1: %%
2: %% Beginning of file 'sample.tex'
3: %%
4: %% Modified 2004 January 9
5: %%
6: %% This is a sample manuscript marked up using the
7: %% AASTeX v5.x LaTeX 2e macros.
8:
9: %% The first piece of markup in an AASTeX v5.x document
10: %% is the \documentclass command. LaTeX will ignore
11: %% any data that comes before this command.
12:
13: %% The command below calls the preprint style
14: %% which will produce a one-column, single-spaced document.
15: %% Examples of commands for other substyles follow. Use
16: %% whichever is most appropriate for your purposes.
17: %%
18: %%\documentclass[12pt,preprint]{aastex}
19:
20: %% manuscript produces a one-column, double-spaced document:
21:
22: %% \documentclass[manuscript]{aastex}
23:
24: %% preprint2 produces a double-column, single-spaced document:
25:
26: \documentclass[12pt, preprint]{aastex}
27: %% Sometimes a paper's abstract is too long to fit on the
28: %% title page in preprint2 mode. When that is the case,
29: %% use the longabstract style option.
30: %% \documentclass[preprint2,longabstract]{aastex}
31: %% If you want to create your own macros, you can do so
32: %% using \newcommand. Your macros should appear before
33: %% the \begin{document} command.
34: %%
35: %% If you are submitting to a journal that translates manuscripts
36: %% into SGML, you need to follow certain guidelines when preparing
37: %% your macros. See the AASTeX v5.x Author Guide
38: %% for information.
39: \newcommand{\vdag}{(v)^\dagger}
40: \newcommand{\myemail}{skywalker@galaxy.far.far.away}
41: \newcommand{\Ibar}{I\hspace{-0.4em}\rule[0.35em]{0.3em}{0.02em}\hspace{0.1em}}
42: %% You can insert a short comment on the title page using the command below.
43: %% \slugcomment{Not to appear in Nonlearned J., 45.}
44: %% If you wish, you may supply running head information, although
45: %% this information may be modified by the editorial offices.
46: %% The left head contains a list of authors,
47: %% usually a maximum of three (otherwise use et al.). The right
48: %% head is a modified title of up to roughly 44 characters.
49: %% Running heads will not print in the manuscript style.
50: \shorttitle{Secular Bar-mode Instability in Neutron Stars}
51:
52: \shortauthors{Ou, Tohline \& Lindblom}
53:
54: %% This is the end of the preamble. Indicate the beginning of the
55: %% paper itself with \begin{document}.
56:
57: \begin{document}
58:
59: %% LaTeX will automatically break titles if they run longer than
60: %% one line. However, you may use \\ to force a line break if
61: %% you desire.
62:
63: \title{{Nonlinear Development of the Secular Bar-mode Instability\\
64: in Rotating Neutron Stars}}
65:
66: %% Use \author, \affil, and the \and command to format
67: %% author and affiliation information.
68: %% Note that \email has replaced the old \authoremail command
69: %% from AASTeX v4.0. You can use \email to mark an email address
70: %% anywhere in the paper, not just in the front matter.
71: %% As in the title, use \\ to force line breaks.
72:
73: \author{Shangli Ou and Joel E. Tohline}
74: \affil{Department of Physics \& Astronomy, Louisiana State
75: University, Baton Rouge, LA 70803}
76:
77: \and
78:
79: \author{Lee Lindblom}
80: \affil{Theoretical Astrophysics 130-33,
81: California Institute of Technology, Pasadena, CA 91125}
82:
83:
84: %% Mark off your abstract in the ``abstract'' environment. In the manuscript
85: %% style, abstract will output a Received/Accepted line after the
86: %% title and affiliation information. No date will appear since the author
87: %% does not have this information. The dates will be filled in by the
88: %% editorial office after submission.
89:
90: \begin{abstract}
91: We have modelled the nonlinear development of the secular bar-mode
92: instability that is driven by gravitational radiation-reaction
93: (GRR) forces in rotating neutron stars. In the absence of any
94: competing viscous effects, an initially uniformly rotating,
95: axisymmetric $n=1/2$ polytropic star with a ratio of rotational to
96: gravitational potential energy $T/|W| = 0.181$ is driven by GRR
97: forces to a bar-like structure, as predicted by linear theory. The
98: pattern frequency of the bar slows to nearly zero, that is, the
99: bar becomes almost stationary as viewed from an inertial frame of
100: reference as GRR removes energy and angular momentum from the
101: star. In this ``Dedekind-like'' state, rotational energy is
102: stored as motion of the fluid in highly noncircular orbits inside
103: the bar. However, in less than 10 dynamical times after its
104: formation, the bar loses its initially coherent structure as the
105: ordered flow inside the bar is disrupted by what appears to be a
106: purely hydrodynamical, short-wavelength, ``shearing'' type
107: instability. The gravitational waveforms generated by such an
108: event are determined, and an estimate of the detectability of
109: these waves is presented.
110: \end{abstract}
111:
112: \keywords{neutron stars --- hydrodynamics --- secular
113: instabilities --- gravitational radiation}
114:
115: \begin{deluxetable}{lllcll}
116: \tablecolumns{6}
117: \tablewidth{0pt}
118: \tablecaption{Initial Model Parameters\label{TabInitial}}
119: \tablehead{
120: \colhead{} & \multicolumn{2}{c}{Model SPH} &
121: &\multicolumn{2}{c}{Model ROT181} \\
122: \cline{2-3}\cline{5-6}
123: \colhead{Units} & \colhead{code} & \colhead{cgs} && \colhead{code} &
124: \colhead{cgs}
125: }
126: \startdata
127: $c$ & 3.122 & $3.00\times 10^{10}$ && 1.634 & $3.00\times 10^{10}$\\
128: $M$ & 1.359 & $2.80\times 10^{33}$ && 0.1716 & $2.80\times 10^{33}$\\
129: $K$ & 0.7825 & $1.83\times 10^{-10}$ && 0.1281 & $1.83\times 10^{-10}$ \\
130: $R_{\mathrm{eq}}$
131: & 0.8413 & $1.25\times 10^6$ && 0.6102 & $1.97\times 10^{6}$\\
132: $R_{\mathrm{pole}}$
133: & 0.8413 & $1.25\times 10^6$ && 0.2746 & $8.86\times 10^{5}$\\
134: $\bar\rho$ & 0.5460 & $3.42\times 10^{14}$ && 0.4922 & $2.39\times 10^{14}$\\
135: $\Omega_{\mathrm{rot}}$ & 0.0 & 0.0 &&0.9705 & $5.52\times 10^3$\\
136: $J$ & 0.0 & 0.0 && 0.01632 & $1.58\times 10^{49}$
137: \enddata
138: \end{deluxetable}
139:
140: \section{Introduction}
141:
142: As \cite{Ch69} and \cite{T78} have discussed in depth, if a star
143: rotates sufficiently fast -- to a point where the ratio of
144: rotational to gravitational potential energy in the star $T/|W|
145: \gtrsim 0.27$ -- it will encounter a dynamical instability that
146: will result in the deformation of the star into a rapidly
147: spinning, bar-like structure. Although originally identified in
148: configurations that are uniformly rotating and uniform in density,
149: more generalized analyses made it clear that the dynamical
150: bar-mode instability should arise at approximately the same
151: critical value of $T/|W|$ in centrally condensed and
152: differentially rotating stars (cf., Ostriker \& Bodenheimer 1973).
153: A number of groups have used numerical hydrodynamic techniques to
154: follow the nonlinear development of this bar-like structure in the
155: context of the evolution of protostellar gas clouds \citep{TDM85,
156: DGTB86, WT88, PCDL98,CT00} and in the context of rapidly rotating
157: neutron stars \citep{NCT00,B00}. Very recently numerical
158: simulations by \cite{CNLB01} and \cite{SKE02,SKE03} have shown
159: that low-order nonaxisymmetric instabilities can become
160: dynamically unstable at much lower values of $T/|W|$ in stars that
161: have rather extreme distributions of angular momentum. Through
162: linear stability analyses, \cite{KE03} and \cite{WAJ04} have
163: attempted to show the connection between these instabilities and
164: the classical bar-mode instability discovered in stars with less
165: severe distributions of angular momentum. In our present analysis
166: we will not be directly investigating the onset or development of
167: these dynamical instabilities.
168:
169: Classical stability studies have also indicated that a uniformly
170: rotating (or moderately differentially rotating) star with $T/|W|
171: \gtrsim 0.14$ should encounter a secular instability that will
172: tend to deform its structure into a bar-like shape if the star is
173: subjected to a dissipative process capable of redistributing
174: angular momentum within its structure. The nonlinear development
175: of this secular instability has not previously been modelled in a
176: fully self-consistent manner, so it is not yet clear whether stars
177: that encounter this type of instability will evolve to a structure
178: that has a significant bar-like distortion. In this paper, we
179: present results from a numerical simulation that has been designed
180: to follow the nonlinear development of the secular bar-mode
181: instability in a rapidly rotating neutron star. A force due to
182: gravitational radiation-reaction (GRR) serves as the dissipative
183: mechanism that drives the secular development of the bar-mode. By
184: following the development of the bar to a nonlinear amplitude and
185: calculating the rate at which angular momentum and energy are lost
186: from the system due to gravitational radiation, we are able to
187: provide a quantitative estimate of the distance to which such a
188: gravitational-wave source could be detected by existing and
189: planned experiments, such as the laser interferometer
190: gravitational-wave observatory (LIGO).
191:
192: \cite{Ch70} was the first to discover that gravitational
193: radiation-reaction forces can excite the secular bar-mode
194: instability in uniformly rotating, uniform-density stars with
195: incompressible equations of state ({\it i.e.}, the Maclaurin
196: spheroids). This work was generalized by \cite{FS78} and
197: \cite{C79a,C79b}, to show that the GRR instability extends to
198: stars with any equation of state, and to other nonaxisymmetric
199: modes with higher azimuthal mode numbers ($m>2$). \cite{M85},
200: \cite{IFD85}, \cite{IL90}, and \cite{LS95} have all shown that the
201: critical value of $T/|W|$ at which the GRR secular instability in
202: the ($m=2$) bar-mode sets in does not depend sensitively on the
203: polytropic index of the equation of state or the differential
204: rotation law of the star. These stability analyses have also been
205: generalized to systems in which the star is governed by
206: relativistic, rather than purely Newtonian, hydrodynamics and
207: gravitational fields \citep{F78,LH83,C91,CL92,SF98,SZ98,DiGV02}.
208:
209: \cite{LD77} first showed that viscous processes within compact stars
210: can act to suppress the GRR-driven secular bar-mode instability.
211: Various physical viscosities have been considered, for example, shear
212: viscosity due to electron and neutron scattering \citep{FI76,CL87},
213: shear viscosity due to neutrino scattering \citep{KS77,LD79,TD93},
214: bulk viscosity due to weak nuclear interactions
215: \citep{J71,S89,IL91,YE95}, or ``mutual friction'' effects in a
216: superfluid \citep{LM95}. In the present work, we will not be
217: investigating the influence of viscous processes on the GRR-driven
218: bar-mode instability, focusing instead on following the nonlinear
219: development of the bar-mode in purely inviscid systems.
220:
221: \cite{DL77} and \cite{LS95} have made efforts to follow the
222: non-linear evolution of rotating neutron stars that are
223: susceptible to the secular (or the dynamical) bar-mode instability
224: by using energy and angular momentum conservation to construct a
225: sequence of quasi-equilibrium, ellipsoidal configurations. Here,
226: we follow the GRR-driven evolution of the bar-mode in an even more
227: realistic way by integrating forward in time the coupled set of
228: nonlinear partial differential equations that govern dynamical
229: motions in nonrelativistic fluids, and by including a
230: post-Newtonian radiation-reaction force term in the equation of
231: motion. Reviews of this and other instabilities that are expected
232: to arise in young neutron stars have been written by \cite{L97},
233: \cite{L01}, \cite{S03}, and \cite{A03}.
234:
235: \section{Methods}
236: Using the \cite{H86} self-consistent-field technique, we
237: constructed two initial equilibrium stellar models governed by
238: Newtonian gravity and an $n=1/2$ polytropic equation of state;
239: that is, the gas pressure $p$ and density $\rho$ were related
240: through the expression $p = K\rho^{1+1/n}$, where $K$ is a
241: constant. Model ``SPH'' was initially nonrotating and, hence,
242: spherically symmetric; model ``ROT181'' was initially
243: axisymmetric, and uniformly rotating with a ratio of rotational to
244: gravitational potential energy $T/|W| = 0.181$. Other properties
245: of these two initial models are detailed in Table
246: \ref{TabInitial}: $M$ is the mass of the star; $R_{\mathrm{eq}}$
247: and $R_{\mathrm{pole}}$ are the star's equatorial and polar radii,
248: respectively; $\bar{\rho}$ is the star's mean density;
249: $\Omega_{\mathrm{rot}}$ is the angular velocity of rotation; and
250: $J$ is the star's total angular momentum. Columns 2 and 4 of Table
251: \ref{TabInitial} give the values of these various quantities in
252: dimensionless code units, where we have assumed the gravitational
253: constant, the star's central density, and the radial extent of the
254: computational grid are all equal to unity ({\it i.e.},
255: $G=\rho_{\mathrm{c}} = \varpi_{\mathrm{grid}}=1$). Columns 3 and 5
256: of Table \ref{TabInitial} give the value of each quantity in cgs
257: units, assuming both stars have $M =1.4~\mathrm{M}_{\odot}$ and $K
258: = 1.83\times10^{-10} ~\mathrm{cm^8~g^{-2}~s^{-2}}$. (This value of
259: $K$ produces a spherical $1.4~\mathrm{M}_{\odot}$ star with a
260: radius of $12.5~\mathrm{km}$, which is characteristic of a neutron
261: star.)
262:
263: \begin{figure}
264: \epsscale{1.0}\plotone {f1.eps} \caption{Velocity vectors in the
265: equatorial plane of model SPH at time $t=0$, but after the
266: nonrotating model was perturbed by the $\ell=m=2$ ``bar-mode''
267: eigenfunction drawn from linear theory.\label{velPerturb}}
268: \end{figure}
269:
270: Each model was introduced into our hydrodynamical code along with
271: a low-amplitude, nonaxisymmetric perturbation that was designed to
272: closely approximate the eigenfunction of the $\ell=m=2$
273: ``bar-mode'' in a spherical, $n=1/2$ polytrope \citep{IL90}. As an
274: illustration, Fig.~\ref{velPerturb} shows the perturbed velocity
275: field that was introduced in the equatorial plane of model SPH
276: along with a low-amplitude, bar-like distortion in the density
277: which oriented the bar along the vertical axis. Then the nonlinear
278: hydrodynamical evolution of each model was followed using the
279: numerical simulation techniques described in detail by
280: \cite{MTF02}. More specifically, we integrated forward in time a
281: finite-difference approximation of the following coupled set of
282: partial differential equations:
283: \begin{eqnarray}
284: \frac{\partial\rho}{\partial t}+\vec{\nabla}\cdot(\rho\vec{v})&=&0,\\
285: \label{motion}\rho(\frac{\partial\vec{v}}{\partial t} +
286: \vec{v}\cdot\vec{\nabla}\vec{v}) &=& -\vec{\nabla}p
287: - \rho\vec{\nabla}(\Phi + \kappa \Phi_{GR}),\qquad\\
288: \frac{\partial{\tau}}{\partial{t}}+\nabla\cdot(\tau \vec{v})&=&0,\\
289: \nabla^2\Phi &=& 4\pi G\rho,
290: \end{eqnarray}
291: where $\vec{v}$ is the fluid velocity, \($$\Phi$$\) is the
292: Newtonian gravitational potential, $\tau \equiv
293: (\epsilon\rho)^{1/\Gamma}$ is the entropy tracer, $\epsilon$ is
294: the specific internal energy, $p=$$(\Gamma-1)\epsilon\rho$ and
295: $\Gamma=1+1/n =3$. Because the models were initially constructed
296: using a polytropic index $n=1/2$ and they were evolved using an
297: adiabatic form of the first law of thermodynamics (Eq. 3) with an
298: adiabatic exponent $\Gamma=3$, the models effectively maintained
299: uniform specific entropy at a value specified by the initial
300: model's polytropic constant, $K$.
301:
302: In the equation of motion, Eq.~(2), we included
303: the post-Newtonian approximation to the gravitational
304: radiation-reaction potential produced by a time-varying,
305: $\ell=m=2$ mass-quadrupole moment \citep{IL91},
306: \begin{equation}
307: \Phi_{GR} \equiv -\sqrt{\frac{2\pi}{375}} \biggl(\frac{G}{c^5}\biggr)\varpi^2e^{2i\phi}D_{22}^{(5)} \,
308: ,
309: \end{equation}
310: where $D_{22}^{(5)}$ is the fifth time-derivative of the
311: quadrupole moment and $c$ is the speed of light. For modeling
312: purposes, a dimensionless coefficient $\kappa$ was affixed to the
313: radiation-reaction potential term in the equation of motion. By
314: adjusting the value of $\kappa$, we could readily remove or
315: artificially enhance the effect of this non-Newtonian GRR force.
316:
317:
318: As implemented on our cylindrical computational mesh
319: ($\varpi,\phi,z$), $D_{22}$ and its first time-derivative were
320: evaluated using the following expressions,
321: \begin{eqnarray}
322: D_{22}&=&\sqrt{\frac{15}{32\pi}}\int\rho\varpi^2e^{-2i\phi}d^{\,3}x \, , \\
323: D_{22}^{(1)}&=&\sqrt{\frac{15}{8\pi}}\int\rho\varpi[v_{\varpi}-iv_{\phi}]e^{-2i\phi} d^{\,3}x \, .
324: \end{eqnarray}
325: Following Lindblom, Tohline, \& Vallisneri (2001, 2002) we have
326: assumed that the quadrupole moment has a time-dependence of the
327: form $D_{22}\propto e^{-i\omega_{22}t}$, hence,
328: \begin{equation}
329: D_{22}^{(n)} = (-i\omega_{22})^n D_{22} \, ,
330: \end{equation}
331: where, at any instant in time, the complex eigenmode frequency
332: $\omega_{22}=\omega_{\mathrm{r}} + i\omega_{\mathrm{i}}$ can be
333: determined by taking the ratio $D_{22}^{(1)}/D_{22}$, so in
334: Eq.~(5) we set
335: \begin{equation}
336: D_{22}^{(5)} = |\omega_{22}|^4 D_{22}^{(1)} \, \label{eqn:D22_d5} .
337: \end{equation}
338:
339: \section{Predictions of Linear Theory}
340:
341: \begin{figure}
342: \epsscale{.90}\plotone{f2.eps} \caption{Solid curves depict the
343: time-evolution of the amplitude $|D_{22}|$ (bottom) and the real
344: (top) and imaginary (middle) components of the $\ell=m=2$ bar-mode
345: frequency from model SPH. Time is shown in units of the predicted
346: pattern period; $|D_{22}|$ has been normalized to
347: $MR_{\mathrm{eq}}^2$; and frequencies are shown in dimensionless
348: code units. Dash-dotted lines show predictions of linear theory.
349: \label{SPHfrequencies}}
350: \end{figure}
351:
352: Using the linear perturbation techniques described by
353: \cite{IL90,IL91} we determined that in model SPH,
354: $\omega_{\mathrm{r}} = \mathrm{Re}(\omega_{22}) = \pm (1.21 \pm
355: 0.01)\Omega_0$ and, if the model is scaled to a mass of $1.4
356: M_{\odot}$ and a radius of 12.5 km, $\omega_{\mathrm{i}}=
357: \mathrm{Im}(\omega_{22}) = -
358: 1.00\pm0.01\times10^{-3}\kappa\Omega_0$, where $\Omega_0 \equiv
359: \sqrt{\pi G\bar{\rho}} = 1.308$ in dimensionless code units. (The
360: uncertainty is estimated from the values that are determined
361: numerically by the linear perturbation method for different radial
362: resolutions.) In model SPH, therefore, we should expect the
363: amplitude of the bar-mode to damp exponentially on a time scale
364: $\tau_{\mathrm{GR}} \equiv 1/|\omega_{\mathrm{i}}| = 193
365: \kappa^{-1}\tau_{\mathrm{pattern}}$, where
366: $\tau_{\mathrm{pattern}}\equiv 2\pi/|\omega_{\mathrm{r}}| = 3.97$
367: in dimensionless code units.
368:
369: Linear perturbation analyses have not yet provided quantitative
370: values of the bar-mode eigenfrequency in rapidly rotating, $n=1/2$
371: polytropes. From the information given in \cite{IL90,IL91},
372: however, we expect that in model ROT181, (a)
373: $\omega_{\mathrm{r}}/\Omega_0$ should be positive but close to
374: zero, as viewed from an inertial reference frame; and (b)
375: $\omega_{\mathrm{i}}/\Omega_0$ should be slightly positive -- that
376: is, the mass-quadrupole moment should {\it grow} exponentially,
377: but on a time scale that is very long compared to the {\it
378: damping} time predicted for the nonrotating model, SPH. More
379: specifically, if $|\omega_{\mathrm{r}}|/\Omega_0$ proves to be an
380: order of magnitude smaller in model ROT181 than it is in model
381: SPH, then we should expect $\tau_{\mathrm{GR}}$ to be $\sim 10^5$
382: times larger because the amplitude of the GRR driving term in Eq.
383: (2) is proportional to $\omega_{22}^5$.
384:
385: \section{Barmode Evolutions}
386: \subsection{Model SPH}
387:
388: Initially the perturbation applied to model SPH had an amplitude
389: $D_0 \equiv |D_{22}(t=0)| \approx 10^{-3}$ and a velocity field
390: (see Fig.~\ref{velPerturb}) designed to excite the ``backward
391: moving'' $\ell=m=2$ bar-mode, that is, the mode for which
392: $\omega_{\mathrm{r}} < 0$. We followed the evolution of the model
393: on a cylindrical grid with a resolution of $66 \times 128 \times
394: 130$ zones in $\varpi$, $\phi$, and $z$, respectively, and with
395: the coefficient of the radiation-reaction force term in Eq.~(2)
396: set to the value $\kappa = 20$. The effect of this was to shorten
397: the timescale for the exponential decay by a factor of 20, to a
398: predicted value $\tau_{\mathrm{GR}} = 9.63
399: \tau_{\mathrm{pattern}}$. By shortening the decay timescale in
400: this manner, we were able to significantly reduce the amount of
401: computational resources that were required to follow the decay of
402: the barmode while maintaining a decay rate that was slow compared
403: to the characteristic dynamical time of the system, $\Omega_0^{-1}
404: \approx 0.2\tau_{\mathrm{pattern}}$. This is the same technique
405: that was successfully employed by \cite{LTV01,LTV02} in an earlier
406: investigation of the r-mode instability in young neutron stars.
407:
408: Figure \ref{SPHfrequencies} displays the key results from our SPH
409: model evolution. The solid curves in the top two frames display
410: $\omega_{\mathrm{r}}$ and $\omega_{\mathrm{i}}$ as a function of
411: time through just over nine pattern periods. Each of these
412: frequencies oscillate about a fairly well-defined, mean value:
413: $\langle\omega_{\mathrm{r}}\rangle\approx - 1.56 = -
414: 1.19\Omega_0$; $\langle\omega_{\mathrm{i}}\rangle\approx -0.03 =
415: -0.023\Omega_0$. Oscillations about these mean values initially
416: had an amplitude $\sim \pm 0.05 = 0.038\Omega_0$, indicating that
417: our initial nonaxisymmetric perturbation did not excite a pure
418: eigenmode, but these oscillations decreased in amplitude somewhat
419: as the evolution proceeded. Our measured values of
420: $\omega_{\mathrm{r}}$ and $\omega_{\mathrm{i}}$ are within 3\% and
421: 15\%, respectively, of the values predicted from linear theory
422: (see \S3). The solid curve in the bottom frame of
423: Fig.~\ref{SPHfrequencies} shows in a semi-log plot the behavior of
424: $|D_{22}|$ with time. (Note that in this figure, $|D_{22}|$ has
425: been normalized to $MR_{\mathrm{eq}}^2$.) There is a clear
426: exponential decay with a measured damping time (given by the slope
427: of the solid curve) $\tau_{\mathrm{GR}}\approx
428: 8.45\tau_{\mathrm{pattern}}$. This decay time is completely
429: consistent with the measured value of
430: $\langle\omega_{\mathrm{i}}\rangle$ that we have obtained from the
431: middle frame of Fig.~\ref{SPHfrequencies} and, again, within 15\%
432: of the predicted value (illustrated by the solid dash-dotted line
433: in the top frame of the figure). The somewhat larger discrepancy
434: in the measured value of $\omega_i$ is most probably due to the
435: fact that the GRR formalism used here was derived under the
436: assumption that $|\omega_i|<<|\omega_r|$. Since $\omega_i$ is
437: caused by the $\Phi_{GR}$ potential which is proportional to the
438: fifth power of the frequency, fractional discrepancies which are
439: of order $5|\omega_i/\omega_r|\approx 0.1$ are not unexpected.
440:
441: The first row of numbers in Table \ref{TabResults} summarizes
442: these simulation results. Specifically, columns 4, 5 and 6 list
443: the values of $\omega_r$, $\omega_i$, and $\tau_{\mathrm{GR}}$
444: that have been drawn directly from Fig.~\ref{SPHfrequencies}; all
445: three numbers are given in dimensionless code units. In the last
446: two columns of this table, the real and imaginary frequencies have
447: been reexpressed in units of the dynamical frequency, $\Omega_0$.
448: In the last column, we also have adjusted $\omega_i$ by the factor
449: of $\kappa$ in order to show the frequency (and associated growth
450: rate) as it would appear in a real neutron star where the GRR
451: force would not be artificially exaggerated.
452:
453: \subsection{Model ROT181}
454:
455: \begin{deluxetable}{lrrrrcrc}
456: \tablewidth{0pt} \tablecaption{Simulation
457: Results\label{TabResults}}
458:
459: \tablehead{ \colhead{Model} & \colhead{$\Omega_0$} &
460: \colhead{$\kappa$} & \colhead{$\omega_{\mathrm{r}}$} &
461: \colhead{$\omega_{\mathrm{i}}$} & \colhead{$\tau_{\mathrm{GR}}$} &
462: \colhead{$\omega_{\mathrm{r}}/\Omega_0$} &
463: \colhead{$\omega_{\mathrm{i}}/(\Omega_0\kappa)$} } \startdata
464:
465: SPH & 1.308 & $2.00\times10^1$ & $-1.56$ & $-0.03$ & 34 & $-1.19$ & $-1.1\times10^{-3}$ \\
466: ROT181 & 1.488 & $1.75\times10^5$ & $+0.27$ & $+0.08$ & 12 & $+0.18$ & $+3.1\times10^{-7}$ \\
467:
468: \enddata
469: \end{deluxetable}
470:
471: \subsubsection{Radiation-reaction with $\kappa = 1.75\times10^5$}
472: Model ROT181 was introduced into our hydrocode with a
473: nonaxisymmetric perturbation in the density that had the same
474: structure as the perturbation that was introduced into model SPH.
475: Because we expected the natural oscillation frequency of the
476: bar-mode to be close to zero (as viewed from an inertial reference
477: frame), however, we did not perturb the velocity field of the
478: model. We followed the evolution of model ROT181 on a cylindrical
479: grid with a resolution of $130\times128\times98$ zones in
480: $\varpi$, $\phi$, and $z$, respectively, and with the coefficient
481: of the radiation-reaction force term set to $\kappa =
482: 1.75\times10^5$. Note that fewer vertical grid zones were
483: required than in model SPH because model ROT181 was significantly
484: rotationally flattened, but more radial zones were used than in
485: model SPH in order to allow room for model ROT181 to expand
486: radially during the nonlinear-amplitude phase of its evolution. A
487: much larger value of $\kappa$ was selected because, as explained
488: earlier, the natural growth rate of the bar-mode in model ROT181
489: was expected to be orders of magnitude smaller than the decay rate
490: measured in model SPH.
491:
492: \begin{figure}
493: \epsscale{.90} \plotone{f3.eps} \caption{The time-evolution of the
494: amplitude $|D_{22}|$ (top) and the $\ell=m=2$ bar-mode frequency
495: $\omega_{22}$ (bottom) from two rapidly rotating models. Time is
496: shown in units of the initial rotation period
497: $\tau_{\mathrm{spin}}=2\pi/\Omega_{\mathrm{rot}}$ of the model;
498: $|D_{22}|$ has been normalized to $MR_{\mathrm{eq}}^2$; and
499: frequencies are shown in dimensionless code units. Curves that
500: terminate at approximately $17\tau_{\mathrm{spin}}$ display data
501: from model ROT181 and curves that extend past
502: $35\tau_{\mathrm{spin}}$ show data from the lower resolution model
503: ROT179. In the bottom panel, both the real (eg., dash-dotted curve
504: for model ROT181) and imaginary (eg., solid curve for model
505: ROT181) components of $\omega_{22}$ are displayed.
506: \label{rot181freq}}
507: \end{figure}
508:
509: Figures \ref{rot181freq} and \ref{rot181ToverW} display some of
510: the key results from this ROT181 model evolution. The bottom
511: frame of Fig. \ref{rot181freq} shows the time-dependent behavior
512: of the real (dash-dotted curve) and imaginary (solid curve)
513: components of $\omega_{22}$, in our code's dimensionless frequency
514: units; the solid curve in the top frame displays the time-dependent behavior of
515: $|D_{22}|$, normalized to $MR_{\mathrm{eq}}^2$. Figure
516: \ref{rot181ToverW} shows how the global parameters $T/|W|$ (solid
517: curve) and $J$ (dashed curve) evolve with time. The behavior of
518: the model can be best described in the context of three different
519: evolutionary phases: {\it Early} [$0 \leq t/\tau_{\mathrm{spin}}
520: \lesssim 7$]; {\it intermediate} [$7 \lesssim
521: t/\tau_{\mathrm{spin}} \lesssim 12$]; and {\it late}
522: [$t/\tau_{\mathrm{spin}} \gtrsim 12$], where
523: $\tau_{\mathrm{spin}}\equiv 2\pi/\Omega_{\mathrm{rot}} = 6.47$ in
524: dimensionless code units.
525:
526: During the model's {\it early} evolution, both components of the
527: frequency $\omega_{22}$ oscillate about well-defined, mean values:
528: $\langle\omega_{\mathrm{r}}\rangle\approx 0.27 = 0.181\Omega_0$;
529: $\langle\omega_{\mathrm{i}}\rangle\approx 0.08 = 0.054\Omega_0$.
530: (Following Ipser \& Lindblom 1991, we define $\Omega_0$ in terms
531: of the mean density $\bar{\rho}_0$ of a spherical star that has
532: the same $M$ and $K$ as model ROT181, that is, $\Omega_0 \equiv
533: \sqrt{\pi G \bar{\rho}_0} = 1.488$ in dimensionless code units;
534: see Table \ref{TabResults}.) During this same phase of the
535: evolution, both $J$ and $T/|W|$ remain fairly constant, but
536: $|D_{22}|$ increases exponentially with a growth time (obtained
537: from the slope of the displayed curve) $\tau_{\mathrm{GR}}\approx
538: 1.85 \tau_{\mathrm{spin}}$. This growth time is completely
539: consistent with the measured value of
540: $\langle\omega_{\mathrm{i}}\rangle$, from which we would expect
541: $\tau_{\mathrm{GR}}/\tau_{\mathrm{spin}} =
542: \langle\omega_{\mathrm{i}}\rangle^{-1}
543: (\Omega_{\mathrm{rot}}/2\pi) = 1.93$. The second row of numbers
544: in Table \ref{TabResults} summarizes these simulation results.
545:
546: After approximately seven rotation periods, the amplitude of
547: $|D_{22}|$ begins to saturate, and the model deforms into a
548: clearly visible bar-like configuration with an axis ratio measured
549: in the equatorial plane of approximately 2:1 (see
550: Fig.~\ref{velocityVectorsEarly}). The bar-like structure is
551: initially spinning with a frequency given by
552: $\langle\omega_{\mathrm{r}}\rangle/2$, as measured during the {\it
553: early} phase of the ROT181 evolution. This pattern frequency of
554: the bar is a factor of 7.2 smaller than the rotation frequency
555: $\Omega_{\mathrm{rot}}$ of the model in its initial, axisymmetric
556: state, so it is not surprising that the bar also exhibits sizeable
557: internal motions -- it has a ``Dedekind-like'' structure. Figure
558: \ref{velocityVectorsEarly} illustrates the structure of the model
559: at this time. Both frames contain the same set of
560: equatorial-plane, isodensity contours delineating the bar, along
561: with a set of velocity vectors depicting the fluid flow inside the
562: bar: on the left-hand-side, the velocity vectors are drawn in a
563: frame corotating with the bar ({\it i.e.}, rotating at the
564: frequency, $\langle\omega_{\mathrm{r}}\rangle/2$) to illustrate
565: the elliptical streamlines of fluid flow within the
566: ``Dedekind-like'' bar; on the right-hand-side, the velocity
567: vectors are drawn in a frame rotating at the frequency
568: $\Omega_{\mathrm{rot}}$. When viewed in this latter frame, one
569: sees a global velocity structure that is very similar to the
570: flow-field depicted in Fig.~\ref{velPerturb}, that is, it
571: resembles the natural eigenfunction of the $\ell = m = 2$ bar-mode
572: that was derived by perturbation analysis for nonrotating
573: spherical stars, such as our model SPH. We note that this
574: velocity structure developed spontaneously in model ROT181, as the
575: initial model contained no velocity perturbation.
576:
577: During this {\it intermediate} phase of the model's evolution, the
578: bar remains a robust configuration, but its pattern frequency
579: slows as the system loses approximately 10\% of its angular
580: momentum (through gravitational radiation) and $T/|W|$ drops to a
581: value $\sim 0.156$. It is particularly interesting to note that,
582: during this phase of the evolution, the GRR driving term in the
583: equation of motion reaches a maximum, then drops as rapidly as it
584: initially rose; this is illustrated in Fig.~\ref{rot181d22}, where
585: we have plotted the time-dependent behavior of the product,
586: $|\omega_{22}|^5 |D_{22}|$. Although the bar maintains a
587: nonlinear structure, {\it i.e.}, $|D_{22}|$ remains large, during
588: this {\it intermediate} phase of the model's evolution
589: $\Phi_{\mathrm{GR}}$ drops quickly in concert with a decrease in
590: the frequency $|\omega_{22}|$.
591:
592:
593: \begin{figure}
594: \epsscale{.90} \plotone{f4.eps} \caption{The time-evolution of the
595: angular momentum $J$ and the energy ratio $T/|W|$ from model
596: ROT181; $J$ is in dimensionless code units, time is shown in units
597: of the initial rotation period of the model. During the
598: {\it intermediate} phase of the evolution, both quantities noticeably
599: drop as angular momentum is lost via the GRR force term in the
600: equation of motion. \label{rot181ToverW}}
601: \end{figure}
602:
603: \begin{figure}
604: \epsscale{1.0} \plotone{f5.eps} \caption{The structure of model
605: ROT181 is shown at time $t = 8\tau_{\mathrm{spin}}$, during the
606: {\it intermediate} phase of its evolution. In both frames, solid
607: curves are isodensity contours in the equatorial plane while
608: vectors illustrate the equatorial-plane, velocity flow field as
609: viewed from a frame rotating with a specific frequency as follows:
610: $\Omega_{\mathrm{frame}}=\langle\omega_{\mathrm{r}}\rangle/2$
611: (left); $\Omega_{\mathrm{frame}}=\Omega_{\mathrm{rot}}$ (right).
612: \label{velocityVectorsEarly}}
613: \end{figure}
614:
615: \begin{figure}
616: \epsscale{.90} \plotone{f6.eps} \caption{From model ROT181
617: (ROT179), the solid (dotted) curve depicts the time-evolution of
618: the product $\omega_{22}^5|D_{22}|$, which indicates the strength
619: of $\Phi_{\mathrm{GR}}$ in the equation of motion. Time is shown
620: in units of the initial rotation period. \label{rot181d22}}
621: \end{figure}
622:
623: During the {\it late} phase of the model ROT181 evolution, the
624: Dedekind-like bar began to lose its coherent structure.
625: Small-scale fluctuations in the density and velocity fields
626: developed throughout the volume of the bar, and these fluctuations
627: grew in amplitude on a dynamical time scale. Even vertical
628: oscillations developed throughout the model, disrupting both the
629: vertically stratified planar flow and reflection symmetry through
630: the equatorial plane that persisted throughout the {\it early} and
631: {\it intermediate} phases of the model's evolution. After
632: approximately $15\tau_{\mathrm{spin}}$, the model was no longer a
633: recognizable bar, although it remained decidedly nonaxisymmetric,
634: showing density and velocity structure on a wide range of scales
635: in all three dimensions. Figure \ref{velocityVectorsLate}
636: provides a snapshot of model ROT181's structure at
637: $t=19.9~\tau_{\mathrm{spin}}$ during the {\it late} phase of its
638: evolution. (Actually, Fig.~\ref{velocityVectorsLate} is drawn from
639: the {\it late} phase of a ``revised'' evolution of model ROT181,
640: which was evolved further in time; see \S\ref{revised} for
641: details.) Isodensity contours reveal a nonaxisymmetric structure
642: that no longer can be described simply as a bar and, when viewed
643: from a frame rotating at a frequency $\Omega_{\mathrm{rot}}$ (the
644: right-hand frame), the flow field is seen to be more complex than
645: in the bar.
646:
647: \begin{figure}
648: \epsscale{1.0} \plotone{f7.eps} \caption{The neutron star's
649: structure is shown at time $t = 19.9~\tau_{\mathrm{spin}}$ during
650: the {\it late} phase of the ``revised'' ROT181 model evolution. In
651: both frames, solid curves are isodensity contours in the
652: equatorial plane while vectors illustrate the equatorial-plane,
653: velocity flow field as viewed from a frame rotating with a
654: specific frequency as follows: $\Omega_{\mathrm{frame}}=0$ (left);
655: $\Omega_{\mathrm{frame}}=\Omega_{\mathrm{rot}}$
656: (right).\label{velocityVectorsLate}}
657: \end{figure}
658:
659: \begin{figure}
660: \epsscale{.90} \plotone{f8.eps} \caption{The solid curve traces
661: the evolution of model ROT181 in a ``strain-frequency'' diagram
662: from $6\tau_{\mathrm{spin}}$ to $11\tau_{\mathrm{spin}}$. As is
663: schematically illustrated by the vertical dotted line, initially,
664: the amplitude $h_{\mathrm{norm}}$ of the gravitational wave signal
665: grows at a constant frequency, $f
666: =\omega_{\mathrm{r}}/(2\pi)\approx 240~\mathrm{Hz}$. As energy and
667: angular momentum are radiated from the system, the frequency drops
668: monotonically, and the strain reaches a maximum amplitude then
669: steadily declines. \label{DongLaiPlot}}
670: \end{figure}
671:
672: \subsubsection{Detectability of gravitational-wave radiation}
673: A rapidly spinning neutron star located in our Galaxy (and perhaps
674: anywhere in our local group of galaxies) that acquires the type of
675: nonlinear-amplitude, bar-like structure that developed in model
676: ROT181 will produce gravitational radiation at a frequency and
677: amplitude that should soon be detectable by gravitational-wave
678: detectors such as LIGO \citep{LIGO,S1burst}, VIRGO \citep{VIRGO},
679: GEO600 \citep{GEOa,GEOb}, or TAMA300 \citep{TAMA}. As our
680: simulation shows, however, both the amplitude and pattern
681: frequency of the bar --- and, hence, the strength and observed
682: frequency of the gravitational radiation --- will vary with time.
683: To illustrate this, Fig.~\ref{DongLaiPlot} depicts the evolution
684: of model ROT181 across a ``strain-frequency'' diagram, which is
685: often referenced by the experimental relativity community when
686: discussing detectable sources of gravitational radiation.
687: Specifically, the dimensionless strain
688: $h_{\mathrm{norm}}\equiv\sqrt{h_{+}^2+h_{\times}^2}$, where
689: $h_{+}$ and $h_{\times}$ are the two polarization states of
690: gravitational waves. For an observer located a distance $r$ along
691: the axis ($\theta=0$, $\phi=0$) of a spherical coordinate system
692: with the origin located at the center of mass of the system, we
693: have $h_{+} = \frac{G}{c^4} \frac{1}{r}
694: (\,\ddot{\!\Ibar}_{xx}-\,\ddot{\!\Ibar}_{yy})$ and $h_{\times} =
695: \frac{G}{c^4} \frac{2}{r} \,\ddot{\!\Ibar}_{xy}$, where the
696: reduced moment of intertia $\Ibar_{lm} \equiv \int \rho (x_{l}
697: x_{m}-\frac{1}{3}\delta_{lm}x_kx_k)dx^3$. To obtain the strain
698: values $h_{\mathrm{norm}}$ shown in Fig.~\ref{DongLaiPlot}, we
699: have assumed $r=10\mathrm{kpc}$, and the time-derivative of each
700: reduced moment of inertia was evaluated numerically using the
701: method recommended by \cite{FE90}. Model ROT181's evolutionary
702: trajectory in this diagram is strikingly similar to the trajectory
703: that was predicted by \cite{LS95} -- see their Fig.~4 -- using a
704: much simpler, approximate model for the development of the secular
705: bar-mode instability in young neutron stars.
706:
707: In order to estimate the distance to which a gravitational wave
708: source of this type would be detectable by a gravitational-wave
709: interferometer, such as LIGO, we could integrate under the curve
710: in Fig.~\ref{DongLaiPlot}, taking into account the amount of time
711: that the source spends in each frequency band. Because we have
712: artificially amplified the strength of the GRR force, however, our
713: model evolves through frequency space along the curve shown in
714: Fig.~\ref{DongLaiPlot} much more rapidly than would be expected
715: for a real neutron star that experiences this type of instability,
716: hence our model cannot be used directly to estimate the length of
717: time that such a source would spend near each frequency. However,
718: \cite{OL01} have outlined a method by which the detectability of a
719: source can be estimated from a knowledge of $\Delta J$, the total
720: angular momentum that is radiated away from the source via
721: gravitational radiation. Specifically, the signal-to-noise ratio
722: $S/N$ that could be achieved by optimal filtering
723: can be estimated from the expression,
724: \begin{equation}\label{noise}
725: \biggl(\frac{S}{N}\biggr)^2 \approx \frac{4G}{5m\pi c^3 r^2}
726: \frac{|\Delta J|} {f S_h (f)} \, ,
727: \end{equation}
728: where $m$ is the azimuthal quantum number ($m=2$ for the
729: bar-mode), $r$ is the distance to the source, and $S_h (f)$ is the
730: power spectral density of the detector noise at frequency $f.$
731: From our model ROT181 evolution, we find $\Delta J = 1.67 \times
732: 10^{48} \mathrm{g}~\mathrm{cm}^2 \mathrm{s}^{-1}$; and when it
733: reaches its design sensitivity, LIGO's $4~\mathrm{km}$
734: interferometer noise curve\footnote{The projected noise curve for
735: LIGO's $4~\mathrm{km}$ interferometers -- published as part of the
736: LIGO Science Requirements Document (SRD) -- and the actual noise
737: curve achieved by the $4~\mathrm{km}$ interferometer at the
738: LIGO-Hanford Observatory (LHO) during the S3 science run can be
739: obtained from
740: \url{www.ligo.caltech.edu/$\sim$lazz/distribution/LSC\_Data/}.}
741: should exhibit $\sqrt{S_h}\approx 3 \times 10^{-23}
742: ~\mathrm{Hz}^{-1/2}$ at $f=220 ~\mathrm{Hz}$, which is the
743: characteristic frequency of the spinning bar in model ROT181. From
744: expression (\ref{noise}), we therefore estimate that a source of
745: the type we are modelling will be detectable by LIGO with a
746: $S/N\gtrsim 8$ out to a distance of $2~\mathrm{Mpc}$. (During
747: LIGO's S3 science run in late 2003, the $4 ~\mathrm{km}$ LHO
748: interferometer had already come within a factor of two of this
749: design sensitivity.\footnotemark[1]) With Advanced LIGO (using
750: sapphire test masses, the projected noise curve\footnote{Projected
751: noise curves for the Advanced LIGO design using either sapphire or
752: silica test masses can be obtained from
753: \url{www.ligo.caltech.edu/advLIGO}.} gives $\sqrt{S_h}\approx 2.0
754: \times 10^{-24} ~\mathrm{Hz}^{-1/2}$ at $f=220 \mathrm{Hz}$) we
755: estimate that this type of source will be detectable with $S/N
756: \gtrsim 8$ out to $32~\mathrm{Mpc}$.
757:
758: Of course the detectability of gravitational waves generated by
759: the secular bar-mode instability will also depend on the frequency
760: with which such events occur nearby. To estimate an event rate we
761: can draw on the discussion of \cite{Kokkotas04} where an estimate
762: was made of the event rate of the dynamical bar-mode instability
763: in young neutron stars. Since the conditions required for the
764: onset of the secular bar-mode instability ($T/|W| \gtrsim 0.14$)
765: are almost as extreme as the conditions required for the onset of
766: the dynamical bar-mode instability ($T/|W| \gtrsim 0.27$), it
767: would be very surprising if the two event rates were not similar.
768: If we assume that only young neutron stars can be rotating rapidly
769: enough to be susceptible to either bar-mode instability, and if we
770: assume that a neutron star can form only from the collapse of the
771: core of massive star, then a reasonable upper limit on the rate of
772: these events will be given by the event rate of Type II
773: supernovae, that is, 1-2 per century per gas-rich galaxy
774: \citep{CET99}. (Another scenario is that rapidly rotating neutron
775: stars form from the accretion-induced collapse of white dwarfs.
776: But according to Liu 2002, the frequency of such events is orders
777: of magnitude lower than the event rate of Type II supernovae.)
778: Adopting a local galaxy density of $n_{\mathrm{g}}\approx 0.01~
779: \mathrm{Mpc}^{-3}$ \citep{KNST01}, we should expect $\lesssim 30$
780: Type II supernovae each year out to 32 Mpc. Now not all Type II
781: supernovae will produce neutron stars (Kokkotas 2004 estimates,
782: for example that 5-40\% of supernova events produce black holes
783: instead), and only a fraction $f_{\mathrm{rot}}$ of neutron stars
784: will be formed with sufficient rotational energy to be susceptible
785: to a bar-mode instability, so the predicted event rate should be
786: reduced accordingly. A naive estimation based on angular momentum
787: conservation during core collapse suggests that virtually all
788: newly born neutron stars will be formed rapidly rotating and,
789: therefore, $f_{\mathrm{rot}} \sim 1$; this is the direction
790: \cite{Kokkotas04} leans. But models of axisymmetric core collapse
791: \citep{Tohline84,DFM02a,DFM02b,OBLW04} indicate that the ratio of
792: energies $T/|W|$ in a newly formed neutron star is quite sensitive
793: to the equation of state of the core during its collapse and it is
794: easy to imagine physical scenarios in which appropriately rapidly
795: rotating neutron stars will rarely be formed; therefore,
796: $f_{\mathrm{rot}} \ll 1$. At the present time it is not clear
797: which picture is more correct, but adopting the more optimistic
798: view it should be possible for LIGO to detect on the order of ten
799: such events each year.
800:
801:
802: \subsubsection{Model Convergence}\label{revised}
803: In an effort to determine whether the Dedekind-like bar structure
804: was destroyed during the {\it late} phase of the ROT181 model
805: evolution as a result of physically realistic, hydrodynamical
806: processes, or by a radiation-reaction force that was artificially
807: too large, we set $\kappa=0$ then re-ran the last segment of the
808: simulation, starting from $t=11\tau_{\mathrm{spin}}$. This
809: ``revised'' evolution produced results that were qualitatively
810: identical to the late phase of the GRR-driven evolution. That is,
811: the bar was destroyed by the dynamical development of velocity and
812: density structure on a wide range of scales in all three
813: dimensions. In an effort to quantitatively describe this
814: relatively complex structure, Fig.~\ref{modeamplitudesLines} shows
815: a representation of the azimuthal Fourier-mode amplitudes of the
816: model's density distribution at two points in time:
817: $t=10\tau_{\mathrm{spin}}$, when the bar was well-developed; and
818: $t=20\tau_{\mathrm{spin}}$, after the higher-order nonaxisymmetric
819: structure was well-developed. (Note that the {\it late} phase of
820: this ``revised'' evolution was followed somewhat farther in time
821: than the original model ROT181 evolution described in \S4.2.1.) At
822: the earlier time, only the $m=2$ amplitude contained a significant
823: amount of power, and all odd amplitudes were smaller than their
824: even neighbors. At the later time the Fourier-mode amplitudes
825: appear to be related to one another by a simple power law,
826: indicating that power has been spread smoothly over all resolvable
827: length scales.
828:
829: \begin{figure}
830: \epsscale{.90} \plotone{f9.eps} \caption{A spectrum of the
831: Fourier-mode amplitude of the azimuthal density distribution is
832: shown at time $t=10\tau_{\mathrm{spin}}$ (filled circles), when
833: the bar was well-developed, and at time $t=20\tau_{\mathrm{spin}}$
834: (open circles), after the higher-order modes destroyed the
835: coherent bar in the ``revised'' evolution of model ROT181. To
836: guide the eye, amplitudes determined for various modes at the same
837: time are connected by straight line segments.
838: \label{modeamplitudesLines}}
839: \end{figure}
840:
841: As an additional test of the reliability of our results, we
842: repeated our rotating model evolution on a computational grid that
843: had a factor of two poorer resolution in each of the three spatial
844: dimensions; specifically, the new simulation was performed on a
845: grid with $66\times 64\times 66$ zones in $\varpi$, $\phi$, and
846: $z$, respectively. On this lower resolution grid, it was not
847: possible to begin the evolution from precisely the same initial
848: state as model ROT181. However, we were able to construct a
849: uniformly rotating, $n=1/2$ polytrope with $T/|W| = 0.179$ (only
850: $1\%$ less than the corresponding initial energy ratio of model
851: ROT181); we introduced a nonaxisymmetric density perturbation that
852: produced approximately the same initial mass-quadrupole moment
853: amplitude $|D_{22}|$ as in model ROT181; and throughout the
854: evolution the coefficient of the radiation-reaction force term in
855: Eq. \ref{motion} was set to $\kappa=1.75\times10^5$, as in model
856: ROT181. Hereafter, we will refer to this lower resolution
857: evolution as model ROT179.
858:
859: The key results of this lower resolution evolution are illustrated
860: by the curves in Figs.~\ref{rot181freq} and \ref{rot181d22} that
861: extend beyond $\tau_{\mathrm{spin}}= 35$. As shown in the bottom
862: frame of Fig.~\ref{rot181freq}, the real and imaginary components
863: of the eigenfrequency $\omega_{22}$ identified during model
864: ROT179's {\it early} evolution were nearly identical to the values
865: measured in model ROT181; as shown in the top of
866: Fig.~\ref{rot181freq}, $|D_{22}|$ grew exponentially up to a
867: nonlinear value, levelled off, then slowly decayed as the
868: bar-mode's pattern frequency slowed; and as shown in
869: Fig.~\ref{rot181d22} the strength of the GRR force grew steadily
870: up to a maximum value that was somewhat lower in amplitude and
871: somewhat delayed in time compared to model ROT181, then almost as
872: rapidly it dropped by an order of magnitude, as in model ROT181.
873: Finally, during the {\it late} phase of model ROT179's evolution,
874: the bar's coherent structure was destroyed by the development of
875: dynamical structure on much smaller scales, just as was observed
876: during the {\it late} phase of model ROT181's evolution. (This
877: phenomenon is evidenced in Fig.~\ref{rot181freq} by the rapid
878: oscillations in both components of $\omega_{22}$ at times
879: $t\gtrsim 32\tau_{\mathrm{spin}}$.) In this lower resolution
880: evolution, however, the small-scale dynamical structure took
881: roughly twice as long to develop as in model ROT181. The somewhat
882: slower initial growth of the bar-mode and the bar's lower peak
883: nonlinear amplitude can both be attributed directly to this
884: simulation's coarser spatial resolution. The delay in development
885: of the smaller scale structure was almost certainly due, in part,
886: to our inability to resolve structure on the smallest scales in
887: model ROT179, but the delay may also have occurred, in part,
888: because the bar, itself, was never as pronounced as in model
889: ROT181. Similar behavior has been observed in simulations that
890: have analyzed the long-term stability of r-mode oscillations in
891: young neutron stars \citep{GLSSF02}.
892:
893: \section{Summary and Conclusions}
894:
895: Using nonrelativistic, numerical hydrodyamical techniques coupled with
896: a post-Newtonian treatment of GRR forces, we have simulated the
897: nonlinear development of the secular bar-mode instability in a rapidly
898: rotating neutron star. In each simulation we have artificially
899: enhanced the strength of the GRR force term in the equation of motion
900: (by selecting values of the parameter $\kappa > 1$) in order to be
901: able to follow the secular development of the bar with a reasonable
902: amount of computing resources. In each case however, $\kappa$ was set
903: to a small enough value that the amplitude of the mass-quadrupole
904: moment changed slowly, compared to the dynamical time scale of the
905: system, thus ensuring that the system as a whole remained in dynamical
906: equilibrium. We first tested our simulation technique by studying the
907: evolution of the $\ell = m = 2$ bar-mode in a nonrotating neutron star
908: model (model SPH). The developing bar-mode exhibited an azimuthal
909: oscillation frequency within 3\% of the frequency predicted by linear
910: theory, and the amplitude of the bar-mode damped (as predicted) at a
911: rate that was within 15\% of the rate predicted by linear theory.
912:
913: Next, we evolved a rapidly rotating model (model ROT181), which
914: was predicted by linear theory to be unstable toward the growth of
915: the bar-mode. From the {\it early} ``linear-amplitude'' phase of
916: this model's evolution, we measured the bar-mode's azimuthal
917: oscillation frequency and its exponential growth rate; the values
918: are summarized in Table \ref{TabResults}. The oscillation
919: frequency $\langle\omega_{\mathrm{r}}\rangle/\Omega_0$ was almost
920: an order of magnitude smaller than in model SPH, and
921: $\langle\omega_{\mathrm{i}}\rangle/(\Omega_0\kappa)$ was four
922: orders of magnitude smaller than (and had the opposite sign of)
923: the value measured in model SPH. Both of these frequency values
924: reflect the fact that model ROT181 was rotating only slightly
925: faster than the marginally unstable model (predicted to have
926: $T/|W| \approx 0.14$), in which both components of $\omega_{22}$
927: should be precisely zero. We watched the unstable bar-mode grow
928: up to and saturate at a sufficiently large, nonlinear amplitude
929: that the bar-like distortion was clearly visible in two- and
930: three-dimensional plots of isodensity surfaces. This nonlinear
931: bar-like structure persisted for several rotation periods and,
932: during this {\it intermediate} phase of the ROT181 model evolution, we
933: tracked the frequency and amplitude of the gravitational radiation
934: that should be emitted from the configuration due to its
935: time-varying mass-quadrupole moment. Our model's evolution in a
936: ``strain-frequency'' diagram closely matches the evolutionary
937: trajectory predicted by \cite{LS95}, lending additional
938: credibility to their relatively simple (and inexpensive) way of
939: predicting the evolution of such systems as well as to our first
940: attempt to model such an evolution using nonlinear hydrodynamical
941: techniques. During the {\it late} phase of our model ROT181
942: evolution, the bar lost its coherent structure and the system
943: evolved to a much more complex nonaxisymmetric configuration. The
944: general features of this {\it late} phase of the evolution were
945: reproduced when the simulation was re-run on a coarser
946: computational grid, and even when the GRR forces were turned off.
947: So while the size and shape of the {\it intermediate} phase
948: ``Dedekind-like'' structure of our model may well have been
949: influenced strongly by the excessive strength of the GRR force
950: used in our simulation, it appears as though the final complex
951: ``turbulent'' phase of the evolution was governed by purely
952: hydrodynamical phenomena.
953:
954: It is not clear what physical mechanism was responsible for the
955: development of the small-scale structure and subsequent
956: destruction of the bar during the {\it late} phase of the
957: evolution of model ROT181. Because the bar's structure was
958: ``Dedekind-like'' -- that is, fluid inside the bar was moving
959: along elliptical streamlines with a mean frequency that was
960: significantly higher than the bar pattern frequency -- it is
961: tempting to suggest that the small-scale structure arose due to
962: differential shear. But, according to \cite{HBW99}, coriolis
963: forces are able to stabilize differentially rotating,
964: astrophysical flows against shearing instabilities even in
965: accretion disks where the shear is much stronger than in our
966: ``Dedekind-like'' bar. (See, however, Longaretti 2002 for an
967: opposing argument.) Furthermore, other models of differentially
968: rotating astrophysical bars \citep{CT00,NCT00} do not appear to be
969: susceptible to the dynamical instability that destroyed the bar in
970: our ROT181 model evolution. We suspect, instead, that the
971: late-time behavior of model ROT181 results either from nonlinear
972: coupling of various oscillatory modes within the star, or from an
973: ``elliptic flow'' instability similar to the one identified in
974: laboratory fluids that are forced to flow along elliptical
975: streamlines. The dissipative effect of mode-mode (actually,
976: three-mode) coupling has been examined in depth by \cite{SAFTW02}
977: and \cite{AFMSTW03} in the context of the r-mode instability in
978: young neutron stars, and \cite{BTW04} have shown the connection
979: between this process and the rapid decay of the r-mode in extended
980: numerical evolutions such as the ones performed by \cite{GLSSF02}.
981: However, this phenomenon has not yet been studied to the same
982: degree in relation to the $\ell=m=2$ f-mode. \cite{LL93},
983: \cite{LL96}, and \cite{LS99} have demonstrated that the ``elliptic
984: flow'' instability seen in laboratory fluids is likely to arise in
985: self-gravitating ellipsoidal figures of equilibrium, especially if
986: they have ``Dedekind-like'' internal flows. Additional analysis
987: and, very likely, additional nonlinear simulations will be
988: required before we are able to determine which (if either) of
989: these mechanisms was responsible for the destruction of the bar in
990: our ROT181 model evolution.
991:
992: Our nonlinear simulation of model ROT181 demonstrates that when a
993: rapidly rotating neutron star becomes unstable to the secular
994: bar-mode instability, the bar-like distortion may grow to
995: nonlinear amplitude and thereby become a strong source of
996: gravitational radiation. However, it will not be a long-lived
997: continuous-wave source, as one might optimistically have expected;
998: in our simulation, the nonlinear-amplitude bar survived fewer than
999: ten rotation periods. In a real neutron star the GRR forces will
1000: be much weaker than those of our simulation, so we expect the bar
1001: mode to grow and persist for many more rotation periods. However,
1002: we also expect the amplitude of the bar mode to saturate at a much
1003: lower amplitude in a real neutron star. Nevertheless, we expect
1004: the bar mode to persist in rapidly rotating neutron stars long
1005: enough to allow gravitational radiation to remove sufficient
1006: angular momentum for them to relax into a secularly stable
1007: equilibrium state. Thus the amount of angular momentum radiated
1008: away in real neutron stars should be comparable to that in our
1009: simulation. While such astrophysical systems may not be the
1010: easiest sources to detect with broadband, gravitational-wave
1011: detectors such as LIGO because the frequency of the emitted
1012: radiation will change steadily with time, our estimates suggest
1013: that gravitational waves arising from the excited secular bar-mode
1014: instability in rapidly rotating neutron stars could well be
1015: detectable in the not too distant future from neutron stars as far
1016: away as $32~\mathrm{Mpc}$.
1017:
1018: After submitting this paper
1019: for publication, we became aware that \cite{SK04} have just
1020: completed an investigation similar to the one presented here in
1021: which they have utilized post-Newtonian simulations to study the
1022: nonlinear development of the secular bar-mode instability in
1023: rapidly rotating neutron stars. Their initial models were
1024: differentially rotating, $n=1$ ($\Gamma=2$) polytropes with $0.2
1025: \lesssim T/|W| \lesssim 0.26$. The {\it early} and {\it
1026: intermediate} phases of their model evolutions agree well with the
1027: results of our model ROT181 evolution, that is, the bar-mode grew
1028: exponentially at rates consistent with the predictions of linear
1029: theory and reached a nonlinear amplitude, producing an ellipsoidal
1030: star of moderately large ellipticity.
1031: The strength of the GRR force used in our simulations was considerably
1032: larger than theirs. This may explain why the bar mode grows to a
1033: larger amplitude and why, in turn, there is a more significant decrease
1034: in the pattern frequency of the bar as it evolves toward a Dedekind-like
1035: configuration in our simulation. This may also explain why the bar
1036: mode structure was ultimately destroyed by short wavelength disturbances
1037: in our evolutions while such turbulence had not yet developed in theirs.
1038:
1039:
1040:
1041: %However, they did not
1042: %follow their evolutions far enough in time to observe a
1043: %significant decrease in the pattern frequency of the bar toward a
1044: %Dedekind-like configuration or to examine the late-time stability
1045: %of the bar against short wavelength disturbances.
1046:
1047: %% If you wish to include an acknowledgments section in your paper,
1048: %% separate it off from the body of the text using the \acknowledgments
1049: %% command.
1050:
1051: %% Included in this acknowledgments section are examples of the
1052: %% AASTeX hypertext markup commands. Use \url without the optional [HREF]
1053: %% argument when you want to print the url directly in the text. Otherwise,
1054: %% use either \url or \anchor, with the HREF as the first argument and the
1055: %% text to be printed in the second.
1056:
1057: \acknowledgments
1058:
1059: We thank Gabriela Gonz\'alez, Peter Saulson, Peter Fritschel and
1060: Kip Thorne for guidance in obtaining the LIGO noise figures used
1061: in our analysis. We also thank an anonymous referee for recommending
1062: several ways in which the presentation of our results could be improved.
1063: This work was supported in part by NSF grants
1064: AST-9987344, AST-0407070, PHY-0326311 and NASA grant NAG5-13430 at
1065: LSU; and by NSF grants PHY-0099568, PHY-0244906 and NASA grants
1066: NAG5-10707, NAG5-12834 at Caltech. Most of the simulations were
1067: carried out on SuperMike and SuperHelix at LSU, which are
1068: facilities operated by the Center for Computation and Technology
1069: whose funding largely comes through appropriations by the
1070: Louisiana state legislature.
1071:
1072: %% The reference list follows the main body and any appendices.
1073: %% Use LaTeX's thebibliography environment to mark up your reference list.
1074: %% Note \begin{thebibliography} is followed by an empty set of
1075: %% curly braces. If you forget this, LaTeX will generate the error
1076: %% "Perhaps a missing \item?".
1077: %%
1078: %% thebibliography produces citations in the text using \bibitem-\cite
1079: %% cross-referencing. Each reference is preceded by a
1080: %% \bibitem command that defines in curly braces the KEY that corresponds
1081: %% to the KEY in the \cite commands (see the first section above).
1082: %% Make sure that you provide a unique KEY for every \bibitem or else the
1083: %% paper will not LaTeX. The square brackets should contain
1084: %% the citation text that LaTeX will insert in
1085: %% place of the \cite commands.
1086:
1087: %% We have used macros to produce journal name abbreviations.
1088: %% AASTeX provides a number of these for the more frequently-cited journals.
1089: %% See the Author Guide for a list of them.
1090:
1091: %% Note that the style of the \bibitem labels (in []) is slightly
1092: %% different from previous examples. The natbib system solves a host
1093: %% of citation expression problems, but it is necessary to clearly
1094: %% delimit the year from the author name used in the citation.
1095: %% See the natbib documentation for more details and options.
1096:
1097: \begin{thebibliography}{}
1098: \bibitem[Abbott et al.(2004)]{S1burst}Abbott, B. et al. (LIGO Scientific Collaboration) 2004, \prd, 69, 102001(21)
1099: \bibitem[Abramovici et al.(1992)]{LIGO} Abramovici, A. et al. 1992, Science, 256, 325
1100: \bibitem[Acernese et al.(2002)]{VIRGO}Acernese, F. et al. 2002, Class. Quant. Grav., 19, 1421
1101: \bibitem[Andersson(2003)]{A03} Andersson, N. 2003, Class. Quant. Grav., 20, 105
1102: \bibitem[Arras et al.(2003)]{AFMSTW03} Arras, P., Flanagan, \'E. \'E., Morsink, S.M., Schenk, A.K., Teukolsky, S. A., \& Wasserman, I. 2003, \apj, 591,
1103: 1129
1104: \bibitem[Brink, Teukolsky, \& Wasserman(2004)]{BTW04}Brink, J., Teukolsky, S. A., \& Wasserman, I. 2004,
1105: gr-qc/0406085
1106: \bibitem[Brown(2000)]{B00}Brown, J. D. 2000, \prd, 62, 084024
1107: \bibitem[Cappellaro, Evans, \& Turatto(1999)]{CET99}Cappellaro, E., Evans, R., \& Turatto, M. 1999, \aa,
1108: 351, 459
1109: \bibitem[Cazes \& Tohline(2000)]{CT00}Cazes, J. E., \& Tohline, J. E. 2000, \apj, 532, 1051
1110: \bibitem[Centrella et al.(2001)]{CNLB01}Centrella, J.M., New, K.C.B., Lowe, L. L., \& Brown, J. D. 2001, \apj, 550, L193
1111: \bibitem[Chandrasekhar(1969)]{Ch69}Chandrasekhar, S. 1969,
1112: Equilibrium Figures of Equilibrium, New Haven, CT: Yale Univ.
1113: Press
1114:
1115: \bibitem[Chandrasekhar(1970)]{Ch70} Chandrasekhar, S. 1970, \apj,
1116: 161, 561
1117: \bibitem[Comins(1979a)]{C79a}Comins, N. 1979a, \mnras, 189, 233
1118: \bibitem[Comins(1979b)]{C79b}Comins, N. 1979b, \mnras, 189, 25
1119: \bibitem[Cutler(1991)]{C91}Cutler, C. 1991, \apj, 374, 248
1120: \bibitem[Cutler \& Lindblom(1987)]{CL87} Cutler, C., \& Lindblom,
1121: L. 1987, \apj, 314, 234
1122: \bibitem[Cutler \& Lindblom(1992)]{CL92} Cutler, C., \& Lindblom,
1123: L. 1992, \apj, 385, 630
1124: \bibitem[Detweiler \& Lindblom(1977)]{DL77} Detweiler, S. L., \& Lindblom,
1125: L. 1977, \apj, 213, 193
1126: \bibitem[Di Girolamo \& Vietri(2002)]{DiGV02} Di Girolamo, T., \&
1127: Vietri, M. 2002, \apj, 581, 519
1128: \bibitem[Dimmelmeier, Font, \& M\"{u}ller(2002a)]{DFM02a}Dimmelmeier, H., Font, J., \& M\"{u}ller, E. 2002, \aa,
1129: 388, 917
1130: \bibitem[Dimmelmeier, Font, \& M\"{u}ller(2002b)]{DFM02b}Dimmelmeier, H., Font, J., \& M\"{u}ller, E. 2002, \aa,
1131: 393, 523
1132: \bibitem[Durisen et al.(1986)]{DGTB86} Durisen, R. H., Gingold, R. A., Tohline, J. E., \& Boss, A. P.
1133: 1986, \apj, 305, 281
1134: \bibitem[Finn \& Evans(1990)]{FE90}Finn, L. S., \& Evans, C. R. 1990, \apj, 351, 588
1135:
1136: \bibitem[Flowers \& Itoh(1976)]{FI76} Flowers, E., \& Itoh, N.
1137: 1976, \apj, 206, 218
1138: \bibitem[Friedman(1978)]{F78} Friedman, J. 1978, Comm. Math. Phys., 62, 247
1139: \bibitem[Friedman \& Schutz(1978)]{FS78} Friedman, J., \& Schutz,
1140: B. F. 1978, \apj, 222, 281
1141: %%\bibitem[Gonz\'alez(2004)]{G04}Gonz\'alez, G. 2004, ``Status of LIGO
1142: %%Data Analysis,'' in Proceedings of the $8^{\mathrm{th}}$
1143: %%Gravitational Wave Data Analysis Worshop, Class. Quant. Grav.,
1144: %%submitted
1145: \bibitem[Gossler et al.(2002)]{GEOb}Gossler, S., et al. 2002, Class. Quant. Grav., 19, 1835
1146: \bibitem[Gressman et al.(2002)]{GLSSF02}Gressman, P., Lin, L.-M., Suen, W.-M., Stergioulas,
1147: N., \& Friedman, J. L. 2002, \prd, 66, 041303
1148: \bibitem[Hachisu(1986)]{H86} Hachisu, I. 1986, \apjs, 61, 479
1149: \bibitem[Hawley, Balbus \& Winters(1999)]{HBW99} Hawley, J. F.,
1150: Balbus, S. A., \& Winters, W. F. 1999, \apj, 518, 394
1151: \bibitem[Imamura, Friedman, \& Durisen(1985)]{IFD85} Imamura, J.
1152: N., Friedman, J. L., \& Durisen, R. H. 1985, \apj, 294, 474
1153: \bibitem[Ipser \& Lindblom(1990)]{IL90} Ipser, J. R., \&
1154: Lindblom, L. 1990, \apj, 355, 226
1155: \bibitem[Ipser \& Lindblom(1991)]{IL91} Ipser, J. R., \&
1156: Lindblom, L. 1991, \apj, 373, 213
1157: \bibitem[Jones(1971)]{J71} Jones, P. B. 1971, Proc. Roy. Soc. London A,
1158: 323, 111
1159: \bibitem[Kalogera et al.(2001)]{KNST01}Kalogera, V., Narayan, R., Spergel, D. N., \& Taylor,
1160: J. H. 2001, \apj, 556, 340
1161: \bibitem[Karino \& Eriguchi(2003)]{KE03}Karino, S., \& Eriguchi, Y. 2003, \apj, 592, 1119
1162: \bibitem[Kazanas \& Schramm(1977)]{KS77} Kazanas, D., \& Schramm 1977,
1163: \apj, 214, 819
1164: \bibitem[Kokkotas(2004)]{Kokkotas04}Kokkotas, K. D. 2004, Class. Quant. Gravity, 21, 501.
1165: \bibitem[Lai \& Shapiro(1995)]{LS95} Lai, D., \& Shapiro, S. L.
1166: 1995, \apj, 442, 259
1167:
1168: \bibitem[Lebovitz \& Lifschitz(1996)]{LL96}Lebovitz, N. R., \& Lifschitz, A. 1996, \apj, 458, 699
1169: %%"New Global Instabilities of the Riemann Ellipsoids"
1170:
1171: \bibitem[Lebovitz \& Saldanha(1999)]{LS99}Lebovitz, N. R., \& Saldanha, K. I. 1999, Physics of Fluids,
1172: 11, 3374
1173: %%"On the weakly nonlinear development of the elliptic instability"
1174:
1175: \bibitem[Lifschitz \& Lebovitz(1993)]{LL93}Lifschitz, A., \& Lebovitz, N. 1993, \apj, 408, 603
1176: %%"Local hydrodynamic instability of rotating stars"
1177:
1178: \bibitem[Lindblom(1997)]{L97} Lindblom, L. 1997, in General
1179: Relativity and Gravitation, edited by M. Francaviglia, G. Longhi, L. Lusanna,
1180: E. Sorace, World Scientific, pp. 237--258.
1181: \bibitem[Lindblom(2001)]{L01} Lindblom, L. 2001, in Gravitational Waves:
1182: A Challange to Theoretical Astrophysics, edited by V. Ferrari, J. C. Miller,
1183: L. Rezzolla, ICTP Lecture Notes Series, pp. 257--275.
1184: \bibitem[Lindblom \& Detweiler(1977)]{LD77} Lindblom, L., \&
1185: Detweiler, S. 1977, \apj, 211, 565
1186: \bibitem[Lindblom \& Detweiler(1979)]{LD79} Lindblom, L., \& Detweiler, S.
1187: 1979, \apj, 232, L101
1188: \bibitem[Lindblom \& Hiscock(1983)]{LH83} Lindblom, L., \&
1189: Hiscock, W. A. 1983, \apj, 267, 384
1190: \bibitem[Lindblom \& Mendell(1995)]{LM95} Lindblom, L., \&
1191: Mendell, G. 1995, \apj, 444, 804
1192: \bibitem[Lindblom et al.(2001)]{LTV01} Lindblom, L., Tohline, J.
1193: E., \& Vallisneri, M. 2002, \prl, 86, 1152
1194: \bibitem[Lindblom et al.(2002)]{LTV02} Lindblom, L., Tohline, J.
1195: E., \& Vallisneri, M. 2002, \prd, 65, 084039
1196: \bibitem[Liu(2002)]{Liu02}Liu, Y. T. 2002, \prd, 65, 124003
1197: \bibitem[Longaretti(2002)]{Long02} Longaretti, P.-Y. 2002, \apj,
1198: 576, 587
1199: \bibitem[Managan(1985)]{M85} Managan, R. A. 1985, \apj, 294, 463
1200: %%\bibitem[Miller(1974)]{M74} Miller, B. D. 1974, \apj, 187, 609
1201: \bibitem[Motl, Tohline, \& Frank(2002)]{MTF02} Motl, P. M.,
1202: Tohline, J. E., \& Frank, J. 2002, \apjs, 138, 121
1203: \bibitem[New, Centrella \& Tohline(2000)]{NCT00}New, K. C. B., Centrella, J. M., \& Tohline, J. E.
1204: 2000, \prd, 62, 064019
1205: \bibitem[Ostriker \& Bodenheimer(1973)]{OB73}Ostriker, J.P., \& Bodenheimer, P. 1973, \apj, 180,
1206: 171
1207: \bibitem[Ott et al.(2004)]{OBLW04}Ott, C. D., Burrows, A., Livne, E., \& Walder, R.
1208: 2004, \apj, 600, 834
1209: \bibitem[Owen \& Lindblom(2002)]{OL01}Owen, B., Lindblom, L. 2002,
1210: Class. Quant. Grav., 19, 1247
1211: \bibitem[Pickett et al.(1998)]{PCDL98}Pickett, B. K., Cassen, P., Durisen, R. H., \& Link, R.
1212: 1998, \apj, 504, 468
1213: \bibitem[Sawyer(1989)]{S89} Sawyer, R. F. 1989, \prd, 39, 3804
1214: \bibitem[Schenk et al.(2002)]{SAFTW02}Schenk, A. K., Arras, P., Flanagan, \'E. \'E.,
1215: Teukolsky, S. A., \& Wasserman, I. 2002, \prd, 65, 024001
1216: \bibitem[Shapiro \& Zane(1998)]{SZ98} Shapiro, S. L., \& Zane, S.
1217: 1998, \apjs, 117, 531
1218: \bibitem[Shibata \& Karino(2004)]{SK04}Shibata, M., \& Karino, S. 2004, \prd, in press
1219: (astro-ph/040816)
1220: \bibitem[Shibata, Karino, \& Eriguchi(2002)]{SKE02}Shibata, M., Karino, S., \& Eriguchi, Y. 2002, \mnras, 334,L27
1221: \bibitem[Shibata, Karino, \& Eriguchi(2003)]{SKE03}Shibata, M., Karino, S., \& Eriguchi, Y. 2003, \mnras, 343, 619
1222: \bibitem[Stergioulas(2003)]{S03}Stergioulas, N. 2003, Living Rev. Relativity, 6, 3
1223: [Online article]: cited on 26 April 2004
1224: http://www.livingreviews.org/lrr-2003-3/
1225: \bibitem[Stergioulas \& Friedman(1998)]{SF98} Stergioulas, N., \&
1226: Friedman, J. L. 1998, \apj, 492, 301
1227: \bibitem[Tassoul(1978)]{T78}Tassoul, J.-L. 1978, Theory of Rotating Stars,
1228: Princeton: Princeton University Press
1229: \bibitem[Tagoshi et al.(2001)]{TAMA}Tagoshi, H. et al. (TAMA300 collaboration) 2001, \prd, 63, 062001
1230: \bibitem[Thompson \& Duncan(1993)]{TD93} Thompson, C., \& Duncan,
1231: R. C. 1993, \apj, 408, 194
1232: \bibitem[Tohline(1984)]{Tohline84}Tohline, J. E. 1984, \apj, 285, 721
1233: \bibitem[Tohline, Durisen, \& McCollough(1985)]{TDM85}Tohline, J. E., Durisen, R. H., \& McCollough, M.
1234: 1985, \apj, 298, 220
1235: \bibitem[Watts, Andersson \& Jones(2004)]{WAJ04}Watts, A. L., Andersson, N., \& Jones, D. I. 2004,
1236: astro-ph/0309554
1237: \bibitem[Williams \& Tohline(1988)]{WT88}Williams, H. A., \& Tohline, J. E. 1988, \apj, 334,
1238: 449
1239: \bibitem[Willke et al.(2002)]{GEOa}Willke, B. et al. 2002, Class. Quant. Grav., 19, 1377
1240: \bibitem[Yoshida \& Eriguchi(1995)]{YE95} Yoshida, S., \& Eriguchi, Y. 1995, \apj, 438, 830
1241: \end{thebibliography}
1242:
1243: \clearpage
1244:
1245:
1246: %% The following command ends your manuscript. LaTeX will ignore any text
1247: %% that appears after it.
1248:
1249: \end{document}
1250:
1251: %%
1252: %% End of file `sample.tex'.
1253: