1: \documentclass[12pt,preprint]{aastex}
2: %\documentclass{iaus}
3:
4: %\usepackage{graphicx}
5:
6: %\input epsf
7:
8: \begin{document}
9:
10: \title{Variations in Stellar Clustering with Environment: Dispersed Star Formation
11: and the Origin of Faint Fuzzies}
12:
13: %Variation of Stellar Clustering with Environmental Pressure}
14:
15: \author{Bruce G. Elmegreen}
16: \affil{IBM Research Division, T.J. Watson
17: Research Center, P.O. Box 218, Yorktown Heights, NY 10598,
18: bge@watson.ibm.com}
19:
20: \begin{abstract}
21: The observed increase in star formation efficiency with average
22: cloud density, from several percent in whole giant molecular clouds
23: to $\sim30$\% or more in cluster-forming cores, can be understood as
24: the result of hierarchical cloud structure if there is a
25: characteristic density as which individual stars become well
26: defined. Also in this case, the efficiency of star formation
27: increases with the dispersion of the density probability
28: distribution function (pdf). Models with log-normal pdf's illustrate
29: these effects. The difference between star formation in bound
30: clusters and star formation in loose groupings is attributed to a
31: difference in cloud pressure, with higher pressures forming more
32: tightly bound clusters. This correlation accounts for the observed
33: increase in clustering fraction with star formation rate and with
34: the observation of Scaled OB Associations in low pressure
35: environments. ``Faint fuzzie'' star clusters, which are bound but
36: have low densities, can form in regions with high Mach numbers and
37: low background tidal forces. The proposal by Burkert, Brodie \&
38: Larsen (2005) that faint fuzzies form at large radii in galactic
39: collisional rings, satisfies these constraints.
40: \end{abstract}
41:
42: \keywords{stars: formation --- ISM: structure --- open clusters and
43: associations: general}
44:
45:
46: \section{Introduction} \label{sect:intro}
47:
48: Stars form in concentrations with a range of densities, from star
49: complexes, OB associations, and T Tauri associations at the low end
50: to compact clusters and super-star clusters (SSC) at the high end.
51: The overall structure is usually hierarchical (Scalo 1985;
52: Feitzinger \& Galinski 1987; Ivanov et al. 1992; Gomez et al. 1993;
53: Battinelli, Efremov \& Magnier 1996; Bastian, et al. 2005, 2007;
54: Elmegreen et al. 2006; see reviews in Efremov 1995; Elmegreen et al.
55: 2000, Elmegreen 2005), and this hierarchy continues even inside the
56: youngest clusters (Testi et al. 2000; Heydari-Malayeri et al. 2001;
57: Nanda Kumar, Kamath, \& Davis 2004; Smith et al. 2005a; Gutermuth et
58: al. 2005; Dahm \& Simon 2005; Stanke, et al. 2006; see review in
59: Allen et al. 2006). Most likely, the hierarchy in stars comes from
60: a hierarchy in the gas (St\"utzki et al. 1998; Dickey et al. 2001),
61: which is the result of turbulence compression and gravitational
62: contraction that is self-similar over a wide range of scales (see
63: review in Mac Low \& Klessen 2004). Clusters form in the densest
64: parts of this gas and lose their initial sub-structure as the
65: stellar orbits mix (for simulations of star formation in clusters,
66: see Klessen \& Burkert 2000; Bonnell, Bate, \& Vine 2003; Li, et al.
67: 2004; Tilley \& Pudritz 2004; Klessen et al. 2005;
68: V\'azquez-Semadeni, Kim, \& Ballesteros-Paredes 2005; Bonnell \&
69: Bate 2006; Li \& Nakamura 2006).
70:
71: Hunter (1999) and Ma\'iz-Apell\'aniz (2001) noted that some massive
72: star-forming regions (which they called ``scaled OB associations'',
73: or SOBA's) do not form dense clusters while others with the same
74: total mass do (e.g., the SSC's). We would like to understand this
75: difference. Obviously the density of the gas is involved, as dense
76: clusters require dense gas, but the distinction between SSCs and
77: SOBAs should also be related to the efficiency of star formation,
78: because clusters forming at low efficiency disperse quickly when the
79: gas leaves (Lada, Margulis \& Dearborn 1984). At very low
80: efficiency, stars form individually without passing through an
81: embedded cluster phase. Recent observations of giant molecular
82: clouds (GMCs) show star formation at both high and low densities,
83: with some embedded stars in dense clusters and others more dispersed
84: (Megeath et al. 2004; J{\o}rgensen et al. 2006, 2007).
85:
86: Larsen \& Brodie (2000) discovered ``faint fuzzies'' at intermediate
87: radii in the disks of the S0 galaxies NGC 1023 and NGC 3384, and
88: suggested they are old clusters with unusually large radii (7-15 pc)
89: and low densities. They are gravitationally bound because of their
90: large ages (8-13 Gyr; Brodie \& Larsen 2002), and they are as
91: massive as SSC's and halo globular clusters. Such clusters appear to
92: represent an intermediate stage between dispersed and bound star
93: formation. They appear to be too low in average density to have had
94: time for core collapse and envelope expansion as in models of
95: globular clusters by Baumgardt et al. (2002). Burkert, Brodie \&
96: Larsen (2005) suggested they formed in a collisional ring
97: interaction between two galaxies.
98:
99: What determines the relative proportion of dispersed and clustered
100: star formation? Larsen \& Richtler (2000) showed that clustering on
101: a galactic scale, measured as the fraction of uv light in the form
102: of massive young clusters, increases as the star formation rate
103: increases. This could be the result of a selection effect if
104: starbursts are active for less than a cluster destruction time. On
105: the other hand, the clustering fraction could depend on pressure.
106: Higher pressure makes the cool phase of gas denser, which promotes
107: more clustering, and locally high pressures trigger star formation
108: on the periphery of GMCs, making clusters in the dense gas (e.g.,
109: Comer\'on, Schneider, \& Russeil 2005; Zavagno et al. 2006). The
110: Larsen \& Richtler correlation could then follow from the mutual
111: correlation between pressure and star formation rate with gas column
112: density. Faint fuzzies are a counter example, however: they formed
113: bound but presumably at low pressure to have such low central column
114: densities now. Can all of these clustering types be understood as a
115: continuum of properties in a universal physical model?
116:
117: There have been several attempts to explain the difference between
118: clustered and distributed star formation based on numerical
119: simulations. Klessen, Heitsch, \& Mac Low (2000) suggested that
120: clusters form in non-magnetic gas when the turbulence driving scale
121: is large. Heitsch, Mac Low \& Klessen (2001) noted that non-magnetic
122: turbulence driven on large scales produces a clustered collapse,
123: while magnetic turbulence in supercritical clouds produces a more
124: distributed collapse. Mac Low (2002) suggested that stars form in
125: clusters when there is no turbulent support and they form disbursed
126: when there is. V\'azquez-Semadeni, Ballesteros-Paredes, \& Klessen
127: (2003) suggest this transition from no global turbulent support to
128: support corresponds to an increase in the Mach number and a decrease
129: in the sonic scale, which is the length where the size-linewidth
130: relation gives a Mach number of unity. Large sonic scale (low Mach
131: number) corresponds to a sonic mass larger than the thermal Jeans
132: mass, which means a lack of global support and the formation of a
133: cluster. Low sonic scale corresponds to the dispersed formation of
134: stars, one for each tiny compressed region where the mass exceeds
135: the local thermal Jeans mass. Li, Klessen \& Mac Low (2003)
136: suggested that the equation of state determines the stellar
137: clustering properties: soft equations produce dense clusters while
138: hard equations produce isolated stars. These suggestions all apply
139: to initially uniform media. External compression of a cloud into a
140: massive dense core can also make a cluster; most young clusters are
141: in high-pressure regions like OB associations.
142:
143: Simulations of turbulent media produce stars in compressed regions
144: that act as seeds for the small-scale gravitational collapses that
145: follow (V\'azquez-Semadeni, Ballesteros-Paredes, \& Klessen 2003;
146: Clark \& Bonnell 2005). These simulations also have probability
147: distribution functions (pdfs) for density that are either log-normal
148: or log-normal with a power-law tail at high density, especially when
149: self-gravity is important (e.g., Li, Klessen \& Mac Low 2003). The
150: efficiency of star formation is then proportional to the fraction of
151: the gas in a dense form. Here we examine variations in this fraction
152: as functions of average density and velocity dispersion, and as a
153: function of the local density inside a cloud. The efficiency is
154: taken to be the ratio of the stellar mass to the gas mass during a
155: complete star-forming event. It generally increases with cloud
156: density from a few percent in GMCs (Williams \& McKee 1997) to
157: several tens of percent in cluster-forming cores (e.g., Lada \& Lada
158: 2003). It may reach $\sim50$\% or more inside the densest
159: star-forming cores. We explain this increase as a result of
160: hierarchical structure, regardless of the dynamics and mechanisms of
161: star formation, and we show that for log-normal or similar density
162: pdf's, as expected in turbulent media, the mass fraction of regions
163: with high efficiency increases with the Mach number and,
164: independently, with the average density. This result may explain the
165: Larsen \& Richtler (2000) correlation as well as the observed
166: variations in clustering properties with pressure. We also show that
167: at high Mach number, bound clusters can form with relatively low
168: average densities, thereby explaining faint fuzzies. These are all
169: consequences of star formation in the dense cores of clouds that are
170: structured by turbulence. They result primarily from the geometry of
171: the gas, which is somewhat universal, and should be nearly
172: independent of the gas dynamics or the strength of the magnetic
173: field.
174:
175: We make an important assumption that gravitational contraction and
176: star formation can occur in regions that are either larger or
177: smaller than the sonic scale. This means we assume that contraction
178: to one or more stars can occur in a supersonically turbulent region.
179: V\'azquez-Semadeni, Ballesteros-Paredes, \& Klessen (2003) suggest
180: that if a cloud is supported by turbulence, then only regions
181: smaller than a sonic scale and more massive than the thermal Jeans
182: mass are unstable to form stars. Padoan (1995) was the first to
183: consider this condition. However, clouds are probably not supported
184: for any significant time by turbulence, and even if they were, it is
185: only necessary that a clump mass exceed the turbulent Jeans mass for
186: self-gravitational forces to exceed inertial forces. GMCs for
187: example, have comparable self-gravitational and turbulent energy
188: densities and yet are much larger than the sonic length. Our
189: assumption is contrary to that in Krumholz \& McKee (2005), who
190: assume the same as V\'azquez-Semadeni et al.. We are consistent
191: with McKee \& Tan (2003), however, as they consider the collapse of
192: a highly turbulent core to make a massive star. Saito, et al.
193: (2006), for example, observe star formation in massive turbulent
194: cores. Thus the sonic length should not provide a threshold for star
195: formation.
196:
197: Our primary condition for star formation is that the gas density
198: exceed some fixed value, taken here to be $10^5$ cm$^{-3}$. This is
199: the density at which HCN gives a nearly constant star formation rate
200: per unit gas mass (Gao \& Solomon 2004a,b; Wu et al. 2005) and at
201: which a variety of microscopic processes conspire to shorten the
202: magnetic diffusion time (Elmegreen 2007). Regions with this density
203: should have a wide range of Mach numbers but a nearly universal
204: efficiency, according to the Gao \& Solomon and Wu et al.
205: observations. The density of $\sim10^5$ cm$^{-3}$, when converted to
206: 5900 M$_\odot$ pc$^{-3}$, is also typical for star clusters, as most
207: of those surveyed by Tan (2007), which span a factor of $10^5$ in
208: mass, have about this average density. The cluster density equals
209: the gas density times the efficiency, and the efficiency has to
210: exceed $\sim10-30$\% for a bound cluster to form. Thus, the gas
211: density for both cluster formation and high efficiency appears to be
212: around $10^5$ cm$^{-3}$ or, possibly, $10^6$ cm$^{-3}$. We note that
213: the long timescale derived for HCN gas as the ratio of the total HCN
214: mass divided by the total star formation rate (Gao \& Solomon 2004b;
215: Wu et al. 2005) is not the duration of star formation in any one
216: place, but is the HCN consumption time. As long as a new HCN region
217: is formed somewhere each time an old HCN region disperses, the HCN
218: consumption time can be long even when each region of star formation
219: lasts for a short time (see Elmegreen 2007 for a discussion of star
220: formation timescales). As a result, the average efficiency of star
221: formation in any one HCN region can be moderately large even if the
222: average HCN consumption time is long. An average efficiency of
223: $\sim5-10$\% would be reasonable considering that most star-forming
224: regions leave unbound clusters after the gas leaves (Lada \& Lada
225: 2003), most regions are observed at half their total ages, and star
226: formation typically accelerates over time (Palla \& Stahler 2000).
227: With constant acceleration, only $\sim1/4$ of the total stars form
228: in the first 1/2 of the total time. The efficiency that determines
229: whether a bound cluster will remain is taken here to be 14.4\% (see
230: below), and the peak efficiency in a single-star core is taken to be
231: 50\% at $10^5$ cm$^{-3}$ density. These values are uncertain and are
232: used here only to illustrate how cluster formation might scale with
233: the velocity dispersion and density of the cool component of the
234: interstellar medium.
235:
236: \section{Cluster Formation in a Hierarchical ISM: Bound and Unbound
237: Clusters} \label{sect:clus}
238:
239: The hierarchical structure of interstellar gas implies that the mass
240: fraction of star-forming cores at high density $n_c$ increases in
241: regions of the cloud that have a higher average density. This
242: density dependence may be illustrated with a simple model. Consider
243: a local star formation rate proportional to $\epsilon(\rho)
244: \rho\left(G\rho\right)^{1/2}$ for $\rho<\rho_{c}$ for efficiency
245: $\epsilon$. In this paper, mass density will be denoted by $\rho$
246: and molecule density by $n$. If we denote the galactic average
247: quantities by a subscript ``0'', then the Kennicutt-Schmidt star
248: formation rate is analogous to $\epsilon_0 \rho_0 (G\rho_0)^{1/2}$
249: for $\epsilon_0\sim0.012$ (Elmegreen 2002a). The same galactic rate
250: would be obtained for observations at any other density, provided
251: the efficiency and volume filling factor at that density are
252: properly scaled. Thus we write:
253: \begin{equation}\epsilon_0 \rho_0 (G\rho_0)^{1/2} =
254: \epsilon(\rho) \rho (G\rho)^{1/2}f_V(\rho) = \epsilon_{c}
255: \rho_{c}(G\rho_{c})^{1/2}f_V(\rho_{c}),\end{equation} where
256: $f_V(\rho)$ is the fraction of the volume having a density larger
257: than $\rho$. This expression with the volume fraction can be
258: converted to one with the mass fraction using the relations $\rho
259: f_V(\rho) = \rho_0f_M(\rho)$ and $\rho_{c} f_V(\rho_{c}) =
260: \rho_0f_M(\rho_{c})$, where $f_M(\rho)$ is the fraction of the mass
261: having a density larger than $\rho$. Subscripts {\it c} denote gas
262: at the threshold density for a constant efficiency. As a result,
263: \begin{equation}\epsilon(\rho) = \epsilon_{c}
264: \left(\rho_{c}/\rho\right)^{1/2} \left[f_M\left(\rho_{c}\right)/
265: f_M\left(\rho\right)\right].\label{eq:eps}\end{equation}
266:
267: Mass fractions and efficiencies for log-normal density pdfs are
268: shown in Figure \ref{fig:fractio3} for $n_c=10^5$ cm$^{-3}$. The
269: decreasing lines are the mass fractions $f_M(\rho)$ at densities
270: larger than the value on the abscissa, and the increasing lines,
271: which are $\epsilon(\rho)$, are $\epsilon_c$($=0.5$) times the mass
272: fractions of the clumps of density $\rho_c$ inside regions of
273: average density $\rho$. For this log-normal case, the normalized pdf
274: is
275: \begin{equation}
276: P(x)={{1}\over{\left(2\pi\right)^{1/2}\sigma}}
277: e^{-0.5\left(x^2-x_p^2\right)/\sigma^2}\end{equation} where
278: $x=\ln\rho$ and $x_p=\ln\rho_p$ at the pdf peak
279: ($\rho_p=\rho_0e^{-\sigma^2/2}$ for average $\rho_0$). The pdf for
280: equal increments in density is $P^\prime\left(\rho\right)=P(x)/\rho$
281: and the mass fraction greater than some density is
282: $f_M(\rho)=\int_\rho^\infty \rho P^\prime\left(\rho\right)d\rho$.
283:
284: The density pdfs in the figure have dispersions and average
285: densities of 2.3 and 1 cm$^{-3}$ for the solid line, 2.3 and 10
286: cm$^{-3}$ for the dashed line, 2.58 and 10 cm$^{-3}$ for the
287: dot-dash line, and 1.98 and 0.5 cm$^{-3}$ for the dotted line. The
288: first of these dispersions comes from galaxy simulations in Wada \&
289: Norman (2001) and from the low density case in Wada \& Norman
290: (2007); the density of 1 cm$^{-3}$ is typical for normal galaxy
291: disks. The second has a higher density, similar to the inner parts
292: of galaxies, and the third has the same high density and a Mach
293: number that is twice the effective value in Wada \& Norman (2001),
294: as determined from the relation between dispersion and Mach number
295: in Padoan \& Nordlund (2002), which is
296: $\sigma=\left(\ln\left[1+0.25M^2\right]\right)^{1/2}$ for Mach
297: number $M$. These latter two cases represent moderately strong
298: starburst regions. The fourth case has a Mach number that is half
299: the Wada \& Norman value and a density that is also small. This case
300: applies to quiescent regions with low-intensity star formation, such
301: as the outer parts of galaxies, parts of dwarf galaxies, and low
302: surface brightness galaxies. Wada \& Norman (2007) simulate
303: self-gravitating disks and find log-normal density pdfs with
304: $\sigma$'s that increase from 2.4 to 3.0 as the initial density
305: (within 10 pc of the midplane) increases from 5 to 50 M$_\odot$
306: pc$^{-3}$. Our $\sigma$ are in this range. The equilibrium peak
307: densities in the Wada \& Norman simulations are all about $\sim2$
308: cm$^{-3}$, which is also comparable to the values used here.
309:
310: The log-normal nature of the assumed density pdf is not critical to
311: the conclusions of this paper. Any density pdf with an extended tail
312: at high density and a breadth that increases with Mach number will
313: give the same correlations between bound cluster fraction, Mach
314: number, and pressure. Similarly, any pdf that has a smaller slope
315: near the average density than at high density (like a log-normal)
316: will give our additional dependence of the bound cluster fraction on
317: the average density. The main point is that when the slope of the
318: density pdf is relatively shallow near the critical density for star
319: formation, bound clusters can form with a wide range of average
320: densities. These types of pdfs will also produce faint fuzzies in
321: the limit of high dispersion as these clusters result from highly
322: efficient star formation at a low average cloud core density. The
323: absolute calibration of the density pdf is also not critical to our
324: model. We could equally well consider pdfs for cluster-forming cores
325: where the average density is $\sim10^2$ cm$^{-3}$ and the critical
326: density for ``final'' collapse is $\sim10^7$ cm$^{-3}$. The
327: essential points are that (1) for all of these pdfs, the mass
328: fraction of the ``final'' collapsed cores, and therefore the
329: efficiency of star formation in a particular region,
330: $\epsilon(\rho)$, increases with the average density of the region
331: (an implication of hierarchical structure), and (2) the density pdf
332: is more or less flat at the critical density of star formation,
333: depending on some physical variable, taken here to be the Mach
334: number and/or average density. Wada \& Norman (2007) point out that
335: there is no single Mach number in their simulations but a range of
336: values, and still the density pdf is log-normal. They find the
337: primary dependence of $\sigma$ to be on the initial disk density.
338: The origin of this $\sigma$ is not important here, only the effect
339: that variable $\sigma$ and $\rho_0$ have on the mass fraction of
340: dense gas.
341:
342: According to the equations which define $\epsilon(\rho)$ for the
343: log-normal model, there is a minimum value in each $\epsilon(\rho)$
344: curve that occurs where $d\ln f_M/d\ln \rho=-0.5$. At lower density,
345: the curve $f_M(\rho)$ flattens and $\epsilon(\rho)$ is dominated by
346: $\rho^{-1/2}$ so it increases with decreasing $\rho$. At higher
347: density, $\epsilon(\rho)$ is dominated by the Gaussian tail in
348: $f_M(\rho)$ so it increases with increasing $\rho$. The low density
349: behavior of $\epsilon(\rho)$ is not realistic because the low
350: density part of $f_M$ is not likely to remain a log-normal. This
351: part corresponds to the low density intercloud medium and the
352: physical processes there are different than in dense gas. It may be
353: controlled by supernova cavities, for example. In Wada \& Norman
354: (2001), the low density part of $f_M$ was not a log-normal. Thus we
355: consider only the increase in $\epsilon(\rho)$ with $\rho$ to be
356: characteristic of dense gas involved with star formation.
357:
358: For each of the cases considered in Figure 1, the efficiency
359: increases with density because hierarchical structure gives $n_{c}$
360: cores a higher filling factor at higher average densities. The mass
361: fraction decreases with density because only a small fraction of the
362: matter is dense. For the fiducial case (solid line), an average
363: galactic efficiency, $\epsilon(\rho_0)=0.01$ (e.g., Kennicutt 1998),
364: is indicated by the cross on the left and a typical efficiency for a
365: dense star-forming core, $\sim50$\% at $\rho_{c}\sim10^5$ $m(H_2)$
366: cm$^{-3}$, is indicated by the cross on the right. The vertical
367: lines show typical ranges for efficiencies and densities in OB
368: associations ($\epsilon\sim1$\% to 5\% at $n\sim10^{3.3}$ cm$^{-3}$)
369: and in the cluster-forming cores of OB associations
370: ($\epsilon\sim10$\% to 30\% at $n\sim10^{4.5}$ cm$^{-3}$).
371:
372: The approximate value of the efficiency separating clusters that end
373: up mostly bound after gas dispersal from clusters that become mostly
374: unbound ($\epsilon_{cluster}$) is placed in the figure at
375: $\epsilon(\rho)\sim0.144$, which is the value of $\epsilon(\rho)$ in
376: the fiducial case at a gas density of $n=10^{4.55}$ cm$^{-3}$. This
377: efficiency is shown by the dotted horizontal line; its exact value
378: is not important here. The efficiency for bound cluster formation
379: depends on the rate of gas dispersal and the initial velocities of
380: the stars. Slower gas dispersal rates and smaller initial speeds
381: compared to virial require lower efficiencies for boundedness (e.g.,
382: Verschueren 1990; Lada \& Lada 2003; Boily \& Kroupa 2003; Goodwin
383: \& Bastian 2006).
384:
385: The basic trends in Figure \ref{fig:fractio3} illustrate the
386: important differences between cluster formation in various
387: environments. First, the threshold efficiency occurs at lower
388: average densities for starburst pdfs and at higher average densities
389: for quiescent pdfs. That is, the density $\rho$ at which
390: $\epsilon(\rho)=0.144$ in the figure is smaller than $10^{4.55}$
391: $m(H_2)$ cm$^{-3}$ in the first case and larger than $10^{4.55}$
392: $m(H_2)$ cm$^{-3}$ in the second. When the average density for
393: $\epsilon>\epsilon_{cluster}$ is much lower than the inner core
394: density $n_{c}$, a high fraction of the gas mass can form stars with
395: $\epsilon>\epsilon_{cluster}$. Long-lived clusters should be easier
396: to form in this case. Because this is the starburst limit, the
397: result is in qualitative agreement with the observations by Larsen
398: \& Richtler (2000).
399:
400: Figure \ref{fig:epsilon} shows this result by plotting as a solid
401: line the density at which $\epsilon(\rho)=0.144$ versus the
402: dispersion $\sigma$ of the log normal density pdf, for an average
403: density of 1 cm$^{-3}$. The threshold density value of $10^{4.55}$
404: cm$^{-3}$ when $\sigma=2.3$ from Figure 1 is one point on the curve.
405: Higher dispersions correspond to lower threshold densities for bound
406: cluster formation. Figure \ref{fig:epsilon} also shows as a dashed
407: line the fraction of the mass, $f_M$, at densities greater than the
408: threshold density as a function of the dispersion. Larger
409: dispersions lead to a larger fraction of the mass in bound clusters,
410: mostly because the threshold density decreases and more gas mass is
411: above the threshold.
412:
413: When the average density for $\epsilon>\epsilon_{cluster}$ is
414: comparable to $n_{c}$, the mass fraction of gas with highly
415: efficient star formation is very low. This case applies to low Mach
416: numbers or low pressures and may explain why in some regions like
417: the giant OB association NGC 604 in M33 there is a lot of star
418: formation and thousands of young stars, but few bound clusters
419: (Ma\'iz-Apell\'aniz 2001).
420:
421: We propose that the difference between the formation of clusters and
422: the formation of loose OB associations or stellar groupings of the
423: same total mass and age depends mostly on the ratio of the
424: characteristic density for stellar core formation, $\rho_c$, to the
425: density at which $\epsilon(\rho)=\epsilon_{cluster}$. High pressure
426: gas has a high density ratio and should form a high proportion of
427: stars in bound clusters. Low pressure gas has a density ratio near
428: unity and should form a low proportion of stars in bound clusters
429: with most forming in loose OB associations.
430:
431: Some GMCs in the solar neighborhood have substantial populations of
432: protostars outside the dense clusters (Megeath et al. 2004).
433: J{\o}rgensen et al. (2006, 2007) found that 30-50\% of the
434: protostars in Perseus are outside the two main clusters. This is to
435: be expected in hierarchical clouds because not all of the
436: star-forming clumps are in the cluster-forming cores. In a high
437: pressure region, a hypothetical Perseus-like collection of
438: protostars would form at a higher average density, putting the
439: individual stars closer together and giving them a higher fraction
440: of the total cloud mass where they form. As a result, they would be
441: more tightly bound to each other when the gas leaves. In a low
442: pressure region, the cores would be further apart with relatively
443: more intercore gas, and they would be more likely to disperse into
444: the field along with the gas. In both cases, a fixed fraction of the
445: turbulence-compressed cores could form stars, but the mass fraction
446: represented by these cores can be high or low, depending on the Mach
447: number and density. The conditions in the Perseus cloud are
448: apparently intermediate between these two limits.
449:
450: Johnstone et al. (2004) found that 2.5\% of the $\rho$ Ophiuchus
451: cloud mass is in dense sub-mm cores, while Kirk et al. (2006) found
452: that 0.4\% of the mass in the Perseus cloud is at $A_V>10$ mag.
453: Enoch et al. (2006) observed $<5$\% of the Perseus cloud mass in 1.1
454: mm cores. These small fractions are consistent with the turbulent
455: fragmentation model of Figure \ref{fig:fractio3}, as shown by the
456: decreasing $f_M(\rho)$ lines; recall that $f_M$ is the fraction of
457: the total mass at density larger than $\rho$ (while $\epsilon(\rho)$
458: is the fraction of the mass having a density greater than fixed
459: $\rho_c$ in regions of density $\rho$). At high density, $f_M(\rho)$
460: is low while $\epsilon(\rho)$ is high. If each sub-mm core evolves
461: on its own dynamical time, and the surrounding cloud does the same
462: as it forms new cores, then the cloud will have a number of
463: core-forming events equal to the square root of the ratio of
464: densities, which is $\sim10$. Thus the net efficiency of core
465: formation can build up to several tens of percent over time, and the
466: final efficiency of star formation might be $\sim10$\% or more,
467: considering the efficiency inside each sub-mm core and the
468: likelihood that some cores will disperse without forming a star
469: (e.g. V\'azquez-Semadeni et al. 2005). A large number of core
470: crossing times is not necessary for this to happen. The increasing
471: $\epsilon(\rho)$ curves in Figure \ref{fig:fractio3} illustrate
472: qualitatively a second point made by these authors, and by
473: J{\o}rgensen et al (2007) as well for Perseus, that the mass
474: fraction of cores becomes high in the generally denser regions of
475: the cloud; i.e., at high extinction. They find an $A_V$ extinction
476: threshold for the occurrence of sub-mm cores equal to 5 to 7 mag (15
477: mag in Johnstone, et al). Figure \ref{fig:fractio3} suggests that it
478: is not the extinction, per se, which produces this correlation, but
479: the hierarchical structure of clouds.
480:
481: A second result from Figures \ref{fig:fractio3} and
482: \ref{fig:epsilon} is that a bound cluster can have a lower average
483: density when the Mach number or ambient density is higher, i.e.,
484: when the pdf is flatter or shifted to higher $\rho$. This suggests
485: an explanation for faint fuzzies: they are normal clusters that form
486: at the limiting lower density for boundedness in high pdf-dispersion
487: regions. The lowest density for a bound cluster is lower when the
488: $\epsilon(\rho)$ curve in Figure \ref{fig:fractio3} is flatter at
489: $\rho_c$. This makes the density at $\epsilon(\rho)=
490: \epsilon_{cluster}$ lower and corresponds to a broader pdf
491: dispersion $\sigma$ in Figure \ref{fig:epsilon}. For an isothermal
492: gas, a broader dispersion corresponds to a higher Mach number. Faint
493: fuzzies also require low background tidal forces or they would have
494: broken apart by now. Usually, high Mach numbers correspond to high
495: critical tidal densities because both occur in starburst regions.
496: But faint fuzzies require an odd combination: high Mach numbers (or
497: broad pdf-dispersions) and low tidal densities. This is possible in
498: highly shocked clouds that occur in remote regions of a galaxy. For
499: example, during a galaxy collision, shocked clouds in the outer
500: regions (or direct ISM impacts in the outer regions) can have Mach
501: numbers approaching the pair orbital speed, but at these locations,
502: the local tidal forces are fairly low. This result fits well with
503: the model of faint fuzzies by Burkert, Brodie \& Larsen (2005), who
504: suggested they formed in galaxy collisional rings. Such rings are
505: indeed highly shocked regions in the outer parts of galaxies, where
506: the tidal density is fairly low. Burkert et al. did not explain how
507: such conditions would produce faint fuzzies, however. Here, they
508: result naturally in a turbulent ISM model as a consequence of the
509: flattening of the density pdf at higher Mach number.
510:
511: \section{Multiple Star Formation Events inside the Hierarchies}
512:
513: The observed hierarchies of star formation, such as subgroups inside
514: OB associations, which, in turn, are inside star complexes, have the
515: property that the duration of star formation increases approximately
516: as the square root of size (Efremov \& Elmegreen 1998) and is always
517: in the range of 1 to 2 crossing times (Elmegreen 2000, 2007). The
518: smaller regions come and go relatively quickly while the larger
519: regions form stars, and there are many small regions within the
520: spatial bounds and during the total time span of the large region.
521: When the duration of star formation is always about a crossing time,
522: then the number of events at the density $\rho$ is proportional to
523: $\left(G\rho\right)^{1/2}$. These small-scale events are not all at
524: the same place, but they move around inside the large scale event as
525: clouds get dispersed and cloud envelopes get triggered.
526:
527: For each large-scale event with cloud mass $M$ ($\sim10^7$ M$_\odot$
528: in main galaxy disks) and average density $\rho_0$, the total mass
529: of all the gas in star forming events at any one time at a density
530: greater than $\rho$ is $Mf_M\left(\rho\right)$ according to the
531: definition of $f_M$ in the previous section. The star formation
532: efficiency in each of these high density regions is
533: $\epsilon\left(\rho\right)$. The number of events of star formation
534: at the density $\rho$ during the lifetime of the larger-scale cloud
535: is $\left(G\rho\right)^{1/2}/\left(G\rho_0\right)^{1/2}$, according
536: to the previous paragraph. The product of these three terms is the
537: total stellar mass formed in all the fragments of the large-scale
538: cloud that ever had a density $\rho$. This product is
539: $Mf_M(\rho)\epsilon(\rho)\left(\rho/\rho_0\right)^{1/2}$.
540: Substituting for $\epsilon(\rho)$ from equation \ref{eq:eps} and
541: using the identity
542: $f_M\left(\rho_c\right)=f_V\left(\rho_c\right)\rho_c/\rho_0$ from
543: the discussion just before equation \ref{eq:eps}, all of the terms
544: involving density cancel and we get simply $\epsilon_0M$. This
545: result confirms the basic nature of hierarchical star formation: the
546: mass of stars forming in each level of the hierarchy is the same and
547: equal to the total mass of stars formed. This is a property of
548: hierarchies because stars which form in the densest level are the
549: same stars as those which form in the lower-density levels that
550: contain the densest levels. The important point for star formation
551: is that this hierarchy is in both space and time. The answer comes
552: out correctly only when the multiple events of star formation at
553: each density are considered.
554:
555: \section{Stellar Mixing and the Formation of Clusters}
556:
557: If all star-forming regions last for about a crossing time before
558: they disperse and reform into other star-forming regions inside the
559: same larger-scale regions (Elmegreen 2000, 2007), then there is an
560: equal opportunity for stars on all scales to mix in their orbits and
561: become smooth clusters. This means that the oldest stars in all
562: groupings, ranging from embedded clusters to giant star complexes,
563: should be fairly well mixed while the youngest stars in these
564: groupings should still be hierarchical. The scale dependence of the
565: mixing fraction has never been observed systematically, but general
566: observations suggest it is approximately true. For example, the
567: youngest objects in the Orion nebula, the proplyds, still cluster
568: near $\theta^1$ Ori C while the slightly older stars are more
569: dispersed (Smith et al. 2005b; Bally et al. 2005). Similarly, the
570: youngest objects in Gould's Belt, which could be the nearest star
571: complex, are clustered into the local OB associations such as Orion,
572: Perseus, and Sco-Cen, while the oldest objects, the Pleiades and
573: Cas-Tau associations, are now dispersed into stellar streams (e.g.,
574: Elias, Alfaro \& Ca\~no 2006). The large regions never get
575: self-bound, unlike the small regions, because the efficiency of star
576: formation is always small on large scales, as shown in Section
577: \ref{sect:clus}. Still these large regions mix through the action of
578: random stellar drift.
579:
580: Galactic shear pulls apart the large-scale groupings over tens of
581: millions of years, which is the crossing time. The shear time is
582: independent of size and equal to the inverse of the Oort A constant,
583: while the crossing time increases with size approximately as the
584: square root. This means there is a sufficiently large size where the
585: crossing time equals the shear time and at larger sizes, the
586: crossing time exceeds the shear time. This critical size is always
587: about the scale height in the galaxy for Toomre $Q\sim1$ (Elmegreen
588: \& Efremov 1996). Larger regions look like star steams or flocculent
589: spiral arms because shear dominates, while smaller regions look
590: clumpy like star complexes and OB associations because the internal
591: dynamics dominates.
592:
593: \section{Discussion}
594:
595: Stars form bound clusters where the total efficiency is high. The
596: efficiency should be proportional to the mass fraction of a cloud in
597: the form of dense cores where the individual stars form. This mass
598: fraction increases with the cloud density in hierarchical clouds. It
599: also increases with the Mach number of the turbulence because
600: stronger shocks at higher Mach numbers compress the gas to a wider
601: range of densities. As a result, a log-normal density pdf becomes
602: flatter below the threshold density of star formation, and this
603: means the mass fraction is higher at each density below this value.
604: When the Mach number is high (really, when the dispersion of the
605: density pdf is high), the efficiency is high even at a fairly low
606: average cloud density, and so a high fraction of star formation ends
607: up bound. An extreme example of this trend is the type of cluster
608: called a faint fuzzie. We propose that faint fuzzies form in
609: moderately low density clouds with moderately high Mach numbers. The
610: galactic tidal ring environment proposed by Burkert et al. (2005) is
611: an example of a region that would have such clouds.
612:
613: The mass fraction of clumps at a particular high density of star
614: formation also increases with the average density of the ISM because
615: then the whole density pdf shifts toward higher values. The
616: combination of high densities and high Mach numbers, characteristic
617: of starburst regions, makes for a high fraction of star formation in
618: bound clusters.
619:
620: On the other hand, relatively low Mach numbers and/or low densities
621: should produce stars in a more dispersed way, as in the low-density
622: regions of molecular clouds or in low surface brightness galaxies
623: and regions of galaxies. This combination of parameters corresponds
624: to a low ISM pressure, so we infer that low pressure regions, which
625: means those with a low gas column densities and low star formation
626: rates per unit area, should produce relatively fewer bound clusters
627: and relatively more unbound associations.
628:
629: There is a physical explanation for the trends discussed here.
630: Consider a moderately low density cloud with a low turbulent Mach
631: number. The compressions inside that cloud will be modest and most
632: of them will not reach a level of density or enhanced magnetic
633: diffusion rate that allows gravitational collapse before the
634: compression ends. Then few stars will form and the efficiency will
635: be low (unless the cloud continues to makes these weak compressions
636: for a very large number of crossing times, which seems unlikely).
637: Now consider this same cloud with the same size and average density
638: but with a higher Mach number (this will require a higher ISM
639: pressure). The stronger compressions will more easily produce
640: high-density cores in which gravity is important and magnetic
641: diffusion is fast. More stars will form and the efficiency will be
642: high. The only difference between these two examples is the Mach
643: number, which affects the range of densities in the compressed
644: regions. In terms of the density pdf, higher Mach numbers produce a
645: broader pdf at the same average density. In terms of star formation
646: in bound clusters, high pressure regions having clouds with high
647: Mach number produce a higher fraction of their stars in bound
648: clusters.
649:
650: The model predicts that young bound clusters in starburst regions,
651: or in regions of high ISM pressures or Mach numbers, will have a
652: wider range of densities than bound clusters in more quiescent,
653: low-pressure regions. This is partly because the cloud densities
654: should be higher in starburst regions, so the resulting clusters can
655: be denser overall, but it is also because the lower-density parts of
656: starburst clouds can form stars with a high efficiency, leaving
657: bound clusters when the gas leaves rather than dispersed OB
658: associations. As there is generally more mass at low density than at
659: high density, the net distribution of cluster density could shift
660: toward lower values when the ISM density pdf is broad. According to
661: the model, this density shift is relative to the mean ISM density,
662: and it should be measured only before significant cluster expansion.
663: Low surface brightness galaxies should make a preponderance of
664: unbound OB associations rather than bound clusters, and the young
665: clusters they form should have a relatively narrow range of
666: densities compared to normal.
667:
668: \section{Summary}
669:
670: Hierarchical structure in both gas and young stars is evident from
671: fractal analysis, correlation functions, and power spectra (see
672: references in Sect. I) and is probably the result of turbulence
673: compression and self-gravity. Stars form in the densest part of the
674: gas. In regions where the individual star-forming clumps represent a
675: high fraction of the mass, the cluster that forms has a good chance
676: of remaining self-bound after the gas leaves. Such regions typically
677: have densities of $\sim10^{4}$ cm$^{-3}$ or more in the solar
678: neighborhood. We suggest here that star formation at a high average
679: density like this automatically has a high efficiency because the
680: individual star-forming clumps are only an order of magnitude or two
681: denser. This follows entirely from the hierarchical ISM. In such a
682: medium, clumps of any density cluster together with clumps of a
683: similar density, and the mass fraction of these clumps increases as
684: the average density of the surrounding region increases. As a
685: result, low density clouds like GMCs form stars with a low total
686: efficiency and high density cloud parts like GMC cores form stars
687: with a high total efficiency, leaving bound or marginally bound
688: clusters.
689:
690: The density where the efficiency first becomes high decreases as the
691: Mach number of the turbulence increases and as the average density
692: of the ISM increases. Consequently, high pressure regions should
693: place a higher fraction of their stars in bound clusters, while low
694: pressure regions should preferentially make unbound stellar
695: groupings. Regions with moderately low density and moderately high
696: Mach number should be able to make faint fuzzies, which are
697: low-density bound clusters. Regions with high densities and low Mach
698: numbers should make extremely dense clusters.
699:
700: The masses of the high density regions and of the star clusters that
701: form in them have not been specified in this derivation of
702: $\epsilon(\rho)$. These masses should vary over a wide range and be
703: distributed approximately as $dN/dM\sim M^{-2}$ because of the
704: hierarchical nature of the gas (Fleck 1996; Elmegreen \& Efremov
705: 1997; Elmegreen 2002b). Thus bound and dispersed stellar groupings,
706: as well as faint fuzzies, should have the usual $M^{-2}$ mass
707: distributions.
708:
709: \acknowledgements
710: My thanks to Mark Gieles for a helpful referee's
711: report.
712:
713: \begin{thebibliography}{}
714:
715: \bibitem[]{637} Allen, L., Megeath, S. T., Gutermuth, R., Myers,
716: P. C., Wolk, S., Adams, F. C., Muzerolle, J., Young, E., \& Pipher,
717: J. L. 2007, in Protostars and Planets V, ed. B. Reipurth, D. Jewitt,
718: \& K. Keil, (Tucson, Univ of Arizona), 361
719:
720: \bibitem[]{642} Bastian, N., Gieles, M., Efremov, Yu. N., \& Lamers,
721: H. J. G. L. M. 2005, A\&A, 443, 79
722:
723: \bibitem[]{645} Bastian, N., Ercolano, B., Gieles, M., Rosolowsky, E.,
724: Scheepmaker, R. A., Gutermuth, R., \& Efremov, Yu. 2007, MNRAS, 379,
725: 1302
726:
727: \bibitem[]{649} Battinelli, P., Efremov, Y., \& Magnier, E.A.
728: 1996, A\&A, 314, 51
729:
730: \bibitem[]{} Baumgardt, H., Hut, P., \& Heggie, D.C. 2002, MNRAS,
731: 336, 1069
732:
733: \bibitem[]{652} Boily, C.M., \& Kroupa, P. 2003, MNRAS, 338, 673
734:
735: \bibitem[]{654} Bonnell, I.A., Bate, M.R., \& Vine, S.G. 2003, MNRAS,
736: 343, 413
737:
738: \bibitem[]{657} Bonnell, I.A., \& Bate, M.R. 2006, MNRAS, 370, 488
739:
740: \bibitem[]{659} Brodie, J.P., \& Larsen, S.S. 2002, AJ, 124, 1410
741:
742: \bibitem[]{661} Burkert, A., Brodie, J., \& Larsen, S. 2005, ApJ, 628,
743: 231
744:
745: \bibitem[]{664} Clark, P.C. \& Bonnell, I.A. 2005, MNRAS, 361, 2
746:
747: \bibitem[]{666} Comer\'on, F., Schneider, N., \& Russeil D., 2005,
748: A\&A, 433, 955
749:
750: \bibitem[]{669} Dahm, S.E., \& Simon, T. 2005, AJ, 129, 829
751:
752: \bibitem[]{673} Dickey, J.M., McClure-Griffiths, N.M., Stanimirovic,
753: S., Gaensler, B.M, \& Green, A.J, 2001, ApJ, 561, 264
754:
755: \bibitem[]{676} Efremov, Y.N. 1995, AJ, 110, 2757
756:
757: \bibitem[]{678} Efremov, Y.N., \& Elmegreen, B.G. 1998, MNRAS, 299,
758: 588
759:
760: \bibitem[]{681} Elias, F., Alfaro, E.J., \& Cabrera-Ca\~no, J. 2006,
761: AJ, 132, 1052
762:
763: \bibitem[]{684} Elmegreen, B.G. 2000, ApJ, 530, 277
764:
765: \bibitem[]{686} Elmegreen, B.G. 2002a, ApJ, 577, 206
766:
767: \bibitem[]{688} Elmegreen, B.G. 2002b, ApJ, 564, 773
768:
769: \bibitem[]{690} Elmegreen, B.G. 2005, in The many scales in the
770: Universe, eds. J. C. del Toro Iniesta, et al. (Dordrecht: Kluwer).
771: p. 99
772:
773: \bibitem[]{694} Elmegreen, B.G. 2007, ApJ, 668, in press
774:
775: \bibitem[]{696} Elmegreen, B. G., Efremov, Y., Pudritz, R. E., \&
776: Zinnecker, H. 2000, in Protostars and Planets IV, eds. V. Mannings,
777: A.P. Boss, \& S.S. Russell, (Tucson: Univ. of Arizona Press), p. 179
778:
779: \bibitem[]{700} Elmegreen, B.G., \& Efremov, Y.N. 1996, ApJ, 466, 802
780:
781: \bibitem[]{702} Elmegreen, B.G., \& Efremov, Y.N. 1997, ApJ, 480, 235
782:
783: \bibitem[]{704} Elmegreen, B.G., Elmegreen, D.M., Chandar, R.,
784: Whitmore, B., Regan, M. 2006, ApJ, 644, 879
785:
786: \bibitem[]{707} Enoch, M.L. 2006. ApJ, 638, 293
787:
788: \bibitem[]{709} Feitzinger, J. V., \& Galinski, T. 1987, A\&A, 179,
789: 249
790:
791: \bibitem[]{} Goodwin, S.P., \& Bastian, N. 2006, MNRAS, 373, 752
792:
793: \bibitem[]{712} Gomez, M., Hartmann, L., Kenyon, S. J. \& Hewett, R.
794: 1993, AJ, 105, 1927
795:
796: \bibitem[]{715} Gutermuth, R.A., Megeath, S.T., Pipher, J.L.,
797: Williams, J.P., Allen, L.E., Myers, P.C., \& Raines, S. N. 2005,
798: ApJ, 632, 397
799:
800: \bibitem[]{719} Heitsch, F., Mac Low, M.-M., \& Klessen, R.S. 2001,
801: ApJ, 547, 280
802:
803: \bibitem[]{722} Heydari-Malayeri, M., Charmandaris, V., Deharveng, L.,
804: Rosa, M. R., Schaerer, D., \& Zinnecker, H. 2001, A\&A, 372, 527
805:
806: \bibitem[]{725} Hunter, D. A. 1999, in IAU Symp. 193, Wolf-Rayet
807: Phenomena in Massive Stars and Starburst Galaxies, ed. K. A. van der
808: Hucht, G. Koenigsberger, \& P. R. J. Eenens (San Francisco: ASP),
809: 418
810:
811: \bibitem[]{730} Ivanov, G. R., Popravko, G., Efremov, Y. N., Tichonov,
812: N. A., \& Karachentsev, I. D. 1992, A\&AS, 96, 645
813:
814: \bibitem[]{733} Johnstone, D., Di Francesco, J., \& Kirk, H. 2004,
815: ApJ, 611, L45
816:
817: \bibitem[]{736} J{\o}rgensen, J.K. et al. 2006, ApJ, 645, 1246
818:
819: \bibitem[]{738} J{\o}rgensen, J.K., Johnstone, D., Kirk, H., \& Myers,
820: P.C. 2007, ApJ, 656, 293
821:
822: \bibitem[]{741} Kennicutt, R.C., Jr. 1998, ApJ, 498, 541
823:
824: \bibitem[]{743} Kirk, H., Johnstone, D., \& Di Francesco, J. 2006,
825: ApJ, 646, 1009
826:
827: \bibitem[]{746} Klessen, R.S., \& Burkert, A. 2000, ApJS, 128, 287
828:
829: \bibitem[]{748} Klessen, R.S., Ballesteros-Paredes, J.,
830: V\'azquez-Semadeni, E., \& Dur\'an-Rojas, C. 2005, ApJ, 620, 786
831:
832: \bibitem[]{799} Lada, C. J., Margulis, M. \& Dearborn, D., 1984, ApJ, 321, 141
833:
834: \bibitem[]{751} Lada, C.J., \& Lada, E.A. 2003, ARAA, 41, 57
835:
836: \bibitem[]{753} Larsen, S.S., \& Richtler, T. 2000, A\&A, 354, 836
837:
838: \bibitem[]{755} Larsen, S.S., \& Brodie, J.P. 2000, AJ, 120, 2938
839:
840: \bibitem[]{757} Li, Y., Klessen, R.S., \& Mac Low, M.-M. 2003, ApJ,
841: 592, 975
842:
843: \bibitem[]{760} Li, P. S., Norman, M,L., Mac Low, M.-M., \& Heitsch,
844: F. 2004, ApJ, 605, 800
845:
846: \bibitem[]{763} Li, Z.-Y., Nakamura, F. 2006, ApJ, 640, L187
847:
848: \bibitem[]{765} Mac Low, M.-M. 2002, in Modes of Star Formation and
849: the Origin of Field Populations, ASP Conf. Ser. 285, eds. E.K.
850: Grebel \& W. Brandner, p. 112
851:
852: \bibitem[]{769} Mac Low, M.-M., \& Klessen, R.S. 2004, RvMP, 76, 125
853:
854: \bibitem[]{771} McKee, C.F., \& Tan, J.C. 2003, ApJ, 585, 850
855:
856: \bibitem[]{773} Ma\'iz-Apell\'aniz, J. 2001, ApJ, 563, 151
857:
858: \bibitem[]{775} Megeath, S.T., et al. 2004 ApJS, 154, 367
859:
860: \bibitem[]{777} Nanda Kumar, M. S., Kamath, U. S., Davis, C. J. 2004,
861: MNRAS, 353, 1025
862:
863: \bibitem[]{780} Nordlund, \AA., \& Padoan, P. 1999, in Interstellar
864: Turbulence, ed. J. Franco \& A. Carrami\~nana (Cambridge: Cambridge
865: Univ. Press), 218
866:
867: \bibitem[]{} Palla, F., \& Stahler, S.W. 2000, ApJ, 540, 255
868:
869: \bibitem[]{784} Saito, H., Saito, M., Moriguchi, Y., \& Fukui, Y.
870: 2006, PASJ, 58, 343
871:
872: \bibitem[]{787} Scalo, J.S. 1985, in Protostars and Planets II, ed.
873: D.C Black and M. S. Matthews, (Tucson: Univ. of Arizona Press), p.
874: 201
875:
876: \bibitem[]{791} Smith, M.D., Gredel, R., Khanzadyan, et al. 2005a,
877: MmSAI, 76, 247
878:
879: \bibitem[]{794} Smith, N., Bally, J., Shuping, R. Y., Morris, M., \&
880: Kassis, M. 2005b, AJ, 130, 1763
881:
882: \bibitem[]{797} Stanke, T., Smith, M.D., Gredel, R., \& Khanzadyan, T.
883: 2006, A\&A, 447, 609
884:
885: \bibitem[]{800} St\"utzki, J., Bensch, F., Heithausen, A., Ossenkopf,
886: V., \& Zielinsky, M. 1998, A\&A, 336, 697
887:
888: \bibitem[]{853} Tan, J.C. 2007, in Triggered star formation in a turbulent interstellar
889: medium, IAU Symposium 237, eds. B.G. Elmegreen \& J. Palo\v{s},
890: (Cambridge: Cambridge Univ Press), p.258
891:
892: \bibitem[]{803} Testi, L., Sargent, A.I., Olmi, L et al. 2000, ApJ,
893: 540, 53
894:
895: \bibitem[]{806} Tilley, D.A., \& Pudritz, R.E. 2004, MNRAS, 353, 769
896:
897: \bibitem[]{808} V\'azquez-Semadeni, E., Ballesteros-Paredes, J. \&
898: Klessen, R.S. 2003, ApJ, 585, L131
899:
900: \bibitem[]{811} V\'azquez-Semadeni, E., Kim, J., Shadmehri, M., \&
901: Ballesteros-Paredes, J. 2005, ApJ, 618, 344
902:
903: \bibitem[]{814} V\'azquez-Semadeni, E., Kim, J., \&
904: Ballesteros-Paredes, J. 2005, ApJ, 630, L49
905:
906: \bibitem[]{817} Verschueren, W. 1990, A\&A, 234, 156
907:
908: \bibitem[]{819} Wada,~K., \& Norman,~C.~A. 2001, ApJ, 547, 172
909:
910: \bibitem[]{821} Wada,~K., \& Norman,~C.~A. 2007, ApJ, 660, 276
911:
912: \bibitem[]{823} Williams, J.P. \& McKee, C.F. 1997, ApJ, 476, 166
913:
914: \bibitem[]{} Wu, J., Evans, N.J., II., Gao, Y., Solomon, P.M.,
915: Shirley, Y.L., \& Vanden Bout, P.A. 2005, ApJ, 635, L173
916:
917: \bibitem[]{825} Zavagno, A., Deharveng, L., Comer\'on, F., Brand, J.,
918: Massi, F., Caplan, J., Russeil, D. 2006, A\&A, 446, 171
919:
920: \end{thebibliography}
921:
922: \clearpage
923:
924: \begin{figure}\epsscale{.75}
925: \plotone{f1.eps}\caption{Left-hand axis. and downward sloping lines:
926: The fraction of the mass in a turbulent region, $f_M$, modeled with
927: a log-normal density pdf, that has a density greater than the value
928: on the abscissa. Right-hand axis and upward sloping lines: the local
929: efficiency $\epsilon$ of star formation as a function of average
930: local density. Observed efficiencies for OB associations and for
931: clusters are shown by the vertical lines. The efficiency is taken to
932: be the fraction of the cloud mass having a density greater than
933: $10^5$ cm$^{-3}$, multiplied by the efficiency of star formation at
934: the threshold, which is 50\% here. The different line types
935: correspond to normal galaxies (solid line, $[\sigma,n_0]=[2.3,1]$),
936: starbursts (dashed $[2.3,10]$ and dot-dashed $[2.56,10]$) and
937: quiescent regions (dotted $[1.98,0.5]$). Here $\sigma$ is the
938: dispersion of the log-normal pdf and $n_0$ is the average density.
939: The mass fraction of gas with large $\epsilon$ ($>0.144$ in this
940: example) is larger in starburst regions than it is in normal
941: galaxies; it is lower in quiescent regions. This implies that
942: starbursts should form most of their stars in bound clusters, while
943: quiescent regions should form most of their stars in unbound
944: associations. The figure also shows that the minimum density for
945: bound clusters decreases with increasing Mach number; faint fuzzies
946: may form as an extreme example of this.}\label{fig:fractio3}
947: \end{figure}
948:
949: \begin{figure}\epsscale{.9}
950: \plotone{f2.eps}\caption{(Left-hand axis, solid line) The density at
951: a fixed threshold efficiency for cluster formation,
952: $\epsilon(\rho)=\epsilon_{cluster},$ is shown as a function of the
953: dispersion $\sigma$ of the density pdf for cloud structure.
954: (Right-hand axis, dashed line) The mass fraction of gas at a density
955: greater than the threshold density for cluster formation, versus
956: $\sigma$. The average gas density is taken to be 1 cm$^{-3}$,
957: consistent with the solid lines in Fig. 1. The trends observed here
958: arise for a wide range of threshold efficiencies, so the precise
959: value is not important; $\epsilon_{cluster}=0.144$ from Fig. 1 was
960: used. The figure indicates that ISM regions with higher dispersions
961: for their density pdfs (e.g., higher Mach numbers) have lower
962: relative density thresholds for bound cluster formation and higher
963: fractions of the total mass forming bound
964: clusters.}\label{fig:epsilon}\end{figure}
965:
966: \end{document}
967: