1: %%
2: %% AASTeX v5.x LaTeX 2e macros
3: %% -------------------------------------------------------------
4: %% OUTPUT OPTIONS
5: \documentclass[12pt,preprint]{aastex}
6: %\documentclass[preprint1]{aastex}
7: %% one-column, double-spaced document:
8: % \documentclass[manuscript]{aastex}
9: %% double-column, single-spaced document:
10: %\documentclass[preprint2]{aastex}
11: %% -------------------------------------------------------------
12: \def\la{Ly$\alpha$}
13: \def\hb{H$\beta$}
14: \def\arcmin {$^{\prime}$}
15: \def\arcm {$^{\prime}$}
16: \def\arcsec {$^{\prime\prime}$}
17: \def\arcs {$^{\prime\prime}$}
18: \def\kms {km\thinspace s$^{-1}$}
19: \def\cms {cm$^{-2}$}
20: \def\etal{{\it et al. }}
21: \def\aa{{\rm A$\,$\&$\,$A}} % A \and\ A
22: \def\araa{{\rm ARA$\,$\&$\,$A}} % Annu. Rev. A \and\ A
23: \def\aar{{\rm A$\,$\&$\,$AR}} % A \and\ A Review
24: \def\aas{{\rm A$\,$\&$\,$AS}} % A \and\ A Sup.
25: \def\apj{{\rm ApJ}} % Astrophys. J.
26: \def\apjs{{\rm ApJS}} % Astrophys. J. Supl.
27: \def\apjl{{\rm ApJ Let}} % Astrophys. J. (Letters)
28: \def\aj{{\rm AJ}} % Astron. J.
29: \def\sva{{\rm SvA}} % Sov. Astron.
30: \def\pasp{{\rm PASP}} % Publ. Astr. Soc. Pacific
31: \def\pasj{{\rm PASJ}} % Publ. Astr. Soc. Japan
32: \def\mnras{{\rm MNRAS}} % Monthly Notices.
33: \def\kmsmpc {km\thinspace s$^{-1}$\thinspace Mpc$^{-1}$}
34: \def\Msol{\thinspace\hbox{$\hbox{M}_{\odot}$}}
35: \def\Zsol{\thinspace\hbox{$\hbox{Z}_{\odot}$}}
36: \def\sol{\thinspace\hbox{$_{\odot}$ }}
37: \def\deg{\hbox{$^\circ$}}
38: \def\kpc{\thinspace\hbox{kpc}}
39: \def\ojo{\fbox{\bf !`$\odot$j$\odot$!}}
40: \def\ie{{\it i.e.} } % i.e. italicized
41: \def\a4{\hsize 17.0cm \vsize 25.cm}
42: \newcommand{\der}[2] { \frac{{\rm d}#1}{{\rm d}#2} }
43: \newcommand{\derp}[2] { \frac{\partial #1}{\partial #2} }
44: \newcommand{\dif} {{\rm d}}
45: \newcommand{\difp} {\partial}
46: \newcommand{\Hii}{\ion{H}{2}}
47: \newcommand{\Sub}[1]{_\mathrm{#1}}
48: \newcommand{\SSub}[1]{_\mathrm{\scriptscriptstyle #1}}
49: \newcommand{\Sup}[1]{^\mathrm{#1}}
50: \newcommand{\thC}{$\theta^1$~C~Ori}
51: \newcommand{\vdag}{(v)^\dagger}
52: \newcommand{\gmtemail}{gustavo@iagusp.usp.br}
53:
54: \shorttitle{2D Model of a Bimodal Wind}
55: \shortauthors{ W\"unsch et al.}
56:
57: \begin{document}
58:
59: \title{Two-Dimensional Hydrodynamic Models of Super Star Clusters
60: with a Positive Star Formation Feedback}
61:
62: \author{R. W\"unsch \altaffilmark{1}}
63: \affil{Cardiff University, Queens Buildings, The Parade, Cardiff. CF24 3AA,
64: United Kingdom; richard@wunsch.cz}
65: \altaffiltext{1}{Astronomical Institute, Academy of Sciences of the Czech
66: Republic, Bo\v{c}n\'\i\ II 1401, 141 31 Prague, Czech Republic}
67: \author{G. Tenorio-Tagle}
68: \affil{Instituto Nacional de Astrof\'\i sica Optica y
69: Electr\'onica, AP 51, 72000 Puebla, M\'exico}
70: \author{J. Palou\v{s}}
71: \affil{Astronomical Institute, Academy of Sciences of the Czech
72: Republic, Bo\v{c}n\'\i\ II 1401, 141 31 Prague, Czech Republic}
73: \and
74: \author{S. Silich}
75: \affil{Instituto Nacional de Astrof\'\i sica Optica y
76: Electr\'onica, AP 51, 72000 Puebla, M\'exico}
77:
78: \begin{abstract}
79: Using the hydrodynamic code ZEUS, we perform 2D simulations to determine the
80: fate of the gas ejected by massive stars within super star clusters. It turns
81: out that the outcome depends mainly on the mass and radius of the cluster. In
82: the case of less massive clusters, a hot high velocity ($\sim 1000$ km s$^{-1}$)
83: stationary wind develops and the metals injected by supernovae are dispersed to
84: large distances from the cluster. On the other hand, the density of the
85: thermalized ejecta within massive and compact clusters is sufficiently large as
86: to immediately provoke the onset of thermal instabilities. These deplete,
87: particularly in the central densest regions, the pressure and the pressure
88: gradient required to establish a stationary wind, and instead the thermally
89: unstable parcels of gas are rapidly compressed, by a plethora of re-pressurizing
90: shocks, into compact high density condensations. Most of these are unable to
91: leave the cluster volume and thus accumulate to eventually feed further
92: generations of star formation.
93:
94: The simulations cover an important fraction of the parameter-space, which
95: allows us to estimate the fraction of the reinserted gas which accumulates
96: within the cluster and the fraction that leaves the cluster as a function of
97: the cluster mechanical luminosity, the cluster size and heating efficiency.
98: \end{abstract}
99:
100: \keywords{Galaxies: star clusters --- ISM: bubbles --- ISM: HII regions ---
101: ISM}
102:
103:
104: \section{Introduction}
105:
106: Due to their stellar mass, which in some cases amounts to several million
107: M$_\odot$, and their compactness, as they only span a few parsecs, young ($<
108: 10$~Myr) super star clusters (SSCs) represent the most spectacular and dominant
109: mode of star formation in starburst and interacting galaxies
110: \citep{1995ApJ...446L...1O,1997RMxAC...6....5H,1995AJ....109..960W,2005ApJ...619..270M}.
111: SSCs have been detected in the optical, UV and X-rays, and also in the radio
112: continuum and IR regimes, as some of them are deeply embedded behind dense
113: obscuring material, leading to powerful ultra-dense HII regions
114: \citep{1999ApJ...527..154K,2007ApJ...668..168G}.
115:
116:
117: On theoretical grounds, it has been inferred that such extreme modes of
118: star formation should lead to several tens of thousands of massive stars, all
119: of them known to rapidly ($\leq 50$~Myr) reinsert, through stellar winds and
120: supernova (SN) explosions, a large fraction of their mass back into the ISM.
121: The first (adiabatic) approach to the hydrodynamics of the matter reinserted
122: within a SSC is due to \citet[hereafter CC85]{1985Natur.317...44C}. In their
123: model, the stellar sources of mass and energy, are assumed to be equally
124: spaced within the SSC volume of radius $R_\mathrm{SC}$. They also assumed that
125: all of the kinetic energy provided by massive stars is immediately, and in situ,
126: thermalized via random collisions of the ejecta from neighboring sources.
127: This results in energy and mass deposition rate densities $q_e =
128: (3L_\mathrm{SC})/(4\pi R_\mathrm{SC}^3)$ and $q_m = (3\dot{M}_\mathrm{SC})/(4\pi
129: R_\mathrm{SC}^3)$, respectively, where $L_\mathrm{SC}$ and $\dot{M}_\mathrm{SC}$
130: are the cluster mechanical luminosity and mass deposition rates. These
131: assumptions lead to a high temperatures gas ($T > 10^7$~K) in which the
132: interstellar cooling law is close to its minimum value, and this justifies
133: their adiabatic assumption. In the adiabatic model of CC85, the thermalized hot
134: gas rapidly settles into almost constant density and temperature
135: distributions, although a slight outward pressure gradient establishes a
136: particular velocity distribution with the stagnation point (\ie zero
137: velocity) at the cluster center. The gas velocity increases then almost
138: linearly with radius to reach the sound speed ($c_\mathrm{SC}$) right at the
139: cluster surface and then streams into the surrounding lower pressure ambient
140: medium to reach its terminal velocity ($v_\mathrm{A\infty} \sim 2 c_\mathrm{SC}$),
141: while the density and temperature of the outflow, the cluster wind, decrease as
142: $r^{-2}$ and $r^{-4/3}$, respectively. The solution of such a stationary outflow
143: depends on three variables: the cluster size ($R_\mathrm{SC}$), the mass
144: deposition rate ($\dot M_\mathrm{SC}$) and the mechanical luminosity of the
145: cluster ($L_\mathrm{SC}$). Knowledge of these three variables allows one to
146: solve the hydrodynamic equations analytically and workout the run of density,
147: temperature and velocity of the stationary outflow. Note that $\dot
148: M_\mathrm{SC}$ is usually replaced by the adiabatic terminal speed ($v_\mathrm{A\infty}
149: = (2 L_\mathrm{SC}/\dot M_\mathrm{SC})^{1/2}$).
150:
151: The model then yields a stationary flow in which the matter reinserted by the
152: evolving massive stars ($\dot M_\mathrm{SC}$) equals the amount of matter
153: feeding the cluster wind ($4 \pi R_\mathrm{SC}^2 \rho_\mathrm{SC}
154: c_\mathrm{SC}$); where $\rho_\mathrm{SC}$ is the reinserted gas density value at
155: the star cluster surface. As $L_\mathrm{SC}$ and $\dot M_\mathrm{SC}$ increase
156: linearly with the cluster mass ($L_\mathrm{SC} \sim M_\mathrm{SC}$,
157: $\dot{M}_\mathrm{SC} \sim M_\mathrm{SC}$), the adiabatic model predicts that the
158: more massive a cluster is, the more powerful is its resultant wind. This latter
159: conclusion breaks down if one relaxes the adiabatic assumption
160: \citep[see][]{2004ApJ...610..226S}. More massive clusters deposit larger
161: amounts of matter and thereby deliver a sufficiently high density,
162: $\rho_\mathrm{SC} = \dot M_\mathrm{SC} / (4 \pi R_\mathrm{SC}^2)$, to provoke
163: strong radiative cooling. Thus, as radiative cooling ($Q$ = $n^2 \Lambda(T,
164: Z)$; where $\Lambda(T, Z)$ is the cooling function) is proportional to
165: $M_\mathrm{SC}^2$ and $L_\mathrm{SC}$ is proportional to $M_\mathrm{SC}$,
166: there is a threshold mechanical luminosity (for given $R_\mathrm{SC}$ and
167: $\dot{M}_\mathrm{SC}$), above which strong radiative cooling takes over despite
168: the large temperatures and the minimum value of the interstellar cooling law at
169: these temperatures.
170:
171: Note that details of the thermalization process have been largely ignored
172: although more recent formulations of the problem \citep{2007A&A...471..579W,
173: 2007ApJ...669..952S} have inferred that a significant fraction of the deposited
174: mechanical energy could be radiated away as soon as it is inserted. In such
175: cases, only a fraction of the deposited energy remains available to heat the
176: thermalized matter. This fraction is called the heating efficiency,
177: $\eta$. It is assumed by different authors to have values between $0.01$
178: and $1$ \citep{1998A&A...337..338B,2004A&A...424..817M}, and shown to acquire
179: small values in the case of massive compact SSCs
180: \citep[see][]{2007ApJ...669..952S}.
181:
182: Figure~\ref{Lcrit} presents the threshold mechanical luminosity found by
183: \citet{2004ApJ...610..226S}, here re-calculated for three different values of
184: the heating efficiency $\eta$. Clusters with a mechanical luminosity (or mass)
185: far below this line evolve in the quasi-adiabatic regime. For these, the
186: Chevalier \& Clegg model provides a good approximation to the structure and
187: hydrodynamics of the star cluster wind
188: \citep{2000ApJ...536..896C,2001ApJ...559L..33R}. Figure~\ref{m1d} displays
189: radial profiles of temperature, particle density, pressure and velocity, typical
190: of such steady state winds.
191:
192: For clusters with a mechanical luminosity close to the threshold value, the
193: temperature distribution within the stationary wind becomes very different from
194: that predicted by the adiabatic model \citep{2004ApJ...610..226S}. This is
195: because, as soon as the temperature of the wind decreases to approximately $\sim
196: 10^6$~K, radiative cooling begins to increase sharply (mainly due to f-b and
197: b-b transitions of heavy elements) and the temperature in the free wind region
198: drops rapidly several orders of magnitude. Nevertheless such clusters manage to
199: eject all the matter deposited by stellar winds and supernovae, and thereby
200: sustain a stationary wind.
201:
202: For clusters above the threshold line, radiative cooling becomes a dominant
203: factor. Radiative cooling affects first the central densest regions causing a
204: sudden drop in pressure. This promotes the shift of the stagnation point out of
205: the cluster center. Such clusters adhere to a bimodal solution
206: \citep{2007ApJ...658.1196T,2007A&A...471..579W} in which the stagnation radius
207: ($R_\mathrm{st}$) defines two different regions within the cluster volume (see
208: Figure~\ref{figbm}). On one hand, there is an outer
209: shell in which the deposited energy, although affected by strong radiative
210: cooling, is still able to drive an outward stationary wind, by making the gas
211: velocity acquire the sound speed exactly at the cluster surface. On the other
212: hand, the matter deposited inside the stagnation radius is strongly affected by
213: radiative cooling and becomes thermally unstable. The instability leads to a rapid
214: and continuous loss of energy from large and small parcels of gas, thereby
215: reducing their temperature and pressure. These events lead immediately to the
216: formation of strong shocks that emanate from the hot, high pressure regions and
217: are driven into the cold parcels of gas in order to restore their pressure.
218: These have been termed re-pressurizing shocks, and have been invoked in several
219: astrophysical circumstances, such as the formation of globular clusters
220: \citep{1995ApJ...442..618V}, and also in the cooling of supernova
221: matter within superbubbles, leading to highly metallic droplets falling onto the
222: galaxy \citep{1996AJ....111.1641T}. In the context of the matter cooling
223: inside the superstar cluster stagnation radius, the re-pressurizing shocks have been
224: first modelled by means of 1D numerical hydrodynamics in
225: \citet{2007ApJ...658.1196T}. The re-pressurizing shocks rapidly compress the
226: cold gas, enhancing its density while reducing its volume, until the cold
227: condensations acquire again the thermal pressure value of the hot gas. Given the
228: initially similar values of density ahead of and behind the shocks, one can show that
229: their velocity is only a function of the temperature $T$ of the hot gas
230: ($V_\mathrm{RP} \approx [(kT)/(\mu m_p)]^{0.5}$, where $k$ is
231: the Boltzmann constant, $\mu m_p$ is the mean gas-particle mass, and $m_p$
232: is the proton mass), and thus cold parcels of gas within the SSC are rapidly driven
233: into small high density condensations.
234:
235: The continuous occurrence of thermal instabilities results in the accumulation
236: of mass in this region and in a very chaotic, highly non-stationary
237: hydrodynamical pattern, with a number of radiative shocks and cooling fronts
238: propagating within the cluster volume. The continuous accumulation of
239: thermally unstable matter must finally lead to its re-processing into further
240: generations of stars \citep{2005ApJ...628L..13T}. Unfortunately, 1D
241: simulations do not allow one to reach an adequate understanding of the physics
242: within massive clusters in the bimodal regime, nor to make realistic
243: predictions regarding the fate of the matter reinserted by massive stars. In
244: order to develop a more realistic model, here we present detailed 2D
245: hydrodynamic simulations of the gaseous flows that result from the
246: thermalization of the kinetic energy supplied by stellar winds and supernovae
247: ejecta inside the volume occupied by the stellar cluster. We focus on clusters
248: evolving in the bimodal regime. The major result from these simulations is that
249: the central zones of clusters evolving in the bimodal regime accumulate large
250: amounts of matter in the form of warm ($T \sim 10^4$ K) high density
251: condensations embedded into a hot plasma of much lower density. The amount of
252: accumulated matter depends on the excess star cluster mechanical luminosity
253: over the threshold value and grows with time, unless the accumulation of
254: reinserted matter inside the stagnation radius is compensated by secondary
255: star formation. Our results also show that the stagnation surface itself has a
256: very complicated dynamic morphology that continuously changes with time.
257: Nevertheless, the average radius of the stagnation surface remains close
258: to that predicted by 1D and semi-analytic calculations.
259:
260: The paper is organized as follows. Section 2 describes the numerical model
261: and discusses the input physics. Section 3 present the results from our
262: numerical simulations and compares them with the semi-analytical results and
263: previous 1D simulations. In section 4 we discuss our results and
264: section 5 gives a summary of our findings.
265:
266: %------------------------------------------------------------------------
267: \clearpage
268: \begin{figure}
269: \plotone{f1.eps}
270: \caption{The threshold lines.
271: Clusters above the threshold line evolve in a bimodal regime in
272: which the injected matter is accumulated inside their densest inner
273: regions while the outer zones develop a strongly radiative stationary
274: wind. The threshold lines were calculated for the adiabatic terminal velocity
275: $V_\mathrm{A\infty} = 1000$~km s$^{-1}$ and three different heating efficiencies
276: $\eta = 1.0, 0.3$ and $0.1$ denoted with solid, dashed and dotted lines,
277: respectively. The 2D simulations (see Table~\ref{modtab}) are represented by
278: symbols. Different symbols denote different heating efficiencies (and thus
279: are to be compared with the corresponding threshold line) $\eta = 1.0$
280: (triangles), $\eta = 0.3$ (circles), and $\eta = 0.1$ (plus signs).
281: The secondary y-axis shows the approximate mass of the cluster obtained using a
282: relation $M_\mathrm{SC} = (L_\mathrm{SC} / 3\times 10^{40}) 10^6$~M$_\odot$
283: \citep{1999ApJS..123....3L}.}
284: \label{Lcrit}
285: \end{figure}
286:
287: \begin{figure}
288: \plotone{f2.eps}
289: \caption{Model 1 ($L_\mathrm{SC}/L_\mathrm{crit} = 0.5$): radial profiles of
290: the wind particle density, $n_w$ (note the units are $m^{-3}$ to
291: fit within the figure), temperature ($T_w$), pressure ($P_w$) and radial
292: velocity ($v_w$). The thin lines represent the semi-analytical solution
293: \citep{2004ApJ...610..226S}, the
294: thick lines are results from the 2D simulation (model 1) at $t = 0.25$~Myr.
295: The simulation, after a short initial relaxation period,
296: stays perfectly stationary and spherically
297: symmetric at all times.}
298: \label{m1d}
299: \end{figure}
300:
301: \begin{figure}
302: \plotone{f3.eps}
303: \caption{The internal structure of SSCs in the bimodal regime.
304: The cluster volume presents two distinct regions. Below the
305: stagnation radius $R_\mathrm{st}$: is the thermally unstable region and
306: above it lies the outer quasi-stationary wind region.
307: The radial profiles of the wind particle density, temperature, pressure and
308: radial velocity were obtained as the axial time averages values of the 2D
309: simulation (model~3) between $0.4$ and $0.8$~Myr. Only the hot gas
310: ($T > 2\times 10^4$) was taken into consideration for this plot. The values
311: of temperature and pressure at the stagnation radius, obtained from the
312: semi-analytical model (thin solid lines across the inner region), are given with
313: horizontal lines.}
314: \label{figbm}
315: \end{figure}
316: %-------------------------------------------------------------------
317: \clearpage
318:
319: \section{The numerical approach}
320:
321: The numerical models presented here are based on the finite difference Eulerian
322: hydrodynamic code ZEUD3D v3.4.2 \citep{1992ApJS...80..753S}. All
323: simulations have been carried out on a spherical ($r$, $\theta$) grid,
324: symmetric along the $\phi$-coordinate. We have set the radial size of grid cells
325: $\Delta r$ proportional to the radial coordinate $r$ which ensures that all
326: grid-cells have $\Delta r \sim r\Delta\theta$. Another advantage
327: of this radially scaled grid is that the resolution is higher at smaller
328: radii (inside the cluster) where thermal instabilities are expected.
329:
330: The cooling routine accounts for extremely fast cooling both in the wind or
331: within the SSC volume. The change of internal energy $e$ due to cooling is
332: \begin{equation}
333: \label{coolrate}
334: \left.\frac{de}{dt}\right|_\mathrm{cool} = - n_i n_e \Lambda(T,Z) \ ,
335: \end{equation}
336: where $n_i$ and $n_e$ are ion and electron densities, respectively. We compute
337: them from the gas density $\rho$ as $n_i = n_e = \rho / (\mu_\mathrm{ion}
338: m_p)$, where $\mu_\mathrm{ion} m_p$ is the average ion mass.
339: We assume $\mu = 0.609$ and $\mu_\mathrm{ion} = 1.27$ for all computed models.
340: $\Lambda(T, Z)$ is the cooling function. We use Raymond \& Cox cooling function
341: which has been supplemented with new elements and tabulated by
342: \citet{1995MNRAS.275..143P}.
343:
344: The RHS of equation~(\ref{coolrate}) is evaluated in the middle of time-steps to
345: maintain the second order accuracy of the code and the energy conservation
346: equation is then solved iteratively using the Brendt algorithm which is more
347: stable and accurate than the Newton-Raphson method originally used in ZEUS.
348:
349: The cooling rate has to be included in the computation of the time-step,
350: otherwise rapid cooling not resolved in time may lead to the occurrence of
351: negative temperatures. A common way of solving this is to limit the amount of
352: internal energy which can be radiated away during one time-step by setting
353: \begin{equation}
354: dt_\mathrm{cool} =
355: \epsilon \left| \frac{e}{\left.\frac{de}{dt}\right|_\mathrm{cool}} \right|
356: \end{equation}
357: where $\epsilon$ is a safety factor smaller than unity
358: (see e.g.~\citealt{1997A&A...318..595S}; where $\epsilon = 0.3$ was used).
359: In this work we use $\epsilon = 0.25$.
360:
361:
362: The global time-step, $dt$, is computed as follows:
363: \begin{equation}
364: dt = \left\{
365: \begin{array}{lcl}
366: dt = dt_\mathrm{HD}, & \mathrm{for} & dt_\mathrm{cool} \ge dt_\mathrm{HD}; \\
367: dt = dt_\mathrm{cool},
368: & \mathrm{for} & dt_\mathrm{HD} > dt_\mathrm{cool} \ge \delta\times
369: dt_\mathrm{HD};\\
370: dt = \delta\times dt_\mathrm{HD}, & \mathrm{for} & \delta\times dt_\mathrm{HD} >
371: dt_\mathrm{cool}\ (\mathrm{local\ substeps}\ dt_\mathrm{sub} = dt_\mathrm{cool});
372: \end{array}
373: \right.
374: \end{equation}
375: where $dt_\mathrm{HD}$ is the "hydrodynamic" time-step resulting from the
376: Courant-Friedrich-Levi criterion and $\delta$ is the minimum fraction of
377: $dt_\mathrm{HD}$ to which the global time-step $dt$ is allowed to drop. If, due
378: to the cooling condition, a certain cell requires an even smaller time-step (\ie
379: $dt_\mathrm{cool} < \delta dt_\mathrm{HD}$), the energy equation is integrated
380: numerically in that cell, using $dt_\mathrm{sub} = dt_\mathrm{cool}$ but
381: assuming that during this time the density and temperature in the cell are not
382: substantially affected by interactions with neighboring cells. This ensures that
383: CPU time is not wasted in cells where a high time resolution is not required. To
384: determine a reasonable value of $\delta$ we ran several tests on a low
385: resolution grid ($150\times 56$) and found that there are no appreciable
386: differences for $\delta \lesssim 0.3$. Therefore, we use $\delta = 0.1$.
387:
388: In order to simulate the effect of the stellar UV radiation field, in some
389: of the simulations (see Table~\ref{modtab}), we do not allow the gas
390: temperature to drop below $T_\mathrm{lim} = 10^4$~K. This is equivalent to
391: assuming that there are sufficient UV photons to ionize the dense thermally
392: unstable matter which otherwise would cool to much lower temperatures.
393:
394: The wind source was modelled by a continuous replenishment of mass and internal
395: energy in all cells within the cluster volume, at rates $q_m =
396: (3\dot{M}_\mathrm{SC})/(4\pi R_\mathrm{SC}^3)$ and $q_e =
397: (3\dot{L}_\mathrm{SC})/(4\pi R_\mathrm{SC}^3)$, respectively. The procedure
398: applied to each cell within the cluster volume at every time-step is:
399: \begin{enumerate}
400: \item the density and the total energy in a given cell are saved to $\rho_\mathrm{old}$ and
401: $e_\mathrm{tot,old}$
402: \item the new mass is inserted $\rho_\mathrm{new} =
403: \rho_\mathrm{old} + (1 +A_\mathrm{noise} \zeta ) q_\mathrm{m} dt$, the velocity is corrected so that the
404: momentum is conserved $\mathbf{v}_\mathrm{new} =
405: \mathbf{v}_\mathrm{old}\rho_\mathrm{old}/\rho_\mathrm{new}$
406: \item the internal energy is corrected to conserve the total energy
407: $e_\mathrm{i,mid} = e_\mathrm{tot,old} - \rho_\mathrm{new}\mathbf{v}^2_\mathrm{new}/2$
408: \item the new energy is inserted in
409: a form of internal energy $e_\mathrm{i,new} = e_\mathrm{i,mid} + (1 + A_\mathrm{noise} \zeta ) q_\mathrm{e}dt$
410: \end{enumerate}
411: where $\zeta$ is a pseudo-random number from the interval $(-1, 1)$ generated
412: each time it is used, and $A_\mathrm{noise}$ is the relative amplitude of the
413: noise. The inclusion of a noise term is necessary to break the artificial
414: spherical symmetry imposed by the initial conditions (see below). The model is
415: very robust with respect to $A_\mathrm{noise}$. Test runs with
416: $A_\mathrm{noise} = 0.01, 0.05, 0.1$ and $0.5$ lead to very similar general
417: properties (mass fluxes at boundaries, number of fragments formed and their
418: approximate sizes) in all models. The only noticeable difference is the
419: duration of the initial relaxation period required to break the initial
420: symmetry. We use $A_\mathrm{noise} = 0.1$ for all models described in this
421: paper.
422:
423: The boundary conditions are set open at both r-boundaries and periodic at both
424: $\theta$-boundaries. The open inner r-boundary allows the dense clumps that
425: are not ejected from the cluster (through its boundary at $R_\mathrm{SC}$) to
426: leave the computational domain. Otherwise, they would accumulate within the
427: cluster and eventually fill its whole volume; this happens for some models with
428: high $L_\mathrm{SC}$ and low $\eta$, even with the open inner r-boundary (see
429: Section~\ref{S_other}). It is unphysical, because the accumulated mass exceeds
430: the amount which can be ionized by the available UV photons; therefore it should
431: cool down, become gravitationally unstable and collapse into stars. Moreover,
432: the accumulated mass ultimately becomes so high that it would be gravitationally
433: unstable even at $T = 10^4$~K. However, determining the fate of the dense mass
434: properly would need a model of star formation which takes into account radiation
435: transfer and self-gravity: this is not included in our current numerical model.
436:
437: Two different initial conditions are used for the two series of models presented
438: here (see Table~\ref{modtab}). Models 1 -- 6 and 12 with $R_\mathrm{SC}=10$~pc
439: have as initial condition the spherically symmetric semi-analytical solution
440: with $L_\mathrm{SC} = L_\mathrm{crit}$ \citep{2004ApJ...610..226S}. Models 7 -
441: 11 and 13, with $R_\mathrm{SC} = 3$~pc, use as initial condition the
442: semi-analytical solution with the appropriate $L_\mathrm{SC}$. As this is only
443: fully defined for $r > R_\mathrm{st}$, at $t = 0$ the central region $r <
444: R_\mathrm{st}$ is filled with a zero velocity, constant density and temperature
445: gas, with values equal to those at the stagnation radius
446: $\rho=\rho(R_\mathrm{st})$ and $T = T(R_\mathrm{st})$, respectively. The
447: advantage of this approach is that such conditions are closer to the
448: quasi-stationary state and therefore it takes a shorter time to reach it.
449:
450:
451: \section{Results}
452:
453: We have calculated two series of models (see Table~\ref{modtab}). Models in the
454: first series all have a cluster radius $R_\mathrm{SC} = 10$~pc and an assumed
455: heating efficiency $\eta = 1$. The cluster mechanical luminosities considered
456: are: $2.5\times 10^{41}$, $10^{42}$, $10^{43}$ and $10^{44}$~erg s$^{-1}$, which
457: result in values of $L_\mathrm{SC}/L_\mathrm{crit} = 0.5$, $2$, $20$ and
458: $200$, respectively. The model with $L_\mathrm{SC}/L_\mathrm{crit} = 2$ was
459: computed with three different numerical resolutions $150\times 56$, $300\times 112$
460: and $600\times 224$ to check and establish convergence. The computational domain
461: extents radially from $R_\mathrm{IB} = 2$~pc (the inner boundary) to
462: $R_\mathrm{OB} = 30$~pc (the outer boundary) and from $\theta_\mathrm{LB} =
463: \pi/2 - 0.5$ (left boundary) to $\theta_\mathrm{RB} = \pi/2 +
464: 0.5$ (right boundary) in the axial direction.
465:
466: For the second series of models, we assume a more compact cluster and more
467: realistic cluster parameters by setting $R_\mathrm{SC} = 3$~pc and $\eta = 0.1$
468: or $0.3$. We ran 6 models with 3 different ratios $L_\mathrm{SC}/L_\mathrm{crit}
469: = 2.5$, $25$ and $250$. In these cases the radial extent of the grid goes
470: from $R_\mathrm{IB} = 0.5$~pc to $R_\mathrm{OB} = 10$~pc, and the axial extent
471: goes from $\pi/3$ to $2\pi/3$.
472:
473: In all cases we assumed an adiabatic terminal velocity is $v_\mathrm{A,\infty}
474: \equiv \sqrt{\frac{2L_\mathrm{SC}}{\dot{M}_\mathrm{SC}}}= 1000$~km/s, and the
475: lower temperature limit was set equal to $T_\mathrm{lim} = 10^4$~K, or 10$^2$
476: K (models~12 and 13). In cases with an $\eta < 1$, $v_\mathrm{A,\infty}$, was replaced by
477: $v_\mathrm{A,\infty} \equiv \sqrt{\frac{2 \eta
478: L_\mathrm{SC}}{\dot{M}_\mathrm{SC}}}$. Note that in all cases, radiative
479: cooling lowers the outflow velocity at the grid outer boundary to somewhat
480: smaller values.
481:
482: \clearpage
483: \begin{table}
484: \begin{tabular}{|l|l|l|l|l|l|l|l|l|l|l|}
485: \hline
486: No. &
487: $\frac{L_\mathrm{SC}}{\mathrm{[erg/s]}}$ &
488: $\frac{R_\mathrm{SC}}{\mathrm{[pc]}}$ &
489: $\eta$ & $\frac{L_\mathrm{SC}}{L_\mathrm{crit}}$ &
490: grid &
491: $T_\mathrm{lim}$ &
492: $\frac{R_\mathrm{st}}{\mathrm{[pc]}}$ &
493: $\frac{T_\mathrm{st}}{\mathrm{[}10^6\mathrm{K]}}$ &
494: $\frac{\dot{M}_\mathrm{out}}{\dot{M}_\mathrm{SC}}$ &
495: $T_\mathrm{min}$ \\
496: %& Gravity \\
497: \hline
498: 1 & $2.5\times 10^{41}$ & 10 & 1 & 0.5 & $300\times 112$ & $10^4$ & 0 & $10.4$ & 1.00 & $10^4$ \\ %& No \\
499: 2 & $10^{42}$ & 10 & 1 & 2 & $150\times 56 $ & $10^4$ & 6.056 & $10.4$ & 0.81 & $10^4$ \\ %& No \\
500: 3 & $10^{42}$ & 10 & 1 & 2 & $300\times 112$ & $10^4$ & 6.056 & $10.4$ & 0.84 & $10^4$ \\ %& No \\
501: 4 & $10^{42}$ & 10 & 1 & 2 & $600\times 224$ & $10^4$ & 6.056 & $10.4$ & 0.79 & $10^4$ \\ %& No \\
502: 5 & $10^{43}$ & 10 & 1 & 20 & $300\times 112$ & $10^4$ & 8.991 & $10.4$ & 0.39 & $10^4$ \\ %& No \\
503: 6 & $10^{44}$ & 10 & 1 & 200 & $300\times 112$ & $10^4$ & 9.696 & $10.4$ & 0.19 & $10^4$ \\ %& No \\
504: \hline
505: 7 & $10^{39}$ & 3 & 0.1 & 2.5 & $300\times 104$ & $10^4$ & 1.872 & $1.10$ & 0.77& $10^4$ \\ %& No \\
506: 8 & $10^{40}$ & 3 & 0.3 & 2.5 & $300\times 104$ & $10^4$ & 1.860 & $3.48$ & 0.84& $10^4$ \\ %& No \\
507: 9 & $10^{40}$ & 3 & 0.1 & 25 & $300\times 104$ & $10^4$ & 2.709 & $1.10$ & 0.98& $10^4$ \\ %& No \\
508: 10 & $10^{41}$ & 3 & 0.3 & 25 & $300\times 104$ & $10^4$ & 2.707 & $3.48$ & 0.83& $10^4$ \\ %& No \\
509: 11 & $10^{41}$ & 3 & 0.1 & 250 & $300\times 104$ & $10^4$ & 2.912 & $1.10$ & 1.00& $10^4$ \\ %& No \\
510: \hline
511: 12 & $10^{43}$ & 10 & 1 & 20 & $300\times 112$ & $10^2$ & 8.991 & $10.4$ & 0.38& $10^2$ \\ %& No \\
512: 13 & $10^{40}$ & 3 & 0.1 & 25 & $300\times 104$ & $10^2$ & 2.709 & $1.10$ & 0.44& $10^2$ \\ %& No \\
513: \hline
514: \end{tabular}
515: \caption{The set of computed models. The values of $R_\mathrm{st}$ and
516: $T_\mathrm{st}$ were determined by means of semi-analytic models, the values of
517: $\dot{M}_\mathrm{out}/\dot{M}_\mathrm{SC}$ are the mass
518: flux through the cluster border in the 2D simulations, averaged over the time
519: interval $0.1 - 0.8$~Myr for $R_\mathrm{SC} = 10$~pc models and $1-2$~Myr for
520: $R_\mathrm{SC} = 3$~pc models.}
521: \label{modtab}
522: \end{table}
523: \clearpage
524:
525: \subsection{Model 1, $L_\mathrm{SC}/L_\mathrm{crit} = 0.5$}
526:
527: This model was computed in order to test the numerical code against the
528: semi-analytical solution which is known for $L_\mathrm{SC} < L_\mathrm{crit}$.
529: Despite the perturbations of the deposited mass and energy, the flow is
530: perfectly stationary. The resultant radial density, temperature and velocity
531: profiles are shown in Figure~\ref{m1d} where they are compared to the
532: semi-analytic solution. The agreement is very good, despite the fact that the
533: 2-D model does not calculate the central sphere of radius $R_\mathrm{IB}$, and
534: this induces a small discrepancy in the velocity in the inner cluster regions.
535:
536: \subsection{Model 5, $L_\mathrm{SC}/L_\mathrm{crit} = 20$}
537:
538: We continue with a detailed description of model~5 which exhibits the typical
539: hydrodynamic behavior for clusters above the threshold line, $L_\mathrm{SC} >
540: L_\mathrm{crit}$. The model starts with the semi-analytical solution for
541: $L_\mathrm{SC} = L_\mathrm{crit}$ and from then onwards, at each time step, the
542: mass and energy input rates are consistent with the selected ratio
543: $L_\mathrm{SC}/L_\mathrm{crit} = 20$. Initially, the density grows slowly all
544: over the cluster volume, and this steadily enhances the radiative cooling,
545: causing lower temperatures. Eventually, as the temperature approaches $T\sim
546: 3\times10^5$~K, the cooling rate increases steeply, and thermal instability
547: occurs. This lowers the temperature rapidly to $T = 10^4$ K, particularly
548: in the densest central regions. On the other hand, the outer regions, where the
549: density is lower, remain hot and thus a large pressure gradient between the
550: two regions leads to the formation of a strong shock wave propagating inwards
551: (see Figure~\ref{rTd43}, left column).
552:
553: The simulations show a very dynamic evolution in which regions of hot gas
554: of different dimensions appear and grow close to the center, as more energy is
555: deposited within the cluster volume. The hot gas expands super-sonically into
556: the low pressure warm ($10^4$~K) surroundings. This decreases locally the
557: density of the hot gas, {preventing its becoming} thermally unstable, while at
558: the same time the parcels of warm gas are compressed from all sides into high
559: density condensations, until they reach pressure equilibrium with the
560: surrounding hot gas. The hot regions grow until they again occupy most of
561: the cluster volume, see Figure~\ref{rTd43}, middle column.
562:
563: While all this is happening, the inward propagating shock wave, overruns the
564: central region, accelerating inwards most of the high density condensations, and
565: these eventually leave the computational grid as they cross the inner boundary.
566: As the high temperatures are reestablished, the overall pressure gradient
567: vanishes and the shock wave decays. Once the hot gas again permeates the
568: cluster volume, the density starts to grow again and the whole process
569: repeats itself, but in a weaker form because of the presence of a few dense
570: condensations surviving from the previous cycle. This model exhibits 2-3
571: subsequent weaker periods of oscillations similar to the ones observed in 1D
572: simulations (\citealt{2007ApJ...658.1196T}; the low-energy model).
573:
574: Finally, the oscillations vanish and a quasi-equilibrium situation is
575: established, in which high density condensations form at an approximately
576: constant rate (see Figure~\ref{rTd43}, right column). The amount of mass which goes into
577: warm dense condensations is just enough to maintain the hot medium close to the
578: thermally unstable regime. It is a quasi-stationary situation: the density
579: tends to grow everywhere, surpassing some threshold value that favours the
580: thermal instability in the densest central regions. This sudden loss of pressure
581: prevents matter within the stagnation boundary from escaping the
582: cluster as a wind. The thermally unstable gas is instead packed into high
583: density condensations, most of which leave the computational grid through the
584: inner boundary. The location of the stagnation surface, which separates the
585: region where the thermal instability occurs from the outer stationary wind,
586: also experiences some oscillations. This leads to a non-spherical surface
587: with an average radius close to that given by the analytic approximation
588: \citep{2007A&A...471..579W},
589: %---------------------------------------------------------------------
590: \begin{equation}
591: \frac{R_\mathrm{st}^3}{R_\mathrm{SC}^3} = 1 -
592: \left(\frac{L_\mathrm{crit}}{L_\mathrm{SC}}\right)^{1/2} \ .
593: \end{equation}
594: %---------------------------------------------------------------------
595: The $R_\mathrm{st}$ radius at which vicinity the orientation of the
596: velocity vectors abruptly changes from an outward to an inward motion has been
597: indicated with a red line in the bottom panels of Figure~\ref{rTd43}.
598:
599:
600: \subsection{Models 2-4, $L_\mathrm{SC}/L_\mathrm{crit} = 2$}
601:
602: In these cases the stagnation radius $R_\mathrm{st}$ is smaller than in model~5,
603: resulting in a smaller thermally unstable region. The lower mass deposition rate
604: results in a smaller amount of high density gas being formed there (see
605: Figure~\ref{rTd42}). Otherwise, the evolution is similar to that of model~5,
606: including the initial relaxation period of intense thermal instability, the
607: appearance of re-pressurizing shocks which lead to high density condensations,
608: and the exit of most of these through the inner boundary, to finally reach
609: equilibrium between the formation of high density condensations and the mass
610: deposition rate.
611:
612: We have computed this model on three different grids to check how the resolution
613: affects the results (see the different rows of Figure~\ref{rTd42}). Although the
614: resultant fragments tend to be smaller and more structured on the higher
615: resolution grids, the global characteristics such as the mass flux through the
616: cluster border (see Table~\ref{modtab}) are in a reasonable agreement.
617:
618: \subsection{Model 6, $L_\mathrm{SC}/L_\mathrm{crit} = 200$}
619:
620: The hydrodynamical behavior of this case is very similar to that of model~5, the
621: only differences are quantitative: In this case many more high density
622: condensations form and occupy a larger fraction of the cluster volume (see
623: Figure~\ref{rTd44}). The quasi-stationary region, above the rapidly varying
624: stagnation surface, becomes at times very narrow and, as it is repeatedly
625: perturbed by high density condensations leaving the cluster, at times it is not
626: even a contiguous region. Nevertheless, most of the thermally unstable gas
627: driven into high density condensations stays within the cluster volume and
628: after some time, as in model~5, goes through the inner boundary and disappears
629: from the computational domain.
630:
631:
632: \subsection{Other Models}
633: \label{S_other}
634:
635: The most important parameter of our cluster model is the ratio
636: $L_\mathrm{SC}/L_\mathrm{crit}$, as this defines the location of the stagnation
637: surface and thus the relative sizes of the thermally unstable region and of the
638: quasi-stationary outer wind region. We have performed many more simulations (see
639: Table 1), all presenting the same general features as in model~5. For example
640: models~7 and 8, ($L_\mathrm{SC}/L_\mathrm{crit} = 2.5$, $\eta = 0.1$ and $0.3$)
641: have $L_\mathrm{SC}/L_\mathrm{crit}$ values close to those of models~2 - 4.
642: However, the smaller heating efficiency $\eta$ assumed in these cases, results
643: in a smaller temperature and pressure of the hot medium, and this leads to a
644: stationary wind that expands with a smaller velocity. The lower temperature of
645: the hot gas leads to slower velocity re-pressurizing shocks. The lower ambient
646: pressure of the hot gas also leads to larger final sizes of the high density
647: condensations and after some time to their accumulation near the central zones
648: of the computational grid. This sooner or later prevents the exit of matter
649: through the inner boundary in case of models~9 and 10
650: ($L_\mathrm{SC}/L_\mathrm{crit} = 25$, $\eta = 0.1$ and $0.3$) and model~11
651: ($L_\mathrm{SC}/L_\mathrm{crit} = 250$, $\eta = 0.1$). We believe this is an
652: artifact promoted by the fact that we do not allow for this matter to go into
653: star formation.
654:
655: Models~12 and 13 are similar to models~5 and 9, respectively, but the gas
656: is allowed to cool to $10^2$~K instead of $10^4$~K. A comparison between
657: model~12 and 5 shows very similar results. This means that the reduced volume
658: of clumps in model~12 (due to their lower temperature and hence pressure) does
659: not affect the ratio of the clumps that are ejected from the cluster to the
660: clumps that stay there and finally leave the computational domain through the
661: inner boundary. Therefore, the outflow from the cluster ($\dot{M}_\mathrm{out}$)
662: remains the same in both models. However, in the case of model~13
663: ($R_\mathrm{SC} = 3$~pc, $\eta = 0.1$), the reduction of the clump sizes
664: prevents accumulating and filling the cluster with a dense warm material, as
665: happens in model~9. As shown in Figure~\ref{rTd_tm2} and in Table~\ref{modtab}
666: the formation of new clumps via thermal instability is compensated with the
667: outflow from the computational zone making, model~13 more realistic.
668:
669:
670: \clearpage
671: \begin{figure}
672: \plotone{f4-bmp.eps}
673: \caption{Model~5 ($L_\mathrm{SC}/L_\mathrm{crit} = 20$, $\eta = 1$) at $t =
674: 0.025$~Myr (left column), $t = 0.044$~Myr (middle column) and $t = 0.4$~Myr
675: (right column). The first two rows of panels show the logarithm of the wind
676: temperature and density, respectively, across the whole computational domain.
677: The third row shows the logarithm of pressure in the cluster central region
678: (below $R_\mathrm{SC} = 10$~pc), and the bottom row shows the wind velocity in
679: the same region, as arrows together with its magnitude coded by the color scale.
680: The red arc is the stagnation radius given by the semi-analytical solution.}
681: \label{rTd43}
682: \end{figure}
683: %------------------------------------------------------------------------
684: \begin{figure}
685: \plotone{f5-bmp.eps}
686: \caption{Models~2-4 ($L_\mathrm{SC}/L_\mathrm{crit} = 2$, $\eta = 1$) at $t = 0.25$~Myr.
687: The left and right columns show the logarithm of the wind temperature and
688: density, respectively, in the cluster central region. The three rows of panels
689: represent the different grid resolutions: the top row is $150\times56$
690: (model~2), the middle row $300\times112$ (model~3) and the bottom row
691: $600\times224$ (model~4).
692: }
693: \label{rTd42}
694: \end{figure}
695: %------------------------------------------------------------------------
696: \begin{figure}
697: \plotone{f6-bmp.eps}
698: \caption{Model~6 ($L_\mathrm{SC}/L_\mathrm{crit} = 200$, $\eta = 1$)
699: at $t = 0.25$~Myr. The top left and right panels show the logarithm of the wind
700: temperature and density, respectively, across the whole computational domain.
701: The bottom panels show only the cluster region (below $R_\mathrm{SC} = 10$~pc),
702: the left panel represents the logarithm of pressure, the right panel shows
703: the wind velocity as arrows together with its magnitude coded by the color
704: scale. The red arc is the stagnation radius given by the semi-analytical
705: solution.}
706: \label{rTd44}
707: \end{figure}
708: %------------------------------------------------------------------------
709: \begin{figure}
710: \plotone{f7-bmp.eps}
711: \caption{Model~13 ($L_\mathrm{SC}/L_\mathrm{crit} = 25$, $\eta = 0.1$)
712: at $t = 0.74$~Myr. The meaning of the panels is the same as in
713: Figure~\ref{rTd44}.
714: }
715: \label{rTd_tm2}
716: \end{figure}
717: %------------------------------------------------------------------------
718:
719:
720: \begin{figure}
721: \plotone{f8.eps}
722: \caption{
723: The solid line represents the time evolution of the mass flux measured at the
724: outer boundary of the computation domain in model~5. The mass deposition
725: rate to the whole cluster $\dot{M}_\mathrm{SC}$ and its fraction deposited
726: between $R_\mathrm{st}$ and $R_\mathrm{SC}$ are represented by dash-dotted line and
727: dashed line, respectively. The dotted line represents the average flux at the
728: outer boundary in the period 0.2 - 0.8 Myr. Those average mass fluxes are shown
729: as symbols in Figure~\ref{dMOB}. The rarefied wind produces the flux close to the
730: value given by the dashed line. The maxima of $\dot{M}_\mathrm{out}$ are due to condensations
731: passing through the outer boundary. They are preceded by short periods in
732: which $\dot{M}_\mathrm{out}$ drops below the dashed line, as if the condensations
733: "cast shadows" to the regions above them -- the mass flux of the rarefied wind
734: is slightly lower then and corresponds to the moments when the outer
735: boundary is shadowed by an approaching condensation(s).
736: }
737: \label{fluxes43}
738: \end{figure}
739:
740: %------------------------------------------------------------------------
741: \begin{figure}
742: \plotone{f9.eps}
743: \caption{The fraction of the matter inserted within the stellar cluster which is
744: able to leave it as a function of the cluster luminosity normalized by its
745: threshold value ($L_\mathrm{SC}/L_\mathrm{crit}$). The solid line represents the
746: fraction of mass which is inserted above $R_\mathrm{st}$ and which leaves the
747: cluster in a form of the hot wind (which may eventually cool down outside the
748: cluster). The $\times$ symbols denote the 1D simulations described in
749: \citet{2007ApJ...658.1196T}. The other symbols represent the 2D
750: simulations: triangles (models 2 -- 6), the plus (model 7),
751: the circle (model~8), the square (model~12), and the diamond (model~13).
752: The values were obtained by averaging over time periods $0.2-0.8$~Myr (models 2
753: -- 6), and $1-2$~Myr (models~7, 8, 12 and 13). The dashed line is the fit to the
754: outflows of 2D models.}
755: \label{dMOB}
756: \end{figure}
757: \clearpage
758:
759: \section{Discussion}
760:
761: Figure~\ref{fluxes43} shows the time evolution of the mass flux through the
762: outer boundary of the computational domain for model~5. The outflow becomes
763: quasi-stationary as an equilibrium is established between the formation of high density
764: condensations and their ejection from the cluster and exit through the inner
765: boundary. The average mass flux across the outer boundary of the
766: computation domain (dotted line in Figure~\ref{fluxes43}) slightly exceeds the rate of mass
767: deposition that occurs between the stagnation radius $R_\mathrm{st}$ and the star
768: cluster edge $R_\mathrm{SC}$. This is because some dense condensations that
769: formed in the thermally unstable inner region ($r < R_\mathrm{st}$) cross the
770: stagnation surface and eventually leave the cluster contributing to the total
771: mass flux across the outer boundary of the computational grid.
772:
773:
774: Figure~\ref{dMOB} shows how the relative outflow from the cluster (measured as
775: the mass flux at the outer boundary, averaged over long time periods) depends on
776: the ratio $L_\mathrm{SC}/L_\mathrm{crit}$. We plot all the models where the
777: equilibrium between the clump formation and their removal through inner or outer
778: boundaries is reached. Other models (Nos 9, 10 and 11), where the volume of the
779: cluster is completely filled with the warm matter (see Sect. 3.5.) are omitted.
780: The outflow from the cluster consists of two components: the hot wind, which
781: originates in the outer region of the cluster above $R_\mathrm{st}$, and the
782: warm dense clumps that are formed below $R_\mathrm{st}$ and that manage to pass
783: to the outer region where they are ejected from the cluster. The figure shows
784: three zones: the bottom zone is the fraction of the deposited mass
785: that goes into the wind, the middle zone represents the mass in ejected
786: high density condensations that stream away with the wind ($\lesssim 20\%
787: \dot{M}_\mathrm{SC}$), and the upper zone is the mass which stays in the
788: cluster (available for star formation). Note that the hot wind also cools down
789: to temperatures $T \sim 10^4$ K at short distances from the cluster surface, so,
790: finally there are also two phases in the outflow: the warm wind and the dense
791: condensations (which should expand unless they cool even further).
792:
793:
794: In order to advance the problem further, apart from an adequate hydrodynamic
795: model that takes into consideration specific characteristics of SSCs (\ie their
796: high masses, small radii, large stellar densities and extreme output of
797: mechanical energy), one would need to couple the hydrodynamics to the UV
798: radiation field. We have assumed here that this is the case and that the UV
799: radiation field generated by the massive stars evolving in the cluster keeps the
800: temperature of the thermally unstable gas at $T \sim 10^4$~K through
801: photoionization. This may be true for very young clusters, before the supernova
802: era starts ($t_\mathrm{SN} \sim $3 Myr). However, older clusters, with a reduced
803: ionizing photon flux, would soon become unable to photoionize all the gas that
804: became thermally unstable. In such cases, (see Figure~\ref{rTd_tm2}) the
805: thermally unstable gas would continue to cool further, while being compressed
806: into correspondingly smaller volumes. As a result, the increasingly high
807: densities would trap the ionization front in the outer skins of the
808: condensations and their interior would remain neutral and at low temperature
809: ($\sim 10$~K). In this way, if a parcel of gas with an original temperature
810: $\sim 10^7$ K, that becomes thermally unstable and is able to cool down to
811: $10$~K, would experience a rapid evolution in which its volume, in order to
812: preserve pressure equilibrium, would be reduced by six orders of magnitude while
813: its density would become six orders of magnitude larger.
814:
815:
816: Another factor not taken into account in the present set of simulations is
817: gravity. The gravitational pull caused by the cluster is perhaps not
818: significant for the high temperature gas, as this has a sound speed of several
819: hundreds of km$\,$s$^{-1}$, much larger that the escape speed from the cluster.
820: However, it should promote a faster exit of low temperature condensations across
821: the inner grid boundary. Indeed, if one considers a condensation that develops
822: at the stagnation radius, its free-fall time to the cluster center will be
823: $\tau_\mathrm{ff} = \pi R_\mathrm{st}^{3/2} / (2GM_\mathrm{st})^{1/2}$ where
824: $M_\mathrm{st}$ is the mass below $R_\mathrm{st}$. In pc and solar mass units,
825: it is $\tau_\mathrm{ff} \approx 16.5 R_\mathrm{SC}^{3/2} /
826: M_\mathrm{SC}^{1/2}$~Myr which leads to time-scales much shorter than the
827: computational time. Thus gravity would lead to an increase in the speed
828: of such condensations as they move to cross the inner boundary. The implication
829: of this is a faster accumulation of the thermally unstable matter near the center
830: of the cluster, where further generations of stars are to take place. Another
831: important factor, also promoting a faster matter accumulation and further
832: stellar generations arises from the self-gravity of the thermally unstable gas.
833:
834:
835: \section {Conclusions}
836:
837:
838: Here we have confirmed by means of 2D hydrodynamic simulations the
839: existence of a bimodal solution for SSCs above the threshold line. We have shown
840: that the evolution within the volume defined by the stagnation surface is very
841: dynamic. The stagnation surface itself has a very dynamic morphology that
842: continuously changes with time. Nevertheless, the average value of the stagnation
843: radius remains close to the value predicted by 1D simulations and
844: semi-analytic solutions. This region suffers a very dynamic evolution in which
845: parcels of gas continuously become thermally unstable and are rapidly driven
846: into very small volumes to compensate their sudden loss of pressure. The number
847: of such regions depends on the excess star cluster mechanical luminosity above
848: the threshold value. The fraction of the cluster volume occupied by the warm
849: medium depends on the balance between the formation of high density
850: condensations, via thermal instability, and their removal via secondary star
851: formation and/or their escape from the cluster. In our model, the secondary star
852: formation is partly accounted for with the exit of high density gas across the
853: inner boundary. However, a better treatment in the future would be to implement
854: a more physical description of secondary star formation.
855:
856: We have also shown that the growth of high density condensations within a SSC
857: volume, is strongly linked to the parameter $\eta$, as this determines the
858: location of the critical luminosity, $L_\mathrm{crit}$, in the star cluster
859: mechanical luminosity vs size plane, and thus determines how far above the
860: critical luminosity a cluster is. It also influences the sound speed (and
861: pressure) in the thermal instability region and as a result it fixes the
862: final high density that condensations, confined by pressure, attain
863: (higher $\eta$ $\Rightarrow$ higher pressure $\Rightarrow$ denser
864: condensations).
865:
866:
867: We have shown that most of the condensations generated within the stagnation
868: volume, are unable to leave the cluster volume and thus accumulate. Eventually
869: this will result in further generations of star formation. The central
870: implication of this result is that most of the metals processed by stars in
871: massive and compact SSCs will not be ejected back into the host-galaxy
872: ISM, an important issue to be taken into consideration by models of galactic
873: chemical evolution and $\Lambda$CDM models of the universe. Careful analysis of
874: observational data of the most massive and compact star clusters is now
875: required to select those which may evolve in the bimodal, catastrophic cooling
876: regime. For all of them we expect a mixture hot X-ray gas, a warm partially
877: photo-ionized plasma, and an ensemble of accumulating cool dense
878: condensations. Thus they should be detectable in the X-ray band, visible,
879: radio and infrared regimes simultaneously.
880:
881:
882: Note also, that the high stellar densities associated with SSCs resemble those
883: of globular clusters, in which star-to-star abundance inhomogeneities have been
884: observed \citep{2007ApJ...665.1164B}. This can be understood if globular
885: clusters have entered the bimodal regime during their early evolution,
886: and this has led to multiple stellar generations forming with the matter
887: reinserted into the cluster volume. Such clusters, having a reduced amount of
888: matter being lost back into the ISM, would have remained more stable against
889: disruption.
890:
891: Other open issues and some more astrophysical consequences of clusters
892: undergoing this bimodal evolution have been discussed by
893: \citet{2008Ap&SS.granada.jan}.
894:
895: \acknowledgements
896:
897: The authors would like to express their thanks to Anthony Whitworth for many
898: helpful discussions leading to a significant improvement of the paper. An
899: anonymous referee also provided valuable comments and suggestions. This study
900: has been supported by CONACYT - M\'exico, research grant 60333 and 47534-F and
901: AYA2004-08260-CO3-O1 from the Spanish Consejo Superior de Investigaciones
902: Cient\'\i{}ficas, by the Institutional Research Plan AV0Z10030501 of the Academy
903: of Sciences of the Czech Republic and by project LC06014 Center for Theoretical
904: Astrophysics of the Ministry of Education, Youth and Sports of the Czech
905: Republic. RW acknowledges support by the Human Resources and Mobility Programme
906: of the European Community under the contract MEIF-CT-2006-039802.
907:
908: \bibliographystyle{aa}
909: \bibliography{winds2d}
910:
911: \end{document}
912: \bye
913:
914:
915: