1: \documentclass[12pt,preprint]{aastex}
2:
3: \slugcomment{Draft as of \today}
4:
5:
6: \shorttitle{Gravitational Waves from 1987A}
7: \shortauthors{Santostasi, Johnson, Frank}
8:
9: \newcommand{\etal}{{\it et al.~}}
10: \newcommand{\bfdelta}{\mbox{\boldmath{$\delta$}}}
11: \newcommand{\bfomega}{\mbox{\boldmath{$\omega$}}}
12:
13: \def\spose#1{\hbox to 0pt{#1\hss}}
14:
15: \def\lta{\mathrel{\spose{\lower 3pt\hbox{$\mathchar"218$}}
16: \raise 2.0pt\hbox{$\mathchar"13C$}}}
17: \def\gta{\mathrel{\spose{\lower 3pt\hbox{$\mathchar"218$}}
18: \raise 2.0pt\hbox{$\mathchar"13E$}}}
19:
20: \received{2002 March 14}
21: \begin{document}
22:
23: \title{ Detectability of Gravitational Waves from SN 1987A }
24:
25:
26: \author{Giovanni Santostasi, Warren Johnson \and Juhan Frank}
27: \affil{Department of Physics and Astronomy, Louisiana State University,
28: Baton Rouge, LA 70803-4001}
29:
30: \begin{abstract}
31: We discuss the potential for detection of gravitational waves from
32: a rapidly spinning neutron star produced by supernova 1987A
33: taking the parameters claimed by \citet{mid00} at face
34: value. Assuming that the dominant mechanism for spin down is
35: gravitational waves emitted by a freely precessing neutron star,
36: it is possible to constrain the wobble angle, the effective moment of
37: inertia of the precessing crust and the crust cracking stress limit.
38: Our analysis suggests that, if the interpretation of the Middleditch
39: data is correct, the compact remnant of SN 1987A may well
40: provide a predictable source of gravitational waves
41: well within the capabilities of LIGO II.
42: The computational task
43: required for the data analysis is within the capabilities of
44: current computers if performed offline and could be accomplished online
45: using techniques such as demodulation and decimation.
46:
47: \end{abstract}
48: \keywords{stars: neutron --- pulsars: general --- gravitational waves --- supernovae: individual (SN1987A)}
49:
50:
51: %\keywords{stars: neutron---pulsars: general---gravitational waves}
52:
53:
54: \section{Introduction}
55: \label{intro}
56:
57: \citet{mid00} have claimed the likely first detection of
58: the compact remnant of supernova 1987A (hereafter SN 1987A).
59: Through fast photometry of a small region around the supernova they
60: were able to find a modulated signal with a main frequency of 467.5 Hz and a
61: modulation period of about 1000 seconds. Assuming that this was the
62: spin frequency of the presumed pulsar, and following the source between
63: 1992 and 1996, they were able to determine the
64: spindown of the pulsar and changes in the precession period. The observations
65: were complicated by times when the pulsation or the modulation were
66: not visible
67: or not so evident. The pulsations seem to have disappeared completely since
68: 1996. While astrophysically plausible explanations for the intermittency
69: of the signal can be devised appealing to the very complex nature of the
70: SN 1987A environment, the reality of a pulsar with the described
71: characteristics is at best very suggestive.
72:
73: In this paper we will simply assume that the pulsar interpretation is
74: correct, adopt the parameters derived by Middleditch \etal
75: at face value and derive some interesting implications for the detection
76: of gravitational waves from this source. We base our discussion
77: on simple free precessing neutron star models \citep{alp85, cut00, jan01}.
78:
79: The general problem of emission of gravitational waves from rotating
80: and precessing neutron stars including pulsars and low-mass X-ray binaries
81: has been recently reviewed by \citet{jon01}. For the particular case of
82: SN 1987A, while other authors \citep{cut00, jan01, nag01}
83: have also examined some of the consequences of
84: the results of \citet{mid00}, none of these papers
85: provides a precise calculation of the intensity and detectability of
86: gravitational waves from this source. The main aim of this paper is
87: to provide these estimates and to discuss the likelihood of detection
88: of gravitational waves from the hypoyhetical pulsar in SN 1987A
89: by LIGO I and II. In Section \ref{strength} we estimate
90: the time required to
91: observe the signal with different types of
92: detectors using coherent integration techniques.
93: We show that, within a plausible range of
94: values of the moment of inertia $I_{0},$
95: the gravitational wave strain is big
96: enough to be detectable by LIGO II within integration times
97: ranging from days to months. Thus, if the interpretation of the
98: periodicities in the optical observations is correct, 1987A
99: {\em should} be a predictable source of gravity waves
100: for ground based observatories. The computational requirements
101: for the data analysis discussed in Section \ref{templates}
102: are non-trivial but within the capabilities of modern computers.
103:
104:
105:
106:
107: \section{Summary of the Observations}
108: \label{observations}
109:
110: \citet{mid00} discuss fast photometry observations
111: of the remnant of the
112: supernova 1987A carried out at different times over the
113: period 1992--1996 from several observatories. During that time interval the
114: pulsar was detected several times at slightly different frequencies.
115: The power in the signal faded since 1993 and was last detected in February
116: 1996.
117: While they found ``no clear
118: evidence of any pulsar of constant intensity and stable timing,'' they did
119: find ``emission with a complex period modulation near the
120: frequency of 467.5 Hz - a 2.14 ms pulsar candidate''. They also point out
121: that: ``the frequency of the signals followed a consistent and
122: predictable spin-down ( $\sim $2-3 x 10$^{-10}$ Hz/s) over the several year
123: timespan. They find evidence for ``modulation of the 2.14 ms period with a
124: $\sim $1,000 s period which complicates its detection.''
125:
126: The observed modulation of the 2.14 ms period can be interpreted as the effect
127: of precession due to some deformation or crustal density
128: distribution which is not symmetric about the axis of rotation, including the case
129: in which the
130: precesing object itself possesses axial symmetry about a body axis which is not
131: aligned with the axis of rotation. In the absence of any external torques,
132: this situation is termed ``free precession". Classical mechanics tells us that
133: the ratio between the precession frequency and the rotation frequency is
134: proportional to the size of the deformation (e.g. \citet{mar95}). The size
135: of the deformation and the frequency of rotation determine the rate of spin-down if
136: the neutron star is assumed to lose energy mainly due to
137: gravitational radiation.
138:
139: A freely precessing neutron
140: star emits gravitational waves (e.g. \citet{zim79},
141: \citet{zim80}).
142: Using the general relativistic energy loss
143: equation and the classical mechanics relationship between ellipticity, rotation
144: and precession frequency, we have that the spin down rate is proportional to the
145: square of the precession frequency under the assumption that all the energy is
146: lost due to gravitational back reaction. If an electromagnetic contribution to the
147: spin down rate is also present, this term would be independent of ellipticity
148: and would be approximately constant during the time span of the observations.
149: The data shown on Fig. 9 of \citet{mid00} are consistent with a linear
150: correlation between spin down rate and the square of the precession frequency
151: going straight through the origin, i.e. with
152: zero contribution from magnetic dipole emission. Thus \citet{mid00}
153: conclude that the characteristics of the
154: 2.14 ms signature and its
155: $\sim $1,000 s modulation are consistent with precession and spindown via
156: gravitational radiation of a neutron star with effective non-axisymmetric
157: oblateness of $\sim $10$^{-6}$. We re-examine some aspects of this problem in
158: Section \ref{model}.
159:
160:
161: \section{A Model for the Precessing Neutron Star}
162: \label{model}
163:
164: \subsection{System Geometry}
165: \label{geometry}
166: Rotating neutron stars are often mentioned
167: as a possible continuous source of gravitational radiation. Usually what is
168: envisioned is that the star has an axissymetric deformation perpendicular to the
169: axis of rotation to allow for a changing mass quadrupole that will generate
170: gravity waves. Such a prolate or oblate star, tumbling about an axis
171: perpendicular to the axis of symmetry, will emit gravity waves
172: at twice the rotation frequency.
173:
174: If the star is deformed on a axis that is
175: at any other angle with the rotation axis then it will precess as a spinning
176: top, and will emit at both twice the rotation frequency and at the rotation frequency.
177: The
178: simplest situation is that the star is a rigid body and has just two non-equal
179: principal moments of inertia. We have then $I_{1}=I_{2}=I_{0}-\Delta I_{ \rm d}/3,$
180: $I_{3}=I_{0}+2/3\Delta I_{\rm d}$, so that $\Delta I_{\rm d} = I_{3}-I_{1}$.
181: $I_{0}$ is the average value of \ the moment of inertia and $\Delta I_{\rm d}\ll I_0$.
182: A more complete and realistic model is
183: considered further below but the simplest case remains the basis for the discussion of
184: precessing neutron stars. The main equations are the same even in the more
185: realistic case with minor modifications. Figure 1 shows our convention in the
186: orientation of the important vectorial quantities involved in the
187: problem.
188:
189: \clearpage
190:
191: \newcommand{\mbf}{\mathbf}
192: \begin{figure}
193: \plotone{f1.eps}
194: \caption{The reference plane for a
195: freely precessing body. This diagram shows the respective orientation of the axis of
196: deformation ${\mbf n}_{\rm d}$, angular momentum ${\mbf J}$, and axis of rotation
197: ${\mbf \omega}$ of the star. The projection of the instantaneous angular velocity
198: vector ${\mbf \omega}$ on to the symmetry axis ${\mbf n}_{\rm d}$ is indicated by
199: ${\mbf \omega}_3$.
200: }
201: \label{fig1}
202: \end{figure}
203:
204: \clearpage
205:
206: We can
207: define the total moment of inertia as:
208: \begin{equation}
209: {\mbf I}=I_{0}\bfdelta + \Delta I_{\rm d}\left(
210: {\mbf n}_{\rm d}{\mbf n}_{\rm d}- \bfdelta/3\right)
211: \label{itensor1}
212: \end{equation}
213: where
214: ${\mbf n}_{\rm d}$ is a unit vector pointing along the body symmetry axis and
215: $\bfdelta$ is the unit tensor. Now define the ``ellipticity"
216: as a small quantity $\epsilon =\Delta
217: I_{\rm d}/I_{0}$, then classical mechanics implies
218: \begin{equation}
219: \epsilon =\frac{\Omega _{\rm p}}{\omega
220: _{3}}=\frac{\Omega _{\rm p}}{\omega \cos \gamma }\approx
221: \frac{\Omega _{\rm p}}{\omega\cos \theta }
222: \label{eps}
223: \end{equation}
224: where $\theta $ is the angle
225: between the total angular momentum and the vector ${\mbf n}_{\rm d}$, $\gamma
226: =\theta +\widehat{\theta }\approx \theta .$ The angle between the rotation axis
227: and the angular momentum is a small quantity of order $\Delta I_{\rm d}/I_0.$ The
228: quantity $\Omega _{\rm p}$ is the precession frequency and $\omega $ is the rotation
229: frequency and $\omega _{3} $ its projection along the 3-axis that coincides in
230: this case with the axis ${\mbf n}_{\rm d}.$ We will proceed from the
231: assumption that we know the parameters \ $\Omega _{\rm p}$ and $\omega$
232: from the observations of SN 1987A by \citet{mid00}.
233:
234: The observed modulation or precession period varied during the span of the
235: observations in the range from approximately 935 s to 1430 s, while $\omega$
236: or the spin period varied measurably but relatively little. Consequently the
237: observed variations in $\Omega _{\rm p}$ must be attributed to variations in
238: $\epsilon$ or $\theta$ or both. Note, however, that the correlation between
239: $\dot\omega$ and $\Omega _{\rm p}$ claimed by Middleditch \etal requires
240: that $\theta$ remain constant. \citet{jan01} and \citet{jon01}
241: have claimed that it is not easy to imagine how significant variations in
242: ellipticity can occur without affecting the wobble angle. We shall return to
243: this question in Section \ref{constwobble} and argue that it is in fact
244: unlikely that variations in epsilon can significantly change the wobble angle.
245:
246:
247: \subsection{Gravitational radiation caused
248: by misalignment}
249: \label{misalign}
250: To determine the size of deformation and
251: consequently the strain carried by the gravitational radiation on earth we need to
252: evaluate the wobble angle $\theta$.
253: This can be done assuming that the star
254: is losing energy solely through gravitational radiation.
255: Then we can use the general
256: relativistic equation for the rate of energy emission by gravitational waves
257: \citep{zim79,zim80}:
258: \begin{equation}
259: \dot{E}=-\frac{2}{5}\frac{G}{c^{5}}\epsilon
260: ^{2}I_{0}^{2}\omega ^{6}\sin ^{2}\theta \left( 16\sin ^{2}\theta +\cos
261: ^{2}\theta \right),
262: \label{dote}
263: \end{equation}
264: where the first and second terms in parenthesis represent the contributions
265: at $2\omega$ and $\omega$ respectively.
266:
267: If the only source of energy for this emission is the neutron
268: star's rotational energy reservoir $E=1/2$ $I\omega ^{2},$ we have then
269: $\dot{E}=\omega \dot{\omega I},$ so that
270: \begin{equation}
271: \dot{\omega
272: }=\frac{2}{5}\frac{G}{c^{5}}\epsilon ^{2}I_{0}\ \omega ^{5}\sin ^{2}\theta
273: \left( 16\sin ^{2}\theta +\cos ^{2}\theta \right).
274: \label{dotomega}
275: \end{equation}
276: Since the change in angular velocity
277: $\dot{\omega }$ is known from observations, it is possible to
278: solve equation (\ref{dotomega}) for the ellipticity as a function of the wobble
279: angle for any given $I_0$. Figure
280: 2 shows the relationship between the ellipticity $\epsilon $ and
281: the wobble angle $\theta $ derived from equation (\ref{eps}) for
282: the observed range of
283: precession periods between 935 s and 1430 s (monotonically increasing curves).
284: Figure 2 also shows the result of solving
285: (\ref{dotomega})
286: (monotonically decreasing curves),
287: for the range of observed values of $2\times 10^{-10} < \dot{\nu}< 3\times 10^{-10}$ Hz/s
288: for an arbitrarily chosen representative value of $I_0=10^{44}$ g cm$^2$ .
289:
290: \clearpage
291:
292: \begin{figure}
293: \plotone{f2.eps}
294: \caption{
295: The general relativistic (dashed) and classical (solid)
296: relationships between $\epsilon$ and $\theta$ given the ranges of
297: observed values of spindown rate and precession frequency. The curves shown
298: correspond to
299: $I_0=10^{44}$ g cm$^2$, a value intermediate between
300: the minimum (just the crust precesses) and the
301: maximum (all the star is involved in the
302: precession).
303: The possible solutions for the adopted value of the moment of inertia
304: lie in the shaded region.
305: }
306: \label{fig2}
307: \end{figure}
308:
309: \clearpage
310:
311: The relativistic equation (\ref{dotomega}) depends on
312: the value $I_{0}$. This value is the average moment of the inertia of the part
313: of the star that actually participates in the precession. If the star has a crust
314: and a liquid interior then $I_{0}$ is the crust's moment of inertia and that of
315: any liquid coupled to the crust. In fact, part of the liquid should be stress
316: free and not influenced by the precession. So we can take $I_{0}$ to be an arbitrary
317: quantity equal or less than entire moment of inertia of the star
318: $I_{\rm star}=\frac{2}{5}MR^{2}=1.12\times 10^{45}\, {\rm g\ cm}^{2}M_{1.4}R_{6}^{2}$,
319: where $M_{1.4}$ is the mass of the star in units of 1.4 solar masses and $R_{6}$ the
320: radius in units of 10$^{6}$ cm. If just the crust participates in the precession
321: then $I_{0}$ $\approx 1/100$ $I_{\rm star}$ according standard neutron star theory.
322: Now, the classical mechanics equation (\ref{eps}) and the
323: relativistic equation (\ref{dotomega}) have to be satisfied at the same time.
324: This means
325: that for given observed $\Omega _{\rm p},\omega $, $\dot{\omega }$ and choice
326: of $I_{0}$ the functions have to meet at a point in the parameter space
327: $\epsilon -\theta .$ If we consider the moment of inertia the unknown parameter
328: of our problem we can determine which wobble angle the star should have
329: according the value of $I_{0}$. This is illustrated in Figure 3.
330: It seems that the 1987A remnant had some
331: relatively big and rapid changes in precession frequency during the first years of
332: observation. The astrophysical explanation for this could be a very active
333: dynamic environment in the young neutron star, that can bring abrupt changes in
334: the density of the crust, fractures and re-arrangement of surrounding material. We
335: already mentioned that Middleditch \etal find a power two relationship
336: between the observed change in $\dot{\omega }$ and $\Omega _{\rm p}.$ This
337: relationship holds exactly if we substitute
338: equation (\ref{eps}) into (\ref{dotomega}), namely:
339: \begin{equation}
340: \dot\omega
341: =\frac{2}{5}\frac{G}{c^{5}}\frac{\Omega _{\rm p}^{2}}{\cos ^{2}\theta }I_{0}\
342: \omega ^{3}\sin ^{2}\theta \left( 16\sin ^{2}\theta +\cos ^{2}\theta \right),
343: \label{correlation}
344: \end{equation}
345: and require that $\theta$ remain constant while $\epsilon$ and hence $\Omega_{\rm p}$
346: vary.
347:
348: \clearpage
349:
350: \begin{figure}
351: \plotone{f3.eps}
352: \caption{
353: Wobble angle $\theta$ as a function of the moment of inertia
354: involved in the precession for the four possible combinations of
355: precession period (935 s and 1430 s) and spin down rate
356: ($2\times 10^{-10}$ Hz/s and $3\times 10^{-10}$ Hz/s).
357: The curves shown, from bottom to top, correspond to the following
358: pairs of parameters: (935 s, $2\times 10^{-10}$ Hz/s),
359: (935 s, $3\times 10^{-10}$ Hz/s), (1430 s, $2\times 10^{-10}$ Hz/s),
360: and (1430 s, $3\times 10^{-10}$ Hz/s), respectively.
361: }
362: \label{fig3}
363: \end{figure}
364:
365: \clearpage
366:
367: \subsection{The constancy of the wobble angle}
368: \label{constwobble}
369: \citet{mid00} display graphically the correlation between $\dot\omega$
370: and $\Omega_{\rm p}^2$. By reading off the values of these variables and applying
371: the method and equations of Section \ref{misalign}, it is possible to determine the
372: values of the wobble angle $\theta$ required for each individual pair of values
373: $\dot\omega$ and $\Omega_{\rm p}$
374: for any assumed moment of inertia involved in the
375: precession. This exercise reveals that despite variations of $\epsilon$ and $\Omega_{\rm p}$
376: exceeding a factor of 1.5, the wobble angle does not change by more than a couple of
377: degrees and appears consistent with remaining constant within experimental errors.
378:
379: In the case of a freely precessing solid body, the wobble angle is largely determined
380: by initial conditions: taking the principal axes introduced in Section \ref{geometry},
381: if the associated moments of inertia remain constant, then $\omega_3=\omega\cos{\gamma}$,
382: also remains constant (see Fig. \ref{fig1}). It is easy to generalize Euler's equations to
383: the case in which the principal moments of inertia change due to unspecified
384: internal forces while the external torques vanish and the total angular momentum is conserved:
385: \renewcommand{\d}{{\rm d}}
386: \begin{eqnarray}
387: \frac{\d I_1 \omega_1}{\d t} &=& \omega_2\omega_3(I_2-I_3)\\
388: \frac{\d I_2 \omega_1}{\d t} &=& \omega_3\omega_1(I_3-I_1)\\
389: \frac{\d I_3 \omega_1}{\d t} &=& \omega_1\omega_2(I_1-I_2)\, .
390: \label{geneuler}
391: \end{eqnarray}
392: When the principal moments of inertia are all of the form $I_i=I_0 + \epsilon_i$, then
393: clearly all the time derivatives are of order $\sim\epsilon_i$ and even if the given
394: $\epsilon_i$ were to change by factors of a few, the result would be a small wobble of the
395: tip of $\omega$ in the body frame. Therefore we conclude that while the variations
396: probably detected by \citet{mid00} in
397: both $\epsilon$ and $\Omega_{\rm p}$ were significant, they do not imply any
398: measurable change in $\theta$. Referring back to Fig. \ref{fig1} we see that
399: $\widehat{\theta}$ may indeed change by amounts comparable to itself, but
400: the wobble angle $\theta$ would change very little. This conclusion is contrary to
401: what \citet{jan01} and \citet{jon01} have claimed regarding the
402: wobble angle, and thus it makes more plausible that the remnant of SN 1987A is indeed
403: freely precessing while undergoing changes in $\epsilon$ and $\Omega_{\rm p}$.
404:
405:
406:
407: \subsection{A more realistic model: allowing for a
408: elastic crust and presence of a fluid interior}
409: \label{realistic}
410: The textbook discussion of a precessing
411: body assumes that the object is perfectly rigid. A more realistic neutron star will have a
412: more or less elastic shell, and a fluid interior. The fluid is supposed to be
413: composed of a electron-proton plasma and a neutron superfluid. The plasma fluid
414: interior can couple to the crust because of friction. Under these conditions the
415: system is not simply described by the rigid body model.
416:
417: Usually the approach taken to explore the
418: properties of such more complicated systems is to understand the effect of one
419: additional complication at the time. The paper of \citet{cut00} addresses these
420: complications and shows how the more realistic model needs to be modified to
421: account for these complications. In this section we summarize these
422: results and apply them to the particular problem of the
423: detection of the 1987A remnant.
424:
425:
426: \subsubsection{The elastic crust}
427: \label{crust}
428:
429: In the case of an elastic crust's shell we
430: have to write the moment of inertia as:
431: \begin{equation}
432: {\mbf I}=I_{0}\bfdelta + \Delta I_{\rm d}\left(
433: {\mbf n}_{\rm d}{\mbf n}_{\rm d}- \bfdelta/3\right) +
434: \Delta I_\omega\left(
435: {\mbf n}_{\omega}{\mbf n}_{\omega}- \bfdelta/3\right)
436: \label{itensor2}
437: \end{equation}
438: this is the sum of a spherical part and two small
439: quadrupole contributions.
440: The first term is the moment of the inertia of the undeformed
441: shell, in the absence of rotation. The second term is a deformation due to
442: Coulomb lattice forces and the third is the deformation due to centrifugal forces. The
443: vector ${\mbf n}_{\rm d}$ determines the axis of symmetry of the deformation
444: $\Delta I_{\rm d}$. The vector ${\mbf n}_{\omega }$ lies along the axis of
445: rotation and determines the direction the axis of symmetry of the centrifugal
446: deformation $\Delta I_{\omega }.$
447:
448: The quantity $\Delta I_{\omega }$ is
449: caused by the deformation due to the centrifugal force, its value is determined
450: by:
451: \begin{equation}
452: \frac{\Delta I_{\omega }}{I}=\frac{I_{0}^{2}\omega
453: ^{2}}{4\left( A+B\right)}
454: \end{equation}
455: where the constants $A$ and $B$ depend on
456: the particular stellar equation of state. The constant $A$ is on the order of
457: the gravitational binding energy and the constant $B$ is on the order of the
458: total electrostatic binding energy of the ionic crustal lattice. The quantity
459: $B$ is much smaller than $A $ so we can make the approximation:
460: \begin{eqnarray}
461: \frac{\Delta I_{\omega }}{I}\approx
462: \frac{I_{0}^{2}\omega ^{2}}{4A}\approx \frac{\omega ^{2}R^{3}}{GM}\nonumber \\
463: \approx 2.1\times 10^{-3}\left( \frac{f}{100\,{\rm Hz}}\right)
464: ^{2}R_{6}^{3}/M_{1.4}
465: \end{eqnarray}
466: where $f$ is simply $\omega /2\pi$.
467:
468: In the general situation of non parallel
469: ${\mbf n}_{\rm d}$ and ${\mbf n}_{\omega}$, the body will
470: precess. As a consequence of ${\mbf n}_{\omega}$ being in the direction of
471: the rotation axis (at any given instant) the body will behave as a axysimmetric
472: top even if the body has a triaxial shape \citep{cut00}.
473: The angular momentum of an arbitrary body ${\mbf J}={\mbf I} \bfomega$,
474: with the inertia tensor ${\mbf I}$
475: given by equation (\ref{itensor2}), can be rewritten as
476: ${\mbf J}= {\mbf I}_{\rm eff} \bfomega$,
477: where ${\mbf I}_{\rm eff}=(I_{0}+2\Delta I_\omega/3)\bfdelta +
478: \Delta I_{\rm d}\left({\mbf n}_{\rm d}{\mbf n}_{\rm d}- \bfdelta/3\right)$.
479: Thus in this case the three moments of inertia in the original body axes
480: are:
481: \begin{eqnarray}
482: I_{1} &=&I_{0}-\Delta I_{\rm d}/3+2\Delta I_{\omega}/3\nonumber\\
483: I_{2} &=&I_{1}\\
484: I_{3} &=&I_{0}+2\Delta I_{\rm d}/3+2\Delta I_{\omega}/3\nonumber
485: \end{eqnarray}
486: The main implication of this is that even in the
487: case of a elastic crust the star will still behave for what concerns precession
488: as a biaxial rigid object. The fundamental equation (2) holds for this situation
489: (with the appropriate inertia moments given above) and this means that the only
490: piece of the moment of \ inertia that contributes to precession is $\Delta
491: I_{\rm d}=I_{3}-I_{1}$.
492:
493: \subsubsection{The presence of a fluid interior}
494: \label{fluid}
495: To further improve our model we consider the effects of the presence
496: of a fluid interior.
497: The shape of the cavity and the viscosity of the fluid contained are
498: important parameters. If the cavity is spherical, the presence
499: of the fluid in absence of viscosity has no influence on the
500: precession. If the cavity is non-spherical, then there will be a reaction force
501: that is generated by the tendency of the fluid to assume axial symmetry
502: around the axis of rotation. The shell will be pushed by the fluid. This problem
503: is solved in the literature (Lamb 1932; Jones \& Andersson 2001)
504: under the assumptions of uniform fluid vorticity,
505: small cavity ellipticity, and small wobble angle. A small wobble angle is
506: adopted in the treatment given by \citet{lam32} only for mathematical convenience,
507: but this assumption can be safely relaxed as long as the ellipticity remains small
508: without altering the result. The upshot is that the usual
509: precession equations described above are still valid. The only modification
510: to take into account is that $\Delta I_{\rm d}$ refers to the difference in moment of
511: inertia along the axis 1 and 3 of the whole star, and $I_{0}$ refers to the
512: average moment of the inertia of the shell only.
513:
514: In the presence of friction between the
515: crust and a part of the interior fluid in contact with the crust we could have
516: some coupling between the motion of the crust and the core. It can be shown that
517: in the case of neutron stars the coupling is very weak and the core does not
518: participate in the precession. If there are frictional forces at work in the
519: interior of the star these will serve just to damp the free precession on time
520: scales between 400 and 10$^{4}$ precession periods \citep{alp85}.
521:
522: \subsubsection{The problem of pinning}
523: \label{pinning}
524: \citet{jan01} following and extending previous
525: work by \citet{sha77} conclude that the presence of pinning of the superfluid to the
526: crust, at least in the simplest possible configuration does not change the form
527: of the equations that describe the precession. The main modification required is
528: that the relevant {\it effective ellipticity} is generated by combination
529: of the lattice deformation and the moment of the inertia $I_{\rm SF}$ of the pinned
530: fluid, as in the following:
531: \begin{equation}
532: \epsilon_{\rm eff}= \frac{\Delta I_{\rm d}}{I_{0}}+\frac{I_{\rm SF}}{I_{0}}
533: \end{equation}
534:
535: The most common theories on pulsar
536: glitches give a precise prediction on the precession behavior in the presence of
537: pinning in a neutron star. The theories require at least a few percent of the
538: total moment of inertia of the star to be in the pinned superfluid.
539: Current understanding of neutron star properties indicates that the moment
540: of inertia of the crust
541: is a few percent of the total moment of inertia of the
542: star. These considerations imply that:
543: \begin{equation}
544: \epsilon_{\rm eff}=\frac{\Omega _{\rm p}}{\omega \cos \theta }=\frac{\Delta
545: I_{\rm d}}{I_{0}}+\frac{I_{SF}}{I_{0}}\approx 1,
546: \end{equation}
547: in the case of small
548: deformations $\Delta I_{\rm d}.$ The precession and rotation frequency should be
549: close in value if there is a sizable quantity of superfluid that is pinned to
550: the crust. These predictions are not confirmed by observations of the three
551: strong cases of precession in neutron stars: PSR\ B1642-03 \citep{jan01}, PSR\ B1828-11
552: \citep{sta00} and the SN
553: 1987A remnant, where the precession is on a time scale much longer than the
554: rotation. The conclusion is that if the free precession interpretation of the
555: modulation of the signal of these pulsars is correct, then there is
556: almost no pinned superfluid in these stars (see \citet{link02} for further
557: discussion of this issue).
558:
559: \section{The wobble angle and crust fracture}
560: \label{fracture}
561:
562: Precession will cause the rotation axis of
563: the star to change its position relatively to the body frame. This means that
564: the centrifugal force distribution will be a function of position and time with
565: a timescale on the order of the precession period.
566: If the star has an elastic crust, then it will change its shape
567: in response to variations in the centrifugal force and
568: cause time dependent stresses in the crust. A simple order of magnitude
569: estimate of the strain on the crust $\sigma$ due to precession yields:
570: \begin{equation}
571: \sigma \approx \left(\Delta I_{\omega}/I\right)\sin{\theta} \approx 0.046\sin{\theta}
572: \ R_{6}^{3}/M_{1.4}.
573: \end{equation}
574:
575: Experiments with crystals suggest an
576: upper limit for the maximum possible strain sustainable by the crust before
577: breaking, i.e. $\sigma _{\max } \approx 10^{-2}.$ This implies that the possible
578: maximum wobble angle for our pulsar must satisfy:
579: \begin{equation}
580: \theta\le\arcsin{(21.8\sigma _{\rm max})}, \nonumber
581: \end{equation}
582: which would require $\theta\le 13^\circ$ for $\sigma _{\max }=0.01$.
583: Since all of our estimated wobble angles are larger than 30$^\circ$, either
584: $\sigma _{\max }$ is larger for the crust or the model is too simple
585: to account for the observations or our interpretation of the observations
586: is incorrect.
587:
588: \section{The strength of the radiation at earth}
589: \label{strength}
590:
591: \citet{zim79} and \citet{zim80} treat the
592: case of a body with two distinct moments of inertia and obtain the following
593: expressions for the strain parameter $h$ of gravity waves from a neutron star at
594: a distance $r$ from Earth and average moment of inertia
595: $I_{0}$:
596: \begin{eqnarray}
597: \lefteqn{h_{+} =\frac{G}{c^{4}}
598: \frac{2I_{0}\omega^{2}\varepsilon\sin{\theta}}{r}\times }\nonumber \\
599: &\left[ (1+\cos^2{i})\sin{\theta} \cos{2\omega t}+
600: \cos i\sin i\cos \theta \cos{\omega t}\right]\nonumber\\
601: & \\
602: \lefteqn{h_{\times} =\frac{G}{c^{4}}
603: \frac{2I_{0}\omega^{2}\varepsilon\sin{\theta}}{r}\times }\nonumber \\
604: &\left[ 2\cos{i}\sin \ \theta \sin{(2\omega t)}+
605: \sin{i}\cos{\theta}\sin{\omega t}\right] \nonumber,
606: \label{hphx}
607: \end{eqnarray}
608: where $i$ is the unknown angle between the
609: angular momentum vector ${\mbf J}$ and the plane of the sky.
610: It is important to notice that the time dependence of the
611: wave forms is sinusoidal with two main frequencies at $\omega $ and $2\omega.$
612: If the object was rotating along its symmetry axis it will emit just at a
613: frequency $2\omega $ (it will have also to be deformed along the axis
614: perpendicular to the rotation axis)$.$
615:
616: We see that the frequency of rotation
617: $\omega $ is one of the important parameters that determine the strength of the
618: gravitational radiation on earth. We know from observation the value of the
619: rotation frequency to be $\omega =2\pi \ 467.5$ Hz. To determine what is the
620: strength of the radiation we need to know also the moment of inertia
621: $I_{0}$ involved in the precession and the wobble angle $\theta$.
622: We showed previously that $\theta $ depends on how much of the moment of inertia of
623: the star is actually involved in the precession, as shown by the general
624: relativistic energy loss equation (\ref{dote}). When this relationship between $I_{0}$ and
625: $\theta$ is factored in the strain equations (17), we can determine the strength
626: of the gravitational radiation on earth as a function of the parameter $I_{0}.$
627: The result is shown in Figure 4.
628:
629: \clearpage
630:
631: \begin{figure}
632: \plotone{f4.eps}
633: \caption{
634: The amplitude of the strain at the Earth given by equation
635: (\ref{hphx}) -- leaving out dependences on $i$ and $t$ --
636: as a function of the moment of inertia
637: involved in the precession, in units of $10^{45}$ g cm$^2$.
638: The dashed curve corresponds to the
639: contribution from the $\omega$ term and the solid curve to the $2\omega$ term.
640: }
641: \label{fig4}
642: \end{figure}
643:
644: \clearpage
645:
646: These small values for $h$ may appear to
647: require an impossible level of sensitivity from the bar detectors or
648: interferometers existing today or soon available. It is important to notice that
649: the source is a continuous source of radiation, of which all the fundamental
650: parameters (besides the phase of the signal) are known. So it possible, in
651: principle, to integrate the detector data over a long period (even years) to
652: extract the signal from the incoherent noise. A detailed calculation of the
653: necessary integration time $\tau $ is required. To do so we use the following
654: equation:
655: \begin{equation}
656: h_{n}=\sqrt{S_h (f_0)}\sqrt{BW},
657: \end{equation}
658:
659: This equation expresses the level of the
660: strain $h_{n}$ of the noise in the data from a detector with characteristic noise
661: spectrum $S_{h}$. The equation evaluates the value $S_{h}(f_{0})$ of the
662: spectrum at the precise frequency $f_{0}$ of the looked for gravitational wave
663: signal. The quantity $BW$ is the bandwidth of the periodic signal. From Fourier
664: Analysis theory in the case of a sinusoidal signal, the value of $BW$
665: =$\frac{1}{\tau },$ where $\tau $ is the observation or ``integration'' time. So
666: the required integration time is:
667: \begin{equation}
668: \tau =
669: \left({\sqrt{S_{h}(f_{0})}\over h_{n}}\right)^{2}
670: \end{equation}
671:
672: Now we require that the noise level in the
673: data from the detector be at least of the same size of the signal (it should
674: less, 4 times less for a 4 $\sigma $ confidence level in the statistics, for
675: example). A typical value for\ the noise strain $h_{\rm d}=\sqrt{S_{h}(f_{0})}$ in
676: the existing bar detectors such as the Louisiana State University's ALLEGRO or first
677: generation light interferometers \ as LIGO\ I, is currently of order $10^{-20}$.
678: The fully optimized LIGO I sensitivity is projected to attain a minimum noise strain
679: $h_{\rm d}\approx 10^{-22}$ in a couple of years (see Fig. 5).
680: We see from Figure \ref{fig4} that a typical value for the signal amplitude strain is
681: approximately $h_{s}=5\times 10^{-27}.$
682: So if $h_{n}$ is chosen to be $\approx
683: 1/4\times h_{s}=1.25\times 10^{-27}$ then we get that:
684: \begin{eqnarray}
685: \tau = \left({\sqrt{S_{h}(f_{0})}\over h_{n}}\right)^{2}=
686: \left({10^{-20}\over 1.25\times 10^{-27}}\right)^{2}
687: \left({h_{\rm d}\over 10^{-20}}\right)^2 {\rm s}\nonumber \\
688: \approx 2\times 10^{6} \left({h_{\rm d}\over 10^{-20}}\right)^2 {\rm yr}.
689: \end{eqnarray}
690: This is a time obviously too long to be useful, even with a fully optimized LIGO I.
691: Thus the remnant of supernova 1987A is
692: undetectable by the existing gravitational wave detectors, but may be
693: detectable if the sensitivity of the detectors planned for the near future
694: reaches the estimated levels.
695:
696: In fact preliminary estimates of the noise
697: spectrum of the second generation Laser Interferometer, LIGO II are very
698: promising (see Fig. 5). LIGO II will be built on the experience of the first LIGO and will be
699: a much better gravitational wave observatory. It will be on line in 4 or 5 years
700: from now. According to Fig. 5, there is a lowest point in the total strain-noise
701: (the sum of different kind of expected noises). This point is about
702: $h(f_{r})=$1.5$\times 10^{-24}$ /$\sqrt{Hz}$ at a frequency of 350 Hz. But the
703: LIGO\ II detector will be able to use narrow banding to shift this lowest point
704: in noise level to higher frequencies. For further discussion of noise levels
705: expected in (Advanced) LIGO II, see \citet{abb02}.
706:
707: \clearpage
708:
709: \begin{figure}
710: \plotone{f5.eps}
711: \caption{
712: Preliminary estimates of the noise spectrum expected for LIGO II.
713: The contributions from various sources are shown separately and added
714: together (See http://www.ligo.caltech.edu/docs/M/M990288-A1.pdf).}
715: \label{fig5}
716: \end{figure}
717:
718: \clearpage
719:
720: So we could take this as the level of
721: noise at the frequency of emission of \ 1987A. Figure 6 shows the required time
722: of integration for LIGO\ II as a function of the parameter $I_{0}.$ We can see
723: that within a certain range of possible values of $I_{0}$ it will be possible to
724: detect the signal from the 1987A remnant in reasonable time. This time is
725: actually few days if most of the moment of the inertia participates in the
726: precession.
727:
728: \clearpage
729:
730: \begin{figure}
731: \plotone{f6.eps}
732: \caption{
733: Estimated integration times for a 4 sigma detection by LIGO II for the signal at
734: $\omega$ (dashed) and at $2\omega$ (solid) as a function of $I_0$
735: in units of $10^{45}$ g cm$^2$. The curves shown above have been estimated
736: using the minimum strain level for an optimized LIGO II, $h_{\rm d} = 1.5\times 10^{-24}$.
737: For higher noise levels, the integration times scale as $(h_{\rm d}/1.5\times 10^{-24})^2$.
738: }
739: \label{fig6}
740: \end{figure}
741:
742: \clearpage
743:
744: \section{Data analysis and templates}
745: \label{templates}
746:
747: The simple estimates for integration times in the previous section assume that
748: the signal is sinusoidal with constant frequency, which clearly is not the case
749: if our interpretation of the properties of the observed optical signal is correct.
750: This raises the question
751: of how detectable would be the signal if one allows for the observed
752: changes in the spin frequency, precession frequency and spin down rate.
753: We show here that simple templates
754: that describe the observed behavior can be constructed, and that the required
755: number of such templates and the total computational effort needed to
756: adequately keep track of the phase and to detect
757: a signal with the presumed properties is
758: within current computational capabilities
759: as long as there is phase
760: stability over a time series of length comparable or longer than the integration time.
761: Since the required integration times are on the order of 10--30 days, and the
762: phase stability in the Middleditch data was comparable or better than that,
763: this requirement is likely to be satisfied.
764:
765: Following standard treatments we write the time-dependent frequency as a Taylor series
766: \begin{equation}
767: \omega(t) =
768: \sum_{n=0}^\infty {\omega_n(0) t^n\over n!},
769: \label{templ1}
770: \end{equation}
771: where $\omega_n(0)$ indicates the $n$-th derivative at some arbitrary reference time
772: taken to be zero without loss of generality. A given choice of the parameters $\omega_n(0)$
773: constitutes a particular choice of template. Then the phase difference between two
774: different templates is
775: \begin{equation}
776: \Delta\varphi (t) =
777: \sum_{n=0}^\infty {\Delta\omega_n(0) t^{n+1}\over (n+1)!}=
778: \Delta\omega(0)\, t + {1\over2}\Delta\dot\omega(0)\, t^2 + \cdots\, ,
779: \label{templ2}
780: \end{equation}
781: where $\Delta\omega_n(0)$ represents the difference between the $n$-th derivatives
782: for a pair of templates.
783:
784: During the observations a typical value for the spin-down was
785: $\dot f\sim 10^{-10}\, {\rm s}^{-2}$ and showing a secular decreasing trend.
786: Since the earliest opportunity for LIGO to observe this source is $T\sim$ 10 years
787: away, we take the uncertainty in the frequency to be on the order of
788: $\dot f T \sim 3\times 10^{-2} \, {\rm s}^{-1}$ or a bandwith of
789: $BW=3\times 10^{-2}$ Hz.
790: This is an estimate for the total range
791: of frequencies to be explored. The standard phase stability
792: requirement \citep{jakr00} $\Delta\omega\tau \lta \pi/4$ over the
793: integration time, yields an estimate of how closely spaced the frequency
794: templates have to be. For $\tau\sim 10\,{\rm d}\sim 10^6\, {\rm s}$, this
795: argument yields $\Delta f = \Delta\omega/2\pi\sim 10^{-7}$ Hz. Consequently the
796: total number of frequencies to be sampled is on the order of
797: $BW/\Delta f\sim3\times 10^5 (\tau/10 {\rm d})$.
798:
799: To calculate how many frequency derivative values need to be considered,
800: we estimate that the total range of values is comparable to $\dot f$ itself.
801: The phase stability requirement then yields the spacing between spin-down
802: values: $\Delta\dot f = \Delta\dot\omega/2\pi \lta \tau^{-2}/4 \sim 2.5 \times 10^{-13}$ Hz/s.
803: And therefore the total number of values of $\dot f$ to be sampled is
804: approximately $\dot f/\Delta\dot f \sim 400 (\tau/10 {\rm d})^2$.
805:
806: Finally, the total number of two-parameter templates $N_f$ we require is
807: given by the simple product of the number of frequency values
808: times the number of frequency derivative values:
809: $N_f \sim 1.2\times 10^8(\tau/10 {\rm d})^3$.
810: The total number of floating-point operations required to carry out the search of
811: these templates over an integration time $\tau$ is approximately given by the
812: formula \citep{br98}
813: \begin{equation}
814: N_{\rm fpo} = 6 f N_f \tau [\log_2{(2 f \tau) +1/2]}\, ,
815: \label{flops}
816: \end{equation}
817: where $f$ is the maximum frequency to be searched (say 500 Hz).
818: With the values derived above,
819: this yields a total computational load of
820: approximately $1.1 \times 10^{19}(\tau/10 {\rm d})^4$
821: floating-point
822: operations, which would require 3 months of calculations for a Teraflop machine.
823: While this load is not trivial, it can be achieved by either processing the data
824: offline or using a machine clocking at least
825: $11 (\tau/10 {\rm d})^3$ Teraflops for online processing.
826: However, the above estimate is an upper limit that makes little
827: use of our prior knowledge of the expected frequency and frequency range
828: of the signal. We need only to search over the $BW$ of $3\times 10^{-2}$ Hz, whereas the
829: standard argument above assumes we are searching for signals over the entire
830: band from 0 to 500 Hz. The
831: computational task can be significantly reduced by first `demodulating' or filtering
832: the signal to the bandwith $BW$ estimated above and then `decimating' or reducing the
833: signal sampling rate to the bandwith. This technique cuts the processing rate
834: essentially by a factor $BW/f\sim 6\times 10^{-5}$ to approximately
835: $0.7 (\tau/10 {\rm d})^3$ Gigaflops,
836: well within the capabilities of current computers.
837:
838:
839:
840:
841:
842: \section{Conclusions}
843: \label{conclude}
844:
845:
846: In this paper we discussed the
847: implications of the observation of a precessing neutron star in the remnant of
848: supernova 1987A for gravitational wave detection. We used the observed data on
849: rotational velocity, spin down and precession rate to determine the value of the
850: possible asymmetric deformation that causes the precession.
851:
852: To estimate the size of
853: deformation it is important also to determine the wobble angle
854: between the axis perpendicular to the deformation and the rotation axis.
855: General relativity gives us an equation of the loss of energy, trough gravity
856: waves. Knowing the rate of spin down, the rotation frequency and the precession
857: frequency allows us to find the wobble angle. This is possible under the
858: assumption that the main mechanism for the loss of rotational energy is due to
859: emission of gravitational radiation. Once we know the wobble angle,
860: we can calculate the
861: strength of the radiation on earth. In fact the value of the dimensionless
862: strain parameter $h$ depends on the value of the wobble angle quite strongly.
863:
864: Our discussion shows that even with a more a realistic model
865: of a precessing neutron star that takes in consideration the presence of a
866: crust, with a certain elasticity and the eventual presence of a fluid interior
867: the precessional behavior is similar to that of the simple biaxial model.
868: The ratio of precesion frequency and spin frequency determines the order of
869: magnitude of the ellipticity, but a complete solution requires an estimate of
870: the wobble angle. The preceding discussion shows that it is possible to obtain
871: such self-conssitent models as a function of essentially one parameter:
872: the moment of inertia $I_{0}$ that is involved in the precession. Given it,
873: the observational data allow to determine the wobble angle, the size of
874: deformation and consequently the strength of the radiation on earth.
875:
876: We saw in the previous section
877: that to avoid crust breaking the wobble angle has to be relatively small.
878: In fact, formally, even the smallest wobble angle among the possible range
879: of solutions violates the maximum crustal strain. Given the uncertainties in the
880: model and in the interpretation of the data, we conclude that even if the
881: limits on the maximum strain $\sigma_{\rm max}$ are relaxed, any viable solution
882: is likely to have a wobble angle near the small end of the range and consequently
883: the moment of inertia must be near the high end of its range.
884: In turn this means that a short integration time on the order of days
885: is required to
886: observe with confidence the gravity wave signal from SN1987A using advanced
887: detectors as LIGO\ II. Unfortunately the presently generation of
888: detectors such as the resonant bars and LIGO I would require
889: observation times of the order
890: of a million years
891: to extract the signal from the noise.
892: Thus if the precession interpretation is correct,
893: the SN 1987A remnant would be among the best candidates for
894: {\em a search for a}
895: continuous source of gravitational waves. In any case, it is clear that
896: a targeted search for gravitational waves from this source is worthwhile
897: since both detection and absence of detection over a relatively short time
898: will yield interesting constraints on models for precessing neutron stars.
899:
900:
901:
902: \acknowledgements
903: This research was supported in part by NSF's Experimental Gravity Program
904: grant 9970742 and by grants AST9720771 and NAG5 8497 to LSU.
905:
906: \begin{thebibliography}{}
907:
908: \bibitem[Abbott \etal (2002)]{abb02}
909: Abbott, R., \etal. (2002), `Seismic isolation for Advanced LIGO',
910: {\em Class. Quantum Grav.} {\bf 19}, 1591.
911:
912: \bibitem[Alpar \& Pines (1985)]{alp85}
913: Alpar, A. \& Pines, D. 1985, `Gravitational radiation from a solid-crust neutron
914: star', {\em Nature} {\bf 314}, 334.
915:
916: \bibitem[Brady \etal (1998)]{br98}
917: Brady, P.R., Creighton, T., Cutler, C. \& Schutz, B.F. (1998),
918: {\em Phys. Rev. D} {\bf 57}, 2101.
919:
920: \bibitem[Cutler \& Jones (2000)]{cut00}Cutler, C. \& Jones, D.I. (2000),
921: `Gravitational Wave damping of Neutron Star Wobble',
922: gr-qc/0008021 (2000).
923:
924: \bibitem[Jaranowski \& Krolak (2000)]{jakr00}
925: Jaranowski, P. \& Krolak, A. (2000), `Data analysis of gravitational-wave
926: signals from spinning neutron stars. III. Detection statistics and
927: computational requirements', {\em Phys. Rev. D} {\bf 61}, 62001.
928:
929: \bibitem[Jones (2001)]{jon01}
930: Jones, D.I. (2001), `Gravitational waves from rotating neutron stars',
931: gr-qc/0111007. Conf. Proc. Amaldi 4, Perth, Australia.
932:
933: \bibitem[Jones \& Andersson (2001)]{jan01}
934: Jones, D.I. \& Andersson, N. (2001), `Freely precessing neutron stars: Models and
935: Observations', {\em M.N.R.A.S.} {\bf 324}, 811 (astro-ph/0011063).
936:
937: \bibitem[Lamb (1932)]{lam32}
938: Lamb, H. (1932), {\em Hydrodynamics}, 6th ed., Dover Publications (1945), New York.
939:
940: \bibitem[Link \& Cutler (2002)]{link02}
941: Link, B. \& Cutler, C. (2002), `Vortex unpinning in precessing neutron stars',
942: {\em M.N.R.A.S.} {\bf 336}, 211.
943:
944: \bibitem[Marion \& Thornton (1995)]{mar95}
945: Marion, J.B. \& Thornton, S. T. (1995), {\em Classical Mechanics of particles and systems},
946: Saunders College, Forth Worth, Philadelphia.
947:
948: \bibitem[Middleditch \etal (2000)]{mid00}
949: Middleditch, J.A. Kristian, W.E. Kunkel, K.M. Hill, R.D. Watson, R. Lucinio,
950: J.N. Imamura, T. Steiman-Cameron, Y. Thomas, A. Shearer, R. Butler, M. Red ,
951: Michael; A.C. Danks, 2000, `Rapid photometry of supernova 1987A: a 2.14 ms
952: pulsar?' {\em New Astronomy} {\bf 5}, 243
953:
954: \bibitem[Nagataki \& Sato (2001)]{nag01}
955: Nagataki, S. \& Sato, K. (2001), `Implications of the Discovery of a Millisecond Pulsar in SN
956: 1987A'. {\em Prog. Theor. Phys.} {\bf 105}, 429. astro-ph/001163 v3 (2001).
957:
958: \bibitem[Shaham (1977)]{sha77}
959: Shaham, J. (1977), {\em Ap.J.} {\bf 214}, 251
960:
961: \bibitem[Stairs \etal (2000)]{sta00}
962: Stairs, H., Lyne, A.G. \& Shemar, L. (2000), `Evidence of free precession in a
963: pulsar', {\em Nature } {\bf 406}, p.484-466.
964:
965: \bibitem[Zimmerman \& Szedentis (1979)]{zim79}
966: Zimmerman, M. \& Szedentis Jr., E. (1979), `Gravitational waves from rotating and
967: precessing bodies. Simple models and applications to pulsars'. {\em Physical
968: Review D} {\bf 20}, 351-355.
969:
970:
971: \bibitem[Zimmerman (1980)]{zim80}
972: Zimmerman, M. (1980), `Gravitational waves from rotating and precessing bodies. II.
973: General solutions and computationally useful formulas', {\em Physical
974: Review D} {\bf 20}, 351-355.
975:
976: \end{thebibliography}
977:
978:
979:
980:
981: \end{document}
982:
983:
984: