astro-ph0002212/ms.tex
1: %Uses AASTeX v5.0
2: %
3: %  The porous atmosphere of eta-Carinae / Nir Shaviv
4: %
5: 
6: 
7: \documentclass[preprint,11pt]{aastex}
8: \usepackage{epsfig}
9: 
10: %----------------------------- My definitions
11: 
12: %
13: \newcommand{\be}{\begin{equation}}
14: \newcommand{\ee}{\end{equation}}
15: \newcommand{\bt}{\begin{table} \begin{center}}
16: \newcommand{\et}{\end{center} \end{table}}
17: \newcommand{\ba}{\begin{eqnarray}}
18: \newcommand{\ea}{\end{eqnarray}}
19: \newcommand{\ie}{{\it i.e.~}}
20: \newcommand{\eg}{{\it e.g.~}}
21: %
22: \newcommand{\citenp}[1]{\citeauthor{#1}~\citeyear{#1}}
23: \newcommand{\sch}{Schwarzchild~}
24: \newcommand{\BV}{Brunt-V\"ais\"al\"a~}
25: \newcommand{\mt}{\mathit}
26: %
27: \newcommand{\ms}{{\cal M}_\odot}
28: \newcommand{\ls}{{\cal L}_\odot}
29: 
30: 
31: % ---- For some reason the AAS definitions for <~ and >~ do
32: % not work properly, so they are defined.
33: 
34: \def\lesssim{\mathrel{\mathchoice {\vcenter{\offinterlineskip\halign{\hfil
35: $\displaystyle##$\hfil\cr<\cr\sim\cr}}}
36: {\vcenter{\offinterlineskip\halign{\hfil$\textstyle##$\hfil\cr
37: <\cr\sim\cr}}}
38: {\vcenter{\offinterlineskip\halign{\hfil$\scriptstyle##$\hfil\cr
39: <\cr\sim\cr}}}
40: {\vcenter{\offinterlineskip\halign{\hfil$\scriptscriptstyle##$\hfil\cr
41: <\cr\sim\cr}}}}}
42: \def\gtrsim{\mathrel{\mathchoice {\vcenter{\offinterlineskip\halign{\hfil
43: $\displaystyle##$\hfil\cr>\cr\sim\cr}}}
44: {\vcenter{\offinterlineskip\halign{\hfil$\textstyle##$\hfil\cr
45: >\cr\sim\cr}}}
46: {\vcenter{\offinterlineskip\halign{\hfil$\scriptstyle##$\hfil\cr
47: >\cr\sim\cr}}}
48: {\vcenter{\offinterlineskip\halign{\hfil$\scriptscriptstyle##$\hfil\cr
49: >\cr\sim\cr}}}}}
50: 
51: 
52: %---------------------------- End of definitions
53: 
54: \begin{document}
55: 
56: %\def\temp{
57: \title{The porous atmosphere of $\eta$-Carinae}
58: \author{Nir J. Shaviv}
59: \affil{Canadian Institute for Theoretical
60:         Astrophysics, University of Toronto \\ 60 St. George St.,
61:         Toronto, ON M5S 3H8, Canada}
62: %}
63: 
64: \begin{abstract}
65:   
66:   We analyze the wind generated by the great 20 year long
67:   super-Eddington outburst of $\eta$-Carinae.  We show that using
68:   classical stellar atmospheres and winds theory, it is impossible to
69:   construct a consistent wind model in which a sufficiently {\em
70:   small} amount of mass, like the one observed, is shed.  One expects
71:   the super-Eddington luminosity to drive a thick wind with a mass
72:   loss rate substantially higher than the observed one.  The easiest
73:   way to resolve the inconsistency is if we alleviate the implicit
74:   notion that atmospheres are homogeneous.  An inhomogeneous
75:   atmosphere, or ``porous", allows more radiation to escape while
76:   exerting a smaller average force.  Consequently, such an atmosphere
77:   yields a considerably lower mass loss rate for the same total
78:   luminosity. Moreover, all the applications of the Eddington
79:   Luminosity as a strict luminosity limit should be revised, or at
80:   least reanalyzed carefully.
81: 
82: \vskip 0.5cm
83: \centerline{\em To appear in the Astrophysical Journal Letters}
84: \vskip 0.5cm
85: 
86: \end{abstract}
87: 
88: \keywords{
89:   Radiative transfer --- hydrodynamics --- instabilities --- stars:
90:   atmospheres --- stars: individual ($\eta$ Carinae) }
91: % to find at:http://www.noao.edu/apj/keywords96.html
92: 
93: %------------------------------------------------------
94: 
95: 
96: \section{Introduction}
97: 
98:   $\eta$-Carinae is probably one of the most remarkable stellar object
99:   to have ever been documented. About 150 years ago, the star began a
100:   20 year long giant eruption during which it radiated a
101:   supernova-like energy of roughly $3 \times 10^{49}~ergs$~
102:   (\citenp{DH97}).  Throughout the eruption it also shed some
103:   $1-2~\ms$ of material carrying approximately $6\times 10^{48}~ergs$
104:   as kinetic energy (\citenp{DH97}), while expanding at a velocity of
105:   $650~km/sec$ (\citenp{HA92}, \citenp{C96}). $\eta$-Carinae can
106:   therefore serve as a good laboratory for the study of atmospheres
107:   at extreme luminosity conditions.
108: 
109:   At first glance, it appears that the star shed a large amount of
110:   material. Indeed, the inferred mass loss rate during the great
111:   eruption of $\sim 0.1~\ms/yr$ is significantly larger than the mass
112:   loss rate inferred for the star today ($\lesssim 10^{-3}~\ms/yr$,
113:   \citenp{DH97} and references therein). However, considering that the
114:   luminosity during the great eruption is estimated to be
115:   significantly above the Eddington limit, we shall show that the star
116:   should have had a much higher mass loss rate. In fact, it should
117:   have lost during the 20 year eruption more mass than its total mass,
118:   giving rise to an obvious discrepancy.
119: 
120:   A review of our current knowledge of $\eta$ Car can be found in
121:   \cite{DH97}. In section \ref{sec:wind} we summarize how a wind
122:   solution for the star $\eta$ Car should be constructed. Since the
123:   luminosity is very high, the effects of convection must be taken
124:   into account. In section \ref{sec:discrepancy} we integrate the wind
125:   equations to show that no consistent solution for $\eta$ Car exists
126:   within the possible range of observed parameters. Section
127:   \ref{sec:bad} is devoted to possible classical solutions to the
128:   discrepancy, showing that no such possibility exists. In section
129:   \ref{sec:good}, we show that a porous atmosphere is a simple and
130:   viable solution to the wind discrepancy.
131: 
132: \section{Solving for the Wind}
133: \label{sec:wind}
134: 
135:   Since the mass of $\eta$-Car is estimated to be of order
136:   $100-120~\ms$ (\citenp{DH97}), the average luminosity in the great
137:   eruption was clearly super-Eddington (of the order of 5 times the
138:   Eddington limit).  That is to say, the radiative force upwards,
139:   assuming the smallest possible opacity (for ionized matter) given by
140:   Thomson scattering, was significantly larger than the gravitational
141:   pull downwards.  Optically thin winds formally diverge at the
142:   Eddington limit (e.g., \citenp{Puls} and references therein).
143:   Consequently, a consistent wind solution requires an optically thick
144:   wind. We thus look for a wind in which the sonic point (which is the
145:   point at which the local speed of the outflow equals the speed of
146:   sound) is below the photosphere.  Moreover, since the duration of
147:   the eruption is longer than the sound crossing time of the star by
148:   about a factor of 50, a stationary wind appears to be a good
149:   approximation.
150: 
151:   In practically all super-sonic wind theories which describe
152:   super-sonic outflows from an object at rest, a consistent stationary
153:   solution is obtained only when the net driving force of the wind
154:   (excluding the pressure gradient) vanishes at the sonic
155:   point\footnote{ The exception is line driven winds in which the
156:   force is explicitly a function of $dv/dr$ which is actually an
157:   approximation to the line transfer equations. If we had written the
158:   proper radiation transfer equations for this case which only {\em
159:   implicitly} depend on $dv/dr$, we would have recovered that the
160:   sonic point coincides with the point at which the total force
161:   vanishes (cf \citenp{MWM84} \S107).  Moreover, line driven winds are
162:   important only under optically thin conditions while we describe the
163:   optically thick part of the wind.  }.  Thus, material experiencing a
164:   super-Eddington flux necessarily has to be above the sonic point.
165:   If most of the envelope carries a super Eddington flux, then no
166:   consistent stationary wind solution can be obtained and in fact, the
167:   object will evaporate on a dynamical time scale. In most systems
168:   however, this need not be the case. For example, in very hot systems
169:   (e.g., hot neutron stars during strong X-ray bursts, \citenp{QP85}),
170:   the opacity in the deep layers is reduced due the reduced
171:   Klein-Nishina opacity for Compton scattering at high
172:   temperatures. Thus, the sonic point in these objects is found where
173:   the temperature is high enough to reduce the opacity to the point
174:   where the flux corresponds to the local Eddington limit.
175: 
176:   Another important effect, which should be taken into account, is
177:   convection.  Deep inside the atmosphere, convection can carry a
178:   significant part (or almost all) of the energy flux, thus reducing
179:   the radiative pressure to a sub-Eddington value. In fact, as the
180:   radiative flux approaches the Eddington limit, convection generally
181:   arises and carries the lion share of the total energy flux (if it
182:   can) to keep the system at a sub-Eddington level
183:   (\citenp{JSO73}). Although the total flux in the entire envelope (or
184:   almost all of it) can be equivalent to a super-Eddington flux, up to
185:   some depth below the photosphere, convection carries most of the
186:   flux so as to reduce the radiative flux alone into a sub-Eddington
187:   value. A consistent wind solution should therefore, have its sonic
188:   point at the location where the most efficient convection cannot
189:   carry enough flux any more.  As we shall soon see, the problem in
190:   $\eta$-Car is that this point is relatively deep within the
191:   atmosphere, where the density is so high that the expected mass loss
192:   is significantly {\it higher} than the observed one.
193: 
194:   To see this in a robust way we integrated numerically the wind
195:   equations starting from the photosphere inwards. The equations are
196:   those that describe optically thick spherically symmetric winds
197:   (\citenp{QP85}; \citenp{Z73}; \citenp{KH94}). The equations of mass
198:   conservation, momentum conservation and temperature gradient are
199: \begin{equation}
200:   4 \pi r^2 \rho v = \dot{M} = {\rm const}
201: \end{equation}
202: \begin{equation}
203:   v {d v\over dr} + {GM\over r^2} +{1\over \rho}{dP_{g} \over dr} -
204:   {\chi {\cal L}_r \over 4 \pi r^2 c}  =  0
205: \end{equation}
206: \begin{equation}
207:    {d T \over dr} = - {3 \chi \rho {\cal L}_r \over 16 \pi r^2 c a T^3} (1 +
208:    {2 \over 3 \chi \rho r}),
209: \end{equation}
210: with standard notation.  The parenthesized term in the last equation
211: is a simple approximate interpolation that has the correct asymptotic
212: limits for optical depths much larger and much smaller than unity
213: (\citenp{QP85}).  The last equation is the integrated form of the
214: energy conservation equation. Unlike the aforementioned references, we
215: specifically include advection by a maximally efficient convection.
216: Thus, the integrated form of the energy conservation equation becomes
217: \begin{eqnarray}
218:     {\cal L}_r + \dot{M}\left({v^2 \over 2} + w - {GM\over
219:    r}\right) + {\cal L}_{conv} &\equiv& \\ {\cal L}_r -{GM \dot{M}  \over r} +
220:    {\cal L}_{adv}+{\cal L}_{conv} &=& \Lambda_{tot} = {\cal L}_{obs}+
221:    {\cal L}_{kin,\infty},
222: \label{eq:fluxes}
223: \end{eqnarray}
224:   were $\Lambda_{tot}$, $\dot{M}$, ${\cal L}_{obs}$ and ${\cal
225:   L}_{kin,\infty}$ are the total energy output of the star, the wind
226:   mass loss rate, the observed luminosity at infinity and the kinetic
227:   energy flux at infinity. On the other hand, ${\cal L}_r, v, w, {\cal
228:   L}_{conv}, {\cal L}_{adv}$ are respectively, the local radiative
229:   luminosity, velocity, enthalpy, convective flux and advected flux
230:   (as internal and kinetic energies).  The expression adopted for
231:   ${\cal L}_{conv}$ is $4 \pi r^2 u v_{s}$ where $u$ is the internal
232:   energy per unit volume and $v_s$ is the speed of sound.  By no means
233:   can convection be more efficient than this expression since highly
234:   dissipative shocks are unavoidable at higher speeds. It is likely
235:   that the maximally efficient convection is somewhat less efficient
236:   than this expression, but this will only aggravate the problem that
237:   we shall soon expose.  Detailed calculations of the wind were
238:   carried out.  The calculations include the latest version of the
239:   OPAL opacities (\citenp{IR96}).  It is found that the total opacity
240:   below the photosphere has comparable contributions from Thomson
241:   scattering and absorption processes. This implies that the {\em
242:   modified} Eddington limit, in which the Thomson opacity is replaced
243:   by the local total opacity, is somewhat lower than the classical
244:   Eddington limit.
245: 
246: 
247:   Since $T_{\mathit{eff}} \sim 9000^{\circ}$K\footnote{This is the
248:   typical observed effective temperature for LBV's during outbursts
249:   (\citenp{HD94}). If the temperature is higher than this value, the
250:   inferred bolometric magnitude of $\eta$ Car during the eruption
251:   would be more negative, increasing the Eddington factor. If the
252:   temperature is lower than $\sim 7000^\circ$K, the opacity at the
253:   photosphere and outwards rises abruptly (\citenp{D87}), thus
254:   reducing the modified Eddington limit. In both cases, the
255:   discrepancy will be aggravated.}, the average luminosity implies a
256:   photospheric radius of $10^{14}~cm$. Note that since it is a thick
257:   wind, the exact definition of the photosphere is ambiguous.
258:   Nevertheless, different definitions do not change the results by
259:   more than $10-20\%$.  Knowing that the observed mass loss rate is
260:   roughly $0.1~\ms/yr$ (which gives the observed $2~\ms$ of shed
261:   material in 20 years, \citenp{DH97}), a specified flow speed at the
262:   photosphere can be translated to a required density. We can
263:   therefore integrate the wind equations inwards.  If a consistent
264:   wind solution can be obtained for some value of the imposed velocity
265:   in the photosphere (which has to be between $v_s$ and $v_{\infty}$),
266:   then the integration inwards should reach a sonic point at which the
267:   total force on the gas vanishes.  This will be attained if the
268:   convective and advective fluxes can carry a significant amount of
269:   the total flux so as to reduce the residual radiative flux to a
270:   sub-Eddington one.
271: 
272: \section{The Discrepancy}
273: \label{sec:discrepancy}
274: 
275:   We define the luminosity needed to be carried by convection and
276:   advection in order to bring about the vanishing of the total local
277:   force as ${\cal L}_{crit}$. If enough energy is advected and
278:   convected then ${\cal L}_r$ will be reduced to the local modified
279:   Eddington flux:
280: \begin{equation}
281: {\cal L}_{Edd,mod}={4 \pi c G M\over \chi}
282: \end{equation}
283:   with $\chi$ the local opacity which can be larger than the Thomson
284:   opacity.   Thus, from eq.~(\ref{eq:fluxes}), the critical
285:   advective+convective flux can be written as
286: \begin{equation}
287: {\cal L}_{crit} = \Lambda_{tot} + {G M {\dot M}\over r}-{\cal L}_{Edd,mod}.
288: \end{equation}
289: 
290:   Figure \ref{fig_1} shows the fraction $\eta\equiv({\cal
291:   L}_{adv}+{\cal L}_{conv})/{\cal L}_{crit}$ at the sonic point. A
292:   consistent solution can be found only if $\eta=1$ at the sonic
293:   point.
294: 
295:   Inspection of the figure clearly shows that the space of possible
296:   observed values does not contain a viable and consistent solution.
297:   This is of course irrespective of whether a solution from the
298:   photosphere outward can or cannot be obtained. The discrepancy
299:   arises because a wind corresponding to the observed low mass loss
300:   rate necessarily has a sonic point that is not deep enough to have
301:   either convection or advection as an efficient mean of transporting
302:   energy.  This can be seen from the optical depth at which the sonic
303:   point is obtained. In all cases, $1 \lesssim \tau < 300$.  However,
304:   convection is efficient only up to an optical depth of $\tau \sim c/
305:   v_s \gg 300$ for $p_{rad} \sim p_{gas}$ (\citenp{S00b}), or even
306:   deeper for larger radiation pressures (i.e., when close to the
307:   Eddington limit).
308: 
309: \section{Unfeasible Solutions to the Discrepancy}
310: \label{sec:bad}
311: 
312:   Can the discrepancy be resolve with a classical assumption?  Since
313:   the discrepancy is rather large, assuming the wind to emerge from an
314:   angular fraction $f$ from the star does not relax the problem (it
315:   actually aggravates the problem because more material will be blown
316:   away from the higher luminosity regions).  Another possibility that
317:   fails is having a higher velocity in the photosphere than the one
318:   observed today for the shed material. This might be the case if the
319:   wind collides with previously ejected slow moving material. Even if
320:   such material did exist, the necessarily reduced mass loss rate
321:   inferred from the present day observed momentum aggravates the
322:   problem.
323: 
324:   The problem is not mitigated if we relax the assumption that the mass
325:   loss rate and the luminosity are assumed to be constant in time
326:   throughout the eruption.
327: 
328:   If one wishes to solve the problem using magnetic fields, then a
329:   solution can be found only if the magnetic energy density at the
330:   photosphere is significantly larger (by several orders of magnitude)
331:   then the equipartition value with the gas pressure. This of course
332:   seems unlikely.
333: 
334:   Another option is to have the distance estimate to $\eta$-Car be
335:   three times smaller than $2300~pc$. A shorter distance will remove
336:   $\eta$-Car out of the cluster Tr16 of massive stars inside which it
337:   is observed and leave it instead roaming the inter galactic arm
338:   space. Considering the short lifetime of the star, just a few
339:   million years, this possibility appears as very unlikely.
340: 
341:   The problem can be solved if the mass of the star corresponds to a
342:   sub-Eddington luminosity. This proposed solution requires $\eta$-Car
343:   to be at least a $\sim 1000~\ms$ star.  However, this suggestion is
344:   at variance with much lower estimates (see for instance
345:   \citenp{DH97} and references therein). Nevertheless, having such a
346:   massive star is in fact not completely unrealistic and would have
347:   far reaching consequences if found to be true.
348: 
349: \section{A Viable Solution to the Discrepancy}
350: \label{sec:good}
351: 
352:   As the title suggests, there is a clear solution to the
353:   discrepancy. As the results show, the sonic point appears to be
354:   between the optical depths of $\sim 1$ and $\sim 300$.  The exact
355:   value cannot be obtained since it requires the integration outward
356:   from the photosphere, which owing to the relatively inaccurate
357:   effective temperature and therefore opacity, yields a wide range of
358:   results. If the {\it mean} radiative force between the point $\eta=1$
359:   and the above found optical depth, is smaller than classically
360:   estimated, then a solution to the discrepancy can be found. Such a
361:   reduction in the mean radiative force is a natural result if the
362:   atmosphere is inhomogeneous.
363: 
364:   \cite{S98} has shown that in an inhomogeneous atmosphere, the
365:   effective opacity used to calculate the average force is reduced
366:   relative to the effective opacity used for the radiation transfer in
367:   a homogeneous medium.  The effective opacity used for the average
368:   force should be a volume {\em flux weighted} average of the opacity
369:   per unit volume\footnote{When the flux is frequency dependant, a
370:   similar average should be taken in order to find the radiative
371:   force.  However, one then takes a flux weighted mean over {\em
372:   frequency space}.}  $\chi_v \equiv \chi \rho$. Namely,
373: \begin{equation}
374:   \chi_{\mt{eff}} = {\left< \chi \rho F\right> \over \left< F \right>
375:   \left<\rho\right>}.
376: \end{equation}
377: 
378:   The effect is universal and arises in inhomogeneous systems that
379:   conduct heat or electricity.  Extensive discussions exist in the
380:   literature under a different terminology (\citenp{I92}). The only
381:   requirement is therefore, that close to the Eddington limit the star
382:   develop inhomogeneities.  The transformation from an homogeneous to
383:   an inhomogeneous atmosphere at luminosities close to but below
384:   Eddington luminosities, was recently found to take place generically
385:   even in Thomson scattering atmospheres (\citenp{S99}; \citenp{ST99};
386:   \citenp{S00}).
387: 
388:   It was found that two different types of instabilities arise
389:   naturally when the luminosity approaches the Eddington limit
390:   (\citenp{S00}). One instability is of a phase transition into a {\em
391:   stationary} nonlinear pattern of ``fingers" that facilitate the
392:   escape of the radiation.  The second type of instability allows the
393:   growth of a propagating wave, from which one expects a {\em
394:   propagating} nonlinear pattern to form.  The two possibilities are
395:   summarized in figure \ref{fig_2}.  Both instabilities bring about a
396:   reduction of the average radiative force on the matter and a
397:   significant reduction of the mass loss rate since the sonic surface
398:   can sit near (or not much below) the photosphere. In both cases, the
399:   nonlinear pattern is necessarily expected to form in the region
400:   between the radius $r_{conv}$ at which $\eta=1$, or in other words,
401:   that ${\cal L}_{conv} + {\cal L}_{adv}$ is large enough to have
402:   ${\cal L}_r \lesssim {\cal L}_{Edd,mod}$, and the photosphere. When
403:   the pattern is stationary, the rarefied regions have a larger than
404:   Eddington flux and the sonic surface in these regions is near
405:   $r_{conv}$. On the other hand, if the pattern is propagating, the
406:   flux may be larger locally than the Eddington limit but the time
407:   average of the force on a mass element is less than Eddington. Since
408:   the instability does not occur above the photosphere, it should be
409:   homogeneous and hence super-Eddington with a super-sonic flow.
410: 
411:   Further analysis of the instabilities is needed to know which
412:   instability will dominate though it is more likely to be the phase
413:   transition since it is dynamically more important.
414: 
415: \section{Summary}
416: \label{sec:summary}
417: 
418:   To summarize, the super-Eddington luminosity emitted by $\eta$-Car
419:   should have generated a much thicker wind with a sonic point placed
420:   significantly deeper than what can be directly inferred from the
421:   observations.  A solution which lives in harmony with observations
422:   and theoretical modeling is a porous atmosphere, which allows more
423:   radiation to escape while exerting a smaller average force. It also
424:   means that the Eddington limit is not as destructive as one would a
425:   priori think it must be, even in a globally spherically symmetric
426:   case. Namely, all astrophysical analyses that employ the Eddington
427:   limit as a strict limit should be reconsidered carefully, even if
428:   they involve only unmagnetized Thomson scattering material. If
429:   $\eta$-Carinae could have been super-Eddington for such a long
430:   duration without ``evaporating'', other systems can display a
431:   similar behavior.
432: 
433: 
434: \begin{thebibliography}{ll}
435: 
436:   \bibitem[Currie et al.(1996)]{C96} Currie, D. G., Dowling, D. M.,
437:   Shaya, E. J., Hester, J., Scowen, P., Groth, E. J., Lynds, R. O., \&
438:   Earl, J. Jr Wide Field/Planetary Camera Instrument Definition Team,
439:   1996, \apj, 112, 1115
440: 
441:   \bibitem[Davidson(1987)]{D87} Davidson, K.  1987, \apj, 317, 760
442: 
443:   \bibitem[Davidson \& Humphreys(1997)]{DH97} Davidson, K.  \&
444:   Humphreys, R. M., 1997, \araa, 35, 1
445: 
446:   \bibitem[Hiller \& Allen(1992)]{HA92} Hillier, D. J. \& Allen,
447:   D. A. 1992, \aap, 262, 153
448: 
449:   \bibitem[Humphreys \& Davidson(1994)]{HD94} Humphreys,
450:   R. M. \& Davidson, K.  1994, \pasp, 106, 1025
451: 
452:   \bibitem[Iglesias \& Rogers(1996)]{IR96} Iglesias, C. A. \& Rogers,
453:   F. J. 1996, \apj, 464, 943
454: 
455:   \bibitem[Isichenko(1992)]{I92} Isichenko, M. B. 1992,
456:   {Rev. Mod. Phys.}, 64, 961.
457: 
458:   \bibitem[Joss et al.(1973)]{JSO73} Joss, P. C., Salpeter, E. E. \& Ostriker,
459:    J. P. 1973, 181, 429
460: 
461:   \bibitem[Kato \& Hachisu(1994)]{KH94} Kato, M., \& Hachisu, I. 1994,
462:   \apj, 437, 802
463: 
464:   \bibitem[Kudritzki et al.(1989)]{Puls} Kudritzki, R. P., Pauldrach,
465:   A., Puls, J. \& Abbott, D. C. 1989 \aap, 219, 205
466: 
467:  \bibitem[Mihalas \& Weibel Mihalas(1984)]{MWM84} Mihalas D., \& Weibel
468:   Mihalas B. 1984, Foundations of radiation hydrodynamics, Oxford univ.
469:   press, Oxford
470: 
471:   \bibitem[Quinn \& Paczynski(1985)]{QP85} Quinn, T. \& Paczynski, B.
472:   1985, \apj, 289, 634
473: 
474:   \bibitem[Shaviv(1998)]{S98} Shaviv, N. J. 1998, \apj, 494, L193
475: 
476:   \bibitem[Shaviv(1999)]{S99} Shaviv, N. J. 1999, in {\it Variable and
477:   Non-spherical Stellar Winds in Luminous Hot Stars, IAU Colloquium
478:   169}, ed.~B.~Wolf, O.~Stahl,~A.~W.~Fullerton, Springer, p. 155
479: 
480:   \bibitem[Shaviv(2000a)]{S00} Shaviv, N. J. 2000a, {submitted to
481:   \apj}
482:   \bibitem[Shaviv(2000b)]{S00b} Shaviv, N. J. 2000b, {submitted to
483:   \apj}
484: 
485:   \bibitem[Spiegel \& Tao(1999)]{ST99} Spiegel, E. A., \& Tao L. 1999,
486:   {Physics Reports}, 311, 163
487: 
488:   \bibitem[\.{Z}ytkow(1972)]{Z73} \.{Z}ytkow, A.  1972, {Acta
489:   Astronomica}, 22, 103
490: 
491: \end{thebibliography}
492: 
493: \begin{figure}[p]
494: \centerline{\epsfig{file=fig1.eps,width=4in,angle=-90}}
495: \caption{
496:   The fraction $\eta = ({\cal L}_{adv}+ {\cal L}_{conv})/{\cal
497:   L}_{crit}$ as a function of the photospheric velocity $v_{ph}$.  A
498:   consistent wind solution requires (a) that the velocity at the
499:   photosphere $v_{ph}$ satisfy: $v_s \le v_{ph}
500:  \le v_\infty$, and
501:   (b) , $\eta(v=v_{s})=1$.  The thick line corresponds to the nominal
502:   observed and inferred values ($M_\star=100~\ms$,
503:   $T_{\mathit{eff}}=9000^\circ$K, $\dot{M}=0.1~\ms/yr$, $\int{\cal
504:   L}_{obs}dt = 3\times 10^{49}~erg$) while the additional lines depict
505:   the result when the values are changed to their reasonable limits
506:   (and even beyond). Clearly, no reasonable choice of parameters can
507:   result with a sonic point that is consistent with a wind solution
508:   (namely, we always find $\eta(v=v_s)\ll 1$). Basically, the
509:   discrepancy arises because the mass loss rate observed is too small
510:   to have the sonic point deep enough in the atmosphere where
511:   convection can be an efficient mean of energy transport. }
512: \label{fig_1}
513: \end{figure}
514: 
515: \begin{figure}[p]
516: \centerline{\epsfig{file=fig2a.eps,width=3.5in}}
517: \centerline{\epsfig{file=fig2b.eps,width=3.5in}}
518: \caption{
519:  The proposed atmospheric structure of $\eta$ Carinae during its great
520:  eruption. A homogeneous atmosphere is unstable as a result of two
521:  generic instabilities that take place even in Thomson atmospheres
522:  when close to the Eddington limit (\citenp{S00}). The effective
523:  opacity is therefore reduced (\citenp{S98}) and with it the average
524:  radiative force. The two panels describe the two types of
525:  possibilities for having a `porous' atmosphere according to the
526:  characteristics of the instability that arises. An instability could
527:  produce a stationary pattern (first panel) if it originates from the
528:  phase transition instability and a moving pattern if it originates
529:  from the finite speed of light instability (second panel).  See
530:  details in the text.}
531: \label{fig_2}
532: \end{figure}
533: 
534: \end{document}
535: 
536: 
537: