1: \documentstyle[12pt,epsf,aasms4]{article}
2: \input{psfig.sty}
3: \begin{document}
4:
5: \title{Convective instability in proto--neutron stars}
6:
7: \author{Juan A. Miralles$^{1)}$, Jos\'e A. Pons$^{1,2)}$ and Vadim A. Urpin$^{3,4)}$}
8:
9: \affil{$^{1)}$ Departament d'Astronomia i Astrof\'{\i}sica, Universitat de
10: Val\`encia, E-46100 Burjassot, Spain \\
11: $^{2)}$ Department of Physics \& Astronomy, SUNY at Stony Brook,
12: Stony Brook, NY 11794-3800, USA \\
13: $^{3)}$ Department of Mathematics, University of Newcastle,
14: Newcastle upon Tyne NE1 7RU, UK \\
15: $^{4)}$ A.F.Ioffe Institute of Physics and Technology,
16: 194021 St. Petersburg, Russia}
17:
18: \begin{abstract}
19:
20: The linear hydrodynamic stability of proto--neutron stars (PNSs) is considered
21: taking into account dissipative processes such as neutrino transport
22: and viscosity. We obtain the general
23: instability criteria which differ essentially from the well-known
24: Ledoux criterion used in previous studies.
25: We apply the criteria to evolutive models of PNSs that,
26: in general, can be subject to the various known regimes such as neutron
27: fingers
28: and convective instabilities. Our results indicate that
29: the fingers instability arises in a more extended region of
30: the stellar volume and lasts a longer time than expected.
31:
32: \end{abstract}
33:
34: \keywords{convection -- instabilities -- stars: neutron}
35: \vspace{1cm}
36:
37: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
38: \newpage
39:
40: \section{Introduction}
41:
42: According to theoretical models, neutron stars are formed in the aftermath
43: of type II and Ib supernova explosions, associated to the gravitational
44: collapse of massive stars (8 to 30$M_{\odot}$) at the end of their lives.
45: A number of authors have recognized
46: that convection in the newly born hot neutron star can play an important
47: role in both enhancing neutrino luminosities and increasing the energy
48: deposition efficiency, which might lead to the explosion of a supernova
49: (Epstein 1979; Livio, Buchler \& Colgate 1980; Colgate \& Petschek 1980;
50: Smarr {\it et al.}, 1981; Lattimer \& Mazurek, 1981;
51: Burrows\& Fryxell 1993; Janka \& Muller 1994).
52: Epstein (1979) pointed out that the negative lepton gradient that
53: naturally arises in the outer layers of the PNS after the
54: shock breaks out of the neutrinosphere can be convectively unstable.
55: Later on, Colgate
56: \& Petschek (1980) suggested that this instability could lead to a complete
57: overturn of the core and a strong enhancement of the neutrino transport needed
58: for a powerful supernova explosion on a dynamical time scale. However,
59: convection in PNSs can be driven not only by the lepton
60: gradient but the entropy gradient as well.
61: The suggestion that dissipation
62: of the shock must result in a negative entropy gradient due to both
63: neutronization and dissociation (Arnett, 1987) has been confirmed in a
64: variety of models by Burrows \& Lattimer (1988). As a matter of fact,
65: despite different equations of state and differences in the consideration
66: of neutrino transport, the development of negative entropy and lepton
67: gradients seems to be common in many simulations of supernova models
68: (Hillebrandt 1987, Bruenn \& Mezzacappa 1994, Bruenn, Mezzacappa \& Dineva
69: 1995) and evolutionary models of PNSs (Burrows \& Lattimer
70: 1986, Keil \& Janka 1995, Sumiyoshi, Suzuki \& Toki 1995, Pons et al.
71: 1999).
72:
73: Concerning numerical simulations, the situation is controversial.
74: Recent hydrodynamic simulations by Keil, Janka \& Muller (1996)
75: demonstrate that convection can arise in a rather extended region of
76: PNSs and may generally last a relatively long time, $>1$s.
77: Bruenn and Mezzacappa (1994), using the mixing length approximation and
78: Mezzacappa et. al (1998) in two dimensional hydrodynamic simulations
79: found only mild convective activity in the region near the neutrinosphere.
80: Note that the two-dimensional hydro code used by Keil, Janka \& Muller
81: is coupled to a radial gray
82: equilibrium diffusion code, that suppresses the neutrino transport in the
83: angular direction, essentially underestimating the stabilizing effect of
84: neutrino transport. On the contrary, Mezzacappa et. al (1998) code works
85: at the limit in which the neutrino transport in the angular direction is
86: fast enough to render the neutrino distributions spherically symmetric,
87: therefore overestimating the stabilizing effect.
88: The answer is obviously at some point in between, that only a future
89: self--consistent multidimensional calculation can determine.
90:
91:
92: Helpful insights into the nature and growth rates of fluid instabilities
93: have been achieved thanks to
94: semi-analytical investigations (Grossman, Narayan
95: \& Arnett, 1993; Bruenn \& Dineva 1996), and this work is in that line.
96: The presence of convection in PNSs is usually argued by the
97: fact that the necessary condition of instability (the Ledoux criterion)
98: is fulfilled in a some fraction of the stellar volume.
99: (Epstein 1979, Livio; Buchler \& Colgate 1980;
100: Keil, Janka \& M\"uller, 1996). However,
101: the Ledoux criterion may have no bearing at all on the physics of convection.
102: If dissipative effects caused by viscosity or
103: energy and lepton transport are taken into account, the Ledoux condition
104: does not apply and has to be substituted by the appropriate criteria.
105: In the present paper, we derived the criteria of instability in
106: PNSs employing the diffusion approximation for neutrino
107: transport (Imshennik \& Nadezhin 1972, Pons et al. 1999) and show that
108: dissipative processes can substantially change the picture.
109:
110:
111: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
112: \section{Basic equations}
113:
114: Consider the condition of instability in a plane-parallel layer
115: between $z=0$ and $z=d$ with the gravity ${\bf g}$ directed in the negative
116: $z$-direction. We neglect a non-uniformity of ${\bf g}$ as well as general
117: relativistic corrections to hydrodynamic equations. Calculations show
118: that initially the instability arises in a surface layer of PNS thus
119: the plane-parallel approximation seems to be accurate, at least, during
120: the initial evolutionary stage and, should give qualitatively correct
121: results during the late stage. We assume that the
122: characteristic cooling time scale of a PNS is much longer
123: than the growth time of instability thus it can be treated in a
124: quasi-stationary approximation. Since convective velocities are
125: typically smaller than the speed of sound, one can describe instability
126: by making use of the standard Boussinesq approximation (see, e.g., Landau
127: \& Lifshitz 1959). We consider the linear instability when the equations
128: governing small perturbations can be obtained by the linearization of
129: hydrodynamic equations. In what follows, small perturbations of hydrodynamic
130: quantities will be marked by a subscript ``1''. The linearized momentum and
131: continuity equations read
132: \begin{equation}
133: \rho \dot{\bf v}_{1} = - \nabla p_{1} + {\bf g} \rho_{1}+ \rho \nu \Delta
134: {\bf v}_{1} ,
135: \label{hyd1}
136: \end{equation}
137: \begin{equation}
138: \nabla \cdot {\bf v}_{1} = 0,
139: \label{hyd0}
140: \end{equation}
141: where $p$ and $\rho$ are the pressure and density, respectively, $\nu$ is
142: the kinematic viscosity. We assume that the matter inside a PNS
143: is in chemical equilibrium thus the density is generally a function of
144: the pressure $p$, temperature $T$ and lepton fraction $Y=(n_{e} +
145: n_{\nu})/n$, with $n_{e}$ and $n_{\nu}$ being the net (particles minus
146: antiparticles) number densities of electrons and neutrinos, respectively, and
147: $n=n_{p}+n_{n}$ is the number density of baryons.
148: Since in the Boussinesq approximation, the perturbations of pressure are
149: negligible because the fluid motions are assumed to be slow and the
150: moving fluid elements are nearly in pressure equilibrium with surroundings,
151: the perturbations of density ($\rho_1$) and entropy per baryon ($s_1$) can
152: be expressed, in terms of the perturbations of temperature ($T_1$) and
153: lepton fraction $Y_1$,
154: \begin{equation}
155: \rho_{1} \approx - \rho ( \beta \frac{T_{1}}{T} + \delta Y_{1} ),
156: \label{rho1}
157: \end{equation}
158: \begin{equation}
159: s_1\approx {m_B c_{p}} \frac{T_{1}}{T} + \sigma Y_{1}
160: \label{t1}
161: \end{equation}
162: where $m_B$ is the mass of the baryon (we neglect the mass difference
163: between protons and neutrons), $\beta$ and $\delta$ are the coefficients
164: of thermal and chemical
165: expansion; $\beta = - (\partial \ln \rho/\partial \ln T)_{pY}$, $\delta =
166: - (\partial \ln \rho/ \partial Y)_{pT}$, $c_{p} = (T/m_B)
167: (\partial s / \partial T)_{pY}$ is the specific heat at
168: constant pressure and $\sigma = (\partial s/ \partial Y)_{pT}$.
169:
170: The above equations should be complemented by the equation driving
171: the evolution of chemical composition and heat balance. We employ
172: the equilibrium diffusion approximation (EDA) which can be reliable
173: during the early stage of evolution of the PNS, when the mean
174: free path of neutrino is short compared to the density and temperature
175: length scales.
176: In this approximation, the diffusion and thermal balance equations
177: read
178: \begin{equation}
179: n \frac{d Y}{d t} = - \nabla \cdot {\bf F},
180: \label{yev}
181: \end{equation}
182: \begin{equation}
183: n T \frac{d s}{d t} + n \mu \frac{d Y}{d t}= - \nabla \cdot {\bf H},
184: \label{sev}
185: \end{equation}
186: where $\mu$ is the neutrino chemical potential and
187: ${\bf F}$ and ${\bf H}$ are the lepton and heat fluxes, respectively.
188: Note that in the equation for the heat balance we have neglected viscous
189: heat production since this term is second order in velocity. In the EDA,
190: these fluxes are given by
191: \begin{equation}
192: {\bf F} = - a_{T} \nabla T - a_{\eta} \nabla \eta,
193: \end{equation}
194: \begin{equation}
195: {\bf H} = - b_{T} \nabla T - b_{\eta} \nabla \eta,
196: \end{equation}
197: where $\eta= \mu/k_B T$ is the degeneracy parameter of the neutrino
198: gas, $k_B$ being the Boltzmann constant; $a_{T}$, $a_{\eta}$ and $b_{T}$,
199: $b_{\eta}$ are the coefficients of diffusion and the thermal conductivities,
200: respectively.
201: These kinetic coefficients can easily be related to the coefficients
202: $D_{2}$, $D_{3}$ and $D_{4}$ introduced by Pons et al. (1999),
203: $a_{T}=\Gamma D_{3}$, $a_{\eta}= T \Gamma D_{2}$, $b_{T}=k_B T \Gamma D_{4}$,
204: $b_{\eta}= k_B T^{2} \Gamma D_{3}$, where
205: $\Gamma = (k_B T)^{2} k_B/ 6 \pi^{2} c^{2} \hbar^{3}$.
206:
207: For hydrodynamic considerations, it is more convenient to express ${\bf F}$
208: and ${\bf H}$ in terms of $\nabla T$ and $\nabla Y$ rather
209: than $\nabla T$ and $\nabla \eta$. Then, neglecting the contribution
210: of the pressure gradient (barodiffusion), we have
211: \begin{equation}
212: {\bf F} = - \left( \chi_{T} \nabla T + \chi_{Y} \nabla Y \right),
213: \label{f1}
214: \end{equation}
215: \begin{equation}
216: {\bf H} = - (\xi_{T} \nabla T + \xi_{Y} \nabla Y),
217: \label{f2}
218: \end{equation}
219: where
220: \begin{eqnarray}
221: \chi_{T} = a_{T}+ a_{\eta} \left(
222: \frac{\partial \eta}{\partial T} \right)_{p,Y} \, , \;\; \chi_{Y} =
223: a_{\eta} \left( \frac{\partial \eta}{\partial Y} \right)_{p,T} \, ,
224: \nonumber \\
225: \xi_{T} = b_{T} + b_{\eta} \left(
226: \frac{\partial \eta}{\partial T} \right)_{p,Y} \, , \;\; \xi_{Y} =
227: b_{\eta} \left( \frac{\partial \eta}{\partial Y} \right)_{p,T} \, .
228: \end{eqnarray}
229: Assuming that
230: unperturbed quantities are homogeneous, equations (\ref{yev}) and
231: (\ref{sev}) can be linearized and written in the form
232: \begin{eqnarray}
233: \dot{Y}_{1} &+& {\bf v}_{1} \cdot \nabla Y =
234: \lambda_{T} \frac{\Delta T_{1}}{T} + \lambda_{Y} \Delta Y_{1}\;,
235: \label{y1}
236: \\
237: \frac{\dot{T}_{1}}{T} &-& {\bf v}_{1} \cdot \frac{\Delta\nabla T}{T} =
238: \kappa_{T} \frac{\Delta T_{1}}{T} + \kappa_{Y} \Delta Y_{1}
239: \label{s1}
240: \end{eqnarray}
241: where
242: \begin{equation}
243: \Delta\nabla T=
244: -\frac{T}{m_B c_p} \left( \nabla s - \sigma \nabla Y \right)
245: = \left( \frac{\partial T}{\partial p} \right)_{s,Y} \nabla p
246: - \nabla T
247: \end{equation}
248: is the superadiabatic
249: temperature gradient that gives the difference between the temperature
250: gradient of a fluid with constant entropy and composition and the
251: actual temperature gradient,
252: and we have introduced the following characteristic conductivities
253: \begin{eqnarray}
254: \kappa_{T} = \frac{1}{m_B c_p} \left[
255: \frac{\xi_{T}}{n} - \lambda_T ( \eta + \sigma ) \right], &\;\;&
256: \lambda_{T} = \frac{T \chi_{T}}{n},
257: \nonumber \\
258: \kappa_{Y} = \frac{1}{m_B c_p} \left[
259: \frac{\xi_{Y}}{n T} - \lambda_{Y} ( \eta + \sigma ) \right], &\;\;&
260: \lambda_{Y} = \frac{\chi_{Y}}{n}.
261: \end{eqnarray}
262:
263: The system formed by equations (\ref{hyd1}), (\ref{hyd0}), (\ref{y1}) and
264: (\ref{s1}), together with the corresponding boundary conditions, completely
265: determines the behaviour of small perturbations. For the sake of
266: simplicity, we consider the case when perturbations are vanishing at the
267: boundaries $z=0$ and $z=d$. Note that other boundary conditions cannot
268: change the main conclusion of our paper qualitatively.
269:
270:
271: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
272: \section{The dispersion equation}
273:
274: The dependence of all perturbations on time and horizontal coordinate
275: can be chosen in the form $\exp(\gamma t - i k x)$, where $\gamma$
276: is the inverse growth (or decay) timescale of perturbations and $k$ is the
277: horizontal wavevector. For such perturbations, we have
278: \begin{equation}
279: \Delta = \frac{d^{2}}{d z^{2}} - k^{2}.
280: \end{equation}
281: The dependence on the vertical coordinate should be obtained from
282: equations (\ref{hyd1}), (\ref{hyd0}), (\ref{y1}) and (\ref{s1}), which
283: can be reduced to one equation of a higher order, say for $v_{1z}$.
284: The coefficients of this equation are constant in our simplified model,
285: therefore the solution for the fundamental mode with ``zero boundary
286: conditions'' can be taken in the form $v_{1z} = \sin (\pi z/ d)$. Then,
287: the dispersion equation for the fundamental mode is
288: \begin{equation}
289: \gamma^{3} + a_{2} \gamma^{2} + a_{1} \gamma + a_{0} = 0,
290: \label{cubic}
291: \end{equation}
292: where
293: \begin{eqnarray}
294: a_{2} = \omega_{\nu} + \omega_{T} + \omega_{Y}, \nonumber \\
295: a_{1} = \omega_{T} \omega_{Y} - \omega_{TY} \omega_{YT} +
296: \omega_{\nu} (\omega_{T} + \omega_{Y}) - (\omega_{g}^{2} +
297: \omega_{L}^{2}), \nonumber \\
298: a_{0} = \omega_{\nu} (\omega_{T} \omega_{Y} - \omega_{TY} \omega_{YT})
299: - \omega_{g}^{2} ( \omega_{Y} - \frac{\delta}{\beta} \omega_{YT})
300: \nonumber \\
301: - \omega_{L}^{2} \left( \omega_{T} - \frac{\beta}{\delta} \omega_{TY} \right).
302: \end{eqnarray}
303: In these expressions, we introduced the characteristic frequencies
304: \begin{eqnarray}
305: \omega_{T} = \kappa_{T} Q^{2}, &\;&
306: \omega_{Y} = \lambda_{Y} Q^{2},
307: \nonumber \\
308: \omega_{YT} = \lambda_{T} Q^{2}, &\;&
309: \omega_{TY} = \kappa_{Y} Q^{2}, \\
310: \omega_{\nu} = \nu Q^{2}, \quad \;
311: \omega_{g}^{2} = \frac{\beta g k^{2}}{Q^{2}}
312: \cdot \frac{(\Delta\nabla T)_z}{T}, &\;&
313: \omega_{L}^{2} = - \frac{\delta g k^{2}}{Q^{2}} \cdot
314: \frac{d Y}{d z}, \nonumber
315: \end{eqnarray}
316: where $(\Delta\nabla T)_z$ is the $z$-component of
317: $\Delta\nabla T$ and $Q^{2} = (\pi/d)^{2} + k^{2}$.
318: The quantities $\omega_{\nu}$, $\omega_{T}$, and $\omega_{Y}$ are the
319: inverse time scales of dissipation of perturbations due to viscosity,
320: thermal conductivity and diffusivity, respectively; $\omega_{YT}$
321: characterizes the rate of diffusion caused by the temperature
322: inhomogeneity (thermodiffusion), and $\omega_{TY}$ describes the influence
323: of the chemical inhomogeneity on the rate of heat conduction;
324: $\omega_{g}$ is the frequency (or, in the case of instability, the
325: inverse growth time) of the buoyant wave; $\omega_{L}$ characterizes
326: the dynamical time scale of the processes associated with the lepton
327: gradient.
328:
329: Equation (\ref{cubic}) describes three essentially different modes which
330: generally exist in a chemically inhomogeneous fluid. The condition that
331: at least one of the roots has a positive real part
332: (unstable) is equivalent to fulfilling one of the following inequalities
333: (see, e.g., Aleksandrov, Kolmogorov \& Laurentiev 1963)
334: \begin{equation}
335: a_{2} < 0, \;\;\; a_{0} < 0, \;\;\; a_{1} a_{2} < a_{0}.
336: \label{3cond}
337: \end{equation}
338: Since $\nu$, $\kappa_T$ and $\lambda_Y$ are positive defined quantities,
339: the first condition $a_2<0$ will never apply, and the discussion
340: will be reduced to the other two conditions.
341:
342: In the particular case of chemically homogeneous plasma with a ``standard''
343: transport ($\omega_{TY} = \omega_{YT} = 0$), we have
344: \begin{equation}
345: (\gamma + \omega_{Y})[(\gamma + \omega_{\nu})(\gamma + \omega_{T}) -
346: \omega_{g}^{2}] = 0.
347: \end{equation}
348: The first root, $\gamma_{1} = - \omega_{Y}$, describes a stable
349: diffusive mode. Two other roots correspond to the ordinary buoyancy
350: modes one of which can be unstable if $\omega_{g}^{2} - \omega_{\nu}
351: \omega_{T}> 0$. Then, the necessary condition of instability reduces
352: to the Schwarzschild condition $(\Delta\nabla T)_z>0$, or in terms
353: of the entropy gradient, $ds/dz < 0$.
354:
355: If $dY/dz \neq 0$ but dissipative effects are negligible
356: ($a_{2}=0$, $a_{0}=0$), we have
357: \begin{equation}
358: \gamma^3 + a_1 \gamma = 0
359: \end{equation}
360: with $a_{1}= -(\omega_{g}^{2} + \omega_{L}^{2})$.
361: The first root is degenerate in this case and the
362: two other roots are given by $\pm \sqrt{-a_{1}}$.
363: The condition of instability is $(- a_{1}) = \omega_{g}^{2} +
364: \omega_{L}^{2} > 0$, or
365: \begin{equation}
366: \left(\frac{\Delta\nabla T}{T}\right)_z - \frac{\delta}{\beta}\frac{dY}{dz}>0,
367: \end{equation}
368: that represents the familiar Ledoux criterion.
369:
370: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
371: \section{Instability criteria in the general case.}
372:
373: Generally, when $dY/dz \neq 0$ and dissipative effects are important,
374: the conditions of instability are more complex and depend on a horizontal
375: wavevector of perturbations, $k$. The temperature and lepton
376: gradients required for instability can be quite different for
377: perturbations with different $k$ and to obtain the criteria we have to
378: find the minimal values of these gradients. From the properties of
379: kinetic coefficients we have $\kappa_{T} \lambda_{Y} - \kappa_{Y}
380: \lambda_{T} > 0$, therefore we can deduce
381: from conditions (\ref{3cond}) that the gradients are minimal for
382: perturbations with the wavevector $k$ minimizing the quantity $Q^{6}/k^{2}$,
383: which is reached at $k^{2} = (\pi/d)^{2}/2$. Then, we have
384: \begin{equation}
385: \left( \frac{Q^{6}}{k^{2}} \right)_{min} = \frac{27}{4} \left(
386: \frac{\pi}{d} \right)^{4}.
387: \end{equation}
388: The instability criteria can now be obtained from
389: the last two conditions in (\ref{3cond}), which read, respectively,
390: \begin{eqnarray}
391: \frac{dY}{dz} (\delta \kappa_{T} - \beta \kappa_{Y} ) &-&
392: \left(\frac{\Delta\nabla T}{T}\right)_z (\beta \lambda_{Y} - \delta
393: \lambda_{T} )
394: \nonumber \\
395: &+& \frac{27 \pi^{4} \nu}{4 g d^{4}} (\kappa_{T}
396: \lambda_{Y} - \kappa_{Y} \lambda_{T}) < 0 \, ,
397: \label{c1}
398: \\
399: \frac{d Y}{dz} (\delta ( \nu + \lambda_{Y} ) + \beta \kappa_{Y})
400: &-&\left(\frac{\Delta\nabla T}{T}\right)_z (\beta ( \nu + \kappa_{T} ) +
401: \delta \lambda_{T}) +
402: \nonumber \\
403: &+& \frac{27 \pi^{4}}{4 g d^{4}} (\kappa_{T} + \lambda_{Y})
404: \left[ (\nu+\kappa_{T})(\nu+ \lambda_{Y}) - \lambda_{T} \kappa_{Y}\right] < 0 .
405: \label{c2}
406: \end{eqnarray}
407:
408: In the case of a negligible diffusivity, the first criterion yields
409: the Rayleigh-Taylor condition of instability, $\delta dY/dz < 0$. This
410: instability can be responsible, for instance, for the salt fingers
411: phenomena in the terrestrial oceans. Therefore, following Bruenn \&
412: Dineva (1996), we can conventionally call the instability associated
413: to condition (\ref{c1}) neutron
414: fingers. We refer to a {\it neutron finger unstable} region as a region where
415: condition (\ref{c1}) is fulfilled but not condition (\ref{c2}). In the case
416: of a ``standard'' transport with small viscosity and
417: diffusivity, the second criterion yields the Schwarzschild criterion of
418: convection, although convection in PNS is, in general, quite different from
419: the Schwarzschild convection and may arise in oscillatory or nonoscillatory
420: regimes. A region is said to be {\it convectively unstable} when both
421: conditions (\ref{c1}) and (\ref{c2}) are satisfied. If condition (\ref{c2}) is
422: satisfied but condition (\ref{c1}) is not, the system is said to be
423: {\it semiconvectively unstable}.
424:
425: Despite the fact that convective instabilities in PNS
426: were originally
427: argued by applying the Ledoux criterion (see, e.g., Epstein 1979,
428: Livio, Buchler \& Colgate 1980), it appears that the Ledoux criterion
429: is not the valid criterion for instability if we allow for conduction of heat,
430: diffusion of particles and/or viscosity. In this case, the true criteria of
431: instability are (\ref{c1}) and (\ref{c2}), and
432: only in two limiting cases
433: our criteria reduces to the Ledoux form.
434:
435: Generally, the region where any of the conditions (\ref{c1},\ref{c2})
436: are fulfilled can be very different to that given by the Ledoux criterion.
437:
438:
439: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
440: \section{Results and discussion.}
441:
442: To obtain the different thermodynamical derivatives, diffusion coefficients
443: and conductivities appearing in the coefficients of the dispersion equation,
444: we used the results from numerical simulations of PNS evolution performed
445: by Pons {\it et al.} (1999). The results discussed in this paper correspond
446: to the model labelled GM3np, with a baryonic mass of 1.6 $M_{\odot}$.
447: The simulation was carried out using a Henyey--like spherically symmetric
448: evolution code, coupled to a 1D neutrino transport scheme in the
449: flux--limited diffusion approximation. Details about the code and
450: the calculation of the diffusion coefficients using opacities
451: consistent with the underlying EOS (Reddy, Prakash \& Lattimer, 1998)
452: can be found in Pons {\it et al.} (1999).
453:
454: Since calculations of viscosity at
455: high densities are unreliable (compare, for instance, van den Horn
456: \& van Weert 1981, Goodwin \& Pethick 1982, Thompson \& Duncan 1993), we
457: calculate the
458: unstable region for different values of $\nu$. In Figure 1 (upper panel)
459: we plot the stability regions of a 1.6 $M_{\odot}$ PNS corresponding to
460: case $\nu=0$. In all the simulations we use a value for $d$ given by
461: the pressure scale, $d=\mid d\ln p/ dr \mid^{-1}$.
462: Three different situations can clearly be distinguished. Initially, there
463: is a small convectively unstable zone (darkest) near the surface that
464: lasts for only a few seconds, surrounded by a neutron finger unstable region
465: (intermediate gray tone). The inner part of the star is stable (lighter).
466: After a few seconds, the neutron finger instability moves inward,
467: occupying a large portion of the PNS, whereas the stable region shrinks.
468: Later on, at $t=12$ s, the innermost core becomes convectively unstable
469: for a few seconds, while the neutron finger unstable region begins
470: to shrink. By about $t=30$ s, the PNS is mostly stable, except a small
471: region near the center that is still subject to the neutron fingers
472: instability.
473: Instability completely disappears after $\approx 40$ s.
474:
475: To study the effect of viscosity, Figure 1 (lower panel) shows the time
476: dependence of the boundaries of unstable regions for $\nu = \eta^{2}
477: \kappa_{T}$ (van den Horn \& van Weert 1981). Qualitatively, the
478: unstable regions evolve in a similar fashion to the inviscid case.
479: The reason why viscosity hardly influences
480: the neutron finger instability region is that the third term on the
481: l.h.s. of (\ref{c1}), which
482: describes dissipation of perturbation due to neutrino transport
483: and depends on $\nu$, is relatively small compared to the first
484: two terms which are independent of $\nu$. Therefore, the neutron finger
485: unstable region is approximately the same as in Fig.1 at any
486: evolutionary stage. On the contrary, convective instability is
487: sensitive to the value of viscosity through condition (\ref{c2}). The third
488: term on the l.h.s.
489: of (\ref{c2}) again does not yield too much contribution, however, two other
490: terms depend on viscosity. Due to this, the region unstable to
491: convection turns out to be more extended at high viscosity.
492: Notice that in the case $\eta \gg 1$ the
493: viscosity is much greater than the diffusivity and thermal diffusivity but,
494: nevertheless, is not sufficiently large to make the term in the r.h.s.
495: important, thus recovering Ledoux criterion from (\ref{c2}).
496: Note, however, that for both models the
497: convectively unstable region always lies inside the region unstable
498: to neutron fingers. In our simulations we have not found any region unstable
499: to semiconvection. Although this result might seem different from Bruenn
500: \& Dineva (1996)
501: we should keep in mind that the thermodynamic conditions used in this work to
502: describe the interior of the PNS are
503: quite different from the conditions used by Bruenn \& Dineva (1996)
504: to describe the matter just below the neutrinosphere of the collapsed core.
505: Bruenn \& Dineva (1996) studied the instability in a region, typically with a
506: density of 3$\times 10^{13}$ g/cm$^3$and with a high entropy per baryon
507: $s=4$, since at
508: such early times ($\approx 50-230$ ms after bounce) the excess of entropy of
509: the shocked mantle has not been yet radiated away. In this work, however, we
510: focus on the long term evolution of a PNS, and most of the
511: region of our interest is at supranuclear densities ($\rho > 3\times 10^{14}$
512: g/cm$^3$) and moderate entropies.
513:
514:
515: Since the superadiabatic temperature gradient and the gradient of lepton
516: number play a fundamental role in the stability criterion we have shown in
517: Figure 2 graphics of these quantities for different times in the evolution.
518: We plot thicker lines for the unstable regions according to the model shown
519: in the lower panel of Figure 1.
520: As we can see from the figures, at early and middle times in the
521: evolution, instability region corresponds approximately to that with a
522: positive value of the superadiabatic temperature gradient.
523: The effect of the negative lepton fraction gradient and/or dissipative
524: processes increase only slightly the unstable region.
525: At late time however, the unstable region almost disappears while the
526: superadiabatic temperature gradient remains positive in a much wider region.
527: This is chiefly caused by the change in the sign of the
528: thermodynamic derivative $\delta$, when the lepton fraction drops some critical
529: value.
530:
531: In the upper panel of Figure 3 we show the maximum of the real parts of the
532: roots
533: of equation (\ref {cubic}) at two different stages in the evolution.
534: A positive value of this quantity
535: implies instability and, in that case, the instability growth time is
536: calculated by taking its inverse (lower panel).
537: As we can contrast in the lower panel of Figure 3 the characteristic
538: growth time of instabilities at the beginning of the evolution (t=1 s) is of
539: the order of
540: tenths of millisecond. Since this time is much shorter
541: than the
542: evolution time scale, convective heat transport might be important in this
543: region.
544: At the middle stages of the evolution (t=20 s), two different unstable
545: regions can be distinguished, a first one with a growth time of the
546: order of 1 ms (convection) and a second region with associated
547: growth times of the order of few tens of millisecond (corresponding to the
548: region unstable to neutron fingers).
549: The growth time in both cases is, again,
550: much shorter than the time evolution and, in principle,
551: instabilities can grow fast enough to influence the heat transport.
552:
553: In summary, we have to emphasize that neutrino transport plays
554: a significant role in the stability properties of PNSs.
555: Criteria (25) and (26) obtained in the present paper correspond
556: properly to the conditions of PNS where heat transport and
557: diffusion can be sufficiently fast. These criteria differ essentially
558: from those used by other authors in this context.
559: Our analysis shows that the unstable
560: region inside the PNS can be more extended than predicted, for
561: instance, by the Ledoux criterion alone. Generally, neutron fingers
562: instability arises in a more extended region of a stellar volume and
563: lasts a longer time than convection. Therefore, this instability
564: seems to be as important as convection for the early evolution of a newly
565: born neutron star.
566: Our main conclusion is that, surprisingly, diffusive
567: effects allow for the existence of new unstable modes and the role
568: of convection in PNS should be carefully reexamined.
569:
570: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
571: \section*{Acknowledgement}
572: This research was supported in part by the Spanish Ministerio of Educaci\'on
573: y Cultura (grant PB97-1432) and the Russian Foundation of Basic Research
574: (grant 00-02-04011). V.U. thanks the University of Valencia for financial
575: support and hospitality.
576:
577:
578: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
579: \section*{References}
580:
581: \noindent
582: Aleksandrov A.D., Kolmogorov A.N., Laurentiev M.A. 1963. Mathematics: Its
583: Content, Methods, and Meaning. MIT press.\\
584: Arnett W. D. 1987, ApJ, 319, 136 \\
585: Bruenn S., Dineva, T. 1996, ApJ, 458, L71 \\
586: Bruenn S., Mezzacappa A. 1994, ApJ, 433, L45 \\
587: Bruenn S., Mezzacappa A., Dineva T. 1995, Phys. Rep., 256, 69 \\
588: Burrows A., Fryxell B. 1993, ApJ, 418, L33 \\
589: Burrows A., Lattimer J. 1986, ApJ, 307, 178 \\
590: Burrows A., Lattimer J. 1988, Phys. Rep., 163, 151 \\
591: Colgate S., Petschek A.1980, ApJ, 238, L115 \\
592: Thompson C., Duncan R. 1993, ApJ, 408, 194 \\
593: Epstein R. 1979, MNRAS, 188, 305 \\
594: Goodwin B., Pethick C. 1982, ApJ, 253, 812 \\
595: Grossman, S.A., Narayan, R., Arnett, D., 1993, ApJ, 407, 284 \\
596: Hillebrandt W. 1987, in High Energy Phenomena around Collapsed Stars
597: (ed. F.Pacini), Dordrecht, Reidel, 73 \\
598: Imshennik V.S., Nadezhin D.K. 1972, Sov. Phys. JETP, 36, 821 \\
599: Janka H.-T., Muller E. 1994, A\&A, 290, 460 \\
600: Keil W., Janka H.-T. 1995, A\&A, 296, 145 \\
601: Keil W., Janka H.-T., Muller E. 1996, ApJ, 473, L111 \\
602: Landau L., Lifshitz E. 1959, Fluid Mechanics, London, Pergamon \\
603: Lattimer, J.M., \& Mazurek, T.J., 1981, ApJ, 246, 955 \\
604: Livio M., Buchler J., Colgate S. 1980, ApJ, 238, L139 \\
605: Mezzacappa, A., Calder, A.C., Bruenn, S.W., Blondin, J.M., Guidry, M.W.
606: Strayer, M.R., and Umar, A.S., 1998, ApJ, 493, 848 \\
607: Pons J.A., Reddy S., Prakash M., Lattimer J., Miralles J.A. 1999, ApJ,
608: 513, 780 \\
609: Reddy S., Prakash M., Lattimer J., 1998, Phys. Rev., D58, 013009. \\
610: Smarr L., Wilson J., Barton R., Bowers R. 1980, ApJ, 246, 515 \\
611: Sumiyoshi K., Suzuki H., Toki H. 1995, A\&A, 303, 475 \\
612: van den Horn L., van Weert C. 1981, ApJ, 251, L97 \\
613:
614: \section*{Figure captions}
615:
616: \vspace*{1cm}
617: \noindent Fig. 1.-- Evolution of the different unstable zones in a $M=1.6 M_{\odot}$
618: PNS. The darker zone corresponds to the convectively unstable regions,
619: the lighter zone to convectively stable regions and the intermediate
620: zone presents the secular instability usually denoted by neutron fingers.
621: The upper and lower panels displays the results of the inviscid and
622: viscous ($\nu =\eta^2 \kappa_T$) models, respectively.
623:
624: \vspace*{1cm}
625: \noindent Fig. 2.-- Superadiabatic temperature gradient (left panels) and
626: lepton fraction gradient (right panels) for three different times in the
627: evolution. Thicker lines correspond to the unstable region according to
628: Figure 1.
629:
630: \vspace*{1cm}
631: \noindent Fig. 3.-- Maximum value of the real parts of the roots of equation
632: (17) at two different stages in the evolution (upper panel) and growth time
633: for the unstable regions (lower panel).
634:
635: \begin{figure}
636: \begin{center}
637: \epsfxsize=6.in
638: \epsfysize=7.in
639: \epsffile{fig1.ps}
640: \end{center}
641: \caption{}
642: \end{figure}
643:
644: \begin{figure}
645: \begin{center}
646: \epsfxsize=6.in
647: \epsfysize=7.in
648: \epsffile{fig2.ps}
649: \end{center}
650: \caption{}
651: \end{figure}
652:
653: \begin{figure}
654: \begin{center}
655: \epsfxsize=6.in
656: \epsfysize=7.in
657: \epsffile{fig3.ps}
658: \end{center}
659: \caption{}
660: \end{figure}
661:
662: \end{document}
663: