1: %\documentclass{emulateapj}

2: 

3: \NeedsTeXFormat{LaTeX2e}[1996/06/01]

4: 

5: \documentclass[12pt,preprint]{aastex}

6: %\usepackage{epsfig}

7: \def\lsim{\raise0.3ex\hbox{$<$}\kern-0.75em{\lower0.65ex\hbox{$\sim$}}}

8: \def\gsim{\raise0.3ex\hbox{$>$}\kern-0.75em{\lower0.65ex\hbox{$\sim$}}}

9: 

10: \begin{document}

11: \title{Magnetically Regulated Star Formation in 3D: The Case of 

12: Taurus Molecular Cloud Complex} 

13:  

14: \author{Fumitaka Nakamura\altaffilmark{1} and Zhi-Yun Li\altaffilmark{2}} 

15: %

16: \altaffiltext{1}{Faculty of Education, Niigata University,

17: 8050 Ikarashi-2, Niigata 950-2181, Japan; fnakamur@ed.niigata-u.ac.jp}

18: \altaffiltext{2}{Department of Astronomy, University of Virginia,

19: P. O. Box 400325, Charlottesville, VA 22904; zl4h@virginia.edu}

20: 

21: \begin{abstract}

22: We carry out three-dimensional MHD simulations of star formation

23: in turbulent, magnetized clouds, including ambipolar diffusion

24: and feedback from protostellar outflows. The calculations focus

25: on relatively diffuse clouds threaded by a strong magnetic field 

26: capable of resisting severe tangling by turbulent motions and 

27: retarding global gravitational contraction in the cross-field 

28: direction. They are motivated by observations of the Taurus 

29: molecular cloud complex (and, to a lesser extent, Pipe Nebula), 

30: which shows an ordered large-scale magnetic field, as 

31: well as elongated condensations that are generally perpendicular 

32: to the large-scale field. We find that stars form in earnest in 

33: such clouds when enough material has 

34: settled gravitationally along the field lines that the 

35: mass-to-flux ratios of the condensations approach the critical 

36: value. Only a small fraction (of order $1\%$ or less) of the 

37: nearly magnetically-critical, 

38: condensed material is turned into 

39: stars per local free-fall time, however. The slow star formation 

40: takes place in condensations that are moderately supersonic; 

41: it is regulated primarily by magnetic fields, rather than turbulence. 

42: The quiescent condensations are surrounded by diffuse halos 

43: that are much more turbulent, as observed in the Taurus complex. 

44: Strong support for magnetic regulation of star formation in 

45: this complex comes from the extremely slow conversion of the 

46: already condensed, relatively quiescent C$^{18}$O gas into 

47: stars, at a rate two orders of magnitude below the maximum, 

48: free-fall value.

49: We analyze the properties of dense 

50: cores, including their mass spectrum, which resembles the 

51: stellar initial mass function. 

52: \end{abstract}

53: 

54: \keywords{ISM: clouds --- ISM: magnetic fields --- MHD --- stars: formation

55: --- turbulence}

56: 

57: 

58: \section{Introduction}

59: \label{intro}

60: 

61: Stars form in molecular clouds that are both turbulent and magnetized. 

62: The relative importance of the turbulence and magnetic field in 

63: controlling star formation is a matter of debate (McKee \& Ostriker 

64: 2007). Early quantitative studies have concentrated on the formation 

65: and evolution of individual (low-mass) cores out of quiescent, 

66: magnetically supported clouds (Nakano 1984; Shu et al. 1987; 

67: Mouschovias \& Ciolek 1999). Recent numerical simulations have 

68: focused more on the role of turbulence in the dynamics of larger 

69: clouds and core formation in them (as reviewed in, e.g., Mac Low 

70: \& Klessen 2004). Ultimately, the debate must be settled by 

71: observations, especially direct measurements of the magnetic 

72: field strength in the bulk of the molecular gas, which are 

73: generally difficult to do (e.g., Heiles \& Crutcher 2005). In the 

74: absence of such measurements, we have to rely on indirect evidence.

75: 

76: The best observed region of active star formation is arguably 

77: the nearby Taurus molecular cloud complex. Although it is not 

78: clear whether the complex is magnetically supported or not as 

79: a whole, the more diffuse regions are probably magnetically 

80: dominated. The strongest evidence comes from thin strands of 

81: $^{12}$CO emission that are aligned with the directions of 

82: the local magnetic field (Heyer et al. 2008), as traced by 

83: polarization of background 

84: star light. The magnetic field in the strands is apparently

85: strong enough to induce a measurable difference between the

86: turbulent velocities along and perpendicular to the field

87: direction. Heyer et al. concluded that the field strength 

88: is probably high enough to render the relatively diffuse 

89: striated region subcritical, with a magnetic flux-to-mass 

90: ratio greater than the critical value $2\pi G^{1/2}$ (Nakano 

91: \& Nakamura 1978). The CO striations are strikingly 

92: similar to those observed in the nearby Riegel-Crutcher HI 

93: cloud, mapped recently by McClure-Griffiths et al. (2006), 

94: using 21cm absorption against the strong continuum emission

95: towards the galactic center region. Its hair-like strands 

96: are also along the directions of the local magnetic field, 

97: again traced by the polarization vectors of the background 

98: star light. The authors estimated a field strength of $\sim 

99: 30$~$\mu$G inside the strands. Kazes \& Crutcher (1986) 

100: measured the line-of-sight field strength at a nearby 

101: location, and found $B_{los}\sim 18~\mu$G, which is 

102: consistent with the above estimate if a correction of $\sim 

103: 2$ is applied for projection effect. These are clear examples 

104: of magnetic domination in, respectively, cold HI

105: % (see also Orion Veil, Brogan et al.??) 

106: and relatively diffuse molecular clouds. 

107: 

108: It is unclear, however, how representative these two clouds are. 

109: In particular, the degree of magnetization is uncertain in 

110: giant molecular clouds (GMCs), where the majority of stars 

111: form. Elmegreen (2007) argued that the envelopes of GMCs, where 

112: most of the molecular gas resides, are probably magnetically 

113: critical or even subcritical, and can be long-lived if not for 

114: disruption by rapid star formation in GMC cores, which are 

115: probably supercritical, with internal dynamics dominated by 

116: turbulence; this supposition needs to be tested by magnetic 

117: field observations on GMC scales (e.g., Novak, et al. 2007). 

118: Nevertheless, the 

119: Taurus cloud complex and the Riegel-Crutcher cloud are probably  

120: best mapped cloud in CO and HI, respectively. How such relatively 

121: diffuse, apparently magnetically dominated clouds proceed to 

122: condense and collapse to form stars is the focus of our paper. 

123: 

124: Diffuse clouds are expected to condense more easily along than 

125: across the field lines as long as they are dominated by ordered 

126: large-scale magnetic fields. Such an anisotropic condensation 

127: can increase the density greatly with little corresponding 

128: increase in the magnetic field strength. This expectation is 

129: consistent with the available Zeeman measurements of field 

130: strength, which is more or less constant below a number density 

131: of order $10^3$~cm$^{-3}$ (Heiles \& Crutcher 2005). Above this 

132: density, the field strength tends to increase with density, indicating 

133: that contraction perpendicular to the field lines becomes 

134: significant. We interpret the break in the observed field 

135: strength-density relation as marking the point where the 

136: self-gravity of the cloud becomes dynamically important in the 

137: cross-field direction. Before this point is reached, one expects 

138: the self-gravity to first become significant along the field 

139: lines (where there is no magnetic support), pulling matter into 

140: condensations, which can be either a sheet, a filament, 

141: or even a knot, depending on the degree of anisotropy in the 

142: initial mass distribution. 

143: 

144: In the case of the Taurus molecular cloud complex, most of the 

145: dense molecular gas traced by $^{13}$CO and especially C$^{18}$O 

146: are distributed

147: in structures elongated more or less perpendicular to the 

148: large-scale magnetic field (e.g., Onishi et 

149: al. 1996, 2002; Goldsmith et al. 2008). It has long been suspected 

150: that they have condensed along the field lines (e.g., Heyer et 

151: al. 1987; Tamura et al. 1987). Cross-field contraction must have 

152: already taken place in the condensed structures, at least locally, 

153: since stars have been forming in these structures for at least

154: a few million years (Kenyon \& Hartmann 1995). Palla \& Stahler

155: (2002) examined the star formation pattern in the Taurus

156: clouds in both space and time, concluding that stars began 

157: to form at a low level at least 10 million years ago, in 

158: a spatially dispersed fashion. The majority of the stars are

159: formed, however, in the last 3 million years, near where the 

160: dense gas is observed today. This pattern of accelerating 

161: star formation is precisely what is expected of a magnetically

162: dominated cloud that condenses in two stages: first along the

163: field line, when little star formation is expected, except 

164: perhaps in some pockets that have an exceptionally weak local 

165: magnetic field to begin with or have their magnetic fluxes 

166: reduced by an exceptionally strong compression-enhanced ambipolar 

167: diffusion, and then across the field, when more active star

168: formation can take place throughout the dense sheets and 

169: filaments that have condensed along the field lines. A strong 

170: support for this supposition comes from the fact that star 

171: formation is slow even in dense gas: the rate of star 

172: formation for the last 3 million years is about ${\dot M}_*

173: \sim 5\times 10^{-5}~M_\odot$yr$^{-1}$ (Goldsmith et al. 2008), 

174: two orders of magnitude below the free-fall rate for the dense 

175: ($\sim 10^4$~cm$^{-3}$) C$^{18}$O gas (Onishi et al. 2002; see 

176: \S~\ref{taurus} for actual numbers); Krumholz \& Tan (2006) found 

177: that a similar result holds for objects of a wide range of density. 

178: Although the observed moderately supersonic motions (${\cal M}\sim 

179: 2$) can provide some support for the dense gas, we believe that 

180: the star formation is slow mainly because the bulk of the dense 

181: gas remains magnetically supported even after cross-field 

182: contraction has begun in localized regions in the condensed 

183: structures. 

184: This requirement can naturally be satisfied if such structures 

185: are marginally magnetically critical (Basu \& Ciolek 2000), 

186: where individual stars can form on a relatively short time 

187: scale, but the overall rate of star formation remains well 

188: below the free-fall value, as we have demonstrated using 

189: simulations in 2D sheet-like geometry (Li \& Nakamura 2004 

190: and Nakamura \& Li 2005; LN04 and NL05 hereafter; see also 

191: Kudoh et al. 2007 and Kudoh \& Basu 2008). The Taurus cloud 

192: complex is discussed further in \S~\ref{taurus}, along 

193: with the Pipe Nebula (\S~\ref{pipe}), which may represent an 

194: earlier phase of star formation in a magnetically dominated 

195: cloud (Lada et al. 2008; F. Alves et al., in preparation).   

196: 

197: The rest of the paper is organized as follows. Section~\ref{setup} 

198: describes the model formulation, including governing equations, 

199: initial and boundary conditions, and numerical code. It is followed

200: by three sections on numerical results. Section~\ref{result} concentrates 

201: on a standard model with a particular combination of initial turbulence 

202: and outflow feedback. It illustrates the essential features of the 

203: magnetically regulated star formation in three dimensions, including 

204: strong stratifications in density and turbulent speed, broad 

205: probability distribution functions (PDFs) of volume and column 

206: densities and, most importantly, low rate of star formation. This 

207: model is contrasted, in \S~\ref{other}, with others that have 

208: different levels of initial turbulence and outflow feedback. We 

209: devote \S~\ref{core} to a detailed analysis of dense cores, including

210: their shapes, mass spectrum, velocity dispersions, flux-to-mass

211: ratios, angular momenta, and whether they are bound or not. In 

212: \S~\ref{discussion}, we outline a general scenario of magnetically

213: regulated star formation in quiescent condensations of relatively

214: diffuse, turbulent clouds, and make connection to observations 

215: of the Taurus molecular cloud complex and Pipe Nebula. The main 

216: results of the paper are summarized in \S~\ref{conclude}. Readers 

217: interested only in the general scenario and their connection to 

218: observations can skip to \S~\ref{discussion}.

219: 

220: \section{Model Formulation}

221: \label{setup}

222: 

223: \subsection{Governing Equations}

224: 

225: The basic equations that govern the evolution of isothermal, weakly 

226: ionized, magnetized molecular gas are (see, e.g., Shu 1991)

227: \begin{equation}

228: {\partial \rho_n \over \partial t} + \nabla \cdot (\rho_n

229:  \mbox{\boldmath{$V$}}_n) = 0 \ ,

230: \end{equation}

231: \begin{equation}

232: \rho_n {\partial \mbox{\boldmath{$V$}}_n \over \partial t} +

233: \rho_n (\mbox{\boldmath{$V$}}_n \cdot \nabla \mbox{\boldmath{$V$}}_n)

234: = -\rho_n \nabla \Psi - c_s^2 \nabla \rho_n + {1 \over 4 \pi}

235: (\nabla \times \mbox{\boldmath{$B$}}) \times \mbox{\boldmath{$B$}} \ , 

236: \end{equation}

237: \begin{equation}

238: {\partial \mbox{\boldmath{$B$}} \over \partial t}

239: =\nabla \times (\mbox{\boldmath{$V$}}_i \times \mbox{\boldmath{$B$}}) \ ,

240: \end{equation}

241: \begin{equation}

242: \nabla ^2 \Psi = 4 \pi G \rho_n  \ , 

243: \end{equation}

244: where 

245: $\rho$, $\mbox{\boldmath{$V$}}$, $\mbox{\boldmath{$B$}}$, $c_s$,

246: and $\Psi$ are, respectively, 

247: density, velocity, magnetic field, isothermal sound speed, 

248: and gravitational potential, with the subscripts $n$ and $i$ 

249: denoting neutrals and ions. The drift velocity between ion 

250: and neutral is 

251: \begin{equation}

252: \mbox{\boldmath{$V$}}_i - \mbox{\boldmath{$V$}}_n = {t_c \over 4 \pi

253:  \rho_n} (\nabla \times \mbox{\boldmath{$B$}}) \times 

254: \mbox{\boldmath{$B$}}  \ , 

255: \end{equation}

256: where the coupling time between the magnetic field and neutral matter,

257: $t_c$, is given by 

258: \begin{equation}

259: t_c = {1.4 \over \gamma C \rho_n^{1/2}},

260: \end{equation}

261: in the simplest case where the coupling is provided by ions

262: that are well tied to the field lines and the ion density $\rho_i$

263: is related to the neutral density by the canonical expression

264: $\rho_i = C \rho_n^{1/2}$. We adopt an ion-neutral drag coefficient 

265: $\gamma = 3.5\times 10^{13}$~cm$^3$~g$^{-1}$~s$^{-1}$ and $C = 3 

266: \times 10^{-16}$~cm$^{-3/2}$~g$^{1/2}$ (Shu 1991). The factor 1.4 

267: in the above equation comes from the fact that the cross section 

268: for ion-helium collision is small compared to that of ion-hydrogen 

269: collision (Mouschovias \& Morton 1991). These governing equations 

270: are solved numerically subject to a set of initial and boundary 

271: conditions. 

272: 

273: \subsection{Initial and Boundary Conditions}

274: 

275: Our simulations are carried out in a cubic box (of size $L$), with

276: standard periodic conditions imposed at the boundaries. The cloud 

277: is assumed to have a uniform density $\rho_0$ initially. The 

278: corresponding Jeans length is 

279: \begin{equation} 

280: L_J= \left({\pi c_s^2/G \rho_0}\right)^{1/2}, 

281: \end{equation}

282: where the isothermal sound speed $c_s=1.88\times 10^4(T/10 

283: K)^{1/2}$~cm/s, with $T$ being the cloud temperature. The 

284: initial density $\rho_0 = 4.68\times 10^{-24} n_{H_2,0}$~g~cm$^{-3}$, 

285: where $n_{H_2,0}$ is the initial number density of molecular 

286: hydrogen, assuming 1 He for every 10 H atoms. 

287: 

288: We consider relatively diffuse molecular clouds, with a fiducial 

289: H$_2$ number density of $250$~cm$^{-3}$. It is the value adopted 

290: by Heyer et al. (2008) for the sub-region in Taurus that shows 

291: magnetically aligned $^{12}$CO striations. Scaling the density by 

292: this value, we have 

293: \begin{equation}

294: L_J=1.22 \left(\frac{T}{10 K}\right)^{1/2} 

295: \left(\frac{n_{H_2,0}}{250\ {\rm cm}^{-3}}\right)^{-1/2} {\rm pc}, 

296: \end{equation}

297: and a Jeans mass 

298: \begin{equation} 

299: M_J= \rho_0 L_J^3 = 31.6 M_\odot \left(\frac{T}{10 K}\right)^{3/2} 

300: \left(\frac{n_{H_2,0}}{250\ {\rm cm}^{-3}}\right)^{-1/2}.

301: \end{equation} 

302: The Jeans length is smaller than the dimensions of molecular cloud 

303: complexes such as the Taurus clouds, which are typically tens of 

304: parsecs. Although global simulations on the complex scale are

305: desirable, they are prohibitively expensive if individual star 

306: formation events are to be resolved at the same time. After some 

307: experimentation, we settled on a moderate size for the simulation 

308: box, with an initial Jeans number $n_J=L/L_J=2$. It represents 

309: a relatively small piece of a larger complex. Scaling the Jeans 

310: number by 2, we have a box size 

311: \begin{equation}

312: L=2.44 \left(\frac{n_J}{2}\right) \left(\frac{T}{10 K}\right)^{1/2} 

313: \left(\frac{n_{H_2,0}}{250\ {\rm cm}^{-3}}\right)^{-1/2} {\rm pc}, 

314: \end{equation}

315: and a total mass 

316: \begin{equation} 

317: M_{tot}= 253\ M_\odot \left(\frac{n_J}{2}\right)^3 \left(\frac{T}{10 

318: K}\right)^{3/2}\left(\frac{n_{H_2,0}}{250\ {\rm cm}^{-3}}\right)^{-1/2}

319: \end{equation} 

320: in the computation domain. Even though the initial Jeans number

321: adopted is relatively small, there is enough material inside 

322: the box to form many low-mass stars, especially in condensations 

323: where the Jeans mass is much reduced. 

324: 

325: We impose a uniform magnetic field along the $x$ axis at the beginning 

326: of simulation. The field strength $B_0$ is specified by the parameter 

327: $\alpha$, the ratio of magnetic to thermal pressure, through 

328: \begin{equation}

329: B_0=3.22~\alpha^{1/2} \left({T\over 10 K}\right)^{1/2}

330: \left(\frac{n_{H_2,0}}{250\ {\rm cm}^{-3}}\right)^{1/2} (\mu~G). 

331: \label{fieldstrength}

332: \end{equation}

333: In units of the critical value $2\pi G^{1/2}$ (Nakano \& Nakamura

334: 1978), the flux-to-mass ratio for the material inside the box is 

335: \begin{equation}

336: \Gamma_0 = 0.45 n_J^{-1} \alpha^{1/2}.

337: \label{centralgamma}

338: \end{equation}

339: Since we are interested in magnetic regulation of star formation,

340: we choose a relatively strong magnetic field with $\alpha=24$, 

341: corresponding to a mildly magnetically subcritical region with 

342: $\Gamma_0=1.1$ (or a dimensionless mass-to-flux ratio $\lambda

343: =0.91$) for $n_J=2$, which is close to the value of 1.2

344: used in the 2D simulations of LN04 and NL05. For the fiducial 

345: values for temperature and density, the initial field strength 

346: is $B_0=15.8\ \mu$G according to equation~(\ref{fieldstrength}); 

347: this value is not unreasonably high (see \S~\ref{taurus} for 

348: a discussion of the magnetic field in the Taurus region). We 

349: postpone an investigation of initially supercritical clouds to 

350: a future publication.  

351: 

352: Molecular clouds are highly turbulent, particularly for the 

353: relatively diffuse gas that we have assumed for the initial 

354: state. It is established that supersonic turbulence decays 

355: rapidly, with or without a strong magnetic field (e.g., Mac 

356: Low \& Klessen 2004). How the turbulence is maintained is 

357: uncertain (e.g., McKee \& Ostriker 2007). For the relatively 

358: small, parsec-scale region that is modeled here, a potential 

359: source is energy cascade from larger scales, which can 

360: in principle keep the turbulence in the sub-region for a 

361: time longer than the local crossing time. Ideally, this 

362: (and may other) form of driving should be included in a 

363: model, although it is unclear how this can be done in  

364: practice. We will take the limit that the turbulence decays 

365: freely, except for the feedback from protostellar outflows 

366: that are associated with forming stars. Other forms of 

367: driving will be considered elsewhere. Following the standard 

368: practice (e.g., Ostriker et al. 2001), at the beginning of 

369: the simulation ($t=0$) we stir the cloud with a turbulent 

370: velocity field of power spectrum $v_k^2 \propto k^{-3}$ in 

371: Fourier space.  

372: 

373: A useful time scale for cloud evolution is the global gravitational 

374: collapse time 

375: (Ostriker et al. 2001)

376: \begin{equation}

377: t_g= {L_J \over c_s} = 6.36 \times 10^6 \left(\frac{n_{H_2,0}} 

378: {250\ {\rm cm}^{-3}}\right)^{-1/2} ({\rm years}), 

379: \label{gravtime}

380: \end{equation}

381: which is longer than the free-fall time at the initial density, 

382: $t_{ff,0}=[3\pi/(32 G \rho_0)]^{1/2}$, by a factor of 3.27. In 

383: the presence of a strong magnetic field, as is the case in our

384: simulation, gravitational condensation is expected to occur 

385: preferentially along field lines, creating a flattened, sheet-like,

386: structure in which most star formation activities take place. 

387: Inside the sheet, the characteristic Jeans length is 

388: \begin{equation}

389: L_{s} \equiv {c_s^2 \over G \Sigma_0} = {1 \over \pi n_J} L_J 

390: =  0.195 \left(\frac{2}{n_J}\right) \left(\frac{T}{10 K}\right)^{1/2} 

391: \left(\frac{n_{H_2,0}}{250\ {\rm cm}^{-3}}\right)^{-1/2} {\rm pc}\  ,

392: \end{equation}

393: corresponding to a local gravitational collapse time 

394: \begin{equation}

395: t_{s} \equiv {L_{s} \over c_s} = {c_s \over G \Sigma_0} 

396: ={1 \over \pi n_J} t_g \ = 1.01 \times 10^6 \left(\frac{2}{n_J}\right) 

397: \left(\frac{n_{H_2,0}} {250\ {\rm cm}^{-3}}\right)^{-1/2} ({\rm years}) ,

398: \end{equation}

399: where the column density $\Sigma_0$ is measured along the initial

400: magnetic field lines, $\Sigma_0 = \rho_0 L = n_J \rho_0 L_J$.

401: The sheet has a Jeans number $n_{s}\equiv L/L_{s} = \pi n_J^2$, 

402: corresponding to $12.6$ for $n_J=2$, which is close to the value 

403: of 10 adopted in the 2D simulations of LN04 and NL05. Note that 

404: the gravitational collapse time $t_s$ is longer than the free-fall 

405: time of the condensed sheet 

406: \begin{equation}

407: t_{\rm ff,s}=\left({3\pi \over 32 G \rho_s}\right)^{1/2},

408: \label{freefalltime}

409: \end{equation}

410: by a factor of 2.31, where the characteristic sheet density is

411: given by  

412: \begin{equation}

413: \rho_s = \frac{\pi G \Sigma_0^2}{2 c_s^2} = \frac{\pi^2 n_J^2}{2} 

414: \rho_0 \ .

415: \end{equation}

416: 

417: \subsection{Numerical Code}

418: 

419: We solve the equations that govern the cloud evolution using a 3D MHD 

420: code based on an upwind TVD scheme, with the ${\nabla\cdot {\bf B}}=0$ 

421: condition enforced through divergence cleaning after each time step. 

422: The base code is the same as the one  

423: used in the cluster formation simulations of Li \& Nakamura (2006) 

424: and Nakamura \& Li (2007), except that we have now included ambipolar 

425: diffusion in the induction equation. The ambipolar diffusion term is 

426: treated explicitly, which puts a stringent requirement on the time 

427: step (since it is proportional to the square of the grid size, Mac 

428: Low et al. 1995), especially in the lowest density regions. To 

429: alleviate the problem, we set a density threshold, $\rho_{AD}=0.1 

430: \rho_0$, below which the rate of ambipolar diffusion is reduced to 

431: zero. A similar approach was taken by Kudoh et at. (2007), and can 

432: be justified to some extent by increased ionization of low (column) 

433: density gas from UV background (McKee 1989). Even with this measure, 

434: it still takes more than a month to run a standard $128^3$ simulation 

435: for two global gravitational collapse time $t_g$ (or $4\pi$ times the 

436: sheet collapse time $t_s$) on the vector machine available to us. 

437: Since our focus is on the evolution of, and star formation in, the 

438: condensed gas sheet, to speed up the simulations, we follow the 

439: initial phase of cloud evolution on a coarser grid of $64^3$ up to 

440: a maximum density of $\sim 40~\rho_0$, before switching to the 

441: $128^3$ grid. To gauge the numerical diffusion of magnetic field on 

442: such a relatively coarse grid, we have followed the evolution of a 

443: model cloud in the ideal MHD limit (with the ambipolar diffusion 

444: term turned off; Model I0 in Table \ref{tab:model parameter}) up 

445: to $2 t_g (=12.6t_s)$, and found that the mass-to-flux ratio is 

446: conserved to within $1.5\%$, 

447: which seems to be much better than the two-dimensional 

448: higher resolution runs with ZEUS ($512^3$, Krasnopolsky \& Gammie 2005). 

449: 

450: A difficulty with grid-based numerical codes such as ours in 

451: simulating star formation in turbulent clouds is that, once 

452: the collapse of a piece of the cloud runs away to form the 

453: first star, the calculation grinds to a halt without special 

454: treatment. Because the bulk of the stellar mass is believed 

455: to be assembled over a rather brief (Class 0) phase of $\sim

456: 3\times 10^4$~yrs (Andre et al. 1993) and the outflows are 

457: most active during this phase (much shorter than the fiducial 

458: characteristic time $t_s\sim 10^6$~yrs), we idealize the 

459: processes of individual star formation and outflow ejection 

460: as instantaneous, and treat them using the following simple 

461: prescriptions. When the density in a cell reaches a threshold 

462: value $\rho_{th}=10^3\rho_0$, we define around it a critical 

463: core using a version of CLUMPFIND algorithm 

464: (Williams, de Geus, \& Blitz 1994; 

465: see \S~\ref{core} for detail). The core identification 

466: routine is specified by a minimum threshold density $\rho_{\rm 

467: th, min}$ and the number of density levels $N$. They are set 

468: to $200~\rho_0$ and 5 respectively. The mean density of the 

469: core so identified is estimated at $\sim 10^5$~cm$^{-3}$, 

470: comparable to those of H$^{13}$CO$^+$ cores in Taurus studied 

471: by Onishi et al. (2002). We extract $30\%$ of the mass from 

472: the core (if its mass exceeds the local Jeans mass), and put 

473: it in a Lagrangian particle located at the cell center; its 

474: subsequent evolution is followed using the standard leap-frog 

475: method. The extracted mass fraction is 

476: motivated by the observations of J. Alves et al. (2007), 

477: who found a mass spectrum for dense cores that is similar to 

478: the stellar IMF in shape but more massive by a factor of $\sim 3$ 

479: (see also Ikeda, Sunada, \& Kitamura 2007). 

480: It is within the range of $25-75\%$ for the star formation

481: efficiency per dense core estimated theoretically by Matzner \& 

482: Mckee (2000). 

483: 

484: The material remaining in the core after mass extraction is 

485: assumed to be blown away in a two-component outflow: a bipolar 

486: jet of $30^\circ$ half-opening angle around the local magnetic 

487: field 

488: direction that carries a fraction $\eta$ of the total momentum, 

489: and a spherical component outside the jet that carries the 

490: remaining fraction. The total outflow momentum is assumed to 

491: be proportional to the stellar mass, with a proportionality 

492: coefficient of $P_*=35$~km/s. 

493: The prescriptions for star formation 

494: and outflow are similar to those used in Nakamura \& Li (2007). 

495: One difference is the use of CLUMPFIND here to define the critical 

496: core. Another is that, when an outflow reenters the computation 

497: box because of periodic boundary condition, we reduce its strength 

498: by a factor of ten if its speed exceeds $50~c_s$, to prevent 

499: it from strongly perturbing or even disrupting the condensed 

500: sheet. 

501: 

502: \section{Standard Model} 

503: \label{result}

504: 

505: We have run four models that include ambipolar diffusion, with 

506: parameters listed in Table \ref{tab:model parameter}. They

507: differ in random realization of the initial velocity field, 

508: relative strength of the jet and spherical outflow components,

509: and level of initial turbulence. An ideal MHD run was also 

510: performed for comparison. In this section, we focus on the 

511: model with initial turbulent Mach number ${\cal M}=3$ and 

512: jet momentum fraction $\eta=75\%$. It serves as the standard 

513: against which other models, to be discussed in the next 

514: section, are compared. 

515: 

516: The initial phase of cloud evolution is similar to those already 

517: presented in Krasnopolsky \& Gammie (2005), who carried out 

518: high-resolution, ideal MHD simulations of magnetically subcritical 

519: clouds in two dimensions, including self-gravity and an initially 

520: supersonic turbulence that decays freely, as in our simulation. 

521: The initial turbulence creates many shocklets that are oriented 

522: more or less perpendicular to the field lines; they appear as 

523: filaments of enhanced density and are the main sites of turbulence 

524: dissipation. As the turbulence decays, the fragments start to 

525: coalesce, eventually forming a single, approximate Spitzer sheet. 

526: In the ideal MHD limit, the sheet formation marks the end of 

527: cloud evolution. With ambipolar diffusion, it signals the beginning 

528: of a new phase---star formation\footnote{For a strong enough 

529: turbulence and/or initially inhomogeneous distribution of 

530: mass-to-flux ratio, stars can form before the formation of a 

531: well-defined sheet (see \S~\ref{discussion} for further

532: discussion).}. 

533: This latter phase is the focus of our discussion next.  

534: 

535: \subsection{Star Formation Efficiency}

536: \label{SFE}

537: 

538: We begin the discussion with a global measure of star formation in the

539: standard model: the efficiency of star formation, defined as the ratio 

540: of the mass of all stars

541: to the total mass of stars and gas. The star formation efficiency (SFE 

542: hereafter) is plotted in Fig.~\ref{fig:sfe} as a function of time in 

543: units of the gravitational collapse time of the sheet $t_s$. Sheet 

544: formation 

545: is finished by the end of the first global collapse time $t_g=2\pi\ 

546: t_s$. Soon after, the first star appears, around $t \approx 7~t_s$. 

547: By the end of simulation at $t=11.4~t_{s}$ (nearly $2~t_g$), 37 stars 

548: have formed, yielding an accumulative SFE of 6.5\%. The average mass 

549: of a star is close to $0.45 M_\odot \left(T/10 K\right)^{3/2}\left(n_{H_2,0}

550: /250 {\rm cm}^{-3} \right)^{-1/2}$, typical of low-mass stars. The 

551: rate of star formation is given by the slope of the SFE curve, which 

552: is estimated at $\sim 1.5\%$ per $t_s$. At this 

553: rate, it would take $\sim 65~t_s$ to deplete the gas due to star 

554: formation. The gas depletion time corresponds to $\sim 10~t_g$ or 

555: $\sim 33$ times the free-fall time $t_{{\rm ff},0}$ at the {\it 

556: initial} (uniform) density $\rho_0$. In other words, 

557: the star formation efficiency per {\it initial} free fall time 

558: (Krumholz \& McKee 2005) is SFE$_{{\rm ff},0}\sim 3\%$. Interestingly,

559: this rate is comparable to that of cluster formation in 

560: protostellar outflow-driven turbulence (Nakamura \& Li 2007). 

561: However, a large fraction of the gas is at a much higher density

562: than the initial density $\rho_0$ after the formation of the 

563: condensed sheet. It is more appropriate to measure the star 

564: formation rate using the free fall time at the characteristic 

565: sheet density $\rho_s=2 \pi^2\ \rho_0$, which is $t_{{\rm ff},s}= 

566: 0.43\ t_s$ (see equation~[\ref{freefalltime}]), yielding 

567: SFE$_{{\rm ff},s}\sim 0.7\%$. To help understand the rather slow 

568: and inefficient star formation, we next examine in some detail 

569: the cloud structure and dynamics at $t=11~t_s$, representative of 

570: the post-sheet formation phase of cloud evolution when stars 

571: are forming steadily, starting with overall mass distribution 

572: as revealed by column density. 

573: 

574: \subsection{Column Density Distribution}

575: 

576: The cloud material is expected to condense along the magnetic field

577: lines into a flattened structure. This expectation is borne out by

578: Figs.~\ref{colden_y} and \ref{colden_critical}, which show the 

579: distributions of column density along, respectively, the $y$- and 

580: $x$-axis (the direction of the initial magnetic field). Viewed

581: edge-on, the structure resembles a knotty, wiggly filament 

582: (Fig.~\ref{colden_y}). The dense filament is sandwiched by a 

583: low-density ``halo,'' which contains a number of spurs more or

584: less perpendicular to the filament. The spurs are roughly aligned 

585: with the global magnetic field. They resemble the diffuse, fuzzy 

586: $^{12}$CO streaks that are frequently observed to stream away 

587: from dense filaments (Goldsmith et al. 2008). Similar spurs are 

588: seen in the recent SPH-MHD simulations of Price \& Bate (2008). 

589: 

590: The face-on view of the condensed structure is quite different 

591: (Fig.~\ref{colden_critical}). It appears highly fragmented, with 

592: low-density voids as well as high-density peaks. Some 

593: over-dense regions cluster together to form a short filament; 

594: others are more scattered. The actual appearance of the cloud 

595: to outside observers will depend, of course, on the viewing 

596: angle. It can be filament-like (as in Fig.~\ref{colden_y}) or, 

597: more likely, patchy (as in Fig.~\ref{colden_critical}). Superposed 

598: on the face-on column density map is the distribution of the 

599: 33 stars that have formed up to the time plotted. With few 

600: exceptions, they follow the dense gas closely. Most of the stars

601: are found in small groups, which is broadly consistent with the 

602: distribution of young stars in Taurus (Gomez et al. 1993). Since 

603: the clouds is initially 

604: magnetically subcritical everywhere, in order for any localized 

605: region to collapse and form stars, it must become supercritical 

606: first. 

607: %

608: % plots recentered, and in units of Lg rather than Ls

609: %

610: 

611: A rough measure of magnetic criticality is provided by the 

612: average mass-to-flux ratio $\bar{\lambda}$ normalized to

613: the critical value,

614: \begin{equation}

615: \bar{\lambda}(y,z) = \int_{-L/2}^{+L/2} 2\pi G^{1/2}

616: \rho/B_x dx.  

617: \label{AverageLambda}

618: \end{equation}

619: Contours of $\bar{\lambda}=1$ are plotted in Fig.~\ref{colden_critical}, 

620: showing that most of the dense regions are indeed magnetically 

621: supercritical, and thus capable of forming stars in principle. 

622: The supercritical regions enclosed within the contours contain 

623: a significant fraction ($34\%$) of the total mass. 

624: The mass fraction of supercritical gas increases gradually with 

625: time initially, and remains more or less constant at later times 

626: ($t\gtrsim 6~t_s$), as shown in Fig. \ref{fig:supercritical}.

627: This value provides an upper limit to the efficiency of 

628: star formation.

629: If the supercritical regions collapse to form stars within a local 

630: free fall time, the star formation rate would be much higher 

631: than the actual value. An 

632: implication is that the bulk of the supercritical material 

633: remains magnetically supported to a large extent and is rather 

634: long lived. Only its densest parts (i.e., dense cores), 

635: which contain a small fraction of the supercritical mass, 

636: directly participate in star formation at any given time. 

637: The longevity is characteristic of the (mildly) supercritical 

638: regions that are produced through turbulence-accelerated 

639: ambipolar diffusion in nearly magnetically critical clouds. 

640: It is in contrast with the highly supercritical material 

641: in weakly magnetized clouds, which is expected to collapse 

642: promptly unless supported by turbulence. This same behavior 

643: was found in 2D calculations (NL05). 

644: 

645: 

646: To quantify the mass distribution further, we plot in Fig.~\ref{colden_PDF} 

647: the probability distribution function (PDF) of the column density 

648: along the $x-$axis, in the direction of initial magnetic field. The 

649: PDF peaks around $\Sigma \sim 1.7~\rho_0 L_J$ (where $\rho_0$ and 

650: $L_J$ are the density and Jeans length of the initial, uniform state), 

651: close to the average value of $2~\rho_0 L_J$. The distribution can be 

652: described reasonably well by a lognormal distribution, although there 

653: is significant deviation at both low and high density end. The 

654: considerable spread in column density may be surprising at the 

655: first sight, since one may expect little structure to develop in 

656: a subcritical sheet dominated by a strong, more or less uniform 

657: magnetic field, especially since the initial mass distribution is 

658: also chosen to be uniform. This expectation is consistent with the 

659: column density PDF of the ideal MHD counterpart of the standard 

660: model (plotted in the same figure for comparison), which is much 

661: narrower. The key difference is that ambipolar diffusion turns an 

662: increasingly large fraction of the initially subcritical cloud into 

663: supercritical material (until saturation sets in at late times, 

664: see Fig.~\ref{fig:supercritical}). Once produced, the supercritical 

665: material behaves drastically differently from the subcritical 

666: background. It is easier to compress by an external flow because 

667: its magnetic field is weaker. The compressed supercritical region

668: also rebounds less strongly, because the magnetic forces driving 

669: the rebound are cancelled to a larger extent by the self-gravity 

670: of the region. More importantly, the supercritical regions can 

671: further condense without external compression, via self-gravity. 

672: The accumulation of supercritical material and subsequent  

673: (magnetically-diluted) gravitational fragmentation in such material 

674: are responsible for the broad column density PDF. 

675: These key aspects of the problem are {\it not} captured by the 

676: ideal MHD simulation, but are well illustrated in 2D simulations 

677: that include ambipolar diffusion (e.g., NL05; Kudoh \& Basu 

678: 2006). Other aspects are better discussed in 3D, however. These 

679: include the dependence of various physical quantities on volume 

680: density, to which we turn our attention next. 

681: 

682: \subsection{Density Stratification} 

683: 

684: Our model cloud is strongly stratified in density. It contains dense 

685: cores embedded in a condensed sheet which, in turn, is surrounded 

686: by a more diffuse halo. The relative amount of mass distributed at 

687: different densities is displayed in Fig.~\ref{lognorm_density},

688: along with that for the ideal MHD case and for a Spitzer sheet. 

689: Whereas the mass distribution for the ideal MHD case resembles 

690: that of the Spitzer sheet closely, the same is not true for 

691: the standard model with ambipolar diffusion: its mass distribution 

692: is considerably broader, with more material at both the low and 

693: high density end. The former indicates that the diffuse material 

694: outside the sheet is dynamically active, as we show below. The 

695: excess at the high density end is produced, on the other hand, 

696: primarily by ambipolar diffusion, which enables pockets of 

697: the sheet material to condense across magnetic field lines 

698: gravitationally. The bulk ($62.5\%$) of the cloud material 

699: resides in the condensed sheet, which we define somewhat 

700: arbitrarily as the material in the density range between 

701: $5~\rho_0$ and $50~\rho_0$, corresponding to a range in 

702: H$_2$ number density between $1.25\times 10^3$~cm$^{-3}$ and 

703: $1.25\times 10^4$~cm$^{-3}$ (assuming $n_{H_2,0}=250$~cm$^{-3}$). 

704: The remaining cloud material is shared between the diffuse halo 

705: ($30.0\%$), defined 

706: loosely as the material with density below $5\ \rho_0$, and 

707: dense cores ($7.5\%$), defined as the material denser than

708: $50 \rho_0$ (see \S~\ref{core} below). 

709: 

710: Our model cloud is also strongly stratified in turbulent speed. 

711: Fig.~\ref{rms_vel} shows the mass weighted rms velocity as a 

712: function of density. At densities corresponding to the 

713: condensed sheet ($5-50~\rho_0$), the turbulent Mach number is 

714: about 2. The moderately supersonic turbulence persists in the 

715: condensed sheet for many local free fall times, despite the 

716: absence of external driving. It is driven by a combination 

717: of outflows (through either direct impact or MHD waves) and 

718: ambipolar diffusion-induced gravitational acceleration. 

719: The modest increase in rms velocity toward the high density 

720: end is because the denser gas is typically closer to star 

721: forming sites and is thus more strongly perturbed by outflows. 

722: The much higher level of turbulence in the diffuse halo is 

723: more intriguing. One may expect little turbulence to remain at 

724: the representative time $t=11~t_s$ 

725: shown in the figure, which is well beyond the time ($t\sim 6\ t_s$) 

726: of sheet formation --- a product of turbulence dissipation. The 

727: expectation is consistent with the ideal 

728: MHD simulation, which shows that, in the absence of external

729: driving, the turbulent motions die down quickly everywhere. 

730: The turbulence decay is changed by ambipolar diffusion 

731: which, accelerated by the initial strong turbulence, enables 

732: stars to form soon after sheet formation. The stars stir 

733: up the halo via protostellar outflows in at least two ways. 

734: First, the outflows are powerful enough to disrupt the 

735: star-forming cores, break out from the condensed sheet, and 

736: inject energy and momentum directly into the halo. Second, a 

737: fraction of the outflow energy and momentum is also 

738: deposited in the sheet, where gas motions can shake the 

739: field lines and send into the halo MHD waves whose amplitudes 

740: grow as the density decreases (Kudoh \& Basu 2003); the waves 

741: help maintain the turbulence in the diffuse region, particularly 

742: in the cross-field direction. 

743: 

744: 

745: Even though the turbulence speed is relatively high in the halo, 

746: it is still sub-Alfv\'enic. This is shown in 

747: Fig.~\ref{ave_MA}, where the average Alfv\'en Mach 

748: number $M_A$ is plotted as a function of density. The flow 

749: is moderately sub-Alfv\'enic at densities between $\sim 0.3$ 

750: and $\sim 10~\rho_0$, with $M_A$ within $20\%$ of $0.5$. The

751: Alfv\'en Mach number 

752: drops slowly towards lower densities, indicating that the 

753: lower-density gas becomes increasingly more magnetically

754: dominated in our model cloud. A caveat is that, in real 

755: clouds such as the Taurus complex, the dynamics of the 

756: low-density halo on the parsec-scale of our modeled region 

757: are expected to be tied to that of the larger complex. It 

758: is likely for the diffuse gas to receive energy cascaded 

759: from larger scales and become more turbulent than suggested 

760: by Fig.~\ref{ave_MA}. Some of this energy may even be fed 

761: through Alfv\'en waves into the condensed sheet and increase 

762: its velocity dispersion as well. The increase is not expected 

763: to be large, however, since the Alfv\'en speed becomes 

764: transonic at high densities (see Fig.~\ref{ave_MA}), where 

765: highly supersonic motions should damp out quickly through 

766: shock formation. In future refinements, it will be 

767: desirable to include external driving of the halo material, 

768: which may modify the way stars form, especially at early 

769: times (see discussion in \S~\ref{discussion}). Nevertheless, 

770: we believe that the decrease of turbulent speed with increasing 

771: density found in our simulation (and earlier in Kudoh \& 

772: Basu 2003) is a generic feature of the magnetically 

773: regulated star formation in three dimensions. It is broadly 

774: consistent with the observed trend (e.g., Myers 1995). 

775: 

776: Another generic feature is the distribution of magnetic field 

777: strength as a function of density. The distribution is shown 

778: in Fig.~\ref{ave_B}. There is a clear trend for the average 

779: field strength to remain more or less constant at relatively 

780: low densities, before increasing gradually towards the high 

781: density end. A similar trend is inferred 

782: in Zeeman measurements of field strength over a wide range of 

783: densities, and has been interpreted to mean that the initial 

784: stage of cloud condensation leading to star formation is 

785: primarily along the field lines, which increases the density 

786: without increasing the field strength (Heiles \& Crutcher 

787: 2005). This is precisely what happens in our model, 

788: where the initial accumulation of diffuse material into the 

789: condensed sheet is mostly along the field lines, and further 

790: condensation inside the sheet is primarily in the cross-field 

791: direction. The cross-field condensation leads to an increase 

792: in field strength with density, although by a factor that is 

793: less than that expected from flux-freezing because of ambipolar 

794: diffusion. 

795: %

796: 

797: \subsection{Sheet Fragmentation: Turbulent or Gravitational?} 

798: %

799: 

800: The star formation activities in our model are confined mostly to 

801: the condensed sheet. Its dynamics hold the key to understanding the 

802: magnetically regulated star formation. An overall impression of

803: the sheet can be gained from Figs.~\ref{colden_y} and 

804: \ref{colden_critical},

805: which showed an edge-on and face-on view of the sheet in column 

806: density. The sheet has a highly inhomogeneous 

807: mass distribution which is, of course, a pre-requisite for active

808: star formation. To illustrate the clumpy mass distribution further, 

809: we plot in Fig.~\ref{den_pot_vel} the distribution of the density  

810: in the ``mid-plane'' of the sheet, defined as the $yz$ plane 

811: (perpendicular to the initial magnetic field along the $x$ axis) 

812: that passes through the minimum of gravitational potential. The 

813: density distribution spans seven orders of magnitude, from 

814: $10^{-4}~\rho_0$ (the density floor adopted in the computation) to 

815: $10^3~\rho_0$ (the threshold for Lagrangian particle generation).

816: The question is: how is the density inhomogeneity created and

817: maintained? 

818: 

819: A widely discussed possibility is turbulent fragmentation, where 

820: regions of high densities are created by converging flows (e.g., 

821: Padoan \& Nordlund 2002; Mac Low \& Klessen 2004). There is indeed 

822: a random velocity field in the mid-plane, as shown by arrows in 

823: the figure. However, the flow speed is typically small, comparable 

824: to the sound speed in most places. 

825: It is unlikely for such a weak turbulence to create the high density 

826: contrast shown in Fig.~\ref{den_pot_vel} by itself. This conclusion 

827: is strengthened by the presence of a strong magnetic field, which 

828: cushions the flow compression. Indeed, the turbulent speeds in the 

829: plane of the sheet are significantly below the approximate

830: magnetosonic speed $v_{ms}=[B_x^2/(4\pi\rho)+c_s^2]^{1/2}$ in

831: most places, and are thus not expected to create large density 

832: fluctuations; they are essentially a collection of magnetosonic 

833: waves. The low level of turbulence points to gravity rather than 

834: flow motions as the primary driver of fragmentation in the sheet. 

835: %

836: % (1) potential loop holes... what if large compression in individual 

837: % regions involving a small fraction of mass??? i.e., tail effect???

838: % 

839: % Indeed the case at freshly shocked sites by active outflows

840: %

841:  

842: 

843: Gravity is expected to be dynamically important in the sheet, a structure 

844: formed by gravitational condensation along field lines in the 

845: first place. The fact that the sheet has a dimension much larger than

846: the thickness indicates that it contains many thermal Jeans masses. 

847: If the thermal pressure gradient was the main force that opposes 

848: the gravity, one would expect the bulk of the sheet to collapse 

849: in one gravitational time $t_s$. And yet, the vast majority of 

850: the sheet material remains uncollapsed after at least $\sim 5~t_s$. 

851: The longevity indicates that the 

852: gravitational fragmentation proceeds at a much reduced 

853: rate. The reduction is due, of course, to magnetic forces, 

854: which cancel out the gravitational forces in the plane of 

855: the sheet to a large extent (Shu \& Li 1997, Nakamura et al. 

856: 1999). The near cancellation 

857: can be inferred from Fig.~\ref{den_pot_vel}, where contours of

858: gravitational potential are plotted along with the density

859: and velocity vectors. If the gravitational forces were 

860: unopposed, one would expect the material to slide down the 

861: potential well, and picking up speed towards the bottom, 

862: with the infall speed reaching highly supersonic values. 

863: For example, for a potential drop of 8~$c_s^2$ (say between 

864: contours labeled  ``$-8$'' and ``$-16$''), an increase in speed 

865: by 4~$c_s$ is expected. These expectations are not met. The 

866: actual velocity field appears quite random, and generally 

867: transonic. The slow random motions in a rather deep 

868: gravitational potential well provide a clear indication 

869: for magnetic regulation of the sheet dynamics. Active 

870: outflows also play an important role in generating the density 

871: inhomogeneities observed in the condensed sheet. Their effects 

872: are discussed in \S~\ref{outflow} below. 

873: 

874: \subsection{Comparison to 2D Calculations}

875: \label{comparison}

876: 

877: %

878: There are a number of differences between the 3D simulation and the 

879: previous 2D calculations. The 2D treatment does not account for the 

880: mass stratification along the field lines, and thus cannot 

881: address the important issue of dynamical coupling between the 

882: condensed material and the diffuse halo. A related difference is

883: that, in 3D, there is an initial phase of mass settling to form the 

884: sheet, which is is absent in 2D. Furthermore, the mass of each star 

885: in the 2D models is held fixed (typically at $0.5~M_\odot$); in 3D, 

886: it is fixed at $30\%$ of the mass of the criticl core that forms 

887: the star. In addition, the prescription for outflow

888: \footnote{We note that the spherical component of the outflow 

889: in the standard 3D model carries a momentum close to  

890: (within $15\%$ of)  that of the 2D outflow in the plane of 

891: mass distribution in the standard model of NL05.}

892:  and the degree 

893: of initial cloud magnetization are also somewhat different. Despite 

894: these differences, the results of the 2D and 3D models are in 

895: remarkably good agreement. 

896: 

897: The most important qualitative agreement is that, once settled 

898: along the magnetic field lines, the condensed sheet material 

899: produces stars at a slow rate. In the standard 2D model of NL05, 

900: about $5.3\%$ of the total mass is converted into stars within 

901: a time interval of $3.35~t_s$ (see their Fig.~1), corresponding 

902: to a star formation rate of $\sim 1.6\%$ per sheet gravitational 

903: collapse time $t_s$. This rate is nearly identical to the value 

904: of $\sim 1.5\%$ estimated in \S~\ref{SFE} for the standard 3D 

905: model (which is also close to that of the 3D ${\cal M}=10$ model, 

906: see \S~\ref{other} below). The (face-on) mass distributions 

907: are also similar in 2D and 3D models. In both cases, dense, 

908: star-forming cores are embedded in moderately supercritical 

909: filaments which, in turn, are surrounded by more diffuse, 

910: magnetically subcritical material (compare 

911: Fig.~\ref{colden_critical} with the lower panels of 

912: Fig.~2 of NL05). More quantitatively, the mass fraction of 

913: supercritical material increases gradually initially and 

914: reaches a plateau of $\sim 30\%$ in both 2D and 3D models. The 

915: similarity extends to the turbulent speed of the condensed 

916: material, which is about twice the sound speed in both cases 

917: at late times (see Fig.~\ref{rms_vel} and Fig.~11 of NL05). 

918: 

919: The similarities indicate that 2D calculations capture the 

920: essence of the slow, magnetically regulated star formation 

921: in moderately subcritical clouds. The same conclusion was

922: drawn by Kudoh et al. (2007) based on 3D simulations of 

923: weakly perturbed magnetically subcritical sheets. The 

924: 2D calculations have the advantage of being much faster to 

925: perform. They can be done with higher spatial resolution 

926: ($512^2$ as compared to the $128^3$ used in this paper) 

927: and in larger numbers. The number of 3D simulations is 

928: necessarily more limited, especially because of the 

929: small time step required by our treatment of ambipolar 

930: diffusion using an explicit method. Nevertheless, we are 

931: able to perform several additional simulations, to check 

932: the robustness of the standard model, and to help elucidate 

933: the effects of two key ingredients: initial turbulence and 

934: outflow feedback.  

935: 

936: \section{Other Models} 

937: \label{other}

938: 

939: \subsection{Initial Turbulence}

940: \label{turbulence}

941: 

942: Even though the initial turbulence dissipates quickly, it seeds 

943: inhomogeneities in mass distribution and mass-to-flux ratio 

944: that are expected to affect star formation later on. To illustrate 

945: the effects of initial turbulence, we focus on a model identical 

946: to the standard model, except for a higher initial turbulent Mach 

947: number ${\cal M}=10$ (instead of $3$). The overall pattern of 

948: cloud evolution follows that seen 

949: in the standard model: dissipation of turbulence leads to 

950: gravitational settling of cloud material onto a condensed sheet, 

951: the densest parts of which collapse to form stars; the stars, 

952: in turn, stir up the sheet and its surrounding halo through 

953: protostellar outflows. One difference is that the stronger 

954: turbulence initiates star formation at an earlier time, as seen 

955: in Fig.~\ref{sfetwo}. The first star forms around $\sim 4.5~t_s$ 

956: in the ${\cal M}=10$ case, compared to $\sim 6.8~t_s$ for 

957: ${\cal M}=3$. Nevertheless, their rates of star formation are 

958: comparable, with values between $\sim 0.5\%$ and $\sim 1.0\%$ 

959: per free fall 

960: time $t_{ff,s}=0.43\ t_s$ in both cases. The earlier start, but 

961: still low rate, of star formation in the more turbulent case 

962: is consistent with the turbulence-accelerated, magnetically 

963: regulated scenario that we advocate based on previous 2D 

964: calculations (LN04, NL05; see also recent 3D simulations of 

965: Kudoh \& Basu 2008). 

966: 

967: The basic structure and kinematics of the initially more turbulent 

968: cloud are also similar to those of the standard model cloud. In 

969: both cases, the condensed sheet is highly fragmented but dynamically 

970: ``cold,'' with moderately supersonic random motions. It is surrounded 

971: by a halo moving at highly supersonic, but sub-Alfv\'enic, speeds. 

972: An interesting difference is that the sheet in the ${\cal M}=10$ 

973: case appears thicker, as illustrated in Fig.~\ref{colden_y_M10},

974: where an edge-on view of the column density distribution (along 

975: the $y$-axis, perpendicular to the initial magnetic field 

976: direction) is plotted for a representative time $t=8~t_s$. It is 

977: to be compared with Fig.~\ref{colden_y} for the standard model. 

978: The main reason for the larger apparent thickness is that the 

979: condensed sheet, while still thin intrinsically, is more warped. 

980: An example of warps is shown in Fig.~\ref{den_Bfield_M10}, where 

981: the density distribution in a plane perpendicular to the $y$-axis 

982: is plotted; superposition of such warps in different parts of 

983: the sheet along the line of sight gives rise to the more 

984: puffed-up appearance.

985: 

986: Associated with the sharp warp of condensed material in 

987: Fig.~\ref{den_Bfield_M10} is a strong 

988: distortion of magnetic field lines. Both are likely driven by 

989: protostellar outflows, which are the main source of energy 

990: that drives the cloud evolution at late times. In particular, 

991: outflows ejected at an angle to the large-scale magnetic field 

992: can excite large-amplitude Alfv\'en waves, which perturb the 

993: field lines in the halo that thread other parts of the condensed 

994: sheet. The large Alfv\'en speed in the diffuse halo allows 

995: different parts of the sheet to communicate with one another 

996: quickly, as emphasized recently by Shu et al. (2007) in the 

997: context of poloidally magnetized accretion disks. The 

998: communication raises the interesting possibility of global, 

999: magnetically mediated, oscillations for the condensed material 

1000: that may have observational implications. This issue warrants 

1001: a closer examination in the future.  

1002: 

1003: We conclude by stressing that condensations produced by gravitational

1004: settling along field lines do not have to appear highly elongated, 

1005: even when viewed edge-on. They can be further distorted by outflows 

1006: from stars formed in neighboring condensations or other external 

1007: perturbations that are not included in our current calculations. We 

1008: have experimented with simulations that allow active outflows to 

1009: re-enter the simulation box, and found that the condensed sheet 

1010: became more puffed up and, sometimes, even completely disrupted. 

1011: 

1012: \subsection{Outflow Feedback}

1013: \label{outflow}

1014: 

1015: Protostellar outflow plays an important role in our model. It 

1016: limits the amount of core mass that goes into a forming star 

1017: and injects energy and momentum back to the cloud. In our 

1018: standard model, we have assumed that $75\%$ of the outflow 

1019: momentum is carried away in a bipolar jet (of $30^\circ$ in

1020: half opening angle), and the 

1021: remaining $25\%$ in a spherical component. Here, we contrast 

1022: the standard model with one that has a stronger spherical 

1023: component (carrying $75\%$ of outflow momentum). The star

1024: formation efficiencies (SFEs) of both models are plotted 

1025: in Fig.~\ref{sfedependm3}. It is clear that the stronger 

1026: spherical component of outflow has reduced the rate (and

1027: efficiency) of star formation significantly, by a factor of 

1028: $\sim 2.3$. This trend is the opposite of what we found in 

1029: the case of cluster formation in dense clumps that are

1030: globally magnetically supercritical (Nakamura \& Li 2007):  

1031: for a given total outflow momentum, more spherical outflows 

1032: tend to drive turbulence on a smaller scale that dissipates 

1033: more quickly, leading to a higher rate of star formation. 

1034: 

1035: There are a couple of reasons for the seemingly contradictory 

1036: result. First, whereas the supercritical cluster-forming 

1037: clumps are supported mainly by turbulent motions, the bulk 

1038: of the star-forming condensations in subcritical or nearly 

1039: critical clouds are magnetically supported (in the cross-field 

1040: direction). As a result, faster dissipation of 

1041: turbulence leads to more rapid star formation in the former, 

1042: but not necessarily in the latter. Second, the jet component, 

1043: while crucial for supporting the more massive, rounder 

1044: cluster-forming clump, has a more limited effect on the flattened 

1045: condensation: it escapes easily from the sheet (and quickly 

1046: out of the simulation box) and its energy is mostly ``wasted'' 

1047: as far as driving turbulence in the star-forming, condensed 

1048: material is concerned. The spherical component, in contrast, 

1049: impacts the condensed material more directly, especially in 

1050: the vicinity of a newly formed star. The dispersal of the 

1051: residual core material after star 

1052: formation to a wider region appears to be the main reason 

1053: for the lower star formation rate in the case of stronger 

1054: spherical component: it takes longer for the more widely 

1055: dispersed material to recondense and form stars. The stars 

1056: in the stronger spherical component model are less clustered 

1057: than those in the standard model, consistent with the above 

1058: interpretation.  

1059: %

1060: 

1061: \section{Dense Cores}

1062: \label{core}

1063: 

1064: \subsection{Identification}

1065: 

1066: A variant of the CLUMPFIND algorithm of Williams et al. (1994) is

1067: ussed for core identification. We apply their ``friends-of-friends'' 

1068: algorithm to the density data cube in regions above a threshold 

1069: density, $\rho_{\rm th, min}$, and divide the density distribution 

1070: between $\rho_{\rm th, min}$ and $\rho_{\rm th, max}$ into 15 bins 

1071: equally spaced logarithmically. The maximum is set to $\rho_{\rm th, 

1072: max}=10^3\ \rho_0$, the density above which a star (Lagrangian 

1073: particle) is created if the core mass exceeds the local Jeans 

1074: mass. The minimum is set to $\rho_{\rm th, min}=50\ \rho_0$, 

1075: corresponding to $n_{H_2}=1.25 \times 10^4$~cm$^{-3}$ for the 

1076: fiducial value of initial density. We include only those 

1077: ``spatially-resolved'' cores containing more than 30 cells in 

1078: the analysis, to ensure that their properties are determined with 

1079: reasonable accuracy. The minimum core mass is thus $M_{c, \rm min}

1080: \equiv \rho_{\rm th, min} \times 30\Delta x\Delta y\Delta z=7.19 

1081: \times 10^{-2} M_{J,s}$, where $M_{J,s}\equiv \rho_s L_s^3 

1082: \simeq2.5M_\odot\left(T/10 K\right)^{3/2}\left(n_{H_2,0}/250 

1083: {\rm cm}^{-3} \right)^{-1/2}$ is the Jeans mass of the Spitzer 

1084: sheet. 

1085: 

1086: In our standard simulation, the cloud material settles into a large 

1087: scale gas sheet by the time $t\sim 6\ t_s$. Soon after, a dense core

1088: collapses in a run-away fashion to form the first star, signaling 

1089: the beginning of active star formation. We will limit our analysis 

1090: of dense cores to this star-forming phase, which complements the 

1091: analyses of, e.g., Gammie et al. (2003) and Tilley \& Pudritz (2007), 

1092: which focus on the phase prior to the runaway collapse. 

1093: 

1094: There are typically a handful of dense cores in our standard model at 

1095: any given time (see Fig.~\ref{colden_critical}), which is too small 

1096: for a statistical analysis. To enlarge the core sample, we rerun the 

1097: standard model with a different realization of the initial turbulent 

1098: velocity field, and include all ``resolved'' cores at six selected 

1099: times $t=6.0$, 7.0, 8.0, 9.0, 10.0, and 11.0~$t_s$ from both runs. 

1100: The total number of cores is 72. Since the majority of them do not 

1101: survive for more than 1~$t_s$ (they do last for several local

1102: free-fall times typically), there is relatively little overlap in 

1103: dense cores from one time to another. We treat all cores independently 

1104: regardless whether they appear at more than one selected time or 

1105: not. They have densities in the range 10$^4\sim 10^5$ cm$^{-3}$ 

1106: and radii of $0.04 \sim 0.12$ pc, corresponding roughly to 

1107: H$^{13}$CO$^+$ or N$_2$H$^+$ cores.

1108: 

1109: \subsection{Core Shape}

1110: 

1111: We compute the shape of a core using the eigenvalues of the 

1112: moment-of-inertia tensor:

1113: \begin{equation} 

1114: I_{ij} \equiv \int \rho x_i x_j dV  \ ,

1115: \end{equation}

1116: where the components of position vector, $x_i$ and $x_j$, 

1117: are measured relative to the center of mass.

1118: We compute the lengths of the three principal axes, $a, b,$ and $c$

1119: ($a\ge b \ge c$) by dividing the eigenvalues of the tensor, 

1120: $M_ca^2$, $M_cb^2$, and $M_c c^2$, by the core mass $M_c$.

1121: The core shape is then characterized by the axis ratios 

1122: $\xi\equiv b/a$ and $\eta \equiv c/a$.

1123: 

1124: Figure \ref{fig:coreshapenew} shows the distribution of axis ratios 

1125: of the dense cores in our sample. It is clear that the cores

1126: are generally triaxial; they show little clustering around either the 

1127: prolate ($b/a=c/a<1$) or oblate ($b/a=1, c<1$) line in the figure. 

1128: This result is perhaps to be expected, since the cores are distorted 

1129: by external turbulent flows and occasionally outflows, neither of 

1130: which are symmetric. 

1131: Following Gammie et al. (2003), we divide the $\xi$-$\eta$ plane 

1132: into three parts: a prolate group which lies above the line

1133: connecting $(\xi, \eta)=(1.0, 1.0)$ and (0.33, 0.0), an oblate group

1134: which lies below the line connecting $(\xi, \eta)=(1.0,1.0)$ and 

1135: (0.67, 0.0), and a triaxial group which is everything else.

1136: Of the 72 identified cores, 43\% (31) are prolate, 44\% (32) are

1137: triaxial, and 13\% (9) are oblate.  These fractions are in general 

1138: agreement with Gammie et al. (2003). One  

1139: difference is that the ratio $\eta=c/a$ tends to be smaller for 

1140: our cores [see Figure 10 of Gammie et al. (2003)]. It is likely 

1141: because dense cores in our model are embedded in a condensed sheet, 

1142: which has already settled 

1143: gravitationally along the field lines, although part of the 

1144: difference may also come from the different methods used in

1145: core identification. Note that more massive cores tend to be more 

1146: flattened along their shortest axis. This trend lends support to 

1147: the idea that gravitational squeezing plays a role in core 

1148: flattening. 

1149: 

1150: Another trend is that more massive cores tend to be more oblate. 

1151: For example, 9 out of 15 (60\%) cores having masses more than 1.5 

1152: times the average core mass 

1153: ($\left<M_c\right>=0.93~M_{J,s}$) 

1154: lie 

1155: below the line connecting $(\xi, \eta)=(1.0,1.0)$ and (0.5, 0.0)

1156: (the dotted line in Fig.~\ref{fig:coreshapenew}). Less massive cores

1157: tend to be more prolate. Their shapes may be more susceptible to

1158: external influences, such as ambient flows. 

1159: 

1160: \subsection{Virial Analysis}

1161: 

1162: 

1163: Virial theorem is useful for analyzing the dynamical states of dense 

1164: cores. We follow Tilley \& Pudritz (2007) and adopt the virial 

1165: equation in Eulerian coordinates (McKee \& Zweibel 1992): 

1166: \begin{equation}

1167: {1 \over 2} \ddot{I} + {1 \over 2} \int _S \rho r^2 

1168: \mbox{\boldmath $v$} \cdot d\mbox{\boldmath $S$} 

1169: = U+K+W+S+M+F \ ,

1170: % = {1 \over 2} \ddot{I'}

1171: \label{eq:virial}

1172: \end{equation}

1173: where

1174: $$

1175: %\begin{eqnarray}

1176: I=\int _V \rho r^2 dV\ , 

1177: U= 3 \int _V P dV \ ,

1178: K= \int _V \rho v^2 dV \ ,

1179: W= -\int _V \rho \mbox{\boldmath $r$} \cdot \nabla \Psi dV \ ,

1180: $$

1181: $$

1182: S= -\int _S \left[P\mbox{\boldmath $r$}+ 

1183: \mbox{\boldmath $r$}\cdot (\rho \mbox{\boldmath $v$} 

1184: \mbox{\boldmath $v$})\right] \cdot d\mbox{\boldmath $S$} \ ,

1185: M= {1 \over 4\pi}\int _V B^2 dV \ ,

1186: F= \int _S \mbox{\boldmath $r$} \cdot \mbox{\boldmath $T$}_M \cdot d

1187: \mbox{\boldmath $S$}  \ .

1188: $$

1189: %\end{eqnarray}

1190: Here, $\mbox{\boldmath $T$}_M$ is the Maxwell stress-energy tensor

1191: \begin{equation}

1192: \mbox{\boldmath $T$}_M={1 \over 4\pi}\left(

1193: \mbox{\boldmath $B$}\mbox{\boldmath $B$}-{1 \over 2} B^2 

1194: \mbox{\boldmath $I$}\right) \  ,

1195: \end{equation}

1196: where $\mbox{\boldmath $I$}$ is the unit tensor.

1197: The terms, $I$, $U$, $K$, $W$, $S$, $M$, and $F$ denote, respectively,  

1198: the moment of inertia, internal thermal energy, internal kinetic

1199: energy, gravitational energy, the sum of the thermal surface

1200: pressure and dynamical surface pressure, internal magnetic energy,

1201: and the magnetic surface pressure.

1202: The internal thermal, kinetic, and magnetic energies are always

1203: positive. The gravitational term is negative for all the 

1204: cores in our sample, although it can in principle be positive, 

1205: especially in crowded environments. Other terms can be either 

1206: positive or negative.

1207: The second term on the left hand side of equation (\ref{eq:virial})

1208: denotes the time derivative of the flux of moment of inertia

1209: through the boundary of the core. 

1210: As is the standard practice (e.g., Tilley \& Pudritz 2007), we 

1211: ignore the left hand side of equation (\ref{eq:virial}) in our 

1212: discussion, and consider a core to be in virial equilibrium if 

1213: the sum $U+K+W+S+M+F=0$. 

1214: 

1215: The equilibrium line is shown in Fig.~\ref{fig:virialparamnew}, where 

1216: the sum of the surface terms ($S+F$) is plotted against 

1217: the gravitational term ($W$), 

1218: both normalized to the sum of the internal terms ($U+K+M$). 

1219: The surface terms (dominated by the external kinetic term) 

1220: are generally important; they are more so for lower mass cores 

1221: than higher mass cores. 

1222: The majority of the cores more massive than 1.5 times

1223: the average core mass lie below the equilibrium line, and 

1224: are thus bound; they are expected to collapse and form stars. 

1225: The majority of lower mass cores lie, in contrast, above the line, 

1226: and may disperse away, if they do not gain more mass through

1227: accretion and/or merging with other cores, or further reduce 

1228: their turbulence support through dissipation and/or magnetic

1229: support through ambipolar diffusion. 

1230: 

1231: Tilley \& Pudritz (2007) found in their ideal MHD simulations 

1232: that as the mean magnetic field becomes stronger, the surface 

1233: terms contribute more to the virial equation.

1234: Indeed, for their most strongly magnetized cloud, a good

1235: fraction of cores have surface terms that are greater than 

1236: the sum of the internal terms ($U+K+M$). This is not the 

1237: case for our non-ideal MHD model, even though our clouds are 

1238: more strongly magnetized than theirs. The reason is that the 

1239: cloud dynamics is modified by ambipolar diffusion, especially 

1240: in dense cores, where the diffusion rate is the highest. A 

1241: related difference is that there are more bound cores in our 

1242: models than in their strong field models, where most of the 

1243: cores are magnetically subcritical (see also Dib et al. 2007).   

1244: Ambipolar diffusion enables these cores to become supercritical, 

1245: and thus more strongly bound. It is a crucial ingredient that 

1246: cannot be ignored for core formation, even for relatively 

1247: weakly magnetized clouds (of dimensionless mass-to-flux ratio 

1248: $\lambda$ of a few). For such clouds, the cores tend to be more 

1249: strongly magnetized (relative to their masses) than the cloud 

1250: as a whole, especially in driven turbulence (Dib et al/ 2007), 

1251: since only a fraction of the cloud mass along any given flux 

1252: tube is compressed into a core. 

1253: 

1254: The boundedness of a core is often estimated using the virial 

1255: parameter, the ratio of the virial mass to core mass, 

1256: particularly in observational studies. 

1257: In Figure \ref{fig:virialrationew}, we plot against core mass 

1258: the virial parameter, defined as 

1259: \begin{equation}

1260: \alpha_{\rm vir}={5 \Delta V_{1D}^2 R_c \over GM_c},

1261: \end{equation}

1262: where $\Delta V_{1D}$ is the one-dimensional velocity dispersion 

1263: including the contribution from thermal motion.

1264: There is a clear trend for more massive cores to have smaller 

1265: virial parameters. Indeed, nearly all cores 

1266: more massive than 1.5 times the average 

1267: ($M_c \gtrsim 1.5 \left<M_c\right>\simeq 1.4 M_{J,s}$)

1268: have virial parameters smaller than unity. 

1269: It is consistent with the fact that most of these 

1270: cores lie below the line of the virial equilibrium in 

1271: Fig.~\ref{fig:virialparamnew},  indicating that they are bound. 

1272: The reason for the agreement is that the surface terms are not 

1273: as significant as the gravitational term for massive cores, and

1274: that their magnetic support is greatly reduced by ambipolar 

1275: diffusion (consistent with the trend for cores with smaller 

1276: flux-to-mass ratios to have smaller virial parameters, see

1277: Fig.~\ref{fig:virialparamnew}).

1278: The reduction of surface terms makes the internal kinetic energy, 

1279: which is directly related to the virial parameter, more important.

1280: We conclude that the virial parameter can be used to gauge the 

1281: importance of self-gravity for dense cores formed in strongly 

1282: magnetized clouds, as long as ambipolar diffusion is treated 

1283: properly. The virial parameter can be fitted by a power law 

1284: $\alpha_{\rm vir} \propto M_c^{-2/3}$, in good agreement 

1285: with the scaling found by Bertoldi \& McKee (1992) for nearby 

1286: GMCs.

1287: 

1288: \subsection{Velocity Dispersion-Radius Relation}

1289: 

1290: The kinematics of molecular clouds and dense cores embedded in them 

1291: are probed with molecular lines. Since the pioneering work of Larson 

1292: (1981), many observational studies have shown that the linewidth of 

1293: a region correlates with the size of the region. The linewidth-size 

1294: relation is often approximated by a power law $\Delta V\propto 

1295: R^{0.5}$. However, this relation may not hold universally, especially

1296: on small scales. There are many cases where no clear correlation is 

1297: found between the two (e.g., Caselli \& Myers 1995; Lada et al. 1991; 

1298: Onishi et al. 2002).

1299: 

1300: In Fig.~\ref{fig:velocitysizenew}, we plot the 

1301: three-dimensional

1302: nonthermal velocity 

1303: dispersion against radius (equivalent to the linewidth-size relation) 

1304: for the cores in our sample. No correlation 

1305: is apparent between the two, consistent with the observations of 

1306: Onishi et al. (2002) for the dense cores in the Taurus molecular 

1307: clouds. There is a large scattering in the velocity dispersion-radius 

1308: plot, especially for cores with small masses. The reason may be that 

1309: some cores (especially the less massive ones) are not in virial 

1310: equilibrium and that their formation is strongly influenced by 

1311: external flows. 

1312: 

1313: \subsection{Angular Momentum Distribution}

1314: 

1315: Angular momentum is one of the crucial parameters that determine the 

1316: characteristics of formed stars or stellar systems. In Fig.~\ref{am1}, 

1317: we plot the specific angular momenta of the cores against their radii.

1318: There is no tight correlation between the specific angular momentum 

1319: and radius, although there is a slight tendency for more massive 

1320: cores to have somewhat larger specific angular momenta. For our 

1321: sample cores, the specific angular momentum ranges from 0.01 to 

1322: 0.4 $c_s L_s$ (where $c_s$ is the isothermal sound speed and $L_s$ the

1323: Jeans length of the Spitzer sheet). The distribution peaks around 

1324: 0.08 $c_s L$, corresponding 

1325: to $0.9 \times 10^{21}$ cm$^2$ s$^{-1}$ (for the fiducial cloud 

1326: parameters). For comparison, we 

1327: superpose in the figure the specific angular momenta of 7 starless 

1328: N$_2$H$^+$ cores deduced by Caselli et al. (2002). They are broadly 

1329: consistent with our results. The average for the N$_2$H$^+$ cores 

1330: is estimated at $\sim 1.6 \times 10^{21}$ cm$^2$ s$^{-1}$, somewhat 

1331: larger than the peak value of our distribution. 

1332: 

1333: An oft-used quantity for characterizing the rotation of a dense core 

1334: is the ratio of rotational energy to gravitational energy. For a 

1335: uniform sphere of density $\rho$ rotating rigidly at an angular 

1336: speed $\Omega$, it is given by $\beta=\Omega^2/(4\pi G\rho)$. 

1337: Even though our cores are not uniform, and do not rotate strictly 

1338: as a solid body, we have computed the value of $\beta$ using the 

1339: average angular speed and mean density. The distribution of  

1340: $\beta$ peaks around $\sim $ 0.02, in good agreement with 

1341: the average value of 0.018 for the starless N$_2$H$^+$ cores of 

1342: Caselli et al. (2002). It is also consistent with the well-known 

1343: result that rotation is generally not important as far as core 

1344: support against self-gravity is concerned (Goodman et al. 1993).  

1345: 

1346: \subsection{Flux-to-Mass Ratio}

1347: 

1348: Ambipolar diffusion enables dense cores to be less magnetized relative

1349: to their masses. In Fig.~\ref{fig:mass2fluxnew}, we plot the magnetic 

1350: flux-to-mass ratios of the cores against their column densities.

1351: The ratio is normalized to the critical value $2\pi G^{1/2}$, and the 

1352: column density to that of the initially uniform state, 

1353: $4\pi\rho_0 L_s$. 

1354: The magnetic flux-to-mass ratio is 

1355: approximated as $\Gamma_c\equiv \pi R_{c}^2 \left< B\right>/M_c$, where

1356: $\left<B\right>$ is the mean magnetic field strength inside a core

1357: (e.g., Dib et al. 2007) and the mean core radius $R_c$ is 

1358: defined as the radius of a sphere enclosing the same volume as 

1359: the core. The column density is calculated from the core mass 

1360: divided by the area of a circle whose radius is equal to $R_{c}$. 

1361: There is a clear trend for cores with higher column densities to have 

1362: lower flux-to-mass ratios. Indeed, all cores with masses greater 

1363: than $1.5\left<M_c\right>$ are magnetically supercritical, with 

1364: flux-to-mass ratios as low as $\sim 0.35$, indicating 

1365: that they have all experienced significant ambipolar diffusion. 

1366: The weakening of magnetic support may have enabled these cores to 

1367: condense to a high column density in the first place. 

1368: 

1369: \subsection{Core Mass Spectrum}

1370: 

1371: In Figure \ref{fig:coremf} we plot the mass spectrum for our sample 

1372: of dense cores. The core mass is normalized to the Jeans mass of 

1373: the Spitzer sheet, $M_{J,s}$. 

1374: The lowest mass at $\sim 0.1~M_{J,s}$ corresponds to the minimum 

1375: core mass determined from the criteria for core identification. There 

1376: is a prominent break around $1~M_{J,s}$ in the mass spectrum. Above 

1377: the break, the spectrum can be fitted roughly by a power law $dN/dM 

1378: \propto M^{-2.5}$, not far from the Salpeter stellar initial mass 

1379: function. We note that cores above the break contain more than 

1380: $\sim 100$ cells; their structures should be reasonably well resolved. 

1381: The core spectrum flattens below the break, which is also

1382: broadly consistent with observations. 

1383: Similar core spectra are also 

1384: obtained in the turbulent fragmentation model of core formation that

1385: involves much weaker magnetic field 

1386: and much stronger turbulence

1387: (e.g., Padoan \& Nordlund 2002).

1388: The characteristic mass below which the core mass spectrum flattens 

1389: apparently coincides with the Jeans mass of the condensed sheet, 

1390: which may result from magnetically regulated gravitational fragmentation.

1391: Caution must be exercised in trying to differentiate various 

1392: scenarios of core formation, such as the magnetically regulated 

1393: gravitational fragmentation and the turbulent fragmentation, based 

1394: on core mass spectrum and, perhaps by extension, stellar mass 

1395: spectrum. The stellar mass spectrum for the 52 stars formed in the 

1396: standard model and its close variant are plotted in Fig.~\ref{fig:coremf}

1397: for comparison. The stellar mass spectrum also resembles the 

1398: Salpeter IMF. 

1399: 

1400: \section{Discussion}

1401: \label{discussion}

1402: 

1403: \subsection{Magnetically-Regulated Star Formation in Quiescent 

1404: Condensations} 

1405: \label{scenario}

1406: 

1407: We have simulated star formation in relatively diffuse, mildly 

1408: magnetically subcritical clouds that are strongly turbulent 

1409: initially. The decay of initial turbulence leads to gravitational 

1410: condensation of the diffuse material along field lines into 

1411: a highly fragmented sheet (see Fig.~\ref{3D}), which produces 

1412: stars in small groups. The bulk of the condensed material 

1413: remains magnetically supported in the cross-field direction, 

1414: however, with typical motions that are moderately supersonic 

1415: and approximately magnetosonic. Only a small fraction of the 

1416: condensed material collapses gravitationally to form stars. 

1417: The calculations lead us to a picture of slow, inefficient, 

1418: star formation in a relatively quiescent, large-scale 

1419: condensation (which can be a filament if the cloud is flattened 

1420: in a direction perpendicular to the field lines to begin with) 

1421: that is embedded in a much more turbulent halo of diffuse gas. 

1422: 

1423: The above picture is an unavoidable consequence of the anisotropy 

1424: intrinsic to the large-scale magnetic support. Before the self-gravity 

1425: becomes strong enough to overwhelm the magnetic support in the 

1426: cross-field direction (to initiate widespread star formation), it  

1427: should be able to condense material along the field lines (where 

1428: there is no magnetic resistance) first. The condensed material 

1429: is expected to be relatively quiescent. Its field-aligned velocity 

1430: component is damped during the condensation process, whereas its 

1431: cross-field motions are constrained by the field lines. The 

1432: lateral magnetic support ensures that the bulk of the condensed 

1433: material is relatively long-lived, allowing more time for the 

1434: turbulence to dissipate. The dissipation is hastened by the high 

1435: density in the condensation, which makes it easier for shock 

1436: formation, because of low Alfv\'en speed. Indeed, when the condensed 

1437: layer accumulates enough material for its mass-to-flux 

1438: ratio to approach the critical value (a condition for widespread 

1439: star formation), its Alfv\'en speed drops automatically to a value 

1440: comparable to the sound speed (NL05, Kudoh et al. 2007). Rapid 

1441: dissipation of super-magnetosonic motions ensures that the bulk of 

1442: the condensed material moves at no more than moderately supersonic 

1443: speeds, as we find here and previously in 2D calculations (NL05). 

1444: It is in this relatively quiescent environment that most star 

1445: formation takes place.

1446: 

1447: The quiescent, star-forming, condensed material is surrounded by 

1448: a highly turbulent, fluffy halo. The stratification in turbulent 

1449: speed is a consequence of the stratification in density (Kudoh \& 

1450: Basu 2003): the amplitude of a wave increases as it propagates 

1451: from high density to low density, and faster motions can survive 

1452: for longer in the lower density region because of higher Alfv\'en 

1453: speed. We believe that this is a generic result independent of 

1454: the source of turbulence. In our particular simulations, the 

1455: saturated, highly supersonic turbulence in the halo at late times 

1456: is driven mainly by protostellar outflows internally. It is 

1457: conceivable, perhaps even likely, that most of the turbulence 

1458: in the diffuse regions of star forming clouds comes from energy 

1459: cascade from larger-scale molecular gas (perhaps even HI envelope). 

1460: Depending on the details of energy injection, the external driving 

1461: of turbulence can slow down the 

1462: gravitational settling along field lines and may even suppress 

1463: it altogether. Even when the large-scale gravitational settling 

1464: is suppressed, star formation can still occur, perhaps in a less 

1465: coherent fashion: stars can form in localized pockets of a 

1466: magnetically subcritical cloud directly from strong compression 

1467: through shock-enhanced ambipolar diffusion (NL05), especially

1468: when the shock is highly super-Alfv\'enic (Kudoh \& Basu 2008). 

1469: Vigorous driving may favor a mode of turbulence-accelerated star 

1470: formation at isolated locations, whereas decay of turbulence may 

1471: allow the gravity to pull the diffuse gas into large-scale 

1472: condensations, in which stars form in a more coherent fashion. 

1473: The first mode is now under investigation, and the results will 

1474: be reported elsewhere. The second corresponds to the picture 

1475: outlined above.  

1476: 

1477: The dichotomy is not unique to star formation in strongly magnetized 

1478: clouds. It is also present when the magnetic field is weak or 

1479: non-existent: whereas stars form rapidly in a clustered mode in a 

1480: decaying non-magnetic turbulence, they are produced more slowly, in 

1481: a more dispersed fashion, if the turbulence is driven constantly, 

1482: especially on small scales (e.g., Mac Low \& Klessen 2004). Different 

1483: modes of star formation may operate in different environments, and 

1484: there is a urgent need for sorting out which mode dominates in which 

1485: environment. In the next subsection, we will concentrate on the Taurus 

1486: molecular clouds, the prototype of distributed low-mass star forming 

1487: regions (see Kenyon et al. 2008 for a recent review), and the Pipe 

1488: nebula, which appears to be a younger version of the Taurus complex 

1489: (Lada et al. 2008). 

1490: 

1491: \subsection{Connection to Observations} 

1492: 

1493: \subsubsection{The Taurus Molecular Cloud Complex}

1494: \label{taurus}

1495: 

1496: The Taurus molecular cloud complex may represent the best case for 

1497: the picture of magnetically regulated star formation investigated 

1498: in this paper. One line of evidence supporting this assertion 

1499: comes from the studies of young stars by Palla \& Stahler (2002). 

1500: They inferred, using pre-main sequence tracks, that the star formation 

1501: in the complex has two phases. It began at least 10 Myrs ago, at 

1502: a relatively low level and in widely dispersed locations (Phase 

1503: I hereafter). In the last three million years or so, stars are 

1504: produced at a much higher rate (Phase II). The majority of the 

1505: stars produced in Phase II are confined within the dense, 

1506: striated, C$^{18}$O filaments (Onishi et al. 2002). Palla \& 

1507: Stahler interpret these results to mean that the Taurus clouds 

1508: evolved from an initial, relatively diffuse, state 

1509: quasi-statically. They envision that, during the early epoch 

1510: of quasi-static contraction, few places have contracted to 

1511: the point of forming dense cores and protostars, and hence only 

1512: scattered star formation activity is seen at isolated locations 

1513: (Phase I). 

1514: Eventually, further contraction of the cloud leads to the formation 

1515: of nearly contiguous, dense filaments where multiple cores can 

1516: form and collapse simultaneously, leading to elevated star formation 

1517: (Phase II). 

1518: % The authors did not speculate on the physical mechanisms 

1519: % that may be responsible for this behavior. 

1520: In view of the discussion 

1521: in the last subsection, it is natural to identify Phase I with the 

1522: shock-accelerated mode of star formation in a magnetically 

1523: subcritical or nearly critical cloud (or sub-region), when the 

1524: turbulence is still 

1525: strong enough to prevent large-scale gravitational condensation 

1526: along the field lines. In this picture, Phase II started when the

1527: turbulence had decayed to a low enough level that the gravity was  

1528: able to collect enough mass along the field lines into condensed 

1529: structures for a faster, more organized, mode of star formation 

1530: to take over. 

1531: 

1532: A salient feature of our picture is that the rate of star formation 

1533: in Phase II should remain well below the limiting free-fall rate 

1534: of the condensed material. There is strong evidence that the 

1535: conversion of dense material in the Taurus clouds into stars is 

1536: indeed slow. The dense, filamentary material is well traced by 

1537: C$^{18}$O, which Onishi et al. (1996) has studied in great detail. 

1538: These authors estimated a mass and average density for the C$^{18}$O 

1539: gas of $M_{{\rm C}^{18}{\rm O}}\sim 2900~M_\odot$ and $n_{H_2}\sim

1540: 4000$~cm$^{-3}$ (comparable to that in the condensed sheet in 

1541: our simulations). The latter corresponds a free fall time 

1542: $t_{{\rm C}^{18}{\rm O},ff}=4.86 \times 10^5$~yrs. The maximum, free-fall, 

1543: rate of turning C$^{18}$O material into stars is therefore 

1544: ${\dot M}_{*,{\rm C}^{18}{\rm O},ff}\sim 6\times

1545: 10^{-3}~M_\odot$~yr$^{-1}$. This 

1546: rate is to be compared with the actual rate of star formation in 

1547: the complex, ${\dot M}_*\sim 5\times 10^{-5}~M_\odot$yr$^{-1}$, 

1548: estimated by Goldsmith et al. (2008) based on an assumed average 

1549: mass of $0.6~M_\odot$ for the 230 stars observed in their

1550: mapped region (Kenyon et al. 2008) and a duration of 3~Myrs for 

1551: Phase II of star formation where most stars are produced (Palla 

1552: \& Stahler 2002). Therefore, the rate of converting the even 

1553: dense (relative to the diffuse $^{12}$CO gas) C$^{18}$O material 

1554: into stars is extremely slow, amounting to $\sim 1\%$ per local 

1555: free fall time. This rate is comparable to those obtained in our 

1556: simulations. 

1557: %

1558: % comparable to Krumholz's estimates for denser gas!

1559: % 

1560: 

1561: One may argue that the slow rate of star formation in C$^{18}$O

1562: gas is due to turbulence support. However, the C$^{18}$O gas 

1563: in the Taurus clouds is observed to be unusually quiescent. 

1564: Onishi et al. (1996) estimated a mean equivalent line-width 

1565: (defined as the integrated intensity divided by the peak 

1566: temperature of the line) of 0.65~km~s$^{-1}$, corresponding 

1567: to a 3D velocity dispersion of 0.45~km~s$^{-1}$ for a Gaussian

1568: profile. The velocity dispersion is $\sim 2.4$ times the 

1569: sound speed for a gas temperature of 10~K, indicating that 

1570: the moderately dense gas traced by C$^{18}$O is not 

1571: highly supersonic in general. Indeed, the velocity 

1572: dispersion is quite comparable to that of the condensed

1573: sheet material in our standard model at similar densities

1574: (see Fig.~\ref{rms_vel}). Nearly 1/3 ($937~M_\odot$ out 

1575: of $2900~M_\odot$) of the C$^{18}$O gas is classified 

1576: into 40 cores, which have a mean line-width (FWHM) of 

1577: 0.49~km~s$^{-1}$, comparable to that estimated above 

1578: for the C$^{18}$O gas as a whole. Onishi et al. concluded 

1579: that most of the cores are roughly gravitationally bound, 

1580: based on their masses, sizes and line-widths. These authors 

1581: argued that the general alignment of the minor axes of the 

1582: cores with the overall field direction indicates that the 

1583: magnetic field plays a role in their formation. We agree 

1584: with this assessment. 

1585: The relatively low velocity dispersion of the C$^{18}$O gas 

1586: and the morphology of dense cores provide indirect support 

1587: for the picture of magnetically regulated star formation. 

1588: 

1589: A more direct test of the picture would come from a 

1590: determination of the mass-to-flux ratio. 

1591: %

1592: % what is predicted? can moderately supercritical clouds 

1593: % work? 

1594: % 

1595: For a given field strength $B$, the critical column density 

1596: along the field line is

1597: \begin{equation}

1598: \Sigma_{\vert\vert,c}=58.7~B_{20}~M_\odot~{\rm pc}^{-2}, 

1599: \label{critical}

1600: \end{equation}

1601: corresponding to an $A_V \approx 2.5~B_{20}$, where $B_{20}$ 

1602: is the field strength in units of 20~$\mu$G. 

1603: Goldsmith et al. (2008) estimated a total mass of 

1604: $2.4\times 10^4 M_\odot$ in their 

1605: mapped area of $28$~pc$\times 21$~pc, yielding an average 

1606: column density of 40.8~M$_\odot$~pc$^{-2}$ along the line 

1607: of sight. Only $\sim 2/3$ of the surveyed area have detectable 

1608: $^{12}$CO emission in individual spectra, however (their 

1609: Masks 1 and 2). For these regions, the total mass and area 

1610: are $1.95\times 10^4 M_\odot$ and $\sim 392$~pc$^2$, yielding 

1611: a somewhat higher average column density of 49.7~M$_\odot$~pc$^{-2}$ 

1612: along the line of sight. If the column density along the field 

1613: lines is comparable to this value, the corresponding critical 

1614: field strength would be $16.9~\mu$G, according to 

1615: equation~(\ref{critical}). About half of the mass (9807~M$_\odot$) 

1616: resides in eight ``high-density'' regions (such as L1495 and 

1617: Heiles Cloud 2) that together occupy $\sim 31\%$ of the area, 

1618: with an average line-of-sight column density of 

1619: 79.5~M$_\odot$~pc$^{-2}$. This value would correspond to 

1620: a critical field strength of $27.1~\mu$G if the magnetic field 

1621: is exactly along the line of sight. There is, however, evidence 

1622: for a large-scale magnetic field in the plane of sky in the 

1623: Taurus cloud complex from the polarization of background 

1624: star light. Indeed, the velocity anisotropy measured by Heyer 

1625: et al. (2008) in a relatively diffuse sub-region points to 

1626: a magnetic field that is mainly in the plane of sky. If this 

1627: magnetic orientation is correct, and if matter condenses along the 

1628: field lines into flattened structures, then projection effects 

1629: may be considerable. Applying a moderate correction factor 

1630: of $\sqrt{2}$ (corresponding to a viewing angle of $45^\circ$) 

1631: would bring the average field-aligned column density of the 

1632: ``high-density'' regions down to 56.2~M$_\odot$~pc$^{-2}$, with 

1633: a corresponding  

1634: critical field strength of $19.2~\mu$~G. It is therefore likely 

1635: that a magnetic field of order 20~$\mu$G is strong enough to 

1636: regulate star formation in the Taurus cloud complex. 

1637: 

1638: The magnetic field strength has been measured for three dense 

1639: cores in Taurus through OH Zeeman measurements using Arecibo 

1640: telescope (Troland \& Crutcher 2008). They are B217-2, TMC-1 

1641: and L1544, with a line-of-sight field strength of $B_{\rm los} 

1642: \approx 13.5$, $9.1$, and $10.8$~$\mu$G respectively. Although 

1643: the true orientation of the magnetic field is unknown, a 

1644: statistically most probable correction factor of 2 would bring 

1645: the total field strength to $\sim 18.2 - 27.0~\mu$G. Given that 

1646: OH samples moderately dense gas approximately as C$^{18}$O 

1647: does, and that the field strength changes relatively little at 

1648: low to moderate densities (see Fig.~\ref{ave_B}), it is plausible 

1649: such a field permeates most of the space in the complex. A 

1650: worry is that such a strong field is not detected in several 

1651: other cores in the Taurus in the same Arecibo survey. Whether

1652: these non-detections can be reconciled with our picture of

1653: magnetically regulated star formation through, e.g., unfavorable 

1654: field orientation, remains to be seen. 

1655: 

1656: Indirect evidence for a relatively strong magnetic field 

1657: comes from the observation that the magnetic field as

1658: traced by the polarization vectors of the background star

1659: light is well ordered over many degrees in the sky, 

1660: indicating that the magnetic energy is larger than the 

1661: turbulent energy.

1662: If we use conservative estimates for H$_2$ number density 

1663: of $10^2$~cm$^{-3}$ (see Table 2 of Goldsmith et al. 2008), 

1664: and turbulent speed of 2~km~s$^{-1}$,

1665: we can put a lower limit to the field strength at $\sim 

1666: 15~\mu$G. The limit agrees with the value of $\sim 14~\mu$G 

1667: estimated by Heyer et al. (2008) based on comparing the 

1668: observed velocity anisotropy in a striated, relatively 

1669: diffuse region with MHD turbulence simulations; they point 

1670: out that 

1671: the value is to be increased if the magnetic field has 

1672: a substantial component along the line of sight. A

1673: strong line-of-sight magnetic field of $\sim 20~\mu$G 

1674: or more is deduced by Wolleben \& Reich (2004) at the 

1675: boundaries of molecular clouds a few degrees away from 

1676: the sub-region studied by Heyer et al. (2008), from 

1677: modeling the enhanced rotation measures inferred from 

1678: polarization observations at 21 and 18~cm. Subsequent 

1679: HI Zeeman observations did not confirm the deduced 

1680: field strength, however (C. Heiles, priv. comm.). 

1681: 

1682: Although none of the pieces of evidence discussed above is 

1683: conclusive individually, taken together, they make a 

1684: reasonably strong case for a global magnetic field of order 

1685: $\sim 20\mu$G in strength. The existence of such a field 

1686: may not be too surprising from an evolutionary point of view. 

1687: It is now established that the cold neutral medium (CNM) of 

1688: HI gas (likely precursor of Taurus-like molecular clouds)   

1689: is fairly strongly magnetized in general, with a median

1690: field strength of $\sim 6~\mu$G (Heiles \& Troland 2005). 

1691: For denser HI clouds, it is not difficult to imagine a 

1692: somewhat higher field strength. 

1693: A case in point is the nearby Riegel-Crutcher HI cloud 

1694: that we mentioned in the introduction. It was mapped 

1695: recently by McClure-Griffiths et al. (2006) using 21~cm 

1696: absorption against strong continuum emission towards the 

1697: Galactic center. They found long strands of cold neutral

1698: hydrogen that are remarkably similar to the CO striations 

1699: in Taurus studied by Heyer et al. (2008): both are aligned 

1700: with the local magnetic field. For the magnetic energy to

1701: dominate the turbulent energy inside the filaments, the

1702: field strength must be of order $30~\mu$G or larger. This

1703: value is consistent with $B_{los}\sim 18~\mu$G obtained 

1704: by Kazes \& Crutcher (1986) through Zeeman measurements 

1705: at a nearby location, if a correction of $\sim 2$ is

1706: applied for projection effects. The physical conditions 

1707: inside the HI filaments of the Riegel-Crutcher cloud are 

1708: not dissimilar to those of diffuse CO clouds: number 

1709: density $n_H\sim 460$~cm$^{-3}$, 

1710: velocity dispersion $\sigma \sim 3.5$~km~s$^{-1}$, and 

1711: temperature $T\sim 40$~K. The main difference is that the 

1712: column density is $\sim 10^{20}$~cm$^{-2}$, which is not 

1713: sufficient for shielding the CO from photodissociation 

1714: (van Dishoeck \& Black 1988). 

1715: 

1716: The above discussion leads us to the following scenario for 

1717: cloud evolution and star formation in the Taurus molecular

1718: clouds: we envision the cloud complex to have evolved from a 

1719: turbulent, strongly magnetized, non-self-gravitating HI cloud 

1720: similar to the Riegel-Crutcher cloud. Once enough material 

1721: has been accumulated, perhaps by converging flows 

1722: (Ballesteros-Paredes et al. 1999) or some other means

1723: (such as differential turbulence dissipation, which may 

1724: lead to a condensation through ``cooling flow'' along the 

1725: field lines, e.g., Myers \& Lazarian 1998), for CO 

1726: shielding, it becomes visible as a molecular cloud. 

1727: Depending on the degree of magnetization, level of 

1728: inhomogeneity, and strength of turbulence, star formation 

1729: can start quickly in some pockets (perhaps even during the 

1730: process of molecular cloud formation), despite of a strong 

1731: global magnetic field, either because they happen to be 

1732: less magnetized to begin with or because their magnetic 

1733: fluxes are force-reduced by shock-enhanced ambipolar 

1734: diffusion. This corresponds to the low-level, Phase I of 

1735: star formation deduced by Palla \& Stahler (2002). Once 

1736: enough material has settled along field lines for the  

1737: mass-to-flux ratio of the condensed material to approach 

1738: the critical value, the rate of star formation increases 

1739: drastically. We identify this phase of more organized star 

1740: formation in condensed structures as the more active, 

1741: Phase II of star formation deduced by Palla \& Stahler 

1742: (2002). Even during this relatively active phase, the star 

1743: formation rate is still well below the free-fall rate of 

1744: the condensed gas, because of magnetic regulation. 

1745: %

1746: % 330 solar masses in H13CO+, SFE rate 10 times slower than free 

1747: % fall rate?? 

1748: %

1749: 

1750: The above picture can be contrasted with that of Ballesteros-Paredes 

1751: et al. (1999) and Hartmann et al. (2001). These authors advocated 

1752: a picture of rapid cloud formation and star formation for the

1753: Taurus-Auriga complex, motivated mainly by a lack of ``post-T Tauri 

1754: stars'' older than $\sim 5$~Myrs in the region; indeed, the 

1755: majority of stars are younger than $\sim 3$~Myrs, as mentioned 

1756: earlier. They argue that it is difficult to understand how star 

1757: formation can occur simultaneously over a spatial extent of $\sim 

1758: 20$~pc (or more) over a timescale as short as a few million years, 

1759: given that the observed turbulent speed is $\sim 2$~km~s$^{-1}$, 

1760: corresponding to a crossing time of $\sim 10$~Myrs (or more), 

1761: unless the star formation is coordinated by some agent other than 

1762: the currently observed turbulence. They suggest that this agent is 

1763: external: a large-scale converging flow with speeds of 5-10~km~s$^{-1}$.  

1764: In our picture, it is the global gravity and a strong, large-scale, 

1765: internal magnetic field that coordinate the star formation: the 

1766: strong field suppresses vigorous star formation until enough 

1767: material has collected along field lines into condensed structures 

1768: for the gravity to start overwhelm magnetic support in the 

1769: cross-field directions. The degree to which the star formation is 

1770: actually synchronized will, of course, depend on how the mass is 

1771: initially distributed along the field lines and rate of turbulence 

1772: dissipation and replenishment, neither of which is, unfortunately, 

1773: well constrained at the present. Indeed, a hybrid scenario may be

1774: possible, with the mass accumulation along the field lines sped 

1775: up by a large-scale converging flow. Nevertheless, we believe 

1776: that, at least in the particular case of the Taurus molecular cloud 

1777: complex, for the large amount of dense, relatively quiescent, 

1778: C$^{18}$O gas to form stars at $\sim 1\%$ of the free-fall rate, 

1779: magnetic regulation is needed. A similar case can be made for the 

1780: Pipe Nebula.   

1781: 

1782: \subsubsection{Pipe Nebula}

1783: \label{pipe}

1784: 

1785: Although less well studied than the Taurus cloud complex, the Pipe 

1786: nebula is an important testing ground for theories of low-mass 

1787: star formation in relatively diffuse dark clouds. It has $\sim 

1788: 10^4~M_\odot$ of molecular material as traced by $^{12}$CO (Onishi 

1789: et al. 1999), not much different from the Taurus complex. The higher 

1790: density gas traced by $^{13}$CO and C$^{18}$O is concentrated in 

1791: a long filament that spans more than 10~pc, again similar to the

1792: Taurus case. The main difference is that there is no evidence

1793: for star formation activity except in the most massive core, B59. 

1794: The lower rate of star formation indicates that the Pipe nebula is 

1795: at an earlier stage of evolution compared to the Taurus complex, as 

1796: pointed out by Muench et al. (2007). The relative inactivity is 

1797: not due to a lack of dense, quiescent material that is a 

1798: pre-requisite of (low-mass) star formation. J. Alves et al. (2007) 

1799: identified 159 dense cores from extinction maps, many of which are 

1800: formally gravitationally unstable, with masses above the critical 

1801: Bonner-Ebert mass (Lada et al. 2008). In the absence of additional 

1802: support, such cores should collapse and form stars on the free-fall 

1803: time scale, which is $t_{ff}=3.7\times 10^5$~yrs for the inferred 

1804: median core density $n(H_2)=7\times 10^3$~cm$^{-3}$. This time 

1805: scale is much shorter than the turbulence crossing time $t_c=L/v 

1806: \sim 5\times 10^6$~yrs, over a region of $L \sim 15$~pc, and for 

1807: a typical turbulent speed of $v\sim 3$~km~s$^{-1}$. The mismatch 

1808: between the turbulent crossing time and star formation time is 

1809: therefore much more severe than that noted above for the Taurus 

1810: region. 

1811: 

1812: There are two possible resolutions to the above conundrum. If 

1813: the dense cores (at least those massive enough to be formally 

1814: gravitationally unstable) are indeed short-lived objects, with 

1815: a lifetime of order the average free-fall time ($\sim 3\times 

1816: 10^5$~yrs), their formation must be rapid and well coordinated. 

1817: It is tempting to 

1818: attribute the coordination to a large-scale converging flow, as 

1819: Ballesteros-Paredes et al. (1999) and Hartmann et al. (2001) 

1820: did for the Taurus case. However, to cover a distance of $15$~pc 

1821: (the size of the cloud, which is comparable to the length of 

1822: the chain of dense cores) in $\sim 3\times 10^5$ years, the 

1823: flow needs to converge at a speed of order $\sim 50$~km~$s^{-1}$, 

1824: which is uncomfortably high. A more likely alternative is that 

1825: the cores live much longer 

1826: than their free-fall times would indicate. While the longevity 

1827: is not necessarily a problem for those non-self-gravitating, 

1828: lower mass cores that can be confined by external pressure, 

1829: it is problematic for the more massive cores that should have 

1830: collapsed on the free-fall time scale, unless they are supported 

1831: by some (so-far invisible) agent in addition to the observed

1832: thermal and turbulent pressures. The most likely candidate is

1833: a magnetic field. There is some evidence for the existence of 

1834: an ordered magnetic field that is more or less perpendicular 

1835: to the long filament that contains the dense cores from stellar 

1836: polarization observations (F. Alves \& Franco 2007; F. Alves et

1837: al., in preparation), although much more work is needed, particularly 

1838: for the diffuse region outside the filament, to firm up or reject 

1839: this possibility. In any case, it is plausible that 

1840: the Pipe nebula represents an earlier stage of the magnetically 

1841: regulated cloud evolution and star formation than the Taurus 

1842: clouds: material has started to collect along the field lines 

1843: into dense, relatively quiescent, structures, but their 

1844: mass-to-flux ratios are yet to reach the critical value, 

1845: except perhaps in the most massive core, B59. In this case, 

1846: the ages of the cores are set by the rate of mass condensation 

1847: along the field lines, which may depend on the ill-understood 

1848: process of turbulence replenishment in the diffuse gas, and 

1849: may be accelerated by large-scale converging flows (Ballesteros 

1850: et al. 1999; Hartmann et al. 2001). 

1851: %

1852: % Southern coalsack??? earlier phase???   

1853: %

1854: 

1855: \subsection{Modes of Star Formation in Strongly Magnetized Clouds}

1856: \label{mode}

1857: 

1858: Strong magnetic fields tend to resist cloud condensation and 

1859: star formation. This resistance can be overcome by either 

1860: turbulence or gravity, leading to two distinct modes of star 

1861: formation. Which mode dominates depends on the relative

1862: importance of gravity and turbulence. When the turbulence is 

1863: strong enough to prevent gravitational settling along field 

1864: lines, star formation is still possible. Turbulence can 

1865: overcome magnetic resistance in localized regions that are 

1866: compressed by converging flows to high densities; such regions 

1867: are expected to have large magnetic field gradients and low 

1868: fractional ionization, both of which enhance the rate of 

1869: ambipolar diffusion. The 

1870: turbulence-accelerated ambipolar diffusion enables dense

1871: pockets to become self-gravitating 

1872: %\footnote{Dib et al. (2007) 

1873: %found that the majority of the dense cores formed in a rather 

1874: %weakly magnetized cloud of dimensionless mass-to-flux ratio 

1875: %of $\lambda=2.8$ with driven turbulence are nearly magnetically 

1876: %critical or even subcritical. Such cores are expected to be

1877: %strongly affected by enhanced ambipolar diffusion as well.} 

1878: and collapse into stars (NL05, Kudoh \& Basu 2008), even when 

1879: the background cloud 

1880: remains well supported. In this mode, the crucial step of 

1881: star formation---core formation---is driven by turbulent 

1882: converging flows but regulated by magnetic fields, with 

1883: self-gravity playing a secondary role. After a bound core 

1884: has formed, its further evolution is driven primarily by 

1885: the self-gravity, but remains regulated by the magnetic 

1886: field. 

1887: 

1888: The second mode corresponds to a turbulence that is not strong 

1889: enough to prevent gravitational settling, either because it 

1890: is weak to begin with or because it has decayed without being 

1891: adequately replenished. Condensation would occur along field 

1892: lines, which would lead to further increase in self-gravity 

1893: and more rapid dissipation of turbulence. In the condensed 

1894: structure, turbulence becomes less important dynamically, and 

1895: star formation is driven mainly by gravity in the cross-field 

1896: direction. The gravitational contraction is resisted primarily 

1897: by magnetic fields, although protostellar outflows play an 

1898: important role in dispersing away the dense gas that remains 

1899: after the formation of individual stars; the dispersal slows 

1900: down the global star formation. In this mode, the gravity and 

1901: magnetic fields are of primary importance, with turbulence 

1902: playing a secondary role. 

1903: 

1904: The above two modes of star formation may be interconnected: it 

1905: is natural to expect the second to follow the first, as a result 

1906: of turbulence dissipation. The transition from one to the 

1907: other should occur when the mass-to-flux ratios in the condensed 

1908: structures approach the critical value. It may be possible for 

1909: the mass-to-flux ratio to stay near the critical value due to 

1910: outflow feedback: if too much mass is added to the structure 

1911: so that its mass-to-flux ratio increases well above the critical 

1912: value, it would form stars at a higher rate, producing more 

1913: outflows that stop further mass accumulation, similar to the 

1914: outflow-regulated scenario of cluster formation envisioned 

1915: in Matzner \& McKee (2000). They have shown that regions of 

1916: outflow-regulated star formation tend to be overstable, 

1917: oscillating with increasingly large amplitudes. Indeed, the 

1918: outflows may unbind the condensed structure completely, 

1919: terminating its star formation quickly. An abrupt termination 

1920: is implied by the dearth of the well-observed star forming 

1921: regions in the declining phase of star formation (Palla \& 

1922: Stahler 2000). 

1923: 

1924: If, on the other hand, outflows fail to retard mass accumulation 

1925: along field lines, the star formation may change over to 

1926: yet another, more active, mode --- cluster formation. The 

1927: transition is expected to take place when the mass-to-flux ratios 

1928: of the condensations become substantially greater than the critical 

1929: value (say by a factor of two), making global (not just local) 

1930: contraction perpendicular to the field lines possible. The

1931: higher rate of star formation should lead to a stronger 

1932: outflow feedback, which could increase the turbulence 

1933: to a level that balances the global collapse (e.g., Nakamura 

1934: \& Li 2007; Matzner 2007). In this mode, protostellar 

1935: outflow-driven turbulence becomes the primary agent that 

1936: resists the global gravitational collapse, with the magnetic 

1937: field playing a secondary role. 

1938: 

1939: In summary, we envision three, perhaps sequential, modes of star 

1940: formation in strongly magnetized clouds, dominated (a) first by 

1941: turbulence and magnetic fields, (b) then by magnetic fields and 

1942: gravity, and (c) finally by gravity and turbulence (again). In 

1943: this sequence, stars form first directly out of turbulent diffuse 

1944: material in a dispersed fashion, then more coherently in 

1945: relatively quiescent, condensed structures, and finally as 

1946: clusters in dense turbulent clumps. The strong turbulence 

1947: is primordial initially. It is transformed into a protostellar 

1948: turbulence by outflows in regions of active cluster formation.

1949: %

1950: % B important in quiescent regions, emphasized by Cox; true for SF as well. 

1951: %

1952: % (b) is unique to strong field case

1953: 

1954: \section{Summary and Conclusions}

1955: \label{conclude}

1956: 

1957: We have performed 3D simulations of star formation in turbulent, 

1958: mildly magnetically subcritical clouds including protostellar 

1959: outflows and ambipolar diffusion. The simulations are motivated

1960: by observations of molecular gas and young stars in the Taurus

1961: cloud complex and the realization that cold neutral HI gas, a 

1962: likely progenitor of molecular gas, is magnetically subcritical

1963: in general. They are natural extensions of our previous 2D 

1964: calculations. The main results are:  

1965: 

1966: 1. The decay of turbulence leads to gravitational condensation 

1967: of material along the field lines. In the ideal MHD limit, the 

1968: resulting structure closely resembles a smooth Spitzer sheet, 

1969: as found previously (Krasnopolsky \& Gammie 2005). Ambipolar 

1970: diffusion changes the structure drastically. It creates 

1971: supercritical material that behaves differently from the 

1972: subcritical background in a fundamental way: it can be condensed 

1973: further by self-gravity. As a result, the distributions of mass 

1974: with respect to volume and column densities are greatly broadened. 

1975: Indeed, the probability distribution function (PDF) of the (more 

1976: easily observable) column density can be fitted approximately 

1977: by a broad, lognormal distribution (see Fig.~\ref{colden_PDF}), 

1978: as is the case for weakly magnetized clouds fragmented by strong 

1979: turbulence (e.g., Padoan \& Nordlund 2002; P. S. Li et al. 2004). 

1980: It may be difficult to distinguish the turbulent fragmentation 

1981: in weakly magnetized clouds and the ambipolar diffusion-enabled 

1982: gravitational fragmentation in strongly magnetized clouds based 

1983: on mass distribution. 

1984: 

1985: 2. The magnetized clouds evolve into a configuration strongly 

1986: stratified in density at late times, with dense cores embedded 

1987: in fragmentary condensations which, in turn, are surrounded 

1988: by a more diffuse halo (see Fig.~\ref{3D}). 

1989: The condensed material can be significantly warped and may appear 

1990: thicker than its intrinsic width, even when viewed edge-on, 

1991: because of superposition of warps along the line of sight. While 

1992: the diffuse halo is highly turbulent, the condensations are 

1993: more quiescent, with moderately supersonic velocity dispersions

1994: in general. One reason is that the bulk of the condensed material 

1995: is magnetically supported, which allows more time for the 

1996: turbulence to decay. Another is their relatively low Alfv\'en 

1997: speed (comparable to the sound speed), which forces highly 

1998: supersonic motions to dissipate quickly, through shock 

1999: formation. The moderately supersonic turbulence is maintained, 

2000: however, for many local free-fall times, through a combination 

2001: of outflow feedback, MHD waves, and gravitational motions 

2002: induced by ambipolar diffusion.

2003: %, even in the absence of any external driving. 

2004:   

2005: 3. Stars form at a low rate. Only one percent or less of the cloud 

2006: material is converted into stars in a free-fall time at the 

2007: characteristic density of the condensed material. The slow star 

2008: formation cannot be due to support by turbulence because the  

2009: star-forming, condensed material is rather quiescent. It is 

2010: regulated by magnetic fields. The low rate of star formation is 

2011: not sensitive to the level of initial turbulence, although more 

2012: turbulent clouds produce stars early, consistent with the picture 

2013: of turbulence-accelerated star formation in strongly magnetized 

2014: clouds. The rate of star formation is further reduced by 

2015: protostellar outflow, particularly the component in the plane 

2016: of mass condensation.   

2017: 

2018: 4. Dense cores formed in our simulations are typically triaxial.  

2019: More massive cores tend to have lower magnetic flux-to-mass

2020: ratios, be more tightly bound gravitationally, and be more 

2021: oblate. Virial analysis indicates that the surface term due to 

2022: external kinetic motions contributes significantly in general, 

2023: especially for less massive cores. We find a larger number 

2024: of bound cores than in previous ideal MHD simulations even 

2025: though our clouds are more strongly magnetized; they are made 

2026: possible by ambipolar diffusion. The cores have specific 

2027: angular momenta comparable to the observed values, and a mass 

2028: spectrum that resembles the Salpeter stellar IMF towards the 

2029: high mass end. The spectrum flattens near the characteristic 

2030: Jeans mass of the condensed material, which is a few solar 

2031: masses for typical parameters. The stellar mass spectrum also 

2032: resembles the Salpeter IMF. 

2033: 

2034: 5. The Taurus molecular cloud complex may represent the best case 

2035: for magnetically regulated star formation. The strongest evidence 

2036: comes from the existence of a large amount of dense, relatively 

2037: quiescent, C$^{18}$O gas, which produces stars at a rate two orders 

2038: of magnitude below the maximum free-fall rate. Most likely, the 

2039: star formation is regulated by a strong, ordered magnetic field. 

2040: Several lines of evidence suggest that a magnetic field of the 

2041: required strength (of order $\sim 20 \mu G$) is indeed present 

2042: in the region, although direct Zeeman measurements are available 

2043: for only three cores, and their interpretations are complicated 

2044: by projection effects. 

2045: 

2046: 6. We speculate that, depending on the (uncertain) rates of 

2047: turbulence replenishment and outflow feedback, there may be 

2048: three distinct modes of star formation in strongly magnetized 

2049: clouds: (a) inefficient star formation through 

2050: turbulence-accelerated ambipolar diffusion at dispersed 

2051: locations in relatively diffuse, magnetically subcritical 

2052: clouds, (b) more coherent, but still inefficient, 

2053: magnetically regulated star formation in relatively 

2054: quiescent, nearly magnetically critical condensations, 

2055: (c) more efficient cluster formation in magnetically 

2056: supercritical clumps that may be regulated by outflow-driven 

2057: protostellar turbulence. Much work remains to be done to 

2058: understand these modes and their possible interconnection. 

2059: 

2060: \acknowledgments 

2061: This work is supported in part by a Grant-in-Aid for Scientific Research 

2062: of Japan (18540234, 20540228),  

2063: and NSF and NASA grants (AST-0307368 and NAG5-12102). 

2064: Parts of this work were performed while the authors were in residence

2065: at the Kavli Institute for Theoretical Physics at UCSB.

2066: The numerical calculations were carried out mainly 

2067: on NEC SX8 at YITP in Kyoto University, 

2068: on Fujitsu VPP5000 at the Center for Computational Astrophysics, CfCA, 

2069: of the National Astronomical Observatory of Japan,

2070: and on Hitachi SR8000 at Center for Computational Science 

2071: in University of Tsukuba.

2072: 

2073: 

2074: 

2075: \begin{thebibliography}{99}

2076: \bibitem[]{}

2077: Alves, F. O. \& Franco, G. A. P. 2007, \aap, 470, 597

2078: \bibitem[]{}

2079: Alves, J., Lombardi, M. \& Lada, C. F. 2007, \aap, 462, 17

2080: \bibitem[]{}

2081: Andre, P., Ward-Thompson, D. \& Barsony, M. 1993, \apj, 406, 122

2082: \bibitem[]{}

2083: Ballesteros-Paredes, J. Hartmann, L., \& Vazquez-Semadeni, E. 1999,

2084:           \apj, 527, 285

2085: \bibitem[]{}

2086: Basu, S. \& Ciolek, G. E., 2004, \apj, 607, 39

2087: \bibitem[]{}

2088: Bertoldi, F. \& McKee, C. F. 1992, \apj, 395, 140

2089: \bibitem[]{}

2090: Caselli, P. \& Myers, P. C. 1995, \apj, 446, 665

2091: \bibitem[]{}

2092: Dib, S., Vazquez-Semadeni, E., Kim, J., Burkert, A., \& Shadmehri,

2093:           M. 2007, \apj, 661, 262 

2094: \bibitem[]{}

2095: Elmegreen, B. G. 2007, \apj, 668, 1064

2096: \bibitem[]{}

2097: Gammie, C. F., Lin, Y.-T., Stone, J. M., \& Ostriker, E. C. 2003, \apj,

2098:           592, 203

2099: \bibitem[]{}

2100: Goldsmith et al. 2008 (astro-ph/arXiv:0802.2206)

2101: \bibitem[]{}

2102: Gomez, M. Hartmann, L., Kenyon, S. J., \& Hewett, R. 1993, \apj, 105, 1927

2103: \bibitem[]{}

2104:   Hartmann, L., Ballesterios-Paredes, J. \& Bergin, E. A. 2001, \apj,

2105:           562, 852

2106: \bibitem[]{}

2107: Heiles, C. \& Crutcher, R. M. 2005, in Cosmic Magnetic Fields, Lecture

2108: Notes in Physics, vol. 664, pp137-182

2109: \bibitem[]{}

2110: Heiles, C. \& Troland, T. H. 2005, \apj, 624, 773

2111: \bibitem[]{}

2112:  Heyer, M., Gong, H., Ostriker, E. Brunt, C. 2008, (astro-ph/arXiv:0802.2084)

2113: \bibitem[]{}

2114: Heyer, M. H., Frederick, J., Snell, R. L., Schloerb, F. P., Strom,

2115:           S. E., Goldsmith, P. F., \& Strom, K. M. 1987, \apj, 321, 855 

2116: \bibitem[]{}

2117: Ikeda, N., Sunada, K., \& Kitamura, Y. 2007, \apj, 665, 1194

2118: \bibitem[]{}

2119: Kazes, I. \& Crutcher, R. M. 1986, \aap, 164, 328

2120: \bibitem[]{}

2121: Kenyon, S. J. \& Hartmann, L. 1995, ApJS, 101, 117

2122: \bibitem[]{}

2123: Kenyon et al. 2008, to appear in Handbook of Star Forming Regions, ASP 

2124: Conference Series, ed. B. Reipurth

2125: \bibitem[]{}

2126:   Krasnopolsky, R. \& Gammie, C. F. 2005, \apj, 635, 1126

2127: \bibitem[]{}

2128: Krumholz, M. R. \& Tan, J. C. 2007, \apj, 654, 304

2129: \bibitem[]{}

2130:         Krumhotz, M. \& McKee, C. F. 2005, ApJ, 630, 250

2131: \bibitem[]{}

2132:    Kudoh, T., Basu, S., Ogata, Y., \& Yabe, T. 2007, \mnras, 380, 499

2133: \bibitem[]{}

2134:    Kudoh, T., \& Basu, S. 2003, \apj, 595, 842

2135: \bibitem[]{}

2136:    Kudoh, T., \& Basu, S. 2006, \apj, 642, 270

2137: \bibitem[]{}

2138: Lada, E. A., Bally, J., \& Stark, A. A., 1991, \apj, 368, 432

2139: \bibitem[]{}

2140: Lada, C. J., Muench, A. A>, Rathborne, J., Alves, J.F., Lombardi,

2141:           M. 2008, \apj, 672, 410

2142: \bibitem[]{}

2143: Larson, R. B. 1981, \mnras, 194, 809 

2144: \bibitem[]{}

2145:   Li, P. S., Norman, M. L., Mac Low, M.-M. \& Heitsch, F. \apj, 605, 800

2146: \bibitem[]{}

2147:        Li, Z.-Y. \& Nakamura, F. 2004, \apj, 609, L83  

2148: \bibitem[]{}

2149:        Li, Z.-Y. \& Nakamura, F. 2006, \apj, 640, L187  

2150: \bibitem[]{}

2151:         Mac Low, M.-M., Klessen, R. S., Burkert, A., \& Smith, M. D. 

2152: 1998, \prl, 80, 2754

2153: \bibitem[]{}

2154:         Matzner, C. D. 2007, \apj, 659, 1394

2155: \bibitem[]{}

2156:         Matzner, C. D. \& McKee, C. F. 2000, ApJ, 545, 364

2157: \bibitem[]{412}

2158:    McClure-Griffiths, N. M., Dickey, J. M., Gaensler, B. M., Green,

2159:           A. J., \& Haverkorn, M. 2006, \apj, 652, 1339

2160: \bibitem[]{}

2161:         McKee, C. F. 1989, \apj, 345, 782 

2162: \bibitem[]{}

2163:   McKee, C. F. \& Ostriker, E. C. 2007, \araa, 45, 565

2164: \bibitem[]{}

2165: McKee, C. F. \& Zweibel, E. G. 1992, \apj, 399, 551

2166: \bibitem[]{}

2167:       Mouschovias, T.  \& Ciolek, G. 1999,

2168:    in The Origins of Stars and Planetary Systems, ed. C. Lada \&

2169:    N. Kylafis (Kluwer), p. 305

2170: \bibitem[]{}

2171: Muench, A. A. Lada, C. J., Rathborne, J. M. Alves, J. F., \& Lombardi,

2172:           M. 2007, \apj, 671, 1820

2173: \bibitem[]{}

2174: Myers, P. C. \& Lazarian, A. 1998, \apj, 507, 157

2175: \bibitem[]{}

2176:    Nakamura, F. \& Li, Z.-Y. 2005, \apj, 631, 411

2177: \bibitem[]{}

2178:    Nakamura, F. \& Li, Z.-Y. 2007, \apj, 662, 395

2179: \bibitem[]{}

2180: Nakamura, F., Matsumoto, T., Hanawa, T. \& Tomisaka, K. 1999, \apj,

2181:      510, 274 

2182: \bibitem[]{}

2183:    Nakano, T. 1984, Fundam. Cosmic. Phys., 9, 139

2184: \bibitem[]{}

2185:         Nakano, T. \& Nakamura, T. 1978, PASJ, 30, 681

2186: \bibitem[]{}

2187: Novak, G., Dotson, J. L. \& Li, H. 2007, \apj, in print (arXiv:0707.2818)

2188: \bibitem[]{}

2189:     Onishi, T., Mizuno, A., Kawamura, A., Ogawa, H. \& Fukui, Y. 

2190:        1996, \apj, 465, 815

2191: \bibitem[]{}

2192:     Onishi, T., Mizuno, A., Kawamura, A., Tachihara, K. \& Fukui, Y. 

2193:        2002, \apj, 575, 950

2194: \bibitem[]{}

2195:         Ostriker, E. C., Gammie, C. F. \& Stone, J. M. 2001, \apj, 546, 980

2196: \bibitem[]{}

2197:        Padoan, P. \& Nordlund, A. 2002, \apj, 576, 870

2198: \bibitem[]{}

2199: Palla, F. \& Stahler, S. W. 2000, \apj, 540, 255

2200: \bibitem[]{}

2201: Palla, F. \& Stahler, S. W. 2002, \apj, 581, 1194

2202: \bibitem[]{}

2203:        Price, D. \& Bate, M. R. 2008, \mnras, in press

2204: \bibitem[]{}

2205:         Shu, F. H. 1991, Physics of Astrophysics, Vol. 2: Gas Dynamics

2206:           (Mill Valley: University Science Books)

2207: \bibitem[]{}

2208:         Shu, F. H., Adams, F. C., \& Lizano, S. 1987, \araa, 25, 23

2209: \bibitem[]{}

2210:         Shu, F. H. \& Li, Z.-Y. 1997, \apj, 475, 251

2211: \bibitem[]{}

2212: Shu, F. H., Galli, D., Lizano, S., Glassgold, A. E. \& Diamond,

2213: P. H. 2007, \apj, 665, 535

2214: \bibitem[]{}

2215: Tamura, M., Nagata, T., Sato, S. \& Tanaka, M. 1987, \mnras, 224, 413

2216: \bibitem[]{}

2217:         Tilley, D. \& Pudritz, R. 2007, \mnras, 382, 73

2218: \bibitem[]{}

2219: Troland \& Crutcher 2008 (astro-ph/arXiv:0802.2253v1)

2220: \bibitem[]{}

2221: van Dishoeck, E. F. \& Black,J. H. 1988 , \apj, 334, 771

2222: \bibitem[]{}

2223: Williams, J. P., de Geus, E. J., \& Blitz, L. 1994, \apj, 428, 693

2224: \bibitem[]{}

2225: Wolleben, M. \& Reich, W. 2004, \aap, 427, 537

2226: \end{thebibliography} 

2227: 

2228: 

2229: 

2230: \begin{deluxetable}{lllll}

2231: %\tabletypesize{\scriptsize}

2232: %\rotate

2233: \tablecolumns{4}

2234: \tablecaption{Model Parameters}

2235: \tablewidth{\columnwidth}

2236: \tablehead{

2237:  \colhead{Model}     & \colhead{${\cal M}$} & \colhead{$\eta$} 

2238: & \colhead{AD}  & \colhead{Note} 

2239: }

2240: \startdata

2241: S0   & 3 & 0.75 & yes & standard model\\

2242: S1   & 3 & 0.75 & yes & different realization of initial velocity field\\

2243: S2   & 3 & 0.25 & yes & stronger spherical outflow component\\

2244: M0   & 10 & 0.75 & yes   & stronger initial turbulence \\

2245: I0   & 3 & N/A & no   & ideal MHD counterpart of standard model\\

2246: \enddata

2247: \tablecomments{

2248: All models have the same initial distributions of mass and magnetic 

2249: field. The field strength is given by the dimensionless ratio of 

2250: magnetic to thermal pressure $\alpha=24$, corresponding to a 

2251: mildly magnetically subcritical cloud of flux-to-mass ratio 

2252: $\Gamma_0=1.1$. The models differ in initial turbulent Mach number 

2253: ${\cal M}$, jet momentum fraction $\eta$, random realization 

2254: of the initial velocity field, and whether ambipolar diffusion 

2255: (AD) is included or not. 

2256: }

2257: \label{tab:model parameter}

2258: \end{deluxetable}

2259: 

2260: \clearpage

2261: \begin{figure}

2262: \plotone{f1.eps}

2263: \caption{Star formation efficiency (SFE) in the standard model as a 

2264: function of time in units of the collapse time of the condensed 

2265: sheet $t_s$. The slope of the curve indicates that less than 

2266: $1~\%$ of the cloud mass is converted into stars per local 

2267: free-fall time of the condensed sheet. }  

2268: \label{fig:sfe}

2269: \end{figure}

2270: 

2271: 

2272: \begin{figure}

2273: \plotone{f2.eps}

2274: \caption{An edge-on view of the condensed sheet at a representative

2275: time $t=11~t_s$, showing the column

2276: density distribution along the $y-$axis, which is perpendicular to

2277: the direction of initial magnetic field. The condensation is surrounded

2278: by a diffuse halo, with some spurs streaking away perpendicular to

2279: the sheet. The length is in units of $L_J$ (the Jeans length of

2280: the initial uniform state), and column density in the color-bar

2281: in units of $\rho_0 L_J$. The three lowest contours have values of 

2282: 0.05, 0.15, and 0.5, respectively, and are not shown in the

2283: color-bar. They are plotted to highlight the structure of the

2284: diffuse gas.} 

2285: \label{colden_y}

2286: \end{figure}

2287: 

2288: \begin{figure}

2289: \plotone{f3.eps}

2290: \caption{Face-on view of the same condensed sheet as in

2291:   Fig.~\ref{colden_y}, showing the column

2292: density distribution along the $x-$axis, the direction of initial 

2293: magnetic field. Superposed are stellar positions (crosses) and 

2294: contours of critical average mass-to-flux ratio (${\bar \lambda}

2295: =1$, see equation~[\ref{AverageLambda}]). The contours divide the 

2296: magnetically supercritical material capable of star formation from 

2297: the ambient subcritical material sterile to star formation.}  

2298: \label{colden_critical}

2299: \end{figure}

2300: 

2301: \begin{figure}

2302: \plotone{f4.eps}

2303: \caption{Mass fraction of magnetically supercritical material

2304: created in the initially subcritical cloud of the standard model

2305: through ambipolar diffusion. The mass fraction of supercritical 

2306: gas increases gradually initially, reaching a plateau at late 

2307: times. 

2308: }  

2309: \label{fig:supercritical}

2310: \end{figure}

2311: 

2312: 

2313: 

2314: \begin{figure}

2315: \plotone{f5.eps}

2316: \caption{The PDF of the face-on column density of the standard model 

2317: shown in Fig.~\ref{colden_critical} ({\it thick solid line}), with a 

2318: fitted lognormal distribution ({\it dashed-dotted line}) superposed.

2319: The distribution is normalized so that the peak value is close to 

2320: unity, and the column density is in units of $\rho_0 L_J$. Also

2321: plotted for comparison is the column density PDF for the ideal MHD

2322: counterpart of the standard model ({\it thin solid line}), which is 

2323: much narrower. The stark contrast between the two PDFs highlights the 

2324: key role of ambipolar diffusion in fragmenting strongly magnetized

2325: clouds.}  

2326: \label{colden_PDF}

2327: \end{figure}

2328: 

2329: \begin{figure}

2330: %\epsscale{0.6}

2331: \plotone{f6.eps}

2332: \caption{Distribution of mass as a function of density (in units 

2333: of initial density $\rho_0$, {\it thick solid line}), along with 

2334: that for the ideal MHD case ({\it thin solid line}) and 

2335: for a Spitzer sheet containing the same amount of mass ({\it 

2336: dashed line}). The mass distribution is much broader in the 

2337: presence of ambipolar diffusion, because of (magnetically-diluted) 

2338: gravitational fragmentation.}  

2339: \label{lognorm_density}

2340: \end{figure}

2341: 

2342: \begin{figure}

2343: %\plotone{rms_vel.eps}

2344: %\epsscale{0.6}

2345: \plotone{f7.eps}

2346: \caption{Distribution of mass-weighted rms velocity (in units of 

2347: sound speed $c_s$) as a function of density (in units of $\rho_0$),

2348: showing that the diffuse halo is much more turbulent than the 

2349: condensed sheet.}  

2350: \label{rms_vel}

2351: \end{figure}

2352: 

2353: \begin{figure}

2354: %\epsscale{0.6}

2355: \plotone{f8.eps}

2356: \caption{Distribution of the average Alfv\'en Mach number $M_A$ ({\it

2357:     solid line}) and Alfv\'en speed (in units of sound speed, {\it 

2358: dashed line}) as a function of density (in units of $\rho_0$), showing

2359: that the motions of the bulk condensed sheet materials are roughly

2360: Alfv\'enic whereas those of the halo materials are sub-Alfv\'enic.} 

2361: \label{ave_MA}

2362: \end{figure}

2363: 

2364: \begin{figure}

2365: %\epsscale{0.6}

2366: \plotone{f9.eps}

2367: \caption{Distribution of average magnetic field strength (in units 

2368: of the initial $B_0$, see eq.~[\ref{fieldstrength}]) as a function 

2369: of density (in units of $\rho_0$).} 

2370: \label{ave_B}

2371: \end{figure}

2372: 

2373: \begin{figure}

2374: %\epsscale{1.0}

2375: \plotone{f10.eps}

2376: \caption{Map of the logarithm of the density (in units of $\rho_0$) in 

2377: the ``midplane'' of the condensed sheet. Overlaid are contours of 

2378: gravitational potential (in units of sound speed squared) and velocity 

2379: vectors (with a unit vector of length equal to sound speed shown 

2380: above the panel for comparison). } 

2381: \label{den_pot_vel}

2382: \end{figure}

2383: 

2384: \begin{figure}

2385: \plotone{f11.eps}

2386: \caption{Star formation efficiencies as a function of time for two 

2387: models with initial turbulent Mach number ${\cal M}=10$ and $3$ 

2388: (standard model), respectively. Stars form earlier in the initially 

2389: more turbulent cloud, illustrating acceleration of star formation 

2390: by turbulence in strongly magnetized clouds.} 

2391: \label{sfetwo}

2392: \end{figure}

2393: 

2394: \begin{figure}

2395: \plotone{f12.eps}

2396: \caption{Same as Fig.~\ref{colden_y} but for an initial turbulent 

2397: Mach number ${\cal M}=10$, showing that the initially more 

2398: turbulent model has an apparently thicker condensed sheet than 

2399: the standard model.} 

2400: \label{colden_y_M10}

2401: \end{figure}

2402: 

2403: \begin{figure}

2404: \plotone{f13.eps}

2405: \caption{Map of the logarithm of the density (in units of $\rho_0$) 

2406: in an $xz$ plane that passes through the minimum of gravitational 

2407: potential. Superposed are unit vectors for magnetic field directions. 

2408: Note the prominent warp at $z\sim 0.3$. Superposition of such warps

2409: along the line of sight can make the condensed sheet appear thicker

2410: than its intrinsic thickness, as illustrated in Fig.~\ref{colden_y_M10}.  

2411: }

2412: \label{den_Bfield_M10}

2413: \end{figure}

2414: 

2415: \begin{figure}

2416: \plotone{f14.eps}

2417: \caption{Star formation efficiencies as a function of time for two 

2418: models with outflow momentum dominated by, respectively, the 

2419: bipolar jet ({\it upper curve}, standard model) and spherical component 

2420: ({\it lower curve}), showing that the spherical component of outflow

2421: is more efficient in slowing down star formation in flattened 

2422: condensations of strongly magnetized clouds. }

2423: \label{sfedependm3}

2424: \end{figure}

2425: 

2426: 

2427: \begin{figure}

2428: \plotone{f15.eps}

2429: \caption{Distribution of axis ratios of dense cores formed after the

2430:   first star was created. The abscissa and ordinate are the axis 

2431: ratios of $\xi=b/a$ and $\eta=c/a$, respectively. A large symbol 

2432: indicates a core with a mass greater than $1.5 \left<M_c\right>$, 

2433: while a small symbol indicates a core with a smaller mass. We divide 

2434: the $\xi$-$\eta$ plane into three parts: a prolate group which lies 

2435: above the line connecting $(\xi, \eta)=(1.0, 1.0)$ and (0.33, 0.0), 

2436: an oblate group which lies below the line connecting $(\xi,

2437: \eta)=(1.0, 1.0)$ and (0.67, 0.0), and a triaxial group which is 

2438: everything else. For comparison we draw the dotted line that connects 

2439: $(\xi, \eta)=(1.0,1.0)$ and (0.5, 0.0). 

2440: }

2441: \label{fig:coreshapenew}

2442: \end{figure}

2443: 

2444: \begin{figure}

2445: \plotone{f16.eps}

2446: \caption{The relationship between the sum of the two surface terms, 

2447: $S+F$, and the gravitational term, $W$, in the virial equation. 

2448: They are normalized to the sum of the internal terms, $U+K+M$.

2449: The solid line indicates the virial equilibrium, $U+K+W+S+M+F=0$.

2450: The large and small symbols are the same 

2451: as those of Figure \ref{fig:coreshapenew}.

2452: For the cores that lie below the line, the left hand side 

2453: of the virial equation (\ref{eq:virial}) is negative and 

2454: thus expected to be bound. 

2455: All others that lie above the line are unbound and expected to 

2456: disperse away, if they do not gain more mass through accretion

2457:  and/or merging with other cores, or reduce internal 

2458: support through turbulence dissipation and/or magnetic flux 

2459: reduction through ambipolar diffusion. 

2460: }  

2461: \label{fig:virialparamnew}

2462: \end{figure}

2463: 

2464: \clearpage

2465: \begin{figure}

2466: %\epsscale{0.6}

2467: \plotone{f17.eps}

2468: \caption{The virial parameter as a function of the core mass.

2469: Large, intermediate, and small symbols denote, respectively, 

2470: cores with $\Gamma_c \le 0.5$, $0.5 < \Gamma_c \le 0.7$, 

2471: and $\Gamma_c > 0.7$, where $\Gamma_c$ is the flux-to-mass ratio 

2472: of a core. The virial parameter can be fitted by a power law of

2473: $\alpha_{\rm vir}\propto M_c^{-2/3}$, indicated by the solid line.

2474: }  

2475: \label{fig:virialrationew}

2476: \end{figure}

2477: 

2478: \begin{figure}

2479: %\epsscale{0.6}

2480: \plotone{f18.eps}

2481: \caption{

2482: The nonthermal velocity dispersion as a function of core radius.

2483: The velocity dispersion is normalized to the sound speed $c_s$,

2484: and radius to the Jeans length of the Spitzer sheet $L_{s}$.

2485: The large and small symbols are the same as those of 

2486: Figure \ref{fig:coreshapenew}.

2487: There is apparently no correlation between the velocity 

2488: dispersion and the radius, in good agreement with 

2489: the observed linewidth-size relation of the dense cores in 

2490: the Taurus molecular clouds (Onishi et al. 2002). 

2491: } 

2492: \label{fig:velocitysizenew}

2493: \end{figure}

2494: 

2495: \begin{figure}

2496: %\epsscale{0.6}

2497: \plotone{f19.eps}

2498: \caption{Distribution of specific angular momentum as a function of core 

2499: radius. The specific angular momentum is normalized to $c_s L_s = 1.1

2500: \times 10^{22}$ cm$^{2}$ s$^{-1}$ (for our fiducial cloud parameters),

2501: and radius to the Jeans length of the Spitzer sheet $L_s$. 

2502: The large and small symbols are the same as those of Figure 

2503: \ref{fig:coreshapenew}. For comparison, the specific angular momenta 

2504: of 7 starless N$_2$H$^+$ cores are plotted as asterisks (data from Table 5 

2505: of Caselli et al. 2002). 

2506: }  

2507: \label{am1}

2508: \end{figure}

2509: 

2510: \begin{figure}

2511: %\epsscale{0.6}

2512: \plotone{f20.eps}

2513: \caption{The magnetic flux-to-mass ratio as a function of the column

2514:  density of the core.

2515: The large and small symbols are the same 

2516: as those of Figure \ref{fig:coreshapenew}.

2517: The cores below the solid line $\Gamma_c=1$ are 

2518: magnetically supercritical.

2519: The flux-to-mass ratio tends to be smaller for more massive cores.

2520: }  

2521: \label{fig:mass2fluxnew}

2522: \end{figure}

2523: 

2524: \clearpage

2525: \begin{figure}

2526: %\epsscale{0.4}

2527: \plotone{f21.eps}

2528: \caption{Core mass spectrum. The core mass is normalized to 

2529: the Jeans mass of the Spitzer sheet, $M_{J,s}$.

2530: The solid histogram indicates the mass spectrum of dense cores.

2531: For comparison, we show the stellar mass spectrum

2532: with the dotted histogram.

2533: The solid line denotes the power law $dN/dM \propto M^{-2.35}$, 

2534: the Salpeter IMF.

2535: There is a prominent break around $1 M_{J,s}$ in the core mass

2536:  spectrum. Above the break, the spectrum is in good agreement 

2537:  with the Salpeter IMF. 

2538: The stellar mass spectrum can be also fitted by the Salpeter IMF.

2539: }  

2540: \label{fig:coremf}

2541: \end{figure}

2542: 

2543: \begin{figure}

2544: %\epsscale{0.65}

2545: \plotone{f22.eps}

2546: \caption{3D view of density distribution and field lines of the  

2547: standard model at the same time as in

2548: Figs.~\ref{colden_y} and \ref{colden_critical}. The isodensity surfaces 

2549: have values $\rho=0.5$

2550: ({\it blue}), 15 ({\it green}) and 60~$\rho_0$ ({\it red}). They 

2551: represent three  distinct cloud components:

2552: turbulent diffuse halo,  fragmented condensed sheet, and dense cores.  

2553: The sheet is nearly magnetically

2554: critical,  whereas the halo and cores are significantly subcritical and 

2555: supercritical, respectively.}

2556: \label{3D}

2557: \end{figure}

2558: 

2559: 

2560: 

2561: \end{document}

2562: 

2563: 

2564: 

2565: 

2566: