1: % mn2esample.tex
2: %
3: % v2.1 released 22nd May 2002 (G. Hutton)
4: %
5: % The mnsample.tex file has been amended to highlight
6: % the proper use of LaTeX2e code with the class file
7: % and using natbib cross-referencing. These changes
8: % do not reflect the original paper by A. V. Raveendran.
9: %
10: % Previous versions of this sample document were
11: % compatible with the LaTeX 2.09 style file mn.sty
12: % v1.2 released 5th September 1994 (M. Reed)
13: % v1.1 released 18th July 1994
14: % v1.0 released 28th January 1994
15:
16: \documentclass[useAMS,usenatbib,referee]{mn2e}
17:
18: % If your system does not have the AMS fonts version 2.0 installed, then
19: % remove the useAMS option.
20: %
21: % useAMS allows you to obtain upright Greek characters.
22: % e.g. \umu, \upi etc. See the section on "Upright Greek characters" in
23: % this guide for further information.
24: %
25: % If you are using AMS 2.0 fonts, bold math letters/symbols are available
26: % at a larger range of sizes for NFSS release 1 and 2 (using \boldmath or
27: % preferably \bmath).
28: %
29: % The usenatbib command allows the use of Patrick Daly's natbib.sty for
30: % cross-referencing.
31: %
32: % If you wish to typeset the paper in Times font (if you do not have the
33: % PostScript Type 1 Computer Modern fonts you will need to do this to get
34: % smoother fonts in a PDF file) then uncomment the next line
35: % \usepackage{Times}
36:
37: %%%%% AUTHORS - PLACE YOUR OWN MACROS HERE %%%%%
38:
39: \usepackage{subfigure}
40:
41: \usepackage{amsmath}
42: \usepackage{amssymb}
43: \usepackage{epsfig}
44:
45:
46:
47: \def\ggrav{\,{g_{\rm grav}}}
48: \def\grad{\,{g_{\rm rad}}}
49: \def\Gamrad{\,{\Gamma_e}}
50: \def\kms{\,{\rm km~s^{-1}}}
51: \def\mso{\,{\rm M}_\odot}
52: \def\rso{\,{\rm R}_\odot}
53: \def\msoy{\,{{\rm M}_\odot~{\rm yr}^{-1}}}
54: \def\vesc{\,{v_{\rm esc}}}
55:
56: \def\apj{\,{ApJ}}
57: \def\apjs{\,{ApJS}}
58: \def\nat{\,{Nat}}
59:
60: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
61:
62: \title[Numerical models of continuum-driven winds]
63: {Numerical simulations of continuum-driven winds of super-Eddington stars}
64: \author[A. J. van Marle et. al.]{A. J. van Marle$^{1}$\thanks{E-mail:
65: marle@udel.edu (AJvM)}, S. P. Owocki$^{1}$\thanks{E-mail:
66: owocki@bartol.udel.edu (SPO)} and N. J. Shaviv$^{2}$\thanks{E-mail:
67: shaviv@phys.huji.ac.il (NJS)}\\
68: % \footnotemark[1]\thanks{This file has been amended to
69: % highlight the proper use of \LaTeXe\ code with the class file.
70: % These changes are for illustrative purposes and do not reflect the
71: % original paper by A. V. Raveendran.}\\
72: $^{1}$Bartol Research Institute, University of Delaware, 104 The Green, Newark, DE 19716, USA\\
73: $^{2}$Racah Institute of Physics, Hebrew University, Giv'at Ram, Jerusalem 91904 Israel}
74:
75: \begin{document}
76:
77: \date{Accepted ?. Received ?; in original form ?}
78:
79: \pagerange{\pageref{firstpage}--\pageref{lastpage}} \pubyear{?}
80:
81: \maketitle
82:
83: \label{firstpage}
84:
85: \begin{abstract}
86: We present the results of numerical simulations of continuum-driven winds of stars that exceed the Eddington limit
87: and compare these against predictions from earlier analytical solutions.
88: Our models are based on the assumption that the stellar atmosphere consists of clumped matter, where the individual clumps
89: have a much larger optical thickness than the matter between the clumps.
90: This `porosity' of the stellar atmosphere reduces the coupling between radiation and matter, since photons tend to
91: escape through the more tenuous gas between the clumps.
92: This allows a star that formally exceeds the Eddington limit to remain stable, yet produce a steady outflow from the region
93: where the clumps become optically thin.
94: We have made a parameter study of wind models for a variety of input conditions in order to explore the properties of
95: continuum-driven winds.
96:
97: The results show that the numerical simulations reproduce quite closely the analytical
98: scalings.
99: The mass loss rates produced in our models are much larger than can be achieved by line driving.
100: This makes continuum driving a good mechanism to explain the large mass loss and flow speeds of giant outbursts, as observed
101: in {$\eta$~Carinae} and other luminous blue variable (LBV) stars.
102: Continuum driving may also be important in population III stars, since line driving becomes ineffective at low metalicities.
103: We also explore the effect of photon tiring and the limits it places on the wind parameters.
104: \end{abstract}
105:
106:
107: \begin{keywords}
108: hydrodynamics -- methods: numerical -- stars: mass loss -- stars: winds, outflows
109: \end{keywords}
110:
111:
112: \section{Introduction}
113:
114: Massive, hot stars continuously lose mass through radiation driving.
115: The most commonly explored mechanism is line driving, wherein the scattering of photons by ions in the stellar atmosphere
116: transfers momentum from the radiation field to the gas.
117: This mechanism results in a quiescent mass loss with mass loss rates ranging up to about $10^{-4}~\mso$~yr$^{-1}$
118: \citep{so06}, which can explain the winds of most massive, hot stars.
119: In fact, for most stars, the observed mass loss rates are
120: considerably less \citep{vk05}.
121:
122: However, some stars, most notably luminous blue variable stars (LBVs) such as
123: {$\eta$~Carinae}, experience outbursts with mass loss rates several orders of
124: magnitude higher than can be explained through line driving \citep{dh97,
125: s02, om08}.
126: In the case of $\eta$~Carinae, the 1840's outburst is inferred to
127: resulted in the ejection of ca. 10-20~$M_{\odot}$ over a time lasting
128: several years, even up to a decade
129: \citep{dh97}.
130: While short compared to evolutionary
131: timescale of millions of years, this is much longer
132: than the typical dynamical timescale of hours, characterized by either
133: free-fall time or interior sound travel time across a stellar radius.
134: Thus in contrast to SN ``explosions'' that are effectively driven by the
135: overpressure of superheated gas in the deep interior,
136: explaining LBV outbursts requires a more sustained mechanism that can
137: drive a quasi-steady mass loss, a stellar wind, from near the stellar surface.
138: In the case of $\eta$~Carinae the outburst was accompanied by a strong
139: increase in radiative luminosity, very likely making it well above the
140: Eddington limit for which {\em continuum} driving by just electron scattering would
141: exceed the stellar gravity.
142: The reason for this extended increase in luminosity is not yet understood, and
143: likely involves interior processes beyond the scope of this paper.
144: Instead, the focus here is on the way such continuum driving can result in a
145: sustained mass loss that greatly exceeds what is possible through line
146: opacity.
147:
148: A key feature of continuum driving is that, unlike line driving, it
149: does not become saturated from self-absorption effects in a very
150: dense, optically thick region.
151: Indeed, since both continuum acceleration and gravity scale with the inverse square of the radius, a star that
152: exceeds the Eddington limit formally becomes gravitationally unbound not only at the surface but throughout.
153: Clearly, this is in contradiction to the steady surface wind mass loss observed for these stars.
154: N.B. Contrary to what is sometimes claimed, this would not automatically destroy the entire star. Although radiative
155: acceleretion might overcome gravity locally, the total energy in the radiation field would not suffice to drive the entire
156: envelope of the star to infinity (this is known as `photon tiring'). Instead, the outward motion of the gas would quickly
157: stagnate and matter would start to fall back. Nevertheless, the net result would not resemble a steady wind.
158:
159: This problem can be resolved by assuming that the stellar material is clumped rather than homogeneous, with the individual
160: clumps being optically thick - and therefore self-shielding from the radiation - whereas the medium in between the clumps is
161: relatively tranparent to radiation.
162: This so called `porosity effect' can lead to a reduced coupling between matter and radiation \citep{s98,s00}.
163: The photons tend to escape through the optically thin material between the clumps without interacting with the matter inside
164: the clumps.
165: This implies that a star that formally exceeds the Eddington limit can remain gravitationally bound and would only exceed
166: the effective Eddington limit at the radius where the individual clumps themselves become optically thin.
167:
168: The structure of such a star should therefore look as follows \citep{s01}.
169: Deep inside the super-Eddington star, convection is necessarily excited \citep{jso73} such that the radiation field remains
170: sub-Eddington through most of the stellar interior.
171: At low enough densities where maximally efficient convection cannot sufficiently reduce the radiative flux, the near
172: Eddington luminosity necessarily excites at least one of several possible instabilities \citep[e.g.,]{a92,s01a} which give
173: rise to a reduced opacity. This `porous' layer has a reduced effective opacity and an increased effective Eddington
174: luminosity. Thus, the layer remains gravitationally bound to the star.
175: At lower densities still, the dense clumps become optically thin and the effective opacity approaches its microscopic value.
176: From this radius outwards, the matter is gravitationally unbound, and is part of a continuum-driven wind.
177:
178: A detailed analytical study of this paradigm was carried out by {Owocki, Gayley \& Shaviv (2004)}, hereafter \citet{ogs04}.
179: This predicted that continuum-driven winds can produce high mass loss rates ($\geq10^{-3}~\msoy$) at intermediate wind
180: velocities ($10^2-10^3~\kms$).
181: Here we test these analytical predictions with numerical simulations of winds from super-Eddington stars.
182:
183: In addition to LBVs, which are the specific objects we study here, other types of astronomical objects can exceed the
184: Eddington limit and therefore experience similar continuum-driven winds.
185: These include for example classical novae \citep{s01} or high accretion rate accretion disks around black holes \citep{b06}.
186: In fact, classical nova eruptions clearly exhibit steady continuum-driven winds, indicating that the mass loss rate is
187: somehow being regulated and likely to be described by the porosity model and the same continuum-driven wind analyzed here.
188:
189:
190: The layout of this paper is as follows. In \S\ref{sec-subEdd} we show the effect of continuum scattering on a
191: sub-Eddington, line-driven wind.
192: In \S\ref{sec-analytic} we summarize the analytic results obtained by \citet{ogs04}.
193: In \S\ref{sec-numeric} we describe the numerical methods that we have used for our simulations.
194: \S\ref{sec-result} shows the results of our simulations and the comparison with the analytical predictions.
195: In \S\ref{sec-phtir} we discuss the effect of photon tiring and show how it influences the results of our calculations.
196: Finally, in \S\ref{sec-disc} we end with a summary and a discussion.
197: %
198: %
199: %
200: \begin{figure*}
201: \centering
202: \resizebox{\hsize}{!}{\includegraphics[width=\textwidth,angle=-90]{f01.eps}}
203: \caption{
204: Left: The effect of continuum driving on a sub-Eddington, line-driven wind for $\Gamrad=0.1$. The solid lines depict the
205: wind parameters for the CAK force only, the dashed lines show the joint effect of both CAK line driving and continuum
206: scattering. Clearly, the influence of the continuum term is negligeable.
207: Right: The same variables , but for $\Gamrad=0.5$. Here the difference caused by continuum driving is significantly more
208: pronounced.}
209: \label{fig:line1}
210: \end{figure*}
211:
212:
213: \section{Sub-Eddington limit}
214: \label{sec-subEdd}
215: As long as a star does not exceed the Eddington limit, continuum interaction alone cannot drive a stellar wind.
216: However, it does influence the parameters of a (primarily) line-driven wind.
217: Figure~\ref{fig:line1} demonstrates the influence of continuum driving on the stellar wind parameters.
218: For these simulations we use the CAK line driving formalism described in \citet{do02}, including both the finite disk
219: correction factor and the modified gravity scaling of the line force (eq.~[19] of \citet{ogs04}). For the same stellar
220: parameters ($50~\mso$, $25~\rso$, 50\,000~K) and CAK parameters (${\bar{Q}=2.0\times10^3}$,
221: ${\alpha=2/3}$) but two different
222: Eddington parameters ($\Gamrad=L_\star/L_{\rm Edd}=0.1$ and 0.5,)
223: we simulate the stellar wind with and without the continuum
224: driving term.
225:
226: Note that unlike the purely continuum-driven winds in the next section, these simulations do not contain any porosity effect.
227: Namely, the continuum driving term in the equation of motion is simply:
228: \begin{equation}
229: g_{\rm cont}~=~ \kappa_e \frac{F_{\rm rad}}{c},
230: \end{equation}
231: with the continuum opacity $\kappa_e$ set to $0.3$~g/cm$^2$.
232: {\bf These parameters give us an Eddington luminosity of ${L_{\rm Edd}=4\pi GcM/\kappa_e=8.33\times10^{39}}$~erg/s.}
233: Figure~\ref{fig:line1} shows both the density (relative to the density at the stellar surface) and wind velocity as a
234: function of radius for these two simulations.
235: Clearly, for a low Eddington factor, the effect of continuum interaction on the wind parameters is negligeable.
236: However, once the star approaches the Eddington limit, continuum driving increases the mass loss rate and decreases the
237: terminal velocity.
238: For $\Gamrad=0.1$ the mass loss rate increases from $1.80\times10^{-6}\msoy$ to $1.92\times10^{-6}\msoy$
239: because of continuum driving. For $\Gamrad=0.5$, these numbers are $2.18\times10^{-6}\msoy$ and $3.22\times10^{-6}\msoy$
240: respectively.
241: %
242: %
243: %
244: \section{Analytical approximation}
245: \label{sec-analytic}
246: \subsection{Wind velocity}
247: For a radiatively driven outflow from a stellar surface, irrespective of the specific driving mechanism involved, the
248: equation of motion can be written as
249: \begin{equation}
250: \biggl[1-\frac{a^2}{v^2}\biggr] v\frac{dv}{dr}~=~-\frac{G M_\star}{r^2} + \grad + \frac{2a^2}{r}- \frac{{\rm d}a^2}{{\rm
251: d}r},
252: \label{eq:mot1}
253: \end{equation}
254: with $a$ the isothermal sound speed, $\grad$ the radiative acceleration and ${G M_\star}/{r^2}$ the inward gravitational
255: acceleration.
256: The last two terms on the right hand side are the result of the gas pressure gradient.
257: In most cases they be neglected as their contribution to the velocity is usually negligeable next to the gravitational and
258: radiative acceleration terms \citep{ogs04}.
259: This assumption becomes incorrect only in those situations where the terminal velocity of the wind is close to the sound
260: speed.
261: If we neglect these gas pressure terms, we can rewrite equation (\ref{eq:mot1}) in a dimensionless form,
262: \begin{equation}
263: \biggl[1-\frac{w_s}{w}\biggr]w'~=~-1+\Gamrad,
264: \label{eq:mot2}
265: \end{equation}
266: with $w=(v/\vesc)^2$, $w_s=(a/\vesc)^2$ and $\Gamrad=\grad r^2/GM_\star$.
267: Note that $\vesc$ is defined as the escape velocity at the surface of the star $R_\star$ rather than as a local escape
268: velocity.
269: The gravitationally scaled inertial acceleration,
270: \begin{equation}
271: w'~=~\frac{r^2 v {\rm d}v/{\rm d}r}{GM_\star},
272: \end{equation}
273: can be written in terms of the inverse radius coordinate $x\equiv1-R_\star/r$, so that $w'={\rm d}w/{\rm d}x$.
274: This inverse radius coordinate makes for a more practical coordinate system than the radius itself, since it scales
275: with the gravitational potential.
276:
277: Since the escape velocity of a massive star is typically at least an order of magnitude larger than the local
278: isothermal sound speed, we can in many cases neglect the bracketed term.
279: For the porous atmosphere, \citet{ogs04} introduced a dimensionless parameter $k_b=\kappa_{\rm eff}/\kappa$,
280: which gives the
281: effective opacity of the local clumped medium vs.\ the opacity of the medium if it were completely homogeneous.
282: Like \citet{ogs04}, we assume that the distribution $(f)$ of optical depths of the individual clumps can be described by a
283: truncated power law,
284: \begin{equation}
285: \tau \frac{{\rm d}f}{{\rm d}\tau}~=~\frac{1}{\Gamma(\alpha_p)}\biggl(\frac{\tau}{\tau_0}\biggr)^{\alpha_p}
286: e^{-\tau/\tau_0},
287: \end{equation}
288: with $\alpha_p \geq 0$ being the power-index and $\tau_0$ the optical depth of the {\it thickest} clump.
289: Here $\Gamma(\alpha_p)$ is the gamma function, rather than the Eddington parameter.
290: This gives us:
291: \begin{equation}
292: k_b~=~\frac{(1+\tau_0)^{1-\alpha_p} - 1}{(1-\alpha_p)\tau_0}.
293: \end{equation}
294: For the full derivation see \citet{ogs04}.
295:
296: As long as the clump characteristics are fixed, the optical depth $\tau_0$ scales with the density as,
297: \begin{equation}
298: \tau_0 = \rho/\rho_0,
299: \end{equation}
300: where $\rho_0$ is the critical density at which the thickest blob has unit optical depth.
301: This is calculated by
302: \begin{equation}
303: \rho_0~=~\frac{1}{\kappa h_0},
304: \end{equation}
305: with $\kappa$ the continuum opacity and $h_0$ the porosity length of the thickest clump; the porosity length $h$ is defined
306: as $L^3/l^2$, with $L$ being the typical separation between clumps and $l$ the size of the clump.
307:
308:
309: We thus rewrite eq.~(\ref{eq:mot2}) by substituting,
310: \begin{equation}
311: \Gamrad \longrightarrow \Gamrad k_b(\tau_0),
312: \end{equation}
313: for the Eddington parameter $\Gamrad=L_\star/L_{\rm Edd}$.
314: % For simplicity, we ignore gas pressure terms associated with the
315: % spherical expansion; but to ensure a smooth transition through the
316: % sonic point, we retain the pressure gradient associated with flow
317: % acceleration.
318: The scaled velocity gradient thus takes the form,
319: \begin{equation}
320: w'(x)~=~ \frac{\Gamrad k_b [\tau_0(x)] -1}{1-w_{s}/w}
321: \, .
322: \label{eq:w}
323: \end{equation}
324: This is a first-order differential equation that can be integrated using standard numerical
325: techniques, starting from the sonic-point initial condition $w(x=0)=w_{s}$,
326: where the vanishing of both the numerator and denominator on the right-hand side
327: requires application of L'Hopital's rule to evaluate the initial gradient,
328: \begin{equation}
329: w'_{s} \equiv w'(x=0) = \frac{\alpha_{p}}{4} \,
330: \left [ 1 + \sqrt{1+32 w_{s}/\alpha_{p}} \right ]
331: \, .
332: \label{eq:wps}
333: \end{equation}
334:
335:
336: By introducing the porosity term into the equation of motion, we have
337: made the radiative acceleration dependent on the density and therefore
338: on the radius.
339: Under these circumstances, the radiative acceleration is only able to drive matter away from the star
340: in the outermost layers of the star,
341: producing a steady wind that can last as long as the star remains
342: about the Eddington limit.
343: In effect, porosity {\em regulates} the mass loss to a level that can be
344: sustained in a nearly steady way throughout an LBV outburst.
345: Here again the word ``outburst'' has to be interpreted in context.
346: LBV outbursts can last years, which, though short compared to stellar evolutionary time,
347: is quite long compared to a SN explosion.
348: Most importantly, this is much longer
349: than a typically dynamical flow time (ca. a free fall time, $R/\vesc$, or a
350: wind expansion time, $R/v_\infty$).
351: As such, the mass loss can be well modeled in
352: terms of steady-state wind solutions that assume the luminosity, etc. are
353: constant over such dynamical timescales.
354:
355:
356: \subsection{Mass loss rate}
357: The mass loss rate induced by continuum driving is directly related to the luminosity of the star and the power index
358: $\alpha_p$.
359: The generic mass loss rate that follows from the porosity is
360: \begin{equation}
361: \dot{M}~=~\biggl(\frac{\rho_\star}{\rho_0}\biggr)\biggl(\frac{H}{h_0}\biggr)
362: % \biggl(\frac{1}{\Gamma}\biggr)
363: \frac{L_{\rm Edd}}{ac},
364: \label{eq:mdotbasic}
365: \end{equation}
366: with $H$ the gravitational scale height of the gas at the sonic point.
367: The sonic point density, $\rho_\star$, can be found through condition
368: \begin{equation}
369: \Gamrad k_b~=~1.
370: \label{eq:gamkbeq1}
371: \end{equation}
372: For the simple case $\alpha_{p} = 1/2$, this has the explicit solution
373: \begin{equation}
374: \biggl(\frac{\rho_\star}{\rho_0}\biggr)
375: = 4 \Gamrad \, (\Gamrad - 1) ~~~ ; ~~~ \alpha_{p} = \frac{1}{2}
376: \end{equation}
377: which leads to an explicit expression for the mass loss rate
378: (cf. eqn. (77) in \cite{ogs04}),
379: \begin{equation}
380: \dot{M} = 4(\Gamrad-1)\frac{L_\star}{\eta_0 a c} ~~~ ; ~~~ \alpha_p=\frac{1}{2}
381: \, ,
382: \label{eq:mdot}
383: \end{equation}
384: with ${\eta_0 \equiv h_0/H}$.
385: For more general cases, \citet{ogs04} also give explicit
386: approximations for the expected mass loss rate, but in quoting values
387: for this ``analytic porosity model'' below, we choose the more
388: accurate approach of solving eqn.~(\ref{eq:gamkbeq1}) implicitly, and
389: using the resulting $\rho_{\star}$ to compute the associated mass loss rate
390: from eq.~(\ref{eq:mdotbasic}).
391:
392:
393: \section{Numerical method}
394: \label{sec-numeric}
395: For our numerical simulations we use the ZEUS hydrodynamics code \citep{sn92, c96}.
396: We do our computations on a 1D spherical grid with an inflow boundary in the center and outflow at the outer boundary.
397: Continuum driving is modeled through an acceleration term $\Gamrad k_b \ggrav$
398: added to the radial component of the equation of motion (see also, \citet{mos08a}).
399:
400:
401: We start our simulation by initializing a 1D radial grid with matter moving away from the origin with a power law velocity
402: distribution.
403: At the inner boundary, we specify an inflow density, higher than the critical density $\rho_s$ to ensure that the sonic
404: point (where the wind velocity equals the isothermal sound speed) is inside the grid.
405: This inflow density remains constant during the simulation.
406: The inflow velocity is time-dependent, and recalculated at each timestep such that the velocity gradient over the boundary
407: is equal to the velocity gradient immediately above the boundary radius.
408: Such ``floating'' boundary conditions allow approach to a stable,
409: steady, flow solution, and have been successfully used for
410: line-driven-wind simulations in both one \citep{ocr88} and two dimensions \citep{ocb94}.
411:
412: The radial grid is not evenly spaced, but rather chosen in such a fashion that the individual grid size at the sonic point
413: is always smaller than the local scale height.
414:
415: For our simulations we assume that the gas is isothermal with a temperature of 50\,000\,K, which implies an isothermal sound
416: speed of approximately $20~\kms$.
417: The opacity of the gas is set to 0.4 gr/cm$^2$ (Thompson scattering opacity for pure hydrogen) and $\eta_0$ to 1.
418: The radius and mass of the central star are set to 50~$\rso$ and 50~$\mso$ respectively.
419: {\bf A star such as this has an Eddington luminosity of $6.24\times10^{39}$~erg/s.}
420:
421: The grid extends from the stellar surface to a distance of ten times the stellar radius.
422:
423: \begin{figure}
424: \centering
425: \resizebox{\hsize}{!}{\includegraphics[width=\textwidth]{f02.eps}}
426: %%\plotone{f02.eps}
427: \caption{Wind velocity and density as a function of radius for a continuum-driven wind. On the horizontal axis, the local
428: radius divided by the sonic point radius. Independent parameters: $\Gamrad=3$, $\alpha_p=0.5$, $w_\star=0.001$}
429: \label{fig:windprofile}
430: \end{figure}
431:
432: \begin{figure}
433: \centering
434: \resizebox{\hsize}{!}{\includegraphics[width=\textwidth]{f03.eps}}
435: % \resizebox{\hsize}{!}{\includegraphics[width=\textwidth]{f04.eps}}
436: %%\plotone{f03.eps}
437: \caption{
438: Wind velocity in units of the sonic point escape radius plotted versus
439: the dimensionless radial coordinate $x=1-R_{sonic}/r$,
440: comparing results for the analytical solution (dashed curves) and numerical
441: simulation (solid curves) for a continuum-driven wind.
442: The upper curves show results for the same parameters as
443: figure~\ref{fig:windprofile}, with $\Gamrad=3$ and $\alpha_p=0.5$,
444: while the lower curves are for the marginal super-Eddington case
445: $\Gamrad=1.1$, with $\alpha_{p}=0.1$.
446: The agreement between numerical and analytical solutions is quite
447: good, with only a slightly stronger differences for the $\Gamrad=1.1$
448: case.
449: }
450: \label{fig:vel12}
451: \end{figure}
452:
453: % \begin{figure}
454: % \centering
455: % \resizebox{\hsize}{!}{\includegraphics[width=\textwidth]{f04.eps}}
456: % %%\plotone{f04.eps}
457: % \caption{Similar to figure \ref{fig:vel1}, except for parameters $\Gamrad=1.1$
458: % and $\alpha_p=0.1$. The near-unity value of $\Gamrad$ means that a
459: % lower net acceleration, making the thermal pressure relatively more important
460: % and leading to differences between the analytical and numerical results
461: % that are more pronounced than in figure~\ref{fig:vel1}.}
462: % \label{fig:vel2}
463: % \end{figure}
464:
465:
466: \section{Results}
467: \label{sec-result}
468: A grid of 30 models was calculated, encompassing a parameter space of $0.1\leq\alpha_p \leq 0.8$ and $1.1\leq\Gamrad \leq
469: 6.0$.
470: Stellar mass, radius surface temperature, opacity and clumping parameter $\eta_0$ were kept constant.
471: A typical result for these calculations is shown in figure~\ref{fig:windprofile}, which shows the density and wind velocity
472: as a function of distance from the star.
473:
474:
475: \clearpage
476: \begin{table*}
477: \centering
478: % \begin{tabular}{|p{1cm}|p{2cm}|r|r|r|r|}
479: \begin{tabular}{p{1cm} p{3cm} p{2cm} p{2cm} c r}
480: \hline
481: \hline
482: \noalign{\smallskip}
483: $\Gamrad$ & $\alpha_p$ & $\dot{M}_{an}$ & $\dot{M}_{num}$ & $\dot{M}_{an}/\dot{M}_{num}$ &v/$\vesc$ \\
484: & & $\mso yr^{-1}$ & $\mso yr^{-1}$ & \\
485: \noalign{\smallskip}
486: \hline \hline
487: % \noalign{\smallskip}
488: 6.0 & 0.8 & 4.86$\times 10^{-2}$ & 4.87$\times 10^{-2}$ & 0.997 & 1.93 \\
489: 6.0 & 0.5 & 1.98$\times 10^{-1}$ & 1.99$\times 10^{-1}$ & 0.997 & 1.80 \\
490: 6.0 & 0.3 & 2.09 & 2.10 & 0.994 & 1.50 \\
491: 6.0 & 0.2 & 3.92$\times 10^{1}$ & 3.99$\times 10^{1}$ & 0.982 & 1.13 \\
492: 6.0 & 0.1 & 2.87$\times 10^{5}$ & 3.00$\times 10^{5}$ & 0.956 & 0.655 \\
493: \hline
494: 4.0 & 0.8 & 2.43$\times 10^{-2}$ & 2.44$\times 10^{-2}$ & 0.995 & 1.53 \\
495: 4.0 & 0.5 & 7.94$\times 10^{-2}$ & 7.97$\times 10^{-2}$ & 0.996 & 1.47 \\
496: 4.0 & 0.3 & 5.23$\times 10^{-1}$ & 5.28$\times 10^{-1}$ & 0.991 & 1.31 \\
497: 4.0 & 0.2 & 5.13 & 5.23 & 0.981 & 1.08 \\
498: 4.0 & 0.1 & 4.97$\times 10^{3}$ & 5.21$\times 10^{3}$ & 0.954 & 0.655 \\
499: \hline
500: 3.0 & 0.8 & 1.42$\times 10^{-2}$ & 1.42$\times 10^{-2}$ & 0.997 & 1.26 \\
501: 3.0 & 0.5 & 3.97$\times 10^{-2}$ & 4.00$\times 10^{-2}$ & 0.992 & 1.23 \\
502: 3.0 & 0.3 & 1.91$\times 10^{-1}$ & 1.93$\times 10^{-1}$ & 0.990 & 1.14 \\
503: 3.0 & 0.2 & 1.20 & 1.22 & 0.985 & 1.00 \\
504: 3.0 & 0.1 & 2.80$\times 10^{2}$ & 2.92$\times 10^{2}$ & 0.959 & 0.654 \\
505: \hline
506: 2.0 & 0.8 & 5.83$\times 10^{-3}$ & 5.92$\times 10^{-3}$ & 0.984 & 0.907 \\
507: 2.0 & 0.5 & 1.32$\times 10^{-2}$ & 1.34$\times 10^{-2}$ & 0.987 & 0.894 \\
508: 2.0 & 0.3 & 4.20$\times 10^{-2}$ & 4.27$\times 10^{-2}$ & 0.983 & 0.862 \\
509: 2.0 & 0.2 & 1.47$\times 10^{-1}$ & 1.50$\times 10^{-1}$ & 0.980 & 0.812 \\
510: 2.0 & 0.1 & 4.83 & 5.06 & 0.955 & 0.633 \\
511: \hline
512: 1.5 & 0.8 & 2.53$\times 10^{-3}$ & 2.57$\times 10^{-3}$ & 0.984 & 0.643 \\
513: 1.5 & 0.5 & 4.96$\times 10^{-3}$ & 5.09$\times 10^{-3}$ & 0.974 & 0.643 \\
514: 1.5 & 0.3 & 1.19$\times 10^{-2}$ & 1.22$\times 10^{-2}$ & 0.978 & 0.632 \\
515: 1.5 & 0.2 & 2.87$\times 10^{-2}$ & 2.95$\times 10^{-2}$ & 0.973 & 0.616 \\
516: 1.5 & 0.1 & 2.61$\times 10^{-1}$ & 2.72$\times 10^{-1}$ & 0.958 & 0.551 \\
517: \hline
518: 1.1 & 0.8 & 4.33$\times 10^{-4}$ & 4.88$\times 10^{-4}$ & 0.888 & 0.301\\
519: 1.1 & 0.5 & 7.27$\times 10^{-4}$ & 8.10$\times 10^{-4}$ & 0.898 & 0.300\\
520: 1.1 & 0.3 & 1.32$\times 10^{-3}$ & 1.48$\times 10^{-3}$ & 0.892 & 0.299\\
521: 1.1 & 0.2 & 2.20$\times 10^{-3}$ & 2.47$\times 10^{-3}$ & 0.891 & 0.298\\
522: 1.1 & 0.1 & 6.08$\times 10^{-3}$ & 6.81$\times 10^{-3}$ & 0.892 & 0.293\\
523: % \noalign{\smallskip}
524: \hline \hline
525: \end{tabular}
526: \caption{ For parameters $\Gamrad$ and $\alpha_{p}$ given in the
527: first two columns, comparison of the mass loss rates predicted by
528: the analytic porosity model (column 3) vs. those obtained in full
529: numerical simulations (column 4), with the fifth column giving the
530: ratio, analytical/numerical. The agreement is remarkably good, but with
531: modest, ca. 10\% differences for lower $\alpha_{p}$ or $\Gamrad$
532: near unity. The last column gives the ratio between the terminal velocity and
533: the escape speed from the sonic radius for the numerical models.}
534: \label{tab:mdot}
535: \end{table*}
536:
537:
538: Table~\ref{tab:mdot} compares mass loss rates for the numerical and analytical solutions for
539: a sample of model parameters $\Gamrad$ and $\alpha_{p}$.
540: As noted above, the analytical mass loss rate is computed from
541: eqn.~(\ref{eq:mdotbasic}) using the sonic point density derived from
542: implicit solution of eqn.~(\ref{eq:gamkbeq1}).
543: The comparison shows that the analytical and numerical results coincide very
544: well, with maximum differences of about 10\% for $\Gamrad$ near unity
545: and small $\alpha_{p}$.
546:
547: The wind velocity can be approximated semi-analytically by integrating equation~(\ref{eq:w}).
548: The results again match the numerical simulation quite well, as plotted in
549: figure~\ref{fig:vel12}.
550:
551: The match is partcularly good for the standard parameter set used in figure \ref{fig:windprofile}, with
552: $\Gamrad=3$ and $\alpha_{p}=0.5$.
553: For the marginal super-Eddington
554: model $\Gamrad=1.1$ with small power index $\alpha_{p} = 0.1$,
555: which represents an almost pathogical case,
556: the differences are greater, but still quite modest.
557: % Typically they are within a few percent of one another.
558: % , with the numerical results always slightly higher due to the
559: % influence of thermal pressure.
560: % This only breaks down at low $\Gamrad$ where the thermal pressure becomes an important part of the
561: % momentum balance, which means that equation (\ref{eq:w}) no longer applies.
562: % For example, in fig.~\ref{fig:vel2} the difference between the analytical and numerical results is much larger than in
563: % figure \ref{fig:vel1}.
564: % In this case the wind velocity is never much higher than the sound speed, which means that the thermal pressure contributes
565: % significantly to the acceleration.
566: % {\bf In both Fig.~\ref{fig:vel1} and \ref{fig:vel2} the difference is largest close to the sonic
567: % point, where the analytical solution has an initial plateau with
568: % nearly constant velocity.
569: % This is a consequence of neglecting gas pressure in the analytical
570: % model.
571: % The neglect of gas pressure in the analytical model means that its
572: % initiation at the sonic point
573: % reflecting the zero velocity gradient.
574: % Since the radiative driving force at the sonic point is equal to gravity there is no net force and therefore no acceleration
575: % until the radiative force exceeds gravity beyond the sonic point.
576: % In the numerical model, the thermal pressure stratification creates an extra acceleration term, which ensures a positive
577: % velocity gradient at the sonic point.}
578:
579: The complete results of the parameter study are shown in fig.~\ref{fig:dM}, which shows the mass loss rate and terminal
580: wind velocity for all simulations.
581: As expected, the mass loss rate increases with $\Gamrad$ and decreases with $\alpha_p$ as predicted in
582: \S~\ref{sec-analytic}.
583: The terminal wind velocity increases both with $\Gamrad$ and with $\alpha_p$.
584: This is hardly surprising.
585: A high $\Gamrad$ means a strong radiative acceleration, which will lead to a higher velocity.
586: A high $\alpha_p$ means a weaker coupling between matter and radiation.
587: This decreases the mass loss rate, which for a given total available energy means an increase in velocity.
588:
589: Mass loss rates in general are very high, though one should remember that photon-tiring was not included in these
590: simulations.
591: Although photon tiring does not change the mass loss rate at the stellar surface, it can prevent that part of the mass
592: escapes the stellar gravity.
593: Wind velocities are usually on the order of the escape velocity.
594:
595:
596: \begin{figure*}
597: \centering
598: %\mbox{\subfigure{\epsfig{figure=f05.eps,width=0.45\textwidth,angle=-90}}\quad
599: % \subfigure{\epsfig{figure=f06.eps,width=0.45\textwidth,angle=-90}}
600: % }
601: \mbox{\subfigure{\includegraphics[width=0.45\textwidth,angle=-90]{f05.eps}}
602: \subfigure{\includegraphics[width=0.45\textwidth,angle=-90]{f06.eps}}}
603: \caption{
604: Left:
605: Total mass loss rate in solar masses per year as a function of $\Gamrad$ and $\alpha_p$. As predicted in the analytical
606: approximations (see \S\ref{sec-analytic}) the mass loss rate increases with $\Gamrad$ but decreases with $\alpha_p$.
607: Right:
608: Terminal wind velocity divided by escape velocity as a function of $\Gamrad$ and $\alpha_p$. The wind velocity increases with
609: both $\Gamrad$ and $\alpha_p$.
610: The white circles indicate the actual datapoints, taken from table~\ref{tab:mdot}.
611: }
612: \label{fig:dM}
613: \end{figure*}
614:
615: \section{Photon tiring}
616:
617: One of the most interesting aspects of continuum-driven winds is their high mass loss rates, which gives rise to the
618: phenomenon of ``photon tired winds" suggested by \citet{og97}. That is, winds in which the energy required to drive the gas
619: out of the gravitational well is comparable or even larger than the energy available in the radiation field. In
620: \S\ref{sec-tired-analytic} we briefly cover the analytical description of these winds, and then continue with a numerical
621: study of the case in which there is sufficient energy in the radiation field. The opposite limit, which gives rise to
622: non-stationary solutions will be dealt in the subsequent publication.
623:
624: \label{sec-phtir}
625: \subsection{Analytical approximation}
626: \label{sec-tired-analytic}
627: If the stellar wind is driven by radiation, conservation of energy limits the maximum mechanical luminosity of the star to
628: the radiative luminosity.
629: This effect is known as photon tiring and was not included in the simulations described in the previous section.
630:
631: The maximum amount of mass that can be driven of by radiation is the mass loss rate for which the mechanical luminosity of
632: the wind matches the radiative luminosity,
633: \begin{equation}
634: \frac{1}{2}\dot{M}\vesc^2~\leq~L_{\rm tir}~=~L_{\star},
635: \end{equation}
636: with $L_\star$ the radiative luminosity of the star.
637: This sets the upper limit for the mass loss rate to,
638: \begin{equation}
639: \dot{M}_{\rm tir}~=~\frac{2 L_\star}{\vesc^2}.
640: \end{equation}
641: Note that this is the same for both line driving and continuum driving, and is unrelated to the composition of the stellar
642: atmosphere.
643: It is a fundamental property of the star.
644: From this limit we calculate the photon tiring number~$m$,
645: \begin{equation}
646: m~=~\frac{\dot{M}}{\dot{M}_{\rm tir}}.
647: \label{eq:mtir}
648: \end{equation}
649: As long as ${m \ll 1}$, photon tiring has no significant influence on the wind properties, but when ${m \lesssim 1}$ the
650: situation changes.
651: The radiation field of the star becomes depleted as much of its energy is used to drive the matter, leaving less energy
652: available to drive the outer layers of the wind.
653: The mass loss rate at the surface remains the same since it is set by the local radiative acceleration, but the wind
654: velocity will decrease.
655:
656: This will also influence observations, since radiative energy that is used to drive the wind will no longer be part of the
657: visible radiation output of the star.
658: For stars with line-driven winds, this is hardly worth considering.
659: Even a Wolf-Rayet star with a powerful wind loses only a few percent of its luminosity to the wind. For continuum-driven
660: winds, the effect can become quite significant.
661:
662: Should $m$ actually exceed unity, the radiation field loses all its energy.
663: As a result, the matter that leaves the stellar surface will not be able to reach the escape velocity and start to fall back
664: onto the star.
665: This will make it impossible to obtain a steady state solution.
666:
667: A analytic analysis of the photon tiring effect was done by \cite{ogs04}, predicting that while the mass loss rate would
668: remain the same, the velocity of the wind would drop.
669: The new equation for the wind velocity becomes:
670: \begin{equation}
671: w'(x)~=~\frac{\kappa_{\rm eff}[\tau_0(x)]}{\kappa} \Gamrad[1-m(w+x)] - 1.
672: \label{eq:wtiring}
673: \end{equation}
674: Note that $\Gamrad$ is no longer a constant but decreases with the radius and the velocity.
675: The photon tiring number can be found by combining equations (\ref{eq:mdot}) and (\ref{eq:mtir}).
676: For the special case where $\alpha_p=1/2$ this implies that,
677: \begin{equation}
678: m~=~0.13\frac{\Gamrad -1}{\eta_0 a_{20}}\frac{M_\star}{\mso}\frac{\rso}{R_\star},
679: \end{equation}
680: with $a_{20}$ the sound speed in units of $20~{\rm km/s}$.
681: The implications of this equation are quite clear.
682: The relative effect of photon tiring increases with $\Gamrad$ and decreases with the porosity length.
683: This is to be expected, as a larger $\Gamrad$ means an increase in mass loss rate, and therefore an increase in the amount
684: of energy necessary to lift the material.
685: A larger $\eta_0$ means a decrease in coupling between radiation and matter, such that the effect of photon tiring
686: diminishes.
687: The same is true for an increase in $\alpha_p$, so we can expect the effect of photon tiring to diminish with higher
688: $\alpha_p$.
689:
690:
691: \subsection{Numerical method}
692: We calculate the effect of photon tiring by calculating the work integral along the radial gridline and subtracting the
693: result from the total luminosity of the star.
694: This provides us with the luminosity available at each radial grid point, which can then be used to accelerate the wind
695: during the next time step.
696:
697: The work integral is given by:
698: \begin{equation}
699: \begin{aligned}
700: W~&=~\int_{R\star}^r \dot{m}(r) \grad {\rm d}r, \\
701: &=~4\pi\int_{R\star}^r \rho(r) r^2 v(r) \grad {\rm d}r.
702: \end{aligned}
703: \end{equation}
704: Therefore, the effective luminosity at a given radius equals:
705: \begin{equation}
706: \begin{aligned}
707: L~&=~L_\star - W, \\
708: &=~L_\star - 4\pi\int_{R\star}^r \rho(r) r^2 v(r) \grad {\rm d}r
709: \end{aligned}
710: \end{equation}
711: \citep{mos08a}.
712: Using this value for the luminosity, rather than the stellar luminosity used in the simulations described in
713: \S\ref{sec-result}, a new set of simulations was made for the same range of $\Gamrad$ values and $\alpha_p=0.5$.
714: All other parameters were kept the same as in \S\ref{sec-result}.
715:
716: \begin{figure}
717: \centering
718: \resizebox{\hsize}{!}{\includegraphics[width=\columnwidth]{f07.eps}}
719: %%\plotone{f07.eps}
720: \caption{The effect of photon tiring on the wind velocity (I). Clearly, the effect is much larger for larger $\Gamrad$.}
721: \label{fig:vtir}
722: \end{figure}
723:
724: \begin{figure}
725: \centering
726: \resizebox{\hsize}{!}{\includegraphics[width=\columnwidth]{f08.eps}}
727: %%\plotone{f08.eps}
728: \caption{The effect of photon tiring on the wind velocity (II). This graph depicts the terminal velocity as a function of
729: $\Gamrad$ for simulations with and without photon tiring included. The effect clearly increases with
730: $\Gamrad$.}
731: \label{fig:vtir2}
732: \end{figure}
733:
734: \subsection{Numerical results}
735: The effect of photon tiring is demonstrated in figure~\ref{fig:vtir}.
736: For $\Gamrad=2$, the effect is small, whereas for $\Gamrad=6$ the effect is very large.
737: Also, for the larger value of $\Gamrad$, the velocity decreases beyond a certain radius.
738: This corresponds to the radius where the luminosity falls below the Eddington luminosity.
739: Note that the mass loss rate is independent of the tiring parameter and will therefore not change if photon tiring is
740: included.
741: This implies that the density of the wind increases, since the velocity decreases due to the photon tiring effect.
742: It therefore also implies that the optical depth of the wind is higher as well.
743:
744: In figure~\ref{fig:vtir2}, one can see a comparison between the values of the terminal velocity obtained with and without
745: photon tiring.
746: The influence of photon tiring clearly increases with $\Gamrad$.
747: For large $\Gamrad$, the terminal velocity itself actually decreases.
748: It is clear from these results that for larger values of $\Gamrad$ the photon tiring effect must be included in the
749: numerical simulations, or the results will quickly become unphysical.
750:
751: \section{Discussion}
752: \label{sec-disc}
753: A series of numerical simulations was carried out in order to test the analytical approximations for the porosity length
754: formalism of continuum driving, published by \cite{ogs04}.
755: The numerical results coincide well with the analytical results and demonstrate that this mechanism allows for powerful
756: radiation-driven winds.
757: This effect can explain the mass loss rates and wind velocities observed in Luminous Blue Variables such as
758: {$\eta$~Carinae}.
759:
760: The simulations also confirm that the photon tiring effect plays an important role in continuum driving as it places an
761: upper limit on the mass loss rate.
762: The effects of actually crossing the photon tiring limit have not yet been explored.
763: This situation is much more complicated, since the simulations will no longer be able to reach a steady state solution.
764: Ideally, such simulations should be done in two, or even three dimensions, to investigate the effect of
765: interactions between different layers of the stellar wind as they move back and forth.
766: Note that at this point we cannot predict how the star itself would react to such a situation.
767: All our simulations have been done under the assumption that stellar parameters do not change significantly over time.
768: It is possible that conditions in the outer layers of the star would change to reduce the driving force.
769:
770: Porosity reduced continuum driving can also be important for the winds of other super-Eddington objects such as Novae
771: \citep{s01}, accretion disks \citep{b06} and transients like {M85OT2006-1} \citep{ketal07}.
772:
773: In a companion paper, \citep{mos08b}, we explore the situation where the star exceeds the photon tiring limit.
774: We also intend eventually to generalize the simulations to multiple dimensions and explore the influence of stellar rotation
775: on continuum-driven winds.
776:
777:
778: \section*{Acknowledgments}
779:
780: A.J.v.M. acknowledges support from NSF grant AST-0507581 and N.J.S. the
781: support of ISF grant 1325/06. We thank A.~ud-Doula and R.~Townsend for helpful discussions and comments.
782:
783:
784:
785: \begin{thebibliography}{}
786:
787: \bibitem[\protect\citeauthoryear{Arons}{1992}]{a92} Arons, J., 1992, \apj, 388, 561
788:
789: \bibitem[\protect\citeauthoryear{Begelman}{2006}]{b06} Begelman, M.\,C., 2002, \apj, 643, 1065
790:
791: \bibitem[\protect\citeauthoryear{Clark}{1996}]{c96} Clark, D.\,A., 1996, \apj, 457,291
792:
793: \bibitem[\protect\citeauthoryear{Davidson \& Humphreys}{1997}]{dh97} Davidson, K., \& Humphreys, R.\,M., 1997, ARA\&A, 35, 1
794:
795: \bibitem[\protect\citeauthoryear{Joss, Salpeter \& Ostriker}{1973}]{jso73} Joss, P., Salpeter, E., and Ostriker, J., 1973,
796: \apj, 181, 429
797:
798: \bibitem[\protect\citeauthoryear{Kulkarni et al.}{2007}]{ketal07} Kulkarni, S. et al., 2007, \nat, 447, 458
799:
800: \bibitem[\protect\citeauthoryear{Owocki, Castor \& Rybicki}{1988}]{ocr88} Owocki, S.\,P., Castor, J.\,I. \& Rybicki, G.\,B., 1988, \apj, 335, 914
801:
802: \bibitem[\protect\citeauthoryear{Owocki, Cranmer \& Blondin}{1994}]{ocb94} Owocki, S.\,P., Cranmer, S.\,R. \& Blondin, J.\,M., 1994, \apj, 424, 887
803:
804: \bibitem[\protect\citeauthoryear{Owocki \& Gayley}{1997}]{og97} Owocki, S.\,P., \&
805: Gayley, K.\,G., 1997, Luminous Blue Variables: Massive Stars in Transition,
806: 120, 121
807:
808: \bibitem[\protect\citeauthoryear{OGS04}{}]{ogs04} Owocki, S.\,P., Gayley, K.\,G. \& Shaviv, N.\,J., 2004, \apj, 616, 525
809:
810: \bibitem[\protect\citeauthoryear{Owocki \& van Marle}{2008}]{om08} Owocki, S.\,P. \& van Marle, A.\,J., 2008, Conference
811: proceedings, Massive Stars as Cosmic Engines, IAU Symp 250, ed. F. Bresolin, P.\,A. Crowther, \& J. Puls, arXiv:0801.2519
812:
813: \bibitem[\protect\citeauthoryear{Shaviv}{1998}]{s98} Shaviv, N.\,J., 1998, \apj, 494, L193
814:
815: \bibitem[\protect\citeauthoryear{Shaviv}{2000}]{s00} Shaviv, N.\,J., 2000, \apj, 532, L137
816:
817: \bibitem[\protect\citeauthoryear{Shaviv}{2001a}]{s01a} Shaviv, N.\,J., 2001a, \apj, 549, 1093
818:
819: \bibitem[\protect\citeauthoryear{Shaviv}{2001b}]{s01} Shaviv, N.\,J., 2001b, MNRAS, 326, 126
820:
821: \bibitem[\protect\citeauthoryear{Smith}{2002}]{s02} Smith, N., 2002, MNRAS, 337, 1252
822:
823: \bibitem[\protect\citeauthoryear{Smith \& Owocki}{2006}]{so06} Smith, N. \& Owocki, S.\,P., 2006, \apj, 645, L45
824:
825: \bibitem[\protect\citeauthoryear{Stone \& Norman}{1992}]{sn92} Stone, J.\,M. \& Norman, M.L., 1992, \apjs, 80, 753
826:
827: \bibitem[\protect\citeauthoryear{ud-Doula \& Owocki}{2002}]{do02} ud-Doula, A. \& Owocki, S.\,P., 2002, \apj, 576, 413
828:
829: \bibitem[\protect\citeauthoryear{van Marle et al.}{2008a}]{mos08a} van Marle, A.\,J., Owocki,
830: S.\,P. \& Shaviv, N.\,J., 2008a, proceedings of: First Stars III, Eds. B.
831: O'Shea, T. Abel, A. Heger, AIPC, 990, 250
832:
833: \bibitem[\protect\citeauthoryear{van Marle, Owocki \& Shaviv}{2008b}]{mos08b} van Marle, A.\,J., Owocki, S.\,P. \& Shaviv,
834: N.\,J., 2008, in preparation
835:
836: \bibitem[\protect\citeauthoryear{Vink \& de Koter}{2005}]{vk05} Vink, J.\,S. \& de Koter, A., 2005, A\&A, 442, 587
837:
838: \end{thebibliography}
839:
840: \bsp
841:
842: \label{lastpage}
843:
844: \end{document}
845:
846: