
1637:
1638: \subsection{Geodesics}
1639: \label{ss:Geodesics}
1640:
1641: We begin by computing the quantities associated with the specified
1642: geodesic for a given set of black hole and orbital parameters
1643: $(a,e,p,\theta_\text{inc})$. This calculation proceeds as follows:
1644: \begin{enumerate}
1645:
1646: \item Compute the constants of the motion $E$, $L_z$, $Q$
1647:
1648: \item Compute the orbital frequencies $\Upsilon_{r,\theta,\phi}$, and
1649: $\Gamma$
1650:
1651: \item Construct routines to evaluate $w_r(\psi)$, $w_\theta(\chi)$ and
1652: their derivatives
1653:
1654: \item Construct the expansions (\ref{expanded x}) of $\Delta t$ and
1655: $\Delta \phi$
1656:
1657: \end{enumerate}
1658: Explicit expressions for the first three of these steps are shown in
1659: Appendix \ref{s:Geodesic details}. The last step requires that we
1660: approximate the sums for the expansions (\ref{expanded x}) of $\Delta
1661: x= \Delta t,\Delta \phi$ as
1662: \begin{align}
1663: &\Delta x^\theta \approx \sum_{k=1}^K \Delta x_k^\theta \sin(k w_\theta),&
1664: &\Delta x^r \approx \sum_{n=1}^N \Delta x_n^r \sin(n w_r),&
1665: \end{align}
1666: where we determine the values of $K$ and $N$ by requiring that 15
1667: consecutive $\Delta x_{k,n}^{r,\theta}$ coefficients make fractional
1668: contributions to the sum which are less than some specified accuracy;
1669: we typically find $K,N\lesssim 50$. Each of these calculations
1670: associated with the geodesics is done to a fractional accuracy of
1671: $10^{-12}$, and the results are made available to the remaining
1672: computations. The calculations described so far are typically
1673: completed in a few seconds (using a single $\sim 1$ GHz processor).
1674: It's worth bearing in mind that these calculations need only be done
1675: once when computing the waveform and fluxes for a given orbit.
1676:
1677: \subsection{Modes}
1678: \label{ss:Modes}
1679:
1680: After computing geodesic quantities, we construct an engine for
1681: calculating the modes $Z^\star_{lmkn}$. This requires solutions to
1682: the angular equation, and to the homogeneous radial equation. (Our
1683: homogenenous solutions have an implicit coupling to the source, since
1684: they depend on the frequencies $\omega_{mkn}$, which are determined by
1685: the orbit.) The routines for solving the homogeneous radial and
1686: angular equations were inherited from the code used in
1687: Ref.~\cite{circular}, where they are described in detail. We solve
1688: the angular equation to nearly machine accuracy, and we solve the
1689: radial equation to a fractional accuracy of $\min(10^{-7},
1690: \varepsilon_\text{flux}/100)$, where $\varepsilon_\text{flux}$ is the
1691: overall fractional accuracy demanded of the waveforms and fluxes.
1692:
1693: When computing (\ref{final Z}), we treat the radial integral as the
1694: outermost integral [the reverse of how we have written Eq.~(\ref{final
1695: Z})]. After experimenting with a variety of numerical methods for
1696: evaluating these integrals, we found a Clenshaw-Curtis quadrature
1697: algorithm to be the most efficient. This method analytically
1698: integrates a numerical series representation (in Chebyshev
1699: polynomials) of the integrand, to produce a series representation of
1700: the integral (for more details, see for example Ref.~\cite{NR}). Each
1701: of these calculations depends on the indices $(l,m,k,n)$, and must be
1702: repeated many times to compute the waveform and fluxes associated with
1703: a given orbit.
1704:
1705: We compute the angular integral in Eq.~(\ref{final Z}) to a fractional
1706: accuracy of $\min(10^{-6},\varepsilon_\text{flux}/10)$. The accuracy
1707: demanded of the radial integral is chosen dynamically. We begin by
1708: asking for a fractional accuracy of
1709: $\min(10^{-5},\varepsilon_\text{flux})$. As each mode is computed, we
1710: store the magnitude of the largest modes encountered so far. Later
1711: modes are then computed to a fractional accuracy of at least 10\%.
1712: After this 10\% accuracy has been achieved the integrator continues to
1713: add more terms to the Chebyshev expansion. At each iteration it
1714: estimates the associated uncertainty in the total energy fluxes (the
1715: energy flux at the horizon for the $Z^\infty_{lmkn}$ integrals, and
1716: the energy flux at infinity for the $Z^\text{H}_{lmkn}$ integrals) by
1717: comparing to the magnitudes of the largest known modes. It then
1718: truncates the Chebyshev expansion as soon as the integral's associated
1719: relative flux error is smaller than $\varepsilon_\text{flux}$. The
1720: reason for this dynamic accuracy control is that many of the modes are
1721: very near zero. If we know that the total flux contains modes which
1722: are $\sim 10^{-3}$, there is no need to compute a mode which is $\sim
1723: 10^{-20}$ beyond the first couple of digits. A single $\sim 1$ GHz
1724: processor usually takes considerably less than a second to compute
1725: both $Z^\infty_{lmkn}$ and $Z^\text{H}_{lmkn}$ (see Fig.~1 in
1726: Ref.~\cite{Hughes et al}), unless the modes are especially badly
1727: behaved in the sense described in Sec.~\ref{ss:Truncation} below. In
1728: these rare cases the modes can take anywhere from 10 seconds to a
1729: minute to compute.
1730:
1731: \begin{figure}
1732: \includegraphics[width = .45\textwidth]{diverge}
1733: \caption{Modal energy flux through the horizon (superscript H, marked
1734: with $\times$'s) and at radial infinity (superscript $\infty$, marked
1735: with $+$'s). Results are for $l = m = 4$, $k = 0$. The orbit has
1736: eccentricity $e = 0.7$, inclination $\theta_\text{inc} = 45^\circ$,
1737: and semilatus rectum $p = 5.1$, while the massive black hole has $a =
1738: 0.9M$.}
1739: \label{f:diverge}
1740: \end{figure}
1741:
1742: When computing a flux, the majority of the numerical work is devoted
1743: to computing modes which are insignificant or zero. The integrator
1744: has little difficulty when computing the dominant modes. However,
1745: modes which are very near zero (near round off error) can cause the
1746: integrator to fail. This typically happens when computing modes with
1747: ``large'' values of the indices ($l,m,k,n$); the meaning of ``large''
1748: is somewhat orbit and index dependent. For example, for fixed
1749: ($l,m,k$), we find that there is some limiting value of $n$ beyond
1750: which the mode calculations cannot be trusted.
1751:
1752: We show an example of this behavior in Fig.~\ref{f:diverge}. Beyond
1753: the dominant modes at $n \simeq 20$, the mode magnitudes fall
1754: (roughly) exponentially until reaching a bottom. For the case shown
1755: in Fig.~\ref{f:diverge}, the horizon modes bottom out at about $n=70$,
1756: after which they gradually diverge, while the infinity modes bottom
1757: out at about $n=90$, after which they appear to be dominated by
1758: numerical error. Both behaviors are unphysical since they indicate a
1759: divergent total energy flux. The different behaviors arise from the
1760: different scalings of these modes with frequency:
1761: \begin{eqnarray}
1762: \left\langle\frac{dE}{dt}\right\rangle^\infty_{lmkn} \sim
1763: \omega_{mkn}^{-2} |Z^\text{H}_{lmkn}|^2\;,
1764: \label{Einf scaling}\\
1765: \left\langle \frac{dE}{dt}\right\rangle^\text{H}_{lmkn} \sim
1766: \omega_{mkn}^2 |Z^\infty_{lmkn}|^2\;.
1767: \label{EH scaling}
1768: \end{eqnarray}
1769: This difference comes from the factor $\alpha_{lmkn}\sim
1770: \omega_{mkn}^4$ in Eq.~(\ref{Horizon fluxes}).
1771:
1772: For reasonable requested accuracies, the modes which are difficult to
1773: compute are always insignificant when compared to the dominant modes
1774: (as is clearly the case in Fig.~\ref{f:diverge}). In the vast
1775: majority of cases, the truncation rules (see Sec.~\ref{ss:Truncation}
1776: below) have been invoked long before these badly behaved modes are
1777: encountered. If this is not the case, these badly behaved modes can
1778: undermine the truncation scheme since, despite their small magnitudes,
1779: they do not regularly decay with increasing index. In order to
1780: account for this, we artificially zero any modes for which the ratio
1781: of the magnitude of the integral $Z^\star_{lmkn}$ to the maximum over
1782: $\zeta$ and $\chi$ of the absolute value of the integrand in
1783: Eq.~(\ref{final Z}) is below some threshold. We have found that this
1784: ratio is a good indicator of when the modes have bottomed out in the
1785: sense shown in Fig.~\ref{f:diverge}. Unfortunately, we have had to
1786: ``hand-tune'' this parameter somewhat. This was particularly so as we
1787: explored orbits with increasing eccentricity. In the future, we
1788: expect to incorporate our code into one which computes inspiral
1789: waveforms. Hand tuning of this parameter will be unacceptable for
1790: such a code, and a more robust technique will be needed.
1791:
1792: \subsection{Truncation}
1793: \label{ss:Truncation}
1794:
1795: Given the routines described above, the remaining task is to evaluate
1796: the four dimensional sums which make up the waveforms and fluxes
1797: (Sec.~\ref{ss:Waveforms and fluxes}). As mentioned above, we do this
1798: by monitoring the sums which represent the fluxes. We assume that
1799: once the fluxes have converged to some requested fractional accuracy
1800: $\varepsilon_\text{flux}$, the sums representing the waveforms are
1801: comparably accurate since they are constructed from the same mode
1802: coefficients as were used for the fluxes. We now discuss the
1803: algorithm for truncating the sums which represent the fluxes.
1804:
1805: Let $F$ be the flux of energy or angular momentum at the horizon or at
1806: infinity
1807: \begin{equation}
1808: F = \left< \frac{dE}{dt} \right>^{\text{H},\infty}
1809: , \quad \left< \frac{dL_z}{dt} \right>^{\text{H},\infty}.
1810: \end{equation}
1811: The algorithm for computing these fluxes is:
1812: \begin{subequations}\label{sum outline}
1813: \begin{align}
1814: F &= \sum_{l=2}^\infty F_l,&
1815: \\
1816: F_l &= \sum_{m=-l}^l F_{lm},&
1817: \\
1818: F_{lm} &= \sum_{k=-\infty}^\infty F_{lmk},&
1819: \\
1820: F_{lmk} &= F_{lmk0} + 2\sum_{n=1}^\infty F_{lmkn}.&
1821: \end{align}
1822: \end{subequations}
1823: We have used the symmetry (\ref{Zsym}), which gives
1824: \begin{equation}
1825: F_{lmkn} = F_{l(-m)(-k)(-n)},
1826: \end{equation}
1827: to simplify the sum over the radial harmonic index $n$. Each of the
1828: sums which has an infinite boundary is then truncated as described
1829: below.
1830:
1831: The terms which make up outermost sum, $F_l$ in the sum over $l$,
1832: always decrease monotonically with increasing $l$. This means that
1833: the $l$-sum can be written
1834: \begin{equation}
1835: F = \sum_{l=2}^L F_l + \delta F,
1836: \end{equation}
1837: where to a good approximation $\delta F \approx F_L$, and $L$ is a
1838: number which is determined by the requested accuracy via
1839: \begin{equation}
1840: |F_L| < \varepsilon_\text{flux} \max_l |F_l|\;.
1841: \end{equation}
1842:
1843: The sum over the $m$-index is computed for all values $-l \le m \le l$
1844: --- we do not truncate it.
1845:
1846: The $k$ and $n$ sums are approximated as
1847: \begin{align}
1848: F_{lm} &\approx \sum_{k=K_-}^{K_+} F_{lmk},&
1849: F_{lmk} &\approx F_{lmk0} + 2\sum_{n=1}^{N_+} F_{lmkn},&
1850: \end{align}
1851: where the constants $K_\pm$ and $N_+$ are determined by requiring
1852: that the following relation
1853: \begin{subequations}
1854: \begin{eqnarray}
1855: |F_{lmK_\pm}| &<& \varepsilon_\text{flux} \max_{l'm'k'} |F_{l'm'k'}|,
1856: \\
1857: |F_{lmkN+}| &<& \varepsilon_\text{flux} \max_{l'm'k'n'} |F_{l'm'k'n'}|,
1858: \end{eqnarray}
1859: \end{subequations}
1860: (where the maximization is done over all previously encountered mode
1861: amplitudes) is satisfied for $B$ consecutive terms. We also require
1862: that the first of these $B$ terms is larger in magnitude than the
1863: last. This extra condition gives some assurance that we are not
1864: truncating during a slow rising up of the modes. For the sums over
1865: the polar harmonic index $k$, we use $B = 2$. For the sums over the
1866: radial harmonic index $n$, we use $B=5$ --- the dominant radial modes
1867: show a less orderly distribution than the polar modes.
1868:
1869: The validity of any given truncation scheme is always subject to doubt
1870: --- it could be the case that modes which haven't been computed are
1871: anomalously large in magnitude. In this sense truncating sums can be
1872: likened to financial investments: past performance is not necessarily
1873: indicative of future results.
1874:
1875: We have found these words of caution are especially applicable for the
1876: sums over the radial harmonic index $n$. We find that the $F_l$
1877: terms, as defined by Eqs.~(\ref{sum outline}), in the outermost sum
1878: over $l$ always decrease monotonically with increasing $l$. Also, the
1879: $F_{lmk}$ terms are always peaked near $k=0$, so that setting the
1880: buffer $B$ to a modestly small number is sufficient for catching the
1881: dominant modes. Unfortunately, the $F_{lmkn}$ appear not to obey any
1882: similar general rules. If either $l$ or the eccentricity $e$ are
1883: increased, the $n$ values of the dominant $F_{lmkn}$ modes drift to
1884: large values of $n$. As a result, the truncation rules are expected
1885: to be limited in a rough sense by large eccentricity, or by requests
1886: for high accuracy (which would require modes with large $l$).
1887:
1888: It may be possible to better anticipate $n$ for the dominant modes.
1889: Peters and Mathews \cite{Peters Mathews} used the quadrupole
1890: approximation to describe the radiation of two point particles moving
1891: along closed Keplerian orbits of arbitrary eccentricity $e$. Under
1892: this approximation, they decomposed the metric perturbation into
1893: tensor spherical harmonics \cite{Thorne} with $l=2$. They then showed
1894: that the power $\left< dE/dt\right>^{\text{PM}}_n$ radiated at a
1895: frequency $n\omega$ (where $n$ is an integer and $\omega$ is the
1896: single unique orbital frequency for a Keplerian orbit) is given by
1897: \begin{align}\label{PM power}
1898: \left\langle \frac{dE}{dt} \right\rangle^{\text{PM}}_n\!\!
1899: &\propto \frac{n^4}{32} \biggl\{ \bigl[ J_{n-2}(ne) - 2eJ_{n-1}(ne) &
1900: \nonumber \\
1901: & + \frac{2}{n}J_n(ne)+2eJ_{n+1}(ne)-J_{n+2}(ne) \bigr]^2&
1902: \nonumber \\
1903: &+ (1-e^2)\left[ J_{n-2}(ne) - 2J_n(ne)+J_{n+2}(ne) \right]^2&
1904: \nonumber \\
1905: & + \frac{4}{3n^2}\left[ J_n(ne) \right]^2 \biggr\} ,&
1906: \end{align}
1907: where $J_n(x)$ are Bessel functions of the first kind. We have found
1908: that this formula (which was derived using only the quadrupole
1909: equation and Keplerian orbits) accurately predicts the $n$ values of
1910: the dominant $F_{lmkn}$ modes for $l=2$. More quantitatively,
1911: define ${\tilde n}$ by
1912: \begin{equation}
1913: \left< \frac{dE}{dt} \right>^{\text{PM}}_{\tilde n} = \max_n \left<
1914: \frac{dE}{dt} \right>^{\text{PM}}_n ,
1915: \end{equation}
1916: so that
1917: \begin{equation}
1918: \tilde n \approx \exp\left[ \frac{1}{2} -\frac{3}{2} \ln (1-e) \right];
1919: \end{equation}
1920: this approximate formula was found by fitting to numerically
1921: determined maxima of Eq.~(\ref{PM power}). Then, even for highly
1922: relativistic, inclined, and eccentric orbits, we have
1923: \begin{equation}
1924: \max_n \left| F_{2mkn}\right| \approx \left| F_{2mk\tilde n} \right|.
1925: \end{equation}
1926: An example is shown in Fig.~\ref{f:PM}.
1927: \begin{figure}
1928: \includegraphics[width = .45\textwidth]{PM_20}
1929: \caption{Modal energy flux through the horizon (superscript H, marked
1930: with $\times$'s) and at infinity (superscript $\infty$, marked with
1931: $+$'s). The indices $l = m = 2$, $k=0$. The orbit has eccentricity
1932: $e = 0.85$, inclination $\theta_\text{inc} = 20^\circ$, and semilatus
1933: rectum $p = 6$, while the massive black hole has $a = 0.9M$. The
1934: horizon modes are peaked at $n=29$, while the infinity modes are
1935: peaked at $n=26$. The dashed line shows the estimated peak location
1936: (at $n=28$) found using the Peters-Mathews power formula (\ref{PM
1937: power}).}
1938: \label{f:PM}
1939: \end{figure}
1940: Reproducing Fig.~\ref{f:PM} under the same conditions but with an
1941: orbital inclination of $\theta_\text{inc} = 70^\circ$ shifts the peaks
1942: occur to $n \simeq 35$. It is somewhat remarkable that the
1943: Peters-Mathews result can be used to predict the peak locations in
1944: this way, since the generic Kerr orbits considered here do not closely
1945: resemble the analogous Keplerian orbits either in their orbital
1946: frequencies or in their shapes. We note that while the Peters-Mathews
1947: power formula (\ref{PM power}) gives reasonable estimates for the
1948: location in $n$ of the dominant modes, it does not accurately predict
1949: the actual flux associated with the dominant modes (for example, the
1950: top panel of Fig.~14 in Ref.~\cite{equatorial} shows a very narrow
1951: peak when compared to the relatively broad peak for the $e=0.7$ orbit
1952: in Fig.~3 from Ref.~\cite{Peters Mathews}).
1953:
1954: It is likely that the $l>2$ analog of the Peters-Mathews power formula
1955: (\ref{PM power}) would be a useful estimator of the values of the
1956: radial harmonic index $n$ for the dominant modes. Such an estimator
1957: would be valuable tool for designing a more accurate truncation scheme
1958: for the sums over $n$. The value of $n$ for the dominant modes
1959: $F_{lmkn}$ increases roughly linearly with $l$, so it may be possible
1960: to ``track'' the peak location starting from the Peters-Mathews power
1961: formula (\ref{PM power}) at $l=2$. Future progress along these lines
1962: may reduce the limitations of our code for computing accurate fluxes
1963: for highly eccentric orbits.
1964:
1965: \subsection{Validation}
1966: \label{ss:Validation}
1967:
1968: We have compared the output of our code with two existing codes. The
1969: first, the ``circular code'' by Hughes \cite{circular}, treats orbits
1970: which have a non-zero inclination and a constant radius. The second,
1971: the ``equatorial code'' by Glampedakis and Kennefick
1972: \cite{equatorial}, treats orbits which are eccentric and confined to
1973: the equatorial plane. Both of these codes have already passed a
1974: variety of tests of their own (described below, but see
1975: Refs.~\cite{circular,equatorial} for more details). Also, the
1976: circular and equatorial codes have been shown to agree with each other
1977: for orbits which are both circular and equatorial \cite{equatorial}.
1978:
1979: The circular code has been shown to agree with analytical
1980: post-Newtonian approximations, found in Ref.~\cite{Minoetalreview},
1981: in the weak field limit. It has also been shown to have the correct
1982: rotational properties in the Schwarzschild limit (see Sec.~V B of
1983: Ref.~\cite{circular} for details). Our code is a direct descendant of
1984: the circular code, and it has some elements in common with that code.
1985: Namely, both use the same routines for solving the homogeneous radial
1986: equation, and also for solving the angular equation. However, they
1987: differ in their methods for solving the geodesic equations, in their
1988: methods for computing the quantities $Z^\star_{lmkn}$, and in
1989: their methods for truncating the sums representing the fluxes. When
1990: computing the individual modes
1991: $\left<dX/dt\right>^{\text{H},\infty}_{lmkn}$ (for $X = E,L_z$) with
1992: our code and the circular code, we find differences $\Delta
1993: \left<dX/dt\right>^{\text{H},\infty}_{lmkn} \lesssim
1994: (10^{-5})\left<dX/dt\right>^{\text{H},\infty} $. Here we are
1995: comparing the difference to the total flux rather than the mode itself
1996: because often there can be large fractional disagreements for
1997: insignificant modes. When computing the total fluxes
1998: $\left<dX/dt\right>^{\text{H},\infty}$ with these two codes, we find
1999: differences $\Delta \left<dX/dt\right>^{\text{H},\infty} \lesssim
2000: (10^{-5})\left<dX/dt\right>^{\text{H},\infty}$.
2001:
2002: The equatorial code has been tested against its own direct ancestors
2003: \cite{ckp,Kennefick}, against the circular code, and also against
2004: Shibata's earlier equatorial code \cite{Shibata equatorial}. Note
2005: that in the case of the test against Shibata's code, Glampedakis and
2006: Kennefick found agreement only at a level of about 1\%. Shibata has
2007: since reported that after improving his code, he now agrees with
2008: Glampedakis and Kennefick to much greater accuracy \cite{Shibata fix}.
2009: When computing the individual modes $\langle
2010: dX/dt\rangle^{\text{H},\infty}_{lmkn}$ (for $X = E,L_z$) our code
2011: always agrees with the equatorial code at the level of the equatorial
2012: code's claimed fractional accuracy of $10^{-3}$. We have tested this
2013: for the top panels of Figs.~14-16 of Ref.~\cite{equatorial}; by
2014: fractional accuracy, we mean the ratio of the differences in modes to
2015: the sum of all the modes shown in each of the figures. When computing
2016: the total fluxes $\left<dX/dt\right>$ we find that our code always
2017: agrees with the equatorial code at the level of their claimed
2018: fractional accuracy of $10^{-3}$. We have tested this for the top
2019: panel, for $a=0.5$, of Table VII of Ref.~\cite{equatorial}. Note that
2020: the fractional accuracy of $10^{-3}$ is the most conservative of the
2021: accuracy claims made in Ref.~\cite{equatorial}. Usually we find
2022: agreement with the equatorial code well beyond this figure; for many
2023: of the total fluxes, we agree with all six of their published digits.
2024:
2025: While we find agreement with the equatorial code when computing the
2026: \emph{total} fluxes $\left<dX/dt\right>$, the two codes show
2027: significant differences when computing the horizon fluxes
2028: $\left<dX/dt\right>^{\text{H}}$ (no such disagreement was found for
2029: the fluxes at infinity $\left<dX/dt\right>^{\infty}$). The
2030: disagreement is especially strong for the energy flux at the horizon,
2031: but it is also present to a lesser extent with the angular momentum
2032: flux at the horizon. For the energy fluxes at the horizon, we
2033: typically only agree with the equatorial code to about 1-10\%, however
2034: in one case (the orbit with $e = 0.4$ and $p = 5$ in Table VII of
2035: Ref.~\cite{equatorial}, which has an exceptionally weak energy flux at
2036: the horizon) our results differ fractionally by about a factor of
2037: three, and they even have different signs. Currently the source of
2038: this disagreement is unknown. We emphasize however that the total
2039: flux of either energy of angular momentum is always dominated by the
2040: flux at infinity, and even in the most extreme case of our
2041: disagreements over the horizon fluxes, we still find acceptable
2042: agreement for the total fluxes.
2043:
2044: Hughes has shown that, in the Schwarzschild limit, under a
2045: transformation from an equatorial to an inclined orbit, the individual
2046: modes must transform as \cite{circular}\footnote{Here we correct two
2047: typos in Eq.~(5.4) of Ref.~\cite{circular}.}
2048: \begin{equation}
2049: \frac{\left< dE/dt\right>^\star_{l(m-k)kn}(\theta_\text{inc})}{\left< dE/dt\right>^\star_{lm0n}(\theta_\text{inc}=0)}
2050: = \left|\mathcal{D}^l_{(m-k)m}(\theta_\text{inc}) \right|^2,
2051: \end{equation}
2052: where $\mathcal{D}^l_{m'm}(\theta_\text{inc})$ is the Wigner D-function,
2053: as defined by Eq.~(4.256) in Ref.~\cite{Arfken}:
2054: \begin{align}
2055: & \mathcal{D}^l_{m'm}(\theta_\text{inc})
2056: = \sum_{q=\max(0,m'-m)}^{\min(l-m,l+m')} (-1)^q &
2057: \nonumber \\
2058: &\times \left(\cos \frac{\theta_\text{inc}}{2} \right)^{2l + m' - m - 2q}
2059: \left(-\sin \frac{\theta_\text{inc}}{2} \right)^{m - m' + 2q}&
2060: \nonumber \\
2061: & \times \frac{ \sqrt{(l+m')! (l-m')! (l+m)! (l-m)!} }{q! (l-m-q)! (l+m'-q)! (m-m'+q)!} .&
2062: \end{align}
2063: For a variety of generic Schwarzschild orbits (which are both
2064: eccentric and inclined) and for a variety of modes, we have checked
2065: that our code satisfies this requirement to a fractional accuracy of
2066: $10^{-5}$ or better. We have also
2067: confirmed that, again for a variety of generic Schwarzschild orbits,
2068: the total fluxes of energy at both the horizon and infinity, are
2069: independent of inclination (we have checked this up to a fractional
2070: accuracy of $10^{-4}$).
2071:
2072: With the help of Norichika Sago, we have also compared the output of our
2073: code with analytical post-Newtonian expressions (essentially
2074: an extension of Ref.~\cite{Minoetalreview}) for the fluxes of
2075: energy and angular momentum at infinity for orbits which are both slightly
2076: eccentric and slightly inclined \cite{Sago in prep}. We compared the
2077: contributions for $l=2$, 3, and 4. With a requested accuracy of $10^{-4}$ in our
2078: numerical code, the two methods of calculation always agreed to $\sim 10^{-3}$ or better
2079: (with slightly better agreement for angular momentum fluxes than for energy fluxes).
2080: Preliminary investigations suggest that this disagreement is due to the
2081: small-eccentricity-expansion used when deriving the post-Newtonian expressions, and is
2082: therefore not unexpected.
2083:
2084:
2085: \section{Results}
2086: \label{s:Results}
2087:
2088: In this section we discuss the results found from using the code
2089: described in the previous section. After describing the general
2090: characteristics of the radiation from a typical generic orbit, we will
2091: discuss how these characteristics vary over a catalog of orbits.
2092:
2093: \subsection{Radial and polar voices}
2094:
2095: Figure \ref{f:generic} compares a snapshot waveform produced by a
2096: generic orbit with the waveforms produced by the circular and
2097: equatorial limits of that orbit.
2098: \begin{figure}
2099: \includegraphics[width = .45\textwidth]{generic}
2100: \caption{A comparison of waveforms associated with a generic orbit and
2101: with the related circular and equatorial orbits. Each orbit has a
2102: semilatus rectum of $p=4$. The magnitude of the massive black hole's
2103: spin angular momentum is $aM=0.9M^2$. Each waveform is plotted for
2104: $1.5 \times M/(10^6 M_\odot)$ hours as seen from a viewing angle of
2105: $\theta = 30^\circ$ away from the symmetry axis of the massive black
2106: hole.}
2107: \label{f:generic}
2108: \end{figure}
2109: It is clear that the generic waveform is not well approximated by
2110: either of the limiting cases at the level of accuracy needed to
2111: produce detection waveforms (fractional phase accuracy of $\sim
2112: 10^{-4}$). We can however understand the generic waveform as a
2113: composition of the effects produced by the different components of the
2114: orbital motion: the polar motion $\theta(\lambda)$ and the radial
2115: motion $r(\lambda)$. To do so, we first define the general waveform
2116: quantity
2117: \begin{equation} \label{modes}
2118: H = h_+ - i h_\times = \sum_{kn} H_{kn} e^{-i\omega_{kn}(t-r)},
2119: \end{equation}
2120: where
2121: \begin{align}
2122: &H_{kn} = -\frac{2}{r}\sum_{lm} \frac{Z_{lmkn}^\text{H}}{\omega_{mkn}^{2}}
2123: S_{lmkn}(\theta) e^{im[\phi-\Omega_\phi(t-r)]},& \\
2124: &\omega_{kn} = k\Omega_\theta + n\Omega_r.&
2125: \label{eq:Hkn_def}
2126: \end{align}
2127: Each term $H_{kn} \exp[-i\omega_{kn}(t-r)]$ in the sum (\ref{modes})
2128: can be likened to a ``voice'' in the GW produced by the orbit. By
2129: using the general expression for the $Z^\star_{lmkn}$ coefficients
2130: (\ref{final Z}) and the symmetry rules (\ref{geodesic sym}) for the
2131: geodesic expansions, we see that in the limiting circular and
2132: equatorial cases, only some voices contribute
2133: \begin{subequations}\label{waveform sym}
2134: \begin{align}
2135: H_{kn} &\propto \delta_{n0},& &\text{(circular orbits)}&
2136: \\
2137: H_{kn} &\propto \delta_{k0},& &\text{(equatorial orbits)}&
2138: \\
2139: H_{kn} &\propto \delta_{n0}\delta_{k0}.& &\text{(circular-equatorial orbits)}&
2140: \end{align}
2141: \end{subequations}
2142: The term $H_{00}$, the sum of all the terms which oscillate at pure
2143: multiples of the azimuthal frequency, is the only nonvanishing term
2144: for orbits which are both circular and equatorial. The polar and
2145: radial voices are made of terms which oscillate at integer linear
2146: combinations of the azimuthal frequency, and either the radial or
2147: polar frequency. The polar voices, with $n=0$ and $k \ne 0$, can be
2148: thought of as being produced by the orbital inclination. Similarly,
2149: the radial voices ($k=0$ and $n \ne 0$) are produced by the orbit's
2150: eccentricity. This suggests the following separation of the voices:
2151: \begin{equation} \label{H split}
2152: H = H_\text{azimuthal} + H_\text{polar} + H_\text{radial} + H_\text{mixed},
2153: \end{equation}
2154: where
2155: \begin{subequations}
2156: \begin{align} \label{H pieces}
2157: H_\text{azimuthal} &= H_{00},&
2158: \\
2159: H_\text{polar} &= \sum_{k\ne0} H_{k0} e^{-i\omega_{k0}t},&
2160: \\
2161: H_\text{radial} &= \sum_{n\ne0} H_{0n} e^{-i\omega_{0n}t},&
2162: \\
2163: H_\text{mixed} &= \sum_{k\ne0} \sum_{n\ne0} H_{kn} e^{-i\omega_{kn}t}.&
2164: \end{align}
2165: \end{subequations}
2166: Figure \ref{f:H split} displays the waveforms for the different voices
2167: of the generic waveform from Fig.~\ref{f:generic}.
2168: \begin{figure}
2169: \includegraphics[width = .45\textwidth]{PolRadAzMix}
2170: \caption{The waveform from Fig.~\ref{f:generic} and its separated
2171: pieces Eq.~(\ref{H pieces}). This orbit has an eccentricity $e=0.5$,
2172: inclination $\theta_\text{inc} = 45^\circ$, and semilatus rectum
2173: $p=4$. The spin of the black hole is $a=0.9M$. We plot the
2174: waveforms as seen from a viewing angle of $\theta = 30^\circ$. A plot
2175: of $h_\times$ shows similar behavior.}
2176: \label{f:H split}
2177: \end{figure}
2178: Since that orbit has eccentricity $e=0.5$ and inclination
2179: $\theta_\text{inc} = 45^\circ$, one might expect that the radial and
2180: polar voices would be comparably loud. Instead (as one might guess by
2181: looking at the waveform from the equatorial limit of this orbit; cf.\
2182: Fig.~\ref{f:generic}), the radial voice alone $H_\text{radial}$ has a
2183: phase history which is similar to that of the complete waveform. The
2184: polar voice $H_\text{polar}$ is comparatively unimportant.
2185:
2186: The fluxes in energy and angular momentum can be similarly separated
2187: into radial, polar, azimuthal, and mixed pieces. Again for this case,
2188: the radial voice dominates. Table \ref{t:split fluxes} shows how the
2189: fluxes are distributed in the case of the orbit used to produce
2190: Figs.~\ref{f:generic} and \ref{f:H split}.
2191: \begin{table}
2192: \begin{ruledtabular}
2193: \begin{tabular}{c|cccc}
2194: & radial & polar & azimuthal & mixed \\ \hline
2195: $\left<dE/dt\right>$ & 62\% & 1.2\% & 0.52\% & 37\% \\ \hline
2196: $\left<dL_z/dt\right>$ & 73\% & 0.77\% & 1.0\% & 25\%
2197: \end{tabular}
2198: \end{ruledtabular}
2199: \caption{\label{t:split fluxes} The results of applying the splitting procedure
2200: described by Eq.~(\ref{H split}) to the fluxes of energy $E$ and angular momentum $L_z$ produced by
2201: the generic orbit from Fig.~\ref{f:generic}. Each number is the ratio of the contribution from a specific voice,
2202: to the total quantity.}
2203: \end{table}
2204: The radial voice carries away more than half of both the energy and
2205: angular momentum fluxes, while the polar voice carries only about
2206: 1\%. This suggests that GWs from this orbit might, for some
2207: applications, be well approximated by its radial voice alone: $H
2208: \approx H_\text{radial}$. Such an approximation would by highly
2209: desirable since it would considerably reduce the computational cost of
2210: generating this waveform. Unfortunately $H_\text{radial}$ is still
2211: quite sensitive to the orbital inclination, as can be seen by
2212: comparing to the waveforms for the circular and equatorial orbits in
2213: Fig.~\ref{f:generic}. This is because the inclination influences the
2214: values of all three orbital frequencies. So even if this approximation
2215: were valid for some range of orbits, it would not likely reduce the
2216: number of detection waveforms needed when searching for events
2217: produced by those orbits, though it may reduce the computational cost
2218: to produce each waveform.
2219:
2220: \subsection{Catalog of orbits}
2221:
2222: We now discuss the waveforms and fluxes for a catalog of generic orbits.
2223: The parameters of these orbits are shown in Table \ref{t:orbits}.
2224: %\input{OrbitTable_medium.tex}
2225: \begin{table*}
2226: \begin{ruledtabular}
2227: \begin{tabular}{ccc||cccc}
2228: $e$ &
2229: $p$ &
2230: $\theta_\text{inc}$ &
2231: $E/\mu$ &
2232: $L_z/(\mu M)$ &
2233: $Q / (\mu^2 M^2)$ \\
2234: \hline
2235: 0.1 & 6 & 20$^\circ$ & 0.923976142907 & 2.65115182126 & 0.944969071904\\
2236: 0.1 & 6 & 40$^\circ$ & 0.926051429284 & 2.22165013064 & 3.52285570065 \\
2237: 0.1 & 6 & 60$^\circ$ & 0.929863620985 & 1.52003893228 & 7.01378240707 \\
2238: 0.1 & 6 & 80$^\circ$ & 0.935903446174 & 0.564178168635 & 10.3350037833 \\
2239: 0.1 & 12 & 100$^\circ$ & 0.962956523141 & $-$0.701069332316 & 15.8653948211 \\
2240: 0.1 & 12 & 120$^\circ$ & 0.963964137935 & $-$2.07109099890 & 12.9112484603 \\
2241: 0.1 & 12 & 140$^\circ$ & 0.964900141027 & $-$3.24526311278 & 7.43835020273 \\
2242: 0.1 & 12 & 160$^\circ$ & 0.965574148307 & $-$4.04295255570 & 2.17176613884 \\
2243: \hline
2244: 0.3 & 6 & 20$^\circ$ & 0.929683266598 & 2.66499623815 & 0.953716658814\\
2245: 0.3 & 6 & 40$^\circ$ & 0.931463021604 & 2.23504604006 & 3.56152665553 \\
2246: 0.3 & 6 & 60$^\circ$ & 0.934742155317 & 1.53133514084 & 7.11166313087 \\
2247: 0.3 & 6 & 80$^\circ$ & 0.939961770755 & 0.569540184106 & 10.5245453291 \\
2248: 0.3 & 12 & 100$^\circ$ & 0.965665301518 & $-$0.704444574360 & 16.0138725377 \\
2249: 0.3 & 12 & 120$^\circ$ & 0.966529280028 & $-$2.08286664248 & 13.0549867002 \\
2250: 0.3 & 12 & 140$^\circ$ & 0.967333822046 & $-$3.26633491912 & 7.53338516325 \\
2251: 0.3 & 12 & 160$^\circ$ & 0.967914357052 & $-$4.07155287011 & 2.20208170741 \\
2252: \hline
2253: 0.5 & 6 & 20$^\circ$ & 0.941309757948 & 2.69327096428 & 0.971725840431\\
2254: 0.5 & 6 & 40$^\circ$ & 0.942555891555 & 2.26253463210 & 3.64161734490 \\
2255: 0.5 & 6 & 60$^\circ$ & 0.944866295222 & 1.55467904187 & 7.31622158779 \\
2256: 0.5 & 6 & 80$^\circ$ & 0.948580242140 & 0.580718189662 & 10.9253034674 \\
2257: 0.5 & 12 & 100$^\circ$ & 0.971221837220 & $-$0.711332909044 & 16.3190869455 \\
2258: 0.5 & 12 & 120$^\circ$ & 0.971826517606 & $-$2.10697070044 & 13.3517251785 \\
2259: 0.5 & 12 & 140$^\circ$ & 0.972392433557 & $-$3.30957999464 & 7.73032698828 \\
2260: 0.5 & 12 & 160$^\circ$ & 0.972802509120 & $-$4.13035470435 & 2.26507351639 \\
2261: \hline
2262: 0.7 & 6 & 20$^\circ$ & 0.959306993445 & 2.73722358422 & 1.00010433263 \\
2263: 0.7 & 6 & 40$^\circ$ & 0.959910828459 & 2.30561707530 & 3.76913698051 \\
2264: 0.7 & 6 & 60$^\circ$ & 0.961041797423 & 1.59171983094 & 7.64712824881 \\
2265: 0.7 & 6 & 80$^\circ$ & 0.962889252972 & 0.598733562204 & 11.5872341739 \\
2266: 0.7 & 12 & 100$^\circ$ & 0.979920978103 & $-$0.722026359625 & 16.7987400685 \\
2267: 0.7 & 12 & 120$^\circ$ & 0.980213741124 & $-$2.14458473041 & 13.8215334684 \\
2268: 0.7 & 12 & 140$^\circ$ & 0.980489898895 & $-$3.37737131644 & 8.04421001469 \\
2269: 0.7 & 12 & 160$^\circ$ & 0.980691338739 & $-$4.22282649751 & 2.36594094475
2270: \end{tabular}
2271: \caption{\label{t:orbits}
2272: Each row above describes an orbit in our catalog. The numbers to the right hand
2273: side of the vertical bar were computed (to an fractional accuracy of $10^{-12}$)
2274: from the numbers on the numbers on the left hand of the bar (using the
2275: method described in Appendix \ref{s:Geodesic details}).
2276: Each orbit has $a = 0.9M$.
2277: }
2278: \end{ruledtabular}
2279: \end{table*}
2280: Tables \ref{t:voices_P} and \ref{t:voices_R} show the relative
2281: contributions of the different voices for these waveforms.
2282: %\input{NewVoiceTable_P.tex}
2283: \begin{table*}
2284: \begin{ruledtabular}
2285: \begin{tabular}{cc||cccc|cccc}
2286: $e$ &
2287: $\theta_\text{inc}$ &
2288: $\mathcal{P}_\text{radial}$ &
2289: $\mathcal{P}_\text{polar}$ &
2290: $\mathcal{P}_\text{azimuthal}$ &
2291: $\mathcal{P}_\text{mixed}$ &
2292: $\mathcal{T}_{z\text{,radial}}$ &
2293: $\mathcal{T}_{z\text{,polar}}$ &
2294: $\mathcal{T}_{z\text{,azimuthal}}$ &
2295: $\mathcal{T}_{z\text{,mixed}}$ \\
2296: \hline
2297: 0.1 & $20^\circ$ & $16$\% & $8.3$\% & $74$\% & $1.8$\% & $14$\% & $ 5$\% & $80$\% & $ 1$\% \\
2298: 0.1 & $40^\circ$ & $13$\% & $28$\% & $52$\% & $6.6$\% & $14$\% & $17$\% & $65$\% & $ 4$\% \\
2299: 0.1 & $60^\circ$ & $9.4$\% & $48$\% & $29$\% & $14$\% & $14$\% & $29$\% & $48$\% & $9.5$\% \\
2300: 0.1 & $80^\circ$ & $6.7$\% & $53$\% & $10$\% & $30$\% & $20$\% & $24$\% & $33$\% & $23$\% \\
2301: \hline
2302: 0.3 & $20^\circ$ & $76$\% & $1.9$\% & $14$\% & $7.9$\% & $75$\% & $1.3$\% & $19$\% & $4.7$\% \\
2303: 0.3 & $40^\circ$ & $58$\% & $5.4$\% & $7.8$\% & $29$\% & $67$\% & $3.6$\% & $12$\% & $17$\% \\
2304: 0.3 & $60^\circ$ & $37$\% & $5.9$\% & $1.8$\% & $55$\% & $58$\% & $2.7$\% & $3.8$\% & $36$\% \\
2305: 0.3 & $80^\circ$ & $17$\% & $1.9$\% & $0.48$\% & $80$\% & $51$\% & $-2.4$\% & $1.9$\% & $49$\% \\
2306: \hline
2307: 0.5 & $20^\circ$ & $90$\% & $0.077$\% & $0.46$\% & $9.5$\% & $93$\% & $0.083$\% & $0.92$\% & $ 6$\% \\
2308: 0.5 & $40^\circ$ & $66$\% & $0.24$\% & $0.64$\% & $33$\% & $77$\% & $0.18$\% & $1.5$\% & $21$\% \\
2309: 0.5 & $60^\circ$ & $39$\% & $0.93$\% & $1.1$\% & $59$\% & $58$\% & $0.68$\% & $3.3$\% & $38$\% \\
2310: 0.5 & $80^\circ$ & $19$\% & $1.6$\% & $0.0055$\% & $80$\% & $53$\% & $-2$\% & $0.029$\% & $49$\% \\
2311: \hline
2312: 0.7 & $20^\circ$ & $90$\% & $0.057$\% & $0.62$\% & $9.3$\% & $92$\% & $0.1$\% & $2.3$\% & $ 6$\% \\
2313: 0.7 & $40^\circ$ & $67$\% & $0.22$\% & $0.36$\% & $33$\% & $77$\% & $0.35$\% & $1.5$\% & $21$\% \\
2314: 0.7 & $60^\circ$ & $40$\% & $0.19$\% & $0.0018$\% & $59$\% & $61$\% & $-0.17$\% & $0.0094$\% & $39$\% \\
2315: 0.7 & $80^\circ$ & $19$\% & $0.16$\% & $0.043$\% & $81$\% & $53$\% & $-0.47$\% & $0.33$\% & $47$\%
2316: \end{tabular}
2317: \caption{\label{t:voices_P} Here we show the results of applying the splitting procedure
2318: described by Eq.~(\ref{H split}) to the fluxes of energy $E$ and angular momentum $L_z$ produced by our
2319: catalog of generic orbits (Table \ref{t:orbits}). The symbol $\mathcal{P}$ represents
2320: a ratio of the average power relative to the total average power $\left< dE/dt\right>$, and the symbol
2321: $\mathcal{T}$ represents a ratio of the average torque to the total average torque $\left< dL_z/dt\right>$.
2322: Each of these orbits has $a = 0.9M$ and $p = 6$.}
2323: \end{ruledtabular}
2324: \end{table*}
2325: %\input{NewVoiceTable_R.tex}
2326: \begin{table*}
2327: \begin{ruledtabular}
2328: \begin{tabular}{cc||cccc|cccc}
2329: $e$ &
2330: $\theta_\text{inc}$ &
2331: $\mathcal{P}_\text{radial}$ &
2332: $\mathcal{P}_\text{polar}$ &
2333: $\mathcal{P}_\text{azimuthal}$ &
2334: $\mathcal{P}_\text{mixed}$ &
2335: $\mathcal{T}_{z\text{,radial}}$ &
2336: $\mathcal{T}_{z\text{,polar}}$ &
2337: $\mathcal{T}_{z\text{,azimuthal}}$ &
2338: $\mathcal{T}_{z\text{,mixed}}$ \\
2339: \hline
2340: 0.1 & $100^\circ$ & $1.1$\% & $76$\% & $7.7$\% & $15$\% & $7.8$\% & $23$\% & $67$\% & $1.3$\% \\
2341: 0.1 & $120^\circ$ & $ 4$\% & $60$\% & $22$\% & $14$\% & $7.4$\% & $35$\% & $51$\% & $6.5$\% \\
2342: 0.1 & $140^\circ$ & $10$\% & $35$\% & $45$\% & $9.6$\% & $12$\% & $20$\% & $63$\% & $4.8$\% \\
2343: 0.1 & $160^\circ$ & $18$\% & $11$\% & $68$\% & $3.3$\% & $17$\% & $5.7$\% & $76$\% & $1.6$\% \\
2344: \hline
2345: 0.3 & $100^\circ$ & $6.1$\% & $17$\% & $2.1$\% & $75$\% & $52$\% & $13$\% & $26$\% & $9.4$\% \\
2346: 0.3 & $120^\circ$ & $20$\% & $10$\% & $4.8$\% & $65$\% & $43$\% & $8.4$\% & $14$\% & $34$\% \\
2347: 0.3 & $140^\circ$ & $47$\% & $4.3$\% & $7.1$\% & $42$\% & $62$\% & $3.2$\% & $13$\% & $22$\% \\
2348: 0.3 & $160^\circ$ & $78$\% & $0.94$\% & $7.7$\% & $13$\% & $81$\% & $0.65$\% & $11$\% & $6.9$\% \\
2349: \hline
2350: 0.5 & $100^\circ$ & $7.5$\% & $0.26$\% & $0.042$\% & $92$\% & $80$\% & $1.2$\% & $0.88$\% & $18$\% \\
2351: 0.5 & $120^\circ$ & $23$\% & $0.37$\% & $0.038$\% & $76$\% & $56$\% & $0.5$\% & $0.18$\% & $43$\% \\
2352: 0.5 & $140^\circ$ & $52$\% & $0.51$\% & $0.34$\% & $47$\% & $72$\% & $0.59$\% & $0.97$\% & $26$\% \\
2353: 0.5 & $160^\circ$ & $84$\% & $0.25$\% & $1.2$\% & $15$\% & $89$\% & $0.27$\% & $2.6$\% & $7.8$\% \\
2354: \hline
2355: 0.7 & $100^\circ$ & $6.8$\% & $0.36$\% & $0.038$\% & $93$\% & $84$\% & $0.66$\% & $1.8$\% & $13$\% \\
2356: 0.7 & $120^\circ$ & $22$\% & $0.32$\% & $0.15$\% & $78$\% & $53$\% & $ 1$\% & $1.5$\% & $44$\% \\
2357: 0.7 & $140^\circ$ & $50$\% & $0.16$\% & $0.28$\% & $49$\% & $70$\% & $0.4$\% & $1.5$\% & $28$\% \\
2358: 0.7 & $160^\circ$ & $84$\% & $0.019$\% & $0.22$\% & $16$\% & $91$\% & $0.043$\% & $0.92$\% & $8.5$\%
2359: \end{tabular}
2360: \caption{\label{t:voices_R} Identical
2361: to Table \ref{t:voices_P}, except that these orbits
2362: are retrograde and have semilatus rectum $p = 12$.}
2363: \end{ruledtabular}
2364: \end{table*}
2365:
2366: It is clear that as inclination is increased, radiation is channeled
2367: into the polar voice, and that likewise the radial voice is amplified
2368: by an increase in eccentricity. Although both voices can be amplified
2369: by such adjustments, the radial voice seems particularly booming while
2370: the polar voice is more subtle. While the polar voice can be made
2371: dominant, it requires especially low eccentricity, below about 0.3.
2372: Also, for orbits with eccentricity as low as 0.1 and inclinations
2373: $\theta_\text{inc} \lesssim 45^\circ$ or $\theta_\text{inc} \gtrsim 135^\circ$,
2374: one should expect to capture at least half of the radiative power and torque in the azimuthal voice alone. This is especially significant since the azimuthal voice is \emph{exceedingly} inexpensive to compute in comparison to the others.
2375: It is composed only of oscillations at pure multiples of the azimuthal frequency with $k = n = 0$, so that computing it alone would mean replacing
2376: all of the four dimensional sums in Sec.~\ref{ss:Waveforms and fluxes} with only two dimensional sums.
2377:
2378: The asymptotic energy and angular momentum fluxes associated with the
2379: catalog (Table \ref{t:orbits}) are displayed in Tables
2380: \ref{t:fluxes_P} and \ref{t:fluxes_R}. The energy and angular
2381: momentum fluxes were requested to have fractional accuracies of
2382: $\varepsilon_\text{flux} = 10^{-4}$. The estimates of the actual
2383: fractional accuracies, shown in square brackets in Tables
2384: \ref{t:fluxes_P} and \ref{t:fluxes_R}, were determined as follows:
2385: First we compute each flux with a requested fractional accuracy of
2386: $\varepsilon_\text{flux} = 10^{-4}$. The fractional accuracy claimed
2387: in Tables \ref{t:fluxes_P} and \ref{t:fluxes_R} is then either
2388: $10^{-4}$, or if larger, the fractional residual that was found when comparing with
2389: the same flux computed with a requested fractional accuracy of
2390: $\varepsilon_\text{flux} = 10^{-3}$. The summation buffer $B$ for the
2391: polar harmonic index $k$ is 3, while $B=5$ for the radial harmonic
2392: index $n$. The largest $l$ value needed for the $e=0.1,~0.3$ orbits
2393: was $l=10$, while for the $e=0.5,~0.7$ orbits the largest $l$ value was
2394: $l=11$. The most complex waveform,
2395: $(e,\theta_\text{inc})=(0.7,80^\circ)$, is made up of about $160,000$
2396: modes (about 80,000 mode calculations due to symmetry). The simplest
2397: waveform, $(e,\theta_\text{inc})=(0.1,20^\circ)$, is made from about
2398: 17,000 modes (about 8,500 mode calculations).
2399: %\input{FluxTable_P.tex}
2400: \begin{table*}
2401: \begin{ruledtabular}
2402: \begin{tabular}{cc||ccccc}
2403: $e$ &
2404: $\theta_\text{inc}$ &
2405: $\left< dE/dt \right>^\text{H} (M/\mu)^2$ &
2406: $\left< dE/dt \right>^\infty (M/\mu)^2$ &
2407: $\left< dL_z/dt \right>^\text{H} M/ \mu^2$ &
2408: $\left< dL_z/dt \right>^\infty M/ \mu^2$ &
2409: $\left< dQ/dt \right> / (\mu^2 M)$ \\
2410: \hline
2411: 0.1 & $20^\circ$ & $-4.25756\times 10^{-6} [10^{-4}]$ & $5.87399\times 10^{-4} [10^{-4}]$ & $-6.72410\times 10^{-5} [10^{-4}]$ & $8.53652\times 10^{-3} [10^{-4}]$ & $6.03753\times 10^{-3}$ \\
2412: 0.1 & $40^\circ$ & $-3.96692\times 10^{-6} [10^{-4}]$ & $6.18500\times 10^{-4} [10^{-4}]$ & $-7.76672\times 10^{-5} [10^{-4}]$ & $7.62823\times 10^{-3} [10^{-4}]$ & $2.39458\times 10^{-2}$ \\
2413: 0.1 & $60^\circ$ & $-3.36171\times 10^{-6} [10^{-4}]$ & $6.83855\times 10^{-4} [10^{-4}]$ & $-1.12143\times 10^{-4} [10^{-4}]$ & $6.07264\times 10^{-3} [10^{-4}]$ & $5.50060\times 10^{-2}$ \\
2414: 0.1 & $80^\circ$ & $-9.78653\times 10^{-7} [10^{-4}]$ & $8.07007\times 10^{-4} [10^{-4}]$ & $-1.90843\times 10^{-4} [10^{-4}]$ & $3.61820\times 10^{-3} [10^{-4}]$ & $1.25570\times 10^{-1}$ \\
2415: \hline
2416: 0.3 & $20^\circ$ & $-5.88185\times 10^{-6} [10^{-4}]$ & $6.80432\times 10^{-4} [10^{-4}]$ & $-7.78419\times 10^{-5} [10^{-4}]$ & $8.62437\times 10^{-3} [10^{-4}]$ & $6.11706\times 10^{-3}$ \\
2417: 0.3 & $40^\circ$ & $-5.88820\times 10^{-6} [10^{-4}]$ & $7.26781\times 10^{-4} [10^{-4}]$ & $-1.00628\times 10^{-4} [10^{-4}]$ & $7.83499\times 10^{-3} [10^{-4}]$ & $2.46493\times 10^{-2}$ \\
2418: 0.3 & $60^\circ$ & $-5.28978\times 10^{-6} [10^{-4}]$ & $8.31504\times 10^{-4} [10^{-4}]$ & $-1.66905\times 10^{-4} [10^{-4}]$ & $6.48662\times 10^{-3} [10^{-4}]$ & $5.86987\times 10^{-2}$ \\
2419: 0.3 & $80^\circ$ & $-1.52779\times 10^{-7} [10^{-4}]$ & $1.08629\times 10^{-3} [10^{-4}]$ & $-3.46171\times 10^{-4} [10^{-4}]$ & $4.36910\times 10^{-3} [10^{-4}]$ & $1.48680\times 10^{-1}$ \\
2420: \hline
2421: 0.5 & $20^\circ$ & $-8.37384\times 10^{-6} [10^{-4}]$ & $7.98857\times 10^{-4} [10^{-4}]$ & $-9.16902\times 10^{-5} [10^{-4}]$ & $8.34425\times 10^{-3} [10^{-4}]$ & $5.95501\times 10^{-3}$ \\
2422: 0.5 & $40^\circ$ & $-9.06408\times 10^{-6} [10^{-4}]$ & $8.74449\times 10^{-4} [10^{-3}]$ & $-1.38234\times 10^{-4} [10^{-4}]$ & $7.80844\times 10^{-3} [10^{-3}]$ & $2.46909\times 10^{-2}$ \\
2423: 0.5 & $60^\circ$ & $-8.31005\times 10^{-6} [10^{-4}]$ & $1.05986\times 10^{-3} [10^{-3}]$ & $-2.70993\times 10^{-4} [10^{-4}]$ & $6.92901\times 10^{-3} [10^{-3}]$ & $6.26645\times 10^{-2}$ \\
2424: 0.5 & $80^\circ$ & $ 5.67035\times 10^{-6} [10^{-4}]$ & $1.67918\times 10^{-3} [10^{-3}]$ & $-7.40721\times 10^{-4} [10^{-4}]$ & $5.87398\times 10^{-3} [10^{-3}]$ & $1.93149\times 10^{-1}$ \\
2425: \hline
2426: 0.7 & $20^\circ$ & $-9.34196\times 10^{-6} [10^{-4}]$ & $7.69363\times 10^{-4} [10^{-2}]$ & $-8.95146\times 10^{-5} [10^{-4}]$ & $6.66378\times 10^{-3} [10^{-2}]$ & $4.80410\times 10^{-3}$ \\
2427: 0.7 & $40^\circ$ & $-1.07089\times 10^{-5} [10^{-3}]$ & $8.70367\times 10^{-4} [10^{-2}]$ & $-1.57713\times 10^{-4} [10^{-4}]$ & $6.48763\times 10^{-3} [10^{-2}]$ & $2.06958\times 10^{-2}$ \\
2428: 0.7 & $60^\circ$ & $-9.23969\times 10^{-6} [10^{-3}]$ & $1.13494\times 10^{-3} [10^{-2}]$ & $-3.62611\times 10^{-4} [10^{-4}]$ & $6.29440\times 10^{-3} [10^{-2}]$ & $5.69964\times 10^{-2}$ \\
2429: 0.7 & $80^\circ$ & $ 2.85367\times 10^{-5} [10^{-1}]$ & $2.69226\times 10^{-3} [10^{-1}]$ & $-1.58165\times 10^{-3} [10^{-2}]$ & $8.25056\times 10^{-3} [10^{-1}]$ & $2.58125\times 10^{-1}$
2430: \end{tabular}
2431: \caption{\label{t:fluxes_P} The fluxes of energy $E$ and axial angular momentum $L_z$ at infinity
2432: (superscript of $\infty$) and through the black hole's event horizon (superscript of H) for some generic orbits.
2433: Each of these orbits has $a = 0.9M$ and $p = 6$. Each number
2434: in square brackets is an order of magnitude estimate for the fractional accuracy of the preceding number.
2435: The rates of change for the Carter constant $\left< dQ/dt \right>$ were computed from the formula (\ref{Qdot}) which follows
2436: from the assumption that radiation does not change the orbital inclination.
2437: Since this is an uncontrolled approximation, the accuracy of the $\left< dQ/dt \right>$
2438: figures is unknown; we include these data for comparison purposes.
2439: }
2440: \end{ruledtabular}
2441: \end{table*}
2442: %\input{FluxTable_R.tex}
2443: \begin{table*}
2444: \begin{ruledtabular}
2445: \begin{tabular}{cc||ccccc}
2446: $e$ &
2447: $\theta_\text{inc}$ &
2448: $\left< dE/dt \right>^\text{H} (M/\mu)^2$ &
2449: $\left< dE/dt \right>^\infty (M/\mu)^2$ &
2450: $\left< dL_z/dt \right>^\text{H} M/ \mu^2$ &
2451: $\left< dL_z/dt \right>^\infty M/ \mu^2$ &
2452: $\left< dQ/dt \right> / (\mu^2 M)$ \\
2453: \hline
2454: 0.1 & $100^\circ$ & $1.10409\times 10^{-8} [10^{-4}]$ & $2.50845\times 10^{-5} [10^{-3}]$ & $-1.85760\times 10^{-6} [10^{-4}]$ & $-1.22779\times 10^{-4} [10^{-2}]$ & $5.64112\times 10^{-3}$ \\
2455: 0.1 & $120^\circ$ & $3.80216\times 10^{-8} [10^{-4}]$ & $2.67639\times 10^{-5} [10^{-4}]$ & $-2.47774\times 10^{-6} [10^{-4}]$ & $-4.91749\times 10^{-4} [10^{-4}]$ & $6.16205\times 10^{-3}$ \\
2456: 0.1 & $140^\circ$ & $7.36704\times 10^{-8} [10^{-4}]$ & $2.83939\times 10^{-5} [10^{-4}]$ & $-3.36034\times 10^{-6} [10^{-4}]$ & $-8.40626\times 10^{-4} [10^{-4}]$ & $3.86894\times 10^{-3}$ \\
2457: 0.1 & $160^\circ$ & $1.06500\times 10^{-7} [10^{-4}]$ & $2.96240\times 10^{-5} [10^{-4}]$ & $-4.23915\times 10^{-6} [10^{-4}]$ & $-1.09777\times 10^{-3} [10^{-4}]$ & $1.18394\times 10^{-3}$ \\
2458: \hline
2459: 0.3 & $100^\circ$ & $2.44837\times 10^{-8} [10^{-4}]$ & $2.92408\times 10^{-5} [10^{-4}]$ & $-2.74782\times 10^{-6} [10^{-4}]$ & $-1.14269\times 10^{-4} [10^{-4}]$ & $5.32021\times 10^{-3}$ \\
2460: 0.3 & $120^\circ$ & $7.99022\times 10^{-8} [10^{-4}]$ & $3.18207\times 10^{-5} [10^{-4}]$ & $-3.86655\times 10^{-6} [10^{-4}]$ & $-4.92316\times 10^{-4} [10^{-4}]$ & $6.21994\times 10^{-3}$ \\
2461: 0.3 & $140^\circ$ & $1.57246\times 10^{-7} [10^{-4}]$ & $3.45503\times 10^{-5} [10^{-4}]$ & $-5.43045\times 10^{-6} [10^{-4}]$ & $-8.65358\times 10^{-4} [10^{-4}]$ & $4.01673\times 10^{-3}$ \\
2462: 0.3 & $160^\circ$ & $2.32108\times 10^{-7} [10^{-4}]$ & $3.67886\times 10^{-5} [10^{-4}]$ & $-6.98894\times 10^{-6} [10^{-4}]$ & $-1.15339\times 10^{-3} [10^{-4}]$ & $1.25517\times 10^{-3}$ \\
2463: \hline
2464: 0.5 & $100^\circ$ & $6.01938\times 10^{-8} [10^{-4}]$ & $3.44049\times 10^{-5} [10^{-4}]$ & $-4.38711\times 10^{-6} [10^{-4}]$ & $-9.51540\times 10^{-5} [10^{-4}]$ & $4.56726\times 10^{-3}$ \\
2465: 0.5 & $120^\circ$ & $1.87174\times 10^{-7} [10^{-4}]$ & $3.87672\times 10^{-5} [10^{-4}]$ & $-6.64495\times 10^{-6} [10^{-4}]$ & $-4.68884\times 10^{-4} [10^{-4}]$ & $6.02679\times 10^{-3}$ \\
2466: 0.5 & $140^\circ$ & $3.77537\times 10^{-7} [10^{-4}]$ & $4.37778\times 10^{-5} [10^{-4}]$ & $-9.86398\times 10^{-6} [10^{-4}]$ & $-8.67384\times 10^{-4} [10^{-4}]$ & $4.09805\times 10^{-3}$ \\
2467: 0.5 & $160^\circ$ & $5.75306\times 10^{-7} [10^{-4}]$ & $4.82937\times 10^{-5} [10^{-4}]$ & $-1.31850\times 10^{-5} [10^{-4}]$ & $-1.20168\times 10^{-3} [10^{-4}]$ & $1.33246\times 10^{-3}$ \\
2468: \hline
2469: 0.7 & $100^\circ$ & $1.16031\times 10^{-7} [10^{-3}]$ & $3.29140\times 10^{-5} [10^{-4}]$ & $-5.69490\times 10^{-6} [10^{-4}]$ & $-6.19594\times 10^{-5} [10^{-3}]$ & $3.14810\times 10^{-3}$ \\
2470: 0.7 & $120^\circ$ & $3.48444\times 10^{-7} [10^{-3}]$ & $3.87999\times 10^{-5} [10^{-2}]$ & $-9.38008\times 10^{-6} [10^{-3}]$ & $-3.67910\times 10^{-4} [10^{-2}]$ & $4.86316\times 10^{-3}$ \\
2471: 0.7 & $140^\circ$ & $7.27850\times 10^{-7} [10^{-4}]$ & $4.62297\times 10^{-5} [10^{-2}]$ & $-1.49671\times 10^{-5} [10^{-4}]$ & $-7.28739\times 10^{-4} [10^{-3}]$ & $3.54271\times 10^{-3}$ \\
2472: 0.7 & $160^\circ$ & $1.16315\times 10^{-6} [10^{-3}]$ & $5.37734\times 10^{-5} [10^{-2}]$ & $-2.12322\times 10^{-5} [10^{-3}]$ & $-1.06961\times 10^{-3} [10^{-2}]$ & $1.22235\times 10^{-3}$ \\
2473: \end{tabular}
2474: \caption{\label{t:fluxes_R} Identical
2475: to Table \ref{t:fluxes_P}, except that these orbits
2476: are retrograde and have semilatus rectum $p = 12$.}
2477: \end{ruledtabular}
2478: \end{table*}
2479:
2480: %Another view of the harmonic structure of these waveforms can be gleamed by considering
2481: %the summed spectra for energy and angular momentum fluxes ($X = E,~L_z$), defined as follows
2482: %\begin{subequations}\label{summed spectra}
2483: %\begin{eqnarray}
2484: %\left< \frac{dX}{dt}\right>_l^{\text{H},\infty} &=& \sum_{mkn} \left<\frac{dX}{dt}\right>_{lmkn}^{\text{H},\infty},\\
2485: %\left< \frac{dX}{dt}\right>_m^{\text{H},\infty} &=& \sum_{lkn} \left<\frac{dX}{dt}\right>_{lmkn}^{\text{H},\infty},\\
2486: %\left< \frac{dX}{dt}\right>_k^{\text{H},\infty} &=& \sum_{lmn} \left<\frac{dX}{dt}\right>_{lmkn}^{\text{H},\infty},\\
2487: %\left< \frac{dX}{dt}\right>_n^{\text{H},\infty} &=& \sum_{lmk} \left<\frac{dX}{dt}\right>_{lmkn}^{\text{H},\infty}.
2488: %\end{eqnarray}
2489: %\end{subequations}
2490: %As shown in Fig.~\ref{f:falloff}, these quantities seem to all decay roughly exponentially with
2491: %falloff rates that are nearly independent of $X$.
2492: %\begin{equation} \label{falloff}
2493: %\left< \frac{dX}{dt}\right>_{\nu}^{\text{H},\infty} \propto \exp\left(-|\nu|\gamma_{\nu}^{\text{H},\infty}\right),
2494: %\end{equation}
2495: %for $\nu = l,m,k,n$.
2496: %\begin{figure*}
2497: %\includegraphics[width = .45\textwidth]{l-spec}
2498: %\includegraphics[width = .45\textwidth]{m-spec}
2499: %\includegraphics[width = .45\textwidth]{k-spec}
2500: %\includegraphics[width = .45\textwidth]{n-spec}
2501: %\caption{Here we show the summed spectra (\ref{summed spectra}) for the generic orbit used
2502: %to produce the top panel of Fig.~\ref{f:generic}.
2503: %The data are marked as follows:
2504: %$\left< dE/dt \right>^\text{H}(M/\mu)$ with $\times$'s,
2505: %$\left< dE/dt \right>^\infty (M/\mu)$ with $+$'s,
2506: %$\left< dL_z/dt \right>^\text{H} / \mu$ with $\square$'s,
2507: %$\left< dL_z/dt \right>^\infty / \mu$ with $\bigcirc$'s.
2508: %The orbit has inclination
2509: %$\theta_\text{inc} = 45^\circ$ ($\iota = 45.2^\circ$), semilatus rectum $p=4$, and eccentricity $e= 0.5$.
2510: %The magnitude of the black hole's spin angular momentum is $aM=0.9M^2$.}
2511: %\label{f:falloff}
2512: %\end{figure*}
2513: %Tables \ref{t:falloff_P} and \ref{t:falloff_R} show the rates of exponential falloff for the summed flux spectra
2514: %(\ref{summed spectra} associated with the catalog of waveforms (Table \ref{t:orbits}).
2515: %These falloff rates $\gamma_\nu^{\text{H},\infty}$ (for $\nu = l,~m,~k,~n$) have varying sensitivity
2516: %to the orbital parameters. For example, a change in eccentricity seems to significantly
2517: %effect $\gamma_n^{\text{H},\infty}$ but not the other rates. This seems reasonable since $n$ is
2518: %the radial harmonic index. So as eccentricity is increased, the radial voice not only dominates
2519: %in amplitude, but it becomes more rich in frequencies as well. To a lesser extent the orbital inclination
2520: %has a similarly enriching effect, as can be seen by its influence on $\gamma_k^{\text{H},\infty}$.
2521: %\input{FallOffTable_P.tex}
2522: %\input{FallOffTable_R.tex}
2523:
2524: Some examples of waveforms are finally shown in the time domain in
2525: Figs.~\ref{f:80deg}. There we plot $h_+$ waveforms for $3 \times
2526: M/(10^6 M_\odot)$ hours as seen from four different viewing angles
2527: $\theta = 0^\circ,~30^\circ,~60^\circ,~90^\circ$. These plots show
2528: that the radial voice has a burst-like character, while the polar
2529: voice is more of a modulated hum. These features are consistent with
2530: the characteristics found in earlier work which focused on circular
2531: orbits \cite{circular} and on equatorial orbits \cite{equatorial}.
2532: \begin{figure*}
2533: \includegraphics[width = .45\textwidth]{fig_0.1_80.eps}
2534: \includegraphics[width = .45\textwidth]{fig_0.3_80.eps}
2535: \includegraphics[width = .45\textwidth]{fig_0.5_80.eps}
2536: \includegraphics[width = .45\textwidth]{fig_0.7_80.eps}
2537: \caption{Snapshot waveforms for orbits with inclination
2538: $\theta_\text{inc} = 80^\circ$, semilatus rectum $p=6$, and
2539: eccentricities $e=$ 0.1, 0.3, 0.5, 0.7. The magnitude of the black
2540: hole's spin angular momentum is $aM=0.9M^2$.}
2541: \label{f:80deg}
2542: \end{figure*}
2543:
2544:
2545: \section{Summary and future work}
2546: \label{s:Summary and future work}
2547:
2548: This work describes the first calculation of gravitational waves, and
2549: asymptotic radiative fluxes of energy and axial angular momentum,
2550: produced by a spinless test mass on a generic bound geodesic of a
2551: spinning black hole. It represents the natural extension of the
2552: earlier works shown in Table \ref{t:history}. The future direction of
2553: this work is rather straightforward: fill in the last checkmark in
2554: Table \ref{t:history} by implementing an evolution scheme for the
2555: constants of the geodesic orbits, and in particular for the Carter
2556: constant. Our anticipated path toward this goal is discussed in more
2557: detail in Ref.~\cite{Hughes et al}. Once completed, we will be able
2558: to compute generic inspiral waveforms which will facilitate the
2559: initial detection of EMRIs. Completing this program will require
2560: implementing Mino's scheme for evolving the Carter constant into
2561: equations which can be entered directly into a code \cite{Mino, Hughes
2562: et al, Drasco Flanagan Hughes, sthn2005}. It will also require a
2563: variety of more computational tasks, such as developing a dynamical
2564: scheme for artificially zeroing badly behaved modes as described in
2565: Secs.~\ref{ss:Truncation} and \ref{ss:Modes}. We are optimistic about
2566: completing this program since most of the remaining work is relatively
2567: straightforward in the sense that there are no known potentially
2568: overwhelming hurdles.
2569:
2570: \begin{acknowledgments}
2571: We thank Marc Favata, \'Etienne Racine, Saul Teukolsky, and especially
2572: \'Eanna Flanagan for helpful discussions.
2573: We also thank Enrico Barausse for identifying several typographical errors in
2574: equations contained in the appendices of an earlier version of this manuscript.
2575: We are particularly
2576: grateful to Kostas Glampedakis and Daniel Kennefick for providing data
2577: that allowed us to compare results with their code \cite{equatorial},
2578: and to Norichika Sago for comparing some of our generic fluxes with
2579: analytical post-Newtonian approximations \cite{Sago in prep}.
2580: We also thank Eric Poisson, Adam Pound, and Bernhard Nickel for
2581: providing an advance draft of Ref.\ {\cite{ppn05}} to us, and Eric
2582: Poisson for extensive discussion regarding the importance of the
2583: conservative self force. S.D.~would like to thank Takahiro Tanaka,
2584: Kyoto University, and the Yukawa Institute for Theoretical Physics,
2585: for hospitality during the final stages of editing this manuscript.
2586: This work was supported at Cornell by NSF Grants PHY-0140209 and
2587: PHY-0457200, and the NASA/New York Space Grant Consortium, and at MIT
2588: by NASA Grant NAGW-12906 and NSF Grant PHY-0244424.
2589: Part of this research was carried out at the Jet Propulsion
2590: Laboratory, California Institute of Technology, under a contract with
2591: the National Aeronautics and Space Administration and funded through the
2592: internal Research and Technology Development program.
2593: \end{acknowledgments}
2594:
2595: \appendix
2596:
2597: \section{Geodesic details}
2598: \label{s:Geodesic details}
2599:
2600: The functions $V_t$, $V_r$, $V_\theta$, and $V_\phi$ appearing in the
2601: geodesic equations (\ref{geodesics}) are given by \cite{MTW,Carter}:
2602: \begin{subequations} \label{detailed geodesics}
2603: \begin{align}
2604: V_t(r,\theta)
2605: &= E \left( \frac{\varpi^4}{\Delta} - a^2\sin^2\theta \right)
2606: + aL_z \left( 1 - \frac{\varpi^2}{\Delta} \right),&
2607: \label{tdot} \\
2608: %
2609: V_r(r)
2610: &= \left( E\varpi^2 - a L_z \right)^2
2611: - \Delta\left[\mu^2 r^2 + (L_z - a E)^2 + Q\right],&
2612: \label{rdot}\\
2613: %
2614: V_\theta(\theta) &= Q - L_z^2 \cot^2\theta - a^2(\mu^2 - E^2)\cos^2\theta,&
2615: \label{thetadot}\\
2616: %
2617: V_\phi(r,\theta)
2618: &= L_z \csc^2\theta + aE\left(\frac{\varpi^2}{\Delta} - 1\right) - \frac{a^2L_z}{\Delta},&
2619: \label{phidot}
2620: \end{align}
2621: \end{subequations}
2622: where we have defined
2623: \begin{equation}
2624: \varpi^2 = r^2 + a^2.
2625: \end{equation}
2626:
2627: We now show some explicit relations which are used to
2628: evaluate functions of the geodesics. As indicated below, these
2629: relations were first derived either by Schmidt \cite{Schmidt}
2630: or in Ref.~\cite{Drasco Hughes}. We reproduce them here for the sake
2631: of completeness.
2632:
2633: When evaluated in order, the following equations determine the energy
2634: $E = \mu\tilde E$, angular momentum $L_z = \mu M \tilde L_z$, and
2635: Carter constant $Q = (\mu M)^2 \tilde Q$ for an orbit of a given
2636: minimum radius $\tilde r_1 = r_{\min}/M$, maximum radius $\tilde r_2 =
2637: r_{\max}/M$, and orbital inclination $\theta_\text{inc} = \pi/2 -
2638: \theta_{\min}$:
2639: \begin{eqnarray}
2640: \tilde E^2 &=&
2641: \frac{\kappa \rho + 2\varepsilon\sigma - 2D\sqrt{\sigma(\sigma\varepsilon^2 + \rho\varepsilon\kappa - \eta\kappa^2)}}
2642: {\rho^2 + 4\eta\sigma},
2643: \\
2644: \tilde L_z &=& -\frac{g_1\tilde E}{h_1}
2645: + D \sqrt{ \frac{g_1^2\tilde E^2}{h_1^2} + \frac{f_1 \tilde E^2 - d_1}{h_1}}.
2646: \\
2647: \tilde Q &=& z_- \left( \beta + \frac{\tilde L_z^2}{1 - z_-}\right).
2648: \end{eqnarray}
2649: These were derived by Schmidt in Appendix B of Ref.~\cite{Schmidt}.
2650: Here the constant $D = \sgn L_z$, so that $D = 1$ for a prograde orbit, and $D = -1$ for a retrograde orbit.
2651: The other constants in these equations are defined in terms of the constants\footnote{Note
2652: that Schmidt defines $z_-$ as $\cos\theta_{\min}$ \cite{Schmidt}.}
2653: \begin{align}
2654: &\tilde a = a/M,&
2655: &z_- = \cos^2\theta_{\min},&
2656: &\beta = \tilde a^2 (1 - \tilde E^2),&
2657: \end{align}
2658: the functions
2659: \begin{eqnarray}
2660: \tilde \Delta(\tilde r) &=& \tilde r^2 - 2\tilde r + \tilde a^2,
2661: \\
2662: d(\tilde r) &=& \tilde \Delta (\tilde r^2 + z_- \tilde a^2),
2663: \\
2664: f(\tilde r) &=& \tilde r^4 + \tilde a^2 \left[ \tilde r(r+2) + z_- \tilde \Delta \right],
2665: \\
2666: g(\tilde r) &=& 2\tilde a \tilde r,
2667: \\
2668: h(\tilde r) &=& \tilde r(\tilde r - 2) + \frac{z_-\tilde \Delta}{1 - z_-},
2669: \end{eqnarray}
2670: and the determinants
2671: \begin{eqnarray}
2672: \kappa &=& d_1h_2 - d_2h_1,
2673: \\
2674: \varepsilon &=& d_1g_2 - d_2g_1,
2675: \\
2676: \rho &=& f_1h_2 - f_2h_1,
2677: \\
2678: \eta &=& f_1g_2 - f_2g_1,
2679: \\
2680: \sigma &=& g_1h_2 - g_2h_1.
2681: \end{eqnarray}
2682: A subscript $1,2$ means the function is to be evaluated at $\tilde
2683: r_{1,2}$.
2684:
2685: Schmidt also derives the following expressions for the coordinate-time
2686: frequencies:
2687: \begin{align}
2688: M \Omega_r &= \frac{\pi p K(k)}{(1 - e^2)\left[ (W + \tilde a^2 z_+ \tilde E X)K(k) - \tilde a^2 z_+ \tilde E X E(k) \right]},&
2689: \label{Momega_r} \\
2690: M \Omega_\theta &= \frac{\pi \beta X\sqrt{z_+}}{2\left[ (W + \tilde a^2 z_+ \tilde E X)K(k) - \tilde a^2 z_+ \tilde E X E(k) \right]},&
2691: \label{Momega_theta}\\
2692: M \Omega_\phi &= \frac{(Z-\tilde L_z X)K(k) + \tilde L_z X \Pi(\pi/2,z_-,k)}{(W + \tilde a^2 z_+ \tilde E X)K(k) - \tilde a^2 z_+ \tilde E X E(k)}.&
2693: \label{Momega_phi}
2694: \end{align}
2695: Here $k = \sqrt{z_- / z_+}$,
2696: \begin{equation}
2697: z_+ = \frac{\tilde L_z^2 + \tilde Q + \beta + \sqrt{(\tilde L_z^2 + \tilde Q + \beta)^2 - 4\beta Q}}{2\beta},
2698: \end{equation}
2699: $K(k)$ is the complete elliptic integral of the first kind, $E(k)$ is the
2700: complete elliptic integral of the second kind, and $\Pi(\phi,n,k)$ is the
2701: Legendre elliptic integral of the third kind \cite{NR}:
2702: \begin{align}
2703: K(k) &= \int_0^{\pi/2} \frac{d\theta}{\sqrt{1 - k^2 \sin^2 \theta}},&
2704: \\
2705: E(k) &= \int_{0}^{\pi/2} d\theta \sqrt{1 - k^2 \sin^2\theta},&
2706: \\
2707: \Pi(\phi,n,k) &= \int_0^{\phi} \frac{d\theta}{(1 - n \sin^2 \theta)\sqrt{1 - k^2 \sin^2 \theta}}.&
2708: \end{align}
2709: The remaining quantities in Eqs.~(\ref{Momega_r})-(\ref{Momega_phi})
2710: are defined by the following integrals:
2711: \begin{align}
2712: W &= \int_0^{\pi} \frac{p^2 F(\chi)d\chi}{(1+e\cos\chi)^2H(\chi)\sqrt{J(\chi)}}
2713: \\
2714: X &= \int_0^{\pi}\frac{d\chi}{\sqrt{J(\chi)}},&
2715: \\
2716: Y &= \int_0^{\pi} \frac{p^2d\chi}{(1+e\cos\chi)^2\sqrt{J(\chi)}},&
2717: \\
2718: Z &= \int_0^{\pi} \frac{G(\chi)d\chi}{H(\chi)\sqrt{J(\chi)}},&
2719: \end{align}
2720: where the functions $F,G,H,J$ are given by:
2721: \begin{align}
2722: F(\chi) &= \tilde E + \frac{\tilde a^2 \tilde E}{p^2}(1+e\cos\chi)^2 &
2723: \nonumber \\
2724: &- \frac{2\tilde a(\tilde L_z - \tilde a \tilde E)}{p^3}(1+e\cos\chi)^3,&
2725: \\
2726: G(\chi) &= \tilde L_z - \frac{2}{p}(\tilde L_z - \tilde a \tilde E)(1+e\cos\chi),&
2727: \\
2728: H(\chi) &= 1 - \frac{2}{p}(1+e\cos\chi) + \frac{\tilde a^2}{p^2}(1+e\cos\chi)^2,&
2729: \\
2730: J(\chi) &= (1-\tilde E^2)(1 - e^2)&
2731: \nonumber\\
2732: &+2\left(1 - \tilde E^2 -\frac{1-e^2}{p}\right)(1+e\cos\chi)&
2733: \nonumber\\
2734: &+\left\{ (1-\tilde E^2)\frac{3+e^2}{1-e^2} - \frac{4}{p} + \frac{1-e^2}{p^2} \right.&
2735: \nonumber \\
2736: &\left. \times \left[ \tilde a^2(1 - \tilde E^2) + \tilde L_z^2 + \tilde Q \right] \right\}(1+e\cos\chi)^2&
2737: \end{align}
2738: The Mino time frequencies are then found using
2739: $\Upsilon_{\phi,\theta,r} = \Gamma \omega_{\phi,\theta,r}$, and
2740: $\Upsilon_\theta = \pi \sqrt{\beta z_+} / [2K(k)]$ (from
2741: Ref.~\cite{Drasco Hughes}).
2742:
2743: The function $w_r(\psi)$ and its derivative were derived in the
2744: Appendix of Ref.~\cite{Drasco Hughes}; however, there we did not write
2745: things explicitly in terms of Schmidt's $J$-function. The results
2746: are:
2747: \begin{eqnarray}
2748: w_r(\psi) &=& \frac{1 - e^2}{p} \int_0^{\psi}
2749: \frac{\Upsilon_r\,d\psi'}{\sqrt{J(\psi')}},
2750: \\
2751: \frac{dw_r}{d\psi}(\psi) &=& \frac{1 - e^2}{p}\frac{\Upsilon_r}{\sqrt{J(\psi)}}.
2752: \end{eqnarray}
2753: The function $w_\theta(\chi)$ and its derivative was derived in the
2754: Appendix of Ref.~\cite{Drasco Hughes}. The results are
2755: \begin{align}
2756: w_\theta(\chi) &= \left\{
2757: \begin{array}{lc}
2758: \Upsilon_\theta \lambda_0(\chi) & 0 \le \chi \le \pi/2
2759: \\
2760: \pi - \Upsilon_\theta \lambda_0(\pi - \chi) & \pi/2 \le \chi \le \pi
2761: \\
2762: \pi + \Upsilon_\theta \lambda_0(\chi - \pi) & \pi \le \chi \le 3\pi/2
2763: \\
2764: 2\pi - \Upsilon_\theta \lambda_0(2\pi - \chi) & 3\pi/2 \le \chi \le 2\pi
2765: \end{array}
2766: \right. ,&
2767: \\
2768: \frac{dw_\theta}{d\chi}(\chi) &= \frac{\pi}{2K(k)}\frac{1}{1-k^2 \cos^2\chi},&
2769: \end{align}
2770: where
2771: \begin{equation}
2772: \lambda_0(\chi) = \frac{1}{\sqrt{\beta z_+}} \left[ K(k) - \mathcal{F}(\pi/2-\chi,k)\right],
2773: \end{equation}
2774: and $\mathcal{F}(\phi,k)$ is the incomplete elliptic integral of the
2775: first kind \cite{NR}:
2776: \begin{equation}
2777: \mathcal{F}(\phi,k) = \int_0^\phi \frac{d\theta}{\sqrt{1 - k^2 \sin^2\theta}}.
2778: \end{equation}
2779:
2780: \section{Perturbation details}
2781: \label{s:Perturbation details}
2782: The differential operators in the master equation (\ref{master}) are
2783: given by
2784: \begin{subequations}
2785: \begin{align}
2786: \widehat{U}_{t \phi r}(r) =
2787: &\frac{(r^2 + a^2)^2}{\Delta} \frac{\partial^2}{\partial t^2}
2788: + 4\left[ \frac{M(r^2-a^2)}{\Delta} - r\right] \frac{\partial}{\partial t}&
2789: \nonumber \\
2790: &+ \frac{a^2}{\Delta} \frac{\partial^2}{\partial \phi^2}
2791: + \frac{4a}{\Delta}(r-M) \frac{\partial }{\partial \phi}&
2792: \nonumber \\
2793: &+ \frac{4Mar}{\Delta} \frac{\partial^2 }{\partial t \partial \phi}
2794: - \Delta^{2} \frac{\partial}{\partial r} \left( \frac{1}{\Delta} \frac{\partial}{\partial r}\right),&
2795: \label{Utphir} \\
2796: \widehat{V}_{t \phi \theta}(\theta) =
2797: &- a^2 \sin^2\theta \frac{\partial^2}{\partial t^2}
2798: - 4ia \cos\theta \frac{\partial}{\partial t}&
2799: \nonumber \\
2800: &- \frac{1}{\sin^2 \theta} \frac{\partial^2 }{\partial \phi^2}
2801: + 4\frac{i \cos\theta}{\sin^2\theta} \frac{\partial}{\partial \phi}&
2802: \nonumber \\
2803: &- \frac{1}{\sin \theta} \frac{\partial}{\partial \theta} \left( \sin \theta \frac{\partial \psi}{\partial \theta}\right)
2804: + 4\cot^2\theta + 2.&
2805: \label{Vtphitheta}
2806: \end{align}
2807: \end{subequations}
2808:
2809: In Boyer-Lindquist coordinates, the Kinnersly tetrad vectors are given by \cite{circular,Chandra}
2810: \begin{subequations}\label{K vectors}
2811: \begin{align}
2812: \vec l &= \frac{\varpi^2}{\Delta} \vec \partial_t
2813: + \vec \partial_r
2814: + \frac{a}{\Delta} \vec \partial_\phi,&
2815: \\
2816: \vec m &= -\frac{\rho}{\sqrt{2}} \left(ia \sin \theta \vec \partial_t
2817: + \vec \partial_\theta
2818: + \frac{i}{\sin \theta} \vec \partial_\phi \right),&
2819: \\
2820: \vec n &= \frac{\varpi^2}{2\Sigma} \vec \partial_t
2821: - \frac{\Delta}{2\Sigma}\vec \partial_r
2822: + \frac{a}{2\Sigma} \vec \partial_\phi,&
2823: \end{align}
2824: \end{subequations}
2825: and the Kinnersly tetrad one-forms are
2826: \begin{subequations}\label{K 1forms}
2827: \begin{align}
2828: \widetilde l &=-\widetilde{dt} + \frac{\Sigma}{\Delta}~\widetilde{dr}+a\sin^2\theta~\widetilde{d\phi},&
2829: \\
2830: \widetilde m &=\frac{\rho}{\sqrt{2}}\left( ia\sin\theta~\widetilde{dt} - \Sigma~\widetilde{d\theta} - i\varpi^2 \sin\theta~\widetilde{d\phi} \right),&
2831: \\
2832: \widetilde n &=-\frac{\Delta}{2\Sigma}~\widetilde{dt} -\frac{1}{2}~\widetilde{dr} + \frac{a\Delta \sin^2\theta}{2\Sigma}~\widetilde{d\phi}.&
2833: \end{align}
2834: \end{subequations}
2835:
2836: The $A_{abc}$ functions are given by
2837: \footnote{Note that the code used in Refs.~\cite{circular,circularII}
2838: had the wrong prefactor for $A_{nn0}$ (it had the factor of $1/\bar\rho$
2839: replaced with a factor of $1/\rho$). Also Ref.~\cite{circular}
2840: has an incorrect expression for $A_{n\bar m 1}$ (though this term is
2841: correct in the code for Refs.~\cite{circular,circularII})}
2842: \begin{subequations} \label{A's}
2843: \begin{align}
2844: A_{nn0}
2845: &= -\frac{2\rho^{-3}\bar\rho^{-1}C_{nn}}{\Delta^2}
2846: \left(L^\dag_1L^\dag_2 S + 2 i a \rho L^\dag_2 S \sin\theta\right),&
2847: \label{eq:Ann0}
2848: \\
2849: A_{n\bar m0}
2850: &= -\frac{2\sqrt{2}\rho^{-3}C_{n\bar m}}{\Delta}
2851: \left[\left(\frac{iK}{\Delta} - \rho - \bar\rho\right)L^\dag_2 S \right.&
2852: \nonumber \\
2853: & \left. + \left( \frac{iK}{\Delta} + \rho + \bar\rho \right)
2854: ia(\rho - \bar\rho)S\sin\theta\right] ,&
2855: \label{eq:Anm0}
2856: \\
2857: A_{\bar m\bar m 0}
2858: &= S\rho^{-3}\bar\rho C_{\bar m \bar m}
2859: \left[\left(\frac{K}{\Delta}\right)^2
2860: + 2 i \rho \frac{K}{\Delta} +
2861: i \partial_r\left(\frac{K}{\Delta}\right)\right],&
2862: \label{eq:Amm0}
2863: \\
2864: %\end{align}
2865: %\begin{align}
2866: A_{n\bar m1}
2867: &= -\frac{2\sqrt{2}\rho^{-3}C_{n\bar m}}{\Delta}
2868: \left[L^\dag_2 S + ia(\rho - \bar\rho) S \sin\theta\right] ,&
2869: \label{eq:Anm1}
2870: \\
2871: A_{\bar m\bar m1}
2872: &= 2S\rho^{-3}\bar\rho\,C_{\bar m\bar m}
2873: \left(\rho - \frac{iK}{\Delta}\right) ,&
2874: \label{eq:Amm1}
2875: \\
2876: A_{\bar m\bar m2}
2877: &= -S\rho^{-3}\bar\rho\,C_{\bar m\bar m} ,&
2878: \label{eq:Amm2}
2879: \end{align}
2880: \end{subequations}
2881: where we have used the shorthand notation $S = S_{lm}(\theta,\omega)$.
2882: Note that the factor of $C_{\bar m \bar m}$ in Eq.~(\ref{eq:Amm0})
2883: was missing in Ref.~\cite{circular} (it was not however missing from
2884: the numerical code associated with Ref.~\cite{circular}).
2885:
2886:
2887: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% BIBLIOGRAPHY %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2888: \begin{thebibliography}{0}
2889:
2890: %\bibitem{begelman2003}
2891: %M.\ C.\ Begelman,
2892: %\emph{Evidence for black holes},
2893: %Science {\bf 300}, 1898 (2003).
2894:
2895: %\bibitem{c22}
2896: %J. Heller,
2897: %\emph{Catch-22},
2898: %(Simon \& Schuster; Reprint edition, New York, 1996).
2899:
2900: %\bibitem{Schodel et al}
2901: %R. Sch\"odel et al.,
2902: %\emph{Closest Star Seen Orbiting the Supermassive Black Hole at the Centre of the Milky Way},
2903: %Nature {\bf 419}, 694 (2002).
2904:
2905: %\bibitem{Ghez et al}
2906: %A. Ghez et al,
2907: %\emph{The First Measurement of Spectral Lines in a Short-Period
2908: %Star Bound to the Galaxy's Central Black Hole: A Paradox of Youth},
2909: %Astrophys. J. {\bf 586}, L127 (2003).
2910:
2911: %\bibitem{Eisenhauer et al}
2912: %F. Eisenhauer,
2913: %\emph{A Geometric Determination of the Distance to the Galactic Center},
2914: %Astrophys. J. {\bf 597}, L121 (2003).
2915:
2916: %\bibitem{Beckwith Done}
2917: %K. Beckwith and C. Done,
2918: %\emph{Iron Line Profiles in Strong Gravity},
2919: %Mon. Not. R. Astron. Soc. {\bf 352}, 353 (2004).
2920:
2921: %\bibitem{AE2}
2922: %http://lheawww.gsfc.nasa.gov/docs/xray/astroe
2923:
2924: %\bibitem{ConX}
2925: %http://constellation.gsfc.nasa.gov
2926:
2927: \bibitem{circular}
2928: S.\ A.\ Hughes,
2929: \emph{Evolution of circular, nonequatorial orbits of Kerr black holes due to gravitational-wave emission},
2930: Phys.\ Rev.\ D {\bf 61}, 084004 (2000);
2931: Phys.\ Rev.\ D {\bf 63}, 049902(E) (2001);
2932: Phys.\ Rev.\ D {\bf 65}, 069902(E) (2002);
2933: Phys.\ Rev.\ D {\bf 67}, 089901(E) (2003).
2934:
2935: \bibitem{equatorial}
2936: K.\ Glampedakis and D.\ Kennefick,
2937: \emph{Zoom and whirl: Eccentric equatorial orbits around spinning black holes and their evolution under gravitational radiation reaction},
2938: Phys.\ Rev.\ D {\bf 66}, 044002 (2002).
2939:
2940: \bibitem{MTW}
2941: C.\ W.\ Misner, K.\ S.\ Thorne, and J.\ A.\ Wheeler,
2942: \emph{Gravitation} (Freeman, San Francisco, 1973).
2943:
2944: \bibitem{Carter}
2945: B.\ Carter,
2946: \emph{Global structure of the Kerr family of gravitational fields},
2947: Phys.\ Rev.\ {\bf 174}, 1559 (1968).
2948:
2949: \bibitem{Mino}
2950: Y.\ Mino,
2951: \emph{Perturbative approach to an orbital evolution around a supermassive black hole},
2952: Phys.\ Rev.\ D {\bf 67}, 084027 (2003).
2953:
2954: \bibitem{Hughes et al}
2955: S.\ A.\ Hughes, S.\ Drasco, {\'E}.\ {\'E}.\ Flanagan, and J.\ Franklin,
2956: \emph{Gravitational radiation reaction and inspiral waveforms in the adiabatic limit},
2957: Phys.\ Rev.\ Lett.\ {\bf 94}, 221101 (2005).
2958:
2959: \bibitem{Drasco Flanagan Hughes}
2960: S.\ Drasco, {\'E}.\ {\'E}.\ Flanagan, and S.\ A.\ Hughes,
2961: \emph{Computing inspirals in Kerr in the adiabatic regime.\ I.\ The scalar case},
2962: Class.\ Quantum Grav.\ {\bf 22}, S801 (2005).
2963:
2964: \bibitem{sthn2005}
2965: N.\ Sago, T.\ Tanaka, W.\ Hikida, and H.\ Nakano,
2966: \emph{Adiabatic radiation reaction to the orbits in Kerr Spacetime},
2967: Prog.\ Theor.\ Phys., in press; gr-qc/0506092.
2968:
2969: \bibitem{fm2000}
2970: L.\ Ferrarese and D.\ Merritt,
2971: \emph{A Fundamental Relation between Supermassive Black Holes and Their Host Galaxies},
2972: Astrophys.\ J.\ {\bf 539}, L9 (2000)
2973:
2974: \bibitem{getal2000}
2975: K.\ Gebhardt et al,
2976: \emph{A Relationship between Nuclear Black Hole Mass and Galaxy Velocity Dispersion},
2977: Astrophys.\ J.\ {\bf 539}, L13 (2000).
2978:
2979: \bibitem{spitzer69}
2980: L.\ Spitzer,
2981: \emph{Equipartition and the Formation of Compact Nuclei in Spherical Stellar Systems},
2982: Astrophys.\ J.\ {\bf 158}, L139 (1969).
2983:
2984: \bibitem{meg2000}
2985: J.\ Miralda-Escud{\'e} and A.\ Gould,
2986: \emph{A Cluster of Black Holes at the Galactic Center},
2987: Astrophys.\ J.\ {\bf 545}, 847 (2000).
2988:
2989: \bibitem{munoetal2005}
2990: M.\ P.\ Muno, E.\ Pfahl, F.\ K.\ Baganoff, W.,
2991: N.\ Brandt, A.\ Ghez, J.\ Lu, and M.\ R.\ Morris,
2992: \emph{An Overabundance of Transient X-Ray Binaries within 1 Parsec of the Galactic Center},
2993: Astrophys.\ J.\ {\bf 622}, L113 (2005).
2994:
2995: \bibitem{freitag2001}
2996: M.\ Freitag,
2997: \emph{Monte Carlo cluster simulations to determine the rate of compact star inspiralling to a central galactic black hole},
2998: Class.\ Quantum Grav.\ {\bf 18}, 4033 (2001).
2999:
3000: \bibitem{LISA}
3001: See the official NASA website http://lisa.nasa.gov .
3002:
3003: \bibitem{lisaesa}
3004: See the official ESA website http://sci.esa.int/science-e/www/area/index.cfm?fareaid=27 .
3005:
3006: \bibitem{Gair et al}
3007: J.\ R.\ Gair et al,
3008: \emph{Event rate estimates for LISA extreme mass ratio capture sources},
3009: Class.\ Quantum Grav.\ {\bf 21}, S1595 (2004).
3010:
3011: %\bibitem{Allen et al F}
3012: %B. Allen, J. D. E. Creighton, \'E. \'E. Flanagan, and J. D. Romano,
3013: %\emph{Robust statistics for deterministic and stochastic gravitational
3014: %waves in non-Gaussian noise I: Frequentist analyses},
3015: %Phys. Rev. D {\bf 65}, 122002 (2002).
3016:
3017: %\bibitem{Allen et al B}
3018: %B. Allen, J. D. E. Creighton, \'E. \'E. Flanagan, and J. D. Romano,
3019: %\emph{Robust statistics for deterministic and stochastic gravitational
3020: %waves in non-Gaussian noise II: Bayesian analyses},
3021: %Phys. Rev. D {\bf 67}, 122002 (2003).
3022:
3023: \bibitem{Barack Cutler}
3024: L.\ Barack and C.\ Cutler,
3025: \emph{LISA Capture Sources: Approximate Waveforms, Signal-to-Noise Ratios, and Parameter Estimation Accuracy},
3026: Phys.\ Rev.\ D {\bf 69}, 082005 (2004).
3027:
3028: \bibitem{ryan95}
3029: F.\ D.\ Ryan,
3030: \emph{Gravitational waves from the inspiral of a compact object into a massive, axisymmetric body with arbitrary multipole moments},
3031: Phys.\ Rev.\ D {\bf 52}, 5707 (1995).
3032:
3033: \bibitem{ryan97}
3034: F.\ D.\ Ryan,
3035: \emph{Accuracy of estimating the multipole moments of a massive body from the gravitational waves of a binary inspiral},
3036: Phys.\ Rev.\ D {\bf 56}, 1845 (1997).
3037:
3038: \bibitem{ch2004}
3039: N.\ A.\ Collins and S.\ A.\ Hughes,
3040: \emph{Towards a formalism for mapping the spacetimes of massive compact objects: Bumpy black holes and their orbits},
3041: Phys.\ Rev.\ D {\bf 69}, 124022 (2004).
3042:
3043: \bibitem{Teukolsky}
3044: S.\ A.\ Teukolsky,
3045: \emph{Rotating Black Holes: Separable Wave Equations for Gravitational and Electromagnetic Perturbations},
3046: Phys.\ Rev.\ Lett.\ {\bf 29}, 1114 (1972);
3047: S.\ A.\ Teukolsky,
3048: \emph{Perturbations of a rotating black hole.\ I.\ fundamental equations for gravitational, electromagnetic, and neutrino-field perturbations},
3049: Astrophys.\ J.\ {\bf 185}, 635 (1973);
3050:
3051: \bibitem{mpryan74}
3052: M.\ P.\ Ryan,
3053: \emph{Teukolsky equation and Penrose wave equation},
3054: Phys.\ Rev.\ D {\bf 10}, 1736 (1974).
3055:
3056: \bibitem{penrose60}
3057: R.\ Penrose,
3058: \emph{A spinor approach to general relativity},
3059: Ann.\ Phys.\ (N.Y.) {\bf 10}, 171 (1960).
3060:
3061: \bibitem{HH}
3062: S.\ W.\ Hawking and J.\ B.\ Hartle,
3063: \emph{Energy and angular momentum flow into a black hole},
3064: Commun.\ Math.\ Phys.\ {\bf 27}, 283 (1972).
3065:
3066: \bibitem{hartle}
3067: J.\ B.\ Hartle,
3068: \emph{Tidal Friction in Slowly Rotating Black Holes}
3069: Phys.\ Rev.\ D {\bf 8}, 1010 (1973);
3070: \emph{Tidal shapes and shifts on rotating black holes}
3071: {\bf 9}, 2749 (1974).
3072:
3073: \bibitem{td1}
3074: C.\ Gundlach, R.\ H.\ Price, and J.\ Pullin,
3075: \emph{Late-time behavior of stellar collapse and explosions. I. Linearized perturbations},
3076: Phys.\ Rev.\ D {\bf 49}, 883 (1994);
3077: \emph{Late-time behavior of stellar collapse and explosions. II. Nonlinear evolution},
3078: {\bf 49}, 890 (1994).
3079:
3080: \bibitem{td2}
3081: C.\ O.\ Lousto and R.\ H.\ Price,
3082: \emph{Head-on collisions of black holes: The particle limit},
3083: Phys.\ Rev.\ D {\bf 55}, 2124 (1997).
3084:
3085: \bibitem{td3}
3086: K.\ Martel,
3087: \emph{Gravitational waveforms from a point particle orbiting a Schwarzschild black hole},
3088: Phys.\ Rev.\ D {\bf 69}, 044025 (2004).
3089:
3090: \bibitem{td4}
3091: K.\ Martel and E.\ Poisson,
3092: \emph{One-parameter family of time-symmetric initial data for the radial infall of a particle into a Schwarzschild black hole},
3093: Phys.\ Rev.\ D {\bf 66}, 084001 (2002).
3094:
3095: \bibitem{Sopuerta et al}
3096: C. F. Sopuerta, P. Sun, P. Laguna, and J. Xu,
3097: \emph{A Toy Model for Testing Finite Element Methods to Simulate Extreme-Mass-Ratio-Binary Systems},
3098: gr-qc/0507112.
3099:
3100: \bibitem{ckp}
3101: C.\ Cutler, D.\ Kennefick, and E.\ Poisson,
3102: \emph{Gravitational radiation reaction for bound motion around a Schwarzschild black hole},
3103: Phys.\ Rev.\ D {\bf 50}, 3816 (1994).
3104:
3105: \bibitem{Finn Thorne}
3106: L.\ S.\ Finn and K.\ S.\ Thorne,
3107: \emph{Gravitational waves from a compact star in a circular, inspiral orbit, in the equatorial plane of a massive, spinning black hole, as observed by {LISA}},
3108: Phys.\ Rev.\ D {\bf 62}, 124021 (2000).
3109:
3110: \bibitem{Shibata circular}
3111: M.\ Shibata,
3112: \emph{Gravitational waves induced by a particle orbiting around a rotating black hole},
3113: Prog.\ Theor.\ Phys.\ {\bf 90}, 595 (1993).
3114:
3115: \bibitem{Shibata equatorial}
3116: M.\ Shibata,
3117: \emph{Gravitational waves by compact stars orbiting around rotating supermassive black holes},
3118: Phys.\ Rev.\ D {\bf 50}, 6297 (1994).
3119:
3120: \bibitem{circularII}
3121: S.\ A.\ Hughes,
3122: \emph{Evolution of circular, nonequatorial orbits of Kerr black holes due to gravitational-wave emission.\ II.\ Inspiral trajectories and gravitational waveforms},
3123: Phys.\ Rev.\ D {\bf 64}, 064004 (2001).
3124:
3125: %\bibitem{Poisson}
3126: %E.\ Poisson,
3127: %\emph{The gravitational self-force},
3128: %gr-qc/0410127.
3129:
3130: \bibitem{ALD}
3131: E.\ Poisson,
3132: \emph{An introduction to the Lorentz-Dirac equation},
3133: gr-qc/9912045.
3134:
3135: \bibitem{Dirac}
3136: P.\ A.\ M.\ Dirac,
3137: \emph{Classical theory of radiating electrons},
3138: Proc.\ R.\ Soc.\ London {\bf A167}, 148 (1938).
3139:
3140: \bibitem{MST}
3141: Y.\ Mino, M.\ Sasaki, and T.\ Tanaka,
3142: \emph{Gravitational radiation reaction to a particle motion},
3143: Phys.\ Rev.\ D {\bf 55}, 3457 (1997).
3144:
3145: \bibitem{QW}
3146: T.\ C.\ Quinn and R.\ M.\ Wald,
3147: \emph{Axiomatic approach to electromagnetic and gravitational radiation reaction of particles in curved spacetime},
3148: Phys.\ Rev.\ D {\bf 56}, 3381 (1997).
3149:
3150: \bibitem{ericlivrev}
3151: E.\ Poisson,
3152: \emph{The Motion of Point Particles in Curved Spacetime},
3153: Living Rev.\ Relativity {\bf 7}, 6 (2004), http://www.livingreviews.org/lrr-2004-6 .
3154:
3155: \bibitem{cqg_special}
3156: A good summary of the current state and recent
3157: progress of this field can be found in a special issue of the
3158: journal Classical and Quantum Gravity: \emph{Gravitational radiation
3159: from binary black holes: Advances in the perturbative approach},
3160: edited by C.\ O.\ Lousto; Class.\ Quantum Grav.\ {\bf 22}, number 15
3161: (2005) (Institute of Physics Publishing, Bristol, 2005).
3162:
3163: \bibitem{ppn05}
3164: A.\ Pound, E.\ Poisson, and B.\ G.\ Nickel,
3165: \emph{Limitations of the adiabatic approximation to the gravitational self-force},
3166: Phys.\ Rev.\ D {\bf 72}, 124001 (2005).
3167:
3168: \bibitem{Drasco Hughes}
3169: S.\ Drasco and S.\ A.\ Hughes,
3170: \emph{Rotating black hole orbit functionals in the frequency domain},
3171: Phys.\ Rev.\ D 69, 044015 (2004).
3172:
3173: \bibitem{Schmidt}
3174: W.\ Schmidt,
3175: \emph{Celestial mechanics in Kerr spacetime},
3176: Class.\ Quant.\ Grav.\ {\bf 19}, 2743 (2002).
3177:
3178: \bibitem{SN}
3179: M.\ Sasaki and T.\ Nakamura,
3180: \emph{Gravitational radiation from a Kerr black hole.\ I},
3181: Prog.\ Theor.\ Phys.\ {\bf 67}, 1788 (1982).
3182:
3183: \bibitem{Wilkins}
3184: D.\ C.\ Wilkins,
3185: \emph{Bound Geodesics in the Kerr Metric},
3186: Phys.\ Rev.\ D {\bf 5}, 814 (1972).
3187:
3188: %\bibitem{Kerr}
3189: %R.\ P.\ Kerr,
3190: %\emph{Gravitational Field of a Spinning Mass as an Example of
3191: %Algebraically Special Metrics},
3192: %Phys.\ Rev.\ Let.\ {\bf 11}, 237 (1963).
3193:
3194: \bibitem{Boyer Lindquist}
3195: R.\ H.\ Boyer and R.\ W.\ Lindquist,
3196: \emph{Maximal Analytic Extension of the Kerr Metric},
3197: J.\ Math.\ Phys.\ {\bf 8}, 265 (1967).
3198:
3199: \bibitem{Wald}
3200: R.\ M.\ Wald,
3201: \emph{On perturbations of a Kerr black hole},
3202: J.\ Math.\ Phys.\ {\bf 14}, 1453 (1973).
3203:
3204: \bibitem{NP}
3205: E.\ Newman and R.\ Penrose,
3206: \emph{An approach to Gravitational Radiation by a Method of Spin Coefficients},
3207: J.\ Math.\ Phys.\ {\bf 3}, 566 (1962).
3208:
3209: \bibitem{Chandra}
3210: S.\ Chandrasekhar,
3211: \emph{The Mathematical Theory of Black Holes} (Oxford University Press, New York, 1983).
3212:
3213: \bibitem{Poisson PDE}
3214: E.\ Poisson,
3215: \emph{Absorption of mass and angular momentum by a black hole: Time-domain formalisms for gravitational perturbations, and the small-hole/slow-motion approximation},
3216: Phys.\ Rev.\ D {\bf 70}, 084044 (2004).
3217:
3218: \bibitem{Khanna}
3219: G.\ Khanna,
3220: \emph{Teukolsky evolution of particle orbits around Kerr black holes in the time domain: Elliptic and inclined orbits},
3221: Phys.\ Rev.\ D {\bf 69}, 024016 (2004).
3222:
3223: \bibitem{Fujita Tagoshi}
3224: R.\ Fujita and H.\ Tagoshi,
3225: \emph{New Numerical Methods to Evaluate Homogeneous Solutions of the Teukolsky Equation},
3226: Prog.\ Theor.\ Phys.\ {\bf 112}, 415 (2004).
3227:
3228: \bibitem{Arfken}
3229: G.\ Arfken,
3230: \emph{Mathematical Methods for Physicists} (third edition, Academic Press, Orlando, FL, 1985).
3231:
3232: \bibitem{Breuer}
3233: R.\ A.\ Breuer, \emph{Gravitational Perturbation Theory and Synchrotron Radiation},
3234: Lecture Notes in Physics Vol.\ 44 (Springer-Verlag, Berlin, 1975).
3235:
3236: \bibitem{Sasaki Tagoshi}
3237: M.\ Sasaki and H.\ Tagoshi,
3238: \emph{Analytic Black Hole Perturbation Approach to Gravitational Radiation},
3239: Living Rev.\ Relativity {\bf 6}, 6 (2003), http://www.livingreviews.org/lrr-2003-6/ .
3240:
3241: \bibitem{Kinnersly}
3242: W.\ Kinnersly,
3243: \emph{Type D Vacuum Metrics},
3244: J.\ Math.\ Phys.\ {\bf 10}, 1195 (1969).
3245:
3246: \bibitem{Minoetalreview}
3247: Y.\ Mino et al.,
3248: \emph{Black hole perturbation},
3249: Prog.\ Theor.\ Phys.\ Suppl.\ {\bf 128}, 1 (1997).
3250:
3251: \bibitem{Blandford Thorne} R.\ D.\ Blandford and K.\ S.\ Thorne,
3252: \emph{Applications of Classical Physics}, unpublished book, available
3253: at http://www.pma.caltech.edu/Courses/ph136/yr2004/, Chapt.\ 26.
3254:
3255: \bibitem{Isaacson}
3256: R.\ A.\ Isaacson,
3257: \emph{Gravitational Radiation in the Limit of High Frequency.\ II.\ Nonlinear Terms and the Effective Stress Tensor},
3258: Phys.\ Rev.\ {\bf 166}, 1272 (1968).
3259:
3260: \bibitem{teukpress}
3261: S.\ A.\ Teukolsky and W.\ H.\ Press,
3262: \emph{Perturbations of Rotating Black Hole.\ III.\ Interaction of the Hole with Gravitational and Electromagnetic Radiation},
3263: Astrophys.\ J.\ {\bf 193}, 443 (1974).
3264:
3265: %\bibitem{Bardeen Press}
3266: %J.\ M.\ Bardeen and W.\ H.\ Press,
3267: %\emph{Radiation fields in the Schwarzschild background},
3268: %J.\ Math.\ Phys.\ {\bf 14}, 7 (1973).
3269:
3270: \bibitem{Cutler private}
3271: C.\ Cutler, private communication.
3272:
3273: \bibitem{gair_private}
3274: J.\ Gair, private communication.
3275:
3276: \bibitem{NR}
3277: W.\ H.\ Press, S.\ A.\ Teukolsky, W.\ T.\ Vetterling, and B.\ P.\ Flannery,
3278: \emph{Numerical Recipes in C++: The Art of Scientific Computing}
3279: (Cambridge University Press, Cambridge, 2002).
3280:
3281: \bibitem{Peters Mathews}
3282: P.\ C.\ Peters and J.\ Mathews,
3283: \emph{Gravitational Radiation from Point Masses in a Keplerian Orbit},
3284: Phys.\ Rev.\ {\bf 131}, 435 (1963).
3285:
3286: \bibitem{Thorne}
3287: K.\ S.\ Thorne,
3288: \emph{Multipole Expansions of Gravitational Radiation},
3289: Rev.\ Mod.\ Phys.\ {\bf 52}, 299 (1980).
3290:
3291: \bibitem{Kennefick}
3292: D.\ Kennefick,
3293: \emph{Stability under radiation reaction of circular equatorial orbits around Kerr black holes},
3294: Phys.\ Ref.\ D {\bf 58}, 064012 (1998).
3295:
3296: \bibitem{Shibata fix}
3297: M.\ Shibata and K.\ Glampedakis, private communications.
3298:
3299: \bibitem{Sago in prep}
3300: N.\ Sago, T.\ Tanaka, W.\ Hikida, H.\ Nakano, K.\ Ganz,
3301: \emph{The adiabatic evolution of orbital parameters in the Kerr spacetime},
3302: gr-qc/0511151.
3303:
3304: %\bibitem{membrane}
3305: %K.\ S.\ Thorne, R.\ H.\ Price, and D.\ A.\ MacDonald,
3306: %{\it Black Holes: The Membrane Paradigm} (Yale University Press, New Haven, 1986), p.\ 38.
3307:
3308: \end{thebibliography}
3309:
3310: \end{document}
3311: