1:
2: \documentclass[12pt,preprint]{aastex}
3: \usepackage{natbib}
4: \usepackage{graphicx}
5:
6: \bibpunct{(}{)}{;}{a}{}{,}
7:
8: \newcommand{\new}{}
9: \newcommand{\newrsk}{}
10:
11: \begin{document}
12:
13: \title{The Formation of Stellar Clusters in Turbulent Molecular Clouds: \\
14: Effects of the Equation of State}
15:
16: \author{Yuexing Li\altaffilmark{1,2}, Ralf S. Klessen\altaffilmark{3} and
17: Mordecai-Mark Mac Low\altaffilmark{1,2}}
18:
19: \affil{$^{1}$Department of Astronomy, Columbia University, New York,
20: NY 10027, USA}
21: \affil{$^{2}$Department of Astrophysics, American Museum of Natural
22: History, 79th Street at Central Park West, New York, NY 10024-5192, USA}
23: \affil{$^{3}$Astrophysikalisches Institut Potsdam, An der Sternwarte
24: 16, D-14482 Potsdam, Germany}
25: \email{yxli@astro.columbia.edu, rklessen@aip.de, mordecai@amnh.org}
26:
27: \begin{abstract}
28:
29: We study the effect of varying the equation of state on the formation of
30: stellar clusters in turbulent molecular clouds, using three-dimensional,
31: smoothed particle hydrodynamics simulations. Our results show that the
32: equation of state helps determine how strongly self-gravitating gas
33: fragments. The degree of fragmentation decreases with increasing
34: \new{polytropic exponent} $\gamma$ in the range $0.2 < \gamma < 1.4$,
35: although the total amount of \new{mass accreted onto collapsed fragments}
36: appears to remain roughly
37: constant through that range. Low values of $\gamma$ are expected to lead to
38: the formation of dense clusters of low-mass stars, while $\gamma>1$ probably
39: results in the formation of isolated and massive stars. Fragmentation and
40: collapse ceases entirely for $\gamma > 1.4$ as expected from analytic arguments.
41: The mass spectrum of overdense gas clumps is roughly log-normal for {\em
42: non}-self-gravitating turbulent gas, but changes to a power-law under the
43: action of gravity. The spectrum of collapsed cores, on the other hand,
44: remains log-normal for $\gamma\le 1$, but flattens markedly for $\gamma >1$.
45: The density PDFs approach log-normal, with widths that decrease with
46: increasing $\gamma$. Primordial gas may have effective $\gamma >
47: 1$, in which case these results could help explain why models of the
48: formation of the first stars tend to produce isolated, massive
49: objects.
50: \end{abstract}
51:
52: \keywords{ISM: clouds --- ISM: hydrodynamics --- stars: formation --- equation
53: of state --- turbulence}
54:
55: \section{INTRODUCTION}
56:
57: Stars form alone and in groups, associations, and clusters (see
58: \citealt{pud02} and \citealt{ward02} for recent reviews). Yet the
59: origin of this diversity in the process of star formation remains
60: poorly understood. Recent work suggests that stars form by
61: gravitational collapse of shock-compressed density fluctuations
62: generated by the supersonic turbulence generally observed in molecular
63: clouds (e.g. \citealt{elm93, pn99, khm00, osg01}; see \citealt{mk03} for a
64: recent review). If so, then star formation is controlled by turbulent
65: fragmentation and dynamical collapse of individual, Jeans-unstable,
66: protostellar clumps.
67:
68: Fragmentation of gas clouds has been studied for more than a century,
69: yet the process remains poorly understood. There have been analytical models
70: \citep{jeans02, low76}, and numerical investigations of the effects of various
71: physical processes on the collapse, such as geometry and rotation of the
72: clouds (see \citealt{bob93} for a review), and magnetic fields
73: \citep{galli01}. Recently, the effects of turbulence have been studied
74: extensively in a series of three-dimensional simulations using
75: both grid-based, Eulerian and particle-based, Lagrangian hydrodynamics
76: (\citealt{kbb98}; \citealt[hereafter Paper I]{kb00}; \citealt[Paper
77: II]{khm00}; \citealt[Paper III]{hmk01}; and \citealt[Paper IV]{rsk01}).
78: These papers show that turbulence strongly influences the fragmentation of
79: molecular clouds. Fast, clustered collapse and star formation occurs
80: in regions with turbulence insufficient to support against
81: gravitational collapse, while slow, scattered star formation results
82: from strong turbulent support. However, these results are based on
83: isothermal models with \new{polytropic exponent} $\gamma=1$ only. The role of
84: the equation of state (EOS), which is essential in understanding the
85: physical structure and stability of the turbulent clouds, remained
86: unexplored.
87:
88: The balance of heating and cooling in a molecular cloud can be
89: approximately described by a polytropic EOS
90: \begin{equation}
91: P = K\rho^{\gamma},
92: \label{eq_eos}
93: \end{equation}
94: where $K$ is a constant, and $P, \rho$ and $\gamma$ are thermal pressure, gas
95: density and \newrsk{polytropic exponent}, respectively. The density
96: structure generated by supersonic motions in highly compressible
97: turbulence depends on the EOS. The stiffness of the EOS can largely
98: control the density probability distribution function (PDF) in strongly
99: compressible turbulence \citep{scalo98, pv98, vg01}. In particular,
100: one-dimensional simulations by \citet{pv98} show the shape of
101: the density PDF varying with $\gamma$. They found that the PDF is log-normal
102: when $\gamma = 1$ (isothermal), and develops a power-law tail for $\gamma \neq
103: 1$. \new{\citet{ss00} showed that radiatively cooling gas can be described by a
104: piecewise polytropic EOS, in which the polytropic exponent $\gamma$ changes with gas
105: density \newrsk{$\rho$}. They conducted a detailed analysis including
106: chemistry, thermal balance and radiative transfer, and found that in the
107: interstellar medium the effective polytropic exponent $\gamma$ has a range of
108: $0.2 < \gamma < 1.4$. \newrsk{They further predicted that}
109: the initial mass function (IMF) of protostellar cores would vary with
110: $\gamma$, yielding a peaked IMF for $\gamma > 1$ and a power-law function for
111: $\gamma < 1$}.
112:
113: In this paper, we perform three-dimensional simulations of
114: self-gravitating, supersonic turbulence over the range of $\gamma$
115: identified by \citet{ss00} to investigate the \newrsk{effect of varying the
116: EOS on the dynamical evolution and fragmentation behavior} of turbulent
117: clouds and the subsequent formation of protostellar clusters.
118: \newrsk{To quantify the dependency on $\gamma$, we consider a true polytropic
119: EOS, with $\gamma$ being strictly constant in each simulation regardless of
120: the density, rather than the piecewise polytrope used by \citet{ss00} and
121: earlier workers.}
122:
123: In \S\ref{sec_com} we describe our
124: computational methods; in \S\ref{sec_result} we give results on
125: fragmentation, the resulting mass distribution of protostellar cores,
126: and the gas density PDF; and in \S\ref{sec_discussion} we conclude and
127: speculate about the implications or our work for the formation of star
128: clusters and primordial star formation.
129:
130: \section{METHODS}
131: \label{sec_com}
132: In analytical work on the stability of a self-gravitating, isothermal medium,
133: \citet{jeans02} derived a relation between the oscillation frequency $\omega$
134: and the wavenumber $k$ of small perturbations,
135: \begin{equation}
136: \label{eq_jeans1}
137: \omega^2 - c^2_sk^2 + 4\pi G\rho =0
138: \end{equation}
139: where $c_s$ is the isothermal sound speed, $G$ the gravitational
140: constant, and $\rho$ is the gas density. The medium is
141: unstable to fragmentation at all wavelengths greater than a critical
142: length $\lambda_J = 2\pi / k_J$, or equivalently all masses exceeding
143: the Jeans mass,
144: \begin{equation}
145: \label{eq_jeans2}
146: M_J = \rho\lambda^3_J = \left(\frac{\pi}{G}\right)^{3/2} \rho^{-1/2} c^3_s
147: \end{equation}
148: will collapse under their own weight. Note that we use a cubic
149: definition of $M_J$.
150:
151: In a polytropic cloud with EOS given by equation~\ref{eq_eos},
152: the sound speed is
153: \begin{equation}
154: c_s = \left(\frac{dP}{d\rho}\right)^{1/2} =
155: (K\gamma)^{1/2}\rho^{(\gamma-1)/2},
156: \end{equation}
157: so the Jeans mass will be given as a function of $\gamma$ and $\rho$,
158: \begin{equation}
159: \label{eq_jeans3}
160: M_J = \left(\frac{K\pi}{G}\right)^{3/2} \gamma^{3/2}
161: \rho^{\frac32(\gamma-\frac43)} \,.
162: \end{equation}
163: \new{The specific internal energy is $u = K\rho^{\gamma-1}/(\gamma_{ad} -1)$, where
164: $\gamma_{ad}$ is the adiabatic index, with $\gamma_{ad} = 5/3$ for ideal gas}. For
165: isothermal gas, $\gamma =1$, and $K = c^2_s$.
166:
167: We carry out simulations to determine when and how fragmentation
168: occurs. We use a smoothed particle hydrodynamics (SPH) code \citep{benz90} in
169: order to resolve several orders of magnitude in density during the collapse of
170: shock-compressed regions. Our version of the code includes periodic boundary
171: conditions \citep{rsk97} and can replace high-density cores with sink
172: particles \citep{bbp95}. Sink particles accrete surrounding gas particles
173: while conserving mass and momentum, but they only interact gravitationally. They
174: prevent time step from becoming prohibitively short in very dense
175: regions. This allows us to follow the dynamical evolution of the system over
176: many free-fall times.
177:
178: We set up a periodic cube with side length $L=2$ and total mass of
179: unity such that initial density $\rho_0 = 0.125$, and sound speed $c_s
180: = 0.1$. The treatment of the polytropic EOS follows \citet{bob91}, where we
181: set $K =(R_g/\mu)T_0\rho^{1-\gamma}_0$. We choose values consistent with the
182: sound speed of $R_g = 2/3$, $\mu = 1$, and $T_0 = 0.015$ for the gas constant,
183: mean molecular weight, and initial temperature of the cloud. Uniform
184: turbulence is driven with the method described by \citet{mm98} and
185: \citet{mm99}, adding energy over a small range of wavenumbers $k$. Driving in
186: two wavenumber ranges is considered here, $1 \le k \le 2$ and $7 \le k \le
187: 8$, corresponding to models $B1h$ and $B3$ in Paper II. In each case,
188: the driving strength is chosen to ensure the same rms velocity despite
189: the different driving wavenumbers, so that the average turbulent Jeans
190: mass $\langle M_{J}\rangle_{turb} = 3.2$ and Jeans Number $N_J = 64$,
191: are the same. In a self-gravitating medium, the maximum resolvable
192: density is determined by the requirement that the local thermal Jeans
193: mass be resolved by the SPH kernel \citep{bab97}; our models all have the same
194: resolution limit.
195:
196: The models presented here are computed in normalized units. If scaled
197: to mean densities of $n({\rm H}_2) \approx 10^2 $cm$^{-3}$, and a
198: temperature of $11.4 $K (i.e.\ an isothermal sound speed $c_s = 0.2
199: \,$km$\,$s$^{-1}$), values appropriate for a dark cloud like Taurus
200: \citep{hart98}, then our simulation cube holds a mass of $M \approx 4
201: \times 10^3\,$M$_{\odot}$ and has a size of $L \approx 9\,$pc. For details on
202: the scaling relations, see Paper II.
203:
204: We computed models with $0.2 \leq \gamma \leq 1.4$, covering the range
205: suggested for interstellar gas by \citet{ss00}. We vary
206: $\gamma$ in steps of 0.1 in otherwise identical simulations for both
207: driving wavenumbers. Each simulation starts with a uniform density
208: $\rho_0 = 0.125$. Driving begins immediately, while self-gravity is
209: turned on at t=2.0, after the turbulence is fully established (see
210: Paper II). Our models use $200\,000$ particles. Twenty-six simulations
211: were performed simultaneously using a serial version of the code on
212: single 1 GHz Pentium III processors of the Parallel Computing Facility
213: of the American Museum of Natural History (AMNH). Each computation
214: took about 6 months.
215:
216: \section{RESULTS}
217: \label{sec_result}
218: \subsection{Turbulent Fragmentation}
219: \label{subsec_frag}
220: Figure~\ref{fig_sinkdis1} and Figure~\ref{fig_sinkdis2} show the
221: density distribution of the gas. To make the figures, a 256$^3$ grid
222: was filled with densities properly computed from the SPH kernel, and
223: the result is displayed in three-dimensional projection (\textit{top
224: row}) and in a slice through maximum density one free-fall time after
225: self-gravity is turned on for selected $\gamma$ and both driving
226: wavenumbers. As found in previous studies (e.g.\ \citealt{mm99}), the
227: driving wavenumber strongly influences the density distribution.
228: Driving with $k=1$--2 produces strong filamentary structure, while the
229: density distribution for $k=7$--8 remains more uniform at large scale.
230:
231: We find that the value of $\gamma$ is as important as the driving in
232: determining collapse behavior. As $\gamma$ increases, the number of
233: collapsed cores replaced by sink particles decreases, and the cores
234: cluster less. For the same $\gamma$, driving with $k=1$--2 produces
235: both far more, and more clustered dense cores than driving with
236: $k=7$--8. The dense cores tend to collapse predominantly in filaments
237: or at intersections of filaments. At $\gamma \geq 1.1$, no
238: fragmentation occurs by the time shown for driving with $k=7$--8.
239:
240: The isothermal case, $\gamma=1.0$, agrees well with previous results
241: in Papers~II and~III. Our current investigation agrees with the
242: results in Paper I--III, that collapse tends to form clusters with
243: high efficiency in regions with weak turbulence, while in regions with
244: strong turbulence, sparse, slow collapse occurs. We now add the
245: additional criterion that collapse and fragmentation depends strongly on
246: $\gamma$.
247:
248: In regions with $\gamma < 1$, fragmentation
249: occurs earlier and more frequently, while in regions where $\gamma > 1$,
250: fragmentation is retarded and less frequent.
251:
252: \placefigure{fig_sinkdis1}
253:
254: \placefigure{fig_sinkdis2}
255:
256: Figure \ref{fig_sinknum} compares the number of collapsed cores at
257: different $\gamma$ for models with different driving wavenumbers.
258: The rate at which new protostellar cores form differs for different
259: $\gamma$. Models with low $\gamma$ form cores quickly, while models
260: with large $\gamma$ form new cores more rarely.
261: Again we see more cores for driving with
262: $k=1$--2 than for $k=7$--8.
263:
264: \placefigure{fig_sinknum}
265:
266: \citet{rb90} found an empirical scaling relation $J_c
267: \propto \gamma^2$ of the critical Jeans number $J_c = M/M_J$ for
268: collapse for elongated clouds. The Jeans number should be related to
269: the number of fragments. However, with higher resolution and more
270: statistics for a wide range of $\gamma$ in our simulations, we find
271: something more like the inverse, but strongly dependent on the details
272: of the driving. Fig \ref{fig_sinkgamma} shows the
273: relation between the value of $\gamma$ and the number of fragments at
274: one free-fall time. The number of fragments drops quickly as $\gamma$
275: increases for driving with $k=1$--2, and more slowly for driving with
276: $k=7$--8, which produces fewer fragments to begin with.
277:
278: \placefigure{fig_sinkgamma}
279:
280: Figure \ref{fig_sinkacc} shows the accretion histories (the time
281: evolution of the combined mass fraction of all protostellar cores) for
282: all the individual $\gamma$ cases and for both driving
283: wavelengths. Note that much less mass is accreted when driving with
284: $k=7$--8. We focus on the $k=1$--2 model. The lower $\gamma$ is, the
285: earlier fragmentation occurs and the larger the number of fragments
286: that form. It also takes longer for the clouds with high $\gamma$ to
287: fragment. The slope of the accretion curves are roughly the same, but
288: the number of sinks is very different for different $\gamma$ at the
289: same time, which suggests that the average core mass is different for
290: different $\gamma$. A low-$\gamma$ environment produces a large number
291: of collapsed cores of relatively low mass, while $\gamma > 1$ results
292: in the formation of considerably fewer, but more massive, cores.
293: Since the overall mass growth rate of collapsing cores is closely
294: related to the expected star formation rate, our result suggests that
295: the overall star formation rate is higher in low-$\gamma$ clouds than
296: in high-$\gamma$ clouds. It remains unclear what the final amount of
297: mass in stars will be, as higher mass stars may drive surrounding gas
298: away more quickly (see also \citealt{vbk03}).
299:
300: \placefigure{fig_sinkacc}
301:
302: We do not find any signs of collapse in the model with $\gamma=1.4$
303: even after several free-fall times. This can be explained with a
304: Jeans mass analysis giving the critical condition for gravitational
305: collapse. From Equation~[\ref{eq_jeans3}], we have
306: \begin{equation}
307: \label{eq_jeans4}
308: \frac{\partial M_J}{\partial \rho} \propto \frac{3\gamma -
309: 4}{2}\rho^{(3\gamma - 6)/2}\,.
310: \end{equation}
311: This implies that for $\gamma < 4/3$ the Jeans mass $M_J$ decreases as
312: density increases during collapse; for $\gamma = 4/3$, $M_J$ remains
313: constant; while if $\gamma > 4/3$, then $M_J$ \textit{increases} with
314: density. Thus, any collapse will choke itself off when $\gamma >
315: 4/3$, as found in our numerical results.
316:
317: Fragmentation is a complex process depending on the local conditions
318: in the cloud, so the above analysis for an isolated spherical
319: perturbation is an approximation. Indeed, fragmentation already ceases
320: for $\gamma > 1.1$ in the model driven with $k = 7-8$, suggesting the
321: importance of the details of the turbulence in determining collapse
322: behavior.
323:
324: \subsection{Mass Spectrum}
325: \label{subsec_mas}
326: Observations suggest that the mass distribution of gas clumps in
327: molecular clouds follows a power law, $dN/dM = M^{\nu}$, with typical
328: values of the exponent being $\nu \approx -1.5$
329: (e.g. \citealt{wbm00}). \citet{salp55} derived a power law for the high mass
330: stellar IMF in the same notation with $\nu = -2.35$ (see also \citealt{scalo98,
331: kroupa02}).
332:
333: We measure clump-mass spectra in our models using a clump-finding
334: algorithm similar to the one described by \citet{wdb94} but working on all
335: three spatial coordinates and adapted to make use of the SPH kernel smoothing
336: procedure (for details see Appendix A in Paper I). Figure \ref{fig_mas} shows
337: the spectra of gas clumps and collapsed cores for models with different values
338: of $\gamma$, driven with $k=1$--2. Three different evolutionary phases
339: are shown: after turbulence has been fully established, but before
340: self-gravity has been turned on, at $t=2$; and when the fraction of
341: mass accumulated in collapsed cores (sink particles) has reached
342: $M_{\ast} \approx 20\%$ and $40\%$.
343:
344: \placefigure{fig_mas}
345:
346: The mass distribution of both clumps and collapsed cores changes with
347: $\gamma$, with the effect being most pronounced for the cores. In
348: low-$\gamma$ models, the core mass spectrum at the high-mass end is
349: roughly log-normal. As $\gamma$ increases, fewer but more massive
350: cores emerge. When $\gamma > 1.0$, the distribution is dominated by a few
351: high mass cores, and the spectrum tends to flatten out. It is no
352: longer fit by either a log-normal or a power-law. The clump mass
353: spectra, on the other hand, do show power-law behavior on the high
354: mass side, even for $\gamma > 1.0$.
355:
356: Our results suggest that massive stars can form in small groups or
357: even in isolation in gas with $\gamma > 1.0$. \citet{ss00}
358: suggested that a stiff EOS with $\gamma > 1.0$ should lead to a peaked IMF,
359: biased toward massive stars, while an EOS with $\gamma < 1.0$ results
360: in a power-law IMF, in general agreement with our simulations.
361:
362: \subsection{Density Probability Distribution Function}
363: \label{subsec_pdf}
364:
365: \placefigure{fig_pdf_int}
366:
367: Figure \ref{fig_pdf_int} shows mass and volume weighted PDFs of gas
368: density $p_m$ and $p_v$ before self-gravity has been turned on for
369: selected values of $\gamma$ and both driving wavenumbers. We also
370: show Gaussian fits to the PDFs together with their error $\epsilon =
371: \sum{|{p - p_{fit}}|/{p}}$, summed over all points above $10\%$ of
372: the peak values.
373:
374: The mass-weighted density PDF $p_m(\rho)$ is calculated directly from
375: the local density associated with each SPH particle, as described by
376: \citet{rsk00}. The volume-weighted density PDF $p_v(\rho)$ is
377: calculated on a cubic grid derived from the SPH density field by
378: taking a cube of $256^3$ cells and using the SPH kernel smoothing
379: procedure to find the density at the center of each cell. Both
380: $p_m(\rho)$ and $p_v(\rho)$ are normalized so that
381: $\int_{-\infty}^{\infty} p_m(\rho)d\rho = \int_{-\infty}^{\infty}
382: p_v(\rho)d\rho = 1$.
383:
384: The density behind shocks in supersonic turbulence depends on the
385: compressibility of the gas,
386: \begin{equation}
387: d\rho \propto \frac{1}{\gamma}P^{\frac{1}{\gamma} - 1}dP\,,
388: \end{equation}
389: So turbulent density fluctuations should increase as $\gamma$
390: decreases. Indeed, we see that in Figure \ref{fig_pdf_int}, the width
391: of $p_m$ and $p_v$ increases as $\gamma$ decreases.
392:
393: In models driven with wavenumber $k=7-8$, there are super-Gaussian
394: tails in the low-density end in the volume-weighted PDFs for all
395: $\gamma$ and in the mass-weighted PDF for $\gamma=0.2$.
396: One-dimensional simulations by \citet{pv98}
397: showed that the volume-weighted density PDF of supersonic turbulent
398: gas displays a power-law tail at high densities for $0 < \gamma < 1$,
399: becomes log-normal for $\gamma = 1$ (isothermal), and develops a
400: power-law tail at low density for $\gamma > 1$, which is only in
401: partial agreement with our results. The discrepancy may be due to the
402: low resolution in low-density regions in our simulations, the
403: difference between one- and three-dimensional simulations, or
404: different driving.
405:
406: A full investigation of the density PDFs under various conditions,
407: with different codes, different dimensions (three-dimensional vs.\
408: one-dimensional), different resolution, different turbulence driving,
409: different initial Mach number, and with and without magnetic field is
410: beyond the scope of this paper; these issues will be addressed
411: elsewhere.
412:
413: \subsubsection{Comparison Between Mass and Volume Weighting}
414:
415: In order to compare the mass and volume weighted PDFs, we will find it
416: useful to define $s=\ln (\rho/\rho_0)$; the mass and volume weighted
417: PDFs of $s$ are then $p_m(s)$ and $p_v(s)$, respectively. We can
418: relate $p(s)$ to $p(\rho)$ if we remember that for any montonic
419: function $y(x)$, it can be shown that the PDF $|p(y)dy| =
420: |p(x)dx|$. Therefore,
421: \begin{equation}
422: p_m(\rho) = p_m(s)\frac{ds}{d\rho} = \frac{p_m(s)}{\rho}\,,
423: \end{equation}
424: and
425: \begin{equation}
426: p_v(\rho) = p_v(s)\frac{ds}{d\rho} = \frac{p_v(s)}{\rho} \,.
427: \end{equation}
428:
429: Since $p_m(\rho) \propto dM/d\rho$, and $p_v(\rho) \propto dV/d\rho$
430: \citep{osg01}, it follows that
431: \begin{equation}
432: p_m(\rho) \propto \frac{dM}{dV} \frac{dV}{d\rho} \propto \rho
433: p_v(\rho).
434: \end{equation}
435: Then we can relate $p_m(s)$ to $p_v(s)$
436: \begin{equation}
437: \label{eq_pm_pv}
438: p_m(s) = \rho p_m(\rho) = C \rho^2 p_v(\rho) = C e^s p_v(s)\,.
439: \end{equation}
440: where C is a normalization constant.
441:
442: If $p_v(s)$ follows a normal distribution,
443: \begin{equation}
444: \label{eq_pv}
445: p_v(s) = \frac{1}{\sqrt{2\pi} \sigma}\exp\left(-\frac{(s -
446: s_v)^2}{2\sigma^2}\right)\,,
447: \end{equation}
448: where $s_v$ is the average (in volume) value of $s$, and $\sigma$ is the
449: dispersion, then
450: \begin{equation}
451: \label{eq_pm_0}
452: p_m(s) = \frac{C}{\sqrt{2\pi} \sigma} \exp\left(-\frac{s^2_v - (s_v +
453: \sigma^2)^2}{2\sigma^2}\right) \exp\left(-\frac{\left[s - (s_v
454: +\sigma^2)\right]^2}{2\sigma^2}\right)\,.
455: \end{equation}
456: Since $\int_{-\infty}^{\infty} p_m(s)ds =1$, the normalization
457: \begin{equation}
458: C = \exp\left(\frac{s^2_v - (s_v +
459: \sigma^2)^2}{2\sigma^2}\right)\,.
460: \end{equation}
461: Equation~\ref{eq_pm_0} can thus be rewritten as:
462: \begin{equation}
463: \label{eq_pm}
464: p_m(s) = \frac{1}{\sqrt{2\pi} \sigma} \exp\left(-\frac{(s -
465: s_m)^2}{2\sigma^2}\right)\,.
466: \end{equation}
467: where $s_m = s_v + \sigma^2$. This is also a normal distribution, with the
468: same dispersion $\sigma$ as that of $p_v$ but shifted average value
469: $s_m$. Using the volume PDF normalization
470: $\int \rho \cdot p_v(s)ds = \int e^s p_v(s)ds = 1$,
471: it can be seen
472: \begin{equation}
473: \label{eq_sm_sv}
474: s_m = -s_v = \frac{\sigma^2}{2}\,.
475: \end{equation}
476: as noted, for example, by \citet{osg01}. This
477: holds only if the density PDF is log-normal. (Note that in this case
478: $C= 1$.)
479:
480: This derivation gives several properties of the density PDFs $p_m$
481: and $p_v$. If one of them is log-normal, so is the other; and if
482: they are log-normal, the width $\sigma$ of both profiles should be equal
483: (Eqs.~[\ref{eq_pv}] and~[\ref{eq_pm}]), and the peaks of the
484: PDFs $p_m(s)$ and $p_v(s)$ should lie symmetrically around zero
485: (Eq.~[\ref{eq_sm_sv}]).
486:
487: In Figure \ref{fig_pdf_int} the volume-weighted PDFs lie to the left
488: of the mass-weighted PDFs, with $s_m \approx -s_v$, as predicted. The
489: decent Gaussian fits to the peaks of the PDFs show that the assumption
490: of log-normal behavior is not bad. Three-dimensional grid-based
491: simulations of decaying turbulence by \citet{osg01} show similar Gaussian
492: PDFs.
493:
494: A detailed examination of the moments of the density PDFs, shown in
495: Fig \ref{fig_mom_k1278}, shows that these PDFs are, in fact, not
496: perfect log-normals. The first moment (mean) shows that $p_m$ and
497: $p_v$ of models with both driving wavenumbers $k=1-2$ and $k=7-8$ are
498: not exactly symmetrically distributed. This could be due to turbulent
499: intermittency in our small boxes producing deviations from a purely
500: Gaussian distribution. The Gaussian behavior expected from a well
501: sampled distribution is more visible in models driven with $k=7-8$
502: compared to $k=1-2$. In fact, $p_m$ and $p_v$ of $k=7-8$ are more
503: equal in width, as seen from the second moment (variance). This is a
504: sign of better sampling -- more modes contribute to the overall
505: velocity field, thus the central limit theorem is more appropriately
506: applied to the $k=7-8$ than to the $k=1-2$ case \citep{rsk00}.
507:
508:
509: \placefigure{fig_mom_k1278}
510:
511:
512: \subsubsection{Effect of Self-Gravity}
513: Figure \ref{fig_pdf_3phases} shows the evolution of the mass-weighted
514: gas density PDF $p_m$ of the model driven with $k=1-2$ for different
515: values of $\gamma$. The same three evolutionary phases as defined in
516: \S\ref{subsec_mas} are shown: the initial distribution, and times at
517: which dense cores have accreted $M_{\ast} \approx 20\%$ and $M_{\ast}
518: \approx 40\%$ of the mass. These PDFs do not include gas accreted
519: into the cores. We again show the best-fit Gaussian for the part of
520: each PDF above 10\% of peak.
521:
522: During the dynamical evolution, the PDFs in the non-isothermal cases develop
523: pronounced positive deviations from the the Gaussian fit, while the PDF in the
524: isothermal case remains well fit by a Gaussian (of increasing width),
525: consistent with the results in \citet{rsk00}. The high-density tails
526: gradually diminish as $\gamma$ increases from 0.2 to 1.0, vanishing almost
527: entirely at $\gamma = 1.0$. At $\gamma > 1.0$, though, the tails develop
528: again and get stronger as $\gamma$ continues to increase. We do not fully
529: understand why the tails only appear in non-isothermal cases.
530:
531: The first four moments of the density PDFs for each of the three
532: evolutionary phases discussed previously are shown in
533: Figure~\ref{fig_mom}. We can see that the moments vary as collapse
534: proceeds, demonstrating that self-gravity plays an important role in
535: shaping the density PDF.
536:
537: The plot of the skewness shows that the PDFs of turbulence
538: without self-gravity are always skewed to low densities, regardless of
539: $\gamma$. Shock compression comes at a cost: the gas swept up in shock
540: fronts has to come from somewhere, and those regions now contain only
541: low-density gas. As the filling factor of shocks is low \citep{smh00} most of
542: the volume is occupied by these low-density regions resulting in PDFs with
543: negative skewness. The situation changes, however, when self-gravity comes
544: into play. Gravitational contraction produces high densities in addition to
545: the turbulent shock compression, the PDF's in the later stages of evolution
546: are thus strongly biased towards high-density gas and the skewness becomes
547: positive.
548:
549: \placefigure{fig_pdf_3phases}
550:
551: \placefigure{fig_mom}
552:
553:
554: \section{SUMMARY AND DISCUSSION}
555: \label{sec_discussion}
556:
557: We have investigated the effect of varying the equation of
558: state (EOS) on the formation of stellar clusters in turbulent
559: molecular clouds using a set of three-dimensional numerical
560: simulations. Our results show that the EOS plays an important role in
561: the fragmentation of the clouds, the determination of the initial
562: mass function (IMF) of the protostellar cores, and the shape of the
563: gas density PDFs.
564:
565: The ability of interstellar gas clouds to fragment under the action of
566: self-gravity {\em decreases} with {\em increasing} \new{polytropic exponent}
567: $\gamma$ in the range $0.2 < \gamma < 1.4$ relevant for Galactic molecular
568: clouds. The total amount of mass in collapsed cores, however, appears to
569: remain roughly constant through that range. At least half of this material is
570: expected to accrete onto the protostellar system at the center of each
571: collapsing core (e.g. \citealt{rsk01}). Small values of $\gamma$ thus lead to
572: the formation of dense clusters of predominantly low-mass stars, while
573: $\gamma>1$ results in the formation of isolated and massive stars.
574: Fragmentation and collapse ceases entirely for $\gamma > 1.4$ as
575: expected from analytic arguments.
576:
577: The mass spectra of both clumps and collapsed cores change with $\gamma$, with
578: that effect being most pronounced for the cores. For $\gamma \le 1.0$, their
579: mass spectrum appears roughly log-normal. As $\gamma$ increases, fewer but
580: more massive objects emerge. When $\gamma > 1.0$, the distribution is
581: dominated by high mass objects only, and the spectrum tends to flatten out. It
582: is fit by neither a log-normal nor a power-law. The clump mass spectra, on the
583: other hand, do show power-law behavior at the high mass end for all $\gamma$.
584:
585: The density PDF changes with $\gamma$ as well. The width of the
586: profile decreases as $\gamma$ increases, as expected from analytic
587: arguments about shock compression. Both mass- and volume-weighted
588: PDFs show imperfect log-normal distributions. Self-gravity helps
589: shape the PDF. During dynamical collapse, the PDF develops a
590: high-density tail, which is the imprint of local collapse of high
591: density regions.
592:
593: \newrsk{In a polytropic gas cloud, any temperature fluctuation corresponds to
594: a density fluctuation, $\Delta\log_{10}(T) \simeq \Delta\log_{10}(\rho)(1 -
595: \gamma) -0.6$. We note that from the gas density PDFs in Figure
596: \ref{fig_pdf_int}, the maximum width of the profiles is about 12, which
597: corresponds to $\Delta\log_{10}(\rho) \simeq 5$. So for the extreme case
598: $\gamma = 0.2$, $\Delta\log_{10}(T) \simeq 3.5$. For a cloud with an average
599: temperature of about 10 K the maximum temperature of the low-density gas
600: therefore could reach values up to $10^4$ K, if $\gamma$ is extremely
601: low. This still falls in the temperature range observed in star forming regions.}
602:
603: \newrsk{The principal motivation of this paper is to investigate how the
604: fragmentation behavior of a turbulent cloud and the subsequent
605: formation of stars are affected by physical state of the cloud, as
606: described by the EOS. Since $\gamma$ reflects the balance between
607: heating and cooling, $\gamma$ is expected bo be low in regions of
608: strong cooling, while $\gamma$ may be large where the cooling
609: mechanism is insufficient \citep{ss00}. Unfortunately, the true chemical state
610: and composition of interstellar gas is difficult to assess, so exact values
611: of $\gamma$ are difficult to estimate observationally.}
612:
613: The possibility that isolated, massive stars form when $\gamma > 1$ is
614: of interest, not only because ionizing radiation, stellar winds, and
615: supernovae from massive stars play an important role in the
616: interstellar medium, but also because massive stars are rarely
617: observed in isolation. Rather, they are usually found as members of
618: rich stellar clusters. Recently, however, \citet{lamers02} has
619: reported observations of massive stars found in isolation or
620: associated only with very small groups of lower-mass stars in the
621: bulge of M51. \citet{mas02} also reported finding apparently isolated,
622: massive, field stars in both the Large and Small Magellanic Clouds.
623: Our simulations suggest that molecular cloud regions forming isolated,
624: high-mass stars have $\gamma > 1$. In this case turbulent cloud
625: fragmentation tends to produce isolated, massive collapsing cores that
626: will feed onto isolated high-mass stars. \citet{ys02}
627: demonstrated that high-mass stars can form in collapsing gas clumps
628: via disk accretion very much the same way as low mass stars are
629: believed to do. There is thus no need for a dense cluster environment
630: where high-mass stars could build up by collisions of lower-mass stars
631: (as suggested by \citealt{bbz98}).
632:
633: A theory of galactic-scale star formation partially based on the
634: universality of log-normal density PDFs has been proposed by Elmegreen
635: (2002). Our results do show a log-normal form at least for the peak of
636: the PDFs, containing most of the gas. However, the dispersion of the
637: PDF varies substantially as the EOS changes
638: (Fig.~\ref{fig_mom_k1278}), so that the fraction of gas available for
639: star formation would depend significantly on the EOS in such a theory.
640: This raises the question of how it could still be able to explain a
641: global Schmidt Law across many different galaxies
642: with different metallicities, radiation fields, interstellar
643: pressures, and so forth.
644:
645: High resolution simulations by \citet{abn00, abn02} of the
646: formation of the first star suggest that fragmentation during the
647: collapse of metal-free pregalactic halos is rather inefficient,
648: resulting in the formation of single, massive stars rather than
649: clusters of lower-mass objects. In the absence of metals, inefficient
650: cooling may result in high $\gamma$, perhaps helping to explain these
651: results.
652:
653: We do note, however, that simulations by \citet{bcl99, bcl02} show greater
654: multiplicity. They use similar chemistry,
655: but examine more massive, more isolated halos, with substantially
656: larger limiting mass resolution. The EOS will have less effect on
657: these larger scales, which are more dominated by gravity. If the
658: metal-free gas indeed shows a high effective $\gamma$, our models
659: would suggest that their collapsing regions will show little further
660: fragmentation if followed down to stellar mass scales.
661:
662:
663:
664: \acknowledgments We thank M. Fall, H. Lamers, H. Zinnecker, \new{and
665: S. Glover} for valuable discussions, and D. Janies and W. Wheeler for their
666: work on the AMNH Parallel Computing Facility, which we used for the
667: computations presented here. \new{We are grateful to the anonymous referee for
668: useful comments and corrections}. YL thanks the AIP for its warm
669: hospitality, and the Kade Foundation for support of her visits there.
670: M-MML acknowledges partial support by NASA ATP grant NAG5-10103, and
671: by NSF CAREER grant AST99-85392. RSK acknowledges support by the Emmy
672: Noether Program of the Deutsche Forschungsgemeinschaft (DFG,
673: KL1358/1).
674:
675: \clearpage
676:
677: \begin{thebibliography}{JUNK}
678:
679: \bibitem[Abel, Bryan, \& Norman(2000)]{abn00} Abel, T, Bryan, G., \& Norman,
680: M. L. 2000, ApJ, 540, 39
681: \bibitem[Abel, Bryan, \& Norman(2002)]{abn02} Abel, T, Bryan, G., \& Norman,
682: M. L. 2002, Science, 295, 93
683: \bibitem[Bate, Bonnell, \& Price(1995)]{bbp95} Bate, M. R., Bonnell, I. A., \&
684: Price, N. M. 1995, MNRAS, 277, 362
685: \bibitem[Bate \& Burkert(1997)]{bab97} Bate, M. R., \& Burkert, A. 1997,
686: MNRAS, 288, 1060
687: \bibitem[Benz(1990)]{benz90} Benz, W. 1990, in \textit{The Numerical Modeling
688: of Nonlinear Stellar Pulsations}, ed. J. R. Buchler (Dordrecht: Kluwer), 269
689: \bibitem[Bonnell \& Bastien(1991)]{bob91} Bonnell, I. \& Bastien, P. 1991,
690: ApJ, 374, 610
691: \bibitem[Bonnell \& Bastien(1993)]{bob93} Bonnell, I. \& Bastien, P. 1993,
692: ApJ, 406, 614
693: \bibitem[Bonnell, Bate, \& Zinnecker(1998)]{bbz98} Bonnell, I. A., Bate,
694: M. R., \& Zinnecker, H. 1998, MNRAS, 298, 93
695: \bibitem[Bromm, Coppi, \& Larson(1999)]{bcl99} Bromm, V., Coppi, P. S., \&
696: Larson, R. B. 1999, ApJ, 527, L5
697: \bibitem[Bromm, Coppi, \& Larson(2002)]{bcl02} Bromm, V., Coppi, P. S., \&
698: Larson, R. B. 2002, ApJ, 564, 23
699: \bibitem[Elmegreen(1993)]{elm93} Elmegreen, B. G. 1993, ApJ, 419, L29
700: \bibitem[Elmegreen(2002)]{elm02} Elmegreen, B. G. 2002, ApJ, 577, 206
701: \bibitem[Galli et al.(2001)]{galli01} Galli, D., Shu, F. H., Laughlin, G., \&
702: Lizano, S. 2001, ApJ, 551, 367
703: \bibitem[Hartmann(1998)]{hart98} Hartmann, L. 1998, \textit{Accretion
704: Processes in Star Formation} (Cambridge: Cambridge Univ. Press), 33
705: \bibitem[Heitsch, Mac Low, \& Klessen(2001)]{hmk01} Heitsch, F., Mac Low,
706: M.-M. \& Klessen, R. S. 2001, ApJ, 547, 280 (Paper III)
707: \bibitem[Jeans(1902)]{jeans02} Jeans, J. H. 1902, Phil. Trans.A., 199, 1
708: \bibitem[Klessen(1997)]{rsk97} Klessen, R. S. 1997, MNRAS, 292, 11
709: \bibitem[Klessen(2000)]{rsk00} Klessen, R. S. 2000, ApJ, 535, 869
710: \bibitem[Klessen(2001)]{rsk01} Klessen, R. S. 2001, ApJ, 556, 837 (Paper IV)
711: \bibitem[Klessen \& Burkert(2000)]{kb00} Klessen, R. S., \& Burkert,
712: A. 2000, ApJS, 28, 287 (Paper I)
713: \bibitem[Klessen, Burkert, \& Bate(1998)]{kbb98} Klessen, R. S., Burkert, A.,
714: \& Bate, M. R. 1998, ApJ, 501, L205
715: \bibitem[Klessen, Heitsch, \& Mac Low(2000)]{khm00} Klessen, R. S., Heitsch,
716: F., \& Mac Low, M.-M. 2000, ApJ, 535, 887 (Paper II)
717: \bibitem[Kroupa(2002)]{kroupa02} Kroupa, P. 2002, Science, 295, 82
718: \bibitem[Lamers et al.(2002)]{lamers02} Lamers, H. J. G. L. M. et al. 2002,
719: ApJ, 566, 818
720: \bibitem[Low \& Lynden-Bell(1976)]{low76} Low, C., \& Lynden-Bell, D. 1976,
721: MNRAS, 176, 367
722: \bibitem[Mac Low et al.(1998)]{mm98} Mac Low, M.-M., Klessen, R. S., Burkert,
723: A., \& Smith, M. D. 1998, Phys.\ Rev.\ Lett., 80, 2754
724: \bibitem[Mac Low(1999)]{mm99} Mac Low, M.-M. 1999, ApJ, 524, 169
725: \bibitem[ Mac Low \& Klessen(2003)]{mk03} Mac Low, M.-M., \& Klessen,
726: R. S. 2003, Rev.\ Mod.\ Phys., submitted (astro-ph/0301093)
727: \bibitem[Massey(2002)]{mas02} Massey, P. 2002, ApJS, 141, 81
728: \bibitem[Ostriker, Stone, \& Gammie(2001)]{osg01} Ostriker, E. C., Stone,
729: J. M., \& Gammie, C. F. 2001, ApJ 546, 980
730: \bibitem[Padoan \& Nordlund(1999)]{pn99} Padoan, P., \& Nordlund, A. 1999,
731: ApJ, 526, 279
732: \bibitem[Passot \& V\'azquez-Semadeni(1998)]{pv98} Passot, T., \&
733: V\'azquez-Semadeni, E. 1998, Phys.\ Rev.\ E, 58, 4501
734: \bibitem[Pudritz(2002)]{pud02} Pudritz, R., 2002, Science, 295, 68
735: \bibitem[Rouleau \& Bastien(1990)]{rb90} Rouleau, F. \& Bastien, P., 1990,
736: ApJ, 355, 172
737: \bibitem[Salpeter(1955)]{salp55} Salpeter, E. E. 1955, ApJ, 121, 161
738: \bibitem[Scalo(1998)]{scalo98} Scalo, J. 1998, in ASP Conf. Ser. 142,
739: \textit{The Stellar Initial Mass Function} (38th Herstmonceux Conference),
740: ed. G. Gilmore \& D. Howell (San Francisco: ASP), 201
741: \bibitem[Smith, Mac Low, \& Heitsch(2000)]{smh00} Smith, M.~D., Mac Low,
742: M.-M., \& Heitsch, F.\ 2000, \aap, 362, 333
743: \bibitem[Spaans \& Silk(2000)]{ss00} Spaans, M., \& Silk, J. 2000, ApJ, 538,
744: 115
745: \bibitem[V\'azquez-Semadeni \& Garcia(2001)]{vg01} V\'azquez-Semadeni, E. \&
746: Garcia, N. 2001, ApJ, 557, 727
747: \bibitem[V\'azquez-Semadeni, Ballesteros-Paredes, \& Klessen (2003)]{vbk03}
748: V\'azquez-Semadeni, E., Ballesteros-Paredes, J., \& Klessen, R.\ S. 2003,
749: ApJ, in press (astro-ph/0301546)
750: \bibitem[Ward-Thompson(2002)]{ward02} Ward-Thompson, D. 2002, Science, 295, 76
751: \bibitem[Williams, De Geus, \& Blitz(1994)]{wdb94} Williams, J. P., De Geus,
752: E. J., \& Blitz, L. 1994, ApJ, 428, 693
753: \bibitem[Williams, Blitz, \& McKee(2000)]{wbm00} Williams, J. P., Blitz, L.,
754: \& McKee, C. F. 2000, in \textit{Protostars and Planets IV},
755: ed. V. Mannings, A. P. Boss, \& S. S. Russell (Tucson: Univ. Arizona Press),
756: 97
757: \bibitem[Yorke \& Sonnhalter(2002)]{ys02} Yorke, H.~W.~\& Sonnhalter, C.\ 2002,
758: ApJ, 569, 846
759:
760: \end{thebibliography}
761:
762: \clearpage
763:
764: \begin{figure}
765: \begin{center}
766: \includegraphics[height=1.5in]{f1a.eps}
767: \hspace{0.1cm}
768: \includegraphics[height=1.5in]{f1b.eps}
769: \hspace{0.1cm}
770: \includegraphics[height=1.5in]{f1c.eps}\\
771: \vspace{-0.5cm}
772: \includegraphics[height=1.5in]{f1d.eps}
773: \hspace{0.1cm}
774: \includegraphics[height=1.5in]{f1e.eps}
775: \hspace{0.1cm}
776: \includegraphics[height=1.5in]{f1f.eps}\\
777: \includegraphics[height=1.5in]{f1g.eps}
778: \hspace{0.1cm}
779: \includegraphics[height=1.5in]{f1h.eps}
780: \hspace{0.1cm}
781: \includegraphics[height=1.5in]{f1i.eps}\\
782: \includegraphics[height=1.5in]{f1j.eps}
783: \hspace{0.1cm}
784: \includegraphics[height=1.5in]{f1k.eps}
785: \hspace{0.1cm}
786: \includegraphics[height=1.5in]{f1l.eps}\\
787: \vspace{0.4cm}
788: \includegraphics[width=3.9in]{f1m.eps}
789: \caption{\label{fig_sinkdis1}Density distribution of the gas shown on
790: a grid in three-dimensional projection (\textit{top row}) and in a
791: slice through maximum density, for driving with $k=1$--2 at 1 $\tau_{ff}$
792: after self-gravity is turned on. Values of $\gamma = 0.2$ (\textit{left
793: column}), $\gamma = 1.0$ (\textit{middle column}), and $\gamma = 1.3$
794: (\textit{right column}) are shown. \new{The color bar for the density scale \new{of the
795: slices} is at the bottom. Note the very high densities in these images apply
796: to the collapsed cores only. The distribution of gas density is shown in Fig.
797: \ref{fig_pdf_int}.}}
798: \end{center}
799: \end{figure}
800:
801: \clearpage
802:
803: \begin{figure}
804: \begin{center}
805: \includegraphics[height=1.5in]{f2a.eps}
806: \hspace{0.2cm}
807: \includegraphics[height=1.5in]{f2b.eps}\\
808: \vspace{-0.5cm}
809: \includegraphics[height=1.5in]{f2c.eps}
810: \hspace{0.2cm}
811: \includegraphics[height=1.5in]{f2d.eps}\\
812: \includegraphics[height=1.5in]{f2e.eps}
813: \hspace{0.2cm}
814: \includegraphics[height=1.5in]{f2f.eps}\\
815: \includegraphics[height=1.5in]{f2g.eps}
816: \hspace{0.2cm}
817: \includegraphics[height=1.5in]{f2h.eps}\\
818: \vspace{0.4cm}
819: \includegraphics[width=2.8in]{f2i.eps}
820: \caption{\label{fig_sinkdis2}Density distribution of the gas shown on
821: a grid in three-dimensional projection (\textit{top row}) and in a
822: slice through maximum density, for selected $\gamma$ and for driving
823: with $k=7$--8 at 1 $\tau_{ff}$ after self-gravity is turned
824: on. Values of $\gamma = 0.2$ (\textit{left column}), and $\gamma = 1.0$
825: (\textit{right column}) are shown. The color bar for the density scale of the
826: slices is at the bottom. \new{Note the very high densities in these images apply
827: to the collapsed cores only}.}
828: \end{center}
829: \end{figure}
830:
831: \clearpage
832:
833: \begin{figure}
834: \epsscale{1.0}\plottwo{f3a.eps}{f3b.eps}
835: \caption{\label{fig_sinknum}Comparison of number of collapsed cores
836: for models with different \new{polytropic exponent} $\gamma$, for driving with
837: wavenumber $k=1$--2 (\textit{left panel}) and $k=7$--8 (\textit{right
838: panel}). Note gravity was ``turned on'' at t = 2.0 $\simeq 1.33\tau_{ff}$. The
839: simulations of low-$\gamma$ cases (0.2 and 0.7 in this plot) of model $k=1$--2
840: were terminated after a few $\tau_{ff}$ due to prohibitively small time step.}
841: \end{figure}
842:
843: \clearpage
844:
845: \begin{figure}
846: \plotone{f4.eps}
847: \caption{\label{fig_sinkgamma}Relation of the value of $\gamma$ to the
848: number of collapsed cores one free-fall time after self-gravity
849: is turned on.}
850: \end{figure}
851:
852: \clearpage
853:
854: \begin{figure}
855: \plottwo{f5a.eps}{f5b.eps}
856: \caption{\label{fig_sinkacc}Comparison of accretion rate of collapsing
857: cores for turbulence driven with wavenumber $k=1$--2 (\textit{left
858: panel}) and $k=7$--8 (\textit{right panel}). Note the different scales
859: in the two plots. Gravity is turned on at t = 2.0 $\simeq 1.33\tau_{ff}$. The
860: simulations of low-$\gamma$ cases (0.2 and 0.7 in this plot) of the model
861: driven with $k=1$--2 were terminated after a few $\tau_{ff}$ due to the
862: prohibitively small time step.}
863: \end{figure}
864:
865: \clearpage
866:
867: \begin{figure}
868: \begin{center}
869: \includegraphics[height=1.5in]{f6a.eps}
870: \includegraphics[height=1.5in]{f6b.eps}
871: \includegraphics[height=1.5in]{f6c.eps}\\
872: \includegraphics[height=1.5in]{f6d.eps}
873: \includegraphics[height=1.5in]{f6e.eps}
874: \includegraphics[height=1.5in]{f6f.eps}\\
875: \includegraphics[height=1.5in]{f6g.eps}
876: \includegraphics[height=1.5in]{f6h.eps}
877: \includegraphics[height=1.5in]{f6i.eps}\\
878: \includegraphics[height=1.5in]{f6j.eps}
879: \includegraphics[height=1.5in]{f6k.eps}
880: \includegraphics[height=1.5in]{f6l.eps}\\
881: \includegraphics[height=1.5in]{f6m.eps}
882: \includegraphics[height=1.5in]{f6n.eps}
883: \includegraphics[height=1.5in]{f6o.eps}
884: \caption{\label{fig_mas}Mass spectra of gas clumps (\textit{open
885: histogram}), and of collapsed cores (\textit{filled histogram}) for
886: models with different $\gamma$ and driving with $k=1$--2. The fraction
887: of mass accreted into collapsed cores (sink particles) is given for
888: the right two columns. The vertical line shows the SPH resolution
889: limit. Also shown are two power-law spectra with $\nu = -1.5$
890: (dashed-line) and $\nu = -2.33$ (dotted line).}
891: \end{center}
892: \end{figure}
893:
894: \clearpage
895:
896: \begin{figure}
897: \begin{center}
898: \includegraphics[height=2.0in]{f7a.eps}
899: \includegraphics[height=2.0in]{f7b.eps}\\
900: \includegraphics[height=2.0in]{f7c.eps}
901: \includegraphics[height=2.0in]{f7d.eps}\\
902: \includegraphics[height=2.0in]{f7e.eps}
903: \includegraphics[height=2.0in]{f7f.eps}
904: \caption{\label{fig_pdf_int}Mass-weighted (\textit{thick solid-line})
905: and volume-weighted (\textit{thick dashed-line}) gas density PDFs
906: $p_m$ and $p_v$ of fully-developed turbulence prior to turning on
907: self-gravity. Models with driving wavenumber $k=1-2$ (\textit{left
908: column}) and $k=7-8$ (\textit{right column}) are shown. Gaussian fits
909: are shown with \textit{thin dotted} and \textit{thin dot-dashed} lines
910: correspondingly, and the errors $\epsilon_M$ and $\epsilon_V$
911: summed over the parts of the curves above $10\%$ of the peak values,
912: are quoted. Note that $s=\ln (\rho/\rho_0)$.}
913: \end{center}
914: \end{figure}
915:
916: \clearpage
917:
918: \begin{figure}
919: \plottwo{f8a.eps}{f8b.eps}
920: \caption{\label{fig_mom_k1278} First moment (mean) $\mu$
921: (\textit{top}) and second moment (variance) $\sigma^2$
922: (\textit{bottom}), of both $p_m$ (\textit{filled}) and $p_v$
923: (\textit{open}) as functions of $\gamma$, for models driven with
924: wavenumbers $k=1-2$ (\textit{circle}) and $k=7-8$ (\textit{square}),
925: where $s=\ln (\rho/\rho_0)$.}
926: \end{figure}
927:
928: \clearpage
929:
930: \begin{figure}
931: \begin{center}
932: \includegraphics[height=2.0in]{f9a.eps}
933: \includegraphics[height=2.0in]{f9b.eps}\\
934: \includegraphics[height=2.0in]{f9c.eps}
935: \includegraphics[height=2.0in]{f9d.eps}\\
936: \includegraphics[height=2.0in]{f9e.eps}
937: \includegraphics[height=2.0in]{f9f.eps}
938: \caption{\label{fig_pdf_3phases}Mass-weighted gas density PDFs $p_m$
939: of models driven with $k=1-2$ for selected $\gamma$. Only gas not
940: accreted into sink particles is included. Three evolutionary phases are
941: shown: the initial state (\textit{thick solid-line}), $M_{\ast}
942: \approx 20\%$ (\textit{thick dashed-line}), and $M_{\ast} \approx
943: 40\%$ (\textit{thick dot-dashed line}). The thin lines (solid, dashed,
944: dot-dashed) are Gaussian fits to the corresponding PDFs above 10\% of
945: peak value. Again $s=\ln (\rho/\rho_0)$. Note the change of the
946: high-density tails with $\gamma$.}
947: \end{center}
948: \end{figure}
949:
950: \clearpage
951:
952: \begin{figure}
953: \begin{center}
954: \includegraphics[height=2in]{f10a.eps}
955: \includegraphics[height=2in]{f10b.eps}\\
956: \includegraphics[height=2in]{f10c.eps}
957: \includegraphics[height=2in]{f10d.eps}
958: \caption{\label{fig_mom}
959: The first four moments of three evolutionary phases of the
960: mass-weighted density PDF (corresponding to Figure
961: \ref{fig_pdf_3phases})of the model driven with $k=1-2$, as functions
962: of $\gamma$. , Shown are the mean $\mu$ (top left), variance
963: $\sigma^2$ (top right), the skewness $S$ (bottom left), and the
964: kurtosis $\kappa$, of $s=\ln (\rho/\rho_0)$. We define the
965: fourth moment $\kappa$ with a value of 3 subtracted, so that for a
966: Gaussian $\kappa =0$, while an exponential has $\kappa =3$.}
967: \end{center}
968: \end{figure}
969:
970:
971: \end{document}
972:
973:
974:
975:
976:
977:
978: