astro-ph0409442/ms.tex
1: \documentclass{emulateapj}
2: \RequirePackage{natbib}
3: 
4: %% manuscript produces a one-column, double-spaced document:
5: 
6: %%%\documentclass[manuscript]{aastex}
7: %%%\documentstyle[12pt,epsf]{article}
8: %\documentclass[12pt,preprint,epsf]{aastex}
9: %\documentstyle[emulateapj,epsf]{article}
10: %\documentstyle[12pt,preprint,epsf]{aastex}
11: %%%\usepackage{emulateapj5}
12: 
13: %% preprint2 produces a double-column, single-spaced document:
14: %\documentclass[preprint2,epsf]{aastex}
15: 
16: \newcommand{\vdag}{(v)^\dagger}
17: \newcommand{\myemail}{skywalker@galaxy.far.far.away}
18: \def\agt{\mathrel{\raise.3ex\hbox{$>$}\mkern-14mu\lower0.6ex\hbox{$\sim$}}}
19: \def\alt{\mathrel{\raise.3ex\hbox{$<$}\mkern-14mu\lower0.6ex\hbox{$\sim$}}}
20: 
21: \newcommand{\beq}{\begin{equation}}
22: \newcommand{\eeq}{\end{equation}}
23: \newcommand{\beqn}{\begin{eqnarray}}
24: \newcommand{\eeqn}{\end{eqnarray}}
25: \newcommand{\pa}{\partial}
26: \newcommand{\vp}{\varphi}
27: \newcommand{\varep}{\varepsilon}
28: \newcommand{\ep}{\epsilon}
29: \newcommand{\comp}{(M/R)_\infty} 
30: 
31: %% You can insert a short comment on the title page using the command below.
32: 
33: %\slugcomment{}
34: 
35: %% If you wish, you may supply running head information, although
36: %% this information may be modified by the editorial offices.
37: %% The left head contains a list of authors,
38: %% usually a maximum of three (otherwise use et al.).  The right
39: %% head is a modified title of up to roughly 44 characters.  Running heads
40: %% will not print in the manuscript style.
41: 
42: \shorttitle{Bar-Mode Instability}
43: \shortauthors{Shapiro}
44: 
45: %% This is the end of the preamble.  Indicate the beginning of the
46: %% paper itself with \begin{document}.
47: 
48: \begin{document}
49: 
50: \title{The Secular Bar-Mode Instability 
51: in Rapidly Rotating Stars Revisited}
52: 
53: \author{Stuart L. Shapiro \altaffilmark{1,2}}
54: 
55: \affil
56: {\altaffilmark{1} 
57: Department of Physics, University of Illinois at Urbana-Champaign,
58: \break
59: Urbana, IL 61801-3080}
60: %
61: \affil
62: {\altaffilmark{2} 
63: Department of Astronomy and NCSA, University of Illinois at
64: Urbana-Champaign,
65: \break
66: Urbana, IL 61801-3080}
67: 
68: 
69: \begin{abstract}
70: Uniformly rotating, homogeneous, incompressible Maclaurin spheroids
71: that spin sufficiently rapidly are secularly unstable to nonaxisymmetric, 
72: bar-mode perturbations when viscosity is present. The intuitive explanation is
73: that energy dissipation by viscosity can drive an unstable spheroid to 
74: a stable, triaxial configuration of lower energy -- a Jacobi ellipsoid.
75: But what about rapidly rotating {\it compressible} stars? Unlike incompressible
76: stars, which contain no internal energy and therefore immediately liberate 
77: all the energy dissipated by viscosity, compressible stars have internal energy and can 
78: retain the dissipated energy as internal heat. 
79: Now compressible stars that rotate sufficiently rapidly {\it and} also manage to liberate this 
80: dissipated energy very quickly are known to be unstable to bar-mode perturbations, like 
81: their incompressible counterparts. But what is the situation for rapidly rotating compressible stars
82: that have very long cooling timescales, so that all the energy dissipated by viscosity 
83: is retained as heat, so that the total energy of the star remains constant on a 
84: secular (viscous) evolution timescale? 
85: Are such stars also unstable to the nonlinear growth of bar modes, or is the viscous heating 
86: sufficient to cause them to expand, drive down the ratio of
87: rotational kinetic to gravitational potential energy $\mathcal {T}/|W| \propto R_{\rm eq}^{-1}$, 
88: where $R_{\rm eq}$ is 
89: the equatorial radius, and turn off the instability before it gets underway? Alternatively, 
90: if the instability still arises in such stars, at what rotation rate do they
91: become unstable, and to what final state do they evolve?  We provide definitive answers to these
92: questions in the context of the compressible ellipsoid model for rotating stars.
93: The results should serve as useful guides for
94: numerical simulations that solve the exact Navier-Stokes equations in $3+1$ dimensions for 
95: rotating stars containing viscosity.
96: \end{abstract}
97: 
98: %% Keywords should appear after the \end{abstract} command. The uncommented
99: %% example has been keyed in ApJ style. See the instructions to authors
100: %% for the journal to which you are submitting your paper to determine
101: %% what keyword punctuation is appropriate.
102: 
103: \keywords{Gravitation---hydrodynamics---instabilites---stars: rotation}
104: 
105: %% From the front matter, we move on to the body of the paper.
106: %% In the first two sections, notice the use of the natbib \citep
107: %% and \citet commands to identify citations.  The citations are
108: %% tied to the reference list via symbolic KEYs. The KEY corresponds
109: %% to the KEY in the \bibitem in the reference list below. We have
110: %% chosen the first three characters of the first author's name plus
111: %% the last two numeral of the year of publication as our KEY for
112: %% each reference.
113: 
114: \section{Introduction}
115: 
116: Rapidly spinning equilibrium stars are subject to nonaxisymmetric
117: rotational instabilities.  An exact treatment of these instabilities
118: exists for incompressible fluid configurations in Newtonian
119: gravity (see, e.g., Chandrasekhar 1969). 
120: For axisymmetric, uniformly rotating Maclaurin spheroids, global
121: rotational instabilities arise from nonradial toroidal modes
122: $e^{im\varphi}$ ($m=\pm 1,\pm 2, \dots$) when $\mathcal{T}/|W|$ exceeds a
123: certain critical value. Here $\varphi$ is the azimuthal coordinate and
124: $\mathcal{T}$ and $W$ are the rotational kinetic and gravitational potential 
125: energies of the star, respectively.  
126: In the following discussion we will focus on the $m=\pm 2$ bar-mode,
127: also known as the $l=m=2$ $f$-mode ($f=$ fundamental),
128: which is the fastest growing mode when the rotation is sufficiently
129: rapid.
130: 
131: There exist two different mechanisms and corresponding timescales for
132: bar-mode instabilities.  Uniformly rotating, incompressible stars in
133: Newtonian theory are {\it secularly} unstable to bar-mode formation
134: when $\mathcal{T}/|W| = 0.1375$.  This instability can
135: grow only in the presence of some dissipative mechanism, like
136: viscosity or gravitational radiation, and the growth time occurs on
137: the dissipative timescale, which is usually much longer
138: than the dynamical timescale of the system.  By contrast, a {\it
139: dynamical} instability to bar-mode formation sets in when 
140: $\mathcal{T}/|W| = 0.2738$.  This instability is independent of any
141: dissipative mechanisms, and the growth time occurs on the
142: hydrodynamical timescale of the system. Here we shall be interested in
143: the {\it secular} bar-mode driven by \mbox{\it viscous} dissipation.
144: 
145: Nonaxisymmetric secular instabilities in uniformly rotating compressible stars, 
146: e.g., polytropes, have been analyzed by several authors 
147: (see, e.g., Managan 1985; Imamura, Friedman \&
148: Durisen 1985; Ipser \& Lindblom 1990, 1991). The $m=2$ bar-mode sets in again at $\mathcal{T}/|W| \simeq 0.14$. However, this mode
149: is reached only when the polytropic index of the star satisfies 
150: $n \lesssim 0.808$ (James 1964). Stars with larger  $n$ are too 
151: centrally condensed to support high enough
152: spin in uniform rotation without undergoing mass-shedding at the
153: equator (see Tassoul 1978 and Shapiro \& Teukolsky 1983 
154: for a discussion and references). This constraint does not apply to
155: differentially rotating stars, which can support significantly
156: more rotational energy in equilibrium, even when the degree of
157: differential rotation is only moderate.
158: The critical value for the onset of the 
159: secular $m=2$ bar-mode is again $\mathcal{T}/|W| \simeq 0.14$ for a wide range of angular
160: momentum distributions and barotropic equations of state
161: (see, e.g., Ostriker \& Bodenheimer 1973; Bardeen et al. 1977; 
162: Friedman \& Shutz 1978a,b). Hence stability criteria derived 
163: for uniformly rotating, incompressible spheroids turn out to 
164: have general applicability to rapidly rotating stars with more 
165: realistic density and angular velocity profiles, at least insofar as
166: the $m=2$ bar-mode is concerned.
167: 
168: The point of onset of the secular bar-mode instability in homogeneous, 
169: axisymmetric Maclaurin spheroids exactly coincides with the 
170: point of bifurcation where the Maclaurin
171: sequence branches off into the nonaxisymmetric Jacobi sequence. 
172: Incompressible Jacobi ellipsoids are homogeneous, uniformly rotating, triaxial equilibria that have 
173: lower total energies than corresponding Maclaurin spheroids of the same
174: mass, density and angular momentum.
175: The nonlinear,
176: secular evolution of a slightly perturbed, unstable Maclaurin spheroid to
177: a triaxial Jacobi ellipsoid was explicitly demonstrated by
178: Press and Teukolsky (1973), who integrated
179: the full set of Riemann-Lebovitz ordinary differential
180: equations for incompressible ellipsoids with viscosity.
181: Sequences of uniformly rotating polytropes with $n \leq 0.808$ 
182: also have a point of bifurcation beyond which they are secularly unstable;
183: for example, this point occurs at $\mathcal{T}/|W| = 0.1298$ for
184: $n=0.75$ (Ipser \& Lindblom 1990), close to the
185: value $\mathcal{T}/|W| = 0.1375$ for $n = 0$ (i.e. incompressible) stars. Viscosity 
186: again is expected to drive unstable models to secularly stable,
187: Jacobi-like ellipsoids in compressible stars.
188: 
189: It seeems almost self-evident why spheroids beyond the bifurcation point 
190: are secularly unstable to 
191: bar-mode perturbations, since with viscosity such configurations can evolve 
192: to equilibrium configurations of lower energy, while 
193: conserving their mass and angular momentum. However, this intuitive explanation
194: is not applicable to a rapidly rotating star
195: in which the energy liberated by viscosity
196: is {\it not} emitted by the star, but instead remains trapped  
197: in the form of thermal energy. In this case
198: the star loses no mass, angular momentum {\it or} energy as it evolves via
199: viscosity,
200: and thus it cannot evolve to a lower-energy equilibrium state. 
201: This problem does not arise in
202: incompressible stars, since they possess no internal (thermal) energy 
203: whatsoever and, hence, all the energy generated by viscosity is liberated
204: instantaneously.  The rate of decrease in the total energy of an 
205: incompressible star is exactly the rate of viscous energy generation;
206: this equality is guaranteed by the Riemann-Lebovitz equations 
207: with viscosity.  By contrast, compressible matter can possess thermal 
208: energy, and the fate of
209: viscous energy dissipation in compressible stars 
210: depends on timescales. If the cooling 
211: timescale, due, e.g., to thermal radiation, is shorter than the 
212: thermal heating timescale due to viscosity, then the thermal energy
213: generated by viscosity  will be radiated away quickly 
214: and the total energy of the star
215: will decrease on the viscous dissipation timescale, as in the incompressible
216: case. If, however, the cooling timescale is longer than the viscous 
217: heating timescale, 
218: there will be an increase in thermal energy, but the total energy of the
219: star will be nearly constant on a viscous timescale. 
220: The possibility for the nonlinear growth of a bar-mode instability is not so 
221: obvious in the later situation.
222: In fact, in a compressible star in which the cooling timescale 
223: is long, any viscosity-generated enhancement in the 
224: thermal energy might cause
225: the star to expand, increasing the equatorial radius $R_{\rm eq}$, 
226: decreasing the value of $\mathcal{T}/|W| \sim 1/R_{\rm eq}$ and potentially 
227: turning off any nonaxisymmetric mode that might be unstable initially. 
228: 
229: In this paper we address the issue of the secular bar-mode instability 
230: in rapidly rotating compressible stars in which the cooling timescale is
231: either much longer, or much shorter, than the viscous heating timescale. 
232: We integrate the dynamical equations for a compressible star with viscosity
233: using the formalism of Lai, Rasio \& Shapiro (1994) (hereafter LRS 1994). 
234: This formalism
235: provides a set of {\it ordinary} differential equations (ODEs) for the evolution of the
236: principal axes and other global parameters characterizing the star, 
237: which is approximated as a triaxial, compressible ellipsoid with a 
238: polytropic equation of state and a velocity field that is a linear function 
239: of the coordinates. The 
240: equations are exact in the incompressible limit $n=0$, where
241: they describe the exact hydrodynamical behavior of the objects and 
242: reduce to Riemann-Lebovitz equations identically.
243: This dynamical model for compressible
244: ellipsoids is formally equivalent to the affine
245: model developed by Carter \& Luminet (1983,1985), although 
246: the formulation is quite different. 
247: 
248: We follow the evolution of rapidly rotating, 
249: compressible spheroids that lie beyond the bifurcation point 
250: and are given small,
251: nonaxisymmetric perturbations at $t=0$. 
252: We track their evolution in the two extreme opposite limits of
253: very rapid and very slow cooling. We find that in {\it both} cases the 
254: configurations are secularly unstable to a bar-mode perturbation, but 
255: they evolve to {\it different}, stable, compressible Jacobi ellipsoids. 
256: In the rapid-cooling limit, the final ellipsoid has lower energy than the 
257: initial spheroid, while in the slow-cooling limit, the final 
258: ellipsoid  has the same energy, but a higher entropy than the
259: spheroid.
260: 
261: Our paper is organized as follows. In Section 2 we describe the compressible
262: ellipsoid model for rotating stars and extend the formalism of LRS to the
263: case in which the entropy of the fluid is allowed to change with time due to heating and cooling.
264: In Section 3 we integrate the dynamical equations with viscosity to show that, both in the case
265: of very rapid and very slow cooling, a uniformly rotating, compressible spheroid that rotates sufficiently
266: rapidly and is subjected to a small triaxial perturbation 
267: is unstable to the bar instability and evolves to a Jacobi ellipsoid. In this section we
268: also show that, in fact, the final Jacobi ellipsoid, which is different in the two cases, can be determined
269: a priori from the conserved quantities without having to follow the evolutionary track.
270: In Section 4 we confirm our numerical result that, in both cases, 
271: the onset of secular instability coincides with
272: the bifurcation point. In particular, we sketch a proof of this result by examining
273: the energy and free energy functionals of perturbed stars along the compressible Maclaurin sequence.
274: Finally, in Section 5 we summarize our results and describe plans for future work.
275: 
276: \section{Basic Equations}
277: \label{Sec2}
278: 
279: The key dynamical equations for the evolution of a compressible
280: fluid in the ellipsoidal approximation
281: are derived in LRS (1994)
282: and will not be repeated here. [See Section 2 of that paper for 
283: an ideal fluid, and Section 4 for a fluid with shear
284: viscosity.] They are obtained from the Euler-Lagrange equations derived
285: from a Lagrangian governing the dynamics of a compressible ellipsoid. 
286: The properties of compressible equilibria satisfying the
287: stationary-state ellipsoidal equations are described in LRS (1993). The 
288: assumptions underlying the compressible ellipsoidal treatment, as well as comparisons between
289: ellipsoidal model calculations and calculations employing the exact fluid
290: equations, are summarized in these papers and the references therein. The basic assumption 
291: is that we assume that the surfaces of constant density can be represented approximately
292: by {\it self-similar ellipsoids}. The geometry is then completely specified by the three principal
293: axes of the outer surface. Furthermore, we assume that the density profile $\rho(m)$ inside each star,
294: where $m$ is the mass interior to an isodensity surface, is identical to that of a {\it spherical}
295: polytrope with the same volume. The velocity field of the fluid in a frame at rest with the star's 
296: center of mass is modeled as a linear superposition
297: of three components: (1) a rigid rotation of the ellipsoidal {\it figure}; (2) an internal fluid circulation
298: with {\it uniform vorticity}; and (3) ellipsoidal expansion or contraction. In our adaptation, where we restrict
299: ourselves to the cases in which both the angular velocity and the vorticity are parallel to one of the principal
300: axes, the exact
301: hydrodynamical equations (the Euler and Navier-Stokes equations) are replaced by a set of ODEs for the
302: time evolution of the principal axes, the angular velocity of the ellipsoidal figure and the angular frequency
303: of the internal circulation. As mentioned above, in the incompressible limit (polytropic index $n=0$) the solutions
304: we derive represent {\it exact} solutions of the true hydrodynamic equations. For $n \neq 0$, our solutions
305: are only approximate, since the isodensity surfaces can no longer be exactly ellipsoidal, and the velocity field
306: of the fluid cannot be described exactly by a linear function of the coordinates. We adopt
307: the notation of LRS in the discussion below.
308: 
309: \subsection{Energy Evolution Equation}
310: \label{Secevo}
311: 
312: The equation of state characterizing the fluid is given by the 
313: polytropic law
314: \begin{equation} \label{eos}
315: P = K \rho^{\Gamma}, ~~~\Gamma = 1 + 1/n,
316: \end{equation}
317: where $P$ is the pressure, $\rho$ is the density, $\Gamma$ is the adiabatic
318: constant and $n$ is the polytropic index.
319: If the polytropic entropy parameter $K$ is assumed to be
320: constant in space {\it and} time, as in LRS (1994), the equation of state is 
321: barotropic with $P = P(\rho)$. In this case, no internal energy
322: equation is needed to describe the evolution of the 
323: thermodynamic state of the star.
324: In general, however, the gas may undergo heating and cooling, so
325: that  the parameter $K$ will be a function of the changing specific
326: entropy $s$, i.e., $K = K(s)$.  In this case the equation of state
327: is not barotropic, since $P = P(\rho, s)$, and it is necessary to solve an
328: energy evolution equation, which we will now derive. Combining 
329: eqn.~(\ref{eos}) with
330: the first law of thermodynamics,
331: \begin{equation} \label{first}
332: T ds = d (\varepsilon/\rho) + P d(1/\rho),
333: \end{equation}
334: where $T$ is the temperature, and adopting the 
335: ideal gas relation for the internal energy density, 
336: $\varepsilon = P/(\Gamma -1)$,  
337: yields an equation for $K$:
338: \begin{equation} \label{K}
339: T ds = \frac{\rho^{\Gamma-1}}{\Gamma-1} dK.
340: \end{equation}
341: Henceforth we shall assume, for simplicity,
342: that $s$, and hence $K$, are independent of position
343: in the star, but we will allow them to vary with time.
344: Integrating eqn.~(\ref{K}) over the entire compressible ellipsoid, 
345: yields an evolution equation
346: for $K$:
347: \begin{equation} \label{kdot}
348: U \frac{d \ln K}{dt} = {\langle T \rangle} \frac{dS}{dt} 
349: = \Gamma_{\rm vis} - \Lambda_{\rm cool}.
350: \end{equation}
351: In eqn.~(\ref{kdot}), $\langle T \rangle$ is the mass-averaged (mean) temperature defined by 
352: \begin{equation} \label{temp}
353: \langle T \rangle = \frac{1}{M} \int T dm,
354: \end{equation}
355: $S = sM$ is the total entropy, $M$ is the total mass, and $U$ is
356: the total internal energy of the star, given by 
357: \begin{equation} \label{u}
358: U = \int \left ( \varepsilon/ \rho \right) dm = k_1 K \rho_c^{1/n} M , 
359: \end{equation}
360: (eqn. 2.9 in LRS 1994), where
361: $k_1$ is a dimensionless polytropic structure constant of order unity 
362: which depends on the polytropic
363: index $n$ ($ k_1 \sim n$) and $\rho_c$ is the central density.
364: As a simple example, for the case of an isentropic
365: ellipsoid described by a Maxwell-Boltzmann equation of state, we have
366: \begin{eqnarray} \label{maxb}
367: P &=& \rho k_B T/\mu, ~~~~ K = \rho_c^{1-\Gamma} k_B T_c/\mu, \nonumber \\
368: U &=& k_1 k_B T_c M/\mu,  ~~~~ \langle T \rangle = k_1 (\Gamma-1) T_c, \\
369: S &=& U \ln K /\langle T \rangle = \frac{ k_B}{\Gamma - 1} \nonumber
370: \frac{M}{\mu} \ln K.
371: \end{eqnarray}
372: In eqn.~(\ref{maxb}) $k_B$ is Boltzmann's constant, $\mu$ is the mean
373: molecular weight and the subscript `c'
374: denotes that the quantity is evaluated at the center of the star.
375: 
376: The quantity  $\langle T \rangle dS/dt$ in eqn.~(\ref{kdot}) 
377: is the {\it net} heating
378: rate of the star, determined by the difference between the 
379: heating rate due to viscosity, $\Gamma_{\rm vis}$, and 
380: the cooling rate, $\Lambda_{\rm cool}$,
381: due to radiation losses. The viscous heating rate is given by
382: \begin{equation} \label{heat}
383: \Gamma_{\rm vis} = \int \sigma_{ij}u_{i,j} d^3 x = -{\cal W},
384: \end{equation}
385: where $u_i$ are the components of the stellar velocity,
386: $\sigma_{ij}$ is the viscous stress tensor and
387: ${\cal W}$ is the viscous dissipation rate. In eqn.~(4.6) of 
388: LRS (1994) the quantity
389: ${\cal W}$ is calculated analytically in terms of the principal axes of the
390: ellipsoid $a_i$, their velocities $\dot a_i$ and the angular frequency
391: of the internal fluid motion $\Lambda$ ($\propto$ vorticity), and is proportional 
392: to the mass-averaged
393: shear viscosity ${\langle \nu \rangle}$ defined by
394: \begin{equation} \label{vis}
395: \langle \nu \rangle = \frac{1}{M} \int \nu dm .
396: \end{equation}
397: Following Press and Teukolsky (1973) we parametrize $\langle \nu \rangle$ according
398: to $\langle \nu \rangle = C_{\rm visc}(\pi G \bar \rho)^{1/2}R^2$, where $R = (a_1 a_2 a_3)^{1/3}$
399: is the mean radius of the ellipsoid, $\bar \rho = M/(4\pi R^3/3)$ is the mean
400: density and $C_{\rm visc}$ is a dimensionless constant.
401: Choosing $C_{\rm visc} \lesssim 1$ guarantees that the viscous damping timescale
402: will be longer than the hydrodynamical timescale of the star.
403: 
404: The form of the cooling rate $\Lambda_{\rm cool}$ will depend on the
405: details of the stellar microphysics, but here we shall only assume that
406: $\Lambda_{\rm cool} = 0$ when $K$ is held at its initial value $K(s) = K_0 = K(s_0)$ in
407: the absence of viscosity.
408: Thus, we assume that the fluid would evolve adiabatically on the timescales of interest
409: were the viscosity to be turned off completely.
410: For example, if we take the initial configuration to be a cold, degenerate neutron star
411: then we would have $K=K_0 > 0$ and no cooling or heating in the absence of viscosity
412: (i.e. no emission from a zero-entropy gas).
413: 
414: Setting $K = K_0$ at $t=0$, we shall  
415: treat the subsequent energy evolution in two extreme, opposite
416: physical regimes. In the ``rapid-cooling'' regime the cooling rate
417: greatly exceeds the viscous heating rate for all $K > K_0$. In this case
418: any thermal energy generated by viscosity will be radiated almost 
419: immediately, and the gas will not be heated much above $K_0$. 
420: In the limit of arbitrarily fast cooling,  eqn.~(\ref{kdot}) establishes
421: a balance between heating and cooling, i.e.,
422: $\Gamma_{\rm vis} = \Lambda_{\rm cool}$, and maintains the value of
423: $K$ arbitrarily close to $K_0$. We thus have 
424: \begin{equation} \label{rcool}
425:  \Gamma_{\rm vis} = \Lambda_{\rm cool}, \ \ \ \  K = K_0 ~{\rm = \text{constant}},\ \ \ \ {\rm (\text{``rapid \ cooling''})}.
426: \end{equation}
427: In the ``no-cooling'' regime, the viscous heating rate
428: greatly exceeds the cooling rate, in which case cooling is unimportant and
429: eqn.~(\ref{kdot}) gives
430: \begin{equation} \label{ncool}
431: \frac{d \ln K}{d t } = \Gamma_{\rm vis}/U = -{\cal W}/U > 0, 
432: ~~~~~~{\rm (\small{``no \ cooling''})}.
433: \end{equation}
434: In this case the energy generated by viscosity will go into thermal
435: energy, which will be reflected in the secular increase of $K$ above its
436: initial value $K_0$.
437: 
438: Combining the general energy evolution eqn~(\ref{kdot}) with the
439: dynamical equations for a compressible ellipsoid 
440: (LRS 1994, eqns. 4.10- 4.14) yields a conserved energy integral:
441: \begin{equation} \label{cons}
442: E(t) + \int_{0}^{t} \Lambda_{\rm cool} dt 
443: = \ E(0) =\ {\rm constant},
444: \end{equation}
445: where the total energy $E(t)$ is given by
446: \begin{equation} \label{E}
447: E(t) = U + W + \mathcal{T}.
448: \end{equation}
449: In the ``rapid-cooling'' regime the total energy $E(t)$ of the star will
450: decrease secularly by exactly the amount of energy dissipated by
451: viscosity, according to eqns.~(\ref{rcool}) and ~(\ref{cons}).
452: This is {\it always} the situation in 
453: incompressible stars, since they
454: contain no internal energy and hence must lose energy whenever there is
455: viscous dissipation. In the ``no-cooling'' regime, by contrast, 
456: the energy $E(t)$ will be
457: strictly conserved, as the differential rotational
458: kinetic energy dissipated by viscosity is entirely converted into internal
459: energy and is not lost to the star. Compressible stars can
460: evolve in either of these extreme opposite regimes, as well as in 
461: intermediate regimes, depending on the physical conditions.
462: In all cases, the mass and angular momentum of a star is conserved in
463: the presence of viscosity, but circulation is not. These quantities are
464: all monitored in our evolution calculations.
465: 
466: In reporting the results of our calculations below, it will be
467: useful to define the following nondimensional
468: ratios for energy and angular momentum: 
469: \begin{equation} \label{dim}
470: \bar E = \frac{E}{G M^2/R_0}, ~~~\bar J = \frac{J}{(G M^3 R_0)^{1/2}},
471: \end{equation}
472: where
473: \begin{eqnarray} \label{rad}
474: R_0 & = & \xi_1 \left(\xi_1^2 |\theta'_1|\right)^{-(1-n)/(3-n)} 
475: \left(\frac{M}{4 \pi}\right)^{(1-n)/(3-n)} \nonumber \\
476: & & \times \ \left[\frac{(n+1) K_0}{4 \pi G}\right]^{n/(3-n)}
477: \end{eqnarray}
478: is the radius of a nonrotating {\it spherical} polytrope
479: of mass $M$, polytropic entropy parameter $K_0$ and polytropic index $n$.
480: The dimensionless quantities $\theta_1$ and $\xi_1$ are the familiar
481: Lane-Emden variables at the surface of a polytrope
482: (see, e.g. Chandrasekhar 1939).
483: 
484: \section{Secular Evolution of Unstable Spheroids}
485: 
486: We now determine the secular evolution of compressible ellipsoids with
487: viscosity in the ``rapid-cooling'' and ``no-cooling'' regimes.
488: Our initial configurations consist of compressible Maclaurin
489: spheroids that are given small, but finite, 
490: nonaxisymmetric (triaxial) perturbations and
491: evolved until they settle down to a final equilibrium state.
492: Typical results for secularly unstable (but dynamically stable) spheroids 
493: are shown in Fig. 1. The parameters $\{ a_i \}$ are the semi-major axes of the configuration, with
494: $a_3$ along the rotation axis; in axisymmetry $a_1 = a_2$.
495: For the models plotted in this figure we took
496: an $n=1$ compressible Maclaurin spheroid 
497: with eccentricity $e = (1 - a_3^2/a_1^2)^{1/2} = 0.94$ 
498: and $\mathcal{T}/|W| = 0.252$. We increased the equatorial axis $a_1$ slightly 
499: to induce a perturbation, setting $a_2/a_1 = 0.999$ at the 
500: start of the evolution. The viscosity parameter was chosen to be
501: $C_{\rm vis}=0.15$;  however, as long as the secular timescale is longer than
502: the dynamical timescale (i.e. as long as $C_{\rm vis}$ is sufficiently small), 
503: the secular (viscous) evolution track is independent of $C_{\rm vis}$ 
504: provided we ``rescale'' the evolution time according to 
505: $t \propto 1/C_{\rm vis}$. For both the ``rapid-cooling'' and
506: ``no-cooling'' cases we find that the evolution proceeds quasi-statically 
507: through a sequence of compressible, differentially rotating, Riemann-S equilibrium ellipsoids, finally
508: arriving at a uniformly rotating, compressible, Jacobi ellipsoid. While the initial spheroid
509: is unstable in either cooling regime, 
510: both the nonlinear evolutionary track and the final Jacobi equilibrium 
511: ellipsoids are different in the two cases. 
512: 
513: %Place Figure one here:
514: %This is the ApJ figure environment.
515: \begin{figure}
516: \plotone{f1.ps}
517: \caption{Viscous evolution of a secularly unstable, uniformly rotating Maclaurin spheroid
518: with $e=0.94$ to a stable, uniformly rotating, Jacobi ellipsoid. The configuration is a compressible polytrope of
519: index $n=1$ that is given a slight nonaxisymmetric perturbation at $t=0$.  The solid lines correspond to
520: ``rapid cooling'' and the dotted lines to ``no cooling''. Time is measured in units of the initial rotation period
521: of the spheroid, $P$, and the semi-major axes $a_i$ are measured in units of $R_0$ defined by eqn.~(\ref{rad}).
522: \label{fig1}}
523: \end{figure}
524: 
525: The difference in evolution in the two cases 
526: can be appreciated by examining Fig. 2, which plots 
527: isentropic (i.e. constant $K$) 
528: equilibrium curves for two sequences of compressible, $n=1$, Maclaurin spheroids 
529: of different entropy, and the corresponding Jacobi ellipsoids that branch off from these sequences. 
530: The initial spheroid for the evolution
531: plotted in Fig. 1 is indicated by the
532: solid dot on the $K=K_0$ Maclaurin curve beyond the 
533: bifurcation point.  For the case with ``rapid cooling'', 
534: eqn.~(\ref{rcool}) applies; accordingly, $M$, $J$ and $K$ all remain constant 
535: with time while the
536: total energy of the configuration decreases, as the energy dissipated by
537: viscosity escapes. The star thus evolves vertically downward in the
538: figure to a Jacobi ellipsoid with lower energy $E$ but
539: with the same angular momentum $J$ and entropy parameter $K=K_0$ as the initial configuration (solid triangle). 
540: By contrast, for the case with``no cooling'', eqn.~(\ref{ncool})
541: applies; thus, while $M$ and $J$ are again conserved, 
542: the entropy parameter $K$ increases with time, since the 
543: energy dissipated by viscosity is retained and heats up the star.
544: Consequently, the star evolves at constant $J$ {\it and} $E$ 
545: and does not move at all from its initial position in Fig. 2. However,
546: the configuration does change -- it evolves to the Jacobi ellipsoid which bifurcates from a
547: Maclaurin sequence with a higher entropy.  
548: The final Jacobi ellipsoid in this case has the
549: same $E$ and $J$ as the initial Maclaurin spheroid, but has a higher entropy parameter
550: $K > K_0$. For the example illustrated in Fig 1., $K/K_0 = 1.197 > 1$
551: for this final ellipsoid. We note that unstable {\it incompressible} spheroids
552: always evolve in the ``rapid-cooling'' regime, since they have no
553: internal energy and cannot retain viscous dissipation as heat. 
554: 
555: %Place Figure 2 here
556: \begin{figure}
557: \plotone{f2.ps}
558: \caption{Energy as a function of angular momentum for two isentropic, equilibrium sequences of compressible spheroids
559: (solid lines) and the compressible Jacobi ellipsoids which branch off from them (dotted lines). The 
560: lower sequences have an entropy constant $K_0$, while the upper sequences have $K = 1.197 K_0$. The initial unstable
561: spheroid evolved in Fig 1. is indicated by the solid dot. For evolution with ``rapid cooling'' it evolves
562: vertically downward to a Jacobi ellipsoid of lower energy (solid triangle). For evolution with
563: ``no cooling'' the star does not move in the figure, but it evolves to a triaxial ellipsoid on the equilibrium
564: branch with higher entropy, $K > K_0$. Although viscosity drives the instability, the final states are independent of its magnitude.
565: \label{fig2}}
566: \end{figure}
567: 
568: Performing numerous simulations of this type, we conclude that compressible 
569: Maclaurin spheroids are unstable to the {\it secular} bar-mode instability
570: whether or not they reside in the ``rapid-cooling'' or ``no-cooling''
571: regimes. Moreover, we find numerically that the onset of 
572: the instability coincides exactly with the bifurcation point along the 
573: compressible, isentropic Maclaurin curve, as in the case of incompressible spheroids. 
574: In the compressible ellipsoid approximation,
575: this point is located at 
576: $e=0.8127$ and $\mathcal{T}/|W| = 0.138$, independent of $n$ (LRS 1993).
577: However, the evolutionary tracks and final
578: Jacobi equilibrium configurations do depend on the cooling regime,
579: as described above. 
580: 
581: %For completeness, we note that 
582: %the onset of the {\it dynamical} bar instability also does not 
583: %depend on the cooling regime.  The dynamical
584: %mode sets in at $e=0.9529$ and $T/|W| = 0.274$, again independent
585: %of $n$ in this compressible ellipsoidal model (LRS 1993).
586: 
587: Finally, we recall that sequences of uniformly rotating polytropes
588: constructed from the exact (as opposed to the ellipsoidal)
589: equations of hydrostatic equilibrium do not
590: have a point of bifurcation when $n \geq 0.808$. We thus expect that
591: the results found here for compressible ellipsoids also apply to exact, uniformly rotating
592: polytropes, but only for configurations with $n \leq 0.808$. But they may also
593: apply to a wide class of differentially rotating polytropes, provided  $\mathcal{T}/|W| \gtrsim 0.14$.
594: 
595: \subsection{Determining the Final Equilibrium State}
596: 
597: Knowing that an unstable, compressible Maclaurin spheroid
598: evolves to a compressible ellipsoid allows us to
599: determine the final equilibrium state without
600: having to track the secular evolution. Once  
601: the appropriate cooling regime is assigned, the final state
602: is uniquely determined by the location of the initial spheroid 
603: along the Maclaurin equilibrium curve, as was illustrated in Fig. 2. 
604: Not only can we determine the size and shape of the final ellipsoid, and 
605: all other global parameters that characterize the final
606: equilibrium configuration, but we also can 
607: calculate its final thermodynamic state, i.e, its
608: entropy parameter $K$, independent of the magnitude of the viscosity.
609: It is the set of conserved quantities that allows us to determine
610: the final state uniquely, given the initial model.
611: 
612: %Place table 1 here.
613: \begin{deluxetable*}{cccccccccc}
614: \tablecaption{Endpoint Jacobi States for $n=1~^{\rm a}$}
615: \tablewidth{8.0in}
616: \tablehead{
617: \multicolumn{4}{c}{Initial Maclaurin Spheroid} & 
618: \colhead{} &
619: \multicolumn{5}{c}{Final Jacobi Ellipsoids} \\
620: \colhead{$e~^{\rm b}$} & \colhead{$\mathcal{T}/|W|$} & \colhead{$\bar J$} & 
621: \colhead{$\bar E$} & 
622: \colhead{} &
623: \colhead{$a_2/a_1$} & \colhead{$a_3/a_1$} &
624: \colhead{$R/R_0$} & \colhead{$\bar E$} & \colhead{$K/K_0$}}
625: \startdata
626: ~~0.8127*     & 0.138         & 0.299        & -0.382       &  & 1.000        & 0.583        & 1.190  & -0.382 & 1.000  \\
627:               &               &              &              &  & 1.000        & 0.583        & 1.190  & -0.382 & 1.000  \\
628: 0.8200        & 0.142         & 0.307        & -0.378       &  & 0.794        & 0.515        & 1.195  & -0.378 & 1.000  \\
629:               &               &              &              &  & 0.795        & 0.515        & 1.196  & -0.378 & 1.000  \\
630: 0.8400        & 0.154         & 0.328        & -0.366       &  & 0.633        & 0.449        & 1.213  & -0.366 & 1.000  \\
631:               &               &              &              &  & 0.635        & 0.450        & 1.215  & -0.366 & 1.004  \\
632: 0.8600        & 0.168         & 0.353        & -0.352       &  & 0.537        & 0.402        & 1.234  & -0.353 & 1.000  \\
633:               &               &              &              &  & 0.539        & 0.403        & 1.238  & -0.352 & 1.008  \\
634: 0.8800        & 0.184         & 0.382        & -0.336       &  & 0.462        & 0.362        & 1.258  & -0.339 & 1.000  \\
635:               &               &              &              &  & 0.468        & 0.366        & 1.274  & -0.336 & 1.029  \\
636: 0.9000        & 0.203         & 0.418        & -0.317       &  & 0.398        & 0.324        & 1.286  & -0.324 & 1.000  \\
637:               &               &              &              &  & 0.407        & 0.330        & 1.318  & -0.317 & 1.058  \\
638: 0.9200        & 0.225         & 0.463        & -0.295       &  & 0.342        & 0.288        & 1.322  & -0.306 & 1.000  \\
639:               &               &              &              &  & 0.354        & 0.296        & 1.381  & -0.295 & 1.107  \\
640: 0.9400        & 0.252         & 0.524        & -0.266       &  & 0.287        & 0.250        & 1.369  & -0.284 & 1.000  \\
641:               &               &              &              &  & 0.304        & 0.262        & 1.479  & -0.266 & 1.197  \\
642: ~~~0.9529**   & 0.274         & 0.578        & -0.244       &  & 0.249        & 0.222        & 1.410  & -0.266 & 1.000  \\
643:               &               &              &              &  & 0.276        & 0.242        & 1.577  & -0.244 & 1.303  \\
644: 0.9600        & 0.288         & 0.615        & -0.229       &  & 0.232        & 0.209        & 1.437  & -0.256 & 1.000  \\
645:               &               &              &              &  & 0.257        & 0.228        & 1.649  & -0.229 & 1.383  \\
646: 0.9800        & 0.340         & 0.789        & -0.173       &  & 0.169        & 0.157        & 1.561  & -0.214 & 1.000  \\
647:               &               &              &              &  & 0.206        & 0.188        & 2.050  & -0.173 & 1.924  \\
648: 0.9900        & 0.381         & 0.989        & -0.129       &  & 0.128        & 0.122        & 1.692  & -0.180 & 1.000  \\
649:               &               &              &              &  & 0.179        & 0.166        & 2.647  & -0.129 & 2.973  \\
650: 0.9950        & 0.413         & 1.221        & -0.094       &  & 0.093        & 0.091        & 1.844  & -0.148 & 1.000  \\
651:               &               &              &              &  & 0.157        & 0.148        & 3.477  & -0.094 & 4.806  \\
652: 0.9990        & 0.458         & 1.892        & -0.045       &  & 0.050        & 0.050        & 2.231  & -0.095 & 1.000  \\
653:               &               &              &              &  & 0.139        & 0.132        & 6.956  & -0.045 & 17.82~~  \\
654: 1.~~~~~~      & 0.5~~~        & $\infty$     &  0.~~~~      &  & 0.~~~~       & 0.~~~~       & $\infty$ & 0.~~~~  & 1.~~~~  \\
655:               &               &              &              &  & 0.~~~~       & 0.~~~~       & $\infty$ & 0.~~~~  & $\infty$~~~ \\
656: \enddata
657: 
658: \tablenotetext{a}{$\bar J$, $\bar E$, and $R_0$ 
659: are defined in eqns.~(\ref{dim}) and ~(\ref{rad}); $R=(a_1 a_2 a_3)^{1/3}$}
660: 
661: \tablenotetext{b}{One asterisk marks the secular instability point, two the dynamical instabilty point.}
662: \end{deluxetable*}
663: 
664: 
665: First consider the determination of  the final state of an unstable
666: spheroid evolving in the ``rapid-cooling'' regime. The configuration
667: evolves at fixed $M$, $J$ and $K = K_0$. Thus the following 
668: dimensionless ratio is also constant:
669: \begin{equation} \label{final0}
670: \frac{1}{\kappa_n} \left( 1-n/5 \right) \left(\frac{J^2}{M^3 R_0} \right)
671: = {\rm constant} = \hat J^2 \hat R^{(n/(3-n))}.
672: \end{equation} 
673: Here $\kappa_n$ is a polytropic structure constant of order unity
674: which depends on the polytropic index $n$ (see LRS 1993 eqn. 3.14 and
675: Table 1).
676: The term on the right-hand side of eqn.~(\ref{final0}) is expressed
677: in terms of the universal dimensionless parameters $\hat J$ and
678: $\hat R$ each of which, like $\mathcal{T}/|W|$,  uniquely specifies a 
679: configuration along 
680: either a compressible Maclaurin {\it or} a Jacobi sequence, independent of $n$.
681: These parameters
682: were introduced by LRS (1993), (see their eqn (3.27)) where they are 
683: defined according to
684: \begin{eqnarray} \label{Jhat}
685: \hat J^2 &=& \frac{1}{\kappa_n} \left( 1-n/5 \right) 
686: \left(\frac{J^2}{M^3 R} \right), \\
687: \hat R &=& \left( \frac{R}{R_0} \right) ^{(3-n)/n},  \nonumber
688: \end{eqnarray}
689: where $R=(a_1 a_2 a_3)^{1/3}$. These universal parameters
690: are tabulated along a Maclaurin sequence in Table 2 and
691: along a Jacobi sequence in Table 4 of their paper. The significance
692: of the product on the right-hand side of eqn.~(\ref{final0}) is
693: that (1) it uniquely specifies a model along either a Maclaurin or
694: Jacobi sequence, and (2) once the initial Maclaurin model is specified,  
695: this product (unlike $\hat J$ or $\hat R$ separately) 
696: remains constant during ``rapid-cooling''  evolution and thereby
697: uniquely determines the final Jacobi model.
698: 
699: 
700: %Place table 2 here
701: \begin{deluxetable*}{cccccccccc}
702: \tablecaption{Endpoint Jacobi States for $n=0.5~^{\rm a}$}
703: \tablewidth{6.0in}
704: \tablehead{
705: \multicolumn{4}{c}{Initial Maclaurin Spheroid} & 
706: \colhead{} &
707: \multicolumn{5}{c}{Final Jacobi Ellipsoids} \\
708: \colhead{$e~^{\rm b}$} & \colhead{$\mathcal{T}/|W|$} & \colhead{$\bar J$} & 
709: \colhead{$\bar E$} & 
710: \colhead{} &
711: \colhead{$a_2/a_1$} & \colhead{$a_3/a_1$} &
712: \colhead{$R/R_0$} & \colhead{$\bar E$} & \colhead{$K/K_0$}}
713: \startdata
714:  ~~0.8127*    & 0.138         & 0.299        & -0.450       &  & 1.000        & 0.583        & 1.072   & -0.450 & 1.000  \\
715:               &               &              &              &  & 1.000        & 0.583        & 1.072   & -0.450 & 1.000  \\
716: 0.8200        & 0.142         & 0.306        & -0.446       &  & 0.796        & 0.516        & 1.074   & -0.446 & 1.000  \\
717:               &               &              &              &  & 0.796        & 0.516        & 1.074   & -0.446 & 1.000  \\
718: 0.8400        & 0.154         & 0.325        & -0.435       &  & 0.637        & 0.451        & 1.080   & -0.436 & 1.000  \\
719:               &               &              &              &  & 0.638        & 0.451        & 1.082   & -0.435 & 1.007  \\
720: 0.8600        & 0.168         & 0.347        & -0.423       &  & 0.541        & 0.405        & 1.087   & -0.423 & 1.000  \\
721:               &               &              &              &  & 0.542        & 0.405        & 1.088   & -0.423 & 1.003  \\
722: 0.8800        & 0.184         & 0.373        & -0.408       &  & 0.469        & 0.366        & 1.095   & -0.412 & 1.000  \\
723:               &               &              &              &  & 0.474        & 0.369        & 1.107   & -0.408 & 1.057  \\
724: 0.9000        & 0.203         & 0.403        & -0.391       &  & 0.406        & 0.329        & 1.105   & -0.397 & 1.000  \\
725:               &               &              &              &  & 0.415        & 0.334        & 1.128   & -0.391 & 1.115  \\
726: 0.9200        & 0.225         & 0.441        & -0.370       &  & 0.350        & 0.293        & 1.116   & -0.381 & 1.000  \\
727:               &               &              &              &  & 0.362        & 0.302        & 1.158   & -0.370 & 1.217  \\
728: 0.9400        & 0.252         & 0.489        & -0.344       &  & 0.297        & 0.257        & 1.130   & -0.361 & 1.000  \\
729:               &               &              &              &  & 0.315        & 0.269        & 1.206   & -0.344 & 1.411  \\
730: ~~~0.9529**   & 0.274         & 0.531        & -0.322       &  & 0.264        & 0.233        & 1.142   & -0.346 & 1.000  \\
731:               &               &              &              &  & 0.285        & 0.248        & 1.254   & -0.322 & 1.647  \\
732: 0.9600        & 0.288         & 0.559        & -0.308       &  & 0.244        & 0.218        & 1.150   & -0.336 & 1.000  \\
733:               &               &              &              &  & 0.268        & 0.236        & 1.290   & -0.308 & 1.849  \\
734: 0.9800        & 0.340         & 0.680        & -0.253       &  & 0.185        & 0.171        & 1.183   & -0.299 & 1.000  \\
735:               &               &              &              &  & 0.220        & 0.200        & 1.482   & -0.253 & 3.377  \\
736: 0.9900        & 0.381         & 0.805        & -0.206       &  & 0.144        & 0.137        & 1.215   & -0.267 & 1.000  \\ 
737:               &               &              &              &  & 0.192        & 0.176        & 1.742   & -0.206 & 7.058  \\
738: 0.9950        & 0.413         & 0.935        & -0.167       &  & 0.118        & 0.113        & 1.246   & -0.241 & 1.000  \\ 
739:               &               &              &              &  & 0.176        & 0.163        & 2.083   & -0.167 & 16.45~~~  \\
740: 0.9990        & 0.458         & 1.254        & -0.104       &  & 0.074        & 0.073        & 1.316   & -0.190 &  1.000 \\
741:               &               &              &              &  & 0.151        & 0.142        & 3.188   & -0.104 & 128.6~~~~~ \\
742: 1.~~~~~~      & 0.5~~~        & $\infty$     &  0.~~~~      &  & 0.~~~~       & 0.~~~~       & $\infty$ & 0.~~~~  & 1.~~~~  \\
743:               &               &              &              &  & 0.~~~~       & 0.~~~~       & $\infty$ & 0.~~~~  & $\infty$~~~ \\
744: \enddata
745: 
746: \tablenotetext{a}{$\bar J$, $\bar E$, and $R_0$ are defined in eqns.~(\ref{dim}) and ~(\ref{rad}); $R=(a_1 a_2 a_3)^{1/3}$}
747: 
748: \tablenotetext{b}{One asterisk marks the secular instability point, two the dynamical instabilty point.}
749: 
750: \end{deluxetable*}
751: 
752: 
753: Next consider how the final state of an unstable
754: spheroid evolving in the ``no-cooling'' regime is determined. In this
755: case the configuration
756: evolves at fixed $M$, $J$ and $E$, while $K$ increases. Suppose we
757: guess the final value of $K$, setting $K = K^{\rm guess} > K_0$.
758: Then from eqn.~(\ref{rad}) this guess yields a new value for
759: $R_0$ given by
760: \begin{equation} \label{rguess}
761: R^{\rm guess}_0 = R_0 \left( \frac{K}{K_0} \right)^{n/(3-n)}.
762: \end{equation}
763: Proceeding as in the previous case, we use $R_0^{\rm guess}$ in 
764: eqn.~(\ref{final0}) to determine the product 
765: [$\hat J^2 \hat R^{(n/(3-n))}]_{\rm guess}$. This parameter specifes
766: a compressible ellipsoid. In general, however, the total energy
767: of the ellipsoid will not be equal to the energy of the initial,
768: unstable spheroid. Improved guesses of $K^{\rm guess}$ are required
769: to guarantee energy conservation. After a few iterations, the process
770: converges to the desired higher entropy, compressible 
771: ellipsoid with the same energy as the original spheroid.
772: 
773: We have applied the above prescription to determine in the compressible 
774: ellipsoidal approximation the endpoint
775: states of all unstable $n=1$ spheroids beyond the point of bifurcation.
776: Results are listed in Table 1. For each unstable spheroid, the top
777: row applies to the ``rapid-cooling'' regime, while the bottom row
778: applies to the ``no-cooling'' regime. We note that the two endpoint states 
779: derived for the unstable spheroid with
780: $e = 0.94$ are consistent with the results found by tracking the full
781: secular evolution of this model (slightly perturbed), as shown in Figs 1 and 2.
782: To assess the dependence on the polytropic index of the star, we determine
783: the endpoint states for unstable spheroids with a different polytropic index, 
784: $n=0.5$, in Table 2.
785: For evolution with ``rapid cooling'', stars with higher $n$ 
786: (i.e. softer equations of state and greater central concentration)
787: lose greater fractional energy than stars
788: with the same initial eccentricity (i.e. $\mathcal{T}/|W|$), but smaller $n$. For evolution with ``no cooling'', stars with 
789: higher $n$ experience less entropy increase (as measured by the 
790: increase in $K$) than stars
791: with smaller $n$. The reason for the entropy dependence is that, 
792: according to eqn.~(\ref{u}),  $U \propto k_1 K$ and
793: $k_1 \sim n$, so that a higher value of $K$ is necessary to heat a star
794: to the same $U$ for smaller $n$. Finally, we note that
795: the expansion of an unstable star evolving with ``no cooling'' 
796: in comparison with one evolving with ``rapid cooling'' (measured by the
797: ratio of their mean radii $R$) is larger for larger $n$, for the same initial
798: eccentricity.
799: 
800: \section{Point of Onset of the Bar-Mode Instability}
801: 
802: Our numerical integrations of the ellipsoidial dynamical equations with
803: viscosity show that the point of onset of the secular bar-mode 
804: instability in compressible spheroids 
805: coincides with the bifurcation point, where $\mathcal{T}/|W| = 0.1375$, 
806: independent of the polytropic index $n$ {\it and} 
807: the cooling regime. LRS (1993) provided a
808: rigorous proof of the first part of this assertion: they  
809: took the entropy parameter $K$ to be strictly constant, effectively assuming
810: ``rapid cooling'',  and 
811: proved that the point of secular instability coincides with the 
812: bifurcation point for all $n$ (see Section 6.2 of LRS 1993).
813: We now sketch how this proof can be generalized to 
814: include the ``no-cooling'' regime as well.
815: 
816: 
817: 
818: 
819: The proof of LRS is based on an energy variational
820: principle and an energy functional for compressible ellipsoids.
821: The functional is of the form 
822: \begin{equation} \label{efunc}
823: E = E(\rho_c, \lambda_1, \lambda_2; M, J, K) = U + W + \mathcal{T},
824: \end{equation}
825: where $U$, $W$ and $\mathcal{T}$ are the internal, gravitational potential and
826: rotational kinetic energies of a compressible ellipsoid, not necessarily
827: in equilibrium. The ellipsoid of index $n$ is  
828: parametrized by the
829: central density $\rho_c$ and the two oblateness parameters
830: $\lambda_1 \equiv (a_3/a_1)^{2/3}$ and $\lambda_2 \equiv (a_3/a_2)^{2/3}$.
831: The equilibrium configuration of fixed $M$, $J$ and $K$ (i.e. fixed entropy) 
832: is determined by extremizing the energy according to
833: \begin{equation} \label{equil}
834: \frac{\partial E}{\partial \alpha_i} = 0, ~~~~i = 1 - 3,
835: \end{equation}
836: where $\{ \alpha_i \} = \{ \rho_c, \lambda_1, \lambda_2 \}$. 
837: Solving these equations simultaneously yields equilibrium relations
838: for compressible Jacobi ellipsoids. Stability requires that an
839: equilibrium configuration correspond to a true {\it minimum} of the
840: total energy, that is, that all eigenvalues of the matrix 
841: $\left ( \partial^2 E / \partial \alpha_i \partial \alpha_j \right)_{\rm eq}$ 
842: be positive. The onset of instability along any one-parameter sequence
843: of equilibrium configurations can be determined from the condition
844: \begin{equation} \label{det}
845: {\rm det} \left ( \frac{\partial ^2 E}{\partial \alpha_i \partial \alpha_j} \right )_{\rm eq}
846: = 0, ~~~~i,j = 1-3 ~.
847: \end{equation}
848: When this condition is first satisfied along the sequence, one of the
849: eigenvalues must change sign. By using the energy functional
850: for a compressible Jacobi ellipsoid given by eqn.~(\ref{efunc}), but
851: evaluating the determinant along the equilibrium Maclaurin sequence
852: (for which $\lambda_1 = \lambda_2$), LRS found that compressible
853: spheroids become secularly unstable to triaxial deformations at the
854: bifurcation point, where $\mathcal{T}/|W| = 0.1375$, independent of $n$.
855: 
856: Now suppose we allow for variations in which the entropy and entropy parameter $K$ is 
857: not fixed, but is allowed to increase, as is the case for evolution with ``no cooling''. 
858: In this situation, the relevant functional that must be minimized in a variational principle is 
859: the effective {\it free} energy of the configuration, 
860: \begin{equation} \label{free}
861: F = E - \langle T \rangle S,
862: \end{equation}
863: where the entropy, or equivalently, the entropy parameter $K$, must now be 
864: treated as another parameter 
865: that is allowed to vary:
866: $E = E(\rho_c, \lambda_1, \lambda_2, K; M, J)$. 
867: The parameter $K$ appears explicitly
868: in $E$ only through the internal energy $U$, given by eqn.~(\ref{u}). However,
869: eqn.~(\ref{kdot}) shows that the free energy $F$ does not change when $K$ changes,
870: while holding the other parameters fixed:
871: \begin{equation} \label{dF}
872: \frac{\partial F}{\partial K}  =  \frac{\partial E}{\partial K} - \langle T \rangle \frac{\partial S}{\partial K} 
873:             =  \frac {U}{K} - U \frac {\partial \ln K}{\partial K}  
874:             =  0 ~. 
875: \end{equation}
876: Similarly,
877: \begin{equation} \label{ddF}
878: \frac {\partial^2 F}{\partial \alpha_i \partial K} = 0 = 
879: \frac {\partial^2 F}{\partial K^2} = 0 , ~~~~i = 1-3 ~.
880: \end{equation} 
881: Hence the set of parameters $\{ \alpha_i \}$
882: found to extremize $E$ and identify its local minimum when holding $K$ fixed 
883: is the same set which extremizes $F$ and identifies its local minimum 
884: as $K$ is allowed to vary.  This set of parameters determines the onset of instability. 
885: Thus, compressible spheroids become secularly unstable 
886: to triaxial deformations at the
887: bifurcation point, where $\mathcal{T}/|W| = 0.1375$, independent of $n$
888: {\it and} independent of the degree of cooling.
889: 
890: \section{Summary and Future Work}
891: 
892: We have analyzed the secular bar instability driven by viscosity in rapidly rotating,
893: Newtonian stars. We have adopted the formalism of LRS, who model rotating stars as
894: compressible ellipsoids governed by a polytropic equation of state. However, we have extended 
895: their treatment to incorporate changes in the fluid entropy as the star evolves:
896: the gas is allowed to heat up by viscous dissipation and cool down by emission. We treated the
897: stars in two extreme opposite limits: one in which the cooling timescale is very much shorter than the
898: heating timescale (``rapid cooling'') and the other in which the timescale inequality is reversed (``no cooling''). 
899: Our numerical integrations of the ellipsoidal evolution equations show that,
900: in both limits, a uniformly rotating star spinning sufficiently rapidly is secularly unstable to
901: nonaxisymmetric perturbations to a bar. The point of onset of the bar instability along an 
902: equilibrium sequence of uniformly
903: rotating, compressible Maclaurin spheroids, parametrized by $\mathcal{T}/|W|$, occurs precisely at the bifurcation point 
904: ($\mathcal{T}/|W| = 0.1375$)  in both regimes. 
905: This numerical finding is confirmed by a stability analysis based on 
906: minimizing the free energy functional of the star. However, the final equilibrium state of an unstable spheroid is different 
907: in the two regimes: for the ``rapid-cooling'' regime, the final configuration is a compressible Jacobi ellipsoid 
908: with the a lower energy but the same entropy as the initial unstable spheroid, while for the ``no-cooling'' regime, 
909: the final state is a Jacobi ellipsoid with the same energy but a higher entropy than the spheroid. The final states can
910: be calculated from the conserved quantities without having to track the evolution of the unstable stars.
911: 
912: As described in the Introduction and in LRS (1993,1994), 
913: rotating compressible ellipsoids mimic many of the same equilibrium and stability properties  
914: of rotating stellar models that obey the exact hydrodynamic equations. Not surprisingly, this concordance
915: is particularly close for stars that are governed by stiff equations of state (e.g. small polytropic
916: indicies $n$) and are not very centrally condensed, since ellipsoids provide exact solutions for homogeneous,
917: incompressible Newtonian stars.  It is therefore relevant
918: to examine the degree to which the new results found in this analysis and summarized above also apply to
919: exact compressible stellar models. We are preparing to study the secular bar-mode instability in 
920: uniformly and differentially rotating, compressible stars by performing numerical simulations  
921: that solve the exact Navier-Stokes equations with viscosity. We will treat the problem 
922: both in Newtonian gravity and in full general relativity. We will follow not only the onset but also the nonlinear 
923: growth of the instability and evolve unstable configurations until they arrive at a final equilibrium state. 
924: A preliminary numerical study that illustrates the numerical approach of tracking 
925: secular (viscous) evolution by performing full hydrodynamical simulations is given in Duez et al. (2004), where 
926: we followed the secular evolution and determined the final fate 
927: of rapidly differentially rotating, ``hypermassive'' neutron stars in general relativity, 
928: both in axisymmetry and in $3+1$ dimensions. Employing a hydrodynamical code to follow secular
929: evolution is necessary, either when the quasi-static configurations
930: traversed along the secular evolutionary track are difficult to construct by solving the hydrostatic
931: equilibrium equations (e.g., due to complicated internal flow patterns), or when the 
932: secular evolution ultimately sends the fluid into rapid hydrodynamical motion (e.g., catastrophic collapse). 
933: Our experience suggests that the technique, though resource intensive, is quite promising, even in general
934: relativity.
935: 
936: \acknowledgments
937: 
938: It is a pleasure to thank M. Duez, Y.T. Liu and B. Stephens for valuable
939: discussions.  This work was supported in part by NSF Grants 
940: PHY-0205155 and PHY-0345151 and NASA Grant NNG04GK54G at the
941: University of Illinois at Urbana-Champaign.
942: 
943: \begin{thebibliography}{}
944: \bibitem{bar77} Bardeen, J. M., Friedman, J. L., Schutz, B. F., \& Sorkin, R. 1977, ApJ, 217, L49.
945: \bibitem{cl1983} Carter, B., \& Luminet, J. P. 1983, A\&A, 121, 97.
946: \bibitem{cl1985} Carter, B., \& Luminet, J. P. 1985, MNRAS, 212, 23.
947: \bibitem{ch39} Chandrasekhar, S. 1939, {An Introduction to the Study of Stellar
948: Structure}, (Dover, New York).
949: \bibitem{ch69} Chandrasekhar, S. 1969, {Ellipsoidal Figures of Equilibrium},
950: (Yale Univ. Press, New Haven).
951: \bibitem{duez04} Duez, M. D., Liu, Y. T., Shapiro, S. L. \& Stephens, B. C. 2004,
952: PRD in press, (astro-ph/0402502). 
953: \bibitem{fs78a} Friedman, J. L., \& Schutz, B. 1978a, ApJ, 221, 937.
954: \bibitem{fs78b} Friedman, J. L., \& Schutz, B. 1978b, ApJ, 222, 281.
955: \bibitem{ifd85} Imamura, J. N., Friedman, J. L. \& Durisen, R. H. 1985, ApJ, 294, 474.
956: \bibitem{il90} Ipser, J. R., \& Lindblom, L. 1990, ApJ, 355, 226.
957: \bibitem{il91} Ipser, J. R., \& Lindblom, L. 1991, ApJ, 373, 213.
958: \bibitem{jam64} James, R. A. 1964, ApJ, 140, 552.
959: \bibitem{lrs93} Lai, D., Rasio, F. A., \& Shapiro, S. L. 1993, ApJS, 88, 205.
960: \bibitem{lrs94} Lai, D., Rasio, F. A., \& Shapiro, S. L. 1994, ApJ, 437, 742.
961: \bibitem{man85} Managan, R. A. 1985, ApJ, 294, 463.
962: \bibitem{ob73} Ostriker, J. P., \& Bodenheimer, P. 1973, ApJ, 180, 171.
963: \bibitem{pt73} Press, W. H., \& Teukolsky, S. A. 1973, ApJ, 181, 513.
964: \bibitem{sha83} Shapiro, S. L., \& Teukolsky, S. A. 1983, 
965: {Black  Holes, White Dwarfs, and Neutron Stars} (Wi1ey interscience, New York).
966: \bibitem{tas78} Tassoul, J-L. 1978 {Theory of Rotating Stars}, (Princeton Univ. Press, Princeton).
967: \end{thebibliography}
968: 
969: \end{document}
970: 
971: 
972: 
973: