1: %\documentclass[preprint2]{aastex}
2: %\documentclass[12pt,preprint]{aastex}
3: \documentclass{emulateapj}
4:
5: %\usepackage{amssymb}
6: %\usepackage{amsmath}
7: %%\usepackage{graphicx}
8:
9: \newcommand{\lO} {\lambda_{\Omega}}
10: \newcommand{\led} {\lambda_{e}}
11: \newcommand{\cAi} {c_{Ai}}
12: \newcommand{\cAn} {c_{An}}
13: \newcommand{\RM} {R_{M}}
14: \newcommand{\mbfB}{\mathbf{B}}
15: \newcommand{\mbfu}{\mathbf{u}}
16: \newcommand{\mbfui}{\mathbf{u}_i}
17: \newcommand{\mbfun}{\mathbf{u}_n}
18: \newcommand{\mbfz}{\mathbf{z}}
19: \newcommand{\mbfn}{\mathbf{n}}
20: \newcommand{\mbfk}{\mathbf{k}}
21: \newcommand{\mbfx}{\mathbf{x}}
22: \newcommand{\mbfv}{\mathbf{v}}
23: \newcommand{\mbfnabla}{\mathbf{\nabla}}
24: \newcommand{\f} {\frac}
25: \newcommand{\ddx}{\partial_x}
26: \newcommand{\ddy}{\partial_y}
27: \newcommand{\ddt}{\partial_t}
28: \newcommand{\lomega}{\lambda_\Omega}
29:
30: \begin{document}
31:
32: \title{Cooling, Gravity and Geometry: Flow-driven Massive Core Formation}
33:
34: \author{Fabian Heitsch\altaffilmark{1}}
35: \author{Lee W. Hartmann\altaffilmark{1}}
36: \author{Adrianne D. Slyz\altaffilmark{2}}
37: \author{Julien E.G. Devriendt\altaffilmark{3}}
38: \author{Andreas Burkert\altaffilmark{4}}
39: \altaffiltext{1}{Dept. of Astronomy, University of Michigan, 500 Church St.,
40: Ann Arbor, MI 48109-1042, U.S.A}
41: \altaffiltext{2}{Oxford University, Astrophysics, Denys Wilkinson Building, Keble Road,
42: Oxford, OX1 3RH, United Kingdom}
43: \altaffiltext{3}{Universit\'e Claude Bernard Lyon 1,
44: CRAL, Observatoire de Lyon, 9 Avenue Charles Andr\'{e},
45: 69561 St-Genis Laval Cedex, France; CNRS, UMR 5574; ENS Lyon}
46: \altaffiltext{4}{Universit\"ats-Sternwarte M\"unchen, Scheinerstr. 1, 81679 M\"unchen, Germany}
47: \lefthead{Heitsch et al.}
48: \righthead{Flow-driven Core Formation}
49:
50: \begin{abstract}
51: We study numerically the formation of molecular clouds in large-scale colliding flows
52: including self-gravity. The models emphasize the competition between the effects of gravity
53: on global and local scales in an isolated cloud. Global gravity builds up large-scale
54: filaments, while local gravity -- triggered by a combination of
55: strong thermal and dynamical instabilities -- causes cores to form.
56: The dynamical instabilities give rise to a local focusing of the
57: colliding flows, facilitating the rapid formation of massive protostellar
58: cores of a few $100$~M$_\odot$.
59: The forming clouds do not reach an equilibrium state, though the motions within
60: the clouds appear comparable to ``virial''. The self-similar core mass
61: distributions derived from models with and without self-gravity indicate
62: that the core mass distribution is set very early on during the cloud
63: formation process, predominantly by a combination of thermal and dynamical
64: instabilities rather than by self-gravity.
65: \end{abstract}
66: \keywords{instabilities --- gravity --- turbulence --- methods:numerical
67: --- stars:formation --- ISM:clouds}
68:
69: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
70: %
71: %\section{Motivation}
72: %
73: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
74: \section{Rapid Star Formation}\label{s:introduction}
75:
76: There is increasing evidence that star formation in the solar neighborhood
77: follows rapidly upon molecular cloud formation
78: (\citealp{2001ApJ...562..852H}; \citealp{2007RMxAA..43..123B} and references therein).
79: This evidence suggests that the density enhancements in which stars form are produced
80: during the cloud formation phase; thus understanding cloud formation
81: is essential to understanding star formation. Moreover, it appears that non-linear
82: density perturbations need to arise quite early in cloud formation,
83: as massive, finite molecular clouds are highly susceptible to large-scale
84: gravitational collapse which could overwhelm small, stellar-mass fragmentation
85: \citep{2004ApJ...616..288B}. While several investigations have adopted various
86: assumed forms of initial and/or driven turbulent motions to produce the necessary
87: small-scale structure (\citealp{2000ApJS..128..287K}; \citealp{2002ApJ...576..870P};
88: \citealp{2002MNRAS.332L..65B,2003MNRAS.339..577B}),
89: it is preferable to have these structures arise naturally.
90: Thus, a close look at instabilities in cloud
91: formation which could lead to strong density fluctuations is needed.
92:
93: \citet{1999ApJ...527..285B} and \citet{2001ApJ...562..852H}
94: proposed that cloud formation as the result
95: of pileup of material by large-scale flows is an essential mechanism for explaining
96: the ``crossing time problem'', i.e. the observation that the typical age spreads in
97: the stellar populations of many large star-forming regions are often substantially smaller
98: than the lateral crossing timescales; in the large-scale flow picture, no information
99: is transmitted laterally, i.e. perpendicular to the large-scale flow.
100: This picture works only if star formation follows closely
101: upon molecular gas formation. Expanding H II regions,
102: supernova bubbles, and spiral density waves are all obvious candidates for large-scale
103: supersonic flows which can sweep up material, and there is considerable direct observational
104: evidence for rapid star formation in these environments (\citealp{2001ApJ...562..852H} and
105: references therein). Thus a plausible place to look for stellar core-forming instabilities
106: is in the post-shock material of the large-scale flows.
107:
108: Several numerical studies relevant to this problem have now been undertaken
109: (see the discussions of the literature in \citealp{2006ApJ...648.1052H}
110: and \citealp{2007ApJ...657..870V}). These calculations typically assume
111: converging flows to keep the shocked gas within the
112: computational volume, but this can easily be extended to describe a more
113: generic situation by recasting the problem in the rest frame of the shock(s).
114: These models overcome the limitations of previous turbulent fragmentation models
115: (see review by \citealp{2004RvMP...76..125M})
116: by avoiding ad hoc assumptions about the source of turbulence and boundary conditions.
117: Indeed, the converging flow models demonstrate with ease that
118: flows provide a natural mechanism for the generation of structure and turbulence in
119: clouds (\citealp{2005A&A...433....1A}; \citealp{2005ApJ...633L.113H}; \citealp{2006ApJ...643..245V};
120: \citealp{2006ApJ...648.1052H}; \citealp{2007A&A...465..445H}; \citealp{2007A&A...465..431H}).
121:
122: Although this study is motivated by the scenario of molecular clouds being
123: transient entities \citep{1999ApJ...527..285B}, forming and dispersing in
124: background flows within a few free-fall times, the results presented here
125: are not restricted to this scenario.
126: Colliding flows can appear even in large scale gravitational instabilities,
127: linking our models to the alternative scenario of ``Giant Molecular Clouds''
128: living for substantially longer than only a few free-fall times. The issue
129: of cloud lifetimes is currently a matter of debate (e.g.
130: \citealp{2007ApJ...654..304K}; \citealp{2007arXiv0707.2252E},
131: \citealp{2007arXiv0707.3514M}), and seems to depend strongly on the
132: galactic environment \citep{2001ApJ...562..852H,2006ApJ...648.1052H}. For
133: reasons discussed in \S\ref{s:answers}, our models cannot predict cloud life
134: times. Thus, the emphasis of this study is on the {\em onset} of star formation.
135:
136: The purpose of this study is to compare the fragmentation processes in simulated converging
137: flows with and without self-gravity in order to show that the rapid
138: onset of star formation is pretty much unavoidable within the scenario where molecular
139: clouds form in converging flows.
140: Recently, \citet{2007ApJ...657..870V}
141: presented a study of star formation in clouds formed by colliding flows, emphasizing the long-term
142: evolution of the system. Here we focus more on the initial development of the cloud,
143: discussing the consequences of cooling, cloud geometry, and gravity
144: for the star formation process.
145:
146: Cooling and gravity both are fragmentation agents, with the difference that gravity can be
147: relevant on all scales which surpass the Jeans length, while the isobaric condensation mode of
148: the thermal instability \citep{1965ApJ...142..531F} is limited to scales set by the sound speed
149: and the cooling time, $\lambda_c = c_s\,\tau_c$
150: (e.g. \citealp{2000ApJ...537..270B}; \citealp{2007A&A...465..431H}). Thus, in the early stages
151: of cloud formation, when only a little mass has accumulated, the thermal instability is the
152: dominant fragmentation agent.
153:
154: {\em Non-linear} density perturbations collapse at a higher
155: rate than the global cloud \citep{2004ApJ...616..288B},
156: thus allowing stars to form locally before the whole-sale collapse of the
157: cloud. We find that the strong dynamical and thermal instabilities
158: generate non-linear density perturbations for rapid {\em local} collapse, while still
159: allowing for the global build-up of the cloud.
160:
161: The physics and methods are summarized in \S\ref{s:physmeth}, followed by the
162: model results in \S\ref{s:results}. Readers solely interested in the discussion
163: of the results and their consequences should directly proceed to
164: \S\ref{s:discussion} and \S\ref{s:answers}.
165:
166: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
167: %
168: %\section{The Physics and their Numerical Realization}
169: %
170: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
171:
172: \section{Physics and Methods}\label{s:physmeth}
173:
174: Our study focuses on the effects of global versus local gravity on the one hand, and
175: on the rapid generation of substructure in the colliding flows on the other.
176: Since we are interested in global gravitational effects, we cannot use the
177: periodic boundary conditions of earlier turbulent fragmentation studies (e.g.
178: \citealp{2000ApJ...535..887K}; \citealp{2001ApJ...547..280H};
179: \citealp{2001ApJ...553..227P}; \citealp{2003ApJ...592..203G}; \citealp{2005ApJ...618..344V}).
180: Neither will we generate or drive turbulence by imposing a randomly chosen velocity or density field,
181: but instead, we will rely on turbulence generated by the dynamical instabilities arising
182: from the collisions of the flows.
183:
184: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
185: %\section{The Models}
186: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
187:
188: \subsection{The Models}\label{ss:models}
189: We ran four models, whose parameters are listed in Table~\ref{t:modparam}.
190: All models are run on a fixed grid with the instreaming gas flowing along the
191: $x$-direction, entering the domain at the $(y,z)$-planes. To trigger
192: the fragmentation of the (otherwise plane-parallel) interaction region,
193: we perturb the collision interface. We chose the
194: perturbations of the collision interface from a random distribution of
195: amplitudes in Fourier space with a top hat distribution restricted
196: between wave numbers $k=1..4$.
197:
198: Model Gs (for ``gravity in shell'') can be interpreted as
199: colliding continuous gas streams in spiral shocks (e.g. \citealp{1990imeg.conf..298T}
200: for observational evidence, and \citealp{2007MNRAS.376.1747D} for numerical models),
201: or as a close-up view of two expanding and colliding super-shells in the LMC.
202: The collision interface is plane-parallel, except for the imposed perturbations.
203: Material is free to leave the box in the lateral (i.e. perpendicular to
204: the inflow) directions.
205:
206: In models Gf1 and Gf2 (for ``gravity in finite cloud''), we restrict the
207: inflow to a cylinder of elliptical cross section with an ellipticity of $3.3$
208: and a major axis of $80$\% of the (transverse) box size,
209: mimicking two colliding gas streams in a more general geometry. Again, the
210: collision interface is perturbed. The motivation here is to generate one
211: finite cloud in order to study global gravitational effects.
212:
213: Finally, model Hf1 is a non-gravitating version of Gf1, to compare the role
214: of gravity versus that of the thermal instability for the fragmentation of the
215: gas streams.
216:
217: The inflow density in all models is $n_0=3$~cm$^{-3}$ at a temperature of
218: $T_0=1800$~K and an inflow velocity of $7.9$~km~s$^{-1}$, corresponding to
219: a Mach number of ${\cal M} = 1.5$. The flows are initially in thermal
220: equilibrium. The models start at time $t=0$ with the collision of the two
221: flows. For models with spatially constrained inflows (Hf1, Gf1, Gf2), the
222: fluid is at rest everywhere except in the colliding cylinders.
223:
224:
225: \begin{deluxetable}{c|ccccc}
226: \tablewidth{0pt}
227: \tablecaption{Model Parameters\label{t:modparam}}
228: \tablehead{\colhead{Name}&\colhead{$n_xn_yn_z$}
229: &\colhead{$L_xL_yL_z$ [pc]}
230: &\colhead{gravity}
231: &\colhead{$t_{end}$ [Myr]}
232: &\colhead{$\eta$ [pc]}}
233: \startdata
234: Hf1 & $256\times 512^2$ & $22\times 44^2$&no & 14.5 & $2.2$ \\
235: Gf1 & $256\times 512^2$ & $22\times 44^2$&yes & 14.5 & $2.2$ \\
236: Gf2 & $256\times 512^2$ & $22\times 44^2$&yes & 14.5 & $4.4$ \\
237: Gs & $256^3$ & $44^3$&yes & 16.0 & $2.2$
238: \enddata
239: \tablecomments{1st column: Model name. 2nd column: resolution. 3rd column:
240: physical grid size. 4th column: gravity. 5th column: end time of run.
241: 6th column: amplitude of interface displacement.}
242: \end{deluxetable}
243:
244:
245: The finite cloud models Gf1, Gf2 and Hf1 have
246: a grid cell size of $\Delta L=8.6\times 10^{-2}$~pc, while model Gs has one of
247: $\Delta L=1.7\times 10^{-1}$~pc (Tab.~\ref{t:modparam}). We note that this
248: does {\em not} constitute the physical resolution power of the simulation.
249: At minimum, the stencils (i.e. the support points) used for the higher-order
250: reconstruction of the fluid states at the cell wall will render the cells within one stencil
251: not independent. In other words, conclusions should not be drawn from structures of $4$
252: cells or less of linear size. Thus, we use only cores with $64$ or more cells for analysis.
253:
254: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
255: %\section{Boundary Conditions}
256: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
257: \subsection{Boundary Conditions}\label{ss:boundcond}
258: The $x$-boundaries are partly or entirely defined as inflow-boundaries,
259: depending on the model.
260: Indeed, the inflow is either defined over the whole
261: $(y,z)$-plane (model Gs), or within an elliptical surface (models Hf1, Gf1, Gf2:
262: see \S\ref{ss:models} for details).
263: The $y$ and $z$ boundaries, -- as well as the part of the $x$-boundaries
264: that is not occupied by the inflow in models Hf1, Gf1 and Gf2 --, are open,
265: meaning material is free to leave
266: the simulation domain through these boundaries.
267: This is bound to cause trouble once material tries
268: to ``come back'' during the later stages of the simulation (note that
269: this material is not actually coming back, but that it is the result of
270: the extrapolation of the last active cells properties, i.e. material
271: with the properties of the last active cell layer within the domain will try
272: to enter the domain).
273: This inevitably will happen once global gravity dominates over the
274: over-pressurized material shooting out of the flow-collision region.
275: However, the ``re-entering'' material does not reach the central
276: cloud region within the simulation time, and in any case contributes only
277: a negligible amount to the total mass within the box.
278:
279: The situation becomes more critical once material is leaving the simulation domain
280: in the $x$-direction, i.e. once it is moving against the inflow. Since the bounding
281: shocks (and the cooling) will set the density and the temperature of the post-shock
282: gas, the physical state of the gas will be undefined once the bounding shocks move
283: off the grid. This will render the ``returning'' material essentially in a
284: hydrodynamically inconsistent state. Thus, once material
285: encounters the $x$-boundaries, the simulation needs to be stopped.
286:
287: In model Gs, the inflow velocities are slightly reduced at the edges of the domain,
288: mimicking the velocity profile of an expanding shell of material driven by two
289: sources at a distance of approximately $100$ pc to the left and to the right of
290: the mid plane of the simulation domain. This reduces the amount of material collected
291: at the edges of the domain, and thus limits the edge effects due to gravity at
292: later stages of the simulation.
293:
294: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
295: %\section{Hydrodynamics}
296: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
297: \subsection{Hydrodynamics and Atomic Line Coolants}\label{ss:hydrocool}
298: As in our previous studies of colliding flows, we used the
299: higher-order gas-kinetic grid method Proteus
300: (\citealp{1993JCoPh.109...53P}; \citealp{1999A&AS..139..199S};
301: \citealp{2004ApJ...603..165H}; \citealp{2005MNRAS.356..737S};
302: \citealp{2006ApJ...648.1052H}, \citeyear{2007ApJ...665..445H}), allowing
303: full control of viscosity and heat conduction.
304: The code evolves the
305: Navier-Stokes equations in their conservative form to second order in time and
306: space. The hydrodynamical quantities are updated in time unsplit form.
307:
308: The heating and cooling rates are restricted to optically thin
309: atomic lines following \citet{1995ApJ...443..152W}.
310: Dust extinction becomes important above
311: column densities of $N(\mbox{HI})\approx 1.2\times 10^{21}$cm$^{-2}$, which are
312: only reached in the densest regions modeled. Thus, we use the unattenuated
313: UV radiation field for grain heating \citep{1995ApJ...443..152W},
314: expecting substantial uncertainties in cooling rates only for the densest regions.
315: The ionization degree is derived from a balance between ionization by cosmic rays and
316: recombination, assuming that Ly $\alpha$ photons are directly reabsorbed.
317: Numerically, heating and cooling is implemented iteratively as a source
318: term for the internal energy $e$ of the form
319: \begin{equation}
320: \ddt e = n\Gamma(T) - n^2\Lambda(T)\,[\mbox{erg}\mbox{ cm}^{-3}\mbox{ s}^{-1}].
321: \label{e:cooling}
322: \end{equation}
323: Here, $\Gamma$ is the heating contribution (mainly photo-electric heating from grains),
324: $n\Lambda$ the cooling contribution (mainly due to the CII HFS line at $158\mu$m).
325: Since the cooling and heating prescription has to be added outside the
326: flux computations, it lowers the time order of the scheme.
327: To speed up the calculations, equation~(\ref{e:cooling}) is tabulated on a $2048^2$ grid
328: in density and temperature. For each cell and iteration, the actual energy change
329: is then bi-linearly interpolated from this grid.
330:
331: The cores forming due to gravitational collapse reach densities of a few
332: $10^5$~cm$^{-3}$, far beyond the applicable range of our cooling
333: curve. Strictly speaking, we should therefore
334: extend the cooling curve to include molecular lines at high densities.
335: However, our cooling curve reaches an equilibrium temperature of approx $12$K for
336: $n>10^3$~cm$^{-3}$, close enough to a realistic temperature for molecular cores.
337: Since we cannot resolve the core structure anyway, we chose to stick
338: to this simplified treatment.
339:
340: The cooling curve is limited to densities of $n\leq 10^5$~cm$^{-3}$, to prevent a
341: ``catastrophic'' collapse which would be generated by a sub-isothermal effective equation of state.
342: The sudden reversal to an adiabatic equation of state helps limit the densities
343: and prevents numerical artifacts caused by single, very high density cells.
344: However, this stiffening of the equation of state -- if introduced at too low densities --
345: could stabilize the cores, prevent their fragmentation and render them
346: more prone to dispersion. We experimented
347: with the density threshold $n_{max}$ and found that a value
348: of $n_{max}=10^5$~cm$^{-3}$ prevents the run-away collapse while allowing
349: the cores to remain small (and dense) enough to stay gravitationally
350: bound once they have formed. See \S\ref{ss:resolution} for a discussion of the
351: resolution limits.
352:
353: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
354: %\section{Gravity}
355: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
356: \subsection{Gravity}
357: Self-gravity is implemented as an external source term in time-unsplit form.
358: The Poisson-equation is solved via a non-periodic Fourier solver, using
359: the (MPI-parallelized) {\tt fftw} (Fastest Fourier Transform in the West)
360: libraries. We tested this against direct summation
361: to assure that the Poisson equation is solved accurately.
362: The non-periodic solver needs twice the grid size for
363: padding the Fourier transforms. This limits the resolution
364: of our simulations to effectively $512^3$ cells.
365:
366: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
367: %\subsection{Gravity}
368: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
369: \subsection{Core Identification}\label{ss:coreident}
370:
371: We use two methods to identify cores in the model data.
372: To find gravitationally bound objects,
373: we employ the CLUMPFIND algorithm \citep{1994ApJ...428..693W}
374: in a modified version \citep{2000ApJ...535..887K}. We then test
375: whether the structures identified by CLUMPFIND are gravitationally bound or collapsing,
376: by checking their Jeans mass, the ratio of thermal and (internal) kinetic
377: energy over gravitational energy, and the velocity divergence.
378: If all three tests are passed, a structure is accepted as a core.
379: We track individual cores by identifying the closest ``neighbor'' to a given
380: core in the next timestep (where ``timestep'' does not mean the CFL-timestep,
381: but the time between writing data sets). The simulations presented here form
382: sufficiently few cores for this simple method to be accurate.
383:
384: The second method is a simple clipping algorithm, motivated by
385: the fact that due to the thermal instability, dense coherent
386: regions are generally well defined in our models. The method
387: selects the maximum density and builds a tree structure around
388: the central cell, thus connecting all cells above the given
389: density threshold of $n_{th}>50$~cm$^{-3}$. Once this threshold is reached,
390: the process restarts with the next-lower density
391: peak not included in the previous structure. The resulting cores
392: are accepted independently of whether they are gravitationally bound or collapsing.
393:
394: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
395: %
396: %\section{Model Results}
397: %
398: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
399: \section{Model Results}\label{s:results}
400:
401: The general signature of fast local fragmentation in colliding flows is most easily
402: recognized in the morphologies of the clouds (\S\ref{ss:morph}). A more quantitative
403: measure can be gleaned from the core mass evolution and the energy distribution
404: (\S\ref{ss:colhist}). Dynamical signatures are discussed in \S\ref{ss:gasdyn},
405: and \S\ref{ss:resolution} sounds two cautionary notes regarding the numerical resolution.
406:
407: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
408: %\subsection{Morphologies}
409: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
410: \subsection{Morphologies}\label{ss:morph}
411:
412: We begin by comparing the morphologies of the clouds forming in
413: colliding flows (models Hf1, Gf1 and Gf2, \S\ref{sss:global}).
414: A new formation mechanism for massive cores is discussed in \S\ref{sss:dynafocus} .
415:
416: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
417: %\subsubsection{Global Collapse and Filament Formation}
418: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
419: \subsubsection{Global Collapse and Filament Formation}\label{sss:global}
420: The top row of Figure~\ref{f:morph-Gf12l} shows three
421: time instances of model Hf1, seen along the inflow direction. The colliding
422: flows cause a big ``splash'', the effects of which are still noticeable $7.6$~Myr
423: after the initial flow collision (left column), however, rapid cooling leads to strong density
424: enhancements in the interaction zone, and in combination with the dynamical
425: instabilities triggered by the perturbed interface, to strong fragmentation.
426: Note the radial filaments and the outermost ``bounding ring'' at $7.6$~Myr. The radial
427: filaments are similar to those seen in the conceptually equivalent models by
428: \citet{2007ApJ...657..870V}: the compressed material is escaping the interaction
429: region by the way of least resistance, i.e. laterally to the inflow. Once
430: an ``escape channel'' has been formed due to small fluctuations in the external
431: pressure, the resulting pressure deficit will ensure that the channel will continue
432: to be used by subsequent material. The ring-like structure is just
433: the shock wave from the initial splash caused by the colliding flows.
434:
435: \begin{figure*}
436: % \includegraphics[width=\textwidth]{../figures/morphGf12l.eps}
437: \includegraphics[width=\textwidth]{f1.eps}
438: \caption{\label{f:morph-Gf12l}Time sequence of logarithmic column density maps
439: for models Hf1 (top), Gf1 (center) and Gf2 (bottom), seen along the inflow direction.
440: At $t=7.6$~Myr (left column) the full domain (44~pc) is shown, while we restrict the field
441: of view to the central 3/4 (33~pc) of the domain at later times, to highlight the small-scale
442: dense structures forming.}
443: \end{figure*}
444:
445: With increasing time, more and more mass is collected in the interaction region,
446: rapidly assembling the cloud. Note that for the sake of greater detail
447: the center and right-hand column of
448: Figure~\ref{f:morph-Gf12l} show only 3/4 (in linear extent) of the simulation domain
449: corresponding to a box length of 33~pc. The non-gravitating model Hf1 also continues
450: to collect mass as time proceeds, however the column densities reached do not exceed a
451: few $10^{21}$~cm$^{-2}$.
452:
453: This changes with the introduction of gravity (center and lower row of Fig.~\ref{f:morph-Gf12l}).
454: At $7.6$~Myr, the structures are still pretty much indiscernible, while at $11.4$~Myr, the first
455: regions of high column density have formed.
456: While the cores forming in the cloud result from local collapse (see below), the filaments
457: at later stages ($14.5$~Myr) are a consequence of the global collapse of the whole cloud. This can be
458: more easily seen in the right-hand column of Figure~\ref{f:morph-Gf12p}. The same models
459: at the same times as in Figure~\ref{f:morph-Gf12l} are shown, but now seen perpendicularly
460: to the inflow. Clearly, at late times, the initially elongated ``red'' structure (the dense
461: part of the cloud) crumples under its own weight. Because of the (generalized) non-circular inflow
462: cross section, this leads to the formation of a filament. This mechanism for filament formation
463: offers a substantial reservoir of mass for further star formation. The subsequent fragmentation
464: of the filaments is possibly enhanced by the thermal instability
465: (see e.g. \citealp{2001Ap&SS.276.1097T}).
466:
467: The larger amplitude of the collision interface perturbation in model Gf2 (see Tab~\ref{t:modparam})
468: leads to a stronger initial fragmentation, mirrored in a more distributed core formation
469: at later stages.
470:
471: Note that the high-density regions (in Fig.~\ref{f:morph-Gf12l}) do not necessarily
472: form at the center of the cloud. On the contrary, there is a tendency for material to collect
473: away from the center, forming filaments (model Gf2 at $14.5$~Myr), or at least extended dense cores.
474: This is a mild version of the edge effect in collapsing finite sheets, as discussed by
475: \citet{2004ApJ...616..288B} and applied to a model of the Orion star forming region
476: by \citet{2007ApJ...654..988H}. Figure~\ref{f:accel3d} is more specific about the actual mechanism:
477: It shows a map of the (projected) gravitational accelerations $|\nabla\Phi|$
478: in the midplane perpendicular to the inflow (to be
479: compared to the bottom row of Fig.~\ref{f:morph-Gf12l}). Contours denote the column density, and
480: the actual potential gradient $-\nabla\Phi$ is indicated by the arrows. The color table is identical
481: to Figure~\ref{f:morph-Gf12l}, i.e. yellow/red indicates strong accelerations. The ring-like structure
482: of strong accelerations towards the center is obvious. Note that the accelerations extend towards larger
483: radii than their accompanying high-density structures: material is accelerated at the edges and is
484: being piled up further down the (radial) flow.
485:
486: The simulations of \citet{2007ApJ...657..870V} show a similar
487: effect as the models discussed here, although much stronger. This quantitative difference
488: might be a consequence of the choice of initial perturbations in the two simulation sets.
489: While \citet{2007ApJ...657..870V} put small-scale perturbations in their inflow velocities,
490: we perturb the collision interface, which in turn leads to a stronger excitation of dynamical
491: instabilities. The predominant instability
492: arising in our setup is the non-linear thin shell instability (NTSI, \citealp{1994ApJ...428..186V}),
493: in combination with shear flow instabilities and the thermal instability \citep{1965ApJ...142..531F},
494: preventing the formation of a more or less uniform slab.
495:
496: \begin{figure}
497: % \includegraphics[width=\columnwidth]{../figures/morphGf12p.eps}
498: \begin{center}
499: \includegraphics[width=\columnwidth]{f2.eps}
500: \end{center}
501: \caption{\label{f:morph-Gf12p}Time sequence of logarithmic column density maps
502: for models Hf1 (top), Gf1 (center) and Gf2 (bottom), seen perpendicular to the
503: inflow direction. The full computational domain is shown, measuring $22\times 44$~pc.}
504: \end{figure}
505: \begin{figure*}
506: % \includegraphics[width=\textwidth]{../figures/accel3d.eps}
507: \begin{center}
508: \includegraphics[width=\textwidth]{f3.eps}
509: \end{center}
510: \caption{\label{f:accel3d}Time sequence of accelerations for model Gf2, projected in plane perpendicular to the
511: inflow. Colors denote $|\nabla\Phi|$, the contours represent density, and the arrows
512: indicate $\nabla\Phi$. This should be compared to the bottom row of Figure~\ref{f:morph-Gf12l}.}
513: \end{figure*}
514:
515: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
516: %\subsubsection{Dynamical Focussing}
517: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
518: \subsubsection{Dynamical Focusing and Massive Core Formation}\label{sss:dynafocus}
519:
520: The NTSI is driven by the ram pressure imbalance at the troughs of
521: the rippled interaction surface: the concave side will have an excess of ram pressure
522: because the flows are focused into the troughs, while the convex side will
523: experience a deficit of ram pressure because the incoming gas is deflected.
524: Thus, the NTSI provides a very efficient mechanism to collect gas locally.
525: \citet{2003NewA....8..295H} demonstrated this effect with
526: the help of self-gravitating two-dimensional models.
527:
528: This focusing effect still holds in three dimensions, as can be seen from
529: the time sequence of logarithmic column density maps of two colliding flows
530: under the effect of self-gravity, as shown in Figure~\ref{f:morph-Gs}.
531: The leftmost panel at $t=0.8$~Myr shows structures still pretty close to the
532: initial conditions. The cooling has not yet led to perceptible fragmentation,
533: and turbulence has not yet developed. At $8.0$~Myr, the NTSI is already in
534: full swing, and on the left edge of the cloud, roughly in the mid-plane, the first
535: core starts to form. This core is located at one of the troughs amplified
536: by the NTSI, so that it had ample opportunity to collect instreaming material.
537: This seems to continue all the way up to $16$~Myr, at which point the cloud
538: has grown globally unstable (see \S\ref{ss:colhist}), indicated by the
539: frenzy of core formation all over the cloud. We will present a more
540: detailed discussion of the dynamical focusing in a subsequent paper,
541: including models at higher resolution.
542:
543: \begin{figure}
544: % \includegraphics[width=\columnwidth]{../figures/morphGsp.eps}
545: \includegraphics[width=\columnwidth]{f4.eps}
546: \caption{\label{f:morph-Gs}Time sequence of logarithmic column density maps
547: for model Gs, seen perpendicular to the inflow direction. The domain measures
548: 44~pc in linear extent.}
549: \end{figure}
550:
551: Isolating the gravitationally collapsing cores with CLUMPFIND (see \S\ref{ss:coreident})
552: and tracking the core masses with time yields Figure~\ref{f:masstime-Gs}.
553: The dynamical focusing seems to be a very efficient
554: mechanism to collect substantial mass in a small volume over a short time: 9~Myr
555: after initial flow contact, the first $100$~M$_\odot$-core has formed. At later
556: times, the mass accretion rates get steeper: there is more mass available, and
557: (concurrently) the potential well deepens, so that more massive cores can form
558: within shorter times. The first core does not participate in the steepened
559: accretion history because it sits at the edge of the potential well so
560: that it does not benefit from the higher densities.
561:
562: \begin{figure}
563: % \includegraphics[width=\columnwidth]{../figures/coremass_3aaww00b0da_g05.eps}
564: \includegraphics[width=\columnwidth]{f5.eps}
565: \caption{\label{f:masstime-Gs}Core mass evolution for model Gs. The global
566: collapse of the cloud is paralleled by a frenzy of core formation
567: at late times.}
568: \end{figure}
569:
570: Obviously, stellar feedback will influence the cloud dynamics and star formation
571: efficiency at the late stages of the cloud evolution.
572:
573: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
574: %\subsection{Collapse History}
575: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
576: \subsection{Collapse History}\label{ss:colhist}
577:
578: \subsubsection{Energy Equipartition}\label{ss:equip}
579:
580: A global measure for the cloud evolution under the effect of gravity is the
581: equipartition parameter
582: \begin{equation}
583: \alpha_{eq} \equiv \frac{\int(\rho\mbfv^2+3P)dV}{\int\rho\Phi dV},\label{e:alphaeq}
584: \end{equation}
585: i.e. the ratio of the total kinetic and thermal energy over half the total potential
586: energy\footnote{We decided to follow \citet{2006MNRAS.372..443B} and replace the term ``virial
587: parameter'', since this usually has the connotation of ``virial equilibrium'', an assumption,
588: which generally holds only for an ensemble of molecular clouds, or for a time-average
589: of one cloud over many dynamical times \citep{1999osps.conf...29M}. Since we are concerned
590: here with a single molecular cloud on short timescales, the notion of virial equilibrium is
591: inapplicable -- while that of energy equipartition may still hold \citep{2006MNRAS.372..443B}.}.
592: We neglect any surface terms in the determination of $\alpha_{eq}$, which only allows us to obtain a
593: rough approximation of the ``true'' energetic state of the cloud. Nevertheless, as a closer study
594: of Figure~\ref{f:virpar} demonstrates, the time evolution of $\alpha_{eq}$ mirrors the
595: cloud morphologies (Figs.~\ref{f:morph-Gs} and \ref{f:morph-Gf12l}).
596: Figure~\ref{f:virpar} shows $\alpha_{eq}$ for all the gas above a given threshold density.
597: A low threshold density means that most of the mass and most of the volume enter
598: the calculation of $\alpha_{eq}$.
599: Increasing the density threshold emphasizes more and more the dense cores that form later.
600: The lowest density threshold is set to $n_{th}=10^{2}$~cm$^{-3}$, since we are interested
601: in the equipartition parameter of the isolated cloud, and not of the total simulation volume
602: including the (highly energetic) inflows. The following items are noteworthy:
603:
604: (1) The dense regions tend to be gravitationally
605: bound ($\alpha_{eq} < 1$), while the global cloud behavior (illustrated by the lowest-density
606: curves) approaches $\alpha_{eq} \sim 1$ by the end of the simulations. Thus, in terms of
607: observables, the cloud exhibits an $\alpha_{eq}$ consistent with ``virial equilibrium'' to
608: within a factor of two, well within observational uncertainties in terms of mass estimates,
609: velocity dispersions derived only from line-of-sight motions, and the elimination of
610: surface terms (\citealp{1999ApJ...515..286B}; \citealp{2006MNRAS.372..443B}).
611:
612: (2) The solid lines indicating the lowest density threshold of $n_{th}=10^2$~cm$^{-3}$
613: reach $\alpha_{eq}=1$ between $10$ and $13$~Myr. A lower density threshold can be interpreted
614: as tracing a larger volume, thus the evolution of $\alpha_{eq}$ for low density thresholds
615: indicates that global collapse lags behind the local collapse -- isolated
616: dense cores form before the cloud can collapse globally. However, the cloud
617: {\em does} show the onset of global collapse.
618:
619: (3) At high density thresholds, $\alpha_{eq}<1$ for all times (at which high
620: densities are available): the massive cores (see the corresponding mass history)
621: are fully gravitationally unstable. Note that $\alpha_{eq}$ does not drop further,
622: although the mass increases: the cut-off of the cooling curve leads to a stabilization
623: of the cores and prevents catastrophic collapse.
624:
625: (4) The clouds do not go through an ``equilibrium'' stage, but start to collapse
626: locally during their formation: stars can form locally without a global collapse
627: of the cloud.
628:
629: \begin{figure*}
630: % \includegraphics[width=\textwidth]{../figures/virpar.eps}
631: \includegraphics[width=\textwidth]{f6.eps}
632: \caption{\label{f:virpar}The equipartition parameter $\alpha_{eq}$ against time for
633: the three gravity models Gs, Gf1 and Gf2 as indicated in the panels. The top half of each panel
634: gives $\alpha_{eq}(n>n_{th})$, the bottom half shows the total mass $M(n>n_{th})$.
635: The line styles stand for the threshold densities $n_{th}$ as indicated in the top panels.}
636: \end{figure*}
637:
638: Figure~\ref{f:virpar} allows only an indirect statement about the scale-wise evolution
639: of $\alpha_{eq}$. A more accurate measure is the
640: ratio of the respective Fourier spectra of
641: the kinetic and internal energy over the gravitational energy. This yields $\alpha_{eq}(L)$,
642: a scale-dependent measure of the cloud's stability against gravity (Fig.~\ref{f:virspec}).
643: As in Figure~\ref{f:virpar},
644: we show the three gravitational models, but now at various times. Note that we do not select for
645: gas in the cloud or gas above a density threshold. This is because masking leads to additional
646: structure, which in turn results in ``noise'' signals in the Fourier spectra. To facilitate
647: an easier comparison to spatial scales we plot the scale-dependent $\alpha_{eq}$ against
648: physical length scale, rather than against wave number. Since $\alpha_{eq}$ is only a rough measure
649: of the system's energetics, the following discussion should be seen as a qualitative analysis.
650:
651: At early times, the system is gravitationally stable on all scales. This is not surprising
652: since we are now looking pretty much at the whole box (we removed the $L=44$~pc mode, since this
653: would just be the mean).
654: The small scales definitely collapse first: they are the first to fall
655: beneath $\alpha_{eq}=1$ with increasing time. The minimum in $\alpha_{eq}$ around
656: $L=1\dots 3$~pc at earlier times stems
657: from a ``conspiracy'' between the kinetic and potential energy scales: On larger scales
658: ($L\gtrsim 3$~pc), the kinetic energy of the large-scale inflow still dominates the energy budget
659: at early times, whereas on the small scales, the potential energy drops faster with decreasing
660: scale than the kinetic energy. The small structures forming due to thermal and dynamical fragmentation
661: did not have enough time yet to collect a significant amount of mass.
662:
663: \begin{figure*}
664: % \includegraphics[width=\textwidth]{../figures/virspec.eps}
665: \includegraphics[width=\textwidth]{f7.eps}
666: \caption{\label{f:virspec}The equipartition parameter $\alpha_{eq}$ against scale, for the three gravity
667: models Gs, Gf1 and Gf2 as indicated in the panels. The line styles stand for
668: the times at which $\alpha_{eq}$ has been measured. All models have been cut at a scale
669: of $0.3$~pc, although the numerical resolution is a factor of approximately 3 (6) higher
670: for model Gs (Gf1, Gf2).}
671: \end{figure*}
672:
673: In summary, we note that in all models local collapse wins over global collapse, i.e. the small scales
674: generated by thermal and dynamical fragmentation collapse first. Global collapse however occurs and
675: feeds more material into the already active ``star forming'' region. The clouds do not
676: reach an
677: ``equilibrium stage'', but proceed from formation directly to local collapse and only then to global
678: collapse. This result does not
679: depend on the presence or absence of stellar feedback, in contrast to the
680: late-stage evolution of our model clouds, where feedback will have a deciding influence.
681:
682: \subsubsection{Total and Core Mass Evolution}
683: We have already seen the core mass history of model Gs (Fig.~\ref{f:masstime-Gs}).
684: Figure~\ref{f:allmasses} shows the total mass evolution of all models.
685: Thick lines refer to gas at $T<100$~K, which can be identified as cloud gas due
686: to the thermal instability. Thin lines denote the gas at $T>100$~K.
687: The symbols stand for the total mass within collapsing cores for each model,
688: i.e. for $M_*$, the mass that constitutes the reservoir for star formation.
689:
690: \begin{figure}
691: % \includegraphics[width=\columnwidth]{../figures/allmasses_grav3d.eps}
692: \includegraphics[width=\columnwidth]{f8.eps}
693: \caption{\label{f:allmasses}Mass history of all models. Thick lines denote
694: $M(T<100\mbox{K})$, thin lines stand for $M(T>100\mbox{K})$. Different
695: line styles represent the four models. Symbols refer to the total mass
696: in collapsing cores identified by CLUMPFIND (see \S\ref{ss:coreident}).}
697: \end{figure}
698:
699: Note that the mass is scaled logarithmically, while the (cold) cloud mass evolves
700: linearly with time. Essentially, despite all the substructure, the colliding flows
701: form one cold slab (see also \citealp{2006ApJ...648.1052H}). Comparing models
702: Hf1, Gf1 and Gf2, gravity does not affect the global cloud mass: the inflows
703: determine the mass in the cold gas phase (i.e. in the cloud), and thus the material available
704: for making stars. Model Gs differs from all others just because of the larger
705: extent of the collision site. Otherwise, its evolution regarding the cloud mass
706: is qualitatively similar.
707:
708: The core masses evolve in parts nearly exponentially. This is a consequence not of
709: the individual mass accretion events, as Figure~\ref{f:masstime-Gs} demonstrates: those
710: are fairly linear with time. Rather, the non-linear evolution is a consequence of the
711: explosion of star formation activity once sufficient mass has
712: accumulated. Comparing the mass in the cores, $M_*$, against the cloud mass $M_{cl}$ -- identified
713: with $M(T<100\mbox{K})$ --, we note that over a large stretch of time, the ``star formation
714: efficiency'' $M_*/M_{cl} < 0.1$ -- the collapse is initially truly local.
715:
716: \subsubsection{Core Mass Distribution\label{sss:comafu}}
717: Since the thermal instabilities are the first fragmentation agent in the
718: colliding gas streams, will they set the core mass distribution in molecular
719: clouds? Figure~\ref{f:coremassfunc} shows the core mass distribution in logarithmic
720: mass intervals for all four models. The (fit) slopes $d\ln N/d(\ln M)= s$
721: are indicated in the legend. The Salpeter IMF would have an exponent of $s=-1.35$.
722: While the slopes are approximately consistent
723: with observed core mass distributions
724: \citep{1996A&A...307..915K,2002A&A...384..225S},
725: we note that there is at most one decade in mass for the
726: fits. Moreover, these mass spectra are most likely affected by
727: the stiffening of the equation of state for $n>10^5$~cm$^{-3}$
728: (\S\ref{ss:hydrocool}), i.e. low-mass objects may be under-represented.
729: This may cause a flattening of the spectra of the gravity-models
730: (Gf1, Gf2 and Gs) -- possibly visible in the lower right panel (13 Myr) of
731: Figure~\ref{f:coremassfunc} --,
732: whereas model Hf1 shows a rather well-defined power law
733: down to the lowest masses at late times.
734: Note that the cores contributing to Figure~\ref{f:coremassfunc} are not
735: necessarily self-gravitating. Over a period of 4~Myr, the spectra roughly keep
736: their shape, another indication that the fragmentation is mainly due to thermal
737: effects rather than gravity.
738:
739: \begin{figure*}
740: % \includegraphics[width=\columnwidth]{../figures/coremassspec_grav3d_17.eps}
741: \includegraphics[width=\textwidth]{f9.eps}
742: \caption{\label{f:coremassfunc}Time sequence of core mass distributions for all four
743: models. Note that the distributions are shifted vertically by the factor
744: indicated in the legend, to make them easier to identify.
745: The Salpeter IMF would have a slope of $-1.35$. The dashed line denotes
746: $d(\ln N)/d(\ln M) = -0.8$.}
747: \end{figure*}
748:
749: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
750: %\subsection{Gas Dynamics}
751: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
752: \subsection{Gas Dynamics}\label{ss:gasdyn}
753:
754: Figure~\ref{f:veldisvt} summarizes the various line-of-sight velocity dispersions
755: (LOSVD) for the cold ($T<100$~K) gas, for all models.
756: All LOSVDs are one-dimensional and density-weighted, e.g. the total LOSVD is
757: \begin{equation}
758: \sigma_v\equiv\left(\frac{\int\mbfv^2\,n\,dV}{3\int n dV}\right)^{1/2}.\label{e:losvd}
759: \end{equation}
760:
761: \begin{figure}
762: % \includegraphics[width=\columnwidth]{../figures/allveldis_vt_all.eps}
763: \includegraphics[width=\columnwidth]{f10.eps}
764: \caption{\label{f:veldisvt}{\em Top:} One-dimensional velocity dispersions against time,
765: along the inflow direction ($\sigma_x$, thin lines), perpendicular to the
766: inflows ($\sigma_{yz}$medium lines), and total ($\sigma_v(T<100\mbox{K})$, thick
767: lines), for all models.
768: {\em Bottom:} One-dimensional velocity dispersions against time.
769: We distinguish between the velocity dispersion within each core,
770: averaged over all cores ($\langle\sigma_{core}\rangle$, thin lines), the
771: velocity dispersion of the gas taken over all cores ($\sigma_{core}$, medium lines),
772: and the total velocity dispersion (as in top panel, thick lines).}
773: \end{figure}
774:
775: The top panel shows the velocity dispersion along the inflow direction ($\sigma_x$),
776: the transverse velocity dispersion $\sigma_{yz}$, and the total dispersion.
777: For all models, $\sigma_x$ is highest, initially increasing slightly with time,
778: and later leveling off. The initial rise is due to the still acting NTSI, i.e.
779: the ripples in the interaction interface are still amplified, while at later
780: times, the NTSI is saturated, and global gravity takes over. The latter can
781: be seen when comparing model Hf1 and Gf1, which are identical except that
782: Gf1 has self-gravity. For both, $\sigma_x$ initially evolves similarly, up
783: to $t\approx 7.5$~Myr. After that, $\sigma_x(\mbox{Gf1})$ starts to level off,
784: while $\sigma_x(\mbox{Hf1})$ still continues to rise: gravity constrains the
785: slab and suppresses the further growth of the NTSI. This is obvious in
786: Figure~\ref{f:morph-Gf12p}.
787:
788: The transverse velocity dispersion $\sigma_{yz}$ (medium lines in top panel
789: of Fig.~\ref{f:veldisvt}) shows the influence of global gravity at later
790: times for models Gf1, Gf2 and Gs, while that of model Hf1 drops with time.
791: For model Hf1, the total energy input of the inflows is balanced by radiative
792: losses and out-streaming material: the total velocity dispersion stays
793: approximately constant with time. Not so for the self-gravitating models:
794: they all have increasing total velocity dispersions with time.
795:
796: In the bottom panel of Figure~\ref{f:veldisvt}, the thick lines stand for
797: the total velocity dispersion again (as in the top panel). The thin lines
798: ($\langle\sigma_{core}\rangle$)
799: denote the velocity dispersion within each core, averaged over all cores.
800: This is the same quantity as the ``internal'' velocity dispersion as
801: discussed by \citet{2006ApJ...648.1052H}. Due to the strong radiative losses,
802: this velocity dispersion is subsonic (the sound speed in the cold gas
803: is $c_s\approx 0.3$~km~s$^{-1}$). In contrast, the velocity dispersion of
804: all gas in the cores, $\sigma_{core}$ is slightly supersonic, and increases
805: with time due to global gravitational collapse.
806:
807: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
808: %\subsection{A Comment on Resolution}
809: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
810: \subsection{Two Comments on Resolution}\label{ss:resolution}
811:
812: The left column of Figure~\ref{f:jeanslength}
813: shows the histograms of the Jeans length for all three
814: self-gravitating models. The dashed vertical line indicates the
815: \citet{1997ApJ...489L.179T} criterion with a safety factor of 4,
816: mandating a minimum number of cells per Jeans length. We evaluate the criterion
817: locally, i.e. per cell. Clearly, with
818: increasing time, more and more cells fall below the resolution limit (to the
819: left of the vertical dashed line). A more detailed view is offered by the set of
820: panels on the right side of Figure~\ref{f:jeanslength}. They show the
821: Jeans length in each cell against the corresponding density. Again,
822: dashed lines denote the Truelove limit, the upper line for 4 cells, the lower
823: one for 2. Although only a minor fraction of cells is unresolved, they
824: do exist. The strict correlation between $n$ and $\lambda_J$ for
825: $n<10^5$~cm$^{-3}$, and the scatter at larger densities, is a direct consequence
826: of the cooling curve: we are far up the isothermal branch, so that the
827: thermal timescales are much shorter than the dynamical timescales.
828: As discussed above, we switch off
829: the cooling for densities $n>10^5$~cm$^{-3}$, to prevent unphysically high
830: densities and subsequent numerical problems. The effective equation of state reverts
831: to adiabatic with $\gamma=5/3$ at that point, so that the temperature
832: increases due to the strong compressions.
833:
834: \begin{figure*}
835: % \includegraphics[width=\textwidth]{../figures/jeanslength.eps}
836: \includegraphics[width=\textwidth]{f11.eps}
837: \caption{\label{f:jeanslength}Histograms of the Jeans length (left column)
838: for all three self-gravitating models, and scatter plots of the Jeans
839: length against density, for all models at three times. The dashed lines
840: denote the \citet{1997ApJ...489L.179T} criterion with a safety factor of 2 and 4.}
841: \end{figure*}
842:
843: There are several resolution criteria for thermally unstable systems (see e.g.
844: \citealp{2007A&A...465..431H} for a discussion).
845: The dynamically most stringent condition is to resolve the cooling
846: length $\lambda_c = c_s\,\tau_c$. If this length scale
847: is not resolved in the cold gas, the (isobaric) condensation mode of the thermal instability will be
848: underestimated (see \citealp{1965ApJ...142..531F} and \citealp{2000ApJ...537..270B}).
849: The bulk of the cold gas resides at $40$~K for our cooling curve
850: (see \citealp{2006ApJ...648.1052H}), and the cooling time scales in the cold gas are on the order
851: of $10^4$~years, so that the cooling length scale is $\lambda_c\approx 6\times 10^{-2}$~pc,
852: slightly beneath the nominal resolution for models Hf1, Gf1 and Gf2 (see \S\ref{ss:models}).
853: Underestimating the condensation mode of the thermal instability will result in fewer low-mass
854: fragments. In that sense, the core mass budgets (\S\ref{ss:colhist}) and the fragmentation
855: history in our models are conservative estimates: at higher resolution, more fragmentation
856: -- and possibly earlier low-mass star formation -- is expected.
857:
858: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
859: %
860: %\section{Discussion}
861: %
862: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
863: \section{Discussion}\label{s:discussion}
864:
865: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
866: %\subsection{Global Collapse into Filaments}
867: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
868: \subsection{Global Collapse and Filament Formation}
869:
870: Galactic star formation seems to prefer filamentary -- at least elongated --
871: rather than spherical molecular clouds (\citealp{1979ApJS...41...87S};
872: \citealp{1997ApJ...474L.135C}; \citealp{2001ApJ...562..852H}; \citealp{2005A&A...440..151H};
873: \citealp{2007A&A...462L..17A}).
874: The finite extent of the forming cloud opens two paths for filament formation. The first -- obvious -- one
875: is global collapse along the shorter axes. The second one -- less obvious -- arises from the fact that
876: the radial accelerations in a two-dimensional elliptical or circular cloud of uniform density diverge
877: at the edges, i.e. material at the edges experiences the strongest accelerations inwards, leading to
878: a pile-up of gas at the edge, i.e to the formation of a filament \citep{2004ApJ...616..288B}.
879:
880: Obviously, ``real'' clouds are three-dimensional. However, if clouds form in (laterally constrained)
881: colliding flows, they will have a finite extent, and more importantly, global gravity will not have
882: had sufficient time to lead to a centrally peaked density profile. The strong thermal instabilities lead
883: to an initially rather thin sheet, only broadened by the dynamical instabilities. Thus, to zeroth
884: order, in such a scenario the sheet-approximation is quite reasonable.
885: The numerical models of \citet{2007ApJ...657..870V}
886: and the global collapse pattern of models Gf1 and Gf2 above support this interpretation. In fact, the
887: dynamical instabilities will enhance a ``crumpling'' of the cloud in the lateral directions once
888: global gravity dominates. As the discussion in \S\ref{sss:global} shows however, stars will have formed
889: locally by then.
890:
891: It is reasonable to assume that the (idealized) inflows implemented in our models will not have
892: a circular cross section. In galaxy mergers or in the collisions of super-shells in the LMC,
893: the thickness of the disk would limit the thickness (vertical extent) of the flow. A similar
894: assumption seems valid for spiral shocks in the Galaxy, where the disk potential would again
895: lead to a flattening of the inflows (e.g. \citealp{2006MNRAS.371.1663D}).
896:
897: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
898: %\subsection{Formation of Massive Cores}
899: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
900: \subsection{Formation of Massive Cores}
901: Ripples in the flow collision interface can focus the instreaming gas,
902: leading to a very efficient mechanism to form massive cores
903: (Figs.~\ref{f:morph-Gs}, \ref{f:masstime-Gs}).
904: $10$~Myr after flow collision, the first core
905: in model Gs has a mass of approximately $150$~M$_\odot$ and a diameter
906: of $1$~pc, corresponding to a column density of $3.6\times10^{22}$~cm$^{-2}$.
907: Likewise, after $10$~Myr, the mean column density in the box along the
908: inflow direction will be $N_{tot} = 3\mbox{ cm}^{-3}\times(44\mbox{ pc}+7.9\mbox{ km}
909: \mbox{ s}^{-1}\times 10\mbox{ Myr}) = 1.1\times 10^{21}$~cm$^{-2}$. The dynamical
910: focusing -- helped along by cooling and eventually gravity -- leads to an
911: excess of column-density by a factor of $\approx 30$.
912:
913: This is as good a place as any to remind the reader that we are leaving out the
914: crucial step of molecule formation in our models. The first appearance of
915: molecules sets the clock for the lifetime of the cloud. We can
916: only mimic this by arguing that once column densities of $\approx 10^{21}$~cm$^{-2}$
917: are reached, we regard the cloud as ``molecular''. \citet{2001ApJ...562..852H} point out
918: that this column density is of the same order as the critical column density
919: for gravity to become dominant.
920:
921: With the NTSI as driving mechanism, the cores will
922: form at a certain distance from the bulk of the cloud (Fig.~\ref{f:morph-Gs}).
923: This distance is basically given by the amplitude the NTSI has reached up to
924: that point. The somewhat peculiar location could have profound repercussions
925: on the effect of feedback from the massive stars forming in the
926: core: the winds and the expanding HII regions or supernovae might
927: lead to a further compression of the already massive cloud next to the young
928: stars, thus triggering further star formation. In contrast, if the
929: stars were located inside the bulk of the cloud, the stellar feedback
930: could be expected to provide partial support to the cloud, or more
931: likely disperse the lower-density regions.
932:
933: \citet{2005ApJ...635.1062K} analyzed the molecular cloud population
934: of the starburst galaxy M82, arguing that in many cases, star formation
935: seems to proceed from the outside inwards, i.e. that the regions of
936: massive star formation (indicated by HII regions and supernova remnants)
937: are to be found at the edges of the molecular clouds. They argue that
938: a sudden increase in external pressure (e.g. a shock wave traveling
939: through the cloud) triggers star formation in a previously existing
940: cloud. The dynamical focusing effect discussed here might offer
941: another explanation, which then would imply that the central molecular
942: cloud is still forming -- an alternative which may be even more attractive in view
943: of the notorious problem of stabilizing an object of many Jeans masses
944: against gravitational collapse (\citealp{2004ApJ...616..288B}; \citealp{2007RMxAA..43..123B}).
945:
946: The role of the ``external pressure increase'' in our models is taken over
947: by the dynamical focusing, i.e. by the excess ram pressure at the troughs
948: of the perturbed sheet. As Figure~\ref{f:pressprof-Gs} demonstrates, the
949: {\em thermal} pressure does not vary strongly.
950:
951: \begin{figure}
952: % \includegraphics[width=\columnwidth]{../figures/pressprof_gs.eps}
953: \begin{center}
954: \includegraphics[width=\columnwidth]{f12.eps}
955: \end{center}
956: \caption{\label{f:pressprof-Gs}Laterally averaged pressure (energy density)
957: profiles for
958: model Gs, at times as indicated in panels. At $11$~Myr, gravity
959: has a noticeable global effect. Note that the internal pressure
960: (dash-3-dot) varies only mildly.}
961: \end{figure}
962:
963: Thus, the gas still
964: evolves sub-isothermally, i.e. the effective adiabatic exponent
965: $\gamma < 1$.
966:
967: At late times (top panel, $15.2$~Myr), kinetic and gravitational pressure
968: are in equipartition. Once again, we emphasize that this does not mean that the
969: system is in equilibrium: the kinetic energy just follows the gravitational energy
970: \citep{2006MNRAS.372..443B}.
971:
972: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
973: %\subsection{Early Fragmentation and Core Mass Spectra}
974: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
975: \subsection{Early Fragmentation and Core Mass Distributions}
976: The core mass distributions (Fig.~\ref{f:coremassfunc}) are consistent
977: with observations \citep{1996A&A...307..915K,2002A&A...384..225S}, although they are flatter than
978: the Salpeter IMF. \citet{2007A&A...462L..17A} quote a core mass distribution
979: for the Pipe nebula close to the Salpeter IMF, however, as they point out, the cores
980: that they are using for analysis are probably the direct progenitors of stars --
981: a stage we cannot hope to reach with the models presented here.
982: Bearing in mind that the distributions
983: for models Gf1, Gf2 and Gs are probably slightly too flat for numerical reasons
984: (\S\ref{sss:comafu}), their similarity to the non-gravitating model Hf1
985: substantiates the claim that the density substructure -- specifically the
986: core mass distribution -- in molecular clouds could very well arise
987: very early on during their formation (\citealp{2007A&A...462L..17A};
988: \citealp{2007A&A...465..445H}) -- driven more by thermal than by gravitational fragmentation.
989:
990: However, we caution that we have not followed the cloud evolution through
991: the stage of molecular hydrogen formation, an investigation we defer to a later paper.
992: Finally, we again emphasize the difficulty of determining mass spectra
993: from simulations with limited dynamic range.
994:
995: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
996: %\subsection{Star Formation Duration and Age Spread}
997: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
998: \subsection{Star Formation Duration and Stellar Age Spread}
999:
1000: From Figure~\ref{f:masstime-Gs} one would infer an
1001: ``age spread'' of the stellar population in the complex of
1002: up to $8$~Myr. This of course assumes that massive, self-gravitating
1003: cores will sit around for several Myr without forming massive stars
1004: whose energy input will tend to disrupt the cloud - a highly improbable
1005: supposition. While this aspect of our simulation is quantitatively
1006: unrealistic, it does suggest the following qualitative points:
1007:
1008: (1) Even in a completely dynamic simulation without magnetic or driven
1009: turbulent support, the overall timescales of star formation can be
1010: longer than the local collapse times. Putting this another way,
1011: the age spread in a star-forming region is an {\em upper limit}
1012: to the timescale of local collapse; it is only equal to the local
1013: collapse time if all star formation is globally synchronized, which
1014: is less and less plausible on larger and larger scales.
1015:
1016: (2) The large-scale flow picture of star-forming cloud accumulation is
1017: attractive in that it allows for the formation of stars over timescales (1-2 Myr)
1018: short compared with lateral crossing times (10-20 Myr), as observed \citep{2001ApJ...562..852H}.
1019: This does not mean that a region of space may not have a significant
1020: age spread. In the paradigm we are pursuing, where non-linear fluctuations
1021: are important, it is plausible that a few large perturbations might collapse
1022: long before more general star formation ensues, as we see in our simulations.
1023: Although we form mostly massive cores, other initial conditions -- and higher
1024: numerical resolution (see \S\ref{ss:resolution}) -- may lead
1025: to the formation of a few low-mass stars initially from the low-probability
1026: high density tail of perturbations, with the smaller, more frequent perturbations
1027: constituting the bulk of the star formation at a later time. Again, such
1028: age spreads provide no constraint on local collapse timescales.
1029:
1030: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1031: %\subsection{The Role of Magnetic Fields}
1032: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1033: \subsection{The Role of Magnetic Fields}
1034:
1035: Magnetic fields are not included in the models presented here.
1036: Based on two-dimensional simulations by
1037: \citet{1995ApJ...441..702V}, \citet{2001ApJ...562..852H} envisaged
1038: the large-scale converging flows preferentially running along
1039: (dynamically dominant, e.g. \citealp{2005ApJ...624..773H}) magnetic
1040: field lines in the diffuse ISM, arguing that if the field lines
1041: were oriented perpendicularly, the stiffening of the equation of
1042: state due to the increased magnetic pressure would counter the
1043: lowering of the effective adiabatic index due to (atomic line)
1044: cooling (see also \citealp{2004ApJ...612..921B}). Although
1045: there are problems with this scenario (for one, substructures
1046: can form even in the case of perpendicular field lines
1047: and lead to further fragmentation [e.g.
1048: \citealp{2007ApJ...665..445H}], and then, this scenario is
1049: clearly motivated by two-dimensional simulations, not allowing
1050: for interchange modes in the instabilities. These usually grow
1051: at least at the hydrodynamical rate, see e.g. \citealp{2007arXiv0707.1022S}),
1052: we will adopt it for the current study. Its consequence is that since
1053: material is being piled up along the field lines, the mass-to-flux ratio
1054: would easily exceed the critical value for magnetical dominance
1055: \citep{1976ApJ...210..326M} after a few Myr, thus rendering the magnetic
1056: fields unimportant for the local gravitational collapse.
1057: It is true that the global dynamics of the cloud especially during
1058: its early evolutionary stages could be fundamentally influenced by the
1059: presence of magnetic fields. We will defer this important problem to
1060: a future paper.
1061:
1062: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1063: %
1064: %\section{Summary}
1065: %
1066: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1067:
1068: \section{Summary}\label{s:answers}
1069:
1070: In an extension of our previous work \citep{2005ApJ...633L.113H,2006ApJ...648.1052H}
1071: we presented three-dimensional models of molecular cloud formation in colliding flows
1072: including self-gravity. At an effective resolution of $512^3$ they currently are the
1073: most highly-resolved models of this kind. We used a fixed-grid code and a non-periodic
1074: Poisson-solver. The initial conditions emphasized the effects of global versus local
1075: gravity on the newly forming cloud.
1076:
1077: The model resolution did not allow us to follow the collapsed cores and to study their properties.
1078: Stellar feedback was not included. Thus, as time increases, the global evolution of
1079: the molecular cloud gets less and less realistic. Furthermore, we cannot
1080: make statements about the lifetime of the molecular cloud.
1081: Bearing these caveats in mind, we found that:
1082:
1083: (1) Any perturbation in the colliding flows is strongly amplified by a combination of
1084: thermal and dynamical instabilities. Specifically, the thermal effects lead to local
1085: high-density fragments that subsequently collapse before the (still forming) cloud
1086: can collapse globally. Thus, cloud formation in colliding flows allows the rapid onset of
1087: local star formation while evading the problem of globally collapsing clouds.
1088: This is consistent with earlier findings \citep{2004ApJ...616..288B}
1089: that local collapse can only
1090: win over global collapse in the presence of early non-linear density perturbations
1091: (Fig.~\ref{f:virspec}).
1092:
1093: (2) Even in the highly dynamical environment of the colliding flows, an
1094: elongated finite cloud under global gravity can form one or more filaments by
1095: collapsing along its shorter axes. This lateral collapse
1096: opens up a further mass reservoir for star formation
1097: (Figs~\ref{f:morph-Gf12l}, \ref{f:morph-Gf12p}).
1098:
1099: (3) Dynamical focusing, i.e. the deflection of incoming gas due to ripples in the
1100: interface between colliding gas streams, leads to very efficient high-mass core formation
1101: (Figs~\ref{f:morph-Gs}, \ref{f:masstime-Gs}): core masses of a few $100$~M$_\odot$ are
1102: reached within $\approx 10$~Myr.
1103:
1104: (4) The clouds are not in any state of equilibrium at any time. Different scales submit
1105: to gravitational collapse at different times, with the small scales going first
1106: (Figs~\ref{f:virpar}, \ref{f:virspec}). Still, the global equipartition parameter
1107: (eq.~\ref{e:alphaeq}) ranges around $1$ within a
1108: factor of a few -- consistent with
1109: observations when taking into account the observational uncertainties.
1110:
1111: (5) The similarity of the core mass distribution for models with and without
1112: self-gravity and their robustness with time
1113: indicate that it might be set very early on during
1114: cloud formation. Thus, the core mass distribution would be a consequence of
1115: fragmentation due to (magneto-)hydrodynamical instabilities and cooling rather
1116: than due to self-gravity (Fig.~\ref{f:coremassfunc}).
1117:
1118: (6) The clouds become globally unstable in the absence of feedback, leading to an exploding
1119: ``star formation efficiency'' (in quotes, since we only can form cores in our models,
1120: but not stars). The turbulence imparted by the (continuing inflows) does not suffice to
1121: balance global gravity. As mentioned in \S\ref{s:introduction} and above, the
1122: absence of feedback in our models does
1123: not allow us to test how long the clouds will survive. However, our models show that
1124: the clouds will not disperse ``on their own accord'' in the background flows, i.e without
1125: the help of stellar feedback.
1126:
1127: %While this might be realistic for a starburst environment,
1128: %it begs the question about the (observationally mandated) short cloud lifetimes.
1129: %Thus, feedback is a crucial ingredient in the scenario of rapid star formation and
1130: %short cloud lifetimes: although the initial increase of the ``star formation efficiency''
1131: %may still be realistic, we cannot model the end of the star formation process and the
1132: %dispersal of the cloud.
1133:
1134: \acknowledgements
1135: Computations were performed at the NCSA (AST040026, AST060031) and on the
1136: local resources at U of M including the 64-processor cluster Star,
1137: perfectly administered and maintained by J.~Hallum.
1138: This work was supported by NASA grant NNG06GJ32G and the University of Michigan.
1139: It has made use of the NASA Astrophysics Data System.
1140:
1141: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1142: %
1143: %\references
1144: %
1145: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1146:
1147:
1148: \bibliographystyle{apj}
1149: \bibliography{./references}
1150:
1151: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1152: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1153: \end{document}
1154: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1155: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1156:
1157:
1158:
1159:
1160: